Human Mammary Cells in a Mature, Stratified Epithelial Layer Flatten and Stiffen Compared to Confluent and Single Cells

The epithelium forms a protective barrier against external biological, chemical and physical insults. So far, AFM-based, micro-mechanical measurements have only been performed on single cells and confluent cells, but not yet on cells in the physiologically relevant, mature epithelial layer. Using a combination of atomic force, fluorescence and confocal microscopy, we determined the changes in stiffness, morphology and actin distribution of human mammary epithelial cells (HMECs) as they transition from single cells to confluency to a mature epithelial layer. Single cells have a tall, round (planoconvex) morphology, have actin stress fibers at the base, have diffuse cortical actin, and have a stiffness of 1 kPa. Confluent cells become flatter, basal actin stress fibers start to disappear, and actin accumulates laterally where cells abut. Overall stiffness is still 1 kPa with two-fold higher stiffness in the abutting regions. Cells in an epithelial layer are flat on top and seven times stiffer (average, 7 kPa) than single and confluent cells. Epithelial layer cells show strong actin accumulation in the regions where cells adjoin and in the apical regions. Stiffness is significantly enhanced in the regions of adjoining cells, compared to the central regions of cells. Physiologically, this previously unrecognized, drastic stiffness increase may be important to the protective function of the epithelium.


Introduction 24
Importance of biomechanical properties of cells. Over the last two decades, microscopic biomechanical 25 analysis techniques, such as micropipette aspiration 1 , AFM indentation 2 , and magnetic and optical 26 tweezers 3, 4 , have progressed to the point where it is possible to study the mechanical properties of 27 individual cells with relative ease. These studies have shown that cell mechanical properties are 28 different for different cell types. The stiffnesses measured for isolated, cultured cells range 29 approximately from 0.1 kPa to 40 kPa 5 , and the stiffness of cells correlates with biological function and 30 the mechanical properties of tissue of origin. For example, neurons, constituents of the brain, which is 31 one of the softest tissues, are soft with cell stiffness ranging from 0.1-2 kPa 6 . On the other hand, cardiac 32 myocytes, which make up the cardiac muscle, are very stiff with elastic moduli in the 35-42 kPa range 7 . 33 These studies also indicated that mechanical properties of cells correlate with their microenvironment 34 and with cellular processes, including cell division 8 , adhesion 9 , migration 10 , motility 11 and 35 differentiation 12 . Moreover, several human diseases closely correlate with abnormal stiffening of cells, 36 e.g., asthma 13 , vascular disorders 14 and aging 7,15,16 ; or softening of cells, e.g., cancer [17][18][19][20] . Therefore, 37 investigating the mechanical properties of cells can provide valuable insights into various cellular 1 processes, disease and cancer progression, and they may be used as a potential biomarker. 2 Actin. Filamentous actin (F-actin) is a semi-flexible polymer, which is assembled from the monomeric, 3 globular form of actin (G-actin). Actin filaments are highly dynamic and responsible for many cellular 4 processes, including cytokinesis, changes in cell shape, cell motility, and intracellular trafficking. 5 Additionally, actin filaments play an important role as a mechanotransducer, translating external 6 physical forces into biochemical signals and leading to various cellular responses [21][22][23][24] . Actin filaments are 7 organized into two types of main structures, stress fibers and cortical actin filaments, which play 8 different mechanical roles within cells. Stress fibers, together with non-muscle myosin II, generate 9 mechanical forces during movement, and they are responsible for the formation and maintenance of 10 cell-to-cell or cell-to-extracellular matrix adhesions [25][26][27] . Cortical actin filaments form 50-200 nm thick 11 networks underlying the inner surface of the plasma membrane. 28 They endow the cell with its 12 mechanical integrity and contractility, which is important for cell shape and deformability. The 13 components of the cytoskeleton -actin filaments, microtubules and intermediate filaments, the plasma 14 membrane, the nucleus, and other cellular organelles are possible determinants of the mechanical 15 properties of cells. Numerous studies, however, have shown a close correlation between actin filaments 16 and cell stiffness, indicating that actin filaments are the primary contributor to cell mechanical 17 properties. This is likely due to their higher-order structures and interaction with crosslinkers and other 18 actin binding proteins. Cells treated with reagents that inhibit actin polymerization, such as cytochalasin 19 D and latrunculin B, showed significantly reduced cell stiffness [29][30][31][32][33] . Conversely, cells treated with 20 nocodazole, which interferes with microtubule polymerization, showed insignificant changes in cell 21 stiffness 29, [32][33][34] . Cancer cells are softer and easier to deform than their healthy counterparts 20 . It has 22 been proposed that the lower stiffness of cancer cells is attributed to a reduction in the amount of actin 23 filaments or an increase in disorganized actin structures [35][36][37] . 24 In vivo, cells continuously exchange signals with their neighboring cells, and they are linked together by 25 cell-to-cell junctions, which are responsible for regulating tissue homeostasis and integrity 38,39 . One of 26 the major cell-to-cell junctions is the adherens junction where cadherin receptors serve as adhesion 27 molecules, and actin filaments interact with them via the catenin protein complex, which form a bridge 28 between the cytoplasmic domains of cadherins and actin filaments. The cadherin-actin interactions also 29 play a role in the reorganization of actin filaments during development of adherens junctions. For 30 example, actin filaments associated with e-cadherins are perpendicularly oriented to the plasma 31 membrane at early stages of development of adherens junctions of epithelial cells. In contrast, actin 32 filaments align parallel to the cell borders in mature epithelial sheets. 40,41 33 Motivation. As outlined above, the relevance of the mechanical properties of cells to a deeper 34 understanding of key cellular processes was recognized over the last several years and a significant 35 amount of work has been done determining the mechanical properties of different cell types. Initially, 36 many studies concerning cell stiffness were performed on single, isolated cells, and the influence of 37 confluency, cell packing, and cell-to-cell contact on cell stiffness has been underappreciated 35,36 . 38 Recently it has been recognized that confluency and cell packing also need to be considered. Cell-cell 39 interactions via adherens junctions underlie long-range correlations in cell stiffness 42 . Human mammary 40 epithelial cells inside a colony (confluent layer) are stiffer than isolated cells 43 . The stiffness of MDCK II 41 cells depends on confluency and cell size 44 . In most of these studies confluent cells were slightly stiffer, 42 by a factor of two or less, than isolated cells 42-46 . 43 However, very few studies, if any, were performed on mature epithelial layers that had been grown for 1 several days, which is the physiologically most relevant structure of epithelial cells. 2 Current work. Using a combination of atomic force microscopy, fluorescence microscopy and confocal 3 microscopy, we investigated the stiffness, overall morphology and actin distribution of human 4 mammary epithelial cells (HMECs) as they transition from single cells to confluency to a mature 5 epithelial sheet conformation. We found that morphology, actin distribution and stiffness change 6 significantly during this transition. Single cells have a tall, round morphology, with actin stress fibers at 7 the base, diffuse cortical actin within the cell, and a stiffness of about 1 kPa. As cells become more 8 confluent, they become flatter, actin stress fibers at the base start to disappear, and actin accumulates 9 on the side where cells adjoin neighboring cells. experiments. The protocol to attach the microsphere is described next. 36 A clean glass coverslip was used as a substrate for the 5.3 µm microspheres and glue. 20 µl of 37 microspheres in 2.5% aqueous suspension were deposited on the glass coverslip. A paper wipe was used 38 to absorb and dry the water after the spheres sank to the coverslip surface, which occurred in a few 39 minutes. Most spheres formed clusters, but some isolated spheres existed on the coverslip; these 1 isolated spheres can be picked up by a cantilever. Using a pipette tip, a small amount (approximately 20 2 µl) of mixed, two-part marine epoxy glue was transferred onto the same coverslip. Using another 3 coverslip, the glue was spread to a thin layer close to where the microspheres were deposited. The glue 4 was dried for 20 minutes before moving on to the next step; drying for this period increased the 5 viscosity of the low viscosity glue to a viscosity level appropriate for the following dip-pen nano-6 lithography step. The AFM mounted on the optical microscope was used as a nanomanipulator for fine 7 positioning, and the micrometers for the movement of the sample stage and the AFM head were used 8 for coarse positioning. After putting the coverslip on the sample stage, the AFM cantilever was 9 positioned in the middle of the field of view of the 40x lens. The laser was focused on the end of the 10 cantilever, as is standard, and the deflection signal was monitored during the following processes to 11 avoid breaking the cantilever. The edge of the spread glue on the coverslip was moved under the end of 12 the cantilever. A small amount of the glue was applied to the end of the cantilever by lowering the AFM 13 head manually while observing the cantilever through the optical microscope. After applying the glue, 14 the cantilever was retracted to a safe distance above the surface (> 20 µm) and then an isolated 15 microsphere was positioned under the cantilever. Using the AFM software, which can control the 16 position of the piezo scanners with nm resolution, the cantilever was finely positioned above the 17 microsphere. The cantilever was lowered by the z-piezo scanner until it contacted the microsphere. A 18 few seconds after the contact, the cantilever was retracted and taken out of the AFM head. The 19 cantilever with the microsphere was cured overnight under ambient conditions. A representative bead-20 modified cantilever is shown in Fig. 1. 21 To measure the spring constant of an AFM probe, we used the 'GetReal' automated probe calibration 22 function in the Asylum Research software, which is based on the thermal noise method 47 and Sader's 23 method 48 . The software first determines the spring constant, using Sader's method with Q factor, the 24 resonance frequency from the thermal noise spectrum, and the dimensions of a cantilever. 25 Subsequently, GetReal determines a parameter, the so-called inverse optical lever sensitivity (InvOLS), 26 which is needed to convert the photo diode signal in volts to displacement of the cantilever in 27 nanometers, using the thermal noise method and the spring constant calculated in the previous step. 28 However, we used the InvOLS determined by a different, more direct way. Specifically, by measuring the 29 slope of the deflection voltage vs. z distance curve on a hard surface. There was usually a 10-20 % 30 discrepancy between this slope method as compared to the thermal noise method. Since the latter 31 slope method is more direct, we used that value. 32 The following equation, derived from the Hertz model, was used to determine the Young's modulus of 33 each cell. 34 Here F is the indentation force, R is the bead radius, E is the Young's modulus, is the Poisson's ratio of 36 a cell (taken to be 0.5), and δ is the indentation depth. Three to five indentations were carried out on 37 the center of each cell to obtain force versus indentation depth curves. Since the Young's modulus of 38 cells depends on the loading rate and indentation depth, the speed of the indentation was fixed at 5 39 µm/sec and the indentation depth was maintained in the range of 0.6 to 1 µm. The obtained force 40 curves were processed using the built-in Hertz model fitting tool of the Asylum AFM software. The 41 Young's modulus values of the curves were arithmetically averaged for each cell. A representative 1 indentation curve and fit, according to the Hertz model, is shown in Fig. 1

(c). 2
A stiffness map, also called a force-volume map, consisting of 32 × 32 pixels within an 80 × 80 µm 2 3 area, was acquired by taking individual force versus indentation curves on each pixel. Like the single 4 indentation, the indentation depth and loading speed for a stiffness map on each pixel were set in the 5 range of 0.6 to 1 µm and 5 µm/sec, respectively. It took approximately 18 minutes, which is short 6 enough to avoid drifts caused by cell movements, to obtain a stiffness map with these conditions. The 7 Young's modulus of each pixel was determined by fitting the Hertz model for the 5.3 µm microsphere to 8 the individual curves, using the analysis tools of the AFM software. 9

Fluorescence microscopy 10
For fluorescence imaging of F-actin and nuclear structures of live cells, F-actin filaments and the nuclei 11 were labeled with 1 µM of SiR-actin (Cytoskeleton, Denver, USA) and 1 µg/ml of Hoechst 33342 12 (Invitrogen, Carlsbad, USA), respectively. Each of the dyes, dissolved in DMSO, were directly added to 13 the culture medium, and the cells were incubated for 30 minutes prior to fluorescence imaging. 14 Epifluorescence images correlating to the AFM topographic images were taken by the X73 Olympus  15 inverted microscope, which is situated under the AFM, with a 60x oil immersion lens (numerical 16 aperture, NA = 1.3) and a Hamamatsu digital camera (C11440, ORCA-Flash2.8, Hamamastsu, 17 Hamamatsu, Japan). Filter sets for DAPI and Cy5 channels were used for Hoechst 33342 and SiR-actin, 18 respectively. 19

1
One of the key functions of epithelial cells is to form a protective barrier between the inside of the body 2 and the outside world. The overall goal of our work was to characterize the global morphology, stiffness 3 and actin distribution of human mammary epithelial cells as they progress from single cells, through 4 confluency, to a mature, epithelial layer. 5 Global morphology and actin formation of HMECs at different degrees of confluency 6 AFM and fluorescence images were taken to determine how cellular morphology and F-actin structure 7 change with increasing cell confluence (Fig. 2). AFM deflection images (error signal) indicate the 8 topography of single HMECs, confluent HMECs, and HMECs of epithelial layers (Fig. 2 a-c). Cell areas 9 tend to decrease as cells reach confluence and then form the basal layer of the epithelium. However, 10 additional cells that form the apical level (second layer) of the epithelium had the largest area compared 11 to single, confluent cells and basal epithelial layer cells. The cross-sectional profiles ( Fig. 2 b), 12 corresponding to the white lines in Fig. 2a, show the height differences between the lowest and the 13 highest topographical features. The measured height differences for the single cell, confluent cells and 14 the epithelial layer shown in Fig.2 are 8.5 µm, 2.5 µm and 300 nm, respectively. This indicates that cells 15 become flatter as they reach confluence and form the epithelium. We took epi-fluorescence microscope 16 images of actin filaments stained with SiR-actin of the cells corresponding to the AFM images (Fig 2. g-l). 17 Using a 60x oil immersion lens (NA = 1.3) and shallow depth of field, it was possible to distinguish actin 18 filaments in the apical focal plane (Fig. 2 g-i) and basal focal plane (Fig. 2 j-l). In the apical plane of the 19 single cell (Fig. 2g), the actin distribution is diffuse, with no apparent, distinct actin filaments. The image 20 of the apical focal plane of the confluent cells (Fig. 2h) shows thin cortical mesh actin filaments in the 21 middle of cells and a strong, blurred distribution of circumferential actin. Some of the circumferential 22 actin is in focus and some out of focus, indicating that it is vertically distributed along the entire height 23 of the cell. In the apical focal plane of the epithelial layer cells (Fig. 2I), we observed dense, distinct, 24 circumferential actin filaments and filaments that stretched out to span cell-to-cell contact regions. Note 25 that the apical plane has a smaller number of cells than the basal plane. Thick and dense dorsal stress 26 fibers were observed in the basal focal plane of the single cell (Fig. 2j), which indicates that the cell was 27 actively migrating toward the upper left of the image. The image in the basal focal plane of the 28 confluent cells (Fig. 2k) shows that thinner, sparse, and unidirectionally aligned stress fibers and actin 29 filaments surrounding the cells were formed, indicating a more stationary and collectively migrating cell. 30 Overall, in the basal focal plane of the cells in the epithelial layers (Fig. 2l), the amount of stress fibers is 31 markedly decreased, and the stress fibers are randomly oriented, indicating non-migrating, spatially 32 stable cells.

Young's modulus of individual cells measured with an AFM. 3
The first two histograms (Figs. 3a & b) show the distributions of Young's moduli measured in the central 4 region of individual cells in the single and confluent states, and the third histogram (Fig. 3c) is the 5 distribution of Young's moduli measured on the top layer of the epithelium. Note that the indentation 6 locations on the top layer were randomly chosen due to the ambiguity in determining the borderlines 7 between cells that is a result of the continuous and flat morphology as shown in Fig. 2c. Also, note that 8 the x-axis scales of Figs. 3a & b are the same, while the scale of Fig. 3c is much larger. The width of the 9 distribution of single cell moduli (Fig. 3a) is slightly wider than that of confluent cells (Fig. 3b), while the 10 averaged Young's modulus, E, of single cells and confluent cells are equal within error, with E = 1.01 ± 11 0.52 kPa (N = 183) and 1.09 ± 0.51 kPa (N = 199), respectively. According to a Student's t-test analysis, 12 the two distributions are not statistically different (p > 0.5). The width of the distribution of Young's 13 moduli of the epithelium (Fig. 3c) is much wider than the other two distributions, ranging from 0.37 kPa 14 to 37 kPa, and the averaged Young's modulus of the epithelium increased about seven-fold to 7.09 ± 15 5.8 kPa (N = 717). Compared to single and confluent cells, the increase in Young's modulus is statistically 16 significant (p < 0.001). 17 To determine if overall cell area and shape affect stiffness, we plotted the Young's modulus vs. aspect 18 ratio and Young's modulus vs. cell area (Fig. 3e & f). Note that the data used for these plots are sub-sets 19 of the histograms, but the data for the epithelium was not included in these plots because the modulus 20 magnitudes differ significantly (see discussion). The aspect ratios of the single cells ranged from 21 approximately 2 to 14, and the Young's modulus of single cells tends to increase with higher aspect 22 ratios. On the other hand, the distribution of the aspect ratios of the confluent cells was localized 23 around 2 with the Young's modulus ranged between 0.3 and 2 kPa. Single cell areas ranged from 250 24 µm 2 to 3800 µm 2 , and the Young's modulus tends to increase as cell area increases. The areas of the 25 confluent cells did not exceed 1800 µm 2 , and there was no correlation between cell area and stiffness. 26 Taken together, although the Young's moduli of both single and confluent cells ranged between 0.3 kPa 27 and 3 kPa and the two averaged values are very similar, single cells saw their stiffness (force/area) rise 28 as each cell stretched out over a larger region. When comparing cells smaller than 1700 µm 2 (see Fig.  29 5f), confluent cells were slightly stiffer than single cells in this small cell area regime.

Stiffness Maps of Cells 7
To investigate the correlation between actin filament formation and cell stiffness, fluorescence 8 microscope images of the stained actin filaments of cells were taken, followed by force mapping of the 9 same cells with the AFM (Fig. 4). It can be seen that single cells have the lowest stiffness, and there were 10 no distinct features in the stiffness map that correspond to actin formation. This is likely because there 11 are no actin filaments directly contributing to the stiffness except for cortical actin; actin is found in the

F-actin Distribution via Confocal Microscopy 2
To investigate how actin filament formation changes due to increased cell-cell interactions as cells reach 3 confluence, images of SiR-actin stained actin filaments were taken by confocal microscopy. Fig. 5 shows  4 actin filament formation in the apical (Fig. 5a-c), mid-height (Fig. 5d-f) and basal planes (Fig.5g-i) of 5 single and confluent cells, and the epithelial layer. In single cells, only the cortical actin mesh network, 6 which underlies the inner surface of the plasma membrane, is observed in both the apical (Fig. 5 a) and 7 mid-height (Fig. 5d) focal planes. In the basal plane of the single cell (Fig. 5g), extensive circumferential 8 stress fibers were observed. Unlike Fig. 2j, in which dorsal stress fibers were seen in a migrating cell, 9 there are no dorsal stress fibers seen here, indicating that this single cell was not migrating in a specific 10 direction. The cross-sectional image of the single cell (Fig. 5j) shows a typical morphology of an adherent 11 cell which has a planoconvex shape in the middle. In confluent cells, except for the central region of the 12 cells, dense actin filament bundles surrounding the cells were observed in the apical (Fig. 5b) and mid-13 height (Fig. 5e) focal planes. In the basal focal plane of the confluent cells (Fig. 5h), randomly oriented 14 stress fibers were observed, and stress fibers have mostly disappeared in some of the cells. In the cross-15 sectional images of confluent cells (Fig. 5k), cells remained roundish, but the actin filaments were 16 prominent, underlying the inner surface of the plasma membranes close to the cell-cell contact regions. 17 Cells in the apical level of the epithelial layer (Fig. 5c) were observed to have actin formation intricately 18 intertwined between the cells. In the mid-height plane (Fig. 5f), the actin filaments forming the 19 boundaries between cells were observed. In the basal plane (Fig. 5i) Their tasks include providing a protective  3 biological, chemical and physical barrier; being able to respond to physical forces and disturbances; and 4 being able to migrate and rearrange in response to external stimuli and insults. The mechanical 5 properties of epithelial cells are, therefore, critical to their function. Moreover, they are also of interest 6 to cancer researchers as many cancers start in the epithelium. 7 Human mammary epithelial cells (HMEC) form a stratified cuboidal epithelium, which differs from a 8 simple epithelium in that it is multilayered. It is typically found in gland linings that are specialized in 9 selective absorption and secretion, that need to withstand mechanical or chemical insults, and where 10 cells might be abraded. Cell strata in these epithelia become flatter as the strata become more apical 49 . 11 Numerous studies have examined the mechanical properties of both single and confluent epithelial cells. 12 However, few, if any, studies have examined the mechanical properties of cells in a mature epithelial 13 layer, and how mechanical properties change as cells transition from single cells to an epithelial layer. 14 The goal of our work was to determine the changes in gross morphology, cell mechanical properties 15 (stiffness) and actin distribution in human mammary epithelial cells as they transition from single cells to 16 confluent cells to an epithelial layer configuration. We used the following methods to characterize the 17 cells: AFM imaging, fluorescence imaging, AFM-based nanoindentation, AFM force mapping with a 5.3 18 µm micro-bead probe, and confocal microscopy. Our key findings were: other cell types, single epithelial cells are soft; they are slightly stiffer than the very soft neurons 6 (0.4 4 kPa), but significantly softer than cardiomyocytes 7 (35 kPa) and osteoblasts 51 (8.3 kPa). The tall, roundish 5 morphology we observed also agrees with the shape of single epithelial cells reported in the 6 literature. 35,36 Pronounced actin stress fibers at the basal plane indicate migrating cells, while the diffuse 7 cortical actin distribution is indicative of a soft apical region. 8 As cells become confluent, but not yet tightly packed and layered, cells are sensing each other, and they 9 are forming adherens junctions 40,52 . In epithelial cells, the formation of adherens junctions triggers a 10 decrease in RhoA activity and an increase in Rac1 and Cdc42 activity; it gives rise to slowed migration, a 11 decrease in stress fibers, the formation of focal adhesions, and actin accumulation in the regions where 12 cells abut 53,54 . Our observations agree with these changes, as shown in Fig. 5e & h. Concurrently, the 13 stiffness in the peripheral cellular regions increases, while the stiffness in the center of the cell remains 14 the same. The modulus we observed for confluent cells largely agrees with values reported in the 15 literature for confluent epithelial cells 43-45 . As epithelial cells become confluent, it was observed that the 16 modulus stays the same or slightly increases (by up to a factor of 2). The slight discrepancy between 17 either staying the same or slightly increasing may be explained by the following factors. Different 18 epithelial cell types or measuring conditions were used across the different experiments. Additionally, as 19 can be seen in Fig. 4  Young's modulus, as calculated from E = 2(1+ν)G' is 4.5 kPa for MCF-10A cells (immortalized, benign 40 cells similar to our HMECs) for indentation speeds and indentation depth similar to those in our 1 measurements. In a series of publications, the Janshoff lab investigated the viscoelastic properties of 2 Madin-Darby canine kidney cells, strain II (MDCK II) using two dynamic AFM-based indentation methods, 3 oscillatory microrheology (OMR, tip performs small, 40 nm oscillations during indentation) and force 4 cycle experiments (FCE, cyclic indentation curves). 44,57-59 Their modulus values for confluent cells agrees 5 with our values (~1 kPa). 6 Notably, the mechanical properties of cells in stratified epithelial layers, which is the natural 7 physiological state of mammary epithelial cells, has not been investigated in the past. We found that 8 cells undergo a significant transformation as they arrange in a mature, epithelial layer. Cells become, on 9 average, stiffer by a factor of 7. Stiffness is not uniform, as it is significantly higher in the regions where 10 cells abut. Moreover, actin is seen in a strong, distinct belt in this region (especially visible in Fig. 5 f). 11 There are, furthermore, flat, actin-rich cells on top of the lower cells, which is expected for HMECs as 12 they form a stratified epithelial layer. This actin distribution and stiffening strongly suggest that thick 13 actin fibers are a major determinant of cell stiffness (see Actin section below). It also suggests that this is 14 physiologically important as the epithelial layer forms a barrier against external insults. There may be additional error sources; however, all of them are smaller than the factor of 7 increase in 39 stiffness we observe for epithelial layer cells. Our indentations were small enough (600 nm to 1000 nm) 40 to not 'feel' the nucleus or the hard glass substrate underneath the cell. In some instances when taking 1 measurements over the thin, peripheral section of single cells, the shape of the indentation curve 2 actually did deviate significantly from the Hertz model, resulting in very high stiffness values. In these 3 instances, the substrate stiffness was measured, and these data were eliminated. Typically, the probe-4 sample contact point in the force-indentation curves was well defined (Fig. 1b), indicating that these 5 cells have a thin glycocalyx. Our experiments were all carried out on a hard, functionalized substrate 6 (collagen-and laminin-treated glass) in temperature-controlled (36.5 o C) and pH-controlled media. 7 Recently, it was shown that cell stiffness is somewhat influenced by the stiffness of the substrate, as 8 cells grown on stiff substrates cells are about 1.5 to 3 times stiffer than cells on soft substrates. 61,62 9 However, the stiffening factor due to the transition from single/confluent cells to a mature epithelial 10 layer is 7 and, thus, much larger than the substrate-dependent factor of 1.5. 11 Role of Actin. A central theme of our work concerning cell mechanical and morphological properties is 12 the role of reorganized actin filaments. Determining the underlying elements that influence cell 13 mechanical properties is currently a highly active research area; though, actin filaments are emerging as 14 a major factor. Our data, and literature reports 43,60 , suggest that cells possess a baseline stiffness of 15 around 500 -1000 Pa that may originate from the cytoplasm, the nucleus, the membrane, and 16 cytoskeletal filaments, or a combination of these factors. Our data, and data in the literature, also 17 suggest that enhanced stiffness -on top of the baseline stiffness -correlates with increased actin 18 density. Fig. 4 shows that enhanced stiffness is measured in the region where cells abut, and the 19 enhanced stiffness strongly co-localizes with dense actin filament bundles in confluent cells and in the 20 epithelial layer cells. This agrees with the observations by Schierbaum et al. 55 . Enhanced stiffness is also 21 seen in epithelial layer cells, in general, and it again strongly co-localizes with dense actin filament 22 bundles. Several studies found that actin filaments play a major role in determining cell stiffness. 23 Notably, treatment of cells with reagents, such as cytochalasin and latrunculin, which promote actin 24 depolymerization, resulted in cells with significantly decreased stiffness. 29 Additionally, diffuse cortical F-actin, as seen in the confocal images of single cell (Fig. 5a & d), does not 31 appear to correlate with the enhanced stiffness, since the cortical F-actin does not vary noticeably with 32 stiffness. 33 between the traction force, cortical membrane tension and the cytoplasmic pressure is required to 38 maintain cell shape. This implies that larger cells maintain a stronger tensional prestress. Since prestress 39 strongly affects cell stiffness, 65-69 it is expected that cells with a large spreading area are stiffer than cells 40 with a small spreading area. This is consistent with our data showing that stiffness increases with cell 41 area for single cells (Fig. 2f). Similar correlations were found in endothelial cells 9 . Cell area of single cells 1 also correlated with their aspect ratio (Fig. S2), which is possibly due to a correlation between aspect 2 ratio and cells stiffness (Fig. 2e). In contrast, in confluent cells, there was no correlation between those 3 geometrical factors and stiffness. (Fig. 2f & g). This is probably because in confluent cells tensional 4 prestress originates from actomyosin filaments linked to adherens junctions, and not from 5 counterbalancing forces against traction forces, which depend on cell area. This view is supported by the 6 observation that basal stress fibers and focal adhesions decrease as cells become confluent. 7 Even though we did not quantitatively measure the spreading area and height of the cells on the top of 8 the epithelial layer, these cells appeared to be 3-4 times wider and thinner than the basal cells (Fig. 1, 9 Fig. 5, Fig. S1). This extremely stretched and flattened morphology possibly indicates that much stronger 10 tension was loaded over the cells on the top layer, making the cells recruit a large amount of actomyosin 11 filaments at the adherens junctions to provide tension and maintain the cell shape, which, in turn, 12 caused the dramatically enhanced stiffness. 13 Conclusions. In summary, we found that HMECs possess significantly different mechanical properties 14 along with different F-actin distributions as they transition from single to confluent to mature epithelial 15 layer cells. The observed significant stiffening of cells in an epithelial layer is likely physiologically 16 important, providing protection against external insults. There are likely additional intermediate 17 transitioning states between confluent cells and mature epithelial cells, in which the stiffnesses 18 progressively increase. Future work will focus on investigating the detailed mechanisms by which the 19 cells on the top of the mature epithelial layer become flat and stiff. Our findings advance the 20 understanding of breast ductal morphogenesis and mechanical homeostasis. 21 Acknowledgements. This work was supported by an instrumentation grant for the AFM from the North 22 Carolina Biotechnology Center (NCBC; 2014-IDG-1012), and grants from the Discover Institute and the 23 NIH (1R15HL148842). We are grateful to Dr. Adam Hall, who is the PI of the NCBC instrumentation grant; 24 to Heather Brown-Harding, who is the head of the Wake Downtown confocal microscope facility at 25 Wake Forest University; to Pierre-Alexandre Vidi for helpful discussions; and to Amanda Smelser for help 26 with cell culture work.

25.
Ridley AJ, Hall A. The small GTP-binding protein rho regulates the assembly of focal adhesions 22 and actin stress fibers in response to growth factors.