Plasminogen Activator Inhibitor 1 (PAI1) Promotes Actin Cytoskeleton Reorganization and Glycolytic Metabolism in Triple-Negative Breast Cancer

This article is featured in Highlights of This Issue, p. 1013

Plasminogen activator inhibitor 1 (PAI1, or serpin family E member 1; SERPINE1) is a serine protease inhibitor and the most prominent negative regulator of the proteolytic urokinase plasminogen activator system. Although originally hypothesized to have tumor suppressor effects, cancers typically upregulate PAI1 with higher levels correlating with worse prognosis in breast and other malignancies (1–3). Given the strong links between PAI1 levels and prognosis in breast cancer, the American Society for Clinical Oncology (ASCO) recommends analysis of PAI1 levels as a biomarker for risk assessment and treatment decisions in lymph node–negative breast cancers (4).

We identified PAI1 as a key driver of cancer cell migration and chemotaxis in triple-negative breast cancer (TNBC) using a high-throughput, single-cell microfluidic device that separates migratory and nonmigratory subpopulations of cancer cells (5). This device enables recovery of migratory and nonmigratory cancer cells for subsequent analysis, which provides a powerful approach to identify cells with highest capacity to migrate within a heterogeneous bulk population. RNA sequencing revealed PAI1 as the most upregulated gene in migratory cells from two different TNBC cell lines. These data reinforce a recent screening study that identified PAI1 as a key gene driving epithelial-to-mesenchymal transition and migration in TNBC (6).

Cells require a large amount of energy to coordinate cytoskeletal reorganization and subsequent migration. Therefore, migrating cancer cells must adopt a metabolic phenotype that allows them to generate energy near sites of rearrangement in the actin cytoskeleton (7). Prior studies indicate cancer cells typically rely on glycolysis rather than oxidative phosphorylation to drive migration and invasion, consistent with studies associating glycolysis with more aggressive malignancies (7–9). Identifying mechanisms promoting glycolytic metabolism in migrating cancer cells offers the potential to block this key process in tumor progression. Recent studies establishing correlations between PAI1 with obesity and metabolic syndrome suggest PAI1 as a molecular regulator of metabolism that may translate to cancer biology (10, 11). In support of PAI1 as a regulator of cancer metabolism, secretion of this molecule by fibroblasts increased mitochondrial mass in cocultured breast cancer cells (12). However, a direct relationship between PAI1 and cancer metabolism remains to be established.

This study investigates mechanisms through which PAI1 drives migration and an aggressive phenotype in TNBC, focusing on actin dynamics and metabolism. Increasing expression of PAI1 in breast cancer cells causes dramatic actin reorganization and drives fragmentation of mitochondria and glycolytic metabolism in 2D and 3D culture environments. Overall, this research for the first time establishes PAI1 as a novel regulator of cancer cell metabolism and highlights the critical interplay between metabolism and motility in breast cancer.

We purchased MDA-MB-231 cells from the ATCC and cultured cells in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin (Thermo Fisher Scientific). We obtained SUM159 cells from Dr. Stephen Ethier (now at The Medical University of South Carolina, Charleston, SC) and cultured cells in F-12 media supplemented with 10% FBS, 1% penicillin/streptomycin, 1% glutamine, 5 μg/mL hydrocortisone, and 1 μg/mL insulin. We authenticated all cells by short tandem repeats analysis and characterized cells as free of Mycoplasma at the initial passage. We used all cells within 3 months after resuscitation, and we maintained all cells at 37°C in a humidified incubator with 5% CO₂.

We cloned human full-length PAI1 fused to NanoBiT (Promega) with a Tev-cleavable linker into the pLVX-Puro vector (ClonTech; PAI1) and verified products by sequencing. We produced recombinant lentiviral vectors and transduced target cells as described previously (13). We first generated cells stably expressing click beetle green luciferase (SUM159-CBG and MDA-MB-231-CBG) as described previously through selection with blasticidin (14). We subsequently transduced cells with the PAI1 viral vector and used puromycin selection for cells stably expressing PAI1. We confirmed expression of PAI1 in these cells through qRT-PCR and luminescence of NanoBiT.

The pLenti.PGK.LifeAct-GFP.W vector was a gift from Rusty Lansford (Addgene plasmid # 51010; University of Southern California, Los Angeles, CA). We transduced wild-type and PAI1 cells with LifeAct-GFP viral vector and sorted cells by flow cytometry to obtain populations with homogeneous expression of LifeAct-GFP. For in vivo fluorescence lifetime imaging microscopy (FLIM) studies, we transduced wild-type and PAI1 cells with mCherry Nuc-FUW viral vector and sorted cells by flow cytometry to obtain a population expressing nuclear mCherry. For 3D spheres, we transduced human mammary fibroblasts (HMF, provided by Daniel Hayes, University of Michigan, Ann Arbor, MI) with mCherry viral vector and sorted for stable cells by flow cytometry.

The pLentiCMV Puro DEST ERK KTRClover was a gift from Markus Covert (Addgene plasmid # 59150; ref. 15). We replaced mClover fluorescent protein with mCitrine and added a nuclear H2B-mCherry and a puromycin selection marker through P2A linker sequences using the NEB HiFi DNA Assembly Kit (New England BioLabs). This vector allows us to visualize ERK activation via a nucleocytoplasmic shuttling event of the mCitrine ERK reporter, while mCherry demarks the nucleus. We cloned the construct into the Piggyback transposon vector (Systems Biosciences) and transfected cells using FuGENE HD (Promega). One week after transfection, we treated cells with puromycin to identify stable integrants and confirmed expression by fluorescence from mCitrine and mCherry.

To analyze levels of PAI1, we performed qRT-PCR for PAI1 and GAPDH using SYBR Green detection as described previously (16). Primers for PAI1 were 5′-CGCAACGTGGTTTTCTC-3′ and 5′-CATGCCCTTGTCATCAATC-3′ and for GAPDH were 5′-GAAGGTGAAGGTCGGAGT-3′ and 5′-GAAGATGGTGATGGGATTTC-3′.

We performed whole transcriptome next-generation sequencing as described previously (17). We deposited these data as GEO accession number GSE125802.

We analyzed effects of PAI1 on cell growth using bioluminescence imaging for CBG with medium binning and 30-second exposure as described previously (14).

We used our previously published microfluidic device and wound-healing assays to verify PAI1 as a regulator of cell migration. We performed microfluidic migration assays and imaged cells as described previously (17). For wound-healing assays, we seeded 1 × 10⁵ cells into 35-mm dishes and allowed cells to form confluent monolayers before creating a linear scratch with a 200-μL pipette tip. We washed dishes once with PBS and then added fresh medium containing the proliferation inhibitor Mitomycin C (1 μg/mL). When indicated, we also added vehicle, the PAI1 inhibitor tiplaxtinin (5 μmol/L) (Selleckchem), or recombinant PAI1 (rPAI1, Sigma-Aldrich, 40 nmol/L) at the time of wounding. We measured the size of the wound in each monolayer of cells at 0 and 17 hours. We used the following formula to calculate wound closure over time: (1−(wound width at 17 hours/wound width at 0 hours)) × 100.

We also analyzed the ability of PAI1 to modulate cell adhesion. We seeded 1 × 10⁵ HMFs expressing mCherry into a 24-well plate 48 hours before the assay. After 48 hours, we seeded 2.5 × 10⁵ MDA-MB-231 and SUM159 wild-type or PAI1 LifeAct-GFP–expressing cells onto the HMFs and incubated at 37°C for 15 minutes. We removed nonadherent cells with PBS and visualized adherent cells by fluorescence imaging. We present data as the number of adherent cells to total number of cells seeded relative to matched wild-type cells.

To visualize the actin cytoskeleton, we performed immunofluorescence staining. We seeded 2.5 × 10⁴ cells on glass cover slips and incubated overnight. We fixed cells with 4% formaldehyde for 15 minutes at room temperature and then washed three times with PBS for 5 minutes each. We permeabilized cells with ice-cold 100% methanol for 10 minutes at −20°C; washed with PBS for 5 minutes; and blocked with 10% goat serum for 1 hour at room temperature. Next, we incubated cells for 1 hour at room temperature with an antibody against phosphorylated paxillin (p-paxillin, Cell Signaling Technology) in 5% goat serum. After incubation, we washed cells with PBS 3 times for 5 minutes each and costained with Alexa Fluor 488–conjugated secondary antibody (Jackson ImmunoResearch Laboratory) for p-paxillin and conjugated Texas Red-X phalloidin (Thermo Fisher Scientific) for actin for 1 hour at room temperature in the dark. We washed the slides and mounted with medium containing an anti-fade reagent and DAPI (ProLong Gold, Thermo Fisher Scientific). We acquired images on an Olympus IX73 microscope with a DP80 CCD camera (Olympus) and we analyzed epifluorescence images with cellSens software (Olympus) and stress fibers with in-house MATLAB code.

We formed spheroids as described previously (14), coculturing MDA-MB-231 or SUM159 wild-type or PAI1 LifeAct-GFP–expressing cells with human mammary fibroblasts (HMF) expressing mCherry. We embedded spheroids or single cells (1 × 10⁶/mL) in fibrin gels (final concentration 4 mg/mL fibrin, 2.5 U/mL thrombin, and 0.01 U/mL aprotinin; ref. 18) and imaged actin and fibroblasts by two-photon microscopy. We analyzed the perimeter of spheroids using in-house MATLAB code. We mechanically dissociated spheroids after 24 hours by pipetting up and down 5 times with a micropipette in the bottom of a 96-well plate. We enumerated dissociated cells by fluorescence microscopy.

To quantify activation of ERK by PAI1, we used cells stably expressing a validated kinase translocation reporter (KTR) for this kinase. This reporter utilizes a known downstream substrate of ERK to drive reversible translocation of the reporter into and out of the nucleus (15). We seeded 1.2 × 10⁵ MDA-MB-231 and 8.5 × 10⁴ SUM159 wild-type cells containing the KTR for ERK and a nuclear marker (H2B-mCherry) on to 35-mm dishes with a 20-mm glass bottom (Cellvis) in FluoroBrite DMEM media (Thermo Fisher Scientific, A1896701). After 2 days, we changed medium to low (1%) serum in Fluorobrite DMEM and added tiplaxtinin (5 μmol/L) where indicated. The next day, we treated cells with either rPAI1 (40 nmol/L) or vehicle control and acquired images at times listed in the figure. We analyzed images for activation of ERK using custom MATLAB code.

We used FLIM to quantify metabolic differences between wild-type and PAI1 cells based on endogenous fluorescence from NADH as described previously (ISS FastFLIM; refs. 19, 20). For both NADH and mCherry, we used 740 nm excitation, 25 × NIR corrected objective, 512 × 512 matrix, 15% laser power, 4 μs dwell time, and 2.5× electronic zoom (Olympus FVMPE-RS upright microscope with Spectra-Physics Insight DS+ laser). We separated light emitted from NADH and mCherry with band pass filters of 410–460 nm and 545–645 nm, respectively.

We exported modulation lifetime and direct counts (DC, intensity) data as “.tiff” files from ISS VistaVision into MATLAB alongside red-channel intensity data obtained by two-photon microscopy from the Olympus software for analysis. We filtered images to remove all data points with NADH lifetimes less than 0.5 ns and greater than 5 ns, as well as points with DC's less than 15 as per the manufacturer's instructions. We created a mask using mCherry to mark locations of cancer cells in images. We analyzed NADH lifetimes of pixels within the mask for each cell and present lifetimes on images based on a pseudo color scale.

To confirm our FLIM data, we performed metabolic flux analysis to measure the glycolytic capacity of cells by measuring the extracellular acidification rate (ECAR) of cells using a Seahorse Bioscience XF^e96 Extracellular Flux Analyzer and XF Cell Energy Phenotype test kit (Agilent) as described previously (19).

The University of Michigan Institutional Animal Care and Use Committee approved all animal procedures (protocol 00006795). The animals used in this study received humane care in compliance with the principles of laboratory animal care formulated by the National Society for Medical Research and Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences and published by the NIH (Bethesda, MD; publication no NIH 85-23, revised 1996). We established orthotopic tumor xenografts in the fourth inguinal mammary fat pads of 17–21-week-old female NSG mice (The Jackson Laboratory; ref. 16), implanting 2 × 10⁵ MDA-MB-231 or SUM159 wild-type or PAI1 cells that express LifeAct-GFP. We quantified tumor growth and metastasis by bioluminescence imaging as described previously (13). We euthanized mice 7–8 weeks after injection (n = 5 mice in each group) and then immediately visualized collagen and actin by two-photon microscopy. We used an 880 nm excitation wavelength and collected second harmonic signal from fibrillary collagen and emission from LifeAct-GFP with 410–460 nm and 495–540 nm filters, respectively (21). For FLIM experiments, we implanted 2 × 10⁵ MDA-MB-231 or SUM159 wild-type or PAI1 cells expressing mCherry Nuc-FUW to demark nuclei of cancer cells. Six weeks after injection, we euthanized mice, removed the tumor and lungs, and immediately imaged metabolism of cancer cells in orthotopic tumors or lung metastases ex vivo by FLIM. We note that previous research demonstrates NADH/FAD metabolism of cancer cells maintained in intact tumors and organs ex vivo remains the same as in vivo for up to eight hours (22, 23). To calculate numbers of cells in an image, we used FIJI (24) to binarize and analyze numbers of nuclei in a region. We then pooled and plotted the density from each image.

MATLAB programs used to measure the kinase translocation reporter (KTR), collagen directionality, stress fiber density, and spheroids are available through a Material Transfer Agreement (MTA).

We used a nonparametric Mann–Whitney U test for comparisons of cell migration and motility with a significance level of 0.05 considered statistically significant. For sphere formation and other experiments, we used two-tailed, unpaired Student t tests. We prepared bar graphs (mean values + SD or SEM as denoted in figure legends) and box plots and whiskers using GraphPad Prism 7 or Origin 9.0. For box plots and whiskers, the bottom and top of a box define the first and third quartiles, and the band inside the box marks the second quartile (the median). The ends of the whiskers represent the 5th percentile and the 95th percentiles, respectively. The “+” inside the box indicates the mean, dots outside the box and whiskers indicate outliers, and the “x” refers to the maximum and minimum of all data.

We recently used our migration-based microfluidic device to identify critical positive and negative regulators of cell migration (ref. 17; Fig. 1A and B). RNA-sequencing data revealed plasminogen activator inhibitor 1 (PAI1) as one of the most highly upregulated genes (∼8-fold relative to nonmigratory cells, P < 0.0001) in both highly migratory SUM159 and MDA-MB-231 triple-negative breast cancer (TNBC) cells. To investigate effects of PAI1 on migration and motility of TNBC cells in vitro and in vivo, we stably expressed PAI1 in SUM159 and MDA-MB-231 cells. qRT-PCR analysis revealed higher expression of PAI1 in stably transduced SUM159 and MDA-MB-231 PAI1 cells compared with wild-type control (Supplementary Fig. S1). RNA sequencing with gene set enrichment analysis (GSEA) identified significant associations with cell movement and organization of the cytoskeleton/cell projections in PAI1 cells (Table 1), confirming the effect of PAI1 on cell migration. Stable expression of PAI1 significantly increased migration of both SUM159 and MDA-MB-231 cell in our migration device (Fig. 1C and D) and a wound-healing assay (Fig. 1E and F). Adding exogenous recombinant active PAI1 (rPAI1) also significantly increased migration of wild-type cells from both cell lines, and a PAI1-specific inhibitor, tiplaxtinin, significantly reduced cell migration (Fig. 1E and F; Supplementary Fig. S1). Together, these data show that elevated expression of PAI1 enhances migration.

Cell migration critically depends on the actin cytoskeleton. Lamellipodia, membrane ruffles, filopodia, and stress fibers constitute the most important and studied actin-based structures for motility. A lamellipodium and membrane ruffles demark actin-rich sheet-like projections of a motile cell, while a filopodium defines a finger-like projection extending from a lamellipodium. Stress fibers form from contractile actomyosin bundles. Effects of PAI1 on each of these actin-based structures remain poorly defined. Phalloidin staining revealed dramatic reorganization of the actin cytoskeleton in PAI1 cells compared with wild type. Stable expression of PAI1 or addition of rPAI to cells increased the formation of actin-rich migratory structures (lamellipodia and membrane ruffles; Figs. 2A and 3A) and reduced stress fiber density (stress fiber area divided by total cell size; Supplementary Fig. S2) compared with control cells. In addition, we found that PAI1 increases, and tiplaxtinin reduces, the aspect ratio (length/width) of cells (Figs. 2B and 3B), which suggests PAI1 supports a more mesenchymal morphology associated with cell migration.

To study actin dynamics in living cells, we stably expressed LifeAct-GFP, a 17-amino-acid peptide that binds to filamentous actin (F-actin) and does not interfere with actin dynamics (25), in both wild-type and PAI1 cells. To confirm our phalloidin staining, we imaged SUM159 PAI1 and control cells expressing LifeAct-GFP as they migrated through our microfluidic device. Consistent with our previous data, we found that PAI1 cells had greater actin-rich migratory structures compared with control cells (Supplementary Video S1 and S2). Next, we uniformly embedded single cells expressing LifeAct-GFP into fibrin gels and allowed them to proliferate. Fibrin promotes adhesion and migration of cells and provides structural integrity for breast tumors (26). Aggregates of SUM159 PAI1 cells displayed a more invasive phenotype with more actin-rich structures extending into the surrounding gel (Fig. 2C). These projections contained concentrated actin-based leading edges, suggesting an actin-rich migratory structure such as lamellipodia or membrane ruffles. We also noted that SUM159 PAI cells formed smaller aggregates than wild-type cells. MDA-MB-231 cells did not form aggregates in gels, independent of PAI1 expression. However, wild-type MDA-MB-231 cells did form grape-like nonadherent structures, whereas PAI1 cells formed adherent structures that displayed actin-based projections (Fig. 3C).

We next formed spheroids using LifeAct-GFP cells and human mammary fibroblasts (HMF) and embedded spheroids into fibrin gels to investigate effects of PAI1 on cancer cells in a physiologically relevant, coculture 3D model. While HMFs showed similar phenotypes between groups, wild-type LifeAct-GFP cells displayed a compact, spherical shape with a defined edge (Fig. 2D and 3D). By comparison, both SUM159 and MDA-MB-231 PAI1 cells demonstrated greater invasion into the adjacent fibrin gel (Figs. 2D and 3D). We also discovered that cancer cells near the edge of PAI1 spheroids showed more actin-rich migratory structures, suggesting actin reorganization and enhanced motility. Using custom MATLAB code we identified the margins of cancer cells in spheroids. Relative to spheroids with control cells, PAI1 spheroids contained more projections from a spheroid, and these projections extended farther from the spheroid core into the adjacent gel (Figs. 2D and 3D). Overall, these data establish that PAI1 promotes actin cytoskeleton reorganization to display a more migratory phenotype in 2D and 3D settings.

In addition to cell migration, GSEA also revealed multiple sets with associations to cell adhesion or attachment (Table 1). We noted that cancer cells invading from a spheroid into the gel tracked closely with HMF cells, suggesting that PAI1 promotes intercellular adhesion. To test this hypothesis, we mechanically dissociated coculture spheroids of cancer cells and HMFs and then enumerated single cancer cells. We found more single wild-type cancer cells following mechanical disruption of spheroids, consistent with reduced adhesion relative to cancer cells with elevated PAI1 (Supplementary Fig. S3). We also identified greater adhesion of PAI1 cancer cells to HMFs in monolayer culture (Supplementary Fig. S3). Together, these data demonstrate that PAI1 promotes intercellular adhesion of breast cancer cells.

To investigate mechanisms through which PAI1 promotes adhesion, we focused on focal adhesions, structures of the actin cytoskeleton that form anchor points to the extracellular matrix (ECM). Immunofluorescence staining for paxillin, a protein component of focal adhesions, showed reduced numbers of focal adhesions in breast cancer cells with elevated PAI1 relative to control (Supplementary Fig. S4). This result suggests that PAI1 increases adhesion of breast cancer cells independent of focal adhesions.

MAPK signaling through ERK defines one prominent pathway that promotes reorganization of actin, turnover of focal adhesions, and cell motility (27). We investigated to what extent PAI1 activates ERK in single cells using a kinase translocation reporter (KTR) system (15). Using cells stably expressing the ERK KTR, we found that exogenous addition of rPAI1 to both wild-type SUM159 and MDA-MB-231 cells increased kinase activity of ERK at 20 minutes (Fig. 4A; Supplementary Fig. S5). In addition, tiplaxtinin inhibited activation of ERK by rPAI1 (Supplementary Fig. S5). Activation of ERK provides one potential mechanism through which PAI drives cell migration.

Recent studies show that ERK promotes mitochondrial fission to drive tumor growth (28) and cellular reprogramming (29), and more migratory cancer cells frequently exhibit greater fission of mitochondria. Staining with MitoTracker Green showed greater fragmentation of mitochondria, indicative of mitochondrial fission, in both SUM159- and MDA-MB-231-PAI cells. Treatment of wild-type parental cells with rPAI1 also shifted mitochondria from more fused, elongated structures to a more fragmented morphology, characteristic of fission (Supplementary Fig. S6).

Having established that PAI1 promotes mitochondrial fission, we next investigated effects of PAI1 on cell metabolism in vitro. We used FLIM of fluorescence from endogenous NADH to measure relative glycolytic versus oxidative metabolism in single TNBC cells (30). Cells predominantly relying on glycolysis have greater amounts of free NADH, which exhibits a shorter lifetime than protein-bound NADH in cells utilizing oxidative phosphorylation (OXPHOS; ref. 31). We determined that PAI1 promotes glycolytic metabolism in both SUM159 and MDA-MB-231 cells in culture as shown by a shorter lifetime for fluorescence from NADH (Fig. 4B). We also found that PAI1 promotes glycolytic metabolism in cells seeded in fibrin gels (Fig. 4C). Consistent with our FLIM data, we found that PAI1 increased the extracellular acidification rate (ECAR) in both SUM159 and MDA-MB-231 cells (Supplementary Fig. S7), suggesting an increase in glycolytic capacity in cells that express PAI1. Together, these data demonstrate that PAI1 shifts cellular metabolism to glycolysis.

To extend our studies to an in vivo setting, we orthotopically implanted SUM159 or MDA-MB-231 wild-type or PAI1 cells that express LifeAct-GFP into the fourth mammary fat pad of NSG mice. Consistent with our data showing that PAI1 cells form fewer (Supplementary Fig. S8) and smaller tumor spheres than wild-type cells (Figs. 2C and 3C), mice injected with PAI1 TNBC cells produced significantly smaller tumors (Fig. 5A). Even though we saw significantly smaller orthotopic tumors, we did not observe any differences in growth of PAI1 versus wild-type cells in standard culture (Supplementary Fig. S8). Interestingly, we found reduced density of cancer cells in PAI1 tumors (number of cancer cells divided by total tumor area; Fig. 5B and C). PAI1 TNBC cells also displayed more actin-rich migratory structures than cancer cells in wild-type tumors (Fig. 5B). In addition to imaging cancer cells, we analyzed the architecture of peritumoral collagen by second harmonic two-photon microscopy. Alignment of collagen fibers perpendicular to the margin of a tumor defines a more invasive phenotype with poor prognosis in breast cancer (32). Tumors formed from SUM159- and MDA-MB-231-PAI cells showed greater alignment of collagen perpendicular to the tumor margin than wild-type, consistent with a more locally invasive phenotype (Fig. 5B and D; Supplementary Fig. S9). We note that both bioluminescent and two-photon imaging revealed no differences in metastasis between PAI and control tumors (data not shown), potentially because we euthanized mice at a relatively late time point (8 weeks) after implantation to facilitate imaging by two-photon microscopy.

To determine effects of PAI on metabolism of cancer cells in vivo, we used FLIM to quantify glycolysis versus OXPHOS in orthotopic tumors and lung metastases produced by SUM159 or MDA-MB-231 wild-type or PAI1 cells. For these studies, we used cancer cells expressing nuclear mCherry to avoid interference from GFP in detecting NADH and distinguish disseminated cancer cells. Unlike studies that recover cancer cells from tumors and analyze metabolism of bulk populations of dissociated cancer cells in standard culture, FLIM allows quantitative analysis of metabolism in single cancer cells in intact tumors and metastases. We analyzed mice six weeks after injection. Both SUM159 and MDA-MB-231 PAI1 cells in orthotopic tumors displayed more glycolytic metabolism (Fig. 6A and B). In addition to orthotopic tumors, we discovered enhanced glycolysis in MDA-MB-231 PAI1 cells in lung metastases (Fig. 6C). We were unable to detect SUM159 metastases independent of PAI1 expression at the time of imaging. These data establish that PAI1 promotes glycolytic metabolism at both the orthotopic tumor and the lung metastatic site.

Developing more effective treatments to block local invasion and metastasis in cancer requires mechanistic understanding of pathways driving cancer cell migration. While well-established as a prognostic and predictive biomarker in breast cancer (1), targeting PAI1 for cancer therapy remains challenging due to limitations of existing drugs and incomplete understanding of functions of PAI1 in tumor progression (33). While other studies have identified PAI1 as an important regulator of actin dynamics and cell migration (34–36), mechanisms for how PAI1 promotes cancer cell migration and tumor progression remain incompletely defined. Here we make the unique discovery that PAI1 activates ERK, connecting PAI1 signaling to glycolytic metabolism and cytoskeletal reorganization.

Building on our discovery that highly migratory TNBC cells express high levels of PAI1, this study established that PAI1 drives cell migration by promoting cytoskeletal reorganization and glycolytic metabolism. Recent studies show migrating cancer cells utilize energy generated through glycolysis rather than OXPHOS (7, 8), suggesting that elevated glycolysis mediated by PAI1 drives cell migration. Mechanistically, we observed that PAI1 activates several processes that potentially link cell migration and glycolytic metabolism. Cells with high expression of PAI1 (i) activate ERK; (ii) exhibit fragmented mitochondria; and (iii) utilize glycolysis to a greater extent than control breast cancer cells in environments from 2D cultures to metastases in mice. Overall, our data suggest a novel metabolic regulatory function for PAI1 in migration and tumor progression.

Dysregulation of cell migration during cancer progression shapes the metastatic capacity of cancer cells. Cell migration is the result of a multi-step process driven by dynamic changes in the actin cytoskeleton. Consistent with other studies (37, 38), we found that PAI1 expression reduces stress fiber density while increasing the formation of actin-rich migratory structures in standard cell culture. We also found that cells expressing PAI1 displayed reduced focal adhesions as determined by paxillin staining. Paxillin is a key component of integrin signaling, and phosphorylation of paxillin is required for integrin-mediated reorganization of the actin cytoskeleton. Previous work demonstrated that phosphorylation of paxillin by ERK promotes translocation of paxillin from focal adhesions to the cytosol, resulting in disassembly of focal adhesions (39, 40). Therefore, one explanation for reduced focal adhesions in cells expressing PAI1 may be through activation of ERK. Although cells displayed a more migratory phenotype in culture conditions, breast cancer cells with elevated expression of PAI1 exhibited greater adhesion to HMFs. Apparent discordance between greater migration and adhesion may arise from measuring these processes in 2D and 3D environments, respectively. Furthermore in orthotopic tumors, we found that PAI1 promotes alignment of tumor-associated collagen perpendicular to the margin of a tumor, a signature associated with tumor invasion and poor prognosis (32). These data suggest that PAI1 enhances migration not only through effects on cancer cells, but also on the surrounding tumor microenvironment.

Reprogramming of cellular metabolism represents a hallmark feature of cancer with malignant cells frequently shifting to glycolytic metabolism (41). Breast cancers upregulate glycolytic enzymes relative to normal breast tissue and benign breast lesions, and glycolytic metabolism correlates with worse outcomes (42, 43). Mechanisms underlying a switch from oxidative to glycolytic metabolism in cancer cells remain incompletely defined. Here, we discovered a previously unknown function of PAI1 to promote glycolysis in TNBC cells, a metabolic switch linked to cell migration (7). Consistent with a previous study showing mitochondrial effects of PAI1 (12), we found that PAI1 promotes mitochondrial fission in TNBC cells. Mitochondrial fission promotes a shift to glycolytic metabolism in both normal and malignant settings (44–46), linking changes in mitochondrial morphology to effects of PAI1 on metabolic reprogramming. ERK signaling drives mitochondrial fission (28, 29), suggesting activation of this kinase as a mechanism through which PAI1 promotes mitochondrial fission and glycolysis. In addition, we found that PAI1 cells retain their enhanced glycolytic phenotype following colonization to lung. This contrasts with other studies demonstrating that breast cancer cells engage OXPHOS to metastasize to the lung (47, 48). Differences among studies may be due to the use of human breast cancer cells in our current study versus murine breast cancer cells in prior publications. In addition, we quantified metabolism in single cells in situ using FLIM of endogenous NADH (49, 50), while studies showing OXPHOS in lung metastases relied on profiles of gene expression and cell culture assays with populations of cancer cells selected for enhanced lung metastasis. Our data establish that PAI1 drives glycolysis in cell-based assays, orthotopic xenografts, and lung metastases, suggesting that PAI1 glycolytic reprogramming as an integral function of PAI1 in breast cancer.

In summary, this study not only provides the first evidence that PAI1 promotes glycolytic metabolism but also suggests an underlying mechanism mediated by ERK and known effects of this kinase on mitochondrial morphology. Our data support recent research proposing that highly migratory cells rely on energy obtained primarily through enhanced glycolysis for migration and chemotaxis (7, 8). Data from our study provide a strong rationale for targeting cancer cell metabolism as a potential therapeutic approach not only for aggressive and highly migratory cancers, such as TNBC, but also for patients with elevated levels of PAI1 in primary tumors and/or circulating tumor cells.

G.D. Luker is a consultant/advisory board member for Polyphor. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.A. Humphries, E. Yoon, K.E. Luker, G.D. Luker

Development of methodology: B.A. Humphries, J.M. Buschhaus, Y.-C. Chen, B. Chiang, Y.-H. Cheng, E. Yoon, K.E. Luker, G.D. Luker

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.A. Humphries, J.M. Buschhaus, Y.-C. Chen, H.R. Haley, T. Qyli, B. Chiang, N. Shen, A. Cutter, K.E. Luker, G.D. Luker

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.A. Humphries, J.M. Buschhaus, Y.-C. Chen, H.R. Haley, T. Qyli, B. Chiang, S. Rajendran, A. Cutter, J. Cong, P.C. Spinosa, K.E. Luker, G.D. Luker

Writing, review, and/or revision of the manuscript: B.A. Humphries, J.M. Buschhaus, Y.-T. Chen, K.E. Luker, G.D. Luker

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-T. Chen, E. Yoon, K.E. Luker, G.D. Luker

Study supervision: E. Yoon, K.E. Luker, G.D. Luker

The authors acknowledge funding from NIH grants R01CA196018, R01CA195655, U01CA210152, and R37CA222563. This study was supported by NIH grants R01CA196018, R01CA195655, U01CA210152, and R37CA222563.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.