Biophysical phenotype mixtures reveal advantages for tumor muscle invasion in vivo

Abstract

Bladder, colon, gastric, prostate, and uterine cancers originate in organs surrounded by laminin-coated smooth muscle. In human prostate cancer, tumors that are organ confined, without extracapsular extension through muscle, have an overall cancer survival rate of up to 97% compared with 32% for metastatic disease. Our previous work modeling extracapsular extension reported the blocking of tumor invasion by mutation of a laminin-binding integrin called α6β1. Expression of the α6^AA mutant resulted in a biophysical switch from cell-ECM (extracellular matrix) to cell-cell adhesion with drug sensitivity properties and an inability to invade muscle. Here we used different admixtures of α6^AA and α6^WT cells to test the cell heterogeneity requirements for muscle invasion. Time-lapse video microscopy revealed that tumor mixtures self-assembled into invasive networks in vitro, whereas α6^AA cells assembled only as cohesive clusters. Invasion of α6^AA cells into and through live muscle occurred using a 1:1 mixture of α6^AA and α6^WT cells. Electric cell-substrate impedance sensing measurements revealed that compared with α6^AA cells, invasion-competent α6^WT cells were 2.5-fold faster at closing a cell-ECM or cell-cell wound, respectively. Cell-ECM rebuilding kinetics show that an increased response occurred in mixtures since the response was eightfold greater compared with populations containing only one cell type. A synthetic cell adhesion cyclic peptide called MTI-101 completely blocked electric cell-substrate impedance sensing cell-ECM wound recovery that persisted in vitro up to 20 h after the wound. Treatment of tumor-bearing animals with 10 mg/kg MTI-101 weekly resulted in a fourfold decrease of muscle invasion by tumor and a decrease of the depth of invasion into muscle comparable to the α6^AA cells. Taken together, these data suggest that mixed biophysical phenotypes of tumor cells within a population can provide functional advantages for tumor invasion into and through muscle that can be potentially inhibited by a synthetic cell adhesion molecule.

Significance

Many epithelial cancers including bladder, colon, and prostate escape the primary site as a cohesive population traversing the smooth muscle surrounding the organ. Preventing invasion of cohesive tumor clusters into and through muscle and therefore preventing metastatic disease would significantly improve cancer-specific mortality. The current study suggests that specific biophysical phenotype mixtures of cells—epithelial (cell-cell) and mesenchymal (cell-extracellular matrix) cooperation—within the tumor population can increase invasion. The different biophysical phenotypes have different therapeutic drug responses. In contrast, targeting cell adhesion with a synthetic integrin ligand showed that tumor invasion into and through a living mouse muscle was blocked and served as an example of a synthetic molecular approach for metastasis prevention.

Introduction

A variety of epithelial cancers originate in organs containing a smooth muscle outer layer, and several rank among the top 10 cancers in incidence and deaths for both men and women, including colon, bladder, prostate, uterine, and gastric cancers (1). Each of these cancers is nonlethal with treatment if they remain confined to the organ, but they become lethal when they escape the primary organ and metastasize. One critical step in aggressive prostate cancer is tumor invasion into and through the smooth muscle surrounding the organ, referred to as extracapsular extension (ECE). ECE defines aggressive disease and informs therapy decisions. The smooth muscle is a hostile biophysical environment for tumors due, in part, to compressive forces and oxidative stress, so it is rarely a metastatic destination (2).

Biophysical challenge of the muscle barrier to invasion by tumor clusters

Tumors can escape the site of origin via invasion through the smooth muscle layer in the exterior of the organ (3). Tumor also escapes the organ via intravasation or extravasation through blood vessels, which requires tumor clusters to migrate through a smooth muscle layer sheathing the vascular wall (4). Smooth muscle constitutes a distinctive biophysical tumor microenvironment, and tumor invasion into and through this barrier is a first step toward metastatic progression.

Layers of smooth muscle, along with connective tissue, constitute the anatomical layers submucosa, mucosa, and serosa surrounding hollow organs or organ systems (3). Smooth muscle is distinguished from skeletal muscle in that it is involuntary and is found in the walls of the urinary bladder (5), uterus, stomach, intestines, and prostate (6), in arteries and veins of the circulatory system, and in the tracts of the respiratory, urinary, and reproductive systems (7,8). Fully differentiated smooth muscle contains the cytoplasmic molecular markers of desmin (an intermediate filament), calponin (an actin-binding protein), and laminin-binding adhesion receptors, including the biosensing α7β1 integrin (9,10,11,12).

Individual smooth muscle cells, called myocytes, are spindle-shaped (wide in the middle and tapered at both ends) and, like striated muscle, can tense and relax in response to signals from the nervous system (8). Smooth muscle tissue exhibits superior elasticity and function within a larger length-tension curve than does striated muscle (13). In the relaxed state, each cell is 20–500 μm in length (thousands of times shorter than skeletal muscle fibers), and these cells produce their own connective tissue, the endomysium (7). Myocytes are covered by a basement membrane and confined by loose collagen-rich fibrillar mesh-works that offer both tissue cohesion and plasticity of movement during contraction, providing a moderately low resistance for migrating cells, whereas the basement membranes provide abundant ligands, including laminin, for cell adhesion (3,13). The biophysical measurements of muscle elasticity and the dynamic physical contraction/relaxation cycles demonstrate a unique microenvironment for invading tumor cells (13). The biophysical features include stiffness, elasticity, and contraction forces with physical cues that likely guide tumor invasion (3). These biophysical factors change as the tumor invades the muscle since the elastic modulus of smooth muscle is roughly 100 times stiffer than that of fat, which surrounds the prostate and bladder (14).

Because of these features of the smooth muscle, we have developed an in vivo mouse model (15) using the mouse diaphragm to serve as a smooth muscle barrier similar to the smooth muscle in the prostate. The diaphragm of the mouse contains, in the central zone, a sympathetic smooth muscle that is flanked by skeletal muscle attached to the ribs. In this model, human tumor cells are injected into the intraperitoneal (IP) cavity where the tumor seeds the undersurface of the diaphragm (inferior side) and invades through the living muscle to appear on the superior surface (lung side).

Integrins as biophysical sensors and coordinators of the force generation of cohesive clusters

Integrins are heterodimeric receptors occurring in at least 24 unique combinations of noncovalently interacting α-subunits (18 types) and β-subunits (eight types), which allows binding to multiple ECM components but also to counter-receptors on other cell types (reviewed in (16)). Integrins traverse the plasma membrane, connecting the extracellular matrix (ECM) to the cell cytoskeleton (conveying information between the extracellular environment and the intracellular pathways), and their function is tightly regulated from both a biochemical perspective (activation) and a mechanical perspective (mechanotransduction) (16). These integrin-mediated connections across the plasma membrane are dynamically linked—the cytoskeleton controls affinity and avidity of the integrin ectodomain, thereby modulating the ECM, and integrins binding to the ECM alter the shape and arrangement of the cytoskeleton (reviewed in (17,18)). Forces created via cytoskeleton contraction or ECM rigidity are conveyed across the plasma membrane through integrins, factoring into the conformational changes in these receptors and their anchoring proteins and to the nature of the biochemical signals generated (17).

Because they provide a connection across the plasma membrane, integrins can transmit forces in either direction between the ECM and the actin cytoskeleton, resulting in a balancing of forces between the ECM and the cytoskeleton (16). Integrins do not bind directly to actin, but the β subunit cytoplasmic tail binds to several actin-binding proteins, including kindlin-2, talin, tensin, vinculin, filamin, or α-actinin, allowing forces from the ECM to be transmitted to the cytoskeleton via integrins (16). Forces applied to actin are conveyed, likewise, to the ECM via adaptor proteins and integrins (through myosin contraction or actin polymerization) (16).

Integrins help cells recognize and respond to ECM stiffening via cytoskeletal filaments capable of generating changes within the cell and increasing cell migration. The increased ECM stiffness stimulates fibronectin production, which binds extracellular collagen, fibrin, and heparan sulfate proteoglycans from one side of the ECM to integrins on the other side (19). In response to ECM stiffening, integrin clustering increases the recruitment of several focal adhesion proteins that promote tumor progression. Taken as a whole, increased ECM stiffness is associated with reduced E-cadherin expression, increased integrin clustering, cancer-associated fibroblast production, and reduced long-term survival (13,19).

Heterogeneity within the tumor for adhesion receptors

Heterogeneity within tumor is most often attributed to epigenetic, transcriptomic, and posttranslational modifications (20), all of which fall under the heading of genomic instability. The heterogeneity of cell types in disseminated tumor clusters correlates with heterogeneity of adhesion receptors, including laminin-binding and collagen-binding integrins, among others. Cell adhesion heterogeneity also has a profound impact on tumor dissemination. Variations of intracellular adhesion receptor concentrations in a cell population reinforce cell dissemination (21). For example, adhesion heterogeneity in a cell population promotes an increase of cells that disseminate from the population, and nondisseminated cells show considerably higher mean adhesion than disseminated cells (21). As discussed in our recent review (13), disseminated tumor cells may benefit from epithelial-mesenchymal cooperation, which allows alternate cell-cell adhesion molecules, such as N-cadherin, to supplement downregulated E-cadherin. The heterogeneity of cell-cell and cell-ECM adhesion molecules has been observed as changing within tumors within a single patient, providing controversy about whether E-cadherin, for example, is truly a hallmark of indolent disease.

The present study

DU145 cells were chosen for study since they are tumorigenic as a mouse xenograft, make low grade type-II, well-differentiated adenocarcinoma similar to a primary tumor, express wild-type (WT) PTEN and p53, and are androgen independent (22). Confirming studies were performed with other prostate cancer cell lines called PC3 (androgen independent, tumorigenic) and 22Rv1 (well-differentiated adenocarcinoma, androgen responsive). Another cell line called LNCaP was not chosen for comparison due to undifferentiated morphology, luminal cell characteristics of PSA secretion, and absence of laminin-binding integrin adhesion functions (23).

Materials and methods

Cells, antibodies, and reagents

DU145, 22Rv1, and PC3 cell lines were obtained from the ATCC. The cell lines were cultured in Iscove’s modified Dulbecco’s medium from Corning supplemented with 10% FBS (Hyclone Laboratories) and incubated at 37°C in a 5% CO₂ humidified chamber. Nonenzymatic Cellstripper (CelGro) was used for cell harvesting. Reagents used for immunofluorescence microscopy include Alexa Fluor 594 conjugated phalloidin (Invitrogen A12381) for F-actin staining and anti-E-cadherin M168 (Abcam, ab76055). MTI-101 was provided by Modulation Therapeutics (96.4% pure by high-pressure liquid chromatography). To generate DU145 cell lines stably expressing H2B-YFP or H2B-RFP, lentivirus was used to transduce the cells. The lentiviral plasmids LV-YFP and LV-RFP were gifts from Elaine Fuchs (Addgene plasmids #26000 and #26001 respectively). Lentivirus packaging vectors pMD2.G (Addgene plasmid #12259) and psPAX2 (Addgene plasmid #12260) were gifts from Didier Trono. The LV-YFP or LV-RFP plasmid and lentivirus packaging vectors were co-transfected into HEK-293T cells using Lipofectamine 3000 (Thermo Fisher Scientific, cat. L3000008). Virus was harvested 24 and 48 h after transfection, filtered through a 0.45-μm filter, and added to the growth medium of cells supplemented with 8 μg/mL polybrene (Sigma-Aldrich: #H9268). Stable cells were sorted in the University of Arizona Flow Cytometry Core using a BD FACSAria III selecting for the median 50% of fluorescent H2B expression.

Prostate cancer samples

The de-identified patient-derived xenograft samples of human prostate cancer were kindly provided by Nora M. Navone, M.D., Ph.D., Department of Genitourinary Medical Oncology at MD Anderson Cancer Center. Samples were freely provided under a Uniform Biological Material Transfer Agreement. The de-identified radical prostatectomy sample was collected by Raymond B. Nagle, M.D., Ph.D., University of Arizona Cancer Center, and processed through the University of Arizona Cancer Center Tissue Acquisition and Molecular Analysis Research Core Service (TACMASR).

Genome editing

Homozygous knockout cell line for ITGA6 gene (DU145 α6^KO) and homozygous amino acid substitutions for ITGA6 R594A and R595A (DU145 α6^AA) were created using CRISPR-Cas9 technologies in the University of Arizona Cancer Center Genome Editing Facility, as previously described (24). Replacement of the endogenous gene with an ectodomain mutation of the α6 integrin, called α6^AA, removes the protease-sensitive site and results in cells unable to invade and metastasize (24). The mutant DU145 cell line (integrin α6^AA) maintains the full-length version of integrin α6 and has a prominent cell-cell phenotype (24).

ECIS measurements

Electrical properties of confluent or wounded epithelium were measured using electric cell-substrate impedance sensing (ECIS) as described previously (24). Cell adhesion measurements, taken every 4 min, were based on changes in resistance/capacitance to current flow applied at different frequencies (Applied Biophysics). The eight-well chamber slide with 10 250-μm electrodes (Applied Biophysics, 8W10E + PET) was first stabilized by coating with 10 mM cysteine in water and then rinsed twice with sterile distilled water. The eight-well chamber slide was then coated with laminin at 37°C for 1 h, cells were inoculated at 200,000 cells per well in 400 μL media in duplicates, and resistance/capacitance was measured at 400 Hz (cell-cell) and 40,000 Hz (cell-ECM), per previous research (25,26), for a period of at least 26 h. For the purposes of clarity, all data have been normalized (divides all values by that obtained at the time listed in the zero-time box); this function defaults to time zero so that all curves would be 1.00 at time zero.

Immunoblot analysis

DU145 cells were lysed in a modified RIPA lysis buffer (150 mM NaCl, 50 mM Tris-HCl, 5 mM EDTA, 1% Triton X-100, pH 7.5) supplemented with Complete Mini Protease Inhibitor Cocktail (Roche) and Phosphatase Inhibitor Cocktails 2 and 3 (Sigma). The sample protein concentrations were determined using the Bradford assay (Bio-Rad) and then prepared using NuPAGE LDS sample buffer (Thermo Fisher Scientific) and PBS. The samples were boiled before being resolved on a NuPAGE 4%–12% Bis-Tris gel (Thermo Fisher Scientific) in nonreducing conditions. The antibodies used for immunoblotting were AA6NT A6 integrin rabbit antibody (1:10,000) (24), Emerin rabbit monoclonal antibody (Cell Signaling Technology D3B9G, 1:4000), and E-cadherin mouse antibody (Abcam M168, 1:2000).

Video microscopy

Matrigel (phenol red-free, growth factor reduced, Corning 356231) was plated at 50 μL per well into a flat-bottom tissue culture-treated 48-well plate (CytoOne CC7682-7548) and allowed to solidify for 30 min at 37°C. DU145 WT and AA cells, expressing H2B-YFP or RFP respectively, were plated at a total of 75,000 cells per well. Images were acquired for at least 16 h using a Nikon Ti2 Inverted Microscope with Perfect Focus to correct for axial focal drift with a 20x objective from Nikon. The microscope is equipped with an environmental chamber that is heat regulated and humidified, and all images were taken at 37°C and 5% CO₂. All fluorescent images utilized fluorescently tagged H2B with YFP or RFP acquired using a SPECTRA X LED light engine. The camera used for image acquisition was CoolSNAP MYO and was run using the Nikon Elements Advanced Research Acquisition and Analysis Duo Package.

Neighbor ratio image processing

In a randomly distributed mixture of N cells plated on a 2D surface of area A, the likelihood of finding one cell within a distance d from another cell is πd²/A, and we expect to find N_exp = πΝd²/A cells within that distance. For cells that are aggregating, the actual number will be larger than this value. Therefore, a measure of how likely cell type A is to aggregate with another cell type B is to determine the ratio of the average number of cells of type B within the distance d from cells of type A divided by the expectation value. If we denote the number of cells of population B that are a distance d away from the cell labeled i in population A as N_B→A,i, then we can define this measure, the neighbor ratio between cells of type A and B, as

$${\mathsf{R}}_{\text{AB}}=\frac{\frac{A}{2}\left(\sum _i{N}_{\mathrm{B}\to \mathrm{A},i}+\sum _j{N}_{\mathrm{A}\to \mathrm{B},j}\right)}{\pi d^2{N}_{\mathrm{A}}{N}_{\mathrm{B}}}\text{,}$$

where N_A and N_B are the total number of cells of type A and B, respectively. The two sums in this expression are used to treat species A and B symmetrically.

Since the cells we are using express fluorescent proteins, instead of counting cells and computing R directly, we can instead treat the cells of a given type as being part of a continuum distribution of density ρ_A,B, respectively, which is given by the fluorescence intensity in the images corresponding with that cell type. We then define a kernel K(x,y) that integrates the intensity in a region of size d about a cell at location (0,0). If our images are of size H×W, the continuum version of the neighbor ratio is

$${\mathsf{R}}_{\text{AB}}=\frac{\frac{HW}{2}\left(\sum _{i,j}\sum _{m,n}\left({\rho }_{\mathrm{A},ij}{\rho }_{\mathrm{B},mn}+{\rho }_{\mathrm{A},mn}{\rho }_{\mathrm{B},ij}\right)K\left(x_i-x_m,y_j-y_n\right)\right)}{\left(\sum _{i,j}{\rho }_{\mathrm{A},ij}\right)\left(\sum _{m,n}{\rho }_{\mathrm{B},mn}\right)}\text{,}$$

where i and m label the pixels horizontally, and j and n label them vertically.

We define K to be a top hat averaging kernel of radius 20 pixels. To remove the overlap of a cell’s own intensity with itself, we also set any distances less than seven pixels in the kernel to be zero. That is,

$$\mathsf{K}=\frac{1}{\sum _{ij}{K}_{ij}}\text{when}7<\sqrt{x^2+y^2}<20=0\text{otherwise}$$

Calculation of the neighbor ratio was carried out using in-house written code in MATLAB (The MathWorks).

Survival assay

Cells were harvested, counted, and added at 9000 cells per well of a flat-bottom tissue culture-treated 96-well plate and incubated at 37°C and 5% CO₂. Approximately 24 h later, the medium was removed and replaced with 100 μL medium containing serial dilutions of a tested compound. The cells were incubated for an additional 72 h. MTT dye metabolism was used to assay cell viability as previously described (27). The IC₅₀ is defined as the concentration of drug required to inhibit cell growth by 50% compared with untreated controls. The compounds tested were bortezomib (Selleckchem, S1013), FK866 (Selleckchem, S2799), and docetaxel (Sigma, 01885).

Mouse xenograft studies of muscle invasion

For mouse tumor xenograft smooth muscle assay, male NOD-SCID mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ, Strain 005,557; Jackson Labs; five mice per group) were injected intraperitoneally with human tumor cells (5 x 10⁶ cells) and tumor colonies were allowed to grow on the undersurface of the diaphragm and within the peritoneal cavity for 8 weeks, as described previously (27). The collected diaphragm from each animal is fixed and embedded so that the tumor colony is oriented on top of the muscle, and transverse sections will detect the tumor invasion into, within, and through the smooth muscle layer. Approximately 4 mm of diaphragm can be analyzed using this method. Similarly, in the MTI-101 experiments, DU145^WT and DU145^AA tumors seed onto the IP surface of the diaphragm for 1 week, followed by weekly intraperitoneal injections of 10 mg/kg of MTI-101. The mice were sacrificed after 8 weeks and the diaphragms harvested as described. The experimental mouse studies were reviewed and approved by the Institutional Animal Care and Use Committee as Protocol Number 07029. The protocol was conducted in accordance with all applicable federal and institutional policies, procedures, and regulations, including the PHS Policy on Humane Care and Use of Laboratory Animals, USDA regulations (9 CFR Parts 1, 2, 3), the Federal Animal Welfare Act (7USC 2131 et seq.), the Guide for the Care and Use of Laboratory Animals, and all relevant institutional regulations and policies regarding animal care and use at the University of Arizona.

Statistical analysis

Statistical analysis was conducted in GraphPad Prism 9.4.1 using their built-in capabilities (unpaired t-test). Standard deviations for ECIS measurements were provided by the ECIS software (Applied Biophysics), v.1.4.18.0 PC (15 February 2023), Runtime: 2017B, 64 bit.

Results

Our previous work modeled cohesive cancer ECE since, as shown in Fig. 1, moderate Gleason sum scores and castrate resistant prostate tumors (28) can invade muscle to escape into the smooth muscle capsule, a prerequisite for ECE. We reported blocking tumor invasion by mutation of a laminin-binding integrin called α6β1, resulting in a switch from a cell-ECM (α6^WT cells) to a cell-cell (α6^AA mutant cells) biophysical phenotype with drug resistance properties (24). Our lab was first to show that in cancer progression, the α6 integrin is cleaved to form a cancer-specific variant called α6p (29).

Figure 1.

Cohesive tumor clusters invading muscle in prostate cancer. (A) Patient-derived xenograft (PDX) of low- to moderate-risk prostate cancer (PCa) (Gleason sum score (GS) and androgen receptor (AR+)) invading smooth muscle detected by H and E staining. (B) PDX of moderate-risk prostate cancer (GS and CRPC (castrate-resistant prostate cancer)) invading smooth muscle detected by H and E staining. (C) De-identified radical prostatectomy tissue specimen showing tumor clusters, stained to detect tumor-specific racemase and high molecular weight cytokeratin (pink), invading the smooth muscle in the prostate capsule and surrounding a nerve. Error bars represent 70 μm. ^∗∗∗ = smooth muscle; ### = tumor clusters, ^^^ = nerve. To see this figure in color, go online.

In this study, we took advantage of these stable phenotypes to determine the functional relevance of cell heterogeneity within the tumor population clusters. Admixtures of the two stable phenotypes in a 3D context (Matrigel) tested the functional heterogeneity requirements for muscle invasion. The endpoints included time-lapse video microscopy of invasive networks in vitro, a mouse xenograft model to track invasion into and through live muscle, and ECIS measurements to detect the biophysical properties of cell-ECM and cell-cell adhesion of the mixtures and the adhesion response in real time to an induced wound. The data suggested that mixed biophysical phenotypes of tumor cells within a population can provide functional advantages for invasion and are potentially targetable using synthetic cell adhesion molecules.

Integrin mutation results in a cell-cell adhesion phenotype distinct from the parental line characterized by unique biochemical, morphological, and biophysical properties and survival responses to chemotherapeutic agents

A tumor-specific form of the laminin-binding α6 integrin, α6p, with the ligand-binding domain removed, is formed on the surface of prostate and breast cancer cells by a specific protease-sensitive site (29,30,31,32). The α6pβ1 heterodimer results in migratory and metastatic tumor cells. Replacement of the endogenous gene with an ectodomain mutation of the α6 integrin, called α6^AA, removes the protease-sensitive site and results in cells unable to invade and metastasize (24). The mutant DU145 cell line (integrin α6^AA) maintains the full-length version of integrin α6 and has a prominent cell-cell phenotype (24). The stable switch from a cell-ECM (α6^WT cells) to a cell-cell (α6^AA cells) adhesion phenotype offered the opportunity to use admixtures of WT and mutant cell types to determine the functional relevance of heterogeneous tumor population for muscle invasion. As a first step, we characterized their biochemical, morphological, and biophysical properties and survival responses to chemotherapeutic agents of α6^WT and α6^AA cells.

Fig. 2A and 2B show cell clusters that highlight the phenotypic differences between the DU145 α6^WT cells (Fig. 2A) and the α6^AA cells (Fig. 2B). The α6^AA cells display a highly organized distribution of filamentous actin (F-actin) with cortical wrapping and typically grow in compact colonies in 2D culture compared with the α6^WTcells. ECIS (Fig. 2C) documented the ability of each cell type under optimal growth conditions to display increasing resistance to an applied electric field, measured at 40,000 Hz (reflecting cell-ECM resistance) as the cells spread onto the electrodes (33) over a period of 16 h. During the initial hours, cells attach to the surface and then spread.

Figure 2.

Characterization of two stable biophysical phenotypes dependent upon integrin function. The morphological appearance of the DU145^WT cells (A) versus the DU145^AA cells (B) is distinct as detected by F-actin staining. The cell adhesion resistance properties as measured by ECIS (electric cell-substrate impedance sensing) (C) were characterized in monolayers of DU-145^WT (WT) and DU-145^AA cells (AA) during the early phase of cell-ECM attachment and the late phase of cell spreading on the monolayer and the response after wounding (red arrow). The total cell lysates (D) confirm the expression of α6 integrin full-length form and the α6p integrin generated by posttranslational modification in the WT cells, whereas the AA cells only express the full-length form of α6 integrin. In (D), DU145^WT (WT), DU145^AA (AA), or DU145^A6−/− (A6^−/−) cells were tested for E-cadherin compared with total protein marker, Emerin (EMD). All panels are representative examples of at least three biological and technical replicates. Scale bar represents 90 μm. To see this figure in color, go online.

At 16 h, an electric pulse wounded the cell monolayer and triggered migration to fill the wound (Fig. 2C). Before wounding, WT cells had approximately 1.5-fold higher cell-ECM resistance compared with the AA cells (2.0 versus 1.3, respectively at 15 h). After wounding, WT cells quickly recovered 90% of resistance within 2 h compared with 5-h recovery time required by the AA cells to reach 90% of their pre-wound resistance value. In-depth kinetic analysis documents the dramatic differences in recovery time and complexity of the biophysical phenotype (see Fig. S1). Further, WT cells returned to their pre-wound state within 10 h and continued to spread. In contrast, the AA cells do not return to their pre-wound spreading levels within the same time frame. The WT cells contain both the full-length α6 integrin and the α6p form of the receptor, whereas the AA cells only contain the full-length, uncleavable form as detected by Western blot analysis (Fig. 2D). Cells without the α6 integrin (A6^−/−) are shown for comparative purposes. As expected, the AA cells expressed elevated levels of E-cadherin compared with the WT cells, consistent with their increased cell-cell adhesion phenotype (Fig. 2D).

Since dramatic changes in cell adhesion can be characteristic of cell adhesion-mediated drug resistance (27,34,35), we tested the cell survival response in the two cell types to structurally unrelated compounds. Three different agents were tested to determine the survival response of the WT and AA cells using the MTT assay after 72 h of exposure to at least 12 different doses of the drug. The complete survival curves are included in the supplementary data (Fig. S2). The data in Table 1 summarizes the resulting IC₅₀ values.

Table 1.

IC₅₀ values for three common drugs using the MTT assay

Drug	Biological target	DU145WT	DU145AA
Bortezomib	proteosome inhibitor	10.83 +/− 7.15 nM	10.44 +/− 10.44 nM
Docetaxel	microtubule stabilizer	12.0 +/− 8.49 nM	14.5 +/− 6.36 nM
FK866	NAMPT inhibitor	85.33 +/− 21.21 nM	28.0 +/− 16.97 nM

The DU145^WT or DU145^AA tumor cells were grown for 18 h under optimal growth conditions and subsequently treated with varying concentrations of drug to determine the survival response. MTT dye metabolism was used to assay cell viability 72 h after drug treatment. The IC₅₀ is defined as the concentration of drug required to inhibit cell growth by 50% compared with untreated controls. Values given are with one standard deviation from the mean.

The WT and AA cells are similar in survival responses to bortezomib and docetaxel, which are structurally and mechanistically unrelated compounds. Interestingly, the AA cells are more vulnerable to disruption of the nicotinamide adenosine dinucleotide salvage pathway requiring nicotinamide adenosine dinucleotide phosphoryl transferase (NAMPT), compared with WT cells. These data indicate that different biophysical phenotypes have differing drug sensitivities.

Although heterogeneity of adhesion and drug resistance phenotypes within tumor populations is a hallmark of cancer progression, it is generally not explored if or how variability contributes a functional advantage for tumor progression. The distinct and stable phenotypes characteristic of these two cell types allowed testing whether a defined mixture of α6^WT and α6^AA cells would provide functional advantages for the tumor population.

Mixture of the two biophysical phenotypes resulted in nonrandom self-assembly and optimal invasive network formation

As previously reported by us, the WT cells and not the AA cells will make extensive invasive networks on Matrigel, as shown in Fig. 3A for comparative purposes. We next mixed the WT and AA cells in a 1:1 ratio (100,000 cells from each lineage) in suspension and applied them to Matrigel (Fig. 3), which resulted in networks with invasive properties as previously reported by us. The networks were tracked by video microscopy over the next 21 h at 37°C (Fig. S3, Video S1). The raw video was colorized with WT cells in cyan and AA cells in magenta (Fig. 3B, 3C, and 3F), allowing network development to be visualized and tracked over time. To visualize the tendency of the 1:1 mixture to form networks, we collected frames of the video at 1 and 16 h, as shown in the figure. The videos were converted for single-cell quantification (Fig. 3D and 3E) and the unique lineages were analyzed for self-assembly (Fig. 3G). We were interested in determining whether WT cells preferentially aggregate with other WT cells or with AA cells, and likewise for the AA cells.

Figure 3.

Mixed tumor cell population resulted in an improved self-assembly of invasion networks. DU145 WT-only cells (A, left) and DU145 AA-only cells (A, right) were grown in 3D for 16 h and visualized by video microscopy. For comparison, a mixture of equal numbers (B and C, 100,000 cells) of each cell phenotype (WT (cyan), AA (magenta)) in suspension were applied to Matrigel, and the resulting networks were tracked by video microscopy. Representative time stamped images (1H and 16H) are shown for the original live cell imaging (B and C), with a conversion to single-cell quantification images at 1, 4, and 16 h (WT (blue) and AA (green)) (D and E). (F) is a higher magnification of cell mixtures from (C). Network self-assembly was measured (G) by the neighbor ratio. Scale bars in (A)–(E) represent 200 μm, and scale bar in (F) represents 50 μm. To see this figure in color, go online.

Video S1. Cohesive clusters of mixtures of WT and AA cells

DU145 WT-only cells and DU145 AA-only cells were grown in 3D (video from Fig. 3). A mixture of equal numbers (100,000 cells) of each cell phenotype (WT (cyan), AA (magenta)) in suspension were applied to Matrigel and the resulting networks tracked by video microscopy for 16 h.

To determine this, we developed a quantitative measure of the likelihood of finding a cell within a certain distance from another cell, relative to the chance that the cells are within that distance randomly. We defined this measure as the neighbor ratio (see materials and methods). The neighbor ratio was calculated for each admixture group (each group contains the same number of total cells) at hourly intervals to determine the best condition for network self-assembly during approximately 20 h. For all pairings (WT:WT, WT:AA, and AA:AA), the neighbor ratio increased in time, indicating overall aggregation (Fig. 3G). The slowest and least efficient network formation over the time interval was observed with the WT to WT group (as judged by requiring 16 h to reach a neighbor ratio of 2) followed by the AA to AA group (requiring 12 h to reach a neighbor ratio of 2), with the most efficient group resulting from the WT:AA mixture (requiring 5 h to reach a neighbor ratio of 2). Since the nearest neighbor ratios have a linear increase with the WT:AA mixture, self-assembly is not a stochastic property (e.g., the result of a random probability), but rather, the heterogeneity appears to be a deterministic property of the tumor population for optimal invasive network formation.

The increase in invasion network formation in vitro suggested that there would be an advantage for invasion of the WT:AA mixed phenotypes. This hypothesis was tested by using an animal model. Further, since AA cells have no ability by themselves to invade into mouse muscle, it was possible that WT:AA mixtures would either be prevented from invasion due to the AA phenotype within the population, or the ability to invade for all cell types would be conferred by the WT phenotype within the population.

The advantage of heterogeneity for successful muscle invasion of xenograft tumor

DU145 cells readily produce tumors in male NOD-SCID mice, which enabled testing the invasive potential of cell lines alone and cell mixtures in vivo. Using an IP injection route, the tumors arising from the two cell lines (DU145^WT and DU145^AA) and a 1:1 mixture were tested for their ability to form colonies on the undersurface of the diaphragm (IP side), invade into the smooth muscle, and traverse to the superior side (lung side). The diaphragm is one of the few anatomical structures where skeletal and smooth muscle co-exist (15). The smooth muscle surface of the diaphragm with the accompanying endothelial lining is an ideal and physiologically relevant model for testing smooth muscle invasion occurring during prostate cancer escape from the gland. In the diaphragm model, the muscle layer is auto-fluorescent and the disruption by tumor can be clearly seen and quantified. Both cell lines and the mixture produced xenograft tumors that increased in volume over an 8-week period. At the end of 8 weeks, the tumors were analyzed for angulated muscle invasive characteristics. All tumors were proliferative as judged by the appearance of mitotic cells within the tumors, and none of the tumors were necrotic. The DU145 α6^WT tumors were able to invade the muscle (Fig. 4A and 4D) compared with no or little invasion in the AA cells or the cells without the α6β1 integrin (A6 KO) (Fig. 4G). Although the 1:1 mixture of the AA:WT cells resulted in approximately 150 μm of invasion, it is remarkable that the composition of the invasive tumor contains both WT and AA cells. In Fig. 4E, white arrows point to the blue nuclei of the WT cells, which in this example of an invasive cluster are outnumbered by the AA cells (brown nuclei). We quantified the AA cells within the mixtures and found that an equivalent number of both WT and AA cells were able to invade the muscle when mixed (Fig. 4H).

Figure 4.

Heterogeneity drives muscle invasion of a specific phenotype in the xenograft model. Tumor invasion of WT (A), AA (B), and WT:AA (C) in a xenograft model by auto-fluorescence imaging of H&E (D is a section of A) tissue sections of mouse diaphragm. The depth of invasion (F), number of invading cells for each cell type, WT, AA, and A6KO (G), and number of cell types in a 1:1 mixture of WT and AA cells (H) was quantified using the entire diaphragm from three mice for each condition. (E) shows that the AA cells within the 1:1 WT:AA mixture breached the muscle barrier (white line) as detected by RFP-H2B reporter (brown nuclei). White arrows point to WT cells (blue nuclei) mixed with the AA cells (brown nuclei) in a muscle-invasive cluster. The statistical analysis, unpaired t-test, was performed using GraphPad Prism 9.4.1 (^∗∗∗∗ = .0001, ^∗∗∗ = .0007, ^∗∗ = .0041, ^∗ = .0276, ns = no significance). Bars in (A)–(D) represent 460 μm. Bar in (E) size represents 70 μm. Asterix in top, center of (C) marks tumor on the superior surface of the diaphragm. (F) Diaphragm surface area invaded: WT, 74.6%; AA, 6.3%; WT:AA, 74.7%. To see this figure in color, go online.

Since the admixture of 1:1 AA:WT dramatically altered both the in vitro and in vivo invasion response, we next determined if the admixtures were conferring specific changes of the biophysical features of the population. Using ECIS measurements, we tested the ability of increasing the proportion of AA cells to WT cells to both decrease cell-ECM and increase cell-cell resistance properties of the population.

Proportional mixing of cell types changed the biophysical performance properties of the tumor cell population

Noninvasive ECIS resistance measurements were collected over a 24-h period for cell populations (250,000 total cells) consisting of either WT only, AA only, or mixtures containing increasing proportions of the AA cell phenotype (Fig. S4). The ability of the cells to attach and spread on the ECM was detected at 40,000 Hz (Fig. S4A) or the strength of cell-cell adhesion by 400 Hz (Fig. S4B) using normalized resistance over time. A consistent reduction in the cell-ECM resistance was observed after 5 h as the proportion of AA cells increased. After 18 h of incubation, the values of the cell-ECM (Fig. S4C) and cell-cell (Fig. S4D) resistance were summarized as an increasing proportion of the AA cells within the population. Increasing the proportion of AA cells continued to decrease the cell-ECM resistance of the population. We note that the biggest drop in cell-ECM adhesion occurred in the 1:1 mixture (50% decrease of suppressor AA, Fig. S4C), which was the condition used in the in vivo experiment (Fig. 4).

The proportional changes in the biophysical performance properties of the tumor population prompted testing if admixtures had a corresponding change in a wound healing response. The rebuilding kinetics of the cell-ECM and cell-cell resistance after a wounding was tested in the admixtures.

Heterogeneity results in increased response to wound healing challenge

ECIS measurements documented the cell-ECM (40,000 Hz) and the cell-cell (400 Hz) adhesion, spreading, and wound responses of both cell phenotypes alone or within 1:1 or 4:1 AA:WT mixtures (Fig. 5A and 5B). As in previous results, the WT cells had cell-ECM attachment normalized resistance values of approximately 2 compared with AA cells of 1.6, whereas the cell-cell resistance values for the AA cells were 20-fold higher than the WT cells just before wounding. The ECIS wound response time for cell-ECM to first reach the pre-wound resistance measurements (2.0 for WT cells; 1.6 for AA cells) was approximately 8 h for both the WT and AA cells (Fig. 5A) compared with approximately 1 h required by the 1:1 and 4:1 AA:WT mixtures. The cell-ECM rebuilding kinetics show that an increased response occurred in the mixtures as the response was eightfold greater in the mixtures compared with populations containing only one cell type. In contrast, the cell-cell rebuilding kinetics show that the ECIS response time to reach the pre-wound levels in the mixtures or the AA cells is not reached during the 45-h time period of the experiment. The WT cells recover the minimal pre-wound cell-cell resistance within 8 h of the wound. Taken together, these data indicate that the improved recovery from wounding afforded by the admixtures is restricted to the cell-ECM response. Additional experiments were done to determine if paracrine effects were a factor in the increased wound healing response. Conditioned media from the WT cells provided to the AA cells did not increase the mutant wound healing ability compared with the AA:WT mixture detected by the ECIS assay (Fig. S5).

Figure 5.

Heterogeneity mixing of the lineages and rebuilding of cell-ECM barrier responses. Unmixed populations (WT or AA) or population mixtures (AA:WT (1:1) or AA:WT (4:1)) were seeded within ECIS chambers in suspension at 37°C and 16 h later wounded (red arrow) and monitored for their response up to 45 h. The populations were tested for the ability to attach, spread, and reform (A) cell-ECM or (B) cell-cell resistance at 400,000 Hz or 400 Hz respectively. Data represent three biological replicates and two technical replicates for each condition (n = 6). Error bars represent standard deviation. To see this figure in color, go online.

Since the ECIS measurements indicated that the cell-ECM phenotype was the primary biophysical response advantaged by the admixtures, we next tested the efficacy of a synthetic ligand mimetic called MTI-101 to alter the biophysical properties of the DU145 population in vitro or in vivo using the mouse xenograft model. We chose the DU145^WT as a test rather than the admixtures since this would model the known heterogeneity of the DU145 population to contain stable phenotype mixtures (36). For comparative purposes, we also determined the cell-ECM wound healing response in 22Rv1 and PC3 prostate cell populations. MTI-101 was used as this is a cyclized version of HYD-1 with improved bioactivity (37). HYD-1 is a synthetic d-amino acid-containing peptide that acts as a ligand mimetic for laminin in epithelial cancers without altering prostate cancer cell survival or cell cycle progression (38,39,40).

A synthetic ligand mimetic (MTI-101) inhibits biophysical response properties and blocks invasion of tumor into the muscle of the mouse

The ECIS wells were either precoated with laminin-332 or MTI-101, and the effects on the cell-ECM resistance properties before and after wounding were investigated (Fig. 6A). Consistent with previous results (Figs. 3 and S4), the cell-ECM normalized resistance of DU145^WT cells reached a value of approximately 2 at 15 h, and approximately 8 h was required to regain pre-wounding resistance after wounding. In marked contrast, whereas the MTI-101 supported cell adhesion and spreading (41), recovery from the wound was not observed during the 45-h course of the experiment. These data suggest that the MTI-101 dramatically affects the ability of the cells to respond to an induced migration response, such as a wound.

Figure 6.

Synthetic ligand mimetic (MTI-101) alters biophysical properties and blocks in vivo invasion of tumor into mouse muscle. The cell-ECM adhesion resistance properties in WT cells as measured by ECIS, 40,000 Hz, up to 45 h (A), were characterized in monolayers coated with or without 1 mg/mL MTI-101 ligand mimetic and the wound response after 16 h (red arrow). The mouse invasion properties of WT and AA cells were measured in the diaphragm model for percentage of surface area invaded (B) or invasion depth (C) with or without treatment by 1 mg/kg MTI-101 ligand mimetic by IP injection. (D) (22RV1 cells) and (E) (PC3 cells) show the response to MTI-101 and the wound response after 60 h (red arrow). The statistical analysis, unpaired t-test, was performed using GraphPad Prism 9.4.1 (^∗∗∗∗ = .0001, ^∗∗∗ = .0002, ^∗∗ = .0086, ns = not determined due to unmatched sample size). Error bars represent standard deviation. To see this figure in color, go online.

We next tested the ability of DU145^WT cells to invade the surface of the diaphragm or the depth of invasion into the contractile muscle using the xenograft mouse model. The design was for DU145^WT and DU145^AA tumors to seed onto the IP surface of the diaphragm for 1 week, followed by weekly intraperitoneal injections of MTI-101. In this way, the design mimics the clinical situation where antitumor strategies are started after tumor detection. As expected, approximately 25% of the diaphragm surface is invaded by DU145^WT cells without MTI-101 treatment, and none of the DU145^AA cells can invade. The MTI-101-treated mice displayed approximately a threefold reduction of the diaphragm surface invasion by DU145^WT cells compared with the untreated mice (to 36%–12%) (Fig. 6B). Unexpectedly, the DU145^AA mice increased surface invasion to 20% in response to the MTI-101 synthetic ligand mimetic (Fig. 6B). For comparative purposes, we analyzed the cell-ECM wound healing response in two other prostate cancer cell populations. MTI-101 as a synthetic ligand mimetic was remarkable since the wound healing response was unable to recover to pre-wounding levels of resistance in the well-differentiated 22Rv1 cell populations (Fig. 6D) while suppressing the maximal cell-ECM adhesion resistance levels in the undifferentiated PC3 cell population (Fig. 6E).

MTI-101 treatment results in nearly undetectable levels of tumor invading deeply into the contractile muscle in the mouse, independent of the phenotype. Taken together, the results show that synthetic ligand mimetic-like MTI-101 can decrease the tumor biophysical responses in vitro and corresponds to preventing the invasion through a contractile muscle layer.

Discussion

The heterogeneity of cell adhesion phenotypes within tumor populations is well recognized as a hallmark of aggressive disease (42). It is a major impediment to effective eradiation and often the source of residual and recurrent disease (43). Lineage tracing experiments in a lung adenocarcinoma animal model by others have highlighted phenotypic heterogeneity and the remarkable plasticity and evolution of the tumor population (44). In this study, our objective was to determine if stable mixtures of cell-ECM and cell-cell phenotype cells, with a defined heterogeneity, had purposeful consequences as predicted by the concept of epithelial-mesenchymal cooperation of tumor phenotypes to invade muscle (45).

Using both in vitro and in vivo approaches, the data suggest that mixed biophysical phenotypes of tumor cells within a population can provide functional advantages for tumor invasion into and through muscle. Of particular importance is that the cell-cell phenotype that was unable to invade through the muscle on its own (DU145^AA) is more resistant to chemotherapeutic agents compared with the DU145^WT and gained the ability to invade into the muscle and reach the superior surface when included in a population containing the DU145^WT cells. This raises the possibility that combination strategies to target both the cell-cell and cell-ECM phenotypes will be a more effective strategy in preventing aggressive disease.

The remarkable cell-cell biophysical properties of the DU145^AA cells and their increased sensitivity to an NAMPT inhibitor compared with the DU145^WT cells is reminiscent of our previous reports of cell adhesion-mediated drug resistance (27,34,46). Future experiments will determine the molecular requirements of survival responses using specific integrin mutations and expression studies. A query of publicly available databases revealed that a strong epithelial phenotype signature positively correlated with cells resistant to inhibitors of HDAC (vorinostat, panobinostat) and topoisomerase I (irinotecan), whereas no correlation was found with bleomycin, doxorubicin, methotrexate, gemcitabine, docetaxel, or bortezomib (47). Future studies will be directed to determine if the biophysical characteristics of the cell-cell adhesion phenotype is a surrogate biophysical marker for drug sensitivity profiles.

We note with interest that the cell-ECM rebuilding kinetics show that an increased response occurred in mixtures since the response was eightfold greater compared with populations containing only one cell type. These data suggest that both cell types have contributory functions for rebuilding the cell-ECM interactions within the population. This is an example of epithelial (cell-cell) and mesenchymal (cell-ECM) cooperation of tumor cells within a population (45) and indicates that targeting the cell-cell phenotype will be an important factor in controlling aggressive tumor populations. These results predict that traction forces of cell-ECM tumor cells will be increased with linkage to tumor cells with a prominent cell-cell adhesion phenotype and will result in an increased tumor population invasion into contractile muscle.

Other groups are identifying cell velocities, cell-cell tractions, and monolayer stresses to analyze spatiotemporal force and motion of specific cells in collective migration (48). This will be applied to increase our understanding functional heterogeneity of biophysical phenotypes and the resulting advantages for invasion into muscle. We note that the appearance of more fluidic types of migration, also known as “unjamming” within the tumor population results in DNA damage and nuclear rupture (49) and can make the tumor cells more susceptible to cytolysis by intraepithelial lymphocytes (50). Understanding the regulation of the biophysical heterogeneity of the tumor population required for both breaching the prostate muscle capsule and reaching Batson’s plexus (51,52) for systemic dissemination will yield important insights into the therapeutic vulnerabilities of lethal prostate cancer.

Conclusion

The distinct nature of the biophysical phenotypes and the ability of ECIS to distinguish these characteristics adds to the potential for in-depth screening of agents to block cell-cell recovery kinetics. In addition, the use of ligand mimetic-like MTI-101 at nontoxic doses to block local invasion into and through the muscle is an alternative strategy to prevent secondary dissemination. With indolent, organ-confined cancers like prostate or bladder cancer, there is a great potential for the usefulness of these approaches to block dissemination using a noncytotoxic approach that would circumvent the known selection pressure of conventional agents that evolve the tumor toward more aggressive disease. We also note that a more complete understanding of cell adhesion and biophysical phenotype signaling within tumor populations holds the potential for local control of tumors to avoid the complexities of systemic tumor control.

Author contributions

A.E.C. conceptualized the paper; A.E.C., J.M.C.G., K.D.M., and L.H. designed the studies; K.D.M. conducted the mouse experiments and created the video imaging in Fig. S3 and the graphs in Fig. S4; K.D.M. and J.M.C.G. conducted the live cell imaging in Fig. 3; W.L.H., J.M.C.G., C.W., and A.E.C. wrote the paper; A.E.C., J.M.C.G., E.J.K., C.W., A.I.P., and W.L.H. created the figures; E.J.K., A.I.P., and J.M.C.G. conducted the ECIS experiments; C.W. provided the nearest neighbor analysis and the equations to make those calculations; B.S.K., R.B.N., and M.S. provided pathology and surgical oncology expertise, respectively. K.D.M., W.L.H., J.M.C.G., A.I.P., L.H., C.W., B.K., and A.E.C. edited the final manuscript.

Editor: Deborah Leckband.