Confocal Endomicroscopy For In Vivo Diagnosis Of Gastric Intraepithelial Neoplasia And Molecular Imaging Of Gastric Cancer

Backgrounds and aimsGastric cancer remains the world's second leading cause of cancer-related death. In1988, the multistep carcinogenesis process of gastric carcinoma was proposed by Correa, and it was described as atrophic gastritis-intestinal metaplasia-dysplasia-intestinal type gastric cancer. This theory has been validated by many follow-up clinical trials todate. Therefore, atrophic gastritis, intestinal metaplasia, and dysplasia are all widely considered as premalinant lesions of the gastric mucosa. Nevertheless, as the penultimate step for the development of intestinal type gastric cancer, gastric dysplasia is the most threatening precancerous changes. Thus the early diagnosis of gastric dysplasia is significant for the prevention and treatment of gastric cancer. In2010, the World Health Organization suggested to use the nomenclature of "intraepithelial neoplasia" to replace "dysplasia".Endoscopy is the most efficient tool for early detection of gastric neoplasia. Modern endoscopic devices such as high-definition endoscopy, magnifying endoscopy, chromoendoscopy and narrow-band imaging (NBI), can provide high resolution endoscopic pictures of color changes and minute structure alterations of gastric mucosa. However, although these endoscopic modalities can improve the detection rate of early gastric carcinoma on a certain extent, they can't enable microscopic observation of epithelial cells, gland architecture and vasculature. Therefore, multiple biopsies are still needed to diagnose endoscopic suspicious lesions, and the biopsy scars may interefere with the following short-term endoscopic therapy of these lesions if diagnosed as gastric intraepithelial neoplasia (GIN) by histopathology. Moreover, the histological processing of biopsy specimens usually takes of2-3days, which not only postpones therapeutic opportunity for patients with high-grade intraepithelial neoplasia (HGIN) and early gastric cancer, but also severes the patients'psychological and financial burdens.Confocal laser endomicroscopy (CLE) is a novel endoscopic technology which can provide real-time cellular and subcellular visualization of gastric mucosa with the aid of intravenous surface application of fluorescent dyes, while at the same time displaying white-light endoscopic images. Previous studies found that CLE is a powerful new tool for early endoscopic diagnosis of intestinal metaplasia and gastric cancer. However, no investigation has yet applied CLE in the diagnosis of GIN.Early diagnosis is essential for improving the prognosis of gastric cancer. Currently, although endoscopic biopsy is the gold standard for diagnosing gastric cancer, it still suffers from several drawbacks such as lesions'morphological alteration, bleeding and perforation. What's more, the delayed histopathological diagnosis will lead to later therapy and increased anxiety of patients. Thus a new endoscopic device is needed which can enable immediate histopathological diagnosis of suspicious lesions during white-light endoscopic examination. Several studies have been investigating the role of CLE in diagnosing gastric cancer in vivo, which showed that according to predefined endomicroscopic diagnostic criteria, gastric cancer can be accurately differentiated from non-cancerous lesions using CLE. However, all those studies applied intravenous fluorescein sodium as the contrast agent to obtain nonspecific endomicroscopic images, which cannot provide functional or molecular imaging of the gastric mucosa. Most recently, several researches revealed that using fluorescently labeled specific antibodies, CLE can achieve intravital molecular imaging of GI cancer. Therefore, the aims of this study were as follows:(1) To create the endomicroscopic classification of gastric lesions, and to evaluate the diagnostic value of CLE for the identification and grading of GIN prospectively;(2) To evaluate the feasibility of CLE for molecular imaging of gastric cancer in vivo.MethodsPart1:The establishment and evaluation of the endomicroscopic diagnostic criteria for GINThis part was divided into two steps. In the first step, CLE was used to examine histologically confirmed normal gastric mucosa, nonneoplastic and neoplastic gastric lesions. CLE images were compared with horizontal sectioned histopathological pictures obtained from the same gastric lesions so as to establish the endomicroscopic classification for nonneoplastic and GIN lesions and the endomicroscpic scoring system for grading GIN. In the second step, consecutive patients were prospectively recruited according to predefined inclusion and exclusion criteria from July2009to January2010in Qilu Hospital. In vivo real-time CLE diagnosis was made for endoscopic suspicious lesions by the performing endoscopist who was blinded to the patients'previous histological diagnosis. CLE images of each lesion were stored in a specific folder, and were reevaluated after the endoscopy by another CLE investigator blinded to the patients'clinical history and endoscopic information. In addition, a post hoc assessment of inter-observer and intra-observer agreements for the CLE findings was performed according to the following protocol. A data set containing50confocal images of medium depth from50enrolled subjects were randomly selected and displayed to3independent endoscopists and one GI pathologist in a blinded fashion. To evaluate the intra-observer agreement, one investigator (R.J) reassessed these50pictures after a7days interval. Moreover, in an attempt to assess the feasibility of CLE for the differentiation between LGIN and HGIN, one confocal image of medium depth from each CLE diagnosed GIN lesion was further evaluated by an investigator according to the predefined scoring system in a blinded fashion.Part2:Confocal laser endomicroscopy for in vivo molecular imaging of gastric cancerThis part was divided into3steps, including tumor cell line characterization, experiments of tumor-bearing mice, and human studies.(1) The gastric cancer cell lines BGC823and SGC7901were incubated with AF488-MG7, isotype control antibody, or PBS respectively. Flowcytometry was firstly carried out, followed by CLE (FIVE1, Optiscan) and fluorescent microscopy. Finally, celluar immunohistochemistry was performed in both gastric cancer cell lines.(2) Xenograft tumors were induced using the gastric cancer cell lines BGC823and SGC7901in BALB/c nu/nu mice. In order to evaluate the optimate imaging time and antibody concentration, full-body fluorescent imaging was firstly carried out using a cooled CCD camera (Kodak2000MM, Kodak Company, USA) after intracardiac administration of AF680-labelled MG7or isotype control antibody. Intravital CLE examination of xenograft tumors was then performed according to the results of full-body fluorescent imaging.(3) Surgical specimens (5-10mm) or endoscopic biopsies of gastric cancerous tissue and non-cancerous gastric tissue were incubated with AF488-labelled MG7antibody, and imaged using CLE (FIVE, Optiscan) immediately. The CLE investigator was blinded to the patients'previous histopathological diagnosis of gastric lesions. MG7-specific fluorescence signal in CLE images was ranked as negative (0), weakly positive (+), moderately positive (++) or strong expression (+++). Then3experienced CLE investigators who were not involved in data collection were invited to give the semi-quantitative grading of MG7-Ag expression of all human specimens observed by CLE probe independently. H&E staing and immunohistochemistry were performed for all the specimens after CLE examinations. Spearman's correlation coefficient was also applied to assess the semi-quantitative results between CLE and IHC findings.ResultsPart1:In total,15093confocal images were obtained from91macroscopic lesions in the75patients. Using the prior described CLE diagnostic criteria for gastric lesions, the sensitivity, specificity, PPV, NPV, PLR and NLR of final CLE diagnosis of GIN were77.8%(95%CI,63.7%-87.5%),81.8%(95%CI,68.0%-90.5%),81.4%(95%CI,67.4%-90.3%),78.3%(95%CI,64.4%-87.7%),4.28(95%CI,2.24-8.16) and0.27(95%CI,0.16-0.48) respectively. There was no statistical difference between preliminary and final CLE diagnostic accuracy for GIN (P=0.549). The inter-observer agreement was substantial for the diagnosis of GIN among3endoscopists (mean kappa=0.70). We also investigated the inter-observer agreement between endoscopist and GI pathologist, and the mean kappa value was0.71. Intra-observer agreement was also graded as substantial (kappa=0.78). The intraepithelial neoplasia score≥5had a sensitivity of66.7%and a specificity of88.5%in discriminating HGIN from LGIN.Part2:(1) The MG7-Ag expression was confirmed in the two gastric cancer cell lines BGC823and SGC7901by using Flow Cytometry. CLE (FIVE1, Optiscan) imaging showed membranous and cytoplasmic specific fluorescent signal. The immediate bench top fluorescence microscopic observation was similar to CLE images. IHC results also demonstrated specific staining of both tumor cell lines in the cell membrane and cytoplasm.(2) In the preliminary study of full-body imaging of tumor-bearing mouse, specific fluorescent signal was observed at24h after intracardiac injection of AF680-labelled MG7antibody (1μg/g body weight). In addition, the tumor fluorescence intensity reached its maximum at48h after antibody administration. Forty-eight hours after AF488-labelled MG7antibody injection, specific fluorescent signal could be visualized in both BGC823and SGC7901tumors; Ex vivo IHC results and fluorescent microscopic imaging of cryosections resembled in vivo endomicroscopic findings.(3) A total of46samples, including18surgical specimens and28biopsy specimens from23patients with gastric cancer, were analyzed prospectively. Specific signals (+to+++) were found in22/23samples diagnosed as gastric cancer, whereas non-cancerous samples revealed no or only weak (0or+, n=18, n=5) fluorescent signal. The correlation between different grades of IHC results of human samples and semi-quantitative assessment of in vivo MG7-Ag expression using CLE was good (Spearman's r=0.87, P＜0.001). The interobserver agreement of the semi-quantitative assessment scheme of CLE images was also excellent (ICC=0.883,95%CI0.819-0.929).Conclusions1. GIN, non-neoplastic gastric lesions, and normal gastric mucosa can be differenciated according to characteristic changes of gland architecture, cell morphology, and vassel architecture in CLE images.2. HGIN can be differenciated from LGIN using CLE according to the predefined endomicroscopic scoring system.3. The interobserver and intraobserver agreement are all substantial for the diagnosis of GIN using CLE, indicating the diagnostic criteria in this study is reliable.4. Using fluorescently labeled MG7antibody, CLE can achive targeted molecular imaging in tumor-bearing mouse in vivo and in the fresh tissue of patients with gastric cancer; there is a good correlation between CLE fluorescent imaging and ex vivo immunohistochemistry results. SignificanceSimilar to the case with histopathologic features, distinct glandular derangements, cellular abnormalities, and microvascular alterations are identifiable for the diagnosis of GIN using CLE. The established endomicroscopic classification for gastric lesions can differentiate GIN lesions from non-neoplastic ones with a substantial inter-and intraobserver agreement, which indicates a good reliability of it. Moreover, HGIN can be differentiated from LGIN with a high specificity using CLE, thus endoscopists can make spot-decision regarding whether to perform endoscopic resection or otherwise. With the aid of fluorescently labeled specific antibody targeting gastric cancer, CLE is able to diagnose gastric cancer in the molecular level in vivo. This technique may have provided us with enhanced diagnosis of gastric cancer and accurate selection of MG7-Ag-positive patients. Although in vivo human studies are needed to further investigate its clinical role, data of the present study verified the feasibility of CLE mediated immunodiagnosis of gastric cancer from bench to bedside.