Combined X-ray diffraction tomography imaging of tension and opposite wood tissues in young hybrid aspen saplings

Combining spatially localized X-ray diffraction (XRD) with X-ray microtomography (XMT) enables the mapping of the micro- and nanoscale structures simultaneously. The combination of these methods results in a powerful tool when considering the structural studies of hierarchical materials, allowing one to couple the relationships and connections of the structures at various scales. In this study, XMT was used to map the anatomy and cellular structures in 3D in tension and opposite wood with 1.5 µm resolution, while XRD was used to determine the cellulose crystallite widths and microfibril orientations with 100 µm spatial resolution within the same tissues. Tension wood (TW) has an important biological function with clearly distinct properties to opposite (OW) and normal wood, e.g. differing cellular structures with a higher cellulose content. This is the first study of very young hybrid aspen saplings (1-month-old) using the combined diffraction tomography method. The TW tissues could be identified from the OW tissues based on both the XMT and XRD results: TW had a higher average size of the cellulose crystallites and smaller mean microfibril angles (mMFA) than those in OW. With the XRD data, we were able to reconstruct the images of the cross sections of the saplings using the structural parameters (cellulose crystallite width and mMFA) as contrast mechanisms. As far as the authors know, there are no previous studies with images on any TW samples using the XRD-based contrast. Home laboratory bench-top set-up offers its advantages for these studies, considering the number of samples characterized, time-dependent studies and larger field of views.


Introduction
The hierarchy of structures, forming from the atomistic to the macroscopic scale, is determinative for all biomaterials. Covering the different size scales of the various structures is crucial due to the definitive nature of the structure itself: it dictates the materials' function and properties. As a key example, the extraordinary mechanical properties of wood arise from the interplay of crystalline and amorphous order and orientation of the cellulose chains on the nanoscale, building up the elementary cellulose microfibrils and further reinforced matrix-fibril structures together with lignin and hemicellulose. These main components with the addition of other, minor compounds organize into various layered cell wall structures, which then define distinct cells and cellular tissues and finally the macroscopic stem structure. Wood structure has been extensively studied with physical, biological and chemical methods, but the structural details still offer an intriguing puzzle due to the complexity and the sensitivity of the biological living organism.
Various X-ray-based methods offer significant advantages for solutions to these questions: they are in principle non-invasive, and by using X-ray scattering and diffraction (XRD), the structures of the studied material can be revealed down to the atomistic scale. Besides the origin of the material properties, the detailed knowledge of the wood structural organization shed light on the fundamental questions on plant growth.
Reaction wood is a biologically important, special type of tissue generated in stems and branches as an adaptation to mechanical and gravitational stimulus, e.g. strong wind loads, leaning or tilting directed on the stem and branches. Deciduous trees generate tension wood (TW) on the upper side of their leaning stem, which pulls the stem upward using tension stress. The TW generation is a widely existing property of the stem growth of trees and similar fibre properties are shared with the bast fibres in, for example, flax and hemp (Mellerowicz and Gorshkova 2012;Gorshkova et al. 2018).
Tension wood xylem tissues have distinctive properties due to the specific chemical compositions, and cellular and cellulose crystallite structures compared to normal wood (NW) (Gardiner et al. 2014;Chang et al. 2015). In TW, the average width of cellulose crystallites is larger (Müller et al. 2006;Clair et al. 2011;Viljanen et al. 2020) and the cellulose content is higher (Foston et al. (2011)). In many hardwood species, these changes often originate from the formation of a cellulose-abundant gelatinous layer or the G-layer, often replacing parts of the other layers in the secondary cell wall of the wood fibres (Gardiner et al. (2014)). The wood tissue on the other side of the stem, directly opposite to the TW side, is called opposite wood (OW).
The special properties of TW make it highly interesting for various scientific purposes and technological applications. Structural studies on TW offer direct insights into the wood structure-properties relationships: by comparing the results on TW structure (cellulose crystallites, amorphous matrix, cell wall characteristics) to NW, conclusions on the structure-function relations can be made.
In addition, the differences in reaction wood formation in stems play a vital role regarding the lignocellulosic yield from the crops (Brereton et al. 2012;Ding et al. 2012) and with that reaction wood may also help in solving the recalcitrance problem (Sawada et al. 2018). Additionally, the possibilities of TW for nanocellulose production have been recently considered (Jonasson et al. 2021).
Reaction wood has been characterized intensively using microscopy and chemical methods (e.g. Gardiner et al. 2014). Microscopy methods offer results in 2D, whereas full 3D information is reached by using tomography methods. X-ray microtomography (XMT) has proven to be an efficient tool for the quantitative analysis of the anatomical structure of various plant and wood tissues (Stuppy et al. 2003;Brodersen and Roddy 2016;Van den Bulcke et al. 2019). Regarding this background, the TW formation and structures have been studied relatively little using the possibilities of XMT. To our knowledge, there exist only a few previous examples of these kinds of studies: Brereton et al. (2015) studied the structure of TW tissue in willow saplings by using microtomography. They conducted the measurements using a resolution of 9 µm, which enabled the analysis of vessel structures in more detail. In our study, we aimed to use significantly higher resolution and thus, very young hybrid aspen saplings were selected. The diameter of the studied saplings was limited to below 2 mm, which enabled the usage of high-resolution tomographic imaging with 1.5 µm resolution. The chosen imaging parameters allowed the coverage of the entire cross section of the single saplings. Another advantage of studying these juvenile saplings is the ability to cover the possible start pattern of the TW formation.
The highest available spatial and time resolution can be achieved by using synchrotron light sources. Schmelzle et al. (2021) were able to study in situ the deformations of TW and OW samples as a function of tensional stress in 3D by using synchrotron X-ray microtomography. However, regarding synchrotron techniques on the biological samples, beam damage might be a severe issue (Schmelzle et al. 2021;Storm et al. 2015). The other limiting factor, regarding the use of the synchrotron techniques, is that the beam time availability is highly limited, both considering the number of samples and time invested. Thus, it is a great advantage that the synchrotron studies can be complemented by using home laboratory bench-top devices. For example, home laboratory experiments enable a wider field of view, larger sample series and longer measurement times when considering experiments including time series or strictly scheduled sampling, e.g. the case of living plants. For these reasons, it is highly important to develop novel home laboratory-based methods for especially plant and wood studies.
In this study, young, 1-month-old hybrid aspen saplings were imaged using simultaneous X-ray diffraction and microtomography utilizing a laboratory benchtop device. The microtomography results were obtained with a resolution of 1.5 µm and the spatial resolution in the localized diffraction was 100 µm. This study represents the first structural study on reaction wood tissues in young wood saplings examined on the micro-and nanometre scales using a simultaneous combination of high-resolution tomography and XRD line scans. Structural details on TW and OW sides were determined both at the nano-and micrometre levels and the relationships of these two structural scales were mapped.

Samples: growth and tilting procedure of the saplings
Four T89 hybrid aspen saplings (Populus tremula x tremuloides) were selected for the study. The trees were in vivo saplings grown at the Viikki Plant Growth Facilities, University of Helsinki. Around 1 month old, the plants were transplanted from the in vivo pots into the soil in a table-top-greenhouse (see Fig. 1). After 22 days in a greenhouse, three of the saplings were tilted to 45 degrees for 16 days: two of them using a tilted stage (samples 1 and 2) and one of the saplings using leaning (sample 3). One of the saplings (sample 4) was left to grow in a straight position, i.e. it was not exposed to physical tilting or leaning. All the samples were exposed to non-isotropic lighting conditions because they grew in a table-top greenhouse in front of a window (opening to the south direction). Thus, sample 4 can be considered as a reference sample of the possible phototropic effect on the samples. All of the saplings and their growing positions and conditions can be seen in Fig. 1. The tilting conditions lasted for 16 days, after which all the samples were cut and prepared from the saplings.

Pretreatments and drying protocol of the samples
To achieve the highest possible resolution for the microtomography images, the wood samples should be studied in their dry state as the contrast between the cell wall materials and the air is much higher than that of the cell wall materials and water. Due to these reasons, the studied samples were freeze-dried prior to any XRD or µCT measurements.
Using a razor blade, the samples were cut from the stems at a height of 10 cm from the soil. The stem samples were approximately 5 mm long cylinders with a diameter of 2 mm, including all the different tissues from pith to bark within the stem. The assumed OW side of the samples was marked with a slight incision.

Experimental
All the scattering and tomography experiments were conducted using a custombuilt combined X-ray scattering-X-ray microtomography set-up of the University of Helsinki (Suuronen et al. (2014)). The combined set-up consists of two independently operated X-ray devices. The microtomography device is based on a conebeam geometry of the X-ray scanner (Phoenix|x-ray Systems and Services GmbH, Waygate Technologies) with microfocus X-ray tube and CMOS flat-panel detector (C7942SK-25, Hamamatsu Photonics, Japan). The scattering device is based on a molybdenum (Mo) anode source (IμS, Incoatec GmbH, Germany) with a pencil beam and an area detector (Pilatus 1 M, Dectris Ltd, Switzerland).
A shared, motorized sample holder stage was used to ensure the spatial alignment of the sample between the scattering and the tomography modalities. The calibration of the tomography coordinates to the scattering coordinates was conducted by measuring a silver behenate sample. Due to this alignment, regions of interest for localized micro-diffraction could be selected directly from the tomographic reconstruction slice. In this experiment, the tomographic reconstruction was used to align the sapling cross section so that the line scans with 100 µm step size included either tension wood xylem or opposite wood xylem (see Fig. 2).

Line scanning WAXS
The wide-angle X-ray scattering (WAXS) patterns were measured with 100 µm step size (i.e. the line scans), using perpendicular transmission geometry with a wavelength ( ) of 0.729 Å (Mo K-α). The measurement time for each WAXS measurement was 10 min and the scattering data was collected onto a Pilatus 1 M area detector. The q-scale calibration was conducted by using a silicon standard sample and the instrumental broadening was determined to be 0.03 Å −1 . The data is presented as a function of the length of the scattering vector (q): where is half of the scattering angle.
The data were corrected for air scattering, and in addition, the geometrical, polarization and transmission corrections were applied. The transmission values were set for the samples to be 1.1 (data/background) based on both the validation of the air scattering correction and the previous transmission experiments on hybrid aspen samples having similar thicknesses .
From the WAXS data, the average crystallite width and the microfibril orientation were determined. To obtain the radial data, the WAXS patterns were integrated from the 40° wide sectors around both maxima of 200 reflections of the cellulose Iβ as observed on the area detector, dictated by the fibre symmetry of the 2D scattering pattern. To improve the statistics of the data, these two integrated intensity curves were summed together.
The radial scattering pattern around the q range 0.75-1.95 Å −1 (corresponding 2θ range 5-13°) was modelled by using four Gaussian curves (each one for cellulose reflections 110, 110, 102 and 200 , respectively). The amorphous contribution Approximative scan lines (1…13) for the localized WAXS measurements for sapling 1 after the alignment of the sample according to the tomography reconstruction so that the scanned areas include either pure TW or OW tissue as much as possible. The corresponding analysis results based on the WAXS patterns for each of these line scans are given in Table 2 was modelled using an experimentally measured scattering curve of a lignin sample (Andersson et al. (2003)). Examples of this fitting procedure are presented in the supplementary figures S1 and S2.
The average crystallite width (B) was determined from the FWHM of the 200-reflection using the Scherrer equation with the assumption of both 200-reflection and instrumental broadening being Gaussians and by setting the shape factor K = 0.9: where INST was the instrumental broadening (0.03 Å −1 ).
The WAXS patterns were integrated around the cellulose 200-reflection corresponding to the q range 1.45-1.55 Å −1 (2θ = 9.6-10.3°) to obtain the azimuthal data. The mean microfibril angle (mmfa) was analysed from these azimuthal curves by using T-parameter values determined as mmfa = 0.6 T (Cave 1966). The value of the T-parameter is simply obtained by fitting a tangent to the azimuthal integral curve of the 200-reflection and then determining the distance from the origo (zero angle) to the intersection of the tangent and the polar angle axis.
The decision to use the more robust T-parameter analysis instead of a more validated MFA analysis method (see, for example, Rüggeberg et al. (2013), Sarén et al. (2006)) was made because in these juvenile cross-sectional samples the beam position related to the cell wall and the cell shapes highly varied across all the samples. Thus, it was observed, that the differences between the samples and different locations could be compared most reliably by using the robust T-parameter analysis. It must be noted that the given mmfa values are for comparative purposes inside this data set, and they are not meant to be read as absolute values of the MFA values in these saplings.

X-ray microtomography (XMT)
X-ray microtomography measurements were conducted with a current of 240 μA and a voltage of 60 kV for the X-ray tube. A total of 10 transmission images of 500 ms exposure were averaged to produce one projection image.
Pixel sizes of 1.2-1.5 µm resulting in a field of view of 1.3-1.7 mm (with a 2 × 2-pixel binning) and a pixel size of 0.8 µm and a field of view of 1.8 mm (without binning) were used. 400-800 projection images were taken over the 360-degree. The parameters for the individual scans of the samples are given in Table 1. The reconstruction of the tomographic data was done with GE phoenix datos|x software using a back-projection algorithm.

X-ray diffraction tomography (XDT)
Sample 1 was measured by X-ray diffraction tomography. A total of 31 rotation steps were taken over 180° (step size of 6°). At each rotation step a line scan was performed with a step size of 100 µm by measuring the scattering pattern for 120 s at each step (a total of 17 line scans), which made the total duration of the whole scan to be 17.5 h. Thus, in the XDT scan, a total of 17 × 31 XRD patterns were measured. From each of these scattering patterns, any XRD-based parameter (chosen intensity, peak position or width) can be determined and used as a contrast for the XDT image. The XDT image was then computed by MATLAB (MathWorks Inc., USA) by using inverse filtered Radon transform (with linear interpolation and Hamming filter). The computed XDT image represents the scanned sample cross section.
In this study, two different contrasts were used to construct the XDT images: 1) the intensity of cellulose 200-reflection (reflecting the cellulose content in the sample) and 2) the MFA determined based on the T-parameter value (reflecting the orientation of the cellulose microfibrils).

Tomography imaging
As can be seen in Fig. 3, the full cross-sectional width of the saplings varied significantly even though all the samples were the same age and grown in the same small table-top greenhouse. The widest cross section was in the case of sample 2 (in the range of 1.4-1.5 mm) and the thinnest cross section was in sample 4 (in the range of 0.8-1.0 mm). Additionally, the differences in the cross-sectional symmetry and the width and shape of the TW zones can be observed in these figures.
The cross sections of all the samples are non-symmetrical in the way that the possible TW side (set to point up in these images) has larger dimensions compared to the OW side. The TW zones which presumably consist of G-fibres are observed in these tomographic images as brighter areas, i.e. more dense tissue areas. In these denser areas, the G-layer containing cells have collapsed which can, for example, be observed from the TW structures for samples 1-3 in Fig. 3. Presumably, even the careful drying protocol used for these samples did not eventually prohibit the collapse of the G-layer structures.
The detachment and collapsing of G-fibres is a well-known phenomenon; it has been studied by microscopy (Clair et al. 2005) but also observed in previous studies on hybrid aspen saplings by microtomography ). With X-ray µCT, it is possible to obtain volumetric data of the sample: the full advantage of this tomographic data is presented in Fig. 4 with the simultaneous cross-sectional, radial-longitudinal (RL) and tangential-longitudinal (TL) sections of the sample. In the longitudinal images (see B and the left part of A), the TW zone can be observed based on the brighter areas in middle.
Using the tomographic data, the relative amount of the formed TW tissues across the whole cross-sectional area was determined for all the samples. This was done by calculating the approximate TW and cross-sectional areas per slice through the whole volume. Additionally, the spread of the TW tissues across the cross section in terms of sector angle was also approximated by calculating the angle between two fitted lines (see Fig. 3, sample 1). For sample 1, the amount of the formed TW tissue was 22% ± 5% of the cross-sectional area, spread at 80° ± 10° angle over the cross section. For sample 2, the values were 15% ± 5% and 90° ± 10°, for sample 3, 18% ± 5% and 95° ± 10° respectively. The errors were obtained using standard deviation. Fig. 3 The cross sections of all samples 1-4 were measured by µCT. The images are in the same scale and the length of the scale bar is 100 um. In 1) the observed TW formation is enclosed with a dashed line and the sector of the observed TW formation is marked with solid lines By studying the X-ray microtomography images, Brereton et al. (2015) reported an increase in the attenuation of the X-rays in the lateral parts of the willow stems. The lateral part is the region formed by vascular cambium and the maturing secondary xylem, which in the work by Brereton et al., was observed to extend to the TW side of the stems. They stated that on the TW side of the stems, the delay of the programmed cell death was present, and this observation was attributed to more striking attenuation of the X-rays in the TW tissues in their study. In our very young hybrid aspen saplings containing TW, we did not detect a similar kind of feature, at least not to the same extent. It must be noted that our sapling material represented different species and younger wood material compared to the study by Brereton et al. so the difference might be explained by these factors. Brereton et al. (2015) also analysed the number frequency of the vessels in TW, OW and NW and observed that the number of the vessels was lesser in TW whereas it was about the same in OW and NW. In addition, they determined the average vessel volumes and vessel surface area to vessel volume ratio in TW, OW and NW. They found out that the vessel volume was greater on the TW side but the vessel Fig. 4. 1) A radial-tangential view of sample 1, A radial-longitudinal (RL) view, B tangential-longitudinal (TL) from the TW side, C TL view from the OW side surface area-volume ratio was smaller in TW in comparison to the average. On the OW and NW sides, the vessel surface-volume ratio was similar to the average. In our very young saplings, this kind of analysis was not reliably conducted due to large variations in the sizes of the individual vessels: some of the vessel diameters were of the same order as the fibre cells, thus making the differentiation of the vessels from other xylem cells futile.
Additionally, for samples 1, 2 and 3, very dense (seemingly white voxels in absorption images) mineral particles of various shapes and sizes were observed. An example of one of the largest, observed mineral deposits is shown in Fig. 5. For sample 4, these deposits were not as prominent as for the other samples. Interestingly, the deposits in samples 1, 2 and 3 were heavily located near the pith of the sapling and within the bark layer structures. The diameters of these deposits varied a lot: for the larger conglomerates (see the upper part of Fig. 5), the diameters were approximately between 17 and 30 µm with lengths up to 55 µm. The smaller deposits (lower part of Fig. 5) ranged in size between 9 and 18 µm in diameter and length. However, all the deposits most likely consist of multiple, separate smaller units which tend to merge into one solid area in the absorption-based tomographic data.
These crystal deposits were also visible in the X-ray scattering data. Silica, calcium-based oxalates and carbonates are very commonly found in all species of plants (Franceschi and Horner 1980), serving as nutritional storage and providing mechanical protection against herbivores (He et al. 2012). Due to the variety of the showing the smaller deposits within the bark layers. The longitudinal and tangential directions are marked with L and T, respectively. In both images, the tangential direction is pointing towards the pith and the scale bar is the same compounds and their allomorphs, tracking the exact compounds behind the peaks in the XRD data is extremely challenging. However, based on the commonness of these compounds in plants and the literature, it is highly likely that the observed compounds in the XRD patterns are compounds of silica-and calcium-based oxalates and carbonates. We have also observed similar deposits in our previous XRD imaging studies of phloem and bark layers in young silver birch saplings (unpublished work) and by XRD in bast fibres, especially in nettle and hemp (Viljanen et al. 2022).

Nanostructural results: cellulose crystallite widths and microfibril orientation
The line scan results measured with 100 µm step size are given in Table 2. The error limits have been estimated from the goodness of the fits in both mmfa and crystallite widths analysis. Sample 2 was the thickest sample; thus, there are the most locational results for that sample (from 15 locations). Based on the nanostructural parameters (size and orientation of the cellulose microfibrils), there might be TW tissue in all these four saplings studied, including sample 4 in which case the TW properties would be induced possibly only by phototropic effects. However, it is also observed that in greenhouse grown juvenile hybrid aspen, TW tissue is often generated in the straight-grown stems (Bjurhager et al. 2010;Svedström et al. 2012).
The examples of the differences in the shapes of the azimuthal profiles are presented in Fig. 6, on which the T-parameter analysis and thus the mmfa values are based. It can be observed that the profiles corresponding to locations 3, 4, and 5 are narrower than the profiles corresponding to locations 8 and 9, which corresponds to the lower and higher mmfa values, respectively.
In Fig. 7, the CT cross sections created using different contrast mechanisms are presented for sample: A) based on the absorption value of the voxel by XMT, B) the intensity of the cellulose 200-reflection by XDT and C) the mmfa value by XDT. Considering the cellulose 200-reflection intensity, the zone corresponding to the same areas as the collapsed cells can be observed. The mmfa values are more spread, so they do not point as clearly to this zone; however, the lowest mmfa values are connected to that same area.
We have previously studied young T89 clone hybrid aspen samples, including 21 plants with ages between 2.5 to 3.5 months . In that study, it was observed, that the mean MFA was significantly lower in juvenile T89 compared to that of juvenile birch wood (Leppänen et al. 2011;Svedström et al. 2012;Bonham and Barnett 2001). It was also observed that the low MFA values in most cases occurred alongside crystallite widths of around 3 nm, which corresponds to the average crystallite width of normal wood (see, e.g., Andersson et al. 2003;Leppänen et al. 2011;Svedström et al. 2012). The property of hybrid aspen to produce fibres with low MFA values was also observed in the current study. As such, this is an interesting property of hybrid aspen, because generally it is expected that juvenile saplings are connected to high MFAs to sustain mechanical adaptivity (Barnett and Bonham 2004).

Table 2
Average crystallite widths and the mean MFA (mmfa) values with the error limits ( ±) of the single XRD line scans conducted on the saplings 1-4 The error margins presented were estimated based on the goodness of fit considering the fitting procedures, and the values reflect the upper limit of the estimated error. In the case of samples 3 and 4, the small stem widths resulted in fewer line scans in comparison to samples 1 and 2. All the mmfa values below 14 degrees are marked in bold along the corresponding crystallite widths from the same line scan location. The location column in the table corresponds to the line scan locations per each sample, see, for example, Figure 6 Sample 1#  Compared to the crystallite width results of Table 2 corresponding to the TW side, the average crystallite width has been determined to be larger in more mature TW (typically around 4-5 nm) based on the literature values for poplar, European aspen, hybrid aspen and many tropical hardwood species (Müller et al. 2006;Leppänen et al. 2011;Viljanen et al. 2020). This observation is valid also when taking into account the proportions of the TW tissues among the whole line scan area. So, it might be possible that in these very juvenile saplings (less than 2 months old), the cellulose microfibrils are not yet fully crystallized and aggregated in TW G-fibre cell walls.
Another observation, which can be made based on the XRD results is that in the samples (samples 1, 2, 3) the crystallite sizes in the positions which correspond to the OW side also are slightly smaller than those of the hybrid aspen NW . This similar kind of slight difference can be also observed in the results represented by Sawada et al. (2018). However, also in their study, the difference is just within the accuracy of the results: the mean crystallite width based on the cellulose reflection 200 is in OW 3.1 and 3.2 nm, whereas in NW they obtained 3.3 nm (Sawada et al. 2018). The slightly smaller crystallite widths of OW are interesting to be compared to the similar crystallite widths (around 2.9 nm) observed for compression wood samples (Andersson et al. 2000) and compression wood-like species such as juniper (Hänninen et al. 2012).
In the future, it would be essential to repeat the experiments for a larger group of samples due to the biological variation between the individual saplings. Additionally, it would be interesting to compare the TW formation between greenhouse-, outdoor-and wild-grown saplings to see if the density and occurrence of the TW are similar between all the studied groups.
Combining the nanostructural results obtained by X-ray scattering with the cellular scale structures observed by XMT, offered intriguing views of the structural features of the juvenile hybrid aspen saplings. Overall, the combination of these methods is a powerful tool for wood characterization.

Conclusion
The combined scattering and tomography techniques reveal tissue-specific structural information both at the nano-and microscales and especially map these structural properties together. Using these methods, the average width of the cellulose crystallites in TW was 3.5 ± 0.1 nm, and in OW it was 3.1 ± 0.1 nm. Thus, a consistent difference was observed between TW and OW. The difference was further confirmed and detected also regarding the orientation, i.e. mean microfibril angle, which was determined to be below 14° in TW and above 20 degrees in OW. In previous studies, the cellulose crystallite width has been observed to be distinctly larger in the mature TW (around 4-5 nm); thus, it might be that in the juvenile saplings (less than 2-month-old and stem width smaller than 2 mm), the cellulose microfibrils are not yet fully crystallized and aggregated in TW G-fibres.
Home laboratory bench-top set-up offers its own advantages for these studies, considering the number of samples characterized, time-dependent studies and larger field of views.