In Vivo Rectal Mucosal Barrier Function Imaging in a Large-Animal Model by Using Confocal Endomicroscopy: Implications for Injury Assessment and Use in HIV Prevention Studies

Abstract

Injury occurring on the surface of the rectal mucosal lining that causes defects in barrier function may result in increased risk for transmission of infection by HIV and other pathogens. Such injury could occur from microbicidal or other topical agents, mechanical trauma during consensual or nonconsensual intercourse, or inflammatory conditions. Tools for evaluation of rectal mucosal barrier function for assessing the mucosa under these conditions are lacking, particularly those that can provide in vivo structural and functional barrier integrity assessment and are adaptable to longitudinal imaging. We investigated confocal endomicroscopy (CE) as a means for in vivo imaging of the rectal epithelial barrier in the ovine model following spatially confined injury to the surface at a controlled site using a topical application of the microbicide test agent benzalkonium chloride. Topical and intravenous (i.v.) fluorescent probes were used with CE to provide subcellular resolution imaging of the mucosal surface and assessment of barrier function loss. A 3-point CE grading system based on cellular structure integrity and leakage of dye through the mucosa showed significant differences in score between untreated (1.19 ± 0.53) and treated (2.55 ± 0.75) tissue (P < 0.0001). Histological grading confirmed findings of barrier compromise. The results indicate that CE is an effective means for detecting epithelial injury and barrier loss following localized trauma in a large-animal model. CE is promising for real-time rectal mucosal evaluation after injury or trauma or topical application of emerging biomedical prevention strategies designed to combat HIV.

INTRODUCTION

The mucosal surface lining the gastrointestinal tract functions as a physical interface between the external environment and the body, playing the essential role as a barrier to luminal foreign substances and pathogens (1, 2). Barrier function is maintained by the physical microstructural organization of the epithelium, the molecular microenvironment (including the immunological milieu), and the layer of secreted mucus lining the epithelium, which provides a first line of defense against inflammation and infection (3). Maintenance of each component is essential in protecting the body from luminal contents. The rectal mucosa is considered a susceptible site for human immunodeficiency virus (HIV) transmission, with characteristics that make this site particularly vulnerable including increased numbers of HIV target cells compared to those in other parts of the gastrointestinal tract and similar to those found in the vaginal tract, with the highest numbers of target cells at the distal rectum (4–6). Defects in rectal epithelial barrier integrity and function are a concern for increased risk of transmission (4, 7), with growing concern that breaches in the epithelium may facilitate pathogen translocation (8) or stimulate an immune environment which facilitates transmission (7). Recent studies indicate that physical or functional epithelial layer defects may exacerbate pathogenesis in the event of an acquired infection, allowing for translocation of products from the lumen into the tissue and driving inflammation. Damage to the gastrointestinal tract prior to intravenous inoculation with simian immunodeficiency virus (SIV) in pigtail macaques was recently shown to be associated with increased progression to AIDS (9).

Among scenarios in which rectal mucosal defects are likely to occur are (i) injury due to external stimuli, including (micro)abrasions during anal intercourse (AI), (ii) injury due to use of topical lubricants, spermicides, or microbicides, and (iii) chronic inflammatory conditions (e.g., inflammatory diseases). Each is expected to result in barrier loss and mucosal leakiness. Mechanical shearing during AI will result in focal loss of surface epithelial cells, and given that a single columnar cell layer covers this mucosa, increased permeability is likely. Lubricants, spermicides, and topical microbicides can result in epithelial denuding (10–13), loss of adherens and tight junctions (14), and increased mucosal permeability (15). In some cases, increased susceptibility to infection by HIV and other sexually transmitted infections (STIs) has been indicated as in the case of the spermicide gel nonoxynyl-9 (16, 17). Chronic inflammatory conditions have been linked with a dysfunctional and/or leaky mucosa and an increase in cellular targets for HIV (7).

The full impact of gastrointestinal dysfunction on disease transmission and pathogenesis is only beginning to be understood. A greater understanding of the effects of injury or trauma on both the structural and functional barriers of the rectal mucosa is needed. Moreover, in vivo models mimicking probable injury scenarios as well as methods to monitor the functional barrier are needed. Previous approaches to the study of agent safety have included a range of methods. In vitro cell culture models have allowed the rapid testing of a variety of agents simultaneously but do not recapitulate the in vivo three-dimensional (3D) microenvironment (18). Other in vitro methods following in vivo examination or treatment include biopsy specimen collection with histology or immunological analysis, collection of fluids by lavage or swabs for analysis of cellular and immunological components in animal models (19) and humans (20), and the use of tissue for short-term ex vivo human rectal tissue explants (21–23). Histological evaluation performed on biopsy specimens or in explants reveals structural defects or may even provide molecular markers of damage but lacks the ability to simultaneously assess the functional barrier through permeability or mucosal leakiness testing. Collection of fluids via swabs or lavage reveals inflammation and indirect indicators of barrier loss which may be obtained rapidly and repeatedly but lack structural information. More recently, methods for mucosal tissue assessment that do provide structural or functional assessment have been used. For structure, in vivo imaging by colonoscopy and optical coherence tomography revealed disruption of the rectal surface with a topical test agent (24). With a resolution on the order of 15 to 20 μm and morphological imaging based on reflected light, it was not possible to obtain cellular-level detail or assess function with this method, though indication of morphological change in depth was possible. With regard to function, Fuchs et al. employed plasma sampling of probes delivered to the rectal compartment for monitoring permeability across this tissue, with the advantage that changes in permeability due to rectally applied gels may be estimated on a global basis (15). In the development of safer topical products or in the goal to better understand barrier defects, there is also an interest in assessing the local tissue effects on structure as well as function. Advanced imaging provided by confocal endomicroscopy (CE) may offer a new approach to study the effect of injury on the structural and functional barrier of the rectal mucosa. CE has been applied clinically to evaluate epithelial microstructure associated with neoplasia (25). Recently, it has been used to investigate gastrointestinal barrier structure and permeability in patients with inflammatory bowel disease (26). This real-time, high-resolution technique could be used to better understand how injury affects the rectal mucosal physical and functional barrier, the loss of which may result in increased risk to infection by HIV and other pathogens.

The purpose of this study was to investigate the in vivo use of CE in the ovine model as a method for quantitative image-based evaluation of barrier structure and function resulting from a localized injury on the rectal mucosa. To model damage, we created a spatially defined local injury resulting in denuding of the surface columnar epithelium by focal application of a positive test agent for injury, benzalkonium chloride (BZK), to the surface using a contact applicator. Tissue at the site of and surrounding the injury was evaluated to assess subtle changes in the microstructure and function of the rectal barrier using in vivo confocal endomicroscopy and histopathology.

MATERIALS AND METHODS

Animal model and imaging.

Six virginal Rambouillet Merino sheep weighing 25 to 35 kg were used, with studies approved by the Institutional Animal Care and Use Committee at the University of Texas Medical Branch. Studies conformed to the Guide for the Care and Use of Laboratory Animals (27). Sheep were kept on a 12-h/12-h light/dark cycle and housed for a minimum of 2 weeks prior to imaging studies. Animals were fasted the evening before imaging, which was performed the following morning. On the day of imaging, sheep were sedated with 10 mg/kg of ketamine intramuscularly (i.m.), followed shortly thereafter by intravenous (i.v.) administration of 10 mg/kg of ketamine and 0.1 to 0.2 mg/kg of diazepam, and placed dorsal supine on an examination V-tilt table.

Imaging was conducted using an Optiscan FIVE1 (Optiscan P/L, Notting Hill, Victoria, Australia) laser point scanning confocal endomicroscope. The endoscopic probe is 5 mm in diameter and consists of a flexible endoscope having a rigid 30-cm endomicroscope probe at the distal end housing a scanning unit for imaging. In this system, light from an argon ion laser having a wavelength of 488 nm is guided through an optical fiber for single-point illumination with collection of fluorescence emission (filters allow for collection of 505 to 750 nm) through the same fiber. Fluorescence emission from the tissue is collected point-by-point in a raster scanned pattern providing a field of view is 475 by 475 μm. The lateral and axial resolutions are 0.7 μm and 7 μm, respectively. The frame rate of image acquisition (1,024 by 1,024 pixels) used was 0.8 frames/s; once scanning is initiated, ongoing video endoscopy is performed. For acquisition of images at depth beyond the surface, the endoscope allows for optical sectioning in depth at increments of 4 μm (measured in air). Acquisition of ∼25 frames, a typical number per site, takes approximately 40 s.

In these studies, two fluorescent labels were used to provide (i) barrier structural (microarchitectural) evaluation and clear delineation of the single cell layer on the surface (acriflavine [AF], 0.05% in saline; Sigma) and (ii) assessment of barrier function using a permeability probe (fluorescein [FL]; AK-Fluor; AKORN, Lake Forest, IL). A unique aspect was the combined use of both dyes for better delineation of cells on the mucosal surface (AF) while also detecting permeability (FL). Pilot assessments were conducted for each individual label and then with the combined use of the two labels as described below. Pilot assessments were also used to determine the dose of 0.01% BZK for inducing focal damage.

A gauze-filled glove was inserted 10 to 15 cm into the rectum to prevent feces from moving into the area of interest in the distal rectum and creating a reservoir for containing the topical fluorophore AF. A phosphate-buffered saline (PBS) rinse (10 ml) of the distal rectum was performed to further clean feces from the site. Baseline background images without the use of fluorescent labels were obtained of the rectal surface by insertion of the handheld endomicroscope into the rectal compartment and placement of the optical probe in contact with the tissue surface to acquire images in a plane parallel to the tissue surface (these background images contained no detectable signal). Next, the rump was raised (∼6 to 7 in.) by placing a wedge beneath it on the V-tilt table and a 10-ml intrarectal bolus of AF was given for topical staining of the rectal lumen. The area was rinsed with PBS after 5 min and baseline images were obtained from the tissue surface and in depth. Sites for imaging were on the posterior, anterior, and lateral rectal walls, and at each wall face, three regions were imaged approximately 20 mm beyond the anorectal junction on the rectal surface. FL was then given by intravenous administration in which 500 mg of FL was delivered to an i.v. saline bag having 250 ml of saline at an infusion rate of 3 ml/min. Images were obtained of the same regions beginning approximately 2 min after the initiation of FL delivery and continuing for approximately 20 min.

Localized injury to the rectal mucosa was accomplished through topical application of BZK, an agent that has been used in contraceptive gels and lubricants (28, 29), as a topical antimicrobial in the vagina (30), and as a positive control in microbicide safety studies (31–33); it is known to result in epithelial disruption in a predictable manner with concentrations in the range of 0.02 to 2% in previous studies (24, 34, 35). A 16-in. sterilizable proctoscopic swab (Fox Converting, Inc., and available from Medline) soaked in 0.01% benzalkonium chloride (BZK) was placed in contact with the rectal mucosa at the posterior wall 20 mm beyond the anorectal junction for 5 min and then removed. The rectum was rinsed with PBS, and images were obtained within and outside the perturbation site on the lateral and anterior walls. During all procedures, a pediatric Graves speculum was used. Imaging time following removal of the swab was approximately 30 min.

Animals treated with 0.01% BZK were euthanized and the rectal tract obtained for histological assessment. Samples were fixed in 10% buffered formalin and processed for hematoxylin and eosin (H&E) staining of sections from imaged sites.

Image scoring.

A scoring system was devised in which metrics were based on both structural and functional measures of integrity from images of the surface and in which the tissue was stained with the two dyes (AF and FL) simultaneously. This differed from CE scoring of gastrointestinal tissue, in which a single dye (fluorescein) was administered (26). Cellular surface features were primarily delineated by the AF stain, which strongly labeled the nuclei of cells, the staining characteristics described in the literature (26). FL provided indication of leakage across the mucosal tissue as well as additional structural features in the lamina propria. Although FL labeled surface cells as well, using this dye alone, it was difficult to consistently determine if surface columnar cells were present in cases of flooding by FL with leakage. The distinct nuclear staining and recognizable ordered pattern of the cell nuclei on the epithelium when using AF removed this ambiguity. Thus, the two dyes were used together despite having similar spectral excitation and emission properties.

In the localized injury study, 370 surface images were evaluated from the six sheep. Images were randomized and grouped into a set for training graders (100 images) and a grading set. Since it was determined in pilot studies that surface images provided the most realistic indication of injury and leaky mucosa, the 3-point CE grading system was based on surface features. Structural changes and permeability changes often coincided. Three general image types were identified and scored as grade 1, 2, or 3, based on surface microstructural (cellular and crypt) features and evidence of FL dye leakage from the mucosa (features described below and in Table 1). Graders trained to identify features of leakage and structural damage and masked to the experimental conditions graded 270 images from the 6 six sheep, providing each image a score of 1, 2, or 3. Fleiss's Κ score was used to evaluate the degree of agreement between graders. Scores were averaged between the three graders. Plots showing the proportion of scores based on condition were created for a final summary of findings. Statistical analysis between the two groups was determined using the Mann-Whitney U statistical test, with a P value of <0.05 considered significant.

TABLE 1.

Grading criteria for CE scoring

Grade	Criteria
1	Ordered arrangement of the gland
	Intact layer of columnar epithelium
	No microstructural damage
	Glands round or elongated
2	One, some, or all lumen of the crypts filled with fluorescein
	No microstructural damage
	Flooding of fluorescein (between the crypts) less than 25%
3	Exfoliation in the cellular structure
	Disruption in crypts
	Excessive flooding

Histological assessment.

Pathological grading was performed on biopsy specimens from untreated and BZK-treated sites with metrics based on inflammation (scored 0 or 1), edema of the lamina propria (0 or 1), epithelial disruption or detachment (0 or 1), presence of microabcesses (0 or 1), and hemorrhage (0 or 1), similar to a previous study (24). Four to six sections per sheep (five fields per section) (10×, 0.85 numerical aperture [NA]) were examined. A cumulative score was obtained by summing the individual scores for the pathological observations. An independent quantitative measure of epithelial disruption was to measure the percentage of epithelial denuding from the mucosal surface. This involved using an image processing software (ImageJ; NIH Image). Line segments were used to manually delineate and measure all surface regions missing the full columnar epithelium, and the sum length of these comprised the denuded epithelium length. A ratio of denuded length to total length was used to determine percent epithelial disruption.

RESULTS

Typical features identified on the surface of the rectal mucosa with CE and the fluorescent labels (AF and FL) are shown in Fig. 1. In Fig. 1A, normal rectal mucosa was stained with topical AF alone, showing the microstructural organization of the surface columnar epithelium; the nuclei of epithelial cells are visible due to the dye. The colonic crypt architecture with surrounding epithelial cells and goblet cells lining the crypt is shown. The crypt lumen is evident as larger dark regions, which are generally round (arrows). While not visible in Fig. 1A, folds may appear across the surface, making crypts appear flattened. The cells of the columnar epithelium are well ordered, closely packed, and without significant gaps (occasionally an epithelial gap typical of cell shedding is visible). A plane in the lamina propria is shown in Fig. 1B to demonstrate staining in this layer. For Fig. 1C, FL was injected following AF topical staining and the image taken at the surface; the result is a more uniform signal throughout the mucosa. However, individual nuclei of cells are still visible, and the closely packed pattern of cells on the surface is evident (asterisk). As an example of what CE reveals in depth, Fig. 1D shows the lamina propria following AF application plus FL infusion where vessels may be identified. Figure 1E and F show the surface following BZK treatment, where leakage of FL from the tissue is evident, resulting in staining patterns in which FL is seen pooling in the crypts as well as in tissue gaps that occur on the surface between the crypts (Fig. 1E). Pooling in the crypt center is visible by small bright circles of dye seen in the center (particularly bright in Fig. 1E). FL pooling at the surface is evident by the bright uniform (smooth, and lacking texture) signal that dominates areas between the crypts. In contrast to the clean appearance of the undamaged tissue (Fig. 1C), free-floating cells and debris can be identified. In Fig. 1F, significant disruption is found in the organization of the mucosal surface, with damage to the epithelial lining evident by large gaps between islands of cells (e.g., note the island of bright cells near the top [asterisk]) in addition to the pooling of FL. Over time, the FL leakage necessitates PBS washing of the surface to maintain clear imaging.

FIG 1.

Typical features seen on the rectal mucosa by CE. (A) Topical staining of the mucosal surface nuclei with AF. Arrow indicates a crypt/gland lumen. (B) Image below the surface showing lamina propria (arrow, crypt lumen). (C) Surface following delivery of intravenous FL making the surface overall brighter; despite a brighter background in the space between nuclei as in panel A, individual nuclei of the columnar surface are still clearly evident (asterisk), as are crypts (arrow). (D) AF- and FL-labeled tissue showing vessels found below the immediate surface. The lamina propria comprises the tissue between vessels and crypts. A region with surface columnar epithelium is shown in the upper left corner (asterisk). (E) Surface following treatment with 0.01% BZK in the AF- and FL-labeled mucosa. FL has leaked and accumulates on the surface between crypts as a bright uniform signal (#) and can be seen in the center of crypts (arrow). (F) In a second example, the surface is highly disrupted, and islands of surface epithelial cells remain with large gaps in between (asterisk). CE micrographs are 475 μm across. Scale bar: 100 μm.

For improved understanding of the image features following injury and loss of barrier function (permeability) in Fig. 1E and F, micrographs of BZK-damaged mucosa labeled only with AF were obtained (Fig. 2). Comparing the two micrographs in Fig. 2 with that of Fig. 1A of intact mucosa labeled only with AF, one can see the structural damage induced by BZK. In the case of Fig. 2A, exfoliation of the surface epithelial cells in the spaces between individual crypts is seen (asterisks). This pattern of missing cells between crypts mirrors the location in which FL pooling is seen in Fig. 1E between crypts. In the second damage case, shown in Fig. 2B, the actual crypt structure is clearly disrupted at the surface and much of the lamina propria is evident between islands of surface columnar cells (labeled “lm”). This damage pattern would be expected to correspond to Fig. 1F, in which the surface is highly disrupted and clumps of surface epithelial cells remain but FL leakage is also evident. Figure 3 is shown as an example of other features that may be revealed by CE but not graded in the CE scoring system. Figure 3A shows disorder in goblet cell organization within the crypt lumen compared to the ordered pattern of Fig. 1A. Inflammatory cells evident within the lamina propria include lymphocytes within the stromal extracellular matrix (Fig. 1B), similar in appearance to those reported elsewhere (36). Additionally, blood vessels within the lamina propria show accumulation of leukocytes indicated in Fig. 3C and D.

FIG 2.

Microstructural organization of the surface following injury using only the AF topical dye without FL. (A) Surface columnar epithelium neighboring the crypts remains present in this example; however, continuity in the epithelium between crypts is missing (dark empty areas, asterisks). (B) Marked disruption to the surface and crypt structure is evident. Asterisks indicate exposed lamina propria (lm) at the luminal surface seen next to neighboring islands of epithelial cells. Scale bar: 100 μm.

FIG 3.

Mucosal cellular responses observed with CE following BZK treatment. (A) Cellular exfoliation was noted, including disruption to the regular goblet cell organization compared to untreated mucosa (compare to Fig. 1A). (B) Perivascular lymphocytes within lamina propria (arrows). (C and D) Intravascular leukocytes (arrows) occluding vessel lumen; the asterisk in panel D indicates a larger vessel having a bright fluorescein signal. Scale bar: 100 μm.

An advantage of CE is the ability to interrogate areas as needed in real time. Figure 4 shows an example in which images were taken across the transition between an untreated area and the BZK-treated site. Figure 4A represents an area outside the injury site, Fig. 4B was taken at the transition between uninjured and injured mucosa, and Fig. 4C was taken at the center of the injury site. In Fig. 4A, the surface is highly intact, resembling uninjured surfaces shown in Fig. 1C. In Fig. 4B, some disorder is evident (e.g., a general crypt shape is seen but the lamina propria is evident at the surface without overlying epithelium); however, intact crypts with a distinct epithelium at the surface can be seen in the upper part of the field. In Fig. 4C, the tissue is clearly disrupted, with distortions in crypt structure evident throughout the field. H&E micrographs taken from similar regions are shown in Fig. 4D, E, and F. In Fig. 4D, histological features of normal tissue are a single cell layer of continuous surface columnar epithelium lining the surface and crypts extending into the tissue (seen in transverse section). In Fig. 4E, partial disruption of the surface is visible. Columnar epithelium remains on one side of the micrograph, but other areas are fully denuded, consistent with the lack of surface cells over much of the corresponding CE in Fig. 4B. In the center of an injury site (Fig. 4F), the columnar epithelium is missing in much of the field, consistent with the cellular exfoliation seen in the confocal images of Fig. 4C as well as those in Fig. 1F and 2A and B. In this histology section, one can see that the underlying lamina propria is exposed to the lumen in treated areas with exposure of the vasculature and immune cells that would normally reside under the surface epithelium. The crypt architecture is largely retained even in the BZK-treated specimens; however, a noted feature is the loss of goblet cells from the lumen of the crypts, obvious in Fig. 4F compared to Fig. 4D. In corresponding CE (Fig. 4C), goblet cells in the crypt lumens are less evident and disordered compared to the uninjured side (Fig. 4A).

FIG 4.

Transition between an untreated site and a site with BZK treatment. (A) Largely intact surface with intact epithelium and goblet cell organization. (B) Transition area with normal structure near the top of the field and indication of altered surface in the lower two-thirds of the image, with missing epithelium. (C) Area with largely disordered crypt and surface; areas of the lamina propria (lm) appear between crypts. When visible, goblet cells in the lumen appear disordered. CE scale bar: 100 μm. Corresponding H&E micrographs are shown. Panel D corresponds to normal case showing an intact surface layer and ordered crypts (arrow). Panel E depicts a transition zone showing columnar epithelium on the left side of the micrograph but disrupted surface layer otherwise. Panel F corresponds to an area clearly in the damage zone; much of the columnar epithelium is disrupted, and in areas, lamina propria is exposed to the surface (lm). Note the absence of goblet cells within crypts. CE micrographs are 475 μm across. H&E scale bar: 50 μm.

Figure 5 shows CE images representing the grades used in the scoring system and typical histology. The criteria for CE grading are in Table 1. In grade 1 the tissue surface appeared normal, having a clear continuous columnar epithelial surface and a regular pattern of intact crypts as described in the literature (26). While crypts generally appeared round, an elongated appearance (evident as folds) was possible (as shown in Fig. 5A), with the defining criteria of grade 1 being the intact epithelium (closely packed continuous layer of epithelial cells), with no leakage and no evidence of surface disruption. In histology, these sites had a continuous layer of columnar epithelial cells lining the mucosa with no evidence of inflammation, hemorrhage, or edema (Fig. 5D and 4A). CE images classified as grade 2 demonstrated no clear evidence of epithelial cell exfoliation but had evidence of FL localized within the crypts (bright round, filling crypt glands) or less than 25% of the image surface area showing leakage between crypts. CE images classified as grade 3 had substantial flooding of the mucosal surface by FL in addition to evidence of microstructure disorganization (cellular disruption or altered crypt shapes). In histology, damage to the mucosa appeared like micrographs of Fig. 5B and C. In Fig. 5B, while there is partial epithelial exfoliation, the lamina propria is not as significantly exposed as in the case of Fig. 5C. In Fig. 5C, there are large regions in which the lamina propria is exposed to the lumen, likely corresponding to a grade 3 CE image. Inflammation is present in both Fig. 5B and C. No evidence of edema or significant hemorrhage was seen in samples by histology.

FIG 5.

Three-point confocal grading system: representative grade 1, 2, and 3 images. (A) Typical grade 1 images in which the surface is intact, with no evidence of structural breaches or barrier loss (no FL leakage). (B) Grade 2 example showing a partially damaged area, no clear evidence of microstructural damage, but some FL leaking (#). (C) Significant structural and functional defects evidenced by exfoliated epithelial cells (asterisk) and significant FL leakage from the tissue (#). Scale bar: 100 μm. (D) Typical H&E micrograph of normal intact epithelium outside the injury zone showing epithelial cells (ep) lining the mucosal surface and crypts with goblet cells (gc) inside crypts. This layer covers a lamina propria having extracellular matrix, vessels, and stromal support cells. (E and F) Damaged region showing exfoliation of epithelial cells and presence of leukocyte infiltration (arrow). (H&E scale bar = 50 μm).

Results of grading by CE and histology are shown together in Fig. 6. Histological grading indicated damage to the BZK-treated sites, which had a mean score of 2.68 (standard deviation [SD], 0.39), versus 1.75 (SD, 0.35) for untreated sites. The quantitative measure of percent epithelial exfoliation evident in micrographs in untreated and BZK-treated mucosae is shown in Fig. 6B, showing the loss of epithelium from which FL could have leaked as seen in CE following BZK treatment. As seen in Fig. 6C, the mean CE grade obtained for untreated tissue was 1.19 (SD, 0.53), and that for BZK-damaged tissues was 2.55 (SD, 0.75). Of the 270 surface images examined by the three trained graders, 168 were from BZK-treated areas and 102 from untreated areas. Results per grader are summarized in Table 2. The BZK-treated images received a grade of 2 or 3 in 141 to 144 of 168 images depending on the grader. Determination of agreement between the three graders was performed using Fleiss's Κ. The resulting Κ value was 0.797, indicating excellent agreement between the three graders.

FIG 6.

Histological and confocal grading results. (A) H&E results of pathology grading for damaged surface, inflammation, and edema (P < 0.005). (B) Measurement of percent epithelial disruption between untreated samples and samples treated with 0.01% BZK (P < 0.005). (C) Average CE grades per condition show a larger CE score in BZK-treated than in untreated samples. Values were obtained by taking the mean value of the scores from three graders. Error bars represent standard deviations. Statistical analysis was by Mann-Whitney U statistical test; P < 0.0001.

TABLE 2.

Results of CE scoring by individual grader

Grader	No. (%) of images with indicated CE score
	BZK treated			Untreated
	1	2	3	1	2	3
I	27 (16)	25 (15)	116 (69)	92 (90)	4 (4)	6 (6)
II	26 (16)	24 (14)	118 (70)	87 (85)	6 (6)	9 (9)
III	24 (14)	26 (16)	118 (70)	88 (86)	9 (9)	5 (5)

DISCUSSION

There is a need for methods providing noninvasive, rapid, and repeated evaluation of the rectal mucosa in vivo to elucidate mucosal responses to injury which may be linked to increased risk of mucosal infection. Scenarios include evaluating safety of HIV prevention agents (microbicides) and lubricants, effects of trauma due to consensual and nonconsensual intercourse, inflammation, and mucosal response during wound healing. In particular need are tools that can provide functional as well as structural indication of loss of barrier function with the added adaptability of longitudinal imaging. We examined the use of CE as a high-resolution tool for assessing the in vivo rectal epithelial barrier in the ovine model. CE provided high-resolution (subcellular) imaging of the mucosal surface, allowing for indication of barrier microstructure, and highlighted loss of barrier function through the use of a permeability probe. With a two-dye approach, CE revealed epithelial layer disruption as well as leakage of a permeability dye from the tissue into the lumen following focal injury resulting in loss of cells from the surface columnar epithelium.

A CE scoring system was developed to grade barrier structure and function together in vivo, and it revealed a structural and functional barrier loss following treatment with BZK. Histological grading similarly found barrier loss specifically due to surface structural compromise without a direct functional measure. Upon grading a masked data set, it was determined that CE identified the majority of BZK-treated sites as injured (grade 3) and the majority of untreated (PBS) sites as normal (grade 1). An excellent level of agreement was found between the three graders, assessed through Fleiss's Κ (value = 0.797), and final CE scores between groups were statistically significant. Comparison to histology allowed for anatomical features to be characterized and aided in the definition of grading criteria for CE. For example, in CE, the unique staining pattern of the epithelium by AF allowed for indication of loss of surface cells denoting a structural injury even in the presence of FL. In histology, distinct exfoliation of the columnar layer was evident with BZK injury and validated the CE observations where surface cells labeled by AF were missing. This was particularly true for CE grade 3, in which it appeared that the lamina propria was exposed and large pools of FL leaked from the surface. The percent epithelial disruption from the surface in histology was a second measure that explained pools of FL leaking from the surface seen in CE. Interestingly, in CE (grade 2), there were cases in which FL leakage occurred but there was not a strong indication of epithelial disruption, potentially either because those areas may have been masked by the FL or the damage occurred due to compromised tight or adherens junctions. Permeability could not be assessed by histology in the current study to confirm leakage in the absence of obvious exfoliation. A corresponding histology approach to complement CE will be to assess tight junctions and/or adherens junctions by immunohistology, which for now in the sheep will necessitate testing of antibodies or reagents for reliable markers.

In the current study, we did not observe consistent trends in lymphocyte and leukocyte numbers, possibly due to excess FL with injury at many sites, but these immune cells could be seen in CE micrographs in damaged regions because they were likely exposed due to surface exfoliation (Fig. 3). Future studies will be designed to examine these components and other features (e.g., vasculature) as we recognize that the rich content that may be contained in CE could lead to additional analyses and an improved scoring system that contains more information about the structural, functional, and immune aspects of injury. This may require optimization of the labeling protocol for structural and permeability assessment. As part of such future efforts, scenarios involving a broader range and type of injury (e.g., denuding with inflammation versus inflammation alone) would be helpful in developing a CE scoring method with more dynamic range. Possible expansion of the histology scoring system dynamic range may also be possible with a broader range of injury or inflammation. Other studies have included metrics for subtypes of inflammatory cells (mononuclear infiltrates, neutrophils, and eosinophils) as well as indicators of epithelial or crypt damage with several subtypes (37, 38) and serve as examples of how both the CE and the histological scoring could be expanded.

These studies used sheep as a model for rectal mucosal injury. Like in the human rectum, the surface of the sheep rectal mucosa consists of a layer of secreted mucus overlying a single columnar layer of epithelium and underlying lamina propria. Crypt microstructure is similar, comprising glands lined with columnar epithelial cells and goblet cells within the lumen. Due to similarity in scale and anatomy, the sheep has been adapted for use in the evaluation of microbicide products in the vagina (39, 40) and as a model for vaginal and rectal topical injury (24). In this study, the distal rectum was chosen for the site of injury, as it is an area believed to be particularly vulnerable to HIV. Inducing controlled injury to a specific focal site on the rectal mucosa was straightforward since the size of the lumen provided accessibility for direct application of the swab at a directed location with visualization. A noted limitation of the sheep model is that there is not currently a manner to induce infection by a pathogen similar to HIV, as in the nonhuman primate model of simian-human immunodeficiency virus (SHIV), limiting its current use to drug and preventive agent safety studies as well as studies based only on mucosal microenvironment features.

CE has been investigated for use in gastrointestinal cancer detection and inflammatory bowel disease. In this study, we found it advantageous to employ a double-staining approach in which nuclei of the mucosal surface were labeled with AF prior to i.v. delivery of the FL solution. This was necessary due to the noted ambiguity of identifying whether FL had penetrated beyond the surface epithelium into the lumen. Adding AF significantly enhanced our ability to define the lumen/tissue interface regardless of the presence of FL.

To produce injury on the rectal mucosal surface, a chemical agent, BZK, found in lubricant products and which has constituted a component of gels designed for vaginal use as a microbicide was used. The effects of BZK on epithelium are known (32–35). Since BZK results in superficial denuding of epithelium, it provided a means to induce structural injury similar to that expected with either mechanical or chemical injury. In our case, BZK resulted in stripping of surface epithelial cells, and since the lamina propria was exposed in areas, the protective mucus layer lining the surface was also affected. The loss of goblet cells observed in histology and CE (Fig. 5) indicated that this protective layer was affected. An interesting finding was that injury could be induced using even the very low concentration of 0.01%, a dose several times smaller than found in many products and one which is much smaller than doses found to result in partial denuding of the superficial vaginal epithelium (33–35). This is likely due to the fact the rectal epithelium only comprises a single layer of columnar cells. It is noted that while treatment led to exfoliation of cells, as might be expected with mechanical injury, differences from a true mechanical injury could include differences in inflammatory response.

Our purpose in applying damage in a localized area was to simulate conditions in which injury resulting in epithelial denuding occurs focally, rather than uniformly across the global surface. A second reason was to demonstrate the ability of CE to visualize transitions between intact and injured epithelium in real time (as shown in Fig. 4). Defects produced by the mechanical mode of injury are likely to be of a focal nature rather than uniform across the entire mucosal surface. Epithelial disruptions in inflammatory bowel disease were also found to occur focally (although across a large surface area) (26). Chemical injury could also result in focal injury even when the agent is applied as a bolus, as has been found in vaginal application of gels resulting in focal damage (33, 41), though defined borders of injury cannot be dictated for controlled injury studies. By inducing a single localized point of injury, it was possible to more specifically define the site of injury and compare characteristics to that of intact, uninjured epithelium at the same time points in the same animal.

The ability of CE to obtain images containing structure and functional epithelial barrier indicators of in vivo mucosa (i) at multiple time points (e.g., before and immediately following injury) and (ii) at multiple sites with time frames of tens of seconds per site without (iii) physical perturbation of the tissue such as required for biopsy highlights the power of this in vivo imaging approach. In contrasting the method to histology, it is noted that while several biopsy specimens may be obtained at any given time, once removed, the tissue at that location is no longer available for subsequent short-term follow-up and inflammatory responses are likely triggered in obtaining the biopsy specimen that affect the surrounding microenvironment. At the same time, CE has limitations. In the current form, CE did not allow for molecular assessment that immunohistochemistry or in vitro lavage sample testing can provide. Development of molecularly targeted probes or translation of in vivo probes used in small-animal models in the future could address this limitation. However, given the complex multifactorial responses of the mucosa to injury, methods which combine the advantages of multiple modalities are desirable. Thus, despite specific advantages of CE for noninvasive, repeated real-time imaging, future scenarios in the use of CE for assessing mucosal injury can likely include traditional methods such as histology, lavage, or even global permeability assessments.

The results of this study only begin to explore the potential use of CE for assessing barrier compromise in the context of safety. Future studies should include investigation of mechanical versus chemical injury as well as subsequent wound repair and more focus on cellular responses, including immune cell responses and effects on goblet cells. This method could be extended to a variety of injury scenarios, including injury from consensual or nonconsensual intercourse and mechanical injury in the context of lubricants or topical product use. There is a continuing need to evaluate the safety of topical agents applied to the mucosa, including HIV biomedical prevention strategies that may be classified as topical microbicides or preexposure prophylaxis (PrEP), considering that previously developed products have progressed through preclinical screening only to fail in clinical trials. Of interest are studies that examine more proximal sites of the colon, as there is indication of localization of microbicide gels tens of centimeters from the anorectal junction (42). Commercially available clinical versions of the CE used in this study having only a 5-cm rigid portion at the distal end coupled with a video colonoscope used clinically to image as far as the terminal ileum in human subjects (36, 43) would allow such studies to be performed. Finally, CE evaluation could be applied in the context of other factors that affect barrier function, including the microbiome (7).

In conclusion, this study demonstrated the potential of confocal endomicroscopy for in vivo evaluation of the colorectal barrier function and structure, introducing a novel method to evaluate both permeability and surface injury simultaneously. Moreover, all assessments were performed in vivo in the sheep, an animal model with anatomical features comparable to those of humans. In future studies, it will be of interest to adapt this method for use in longitudinal evaluations, examining wound healing, and expanding to other injury scenarios of the rectal mucosa that could affect safety, as well as translation of the method for use in clinical studies.