Hybrid AI models allow label-free identification and classification of pancreatic tumor repopulating cell population.

Human pancreatic cancer cell lines harbor a small population of tumor repopulating cells (TRCs). Soft 3D fibrin gel allows efficient selection and growth of these tumorigenic TRCs. However, rapid and high-throughput identification and classification of pancreatic TRCs remain technically challenging. Here, we developed deep learning (DL) models paired with machine learning (ML) models to readily identify and classify 3D fibrin gel-selected TRCs into sub-types. Using four different human pancreatic cell lines, namely, MIA PaCa-2, PANC-1, CFPAC-1, and HPAF-II, we classified 3 main sub-types to be present within the TRC population. Our best model was an Inception-v3 convolutional neural network (CNN) used as a feature extractor paired with a Support Vector Machine (SVM) classifier with radial basis function (rbf) kernel which obtained a test accuracy of 90%. In addition, we compared this hybrid method of supervised classification with other methods of supervised classifications and showed that our working model outperforms others. With the help of unsupervised machine learning algorithms, we also validated that the pancreatic TRC subpopulation can be clustered into 3 sub-types. Collectively, our robust model can detect and readily classify tumorigenic TRC subpopulation label-free in a high-throughput fashion which can be very beneficial in clinical settings.

Reduced Cell-ECM Interactions in the EpiSC Colony Center Cause Heterogeneous Differentiation.

Mechanoregulation of cell-extracellular matrix (ECM) interactions are crucial for dictating pluripotent stem cell differentiation. However, not all pluripotent cells respond homogeneously which results in heterogeneous cell populations. When cells, such as mouse epiblast stem cells (EpiSCs), are cultured in clusters, the heterogeneity effect during differentiation is even more pronounced. While past studies implicated variations in signaling pathways to be the root cause of heterogeneity, the biophysical aspects of differentiation have not been thoroughly considered. Here, we demonstrate that the heterogeneity of EpiSC differentiation arises from differences in the colony size and varying degrees of interactions between cells within the colonies and the ECM. Confocal imaging demonstrates that cells in the colony periphery established good contact with the surface while the cells in the colony center were separated by an average of 1-2 µm from the surface. Traction force measurements of the cells within the EpiSC colonies show that peripheral cells generate large tractions while the colony center cells do not. A finite element modeling of EpiSC colonies shows that tractions generated by the cells at the colony periphery lift off the colony center preventing the colony center from undergoing differentiation. Together, our results demonstrate a biophysical regulation of heterogeneous EpiSC colony differentiation.

Quantifying stiffness and forces of tumor colonies and embryos using a magnetic microrobot.

Stiffness and forces are two fundamental quantities essential to living cells and tissues. However, it has been a challenge to quantify both 3D traction forces and stiffness (or modulus) using the same probe in vivo. Here, we describe an approach that overcomes this challenge by creating a magnetic microrobot probe with controllable functionality. Biocompatible ferromagnetic cobalt-platinum microcrosses were fabricated, and each microcross (about 30 micrometers) was trapped inside an arginine-glycine-apartic acid-conjugated stiff poly(ethylene glycol) (PEG) round microgel (about 50 micrometers) using a microfluidic device. The stiff magnetic microrobot was seeded inside a cell colony and acted as a stiffness probe by rigidly rotating in response to an oscillatory magnetic field. Then, brief episodes of ultraviolet light exposure were applied to dynamically photodegrade and soften the fluorescent nanoparticle-embedded PEG microgel, whose deformation and 3D traction forces were quantified. Using the microrobot probe, we show that malignant tumor-repopulating cell colonies altered their modulus but not traction forces in response to different 3D substrate elasticities. Stiffness and 3D traction forces were measured, and both normal and shear traction force oscillations were observed in zebrafish embryos from blastula to gastrula. Mouse embryos generated larger tensile and compressive traction force oscillations than shear traction force oscillations during blastocyst. The microrobot probe with controllable functionality via magnetic fields could potentially be useful for studying the mechanoregulation of cells, tissues, and embryos.

LAP2ß transmits force to upregulate genes via chromatin domain stretching but not compression.

There is increasing evidence that force impacts almost every aspect of cells and tissues in physiology and disease including gene regulation. However, the molecular pathway of force transmission from the nuclear lamina to the chromatin remain largely elusive. Here we employ two different approaches of a local stress on cell apical surface via an RGD (Arg-Gly-Asp)-coated magnetic bead and whole cell deformation at cell basal surface via uniaxial or biaxial deformation of a fibronectin-coated flexible polydimethylsiloxane substrate. We find that nuclear protein LAP2ß mediates force transmission from the nuclear lamina to the chromatin. Knocking down LAP2ß increases spontaneous movements of the chromatin by reducing tethering of the chromatin and substantially inhibits the magnetic bead-stress or the substrate-deformation induced chromatin domain stretching and the ensuing dihydrofolate reductase (DHFR) gene upregulation. Analysis of DHFR gene-containing chromatin domain alignments along or perpendicular to the direction of the stretching/compressing reveals that the chromatin domain must be stretched and not compressed in order for the gene to be rapidly upregulated. Together these results suggest that external-load induced rapid transcription upregulation originates from chromatin domain stretching but not compressing and depends on the molecular force transmission pathway of LAP2ß. STATEMENT OF SIGNIFICANCE: How force regulates gene expression has been elusive. Here we show that the orientation of the chromatin domain relative to the stress direction is crucial in determining if the chromatin domain will be stretched or compressed in response to a cell surface loading. We also show that nuclear protein Lap2b is a critical molecule that mediates force transmission from the nuclear laminar to the chromatin to regulate gene transcription. This study reveals the molecular force transmission pathway for force-induced gene regulation.

Additive Manufacturing of Viscoelastic Polyacrylamide Substrates for Mechanosensing Studies.

Polymerized polyacrylamide (PAA) substrates are linearly elastic hydrogels that are widely used in mechanosensing studies due to their biocompatibility, wide range of functionalization capability, and tunable mechanical properties. However, such cellular response on purely elastic substrates, which do not mimic the viscoelastic living tissues, may not be physiologically relevant. Because the cellular response on 2D viscoelastic PAA substrates remains largely unknown, we used stereolithography (SLA)-based additive manufacturing technique to create viscoelastic PAA substrates with tunable mechanical properties that allow us to identify physiologically relevant cellular behaviors. Three PAA substrates of different complex moduli were fabricated by SLA. By embedding fluorescent markers during the additive manufacturing of the substrates, we show a homogeneous and uniform composition throughout, which conventional manufacturing techniques cannot produce. Rheological investigation of the additively manufactured PAA substrates shows a viscoelastic behavior with a 5-10% loss moduli compared to their elastic moduli, mimicking the living tissues. To understand the cell mechanosensing on the dissipative PAA substrates, single live cells were seeded on PAA substrates to establish the basic relationships between cell traction, cytoskeletal prestress, and cell spreading. With the increasing substrate moduli, we observed a concomitant increase in cellular traction and prestress, but not cell spreading, suggesting that cell spreading can be decoupled from traction and intracellular prestress in physiologically relevant environments. Together, additively manufactured PAA substrates fill the void of lacking real tissue like viscoelastic materials that can be used in a variety of mechanosensing studies with superior reproducibility.

Effects of forces on chromatin.

Chromatin is a unique structure of DNA and histone proteins in the cell nucleus and the site of dynamic regulation of gene expression. Soluble factors are known to affect the chromatin structure and function via activating or inhibiting specific transcription factors. Forces on chromatin come from exogenous stresses on the cell surface and/or endogenous stresses, which are regulated by substrate mechanics, geometry, and topology. Forces on chromatin involve direct (via adhesion molecules, cytoskeleton, and the linker of nucleoskeleton and cytoskeleton complexes) and indirect (via diffusion and/or translocation processes) signaling pathways to modulate levels of chromatin folding and deformation to regulate transcription, which is controlled by histone modifications and depends on magnitude, direction, rate/frequency, duration, and modes of stresses. The rapid force transmission pathway activates multiple genes simultaneously, and the force may act like a "supertranscription factor." The indirect mechanotransduction pathways and the rapid force transmission pathway together exert sustained impacts on the chromatin, the nucleus, and cell functions.

Machine Learning Approach to Raman Spectrum Analysis of MIA PaCa-2 Pancreatic Cancer Tumor Repopulating Cells for Classification and Feature Analysis.

A machine learning approach is applied to Raman spectra of cells from the MIA PaCa-2 human pancreatic cancer cell line to distinguish between tumor repopulating cells (TRCs) and parental control cells, and to aid in the identification of molecular signatures. Fifty-one Raman spectra from the two types of cells are analyzed to determine the best combination of data type, dimension size, and classification technique to differentiate the cell types. An accuracy of 0.98 is obtained from support vector machine (SVM) and k-nearest neighbor (kNN) classifiers with various dimension reduction and feature selection tools. We also identify some possible biomolecules that cause the spectral peaks that led to the best results.

COL2A1 Is a Novel Biomarker of Melanoma Tumor Repopulating Cells.

Soft 3D-fibrin-gel selected tumor repopulating cells (TRCs) from the B16F1 melanoma cell line exhibit extraordinary self-renewal and tumor-regeneration capabilities. However, their biomarkers and gene regulatory features remain largely unknown. Here, we utilized the next-generation sequencing-based RNA sequencing (RNA-seq) technique to discover novel biomarkers and active gene regulatory features of TRCs. Systems biology analysis of RNA-seq data identified differentially expressed gene clusters, including the cell adhesion cluster, which subsequently identified highly specific and novel biomarkers, such as Col2a1, Ncam1, F11r, and Negr1. We validated the expression of these genes by real-time qPCR. The expression level of Col2a1 was found to be relatively low in TRCs but twenty-fold higher compared to the parental control cell line, thus making the biomarker very specific for TRCs. We validated the COL2A1 protein by immunofluorescence microscopy, showing a higher expression of COL2A1 in TRCs compared to parental control cells. KEGG pathway analysis showed the JAK/STAT, hypoxia, and Akt signaling pathways to be active in TRCs. Besides, the aerobic glycolysis pathway was found to be very active, indicating a typical Warburg Effect on highly tumorigenic cells. Together, our study revealed highly specific biomarkers and active cell signaling pathways of melanoma TRCs that can potentially target and neutralize TRCs.

A quartz crystal microbalance based study reveals living cell loading rate via αvß3 integrins.

Cellular interactions with the microenvironment are mediated by ligand-receptor bonds. Such ligand-receptor bond dynamics is known to be heavily dependent on the loading rate. However, the physiologically-relevant loading rate of living cells remains unknown. Here, using a quartz crystal microbalance (QCM), we developed a bulk-force sensing platform to semi-quantitatively detect the rate of cellular force application during early stages of cell adhesion and spreading. Atop an Au-coated quartz crystal, covalently linked self-assembled monolayers (SAM) were used to immobilize cyclic-RGDfK peptides that can interact with the αvß3 integrins on cells. The QCM detects the changes in resonant frequency of the vibrating crystal due to the cellular activity/probing (force application) on the QCM surface. The corresponding changes in mass on the surface, proportional to the rate of force application, arise from the cellular interactions with the functionalized surface were calculated. The loading rate of living cells was found to be ∼80-115 pN/s. Collectively, our results revealed a fundamental feature of cell adhesion and spreading providing valuable information regarding cellular interactions with the extracellular matrix.

Cdc42-dependent modulation of rigidity sensing and cell spreading in tumor repopulating cells.

Recently, a robust mechanical method has been established to isolate a small subpopulation of highly tumorigenic tumor repopulating cells (TRCs) from parental melanoma cells. In order to characterize the molecular and mechanical properties of TRCs, we utilized the tension gauge tether (TGT) single-molecule platform and investigated force requirements during early cell spreading events. TRCs required the peak single molecular tension of around 40 pN through integrins for initial adhesion like the parental control cells, but unlike the control cells, they did not spread and formed very few mature focal adhesions (FAs). Single molecule resolution RNA quantification of three Rho GTPases showed that downregulation of Cdc42, but not Rac1, is responsible for the unusual biophysical features of TRCs and that a threshold level of Cdc42 transcripts per unit cell area is required to initiate cell spreading. Cdc42 overexpression rescued TRC spreading through FA formation and restored the sensitivity to tension cues such that TRCs, like parental control cells, increase cell spreading with increasing single-molecular tension cues. Our single molecule studies identified an unusual biophysical feature of suppressed spreading of TRCs that may enable us to distinguish TRC population from a pool of heterogeneous tumor cell population.