If you picture the immune system as a disease-fighting army, then antibodies are that army’s foot soldiers. These tiny protein machines are tasked with a mission to seek and destroy invaders, and they employ a variety of tools to do so against a range of threats. Although the immune system is often discussed in the context of external threats (like viruses or bacteria), you may be surprised to learn that the immune system can also respond to threats from within—prominent among these is cancer. As a tumor develops, it acquires molecular changes that distinguish the tumor’s cells from those of the surrounding tissue. Sometimes—through mechanisms we don’t fully understand—these differences are enough for the immune system to begin producing antibodies against these cancer-associated signatures. These antibodies—termed autoantibodies since they recognize the body’s own tissues—are the subject of a recent study from the labs of Dr. A. McGarry Houghton and Dr. Paul Lampe at the Fred Hutch Cancer Center.
Recent attention on cancer-associated autoantibodies is well motivated: although they are meant to target the tumor, they can sometimes cross-react with normal tissue and cause serious autoimmune disease in patients. Furthermore, these antibodies tend to appear early in cancer development, and since they circulate in the bloodstream, detecting them holds real promise as a method of early cancer screening. This study, led by Dr. Kristin Lastwika of the Lampe lab, focuses on autoantibodies in small cell lung cancer (SCLC)—a particularly deadly cancer, partly because most patients are diagnosed only once the cancer has already metastasized. In particular, Lastwika and colleagues set out to answer a fundamental question about SCLC-associated autoantibodies: how numerous are these autoantibodies, and what molecules do they target?
To do so, the team used a large-format screening platform previously developed in Dr. Lampe’s lab with several advantages over existing methods of autoantibody detection. By running SCLC patient plasma samples over an array containing roughly 3,600 diverse antibodies, they were able to isolate the specific antigens recognized by these antibodies—not unlike a sophisticated fishing rod (with 3,600 different baits attached!). Performing this analysis on a cohort of SCLC patient plasma samples and matched controls yielded 319 up-or down regulated antibodies of diverse types. To test whether these antibodies were more broadly associated with SCLC, the team pared this list down to those antibodies which were upregulated in cancer patients and constructed a new, more compact antibody array before using this array to test plasma samples from two independent cohorts of SCLC patients. To their amazement, many of the same antibodies appeared in these patients as well!
Having constructed a high-confidence list of SCLC-associated autoantibodies, Lastwika and colleagues asked an important follow-up question: what exactly were these antibodies recognizing? Searching through their list of antibody hits, they selected five targets—CD133, PLD3, TFRC, CA9, and SPINT2—with potential clinical relevance. Staining for these proteins on samples of SCLC tissue, the team discovered that indeed, these proteins were overexpressed in tumors, but largely absent from adjacent tissue. Digging even deeper into how exactly these proteins were recognized, the team made an even more surprising discovery: rather than simply recognizing the proteins in their native states, it appears that autoantibodies were specifically targeting cancer-associated chemical modifications of these proteins: sugar chains added to CD133, a modified aspartate amino acid tacked onto SPINK1, and the nonstandard amino acid citrulline displayed on TFRC, for example.
This exquisite antibody specificity led the team to their ultimate question: could these autoantibodies be used to detect SCLC in patients, before it spreads? Selecting a subset of their autoantibody hits, Lastwika and colleagues used one of their sample cohorts to train a computational algorithm to sort samples into ‘high risk’ and ‘low risk’ categories according to autoantibody levels. In one cohort, their model using five autoantibody biomarkers alone performed admirably, with an AUROC metric (a measure of the model’s predictive accuracy) of 0.87 out of 1. And when the team incorporated information about patient smoking into their model, the AUROC maxed out at 1 for two other cohorts—that is, the model was perfectly predictive, binning all cases into ‘high risk’ and all controls into ‘low risk’ categories (notably, this predictive accuracy could not be achieved by considering smoking habits alone).
Thus, not only do Lastwika et al. leverage new technologies to expand the scope of known SCLC-associated autoantibodies, but they directly demonstrate the value which such knowledge brings towards detecting—and ultimately treating—this deadly cancer in patients. Their work creates fascinating follow-up questions spanning basic and translational biology, and while they acknowledge their methods’ limitations, they are confident that what they found—in relatively small cohorts of patients with a single cancer type, examining a small subset of preselected antibody targets—is almost surely an underestimate of autoantibody diversity in cancer.
The spotlighted research was funded by the National Institutes of Health.
Fred Hutch/University of Washington/Seattle Children’s Cancer Consortium members Dr. Paul Lampe, Dr. A. McGarry Houghton, and Dr. David MacPherson contributed to this study.
Lastwika, K. J., Kunihiro, A., Solan, J. L., Zhang, Y., Taverne, L. R., Shelley, D., Rho, J. H., Randolph, T. W., Li, C. I., Grogan, E. L., Massion, P. P., Fitzpatrick, A. L., MacPherson, D., Houghton, A. M., & Lampe, P. D. (2023). Posttranslational modifications induce autoantibodies with risk prediction capability in patients with small cell lung cancer. Science Translational Medicine, 15(678). https://doi.org/10.1126/scitranslmed.add8469