Endoscopy is widely used to detect and remove premalignant lesions with the goal of preventing gastrointestinal (GI) cancers. Because current endoscopes do not provide cellular resolution, all suspicious lesions are biopsied and subjected to histologic evaluation. Technologies that facilitate directed biopsies should decrease both procedure-related morbidity and cost. Here we explore the use of multiphoton microscopy (MPM), an optical biopsy tool that relies on intrinsic tissue emissions, to evaluate pathology in both experimental and human GI specimens, using hematoxylin and eosin (H&E)-stained sections from these tissues for comparison. After evaluating the entire normal mouse GI tract, MPM was used to investigate disease progression in mouse models of colitis and colorectal carcinogenesis. MPM provided sufficient histologic detail to identify all relevant substructures in ex vivo normal GI tissue, visualize both acute and resolving stages of colitis, and show the progression of colorectal carcinogenesis. Next, ex vivo specimens from human subjects with celiac sprue, inflammatory bowel disease, and colorectal neoplasia were imaged by MPM. Finally, colonic mucosa in live anesthetized rats was imaged in vivo using a flexible endoscope prototype. In both animal models and human specimens, MPM images showed a striking similarity to the results of H&E staining, as shown by the 100% concordance achieved by the study pathologists' diagnoses. In summary, MPM is a promising technique that accurately visualizes histology in fresh, unstained tissues. Our findings support the continued development of MPM as a technology to enhance the early detection of GI pathologies including premalignant lesions. Cancer Prev Res; 5(11); 1280–90. ©2012 AACR.
Colorectal cancer (CRC) is the second leading cause of cancer-related deaths in the United States with an estimated 140,000 new cases in 2011 (1). The use of colonoscopy to identify and remove premalignant colorectal adenomas reduces the risk of developing CRC (2). For average-risk individuals, colonoscopic screening is recommended every 10 years, beginning at age 50 (3). In individuals with a family history of CRC, earlier screening is recommended (3). In addition, individuals with inflammatory bowel disease (IBD) for more than 8 years are at increased risk for developing CRC and surveillance colonoscopies including multiple random biopsies are recommended every 1 to 2 years (4, 5). Upon discovering dysplasia in an IBD patient, colectomy is often advised to prevent potential life-threatening CRC (6). Other gastrointestinal (GI) diseases that increase the risk of malignancy include Barrett's esophagus, chronic gastritis, and celiac sprue (7–9). Improved imaging methods are needed for the early detection of premalignancy in both the upper and lower GI tract.
Current endoscopes, even with adjuncts to white light, i.e., pre- or post-processing image modulation or chromoendoscopy, do not provide cellular resolution that permit routine distinction between benign and dysplastic lesions. Consequently, the focus of endoscopic screening today is to biopsy or resect any suspicious lesions. The problems associated with such an approach include sampling error, removal of benign lesions, prolonged procedure times, and multiple negative samples for pathologic interpretation leading to additional cost. Therefore, improved early detection methods, which can facilitate targeted biopsies are needed.
Novel “optical biopsy” techniques, such as confocal microendoscopy and optical coherence tomography (OCT), can provide high-resolution imaging in situ, but each approach has significant limitations (10, 11). Here we explore the use of multiphoton microscopy (MPM) as a next-generation optical biopsy tool. MPM relies on the simultaneous absorption of 2 or 3 low-energy (near-infrared) photons to cause a nonlinear excitation equivalent to that created by a single photon of bluer light. Excitation only occurs where there is sufficient photon density, i.e., at the point of laser focus, providing intrinsic optical sectioning. Tissue penetration is greater than with standard confocal microscopy because absorption and scattering of the laser excitation is reduced at near-infrared wavelengths (12, 13). Most importantly, by using 2-photon excitation in the 700 to 800 nm range, MPM enables both in vivo and ex vivo imaging of fresh, unprocessed, and unstained tissue via intrinsic tissue emissions, which includes both autofluorescence and second harmonic generation (SHG) (12–14). MPM imaging, using intrinsic tissue emissions, is capable of generating distinct optical signals from different tissue components that enable imaging of animal (13, 15, 16) and human (17–22) tissues at submicron resolution in 3 dimensions to a depth of up to 0.5 mm below the specimen surface. These imaging parameters enable real-time detailed visualization of cellular and subcellular changes that accompany the development of disease.
Recent studies of MPM in multiple organ sites including the GI tract underscore the potential utility of this technology (18, 23–30). Although MPM has been previously used to assess GI tissue architecture in both animal models and human biopsies (18, 23–32), we posit that the current study is the most comprehensive evaluation of MPM imaging of the GI tract, carried out in 4 phases: (i) generation of an MPM atlas of the entire normal mouse GI tract, including the hepatobiliary system and pancreas, (ii) imaging disease progression in mouse models of colitis and colorectal neoplasia, (iii) visualization of human tissues from subjects with celiac sprue, IBD, and colorectal neoplasia, and (iv) in vivo imaging of the normal rat colon.
Materials and Methods
Procurement of mouse tissues for generating the normal GI atlas
The esophagus, stomach, duodenum, ileum, and colon were harvested from 8-week-old C57BL6/J male mice (The Jackson Laboratory). All tissues were flushed with ice-cold PBS, cut open longitudinally and immediately imaged by MPM. Organs of the hepatobiliary system and pancreas were harvested and immediately imaged. Following imaging, tissues were fixed in 10% formalin overnight then paraffin embedded, sectioned, and stained with hematoxylin and eosin (H&E).
Dextran sodium sulfate-induced colitis
For studies of acute inflammation, 8-week-old male C57BL6/J mice were administered 2% dextran sodium sulfate (DSS; MP Biochemical) dissolved in drinking water for 1, 3, or 7 days (n = 3/time point). For studies of chronic inflammation, mice were given 2% DSS for 7 days, followed by 7 or 14 days of plain drinking water (n = 3/time point). A separate control group of mice was given plain drinking water for the indicated experimental periods (n = 2/time point). This experimental design is depicted in Supplementary Fig. S1A. All mice were maintained on 5053 rodent diet 20 (PMI Nutrition, Inc.). At the end of the experimental period, mice were sacrificed and colons harvested, flushed with ice-cold PBS, cut open longitudinally, and immediately imaged by MPM. Following imaging, colons were fixed in 10% formalin overnight then Swiss-rolled and paraffin-embedded, sectioned, and stained with H&E.
Azoxymethane-induced colorectal carcinogenesis
Male A/J mice were placed on AIN93G purified diet (Research Diets, Inc.) and given 6 weekly intraperitoneal injections of azoxymethane (AOM; 10 mg/kg; Sigma Aldrich) or 0.9% saline beginning at 5 weeks of age. At 1, 3, or 7 weeks after the last injection, mice were sacrificed and colons were harvested and treated as described above. Three mice from the AOM-injected group and 2 mice from the saline-injected group were imaged at each time point. The experimental design is shown in Supplementary Fig. S1B. The animal studies were approved by the Institutional Animal Care and Use Committee at Weill Cornell Medical College.
Procurement and imaging of human samples
Biopsy specimens from subjects with celiac sprue, Crohn's disease, ulcerative colitis, and their respective normal controls were collected during endoscopy. Immediately after biopsy, samples were placed in PBS on ice then imaged by MPM within 1 hour. Adenoma and adenocarcinoma specimens were collected from patients who underwent colectomy. After removal, colons were bisected in surgical pathology then imaged by MPM within 1 hour. After MPM imaging, biopsy and colectomy specimens were formalin-fixed, paraffin-embedded, sectioned, and stained by H&E. Stained slides were evaluated by the attending pathology staff at New York-Presbyterian Hospital/Weill Cornell Medical College and reported using the World Health Organization guidelines (33). The study was approved by the Weill Cornell Medical College Institutional Review Board and all subjects provided written informed consent for participation.
Olympus FluoView FV1000 MPE imaging system was used for all MPM imaging, except the colectomy specimens, for which a home-built MPM system, described in detail previously (17), was used. Each specimen was placed at the center of a glass-bottomed 35-mm dish or a glass slide, with the luminal surface facing up, hydrated with PBS, and overlaid with a coverslip. The whole dish was then placed under the objective of the upright microscope for imaging. Specimens were imaged first at low magnification using a 4 ×/0.28NA dry objective, and then at high magnification using a 25 ×/1.05NA or a 20 ×/0.95NA water immersion objective. Specimens were excited using 780 nm light from a tunable femtosecond pulsed Ti-Sapphire laser (Mai Tai DeepSee, Spectra-Physics, Newport Corporation). Three distinct intrinsic tissue emission signals were collected using photomultiplier tubes in nondescanned configuration: (i) SHG (360–400 nm, color coded red), a nonlinear scattering signal originating from tissue collagen, (ii) short wavelength autofluorescence (420–490 nm, color coded green), originating in part from reduced NADH and flavin adenine dinucleotide in cells, and elastin in the connective tissue, and (iii) long wavelength autofluorescence (550–650 nm, color coded blue), originating in part from lipofuscin in macrophages, and from mucin and bile. The signals in individual channels were collected as separate grayscale images, color-coded and merged to produce the final image. Minor adjustments of brightness and color balance were carried out using Adobe Photoshop CS4.
For in vivo imaging, a prototype assembled at Cornell University, Ithaca (34; Fig 6A) was used. Briefly, the device consists of a miniaturized cantilever fiber raster scanner and a 0.8 numerical aperture gradient index (GRIN) lens, which are packaged into waterproof stainless steel housing. The entire rigid length of the prototype is 4 cm, with an outer diameter of 3 mm. Beyond this 4 cm rigid length, the rest of the device is flexible. The system is able to obtain raster scanned 512 × 512 pixel images, with a field of view of 115 × 115 μm, at 4.1 frames per second. The lateral and axial resolutions are 0.8 and 10 μm, respectively, which are comparable with the state-of-the-art commercial bench-top systems. For in vivo imaging (35), we used adult male Sprague Dawley rats (Charles River Laboratories International), because they are significantly larger, and thus more appropriate for imaging with the human endoscope prototype as compared with mice. To conduct imaging, rats were placed under isofluorane anesthesia and a laparotomy was conducted using a standard ventral midline abdominal approach. The colon was exteriorized, a section was placed on a flat platform and a small incision was made to expose the colonic mucosa and allow for the insertion of the endoscope. The colonic mucosa was visualized using femtosecond pulsed 800 nm light, and the endoscope was focused 20 to 30 μm below the mucosal surface (Fig. 6B–C). The resultant emission was collected in 2 photomultiplier channels, one collecting SHG (≤405 nm) and another collecting broadband autofluorescence (≥405 nm). Images were acquired for several seconds at a time, and collected as movies. One of the frames of the movie was later extracted, color-coded green, and minimally processed (adjustment of brightness and contrast, and Gaussian smoothing to remove single pixel noise). To compare the performance of endoscopic imaging with that of the commercial bench-top Olympus FV1000MPE system, a section of colon was excised from an adult male Sprague Dawley rat after sacrifice, flushed with PBS, cut open longitudinally, and imaged using the commercial system under comparable imaging parameters. This tissue was subsequently fixed in formalin, and processed for routine H&E histology, as described above.
The 3 study pathologists first reviewed H&E-stained slides from the DSS and AOM models along with corresponding control tissues, to familiarize themselves with the morphologic changes at various time points in the disease. Subsequently, each pathologist was independently shown representative MPM images of the same specimens in a blinded fashion. For the DSS model, each pathologist scored the morphologic changes in the following categories: (i) crypt architecture (normal, atrophic, regenerative, or dysplastic), (ii) presence or absence of inflammation, and (iii) ulceration (n = 3 mice/group). On the basis of these morphologic changes, each image was assigned either to the acute or the resolving phase of colitis. For the AOM model, pathologists were asked to identify the following lesions: (i) aberrant crypts, (ii) aberrant crypt foci (ACF), (iii) microadenoma, and (iv) adenoma (n = 3 mice/group). They also noted the degree of dysplasia (low or high) for each lesion.
For histopathologic analysis of human specimens, MPM images of normal biopsies from the duodenum, terminal ileum, and left colon were shown to 2 study pathologists. These locations were specifically chosen as controls for the diseased specimens, which came from these parts of the GI tract. The pathologists were then shown MPM images from the suspected abnormal lesions in a blinded fashion, and asked to categorize them as lesions of malabsorption syndrome (celiac sprue) or IBD. These results were then compared with the diagnosis given on H&E-stained sections by the attending pathologists. The neoplastic lesions (adenomas and adenocarcinomas of the colon) were not included in the blinded analysis of MPM images because these images were generated from intact colectomy specimens and it could not be confirmed that H&E-stained slides used for clinical diagnosis were prepared from precisely the same locations as those imaged by MPM.
MPM generates images of the mouse GI tract that are comparable to H&E-stained tissue
To determine whether MPM is useful for identifying pathology in the GI tract, we first created an atlas of the entire normal mouse GI tract, visualized by both MPM, and concordant H&E staining. Beginning with the esophagus, Fig. 1A and B shows typical morphology including keratinized stratified squamous epithelium lining the mucosa, as well as collagen bundles and elastin fibers within the submucosa. Next, the stomach was visualized showing gastric glands and connective tissue (Fig. 1C and D). Supplementary Fig. S2A–S2D shows images of the gastroesophageal junction. The small intestine was next imaged including the duodenum and ileum, which showed villi lined by enterocytes with interspersed goblet cells (Fig. 1E–H). MPM images of the ileum showed microvilli, which were only visible by MPM (Fig. 1G, inset). Finally, we imaged the cecum and proximal colon which showed surface epithelium and crypts lined by columnar and goblet cells (Fig. 1I–L). Supplementary Fig. S2E–S2H shows images of the ano-rectal junction. To complement these images of the tubular GI tract, the hepatobiliary system, including the gall bladder, liver, and pancreas were also imaged (Supplementary Fig. S3).
MPM can visualize stages of DSS-induced colonic injury
Studies were next undertaken to determine the utility of MPM to visualize colonic injury at different stages of experimental colitis. Colons from untreated mice had intact surface epithelium and normal appearing crypts (Fig. 2A–D). Neither MPM nor H&E revealed significant morphologic changes after 1 or 3 days of DSS exposure. However, in mice given 7 days of DSS, we observed morphologic changes associated with acute colitis including focal ulceration, atrophic crypts in areas of dropout, edema, and an inflammatory cell infiltrate in the lamina propria (Fig. 2E–H). Supplementary video SV1 highlights the 3-dimensional imaging capability of MPM, which allows for better appreciation of histology in various tissue planes. Next, we visualized the resolving phase of colitis in mice administered DSS for 7 days followed by 7 days of plain drinking water (Fig. 2I–L). These images showed enlarged regenerative crypts within the mucosa and increased mononuclear inflammatory infiltrate in the lamina propria (Fig. 2I–L). Similar effects were seen 14 days after stopping DSS exposure.
The progression of colorectal carcinogenesis can be observed by MPM
To extend our preclinical studies of GI pathology, we next imaged the progression of experimental colorectal neoplasia in AOM-administered mice. Colons from saline-administered control mice had normal appearing crypt architecture and easily identifiable lymphoid follicles (Fig. 3A–D). In contrast, in AOM-administered mice, we observed ACF composed of hyperplastic and/or dysplastic crypts (Fig. 3E–H) as well as microadenomas, 1 week after the last of 6 weekly injections. Three weeks after the last AOM injection, a larger number of ACF with increasing degrees of dysplasia and more frequent microadenomas were found (Fig. 3I–L). At the final time point (7 weeks following the last injection), in addition to these early lesions, we also found macroadenomas lining the colonic mucosa with low- and high-grade dysplasia (Fig. 3M–P).
Human GI disease can be visualized with MPM
We next determined whether our examination of GI pathology in mouse models could be translated to human disease. Therefore, we first examined endoscopic biopsies from small intestinal diseases that predispose to malignancy including celiac sprue and Crohn's Disease. Duodenal biopsies from healthy subjects showed normal histology by MPM and H&E with villi lined by enterocytes and goblet cells (Fig. 4A–D). In contrast, duodenal biopsies from a celiac sprue subject revealed blunting of villi with flattened mucosal surface and lamina propria with increased lymphoplasmacytic infiltrate (Fig. 4E–H).
Next, we compared tissues from subjects with Crohn's Disease versus healthy controls. Figure 4I–L shows MPM and H&E images taken from the terminal ileum of a healthy subject showing normal histology. The biopsy taken from a subject previously diagnosed with Crohn's Disease however, showed features of chronic active ileitis with blunted villi and focal ulcerations containing inflammatory cells (Fig. 4M–P).
We extended our examination of human GI disease to visualize abnormalities of the colon. Figure 5 shows images of endoscopic biopsies from the colons of a healthy subject showing normal-appearing crypts (Fig. 5A–D) versus clinically diagnosed chronic active ulcerative colitis (Fig. 5E–H). In the ulcerative colitis specimen, pseudopolyp formation, distortion, and loss of crypts, and an inflammatory cell infiltrate are seen (Fig. 5E–H).
Finally, human colectomy specimens from subjects with colorectal neoplasia were imaged. Figure 5I–L shows a tubular adenoma with crowded glands lined by cells with elongated nuclei indicative of low-grade dysplasia. Figure 5M–P shows a well differentiated adenocarcinoma with complex (back-to-back) arrangement of the tubular glands and cytologic features of dysplasia. Taken together, the results obtained in both experimental models and humans indicate that MPM can be used to visualize GI diseases that predispose to cancer.
MPM-generated images allow for accurate pathologic diagnoses
We next determined whether images generated by MPM alone could be used to identify and categorize GI diseases. H&E-stained sections from the same specimens imaged by MPM were used to confirm diagnoses. For mouse studies, colonic inflammatory lesions from mice exposed to DSS (n = 3 mice/time point) were shown to 3 pathologists in a blinded fashion. Each viewer was able to identify normal, atrophic and regenerative crypts, and regions of ulceration and inflammation with 100% concordance with H&E-stained slides, as well as with 100% inter-observer agreement. Neoplastic lesions of the colon from AOM-exposed mice (n = 3 mice/time point) were shown to pathologists to identify dysplastic crypts, ACF, microadenomas, and macroadenomas in a blinded manner with 100% concordance. For both the DSS and AOM models, 3 pathologists were shown one low-magnification image and 3 to 6 high-magnification images generated by MPM. The low-magnification images were tiles covering a relatively large area (10s of mm) in an effort to capture the heterogeneity of histologic features seen at various stages of disease. High-magnification images focused on structures of diagnostic importance.
For human samples, MPM images from clinically diagnosed subjects with Crohn's disease (n = 4), ulcerative colitis (n = 2), or celiac sprue (n = 2) were examined by study pathologists. Images from IBD specimens (Crohn's disease or ulcerative colitis), were diagnosed as chronic active ileitis or chronic active colitis and samples from sprue cases were identified as such by MPM. These diagnoses were identical to the final pathology report given by attending pathologists based on H&E-stained slides prepared from the same specimens. As stated in the Methods section, neoplastic lesions from human samples were not included in the blinded analysis of MPM images as these images were generated from intact colectomy specimens and it could not be confirmed that H&E-stained slides used for clinical diagnosis were prepared from precisely the same locations as those imaged by MPM.
Imaging of colonic mucosa in vivo using a flexible MPM endoscope
We next assessed our ability to translate the ex vivo visualization of the animal and human GI tracts to in vivo. To this end, we used a miniaturized MPM endoscope prototype assembled at Cornell University, Ithaca (34, 35; Fig 6A–C). Imaging of the exposed colon in a live rat using this device showed all significant histologic features including crypts lined by enterocytes and goblet cells (Fig. 6D). To compare the imaging capability of the prototype endoscope to the standard commercial bench-top system, excised rat colon tissue was also visualized using the bench-top microscope (Fig. 6E). Both systems showed striking similarity in their ability to reveal histologic features (Fig. 6D and E). Standard H&E staining was also conducted on the same tissue (Fig. 6F).
The overall goal of this study was to assess the potential of MPM imaging for real-time, high-resolution contrast-free visualization of both the normal GI tract and GI diseases linked to malignancy. To accomplish this, an MPM atlas of images of the entire normal mouse GI tract from the esophagus to the anus, including the organs of the hepatobiliary system and pancreas was generated. Next, mouse models of experimental colitis and colorectal neoplasia were imaged. MPM imaging was then used to visualize tissues from several diseases linked to GI malignancy including celiac sprue, IBD, and colonic neoplasia. MPM images generated from tissues of both experimental models and human subjects recapitulated all essential aspects of histology found in H&E-stained sections. Finally, the feasibility of carrying out MPM imaging of the GI tract in vivo was established.
Previously, MPM imaging has been used to visualize the normal and diseased mouse GI tract (23, 24, 29, 30, 36). For example, DSS-induced colitis was imaged in green fluorescent protein transgenic mice (23, 24). Although these studies showed clear morphologic changes in the colon as a result of DSS exposure, exogenous fluorescence was used to visualize these changes. Importantly, our study shows the ability of MPM to use endogenous tissue emission signals alone to image similar pathologic changes. We posit that such imaging will be easier to translate to a clinical setting for future real-time endoscopic evaluation of tissue. Two-photon microscopy has also been used to show the aberrant morphology of a small intestinal polyp in APCMin/+ mice (30). Our study builds upon this work and uses MPM to show the progression of morphologic changes during colorectal carcinogenesis. One potential implication is that MPM-derived images may prove useful for determining the ability of chemopreventive agents to inhibit colorectal carcinogenesis.
After establishing the MPM signatures of colitis and colorectal neoplasia in the mouse model, we investigated whether similar changes could be identified in human disease. In each case, we found that MPM can reliably identify all relevant tissue substructures in the normal as well as the diseased human GI tract. Rogart and colleagues have used MPM to assess ex vivo specimens from several parts of the normal human GI mucosa identifying structures such as epithelial cells, goblet cells, and interstitial fibers (18). In addition, human esophageal, gastric, and colon cancers have been previously investigated with MPM (18, 25, 26, 28, 30). To our knowledge, the current study provides the first evidence that MPM can be used to visualize and delineate morphologic changes associated with celiac sprue, Crohn's ileitis, and ulcerative colitis. Importantly, pathologists were able to use MPM to accurately diagnose each of these diseases that predispose to malignancy.
Confocal microendoscopy and OCT are 2 competing technologies for generating “optical biopsies” of diseased tissue. Endoscopes equipped with confocal imaging capabilities have been shown to be useful for detecting neoplastic changes in ulcerative colitis and Barrett's esophagus (37, 38). In another recent human study involving confocal microendoscopy, dysplastic colonocytes were shown to be preferentially labeled with a fluorescein-conjugated heptapeptide (39). However, in spite of the great promise, current clinical confocal systems suffer from several significant limitations including the need for exogenous contrast administration, shallow depth of imaging, and only being able to detect the exogenously administered contrast agent. In vivo colonoscopic OCT has been used in clinical trials to distinguish between colonic polyps and normal mucosa and between ulcerative colitis and Crohn's disease (40, 41). The advantage of this technology is deeper tissue penetration (up to 2 mm) relative to both confocal and MPM, allowing access to the submucosa. However, the lateral resolution of current clinical systems (∼10 μ) is insufficient to obtain the cellular and subcellular resolution that is necessary to distinguish between benign lesions and early stage neoplasms. Furthermore, unlike MPM, OCT does not generate distinct signals from specific tissue components. Thus, MPM imaging may overcome some of the limitations associated with confocal microendoscopy and OCT.
There are potential real-time clinical applications for MPM technology in human disease. In fact, a commercial MPM system (DermaInspect, JenLab) is currently approved in Europe for diagnosis of dermal neoplasms (42, 43). Furthermore, rigid microprobe objectives for use with bench-top MPM systems are commercially available but have only been used to image tissues ex vivo (18). Next generation technologies where MPM devices are miniaturized for use in rigid and flexible endoscopes, including the prototype used in this study, are currently under development in several laboratories (34, 44–50). Future uses for a rigid probe could include examination of the oral and vaginal mucosa and the anal/perianal region to identify dysplastic changes and exclude malignant lesions. Rigid probes could also be useful for evaluating surgical margins. Flexible endoscopes (similar to the one used in this study) could be used during colonoscopy and upper GI endoscopy to diagnose polyps and other suspicious regions within the mucosa as benign or malignant, thereby potentially minimizing biopsies of benign tissues. It is likely that using this technology will save time and reduce both patient morbidity and cost. Given the necessarily small field-of-view of any high-resolution optical biopsy technique, we visualize their role in the near future primarily as a means to confirm the identities of suspicious lesions found with low-resolution screening techniques, such as high-definition white light endoscopy, narrow band imaging, or chromoendoscopy. In summary, our findings show the utility of MPM for imaging inflammatory and neoplastic lesions of the GI tract and support the continued development of MPM. Taken together, MPM may enhance the identification of GI pathology in situ during an endoscopic or surgical procedure, and thereby play a future role in preventing the development and progression of disease.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: T. Makino, D.C. Montrose, J.W. Milsom, C. Xu, A.J. Dannenberg, S. Mukherjee
Development of methodology: T. Makino, D.C. Montrose, J. Sterling, N. Zhang, C. Xu, S. Mukherjee
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Makino, M. Jain, A. Aggarwal, J. Sterling, B. Bosworth, M.M. Shevchuk, C.M. Brown, D.R. Rivera, W.O. Williams, C. Xu, A.J. Dannenberg, S. Mukherjee
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Makino, M. Jain, D.C. Montrose, J. Sterling, B.D. Robinson, M.M. Shevchuk, K. Kawaguchi, C.M. Brown, D.R. Rivera, S. Mukherjee
Writing, review, and/or revision of the manuscript: T. Makino, M. Jain, D.C. Montrose, B. Bosworth, J.W. Milsom, B.D. Robinson, M.M. Shevchuk, K. Kawaguchi, C.M. Brown, D.R. Rivera, A.J. Dannenberg, S. Mukherjee
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Makino, M.M. Shevchuk
Study supervision: M. Jain, C. Xu, A.J. Dannenberg, S. Mukherjee
The authors acknowledge Dr. Parul Shukla, GI surgeon, for many useful discussions regarding the wider implications of MPM in surgical management of GI diseases, and Dr. Nicole Panarelli, GI pathologist, for review of some H&E slides which had ambiguous diagnoses.
This project was supported by the New York Crohn's Foundation (A.J. Dannenberg), The Uehara Memorial Foundation (T. Makino), NIH T32 CA062948 (D.C. Montrose), and NIH/NIBIB R01-EB006736 and NIH/NCI R01-CA133148 (C. Xu).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.