Blood Flow Dynamics Has a Major Influence on the State of Circulating Tumor Cells

Cancer mortality mainly arises from metastases, due to cells that escape from a primary tumor, circulate in the blood as circulating tumor cells (CTCs), leave blood vessels and nest in distant organs. It is still unclear how CTCs overcome the harsh conditions of fluid shear stress and mechanical constraints within the microcirculation. A model of the blood microcirculation was established through the fabrication of microfluidic channels comprising constrictions. Metastatic breast cancer cells of epithelial-like and mesenchymal-like states were flowed into the microfluidic device. These cells were circulated, visualized while in circulation, and analyzed for their dynamical behavior. They were also retrieved post-circulation for staining of DNA damage response and gene expression analysis of key regulators of epithelial-to-mesenchymal transition. Altogether, changes from molecular to cellular scales were observed. Therefore, it is pivotal to consider the full consequences of the circulation in order to understand the metastatic cascade as a whole.


Introduction
Understanding the process of metastasis is a major challenge in the fight against cancer. This process is a multi-step one that often involves cells migrating from a primary tumor (at 90% of epithelial origin, i.e. carcinomas) into the blood stream (intravasation), where they reach a distant organ by re-crossing the endothelial barrier (extravasation). At both cellular and molecular levels, and in most cases, these events involve the ability of cells to undergo complex reprogramming processes named the epithelial-to-mesenchymal transition (EMT) at the primary tumor site, followed by the reverse process called MET (mesenchymal-toepithelial transition), which may help in establishing distant micro-metastases after the circulation step. [1][2][3][4] Blood circulation, together with gene expression reprogramming, therefore plays a central role in the metastatic cascade.
Once in the bloodstream, circulating tumor cells (CTCs) experience shear stress. In addition, considering that capillaries do not have uniform diameters but are regularly constricted instead 5 , CTCs encounter various constraints within the microvascular network, leading to repeated mechanical deformation and eventually to arrest. Although numerous studies have revealed important influences of mechanical constraints in cancer 6 , surprisingly only a few works explored the responses of cancer cells to the mechanical stressors provided by the blood circulation, and most of them were limited to the study of the effect of shear stress.
Malignant cells appeared more resistant than non-malignant cells to shear stress 7,8 yet they still underwent significant apoptosis 9 , in particular as compared to hematopoietic cells. 10 Altered cytoskeleton organization in suspended and sheared primary ovarian cells was observed. 8 Interestingly, the process of suspending cells in and of itself triggers myosin-II inhibition, leading to increased cell stiffness. 11 All these studies, however, involved flowinduced shear without topographical constraints, and relied on macroscopic tools that limited the control of the cell treatment uniformity, preventing single cell studies. The investigated biomarkers were also rather limited, targeting essentially cell survival.
Nonetheless, several works have used microfluidic systems to address the effect of the geometry of the microvasculature on cells. Pioneering work by Preira et al. investigated the role of cell deformability in the pathological arrest of leukocytes in the blood microcirculation of the lungs, by devising a microfluidic system including evenly spaced constrictions. 12 In the field of cancer, using circulated glioblastoma and normal glial cell lines, Khan et al. demonstrated the cell entry time into the confined space provided by 11 µm-wide diameter microchannels as a better marker of malignancy than deformability. 13 Au et al. showed that clusters of CTCs could reorganize reversibly in order to traverse microchannels of 5-10 µm-wide. 14 Nath et al. flowed HeLa cells across 7 µm-wide constrictions, and demonstrated that cell viability was reduced by 50%, but that the expression of MMP2, a metalloproteinase involved in stromal tissue degradation, was unchanged. 15 In a very interesting study, Xia et al. flowed leucocytes, MDA-MB-231 and MCF-7 cell lines into arrays of pores. 16 They reported a pressure dependence of the deformation of cells and nuclei, and proposed that such studies could guide the optimization of CTC sorting devices. Overall, microfluidic attempts at mimicking cancer cells in the blood circulation are still sparse, and are mostly focused on a phenomenological investigation of their mechanical properties.
The question of the effect of the blood circulation on the cell phenotype and its genome, and ultimately aggressiveness has been hardly addressed. The acquisition and maintenance of the key mesenchymal phenotype for the metastatic process, involving i.e. up-regulation of vimentin and down-regulation of E-cadherin expressions, require important cellular reprogramming by the activation of master regulators including transcription factors such as Snail, Twist and zinc-finger E-box-binding (ZEB), and transforming growth factor, TGFβ. 17,18 These phenotypic changes, however, are not bimodal, and recent studies suggest that disseminating tumor cells may present diverse and heterogeneous combinations of epithelial and mesenchymal phenotypic traits. 14,19,20 In this article, the main aim is to tackle the least studied component of the metastatic cascade to date, which is the transient circulation step in the blood stream. More precisely, we will explore the combined influence of shear stress and physical constraints on the characteristics of circulating cancer cells. The main questions that will be addressed are how do circulation and constrictions regulate the phenotype, genome integrity and gene expression of CTCs. To address these questions, cells of two different origins, i.e. epithelial (SK-BR-3) and mesenchymal (MDA-MB-231) breast cancer cell lines, were circulated in microfluidic channels using a flow control system to recapitulate the pressure-velocity patterns of the microcirculation blood flow. While under circulation, the tumor cells were constrained due to the presence of evenly-spaced multiple constrictions to mimic the mechanical condition in the blood capillary bed. The ways in which shear flow and mechanical constraints can promote changes in gene expression were investigated and elaborated based on the common framework for EMT and its transcription factors (EMT-TFs). Overall, we found that circulation affects the cells at several scales, i.e. at cellular, sub-cellular and molecular levels, and that some of these changes are modulated by the epithelial or mesenchymal initial cell type. Therefore, the role of the circulation step seems to go beyond a simple disseminating function of cancer cells to distant organs. For this reason, it should be considered as an active step that is likely to modify gene expression of CTCs and possibly their mechanical properties.

Constrictions
To test the effects of different geometric mechanical constrictions on the flow-induced migration of single tumor cells, five types of geometric microfluidic models were fabricated (Fig. 1a). All five types comprised channels 420 µm in length. The "unconfined" type had a channel height of 20 µm, a value reduced to 15 µm in the "confined" type. Confined type 1, 2 and 3 comprised channels with multiple (11, 7 and 5 respectively) 6 µm-wide constrictions spaced apart by 26 µm-wide chambers that were 20 µm in length. These microfluidic designs were calculated for their flow resistance based on Newtonian fluid flow resistance (Fig. 1b).
A pressure set-point of 10 kPa applied across the whole microfluidic circuit was chosen in order to achieve the same order of magnitude of flow rate (i.e. approximately 1 µl/min) as the blood flow rate reported in micron-sized capillaries in vivo. 21 Either poorly (SK-BR-3) or highly (MDA-MB-231) metastatic breast cancer cell lines were delivered into these five types of geometric microfluidic models for single cell mechanical phenotyping by flowinduced migration. The velocity of these cells through the micro channels was measured ( Fig.   1c), with values in the range of a few mm/s in agreement with recent in vivo values reported in the literature. 22 More precisely, mean values close to 18 mm/s were found in the unconfined condition for both cell lines, reduced to about 10 mm/s and 8 mm/s in the confined situation for SK-BR-3 and MDA-MB-231 cells, respectively. Then, velocity data of both cell lines displayed a general trend, in which a longer constriction tends to systematically induce a lower mean velocity, with mean values of 2.8 mm/s and 5.2 mm/s in the 60 µm long constriction for SK-BR-3 and MDA-MB-231 cells, respectively. The cell path trajectories through the micro channels with constrictions were macroscopically scrutinized and similar behavioral patterns were observed for both cell lines and for two different pressure set points, as illustrated in Fig. 2a for SK-BR-3 cells. A first observation is the large dispersion (i.e. over two orders of magnitude) of the "total transit time", i.e. the total time spent in the constricted channels, whatever the applied pressure set point (Fig. 2b). Quite interestingly, these "position versus time graph" curves revealed that the main factor limiting migration is the arrest in the first constriction. Once this constriction is passed, the subsequent ones are crossed smoothly with minimal arrest (Fig. 2b). Both cell types undergo a strong deformation in the first constrictions (Fig. 2d), which increases with the constriction length as expected from a crude volume conservation hypothesis. This initial deformation is partly maintained in the rest of the cell journey through the subsequent constrictions (  Table 1). The ratio between the cell residence time in the first constriction and the total transit time is plotted in Fig. 2c, showing that around 50% of the total transit time is spent in the first constrictions in all constricted geometries. Let us note that, however, only MDA-MB-231 cells display a significant increase of residence time as a function of the length of the constriction. Overall, these results suggest that cancer cells crossing constrictions retain a persistent memory of such an event, which also seems to depend on the initial cell type. We then explored the consequences of this morphological plasticity observed at the cellular, subcellular and molecular levels. We focused our efforts on the nucleus, the biggest organelle in the cell, which was previously identified as the limiting one for cells in cells were isolated from the microfluidic system, fixed while in suspension, and their nuclei stained with DAPI (Fig. 3a). Approximately, one thousand cells were collected in one hour.
Therefore, the maximum time over which cells are left in suspension, including the time spent before and after crossing the constrictions, is under an hour. Since each of the five geometric microfluidic models could impose unique mechanical constraint on the cells, their cell area, nucleus area, nuclear to cytoplasmic (N:C) ratio, and nucleus aspect ratio (AR) were quantified for each geometry (Fig. 3b). Morphological analysis showed significant increases of nucleus areas in SK-BR-3 cells in most conditions, with the highest changes observed as a result of circulation without confinement (channel height of 20 µm) and with confinement (channel height of 15 µm). This suggests that these cells respond strongly to being suspended and submitted to shear stress, and relatively less to the additional strain provided by constrictions. Conversely, MDA-MB-231 cells appeared quite insensitive to the transition from an adherent to a suspended state, and from being circulated. They however display significant changes in nucleus size when constricted. These trends were similar regarding cell size. This led to almost unchanged nuclear to cytoplasmic size ratio for both cell types and conditions, except for the unconfined and confined conditions for SK-BR-3 cells reflecting the large observed relative changes of nucleus area in this cell type. Finally, the nucleus aspect ratio (computed from an ellipsoidal fit, see Methods) appeared significantly higher in MDA-MB-231 cells, as already reported by Xia et al. 16 , but was not affected by any of the stress encountered by the cells. Subsequently, to further investigate how the morphological changes observed in most circulated conditions as compared to the adherent control condition affect the genome integrity of the cells, the extents of DNA damage and repair activity due to circulation, confinement and constrictions were investigated by staining the cells with γ-H2AX (Phospho-Ser139) antibody, a marker for DNA damage response (Fig. 4a). The integrated densities of γ-H2AX in the nucleus and cytoplasm fractions were measured and their ratio were represented (N:C ratio, see Fig. 4b).
These results showed that DNA damage and repair activity is significantly activated in SK-BR-3 cells in all circulated groups, as compared to the uncirculated (control) group. In MDA-MB-231 cells, DNA damage and repair activity is constitutively activated (N:C ratio > 1) even in the control, and appears to be increased with a statistical significance only in the "unconfined" circulated case. Further analysis will be needed to understand this point but we can note that a similar trend (also not statistically significant) is observed for the N:C ratio in size.  (Fig. 6). Since Twist2 is known to play a role in tumor progression in breast cancer 24 , even though its expression has been less explored than that of Twist1, we decided to measure its expression and cell localization.

Co-localization of Twist2 Protein
The co-localization of Twist2 protein in the nucleus and cytoplasm was determined in SK-BR-3 and MDA-MB-231 cells in response to fluid shear stress alone, and to the combination of fluid shear stress and mechanical constraints by staining with Twist2 antibody (Fig. 7a).
The integrated densities of Twist2 in the nucleus and cytoplasm fractions were measured and represented as N:C ratio (Fig. 7b). SK-BR-3 cells in the circulated groups exhibited significant increases of nuclear co-localization of Twist2 as compared to the control group.
On the other hand, MDA-MB-231 cells in the circulated groups exhibited either maintained N:C levels of Twist2 localization or significant increases of cytoplasmic co-localization of Twist2 in two of the constricted conditions as compared to the control group.

Discussion
Circulation of detached cancer cells within PDMS microchannels provides a foremost model to comprehend the fate of CTCs during their transient existence into the blood circulation. In the present work, we have studied several aspects of this fate, starting from the dynamical behavior of cells that were flow-driven into constrained microchannels. Our data highlight differential mechanical characteristics of poorly versus highly metastatic cancer cell lines in terms of deformability and plasticity. In the literature, the mechanical phenotype of cells is emerging as a potential biomarker for cell types, particularly cancer cells. Many studies have reported the usage of microfluidic devices to measure the mechanical phenotypes of cells including transit time, entry time, cell size, elastic modulus and cell fluidity. 12,[25][26][27] This was applied to dissociated cells from tumors, in line with previously proposed deformation assays. 28,29 We believe that this is particularly interesting for CTCs, since they are expected to undergo large mechanical constraints during their crossing of microcapillaries in vivo. In the present study, not surprisingly, higher confinements led to lower average velocities for cells transiting across a channel with micro constrictions (Fig. 1). Cells undergoing flowinduced migration under 3D confinement with multiple constrictions generally underwent a significant arrest at the first constriction (Fig. 2). Subsequently, they used up a significant amount of time to cross this first encounter. We interpret these high residence times in the first constriction as the time needed by the cells to deform in order to conform to the mechanical constraint of the constriction and to cross it. Once deformed by this first constriction, their subsequent crossings of constrictions of the same type were significantly faster. This suggests a capacity to store mechanical deformation on timescales comparable to  Table 1). This might suggest that the mechanical response of the cell, considered as a viscoelastic body, is strongly nonlinear, and display "hardening" (higher apparent modulus) when deformation increases. This raises intriguing and, to our knowledge, not fully understood questions regarding the mechanics of the nucleus. 25,27 Different experimental tools would be needed to address this question in detail, as it is difficult to characterize different cell samples based on transit time data only. 30 Studies on DNA damage response (DDR) to mechanical stress, particularly fluid shear stress, mechanical confinement and constrictions are still scarce, 31 yet they are slowly emerging within the last years. DDR is a series of coordinated responses designated to remove damage incurred to the genome. 32 Using transformed cell lines, Singh et al. reported that lamins A/C, essential nuclear envelope proteins, are required for maintaining genomic stability and that their depletion stalls DDR. 33 Later, Davidson et al. associated the role of lamins A/C and DNA damage with migration-induced nuclear deformations in fibroblasts. 23 Several additional studies emerged, reporting that the genomic instability caused by migrationinduced nuclear deformation and DNA damage promotes cancer heterogeneity. [34][35][36] Furthermore, Jacobson et al. reported and identified changes in transcripts from RNA-sequencing of neutrophil-like cells in response to migration and constriction stress. 37 This establishes an interesting interplay between nuclear mechanics, genome integrity and phenotypic transformation. Here, tumor cells that underwent circulation, confinement and constrictions indeed were found to present increased levels of DNA damage marker, γ-H2AX (Fig. 4). As compared to MDA-MB-231 cells, SK-BR-3 cells exhibited more significant increases in nuclear co-localization of γ-H2AX. Simplistically, this could suggest that DDR The plasticity of EMT in cells from aggressive tumors enables them to switch from proliferative to invasive phenotypes and vice versa. 39,40 It has been vastly observed in mouse mammary tumor models that the initiation of EMT can occur during the early stages of tumorigenesis and progress during its later stages. [41][42][43][44] The EMT process is considered as a hallmark-facilitating program as well as having influence on pathways linked to tumor progression and metastatic dissemination. This raises the question of the status of CTCs amongst the spectrum of functional and morphological characteristics, and notably the epithelial and mesenchymal ones. It was proposed that CTCs have had undergone EMT or are still continuously undergoing the transition while in the circulation. 45 While it has been reported at the cellular level that the functional roles of the cytoskeleton modulate suspended cell mechanics and that of substrate-adhered cells, 11  Snail2, and ZEB1 as key EMT-TFs that are involved in metastatic breast cancer through different signaling cascades. 46 Here, we showed that, in SK-BR-3 and MDA-MB-231 cells, there is a substantial relationship between EMT-TFs and mechanical stress from fluid shear stress and mechanical constraints mimicking those present in the microcirculation -the route that CTCs take in order to disseminate (Fig. 5). Amongst the EMT markers tested, the ZEB family, Twist1 and N-cadherin did not show any significant difference, while Vimentin, Snail2 and E-cadherin genes were significantly down-regulated. Further work is necessary for a coherent explanation of these first results. However, let us reflect on the coherent overexpression of Twist2 in all confined conditions, a nuclear protein playing the role of distinct tissue-restricted transcription factor in MDA-MB-231 cells (Fig. 6). 47 The central role of Twist2 in embryogenesis and mesodermal development, and its targeting of multiple genes coding for cell-fate proteins inevitably links them to cancer and oncogenesis. 48 This gene has also been reported to be involved in metastasis formation through EMT, thereby facilitating cancer cell invasion in epithelium-based cancers. 49,50 It was also reported that the EMT program might be activated transiently through nuclear Twist2 in the tumor invasion front to facilitate cancer cell invasion and metastasis. 24 An overexpression of Twist2 in circulated MDA-MB-231 shed light on the singularity of Twist2 as compared to Twist1, and its possible activation under certain conditions of stress. In our case, this refers to shear stress. In addition to this gene expression pattern, we observed heterogeneous co-localizations of Twist2 protein ( Fig. 7). At baseline, Twist2 was co-localized much more in the cytoplasm of SK-BR-3 In conclusion, flowing cancer cells from cell lines in micro channels, in conditions tailored to mimic those encountered in the blood flow, has shown significant impacts on cellular processes that can lead to changes in morphological features and genetic expressions. More specifically, this circulation process encompasses (1) the hydrodynamic shear stress from blood flow and (2) the mechanical deformations from the geometrical constrictions of the microvasculature. It was shown that changes in morphological features and genetic expressions from this circulation process differ from one cell type to another. Notably, two cell lines presenting an epithelial-like and mesenchymal-like state, respectively, have different responses to different mechanical stress. Therefore, this work strongly suggests that the mechanical process of circulation in the blood flow may have a significant effect on cells at morphological and gene expression level, and that this effect is cell-type dependent.
Since circulation in the blood is a necessary step in a large fraction of metastatic dissemination events (except for dissemination through the lymphatic system), these findings suggest that this step should not be ignored when trying to comprehend the metastatic process as a whole. Future research efforts should involve (1) an increase in the number of constriction encounter events, through the courtesy of a more elaborate microfluidic design, (2) a more detailed study of the morphological recovery of cells post-constrictions, (3) potential DNA damage repair processes, (4) a comparison of the migratory potential of cells before and after transformation by the circulation and constrictions, and (5) the use of CTCs from patients. This latter study is also challenging, due to the rarity of CTCs. Of course, these studies should be performed in light of and in parallel with the current evolution of the general understanding of the instability of cancer cells and CTCs, and notably the EMT and MET. This should help to achieve a more in-depth understanding of the molecular mechanisms activated or repressed in CTCs in the blood microcirculation from hydrodynamic shear stress and mechanical deformations.

Photo-Masks Design
A computer-aided design software application for two-dimensional design and drafting called QCAD was used to design different types of geometry ( Supplementary Fig. 1). The total length of one geometrical channel is 420 µm. Each type of design consists of four 420 µm-  Table 2a). An optical lithography mask aligner called MJB4 (SÜSS MicroTec AG, Germany) was used to expose UV light unto the chrome mask that was aligned on the photoresist-coated substrate. The UV light exposure dosage was optimized for each of the desired thickness (Supplementary Table 2b).
Post exposure bake at 95°C is carried out directly after exposure. The exposed photoresist was then developed in a shaking bath of photoresist developer (Supplementary Table 2c).
After development, the developed image was rinsed with fresh developer solution for approximately 10 seconds, followed by a rinse with Isopropyl Alcohol (IPA) for another 10 seconds. Then, it was air-dried with pressurized air and hard baked for 30 minutes at 300°C.
The fabrication of a silicon master mold is now complete.

Soft Lithography
From this silicon master mold, a microfluidic polydimethylsiloxane (PDMS) chip is fabricated through soft lithography ( Supplementary Fig. 3). PDMS silicon elastomer, Sylgard 184 was purchased from Dow Corning (Michigan, USA). It was mixed with the provided curing agent at a ratio of 9:1 and then poured over the silicon master mold. They were cured at 70°C and left to harden for 4 hours. After curing, the PDMS stamp was separated from the silicon master mold. 1.5 mm holes were punctured on the designated areas of the PDMS stamp in order to create reservoirs. Oxygen plasma treatment was performed to induce permanent bonding and the chip was closed by a glass coverslip. From this, PDMS-bonded slides were used as chips for the microfluidic system.

Microfluidic System
All reagents and cell suspensions used during the experiments (Supplementary Fig. 4a) were stored in pressurized containers that were connected to the microfluidic chip through a manifold valve (Fluigent, France). An MFCS-8C Flow Controller (Fluigent, France) was used to control the pressure independently in each container to drive all reagents and cell suspensions into the microfluidic chip. A heat generator was used to warm the microfluidic device at 37°C throughout the experimental duration in order to achieve optimal culture conditions for experiments with live cells. Each microfluidic chip experiment experienced three major stages (Supplementary Fig. 4b). The first stage is the priming of the chip in which the microfluidic channels were sterilized with 70% Ethanol (EtOH) followed by a washing step with 1X Phosphate-Buffered Saline (PBS). In order to have non-adhesive surfaces, the channels were coated with 100% of 1 mg of PLL-g-PEG (Susos AG, Switzerland) solution that was dissolved in 100 mM of sodium bicarbonate (NaHCO3) buffer Total RNA extraction was performed using the ARCTURUS® PicoPure® RNA Isolation Kit (Applied Biosystems, USA). The kit was designed to recover high-quality total RNA consistently from fewer than ten cells and even from a single cell. Total RNA extraction was executed as directed by the manufacturer. Total RNA was measured using Qubit RNA reagents (Invitrogen, USA). Measurements were conducted using the Qubit 3.0 Fluorometer (Invitrogen, USA). All the RNA samples did not undergo more than two freeze-thaw cycles to avoid any potential nucleic acid degradation. Total RNA reverse transcription (RT) was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, USA). After reverse transcription, the cDNAs were diluted by 10-fold before amplification.
cDNA amplification was performed using KiCqStart® SYBR® Green qPCR ReadyMix™ (Merck, Germany). The mRNA primer designs (Sigma-Aldrich, USA) are listed in the Supplementary Table 3a. All real-time PCR reactions were performed in triplicates using a SmartCycler® automated real-time PCR system (Cepheid Inc., USA). Supplementary Table   3b sets the thermal cycling conditions for amplification. All target mRNA expressions were normalized to reference gene GAPDH. Relative mRNA levels were calculated using the -ΔΔCt method and were expressed as 2 (−ΔΔCt) .
Immunofluorescence 10 µL of cell suspensions that were retrieved from the microfluidic chip device as aforementioned were fixed with equal volume of 4% paraformaldehyde (PFA) and then dispensed on a poly-L-lysine coated glass slide (Sigma-Aldrich, USA). The cells were incubated at room temperature for 15 minutes. The fixed cells were washed with 1X PBS and then permeabilized with 0.1% Triton-X 100 at room temperature for 10 minutes. The cells were washed with 1X PBS and then blocked with 4% Bovine Serum Albumin (BSA) for 1 hour at room temperature. The cells were washed with 1X PBS and then incubated with primary antibody cocktail in 2% BSA at room temperature for 2 hours (Supplementary Table   4). The cells were washed twice with 1X PBS and then incubated with secondary antibody cocktail in 2% BSA at room temperature for 1 hour (Supplementary Table 4). The cells were washed twice with 1X PBS and then counterstained with ProLong Gold Antifade Mountant with DAPI (Invitrogen, USA). The slide was mounted with a cover slip and the cells were visualized using the Leica TCS SP8 confocal laser scanning microscopy platform (Leica Microsystems, Germany).

Transit Time Analysis
Videos from microfluidic experiments were analyzed with ImageJ Version 1.51. The Mtrack2 plugin was installed on ImageJ. It was used to track the positions of a cell cross a channel with constrictions (Fig. 2a). First, the entire image sequences were converted to mask. Then, the outline of the channel was subtracted from each image sequence. Finally, the positions of the cell were measured from each image sequence. This method automatically churned the measurements of the position vector of the cell that allowed the calculations of cell velocity and cell residence time. This method is only effective when there is only one cell crossing a channel from beginning to end. Thus, manual tracking was done on image sequences that had more than one cell crossing the channel within a single time frame.

Cell Morphological Analysis
All acquired images were analyzed with ImageJ Version 1.51. Regions of interest (ROIs) were used to define specific parts (cell and nuclear boundaries) of an image that was processed independently or measured. Only the pixels within any defined ROI were included in the calculations when measured. ImageJ's set measurements function allowed the generation of the following data: area, aspect ratio and integrated density. This method was also used to measure the protein presence and its localization in a cell. Integrated Density is the product of "Area" and "Mean Gray Value".

Statistical Analysis
All statistical hypothesis testing was conducted using the GraphPad Prism 7 software (GraphPad Software Inc., USA). An alpha of 0.05 was used as the cut-off for significance.
The Kruskal-Wallis test followed by Dunn's multiple comparisons post hoc test was used to compare three or more independent samples of equal or different sample sizes. This method was used to compare data on velocity, transit time and residence time of cells in the microcirculation on five independent samples from the same applied pressure force. This method was also used to compare data on nucleus area, nucleus length, nuclear to cytoplasmic area ratio, nucleus aspect ratio, cell area and nuclear to cytoplasmic ratio of and Dr B. Ladoux for constructive comments and fruitful insights. We thank the IPGG technological platform for technical assistance and support.

Author Contributions
Experiments and data analysis were performed by HC. Experimental design and conception were done by HC, CV and JLV. HC and CV co-wrote the manuscript. CV supervised the research. JLV contributed to supervision, results discussion and manuscript writing.