Structural heterogeneity of tumor spheroids using quantitative assessment of autofluorescence

Introduction: Intratumor heterogeneity is a well-recognized contributor to tumor progression and the development of treatment resistance. Procedures like traditional sampling and bulk sequencing often fall short of capturing the vascular intratumoral landscape. Here, we introduce a novel analytical workflow that uses quantitative 3D autofluorescence imaging to examine the structure and cellular organization of breast cancer cells (BCC) spheroids generated from BCCs and fibroblasts (MEFs and CAFs). This innovative approach enables characterization of structure in multicellular tumor spheroids via fluorescence-based volumetric imaging and computational tools. Methods: Tumor spheroids were created by co-culturing 4T1 (mouse BCC) with MEFs or CAFs in a 2:1 ratio. The spheroids were generated using 24-well AggreWell plates (STEMCELL Technologies), and Z-stack images were taken using a confocal microscope. Custom software was developed using the Google Colab cloud platform and various Python libraries, including OpenCV and scikit-image. Z-stack images were acquired with the EGFP green and red fluorescence channels, and combined at each z-level using the maximum intensity of the two channels. To focus on large-scale objects, the image was filtered using a Gaussian filter and converted to a binary image using the Otsu algorithm. The fluorescent-intensity-weighted centroid of the spheroid was then estimated. With respect to this centroid, three radial functions were computed: mean intensity, standard deviation, and coefficient of variation. Using a log-log regression fit, the power-law relationship between the mean and standard deviation was computed, and Taylor's exponent for each type of spheroid was determined. Results and Conclusion: Our results identify a significant difference in intra-spheroidal heterogeneity between spheroids grown with MEFs, CAFs, or BCCs. The Taylor's exponent, b, for group A (4T1 and MEFs) spheroids was about 1, indicating a Poisson-like distribution. However, for Group B spheroids (4T1 and CAFs), the exponent was larger, suggesting a super-Poisson aggregation. The fitted exponents (mean ± SD) were b = 1.12 ± 0.05 for the A group and b = 1.38 ± 0.04 for the B group. These higher-exponent spheroids indicate greater radial heterogeneity and uneven distribution of the fluorescent signal. A Taylor's exponent b > 1 signifies that the standard deviation grows faster than the mean. i.e., the greater the mean intensity, the more extreme were the denser and rarer regions of the spheroid, potentially creating drug-resistance niches within the spheroid. Our findings are significant and indicate that this label-free, noninvasive method can be used to detect intertumoral heterogeneity in the near future. These insights can inspire future studies to explore changes within the tumor microenvironment and design new strategies targeting treatment-resistant niches in solid tumors.