Creation of a spatially complex mucus bilayer on an in vitro colon model

Abstract

The colonic epithelium is comprised of three-dimensional crypts (3D) lined with mucus secreted by a heterogeneous population of goblet cells. In this study, we report the formation of a long-lived, and self-renewing replica of human 3D crypts with a mucus layer patterned in the X–Y–Z dimensions. Primary colon cells were cultured on a shaped scaffold under an air–liquid interface to yield architecturally accurate crypts with a mucus bilayer (605 ± 180 μm thick) possessing an inner (149 ± 50 μm) and outer (435 ± 111 μm) region. Lectins with distinct carbohydrate-binding preferences demonstrated that the mucus in the intercrypt regions was chemically distinct from that above and within the crypts replicating in vivo chemical patterning. Constitutive mucus secretion ejected beads from crypt lumens in 8–10 days, while agonist-stimulated secretion increased mucus thickness by 17-fold in 8 h. The tissue was long-lived, > 50 days, the longest time assessed. In conclusion, the in vitro mucus replicated key physiology of the human mucus, including the bilayer (Z) structure and intercrypt-crypt (X–Y) zones, constitutive mucus flow, spatially complex chemical attributes, and mucus secretion response to stimulation, with the potential to reveal local and global determinants of mucus function and its breakdown in disease.

Introduction

Colonic mucus is present as a bilayer, the outer layer containing commensal bacteria and the inner layer being sterile. Each layer has unique physicochemical characteristics and barrier properties that together protect the epithelial cells from colonic luminal contents. Further heterogeneity in the spatial organization of the mucus is contributed by the presence of distinct subtypes of goblet cells in different regions of the epithelium: inter-crypt goblet cells, inner-crypt goblet cells, and sentinel (crypt-opening) goblet cells that secrete chemically distinct mucus¹. Inter-crypt goblet cells are characterized by their high–mucus production replenishing the mucus across the surface epithelium between the crypts and forming a mucus layer nearly impenetrable to bacteria but through which small molecules can readily diffuse¹. Inner-crypt goblet cells produce a denser mucus within the crypt lumen that serves to shield the stem cell niche. Sentinel goblet cells surround the entry to the crypt and act to expel bacteria that have penetrated the inner mucus layer². Goblet cells constitutively secrete mucus but can also respond to stimuli by exocytosis of a bolus of mucus to rapidly replace surface mucus or flush the crypts of their contents, adding further spatial complexity to the mucus layer covering the colonic epithelial-cell surface^1,3,4,5.

A variety of models have been established to study human intestinal mucus. Mucus obtained from ex vivo intestinal tissue of healthy individuals as well as those with intestinal diseases has been extensively studied^{4,5,6,7,8,9,10,11}. Using these models, differences in ex vivo goblet cell characteristics have been found in the continuous and stimulated secretion of mucus based on the goblet cell location within the crypt^{1,2,3,4,12,13}. Spatially abnormal chemical composition or mucus volume and viscosity have been identified in chronic intestinal diseases, such as Crohn's disease and ulcerative colitis, and may contribute to the breakdown of intestinal barrier properties¹². However, mucus obtained from ex vivo human tissue is quick to degrade and limited in availability. For this reason, in vivo studies often utilize a murine model to examine mucous in the different regions of the intestinal epithelium⁴. Selective lectin staining of mucus combined with single-cell mRNA and protein expression profiles have shown differences in mucus characteristics between inter-crypt and inner-crypt goblet cells¹. Murine models with dysfunctional inter-crypt goblet cells lack normal inter-crypt mucus and exhibit an inadequate mucus barrier. These animals are more susceptible to dextran sodium sulfate-induced and spontaneous colitis, lending credence to the concept that mucus dysfunction may contribute to chronic intestinal diseases. Despite this elegant work, differences in murine and human systems limit the ultimate relevance of the murine models to the living human colon, particularly concerning the topological production and other physiological properties of native human mucus.

In vitro, human mucus models have been generated from tumor cell lines (e.g., HT29-MTX)¹⁴, but due to genetic and phenotypic differences from primary cells, the generated mucus layer has a different composition and lacks the spatial properties of native human mucus¹⁵. Mucus can be produced from primary goblet cells differentiated from intestinal epithelial stem cells cultured in organoid¹⁶, monolayer^8,17, or gut-on-chip formats¹¹. Intestinal organoids are robust systems with all lineages of the epithelium represented, and significant advances have been made in these systems with respect to elucidating factors that impact goblet cell differentiation and mucus production. However, shortcomings such as the inaccessible lumen into which mucus is secreted, very low numbers of goblet cells, and the surrounding hydrogel barrier remain significant limitations with organoid model systems. To address the limited number of goblet cells formed, forced differentiation has been used to increase the number of goblet cells by withdrawing growth factors and adding notch signaling inhibitors (e.g., gamma-secretase inhibitors)^18,19,20. Topologically, the 3D architecture of the organoids is rudimentary, and the mucus lacks spatial complexity, limiting studies of regional differences in mucus production. Monolayer epithelial cell culture systems and colon-on-chip platforms can be induced to form a mucus layer and have proven to be excellent models for imaging and experimental manipulation^8,21,22,23. Some of these systems have demonstrated mucus strikingly similar in its chemical and biophysical properties to ex vivo human mucus^8,24. However, these systems also lack the architecture of the colonic crypt. Therefore, the produced mucus is significantly different from native human intestinal mucus in terms of thickness, bilayer structure, and other spatial attributes. Additionally, these tissues are short-lived, so that goblet cell physiology or phenotype cannot be studied over time to better understand mucus turnover, mucus localization, and goblet cell distribution.

An architecturally accurate in vitro human 3D crypt model based on human adult stem-cell culture has been reported. That system replicated many features of the in vivo colonic epithelium, including crypt structure polarized stem cell niche and differentiated cell zone^25,26. Goblet cells were present in the system, and some crypts possessed luminal mucus. However, goblet cell numbers and mucus production in the model were well below that present in vivo. Nevertheless, the 3D nature of this in vivo crypt model has the potential to support appropriate spatial and chemical mucus properties as well as distinct goblet cell populations. This 3D colon model was used as a starting point for optimization of the culture components and conditions to build a 3D crypt model with a physiologic and spatially complex mucus layer. In the current work, the culture format and media composition were optimized for goblet cell formation. Global and regional mucus properties were characterized based on microbead penetration, lectin binding, and immunohistochemistry. Bead clearance was used to study the self-renewal of crypt mucus and response to goblet cell stimulation. Fluorescence staining with confocal imaging was employed to define the 3D polarized crypt architecture and lifetimes of stem and differentiated cells, as well as goblet cell distribution and proportion.

Results and discussion

Use of an air–liquid interface culture to increase goblet cell number in a 3D human colon crypt model

In prior reports, air–liquid interface (ALI) culture of monolayers of primary human colonic epithelial cells increased the formation of secretory lineage cells, including goblet cells^8,27. To determine whether ALI culture might also increase goblet cell proportion in the 3D structured epithelium, a collagen-based hydrogel was micromolded into the shape of human colonic crypts with a physiologic crypt density (Supplementary Fig. 1). The hydrogel scaffolding was formed in the base of a hanging basket fabricated from a modified Transwell insert (Supplementary Fig. 1). Primary colonic epithelial stem cells were then seeded onto the scaffold and maintained in a submerged culture, i.e. with WRN (Wnt, R-spondin and Noggin) supplemented medium fully covering the luminal and basal aspects of the scaffold. After 2 days of culture, the epithelial cells partially covered the scaffold and the liquid medium was removed from the luminal surface leaving WRN-supplemented medium (50%) in the basal reservoir. The cells were cultured under this luminal-air, basal-liquid condition for 10 days, at which time the cells grew to cover the luminal surface and inner walls of the crypt microwells (Supplementary Fig. 1).

To determine whether crypts that were formed under ALI conditions (Fig. 1a–c) might possess stem/proliferative cell zones spatially separated from differentiated cell regions, the cultures were pulsed with EdU to identify stem/proliferative cells and immunostained with Muc-2 to identify goblet cells (Fig. 1d–g). The average location of the EdU⁺ cells was 111 ± 58 μm from the crypt base (450–500 μm long crypt), with less than 5% of EdU⁺ cells in the luminal or upper 60% of the crypts (Fig. 1d, e). Goblet cells were located throughout the crypt length and along the luminal surface (Muc2⁺ area normalized to Hoechst 33342⁺ area = 57.6% ± 6.2%, Fig. 1f, g). Living colonocytes were assayed for alkaline phosphatase (ALP), and these cells were localized to the differentiation zone, which covered the upper 80% of the crypt length and also the luminal surface (Fig. 1e). These data demonstrated that it was possible to form polarized crypts under the ALI culture conditions and suggested that a gradient of WRN was formed along the long axis of the crypt in the presence of ALI culture²⁵.

Figure 1

Generation of an in vitro human crypt array. (a) Brightfield image of a single row of 5 crypts after cell seeding (days 2–10). The array was under ALI culture. (b) Brightfield image of the side view of a single crypt dissected from an array on day 10. (c) Brightfield top view of a crypt array on day 10 cultured under ALI. (d) Superimposed brightfield and fluorescence image (projected image Z-stack with maximum intensity projection over 31 slices separated by 15 μm each) of an array section with 21 crypts on day 10 under ALI culture. EdU incorporation 24 h (green). (e) Side view (Projected image Z-stack with maximum intensity projection over 25 slices with 5 μm separation) of a single excised crypt cultured in ALI and stained for Muc2, ALP, EdU, and DNA (Hoechst 33342). (f) Shown is the normalized area positive for EdU incorporation and mucin 2 in submerged (Sub) and ALI cultures after 10 days. The normalized area was calculated as the fluorescence area positive for the stain divided by the area positive for Hoechst 33342 stain. (g) Fluorescence image (Projected image Z-stack with maximum intensity projection over 31 slices separated by 15 μm each) of the top view of a crypt array at day 10 in ALI and submerged culture. EdU (green) and Muc2 (yellow). Data is shown as mean ± SD, analyzed by two-way ANOVA, with multiple comparisons and N = 3. *P < 0.05, ***P < 0.0001.

The crypts formed under ALI conditions were compared to those formed under submerged conditions with a luminal medium in which WRN was present in the basal but not luminal medium. Compared to the submerged culture, the ALI culture significantly increased the Muc2+ area (goblet cells) compared to the area positive for Hoechst 33342 (submerged: Muc2⁺ area normalized to Hoechst 33342⁺ area = 26.8 ± 1.7%; ALI: normalized Muc2⁺ area = 57.6% ± 6.2, p = 0.0004, Fig. 1f,g). The relative area of proliferative cells was also increased under ALI (EdU+ area normalized to Hoechst 33342⁺ area = 31.8 ± 6.3% in ALI), compared to that in the submerged cultures (18.5 ± 7.2%, p = 0.0431) (Fig. 1g). These data demonstrate that ALI culture increased the proportion of goblet cells (compared to all cells), as was demonstrated previously for colonic epithelial monolayers under ALI culture⁸. This difference may be due to an increased O₂ concentration near the cells as a result of the absence of a fluid column above the epithelium²⁷ or the altered shape of the WRN gradient formed under the ALI conditions in the absence of a luminal fluid reservoir)^19,28,29. Although ALI culture increases goblet cell generation in vitro, the spatial pattern of goblet cell number/location was still reduced compared to the number of goblet cells in vivo (from 57 ± 10 current to the in vivo ~ 200 to 500 goblet cells/crypt)^30,31. Additionally, the stem/proliferative compartment extended farther up than the crypt than is physiologic. Further optimizations were required, with a particular focus on increasing intra-crypt goblet cells as well as limiting the stem/proliferative cell region to the crypt base.

Optimizing the culture conditions to increase the number of goblet cells

Several strategies were investigated to enhance the proportion of goblet cells (as measured by the relative area compared to Hoechst 33342 area) in the human 3D crypt arrays. These included: (1) optimization of medium additives, (2) adjustment of the concentration of WRN in the basal compartment, and (3) determination of the ideal culture duration. Both interleukin-22 (IL-22) and vasoactive intestinal peptide (VIP) have been reported to increase goblet cell numbers in intestinal epithelial cell cultures^32,33. For these assays, cells were pulsed with an EdU analog during ALI culture, and the tissues were fixed with paraformaldehyde, followed by removal of the secreted mucus layer, leaving only the retained intracellular mucins for quantification of goblet cell proportion. The cells were then immunostained for Muc2, incubated with Hoechst 33342 to label DNA, and the EdU analog was fluorescently labeled. When IL-22 was added to the basal medium, the proportion of goblet cells (as represented by the Muc2⁺ area normalized to the Hoechst 33342⁺ area) was significantly decreased relative to ALI cultures without IL-22, suggesting that under these conditions, IL-22 decreased the proportion of goblet cells in the crypts (Fig. 2a, b). In contrast, the addition of VIP to the basal reservoir significantly increased the normalized Muc2⁺ area relative to that of ALI cultures without VIP (Fig. 2b, c), suggesting that VIP increased goblet cell proportion consistent with prior reports of VIP’s actions^34,35,36. The addition of basal IL-22 increased EdU-incorporation (p < 0.05) compared to that of crypts treated with basal VIP (Fig. 2b, d) suggesting that IL-22 supports stem/proliferative cells consistent with previous reports^37,38. Since VIP enhanced goblet cell formation without eliminating the stem/proliferative cell compartment in the crypts (Fig. 2e), all subsequent cultures included VIP in the basal reservoir. An additional benefit of VIP is its capacity to hydrate the mucus layer due to its effect on ion fluxes and, consequently, water movement^8,39.

Figure 2

Effect of VIP and IL-22 in goblet cell formation during ALI culture. (a) Normalized fluorescence area of EdU or Muc 2 with and without VIP or IL-22 under ALI culture on day 10. The normalized area was calculated as the fluorescence area positive for the stain divided by the area positive for Hoechst 33342 fluorescence. Data was obtained from a project image stack with maximum intensity projection (MIP) of 25 slices (5 μm separation). (b) Normalized location along the crypt length of Muc2⁺ cells with and without VIP or IL-22. Zero marks the crypt base while 1 is the crypt luminal end. (c) Normalized location along the crypt of EdU⁺ cells with and without VIP or IL-22. Shown in (a–c) are the mean ± SD, analyzed by two-way ANOVA, with multiple comparisons and N = 4. ns non-significative difference, **P < 0.01, ***P < 0.0001. (d) Top view (Projected image Z-stack with maximum intensity projection over 26 slices with 20 μm separation) with or without VIP or IL-22. EdU⁺ (green), Muc 2 (yellow), Hoechst 33342 (blue). (e) Side view (projected image Z-stack with maximum intensity projection over 25 slices) of a pair of crypts showing EdU (green) and Muc2 (yellow) staining with or without VIP or IL-22.

Previous investigators have demonstrated that Wnt3a, Noggin, and R-spondin concentrations modulate lineage allocation and impact the number of goblets present⁴⁰. To assess the impact of the WRN concentration on the number of goblet cells as well as the number of proliferative cells, five different dilutions (0, 15, 30, 40, 50%) of a WRN-conditioned medium^41,42 were assessed in the basal medium during the 10 d of ALI culture (Fig. 3). The largest dilutions (0, 15%) of the WRN-conditioned medium yielded significantly decreased proliferative cell numbers (as measured by the normalized EdU⁺ area) but significantly increased goblet cell proportion (as measured by the normalized Muc2⁺ area) (Fig. 3a, b). Dilutions of the conditioned medium at > 30% significantly reduced the number of proliferative cells, and increased the number of crypts possessing no EdU⁺ cells or no cells at all. For example, > 60% of the crypts possessed no EdU⁺ cells when WRN was ≤ 15% (Fig. 3c, d). Conditioned medium dilutions of < 30% possessed both EdU⁺ and Muc2⁺ cells in the majority of crypts. Since lower WRN concentrations promoted goblet cell formation, 30% L-WRN conditioned medium was selected as the optimal W-R-N concentration corresponding to WRN concentrations of 120 ng/mL (Wnt-3A), 900 ng/mL (R-spondin), and 240 ng/mL (Noggin).

Figure 3

Effect of basal-WRN concentration on proliferative and goblet cells. (a) Normalized location of EdU⁺, Muc2⁺ and ALP⁺ cells at different basal WRN after 10 days of ALI culture. Zero marks the crypt base while 1 is the crypt luminal end. (b) Shown is a side view of single excised crypts after 10 days under ALI culture and at varying basal WRN concentrations. The 3D rendering was created from 14 Z-stack slices over a 140 μm depth. Muc2 (yellow), ALP enzymatic reaction (red), EdU (green), Hoechst 33342 (blue). Grid scale = 50 μm intervals. (c) Percentage of crypts with EdU⁺ cells after 10 days of ALI culture under varying WRN concentrations. Shown is the mean ± SD analyzed by two-way ANOVA with multiple comparisons and N = 3. ns: non-significative difference, *P < 0.05, **P < 0.01, ***P < 0.0001. (d) Top view of representative crypt array sections (Projected image Z-stack with maximum intensity projection over 26 slices with a 20 μm separation) cultured under different concentrations of basal WRN. Fluorescence (EdU, green) is superimposed over a brightfield image.

When the crypt arrays are cultured under submerged conditions, cells in the crypt base proliferate, and a subset of these cells migrate up the crypt walls and differentiate, replacing the short-lived cells lining the luminal surface (which slough from the luminal surface upon apoptosis). This self-renewing aspect in which terminally differentiated cells are continually replaced yields a long-lived tissue construct^43,44. To determine whether cells in the crypt base formed under the ALI microenvironment and optimal WNR concentrations might similarly act in a self-renewing manner, an EdU pulse-chase experiment was performed. An EdU analog was incubated with the cells for 3 h and then washed from the medium. The incorporated EdU analog was then labeled at 0, 24, 48, 72, and 96 h after the washout (Fig. 4a, b). At 0 h after EdU washout, all EdU⁺ cells resided at the basal end of the crypt. By 96 h, EdU⁺ cells were readily observed throughout the length of the crypts, demonstrating that some of the cells incorporating the EdU analog were migrating up the crypts toward the lumen. On average, the cells moved with a velocity of 44 ± 15 μm/day along crypts. Next, the impact of ALI culture duration on the proportion of goblet cells formed was measured by culturing the crypt arrays under ALI for 10–20 days and then quantifying the area of Muc2⁺ in different regions of the array. On days 16, 18, and 20, the normalized Muc2⁺ area at the inter-crypt region or luminal surface of the array was significantly greater (> 24.08%, p < 0.05, > 53.12%, p < 0.0001 and > 66.11%, p < 0.0001, respectively) than that on day 10 (Fig. 4c,d). The proportion of goblet cells (measured as the area positive for Muc2⁺ compared to that positive for Hoechst 33342) in the inner-crypt zone also increased significantly on day 12 (P = 0.0027), 14 (P = 0.0010) and 16 (P < 0.0001) compared with day 9 suggesting that goblet cell proportion within the crypt were increasing over time (Fig. 4e, f). However the normalized Muc2⁺ area within the crypt lumen on days 18 and 20 was not significantly different from that on day 16. In this model system, the area occupied by Muc2+ cells (goblet cells) relative to that occupied by Hoechst 33342 was 36.6% within the crypt on day 14, and 38.6% in the intercrypt area on day 16 consistent with that measured in living humans. In the human colon, 20–40% of cells that form the axis between the crypt-base and surface epithelium are goblet cells^31,45. Our data are consistent with the literature since our metric of relative Muc2+ area (to Hoechst 33342+ area) likely yields an overestimate of the percentage of goblet cells. The number of EdU⁺ cells did not significantly change over time (295 ± 18 cells/crypt at day 20). The proportion of EdU⁺ cells is consistent with the number of proliferative cells/crypt in vivo (~ 15.8% of total cells in a human colon crypt⁴⁶ with an average of 2,000 cells/crypt)^30,47.

Figure 4

Characterization of crypt arrays formed under ALI culture. (a) Pulse-chase EdU labeling experiment showing the location of EdU (green) labeled cells along the crypts at 0, 24, 48, 72, and 96 h after EdU washout. Shown is a projected image Z-stack with maximum intensity projection over 15 image slices separated by 10 μm. Hoechst 33342 (blue). (b) Normalized migration distance of EdU⁺ cells in the crypt after the EdU pulse. Zero marks the crypt base, while 1 is the crypt luminal end. (N = 3). Each data point is shown as a green circle. (c) Top view (Projected image Z-stack with maximum intensity projection over 9 slices separated by 10 μm) of crypt arrays at varying days of culture. All slices were through the epithelium and not through the overlying mucus layer. (d) Normalized Muc2⁺ fluorescence area after varying culture times. The normalized area was calculated as the Muc2⁺ fluorescence area positive divided by the area positive for Hoechst 33342 stain. (e) Projected image Z-stack with maximum intensity projection over 15 slices separated by 10 μm of single excised crypts. Images are shown at varying times of culture. (f) Normalized Muc2⁺ fluorescence area within the lumen of excised crypts. (g) An array of crypts was cultured for 50 days and then stained and imaged (N = 3). The left panel is a top view of EdU incorporation [projected image Z-stack with maximum intensity projection over 15 slices separated by 36 μm and acquired from along the crypt long axis and luminal surface (inset)]. Similarly, the left middle panel shows a top view of the fluorescence of Hoechst 33342 fluorescence along the entire array depth (inset). The right middle panel shows the Muc2, Edu, and Hoechst 33342 fluorescence (projected image Z-stack with maximum intensity projection over 9 slices) from the crypt region only and not the luminal-most cells (inset). The right panel shows an image (projected image Z-stack with maximum intensity projection over 10 slices) of the Muc2 fluorescence along the luminal surface of the array (inset). For (b,d,f), shown is the mean ± SD, analyzed by two-way ANOVA with multiple comparisons. ns non-significative difference, *P < 0.05 **P < 0.01, ***P < 0.0001.

These results showed that longer ALI culture durations increased the proportion of inter-crypt and inner-crypt goblet cells with no changes in EdU⁺ cells; thus, all subsequent experiments used ALI culture durations of > 18 days. In these ALI cultures, the crypt-base cells proliferated with their offspring migrating up the crypts to replenish the short-lived luminal cells, as demonstrated previously for the fully submerged constructs. Given this self-renewing behavior, the length of time that the polarized crypt arrays with confluent epithelial cells could be cultured was assessed. The basal medium was exchanged daily for the ALI cultures, and the polarized 3D crypts were cultured for up to 50 days (the longest duration assessed) (Fig. 4g). During this time interval, the differentiated cells are expected to have been replaced ~ 12 times since the cells require 4 d to migrate from the crypt base to the luminal surface. Human colon crypts in vivo regenerate every 4–7 days⁴⁴ so that this in vitro system was remarkably similar to the in vivo crypt behavior.

Formation of a bilayer mucus barrier

To reveal the mucus layer, the crypt arrays were cultured under ALI and then fixed using Carnoy’s solution, preserving the mucus. The location of mucus within and on the top of the luminal surface was revealed by microscopic inspection, following immunostaining for Muc2. On day 9, mucus began to appear in the interior region of the crypts. On day 10, a thin layer of mucus could be seen on the luminal surface. Mucus secretion and accumulation continued to increase with the increase of ALI culture duration, potentially due to the maturation of goblet cells and/or the increased number of goblet cells (Fig. 5). On days 18 and 20, mucus filled the entire interior region of crypts and covered the luminal surface. The growing thickness of the mucus layer (from 5 μm, day 10 to 600 μm, day 18) suggested that the mucus was being constitutively produced likely due to a combination of intercrypt and intracrypt goblet cell secretion. On day 20 of ALI culture, the mucus layer could be stretched with forceps, revealing its adhesive and viscoelastic properties (Fig. 5a). At day 20 in ALI, the mucus was firmly anchored to the crypts and covering the entire 11.2 mm-diameter hanging basket surface (even areas without crypts) with a thickness of 605 ± 180 μm.

Figure 5

Characterization of the mucus bilayer on human 3D-crypts in vitro. (a) Shown is a photograph of mucus above a crypt array (20 days of culture) grasped with forceps. (b) Diagram of the upper portion of an excised crypt (20 days of culture) with attached plume of mucus. The middle panel shows the crypt stained with Hoechst 33342 (blue) with a dotted white line tracing the opening of the crypt lumen. The right panel is an image made from 22 summed Z slices (total thickness of 330 μm). Muc2 (yellow) and Hoechst 33342. The white dotted line outlines the crypt. (c) Diagram of the experiment overlaying different sized fluorescent beads onto a crypt array with a mucus layer. (d) Side view of 3 crypts with mucus (not stained). Left panel shows green beads (0.05 μm diameter). Left middle panel shows red beads (1 μm). Right middle panel demonstrates Hoechst 33342 fluorescence. Right panels are comprised of 29 summed Z stack images (145 μm thickness). Scale bar = 200 μm N = 3, n = 14.

One feature of the in vivo colonic mucus is its bilayer structure: the inner mucus layer is dense, firmly anchored to the epithelial cells, and free of bacteria. The outer mucus layer is looser and colonized with commensal bacteria⁴⁸. This bilayer structure maintains epithelial cell hydration and reduces mechanical shear forces associated with the movement of intestinal contents across the colonic epithelium. The mucus bilayer discussed in this report is a tissue-associated mucus barrier rather than fecal-associated mucus⁴⁹. When single crypts were excised from the crypt array, large quantities of mucus remained attached suggesting that the mucus was anchored within the crypt (Fig. 5b). To determine whether the two mucus layers were present above the in vitro crypt array, a mixture of fluorescence beads of two diameters, 1 μm (red) and 0.05 μm (green), was overlaid onto the mucus surface after 20 days of ALI culture (Fig. 5c). The 1 μm beads mimicked the size of bacteria, while the 0.05 μm beads mimicked the size of a large protein⁵⁰. As expected, the very small green beads penetrated the mucus to reach the cells, occupying the entire depth of the mucus layer and diffusing into the crypt lumens (Fig. 5d). In contrast, the larger red beads penetrated only the top layer of the mucus (435 ± 111 μm), and a gap (149 ± 50 μm) between the red beads and epithelial cells remained. These data suggested that both an outer and inner layer of mucus was present above the epithelium. The measured inner and outer layers were close to that reported for in vivo mucus, 110–200 μm and 650–800 μm, respectively¹⁸. The mesh size of the outer layer was sufficiently large to admit bacteria-sized objects; however, the inner layer pore size excluded these larger objects but enabled protein-sized entities to diffuse to and contact the epithelial cells. Notably, Hoechst 33342-positive debris was observed within the mucus layers (Fig. 5b,d). Given the short lifespan of the differentiated epithelial cells (~ 5 to 7 days), these objects were most likely remnants of dead cells shed from the luminal epithelial surface. This buildup of debris may also result from the absence of fecal movement across the epithelial surface.

Formation of intra-crypt mucus

To understand whether mucus might be produced within the crypts, i.e., by cells lining the crypt walls, the ability of fluorescent beads to be expelled from the crypt lumen was measured (Fig. 6a). After 6 days under ALI culture, fluorescent beads (1 μm) were added to the luminal crypt array before the accumulation of significant surface mucus. The beads settled onto the array surface by gravity and filled the lumens of the crypts since the crypts were devoid of mucus at that point in time. ALI culture was then continued for a further 10 days, and the percentage of beads remaining within the crypts was measured over time (Fig. 6b). Over the next 2–10 days, the number of beads remaining in the crypt decreased to near zero (Fig. 6a, c). The drop in the bead number paralleled the increase in Muc2 immunofluorescence (Fig. 6a, d) and mucus layer thickness, suggesting that mucus produced within the crypts was expelling the beads from the crypt lumen, i.e., luminal surface mucus was not merely flowing down into the crypts. These data suggest that the intra-crypt mucus was constitutively produced at a sufficient rate to eliminate contaminants from the crypt lumen within 8 days, protecting the stem/proliferative cells from microbial intruders.

Figure 6

Mucus flow within human colon crypts. (a) Diagram of the experiment of displacement internal beads by constitutive mucus production. (b) Time course of bead expulsion from the crypt lumen (left axis) and normalized Muc2 area within the crypt lumen (right axis). The Muc2 positive area was normalized to the Hoechst 33342 positive area. (c) Red fluorescent beads (1 μm) were placed on an array on day 9 of ALI culture prior to significant mucus accumulation. Shown are overlaid brightfield and red fluorescence images of a single crypt excised at varying culture times after bead overlay on day 9 of culture (d). Crypts were excised from an array after varying times of culture. Shown is a cross-sectional image through the crypts created from a projected image Z-stack with maximum intensity projection over 15 image slices separated by 10 μm. Muc2 (yellow), EdU (green), Hoechst 33342 (blue).

Secretagogue-responsive mucus secretion

In vivo, goblet cells lining the intestinal lumen can respond to secretagogues and rapidly exocytose large quantities of mucin to increase the luminal mucus thickness or repair the mucus barrier in response to bacterial infection^2,3. To determine whether the in vitro crypt array might also possess this rapid, protective response, carbachol (cholinergic agonist) was added to the basal reservoir on day 10 of culture. The crypt array was then fixed with Carnoy’s solution and imaged. Prior to carbachol addition, the mucus thickness was 5.57 ± 2.28 μm (Fig. 7a, b). Eight hours after stimulation, the mucus thickness was significantly increased (88 ± 40 μm) compared to that of an unstimulated control (Fig. 7b, c). The in vitro tissue replica of the colonic epithelium was capable of discharging mucus in response to carbachol stimulation, similar to that of the ex vivo human colon, which has been demonstrated to increase the mucus layer by 140 μm upon stimulation with carbachol⁵.

Figure 7

Crypt goblet cells release mucus in response to carbachol. (a) Shown are photographs of crypt arrays within the hanging basket before and after carbachol application. (b) Mucus thickness over a crypt array after application of carbachol for varying times. N = 3, n = 36. Analyzed by one-way ANOVA with Tukey’s multiple comparisons test. ns non-significative difference, *P < 0.05, ***P < 0.0001. (c) Representative side views of crypt arrays (projected image Z-stack with maximum intensity projection over 35 slices separated by 5 μm each) at 0, 4, and 8 h after carbachol stimulation. Scale bar = 200 μm.

Spatial organization of mucus

Spatial complexity, in addition to that of the inner and outer layers, is present within the mucus covering the colonic epithelium^1,12. To identify spatial differences in mucus composition, the crypts were stained with three different lectins: LEL, which binds N-acetylglucosamine, LTL which binds fucose, and WGA, which binds N­acetylglucosamine/sialic acid. LEL, LTL, and WGA have been shown to bind to different regions of the mucus layer in the crypts (Fig. 8a)^51,52. When intact arrays were imaged by confocal microscopy, WGA and LTL lectin were identified in the interior of crypts (XZ cross-sections) (Fig. 8b–e). In contrast, LEL was found only near the luminal opening of the crypts (Z = 310–500 μm) but not within the crypt lumen (Z < 310 μm) (Fig. 8b–e). XY cross-sections through the middle of the crypt arrays demonstrated LTL and WGA, but LEL was relatively sparse (Fig. 8e). These results are similar to the mucus in the intercrypt region observed in vivo for human crypts [WGA⁺/LEL⁺/LTL⁻]¹. In healthy humans, LTL primarily stains the crypt lumen, WGA-binding is detected at the crypt lumen and intercrypt region, and LEL binds to the intercrypt region but not at the crypt lumen¹. These results suggested that this in vitro intestinal model possessed additional spatial complexity beyond that of an inner and outer layer of mucus.

Figure 8

Spatial patterning of mucus on the crypt arrays. (a) Shown is a 9.07 μm thick confocal cross-sectional image (side view) through the lumen of a crypt. Content stained with fluorescent WGA (pink), LTL (green), LEL (red), and Hoechst 33342 (blue). (b) Projected image Z-stack with maximum intensity projection over 15 slices (XZ) separated by 10 μm and located near the top (Z = 400–500 μm) of the crypts stained with the conjugated lectins (c). A single 9.07 μm thick confocal slice at the luminal entryway into the crypt (Z = 500 μm) stained for the various lectins (d). Graphical analysis of lectin content vs location along the crypt length. The area positive for the lectin fluorescence was divided by the area positive for the Hoechst 33342 fluorescence. The areas were calculated using a MIP from 7 XZ slices through a half of the crypt. N = 3. (e) Schematic of a crypt with the observed lectin staining.

Conclusion

An arrayed human in vitro 3D crypt model was optimized to enhance cell differentiation towards the goblet cell lineage with production of a mucus bilayer. The optimal culture conditions were identified as ALI culture, medium with 30% L-WRN conditioned medium, and including VIP with a culture duration of > 18 days. These conditions not only created an in vitro self-renewing array of colon crypts with inter-crypt and inner-crypt goblet cells but also produced a functional mucus layer overlying the luminal surface that mimicked key features of the in vivo colonic mucus. The in vitro mucus was dynamic, with a self-renewing cycle of 8 days to replenish the mucus inside the crypts, as demonstrated by bead clearance. The presence of functional goblet cells in the 3D crypts was verified by their ability to rapidly discharge stored mucins in response to agonist stimulation. The in vitro mucus exhibited a bilayered structure above the epithelial cells as well as spatially distinct lectin-staining, indicating that both the density and composition of the mucus were patterned in this 3-D crypt model. These features distinguish this in vitro mucus model from previously reported models (e.g., planar monolayers, microfluidic organ-on-chips) with replication of key physiological attributes thought to be important in health and disease. The crypts are self-renewing and long-lived for > 50 days. This human colon with physiologic mucus dynamics has a multitude of applications in mucus biology including examination of the impact of genetics, diet, microbiota, and chemical environment on mucus properties. Other potential applications include colonic mucosal drug delivery and barrier function enhancement. This in vitro 3D crypt system will accelerate knowledge critical to understanding the etiology of diseases thought to have origins of mucus barrier dysfunction, e.g., inflammatory bowel disease.

Methods

Ethics statement

All methods were carried out in accordance with relevant guidelines and regulations stipulated by the National Institutes of Health. The experimental protocols were approved by the University of Washington. The cell line used in this work was derived from de-identified cadaveric transplant donor intestines made available through an NIH-funded biobank. The original tissues to create this biobank were obtained from a federally designated organ procurement organization. Transplant donors were consented and de‐identified by the procurement organization.

Human colon cells

Stem cells used in this research are reported as isolated and expanded from the transverse colon of a cadaveric donor organ. Available data of the cell line is in accession: CVCL_ZR41 with RRID: CVCL_ZR41. https://web.expasy.org/cellosaurus/CVCL_ZR41).

Expansion of human colonic epithelial stem cells

The stem cells were maintained as proliferative monolayers on a 1-mm thick collagen hydrogel matrix in six-well plates using maintenance medium (MM, Supplementary Table S1). The process for expansion, maintenance, passage, and cryopreservation of stem cells is described in previously published protocols^26,53 (See Supplementary material). The stem cells were cryopreserved at passage number 8 (p8). For each experiment, one vial of cryopreserved cells (p8) was thawed and propagated as stem cells in 1 well of collagen hydrogel matrix of a 6-well plate in MM. When the monolayers reached ~ 70% confluency, they were passaged (p9) to 3 new wells (passage ratio 1:3). After 3 days, the collagen hydrogel (with attached cells) was dislodged from the wells and transferred to a 15 mL tube containing 500 U/mL collagenase-IV in 3 mL MM for passage (p10). The intact mass of collagen hydrogel was broken up into small pieces by repetitive pipetting with a 5-mL serological pipet, followed by 1-mL pipet tip, and then incubated at 37 °C for 10 min to digest the collagen hydrogel completely. The cell monolayers were pelleted by centrifugation at 600×g for 1 min, washed with 10 mL phosphate-buffered saline (PBS), and pelleted again by centrifugation. The pellet was incubated with 500 μL of TrypLE (Gibco, 12605028) for 10 min at 37 °C, followed by repetitive pipetting using a 200 µL pipet tip to break the monolayers into small cell fragments. The protease reaction was then stopped by adding 5 mL of expansion medium (EM, Supplementary Table S1), which contains 10% fetal bovine serum, and then centrifuged at 600×g for 1 min. The final pellet was resuspended in EM and plated on 3D scaffolds. All experiments were performed with the same cell passage number (p10) to eliminate the possible passage-to-passage variation.

In vitro generation of self-renewing human 3D crypt array

Human 3D crypt arrays were generated by plating stem cells on micromolded and chemically crosslinked 3D collagen hydrogel scaffolds according to previous publications^25,26(see Supplementary material). The scaffolds were fabricated on modified 12-well inserts (Transwells), possessed a diameter of 3 mm, and were covered with an array of microwell structures (~ 210 crypts of 450–500 μm length and 100–125 μm crypt opening). The suspension of cell fragments (see above) from three wells of the six-well plate was dispersed into nine separate inserts (0.5 mL EM with cells in the upper reservoir and 1.5 mL EM without cells in the lower reservoir). After two days, when the stem cells were attached to the scaffolds, two culture methods were tested to generate the polarized crypts: submerged and ALI. In submerged culture, luminal and basal EM were exchanged every two days after plating stem cells. During the ensuing 6–7 days, the cells proliferated and covered the entire scaffold. To polarize the crypts, the medium in the upper reservoir was then switched to 0.5 mL differentiation medium (DM, Supplementary Table S1), and that in the lower reservoir was maintained as 1.5 mL EM (containing growth factors)²⁵. The media in both reservoirs was exchanged every 24 h thereafter.

For ALI culture, the cells were grown for 2 days, and then EM was aspirated from both reservoirs (Fig. 1). ALI was initiated by removing liquid from the luminal reservoir and adding 1 mL modified expansion medium (M-EM, Supplementary Table S1) to the lower reservoir. The WRN concentration of M-EM was modified from that of EM in order to enhance the differentiation of stem cells towards goblet cell lineage and maintain a polarized crypt. The WRN concentrations of undiluted L-WRN conditioned medium were 400 ng/mL (Wnt-3A), 3 μg/mL (R-spondin), and 800 ng/mL (Noggin), respectively^41,42. The M-EM in the lower reservoir was exchanged every 24 h thereafter, up to 50 days (the longest time tested). The compositions of media (MM, EM, M-EM, DM) are listed in Supplementary Table S1.

Fluorescence staining, dissecting, and imaging of 3D crypts

An EdU-ALP-Muc2-DNA staining protocol^36,53 was used to reveal the distribution of proliferative and differentiated cells on the 3D crypt array, with each stain quantified using a different fluorophore. EdU-incorporation (5-ethynyl-2-deoxyuridine, A10044, ThermoFisher Scientific) marked the proliferative (S-phase) cells, an ALP (alkaline phosphatase, SK-5100, Vector Laboratories) enzyme product stained the colonocytes, Muc2 antibody (mucin-2, 1:200, Sc-7314, Santa Cruz Biotechnology) stained the goblet cells. Hoechst 33342 (Hoechst, B2261, Millipore-Sigma) stained the DNA of all cells. Crypts were labeled with an EdU pulse during the 24 h prior to fixation (unless otherwise specified). The immunodetection or fluorescent staining of crypts was performed after fixation with glyoxal fixative (Prefer, Anatech Ltd, 20 °C, 15 min), 4% paraformaldehyde (20 °C, 20 min), or Carnoy’s solutions (60% absolute ethanol, 30% chloroform, 10% glacial acetic acid, 4 °C, 30 min) as specified in each experiment. The secondary antibody was (Alexa Fluor 488-conjugate goat anti-mouse (1:500, Jackson Immunoresearch, 115-545-003). The upper or luminal surface (X–Y dimension) of the epithelium with crypts was imaged using a confocal laser scanning microscope (Olympus, Fluoview FV3000) for EdU-incorporation (CY5 channel), an ALP (Texas Red channel), Muc2 (Alexa Fluor 488 channel), and Hoechst 33342 (Hoechst 33342 channel) with laser-based excitation and emission wavelengths selected using a holographic transmission diffraction grating.

To obtain high-quality side view (X–Z dimension) of the crypts and secreted mucus layer, two methods were used (Supplementary Fig. 2). In the first method, crypts (individual crypts, pairs, triplets, and up to a maximum of 5 crypts in a row) were excised from the array using a micro-dissecting needle (Roboz Surgical Instrument, RS-6065). When transferred to a 96-well plate, the released intact crypts were laid on their side and imaged along their long-axis. In the second method, the membrane below the collagen scaffold was detached from the transwell insert. The array of crypts was cut into halves using micro-dissecting scissors (Roboz Surgical Instrument; RS-5910). Each half was perpendicularly placed on a glass slide to enable the crypts (oriented on their side) nearest the glass surface to be imaged by confocal microscopy. This method permitted imaging of up to 17 crypts in a row, the crypt contents, and the mucus layer adjacent to the crypts. Image analysis was performed using the open-source image processing software ImageJ/Fiji v2.14.0⁵⁴ and Imaris X64 v9.8.2 (imaris.oxinst.com, Oxford instruments). The 3D crypt structure was reconstructed and analyzed in Fiji using a maximum intensity projection of the optical sections (z-slices) of each image collected in the x,y-plane at different z (depth) planes by confocal microscopy. The maximum intensity projection method was applied to the pixels within the stack for each fluorescence channel, followed by manual thresholding for a defined region of interest (ROIs). The visualization of the crypt lumen (half crypt, cross-sectional view) or the entire crypt was obtained using the maximum intensity projection of an overlaid image created with a manually defined range of the Z stack slices⁵⁴.

Characterization of the upward flow of mucus from crypts

Cells were seeded on 3D scaffolds, and the culture was submerged in EM (luminal and basal reservoir) for 4 days. Starting on day 5, the crypts were converted to ALI by removing all liquid with growth factors from the luminal reservoir and initiating crypt polarization under these conditions. On day 9, the crypts were polarized, but the mucus layer was not yet formed. To load 1 μm beads into the crypt lumens, a 300 μL suspension of fluorescent beads (Thermo Scientific, F13083) was diluted 1000× in M-EM and added on the luminal side, and incubated for 24 h. On day 10, the luminal medium was aspirated, and the culture was returned to either ALI or submerged culture. On days 0, 2, 4, 6, 8, and 10 after loading with beads, the crypts were fixed with 4% paraformaldehyde for 20 min at 20 °C, excised, and the location of fluorescence beads within the crypt lumen was imaged. The mucus within the crypts was also imaged after staining with anti-Muc2 antibody.

Mucus secretion in response to carbachol

The cells were cultured in ALI for 10 days to generate mucus-covered, polarized crypt arrays. The cells were pulsed with EdU for 24 h prior to the assay. To study stimulated mucus secretion, 10 μM carbachol (Sigma, C4382) was added to the medium in the basal reservoir for up to 8 h. At 1 h intervals after stimulation, crypts were fixed with Carnoy’s solution for 30 min at 4 °C, then incubated for 5 min with 50% ethanol, 3 min with 80% ethanol, and 2 min with 96% ethanol (v/v in water) at 20 °C. The fixed samples were then stained for EdU, Muc2, and Hoechst. The samples were cut with microdissection scissors, and the side view of crypts and secreted mucus layer was imaged.

Bi-layered structure of mucus layer visualized using bead penetration

The diffusion of particles into mucus depends on both their size and surface chemistry. To demonstrate the size exclusion effect only, two polystyrene-based fluorescent beads with different sizes (1 and 0.05 µm) but similar surface properties were used. A 100 μL mixture of 0.05 μm green beads, 3.64 × 10¹⁴ particles/mL (Polysciences, 17149), and 100 μL of 1 μm red beads, 1 × 10⁹ particles/mL (ThermoScientific, F13083) in Hanks balanced salt solution was added to the luminal reservoir (600 μL total) and incubated with Hoechst (0.02 μg/mL) for 2 h at 37 °C. Tissue was cut by microdissection scissors and imaged by confocal microscopy.

Spatial organization of mucins in 3D crypts

Fully polarized crypts cultured for 20 days under ALI were fixed for 20 min at 20 °C with glyoxal fixative and permeabilized with 200 μL of 0.5% Triton-X 100 in PBS for 20 min and washed 2× with 200 μL of 1% BSA in PBS. The spatial organization of mucins was assessed by adding three different fluorophore-labelled lectins: Lycopersicon esculentum lectin (LEL, Vector Lab, DL-1177), Lotus tetragonolobus lectin (LTL, Vector Lab, FL-1321), and wheat germ agglutinin (WGA, Invitrogen, W11261) lectin. The tissues were incubated in 10 μg/mL lectin at 4 °C for 24 h and extensively washed with PBS. The crypts were cut with microdissection scissors and imaged. Image acquisition by a laser scanning confocal microscope (Olympus FluoView FV3000, Waltham, MA) using a high magnification objective (20x) was used to acquire XY and XZ cross sections.

Statistical analysis

Data are presented as mean ± standard deviation. Differences between means from separate groups were determined using two-way ANOVA multiple comparisons by Tukey. The level of significance is indicated as the P value in each experiment. Unless otherwise specified, asterisks in figures indicate: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns (not significant), P value > 0.05. N = number of arrays, n = number of crypts indicated in figures. Confocal images were created by maximum intensity projection (MIP) of Z stack slices. Statistical analysis and graphical illustrations were performed using GraphPad PRISM 9 software, version 9.5.0. Schemes were designed using Adobe Illustrator software, version 25.2.1, 2021.