Nanomotors Sense Local Physicochemical Heterogeneities in Tumor Microenvironments**

Abstract The invasion of cancer is brought about by continuous interaction of malignant cells with their surrounding tissue microenvironment. Investigating the remodeling of local extracellular matrix (ECM) by invading cells can thus provide fundamental insights into the dynamics of cancer progression. In this paper, we use an active untethered nanomechanical tool, realized as magnetically driven nanomotors, to locally probe a 3D tissue culture environment. We observed that nanomotors preferentially adhere to the cancer‐proximal ECM and magnitude of the adhesive force increased with cell lines of higher metastatic ability. We experimentally confirmed that sialic acid linkage specific to cancer‐secreted ECM makes it differently charged, which causes this adhesion. In an assay consisting of both cancerous and non‐cancerous epithelia, that mimics the in vivo histopathological milieu of a malignant breast tumor, we find that nanomotors preferentially decorate the region around the cancer cells.


Fabrication of thin nanomotors:
To fabricate thin nanomotors, it was necessary to reduce the seed layer size while maintaining enough distance between subsequent seeds to allow shadowing during evaporation. This was achieved by using Langmuir Blodgett layers of 700 nm polystyrene beads. After monolayer formation, the beads were etched down to 500 nm by air plasma etching. The etched sample was further subjected to reactive ion etching to create 1 µm pillars on the silicon wafer (see SI: Figure S1(b)). The top of the 1 µm pillars was coated with a thin film of 18 nm silver which was subsequently annealed at 300 0 C for 15 minutes to form silver balls of diameter 200 nm. This was used as the seed layer for Glancing Angle Deposition (GLAD) of silica in which the magnetic material made of iron and cobalt powder mixed in 1:1 (w/w) ratio was integrated inline during the shadow growth. This ensured encapsulation of the magnetic material by silica, thus shielding it from the external environment and preventing etching and degradation under extreme conditions.

Statistics:
Data was analyzed using Origin 9.1, MATLAB 2015b and ImageJ. All data are presented as mean ± standard deviations (SDs). Comparison of counts between distinct groups was made by t-test of proportions.

3D culture and experimental procedure:
In case of 3D monoculture, 5×10 4 cells were mixed in 50 μl of rBM (Corning, 354230) and allowed to solidify at 37 0 C temperature and 5% carbon dioxide in a humidified chamber. The concentration of rBM was ~9 mg/ml. For co-culture, 2.5×10 4 of HMLE constitutively expressing GFP and 2.5×10 4 MDA-MB-231 constitutively expressing RFP cells were mixed in 50 μl of rBM and allowed to solidify at 37 0 C temperature and 5% carbon dioxide in a humidified chamber. Cultures were grown in a defined medium 1 for 72 hours before injecting the nanomotors. An area of 0.5 mm 2 of a wafer containing nanomotors was sonicated into a microcentrifuge tube containing 50 l deionized water. A 10 µl solution containing 10 5 nanomotors suspended in deionized water was injected into the 3D matrix using a 26-gauge syringe. The sample was placed in a triaxial Helmholtz coil mounted on an optical microscope (Olympus IX71) and imaged through a 50x (or 100x) objective lens. The nanomotors were observed while the field was on and recorded using a CMOS camera. It was observed that at the site of injection of nanomotors many of them got adhered to the glass slide. The presence of local pockets of injected fluids has been reported in a previous paper from our group 2 , where we showed that within a time scale of approximately 30 minutes, the system gains its structural uniformity. To be completely sure that the experiments are carried out in native rBM-cell environment, we drive the nanomotors for 30 minutes; which is at least 1000 m away from the site of injection. We observed 0-5 nanomotors per cell at the region of experiment.
For confocal imaging, the nanomotors were actuated at 50 Gauss field rotating with a frequency of 3Hz for 30 minutes. The sample was subsequently fixed using sucrose solution after actuation. The experiments for measuring the adhesive force was generally done using 5 different field strengths (70, 100, 150, 200 and 250 Gauss) and the actuating frequency was kept either at 3Hz or 5Hz.

Laser scanning confocal microscopy:
All the fluorescence images were captured using either Zeiss LSM 880 or Leica TCS SP8 confocal microscope with system optimized Z intervals. DAPI, Alexa-488 and Alexa-568 dyes were excited with 405 nm diode, 488 nm Argon laser and 543 nm He-Ne lasers with appropriate filters and beam splitters.
Co-culture images were acquired using a Plan-apochromat 40X oil immersion objective with no digital offset and digital gain. Lectin cytochemistry images were acquired using 20X Plan-apochromat objective with no digital gain and digital offset. 2D projections of 3D images were generated using maximum intensity projection algorithm. At least three random fields were imaged in each experiment. Laser intensity and detector gain threshold were decided using a negative control for each experimental condition. Images were processed/analyzed using either Zen lite or Fiji software 3 .

Estimation of the magnetic moment:
The standard method of calculating the magnetic moment of a nanomotor has been described in previous literature in great details. A microfluidic chamber was made using a coverslip on a glass slide, in which nanomotor solution was placed. The chamber was placed under the microscope in a Helmholtz coil and subjected to 30 Gauss field while the frequency was varied from 1Hz to 10Hz. The precession angle of the nanomotor was recorded and a frequency vs precession angle curve was plotted. The frequency, where the nanomotor changed its precession angle from 90 0 (tumbling state) to a lower angle (precession state) was calculated by proper fit. This was done at regular intervals over the time period when the experiments were conducted. Initial magnetic moment was calculated to be ~1.01 × 10 −16 2 . However, since the same sample was used for all our experiments involving force calculations, the magnetic moment was found to have weakened to a value of 5 × 10 −17 2 after a period of 12 months.

Coating of PFO on nanomotors:
To coat nanomotors with PFO the wafer containing nanomotors was placed in vacuum with about 20 µl of PFO and left overnight. This was enough to coat nanomotors with PFO as seen from the FTIR data (SI: Figure S5). We also confirmed the PFO coating on glass surface using similar coating protocol. The contact angle for PFO coated glass surface was found to be 145 0 .

Supporting information
Section S1: Motion of helical nanomotors Helical magnetic nanomotors can be maneuvered using a triaxial Helmholtz coil as shown in the following Figure S1(a). For our experiments, the triaxial coil was controlled by a LabVIEW program through a DAQ connected to 3 amplifiers. This setup was placed inside a fluorescent inverted microscope. An SEM of the Glancing Angle Deposition film grown on silicon pillar seed layer is shown in Figure S1(b). Figure S1(c) shown snapshots from a movie where the nanomotors was seen moving through the reconstituted Basement Membrane (rBM) material. .

Section S3: Dynamics of a nanomotor
The torque used for calculating the force needed by the nanomotor to drive it to a point in the ECM where it gets adhered has been derived from two parameters , ( = × ). represents the magnetic moment vector of the nanomotor. θ is the angle between the vector and the short axis of the helix (see Figure S3). As mentioned in the main text there is a characteristic cutoff frequency Ω 1 which is needed to measure the torque = × = Ω 1 sin(θ ) . This was calculated by observing the dynamics of nanomotor for a range of frequency and measuring the precession angle at the imaging plane. The frequency where the nanomotor transitions from tumbling ( Figure S3(a)) to precession ( Figure S3(b)) was found by fitting the experimentally observed angle of rotation to sin −1 ( ) as shown in Figure S3(c).   The 3D view is generated from the confocal stacks.

Section S4: Calculation of diffusion of helical nanomotors in rBM .
An elongated object like a cylindrical rod is a good approximation for the geometry of nanomotors. Such elongated bodies will have anisotropy in their diffusivities. We estimated the distance from the point of injection at which nanomotors were found in the rBM to be greater than the diffusive length scale by an order of magnitude. The translational diffusivity of a rod is given by = (ln + ) 3 where is the Boltzmann constant, is the ambient temperature, is the aspect ratio given by =