Matching patterns of gene expression to mechanical stiffness at cell resolution through quantitative tandem epifluorescence and nanoindentation

Cell differentiation has been associated with changes in mechanical stiffness in single cell-systems, yet it is unknown whether this association remains true in a multicellular context, particularly in developing tissues. In order to address such questions, we have developed a methodology, termed quantitative tandem epifluorescence and nanoindentation (qTEN), wherein we sequentially determine cellular genetic identity with confocal microscopy, and mechanical properties with atomic force microscopy. We have applied this approach to examine cellular stiffness at the shoot apices of plants carrying a fluorescent reporter for the CLAVATA3 ( CLV3 ) gene, which encodes a secreted glycopeptide involved in the regulation of the centrally-located stem cell zone in inflorescence and floral meristems. We found that these CLV3- expressing cells are characterized by an enhanced stiffness. Additionally, by tracking cells in young flowers before and after the onset of GFP expression, we observe that an increase in stiffness coincides with this onset. This work illustrates how qTEN can reveal the spatial and temporal dynamics of both gene expression and cell mechanics at the shoot apex, and by extension in the epidermis of any thick tissue. the same flower, with white frames delimiting the individual stiffness maps. G-H, Pixel-level quantification of the individual maps, summarizing the Ea distributions of anticlinal pixels from GFP-, GFP+, and 0h mother cells of 24h GFP+ cells. The numbers of analyzed pixels and median Ea ratios are indicated in the box plots, which are constructed as in previous figures. The scale bars indicate 50 µm in A-B and 20 µm in C-F.

Morphogenesis is a complex process that results from the coordinated actions of many genes and gene products across developing tissues and organs. Because shape is a function of the structural elements of cells, the molecular and genetic control of growth and morphogenesis must rely on the regulation of the mechanics of these elements. In this context, cell differentiation has been linked with mechanical stiffness in animal single-cell systems (Collinsworth et al., 2002;Balland et al., 2006;Engler et al., 2006;Darling et al.;2008), though the direct measurement of cell mechanics in growing animal tissues remains elusive (Davidson, 2011;Blanchard and Adams, 2011).
However, these measurements have not been directly associated with cell identity. This association would become feasible if mechanical measurements were combined with optical imaging of fluorescent reporters. Such a combination has already been developed for single animal cells and for thin plant tissues -nanoindentation coupled to inverted optical microscopy (e.g. Rotsch et al., 2000;Routier-Kierzkowska and Smith, 2014) -but it cannot be extended to thick tissues because they are opaque, making it impossible to simultaneously observe the tissue surface optically with an inverted microscope and probe it mechanically. To circumvent this difficulty, we have developed a methodology involving the use of three microscopes to image the same sample: (i) an atomic force microscope (AFM), which is a nanoindentation system for obtaining stiffness maps of the surface of a sample; (ii) an AFM-coupled upright epifluorescence macroscope to precisely identify the points to be probed; and (iii) a confocal microscope to determine cell fate at cellular resolution, which may in turn be correlated with the stiffness maps.
We call this methodology qTEN (quantitative tandem epifluorescence and nanoindentation) and 6 we use it to probe the shoot apical meristem (SAM) of Arabidopsis thaliana, which is a good model system to investigate morphogenesis.
The SAM is located at the growing tip of the shoot and consists of distinct functional zones (Ha et al., 2010). One of these zones is the slow-dividing central zone (CZ), which can be defined by the expression of the CLAVATA3 (CLV3) signaling glycopeptide. Through cell division, cells exit the central zone into the surrounding peripheral zone (PZ). In the PZ, cells proliferate rapidly, and some become incorporated into organ primordia, thus yielding all aerial organs of the plant. Recent work on the SAM has revealed patterns of mechanical properties (Milani et al., 2011;Peaucelle et al., 2011;Kierzkowski et al., 2012;Braybrook and Peaucelle, 2013), but it is still unclear how these patterns are related to the activity of the SAM or to its functional zonation. Here, we analyze the dynamics of such a mechanical pattern in vivo, and show that it is spatially and temporally related to stem cell fate.

Measurement of mechanical properties in a cellular domain using an atomic force microscope coupled to an epifluorescence-macroscope
In order to simultaneously observe cell fate and measure stiffness at cellular resolution, we imaged living SAMs of plants carrying a pCLV3::GFPer reporter construct (Reddy and Meyerowitz, 2005) with both fluorescence and atomic force microscopy. We first confirmed that the observed temporal patterns of GFP expression match CLV3 RNA expression using fluorescent whole-mount in situ hybridization (Supplemental Fig. S1A, B). We designed a customized coupling between an AFM and an upright epifluorescence macroscope in order to position both the mechanical probe and the lens specifically over the GFP domain of shoot apices ( Fig. 1A-C). We used a sharp pyramidal tip and a maximal force of ca. 10 nN, as previously described (Milani et al., 2011;Sampathkumar et al., 2014), to very locally measure the apparent elastic modulus of the cell wall at a set of user-defined locations, e.g. regularly spaced along a line (Fig. 1D). We found that the expression of GFP correlated with a greater apparent elastic modulus (Fig. 1E), consistent with our earlier findings that the shoot summit is stiffer (Milani et al., 2011). Similar results were obtained on floral meristems of stage 3 flowers, 7 which also express CLV3 (Supplemental Fig. S1C,D). However we were unable to associate mechanical properties with identity at a cellular resolution, because the relatively small indentation depths (50 to 100 nm) used in this protocol do not reveal cell outlines, and because the macroscope does not provide cellular expression data. We therefore sought another approach, such that cell fate could be ascertained in individual cells and related to their stiffness maps.

qTEN enables the identification of genetic and mechanical properties at cell resolution
In order to further investigate the link between gene expression and stiffness patterns at cell resolution, we switched to another protocol coupling confocal and mechanical imaging, which we termed qTEN (quantitative tandem epifluorescence and nanoindentation). We employed larger indentations (200-500 nm) that allowed cell geometry delineation, using a cantilever with a spherical tip (diameter of 800 nm) and a force of ca. 1.0 μN. The probe was oscillated at a low frequency while horizontally scanning the sample, and a force curve generated each time the AFM probe made contact with the sample. The apparent elastic modulus (Ea) was then extracted from each force curve using the DMT model (  (Milani et al., 2011;Peaucelle et al., 2011;Kierzkowski et al., 2012;Braybrook and Peaucelle, 2013). Additionally, we also observed that the anticlinal walls corresponding to recent cell divisions often appeared softer than other anticlinal walls.

Data analysis demonstrates a correlation between gene expression and cell mechanics
To go beyond these qualitative assessments, we developed a pipeline to quantitatively analyze the data collected (Materials and Methods). Briefly, we generated cell outlines from the stiffness www.plantphysiol.org on August 20, 2017 -Published by Downloaded from Copyright © 2014 American Society of Plant Biologists. All rights reserved. 8 maps and then used the confocal data to determine which cells expressed GFP. Every pixel in the AFM map was then annotated for two criteria: wall type (anticlinal or periclinal) and, when applicable, cell fate (presence or absence of GFP). Quantifications of these data were then carried out and an estimate of the relative stiffness of the GFP+ and GFP-groups calculated as the ratio of their median apparent elastic modulus (Ea) values. The Wilcoxon test was used to assess the statistical significance of observed difference. In this way, we analyzed one of the individual stiffness maps from the SAM. We found that the GFP+ group is stiffer than the GFPgroup, both for anticlinal and periclinal walls (e.g. Fig. 2E-G; n = 15,269 pixels from about 35 cells), with relative stiffness values of 128% and 109%, respectively (p<0.001). Overall, these results show that genetic and mechanical identities in the Arabidopsis SAM can be accurately measured at cellular resolution.

In flowers, enhanced stiffness is correlated with GFP expression
In order to test whether stiffness is temporally correlated with GFP expression, we turned to floral meristems, where GFP appears transiently during flower development (Fletcher et al., 1999;Brand et al., 2000). Whereas no reporter expression is observed in young (stage 1) flower buds that have just initiated on the flanks of the SAM (Fig. 2D), GFP becomes visible in a small group of cells from mid stage 2 onwards, once the flower has separated from the SAM.
Expression persists until stage 6, when this region is incorporated into the female reproductive organs, the carpels.
We next asked whether GFP-positive cells in the flower also display the higher stiffness seen in SAM GFP-expressing cells (Fig. 3A). We first considered flowers at stage 3, which is characterized by the emergence of sepals. As in the SAM, we observed zonal differences of mechanical properties, which correlate with the presence or absence of GFP ( Fig. 3A CLV3 expression do not have such a stiffness pattern (n = 4). Thus in the floral meristem, like in the SAM, CZ fate correlates with greater stiffness.

The stiffness pattern is temporally correlated with gene expression
The absence of a stiffer region in stage 1 flowers suggests that the onset of CZ identity is temporally coincident with an increase in stiffness. To test this more directly, we performed  4F). The relatively short time interval also allowed us to trace cell lineages between the two time points. We then carried out a pixel-level quantification of the stiffness maps, additionally using cell lineages to identify the mother cells (at 0 h) of the cells that express GFP at 24 h. We find that at 24 hours, the GFP+ cells have a relative stiffness of 110 -127% (n = 8005 pixels, p<0.001), whereas at 0 h, their mother cells are less stiff than their neighbors (relative stiffness = 88%, n = 8975 pixels, p<0.001). Thus the pattern of stiffness is established concomitantly with CZ fate, as indicated by CLV3 expression.

DISCUSSION
We have developed a methodology to measure cell mechanics in specific domains and at cellular resolution. In a first step, we use confocal microscopy to obtain high-resolution images of a fluorescent marker in a whole tissue and to identify an appropriate region of interest. In a second step, we use an upright macroscope with epifluorescence to position the AFM cantilever over the desired region, where we then measure tissue mechanics using a cantilever with a micrometric spherical tip, which allows us to detect cell contours. The order of these first two steps was unimportant in the experiments we performed, although we cannot exclude that laser exposition or nanoindentation may have an impact either on the fluorescent level of other markers or on stiffness in other tissues respectively. In a third step, we associate optical and mechanical images based on cell contours, in order to match cell identity and mechanics. Finally, we quantitate the results to identify any significant cellular trends in the measurements.
We applied the qTEN methodology to the central zone of the shoot apical meristem and floral meristems, using pCLV3::GFP as a reporter of cell fate. In principle, qTEN can be applied to other reporters, such as pAHP6:: GFP (Besnard et al., 2014) which is expressed in incipient floral and sepal primordia (Supplemental Fig. S5 A-B). More generally, qTEN could be used to address questions concerning epidermal cells, in their organ context. For instance, we could associate optical images and AFM stiffness maps of Arabidopsis pavement cells in whole cotyledons to identify specific cells (Supplemental Fig. S5 C-D).
Using the qTEN methodology, we demonstrate that CLV3-expressing cells exhibit greater stiffness than their neighbors in the Arabidopsis shoot apical meristem. These results are consistent with previous studies showing that stiffer regions were located roughly at the shoot summit (Milani et al., 2011;Peaucelle et al., 2011;Kierzkowski et al., 2012;Braybrook and Peaucelle, 2013). We now show that this enhanced stiffness is correlated with CLV3 expression.
Alternatively, enhanced stiffness may also serve as an indicator of cell fate at the shoot apex.
Formally, there are alternative scenarios that might account for the apparent greater stiffness of the CZ in SAMs and in flowers. Our results exclude tissue shape as an explanation for this phenomenon, because young WT flowers do not exhibit tissular stiffness patterns, and that we found no correlations between apparent elastic moduli and geometric features such as height, slope or curvature (Supplemental Fig. S6). Other explanations for the phenomenon could be either a strain-stiffened CZ (Kierzkowski et al., 2012), or an enhanced turgor pressure in the CZ.
While we do not rule out a partial role for these two factors, it is clear that cell walls contribute significantly to the tissular stiffness pattern that we detect. Indeed, we also measured wall stiffness very locally using a sharp tip, and found the CZ to be stiffer than the PZ (Fig. 1E and Supplemental Fig. S1).
Though the biological mechanisms underlying this close correlation remain to be established, the functional importance of an enhanced stiffness could be multifold. Stiffer cell walls could reduce growth and proliferation rates, a known characteristic of the central zone of the wild-type SAM   Laufs et al., 1998;Reddy et al., 2004). This may also provide CZ cells with protection from fluctuating differentiating signals, as the slow growth dynamics of cells in the CZ may make them react only to persistent signals.
We observe a differential stiffness between the CZ and the PZ in all the flowers that display CLV3 expression. Similarly, cells in the boundary regions between the floral dome and the sepals also display an enhanced stiffness (Fig. 3D, map 3), which we would predict to be correlated with low proliferation rates, and with slow growth, as well as with the expression of boundary genes (Breuil-Broyer et al., 2004). These results hint to a framework in which patterning genes would control growth and morphogenesis through cell wall stiffness. Indeed stiffness has been observed to be inversely correlated with growth rates, suggesting that cell wall extensibility is also correlated with stiffness (Milani et al., 2013). However, we observed some variability in absolute stiffness, while the overall trend observed in the quantified floral data is that the average stiffness gradually increases as the flower develops (Fig. 5). This suggests that stiffness ratios across the tissue might be more important than absolute stiffness values for proper morphogenesis to occur. Indeed, the shape of a growing organism only depends on the ratios of growth rates: a proportional modulation of all growth rates would only slow down or accelerate morphogenesis, while an outgrowth requires a locally higher growth rate.
More generally, our methodology provides a fresh perspective in the study of the link between cellular identities and tissular growth, which is fundamental to all morphogenetic processes, and which is beginning to be examined in detail by several groups (Green et al., 2010;Schiessl et al., 2012

Plant growth and microscopy
pCLV3::GFPer and wild type plants (L-er ecotype) were grown on soil under continuous light.
pAHP6::GFP plant (Col-0 ecotype) were grown in short day conditions (8 hours light / 16 hours dark) for 4-5 weeks and were then transferred to long day conditions (16 hours light / 8 hours dark). Immediately after bolting, and usually before any flowers were open, the stems were cut and placed in dissection boxes containing growth medium as described (Fernandez et al., 2010).
Flower buds were removed from the inflorescence meristem until the desired stages were readily accessible for either atomic force or confocal microscopy (for local stiffness measurements and tissue/cell shape determination, respectively). Cotyledons were obtained as previously described (Sampathkumar et al., 2014). In most cases, confocal images were obtained before measurements on the AFM were taken. For the optical imaging of cell outlines, samples were stained with FM4-64 (approx. 1 μ l of a 330 μ g ml −1 stock) for about 10 minutes and then observed using either a Leica macroscope (MacroFluo TM ), or with a Zeiss LSM 510 or a Zeiss LSM 700 system with a water-dipping 40x objective.

Fluorescent whole-mount RNA in situ hybridization
Whole-mount in situ hybridization was performed as previously described (Besnard et al., 2014), except that the revelation step was carried out with fluorescent signals using Tyramide Signal Amplification (TSA) technology (Invitrogen). In this step, the samples were placed in the following solution for 30 minutes: 3 uL Alexa in 100 uL of amplification buffer containing a 1/20,000 dilution of H 2 O 2 . These were washed 3 times in 1x PBS for 5 minute each, and then incubated in 0.1% fluorescent brightener (Sigma) for 1 hour. The samples were then mounted and imaged on an upright confocal microscope (Leica SP5) with a 40x waterimmersion objective.

13
AFM indentation experiments were carried out with a Catalyst Bioscope (Bruker Nano, Inc.) mounted under an optical epifluorescence macroscope (MacroFluo TM , Leica), enabling the use of a 2x plano objective to observe the apex (Leica, Figure 1A). To record surface topology and stiffness modulus maps, PeakForce QNM TM AFM mode (Bruker Nano/Veeco, Inc.) was used (Nanoscope V controller and Nanoscope software versions 8.1). All quantitative measurements were performed using special 0.8 µm diameter spherical probes (SD-Sphere-NCH, Nanosensors TM ). The spring constant of cantilevers was measured using the thermal tuning method (Levy and Maaloum, 2009;Hutter and Bechhoefer, 1993) and ranged from 35-45 N/m. The deflection sensitivity of cantilevers was calibrated against a clean sapphire wafer. We also used cantilevers with pyramidal tips for more local measurements, as previously described 8 . All measurements were made under water at room temperature and the standard cantilever holder for operating in liquid environments was used. The 30 mm petri-dishes containing the sample (5 mm of stem, stabilised in agarose in a petridish) was placed on an XY stage in a specially designed sample holder. Then the AFM head was mounted onto the stage and positioned so that the cantilever was in the vicinity of the GFP-expressing region.
Mechanical property mapping with PeakForce QNM hinges on the ability of the system to acquire and analyze the individual force curves from every tap of the cantilever and the attached probe on the sample that occurs during the imaging process. The requested applied force (PeakForce setpoint) during imaging was 1 µN, which corresponded to indentation depths generally in the range 200-500nm above anticlinal walls. The topography and stiffness images were usually collected over rectangular areas of side 30 µm to 50 µm and at a digital resolution of 128 x 128 pixels (see details in Supplemental Fig. S4). A scanning rate of 0.3 Hz per line was used, corresponding to 38Hz per pixel; the approach and retract velocity were equal, and ranged 30-80µm/s. Ocasionnally, due to the high scanning speed and to instabilities in the control loops of the AFM system, the force applied by the system on the sample differed significantly from the requested one. The difference between the PeakForce setpoint and the real applied force is indicated in a "PeakForce Error" map (Ferr), which we used to assess the quality of mechanical measurements. Control experiments where different forces were applied to our samples have shown that the stiffness maps obtained are very similar for a force range of 0.5 -1.5μN. Based on this, pixels (measures) for which the applied force was less than 0.5 µN or greater than 1.5 μ N (alternatively the absolute value of Ferr was greater than 0.5µm) were automatically discarded from subsequent quantification. For the extraction of elastic moduli from the force-indentation 14 retract curves, we used the Derjaguin-Muller-Toporov (DMT) model (Eq. 1), which applies to the (possibly) adhesive contact between two elastic, thick isotropic solids (Derjaguin et al., 1975) where F is the force on the cantilever relative, F adh is the adhesion force between the two solids, R is the known tip end radius (0.4µm), and d -d 0 is the deformation undergone by the sample, relative to the contact height d 0 . The result of the fit yields the adhesion force, the contact height and the reduced modulus E*. If the Poisson's ratio, ν, is known, the Young's Modulus of the sample (E) is related to the sample modulus by E*=E/(1-ν 2 ) when the tip modulus is considered to be much bigger than that of the sample. As generally assumed for biological samples, we considered our sample as perfectly incompressible taking the value 0.5 for the Poisson's ratio.
We sought to minimize the potential influence of turgor on moduli by focusing on anticlinal walls, reasoning that the greater depth of wall material below the probe increased the applicability of the DMT model. However, because of the potential anisotropy of cell walls, we report in this work only an "apparent modulus" (Ea).

Data analysis
Stiffness maps provided by the AFM software were converted into two-dimensional matrices. In order to determine cell contours, maps were segmented manually using the signal corresponding to anticlinal walls, assuming 3 pixels widths for the anticlinal walls. The periclinal part of the image was simply the complementary of all pixels associated with anticlinal walls. Areas surrounded by pixels corresponding to anticlinal walls were defined as cells and given a label.
Cells present in multiple stiffness maps of the same tissue were labeled identically in segmentations. Maps were filtered solely to remove regions where (a) the whole cell could not be enclosed by the segmentation, and (b) the AFM indicated an absolute peak force error higher than 0.5 µN. These criteria led to the exclusion of between 3% and 28% of pixels, depending on the map.
Cell shapes in the L1 epidermal cell layer were extracted from the confocal optical stacks using the Merryproj software (Barbier de Reuille et al., 2005), which creates a 2D projection of the tissue surface. Cells from the optical images and from the stiffness maps were associated manually based on their shape. Cells were subsequently defined as CLV3+ or CLV3-using www.plantphysiol.org on August 20, 2017 -Published by Downloaded from Copyright © 2014 American Society of Plant Biologists. All rights reserved. pCLV3::GFPer expression. In series obtained from time-lapse imaging, cell lineages were defined manually based on cell shapes. This allowed us to label mother cells in young stage-2 flowers according to whether their daughters did or did not express CLV3+or CLV3-.
Every pixel in a stiffness map corresponds to one data point. It is labeled with the following information: sample number, map number, anticlinal or periclinal, cells with which it is associated (1 cell if periclinal, 2 or 3 cells if anticlinal), CLV3+ or CLV3-status of associated cells, CLV3+or CLV3-status of the daughters of associated cells (in time series). We defined the status of the pixel as CLV3+ if one of the cells to which it is associated was CLV3+, otherwise it was labelled as CLV3-. Datasets were extracted based on the labels. Except for the SAM in Figure 1, we have only presented data corresponding to anticlinal walls, because our experimental setup is better suited to examining such walls (Peaucelle et al., 2011;Routier-Kwierzkowska et al., 2012).
The quantitative representations of datasets use a box-and-whisker plot derived from the boxplot function from the R software (R Development Core Team, 2008). The box limits are the first and third quartile of the values, the median line represents the median value. In this modified version, whiskers extend from the first 10% of values to 90%. We assessed the statistical relevance of the difference between modules measured on CLV+ and CLV-anticlinal walls. We implicitly considered that each pixel corresponds to an independent measurement. As a Shapiro-Wilk test showed that data are not distributed according to normal distributions, we decided to resort to non parametrical tests. We used a Wilcoxon test on each stiffness map and obtained the corresponding p-values. Both the Shapiro-Wilk and the Wilcoxon tests were performed using the standard statistical package from the R distribution (R Development Core Team, 2008). All quantifications are shown in Figure 5 and the corresponding flowers appear in Supplemental Figure S4.
In order to test whether sample geometry had an influence on apparent elastic moduli, we determined slopes and mean curvature from the height maps, which were computed respectively from the norm of the gradient of the height and from the Laplacian of the height (the Laplacian is a good approximation of the mean curvature when slopes are small as here). For each stiffness map, we performed linear regressions between the apparent elastic modulus and the geometric variables: height, slope (Supplemental Fig. S6A) or mean curvature (Supplemental Fig. S6B).
We computed the corresponding coefficients of determination, R 2 (Supplemental Fig. S6C) correlation coefficients (Supplemental Fig. S6D). We concluded that the apparent elastic modulus is not correlated with sample geometry.

Routier-Kierzkowska AL, Smith RS (2014) Mechanical Measurements on Living Plant Cells
by Micro-indentation with Cellular Force Microscopy.   the whiskers from 10% to 90% of all the data set. Pixels were separated into two groups, GFP+ and GFP-. The number of pixels analyzed in each group is indicated. The ratio of Median Ea values of GFP+ and GFP-is shown. The scale bars indicate 20 µm in C and D and 5 µm in E and F.  25 Figure 5. Distribution of apparent elastic moduli for all the flowers that were analyzed quantitatively. Box plot representation of the Ea distribution for all maps and groups (GFP+ and GFP-) of each flower. The boxes extend from the first quartile to the third quartile and the whiskers from 10% to 90% of each data set; the number of analyzed pixels is indicated. The ratio between Median apparent elastic modulus of GFP+ and GFP-and the corresponding p-values (Wilcoxon test, see Materials and Methods) are shown where appropriate; the Ea ratio of mother cells of GFP+ cells to mother cells of GFP-cells is also specified for flower II -4 (Fig. 4G).