NEW YORK – Purdue University researchers are advancing a new technology that uses light waves to detect chemotherapy resistance in cancer cells into interventional clinical trials following positive results in a human pilot study.
In February, David Nolte, a professor of physics at Purdue, and his colleagues published in Scientific Reports the results of a study of digital holographic optical coherence imaging (DHOCI) as a method to detect treatment resistance in tissue biopsy samples from humans and dogs given chemotherapy.
Approximately half of all cancer patients across disease types and stages do not respond to first-line cytotoxic chemotherapy. Methods that can identify these groups in advance could give non-responders a chance to pursue an alternate course of therapy or save them the side effects and toxicity of an ineffective therapy. In the case of neoadjuvant treatment, sparing patients ineffective chemotherapy prior to surgery could also reduce surgical risks and promote better outcomes.
Nolte first introduced DHOCI, a form of biodynamic imaging, as a technique in Optics Express in 2007. In this approach, a light beam penetrates deep within tumor tissue to measure the movements of cellular machinery based on the Doppler effect of light. The DHOCI system comprises a light source and interferometer, and scattered light is captured by a camera to produce a hologram.
During mitosis, some of the most dramatic movements within the cell occur as the cellular structure is reorganized before dividing. Nolte and his collaborators theorized that the effects of anti-cancer agents could be visualized using DHOCI because those drugs target and arrest cell division. Compared with other types of biodynamic imaging such as 2D and 3D confocal microscopy or optical projection tomography, DHOCI can detect intracellular motion at greater depths in the tissue.
Studying resistance to chemotherapy wasn't his team's "original plan," Nolte said, but instead evolved out of a collaboration with John Turek, a professor of basic medical sciences at Purdue's College of Veterinary Medicine. They found that DHOCI could capture sensitive measurements of movement inside living tissue, allowing them to observe in real time the effects of applying chemotherapy to cells and see differences in responses between tissue types.
The natural next step, Nolte said, was to attempt the same type of study in a biopsy sample from a dog with B-cell lymphoma. However, he and his collaborators were surprised when they didn't see any effects within the cells. When they investigated why they did not see the expected effect, they realized that the dog chosen for the study was highly resistant to chemotherapy. Nolte's team then repeated the same experiment with another dog, which responded well to chemotherapy, and saw a difference in the observable motion within the cells between the chemosensitive dog and the chemoresistant dog.
Nolte and his team subsequently studied 19 dogs with B-cell lymphoma, 13 of which were in remission following chemotherapy and six of which had progressed within 100 days of treatment. The DHOCI signatures collected from the dogs' tumor tissues correlated with their survival times, and Nolte's team developed a chemoresistance classifier using machine learning.
This classifier, based on a set of spectrograms from a single patient, could predict whether that patient would have a longer or shorter progression-free survival for a selected treatment regimen. They validated the classifier using a hold-out procedure in which 18 dogs were included in the training set and the 19thdog was classified, repeating 19 times so every dog was classified. All but three dogs were correctly classified according to their predicted survival time.
Based on these results in a canine model, Nolte and his team then turned to human cancer. His group enrolled 28 patients for a clinical study who presented at the Indiana University School of Medicine Hospital with esophageal adenocarcinoma and who had received carboplatin and paclitaxel or cisplatin and 5-fluorouracil.
Using biopsy samples taken for routine diagnostic purposes, they placed 1-cubic-millimeter sections in the wells of a 96-well plate and kept them alive using buffered growth media. Then they acquired 2,000 digital holograms for each sample at a rate of 25 frames per second. The scans were repeated every 40 minutes for up to 16 hours. The time series of holograms showed the movement of molecules within the cells, and from that, the researchers were able to capture the Doppler light shift from each sample. As with the dog study, they developed a classifier to predict response to treatment and used a hold-out method to validate the results.
Nolte's team found that the biodynamic spectra generated from both the dog and human samples could be grouped roughly into four phenotypes reflecting chemotherapy drug sensitivity across species and cancer type: blue-shifted, red-shifted, mid-frequency enhanced, and mid-frequency suppressed. A Doppler shift toward red, or lower frequency light, represented decreasing intracellular activity over the time of the assay. A blue shift was a sign of increasing metabolic health.
The researchers surmised that the red-shifted samples had compromised health prior to the experiment, possibly due to damage during the biopsy or in sample handling. When those samples were excluded, this approach showed potential to predict the effectiveness of chemotherapy drugs prospectively.
Nolte emphasized also that because the results were consistent in dogs and in humans in different types of cancer, the findings could be broadly applicable to many different types of cancer. "It's not like we would have to start from scratch at every new cancer," he said. "We can build a library or dataset where we can just start including additional cancers, building on what we already know because it looks like this is a general phenomenon that crosses species and diseases."
The next step will be a new clinical study, Nolte said, that will compare a group of patients receiving standard chemotherapy with another group whose treatment will be guided using biodynamic imaging. They are hoping to see a better clinical outcome for the patients in the guided treatment group.
Nolte's team is not currently partnered with any drugmakers as yet because their work has focused on chemotherapy. However, Nolte pointed out that the technology has potential to be developed as a companion diagnostic. He said his group is open to partnering with companies and developing a commercial test based on the technology, but they are primarily pursuing development of holographic biodynamic imaging as a laboratory-developed test. Operating as a reference lab, they would offer a kit that doctors could use to collect samples and keep them alive for analysis, he explained.
Comparing biodynamic imaging to other methods based on genomic profiling, Nolte said that "most of the time, a genetic profile doesn't give you any specific information on how you should treat that patient," whereas holographic biodynamic imaging is a technique with the potential to identify patients who aren't going to benefit from chemotherapy and redirect them to alternate therapies sooner.
"I think we really have a shot at improving cancer care," he added.