GAGomes for multi-cancer early detection

Early detection of cancer is generally considered an effective strategy to potentially save subject lives. Screening programs in breast, prostate, colorectal and cervical cancer have resulted in meaningful reductions in mortality rates. It remains that in most cancer types, there are currently no approved biomarkers for early cancer detection.
    Significant advances have been made towards developing liquid biopsy platforms designed for universal multi-cancer screening, leveraging non-invasive biofluidic biomarkers typically relying on sequencing and detecting cancer-derived fraction of circulating free DNA (cfDNA). However, considerable challenges hamper such liquid biopsies. First, they mainly interrogate a specific layer of biological information about cancer, namely genomics. Second, some cancer types do not shed measurable levels of cfDNA. In a recent study, only 12 cancer types were classified as “high-signal” while the remaining 30+ cancer types were found to have a lower signal; these low signal cancers are in fact estimated to account for 50% of global cases and responsible for one third of all cancer deaths. Third, the sensitivity for stage I cancer, i.e. small cancers that do not invade nearby tissues nor lymph nodes or other body parts, is still far from ideal at the very high specificity (i.e. 95% to 99%) required in a screening setting. Some of these challenges have been addressed. Improved versions of the tests have shown to have a higher stage I sensitivity (from 10% to 30%) at a meaningful specificity for multi-cancer screening (95% to 99%). Nonetheless, the improved performance increased the complexity and costs of the required analytical methods, which may become prohibitively expensive for screening of the general population. In addition, cancer types like non-prostate genitourinary and brain tumors remained largely undetectable using these approaches.
    There is therefore an unmet need for an effective IVD MCED test especially to detect cancer at an early stage.
    Instead of genomics and proteomics, Elypta turned towards investigating cancer metabolism as an identifiable hallmark of cancer that could fill the information gap left open by current liquid biopsy platforms. A systems biology multi-cancer analysis of tumor metabolism helped identify cancer-specific reprogramming of glycosaminoglycan (GAG) biosynthesis. GAGs are long unbranched polysaccharides consisting of a repeating disaccharide unit. Depending on the disaccharide unit, different classes of GAGs have been identified in humans. GAGs have remarkable structural diversity, which is determined by complex sulfation and epimerization patterns arising during the template-free biosynthesis of GAGs. The biological functions of GAGs include modulation of the extracellular matrix, cell proliferation and metabolism, and immune supervision. Initial studies were performed in renal cell carcinoma (RCC), a genitourinary tumor with no approved liquid biomarkers. There, Elypta developed GAG scores based on plasma and urine GAGomes that were significantly altered in any-stage RCC, i.e. from organ confined disease up to metastatic disease, with AUC > 0.993 vs. healthy controls.
    The GAG score is a metric that aggregates measurements of the concentration and structural composition of GAGs – the so-called GAGome in a biospecimen. GAGs are measured from freely circulating GAGs, and not protein-bound, and termed free GAGomes. The index test for the present study is based on the GAG score computed from GAGome features measured in plasma and a cut-off value for test positivity. Two other index tests will also be developed and validated in this study, using urine GAGome features or combined plasma and urine features.
    To further explore GAGomes across different cancer types besides RCC, Elypta developed a standardized ultra-high-performance liquid chromatography coupled with triple-quadrupole mass spectrometry (UHPLC-MS/MS) method to measure free GAGomes with high-throughput. In a recent study, plasma and urine free GAGomes were found to change from physiological levels in 14 cancer types and could therefore serve as metabolic biomarkers of cancer. In this study, the largest compendium of biofluidic GAGomes to date was generated on a total of 1529 samples across 1006 healthy donors and cancer patients ranging from stage I/low grade (~33.8%) to stage IV/high-grade disease. Remarkable alterations were found in both plasma and urine GAGomes of cancer compared to physiological levels. Bayesian modeling was used to develop three GAG scores, one for plasma, one for urine, and one combined. In the validation set, the scores were able to detect any-stage cancer with respectively 34.6%, 38.6%, and 40.5% sensitivity at a pre-specified 98% specificity. Notably, the sensitivities ranged from 30.9% to 33.3% for stage I/low grade cancer, including cancer types such as brain and genitourinary tumors. Combined GAGomes were informative of the tissue of origin. A machine learning model was trained, that achieved 74.3% accuracy to detect the correct tissue of origin across 5 cancer types.
    Importantly, survival analyses found that GAG scores independently correlated with overall survival, indicating that cancer subjects undetected using GAGomes may have a better prognosis than detected cases, thereby limiting the risk for false negative cases. In summary, the current performance characteristics, and the ability to detect historically undetectable cancer types indicate that free GAGomes have substantial potential for MCED.
    To assess the translational potential of the GAG scores in MCED, Elypta conducted the project denoted OV19_0526 at Lifelines. In that project, Elypta sought to test GAG scores at baseline and see if elevated GAG scores predicted cancer. By linkage with the NCR, we performed an initial external validation of our GAG scores. We observed that otherwise healthy subjects could be detected with cancer within 3 months using our score. In this study effectively mimicking MCED in asymptomatic adults (n = 371), we demonstrated that plasma GAGomes had an AUC = 0.75 to predict a first cancer diagnosis within 3 months. Combining plasma and urine GAGomes resulted in diagnostic performance characteristics meeting the requirements for clinical implementation of MCED. At 99% specificity, the sensitivity was 46% to identify subjects with any-cancer stage 0 to II within a 3-month window in the subset of 50-69 years old adults (N = 94). 
    To translate these findings into clinical practice, more clinical validation and clinical utility studies in the intended population for a GAGome-based MCED test are required. Our new proposed study, denoted LEVANTIS-0087A (LEV87A) at Elypta, is designed as a prospectively planned retrospective IVD-test clinical validation case-control study to develop and validate a MCED test based on plasma and urine free GAGomes in a large representative sample of the population intended for such test (target population). Based on results from the OV19_0526 pilot study, we believe that it is realistic to develop a clinically useful GAG score for MCED based on free GAGomes with a concrete chance for a rapid implementation in the