Mechanical Suppression of Breast Cancer Cell Invasion and Paracrine Signaling Requires Nucleo-Cytoskeletal Connectivity

Exercise benefits the musculoskeletal system and reduces the effects of cancer. The beneficial effects of exercise are multifactorial, where metabolic changes and tissue adaptation influence outcomes. Mechanical signals, a principal component of exercise, are anabolic to the musculoskeletal system and restrict cancer progression. We examined the mechanisms through which cancer cells sense and respond to mechanical signals. Low-magnitude, high-frequency signals were applied to human breast cancer cells in the form of low-intensity vibration (LIV). LIV decreased invasion through matrix and impaired secretion of osteolytic factors PTHLH, IL-11, and RANKL. Furthermore, paracrine signals from mechanically stimulated cancer cells, reduced osteoclast differentiation resorptive capacity. Physically disconnecting the nucleus by knockdown of SUN1 and SUN2 impaired the ability of LIV to suppress invasion and production of osteolytic factors. LIV also increased cell stiffness; an effect dependent on an intact LINC complex. These data show that mechanical signals alter the metastatic potential of human breast cancer cells, where the nucleus serves as a mechanosensory apparatus to alter cell structure and intercellular signaling.


Introduction 72
Physical activity has beneficial effects on nearly every organ system. In addition to the 73 positive effects of exercise on cardiovascular (Sattelmair et al., 2011) and musculoskeletal 74 health(Warden and Thompson, 2017), regular physical activity is associated with a reduced risk 75 of colon, endometrial, and breast cancers (Moore et al., 2016;Thune and Furberg, 2001). 76 Low magnitude mechanical forces can be introduced to the musculoskeletal system through 102 platforms that emit low intensity vibration (LIV) signals, serving as an effective "exercise 103 surrogate" by delivering mechanical input similar to that of exercise (Rubin et al., 2001). When 104 applied to mesenchymal progenitor cells LIV promotes proliferation and 105 differentiation (Pongkitwitoon et al., 2016). At the molecular level, LIV signals initiate a signaling 106 cascade resulting in increased phosphorylation of focal adhesion kinase (FAK) and Akt, 107 resulting in downstream activation of RhoA and formation of filamentous actin 108 structures( Thompson et al., 2013). The effects of LIV are additive, with a second bout of LIV 109 enhancing FAK phosphorylation and F-actin contractility (Uzer et al., 2015). 110

Previous work in non-cancerous cells demonstrated that the anabolic effects of mechanical 111
stimuli are enhanced when a refractory period is introduced between loading bouts (Sen et al., 112 2011). As such, the responses produced with two 20-minute bouts separated by three hours 113 were greater than that of a single 40-minute mechanical stimulus. Similar effects were observed 114 in animal models (Patel et al., 2017). These data suggest that low magnitude signals are 115 sufficient for anabolism, but also that proper dosing primes the cells to generate a more robust 116 response with subsequent mechanical stimuli. 117 The mechanical compliance of tumor cells dictates cell behavior, where stiffness of the 118 plasma membrane is inversely proportional to metastatic potential (Mohammadi and Sahai,119 2018). Cells with decreased stiffness display increased migration and invasion, which is 120 regulated by the organization of the actin cytoskeleton (Zhou et al., 2017). Exogenous 121 mechanical input enhances actin cytoskeletal structure (Thompson et al., 2013), and work in 122 non-cancerous cells demonstrates that the nucleus serves as a critical mechanosensory organ 123 where direct connections between the nucleus and the cytoskeleton enable transmission of low 124 magnitude, oscillatory mechanical signals (Uzer et al., 2015). Attachment of the nucleus to the 125 cytoskeleton is enabled by the LINC complex, containing Nesprin and Sun proteins, and may be 126 a means by which exercise influences the metastatic properties of cancer cells. Further, cells 127 from human breast tumors have decreased expression of Nesprin and SUN (Matsumoto et al., 128 2015), suggesting that the LINC complex may regulate tumorigenicity. In this work, we 129 subjected human breast cancer cells to mechanical signals and examined direct biochemical 130 changes of the cancer cells, indirect paracrine signaling alterations, and the biophysical 131 mechanisms enabling transmission of mechanical signals to breast cancer cells. 132 Results 133

LIV does not directly alter cell death but increases the susceptibility to Fas ligand-134 induced apoptosis. 135
Mechanical signals regulate cell death of several cancer types (Lien et al., 2013). To determine if 136 direct application of LIV to breast cancer cells influences cell death, MDA-MB-231 human breast 137 cancer cells were exposed to twenty-minute bouts of LIV (0.3 g, 90 Hz) once-or twice-daily for 138 three days, in the presence or absence of TGF-b1. Control cells were placed on the vibration 139 platform with no transmission of LIV signal. Cell viability was assessed using the 3-(4,5-140 Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) assay. No changes in cell viability 141 were observed with LIV or TGF-b1 treatment (Fig. 1a). mRNA expression of FAS, a cell 142 membrane death receptor, was measured by quantitative polymerase chain reaction (qPCR). 143 Once-daily LIV increased FAS expression by 2-fold, but this change was not significant (Fig.  144 1b). Expression of FAS was significantly (p<0.05) increased by 3-fold following twice-daily LIV 145 (Fig. 1b). Expression of CD95, the gene encoding for Fas ligand was not altered with LIV (Fig.  146   1c). As such, MDA-MB-231 cells were exposed to twice-daily LIV or placed on LIV platforms with no 162 signal transmitted (control) and treated with recombinant Fas ligand (75 ng/ml) for 24 hours prior 163 to staining with Annexin V and subsequent sorting by flow cytometry (Fig. 1d). No changes in 164 apoptosis were observed following LIV in vehicle-treated cells (Fig. 1e). Treatment with Fas 165 ligand did not alter apoptosis in control cells; however, addition of Fas ligand to cells treated 166 with LIV increased (p<0.05) the percent of apoptotic cells compared to control cells by ~2-fold 167 (Fig. 1e). These data demonstrate that application of LIV does not directly induce cell death, but 168 that MDA-MB-231 cells are more susceptible to Fas ligand-mediated apoptosis following 169 exposure to mechanical signals. 170

Low magnitude mechanical signals suppress invasion 171
Extravasation and subsequent metastasis of cancer cells requires invasion through matrix-172 dense borders. To determine if exogenously applied LIV influenced the ability of MDA-MB-231 173 cells to invade through ECM, trans-well invasion assays were performed. Cells were treated 174 with LIV for twenty minutes per bout, once-or twice-daily for three days. Cells then were 175 trypsinized and seeded onto trans-well membranes containing Matrigel ® and visualized using 176 crystal violet (Fig. 2a). Once-daily LIV induced a 1.4-fold reduction in invasion; however, cells 177 exposed to LIV twice-daily had a significant (p<0.01) 3-fold reduction (Fig. 2b). The area of cells 178 that invaded through the trans-well membrane was also quantified, showing a significant 179 (p<0.05) reduction in invasion following twice-daily LIV (Fig. 2c). 180 As invasion through the ECM requires matrix metalloproteinases (MMPs) (Duffy et al., 2000), 181 MMP mRNA levels were quantified by qPCR. MMP1 expression was decreased by 4-fold 182 (p<0.01) following once-daily LIV and by 3-fold (p<0.01) with twice-daily LIV (Fig. 2d). 183 Expression of MMP3 was reduced by 3-fold (p<0.05) following once-daily LIV and by 2.7-fold 184 (p<0.05) with twice-daily LIV (Fig. 2e). No significant changes in MMP9 expression were 185 observed (Fig. 2f). whereas treatment with LIV, once-or twice-daily restricted osteoclast formation (Fig. 3a). 214 Quantification of TRAP stained cultures demonstrated a significant (p<0.0001) reduction in the 215 number of multinucleated cells (³3 nuclei) following exposure of RAW 264.7 cells to conditioned 216 media from MDA-MB-231 cells vibrated once-(7-fold) or twice-daily (3.5-fold) LIV (Fig. 3b). 217 To assess the activity of osteoclasts exposed to conditioned media from MDA-MB-231 cells, 218 RAW 264.7 cells were seeded to osteoassay wells (Corning, Corning NY) containing 219 hydroxyapatite. Cells were treated with conditioned media from MDA-MB-231 cells exposed to 220 once-or twice-daily LIV. RAW 264.7 cells exposed to conditioned media from once-and twice-221 daily LIV treated MDA-MB-231 cells reduced (p<0.01) the resorption pit area by 113-fold and 222 129-fold respectively (Fig. 3c). These data demonstrate that mechanically stimulating MDA-MB-223 231 breast cancer cells reduces their ability to support both osteoclast formation and activity.  plated on Osteoassay wells containing hydroxyapatite and exposed to conditioned media from 246 MDA-MB-231 cells, as described above. The area of hydroxyapatite that was resorbed away 247 from the dish was quantified using ImageJ and normalized to the total area (n=10). replicates. One-way ANOVA p-values: *p<0.05, **p<0.01, ****p<0.0001. 254

The LINC nuclear complex is required for mechanically-induced cellular stiffness 435
The mechanical integrity of cancer cells is inversely proportional to metastatic potential (Zhou 436 et al., 2017). We hypothesized that the suppression of invasion and decreases in 437 osteoclastogenesis, observed following LIV, may be the result of increased cell membrane 438 stiffness. To quantify membrane stiffness, MDA-MB-231 cells were exposed to LIV twice-daily 439 for 3 days, or to non-vibrated conditions. Cells were transfected with either control siRNAs or 440 siRNAs targeting SUN1 and SUN2, as described. Cellular stiffness (elastic modulus) was 441 measured by atomic force microscopy, as in Fig. 8a. In cells treated with control siRNA, LIV 442 increased cellular stiffness (1.2-fold, p<0.05), while no differences were found in cells after 443 knockdown of SUN1/2. 444 As part of the LINC complex, SUN proteins enable connection between the actin 445 cytoskeleton and the nucleus. Low magnitude mechanical forces enhance actin cytoskeletal 446 structure, which may account for the increased cellular stiffness. Exposure of MB-MD-231 cells 447 to LIV twice daily for 3 days resulted in increased actin cytoskeleton stress fiber formation, as 448 stained by Alexa 488-conjugated phalloidin, compared to non-vibrated controls (Fig. 8c). The 449 increased cytoskeletal structure observed with LIV was negated with knockdown of SUN1/2, 450 resulting in phalloidin signal similar to that of non-LIV controls (Fig. 8c). 451

Real Time PCR 483
Total RNA was isolated from cells using RNeasy kit (Qiagen, Germantown, MD), reverse 484 transcribed, and genes were amplified with a BioRad CFX Connect TM qPCR machine, using 485 gene-specific primers, as previously described (Thompson et al., 2015). PCR products were 486 normalized to GAPDH and quantified using the ΔΔCT method (denoted as ΔΔCq).

Low Magnitude Mechanical Force 502
Low magnitude mechanical forces were applied in the form of LIV using a custom-designed 503 platform, as previously described (Pagnotti et al., 2016). Individual culture dishes were placed on 504 the vertically oscillating platform at room temperature (RT). Cells were stimulated at a frequency 505 of 90 Hz at a magnitude of 0.3 g ± 0.025, where 1 g is equal to the earth's gravitational field or 506 9.8 m/s 2 . LIV was applied in 20-minute bouts, once-or twice-daily for 3 days. Twice-daily bouts 507 were separated by 3 hours of rest. Non-vibrated control cells were placed on the vibration 508 platform that was not turned on. 509

Transwell Assays 518
Invasion of MDA-MB-231 cells was quantified using Transwell ® membrane inserts with a 519 pore size of 8.0 μm (Corning ® , Corning, NY). Cells first were seeded to 6-well dishes, exposed cat# 354236, 40 ng/ml). Cells within the transwell insert were incubated at 37°C for 24 hrs, then 527 fixed with methanol (100%, v/v) for 10 min and stained with crystal violet (0.5%, w/v) for 10 min. 528 Cell number and area were quantified under light microscopy. 529

Osteoclast Differentiation and Pit Assays 550
Osteoclast differentiation was determined using murine RAW 264.7 macrophage cells, 551 which were seeded onto 6-well dishes (50,000 cells per well). Cells were maintained in DMEM 552 containing RANKL (75 ng/ml), which was replaced with conditioned media from MDA-MB-231 553 cells after 24 hrs. A 50% mixture of conditioned media and DMEM was added from each group 554 (+/-LIV and +/-siSUN1/2) with RANKL (75 ng/ml). Following 4 days of differentiation, cells were 555 stained with tartrate resistant acid phosphatase (TRAP, Sigma, Cat# 386A-1KT) according to 556 the manufacturer's protocol. Cells were considered to be osteoclasts if they stained for TRAP 557 and had 3 or more nuclei. 558 Resorption activity of osteoclasts, in the presence of conditioned media from MDA-MB-231 559 cells, was quantified using the Osteo Assay (Corning, Corning NY) containing hydroxyapatite. 560 Briefly, 5,000 RAW 264.7 cells were seeded onto the 96-well Osteo Assay plate containing 200 561 μl of DMEM with RANKL (75 ng/ml). The next day, half the volume of media was replaced with 562 conditioned media from MDA-MB-231 cells for each condition. Media was replaced every two 563 days. At day 4, media was removed and 100 μl of bleach (10%, v/v) was added for 5 min at RT. 564 Bleach was removed and cells were washed twice with water. Formation of pits in the 565 hydroxyapatite were visualized under light microscopy and quantified with ImageJ. 566

Determination of Elastic Modulus using Atomic Force Microscopy 567
Cells were indented in media at RT using a BioScope Catalyst AFM (Bruker, Santa 568 Barbara, CA). The AFM was mounted on a Leica DMI3000 inverted microscope (Leica 569 Biosystems Inc., Buffalo Grove, IL), facilitating accurate placement of the AFM probe over 570 individual cells. Indentations were carried out using a spherical borosilicate bead (5 µm 571 diameter) mounted on a gold-coated silicon nitride cantilever (Novascan Technologies, Inc., 572 Boone, IA). Prior to indenting, probes were pushed onto a glass surface and the deflection of 573 the cantilever was used to measure the cantilever's deflection sensitivity (nm/V). The 574 cantilever's spring constant (~0.07 N/m) was then determined using the thermal tuning method. 575 The light microscope was used to navigate to randomly-selected cells. The apex of the 576 cantilever (where the bead is attached) was placed directly above the nucleus, then the AFM 577 was engaged. A single ramp was performed using a trigger force of 1 nN at a speed of 0.5 Hz. 578 Analysis was performed using Nanoscope Analysis v1.70 (Bruker). A linear baseline 579 correction was fit from 30-80% of the retraction curve. Since Poisson's ratio is not fully 580 understood for cells, reduced elastic modulus (Er) was fit for each unloading curve in a region 581 spanning from 20-75% of the maximum force using the Hertz model of contact between a rigid 582 sphere and an elastic half space: 583 ( 1) 584 In equation 1, F is force exerted on the cell, r is the radius of curvature of the probe and d is 585 sample deformation. Only indents with goodness of fits (i.e. r 2 ) greater than 0.97 were included 586 for analysis. 587

Statistical Analysis 588
Statistical variance was expressed as the means ± standard error of the mean (SEM). 589 Evaluation of statistical significance performed by one-way analysis of variance (ANOVA), two-590 way ANOVA, or student's t-test, as appropriate (Prism GraphPad, La Jolla, CA). For AFM 591 assays, two-way ANOVA was used followed by two-stage linear step-up procedure of 592 Benjamini, Krieger, and Yekutieli to control for false discovery rate. Values were considered 593 significant if p £ 0.05. Experiments were replicated at least three times to assure reproducibility. 594 Densitometry data from Western blots were compiled from four biological replicates. replicates. qPCR analyses of (d) sun 1 (SUN1) and (e) sun 2 (SUN2) mRNA from MDA-MB-231 926 cells treated with PBS (Veh) or TGF-β1 and exposed to non-vibration control conditions (Con) or 927 LIV twice-daily for 3 days. qPCR analyses were normalized to GAPDH (n=5). Western blots of 928 whole cell lysates (20 μg per lane) from MDA-MB-231 cells transfected with a control siRNA 929 (siCon) or (f) siRNA targeting SUN1 (siSUN1) or (g) siRNA targeting SUN2 and exposed to non-930 vibration control conditions (Con) or LIV twice-daily for 3 days. PVDF membranes were blotted 931 with an antibody recognizing SUN1 and β-actin as a loading control. Densitometry was 932