Disruption of the Cox-1 gene slows repair of microscopic lesions in the mouse gastric epithelium

Olga T. Starodub, Elise S. Demitrack, Heidi K. Baumgartner, Marshall H. Montrose


Cyclooxygenase-1 (Cox-1) contributes to gastric defense of healthy tissue, but the role in the protection of the gastric epithelium after minor, acute damage has been difficult to study in vivo. Using 710-nm two-photon light absorption to create microscopic gastric damage in anesthetized mice with the gastric mucosal surface surgically exposed and perfused on the microscope stage, the acute response of surface cells to injury could be monitored using in vivo microscopy within seconds after injury. Using exogenous (Cl-NERF) and endogenous fluorophores, extracellular pH and cell death were monitored in real time during the entire damage and repair cycle. Two-photon damage was initiated by scanning ∼200 μm2 of gastric surface cells with high laser intensity, causing rapid bleaching of NAD(P)H fluorescence in optically targeted cells. In both Cox-1+/− and Cox-1−/− mice, a similar initial damage area expanded to include bystander epithelial cells over the next 2–5 min, with larger maximal damage noted in Cox-1−/− mice. The maximal damage size seen in Cox-1−/− mice could be reduced by exogenous dimethyl-PGE2. All damaged cells exfoliated, and the underlying epithelium was coincidently repaired over a time interval that was briefer in Cox-1+/− (12 ± 2 min, n = 12) than in Cox-1−/− (24 ± 4 min, n = 14) mice. Directly after damage, pH increased transiently in the juxtamucosal layer (maximal at 3–6 min). A smaller peak pH change was noted in Cox-1−/− mice (ΔpH = 0.3 ± 0.04) than in Cox-1+/− mice (ΔpH = 0.6 ± 0.2). Recovery to normal surface pH took longer in Cox-1−/− mice (27 ± 5 min) than in Cox-1+/− mice (12 ± 1 min). In conclusion, constitutive loss of Cox-1 leaves the gastric mucosa more prone to damage and slowed repair of microlesions.

  • autofluorescence
  • photodamage
  • laser scanning microscopy
  • two-photon microscopy
  • reduced nicotinamide adenine dinucleotide
  • surface pH
  • cyclooxygenase

the stomach is poised to protect itself from a wide variety of intrinsic and exogenous stress factors, aggressors, or irritants. The healthy gastric mucosa is sustained through a multilayered defense that includes regulation of the juxtamucosal environment via mucus and bicarbonate secretion, the epithelial barrier, and subepithelial blood flow and nerves with defensive mediators (1).

The most imperiled component of the gastric mucosa is the layer of epithelial cells directly exposed to the gastric lumen. Injury and repair of this epithelium have been extensively studied in vivo and in vitro (9, 10, 12, 1618, 25, 26). Animal models are favored for studying gastric repair processes as they can provide responses of physiological significance for understanding similar events in human ulcer disease (26). Gastric injury responses have been studied in mammalian and amphibian (frog) models (11, 15, 2729). Although these studies of acute damage have helped to resolve early response mechanisms affecting lesion development and repair, they often required the creation of macroscopic and/or pervasive damage to quickly identify regions of injury. Thus, acute damage is often imposed by harsh treatments like pure ethanol, hypertonic saline solution, boiling water, cold probe, or ischemia-reperfusion (15, 17, 18, 29, 33). Conversely, the response to more physiologically relevant or modest stressors [e.g., acetic acid, nonsteroidal anti-inflammatory drugs (NSAIDs), Helicobacter pylori, or stress-inducing procedures like cold restraint] requires longer times to develop small sporadic lesions in which repair processes are difficult to study (19, 27, 28, 34, 37). In models of severe damage, the creation of transmural wounds cross all lines of defense, inflammatory processes are activated, and repair is slow (28). However, superficial injury of the epithelium has been shown to recover more quickly (12, 1618). Despite the availability of numerous gastric damage models, the field lacks a model in which the early events of damage and repair can be tracked in isolated sites of microscopic, superficial damage.

Much of our knowledge about gastric repair after superficial injury comes from a particularly detailed microscopic examination of the acute response to brief exposure to absolute ethanol (12, 1618). After virtually the entire surface epithelium had been selectively damaged with this insult, tissue was fixed and examined 30 s to 60 min later. It was observed that the necrotic epithelium rapidly detached, forming a “mucoid cap” over the epithelium composed of mucus, cell debris, plasma, and fibrin. Surviving cells in the gastric pits migrated out to the surface to restitute the epithelium (12). Although the restitution process started within 3 min, it was concluded that epithelial repair could start only after the necrotic cells detached and lifted off the basal lamina (18). Since a higher pH was observed under the mucoid cap (7, 36), it was proposed that one function of the cap was to trap secreted bicarbonate to allow a favorable environment for restitution. Expansion of these findings also included the observation that the exogenous addition of PGE2 reduced the depth of ethanol-induced damage (17).

Prostaglandins confer gastric cytoprotection by preventing cellular necrosis produced by strong and mild irritants (including NSAIDs such as indomethacin) through a mechanism distinct from inhibition of acid secretion (23). In the gastric mucosa, most prostaglandins are generated via the cyclooxygenase-1 (Cox-1) isoform, and only a small part are Cox-2 derived (8, 19, 21, 3134, 37). Prostaglandins produced by Cox-1 are known to regulate both gastric surface pH and blood flow (3, 4, 31, 34).

Recent developments in in vivo imaging of the gastric mucosa (35) have led to the development of a new model of gastric damage that makes possible selective injuries to the surface epithelium of the corpus, real-time observations of expansion and restitution of microscopic lesions, and the ability to simultaneously measure extracellular pH at the gastric surface. Two-photon microscopy imaging is performed using low light levels that are nondestructive to both the fluorophore and cell (38, 39). However, brief exposure to focused higher light intensity can lead to destruction of the light-absorbing molecule (photobleaching), which results in the release of toxic side products that induce cell damage and death. This is the mechanism by which photodynamic therapy can kill cancer cells that have selectively absorbed photosensitizing chemicals (2, 6, 22). Recently, the use of two-photon absorption has been applied to cause focal cell damage (13, 20, 22, 35). Endogenous cytosolic fluorophores, like NAD(P)H, can be visualized by both conventional and two-photon fluorescence imaging, providing insights into cell and tissue bioenergetics (22, 24, 38, 40). Using NAD(P)H as the optical target of laser light to generate a mouse model of superficial gastric damage, we found that resolution of microscopic damage proceeds by a different sequence of events than previously reported and requires, at least in part, Cox-1 activity.



Cox-1−/− and Cox-1+/− mice (B6;129P mixed background) were purchased from Taconic (Germantown, NY) and bred as female heterozygotes (Cox-1+/−) mated with male heterozygotes or male homozygote nulls (Cox-1−/−). Animals were toe and tail clipped for genotyping before 14 days and used for experiments at ∼3 mo of age. All animals were fed a standard rodent diet and had free access to water. The surgical preparation for these experiments has been previously described (3, 4). Briefly, the stomachs of anesthetized mice (Inactin, Sigma, St. Louis, MO) were exteriorized and everted to expose the gastric mucosa. The mouse was placed supine on the stage of an inverted confocal/two-photon microscope (Zeiss LSM 510 NLO) such that a portion of the exposed mucosa protruded into a perfusion chamber that was warmed to keep the animal's body temperature at 37°C. All experimental procedures were approved by the Animal Care and Use Committee of the University of Cincinnati.

Live Tissue Microscopy

Time course experiments were performed to image the gastric mucosa of the corpus region of the stomach using a Zeiss C-Apo ×40 objective. At each time point, a set of four images was collected as described below, with time points spaced by variable intervals of 30 s to 3 min. The mucosal surface was superfused (0.2 ml/min, KD Scientific push/pull syringe pump) with a pH 3 slightly buffered solution containing 150 mM NaCl, 4 mM Homopipes (Research Organics, Cleveland, OH), and 10 μM Cl-NERF as the pH indicator (Invitrogen) to report extracellular pH at the gastric surface, as previously described (35). Cl-NERF was alternately excited at 514 and 458 nm, and confocal fluorescence emission (>530 nm) images were obtained; 514-to-458-nm ratio images were calibrated to pH as previously described (3, 4). To image autofluorescence [putative NAD(P)H], tissue was excited at 710 nm with a femtosecond-pulsed titanium sapphire (Ti-Sa) laser for two-photon excitation (Mai Tai, Spectra-Physics) using 60- to 70-mW light input to the scan head. NAD(P)H fluorescence was collected at 390–485 nm. Confocal images of 710-nm reflectance were collected simultaneously to landmark cellular structures (5).

Two-Photon Microlesions

A small rectangular region of the gastric surface (a 150- to 200-μm2 area placed initially over 3–5 cells) was repetitively scanned with 300–340 mW of 710-nm light into the scan head for 5–10 s. The light dose at the surface was calculated as milliWatts per second per micrometer squared, knowing the power of the illuminating light (300–340 mW), the size of an individual pixel (0.102 μm2/pixel, 0.32-μm × 0.32-μm pixel size), and how long each pixel was illuminated to induce photodamage (128–192 μs total when 1.28-μs laser dwell time/pixel was multiplied by 100–150 scan iterations over each individual pixel) (6).

The complete damage-repair time course was followed for an individual lesion, and the process was then repeated one to eight times per animal using separated lesions. All experimental outcomes were evaluated in 8–10 animals, and there was no trend in the outcomes that suggested that tissue responses were different in initial versus later lesions in the same animal (data not shown).

Cox-1−/− Rescue With Dimethyl-PGE2

To ask if exogenous PGE2 rescued the Cox-1−/− phenotype, tissue was first measured during a damage-repair cycle according to our routine protocol, and 0.1 mg/kg 16,16-dimethyl-PGE2 (dmPGE2; Sigma) was then added topically to the surgically exposed gastric mucosal surface for 5 min. Tissue superfusion was then resumed with solution lacking dmPGE2, and a second round of damage and repair was measured in a region of the corpus distant from the site of the first damage. dmPGE2 was solubilized in 95% ethanol and then diluted to a working concentration in saline (final ethanol: <2%).

Image Analysis

All image analysis was performed using Metamorph software (Universal Imaging, West Chester, PA). Using ratio images of Cl-NERF fluorescence acquired during the time course of damage and repair, extracellular pH was measured in a square region (16–27 μm2) adjacent to the site of damage and routinely within 20-μm distance from the gastric surface. The distance from surface varied modestly (up to 50 μm when tissue was very mobile) during measurements between images due to the gradual expansion of the wound and detachment of injured cells.

To measure the extent of damage, the area of surface epithelial cells that lost NAD(P)H autofluorescence was measured over time. Since loss of NAD(P)H resulted in a dark region that could be devoid of any material, simultaneously collected confocal reflectance images were used to identify regions retaining cellular material. A binary (inverted) mask of autofluorescence was used to identify regions with viable cells and was compared pixel by pixel with the reflectance image. Multiplying these two images on a pixel-by-pixel basis allowed the definition of damaged cells as the area lacking NAD(P)H but still attached and continuous with the epithelial layer. A second independent assay of cell viability was also used to quantify the damage area. Damaged cells became permeable toward the Cl-NERF dye in the perfusate and could be quantified by measuring the area of intracellular Cl-NERF fluorescence.


Data are presented as means ± SE, and statistical comparisons were made with an unpaired two-tailed Student's t-test, with a P < 0.05 significance.


Our first goal was to develop two-photon microscopy as a tool to selectively image or photodamage the gastric epithelium of anesthetized mice. Using a surgically exposed gastric mucosa as previously described (35), we first evaluated the effects of low versus high laser power on the gastric mucosa.

Imaging Cellular Autofluorescence

Using 710-nm excitation, two-photon fluorescence was observed at 390- to 485-nm emission. Similar to other laboratories who have imaged NAD(P)H (peak 465-nm emission), fluorescence was observed in the cytosol (Fig. 1A) (13, 22, 24). Our working hypothesis is that the fluorophore we observe is NAD(P)H, but for current purposes the important observation is of a cytoplasmic fluorescent molecule that can act as a substrate for two-photon light absorption.

Fig. 1.

Time course of damage and repair of microlesions in cyclooxygenase-1 (Cox-1)+/− (A–H) and Cox-1−/− (I–P) gastric epithelium. Representative combined two-photon autofluorescence (green) and confocal reflectance (red) images were collected over time, with time 0 indicating the time of photodamage. A and I: intact epithelium, with the arrow showing the position of the impending high light dose. B and J: initial damage. C and K: the damage expands, and injured cells lose autofluorescence but are still present (red cytoplasm). D and L: damage area maximum. E–H and M–P: formation of the mucoid cap within the mucus layer (double arrow) and exfoliation of cells leaving the epithelium intact (green cytoplasm, arrow).

Microlesions Created With Two-Photon Absorption

For the two-photon laser to damage the gastric epithelium, a small region (150–200 μm2) of the epithelial surface was exposed to high laser power and repetitive scanning cycles (see methods). This increased the light dose ∼600-fold compared with settings used for autofluorescence imaging. Completing this high light exposure required 5–10 s and caused a prompt decrease in cytoplasmic autofluorescence in the targeted region (see Supplemental Fig. 1 and Fig. 1).1 In nine experiments, within 5–8 s after repetitive scanning, we observed that autofluorescence in the targeted region decreased to 38 ± 8% of the starting autofluorescence. In contrast, adjacent areas in the same microscope field that were subjected to only low power laser scanning had 103 ± 1% autofluorescence over the same time course and 87 ± 7% of the starting autofluorescence even after 1 h of imaging. In the adjacent areas, there was no evidence of photodamage to tissue. The loss of autofluorescence in the targeted region of the gastric surface was sustained in the targeted region (Fig. 1D). The decreased fluorescence preceded a notable deterioration of cell morphology. When the focal plane was favorable for observation, intracellular changes of morphology were noticed (data not shown) that resembled necrosis (the cell polygonal contour became oval while loosening contact with neighboring cells, vacuolation of the cytoplasm). In some cases, cells in targeted regions also showed spherical cavitation lesions (likely caused by plasma formation from high laser exposure) and increased fluorescence of cell membranes (likely from oxidized membrane lipids from high laser exposure).

Injury and Restitution of Surface Epithelium

As shown in Fig. 1, the progression of injury could be qualitatively monitored on combined reflectance (red) and autofluorescence (green) images, and the same progression of events occurred for both Cox-1+/− (Fig. 1, A–H) and Cox-1−/− (Fig. 1, I–P) mice. Healthy surface epithelial cells are evident as a yellow overlay, indicating an autofluorescent and reflective cytoplasm (arrow, Fig. 1, A and I). Nuclei are not autofluorescent and were therefore visible in reflectance images only. Immediately after high power laser scanning, the targeted cytoplasm loses autofluorescence but is still visible in reflectance (arrow, Fig. 1, B and J).

Interestingly, we observed that the wound expanded beyond the region directly exposed to light. Over time, adjacent cells lose cytoplasmic autofluorescence and were visible in reflectance only (Fig. 1, C, D, K, and L). Associated with changing volume and shape of deteriorating cells, distances between cells changed as stretching, shrinking, and detaching occurred. Injured cells lacking autofluorescence gradually shed and overlayed the surface, forming a mucoid cap that appeared only in reflectance (double arrow, Fig. 1, E and M).

In contrast to observations after damage to the entire surface epithelium (18), at the time when injured cell detachment completed, the underlying epithelium was left apparently intact with no evident discontinuity in the cell layer by either reflectance or autofluorescence measures (Fig. 1, F–H and N–P). These initial qualitative observations were evaluated quantitatively, as described below.

Measuring the Size of Damage

As described in methods, we simultaneously collect both autofluorescence and confocal reflectance images of the gastric mucosa at each time point. The autofluorescence image includes all cell types that contained the responsive fluorophore (viable cells), as well as dark regions that contained dead or dying cells, nuclei, and noncellular regions (Fig. 2, A and A ′). The reflectance image simply reports all structures (cells, mucus, and debris) that reflected 710-nm light (Fig. 2B). An overlay of autofluorescence and reflectance images illustrating the different elements reported in these two modalities is shown in Fig. 2C. Observations made in Fig. 2 suggest that dead cells can be observed as regions lacking autofluorescence but still being reflective. Regions thus defined can include more than unhealthy cells (e.g., it also includes nuclei and mucus/debris), so identifying dead cells by this protocol also required subjective analysis of images to identify regions that were within the epithelial layer adjacent to the original site of photodamage. In this way, measurement of noncellular (but reflective) areas was excluded. To perform this analysis, a binary mask created from the autofluorescence image (to identify regions either with or without fluorescence; Fig. 2D) was multiplied by the reflectance image (Fig. 2E). The size of damage was measured on the resultant image by drawing a region of interest around the reflective and nonautofluorescent area adjacent to the optically targeted region. The same region was independently verified on the original autofluorescence images and also sporadically on combined reflectance and autofluorescence images.

Fig. 2.

Strategy for measurement of the damaged area. A representative set of images from a single time point are shown. A and A′: reflectance images containing live, dead, and dying cells. B: the autofluorescence image includes nonfluorescent regions (gland lumens/pits, dead and dying cells, and nuclei). C: the merged image includes all elements (red, reflectance; green, autofluorescence; and yellow, overlap). For the quantification of the damage area, a binary mask of autofluorescence (D) was combined by multiplication with the reflectance image, resulting in a masked image (E) that marked the limits of the damaged area. Arrows indicate areas of photodamage, and the yellow outline highlights the identified region of damaged tissue.

Results from the quantitative analysis of damaged area confirmed that the three to five cells initially targeted by the high laser power were the first to lose NAD(P)H autofluorescence. This is shown qualitatively in representative experiments in Fig. 1 and quantitatively in Fig. 3. As shown in compiled outcomes in Table 1, the initial damaged areas in Cox-1−/− mice (∼233 μm2) and Cox-1+/− mice (∼186 μm2) were indistinguishable and similar to the area of high intensity light that was scanned across the tissue (150–200 μm2).

Fig. 3.

Time course of damage area expansion and epithelial repair. Time course experiments following events after photodamage (indicated by vertical gray bar) were performed as in Fig. 1 and analyzed as described in methods and Fig. 2. Total imaged areas that were identified as damaged by these criteria were evaluated in representative experiments using Cox-1+/− (▴) or Cox-1−/− (▵) mice at the indicated time points. The horizontal dashed line identifies the size of initial damage.

View this table:
Table 1.

Response to creation of microlesions in Cox-1+/− and Cox-1−/− mice

Subsequent to the initial damage, cells adjacent to those initially targeted gradually lost autofluorescence, increasing the total damage area four- to sixfold, as qualitatively shown in Figs. 1, D and L, and 3 and in the compiled data shown in Table 1. As shown in Table 1, there was a significantly larger maximal damage area in Cox-1−/− versus Cox-1+/− mice, suggesting that damage was more pronounced in the absence of Cox-1.

A second independent measure of cell damage was also used to quantify the damage area. As described in methods, we simultaneously measured extracellular Cl-NERF fluorescence in the juxtamucosal spaces. This led to the unexpected observation that the damaged cells that lost autofluorescence also became permeable toward the Cl-NERF dye in the perfusate. This provided an independent assay of cell viability compared with loss of autofluorescence. We compared kinetics of Cl-NERF uptake and loss of autofluorescence in cells adjacent to the photodamaged area using images acquired sequentially at intervals of ∼2 s (Fig. 4). As shown in three different cells, NAD(P)H autofluorescence decreased gradually over 3 min (not instantaneously, as occurs in directly photodamaged cells; Supplemental Fig. 1), and the cellular uptake of Cl-NERF increased over approximately the same time interval. Based on the use of Cl-NERF uptake as an independent measure of cell injury, the area of damage assayed as Cl-NERF-positive cells was compared versus the NAD(P)H/reflectance measure. In 10 experiments, the damaged area reported by intracellular Cl-NERF at 5 min after photodamage was 133 ± 14% of the area of decreased NAD(P)H autofluorescence. The Cl-NERF assay was qualitatively similar but is predicted to lead to overestimation caused by the inclusion of dye-filled pits and the uncertain boundary when the edge of tissue merges with exfoliating cells.

Fig. 4.

Assay of the expansion of damage by a drop in autofluorescence or an increase in permeability to extracellular Cl-NERF. After photodamage was imposed at time 0, images were collected at the indicated times for Cl-NERF fluorescence (grayscale images of 514-nm fluorescence) and overlayed autofluorescence (green) and confocal reflectance (red) images. A: images indicating that cells adjacent to optically targeted cells become gradually membrane permeable toward Cl-NERF as well as lose autofluorescence; these are among the cells that are shed. Three cells were identified and outlined (1–3) for further analysis. B: each cell was analyzed separately for intensity (arbitrary units) of autofluorescence (▵) and Cl-NERF (▴) over time in the early phase of cell damage, showing simultaneous changes in both hallmarks of cell injury.

Time to Maximal Damage

The times to reach the largest area of damage were distributed from 1 to 9 min in the Cox-1−/− group and from 1 to 6 min in the control group (Figs. 2 and 4), with mean values of the time to peak damage for Cox-1−/− mice significantly longer than for Cox-1+/− mice (Table 1). This may have a trivial explanation because the larger maximal damage area observed in Cox-1−/− mice may simply take longer to manifest. However, the large variability in the time to reach this peak damage weakens our ability to compare rates of damage expansion to test this hypothesis between the two genotypes.

Time to Exfoliate

Injured epithelial cells were gradually pushed from the surface and trapped in the mucoid layer, while the area of damage within the epithelium correspondingly decreased. To quantify the time needed to undergo repair, we selected the time point at which the damage area shrank to the size of the initial damage. Using this time point, the compiled results shown in Table 1 showed that exfoliation and subsequent restitution took significantly longer for Cox-1−/− mice versus Cox-1+/− mice.

Extracellular pH at the Site of Damage

It has been shown that alkali egress into the lumen occurs in response to global gastric damage (1, 11, 29, 30). We sought to test if this was also observed in response to microscopic lesions. Figure 5 shows ratio images of extracellular Cl-NERF adjacent to the gastric surface from a representative experiment. A region of increased pH transiently appeared over the site of damage created by the two-photon laser and spread over the surrounding area as the wound progressed. Figure 6 shows representative time courses of gastric surface pH measurements during the response to photodamage in Cox-1+/− (A) and Cox-1−/− (B) mice. Results are compiled in Table 1, which showed that compared with Cox-1+/− mice, the absence of Cox-1 was associated with 1) initial surface pH values that were significantly lower, 2) maximal pH increases that were significantly smaller after damage, and 3) a slower recovery to the starting surface pH after damage.

Fig. 5.

Time course of extracellular pH imaging with Cl-NERF following photodamage. Photodamage was imposed at time 0, and images of extracellular Cl-NERF fluorescence were collected at the indicated times thereafter. Cl-NERF 514-to-458-nm ratio images (see methods) are presented from a representative experiment, after tissue regions were masked out by thresholding of fluorescence. The rectangle in the “Before” image approximates the size and position of the photodamage pulse, and the orientation of tissue and superfusion fluid (the latter containing Cl-NERF) are indicated as well.

Fig. 6.

Time course of extracellular pH measurement at the gastric surface over the site of damage. Ratio images of Cl-NERF, collected as described in methods and shown in Fig. 5, were analyzed for a 27 × 27-μm2 region adjacent to the site of damage. The resultant time course is shown for representative experiments with either Cox-1+/− (A) or Cox-1−/− (B) mice. As shown, a transient pH increase was observed in both cases. Outcomes from the analysis of the damaged epithelial area (as shown in Fig. 4) showed that the time of maximal damage area (solid arrow) correlated with the maximal pH increase and that the time of exfoliation and epithelial repair (hatched arrow) correlated with the recovery of surface pH.

As shown in Table 1, the average times for surface pH recovery were similar to the average times for exfoliation and epithelial repair. This correlation was also observed to a limited extent in individual experiments. For instance, Fig. 6 shows the time when the maximal damage area was measured in two individual experiments (solid arrow) and the time when exfoliation and epithelial repair were observed (hatched arrow). In individual experiments, surface pH reached baseline values within 5 ± 4 min of the observed exfoliation in 10 of 14 Cox-1−/− mice (range: 18 min before to 22 min after exfoliation) and in 1 ± 2 min of the observed exfoliation in 8 of 11 Cox-1+/− mice (range: 12 min before to 5 min after exfoliation). In the compiled outcomes, there were two cases of brief increase and abrupt return of surface pH within 5 min of damage before any sign of exfoliation/restitution occurred (1 Cox-1−/− and 1 Cox-1+/− mouse). There were also notable exceptions in 4 of 14 Cox-1−/− mice and 3 of 11 Cox-1+/− mice, when surface pH did not recover (sustained increase) within the experimental interval, although exfoliation occurred. The correlation of the time to exfoliate and time to recover surface pH in individual experiments is presented in a scatter plot in Fig. 7. A linear least-squares fit for these n = 18 observations yielded a slope of 0.8 ± 0.2 and R2 = 0.46. The slope was significantly greater than zero, suggesting at least a partial linkage between the two events. We conclude that surface pH recovery is commonly, but not obligatorily, linked to the timing of epithelial restitution.

Fig. 7.

Correlation in individual experiments between the time to recover normal surface pH values versus the time to exfoliate damaged cells. Outcomes such as those shown in Figs. 3 and 6 were compiled from the same experiments on Cox-1+/− (▴) and Cox-1−/− (▵) mice. Each data point (n = 18) is the result of an independent assay of the cycle of damage and repair time course. The dotted line is the ideal correlation of the times for the two processes (slope of 1).

Partial Cox-1−/− Rescue With PGE2

Cox-1 produces prostaglandins that are robust activators of bicarbonate and mucus secretion, so we asked if the addition of exogenous PGE2 (using nonmetabolizable dmPGE2) could reverse the deleterious effects seen in the Cox-1−/− phenotype. Tissues were evaluated before and after a 5-min transient exposure to dmPGE2, but in this long protocol not all experiments could be paired. For this reason, unpaired statistics compared outcomes before and after PGE2 (n = 4–8 experiments for each value). There were no significant differences between the initial damage sizes in Cox-1−/− mice (202 ± 34 μm2) versus Cox-1−/− mice after dmPGE2 treatment (258 ± 67 μm2), but the maximum size of damage in Cox-1−/− mice after dmPGE2 treatment (1,050 ± 106 μm2) was significantly smaller than in the absence of dmPGE2 (1,790 ± 230 μm2). In contrast, when surface pH values before versus after dmPGE2 were compared, there were significant differences before imposing damage (pH 3.4 ± 0.09 vs. 4.0 ± 0.2, respectively) as well as the maximal pH observed after photodamage (pH 3.6 ± 0.08 vs. 4.3 ± 0.3, respectively). Outcomes suggested that dmPGE2 was effective at limiting damage and raising surface pH in the short term. Measurements of the times to exfoliation and restoration of normal surface pH showed no significant differences due to dmPGE2 (data not shown), potentially because the efficacy of the drug declined during extended superfusion of tissue in the absence of dmPGE2 (a financial necessity).


Two-photon microscopy was successfully used to create microlesions at the gastric surface epithelium through photodamage produced by light energy absorption. An advantage of using two-photon absorption for this purpose is that photodamage is restricted to a thin optical section where two-photon absorption occurs (38). This method made possible in vivo visualization in real time of damage expansion, exfoliation of injured cells, and restitution of the underlying epithelium together with the detection of a transient pH increase in the juxtamucosal area over the damaged mucosa.

In the three- to five-cell area targeted for high-power laser scanning, endogenous autofluorescence abruptly decreased and stayed low over time. The mechanism of photodamage by two-photon light absorption was not studied in this initial work, where we focused on evaluating the tissue repair response. However, based on studies of cellular mechanisms of photodynamic therapy, cell death is potentially via absorption of light energy by cytosolic and mitochondrial NAD(P)H, consequent disruption of membranes, and cytotoxicity by reactive oxygen species (ROS) (35). In photodynamic therapy, necrosis occurs when membranes are affected first, and apoptosis when mitochondria are affected first (followed by rapid ROS formation) (2, 6, 22). Our variable observation (data not shown) of cavitation vacuoles and induced membrane fluorescence suggest that other effects of high laser light power could also play a role.

Ito and Lacy (12) used a pervasive superficial ethanol-injury model, and some of the stages and elements of expansion and repair of superficial injury that they describe are also present in our study of isolated superficial microlesions. In both models, cells that become compromised (e.g., lose autofluorescence and become permeable to Cl-NERF) will exfoliate, and the remaining viable cells then restore the epithelial layer (12, 1618). Similar to Ito and Lacy's study (12), we found a mucoid cap also formed over lesions, affecting a small number of cells. However, our model reveals new facets that could not be discovered previously. Initially during injury, cells keep the same position within the surface monolayer, presumably because tight junctions and the basolateral membrane of affected mucous cells remain intact in the damage expansion phase. Damage then expands to surrounding cells that also exfoliate. Exfoliation occurs in very tight time synchrony with restoration of the epithelial layer and somewhat less tightly with restoration of normal surface pH control. Most importantly, repair of the epithelial layer started while only a few injured cells lifted off, not after basal lamina was totally denuded, as in Ito and Lacy's model.

Movement of epithelial cells adjacent to the damaged cells was noticed over time (data not shown), eventually helping to extrude them, but whether these adjacent cells filled the space left by the exfoliated cells or whether cells came from another source, deeper inside glands or from the lamina propria, could not be determined in our analyses because the imaging plane does not image all cells surrounding the damage, and extensive tissue motion compromises analysis of individual cells.

The disruption of the Cox-1 gene slowed down the process of repair/restitution. We chose to compare Cox-1+/− with Cox-1−/− mice so that only one allele was different between animals, because our prior work showed that there is no strong compensatory upregulation of Cox-2 in Cox-1−/− mice and because the physiological response after exogenous prostaglandin (at least in terms of surface pH control) was equivalent among Cox-1+/+, Cox-1+/−, and Cox-1−/− mice (4). We observed that exfoliation took significantly longer for Cox-1−/− tissue compared with Cox-1+/− tissue and that the maximal size of damage was larger in Cox-1−/− mice, although the initial size of damage was similar. Our data support that Cox-1 mediates increased resistance of the mucosa to damage. Cox-1-derived prostaglandins may protect the mucosa by suppressing the release of proulcerogenic factors (nitric oxide, TNF-α, and platelet-activating factor). Among other potential regulatory effects of Cox-1, inhibition of bicarbonate secretion as a consequence of Cox-1 elimination could result in increased gastric damage or inhibited repair (3, 4). When we transiently exposed the gastric mucosa of Cox-1−/− animals to exogenous PGE2, we saw that surface pH values were significantly increased before and after damage and that the maximal damage size was smaller. Results suggest that Cox-1-derived PGE2 is a potential mediator of gastric protection against the early events that lead to damage expansion.

After the acute application of mild irritants to the stomach, Takeuchi and colleagues (14, 3134) observed that the decrease in acid secretion and increase in gastric mucosal blood flow was attenuated under the administration of indomethacin and concluded that Cox-1 is responsible for maintaining the mucosal integrity after barrier disruption. Alternatively, Wallace and colleagues (8, 37) demonstrated that only modest damage was observed in response to selective inhibition of a single Cox isoform in a model of prolonged NSAID-induced gastric injury, whereas inhibition of both isoforms produced more extensive damage. The basis for these differences among damage models remains speculative, but in all cases (including the microlesion photodamage model) a protective role of Cox-1 was identified. It is important to recognize that the source of the transient pH increase after photodamage is unknown. Results suggest that Cox-1 is likely to regulate the process, at least in part, and that a known activator of gastric bicarbonate secretion (PGE2) is at least competent to alter the surface pH response after injury. At this point, it cannot accurately be determined whether the pH increase is due to a decrease in acid secretion, an increase in bicarbonate secretion/leak, or a combination of both.

In the pervasive superficial damage models, a mucoid cap (containing exfoliated cells) and surface pH increases were observed soon after damage (8, 12, 1618, 36). For this reason, the mucoid cap was proposed as a structure that acted to hold secreted bicarbonate at the surface (36). However, we observed that the surface pH increase occurs prior to mucoid cap formation (i.e., cell exfoliation). Thus, our results demonstrate that the mucoid cap is not required to generate a region of raised pH over the site of damage. The simplest model consistent with all observations is that the increased pH is a function of the damaged or denuded epithelium and not the presence of an overlying mucoid cap. It remains to be determined whether the loss of prompt epithelial repair that is observed in the absence of Cox-1 has its basis in the loss of surface pH control or another regulatory function of Cox-1.

In summary, our results in a microscopic damage model suggest that the gastric epithelium is capable of rapid repair of the epithelial layer through a process that allows epithelial integrity to be restored concurrent with cell exfoliation. During most of the process of expanding and resolving cell damage, a surge of pH increase is observed over the damage area that does not appear to be a direct consequence of the egress of damaged cells from the epithelial layer.


This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant RO1-DK-54940 and by a graduate student fellowship (to H. K. Baumgartner) from the American Heart Association.


  • 1 The online version of this article contains supplemental data.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract