Gradient of structural traits drives hygroscopic movements of scarious bracts surrounding Helichrysum bracteatum capitulum

Abstract

Background and Aims The capitulum of Helichrysum bracteatum is surrounded by scarious involucral bracts that perform hygroscopic movements leading to bract bending toward or away from the capitulum, depending on cell wall water status. The present investigation aimed at explaining the mechanism of these movements.

Methods Surface strain and bract shape changes accompanying the movements were quantified using the replica method. Dissection experiments were used to assess the contribution of different tissues in bract deformation. Cell wall structure and composition were examined with the aid of light and electron microscopy as well as confocal Raman spectroscopy.

Key Results At the bract hinge (organ actuator) longitudinal strains at opposite surfaces differ profoundly. This results in changes of hinge curvature that drive passive displacement of distal bract portions. The distal portions in turn undergo nearly uniform strain on both surfaces and also minute shape changes. The hinge is built of sclerenchyma-like abaxial tissue, parenchyma and adaxial epidermis with thickened outer walls. Cell wall composition is rather uniform but tissue fraction occupied by cell walls, cell wall thickness, compactness and cellulose microfibril orientation change gradually from abaxial to adaxial hinge surface. Dissection experiments show that the presence of part of the hinge tissues is enough for movements.

Conclusions Differential strain at the hinge is due to adaxial–abaxial gradient in structural traits of hinge tissues and cell walls. Thus, the bract hinge of H. bracteatum is a structure comprising gradually changing tissues, from highly resisting to highly active, rather than a bi-layered structure with distinct active and resistance parts, often ascribed for hygroscopically moving organs.

INTRODUCTION

Hygroscopic movements are passive movements of plant organs due to uneven deformation caused by changes in water content in dead tissues (Elbaum and Abraham, 2014). Their mechanism is generally based on swelling or shrinking of a cell wall matrix that fills the space between reinforcing cellulose fibrils (Bresinsky et al., 2013). If such swelling or shrinking is not the same in different parts of the organ actuator, stress develops and its relief results in organ deformation (Elbaum et al., 2008). Identification of the movement mechanism is an important aspect of plant cell wall mechanics, and is potentially instructive for biomimetic materials science (Fratzl and Barth, 2009; Burgert and Fratzl, 2009a; Reyssat and Mahadevan, 2009).

In the majority of hygroscopically moving plant organs the actuator comprises active and resistance tissues. Different response of these tissues to wetting and drying that drives the movements is strongly dependent of the difference in directions of cell wall reinforcement by cellulose fibrils (Fratzl et al., 2008; Burgert and Fratzl, 2009b; Elbaum and Abraham, 2014). One exception to this rule are movements of awns in some Geraniaceae species that are driven by a single tissue with cellulose fibrils forming a special tilted helix, rather than the collaboration of active and resistance tissues (Abraham et al., 2012; Elbaum and Abraham, 2014). Another exception are movements of a pod valve in Bauhinia variegata (Armon et al., 2011) that is built of two layers of sclerenchyma fibres, which are oblique with respect to the valve longitudinal axis but nearly orthogonal one to the other. Fibre shrinking in both layers results in pod opening and helical turning of the valve.

Among organ shape changes, such as twisting, coiling or bending, accompanying the hygroscopic movements (Elbaum and Abraham, 2014), bending is the simplest from a geometrical perspective. It is common in organs participating in seed dispersal such as conifer cone scales, awns of a wild wheat Triticum turgidum and of some Geranium species, as well as umbel rays in carrot Daucus carota (Lacey et al., 1983; Dawson et al., 1997; Elbaum et al., 2008; Abraham and Elbaum, 2013). Hygroscopic bending of involucral bracts surrounding a capitulum, in turn, contributes to flower protection and pollination, as in the monocot species Syngonanthus elegans (Oriani and Scatena, 2009). The role in flower protection is also attributed to scarious bracts surrounding the capitulum of some dicot species from Asteraceae family, often used as decorative dried ‘flowers’, including Helichrysum bracteatum (Uphof, 1924).

The actuator of hygroscopically bending organs is in most of the cases a bi-layered structure. In the resistance layer the cellulose microfibrils are usually parallel to the organ axis. An exception is the sesame Sesamum indicum capsule where the resistance tissue comprises two sublayers of mutually orthogonal sclerenchyma fibres, the walls of which are reinforced along the cell axis (Shtein et al., 2016). The cell wall structure in the active layer differs among species. In pinecone scales, active tissue is built of sclereids, the cell walls of which have anisotropically arranged cellulose microfibrils that are predominantly perpendicular to the scale axis (Dawson et al., 1997; Burgert and Fratzl, 2009b). In wheat awns, the cell wall of the active tissue is a complex structure with different microfibril arrangements in adjacent cell wall layers (Elbaum et al., 2008), while in sesame capsules active tissue is a foam-like parenchyma with no preferred direction of cell wall reinforcement (Shtein et al., 2016). In Asteraceae species, although their scarious involucral bracts have for a long time been known as a textbook demonstration of hygroscopic movements (Uphof, 1924; Brauner and Rau, 1966), to our knowledge the mechanism of actuation has not yet been studied. Although the cell structure and programmed cell death that takes place during bract development were investigated for the bract of H. bracteatum (Nishikawa et al., 2008), the cell wall properties were not related to the movements.

Therefore, our objective was to elucidate the mechanism of hygroscopic movements of the H. bracteatum bract. First, using the replica method we quantified the strain and shape changes accompanying the movements, which allowed us to identify two regions undergoing different deformation due to wetting: the bract actuator (the hinge) exhibiting drastic shape change and non-uniform strain, and the bract blade also undergoing deformation but with almost uniform strain and much smaller shape change. Next, performing dissection experiments we assessed the contribution of different hinge tissues in the hinge deformation. We then aimed to relate specific cell wall structure and composition of the hinge and blade tissues to their deformation, using light and electron microscopy combined with confocal Raman spectroscopy.

MATERIALS AND METHODS

Plant material

Plant material was obtained from potted plants of Helichrysum bracteatum (Vent.) Andrews [synonym to Xerochrysum bracteatum (Vent.) Tzvelev; Bayer, 2001] monstruosum fl. pl. grown for 4 months in a glasshouse with temperature 20–22 °C and 16 h of light. Plants with white bracts were examined exclusively. For all the analyses, middle involucral bracts (such as one marked with an asterisk in Fig. 1A), 15–20 mm long and 4–5 mm wide, were isolated from mature open capitulum.

Fig. 1.

Capitulum and involucral bract of H. bracteatum in the dry and wet state. (A) Dry capitulum viewed from above. An exemplary bract like those used for analysis is marked by an asterisk. (B,C) The same capitulum in the wet state, viewed from above (B) and in side view (C). (D,E) Side views of an individual isolated bract in the dry (D) and wet (E) state, on the millimetre scale. Adaxial bract surface (facing the florets) is on the upper side. Bract regions used in the analysis are marked. Note that the bract is held by forceps (on left) by its tip so the base is moving rather than the blade.

Transmission and scanning electron microscopy

For transmission electron microscopy (TEM) examination, central bract fragments, 2–3 mm wide, were fixed in 3 % glutaraldehyde in 50 mm phosphate buffer (pH 7·2) overnight at 4 °C. After rinsing in the phosphate buffer, specimens were post-fixed in 2 % osmium tetroxide in the phosphate buffer for 2 h at room temperature, and rinsed again in the same buffer. They were then dehydrated through a graded ethanol series followed by propylene oxide, and embedded in Epon resin (Polysciences, Inc., Hirschberg, Germany). Ultrathin sections were obtained with the aid of a Leica EM UC6 ultramicrotome (Leica Microsystems, Wetzlar, Germany), collected on copper grids (200 mesh), stained with 2 % uranyl acetate in 50 % ethanol and Reynold’s lead citrate, and examined in a Hitachi H500 TEM device (Hitachi, Tokyo, Japan).

For scanning electron microscopy (SEM) examination, hand-cut longitudinal sections through different regions of dry bracts were sputter coated and observed in a Philips XL 30 TMP ESEN microscope.

Atomic force microscopy

Atomic force microscopy (AFM) measurements were performed using a NanoWizard3 BioScience (JPK Instruments, Berlin, Germany) operating in intermittent contact mode. Images were acquired with the aid of HQ:NSC15 rectangular Si cantilevers (MicroMasch, Estonia) with spring constant specified as 40 N m^–1, cantilever resonant frequency of about 325 kHz and tip radius of 8 nm. All scans were conducted in air in laboratory conditions (22 °C, constant humidity of 45 %). Image analysis was performed using the JPK Data Processing software (JPK Instruments).

Raman spectroscopy

Thin samples (hand-cut cross-sections several dozen micrometres thick) were put on glass slides (1 mm thick), immersed in water to preserve environment conditions, and covered by glass coverslips (0·13–0·16 mm thick) sealed on the margins with nail polish. Raman data were collected with the aid of a WITec confocal Raman microscope CRM alpha 300, equipped with an air-cooled solid-state laser (λ = 532 nm) and an electron multiplying CCD (EMCCD) detector. An Olympus MPLAN (50×/0·76 NA) air objective was used. The excitation laser radiation was coupled into a microscope through a single-mode optical fibre (50 mm diameter). Raman scattered light was focused onto a multi-mode fibre (50 mm diameter) and monochromator with a 600 line mm^–1 grating. Instrument calibration was verified by checking the position of the first-order Si mode (520·7 cm^–1). For the hinge section, Raman imaging was performed in 160 × 400-μm areas using 80 × 200 pixels (= 16 000 spectra), and for the blade in 85 × 50-μm areas using 255 × 150 pixels (= 38^250 spectra). In both cases Raman imaging maps were collected with an integration time of 100 ms per spectrum, precision of ± 1 cm^–1 and resolution of 3 cm^–1. All spectra were collected in the range 200–4000 cm^–1 at 40 mW on the sample.

All the output data were manipulated by performing a baseline correction and cosmic ray removal. For each spectrum band fitting analysis was performed with the aid of the GRAMS\AI software package, using a Voigt function with the minimum number of components, in order to separate overlapping bands originating from different polysaccharides and to identify the cell wall polymer components. Integration over these bands was performed to create Raman images and present the distribution of the selected polysaccharide through the cross-section.

Raman images were generated using a sum filter by integrating over defined wavenumber regions in the spectrum, at the intensities calculated within the selected limits: band I within 1090–1130 cm^–1; band III, 1580–1620 cm^–1; band IV, 2871–2911 cm^–1; and band V, 2928–2968 cm^–1. Bands assignment, based on literature, is discussed and specified in the Supplementary Notes S1. For each wall type similar spots were identified based on the Raman spectra for calculation of average spectra, obtained for polarization angle 0° (direction parallel to the x-axis of images, e.g. Fig. 8). For selected sections, Raman spectra were obtained at two polarization angles, 0° and 90° (Figs 9–11). The incident excitation laser beam was parallel to the z-axis of the cross-section positioned with the (001) face parallel to the xy plane of the microscope stage. For a detailed analysis of the intensity and the shape of the spectra for various cell wall layers, line segments were drawn across the selected walls and band intensities of cellulose and lignin were analysed along these lines.

Fig. 8.

Blade structure. Micrographs were obtained by SEM (A, B) or TEM (C–F). (A) Longitudinal section with adaxial epidermis on the upper side. (B) Longitudinal section of adaxial epidermis (Ep) with large rounded pit apertures (arrowhead) in anticlinal walls, and underlying parenchyma cells (Pa) with flange-like ingrowths of secondary walls and pits (arrowhead). (C–F) Cross-sections. (C) Adaxial epidermis with non-uniformly thickened outer periclinal wall covered by cuticle (Cu), and a pair of pits (arrowhead) in anticlinal walls. (D) Parenchyma cell walls with flange-like secondary wall ingrowths (asterisk). (E) Abaxial epidermis with uniformly thickened outer periclinal wall and a pair of pits (arrowhead) in the anticlinal wall. (F) Fragment of anticlinal walls of abaxial epidermis with electron-dense ‘joints’, and outer periclinal wall with thin cuticle (Cu). Cu, cuticle; Ep, epidermis; Pa, parenchyma; PW, compound primary wall; SW, secondary wall.

Fig. 9.

Raman images and spectra. Hinge (A–C) and blade (D, E) cross-sections are analysed at the polarization angle 0° (parallel to the x-axis). Images are based on the integrated intensity of selected bands (A, D): band I integrating from 1090 to 1130 cm^–1 (cellulose); band III, 1580–1620 cm^–1 (lignin); band IV, 2871–2911 cm^–1 (cellulose); and band V, 2928–2968 cm^–1 (hemicellulose and pectin). Red colour means the strongest relative intensity for an individual band, and black the lack of Raman signal. Abaxial bract surface is on left. Cross-sections (B, E) with numbers (for hinge) or letters (for blade) pointing to exemplary locations for which the average spectra are shown in C and F. (C) Average Raman spectra (each average is from ten spectra) obtained for a reference spectrum of natural cotton fibre (0) and for various locations at the hinge section: 1, outer periclinal wall of sclerenchyma-like abaxial epidermis; 2, anticlinal wall of the same cells; 3, anticlinal wall of subepidermal sclerenchyma-like cell; 4, xylem cell wall; 5, phloem; 6, periclinal wall of parenchyma; 7, outer periclinal wall of adaxial epidermis. (F) Average Raman spectra (each average is from ten spectra) for various blade section locations: a, outer periclinal wall of abaxial epidermis; b, anticlinal wall of parenchyma cell; c, periclinal walls of parenchyma, inner wall of adaxial epidermis and non-swelling part of the outer periclinal wall of adaxial epidermis; d, swelling part of periclinal wall of adaxial epidermis. I–V mark the location of bands shown in A and D. Band II is the putative callose signal, enhanced only in phloem. C, cellulose; L, lignin; H & P, hemicellulose and pectin.

Fig. 10.

Raman images and integrated intensity of cellulose and lignin bands for the hinge cross-section at different polarization angles. (A) Relative intensity for two orientation-sensitive cellulose bands I and IV: left images at polarization angle 0° (parallel to the x-axis); right at an angle of 90° (parallel to y). Red colour indicates the strongest intensity for the selected band, and black the lack of Raman signal. Abaxial hinge surface is on the left. (B) Cross-section with marked line segments (a–b and c–d) used for signal integration. (C) Integrated intensity distribution of cellulose (band I) and lignin (band III) at polarization angles of 0° (left panels) and 90° (right), along the line segments marked in B: segment a–b (green) across two compound anticlinal walls of abaxial epidermis cells; and segment c–d (grey) across subepidermal sclerenchyma-like cells.

Fig. 11.

Raman images for the blade cross-section at different polarization angles. Relative intensity for orientation-sensitive cellulose bands I and IV is shown at polarization angles of 0° (left; parallel to the x-axis) and 90° (right; parallel to y). Red colour indicates the strongest intensity for the selected band, and black the lack of Raman signal. Abaxial blade surface is on the left.

The obtained Raman spectra were presented in full spectral range and magnified to illustrate two main regions: cellulose and lignin, and cell wall material in general. Since in all collected Raman spectra signals originating from cellulose, lignin and polysaccharides such as hemicellulose/pectin are overlapping, and it is difficult to distinguish signals from individual components, the spectra were compared with natural cotton fibre to identify bands originating from cellulose (spectrum 0 in Fig. 8C;Supplementary Data Fig. S5).

Five sections of hinge and five of blade, obtained from bracts of the same capitulum as used for other experiments, were examined.

Immunolabelling technique and epi-fluorescence microscopy

A modified immunolabelling technique of Willats et al. (2001) was performed on hand-cut cross- and longitudinal sections, several dozen micrometres thick. Briefly, the sections were incubated first in phosphate-buffered saline (PBS, pH 7·2) containing 3 % (w/v) bovine serum albumin (BSA; Sigma-Aldrich) for 1 h and next in a monoclonal antibody (either LM5, LM10, LM19 or LM25; PlantProbes, UK) diluted to 1:20 in PBS, for 1–1·5 h in darkness. The samples were then washed three times in PBS and incubated again for 1 h with the secondary antibody Alexa Fluor 488 (Invitrogen) diluted to 1:100 in PBS. After washing three times in PBS the sections were mounted in anti-fade medium Fluoromount (Sigma-Aldrich, St Louis, MO, USA) and observed in an epi-fluorescence microscope (Olympus BX41) equipped with digital camera (Camedia 3040 ZOOM).

Microfibril angle assessment in polarized light microscopy

Microfibril angle (MFA) of cell walls in hinge tissues was assessed in macerates obtained by incubation of hinge fragments in a mixture of glacial acetic acid and 30 % hydrogen peroxide (volume ratio 1:1) at 60 °C for about 18 h (Leal et al., 2006). In the case of blade tissue, MFA was assessed for outer periclinal cell walls of adaxial epidermis in hand-cut paradermal sections. All measurements were performed in a polarized light microscope (Nikon Eclipse 50i Pol) equipped with digital camera (Nikon DS-Fi2) and NIS-Elements Br software, exclusively on exposed fragments of individual cell walls (Leney, 1981; Donaldson, 2008). First with the red wave plate introduced below the analyser, the specimen was rotated clockwise such that the colour of the analysed cell wall fragment changed from yellow to red to blue. When in the red position, the red plate was removed, a slight correction was made for the strongest extinction, and the angle of the rotary microscope stage representing the maximum extinction position was noted. Next, the specimen was rotated to align the long cell axis with the vertical cross line in the eyepiece, and the angle noted again. Microfibril angle was then computed as the difference between these two angles (Leney, 1981).

Measurements of tissue thickness, cell wall thickness and relative cell wall area

Tissue thickness was measured in images of individual hand-cut transverse sections first in the dry state and next after placing in water, examined in a light microscope (Nikon ECLIPSE 80i) equipped by digital camera (DS-Fi2). The thickness of the same cells or groups of cells recognized in two images was measured with the aid of ImageJ software (https://imagej.nih.gov/ij/). Six sections from the hinge and six from the blade obtained from three bracts were used for measurements.

The images of wet sections were used also to assess cell wall thickness and a relative cell wall area, i.e. the percentage of a cell surface area occupied by the wall, as visible in the section. From each tissue layer (abaxial epidermis and subepidermis, three to four parenchyma layers, and adaxial epidermis), five cells with relatively large lumen, i.e. those that were sectioned at the central portion, not near the tip, were selected for measurements. For each cell the centre of mass, the surface area of the whole cell (including the wall) and the surface area occupied by the wall were measured, using the ‘analyze particles’ function of ImageJ. Cell wall thickness was measured for the same cells: at three locations for inner walls of each cell, and additionally at three locations for outer periclinal walls for epidermal cells. Three sections from different bracts were analysed in this way.

Quantification of bract surface geometry and strain based on replicas

To quantify deformation of the bract surface that accompanies the hygroscopic movements, dental polymer (Kerr impression materials: Take 1 – the hydrophilic vinyl wash material, regular set) replicas were taken from its abaxial and adaxial surface in the dry and wet state (Kwiatkowska and Burian, 2014). First, the replicas were taken from the two surfaces of the dry bract isolated from the capitulum. Fast bract movement takes place when water is applied to its abaxial surface, not adaxial, and the full deformation from dry to wet state or vice versa takes only 1–3 min. Thus, next the abaxial bract surface was placed on the surface of a water drop and, when the bract shape was no longer changing, the polymer was applied on the opposite, i.e. adaxial, surface. During setting of the dental polymer, which takes approx. 5 min, the bract shape was not changing given that the abaxial surface was constantly in contact with water. Next, with the already set polymer remaining on the adaxial surface, excess water was quickly removed from the abaxial surface by filter paper, and the fresh polymer was applied immediately to obtain a replica of the abaxial surface in the wet state. During polymer setting on the abaxial surface, water evaporation was limited because the whole bract surface was covered by polymer, and bract movement was halted by the presence of set polymer on the adaxial side. Six bracts from three different inflorescences were used in this experiment.

For strain and shape assessment of the hinge, the replicas were used as moulds to make epoxy resin (Devcon 2 ton epoxy) casts, which after sputter-coating were observed by SEM (Philips XL 30 TMP ESEN). Two images, tilted with respect to each other by 10° around the x image axis, referred to as stereopairs, were taken from each patch of epidermal cells recognized in the casts of bract surface examined in the dry and wet state. The stereoscopic reconstruction protocol was used to reconstruct the bract surface from the stereopairs (Routier-Kierzkowska and Kwiatkowska, 2008). The reconstructed surfaces were then used to compute the surface geometry in the dry and wet state and strain accompanying the movement, with the aid of adjusted protocols previously used for meristem growth analysis (Dumais and Kwiatkowska, 2002). Briefly, local bract shape was estimated as curvature directions of a surface approximating 3D coordinates of all the vertices identified in the examined epidermal patch. Bract surface strain was quantified for the transition from the dry to wet state, by means of: principal directions of strain computed as in Dumais and Kwiatkowska (2002); cell strain in an area; and relative change in cell length or width (linear strain in longitudinal or transverse direction, respectively). Cell surface area was computed as the sum of surface areas of triangles comprising the cell surface (each triangle described by two adjacent cell vertices and the cell centroid). Cell length was the distance between the two most distant cell vertices. To compensate for possible measurement imprecisions, linear strain in the transverse direction was computed based on distances between cell vertices at least four times more distant than a cell width. Strain, areal or linear, was computed as the ratio of cell area or length change and initial area or length, in per cent. All computation protocols were written and performed in Matlab (The Mathworks, Natick, MA, USA).

For strain assessment of the blade, nail polish replicas of the blade surface excluding the curved regions near margins were obtained from dental polymer moulds, and observed in water under light microscopy (Elsner et al., 2012). Images of large regions were acquired under the microscope (Nikon ECLIPSE 80i) with the aid of a digital camera (Nikon DS-Fi2) and Large Image module of NIS-Elements Br software, and the same cells were recognized in images of dry and wet states. Linear strains in the longitudinal and transverse direction were then assessed based on measurements of distances between cell vertices performed in Adobe Photoshop CS4 Extended. To compensate for possible imprecisions in vertex positioning in nail polish replicas, vertices at least twice more distant than a cell length and at least ten times more distant than a cell width were chosen for measurements.

Preparation and analysis of dissected hinge samples for identification of active and resistance tissues

Tissues from the adaxial surface (either adaxial epidermis or epidermis and parenchyma) were removed with the aid of a razorblade from the hinge of an entire bract in the dry state; sclerenchyma-like cells were removed from the abaxial surface of a bract in the wet state (as the hinge is strongly curved in the dry state, tissue on this side is not easily accessible). Prior to dissection the surface of interest was carefully painted with a waterproof marker to ensure that all the marked tissue is removed. After tissue removal, the hinge was isolated and longitudinal sections were obtained from its central portion in the wet state. Care was taken to obtain the sections with no vascular bundles included. All these were performed and documented under a stereoscopic microscope (Nikon SMZ 288) equipped with digital camera (DS-Fi1).

Next, images of the entire sections were acquired under the microscope (Nikon ECLIPSE 80i) with the aid of a digital camera (DS-Fi2) and Large Image module of NIS-Elements Br software, first of the section in the dry state, and next in the wet state after water application onto the section. Measurements of section length were performed from these images with the aid of CorelDRAW Graphics Suite X4 (Corel Corp.).

Three bracts with dissected hinge were examined for each dissection type. Two sections of each kind (intact hinge, hinge without adaxial epidermis, without adaxial epidermis and parenchyma, and without sclerenchyma-like cells) obtained from different bracts were analysed.

Statistical analysis

All the analyses were performed using the Statistics and CircStat (Berens, 2009) toolboxes of Matlab.

RESULTS

Prominent bract shape changes and differential strain due to wetting are restricted to the hinge

Involucral bracts surrounding the H. bracteatum capitulum are bent outward from the capitulum centre when dry (Fig. 1A) and inward when wet (Fig. 1B, C), so that the florets or achenes are exposed only in the dry environment. Also, an isolated bract has a different shape in the dry and wet state (Fig. 1D and E, respectively). Macroscopic examination of the isolated bract shows that in both the states its base has a similar curved shape (Fig. 1D, E). The major shape changes take place adjacent to this region, at about one-quarter of the bract length measured from its basal end, within the narrow zone referred to as a hinge. The bract region distal to the hinge, a blade, is displaced due to hinge deformation but its shape is not greatly affected by wetting: it is gutter-shaped both in the wet and dry state because the bract margins are slightly bent in the adaxial direction. Shape change of a dry bract is quite fast if water is applied to its abaxial (outer) surface (full bract deformation takes only 1–3 min). If only the adaxial surface (facing the capitulum) is wetted the change is much slower.

To characterize the hygroscopic movements in more detail, we used a replica method to measure deformation at both sides of the hinge and to compare it with blade deformation. In the hinge, because its shape is complex, we assessed local surface curvature and strain with cellular resolution for patches of hinge epidermis, at abaxial or adaxial sides, comprising 15–60 neighbouring cells (Fig. 2). In the wet state all the examined patches are nearly flat while in the dry state they are strongly curved along the bract axis (Supplementary Data Fig. S1). Local surface strain, taking place due to wetting, differs profoundly between the abaxial and adaxial hinge surfaces (compare A and B with C and D in Fig. 2). Both the relative changes in cell length (linear strain) and surface area (areal strain) are much higher on the abaxial than adaxial hinge surface, especially in its proximal part where the abaxial epidermal cells extend by over 20 % due to wetting, whereas adaxial epidermal cells extend by less than 10 % or slightly shrink (Figs 2B, D and 3A). The sides differ also in strain anisotropy. On the abaxial surface the direction of maximal strain (extension) is longitudinal, i.e. along the bract axis, while in the transverse direction lower extension or slight contraction take place (Fig. 2A, B). On the adaxial surface, by contrast, the direction of maximal strain is transverse or oblique with slight extension in this direction. In the direction of minimal strain (longitudinal or oblique), the adaxial surface is usually shrinking (Fig. 2C, D).

Fig. 2.

Strain of hinge surface patches accompanying the change from the dry to wet state. Patches of abaxial (A, B) and adaxial (C, D) epidermis of an individual bract hinge are shown in SEM micrographs in the dry and wet state (first and second columns). The third column shows the linear strain of cells in the longitudinal direction (colour code), and the fourth, the areal strain (colour code) and principal strain directions (crosses). Cross arm lengths are proportional to the maximal and minimal strains; arms are red if in this direction the surface is shrinking, and black if it is expanding. Patches are located in distal (A, C) or proximal (B, D) hinge parts. Bract axis orientation is vertical, with base on the bottom.

In the blade we assessed strain for its central part, which is nearly flat before and after wetting, using landmarks further apart than individual cell vertices (twice the cell length in the longitudinal direction, and ten times the cell width in the transverse direction). In this way we were able to assess small strains with acceptable precision. We found no statistically significant difference in strains between the abaxial and adaxial surfaces (Fig. 3B). On both blade sides the highest linear strain (extension by 5–10 %) is in the transverse direction, while in the longitudinal direction a minor shrinking, by circa 3 %, occurs.

Fig. 3.

Linear strains accompanying the change from the dry to wet state, measured for adaxial and abaxial surfaces of bract hinge (A) and blade (B). Within each plot the thick line represents median, box delimits the first and third quantiles, and whiskers extend from each end of the box to the adjacent values in the data as long as the most extreme values are within 1·5 times the interquartile range from the box limits. Lower-case letters below boxes are different if differences between means are statistically significant (P < 0·05; Kruskal–Wallis test was used for multiple comparisons since Kolmogorov–Smirnov test showed a non-normal distribution). (A) For the hinge, numbers of measurements are: 49 and 54 for longitudinal direction on adaxial and abaxial surface, respectively; 25 and 20 for transverse. (B) For the blade, numbers of measurements are: 23 and 35 for longitudinal direction on adaxial and abaxial surface, respectively; 21 and 31 for transverse.

In summary, bract deformation due to wetting takes place mainly at the hinge, distinguished by the strongly non-uniform strain and the largest shape changes. Thus, the hinge is the bract actuator. The difference in strain between abaxial and adaxial sides of the hinge suggests that active tissue (strongly extending when wetting) is located on the abaxial side, while resistance (almost not extending) tissue is on the adaxial side. Therefore, we next examined bract anatomy in the dry and wet state, searching for traits that would distinguish the hinge among other bract regions.

Bract regions differ in type of epidermal and subepidermal tissues

Involucral bracts of H. bracteatum are scarious. Both hinge and blade (Fig. 4A, B) comprise exclusively dead cells (Nishikawa et al., 2008). All along the bract there are several layers of parenchyma with slightly thickened walls (Fig. 4A, C, G), in which about dozen parallel vascular bundles are embedded (e.g. Fig. 4A, G). The most characteristic feature of the hinge is the presence of two to three layers of sclerenchyma-like cells on the abaxial side (Fig. 4A;Supplementary Data Fig. S2J, K). The outermost layer constitutes the abaxial epidermis of the hinge (labelled Ep-Sc in Figs 4–6). In the dry state there are apparent gaps between these cells (Fig. 4A; Fig. S2E, G) but they close when cells become more rounded in the wet state (Fig. 4B;Fig. S2F, H). Intercellular spaces between the sclerenchyma-like cells are apparent in the wet state (Fig. S2K). The adaxial hinge epidermis is in turn continuous and only its outer periclinal cell walls are thickened (Fig. 4A, B;Fig. S2C, D).

Fig. 4.

Bract anatomy. (A–D) Cross-sections of the hinge (A, B) and blade (C, D) of the same bract in the dry (left) and wet (right) state. Note closing of gaps between sclerenchyma-like cells of abaxial hinge epidermis (arrowheads in A and B) and the profound change in the outer periclinal wall of adaxial epidermis of the blade (dot in C and D), taking place due to wetting. The presented fragment of the blade section does not include vascular bundles but they were present in other section portions. (E–G) Cross-sections of other bracts stained with Toluidine Blue (E and F) or Neutral Red (G). (E,F) Sclerenchyma-like cells of the hinge (E) and base (F). (G) Base. In all the cross-sections adaxial surface is on the upper side, abaxial surface on the lower. (H) SEM micrograph of the surface of abaxial epidermis at the border between base and hinge, with leaf-like epidermis comprising collapsed epidermal cells proper (Ep) and stomata (St). Chl, chlorenchyma; Ep, epidermis; Ep-Sc, sclerenchyma-like epidermis; Pa, parenchyma; Sc, sclerenchyma-like cells; St, stomata; VB, vascular bundle.

Fig. 5.

Longitudinal sections of hinges from which various layers were removed in the dry and wet state. (A) Schematic representation. Frames point to the location of the fragments used for experiments. Bract base is always at the left section end; in the dry state the upper surface is adaxial. (B–E) Sections in the dry (left) and wet (right) state. Relative changes of section length measured on the adaxial (upper) and abaxial (lower) sides are given in per cent of dry state length. Boxes delimit fragments shown at higher magnification. (B) Section of the intact hinge. (C) Two layers of sclerenchyma-like cells (epidermis and subepidermis) were removed from the abaxial side. (D) Adaxial epidermis removed. (E) Adaxial epidermis and most of the parenchyma cells were removed so that only sclerenchyma-like epidermis and subepidermis are left. Ep-AD, adaxial epidermis; Ep-Sc, abaxial sclerenchyma-like epidermis; Pa, parenchyma; Sc, sclerenchyma-like cells.

Fig. 6.

Changes in tissue morphology and microfibril angle (MFA) across the hinge. (A,B) Relative cell wall area (A) and cell wall thickness (B) plotted against the normalized distance of cell mass centre from the abaxial hinge epidermis. Measurements for cells from different bract sections are in different colours. The same colours are used to label bars representing the distances within which cell centres of abaxial (Ep-AB) and adaxial (Ep-AD) epidermis are included in different sections. (B) Circles represent outer periclinal walls of epidermis, while dots represent inner walls of all tissues. Each value of cell wall thickness is the mean of three measurements. (C) MFA measured under polarized light microscopy for sclerenchyma-like abaxial epidermis, parenchyma cell walls and outer periclinal wall of adaxial epidermis of the hinge. Dots represent individual measurements. For each tissue the mean resultant vector, marked in red, is plotted such that the longer the vector, the lower the circular spread of the data. Zero MFA corresponds to alignment parallel to the cell axis (longitudinal).

Blade anatomy is simpler (Fig. 4C, D;Supplementary Data Fig. S3). Thin-walled parenchyma is covered by abaxial and adaxial epidermis with thickened outer periclinal cell walls, and except for vascular tissues (not shown) no other cell types are present. The structure of the adaxial epidermis is notable in that the inner portion of its outer periclinal cell wall swells profoundly due to wetting (compare Fig. 4C, D;Fig. S3E, F). This is not the case in the abaxial epidermis (Fig. 4C, D;Fig. S3L, M).

The bract base is distinguished by the presence of leaf-like abaxial epidermis with stomata and underlying chlorenchyma (Fig. 4G, H). These are the only cells that remain alive at bract maturity but die and collapse after capitulum collection for experiments. The leaf-like epidermis is continuous with the outermost layer of sclerenchyma-like cells at the hinge (Fig. 4H). Beneath chlorenchyma there is a sclerenchyma-like tissue that differs from the hinge by less rounded cell shapes and thinner walls (compare Fig. 4E and F).

Except for leaf-like abaxial epidermis of the bract base all the epidermal cells are elongated along the bract axis (Figs S2 and S3), but the regions differ in epidermal cell length. Cells of the central blade portion (Fig. S3C, J) are the longest both among adaxial and among abaxial epidermal cells, while hinge cells are the shortest, with those on abaxial side shorter than on the adaxial side (Fig. S2A, G). This trait probably makes it easier to accommodate for curvature changes accompanying the movements, which are highest on the abaxial side of the hinge, where the radius of curvature in the dry state is the smallest. Moreover, the distal ends of short abaxial hinge cells slightly overlap, i.e. are raised above, the proximal ends of neighbouring cells, which results in a roof tiles-like appearance. After wetting the surface becomes smooth (compare micrographs in Fig. 2B;Fig. S2G, H).

As changes in bract thickness due to wetting may also contribute to the movement, we assessed relative changes in the thickness of different tissues comparing distances between the same cell walls in transverse sections of the hinge or blade, in the dry and wet state (Table 1; Fig. 4A–D). Unlike the surface deformation, the significant changes, i.e. increase in tissue thickness due to wetting, take place mainly at the blade. The largest changes are that of the adaxial epidermis, where the thickness increase accompanies profound swelling of outer cell walls. In the hinge, only the thickness of the adaxial epidermis increases. Of note, the surface strain of this tissue is the lowest. Sections reveal that some air bubbles remain in the cell lumen after tissue deformation due to wetting (asterisks in Fig. 4D), suggesting that filling of the lumen with water is not driving the deformation.

Table 1.

Bract tissue thickness in the dry and wet state

Bract regions and tissues	n	Dry state thickness (µm), mean (SD)	Wet state thickness (µm), mean (SD)	Relative change (% dry state)
Blade
Adaxial epidermis	20	13·45 (2·62)	18·94 (2·31)	40 ·8 *
Parenchyma	18	51·13 (5·88)	56·40 (4·81)	10 ·3 *
Abaxial epidermis	41	13·00 (2·12)	13·90 (1·88)	6 ·9 *
Hinge:
Adaxial epidermis	41	13·72 (2·21)	15·67 (2·72)	14 ·3 *
Parenchyma	23	60·17 (7·20)	59·26 (7·10)	–1 ·5
Abaxial subepidermis	37	16·74 (2·95)	17·09 (2·96)	2 ·1
Abaxial epidermis	51	17·99 (2·58)	18·71 (2·60)	4 ·0

Statistically significant differences between the two states (t-test, P <  0·05) are marked by asterisks. n, number of measurements; SD, standard deviation.

Summarizing, the unique trait of hinge anatomy is the presence of sclerenchyma-like tissue, comprising abaxial epidermis and subepidermis. The sclerenchyma-like cells undergo profound size and shape changes due to wetting, unlike cells of the adaxial epidermis. These data combined with strain measurements suggest that the abaxial sclerenchyma-like cells are the active tissue while adaxial epidermis is the resistance tissue. We next performed dissection experiments to verify identification of such active and resistance tissues.

Hygroscopic movements take place even after adaxial or abaxial tissue removal

The identity of active and resistance tissues was verified by removing abaxial or adaxial cell layers from the hinge. We assumed that if the tissue identification was correct the dissection would disable the movements: after removal of an active (sclerenchyma-like) tissue the remaining tissues would not change size or shape due to wetting, while after removal of resistance (adaxial epidermis) tissue the remaining tissues would undergo strain but without any shape change. Most surprisingly, these dissections did not disable the bract movement. To ensure that the tissues were indeed removed and to assess the strain, we examined individual longitudinal sections of the hinge with various tissue removed (Fig. 5). The sections were obtained from wet bracts, and then dried and wetted again to assess the deformation. Particular care was taken not to include vascular bundles in the sections to ensure that only the sclerenchyma-like tissue, parenchyma or adaxial epidermis participate in movement generation. As expected, the intact hinge sections (Fig. 5B) underwent differential strain due to wetting, with the abaxial side extending by nearly 30 % and adaxial side exhibiting almost no length change. In agreement with the macroscopic observations of dissected bracts, the deformation took place also in sections lacking either abaxial (Fig. 5C) or adaxial (Fig. 5D) tissues, i.e. the postulated active and resistance tissue, respectively (we examined two sections from different bracts for each type of dissection; shape change and strain were very similar in sections representing the same dissection type). In sections comprising adaxial epidermis and parenchyma (after sclerenchyma-like tissue removal) nearly zero extension took place on the adaxial surface while the abaxial surface extended by 14 % (Fig. 5C), i.e. less than the sclerenchyma-like epidermis extension in the intact hinge section. In sections comprising sclerenchyma-like tissue and parenchyma (after removal of adaxial epidermis) both surfaces were extending (Fig. 5D) but the difference in their extension was similar to the previous case. Despite different strains in the two experiments, the general shapes of sections in the dry or wet state were very similar to that of intact hinge sections. This can be explained by the decrease in hinge thickness due to dissection (see Discussion). Surprisingly, even the hinge section comprising only two layers of sclerenchyma-like cells (Fig. 5E) underwent the shape change, although much smaller than in previous cases. The accompanying strain of the abaxial surface exceeded 30 %.

In summary, the presence of all the hinge tissues is not necessary for movements. Rather, cell walls of two adjacent tissues, i.e. either abaxial sclerenchyma-like cells together with parenchyma, or parenchyma together with adaxial epidermis, are sufficient to generate shape changes similar to those generated by all the hinge tissues working together. Nevertheless, the adaxial epidermis halts extension on the adaxial side as after its removal the extension of the adaxial parenchyma due to wetting increases. Even the isolated sclerenchyma-like tissue layers undergo some shape change. All these observations suggest that movements can be explained by gradual differences (an adaxial–abaxial gradient) in some cell wall traits between consecutive hinge tissue layers, rather than by the presence of well-defined active and resistance parts. We therefore next searched for such traits. Since the bract blade undergoes very different strain due to wetting, we compared hinge cells with those of the blade. Vascular bundles were not considered in this analysis because the experiments showed that they are not necessary for the movements.

Bract tissues undergoing different strains have a specific morphology and ultrastructure

The morphology of hinge cells changes gradually from the abaxial to adaxial side (Figs 6 and 7). The tissue fraction occupied by cell walls, represented by the relative cell wall area, is the highest for the abaxial sclerenchyma-like epidermis and decreases gradually with distance from the abaxial bract surface (Fig. 6A). Starting from the adaxial layer of parenchyma the area again increases, but it is much lower for the adaxial epidermis than for the abaxial epidermis. Of note, this tendency is the same for all the sections examined (different colours in Fig. 6A represent different sections). The tissue fraction occupied by cell walls is related to the cell lumen and cell wall thickness. The cell lumen increases from the smallest for the abaxial cells to the largest for adaxial parenchyma, but in adaxial epidermis cells it is again smaller (Figs 4A and 7A). The variation in cell wall thickness is similar to that of relative cell wall area (Fig. 6B). Outer periclinal walls on the abaxial side are the thickest (Fig. 6B). The thickness of inner walls gradually decreases (compare epidermis and subepidermis in Fig. 4A, E;Fig. S2K; parenchyma cells in Fig. 7B, C). Outer periclinal walls of the adaxial epidermis are again thicker (Fig. 7B, J; Fig. S2D).

Fig. 7.

Hinge structure. Micrographs were obtained under SEM (A–E) or TEM (F–K). (A) Longitudinal section with adaxial epidermis on the upper side. Note the sclerenchyma-like cells (Sc) on the abaxial side. (B) Longitudinal section of adaxial epidermis (Ep) with thickened outer periclinal wall (asterisk), and an underlying parenchyma cell (Pa), both with rounded pit apertures (arrowheads) in anticlinal walls. (C) Abaxial parenchyma cell adjacent to sclerenchyma-like cells, with slit-shaped pit apertures (arrowhead). This cell is located in the third cell layer counted from the abaxial surface. (D) Sclerenchyma-like cells of abaxial epidermis (Ep-Sc) and underlying cells (Sc), all with slit-shaped pit apertures (arrowheads). (E) Strips formed during oblique longitudinal sectioning of the secondary wall of sclerenchyma-like cell, obtained from the location similar to that framed in A. (F–H) Walls of abaxial epidermis in cross-section. (F) Fragment of anticlinal walls with electron-dense inclusions and ‘joints’ through the compound primary wall. (G) Electron-dense inclusion in the secondary wall. (H) Outer periclinal wall with an inclusion (arrowhead) and very thin cuticle (Cu). (J) Fragment of adaxial epidermis with uniformly thickened outer periclinal walls. (K) Close-up of outer periclinal walls and compound anticlinal wall of adaxial epidermis. Cu, cuticle; Ep, adaxial epidermis; Ep-Sc, abaxial sclerenchyma-like epidermis; Pa, parenchyma; PW, compound primary wall comprising middle lamella and primary walls of the adjacent cell; Sc, sclerenchyma-like cells; SW, secondary wall.

All the cells have secondary walls deposited over thin compound primary walls (Fig. 7F, J). The orientation of cellulose fibrils in the walls also changes from the abaxial to adaxial surface as manifested in pit morphology and the cellulose MFA measured in macerated cells using polarized light microscopy (Fig. 6C). All the hinge cell walls have numerous simple pits although they differ in aperture shapes. Pits in sclerenchyma-like cell walls have slit-like apertures elongated in the transverse direction (Fig. 7D). Such aperture shape and the MFA (mean = 87·4°; SD = 7·3°; number of measurements n = 19) indicate that cellulose microfibrils in the secondary walls are highly ordered and their orientation is nearly transverse. Pit apertures in abaxial parenchyma cell walls are similar to those in sclerenchyma-like cells (Fig. 7C) but in adaxial parenchyma they are rounded and oblique (Fig. 7B). Accordingly, the MFA in the parenchyma is much less uniform than in other hinge tissues (mean = 75·1°; SD = 41·9°; n = 10), i.e. microfibril orientation ranges from nearly transverse to oblique. The former is probably in abaxial parenchyma cells while the latter is in adaxial cells. In adaxial epidermis, the anticlinal cell walls have pits with rounded apertures (Fig. 7B), suggesting microfibril orientation similar to adaxial parenchyma cell walls, while in the outer periclinal walls MFA measurements show that microfibril orientation is uniformly longitudinal (mean = 6·1°; SD = 3·1°; n = 27).

The specific trait of abaxial hinge cells is the non-compact wall structure: numerous layers and strips become separated when the wall is cut at an oblique angle with the razor blade (Fig. 7E). The non-compactness is manifested probably also by the presence of inclusions between the wall layers (Fig. 7F–H). AFM examination of hinge cross-sections in the dry state (Supplementary Data Fig. S4) shows that both sclerenchyma-like cell walls and outer periclinal walls of adaxial epidermis exhibit a layered structure, apparent in topography and amplitude images. However, in walls of sclerenchyma-like cells, the layer cross-section has a rather rough surface and layers are not closely attached one to the other, but locally are separated by free spaces (Fig. S4A). This is unlike the outer periclinal walls of adaxial epidermis, which are more smooth and compact (Fig. S4E, F). At least some portions of sclerenchyma-like wall layers seem to be built of stripes, probably to separate during sectioning (Fig. S4B, D). The two epidermal layers also differ in that the abaxial epidermis is covered with a thin cuticle (Fig. 7H), probably discontinuous (in dry state gaps open between cells; Fig. 4A), while the adaxial epidermis is covered with cuticle (Fig. 6K) with no discontinuities either in the dry or the wet state (Fig. S2A–D).

The blade structure (Fig. 8) is more uniform than that of the hinge. Only the adaxial and abaxial epidermis are distinguished by cell morphology or wall structure: abaxial epidermal cells have a relatively small diameter (Fig. 4D) while the adaxial epidermis is distinguished by the already mentioned swelling outer periclinal walls (visible also under TEM; Fig. 8C) and also by the smooth inner wall surface (Fig. 8A–C). Cellulose microfibrils in the outer periclinal walls of the adaxial epidermis are well aligned and of nearly longitudinal orientation (mean MFA = 6·7°; SD = 3·6°; n =  14). In all blade cells except for adaxial epidermis, the secondary walls form flange-shaped ingrowths oblique with respect to the cell axis (Fig. 8A, B, D). Their presence obstructed MFA measurements. Anticlinal walls of the adaxial epidermis have numerous simple pits with rounded apertures (Fig. 8B, C). Pits are present but much less often also in other tissue cell walls (Fig. 8B, E). The outer periclinal walls of abaxial epidermis are thickened (Fig. 8E). A thin but distinct layer of cuticle covers the adaxial surface while a less distinct and thinner cuticle is present on the abaxial side (compare Fig. 8C with Fig. 8E, F).

In summary, within the hinge the tissue fraction occupied by cell walls, cell wall thickness and MFA decrease gradually from the abaxial to adaxial side. However, the cell walls on the adaxial surface are again thickened and more compact than on the abaxial side. This spatial variation of tissue traits probably corresponds to different deformation that takes place due to wetting. The blade tissues are much more uniform, and only the two epidermis tissues are distinguished. We next analysed the chemical composition of the bract cell walls.

Differential hinge tissue strain is not related to cell wall composition but cell wall composition in hinge differs from blade

We examined cell wall composition by applying confocal Raman imaging to unstained hand-cut cross-sections in the wet state (Figs 9 and 10; Figs S5 and S6). We compared Raman spectra for different measurement spots at cell walls (marked in Fig. 9B, E) and, based on the literature, selected four Raman bands (Notes S1; Figs S5 and 6) to generate images on the basis of band integration (Fig. 9A, D): band I centred at 1097 cm^–1 (assigned to cellulose); band III at 1600 cm^–1 (lignin); band IV at 2895 cm^–1 (cellulose); and band V at 2943 cm^–1 (polysaccharides including hemicellulose and pectin). A signal from band II, centred at 1458 cm^–1, usually assigned to cellulose and lignin (Gierlinger and Schwanninger, 2006), was specifically enhanced in the phloem spectrum (5 in Fig. 9C and Fig. S5). We assume that this enhancement is due to the presence of callose, since aniline blue staining showed that it occurs exclusively in phloem (data not shown).

In the hinge section the most heterogeneous distribution over the Raman image is that of lignin (Fig. 9A). Lignin is present in all the walls except for phloem (band III in spectrum 5, Fig. 9B, C;Fig. S5) but signals from xylem (spectrum 4, Fig. 9B, C;Fig. S5) and from the adaxial (spectrum 7) and abaxial (spectra 1 and 2) epidermis are increased. The hemicellulose and pectin signal (band V) has the most homogeneous distribution among the bands analysed (Fig. 9A). The stronger signal comes from compound periclinal walls of the adaxial epidermis and parenchyma where the increased content of hemicellulose/pectin may be crucial for adherence between the walls of cells that differ much in strain. The presence of hemicellulose xylan epitope LM10 and xyloglucan LM25 (Fig. S7C, D and Fig. S7F, G, respectively) as well as pectin, galactan epitope LM5 and homogalacturonan LM19 (Fig. S8B, C and Fig. S8E, F, respectively), in cell walls of all the hinge tissues was further confirmed by immunolabelling. Signals from the two cellulose Raman bands I and IV are heterogeneous, especially in abaxial sclerenchyma-like cell walls (Fig. 9A), but they depend on wall orientation. The intensities of cellulose bands I and IV are strongly dependent on the orientation of the polymer chains, i.e. microfibril orientation, with respect to polarization of the incident laser light (Edwards et al., 1997; Gierlinger et al., 2010). Integrating over band I gives the highest signal for cellulose chains orientated at a polarization angle of 0° (i.e. along the x-axis in Fig. 9) whereas integrating over band IV was for 90° (along the y-axis). In the hinge section under polarization angle of 0°, the strongest signal is in the walls orientated along x-axis for band I, and along y for band IV, and the higher the band I signal in the wall, the lower that for band IV. The signal distribution is reversed when the orientation of the section with respect to laser polarization direction is changed by 90° (compare right and left panels in Fig. 10A). This suggests nearly transverse cellulose orientation in sclerenchyma-like cell walls, in agreement with MFA measurements. Less apparent orientation-dependent changes in the signal take place in parenchyma cell walls orientated along the x- or y-axis (Fig. 10A). Unlike in other walls, in the adaxial epidermis the cellulose signals for both x- and y-orientated walls are similar, and the signal is virtually not orientation-sensitive: the change in band I signal accompanying the change in polarization direction is very similar to the change in band IV signal (compare right and left panels in Fig. 10A). This is similar to cross-sections of layer S2 of xylem fibre walls (Gierlinger et al., 2010). Both in adaxial bract epidermis and in xylem, cellulose orientation is nearly perpendicular to the xy plane, i.e. longitudinal, again in agreement with MFA measurements.

Because the isolated sclerenchyma-like abaxial epidermis and subepidermis of the hinge perform some hygroscopic movements, we compared them in more detail. Both band integration images (Fig. 9A) and plots of signal accumulated along transects (green and grey line segments in Fig. 10B, C) show that the main difference is in the stronger lignification of epidermal than subepidermal cell walls. In both layers lignification decreases toward the cell lumen, probably related to lower lignification of secondary than compound primary walls (Donaldson, 2001). Also in both cell layers the intensity of the cellulose band I signal, but not the lignin signal, is strongly orientation-dependent (compare right and left panels in Fig. 10C).

In blade sections the only tissue-specific signal is that of lignin (band III), restricted to walls of abaxial epidermis (Fig. 9D–F;Supplementary Data Fig. S6 – spectrum a). Raman images for other polysaccharide bands (I, IV, V) are much more homogeneous than those of the hinge (compare Fig. 9D and A). The presence of hemicellulose epitope LM10 in secondary cell walls of all the blade tissues was further confirmed by immunolabelling (Fig. S7A, B). However, hemicellulose epitope LM25 and pectin epitopes LM5 and LM19 were detected mainly in vascular bundles and adjacent parenchyma cells (Figs S7E and S8A, D). Interestingly, we were not able to identify any signal specific for the strongly swelling inner portion of outer periclinal walls of adaxial epidermis (marked with a dot in Fig. 4D). Only the orientation-dependent cellulose signal (band IV) was relatively stronger (Fig. 8D–F;Fig. S6 – spectrum d) but weaker than the signal from the outer portion of the same wall, similar to the band I cellulose signal (Fig. 9F;Fig. S6 – compare spectra d and c). This, together with the fact that the swelling wall portion shrinks to a very thin layer in dry state, implies low cellulose content in the swelling layer and probably similar cellulose microfibril orientation in both the swelling and the non-swelling outer wall portions.

Since in the blade the MFA could not be measured for cell walls with flange ingrowths, we tried to get additional information on cellulose microfibril orientation by comparing images for the two orientation-sensitive cellulose bands (I and IV). However, the signal distribution differs between individual blade sections. In some sections signals are orientation-dependent (compare band I and IV signal in periclinal or anticlinal walls in Fig. 9D) while they are not in others (compare images of the other section obtained for different orientations with respect to laser polarization in Fig. 11). These differences may be related to the helical orientation of the flange wall ingrowths. Cellulose microfibrils are most likely orientated along the ingrowths (Talbot et al., 2007) and the ingrowths may be included in some measurement spots while others include only the remaining cell wall layers. Concluding, in the adaxial blade epidermis cellulose microfibrils are nearly longitudinal, while in other cell walls they are probably rather steep but the orientation differs between wall layers and between sections, unlike in the hinge cells.

In summary, the specific trait of hinge cell wall composition is the presence of lignin in sclerenchyma-like cells, parenchyma and adaxial epidermis. However, the distribution of lignin is heterogeneous: highest in epidermal cell walls and lower in inner cell walls. The non-uniform distribution of signals from other bands results mainly from their orientation dependence. Therefore, differences in hinge tissue strain due to wetting are probably not related to differences in cell wall composition. The blade is in turn distinguished by the tissue-specific lignin signal, virtually restricted to the superficial abaxial walls. Other signals are uniformly distributed.

DISCUSSION

Scarious involucral bracts of H. bracteatum perform hygroscopic movements that lead to bract bending toward or away from the capitulum centre, depending on water status of cell walls. We identified two bract regions which react in a different way to water status changes. The hinge is characterized by a large difference between longitudinal strains at opposite bract surfaces taking place due to wetting, which results in a dramatic curvature change that drives movement of the blade. At the blade some transverse strain occurs due to wetting but differences in strain at the opposite surfaces are negligible and, in consequence, shape changes are much less pronounced. These two regions exhibit traits of cell wall structure and composition that can be related to the region-specific deformation.

Unique construction of hinge – the H. bracteatum bract actuator

In search of active and resistance tissues at the hinge of H. bracteatum bract we discovered that it does not work like a bi-layered structure with well-defined active and resistance tissues, as reported for many other hygroscopically bending organs. Movements similar to the intact hinge are also generated by the hinge depleted of abaxial sclerenchyma-like cell layers or the hinge from which adaxial epidermis was removed. Thus, the active and resistance parts of the actuator are not equivocally defined. Also parenchyma is not simply an intermediate unit optimizing the movement, like the soft parenchyma between active and resistance parts of wheat awn (Fratzl et al., 2008), but actively participates in movement generation, as in the sesame capsule (Shtein et al., 2016). Therefore, we propose that the hinge construction, instead of comprising the active and resistance tissues separated by passive parenchyma, is a layered structure characterized by an abaxial–adaxial gradient of tissue traits, such as the fraction occupied by cell walls, wall thickness, compactness and cellulose fibril orientation. Thus, as long as several tissue layers are present the gradient of mechanical stress across the layers, with highest compression on the abaxial surface, is generated due to wetting, which leads to a strain gradient causing bract bending at the hinge. Differences between strains measured for opposite hinge surfaces after tissue removal are lower than for the intact hinge although shape changes are similar. This is because the thinner the organ, i.e. the closer the adaxial and abaxial surfaces, the smaller is the difference in strain between the surfaces sufficient for organ bending. Presumably, also the necessary stress generated in cell walls by wetting may be lower than in intact hinge and the smaller structural differences between adjacent tissues are enough to generate the necessary stress. Such hinge construction enables movements even after mechanical damage of its surface. This is probably advantageous for H. bracteatum reproduction as the bracts need to function from flowering to achene dispersal.

The bract surface strain is anisotropic. In particular, on the abaxial surface upon wetting the cells extend (longitudinal strain) by as much as over 20 % while the transverse strain is much lower or nearly zero. Such anisotropic cell wall deformation during wetting is probably due to swelling of a polysaccharide matrix between anisotropically arranged cellulose microfibrils of the secondary cell wall. The increase in microfibril spacing upon hydration has been shown for collenchyma of celery Apium graveolens petioles where pectins are the interfibrillar matrix (Jarvis, 1992; Kennedy et al., 2007). Theoretical considerations (Burgert et al., 2007; Fratzl et al., 2008) show that isotropic swelling of a matrix filling the space between aligned fibres of much higher or infinite stiffness results in anisotropic stress generation: tension appears along the long cell axis for low MFA values (steeply orientated microfibrils) and compression for higher values. The threshold MFA value for compressive stress generation depends on the tissue compactness: if a slight cell torsion is possible lower MFA may lead to compressive stress. Accordingly, in conifers the nearly zero MFA of cell walls of the compact normal wood and higher MFA in walls of rounded cells of compression wood are enough to generate compressive stress in the latter that drives branch movements (Burgert et al., 2007; Burgert and Fratzl, 2009b). In H. bracteatum bract hinge, abaxial sclerenchyma-like cells, which undergo the highest longitudinal strain due to wetting, fulfil these requirements for compressive stress generation with excess. In their thick secondary walls cellulose microfibrils are highly ordered and nearly transverse to the cell axis (high MFA), i.e. to the direction of maximal extension. The cells are rounded, and between them there are some intercellular spaces. Although the lignin content at contacts between adjacent cells is high, unlike the compression wood (Burgert et al., 2007), the spaces may facilitate some torsion. The traits crucial for compressive stress generation, such as rounded cell shape, high MFA and thickened walls, are also expressed but to a lesser extent in abaxial parenchyma. However, towards the adaxial epidermis, parenchyma cells are larger, their walls are thinner and MFA is probably lower. Finally, the adaxial epidermis exhibits traits specific for the resistance tissue of bi-layered actuators: the thick outer periclinal walls have aligned and almost longitudinal cellulose microfibrils (MFA nearly zero), and the tissue is compact. Such a large difference between the two actuator sides is the same as in the bi-layered bending apparatus of Geranium pusillum awn (Abraham and Elbaum, 2013). However, in pinecone scale, which is also bi-layered, the smaller difference in MFA between resistance sclerenchymatous fibres and hygroscopically active sclereids (approx. 30° vs. 75°) is enough to drive hygroscopic movements (Dawson et al., 1997; Burgert and Fratzl, 2009b), and active tissue cells of the scale elongate as much as 10–30 % while nearly no extension or slight contraction takes place on the opposite side (Reyssat and Mahadevan, 2009). Of note, both the G. pusillum awn and the H. bracteatum bract are ‘lighter’ structures than the pinecone scale.

The gradually changing structure of the H. bracteatum hinge may also explain the somewhat puzzling trait of its structure: if the bract hinge were the bi-layered actuator, the outer periclinal cell walls of the single layer of adaxial epidermis (putative resistance tissue) would have to resist the large tensile forces generated due to swelling of thicker secondary cell walls in two or more layers of the sclerenchyma-like cells (putative active tissue). In this respect the H. bracteatum hinge resembles the involucral bract of S. elegans (Oriani and Scatena, 2009). In this latter species hygroscopic abaxial epidermis cells also have very thick walls, especially the outer periclinal wall, while parenchyma and adaxial epidermis cell walls are thin. It suggests that also in S. elegans the hinge structure is not bi-layered. The unusual construction of organ actuator was shown also for the sesame capsule where it is graded rather than simply bi-layered (Shtein et al., 2016). There the bending direction depends on the local contribution of two differently reinforced fibre sublayers in the resistance tissue that changes gradually along the capsule circumference.

Other structural adaptations of the H. bracteatum hinge to the movements may be discontinuities between the abaxial epidermal cells and cuticle. They probably enable fast water penetration into the bract, further facilitated by the presence of numerous pits in hinge cell walls. The cuticle discontinuities may also be crucial for fast water evaporation directly from the abaxial hinge surface, facilitating a rather quick response to a decline in humidity of the surrounding environment. The discontinuities in active tissue surface may be an important trait because they are present also in the hygroscopic abaxial epidermis cells of S. elegans bracts where channels (referred to as pits) in outer periclinal walls were reported (Oriani and Scatena, 2009). Water penetration of the abaxial H. bracteatum walls may be assisted by a non-compact stripe-like wall structure. The porous structure (nano-sized gaps) was postulated to accelerate water absorption in active tissue of T. turgidum awns (Elbaum et al., 2008). By contrast, the adaxial epidermis of the H. bracteatum bract has compact walls and its surface is covered by thicker cuticle. Water movement between its cells is facilitated by pits but the surface absorbs less water than on the abaxial side.

We have not found any cell wall component in the bract hinge, the distribution of which would correspond to different cell wall extension due to wetting. However, also in pinecone scale and wheat awn, the cell wall composition in the active and resistance tissues is similar and the difference in MFA is regarded as sufficient to drive hygroscopic movements (Dawson et al., 1997; Elbaum et al., 2007; Burgert and Fratzl, 2009b; Dunlop et al., 2011). In both cases hemicelluloses were pointed as the swelling agent, which may also be the case in the H. bracteatum bract hinge. Raman spectroscopy and immunolabelling showed that hemicellulose and pectin are most homogeneously distributed in hinge cell walls, and the main differences between walls are in lignin and orientation-sensitive cellulose signals.

Of note, the H. bracteatum bract hinge, the region undergoing the strongest deformation, is also the most lignified. Cell walls in both abaxial sclerenchyma-like and adaxial epidermis are strongly lignified while the extent of lignification gradually decreases inwards. On the abaxial side the higher the lignin content, the stronger the tissue deformation due to wetting. Lignin is present also in active and resistance tissues of wheat awns (Elbaum et al., 2007), pinecone scales (Dawson et al., 1997; Dunlop et al., 2011) and thick walls of hygroscopic abaxial epidermis cells of S. elegans bracts (Oriani and Scatena, 2009). This common increased lignification of the actuating tissues may have two functions. Lignification may be important for structural protection of the walls that undergo numerous, repetitive and quite fast shape changes. Another lignin function may be in movement generation: recent atomistic models show that lignin can assist swelling of the secondary cell wall (Kulasiński, 2015).

Structure of the H. bracteatum blade is related to its deformation due to wetting

Adaxial and abaxial surfaces of the central portion of the bract blade do not differ significantly in deformation due to wetting, and as a result its shape is not changing. Instead, it thickens and widens by several to dozen per cent, and slightly shortens. At the cellular level, the width and thickness of all the blade cells increase and cells shorten due to wetting. The adaxial epidermis is the most deformed tissue (the reverse situation in comparison with the hinge). The observed uniform strain is probably related to the low MFA in all the tissue walls but not their similar composition. In particular, lignin distribution in blade walls is tissue-specific, i.e. restricted to the abaxial epidermis, while the strain of this epidermis surface is like that on the adaxial side lacking lignin. Lignification of the abaxial epidermal walls may have a protective role as these walls are exposed when the bracts close over the capitulum.

In the adaxial epidermis the observed cell shortening and widening can be explained directly by the anisotropic reinforcement of the outer periclinal walls (low MFA). The profound swelling of the inner layer of these walls may to some extent participate in the necessary stress generation, similar to the inner wall layer of hygroscopically active cells of abaxial bract epidermis in S. elegans (Oriani and Scatena, 2009) or the G-layer of tension wood fibres, G-fibres of Trifolium pratense roots or active tissue fibres of seed capsules in Acanthoideae (Witztum and Schulgasser, 1995; Gierlinger and Schwanninger, 2006; Schreiber et al., 2010). The swelling wall of bract epidermis, like the G-layer, consists mainly of cellulose. In the G-layer the MFA is very low and the cellulose is highly crystalline, which may also be the case of bract blade epidermis. However, in the bract blade, swelling is mainly into the free space of the cell lumen, which may diminish the stress generation effect. Of note, the outer periclinal wall of adaxial epidermis was the only hygroscopic element of the H. bracteatum bract well recognized in early studies (Uphof, 1924; Brauner and Rau, 1966).

The mechanism of shortening and widening of the abaxial epidermis and parenchyma cells is not as clear. The structure of their cell walls is more complex due to the presence of unique flange ingrowths of the secondary wall. Nishikawa et al. (2008) postulated that these cells represent a new tissue type. Mechanical function is attributed to a similar type of thickenings in transfer cells (Andriunas et al., 2013) with strengthening probably along the flanges, since cellulose microfibrils in the ingrowths are aligned (Talbot et al., 2007). The oblique orientation of the flange ingrowths and probably more steep orientation of cellulose microfibrils in remaining wall layers (as suggested by Raman images) may explain the observed cell strain. However, except for the mechanical role, the ingrowths may enable effective wetting of the blade tissues, walls of which are not as densely pitted. Modelling and observations show that in the secondary xylem, roughness of the inner walls, such as warts or helical thickenings, significantly increases the wettability of the cell lumen (Kohonen and Helland, 2009).

CONCLUSIONS

At the hinge of the H. bracteatum bract, the actuator of this hygroscopically moving organ, longitudinal strains at opposite surfaces differ profoundly. This difference results in changes of hinge curvature due to wetting or drying that drive passive displacement of distally located bract blade. The blade in turn undergoes nearly uniform strain on both surfaces and thus only minute shape changes. Differential strain at the bract hinge is due to an adaxial–abaxial gradient in structural traits of tissues and cell walls. The hinge is built of sclerenchyma-like abaxial tissue, parenchyma and adaxial epidermis with thickened outer walls. The cell wall composition of all the tissues is rather uniform but the tissue fraction occupied by cell walls, cell wall thickness, compactness and cellulose microfibril orientation change gradually from the abaxial to adaxial hinge surface. Moreover, dissection experiments show that the presence of part of the hinge tissues is enough for movements. Thus, the hinge is a structure comprising gradually changing tissues, from highly resisting to highly active, rather than a bi-layered structure with distinct active and resistance parts.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Figure S1: side views of surfaces approximating cell vertices of hinge epidermis patches in the dry and wet state. Figure S2: bract hinge: adaxial and abaxial epidermis in the dry or wet state. Figure S3: bract blade: adaxial and abaxial epidermis in the dry or wet state. Figure S4: AFM images of hinge cell walls. Figure S5: fitted Raman spectra for the hinge. Figure S6: fitted Raman spectra for the blade. Figure S7: immunolabelling with hemicellulose epitopes. Figure S8: immunolabelling with pectin epitopes. Notes S1: assignment of Raman bands to cell wall components. Table S1: raman bands used in the analysis of bract cell wall spectra.