Transcriptional activators YAP/TAZ and AXL orchestrate dedifferentiation, cell fate, and metastasis in human osteosarcoma

Osteosarcoma (OS) is a molecularly heterogeneous, aggressive, poorly differentiated pediatric bone cancer that frequently spreads to the lung. Relatively little is known about phenotypic and epigenetic changes that promote lung metastases. To identify key drivers of metastasis, we studied human CCH-OS-D OS cells within a previously described rat acellular lung (ACL) model that preserves the native lung architecture, extracellular matrix, and capillary network. This system identified a subset of cells—termed derived circulating tumor cells (dCTCs)—that can migrate, intravasate, and spread within a bioreactor-perfused capillary network. Remarkably, dCTCs highly expressed epithelial-to-mesenchymal transition (EMT)-associated transcription factors (EMT-TFs), such as ZEB1, TWIST, and SOX9, which suggests that they undergo cellular reprogramming toward a less differentiated state by coopting the same epigenetic machinery used by carcinomas. Since YAP/TAZ and AXL tightly regulate the fate and plasticity of normal mesenchymal cells in response to microenvironmental cues, we explored whether these proteins contributed to OS metastatic potential using an isogenic pair of human OS cell lines that differ in AXL expression. We show that AXL inhibition significantly reduced the number of MG63.2 pulmonary metastases in murine models. Collectively, we present a laboratory-based method to detect and characterize a pure population of dCTCs, which provides a unique opportunity to study how OS cell fate and differentiation contributes to metastatic potential. Though the important step of clinical validation remains, our identification of AXL, ZEB1, and TWIST upregulation raises the tantalizing prospect that EMT-TF-directed therapies might expand the arsenal of therapies used to combat advanced-stage OS.


Treatment of OS-D cells on ACL model with doxorubicin
To compare the effect of doxorubicin on OS-D cells grown as monolayer-2D and on dCTC from the 4D lung model, we seeded 25,000 cells/well in triplicate in 96-well plates with 200 µl cultured media and treated it with 3 µM doxorubicin for 48 hours. The control wells with cells received a similar volume of ethanol. After 48 hours, the total live cells were counted using the trypan blue exclusion method.

Histology and Immunohistochemistry analyses
Lung tissues from the MG63.2 xenograft group of mice treated with AXL inhibitor or placebo were fixed in 10% formalin, embedded in paraffin, and then sliced in 5μm sections. The EZ-retriever microwavebased pretreatment and antigen retrieval system (Biogenex, CA) used for dewaxing, rehydration, and antigen retrieval of these FFPE lung tissue sections. AXL, YAP, TAZ, and vimentin protein expression by IHC were evaluated on Leica Bond MAX Auto-stainer by using primary antibodies, rabbit polyclonal antibody anti-human AXL (Cell Signaling Technology, #8661), rabbit polyclonal antibody anti-human YAP (Cell Signaling Technology, #14074), mouse monoclonal antibody anti-human TAZ (Abcam, #ab242313), and mouse monoclonal antibody anti-human vimentin (Leica, #PA0640); respectively. These primary antibodies were detected using Bond polymer refine detection kit according to the manufacturer protocol (Leica, #DS9800). Then, the slides were dehydrated in grade alcohols, cleared in xylene, cover-slipped, and imaged with Keyence Microscope (Keyence, Tokyo, Japan) at 20X resolution. The digital images were processed and quantified using the Visiopharm software version (2020.04) (Hoersholm, Denmark). An APP (Analysis Protocol Package) was designed to quantify cell-based DAB staining in MG63.2 preclinical samples using traditional thresholding methods. The algorithm was built around a three-step approach 1) Pre-Processing: HDAB-DAB and HDAB-Hematoxylin features were used to detect AR Positive cells (Green masking) and Negative cells (Blue masking) respectively by setting pixel values. A median filter of size 5X5 was used for the proper segmentation of cells. 2) Post Processing: Additional steps were designed to enhance the performance of the APP. Change by shape excludes the artifacts; merged cells were separated by separate labels step, and certain clear areas in the cells were filled with fill holes step to fully mask the cells. 3) Output Variable: Mean intensity of each cell and from each sample was extracted and exported into a spreadsheet. Finally, the data were plotted on a Scatter Plot using GraphPad Prism software, version 8 showing the AR Mean Intensity of each cell from all the analyzed samples, and the AR Average mean intensity from each preclinical group of treated mice.
Cell migration (wound healing) assay MG63 and MG63.2 cells were pre-seeded in 6-well plates at the density of 300 X 10 3 /well. 24 h later, a new 200-µl pipette-tip was used to gently and slowly scratch the cell monolayer across the center of the wells. The detached cells were gently washed away with culture medium. After replenishing the wells with fresh medium, the gap in the monolayer was captured on Keyence microscope (Keyence, Tokyo, Japan). The scratched cell monolayer was then cultured for another 24 h in the presence of 100nM or 1µM of AXL inhibitor (SGI-7079). The gap in the monolayer was captured again by the same microscope at 9, 15, and 24 h, and the images were analyzed by Image J software (NIH, Bethesda, MD, USA).

RNA Extraction and Quantitative Gene Expression
The total RNA was extracted from monolayer cell cultures, recellularized lung matrices, and collected dCTCs using Isol-RNA lysis reagent (5 PRIME, Gaithersburg, MD, USA), followed by a DirectZol RNA miniprep (Zymo Research, USA) per the manufacturer's instructions. The RNA quality and quantity were determined using Nanodrop 1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA). The cDNA was prepared using a high-capacity cDNA Reverse Transcription kit (Applied Biosystems, NY, USA) with 100-500 ng of total RNA and a real-time PCR assay was performed with sensiFAST SyBR No-ROXreagent (Bioline USA, Taunton, MA). The primers were designed using Primer3Plus online too 1 . Equal amounts of each RNA sample were used as PCR templates in reactions to obtain the threshold cycle (Ct). The Ct was normalized using the known Ct from the housekeeping gene (beta microglobulin) RNAs to obtain DCt. To compare the relative levels of gene expression in different samples (2D, 4D, and CTC), DCt values were calculated by using the DCt values associated with the expression levels in 2D as the basis. DCt values were then transformed to the real fold increase in expression by 2 ΔCt . A real fold change was calculated using the 2 (ΔΔCt) formula.

MG63 and MG63.2 OS in vivo models and pulmonary metastasis assessment
For experimental lung metastasis experiments, male non-obese diabetic (NOD)-SCID-IL-2Rg null mice (NSG mice; The Jackson Laboratory; Bar Harbor, ME) were used to generate tail vein (5X10 5 cells injected/animal) MG63 or MG63.2 xenografts with parental luciferase-expressing MG63 or metastatic MG63.2 cell lines. One day before the IV injection of OS cell lines, the mice were randomized into treatment and placebo groups (3 to 4 mice per group), received a daily single oral gavage dose of SGI-7079 (AXL inhibitor; Selleck) at 100mg/kg or placebo for up to 25 days through a blinding procedure.
Tumor formation was followed by bioluminescence imaging on an IVIS spectrum instrument (Caliper Life Science) after Isoflurane-anesthetization of the animals, IP injection of d-luciferin (PerkinElmer), and quantified with Living Image software (PerkinElmer). The lungs of NSG mice injected with luciferase-expressing cells were also checked for metastases by counting the number of tumor nodules after Bouin's staining of lungs.

Supplemental Figure Legends
Supplemental Fig. 1