The cells used in this study are the human hepatoma cell lines Huh-7 cells, obtained from the Type Culture Collection of the Chinese Academy of Sciences, Shanghai, China. The cells were cultured in Dulbecco’s Modified Eagle Medium (Gibco, Life Technologies, China) supplemented with 10% fetal bovine serum (Gibco, Life Technologies, Australia) and 1% penicillin-streptomycin (Gibco, Life Technologies, United States), at 37°C and 5% CO₂ in humid conditions. The pH level of normal culture medium is set as 7.4. One day prior to the AFM experiments, the cells were seeded onto 25 mm × 25 mm glass slides. After the cells had enough time to adhere to the substrate, the existing medium was replaced with the fresh medium to remove dead and loosely attached cells every 2 h. The fresh medium was prepared in the normal or acid condition to mimic microenvironments at different pH levels. HCl (0.5 mol/L) was added precisely to the normal culture medium for the acid culture medium (pH 6.5) by using a pH meter measurement to verify the targeted pH value of 6.5. Phosphate buffer saline (PBS) solution was used to rinse the substrate every time before adding and changing the culture medium. The cells were cultured with either pH microenvironment for at least 6 h in CO₂ incubator for subsequent characterization and measurement in AFM experiments. To maintain the pH level in the cell cultivation, the new acid culture medium of pH 6.5 was replaced every 2 h. By pH monitoring, we found that the acid culture medium of pH 6.5 in the CO₂ incubator can be kept well within 2 h. The time of AM-FM AFM measurement was limited to a maximum of 2 h. During the AM-FM AFM measurement, the pH monitoring was employed constantly and the acid culture medium of pH 6.5 was added when needed.

AFM experiments were performed with an Asylum Research MFP-3D Infinity AFM system. The AM-FM bimodal imaging mode was performed by using BL-AC40TS cantilever (Olympus, Japan) with 0.11 nN/nm nominal force constant (k ₁). The spring constant for the first eigenmode k ₁ of the cantilever was obtained by force constant calibration. In AM-FM method that employs the tapping mode, the cantilever is excited near the first and second resonant frequencies simultaneously. To ensure the physiological conditions of cells, the cantilever was completely immersed in the liquid of culture medium during the experiments. The cantilever tuning was conducted in liquid by which the first cantilever eigenmode was excited to the resonant frequency ( f 1 ≈ 30   kHz ) by the AFM software (Asylum Research) and the second eigenmode to the second-order resonant frequency ( f 2 ≈ 110   kHz ) . And the spring constant for the second eigenmode k 2 of the cantilever can be calculated by k 2 = ( f 2 f 1 ) 2 k 1 .

During the scanning, the first order amplitude A 1 is adjusts to maintain constant that is equal to amplitude setpoint A 1 , set , so that the displacement of scanner in z-direction can be used for living cell topography tracking. Meanwhile, the cantilever is also excited at the second resonant frequency f 2 . Here, a frequency feedback loop adjusts the frequency by a small amount Δ f 2 to maintain the second eigenmode on resonance as the AFM tip interacts with the cell. The AM-FM signals were recorded by the lock-in and feedback system and can be used to extract the effective storage modulus E storage for elastic property [42]: E storage = π R 1 6 ( k 1 Q 1 A 1 , free A 1 , set cos ∅ 1 ) − 1 2 ( 2 k 2 Δ f 2 f 2 ) 3 2 , where Q 1 are the quality factor, A 1 , free (4.0 V) is the free amplitude tuned for the first eigenmode in liquid, A 1 , set (2.0 V) is the setpoint used for imaging, and R (∼10 nm) is the radius of AFM tip provided by the manufacturer. As described in details [43], the mapping of loss tangent E loss can be derived by the amplitude ( A 1 ) and phase signals ( ∅ 1 ) in the first resonance mode: E loss = π ( A 1 , set A 1 , free − sin ∅ 1 ) cos ∅ 1 .

The loss tangent is equivalent to the ratio of the dissipated energy to stored energy: Loss   tangent = E loss E storage , which allows us to calculate the loss modulus for viscous property. Correlated topography and viscoelastic images were generated with a resolution of 256 × 256 pixels. And the scan speed was set at 0.5 line/s. As the modulus of the AFM tip used defers greatly from that of the glass substrate, the mechanical signal from the background part of glass substrate is meaningless and not considered, and is shown as the gray background in Figures 4, 5. The same AFM tip and imaging configuration were used for all samples to maintain consistency between measurements of cell viscoelasticity. Repetitions of experiments with more living Huh-7 cancer cells under different pH states are performed to verify the robustness of results and mechanisms in our work.

Confocal fluorescence microscopy was used to observe the F-actin microstructures and validate the AFM imaging. The cells for the imaging by confocal fluorescence microscopy were seeded onto 20 mm cell slides and incubated with the same culture conditions as for AFM measurements. The cells were washed with PBS, fixed with 4% paraformaldehyde in PBS and kept at 4°C for around 15 min. Afterwards the cells were permeabilized with 0.5% Triton X-100 in PBS at room temperature for 15 min and rinsed with PBS. The nucleus of cells was stained with DAPI (Solarbio, China) for 15 min in dark. The F-actin cytoskeleton of cells was stained in dark using Alexa-Fluor 488 Phalloidin (Solarbio, China) for 30 min. Finally, the slides were mounted with the DAPI mounting medium (Solarbio, China). The stained cells were examined by using a confocal microscope setup (Leica SP8X, Germany) to visualize the F-actin microstructures. The fluorescent F-actin cytoskeleton images were acquired with a 63x/1.40 oil immersion objective lens (Leica Microsystems, Germany).

To simultaneously obtain the images of topography and viscoelastic properties of living cells under physiological conditions, we have improved the original AM-FM AFM assembly. In the AM-FM method, two driving signals with different frequencies (black and red signals) are superposed to excite two independent vertical oscillation modes of the cantilever (Figure 2A). The first resonance operates in normal tapping mode (AM mode). The amplitude A 1 controls the vertical feedback loop of the standard tapping mode for topography, and the phase ∅ 1 gives the values for the loss tangent (black box). The second resonance operates in the FM mode. The change in resonance frequency f 2 determines the elasticity, while the change in the amplitude A 2 gives the dissipation information (red box). Therefore, the AM-FM method strictly relies on the resonance information of the cantilever, and the stability of the first-order and second-order resonance peaks limits the signal-to-noise. Although AFM with AM-FM mode is competent for the simultaneous mapping of topography and viscoelasticity, the original AM-FM AFM of MFP-3D Infinity AFM system is inappropriate for the in situ environment of cell culture medium, as the build-in high damping material that is required to optimize the resonance of the cantilever cannot work in the liquid of culture medium. Without the high damping material, it is difficult for AM-FM AFM to achieve high-resolution imaging of F-actin microstructures [44]. In our work of in situ AM-FM AFM imaging, we unloaded the original high damping material and naturally used the liquid of cell culture medium as the high damping to dissipate unnecessary vibration in the oscillation process of AFM cantilever that deviates from the standard resonance. Figure 2B shows the first resonance signal of the cantilever without high damping in air, while Figure 2C shows the first resonance signal of the cantilever with the high damping in liquid. The introduction of the high damping eliminates spurious peaks of the vibration information from unnecessary media and optimizes the resonance of the cantilever to form a clean standard peak in liquid. Similarly, the cluttered signal of the second-order resonance is greatly reduced due to the presence of high damping in liquid (see Supplementary Figure S1).

In addition, the ratio A 1 , set / A 1 , free is set in the range of 0.3–0.5 which ensures the optimal tracking of surface topography for living cell imaging in AM-FM AFM. By introducing the high damping and setting the appropriate parameters, we successfully stabilize the cantilever oscillation, efficiently increase the signal-to-noise ratio, and achieve a high-resolution mapping in AM-FM AFM. As shown in Figure 3, we are able to simultaneously measure the cell topography (Figures 3A,B), associated with the distributions of storage modulus E storage (Figure 3C), loss modulus E loss (Figure 3D), and loss tangent (Figure 3E) of living cancer cells under physiological conditions. By taking advantage of the high contrast in the mapping of loss tangent, we have achieved an ultrahigh resolution down to 50 nm (Figure 3F) and captured cytoskeleton microstructures under in situ microenvironment.

Figure 3 shows the topography and viscoelasticity of living cells. It is important to note that the string microstructure observed in the images of viscoelasticity (Figures 3C–E) are highly correlated to the features observed in the topography image (Figures 3A,B). The widths of the strings are tens of nanometers. The storage moduli E storage of the strings are higher than the gaps in between as shown in Figure 3C, while the loss tangents of the strings in Figure 3E are lower. These results show that the string microstructure takes the elasticity as the dominant property, which may correspond to the mechanical characteristics of cytoskeletons reported in previous study [45]. Hence, the same topography and viscoelasticity give us a strong implication that the microstructures shown by AM-FM AFM are cytoskeletons, which will be confirmed as F-actin by the latter section of immunofluorescence confocal microscopy. This AM-FM AFM method allows a direct observation in correlation of cytoskeletal structure with both topography and viscoelasticity for living cells, which shows better capability and application than the optical microscope that is limited by the diffraction limit resolution [31], the transmission electron microscope that cannot maintain the physiological state of living cells [32], and other methods for measuring mechanical properties of cells including the traction force microscope, optical tweezers and magnetic twisting cytometry that cannot obtain the topography simultaneously [35–37]. Our improvement of AM-FM AFM has increased the resolution of cell mechanical imaging, and achieved simultaneously characterization of cytoskeleton topography, viscoelasticity, which are crucial for studying the relationship between the organization of cytoskeleton microstructure and the distribution of mechanical properties associated with their effects on cellular behaviors and functions.

Based on the in situ AM-FM AFM method we improved for simultaneous mapping topography and viscoelasticity of living cells, it is feasible to study the relationship between the organization of cytoskeleton microstructures and the distribution of viscoelastic properties under different pH microenvironments. The microenvironment of cancer cells in tumor tissue is constantly acidified [23], with a lower extracellular pH value (∼6.5) than the basic value (∼7.4) in normal tissue [22]. The acid microenvironment enables the progression of cancer cells by promoting proliferation, the evasion of apoptosis, metabolic adaptation, migration and invasion that are involved with the malignant transformation and metastasis [46]. During this stage, the F-actin and the lamellipodium are the two crucial media linking the cellular behaviors and the mechanical properties. It has been reported that the acid extracellular pH induces the reorganization of F-actin cytoskeleton and facilitates the formation of lamellipodial protrusion [25], which can modulate the morphology and rigidity against the forces needed for cell migration [47]. It would be a key challenge and a great significance to study the correlation between the organization of F-actin structures and the viscoelasticity distribution in lamellipodium of living cancer cells under normal (pH 7.4) and acid (pH 6.5) microenvironments for revealing the nanomechanical mechanism of the cellular migration in metastasis.

In the normal microenvironment (pH 7.4), we use the improved AM-FM AFM method to characterize the topography and the viscoelasticity of living Huh-7 cells as shown in Figure 4. Figure 4A shows an overall topography of the cell edge, with the region of lamellipodium in the red box measured by larger contact force and scanning density for higher resolution in topography (Figure 4B), storage modulus E storage (Figures 4C,D), loss modulus E loss (Figures 4E,F), and loss tangent (Figures 4G,H). The topography image of Figure 4B represents the shot and thin cytoskeletal structures randomly dispersed in lamellipodium, which is also reflected by the homogeneous patterns of viscoelasticity (Figures 4C,E,G and Supplementary Figure S2). The histogram analyses of storage modulus, loss modulus and loss tangent for Figures 4C,E,G were performed as shown in Figures 4D,F,H, respectively. The storage modulus in the lamellipodium region (Figure 4D) indicates a single Gaussian peak for elasticity with the main range of 50–80 kPa and the mean value of ∼65 kPa, which are in good agreement with previous studies [29, 48]. Similarly, the viscous properties measured by loss modulus and loss tangent exhibit monomodal Gaussian peaks, which further indicates the homogeneous distribution of cytoskeletons (Figures 4F,H). The Gaussian peak of loss modulus has a mean value of ∼18 kPa with the range varying from 14 to 24 kPa, while that of the loss tangent has a mean value of ∼0.28. The mappings of storage modulus and loss modulus demonstrate that the cytoskeletons have slightly higher elasticity and viscosity than the background of intracellular fluid or other organelles (Figures 4C,E) [48]. Note that the greater difference between the cytoskeletons and the intracellular background is shown in the mapping of loss tangent (Figure 4G), owing to the dominant elasticity of cytoskeletal filaments [28, 29]. As a result, the mapping of loss tangent can show higher contrast and resolution with the patterns greatly corelated to the cytoskeletal structures in the topography image, which further proves the capability in characterizing the detailed organization of cytoskeletal structures by the improved AM-FM AFM. Repetitions of experiments for more living Huh-7 cancer cells at pH 7.4 (Supplementary Figure S2) show similar cytoskeleton organization and cell viscoelasticity, which verify the robustness of our results.

To study the effect of acid microenvironment on living cancer cells, we change pH value of the culture medium to pH 6.5 [22], and measure the topography and the viscoelasticity of Huh-7 cells by the improved AM-FM AFM as shown in Figure 5 and Supplementary Figure S3. Unlike the homogeneous dispersion of cytoskeletons in lamellipodium at pH 7.4, the topography of Figures 5A,B show highly oriented and organized patterns of cytoskeletal structures at pH 6.5. The cytoskeletons are polymerized to long filaments and woven into thick bundle-like structures directed to the protruding direction of lamellipodium. The viscoelasticity, including storage modulus, loss modulus and loss tangent, of the same lamellipodium region in Huh-7 cells are measured and analyzed in Figures 5C–H in which the polymerized filaments and bundles are visible. The strings that have higher storage modulus, higher loss modulus and lower loss tangent than the intracellular background indicate the cytoskeletal filaments in Figures 5C,E,G, respectively. Interestingly, along with the convergence of cytoskeletal filaments towards the protruded lamellipodium, the storage modulus and loss modulus increase while the loss tangent decreases. The gradient of viscoelasticity in the lamellipodium region with the relationship to the cellular migration will be discussed in details in the latter section.

The histogram analyses of storage modulus, loss modulus and loss tangent were also performed as shown in Figures 5D,F,H, respectively. Unlike monomodal Gaussian peaks at pH 7.4, all of the storage modulus, loss modulus and loss tangent show bimodal distributions at pH 6.5, corresponding to the heterogeneous organization of cytoskeletons. In Figure 5D for elasticity, the lower peak of storage modulus with the mean value of ∼35 kPa represents the thin and unwoven cytoskeletal filaments, while the higher peak (∼60 kPa) the thick cytoskeletal filaments that have been bundled together. These results of elasticity distribution are consistent with the previous studies that the increased density of actin filaments can lead to the stiffening of cells [28, 29, 49]. The bundled actin filaments are bound by crosslinking proteins and molecular motor, such as fascin and myosin-X, providing strong viscosity [50, 51], which results in an elevated loss modulus as the higher peak (∼40 kPa) in Figure 5F. Our work visualizes the organization of cytoskeletal filaments and reveals the structural origins of the viscoelasticity distribution in the lamellipodium region of living cancer cells under acid microenvironment. Repetitions of experiments for more living Huh-7 cancer cells at pH 6.5 (Supplementary Figure S3) show similar cytoskeleton organization and cell viscoelasticity, which verify the robustness of our results. We also performed the experiments at the pH level within the range from 6.5 to 7.4, and found that the more cells with F-actin polymerization and viscoelasticity gradient can be seen at the lower pH state.

To identify the type of cytoskeletal structures shown by the improved AM-FM AFM as in Figures 4, 5, the immunofluorescence confocal microscopy was adopted and the fluorescent phalloidin that can selectively bind to F-actin was used to image the distribution of F-actin in the living Huh-7 cells (Figure 6) [52]. The Huh-7 cells were cultured in the two different pH environments (pH 7.4 and pH 6.5) as in AFM experiments and stained with Alexa-Fluor 488 Phalloidin and DAPI. The imaging of confocal microscopy shows that the F-actin filaments dyed by green are short, thin and randomly dispersed within the whole Huh-7 cell at pH 7.4 (Figure 6A). Although becoming slightly thicker near the periphery, the F-actin filaments are loosely arranged with the same homogeneous feature observed in AFM experiments (Figure 6B). On the contrary, at pH 6.5, the F-actin filaments that are much thicker and longer form highly oriented bundle-like structures (Figure 6C) and converge toward the front of lamellipodium (Figure 6D), which is in good agreement with the AFM imaging as well. The consistency of intercellular structures confirms that the cytoskeletal filaments visualized by the improved AM-FM AFM are the F-actin.

In previous sections, by using the improved AM-FM AFM, we have successfully characterized the viscoelasticity distribution of the lamellipodium regions in living Huh-7 cells at pH 7.4 and 6.5 (Figures 4, 5). To compare them quantitatively, we rescale the mapping of storage modulus (Figure 7A) and loss modulus (Figure 7B) to the same color range for studying the effect of viscoelasticity distribution on cell adhesion and migration. Figure 7A shows that along the direction of lamellipodium protrusion (as the arrows), the storage modulus of cell elasticity increases from ∼50 kPa to over 60 kPa at the acid microenvironment and maintains at a plateau (∼60 kPa) at the normal microenvironment. For cell viscosity following the same arrows, the loss modulus increases from 35 to over 50 kPa at pH 6.5, while staying at a low level of ∼10 kPa at pH 7.4 (Figure 7B). Combining the AFM images of cytoskeletons, we can conclude that the acid microenvironment may result in the organization of F-actin filaments and the gradient of both elasticity and viscosity in living cancer cells.

Recent researches of cell migration and cancer metastasis have reported that the living cancer cells prefer to migrate toward the substrate with higher stiffness [19, 53]. However, few studies focus on the distribution and alteration of cell viscoelasticity during migration associated with the effects of cell viscoelasticity on migration. According to literatures, the higher elasticity and viscosity enable stronger adhesion to substrate [13, 54]. In our work, by using the improved AM-FM AFM, we found that the F-actin cytoskeleton aggregate and form the thick bundle-like filaments directed to the protruding direction of lamellipodium, leading to gradient increases in elasticity and viscosity. The gradient of viscoelasticity suggests a gradient of adhesion strength, which means that the lamellipodium can achieve stable adhesion at the front but have high possibility to detach the substrate at the back, resulting in the driving force for cell migration towards the direction of lamellipodium protrusion. Moreover, the aggregation and the stiffening of F-actin cytoskeleton allow focal adhesions to enhance cell adhesion, and provide sufficient cell rigidity against the forces needed for migration [55, 56]. In all, the homogeneity of F-actin cytoskeleton and viscoelasticity in the lamellipodium at pH 7.4 suppresses the migration of cancer cells, which consistent with the fact of uneasy malignant transformation of cancer cells under normal environment [47, 57]. Conversely, the aggregation of F-actin cytoskeleton and the gradient of viscoelasticity at pH 6.5 facilitate cell adhesion and migration, which explains the strong tendency in migration and metastasis of cancer cells under the acidified microenvironment in tumor tissues [47, 58]. These findings may help us to understand the structural and nanomechanical mechanism of malignant behavior of cancer cells under different pH microenvironments.