Atomic force microscopy reveals the mechanical properties of breast cancer bone metastases

Mechanically dependent processes are essential in cancer metastases. However, reliable mechanical characterization of metastatic cancer remains challenging whilst maintaining the tissue complexity and an intact sample. Using atomic force microscopy, we quantified the micro-mechanical properties of relatively intact metastatic breast tumours and their surrounding bone microenvironment isolated from mice, and compared with other breast cancer models both ex vivo and in vitro. A mechanical distribution of extremely low elastic modulus and viscosity was identified on metastatic tumours, which were significantly more compliant than both 2D in vitro cultured cancer cells and subcutaneous tumour explants. The presence of mechanically distinct metastatic tumour did not result in alterations of the mechanical properties of the surrounding microenvironment at meso-scale distances (>200 μm). These findings demonstrate the utility of atomic force microscopy in studies of complex tissues and provide new insights into the mechanical properties of cancer metastases in bone.


Mechanical data represented by mean ± std
Application of a normality test shows that all of the mechanical data are non-parametric, so we have presented data as median value with a range in the main text.

Validation of the mechanical heterogeneity
Due to the hierarchical factors (i.e. number of repeated force curves, measured positions, pieces of sample and animals) and non-parametric data involved in this study, it is challenging to calculate the sample size without excessive assumptions. However, some estimations can help to validate the statistical reliability. Firstly, the number of animals was above the minimum required for statistical comparison. 1 Moreover, the median and mean values of the mechanical properties were calculated for each sample (i.e. one piece of split bone surface or one slice of subcutaneous tumour) from all measured positions within the regions of interest in the sample.
Histograms representing the median and mean data and hence the variability, for the properties quantified in all of the samples are shown in Figure S2. Although the distributions cover approximately 1 to 2 order of magnitude, they are significantly narrower than the corresponding mechanical distributions, based on each measured position (as in Figure S3). This indicates that the data presented in the main text reveals the mechanical heterogeneity within each region of interest in tissues, as opposed to solely reflecting the variations between samples.

Effect of finite sample thickness correction
Large radius spherical probes were selected for use in this study to provide sufficient force sensitivity while indenting the extremely soft samples. Hertz-Sneddon analysis for analysing the force-indentation curves has appropriately corrected the effect of a large contact area in the resultant modulus. 2 Moreover, the effect of finite sample thickness (compared to the large probe radius) on the elasticity measured by indentation (i.e. EH-S) should be taken into account,

S3
which is estimated by a commonly used method. 3 Such correction is negligible for measurements on tissues, because the sample thickness of tissues is typically about 1 mm or above that is far greater than the radius (12.5 µm) of our AFM probe. The finite thickness correction, if applied, gives less than 1% reduction in EH-S at our largest detected indentation depth. For single cultured cells (assume cell height = 6 µm, as often observed 4 ), the finite thickness correction leads to a moderate reduction of the modulus (e.g. 39% at the indentation depth corresponding to median EH-S). The histogram of corrected EH-S is shown in Figure S4.
It is notable that the finite thickness correction has no impact on the statistical comparisons between MDA-MB-231 cells and tumour tissues. Cryo-sections from subcutaneous tumours were defrosted and fixed in 4% PFA solution at ambient temperature for 10 minutes, permeabilised with 0.1% Triton X-100 for 5 mins, then washed in PBS. Subcutaneous tumour cryo-sections were then immuno-fluorescently labelled as described above.

Immunofluorescent imaging and image analysis
Images of stained tissue sections were acquired using a Zeiss LSM 980 (Zeiss Group, Germany) confocal system. Sections from more than 3 tissues for each sample type were imaged. For bone sections, imaging was focused on the metaphysis region.

S4
Images were analysed using ImageJ (National Institute of Health, US) to quantify the fluorescently labelled extracellular components. Statistical analyses were performed using Prism software (v.8.0, GraphPad Software, La Jolla, CA, USA). All data are presented as sample means ± SEM and were analysed using one-way ANOVA. A statistically significant difference was defined as p < 0.05.

Top view optics combined with AFM
The opaque samples studied here required top view optics capable of both conventional and fluorescence imaging with a wide field of view. A bespoke mount was used to secure a simple optical microscope onto the condenser stage of a Nikon Eclipse Ti. 25.4 mm diameter cemented achromatic doublets (Thorlabs) with focal lengths of 75 mm and 200 mm were used for the objective and tube lenses, respectively, to give a 2.67x magnification and a field of view of approximately 1.8 mm x 1.35 mm with the camera used (Imagingsource DFK31BF03, 1/3" sensor). The ~67 mm working distance of the objective lens was sufficient to allow imaging through the AFM head. A filter holder was placed in the infinity space between the objective and tube lenses for the emission filter. To minimise contrast loss from scattered light from the optical components in the AFM head, oblique illumination was used, provided by a CooLED pE300 light source and delivered via a liquid light guide whose output was passed through the excitation filter, then weakly focused and steered onto the sample surface through the gap between the AFM head and stage using a kinematic tip/tilt mount. A photograph of the top view optics and AFM setup is shown in Figure S6.

AFM long Z scanner
Due to the large topography and extremely compliant nature of the samples, the 15 μm range of the Z piezo in the AFM head was not sufficient. In our setup, the Z piezo in the AFM head was disabled, and the sample was mounted on Z piezo stage with 100 μm of range (P-621, Physik Instrumente), which was controlled by the low voltage Z signal from the AFM passed through a high voltage amplifier (PI E500 with E505 HVA and E509 signal conditioner/piezo servo module, Physik Instrumente). Closed loop control was implemented in the E509, and the sensor monitor signal was passed to the AFM controller through the signal access module.
The long Z stage was calibrated by scanning 3 separate silicon calibration gratings in tapping mode, using the Z piezo in the AFM head and then the long Z stage immediately afterwards.
A calibration factor to fix the calibration of the long Z stage such that the height sensor signal from the long Z stage matched that from the (factory calibrated) height sensor in the AFM head S5 was calculated from height histograms of these measurements and set in the software. The measurement of the grating with the Z piezo within the AFM head agreed with the height specifications of the gratings within the manufacturer's specified tolerances, and measurements after calibration showed agreement to within better than 0.5% between the calibrated long Z stage and the Z piezo in the AFM head. and immersed in buffer. A commercial temperature controller (HTHS, JPK, which used a Eurotherm 2216e controller) was re-tuned to control the system, and found to maintain the temperature to within +/-0.5 degrees of the set temperature by independent measurement with a type K thermocouple and DAQ card. A photograph of the constructed stage is shown in Figure S7.       Figure S7. Annotated photograph of the long range Z stage, petri dish heater and coarse sample positioning stage. The AFM head has been removed for clarity. The long range z scanner is largely obscured by the acrylic mount/splash guard, and the resistive heater is obscured under the copper petri dish mount.