Wood samples

We used samples from two different screens of compression wood formation saplings to assess spiral grain formation. One set of experiments used 8-month-old stems from greenhouse-grown, clonally propagated trees. After the age of 2 months, trees were either grown vertically (controls) or tilted and staked at an angle of 30 to 45° to produce compression wood on their lower side of the lean (Apiolaza et al. 2011; Thomas & Collings 2017a). Stem sections (about 25 mm in length, and 8 to 10 mm in diameter) were collected, debarked and then fixed and stored in 10% (v/v) formaldehyde/5% (v/v) acetic acid/50% (v/v) ethanol (FAA). Tilted samples had a vertical line scored along their lower surface to act as a reference mark during image reconstructions.

A second set of experiments used trees derived from 10 different clones commonly used in New Zealand forestry. Trees (9 months old) were transplanted into 100-litre planter bags containing potting mix and slow-release fertiliser and grown outdoors with trickle irrigation. After 3 months, the trees were staked and tied at 10 to 20° from the vertical and grown for a further 18 months. Samples (60 to 100 mm long, diameter 25 to 40 mm) from near the base of the tree where growth was angled were debarked and fixed in FAA. After washing in water, samples were cut to a 25 mm length and while still wet, were split with a hammer and chisel through the central pith and parallel to the gravity vector. The ends and sides of the split samples were scanned with a flatbed scanner operating at 2400 dpi (Epson Perfection V700, Epson, Suwa, Japan) to record dimensions and angles.

Light microscopy

Sections were viewed by confocal microscopy (Leica SP5 on a DMI6000 inverted microscope, Leica Microsystems, Wetzlar, Germany) using a 20× NA 0.7 glycerol-immersion lens. Sequential line scanning was used to image lignin autofluorescence using excitation at 405 and 488 nm, and emission windows of 420-460 nm and 500-550 nm respectively. Threefold line averaging was used, and Z-stacks were recorded with a 2 μm step-size. Transmitted light and lignin autofluorescence images were also collected with a stereo-fluorescence microscope (model MZ10F Fluo, Leica Microsystems, Heerbrugg, Switzerland), with fluorescence using both UV and blue excitation, and imaging blue and green emissions. Samples were photographed with a Leica DFC310 FX camera.

Grain reconstructions using images of resin canals

Serial sledge microtome sections (thickness 60 μm) (Reichert, Vienna, Austria) were cut across entire stems of eight-month-old trees, with a minimum of 72 sections, covering at least 4.5 mm of the stem collected from each of 5 stems from vertical and tilted trees. Sections were washed in distilled water and mounted in glycerol. Sections were scanned at 2400 dpi in transmitted light mode on a flatbed scanner, and using circularly polarised light generated by crossed pairs of linear polarising film and crossed pairs of quarter wave-retarder film (Edmund Optics, Singapore) oriented at 45° to the polarisers (Arpin et al. 2002; Higgins 2010; Thomas & Collings 2016, 2017a). As resin canal cells have only primary cell walls, they did not rotate the polarised light and appeared dark against the bright background of tracheids whose secondary cell walls were strongly birefringent.

Images were manually aligned in Adobe Photoshop CS4 (version 11.0.1, Adobe Systems, San Jose, CA, USA) with subsequent image processing conducted in ImageJ (FIJI installation, version 1.50, National Institutes of Health, Bethesda, MD, USA). Individual images were combined into a stack, brightness and contrast standardised, and the image shape adjusted to a square. Images were aligned with the ImageJ plug-in StackReg and because this plug-in defaulted to aligning items within the image, and because the non-birefringent resin canals dominated the image, a custom-written plug-in straightened the alignment of the score mark within the stack. Once aligned, the stack was thresholded and the Analyse particles function used to detect and measure the location (centroid and stack position) of the resin canals. The Show masks option generated an output image showing the identified canals. Matlab (version R2102a, MathWorks, Natick, MA USA) codes, previously used to measure vessel alignments in X-ray computed microtomography experiments of mahogany (Collings et al. 2021), were used to measure the orientation of the resin canals, with left-handed grain being assigned negative angles. For vertical trees, canal orientations (and thus grain) were analysed as a single group and related to the distance from the centre of the stem. For tilted trees, however, canals were divided into four quadrants, based on their location relative to the stem centre. The orientations of resin canals in the upper and lower quadrants were then compared to the orientations of the pooled canals from the stem sides.

Grain reconstructions using X-ray computed microtomography

A high-resolution μCT system (SkyScan 1172, Bruker, Kontich, Belgium) was used to acquire X-ray images of wood specimens from which tomograms could be calculated with resolutions as low as 2 μm per pixel. A total of 12 eight-month-old saplings (8 vertical controls, 8 tilted trees) were processed. Whole stems were air-dried to generate sufficient contrast between cell walls and cell spaces, and impregnation with heavy metals was not required. The samples were mounted vertically in the scanner with X-ray images collected at 40 kV using a high sensitivity (10 megapixel) digital CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan) as the sample was rotated through 180° at 0.3° intervals. These images were processed through image reconstruction software (NRecon, version 1.5.1.4, Bruker) to produce a stack of transverse sections.

The full processing protocol for μCT data is explained diagrammatically in Fig. A1 in the Appendix and is similar to grain measurement procedures previously used (Collings et al. 2021). Batch processing in Photoshop was used to reduce the size of the data sets by converting the 16-bit output from NRecon to 8-bit images, and by the cropping of excess background. Any slight misalignment from the vertical of the wood samples during the initial X-ray imaging was initially corrected in a two-step procedure (Fig. A1E–F in the Appendix). Radial longitudinal sections through the pith were recalculated using the reslice plugin, and the orientation of tracheids was directly measured with the directionality plugin which uses a Fast Fourier Transform (FFT) to determine the pattern. This value was then used to rotate the entire data set so that the tracheids in this radial plane were vertical. This procedure was then repeated using radial longitudinal sections recalculated through the pith perpendicular to the first set followed by another full-stack rotation. This two-step procedure, with rotations typically in the range of 2° or less, ensured that the pith ran vertically within the samples and provided a baseline for all grain measurements. Stacks covering 2 mm of the tree were generated in ImageJ and Smoothing was applied to the images. The cortical pith was cleared and a white dot was added to mark the stem centre. The stack was saved, rotated through 15° and re-saved, and the rotation repeated a further ten times so that a collection of data sets at 15° intervals was produced (Fig. A1I in the Appendix). From these data stacks, 1 mm-wide strips centred on the white dot were cropped and trimmed (Fig. A1J, K in the Appendix). The reslice function recalculated tangential longitudinal sections through each of these twelve strips, and grain was measured using directionality.

Processing X-ray computed microtomography data

Data processing was conducted in an Excel spreadsheet which converted the directionality output from each of the twelve strips across the sapling into twenty-four radii by identifying the location of the white dot that marked the centre of the pith and then determining the distance from the pith for each location. To allow comparisons between samples collected at different image resolutions, data were binned into 10-μm-wide pools. A final round of grain angle correction was then conducted (Fig. A1O in the Appendix). In this normalisation step, the grain pattern for each tree was determined by averaging the 24 radii, and the curve representing the difference between each radius and the average calculated. These curves showed systematic differences from zero and plotting the average difference between each radius and the average curve against the angle of the radius gave a sine curve. This indicated that there was a systematic grain error caused by the non-vertical orientation of the sample. Fitting of a sine curve to this data in Excel allowed a final grain correction value which differed for each radius, which was typically less than 1°. While this final normalisation step did not change the average grain pattern for each tree, it reduced the variation between the individual radii. Tilted trees were processed in a slightly different manner, with this final grain normalisation step using only the wood formed prior to tilting, as judged by the location where strong compression wood first formed, and only for the upper side of the tree.

For vertical trees, grain patterns were analysed as a single average of all twenty-four radii, and two-dimensional grain maps were calculated with a second Excel spreadsheet that applied with angle-dependent colouring. For tilted trees, a process analogous to the quadrant approach was used. Average grain angles were calculated separately for the top of the stem for three radii (angles 0°, ±15°), for three radii at the bottom of the stem (180°, ±165°) and six radii at the stem sides (±75°, ±90°, ±105°).

The formation of compression wood limits the development of spiral grain

Wood splitting experiments were conducted in 25 mm long wood discs from 96 two-year-old trees that had developed compression wood during 18 months of tilting, and that had been debarked, fixed but not dried. Measurements demonstrated that tilting strongly promoted growth on the lower side of stems with the distance from the pith to bark increasing by approx. 50% because of the induction of compression wood. Surface grain, visible because of long resin canals (Fig. 1C; arrows), was significantly more left-handed on the upper side than the lower side by 1.95° (Fig. 1E). When the wood discs were split through the central pith, the fracture propagated through the disc following the grain (Fig. 1C, F). On the lower side, the fracture ran nearly vertical through the compression wood, but on the upper side, the pronounced left-handed twisting caused an inclined fracture face. These different fracture angles meant that the fracture line on the opposite surface was hooked, with the change between twisting wood and straight wood often associated with the strong compression wood formed when the tree was tilted (Fig. 1C, D; SCW). The average deviation of almost 9° in the direction of the fracture on the opposite face was consistent with the upper surface of the wood showing stronger left-handed spiral grain. These observations suggest a link between compression wood formation and a reduction in the progressive formation of spiral grain in radiata pine saplings. They do not, however, provide definitive evidence for this link because compression wood is more brittle than normal wood (Chauhan et al. 2006) and might split differently.

The anatomy of compression wood and spiral grain

To determine whether links exist between spiral grain and compression wood, we needed to distinguish compression wood from wood found on the opposite side of the tree (normal wood) in both serial sections and X-ray computed microtomography (μCT). A two-year-old tree tilted for 18 months to induce compression wood was analysed by both approaches. Darker compression wood was present in an area of asymmetric growth (Fig. 2A; CW) that surrounded the initially round stem (bracket) and material from the compression/normal wood boundary was assessed by physical sectioning and light microscopy (Fig. 2D-Ii) and by μCT (Fig. 2C, J–L). Scanner images using transmitted light showed a colour division between normal and compression wood (Fig. 2D) which was confirmed by fluorescence. While both normal and compression wood fluoresced with UV excitation (Fig. 2E), the compression wood generated stronger fluorescence with blue excitation (Fig. 2E) consistent with previous spectroscopic observations (Donaldson et al. 2010). Confocal microscopy (Fig. 2G-I) confirmed the distinction between normal and compression wood showing changes in tracheid shape from polygonal to rounded, increased lignification of the S2 layer of the secondary cell wall, the presence of gaps between adjacent tracheids and differences in the relative emissions of the normal and compression wood. These observations were all consistent with previous descriptions of radiata pine wood (Donaldson et al. 2004, 2010). These observations demonstrate that blue light excitation can be used to clearly identify compression wood in samples.

Differentiation of compression wood in 2-year-old trees tilted for 18 months. Scale bars = 10 mm in (A), 1 mm in (B) for (B, C); 1 mm in (D) for (D-F); 0.1 mm in (I) for (G–I); 0.5 mm in (L) for (J–L). CW, compression wood; NW, normal wood. (A) Initial growth was symmetric (brackets), but the tilted tree developed asymmetric compression wood. (B, C) Wood block sectioned from the region boxed in (A) photographed (B) and imaged by X-ray computed microtomography (μCT) (C). (D–F) Transverse section scanned (D) and imaged by fluorescence with ultraviolet (E) and blue excitation (F). (G–I) Confocal sections of the boundary between normal and compression wood showing 405 nm excitation (G) and, 488 nm excitation (H) with these shown in cyan and red, respectively, in an overlay (I). (J–L) Reconstructed μCT images of the region boxed in (C) in transverse (TS) (J), radial longitudinal (RLS) (K) and tangential longitudinal sections (TLS) (L). Structural elements within the wood include tracheids (t), resin canals (rc) and rays (r). Fibre orientations were 5° to the left and right of vertical.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Differentiation of compression wood in 2-year-old trees tilted for 18 months. Scale bars = 10 mm in (A), 1 mm in (B) for (B, C); 1 mm in (D) for (D-F); 0.1 mm in (I) for (G–I); 0.5 mm in (L) for (J–L). CW, compression wood; NW, normal wood. (A) Initial growth was symmetric (brackets), but the tilted tree developed asymmetric compression wood. (B, C) Wood block sectioned from the region boxed in (A) photographed (B) and imaged by X-ray computed microtomography (μCT) (C). (D–F) Transverse section scanned (D) and imaged by fluorescence with ultraviolet (E) and blue excitation (F). (G–I) Confocal sections of the boundary between normal and compression wood showing 405 nm excitation (G) and, 488 nm excitation (H) with these shown in cyan and red, respectively, in an overlay (I). (J–L) Reconstructed μCT images of the region boxed in (C) in transverse (TS) (J), radial longitudinal (RLS) (K) and tangential longitudinal sections (TLS) (L). Structural elements within the wood include tracheids (t), resin canals (rc) and rays (r). Fibre orientations were 5° to the left and right of vertical.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Differentiation of compression wood in 2-year-old trees tilted for 18 months. Scale bars = 10 mm in (A), 1 mm in (B) for (B, C); 1 mm in (D) for (D-F); 0.1 mm in (I) for (G–I); 0.5 mm in (L) for (J–L). CW, compression wood; NW, normal wood. (A) Initial growth was symmetric (brackets), but the tilted tree developed asymmetric compression wood. (B, C) Wood block sectioned from the region boxed in (A) photographed (B) and imaged by X-ray computed microtomography (μCT) (C). (D–F) Transverse section scanned (D) and imaged by fluorescence with ultraviolet (E) and blue excitation (F). (G–I) Confocal sections of the boundary between normal and compression wood showing 405 nm excitation (G) and, 488 nm excitation (H) with these shown in cyan and red, respectively, in an overlay (I). (J–L) Reconstructed μCT images of the region boxed in (C) in transverse (TS) (J), radial longitudinal (RLS) (K) and tangential longitudinal sections (TLS) (L). Structural elements within the wood include tracheids (t), resin canals (rc) and rays (r). Fibre orientations were 5° to the left and right of vertical.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

μCT correlated well with transmitted light scans of sections from nearby locations, producing a stack of computed transverse sections. While details of individual cells were not clear in images of the entire cross-section (Fig. 2C), the change from normal to compression wood was evident. When a smaller region was scanned (boxed region in Fig. 2C shown as Fig. 2J), anatomical details including tracheids (t), resin canals (rc) and rays (r) were visible although the distinction between compression wood and normal wood was less clear. Nevertheless, both fluorescence and μCT images could be used to recognise compression wood distributions. μCT data sets allow recalculation of different image planes: when resliced as a radial longitudinal section, the tracheid orientation was not vertical which confirmed that the sample had not been mounted with the pith fully vertical. The directionality plugin in ImageJ measured tracheid deflection from vertical at approx. 5° (Fig. 2K). In tangential longitudinal sections, the tracheids were also non-vertical, showing a deflection of approx. 5° (Fig. 2L). This orientation would be consistent with either the sample being mounted non-vertically or with left-handed grain but without the pith to act as a baseline, these alternatives could not be distinguished.

Serial sectioning demonstrated symmetrical spiral grain in vertical trees

We have previously identified and measured resin canals in scanned wood discs from eight-month-old trees, imaged with circularly polarised light (Thomas & Collings 2017a) and provided preliminary information on the identification of canals in serial sections to view grain patterns (Thomas & Collings 2016, 2017b). Here, we report a quantitative analysis of grain using resin canals as a proxy, measuring the orientation of resin canals using the same approach in ImageJ and Matlab that we previously used to measure the orientation of vessels in African mahogany (Collings et al. 2021). In all five vertical trees measured, the grain became increasingly left-handed moving away from the pith, with the average grain angle about 0° near the pith increasing to −4° towards the periphery (with left-handed grain being assigned as a negative angle) (Fig. 3B). In these vertical trees, bands of compression wood were often present, as seen in the tree indicated in Fig. 3A. In this tree, the resin canals associated with the compression wood bands, and those that formed immediately afterwards (asterisks), showed a relative decrease in angle towards the vertical and were less left-handed than those canals that had been formed earlier.

Measurements of resin canal orientation confirmed increasingly left-handed grain in a vertical tree, but grain development was disrupted in tilted trees. (A, C) Scanned and fluorescence (blue excitation) images of cross-sections from a vertical control (A) and tilted tree (C). Compression wood (CW); asterisks in (A) indicate a band of canals. The tilted tree cross-section (C) shows quadrants, and the circular stem from before the tree was tilted. Scale bar in (A) = 1 mm for (A, C). (B) Measured canal orientations and grain in vertical trees. Average grain + standard errors (black line) included the tree in panel (A) (red line). The location of the compression wood (CW) bands in panel (A) are highlighted (yellow band). (D) Measured canal orientations and grain in tilted trees. Average curves with standard errors are shown for the upper (black) and lower quadrants (green) along with the vertical control (blue). (E) Grain comparisons in the 1 mm prior to and following tilting. Whereas the upper side became more left-handed, the lower side became less left-handed. (F) The differences between before and after are plotted (right-hand side) and the lower side where grain became more right-handed showed a significantly different response to the upper side where grain became more left-handed (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Measurements of resin canal orientation confirmed increasingly left-handed grain in a vertical tree, but grain development was disrupted in tilted trees. (A, C) Scanned and fluorescence (blue excitation) images of cross-sections from a vertical control (A) and tilted tree (C). Compression wood (CW); asterisks in (A) indicate a band of canals. The tilted tree cross-section (C) shows quadrants, and the circular stem from before the tree was tilted. Scale bar in (A) = 1 mm for (A, C). (B) Measured canal orientations and grain in vertical trees. Average grain + standard errors (black line) included the tree in panel (A) (red line). The location of the compression wood (CW) bands in panel (A) are highlighted (yellow band). (D) Measured canal orientations and grain in tilted trees. Average curves with standard errors are shown for the upper (black) and lower quadrants (green) along with the vertical control (blue). (E) Grain comparisons in the 1 mm prior to and following tilting. Whereas the upper side became more left-handed, the lower side became less left-handed. (F) The differences between before and after are plotted (right-hand side) and the lower side where grain became more right-handed showed a significantly different response to the upper side where grain became more left-handed (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Measurements of resin canal orientation confirmed increasingly left-handed grain in a vertical tree, but grain development was disrupted in tilted trees. (A, C) Scanned and fluorescence (blue excitation) images of cross-sections from a vertical control (A) and tilted tree (C). Compression wood (CW); asterisks in (A) indicate a band of canals. The tilted tree cross-section (C) shows quadrants, and the circular stem from before the tree was tilted. Scale bar in (A) = 1 mm for (A, C). (B) Measured canal orientations and grain in vertical trees. Average grain + standard errors (black line) included the tree in panel (A) (red line). The location of the compression wood (CW) bands in panel (A) are highlighted (yellow band). (D) Measured canal orientations and grain in tilted trees. Average curves with standard errors are shown for the upper (black) and lower quadrants (green) along with the vertical control (blue). (E) Grain comparisons in the 1 mm prior to and following tilting. Whereas the upper side became more left-handed, the lower side became less left-handed. (F) The differences between before and after are plotted (right-hand side) and the lower side where grain became more right-handed showed a significantly different response to the upper side where grain became more left-handed (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Serial sectioning demonstrated asymmetric grain patterns in tilted trees

While vertical trees had symmetrical grain patterns, large amounts of compression wood were generated in trees tilted for 6 months making stem shape asymmetric (Figs 1E, 3C), and grain patterns were no longer symmetrical. Unfortunately, using resin canals as a proxy for grain has limitations for investigating compression wood, because few resin canals are formed in compression wood (Cown et al. 2003, 2011; Thomas & Collings 2017a). Nevertheless, quantification demonstrated that the induction of compression wood modified normal spiral grain development, and suggested that canals formed after compression wood formation on the lower side of the stem were straighter than the helically-organised canals found in normal wood on the stem’s upper side. To measure grain asymmetries, the stem cross-sections were divided into 4 quadrants (upper, lower, left and right) (Fig. 3C). In the five tilted trees measured, grain in the lower quadrant did not show the trend of increasingly left-handed grain seen in all vertical control trees. Instead, grain on the lower side stayed the same over the 5 millimetres of radial growth (Fig. 3D, green line). In the upper sector, however, grain did become more left-handed (Fig. 3D, black line) although the pattern was less clear than in vertical samples. These measurements do not, however, provide a conclusive result because of the variability between samples, the low number of resin canals and the fact that the diameter of the saplings on tilting was variable (Fig. 3D, magenta dots).

To compare the orientation of grain in tilted samples using resin canals as a proxy, an alternative approach was developed. Comparisons were made of the grain in quadrants in the 1 mm prior to tree tilting and the 1 mm after tilting. While grain for the quadrants started at similar values (Fig. 3E, solid bars), wood in the upper quadrant was increasingly more left-handed grain, whereas wood in the lower quadrant showed a decrease. Comparing the upper and lower quadrants, the differences from before and after tilting were significantly different (Fig. 3F). This result suggests that the induction of compression wood modifies grain patterns in young saplings, but the limitations in resin canal analysis mean that this cannot be conclusively demonstrated.

Microtomography confirmed the development of left-handed grain in vertical trees

μCT analysis was conducted on the same set of eight-month-old saplings, including several in which X-ray imaging was conducted on the stem stubs from which serial sections were cut. Cross-sections showed the central pith and resin canals, as well as circular bands of different wood densities (Fig. A2B) in the Appendix in identical patterns to polarised light imaging as determined by the location of resin canals (Fig. A2A in the Appendix) even though the sections were separated by several millimetres, and despite the samples having to be dried prior to μCT imaging.

The superior resolution of μCT allowed tracheids to be observed in tangential longitudinal sections (Fig. 2J), and thus for direct measurements of grain to be made with the directionality plug-in in ImageJ. From a stack of cross-sections (Fig. 4A), grain was measured in a series of 12 1-mm-wide strips arranged at 15° intervals. Locations near the pith showed weakly right-handed grain (Fig. 4D, E) whereas locations near the stem periphery showed grain that was more left-handed (Fig. 4C, F). By combining the observations at 15° intervals, a grain map for the entire stem was generated which showed that the spatial distribution of grain was generally symmetric, and that grain became progressively more left-handed away from the pith (Fig. 4B). These measurements were plotted and showed that the average grain (Fig. 4G; black line), as well as all twenty-four radii (Fig. 4G; grey lines), all became progressively more left-handed away from the pith. A further seven vertical trees were measured, and these all demonstrated weakly right-handed grain that became progressively left-handed within the first several millimetres of growth (Fig. 4H). While the magnitude and direction of the grain change matched that determined by serial sectioning, the baseline from where the changes began was slightly different as serial sectioning showed a change from 0° to −4° (Fig. 3B). While it is possible that sample drying, as required for X-ray computed microtomography analysis, might explain the difference between the samples, this may not be the case. We compared the results obtained from the two saplings analysed by both approaches, and these gave quantitatively similar results (Fig. A2D in the Appendix). Colour-coded grain charts for all eight vertical trees confirmed the symmetry of the grain patterns (Fig. A3 in the Appendix). The colour-coded grain charts also demonstrated that discrete grain bands, continuing in wide arcs, were present in many of these trees, although these did not correspond to visible changes within the wood structure (Fig. A3 in the Appendix; asterisks). Furthermore, quantification of grain data showed that in all eight trees, grain became increasingly left-handed by between 3 and 7° during the first 4 mm of growth (Fig. A3C in the Appendix).

μCT showed that grain in vertical trees became progressively more left-handed. Scale bar in (A) = 1 mm for (A, F); bar in (B) = 200 μm for (B–E). (A) Cross-section with a compression wood band (CW) near the central pith. (B) Colour-coded grain map showing symmetric grain that becomes increasingly left-handedness. (C–F) Four individual, 1 mm-wide tangential longitudinal sections were taken from locations shown in the strip boxed in (A). Arrows and numbers indicate grain measurements. (G) Average grain (black) with standard errors (red) from 24 separate strips (grey lines) from a single tree. (H) Average grain (black) with standard errors (red) for 8 vertical trees (grey lines).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

View Full Size

μCT showed that grain in vertical trees became progressively more left-handed. Scale bar in (A) = 1 mm for (A, F); bar in (B) = 200 μm for (B–E). (A) Cross-section with a compression wood band (CW) near the central pith. (B) Colour-coded grain map showing symmetric grain that becomes increasingly left-handedness. (C–F) Four individual, 1 mm-wide tangential longitudinal sections were taken from locations shown in the strip boxed in (A). Arrows and numbers indicate grain measurements. (G) Average grain (black) with standard errors (red) from 24 separate strips (grey lines) from a single tree. (H) Average grain (black) with standard errors (red) for 8 vertical trees (grey lines).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

μCT showed that grain in vertical trees became progressively more left-handed. Scale bar in (A) = 1 mm for (A, F); bar in (B) = 200 μm for (B–E). (A) Cross-section with a compression wood band (CW) near the central pith. (B) Colour-coded grain map showing symmetric grain that becomes increasingly left-handedness. (C–F) Four individual, 1 mm-wide tangential longitudinal sections were taken from locations shown in the strip boxed in (A). Arrows and numbers indicate grain measurements. (G) Average grain (black) with standard errors (red) from 24 separate strips (grey lines) from a single tree. (H) Average grain (black) with standard errors (red) for 8 vertical trees (grey lines).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Microtomography confirmed that compression wood formation inhibits spiral grain development

We analysed eight tilted trees by μCT, and instead of averaging the full 24 strips separated by 15°, comparisons were made between strips from the upper and lower sides and along the flank of the stem in a manner comparable to the quadrant approach used for serial sectioning (Fig. 5A). Compared to symmetrical grain maps measured in vertical samples (Fig. 3, Fig. A3 in the Appendix), grain in a tilted tree was more variable but clearly asymmetric, with a wedge of compression wood on the lower side of the tree showing a strikingly different pattern (Fig. 5B). Quantification demonstrated these differences and showed that while the upper side of the tree and the flanks were similar to vertical controls, showing increasingly left-handed grain, grain on the lower side of the tree started more left-handed but did not become any more left-handed (Fig. 5C). This pattern matched the observations of grain made on the lower side of tilted trees using serial sectioning (Fig. 3D) but provided a higher resolution view of the pattern. We used the same approach of quantifying grain in the 1 mm prior to and following tilting, and this confirmed that on the upper side of the tree and the flanks, grain did become progressively more left-handed whereas, on the lower side of the tree, grain became more right-handed (Fig. 5D).

μCT demonstrated that compression wood induction disrupted development of left-handed grain in a single tree. (A) Cross-section showing the circular central region formed before tilting. (B) The colour grain map demonstrated the complex grain on the tree’s lower side. Scale bar in (A) = 1 mm for (A, B). (C) Grain measurements demonstrated increasingly left-handed grain on the upper side of the tree and along the sides, but a different pattern on the lower side. (D) Grain in the 1 mm before and after tilting: the upper side of the tree became more left-handed the lower side became significantly less left-handed. (E) Grain differences between before and after tilting.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

View Full Size

μCT demonstrated that compression wood induction disrupted development of left-handed grain in a single tree. (A) Cross-section showing the circular central region formed before tilting. (B) The colour grain map demonstrated the complex grain on the tree’s lower side. Scale bar in (A) = 1 mm for (A, B). (C) Grain measurements demonstrated increasingly left-handed grain on the upper side of the tree and along the sides, but a different pattern on the lower side. (D) Grain in the 1 mm before and after tilting: the upper side of the tree became more left-handed the lower side became significantly less left-handed. (E) Grain differences between before and after tilting.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

μCT demonstrated that compression wood induction disrupted development of left-handed grain in a single tree. (A) Cross-section showing the circular central region formed before tilting. (B) The colour grain map demonstrated the complex grain on the tree’s lower side. Scale bar in (A) = 1 mm for (A, B). (C) Grain measurements demonstrated increasingly left-handed grain on the upper side of the tree and along the sides, but a different pattern on the lower side. (D) Grain in the 1 mm before and after tilting: the upper side of the tree became more left-handed the lower side became significantly less left-handed. (E) Grain differences between before and after tilting.

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Asymmetric grain patterns associated with tilting were seen in a further seven trees analysed by microtomography (Fig. A4 in the Appendix), with these patterns clearly visible both in average grain patterns (Fig. 6A) and in comparisons of grain in the 1 mm before and after tilting (Fig. 6B, C). The measurements showed that whereas grain becomes increasingly more left-handed on the upper side of the stem, grain changes on the lower side of the stem were significantly different with the grain becoming more right-handed (Fig. 6E). The μCT images of tilted samples contain further information, with discrete grain patterns that matched the distribution of normal and compression wood. In numerous trees, bands of normal wood were interspersed between the induced compression wood, and these were associated with an increase in left-handed grain (Fig. A4A, B; in the Appendix asterisks).

μCT demonstrated that compression wood induction disrupted the development of left-handed grain in 8 tilted trees. (A) Averages and standard errors for the upper, lower and combined left and right (L+R) sides. Distances from the pith when trees were tilted are marked with magenta dots. (B) Average grain in the 1 mm prior to and following tilting. (C) Grain change following tilting. While the upper side became more left-handed, the lower side became significantly less left-handed. (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

View Full Size

μCT demonstrated that compression wood induction disrupted the development of left-handed grain in 8 tilted trees. (A) Averages and standard errors for the upper, lower and combined left and right (L+R) sides. Distances from the pith when trees were tilted are marked with magenta dots. (B) Average grain in the 1 mm prior to and following tilting. (C) Grain change following tilting. While the upper side became more left-handed, the lower side became significantly less left-handed. (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

μCT demonstrated that compression wood induction disrupted the development of left-handed grain in 8 tilted trees. (A) Averages and standard errors for the upper, lower and combined left and right (L+R) sides. Distances from the pith when trees were tilted are marked with magenta dots. (B) Average grain in the 1 mm prior to and following tilting. (C) Grain change following tilting. While the upper side became more left-handed, the lower side became significantly less left-handed. (t-test, $\mathit{p}\mathrm{<}0.02$).

Citation: IAWA Journal 44, 1 (2023) ; 10.1163/22941932-bja10088

• Download Figure
• Download figure as PowerPoint slide

Polarised light imaging of serial sections generated three-dimensional models of spiral grain development

Understanding the interactions between spiral grain and compression wood requires an understanding of the organisation of grain at high-resolution and in three dimensions. We avoided the use of confocal microscopy which, while it can provide high-resolution grain images (Ogata & Fujita 2005), is limited in the depth to which samples can be imaged, and the speed at which large samples can be processed. Similarly, high-resolution reflection measurements based on the tracheid effect are also limited in the depth to which samples can be measured (Ma et al. 2019). Instead, two independent methods were devised to investigate spiral grain inside the stem at medium to high resolution. In the first, medium-resolution approach, resin canals were visualised by circularly polarised light using a scanner, with serial sectioning then used to build a three-dimensional map of canal locations (Thomas & Collings 2016, 2017a). Because the resin canals in Pinus are formed from the same cambial cells that form the tracheids (Bannan 1936), they run parallel to the grain and canal orientation has been used extensively as a surface marker for grain (Noskowiak 1963; Werker & Fahn 1969; Harris 1989). Our serial sectioning and three-dimensional modelling of grain throughout the stem confirmed a steady increase in left-handed grain during the first year of growth in vertical trees. The limitation of this approach for studying the link between spiral grain and compression wood was, however, that compression wood contains few resin canals (Cown et al. 2003, 2011; Thomas & Collings 2017a) and thus observations of grain within compression wood were limited.

X-ray computed microtomography provides high-resolution maps of grain patterns

Our second new approach to measuring grain in radiata pine used μCT to generate high-resolution three-dimensional data sets, with image analysis in ImageJ directly measuring the orientation of tracheids in tangential longitudinal sections, an approach that we have previously used to analyse the development of interlocked grain in African mahogany (Collings et al. 2021). Several methods related to μCT have previously been used to observe spiral and interlocked grain. X-ray imaging of interlocked grain in Acacia covered large areas of tissue but no 3-dimensional reconstructions were attempted (Ogata et al. 2003). More importantly, an X-ray tomography investigation of Picea abies logs using an industrial log scanner was used to measure overall spiral grain patterns in logs (Sepúlveda 2001; Sepúlveda et al. 2002) but did not provide high-resolution images. μCT analysis has also been used to identify tension wood formation and programmed cell death in Salix viminalis (willow) (Brereton et al. 2015) and to identify the pith within trees to add in wood processing in timber mills (Gazo et al. 2020) but in neither case were grain measurements made.

Our X-ray-based methods provide not only high-resolution images but also cover large amounts of tissue. By contrast, earlier methods to assess spiral grain at the level of single cells were based on three-dimensional reconstructions from serial sectioning. These experiments revealed the spatial arrangement and development of individual cells, and have suggested the importance of intrusive growth and anticlinal cell divisions in the development of spiral grain in a range of different gymnosperms including Pinus radiata (Harris 1973), P. contorta (lodgepole pine) (Savidge & Farrar 1984) and P. sylvestris (Scots pine) (Jura et al. 2006; Włoch et al. 2002), Picea abies (Włoch et al. 2001) and Picea glauca (white spruce) (Bannan 1936; Savidge & Farrar 1984), Abies balsamea (balsam fir) (Zagórska-Marek & Little 1986), and Larix europaea (European larch) (Hejnowicz 1961). Conventional serial sectioning is, however, fundamentally limited by the slow speed at which data can be processed, meaning that sample areas are limited, and the number of replicates is low.

One limitation of our μCT approach is that the samples needed to be dry to generate sufficient image contrast to visualise the tracheids and measure grain. When measurements were collected from the same samples using both serial sectioning of wet material and μCT of dried wood, similar grain patterns were observed, and observations in both wet and dry samples demonstrated that there was an interaction between the induction of compression wood and changes in spiral grain development. Thus, although some differences in average grain were observed, the comparisons suggested that the drying process had not caused major structural changes to the wood. Higher-resolution imaging of timber that remains moist, and where the cell walls have been impregnated with heavy metals has been conducted by others (Staedler et al. 2013; Wang et al. 2017). Of particular interest is the coupling of metal impregnation with resin embedding (Duncan et al. 2021) which has previously been demonstrated to preserve the vascular cambium (Dickson et al. 2016).

Spiral grain develops rapidly in Young radiata pine trees

Radiata pine shows the typical grain development pattern of pine trees, with the rapid development of left-handed grain in the corewood, followed by a reduction in the angle of the grain and sometimes the formation of right-handed wood. In replicating earlier work in radiata pine (Chattaway 1959), Cown and colleagues demonstrated that grain near the pith is aligned almost vertically, but that a left-handed spiral develops in the youngest growth ring, subsequently reaching a maximum in the second or third ring, before returning towards vertical and sometimes becoming right-handed (Cown et al. 1991). This pattern is typical of most gymnosperms and has been widely observed (Harris 1989; Skatter & Kucera 1998) although the mechanisms that underlie the increasingly left-handed grain in young wood and then allow the grain to become more right-handed in older wood remain unknown.

Our data is consistent with these previously published observations of radiata pine, but instead of one or several measurements per growth ring, measurements were made from pooled data at 10 μm intervals in the radial direction. This meant that we could demonstrate that grain begins becoming increasingly left-handed beginning within the first several millimetres of growth, and that grain patterns were generally symmetric. This pooled data was at intervals smaller than the typical diameter of the tracheids (20–30 μm), and while the vertical resolution of the data and grain maps was only 2 mm, with calculations based on images of this height, this distance is about the same as the length of radiata pine tracheids in early growth rings (Cown 1975). The resolution of μCT images was, however, sufficient for the structure to be observed within samples with, for example, bands of normal wood in amongst compression wood arcs being associated with increasingly left-handed grain.

The induction of compression wood reduces spiral grain development

We have provided numerous pieces of evidence that suggests a link between the induction of compression wood caused by leaning, and a reduction in the formation of spiral grain in young radiata pine stems. In wood quality experiments, samples from the upper side of two-year-old trees that contained normal wood, and where no compression wood was present, often showed severe twisting consistent with the presence of spiral grain. However, the compression wood from the lower sides of the same trees showed warping and bending and did not twist. This effect was quantified in wood splitting experiments from leaned trees in which twisting was more evident on the upper side of the samples than the lower sides that contained compression wood.

We subsequently measured spiral grain development in eight-month-old leaned trees using two independent tests, resin canal alignment and μCT. While resin canals measurements were hampered by the reduction in resin canal numbers associated with the formation of strong compression wood (Cown et al. 2003, 2011; Thomas & Collings 2017a), results suggested that there was an interaction between compression wood formation and the inhibition of spiral grain development. Subsequent μCT experiments confirmed this observation and demonstrated that while grain became progressively more left-handed in all the vertical controls and on the upper side of the tilted stems, grain did not develop in an increasingly left-handed fashion in the compression wood zone on the lower side of the stem. Importantly, however, there were no indications that the development of normal wood on the upper side of the stem was modified by tilting, with grain becoming increasingly left-handed as in vertical stems (Fig. 6A).

This result is consistent with there being an interaction between the induction of compression wood, and the inhibition of spiral grain development, although our quantification of this effect is limited to the first several years of growth. There is, however, a lack of information on this very important wood quality issue in the literature. While Pawsey (1965) reported that grain straightening occurred in timber that contained compression wood, an effect similar to that observed in this study, these observations were not replicated when the statistical analysis was conducted (Zobel et al. 1968). Further, Timell’s comprehensive 3 volume review of compression wood in gymnosperms did not list any evidence for such a link (Timell 1986). It is not unsurprising, however, that this interaction has not previously been demonstrated: this documentation requires the ability to see compression wood and measure wood grain through an entire sample and at high-resolution, and it is only through the use of μCT that the interaction can be confirmed. Furthermore, while the current study is limited to growth within the first year, a time when bands of compression wood can form within the tree as it tries to maintain its vertical growth. However, whether the link between compression wood and spiral grain might exist in older trees has not been investigated by our study.

Mechanisms through which spiral grain might develop

Understanding the mechanism(s) through which spiral grain and compression wood formation are linked requires knowledge as to how these components of wood develop. The cellular mechanisms through which spiral grain develops have been extensively debated. As the presence of the thick secondary wall around mature tracheids would seem to preclude subsequent cell realignments, spiral grain is thought to develop during cell division within the vascular cambium and maturation of cells into xylem. Different processes have been suggested to cause spiral grain development, either acting alone or in combination with each other and have been reviewed extensively (Harris 1989; Barlow 2005; Kramer 2006; Schulgasser & Witztum 2007). These processes include asymmetries in the frequency and orientation of pseudo-transverse periclinal divisions of the vertically aligned cambial cells that would skew the subsequent intrusive growth of the daughter cells, asymmetries in the intrusive growth that occurs during tracheid development, and in which the pointed tips of the cambial derivatives grow between and past one another, and asymmetries in radial expansion during tracheid development.

One possibility that links spiral grain and compression wood may be through seeing these structures as responses to the mechanical stresses present within wood. Compression wood forms as a reaction wood on the lower side of gymnosperm where the wood is subject to bending stresses (Donaldson et al. 2004). Spiral grain, on the other hand, has been suggested to be a response to torsional stresses present within stems, being a reaction to asymmetric stresses induced by wind acting on an asymmetric canopy (Skatter & Kučera 1997). While there is experimental evidence suggesting that interlocked grain can help angiosperms resist torque (Bassu et al. 2018), how these mechanical effects would be sensed and then transduced into cellular responses remains unclear.

One possibility that we wish to highlight is a role for plant hormones and, in particular, for auxin. While the roles of auxin have been investigated in the formation of both compression wood and spiral grain, a direct link between spiral grain, compression wood formation and auxin has not been proposed. Several observations, however, suggest the possibility of such a link. First, auxin flows occur in the vascular cambium, where both the orientation of grain and the developmental pathways that lead to compression wood formation are determined. Second, auxin has also been linked to spiral grain formation based on experiments showing that the orientation of fusiform initials is parallel to auxin flow (Zagórska-Marek & Little 1986), and through modelling of auxin’s role in grain development (Kramer 2006). We have also suggested that auxin and auxin flows might be critical for the formation of interlocked grain in African mahogany (Collings et al. 2021). However, how auxin would generate grain changes remains unclear. And third, exogenously applied auxin induces compression wood in Pinus thunbergii (Japanese black pine) (Onaka 1940), an observation repeated in numerous other systems (Timell 1986) including radiata pine (Cown et al. 2003), although it is unclear whether endogenous auxin asymmetries or changes in auxin sensitivity might be responsible for compression wood development. Thus, we consider it plausible that the link between the induction of compression wood with the inhibition of spiral grain development might be related to auxin signalling.

Conclusion and future directions

This study has demonstrated that a link exists between the formation of compression wood and the inhibition of the increase in spiral grain on the lower side of young trees subject to tilting. However, the mechanisms that drive this interaction remain unknown. While the use of μCT limits the speed at which samples can be assessed for the presence of spiral grain, the development of other approaches in which both spiral grain and compression wood can be rapidly assessed would open up investigations of the genetics that underlie these processes, perhaps along the lines of the genome-wide association study (GWAS) recently conducted into wood formation in Norway spruce (Baison et al. 2019).