Adaptive metabolic pattern biomarker for disease monitoring and staging of lung cancer with liquid biopsy

Abstract

In this manuscript, we demonstrate the applicability of a metabolic liquid biopsy for the monitoring and staging of patients with lung cancer. This method provides an unbiased detection strategy to establish a more precise correlation between CTC quantification and the actual burden of disease, therefore improving the accuracy of staging based on current imaging techniques. Also, by applying statistical analysis techniques and probabilistic models to the metabolic status and distribution of peripheral blood mononuclear cell (PBMC) populations "perturbed" by the presence of CTCs, a new category of adaptive metabolic pattern biomarker (AMPB) is described and unambiguously correlated to the different clinical stages of the patients. In fact, this strategy allows for classification of different categories of disease within a single stage (stage IV) before computed tomography (CT) and positron emission tomography (PET) scans and with lower uncertainty.

Fig. 1

Effects of oxygen on human PBMCs spiked with a tumor cell line. Flow cytometry of PBMCs obtained from human blood from a healthy donor sample spiked with A549 cells (1:100 ratio, cancer:healthy) as a function of the oxygen conditions. a normoxia; b hypoxia; and c hyperoxia. d Average 2NBDG fluorescence emission per cell for the different oxygen conditions. e Ratiometric differences in 2NBDG fluorescence emission for cancer cells (A549) and PBMCs (lymphocytes and monocytes) under the three oxygen conditions. FSC: forward-scattered light

Fig. 2

Gating strategy and correlation between detected and expected CTCs in human PBMCs. Flow cytometry distribution of cells PBMC sample obtained from human blood from a healthy donor sample spiked with A549 cells (1:10⁴ ratio, cancer:healthy) as a function of the a. cell complexity and size; b 2NBDG fluorescence emission and size; and c 2NBDG and CD45 fluorescence emissions. d Correlation between detected and expected A549 cells per million of cells for different cancer:healthy cells ratios (1:10², 1:10³, 1:10⁴, 1:10⁵, and 1:10⁶). SSC: side-scattered light

Fig. 3

Computed tomography Positron emission tomography scanners. a Cross-sectional CT and PET-CT scans corresponding to the maximum size of the primary tumor from five patients. All patients are in stage IV (metastatic) but with different extensions and size of the primary tumor. CP1, adenocarcinoma patient with a non-bulky primary tumor, but with a very wide metastatic extension (subcutaneous); CP2 and CP3, CA patients with large primary tumors (bulky) but with limited extensions of metastasis; CP4, a priori, patient with a large primary tumor (bulky) with limited extension of metastasis (b); CP5, patient with a non-bulky primary tumor with limited extension of metastasis. b CT scan cross-sections showing the tumor progression of CP4 with treatment

Fig. 4

Flow cytometry data of samples of healthy donors and patients. a Healthy donors and b Lung cancer patients, as a function of 2NBDG and CD45 fluorescence emissions. c Number of CTCs per 10⁷ PBMCs

Fig. 5

a Uncategorized, kernel density, and Gaussian mixture models of the flow cytometry row data. Uncategorized flow cytometry raw data, as a function of 2NBDG and CD45 fluorescence emissions; b kernel density estimate of the distribution and c Gaussian mixture models, of PBMC samples obtained from healthy donors (HD) and lung cancer patients (CP)

Fig. 5

a Uncategorized, kernel density, and Gaussian mixture models of the flow cytometry row data. Uncategorized flow cytometry raw data, as a function of 2NBDG and CD45 fluorescence emissions; b kernel density estimate of the distribution and c Gaussian mixture models, of PBMC samples obtained from healthy donors (HD) and lung cancer patients (CP)

Fig. 6

Two-dimensional KS statistic and unsupervised hierarchical clustering of healthy donors and cancer patients. From the distribution of the raw data, distances between all pairs of samples were calculated using the two-dimensional KS statistic. Then individuals where grouped using hierarchical clustering