## Introduction

Nature is a perpetual source of inspiration for biomimetic and actuating devices1. Current biomimetic research involves a multi-scale approach to investigate how structural and mechanical properties at various length-scales determine organism function2. As more hierarchical models are studied, and their micro- and nanoscale properties better understood, more complex biomimetic and actuating devices can be designed2,3. Of the many organisms studied, plants are particularly interesting because of the wide range of functions and structures produced from a limited set of constituent materials, i.e. cell wall components4,5. Juxtapositions of different tissues and/or cell wall structures and compositions can drive organ movements in response to water or humidity. These include both rapid, ‘active’ actuation in response to changes in cellular turgor pressure in living tissues, or slower, ‘passive’, hygroscopic movements resulting from the shrinking and swelling of the cell walls themselves in living or dead organs6,7,8.

## Materials and Methods

### Plant materials

Mature S. lepidophylla plants were purchased from Canadian Air Plants (New Brunswick, Canada), and maintained in a desiccated state at 25 °C and ~30–50% relative humidity until use. Prior to experimentation, S. lepidophylla plants were placed in a plate of water and allowed to rehydrate for three consecutive days to achieve 100% relative water content.

### Time-lapse video capture

Video capture was adapted from the protocol in Rafsanjani et al.10. Time-lapse videos were captured using a Logitech C920 HD Pro Webcam (1080p, Carl Zeiss optics) and Video Velocity Time-Lapse Studio software (Candylabs). S. lepidophylla plants were allowed to rehydrate for 24 hours. Individual inner stems were cut from the plant at the root-stem interface and secured in a metal clamp, which was affixed to the base of a square Petri dish. Stems were then allowed to air-dry for approximately 6 hours, during which time changes in their curvature were captured via time-lapse filming at frame rate of 1 min 21 s. Stems were taken from four different S. lepidophylla plants. In total, four inner stems were tested and displayed similar curling/uncurling patterns.

### Histochemical and immunological analysis of cell wall composition

A total of ten inner stems from five individual plants were isolated and cut into five regions (top, top-middle, middle, middle-base, and base). For lignin detection and cell wall counter-staining, samples were fixed with FAA (formaldehyde:acetic acid:alcohol) solution for one week, dehydrated through an ethanol series, and embedded in paraffin. Following paraffin removal, thick sections (1 μm) were stained with safranin O and alcian blue as outlined in Ruzin19.

A second set of ten inner stems from five individual plants were isolated and cut into five regions for immunohistochemical investigation. For pectin and hemicellulose detection, samples were fixed and embedded in London Resin White following the method outlined in Young et al.20. Samples were incubated at room temperature in a blocking solution [5% (w/v) normal goat serum (NGS) in 1× Tris-buffered saline/0.2% Tween (v/v) (TBST)] in a homemade humidity box for 40 minutes. Blocking solution was washed off using 1× TBST. Slides were then incubated with primary antibodies at 1:10 dilution (v/v) in 1% (w/v) NGS blocking solution in the humidity box for 1 hour (antibodies are described below). Slides were washed 2 × 20 minutes in 1× TBST. Secondary antibodies were diluted at 1:100 (v/v) in 1% (w/v) NGS blocking solution in the dark for 45 minutes. Slides were washed 2 × 20 minutes in the dark and mounted in 90% (v/v) glycerol. Control slides were used to test the specificity of the secondary antibodies and also to test for autofluorescence. Blocked slides that were not incubated with primary or secondary antibodies were imaged, as well as slides blocked and incubated with only secondary antibody. LM10 (Rat IgG2c) and LM11 (Rat IgM) antibodies were obtained from Plant Probes, UK21. LM10 binds to unsubstituted or low-substituted xylan backbone chains, while LM11 is able to additionally bind to wheat arabinoxylan. JIM7 (Rat IgA) and JIM13 (Rat IgM) were obtained from the Complex Carbohydrate Research Centre22,23. JIM7 binds to partially methyl-esterified homogalacturonan and JIM13 binds to arabinogalactan and arabinogalactan protein. Alexa-fluor 488 secondary antibody was obtained from Invitrogen (goat anti-Rat IgG (H + L) polyclonal, CAT# A-11006).

For cellulose detection, a total of five, fully hydrated S. lepidophylla stems were isolated from three different plants and embedded in polyethylene glycol (PEG) using the protocol from Gierlinger et al.24. Embedded samples were sectioned (10 μm thickness) using a Leica RM2245 semi-automated rotary microtome. Solidified PEG was then removed using washes of ddH2O. Samples were then prepared following the method outlined in Ruzin19 for calcofluor white staining.

All samples were imaged using a Leica DM6000B epifluorescence microscope (Leica Microsystems, Wetzlar Germany) and QImaging Retiga CCD camera with Openlab imaging software (QImaging, British Columbia Canada).

### Scanning electron microscopy

Fresh, hand-cut sections (~1 mm thick) from the apical stem region were mounted on SEM stubs and allowed to air-dry (at a relative humidity ~30%) for 24 hours prior to imaging. Dry sections were imaged using a Hitachi TM3030Plus SEM in backscatter electron mode.

### Atomic force microscopy

A JPK Atomic Force Microscope (JPK Nano-wizard@3 Bio Science, Berlin, Germany) was used for imaging and force spectroscopy. S. lepidophylla stem samples (three stems, each from separate plants, with five regions per stem, and three replicate sections per region) were cut transversally using a sharp blade in a wet state. Sections (1 mm in thickness) were air-dried to ambient humidity (~30% relative humidity [RH]) and placed on double-sided clear tape on a microscope slide. Cortical stem tissue in adaxial and abaxial regions was located to perform force measurements. All the measurements were performed on tissue in ~30% RH. Using the QI imaging mode of the JPK AFM, a number of force maps were created within the area of 30 μm × 30 μm on each sample. Reduced scan areas were then selected to obtain the structural details of the cell walls in which 128 × 128 indentation points were tested in areas of 1–10 μm2. The force maps averaged 3 μm2 in size. For consistency we prescribed the indentation parameters (e.g. window size, number of points, indentation depth) within the cortical tissue area across stem sections and between adaxial and abaxial sides.

Three types of AFM probes were used for contact mode imaging: (1) Non-conductive silicon nitride cantilevers with integrated conical tips of radius 20 nm (MLCT Micro-cantilever, Bruker, Mannheim, Germany) [a spring constant of 0.06 N/m and a resonance frequency of 22 kHz]; (2) Biotool high resolution qp-BioAC/Quartz cantilevers with a 2 nm defined conical tip (Nanotools USA LLC, Henderson, NV) [a 60 μm length, a spring constant of 0.1 N/m and a nominal resonance frequency of 50 kHz]; and (3) Super-sharp standard Force Modulation Mode Reflex Coating (FMR) cantilevers with diamond-like carbon nano-tip of radius 2–3 nm (Nanotools USA LLC, Henderson, NV) [a spring constant of 2.8 N/m, a nominal resonance frequency of 75 kHz in air]. Non-Contact High Resonance (NCHR) cantilevers (Nanotools USA LLC, Henderson, NV) with a nominal spring constant of 40 N/m and integrated spherical tip of radius 100 nm (±10%) were applied for indentation measurements. The indentation frequency was 1–500 Hz. The AFM measurements were performed in an ambient environment of 20–25 °C and ~30% RH. Prior to each indentation test, the deflection sensitivity of the AFM cantilever was calibrated by engaging the cantilever on the surface of a clean microscope slide25. The spring constant of the cantilever was then determined from the power spectral density of the thermal noise fluctuations in air by fitting the first free resonance peak of the AFM cantilever to that of a simple harmonic oscillator using the JPK software26.

Indentations were repeated at given locations to ensure that no permanent deformation occurred at the surface of the sample. Force maps containing 128 × 128 indentation points were created on each indentation area. The elastic modulus E of a sample was obtained from the retracting force-indentation depth curve through Hertzian contact mechanics, where $$E=3F(1-{\nu }^{2})/4\sqrt{R{\delta }^{3}}$$ is the relation between the elastic modulus E and the applied indenting load F with $$\nu$$ being the Poisson’s ratio of the sample, R the radius of the AFM probe, and δ the indentation depth27. A number of assumptions were considered. The deformation of the sample relative to its thickness and also relative to the radius of the probe was assumed very small28. Any strain below the elastic limit was also assumed infinitesimal, a condition satisfied with the use of an indentation depth below 50 nm that rules out the influence of the glass substrate as well as any nonlinear and inelastic behaviour of the tissue at higher strains29,30. The Poisson’s ratio $$\nu$$ was selected to be 0.5. Data analysis was performed with the JPK data processing software. Statistical significance was determined by a paired Student’s t-test, when applicable. Differences were considered significant at p < 0.05.

## Results and Discussion

### Cell wall appearance differs between adaxial and abaxial stem sides

Dried, transverse sections from a set of regions along the length of inner S. lepidophylla stems were scanned with AFM to generate topographical images of the adaxial and abaxial cortical cell walls (Fig. 2A,B; Supplementary Fig. 1). Overall, cortex cells appear round to oval in geometry, with no obvious change in cell shape between adaxial and abaxial stem sides or between tip and basal stem segments. Differences arise when comparing cell wall layering between adaxial and abaxial cortical cells. Abaxial cell walls show very distinct secondary cell wall layers, while those in the adaxial region appear relatively smooth. This pattern is observable in sections along the length of the stem and is also visible in cortical cell walls imaged with transmission electron microscopy (Supplementary Fig. 2). Typical dicot and gymnosperm secondary cell walls contain three layers (S1–S3, with S2 being the thickest layer)31,32,33. In some monocots, such as bamboo species, more layers are visible. Bamboo fibers can have up to six to eight distinct cell wall layers33,34,35,36. Qualitatively, S. lepidophylla cortical cells walls, especially those on the abaxial stem side, appear to have more than three SCW layers, and therefore resemble bamboo with respect to cell wall layer morphology. Compared to adjacent cell types, fiber cells in Arabidopsis thaliana, wood, and bamboo are stiffer and mechanically support the stem/trunk/culm37,38,39,40. Thus, given their secondary cell wall morphology, S. lepidophylla stem cortical cells most likely act like fiber cells, providing structural support and reinforcement to the stem.

### Cell wall elasticity differs between adaxial and abaxial stem sides

A number of considerations need to be made to interpret the results obtained from AFM and other indentation techniques17,46,47. Of particular importance is the indentation depth and force relative to the thickness of the specimen. If the indentation is too deep or the forces are too large, there is a risk that the sample could be significantly damaged and/or that the observed results were influenced by the properties of the substrate underlying the section (i.e., a glass slide). To minimize these factors, the samples were indented with low magnitude forces, below 500 nN, and the indentation depth chosen (~20–30 nm) was smaller than both the thickness of the sample (~1 mm) and the radius of the indenting sphere (~100 nm). Another factor that must be considered when indenting living plant cell walls is the contribution of turgor (internal cell pressure on the cell wall from the fluid-filled vacuole) to the measured mechanical properties47,48,49. Inner S. lepidophylla stems, still undergoing development, are living9,11. Thus, to reduce the influence of turgor on the measured cell wall E of cortical cells, samples were air-dried to a relative water content of ~5% prior to indentation.