Biomarker identified for predicting inflammatory bowel disease treatment success

Inflammatory bowel disease (IBD) takes multiple forms, including Crohn’s disease and ulcerative colitis. It is caused by an overactive immune response in the gastrointestinal tract. People with the disease experience abdominal cramping, diarrhea, and fatigue. There is no cure; so far, the only treatment has been to alleviate symptoms and control inflammation.

“As a clinical scientist, I am actively involved in patient care,” says Prof. Ahmed Hegazy from the Department of Gastroenterology, Infectious Diseases and Rheumatology at Charité.

“This disease involves episodes known as flares, which are often unpredictable, so the treatment is constantly being adjusted. So far, it has not been possible to predict the individual course of the disease or how patients will respond to various treatment options, which is what makes treatment so challenging.”

One highly effective treatment option with only minor side effects is known as integrin-blocking therapy. It prevents certain immune cells from entering the gastrointestinal tract and triggering inflammatory processes there. Vedolizumab, a specific antibody to a specific integrin, has a blocking effect: It binds to T helper cells, keeping them from entering the gastrointestinal tract.

“Integrin-blocking therapy is highly effective in about two-thirds of patients. But for the other third, it doesn’t work at all. Previously, figuring out who would respond to the treatment was a matter of trial and error. That is tedious, time-consuming, and costly, plus it is often quite frustrating for patients,” Hegazy says.

“It would be helpful to have a biomarker that can show in advance whether or not the treatment is promising. That’s exactly what we set out to find with our study.”

Machine learning helps with pattern recognition

The researchers’ extensive studies were based on 47 patients with chronic IBD. Blood samples were taken before they started treatment with vedolizumab and six weeks after initiation of treatment. The researchers used advanced analytical methods such as mass cytometry, single-cell RNA sequencing, and serum proteomics to examine the samples.

“We zeroed in on different kinds of immune cells and certain proteins and looked for potential changes caused by the treatment,” Hegazy explains. “This generated extensive data, which we then analyzed using machine learning.

“Machine learning is a field of artificial intelligence that uses algorithms and statistical models to allow computers to learn from data and recognize patterns without needing to be explicitly programmed in advance. This allowed us to identify patterns that help to predict which patients are more likely to respond to this form of treatment.”

The interdisciplinary team made up of researchers from the fields of medicine, bioinformatics, mathematics and biology, which also included researchers from the Berlin Institute of Health at Charité (BIH), the German Rheumatism Research Centre Berlin (DRFZ), and the University of Bonn, identified the same patterns in studies of another group of patients. That group of 26 participants helped the researchers to validate the results of their study.

High biomarker levels, low treatment response

One especially meaningful molecule was a cell division protein called Ki67, which is produced at elevated levels when T helper cells divide. Patients with high levels of these cells in their blood prior to treatment did not respond to vedolizumab.

“We were able to figure out the molecular phenomenon behind this: These T helper cells do not have binding sites for vedolizumab, so they are able to pass unimpeded into the gastrointestinal tract and continue to contribute to inflammation,” Hegazy explains.

“These cells have different trafficking molecules that allow them to move into the gastrointestinal tract. This makes Ki67 a good indicator of the presence of vedolizumab-resistant T helper cells.”

Moving into clinical practice

The researchers plan to verify their findings through large multicenter studies and study the reliability of the biomarker they have identified in detail. Plans also call for further developing their detection and measurement methods so they can be incorporated into routine clinical practice.

“Reliable biomarkers are the key to individualized therapy, and thus better treatment, for our patients with chronic inflammatory bowel disease,” says Prof. Britta Siegmund, the department’s director.

Once that condition is met, decisions regarding the right individual form of treatment can be made faster and with greater accuracy. And that represents a step toward personalized medicine that will bring clarity for patients early on.