A new study published in the journal Nature on March 11, 2020, could transform the currently held views on cancer causation and diagnosis. The researchers have come up with a new technique to detect the presence and type of cancer by looking at microbial DNA circulating in the blood of the person being screened.
Cancer is a disease due to mutations in the genes of the organism affected by it. These mutations result in dysregulation of the cells involved, turning off programs that modulate cell growth and proliferation, or which form tumors.
However, all human cancers aren't caused by mutations in the human genome alone. For instance, pancreatic cancers are most often host to microbes that break down the most commonly used chemotherapy drugs in these patients, as shown by a 2017 study in the journal Science. This led researcher Gregory Poore to pursue the role of microbes in various cancers.
Teaming up with experts across various disciplines, the researchers explored the multi-layered interactions that occur between tumors and the host of microbes that live on and in the human body. Senior researcher Rob Knight said, "Almost all previous cancer research efforts have assumed tumors are sterile environments and ignored the complex interplay human cancer cells may have with the bacteria, viruses and other microbes that live in and on our bodies. The number of microbial genes in our bodies vastly outnumbers the number of human genes, so it shouldn't be surprising that they give us important clues to our health."
Detecting cancer at its earliest stages from a simple blood draw is the goal of several companies currently developing 'liquid biopsies' to detect circulating human tumor DNA. Now UC San Diego researchers have demonstrated they can determine who has cancer, and what type, based on a readout of microbial DNA found in their blood. Image Credit: Szabolcs Borbely / Shutterstock
Microbes associated with tumors
The first step was to examine the data on microbial genes filed in The Cancer Genome Atlas, which is a massive database of tumor-associated genomic data collected from thousands of tumors and maintained by the National Cancer Institute. The researchers say theirs is the first time scientists have attempted to identify microbial DNA from human genomic data on such an impressive scale.
They examined over 18,000 samples of human tumors, taken from almost 10,500 patients with 33 different types of tumors. They tried to find distinctive patterns of microbial genes associated with specific tumors. Some are already known, such as the linkage between HPV (human papillomavirus) and cancer of the cervix, head, and neck, or that of Fusobacterium species with cancers of the gut.
The scientists also picked up many new microbial patterns which were capable of distinguishing one cancer type from another. One example is that of Faecalibacterium, a species that can pick out colon cancer from other tumor types.
The researchers thus arrived at an understanding of how the microbiome of thousands of cancer samples looks. The next step was to use the power of machine learning to see if any of these microbial patterns were specific for cancers of certain types. They trained their machine learning programs to pick up such specific associations and use them to diagnose the type of cancer. They then tested out their hundreds of models and found, to their gratification, that they could identify the type of cancer-based solely on the microbial DNA in the blood.
Early detection of cancers
The next step, for the scientists, was to eliminate advanced cancers from the dataset. The program was still able to distinguish many types of cancer at stages I and II based only on microbial data. The researchers then carried out stringent procedures to decontaminate the bioinformatics data, which means that over 90% of microbial data was removed. Even so, the machine correctly identified many early cancers by type.
Testing it out
In a pilot study, the team of researchers took blood samples from about 100 patients with three types of cancer, including approximately 60, 25, and 15 patients, each with prostate, lung, and melanoma cancer. They used their custom-made tools to ensure the data was as reliable as possible by eliminating contaminating readings. They then took readouts of the microbial profile for each of the blood samples.
Comparing these profiles with each other and with the plasma samples taken from about 70 disease-free volunteers, the team found that their machine learning models could differentiate people with cancer from those without it, in the majority of cases. That is, they could pick up 86% of people with lung cancer while ruling out lung disease correctly in all individuals negative for this condition. They could also distinguish between each of the types of cancer. For instance, they could rightly tell in 81% of cases whether the patient had lung cancer or prostate cancer.
Co-author Sandip Pravin Patel says, "The ability, in a single tube of blood, to have a comprehensive profile of the tumor's DNA (nature) as well as the DNA of the patient's microbiota (nurture), so to speak, is an important step forward in better understanding host-environment interactions in cancer." He explains that it is possible to use this approach to keep track of how the genome of the tumor and of the microbiota change with time, thus using it as not just a diagnostic but a monitoring tool for long-term follow-up of the efficacy of treatment.
He continues, "This could have major implications for the care of cancer patients, and in the early detection of cancer if these results continue to hold up in further testing."
Better than the old?
Patel explains that most cancers today can be diagnosed only by surgical biopsy, which involves removing a piece of the suspected tumor and getting it examined by experts. The diagnosis is based on the molecular markers specific to each type of cancer. The issue with this approach is that it is invasive, takes time, and is expensive.
Recently, liquid biopsies have been developed for many of these cancers. Here a blood sample is used to detect specific mutations in human DNA found in circulation to diagnose specific tumors. While blood markers (mostly molecules found on the surface of tumor cells) are already being used to diagnose and monitor the course of some tumors, genetic markers are relatively new. They are used in some tumors to trace the course of the disease, but their use for tumor diagnosis is not approved by the Food and Drug Administration (FDA).
The reason is the poor accuracy of these tests in discriminating between normal variations in human genes and the true presence of mutations in early cancer. Not only so, says Patel, "they can't pick up cancers where human genomic alterations aren't known or aren't detectable."
This means that liquid biopsies are likely to have a high rate of false-negatives, especially for rare tumors. If the rate of the mutation is low, and the number of shed cells is also low, the chances of picking up the mutation are few indeed. Thus, the patient might well be told there is no evidence of cancer despite the presence of a tumor.
The current test which is based on the detection of microbial DNA rather than tumor DNA, is that it reflects more noticeable changes because, unlike the uniformity of human DNA across all tissues of the human body, microbial DNA profiles vary immensely from tissue to tissue. Thus, rather than hoping to pick up one of the relatively rare changes in human DNA to diagnose a tumor, it is easier to pick up changes in the microbial DNA that could accurately reflect the presence of cancers as well as their type, earlier in the process of tumor formation, compared to liquid biopsies, at least at their current stage. This is also true about cancers that don't have genetic mutations that are currently detectable using liquid biopsies.
The current platform may still return false-negatives, say the researchers, but they are refining their approach with more data to increase the accuracy of prediction. Another danger with this approach is the overdiagnosis of tumors or a high false-positive rate.
And thirdly, many mutations are not cancerous but are related to age or are of the type that resolves spontaneously. In fact, without the microbial DNA testing, the individual would never know about them at all. Even some early cancers are not really deserving of treatment. As a result, it is essential to remember that screening and diagnosing cancer early is not always required. The need for such procedures should be decided by a clinician.
Positive microbial readouts also should not be taken as meaning cancer unless additional tests are done to confirm that a tumor is present, determine its type and its location.
The team is looking forward to developing an FDA-approved test to diagnose cancer.
To achieve this, they need to profile the 'normal' microbial patterns among healthy people of many different populations. Secondly, they must decide if the microbial signatures found in dead blood reflect the presence of dead, live, or burst-open microbes. This is important to evolving a more accurate approach.
The preliminary findings must be validated in a larger population with a much greater variety of patients, which involves a high-cost upfront. To help achieve their goals, the team has filed patent applications and started up a company called Micronoma.
The researchers point out that though microbiologists commonly use decontamination protocols in their work, it is not a frequent practice in cancer studies. They hope this study will change the emphasis in the field of cancer biology, by making cancer scientists more conscious of the microbes in the human body.
Secondly, it could help thrust the new field of cancer-associated blood microbiomes forward into therapeutics, helping to understand what the microbes are doing in cancer and whether they can be used to treat these conditions. And if so, the next question is whether they can be supplied or mimicked to treat cancer more successfully.
Poore, G.D., Kopylova, E., Zhu, Q. et al. Microbiome analyses of blood and tissues suggest cancer diagnostic approach. Nature (2020). https://doi.org/10.1038/s41586-020-2095-1