Research ArticlesImmunology

A sentinel goblet cell guards the colonic crypt by triggering Nlrp6-dependent Muc2 secretion

See allHide authors and affiliations

Science  24 Jun 2016:
Vol. 352, Issue 6293, pp. 1535-1542
DOI: 10.1126/science.aaf7419

Mounting the intestinal barricades

Gut microbiota are important for health and well-being, but they need to be kept under control and prevented from doing any harm. Birchenough et al. investigated the microbial molecules that trigger protective mucus secretion from a class of goblet cells in the colon. Once the molecules are detected, an alarm signal is transmitted from these cells via innate immune signal mediators and inflammasome components to adjacent cells, generating more mucus and repelling the invaders. Subsequently, the sentinel goblet cells are expelled from the epithelium and their remains may also add to the protective barricade.

Science, this issue p. 1535


Innate immune signaling pathways contribute to the protection of host tissue when bacterially challenged. Colonic goblet cells are responsible for generating the two mucus layers that physically separate the luminal microbiota from the host epithelium. Analysis of colonic tissues from multiple mouse strains allowed us to identify a “sentinel” goblet cell (senGC) localized to the colonic crypt entrance. This cell nonspecifically endocytoses and reacts to the TLR2/1, TLR4, and TLR5 ligands by activating the Nlrp6 inflammasome downstream of TLR- and MyD88-dependent Nox/Duox reactive oxygen species synthesis. This triggers calcium ion–dependent compound exocytosis of Muc2 mucin from the senGC and generates an intercellular gap junction signal; in turn, this signal induces Muc2 secretion from adjacent goblet cells in the upper crypt, which expels bacteria. Thus, senGCs guard and protect the colonic crypt from bacterial intruders that have penetrated the inner mucus layer.

The intestinal epithelium separates the luminal microbiota from the systemic tissues, but it is not the first line of defense. That function falls to the colonic mucus layers, which are composed of polymeric sheets of Muc2 mucin (1, 2). The colonic mucus has an inner layer formed of closely packed Muc2 sheets and a less organized outer layer. The inner layer is impenetrable to bacteria, whereas the outer layer serves as a habitat for the microbiota (1). Deletion of Muc2 in mice eliminates the mucus, causing bacterial colonization of the crypts, inflammation, and carcinogenesis (1, 3). The mucus layers are vital for preventing colitis, but they must function cooperatively with other elements of the epithelial innate immune system. These include TLR-MyD88 (Toll-like receptor–myeloid differentiation primary response gene 88) signaling and the Nlrp6 (NOD-like receptor family pyrin domain–containing 6) inflammasome; deletion of either TLR-MyD88 signaling or the Nlrp6 inflammasome renders the host susceptible to severe experimental colitis (47).

Mucus components, including Muc2, are produced by intestinal goblet cells (GCs). As GCs migrate up the colonic crypt, they manufacture, store, and release Muc2 by regulated secretion. The colonic surface GCs continuously secrete Muc2 and maintain the mucus layers (8). Colonic GCs express TLRs, MyD88, and the Nlrp6 inflammasome (9, 10). Thus, the crucial platforms for tissue preservation under challenge are expressed by the GCs. Endogenous factors (e.g., acetylcholine) trigger Muc2 secretion from colonic GCs (11, 12); however, it is unknown whether microbial TLR ligands—for example, lipopolysaccharide (LPS) or lipoteichoic acid (LTA)—might regulate mucin secretion.

Bacterial TLR ligands induce Muc2 secretion

We first investigated the capacity of bacterial TLR ligands to stimulate Muc2 secretion from murine colonic GCs (Fig. 1A). Using dynamic changes in explant mucus thickness as an indicator of secretion (12, 13), we observed that LTA, bacterial DNA, and the peptidoglycan subcomponents muramyl dipeptide (MDP) and γ-d-glutamyl-meso-diaminopimelic acid (iE-DAP) failed to elicit any secretory response. Conversely, LPS, its subcomponent lipid A, the synthetic triacylated lipopeptide P3CSK4, and flagellin all stimulated an increase in mucus thickness similar to that observed with the acetylcholine analog carbachol (CCh). LPS and P3CSK4 treatment of intact colonic tissue verified that the effect was not related to ex vivo muscle removal (fig. S1A). Ileal GCs are also responsive to CCh treatment (12), but we found them to be insensitive to LPS, P3CSK4, and flagellin (Fig. 1B).

Fig. 1 Select TLR ligands induce Muc2 secretion in distal colon.

In (A) to (G), intestinal explants were treated with TLR ligands or CCh. (A and B) Quantification of mucus growth from distal colonic (A) or ileal (B) explants. Unt., untreated; Flag., treated with flagellin. (C) Potential difference (PD) measurement: solid lines, mean; dashed lines, SEM. (D) Confocal micrographs of RedMUC298trTg tissue. Gray, actin; red, mCherry-MUC2. (E) Confocal z-stacks of tissue and 1-μm beads. Blue, tissue; red, beads; im, impenetrable mucus. (F) Impenetrable mucus thickness. (G) Concentration-response curves. Red dashed line, lipid A EC50; black dashed lines, estimated mucus and stool lipid A concentrations. (H) LAL reactivity of stool, mucus, and lipid A EC50 (0.85 μM). (I) Quantification of 16S rRNA genes in stool and mucus by quantitative polymerase chain reaction. Data are means ± SEM (n = 4 or 5); *P < 0.05, Dunnett test [(A), (B), (H)], Sidak test (F), or Mann-Whitney test (I). Scale bars, 50 μm.

Mucus quantified in the tissue explant system is dependent on Muc2 (i.e., completely absent in Muc2−/− tissue); alterations in mucus thickness are primarily reflective of Muc2 secretion. However, mucus thickness alterations may also be affected by volume expansion. Transepithelial potential difference recordings indicated transient ionic flux after treatment with LPS (Fig. 1C). A similar effect has been linked to cholinergic Muc2 secretion (12). Imaging of colonic explants from an mCherry–human MUC2–expressing transgenic mouse (RedMUC298trTg) revealed secreted Muc2 only in treated tissues (Fig. 1D). As mucus expands in volume, it becomes more penetrable to bacteria-sized beads. The thickness of mucus increased after LPS treatment, but no evidence of increased penetrability was observed (Fig. 1, E and F). Thus, bacterial TLR ligands induced Muc2 secretion.

Colonic GCs are likely routinely exposed to high concentrations of the TLR ligands that stimulate Muc2 secretion ex vivo. To examine the in vivo activity of TLR ligand–induced secretion, we measured the sensitivity of the response (Fig. 1G). Similar response curves were produced for lipid A, P3CSK4, and flagellin, with half-maximal effective concentration (EC50) values between 0.4 and 0.85 μM. To assess this sensitivity in the appropriate biological context, we estimated the concentration of soluble lipid A in two colonic luminal compartments, the stool and mucus, by Limulus amebocyte lysate (LAL) assay (Fig. 1H). LAL reactivity of stool was higher than that of mucus by a factor of 360, indicating a steep gradient in lipid A concentration between the mucus and the stool. This reflected the factor of 210 difference in bacterial 16S rRNA gene copy number between these compartments (Fig. 1I). LAL reactivity to 0.85 μM (EC50) lipid A was less than LAL reactivity to stool by a factor of 20 and was higher than LAL reactivity to mucus by a factor of 18 (Fig. 1H). Estimated in vivo lipid A concentrations were plotted on the lipid A response curve and closely mirrored the response window (Fig. 1G). These results show that the TLR ligand–triggered Muc2 secretion in the distal colon was inactive under normal conditions, as the mucus layer close to the colonic GCs did not harbor sufficient lipid A concentrations to trigger secretion.

TLR ligands induce Muc2 secretion in specific GCs

Intestinal GCs are more functionally heterogeneous than previously expected (8, 9). We therefore tested whether TLR ligands induced Muc2 secretion from a specific population of GCs. RedMUC298trTg colonic explants were treated with LPS, P3CSK4, flagellin, or CCh individually or in combination, and tissue sections were imaged (Fig. 2, A and B). Untreated tissue had large Muc2-filled GCs in the upper crypt and smaller GCs containing less Muc2 in the lower crypt. Upper-crypt GCs were emptied in tissues treated with TLR ligands. Conversely, CCh treatment resulted in secretion of Muc2 at the lower crypt without affecting the upper-crypt GCs. Combined treatment with LPS and CCh emptied both upper- and lower-crypt GCs. Mucus secretion was quantified from explants treated with LPS in combination with P3CSK4, flagellin, or CCh (Fig. 2C). Treatment with LPS combined with P3CSK4 or flagellin did not induce any additive secretory response, whereas combined LPS and CCh treatment did. Reversing the direction (apical versus serosal) of LPS and CCh treatment did not induce Muc2 secretion, indicating that responses were compartmentalized (Fig. 2D). Together, these observations indicated that LPS, P3CSK4, and flagellin all induced Muc2 secretion from the same upper-crypt GCs and that these cells were distinct from the GCs targeted by CCh.

Fig. 2 TLR ligand–responsive GCs are localized to the upper crypt.

Colonic explants were treated with TLR ligands or CCh. (A) Confocal micrographs of cryosections from RedMUC298trTg tissue. Red, MUC2; blue, DNA. (B) Upper crypt (yellow boxes) or lower crypt (green boxes) magnified from (A); a dashed gray line shows the epithelial surface. (C and D) Quantification of mucus growth rates. (E) Upper-crypt GCs (yellow arrowheads) and intercrypt GCs (green arrowheads) in RedMUC298trTg colon. Upper left: Confocal z-stack of tissue surface. Upper right: x/y-axis cross section. Lower left and right: x/z-axis cross sections. Blue, DNA; red, mCherry-MUC2; gray, actin. Data are means ± SEM (n = 3 or 4); *P < 0.05, Tukey multiple-comparisons test; ns, not significant. Scale bars, 50 μm (A), 20 μm [(B) and (E)].

The intercrypt GCs are another apically exposed GC population (8). These cells do not store large amounts of mature glycosylated Muc2 and are frequently not observed by conventional mucin staining. However, they can be identified in RedMUC298trTg tissues and are distinguishable from upper-crypt GCs (Fig. 2E). Comparison of untreated and LPS-treated tissue identified upper-crypt GCs secreting Muc2 in LPS-treated tissue but did not show any alterations to the intercrypt GCs. These results confirmed that only the upper-crypt colonic GCs were responsive to LPS.

The Muc2 secretory signaling pathway

Colonic GCs express TLRs that recognize LPS, P3CSK4, and flagellin (9) and also express the Nlrp6 inflammasome, which has been suggested to regulate GC Muc2 secretion (10). To examine the role of these mediators in controlling TLR ligand–stimulated secretion, we quantified mucus secretion in colonic explants from a range of knockout (KO) mice (Fig. 3A).

Fig. 3 TLR ligand–driven Muc2 secretion requires endocytosis, signaling, and ROS synthesis upstream of inflammasome activation.

Colonic explants [(A) to (C)] or cell suspensions [(D) and (E)] were treated with TLR ligands or CCh. (A) Quantification of mucus growth in wild-type (WT) or KO explants. (B) Quantification of mucus growth in explants pretreated with inhibitors; IP, inhibitory peptide. (C) Quantification of mucus growth in explants pretreated with ROS scavengers. (D) DCFDA fluorescence in epithelial cells pretreated with inhibitors. (E) Confocal micrographs of RedMUC298trTg epithelial cells with caspase inhibitory peptide. Purple arrowheads, nonendocytotic GCs; yellow arrowheads, endocytotic GCs; red, mCherry-MUC2; blue, DNA; gray, actin. Data are means ± SEM (n = 4 or 5); *P < 0.05, Dunnett multiple-comparisons test of WT versus KO (A) or no inhibitor versus inhibitor [(B) to (D)]. Scale bars, 50 μm.

P3CSK4, LPS, and flagellin are recognized by TLR2/1, TLR4, and TLR5, respectively, which activate signaling via proximal mediators such as MyD88 and TRIF. Loss of their cognate receptors and MyD88, but not TRIF, prevented both LPS- and P3CSK4-induced secretion (Fig. 3A). Tlr5−/− tissue was not examined; however, the response to flagellin was not lost in MyD88−/− or Trif−/− tissues, which indicates that neither TLR-MyD88 nor TLR-TRIF signaling was essential to flagellin-induced secretion. The response to all three TLR ligands was preserved in Rag1−/− tissue, ruling out a role for mucosal lymphocytes in mediating Muc2 secretion (fig. S1B). CCh responses were unaltered in the Tlr2−/−, Tlr4−/−, MyD88−/−, Trif−/−, and Rag1−/− tissues.

Examination of LPS-, P3CSK4-, and flagellin-induced mucus secretion in Nlrp-deficient tissues precluded the involvement of Nlrp3 and Nlrc4 but indicated that Nlrp6 signaling was essential for the response to all three TLR ligands (Fig. 3A). Similarly, no response was observed in Caspase-1/11−/− and Caspase-11−/− tissue. Caspase-1–mediated maturation of interleukin-1β (IL-1β) and IL-18 is the conventional purpose of activated inflammasomes; the response to all three TLR ligands was unaltered in IL1α/β−/− and IL18−/− tissue, indicating that Nlrp6 inflammasome signaling did not rely on IL-1β and IL-18.

It has been suggested that Nlrp6−/− and Caspase-1/11−/− colonic GCs lack normal Muc2 secretion and an intact inner mucus layer (10), which could account for an inability to secrete Muc2. However, all of the different inflammasome KO tissues responded normally to CCh stimulation, and all strains had an intact inner mucus layer when housed in our facility (fig. S2).

To dissect the cellular responses involved in TLR ligand–induced Muc2 secretion, we tested inhibitors targeting various cellular processes (Fig. 3B). Dynasore, which inhibits endocytosis, blocked the response to both LPS and P3CSK4 but not the response to flagellin. Identical results were obtained by using diphenyleneiodonium (DPI) to inhibit nicotinamide adenine dinucleotide phosphate/dual oxidase (Nox/Duox) reactive oxygen species (ROS) synthesis. Peptide-mediated inhibition of caspase-1 and caspase-11 activation mimicked results from caspase KO tissue, confirming the necessity for inflammasome activation. Atropine, a pan-muscarinic acetylcholine receptor inhibitor, blocked CCh-induced secretion but had no effect on tissue responses to TLR ligands.

ROS synthesis by Nox/Duox can be activated downstream of TLR signaling and can act as an upstream stimulator of inflammasome activation (14, 15). DPI is a potent Nox/Duox inhibitor but can also inhibit mitochondrial ROS release. We clarified this element of the pathway using N-acetylcysteine (NAC), a nonspecific ROS scavenger, and Mito-TEMPO, a ROS scavenger that accumulates in mitochondria (Fig. 3C). NAC prevented LPS-stimulated Muc2 secretion, whereas Mito-TEMPO had no effect; thus, Nox/Duox was the likely source of LPS-induced ROS synthesis. We used the fluorogenic ROS probe DCFDA (2′,7′-dichlorofluorescin diacetate) to quantify LPS-induced ROS production in isolated colonic epithelial cells (Fig. 3D). ROS production was observed when cells were treated with LPS, and was inhibited by Dynasore, a MyD88 inhibitory peptide, DPI, and NAC, but not by a caspase inhibitory peptide. These results indicate that LPS-induced ROS synthesis is dependent on endocytosis and MyD88 signaling but takes place upstream of inflammasome activation. Intriguingly, flagellin treatment also induced ROS synthesis, which could be inhibited by both the endocytosis inhibitor Dynasore and the MyD88 inhibitory peptide (Fig. 3D); this finding shows that flagellin could activate a pathway similar to that activated by LPS.

Cotreatment of RedMUC298trTg tissue with caspase inhibitory peptide and fluorescent LPS allowed imaging of endocytotic GCs without triggering Muc2 secretion (Fig. 3E). Intracellular LPS was observed in only a small population of GCs. Thus, the endocytosis-dependent ROS synthesis observed in DCFDA-loaded cells was likely generated by these GCs, which supports the idea that a distinct population of GCs was responsive to the TLR ligands.

Together, these results suggested that LPS, P3CSK4, and flagellin stimulate Muc2 secretion from responsive GCs via activation of the Nlrp6 inflammasome. LPS- and P3CSK4-induced inflammasome activation occurred downstream of endocytosis and TLR-MyD88–induced ROS synthesis. Although it was apparent that endocytosis of flagellin could also induce TLR-MyD88–dependent ROS synthesis, it was also evident that flagellin-induced Muc2 secretion was not contingent on this element of the pathway, and we therefore focused our subsequent studies on the response to LPS and P3CSK4.

TLR ligands are endocytosed by senGCs

To identify LPS- and P3CSK4-sensitive GCs in intact colonic tissue, we treated RedMUC298trTg colonic explants with fluorescent LPS or P3CSK4 and imaged the tissue by confocal microscopy (Fig. 4, A to D). GCs that had taken up LPS or P3CSK4 were found exclusively in the apical regions of the crypt. Intriguingly, there appeared to be very few of these cells, with less than one cell per crypt (Fig. 4E). These cells were not observed in tissues pretreated with Dynasore, confirming the requirement for endocytosis. Inhibition of inflammasome activation significantly increased the number of endocytotic cells detected, as did loss of TLR or MyD88 signaling (Fig. 4, B and E). This result suggested that LPS- and P3CSK4-triggered TLR signaling and inflammasome activation resulted in loss of detectable endocytotic cells.

Fig. 4 TLR ligands are endocytosed by sentinel GCs (senGCs).

In (A) to (D), colonic explants were treated as indicated, whole-mounted, and visualized by confocal microscopy; in (A) to (C), x/y-axis cross sections are at top; x/z-axis cross sections are in yellow boxes; blue, DNA; gray, actin. (A) RedMUC298trTg tissue; regions of x/z-axis cross sections magnified at bottom are indicated by white boxes. Green, LPS/P3CSK4; red, mCherry-MUC2. (B) WT with or without caspase inhibitory peptide, Tlr2−/−, and MyD88−/− tissue. Green, P3CSK4. (C) RedMUC298trTg tissue. Yellow arrowheads, endocytotic GCs; white arrowheads, nonendocytotic GCs; red, mCherry-MUC2; green, dextran. (D) WT colon cotreated with P3CSK4 and dextran. Yellow arrowheads, endocytotic GCs. (E) Quantification of endocytotic cells. Data are means ± SEM (n = 3 or 4); *P < 0.05, Tukey multiple-comparisons test. Scale bars, 20 μm.

We investigated this possibility by treating explants with nonimmunogenic dextran. Dextran was also taken up by a small number of GCs, but without inducing the same loss of endocytotic cells seen with LPS and P3CSK4 treatment (Fig. 4, C and E). GCs that endocytosed dextran were identified as the same cells that endocytosed P3CSK4 (Fig. 4D), and examination of sections obtained from dextran-treated tissue demonstrated that endocytosis was entirely restricted to GCs in the upper-crypt region (fig. S3A). To ensure that GC endocytosis observed ex vivo reflected processes occurring in vivo, we administered dextran intrarectally to live RedMUC298trTg mice. Uptake of dextran was again observed in a limited number of upper-crypt GCs, confirming ex vivo observations (fig. S3B).

Thus, luminal material was nonspecifically endocytosed by a small subpopulation of upper-crypt GCs. These cells were lost after TLR-MyD88 signaling and activation of the inflammasome. Crucially, only one or two such cells were typically detected per crypt. Because this represents only a fraction of the GCs secreted in response to TLR ligands, this finding suggested that activation of endocytotic GCs must control secretion from other responsive, but not endocytotic, upper-crypt GCs. The sampling and recognition of nonendogenous material is vital to the surveillance function of many types of host sentinel cell. Because this process was apparent in the endocytotic GCs, this cell was named sentinel GC (senGC) to distinguish it from other upper-crypt GCs.

Activated senGCs control Muc2 secretion

We examined the fate of senGCs lost after treatment and identified shed cells containing LPS or P3CSK4 and residual Muc2, which suggests that these were expelled senGCs (Fig. 5A). These were not present in tissue treated with dextran, indicating that expulsion was dependent on TLR engagement. Expelled cells were associated with diffuse extracellular Muc2 that was continuous with the remaining intracellular material (Fig. 5B). This implied that senGCs had undergone compound exocytosis (16) as well as expulsion. Intrarectal administration of P3CSK4 to live wild-type and RedMUC298trTg mice demonstrated that epithelial expulsion of GCs that had endocytosed P3CSK4 could also be observed in vivo (fig. S4, A and B). Intrarectal treatment of Nlrp6−/− mice identified cells that had endocytosed P3CSK4 but did not induce epithelial expulsion (fig. S4C). This confirmed that epithelial expulsion of senGCs was driven by Nlrp6 inflammasome activation.

Fig. 5 Activated senGCs are expelled from the epithelium and trigger gap junction– and Ca2+-dependent Muc2 secretion from responsive upper-crypt GCs.

Colonic explants were treated with TLR ligands or CCh. (A and B) RedMUC298trTg tissue whole mounts visualized by confocal microscopy. Blue, DNA; gray, actin; red, mCherry-MUC2; green, LPS, P3CSK4, or dextran. (A) x/z-axis cross sections showing expelled senGCs (yellow arrowheads). (B) Expelled senGC overview (left) and isosurface enhanced view (right, yellow box). (C) Localization of inflammasome activity after LPS treatment of WT, KO, and RedMUC298trTg tissue. Green, caspase-1/caspase-11 fluorogenic probe (Casp1/11 FP); blue, DNA; yellow arrowheads, senGCs. (D to F) RedMUC298trTg tissue imaged by confocal microscopy. (D) x/y-axis view (top) and x/z-axis view (bottom) of a crypt at different time points after LPS treatment. Green arrowheads and green dashed lines, senGC; yellow arrowheads, responsive GC; blue arrowheads, nonresponsive GC. (E) Quantification of mCherry-MUC2 fluorescence of individual GCs from crypt in (D). (F) Fluorescent isosurfaces used to generate data in (E); responsive GC numbering corresponds to numbered plots in (E). Red arrow shows sequence of responsive cell secretion. White, DNA; green, senGC; orange, responsive GCs; blue, nonresponsive GCs. (G and H) Quantification of mucus growth with or without CBX, TTx, or Ca2+ signaling inhibitors. (I and J) RedMUC298trTg tissue loaded with fluorogenic Ca2+ indicator (Fluo4) and imaged by confocal microscopy. Green, Fluo4; red, mCherry-MUC2. (I) x/y-axis view of tissue before and after treatment. (J) x/y-axis view of a crypt at different time points after LPS treatment. White arrow shows sequence of GC secretion. White dashed line, crypt opening; green arrowheads, senGC; yellow arrowheads, responsive GCs. Data are means ± SEM (n = 4 or 5); *P < 0.05, Dunnett multiple-comparisons test. Scale bars, 20 μm.

Inflammasome activity was assessed using a fluorogenic peptide probe targeting active caspase-1 and caspase-11 (Casp1/11 FP). Tissue was loaded with Casp1/11 FP and then treated with LPS (Fig. 5C). Inflammasome activity was exclusively localized to cells that had been expelled from the upper crypt. Imaging of RedMUC298trTg tissue demonstrated that these cells were senGCs. Expelled senGCs were associated with long plumes of secreted Muc2, which confirmed that LPS treatment had triggered compound exocytosis of Muc2. No inflammasome activity or cellular expulsion was observed after LPS treatment of Caspase-1/11−/− or Nlrp6−/− tissue, but both were detected in treated Nlrp3−/− tissue. This demonstrated that activity observed in these cells was specifically dependent on a functional Nlrp6 inflammasome.

senGC-mediated secretion was studied by live imaging of RedMUC298trTg colonic explants. This revealed loss of Muc2 from upper-crypt GCs in LPS- or P3CSK4-treated tissues but not in vehicle-treated tissues (movies S1 and S2). Rapid Muc2 degranulation and epithelial expulsion were observed in a limited number of GCs per crypt (Fig. 5D and movies S3 and S4). Expulsion was used to designate the senGC, and the remaining upper-crypt GCs could be classified as “responsive” or “nonresponsive” according to whether they secreted Muc2 (Fig. 5D). Muc2 fluorescence was subsequently quantified from different classes of GC and tracked over time (Fig. 5, E and F, and movies S5 and S6). Strikingly, this revealed that LPS- and P3CSK4-induced Muc2 secretion was sequential in nature. Degranulation occurred initially in senGCs and then in responsive GCs. Responsive GC secretion occurred in sequence, starting with cells closest to the senGC and progressing around the crypt (Fig. 5F). This indicates that activated senGCs triggered an intercellular signal that induced Muc2 secretion in adjacent responsive GCs.

Intercellular senGC signaling

Localized intercellular signals can use paracrine or juxtacrine mechanisms. The mammalian intestine is highly innervated, and paracrine signals can be generated by the enteric nervous system (ENS). Juxtacrine signals can be transmitted directly via intercellular cytoplasmic bridges formed by gap junction (GJ) connexons. We measured mucus secretion in colonic explants treated with inhibitors targeting each process. Tetrodotoxin (TTx) blocks ENS signaling, and carbenoxolone (CBX) can reduce GJ conductivity. Neither inhibitor affected CCh-induced secretion; however, although TLR ligand–induced secretion was unaffected by TTx, the response was blocked by CBX (Fig. 5G). To verify GJ decoupling, we used a FRAP (fluorescence recovery after photobleaching) assay to measure GJ conductivity. Vehicle- or CBX-treated explants were loaded with a GJ-permeable dye, and areas of epithelium were photobleached (fig. S5, A and B). Fluorescent recovery was significantly reduced in CBX relative to vehicle-treated explants, indicating that GJ conductivity had been inhibited and supporting a role for GJ signaling in controlling responsive GCs.

A recognized form of GJ signaling is mediated by calcium. Increased cytosolic Ca2+ is derived from extracellular influx or release from endoplasmic reticulum stores via inositol 1,4,5-trisphosphate receptor (IP3R) or ryanodine receptor (RyR) channels. To investigate the role of Ca2+ signaling in TLR ligand–induced secretion, we pretreated explants with inhibitors before treatment with LPS or CCh (Fig. 5H). The cell-permeable Ca2+ chelator BAPTA-AM blocked secretion induced by either molecule, demonstrating the requirement for Ca2+ release in both types of secretion. The cell-impermeable Ca2+ chelator EGTA inhibited only the response to LPS. Intracellular Ca2+-store release was targeted by the RyR inhibitor ryanodine and the IP3R inhibitor xestospongin C (XeC). Inverse results were obtained for LPS and CCh, with LPS dependent on IP3R-mediated Ca2+ release and CCh dependent on RyR-mediated Ca2+ release. The requirement for Ca2+ influx and intracellular store release indicated that Ca2+-induced Ca2+ release was active in LPS-induced secretion. To probe the sequence of Ca2+ signaling events, we used the Ca2+ ionophore ionomycin (fig. S5C). Ionomycin transports Ca2+ across membranes mimicking the Ca2+ influx required for LPS-induced secretion. Ionomycin induced Muc2 secretion and was insensitive to inhibition of ROS synthesis and inflammasome activation, but was inhibited by EGTA and XeC. These findings suggest that Ca2+ influx has a role upstream of Ca2+ store release and that direct induction of Ca2+ signaling negates the requirement for inflammasome activation.

We used the fluorogenic Ca2+ indicator Fluo4 to detect Ca2+ signaling in LPS-treated RedMUC298trTg colonic explants (Fig. 5, I and J). Fluo4-positive GCs were observed after LPS treatment but not vehicle treatment, indicating that LPS-induced Ca2+ signaling could be visualized ex vivo. senGCs appeared to have higher Fluo4 signal than the responsive GCs where the signal was relatively weak (Fig. 5J and movie S7). Sequential increases in cytoplasmic Ca2+ starting at the senGC appeared to follow the same pattern as observed for Muc2 secretion.

SenGCs flush bacteria away from crypt openings

Induced Muc2 secretion may remove bacteria from the crypt opening, and we tested this in our ex vivo tissue system. The inner mucus layer normally separates bacteria from the colonic tissue surface; therefore, this was first mechanically removed. Fluorescent bacteria were then applied to the tissue surface. Muc2 secretion was triggered by treatment with LPS, and images of tissue and bacteria were acquired and bacterial spatial distribution quantified (Fig. 6, A and B). Initially bacteria were identified at the tissue surface and close to the crypt openings. Treatment with LPS, but not vehicle, caused bacteria to be displaced from the crypt openings. Most bacteria remaining at the tissue surface after LPS treatment were in the intercrypt regions, thus supporting the notion that this mechanism functions to specifically protect the crypts in vivo.

Fig. 6 senGC activation removes bacteria from crypt openings ex vivo and is triggered by disruption of the inner colonic mucus layer in vivo.

(A) Fluorescent bacteria (red) applied to colonic explant tissue (green) and imaged before (top) and after (bottom) LPS or vehicle treatment. Left, x/z-axis view; right, x/y-axis view. Yellow arrowheads and dashed circles denote crypt openings. (B) Distribution of bacteria in relation to crypt openings as in (A). Data are representative of four independent experiments. In (C) to (F), mice were given 3% DSS in drinking water; samples were collected after 0 (no DSS), 12, 36, and 84 hours. (C) Confocal z-stacks of colonic explants and 1-μm beads. Blue, tissue; red, beads; im, impenetrable mucus. (D) Quantification of cells with inflammasome activity (Casp1/11 FP+) detected in live tissue. (E) Casp1/11 FP+ cells (pink arrows) with GC morphology in the upper crypt (left) and shed from tissue (right). Top, x/y-axis view; bottom, x/z-axis view. (F) Colonic explants were treated with fluorescent dextran, fixed, whole-mounted, and visualized by confocal microscopy. Top, x/y-axis view; bottom, x/z-axis view. Blue, DNA; green, dextran; gray, actin. Data are means ± SEM (n = 4); *P < 0.05, Dunn multiple-comparisons test. Scale bars, 20 μm.

Oral administration of dextran sodium sulfate (DSS) allows colonic bacteria to penetrate the inner mucus layer and contact the epithelium prior to the onset of inflammation (17). We hypothesized that if senGCs protect the crypt from bacteria, DSS treatment should trigger senGC activation in vivo. Short-term DSS treatment (84 hours) did not cause weight alterations (fig. S6A). However, the inner mucus layer became partially penetrable to bacteria-sized beads after 12 hours of DSS treatment and was largely disrupted after 84 hours (Fig. 6C). The Casp1/11 FP probe was used to quantify epithelial inflammasome activation (Fig. 6D and fig. S6B). This was detected in multiple cells after 12 hours of DSS treatment, but the number of these cells then declined. Inflammasome activity was visualized in both upper-crypt and shed cells with GC-like morphology (Fig. 6E), indicating that DSS had resulted in activation and epithelial expulsion of senGCs. That DSS treatment caused depletion of senGCs was confirmed by the loss of dextran endocytosis by upper-crypt cells after 36 hours (Fig. 6F). Endocytotic cells were clearly identified in untreated tissue but largely absent in tissue exposed to DSS, indicating that the senGCs had been activated and depleted.

Normally, colonocytes are separated from the microbiota by the inner mucus layer, and disruption of this process allows bacteria to approach the crypts (1, 1820) (fig. S7, A and B). Influx of bacteria increases the concentration of TLR ligands near the crypt and activates senGCs (fig. S7C). Our findings suggest that this sequence is initiated by TLR-MyD88 signaling downstream of endocytosis, preceding Nox/Duox ROS synthesis and Nlrp6 inflammasome activation. This causes Ca2+ signaling that drives compound secretion of Muc2, generation of GJ-dependent intercellular signaling, Muc2 secretion from responsive GCs, and the expulsion of the activated senGC. These events clear bacteria from the crypt opening (fig. S7D), thereby protecting the lower crypt and intestinal stem cells from bacterial intrusion. Mouse strains unable to trigger senGC activation had a functional inner mucus layer, which suggests that senGCs are less important for normal mucus layer generation.

The pathway activating senGCs is supported by other observations. Colonic GCs express the expected TLR repertoire and can take up luminal material (9, 21, 22). MyD88 signaling can induce Nox/Duox ROS synthesis (23) that has been coupled to inflammasome activation (14, 15) and was recently linked to GC endocytosis and autophagy (24). The senGC Nlrp6 inflammasome activation resulted in expulsion of these cells; this mechanism could be important for clearing infected cells (25, 26) and blocking the crypt.

Administration of DSS to mice is a commonly used colitis model. DSS disrupts the inner mucus layer, allowing bacteria to access the epithelial surface before the onset of inflammation (17). Increased sensitivity to DSS treatment in animals lacking senGC activation mediators may be related to loss of senGC protection (4, 5, 27). DSS triggered inflammasome activity localized to upper-crypt cells and led to depletion of senGCs (Fig. 6, C to F). Ex vivo experiments confirmed that senGC-mediated mucus secretion displaced bacteria from crypt openings (Fig. 6, A and B), and senGC activation after inner mucus layer disruption likely generates a similar response. Depletion of senGCs by repeated challenge would leave the crypt without defense—an event that may be important in understanding the development of chronic colitis.


Materials and Methods

Figs. S1 to S7

Movies S1 to S7

Reference (28)


  1. See supplementary materials on Science Online.
Acknowledgments: Supported by the Swedish Research Council, Swedish Cancer Foundation, Knut and Alice Wallenberg Foundation, Lundberg Foundation, Sahlgren’s University Hospital (ALF), Torsten Söderbergs Stiftelse, National Institute of Allergy and Infectious Diseases grant U01AI095473, and Swedish Foundation for Strategic Research. We thank the Gothenburg Centre for Cellular Imaging for technical help, F. Svensson for generating the RedMUC298trTg mice, and W.-D. Hardt and the Mucosal Immunobiology and Vaccine Center at the University of Gothenburg for mouse strains.

Stay Connected to Science

Navigate This Article