Predictive endoscopy: Form follows function

Advances in endoscopic imaging have extended our traditional role from diagnostic endoscopists to therapeutic endoscopists and now to prognostic endoscopists. Endoscopes were historically developed to allow flexible, reasonably comfortable access to the lumen of the gastrointestinal tract. The role was to inspect, often to biopsy for ex vivo microscopy, and to treat lesions such as polyps, ulcers, pancreatobiliary obstructions, and more recently flat mucosal neoplasia and even submucosal diseases.

The role of prediction has been largely relegated to the pathologist using morphological (tumor size, stage, grade, degree/type of inflammation) and immunohistochemical patterns. Characterizing molecular changes in tissues has traditionally been the purview of the basic scientist. However, an emerging era of molecular medicine, with characterization of somatic and germline genomic and downstream changes, has allowed highly precise diagnostic, prognostic, and predictive algorithms to enter the clinical arena.

Classic examples include breast cancer with hormonal receptor characterization. In gastrointestinal cancers, ras mutations and epidermal growth factor receptor (EGFR), and c-kit mutations have played an increasing role in the decision to treat patients with targeted chemotherapy. The implications of molecular diagnostics and targeted therapies extend well beyond oncology to inflammatory bowel disease and rheumatoid arthritis among others. As we increasingly target therapies to specific molecules and characterize the molecular features of individual patients’ disease, there is a growing demand to develop methods of predicting which patients (and particularly which disease processes) will respond to which treatments.

In this issue of Endoscopy, Goetz et al. [1] present the first foray into molecular endoscopy as a predication tool for molecular therapy in colorectal cancer. These investigators have taken advantage of the rapid progression in targeted therapies for cancer, in this case by choosing the anti-EGFR monocloncal antibody (MAb) cetuximab. They have attached a fluorescent reporter to the drug, and used commercially available in vivo confocal laser endomicoscopy (CLE) systems to characterize the drug binding to tumor cells in a mouse model of human colorectal cancer, and then shown that binding density is predictive of drug response.

The study is both elegant and clinically relevant. MAbs are increasingly used for targeting chemotherapy, and are highly amenable to imaging as they are specific and can be easily labeled for endoscopic imaging. Essentially all MAbs can be altered to allow in vivo molecular imaging.

In the study, an immunodeficient mouse model was used and implanted with human colorectal cancer cells that did or did not express the EGFR target of cetuximab. Using the fluorescent-labeled cetuximab, the authors then imaged tumors using a CLE probe similar to the one used in standard human CLE studies. They validated several key assumptions in order to support their conclusions: 1) the antibody was binding the intended target, 2) the target was EGFR, 3) the attachment of the fluorescent label to cetuximab itself did not have a chemotherapeutic effect, and 4) binding was associated with standard measures of EGFR staining such as immunohistochemistry.

The most important finding of the study was that in vivo CLE binding was predictive of response to cetuximab therapy as measured by survival, tumor response, and animal measures of wellbeing. A second major finding was that binding of labeled cetuximab can be measured very early on in the treatment process. In the study, the binding density as measure by CLE on Day 0 (day of injection) was predictive of response at Day 30. Furthermore, the test dose of labeled drug used to assess binding density was small. Together, these suggest the potential to give small test doses of a variety of agents (perhaps even multiplexed with different reporter fluorophores) to assess a broad range of chemotherapeutic sensitivities prior to choosing a final drug and full-dose therapy. Overall, the study provides key evidence regarding the feasibility and efficacy of measuring molecular targets and using these to guide therapy.

The general concept of the study does have some limitations. Colon cancers, and less so rectal cancers, require invasive (colonoscopy or proctoscopy) procedures to access the tumor directly. Practically speaking, most patients do not undergo repeat lower endoscopy after the initial diagnosis of cancer is made, with the exception of rectal cancer which is often staged by endorectal ultrasound. Thus, patients would need to undergo repeat procedures to evaluate in vivo molecular targets. Although invasive, this may still be an acceptable procedure if proven to accurately guide therapy.

In the study, the authors used whole animal fluorescent imaging systems to visualize the optical reporter. These systems provide whole body imaging similar to positron emission tomography (PET) scans used commonly in human cancer staging. Although appealing, optical reporters generally have very shallow depths of transmission and thus cannot be visualized in larger animals (such as humans) due to lack of emission of the reporter from the tumor to an extra-corporeal imaging system. Attaching positron emitting reporters is appealing but, to date, these have not been well validated to directly image drug targets.

Another limitation of the study is the single agent model. Most human cancers require treatment with multiple agents for the therapy to be effective. Ideally, a multi-target approach should be evaluated, which could be done either sequentially, or preferably as a multiplexed reporter with different fluorescent color reports for each agent. Finally, the targets should extend beyond MAb therapies to include the many small molecules that are being developed for targeted chemotherapy. There will be fundamental challenges to these as many small molecules are not bound to surface antigens or receptors as in the case of EGFR. It remains to be seen whether molecules that are rapidly internalized into the cell, can be imaged with such reporters. Recent studies using a short peptide and reporters multiplex have shown feasibility to detect adenomatous polyps in the colon after topical application of the agent [2] [3]. Such systems have further opened the potential for molecular imaging to facilitate detection of subtle lesions such as small or flat adenomas and Barrett’s neoplasia [4].

The great American architect, Louis Sullivan, adopted the principal that the form (of buildings) must follow the function of its users [5]. We are now entering an era where our endoscopic imaging can move beyond simply observing forms (tumor size, color, and shape) to predict behavior, and can now measure functions of the cells at a molecular level and use these to guide the treatment of our patients.

• References

• 1 Goetz M, Hoetker MS, Diken M et al. In vivo molecular imaging with cetuximab, an anti-EGFR antibody, for prediction of response in xenograft models of human colorectal cancer. Endoscopy 2013; 45: 469-477
• 2 Joshi BP, Miller SJ, Lee CM et al. Multispectral endoscopic imaging of colorectal dysplasia in vivo. Gastroenterology 2012; 143: 1435-1437
• 3 Joshi BP, Liu Z, Elahi SF et al. Near-infrared-labeled peptide multimer functions as phage mimic for high affinity, specific targeting of colonic adenomas in vivo (with videos). Gastrointest Endosc 2012; 76: 1197-1206 e1-5
• 4 Goetz M. Molecular imaging in GI endoscopy. Gastrointest Endosc 2012; 76: 1207-1209
• 5 Sullivan LH. The tall office building artistically considered. Lipppincott’s Magazine 1896; 57: 403-409