Frontiers of Cancer Diagnostics: From Photoacoustic Chemical Imaging to Cellular Morphodynamics

Abstract

While terrific progress has been made over the last century, cancer continues to be a prevalent, lethal disease and is responsible for millions of deaths each year. The advent of personalized medicine has brought great strides in the treatment of cancer, as clinicians are able to select therapeutic courses that have been tailored to patients specific set of biomarkers. This selection, in principle, maximizes the chances of cancer remission while minimizing overall patient harm. In this spirit, we have focused on developing diagnostic techniques for two separate cancer biomarkers: tumor potassium concentration, and cell morphology.

We first developed an ionophore-based potassium sensing nanoparticle. The sensor works on the principle of Donnan exclusion in which the overall charge of the carrier remains constant. The hydrophobic interior of the nanoparticle holds a pH-sensitive dye and a potassium ionophore. As the potassium concentrations rise, the ionophore chelates potassium from the solution which results in a proton being removed from the pH dye to maintain charge neutrality. The deprotonation event can be calibrated for quantitative measurement and this sensor was developed for use in diverse imaging modes, which include UV-VIS absorption, fluorescence, and photoacoustics. At physiological pH and in the presence of interfering ions, we were able to quantitatively measure potassium concentrations using each of these readouts.

We modified the potassium sensor to enable in vivo measurements of potassium. This formulation makes use of a solvatochromic dye that transitions from the particle's interior to its surface as potassium is chelated, and thus avoids inherent pH-cross sensitivity. Using photoacoustic chemical imaging, we are able to quantitatively measure the potassium concentration in the tumor microenvironment. As predicted, it was shown that the TME is hyperkalemic, having a potassium concentration of 29mM. The results of the in vivo photoacoustic analysis were verified with ICP-MS measurements of TME potassium.

Finally, we combined cell magneto-rotation and machine learning to develop a technique to measure the metastatic potential of a cancer cell population. This technique aims at avoiding the use of expensive and difficult to produce biological labels. By magnetically activating cells, we are able to suspend them in an oscillating magnetic field where they are free to explore their morphological shape space. By collecting fluorescence images of these cells, we are able to train a classifier to recognize cells of a given type. A proof of concept for the technique is provided here, where MCF-7 and MDA-MB-231 cells, both breast cancer but of different metastatic potential, were classified. A random forest classifier trained on cell images was able to correctly identify the cell type with 86.9% accuracy.