Targeting Cancer Cell Adhesion Molecule, CD146, with Low-Dose Gold Nanorods and Mild Hyperthermia Disrupts Actin Cytoskeleton and Cancer Cell Migration

Abstract

Cluster of differentiation 146 (CD146), a cancer cell adhesion molecule, is over-expressed on the surfaces of melanoma, breast, ovarian, and prostate cancer cells, and its high expression indicates the migration tendency of these cancer cells and poor patient prognosis. Here, we hypothesize that targeting the CD146 with low-dose gold nanorods combined with mild hyperthermia can stop the migration of these cancer cells. Two metastatic cancer cells including a melanoma and a breast cancer cell line are selected as the model systems. Cell migration assays show that the migration of both cell lines can be completely stopped by the treatment. Atomic force microscopy and super resolution fluorescence microscopy reveal the alterations of actin cytoskeleton and cell morphology correspond to the inhibited cell migration. Further mechanistic analysis indicates the treatment disrupts the actin cytoskeleton by a synergistic mechanism including depleting membrane CD146 and interfering ezrin-radixin-moesin phosphorylation. As a result, we believe targeting CD146 with low-dose gold nanorods and mild hyperthermia could be a versatile, effective, and safe approach for stopping cancer metastasis. More broadly, the concept of targeting cancer cell surface markers that connect the underlying actin cytoskeleton, offers enormous potential in treating cancer metastasis, which accounts for more than 90% of cancer-associated mortality.

Graphical Abstract

Introduction

Metastasis accounts for more than 90% of cancer-associated mortality.[1, 2] In this process, cancer cells originating from a primary tumor begin invading surrounding tissues, then enter the blood or lymphatic vessels (i.e., intravasation) and translocate to distant sites, finally exiting from the vessels (i.e., extravasation) and adapting to the microenvironment of the new sites to facilitate cell proliferation and form secondary tumors.[3, 4] To invade surrounding tissues, cancer cells should gain the migration ability, which is mediated by the actin cytoskeleton.[5–7] Typically, two characteristic structures of the actin cytoskeleton are critical to the migration of cancer cells: (i) Lamellipodium: the protrusion of a broad leading membrane that initiates the directed movement, consisting of a dense meshwork of actin filaments underneath the flat, thin membrane;[8] and (ii) Stress fibers: actin filament bundles along the cell body that mediate contractile forces to pull the cell body forward.[9] To date, studies have attempted to develop anti-metastatic drugs to disrupt the actin cytoskeleton of cancer cells and prevent cancer cell migration.[10] For example, several candidates including cytochalasins, latrunculins, and geodiamolide H have been shown to destabilize the actin cytoskeleton and inhibit cancer cell migration.[11–13] Unfortunately, these actin-binding molecules had only modest success in clinical trials, mainly because of two limitations: First, their nonspecific binding to actin in healthy cells could cause toxicity to healthy cells such as cardiovascular cells, thereby raising the safety concerns in terms of cardiovascular risks.[14, 15] Second, after several months’ treatment, cancer cells develop drug resistance by protein mutation, causing a loss of drug efficacy.[16, 17] Therefore, it is important to look for alternative approaches to inhibit cancer metastasis.

In recent years, a growing number of researches have demonstrated that nanomaterials internalized by cancer cells inhibit cancer cell migration.[18–24] Among these nanomaterials, gold nanoparticles have received increased interest owing to their good biocompatibility, tunable size, shape, and surface chemistry, as well as their plasmonic photothermal effect.[25–30] For instance, Murphy and co-authors in 2013 showed that gold nanoparticles with different charges and sizes decrease the migration of cancer cells.[31] One year later, Chen, Wei and co-authors reported that serum-protein coated gold nanorods can inhibit cancer cell migration through decreasing ATP production, thereby impairing the actin assembly to inhibit the cancer cell migration.[32] While these earlier studies mainly focused on nonspecific targeted gold nanoparticles, recent studies conducted by El-Sayed and co-authors in 2017 and 2018 demonstrated that cancer cell integrin targeted gold nanorods combined with photothermal treatment can inhibit cancer cell migration by introducing broad perturbations in integrin regulated signaling pathways, including actin, microtubule, Rho GTPases, and kinase-related pathways; all these migration-related pathways are targeted toward the anti-migration direction.[33, 34] However, summarizing these pioneering work (as shown in Table S1), there are several drawbacks that are noteworthy: (i) All these previous studies using gold nanoparticles can only partially inhibit or slow the migration of cancer cells, rather than completely stop their migration; (ii) Despite its effectiveness, targeting integrin may not be the optimal solution since integrin is an upstream surface receptor that controls cell migration in many different ways (directly or indirectly), and the inhibition of cancer cell migration by integrin targeted gold nanorods is the integrated outcome of global changes in several migration-related pathways.[33, 34] In other words, the exact molecular mechanism remains elusive; (iii) The dose of gold nanoparticles was relatively high in these studies (e.g., 50 μM for serum-protein coated gold nanorods, and 2.5–5 nM for integrin targeted gold nanorods),[32–34] which could be a safety concern, as recent studies showed that exposure of low-dose gold nanoparticles for only once can induce gene expression and morphological changes in human cells in a long term;[35, 36] and (iv) The laser irradiance (e.g., 5.8 W/cm² for integrin targeted gold nanorods) significantly exceeds the limit of skin tolerance (0.4 W/cm² at 808 nm, by American National Standards Institute),[33, 37] which may cause damage to surrounding normal tissues. Based on these considerations, we envision that an ideal approach and outcome would involve using a minimal dose of gold nanoparticles with low laser irradiance to specifically target cancer cell surface markers that are directly related to the actin cytoskeleton and metastatic ability, and in so doing, completely stop cancer cell migration.

In this work, we focus on targeting a metastatic cancer cell adhesion molecule, CD146. This transmembrane protein was initially named as a melanoma-specific cell-adhesion molecule (i.e., MCAM) and is over-expressed on the surface of metastatic human melanoma cells.[38, 39] Indeed, studies indicated that CD146 is also over-expressed on the surfaces of metastatic breast cancer, ovarian cancer, and prostate cancer cells, and that CD146 overexpression is directly related to enhanced motility of these cancer cells and poor patient prognosis.[40–43] The molecular mechanisms of how CD146 promotes cancer cell migration have been well-studied by Yan and co-authors.[44, 45] In contrast to integrin, which involves many adaptor proteins and migration-related pathways, CD146 physically interacts only with three adapter proteins (ERM: ezrin-radixin-moesin) and further links the actin cytoskeleton to the cell membrane.[44] As shown in Figure 1, the transmembrane CD146 not only serves as an “anchor” to recruit actin to the cell membrane through ERM proteins (i.e., the “chain”), mediating cancer cell migration, but also activates the highly conserved ERM proteins in cytosol by phosphorylation, allowing phosphorylated ERM (p-ERM) proteins to translocate from the cytosol to the membrane-cytoskeleton interface. According to the roles of CD146 in regulating the actin cytoskeleton and cell migration, we propose two hypotheses: (i) The uptake of CD146 targeted gold nanorods will deplete the cell membrane CD146, and thereby disrupt the actin cytoskeleton and cell migration (i.e., remove the “anchor”); and (ii) Mild hyperthermia will interfere with the phosphorylation of ERM proteins to further impair migration (i.e., break the “chain”).

Figure 1.

The concept of using CD146 target gold nanorods (AuNRs) combined with mild hyperthermia to disrupt the actin cytoskeleton and migration of the cancer cells. On the surfaces of metastatic melanoma and breast cancer cells, transmembrane CD146 not only serves as an “anchor” to recruit actin cytoskeleton to cell membrane through ERM proteins (i.e., the “chain”) to modulate the cancer cell migration, but also activates the highly conserved ERM proteins in cytosol by phosphorylation, allowing activated ERM (phosphorylated ERM, p-ERM) proteins to translocate from the cytosol to the membrane-cytoskeleton interface. Two hypotheses will be tested in this work: (i) The uptake of CD146 targeted gold nanorods will deplete the cell membrane CD146, and thereby disrupt the actin cytoskeleton and cell migration (i.e., remove the “anchor”); and (ii) Mild hyperthermia will interfere with the phosphorylation of ERM proteins to further impair migration (i.e., break the “chain”).

To test our hypotheses, we first synthesized CD146 antibody coated gold nanorods. The dual functions of the CD146 antibody lie in specifically targeting nanorods to the cancer cells and disrupting the downstream regulators of CD146 including ERM proteins and actin. Therefore, our approach is distinct from the dual-targeted strategy that requires two different ligands on nanoparticle surfaces.[46, 47] As a control, nonspecific targeted BSA-coated gold nanorods were also prepared. Strikingly, our cell migration assays showed that low-dose CD146 targeted gold nanorods with mild hyperthermia can almost completely stop the migration of two metastatic human cancer cell lines (SKMEL-2, a melanoma cell line, and MDA-MB-231, a breast cancer cell line) without affecting the cell viability and membrane integrity. We employed atomic force microscopy and super resolution fluorescence microscopy, to investigate the cell morphology, cell mechanics, and actin cytoskeleton at the single cell level. The images showed the cells changing from an elongated shape to a rounded morphology, accompanied by a reduced Young’s modulus, disappeared stress fibers at both apical and basal surfaces, and the formation of thin protrusions containing a meshwork of peripheral actin filaments surrounded the cells. Western blot analysis further verified our two hypotheses on both cell lines, where CD146 targeted gold nanorods and mild hyperthermia generated a synergistic effect on actin cytoskeleton by depleting membrane CD146 and interfering ERM phosphorylation. Based on these results, we believe targeting CD146 with low-dose gold nanorods and mild hyperthermia could be a versatile, effective, and safe approach for stopping cancer metastasis.

Experimental section

Materials:

Chloroauric acid, sodium borohydride, silver nitrate, ascorbic acid, cetyltrimethylammonium bromide (CTAB), sodium chloride, magnesium chloride, glucose, glucose oxidase, catalase, EGTA, MES hydrate, Triton X-100, cysteamine (MEA), glutaraldehyde, bovine serum albumin (BSA), and goat anti-rabbit IgG H&L (HRP) were purchased from Sigma-Aldrich. Streptavidin was purchased from Leinco Technologies. AlexaFluor594 conjugated human MCAM/CD146 antibody and biotinylated human MCAM/CD146 antibody were purchased from R&D system. DMEM/F-12 culture medium, fetal bovine serum (FBS), DPBS buffer, and PBS buffer were from Gibco. L-15 medium, BlockAid buffer, phalloidin conjugated with AlexaFluor568, Thermo Scientific™ Halt™ protease and phosphatase inhibitor cocktail (100X), Thermo Scientific™ RIPA lysis and extraction buffer, Mem-PER™ plus membrane protein extraction kit, and Pierce™ BCA Protein Assay Kit were purchased from Thermo Fisher Scientific. Phospho-ezrin(Thr567)/radixin(Thr564)/moesin(Thr558) antibody from rabbit, ezrin/radixin/moesin antibody from rabbit, β-actin (13E5) rabbit mAb, and MCAM antibody from rabbit were purchased from Cell Signaling Technology. Anti-alpha 1 sodium potassium ATPase antibody from mouse, goat anti-mouse IgG H&L (HRP) were from Abcam. 96-well plates and glass-bottom dishes were obtained from Corning and WPI, respectively.

Synthesis of CTAB-capped gold nanorods (AuNRs):

Gold nanorods were synthesized using a seed-mediated approach.[48, 49] Seed solution was prepared by adding ice-cold sodium borohydride solution (10 mM, 0.6 ml) into CTAB (100 mM, 9.75 ml) and chloroauric acid solution (0.25 mM, 0.25 ml) under vigorous stirring at room temperature. The color of the seed solution changed from yellow to brown. Growth solution was prepared by mixing CTAB (100 mM, 95 ml), silver nitrate (10 mM, 1 ml), chloroauric acid (10 mM, 5 ml), and ascorbic acid (100 mM, 0.55 ml) in the same order. The solution was mixed by gentle stirring. To the resulting colorless solution, freshly prepared seed solution (0.12 ml) was added and set aside in dark for 14–16 h, followed by three times of centrifugation (each time: at 10,000 rpm for 10 min) to remove excess CTAB. The synthesized CTAB-capped AuNRs was stored in deionized water (18.2 MΩ-cm).

Preparation and characterization of CD146 targeted AuNRs (AuNRs@CD146Ab) and nonspecific targeted AuNRs@BSA:

The surface functionalization of AuNRs was achieved by a previously reported ligand exchange method with minor changes. [50, 51] The CTAB-capped AuNRs was added into 2 mg/mL streptavidin solution (0.02% citrate, pH 7.0). The above solution was sonicated for 20 min and then centrifuged at 5,000 rpm for 20 min. Then the supernatant was replaced by 0.5 mg/mL streptavidin solution (0.02% citrate, pH 12.0) and stored at 4 °C for overnight. Then the AuNRs@streptavidin were washed with alkaline water (pH 11) for three times using centrifugation at 5,000 rpm. Then 100 μL (50 μg/mL) biotinylated CD146 antibody was add to the above synthesized AuNRs@streptavidin (500 pM) in 1 mL PBS (pH 7.4) and then incubated for 30 min to acquire AuNRs@streptavidin@CD146 antibody (AuNRs@CD146Ab, for short). The concentration of gold nanorods here is determined from a well-established relationship by Murphy’s group between aspect ratio of gold nanorods and their extinction coefficient.[52] According to this relationship, the gold nanorods with aspect ratio of 4.3 (48 nm/11 nm here) have extinction coefficient at 5 × 10⁹ (M⁻¹cm⁻¹). Therefore, 500 pM gold nanorods here is calculated from [Extinction 2.5] / [5 × 10⁹ (M⁻¹cm⁻¹) × 1 cm] based on the Beer’s law. The amount of biotinylated CD146 antibody was optimized to obtain maximum coverage of CD146 antibody on the AuNRs@streptavidin surfaces (Figure S1). The mixture solution was washed by PBS (pH 7.4) using centrifugation at 5,000 rpm and stored in PBS solution (pH 7.4) at 4 °C. The protocol of BSA coating is the same as streptavidin coating. Zeta potential measurements were conducted using a Malvern Zetasizer Nano ZS system. CTAB-capped AuNRs and AuNRs@CD146Ab were images by TEM-JEOL JEM 2100 LaB₆ TEM at 200 keV. UV-1900 UV-vis spectrophotometer (SHIMADZU) was used to monitor the UV-vis spectra of AuNRs after each step.

Cell culture and cytotoxicity assays:

SKMEL-2 and MDA-MB-231 cells were cultured in DMEM growth medium with 10% FBS and 1% penicillin and streptomycin at 37 °C under 7.5% and 5 % CO₂, respectively. To prepare cytotoxicity assays, 1 × 10⁴ SKMEL-2 or MDA-MB-231 cells were seeded into each well of a 96-well plate with 0.1 mL DMEM growth medium and cultured for overnight. Then the medium was replaced by a fresh medium containing 500 pM of AuNRs@CD146Ab or AuNRs@BSA for 12 h, with or without a subsequent NIR laser irradiation for 10 min (808 nm, 0.4 W/cm², Sunshine-electronics). The temperature of the photothermal effect mediated by the AuNRs was set to 43°C, confirmed by an infrared thermometer (Fisher). For the XTT assay, after the treatment, the medium in each well was discarded and cells were washed by DPBS. Subsequently, to each well, the XTT detection solution (150 μL medium containing 50 μL XTT) was added and the plates were placed at 37 °C for 4 h. Then, the absorbance at the wavelength of 450 nm was measured using an EPOCH2 microplate reader (BioTek). In the case of the LDH assay, from each well after the treatment, 50 μL medium was pipetted to a new 96-well plate and quantified for the LDH release using CyQUANT™ LDH Cytotoxicity Assay Kit (Thermo). For each of the experimental conditions, mean and standard deviation of the quadruplicate wells were reported.

Immunofluorescence staining of CD146:

4% paraformaldehyde (10 min at room temperature) was used to fix the SKMEL-2 or MDA-MB-231 cells. Then the cells were washed with DPBS for 3 times and blocked with 1 mL of 1% BlockAid blocking solution in PBS at 37 °C for 30 min. After that, AlexaFluor594 conjugated human MCAM/CD146 antibody (1:1000 in 1% BlockAid solution) was used to stain the cells at 4 °C for overnight, followed by 3 times of DPBS washing. Finally, DAPI solution was used to stain the cell nucleus. Fluorescence imaging was performed using an Olympus IX71 inverted epifluorescence microscope.

Dark-field and TEM imaging of cell uptake of AuNRs:

For the dark-field imaging, 1×10⁴ SKMEL-2 or MDA-MB-231 cells were plated in a 25 mm cover glass for 12 h, and then incubated with 500 pM of AuNRs@CD146Ab or AuNRs@BSA suspended in culture media for 12 h. Then 4% paraformaldehyde was used to fix cells at room temperature for 10 min, followed by 3 times of DPBS washing. An Olympus BX51M reflected metallurgical microscope with an Olympus camera (KCC-REM-OLY-041600001) was used for the dark-field imaging. For the TEM imaging, after 12 h incubation with 500 pM of AuNRs@CD146Ab suspended in the culture medium, SKMEL-2 cell sections were prepared by fixing cell pellets in 2% paraformaldehyde, 2.5% glutaraldehyde and 0.1 M cacodylate buffer, postfixed in 1% osmium tetroxide, dehydrated in an ethanol series (50%, 70%, 90%, 95%, and 100%) followed by propylene oxide (100%), embedded with propylene-oxide resin, and left in oven (60–70 °C) for 2 days. Sections that were 60 nm thick were cut and stained with uranyl acetate. The images were collected using a JEOL JEM 2100 LaB₆ TEM at 200 keV.

Wound healing assay:

500,000 SKMEL-2 or MDA-MB-231 cells were cultured in a 12-well plate for 12 h, followed by incubation with 500 pM of AuNRs@CD146Ab or AuNRs@BSA suspended in culture media for another 12 h. Subsequently, the cells were irradiated by the NIR laser (808 nm, 0.4 W/cm²) for 10 min. Then a p200 pipet tip was used to scratch the cell monolayer to create a gap. The cells were imaged at 0 h and 24 h after scratching.

3D invasion assay:

3D invasion assays were conducted on 24-well inserts (8 μm pore size) with a Matrigel coating (Corning). After 12 h cell incubation with AuNRs@CD146Ab or AuNRs@BSA (500 pM) followed by 808 nm laser treatment for 10 min at 0.4W/cm², 50,000 SKMEL-2 or MDA-MB-231 cells in 100 μL FBS-free medium with 2% Matrigel were added to the insert. Subsequently, 1000 μL of DMEM medium with 10% FBS was placed in each well. After 24 h, the number of cells that moved across the membrane was counted and normalized to the number of control cells.

AFM Imaging of SKMEL-2 and MDA-MB-231 Cells:

A BioScope Resolve AFM (Bruker) in the PF-QNM mode (PeakForce quantitative nanomechanical mapping) was used to image the untreated and AuNRs@CD146Ab with mild hyperthermia treated SKMEL-2 and MDA-MB-231 cells at 37 °C (on a heating stage) in the CO₂-independent L-15 medium. PF QNM-Live Cell cantilevers with a ~0.1 N/m spring constant and a 70 nm tip radius were selected. The exact spring constant was calibrated using thermal noise method prior to imaging. Before AFM imaging of treated cells, SKMEL-2 or MDA-MB-231 cells were incubated with AuNRs@CD146Ab suspended in DMEM medium for 12 h, followed by 10 min laser treatment (808 nm, 0.4 W/cm²) and then were put back to 37 °C incubator for another 12 h. The imaging parameters are listed as below: 600 pN in peak force, 300–600 nm in amplitude, 0.25 kHz in piezo movement, 256 and 256 in line and pixel, and 0.12 Hz in imaging frequency. As the tip shape is spherical, the Hertz model was used to fit the retraction curve in the NanoScope analysis software (Bruker):

$$F=\frac{4E}{3(1-v^2)}{R}^{\frac{1}{2}}\cdot {\delta }^{\frac{3}{2}}$$

E: Young’s modulus, ν: Poisson ratio (0.5), R: tip radius curvature (70 nm), and 𝛿: indentation depth.

dSTORM imaging of actin cytoskeleton of SKMEL-2 and MDA-MB-231 Cells:

SKMEL-2 or MDA-MB-231 cells were incubated with 500 pM of AuNRs@CD146Ab suspended in DMEM medium for 12 h, followed by 10 min laser irradiation (808 nm, 0.4 W/cm²) and then were put back to 37 °C incubator for another 12 h. 4% PFA was used to fix the cells for 10 min, followed by 3 times of DPBS washing. Subsequently, Triton-X 100 (at 0.1%) prepared in a cytoskeletal buffer was used to permeabilize the cells for 10 min. The cytoskeletal buffer includes 5 mM magnesium chloride, 150 mM sodium chloride, 5 mM glucose, 5 mM EGTA, and 10 mM MES at pH 6.1. Then the cells washed with PBS for 3 times and treated with 0.1% sodium borohydride in DPBS to minimize the background from autofluorescence. After 3 times of DPBS washing, BlockAid buffer was added into cells for 1 h to block the unspecific interactions. Then the cells were incubated with phalloidin AlexaFluor568 (66 μM) for overnight at 4°C, followed by 3 times of DPBS washing. Right before the dSTORM imaging, the imaging buffer was added into the imaging dish right. The imaging buffer recipe used here is: 1% MEA, 10% glucose (w/v), 40 μg/ml catalase, 0.5 mg/ml glucose oxidase, 10 mM NaCl, and 100 mM Tris (always maintaining buffer pH at 8 during imaging). The dSTORM was performed using an Olympus IX73 microscope at a TIRF illumination mode. The detailed instrumentation was described earlier by our group.[53] The images were reconstructed by ThunderSTORM in ImageJ.

Western blot analysis of CD146, ERM and p-ERM proteins:

Western blot analysis of these proteins were conducted only 30 min after laser irradiation to capture early signaling change, which is different from AFM and dSTORM imaging (12 h after laser irradiation allowing the development and stabilization of the cell morphology). For the membrane CD146, SKMEL-2 or MDA-MB-231 cells were lysed by Mem-PER™ Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific) to extract CD146 from the cell membrane. Before loading protein into a 4–20% precast SDS-PAGE gel, BCA assay was used to quantify the protein concentration to ascertain the equal protein amount in each channel. After running under 100 mV for 90 min, the resulting gels were transferred to nitrocellulose membrane (0.2 μm) by Bio-Rad trans blot turbo (Bio-Rad). Then the EveryBlot buffer (Bio-Rad) was used to block the membrane for 5 min. Then the membrane was incubated with the primary antibodies, MCAM (CD146) antibody and anti-α1 sodium potassium ATPase (1:1000 dilution in blocking buffer) for overnight at 4 °C while shaking. Then the secondary antibody (goat anti-rabbit or anti-mouse IgG, HRP conjugate (H+L), 1:5000 dilution in blocking buffer) was added, followed by washing the blot for 3 times (10 min each) in TBST. ChemiDoc (Bio-Rad) was used to image the immunoblots.

For ERM and p-ERM proteins, SKMEL-2 or MDA-MB-231 cells were lysed by RIPA buffer supplemented with protease inhibitors and phosphatase inhibitors cocktail (Sigma-Aldrich). BCA assay was used to quantify the protein to make sure equal amount of protein was loaded into a 4–20% precast SDS-PAGE gel. After running under 100 mV for 90 min, the resulting gels were transferred to nitrocellulose membrane (0.2 μm) by Bio-Rad trans blot turbo (Bio-Rad). EveryBlot buffer (Bio-Rad) was used to block the membrane for 5 min. The primary antibodies, ezrin (Thr567)/radixin (Thr564)/moesin (Thr558), phospho-ezrin (Thr567)/radixin (Thr564)/moesin (Thr558), and β-actin (1:1000 dilution in blocking buffer) were incubated with the membranes at 4 °C for overnight while shaking. Then the secondary antibody (goat anti-rabbit IgG antibody, HRP conjugate (H+L), 1:5000 dilution in blocking buffer) was added. Finally, blots were washed three times for 10 min in TBST after primary and secondary antibodies. ChemiDoc (Bio-Rad) was used to image the immunoblots.

Results and Discussion

Preparation and characterization of CD146 targeted gold nanorods

Figure 2A summarizes the strategy for preparation of CD146 targeted gold nanorods (AuNRs). The surface functionalization of AuNRs is achieved via a ligand exchange method, which provides AuNRs with excellent colloidal stability and biocompatibility through a uniform, thin protein shell surrounding the AuNR surfaces.[50, 51] Briefly, cytotoxic cetyltrimethylammonium bromide (CTAB), the capping reagent on the surfaces of AuNRs, is replaced by a layer of streptavidin, followed by the conjugation of biotinylated CD146 antibody. Here, three critical steps were involved in the streptavidin coating: First, the concentration of streptavidin was higher than that of CTAB, leading to a shift of equilibrium toward streptavidin coated AuNRs. Second, streptavidin, as a protein, is a multivalent ligand, whereas the CTAB is a monovalent ligand. Thus, streptavidin has a higher binding affinity to surface of AuNRs compared to CTAB. Third, the replaced positive-charged CTAB interacted with the free negative-charged streptavidin in the solution and was removed after washing steps. As each streptavidin molecule contains four biotin binding sites (a tetramer), the coverage of CD146 antibody on AuNR surfaces can be maximized, thereby minimizing the nonspecific interactions.

Figure 2.

Preparation and characterization of CD146 targeted gold nanorods. (A) Schematic illustrating the surface functionalization of gold nanorods (AuNRs) with CD146 antibody. (B) TEM images of CTAB-capped AuNRs. (C) Extinction spectra of AuNRs following surface functionalization steps. (D) Zeta potential of AuNRs following surface functionalization steps. Error bars are standard deviation of three independent experiments (n=3). (E) TEM images of CD146 antibody functionalized AuNRs. A clear coating can be observed surrounding the AuNR surfaces. (F) Quantitative analysis of length and diameter of AuNRs before and after the surface functionalization, confirming the successful surface functionalization of AuNRs.

A seed-mediated approach was used to synthesize the AuNRs with a length of 48.0 ± 3.1 nm and a diameter of 11.0 ± 1.8 nm,[48, 49] as shown in the transmission electron microscope (TEM) images (Figure 2B). After replacing CTAB with a streptavidin layer on the AuNR surfaces, the longitudinal localized surface plasmon resonance (LSPR) wavelength showed a 15 nm redshift (from 794 to 809 nm) because of the increase in the refractive index of the medium surrounding the AuNRs (Figure 2C). The subsequent binding of biotinylated CD146 antibody onto the streptavidin layer was confirmed by another 7 nm redshift (from 809 to 816 nm). Notably, the decreased redshift (7 nm versus 15 nm) can be explained by the decreased refractive index sensitivity with increasing distance from the surfaces of the AuNRs.[54] Meanwhile, albumin coated AuNRs as nonspecific targeted AuNRs (as a control) were prepared by replacing CTAB with a bovine serum albumin (BSA) layer, as confirmed by a 14 nm redshift (from 794 to 808 nm) in the LSPR wavelength. The 14 nm redshift is reasonable considering the size of the BSA (Mw: 57 kDa) is comparable with the size of the streptavidin molecule (Mw: 58 kDa). Apart from the LSPR wavelength shifts, the zeta potential of the AuNRs after each step was analyzed to verify the surface functionalization (Figure 2D). The as-synthesized CTAB-capped AuNRs showed a surface charge of +37.6 mV since the CTAB is a positive-charged surfactant. Following replacing CTAB with streptavidin or BSA coating, the AuNR surfaces became negatively charged in the pH=7 water (−31.5 mV after streptavidin coating, and −28.3 mV after BSA coating), as streptavidin and BSA have close isoelectric points at 5.0 and 4.7, respectively.[55, 56] Furthermore, the binding of biotinylated CD146 antibody onto the streptavidin layer was evidenced by the increased zeta potential from −31.5 to −21.7 mV. The final product, CD146 antibody functionalized AuNRs (hereafter referred as AuNRs@CD146Ab), was further characterized by TEM imaging. As shown in Figure 2E, there is a clear protein coating (streptavidin plus biotinylated antibody) surrounding the AuNR surfaces. The average size of the AuNRs@CD146Ab had a 5.6 nm and a 6.1 nm increase in length and diameter, respectively, compared to CTAB-capped AuNRs (Figure 2F). The protein quantification suggested that a monolayer of streptavidin and CD146 antibody formed on each AuNR, leading to 80–160 CD146 antibodies conjugated with each nanorod (supporting information). Overall, these results demonstrate the successful preparation of CD146 targeted AuNRs, with a suitable LSPR wavelength peak (816 nm) falling into the near-infrared (NIR) window.

Cancer cell targeting ability and cytotoxicity assessment of AuNRs

Two metastatic human cancer cell lines (SKMEL-2, a melanoma cell line, and MDA-MB-231, a breast cancer cell line) were selected to evaluate the targeting ability of AuNRs@CD146Ab. The binding of CD146 antibody to the cell surface CD146 was expected to enhance the endocytosis of AuNRs@CD146Ab, whereas the nonspecific targeted AuNRs@BSA, were used as a control. First, the expression of surface CD146 on these two cell lines were verified by immunofluorescence staining (Figure S2 and S3). Subsequently, cell uptake of AuNRs@CD146Ab and AuNRs@BSA were validated by a dark field microscopy as the intensity of the scattered light by AuNRs indicated the internalized amount of AuNRs.[57] As shown in Figure 3A, Figure S4 and S5, after incubation with 500 pM of AuNRs@CD146Ab for 12 h, both cell lines exhibited brightest scatted light, compared to the cells after incubation with the same concentration of AuNRs@BSA and the untreated cells, suggesting higher uptake of AuNRs@CD146Ab by the cells. The cell uptake efficiency was further quantified by monitoring the UV-vis spectral change of AuNRs in the culture media before and after 12 h incubation with cells. This is based on Beer’s law which states that the concentration of AuNRs has a linear relationship with the absorbance at the LSPR wavelength.[18] The results (Figure S6) revealed that the absorbance of AuNRs@CD146Ab decreased by ~40% after 12 h incubation with cells, whereas the AuNRs@BSA only had a ~10% drop in the absorbance at the LSPR wavelength, implying that more AuNRs@CD146Ab were internalized by the cells. It is worth noting that the AuNRs@CD146Ab still possessed excellent colloidal stability after suspension in cell media for 12 h, evidenced by the retention of symmetric LSPR peaks. It was also confirmed that the colloidal stability and targeting ability of the AuNRs@CD146Ab were not compromised after subjected to a slightly acidic environment for 24 h (pH=6.4), a condition to mimic the acidic microenvironment of tumors (Figure S4–S5 and S7). In addition, TEM imaging was employed to confirm the internalization of AuNRs@CD146Ab by the cells (Figure 3B and S8). It was found that most of AuNRs, after 12 h incubation with cells, were located in lysosomes, displayed as electron-dense vesicles, consistent with previous reports.[58, 59] Clearly, these results demonstrate the excellent targeting ability of CD146 antibody functionalized AuNRs (i.e., AuNRs@CD146Ab) to the CD146 over-expressed cancer cell lines.

Figure 3.

Cancer cell targeting ability and cytotoxicity assessment of AuNRs. (A) Dark field images of MDA-MB-231 cells after treated with 0 (control), or 500 pM of AuNRs@BSA or AuNRs@CD146Ab for 12 h. Scale bars: 50 μm. (B) TEM imaging confirming the internalization of AuNRs@CD146Ab by the cells after 12 h incubation. Most AuNRs can be found in the lysosomes. (C) XTT assay assessing the viability of both cell lines after different treatments. (D) LDH cytotoxicity assay assessing the membrane integrity of both cell lines after different treatments. Error bars: n=4.

Following the cell uptake experiments, we assessed whether the AuNRs and mild hyperthermia affect the viability of both cell lines. The mild hyperthermia (43 °C) was induced by applying an 808 nm NIR laser (0.4 W/cm²) on the cell culture medium containing 500 pM of AuNRs@CD146Ab for 10 min (Figure S9A). The 500 pM here could be the minimal dose since 100 pM and 250 pM upon the laser treatment cannot reach the required mild hyperthermia (Figure S9B). The mild hyperthermia is critical for the clinical applications because the low laser power density used here is within the skin tolerance limit suggested by the American National Standards Institute.[60] More importantly, it could be extremely valuable to stop cancer metastasis by mild hyperthermia without killing the cancer cells, considering some tumor areas may not generate sufficient heat energy to trigger cell apoptosis/necrosis (i.e., the photothermal ablation of tumors typically requires 50 °C), due to inhomogeneous distribution or low dose of AuNRs, or weak laser penetration.[34, 61–63] Here, two commonly used cytotoxicity assays, XTT and LDH assay, were used to test the impact of AuNRs and mild hyperthermia on the cell viability and membrane integrity. The results (Figure 3C and 3D) indicated that both cell lines after incubation with 500 pM of AuNRs@CD146Ab or AuNRs@BSA for 12 h, with or without subsequent laser irradiation, maintained over 95% of cell viability and had negligible membrane damage. Therefore, these mild treatments did not kill the cancer cells, and thus were applied in subsequent experiments to investigate how they affect cancer metastasis.

CD146 targeted AuNRs with mild hyperthermia stop cancer cell migration

To evaluate the effect of AuNRs and mild hyperthermia on the cancer metastasis, a wound healing assay was performed on the two cell lines to quantify the cell migration ability upon different treatments. This assay is also named the “scratch assay” as it is conducted by making a scratch on a cell monolayer and observing the scratch at different time intervals by time-lapse microscopy. The cancer cell migration is then quantified by monitoring how fast these cells can close the gap formed by the scratch. As shown in Figure 4, without any treatment (the control groups, Figure S10 and S11), both cell lines were able to close the gap within 24 h (equivalent to 100% wound healing), owing to their strong migration ability. The cells after being treated with AuNRs@BSA also exhibited 100% wound healing, and the introduction of mild hyperthermia (AuNRs@BSA+Laser) just slightly decreased the wound healing to 80–90%, which can be attributed to the low uptake efficiency of the nonspecific targeted AuNRs@BSA (Figure S12–S15). In stark contrast, both cell lines after AuNRs@CD146Ab treatment showed a 30–40% decrease in wound healing (60–70% healing), suggesting the migration ability of both cell lines had been significantly inhibited (Figure S16–S17). Strikingly, the mild hyperthermia (AuNRs@CD146Ab+Laser) further reduced their wound healing to less than 10%, meaning that the migration of the cells was almost completely stopped (Figure S18–19). We confirmed 500 pM AuNRs@CD146 could be the minimal dose here as 250 pM AuNRs@CD146+Laser can only partially stop the cell migration (Figure S20). It should be noted that the laser irradiation here was performed on unwashed cells to ensure the consistent temperature increase (to 43 °C) for both AuNRs@BSA and AuNRs@CD146Ab treated cells. Similar wound healing results were obtained on both cell lines when laser irradiation was conducted after washing the un-internalized nanorods. As shown in Figure S21–S24, the cells with internalized AuNRs@CD146Ab and laser treatment showed only 20% wound healing, compared to 100% wound healing for cells with internalized AuNRs@BSA and laser treatment. To further confirm the un-internalized nanorods had no impact on cell migration, the cells were added with AuNRs@ CD146Ab, treated with laser immediately, and then washed. In this case, almost all nanorods were un-internalized during the laser treatment. The results showed that the cell migration was not affected by these un-internalized nanorods (shown as 100% wound healing, Figure S25). As a control experiment, laser irradiation only showed no effect on wound healing, suggesting the importance of the internalized AuNRs@CD146Ab (Figure S26–S27). Apart from the wound healing assay, a 3D invasion assay was performed to mimic the migration of cancer cells in extracellular matrix in vivo upon different treatments (Figure S28). Different from 2D wound healing assay, 3D invasion assay involves a trans-well cell culture insert with a Matrigel coating.[64, 65] Cell suspension (with and without treatments) added with Matrigel was initially placed in the upper chamber. After 24 h incubation, the invasive cells were able to pass through the matrix basement and adhered to the bottom of the insert membrane, while the non-invasive cells stayed in the upper chamber. The number of cells that moved across the membrane was counted and normalized to those of control (untreated) cells. The results showed that AuNRs@CD146Ab with laser treatment can also stop the migration of both cell lines in the 3D matrix. Overall, this is an exciting result since previous studies could only partially inhibit the migration of cancer cells using gold nanoparticles.[31–34] Our strategy of a combination of picomolar level AuNRs (i.e., the lowest dosage reported so far) with mild hyperthermia induced by low laser irradiance achieved almost complete inhibition of cancer cell migration.

Figure 4.

Wound healing assay evaluating migration ability of SKMEL-2 (A, B) and MDA-MB-231 (C, D) cells after different treatments including: Control (untreated), AuNRs@BSA, AuNRs@BSA with laser, AuNRs@CD146Ab, and AuNRs@CD146Ab with laser. Error bars in B and D represent standard deviation of three independent experiments (n=3). Student’s t-test, ***P < 0.001. Both cell lines after treated by AuNRs@CD146Ab with laser almost completely lost the migration ability.

CD146 targeted AuNRs with mild hyperthermia disrupt the actin cytoskeleton of cancer cells

The structure-function relationship is a key concept for biomechanics.[66–69] Considering the cancer cell migration is highly dependent on the actin cytoskeleton, we set out to investigate how CD146 targeted AuNRs with mild hyperthermia treatment affect the actin cytoskeleton of both cell lines used in this study.

We employed a nanoscale tool, quantitative nanomechanical imaging based on atomic force microscopy (AFM), to characterize single living cells with and without the treatment. This technique is able to generate a Young’s modulus map on individual cell with correlated cell surface topography.[70, 71] Importantly, the F-actin at cell apical layers can be visualized on modulus images as these actin filaments are the major contributors to cell stiffness (Figure 5A).[72] Figure 5B shows the correlated height, topography, and modulus images of individual cells. Each pixel in modulus images correlates to a Young’s modulus value acquired by fitting the indentation curve collected at that spot (Figure 5C–5D). Hertz model was used here because of the spherical shape of the AFM tip. For the untreated cells (the control groups, Figure 5B: column I and III, and Figure S29–S30), both cell lines showed an elongated morphology with obvious leading and trailing edges, suggesting the cells were in the process of movement. By correlating surface topography with Young’s modulus images, two characteristic actin cytoskeleton structures that drive the cancer cell migration can be identified on both cell lines: (i) Lamellipodium (marked by green arrows), the thin membrane located at the leading edge of the cells that initiates directed movement, exhibited the highest modulus on the cells owing to the dense meshwork of actin filaments underlying the membrane and a combined substrate effect; and (ii) Stress fibers (marked by red arrows), the actin filament bundles along the cell body that drives cell contraction to pull the cell body forward, identified as grooved patterns on both topography and modulus images with excellent correlation. Notably, the elevated surface topography perfectly correlates with the higher modulus, owing to the denser apical F-actin beneath the cell membrane. These stress fibers, well-aligned with the cell long axis, produce the contraction forces during cancer cell migration.[7, 73] However, in the case of the treated cells (AuNRs@CD146Ab+Laser, Figure 5B: column II and IV, Figure S31–S32), both cell lines displayed a rounded morphology with the appearance of a thin membrane around the cell periphery, consisting of a dense meshwork of actin filaments. This is in stark contrast with the untreated cells that possessed a lamellipodium only at the leading edge of the cells. Interestingly, the modulus images showed that the main body of the treated cells had no well-aligned stress fibers. Here, the “main body” is defined as the areas higher than 1 μm to exclude the substrate effect. This was further validated by quantitative analysis which showed that the average Young’s modulus of the main body of the treated cells was ~5 times lower than that of the untreated cells (8 kDa versus 40 kDa, Figure S33–S34), suggesting the disappearance of apical stress fibers after the AuNRs@CD146Ab+Laser treatment.

Figure 5.

AFM-based quantitative nanomechanical imaging of single cells with and without treatment. (A) Schematic showing nanomechanical imaging of cell apical actin. (B) Height, surface topography and Young’s modulus maps of single SKMEL-2 and MDA-MB-231 cell with and without treatment (control). Column I and III: The untreated cells show an elongated morphology with obvious lamellipodia (green arrows) and well-aligned stress fibers on the apical surfaces (red arrows). Column II and IV: The treated cells display a rounded morphology with the appearance of a thin membrane around the entire cell periphery, and the main body of the treated cells (defined as the areas higher than 1 μm) had no well-aligned stress fibers. Scale bars: 10 μm. (C) Raw force-displacement curves on three different areas. including glass substrate, cell surface with underlying actin, and cell surface without underlying actin. (D) The force-indentation curve was fitted by the Hertz model to generate Young’s modulus value at single pixel.

To further confirm the disrupted actin cytoskeleton after treatment, a super resolution fluorescence microscopy technique, direct stochastic optical reconstruction microscopy (dSTORM), was applied. Since this technique utilizes a total internal reflection fluorescence (TIRF) mode to excite a 100–200 nm evanescent field above the coverslip, only F-actin at cell basal surfaces can be imaged with reduced out-of-focus background (Figure 6A). The dSTORM enables super resolution imaging by single molecule localization of a sparse subsets of fluorophores in a stochastic manner (Figure 6B and 6C).[74–76] As shown in Figure 6D, the dSTORM clearly improves the spatial resolution of actin cytoskeleton at cell basal surfaces, compared to the conventional fluorescence microscopy. For the untreated cells (control groups in Figure 6D, Figure S35–S36), lamellipodium consisting of a dense meshwork of actin filaments was found at the leading edges of the cells. Meanwhile, actin filaments aligned with the cell long axis were also observable, indicating that both cell lines possessed basal stress fibers. In contrast, for the treated cells (Figure 6D, Figure S37–S38), consistent with our AFM results, a dense meshwork of actin filaments can be seen around the entire cell periphery, while actin filaments aligned with the cell long axis disappeared after the treatment. Together, the AFM and dSTORM images demonstrate that CD146 targeted AuNRs with mild hyperthermia significantly disrupt the cell morphology, mechanical properties, and actin cytoskeleton of both cancer cell lines. The alterations of the cell morphology and actin cytoskeleton are a direct consequence of the AuNRs@CD146Ab+Laser treatment, causing the observed loss of cell migration function.

Figure 6.

(A) Schematic illustrating imaging cell basal actin using a TIRF mode. (B) and (C): dSTORM images are obtained by stochastically activating and localizing individual molecules in a stack of >10,000 frames. Each frame captures a subset of single molecules at different locations and time points. (D) Epifluorescence and dSTORM images of actin cytoskeleton of two cell lines with and without the treatment (AuNRs@CD146Ab with laser). Scale bars: 10 μm.

CD146 targeted AuNRs with mild hyperthermia disrupt the actin cytoskeleton by down-regulation of CD146 and ERM phosphorylation

Based on our observations of actin cytoskeleton structures, we sought to study the mechanism of how the actin cytoskeleton of both cell lines was disrupted by the AuNRs@CD146Ab+Laser treatment. It is known that CD146 is over-expressed on the surfaces of metastatic melanoma and breast cancer cells, and its expression level correlates with malignant progression and metastatic potential of these cells.[40, 45] In essence, membrane CD146 controls cancer cell motility through regulating the underlying actin cytoskeleton.[44] As shown in the schematic (Figure 1), the transmembrane CD146 has two functions: (i) serves as an “anchor” to recruit actin cytoskeleton to cell membrane through ezrin-radixin-moesin proteins (i.e., ERM proteins, the “chain”) to modulate the cancer cell migration; and (ii) activates the highly conserved ERM proteins in cytosol by phosphorylation, allowing phosphorylated ERM proteins (p-ERM) to translocate from the cytosol to the interface of the membrane CD146 and actin cytoskeleton.[44, 45] Based on the roles of CD146 in regulating actin cytoskeleton and cell migration, two hypotheses are proposed here: (i) The uptake of CD146 targeted gold nanorods will deplete the cell membrane CD146, and thereby disrupt the actin cytoskeleton and cell migration (i.e., remove the “anchor”); and (ii) Mild hyperthermia will interfere with the phosphorylation of ERM proteins to further impair cell migration (i.e., break the “chain”).

To test the first hypothesis, membrane CD146 of both cell lines was analyzed using the Western blot. For the melanoma cells (SKMEL-2), after AuNRs@CD146Ab treatment (with or without the laser), the membrane CD146 level showed a ~50% decrease, in comparison to the untreated cells (Figure 7A). The breast cancer cells (MDA-MB-231) also exhibited a ~20% decrease in the membrane CD146 after the AuNRs@CD146Ab treatment (with or without the laser, Figure 7B). Considering the membrane CD146 level correlates with the metastatic potential of both cancer cell lines, these results explain our earlier observations in the wound healing assays, where the AuNRs@CD146Ab treatment alone can partially inhibit the cancer cell migration. We propose that the uptake of AuNRs@CD146Ab down-regulates the membrane CD146 through depleting the membrane CD146 into a lysosomal degradation pathway,[77] as most AuNRs@CD146Ab were found in the lysosomes after 12 h incubation with the cells (Figure 3B). Therefore, it can be concluded that the uptake of AuNRs@CD146Ab down-regulates the membrane CD146 (analogous to removing the “anchor” that immobilizes the actin cytoskeleton to the cell membrane), and partially disrupts the actin cytoskeleton organization and cancer cell migration.

Figure 7.

Mechanistic investigation of the effect of CD-146 target gold nanorods with mild hyperthermia on the actin cytoskeleton structures of cancer cells. (A) and (B) Western blot analysis of membrane CD146 on both cell lines with and without treatments. Sodium potassium ATPase is used as a membrane housekeeping protein. (C) and (D) Western blot analysis of total ERM and p-ERM levels in both cell lines with and without treatments. β-actin is used as a house-keeping protein in cytosol. Data is mean ± standard deviation, n=3. Student’s t-test, *P< 0.05, ***P < 0.001.

Next, we focused on how the introduction of a mild hyperthermia completely stopped the migration of both cell lines as observed in the wound healing assays. Since the mild hyperthermia treatment had almost no effect on the membrane CD146 levels, we turned our attention to the ERM proteins, the adaptor proteins that connect actin filaments to the membrane CD146. It is known that ERM proteins in cytosol have a highly conserved conformation,[78] and they are activated by the CD146 through phosphorylation to switch to an active form (namely, phosphorylated ERM, p-ERM). The p-ERM can then translocate to the membrane CD146, serving as the “chain” that links the actin filament to the membrane CD146.[44] Here, we posit that the mild hyperthermia could interfere the phosphorylation of ERM proteins in the cytosol, as previous studies suggested that heat stress could cause the phosphorylation changes inside various cells.[34, 79–81] Therefore, a Western blot was performed to quantify the ratio of p-ERM to total ERM (p-ERM/total ERM) in the cytosol after different treatments. For the cells after AuNRs@CD146Ab treatment alone, SKMEL-2 and MDA-MB-231 showed a ~50% and a ~15% decrease in the p-ERM/total ERM, respectively, compared to the untreated cells (Figure 7C and 7D), consistent with decreased levels of membrane CD146 in the two cell lines. A more careful examination of individual proteins showed that p-ERM mainly decreased while the total ERM remained constant (Figure S39). This is reasonable as previous studies showed that membrane CD146 level regulates and correlates with the amount of phosphorylated ERM proteins in cytosol.[44] However, for the cells after the AuNRs@CD146Ab with mild hyperthermia treatment (AuNRs@CD146Ab+laser), SKMEL-2 and MDA-MB-231 showed a ~90% and a ~50% decrease in the p-ERM/total ERM, respectively, compared to the untreated cells (Figure 7C and 7D). In comparison with the AuNRs@CD146Ab treatment alone, the introduction of the mild hyperthermia (AuNRs@CD146Ab+laser) further reduced the p-ERM levels in cytosol (on top of the decrease in p-ERM caused by CD146 down-regulation due to the uptake of AuNRs@CD146Ab). Therefore, the available “chain” to connect actin filaments to the remaining membrane CD146 was further decreased by the mild hyperthermia, leading to further disrupted actin cytoskeleton and completely stopped cell migration. In summary, the CD146 targeted AuNRs with mild hyperthermia treatment disrupts the actin cytoskeleton through a synergistic mechanism: (i) Depleting the membrane CD146 (i.e., removing the “anchor”), and (ii) Disrupting ERM phosphorylation (i.e., breaking the “chain”).

Here, it should be noted that the Western blot analysis of proteins were conducted only 30 min after laser irradiation to capture early signaling change, and AFM and dSTORM imaging were performed 12 h after laser irradiation, allowing the development and stabilization of the cell morphology. Within the short period of time, it is hard to expect any genotypic change in these cells. An interesting future work could be the investigation of how the treatment will change the gene expression in a long term. Apart from cancer cells, it will be valuable to see if this concept can be applied to non-cancerous cells in other diseases. For example, in case of vascular smooth muscle cells, the migration of these cells is always associated with atherosclerosis.[82] A potential future work is to see whether targeting a specific marker on migrating vascular smooth muscle cells using the approach here can stop their migration, thereby inhibiting the process of atherosclerosis.

Conclusions

We have demonstrated that targeting a cancer cell adhesion molecule, CD146, with low-dose gold nanorods combined with a mild hyperthermia completely stopped the migration of two metastatic cancer cells, including a melanoma (SKMEL-2) and a breast cancer cell line (MDA-MB-231). We hypothesized that targeting the cancer cell surface marker (CD146, a membrane protein that is linked to the actin cytoskeleton) would disrupt the actin cytoskeleton of the cancer cells and their migration. Atomic force microscopy and super resolution fluorescence microscopy revealed that the cancer cells after the treatment changed from an elongated shape to a rounded morphology, accompanied by a reduced Young’s modulus, disappeared stress fibers at cell apical and basal surfaces, along with the formation of thin protrusions surrounding the cells that contain a meshwork of peripheral actin filaments. The alterations of the actin cytoskeleton and cell morphology caused the loss of cell migration, as demonstrated by cell migration assays. Furthermore, Western blot analysis verified our hypothesis that CD146 targeted gold nanorods and mild hyperthermia generated a synergistic effect on actin cytoskeleton through depleting membrane CD146 and disrupting ERM phosphorylation, analogous to removing the “anchor” that immobilizes the actin cytoskeleton to the cell membrane and breaking the “chain” that links the actin cytoskeleton to the membrane CD146, respectively. Compared to previous literature that cancer cell migration was only partially inhibited by nanomaterials, the approach presented here completely stops the migration of cancer cells and pushes the dose of gold nanorods to picomolar level (the lowest level reported so far), combined with safe laser irradiance.[32–34] Considering CD146 is over-expressed on several metastatic cancers, and its overexpression is directly associated with enhanced motility of these cancer cells and poor patient prognosis, we envision that targeting CD146 with low-dose gold nanorods and mild hyperthermia could be a versatile, effective, and safe approach for stopping cancer metastasis.