Micro-CT Imaging of Low-density Plant Stems

Abstract


Background
Small grain cereals lose a high percentage of their yield through lodging (buckling of the stems near ground level) before harvesting due to the weak structure of the stem, especially under heavy rain and wind (1).Computational models are being used to simulate stem collapse under loads to better understand the relationship between the stem's structure and its strength (2).With this understanding, the stem's structure can then be modi ed through conventional breeding or genetic engineering to increase its resistance against lodging.However, to construct these models, three-dimensional (3D) structural images of the stem are required.3D images can be carefully built from a limited number of individual 2D cross sections (3) of various types of microscopy.These resulting 3D models assume that the cross sections capture all the internal structure variation present and that those structures are continuous.If the assumption is incorrect, error is introduced into the computational models, resulting in large variations that do not mimic real-world results.Micro-CT however takes 2D image stacks from micro slices continuously through the plant stem to build a 3D model by measuring tissue density.This process allows for the 3D visual reconstruction of structures such as anatomical vascular systems.Researchers can now study entire cellular structures to understand how cellular systems are constructed and morphometric parameters can be extracted.At this time, whole potted plants in-situ or intact plants cannot be used to achieve high resolution which is complicated by auto-digestion and liquid evaporation that causes blurry images due to internal movements within the plant (4).Therefore, adjusting sample size is needed to achieve high resolution.Micro-CT imagery has been used to visualize plant tissues of fruits (5), leaves (6) and pollen development (7) but not larger plant structures such as whole stems.
One of the largest challenges to using micro-CT imagery on larger plant and animal tissues is the low contrast between tissues and the environment (4).To some extent, image contrast is dependent on the micro-CT machine parameters, such as over exposure and how the x-ray attenuates the tissue (i.e.signal to noise ratio).However, due to the high radiolucency of organic matter or low density of plants, they do not readily lend themselves to micro-CT imaging.To enhance the contrast between solid and air in the images, absorbable contrast agents can be used to increase the radiopacity of the organic matter (8).In order for some contrast agents to penetrate and bind to the tissues, proper prior xation needs to occur.
There are three main ways xatives can be divided, coagulant vs non-coagulant, additive vs non-additive or acidic vs basic vs aldehyde (9).The key to all these xatives is how they interact with the plant tissue.
Research goals as well as the plant biology dictate which xative type is most applicable for the study.In the case of micro-CT imagery, only cell wall boundaries are of interest.Consequently, xatives that retain cytosolic microstructures are not needed.Conversely, xatives that increase shrinkage due to water loss might alter certain plant cell wall structures and therefore be inadvisable.These chemicals remove and replace free water in cells and tissues and cause a change in the tertiary structure of proteins by destabilizing hydrophobic bonding (10).Fixatives that are coagulants or have additive properties (stay embedded into the cell wall) should be tested with any contrasts to determine if there are any contrast agent clumping or interfering with clear structural cell wall differentiation.With the correct combination of xative and contrast, a clear distinction between cell walls can be determined.Information resulting from these techniques can be applied to computational models such as to theorize how and why structures bend and some break (11).
In this paper, we developed a method to evaluate low-density tissue of herbaceous stems to determine structural differences that might lead to lodging resilience.Outlined is the process to determine the correct combination of xing agent and contrast to visualize two distinct segments (Fig. 1) of the lowdensity wheat and oat stems along with the use of a chemical drying agent to prevent autodigestion and dehydration during scanning.

Results/discussion
The goal of this study was to determine a more direct time saving method to visualize stems of cereals for cellular wall structure comparisons using 3D micro-CT imaging.

Optimization
Micro-CT imaging of plants can be challenging due to the low level of x-ray absorption of the tissues (12).
Our initial micro-CT (XT H 225, Nikon Metrology Inc., Brighton, MI, USA) scanning parameters were 75 kV tube voltage, 75 µA tube current, 708 ms exposure time, 720 projections and 4 frames per projection.The total scanning time was approximately 35 min for each specimen.The resulting images were grey, blurry, and unde ned with no cell boundary details and limited contrast (Fig. 2: B & C).Only one peak was formed (Fig. 2A).Lowering the x-ray energy and exposure time for each sample improved the resolution in the low-density herbaceous plant tissue.Adjusting scanning parameters to optimal settings of 100 kV, 40 uA and exposure time of 354 with a magni cation of 33.05 gave us a voxel size of 4.05 microns and resulted in a second peak in the grayscale (Fig. 2D).This second peak indicated a successful separation of the sample from the background (Fig. 2D) which would allow the tissues to be isolated using a suitable grey-value threshold (Fig. 2: A & D).The number of projections used was 360, with 16 frames per projection.These settings provided a good signal-to-noise ratio.No lter was used for the X-ray beam because it did not improve the quality of the scans.
To further widen the grey scale and to improve visibility of details, we introduced a commonly used contrast agent (Hexibrix 12.5% (v/v) Guerbet LLC, Bloomington, IN, USA) through a fresh stem using capillary action.Stem segments (3-5 cm) were partially embedded into the foam and lled with 400 ml of Hexibrix.The foam helped stabilize the stem for scanning.However, the contrast agent did not penetrate the stem section; contrast only coated and clumped on the outside surface.In addition, the physical stem length needed to be reduced for higher resolution.Optimized machine parameters included bringing the sample closer to the target resulting in a magni cation of 33.05 and a voxel size of 4.05 µm.A straight stem segment with a diameter of 0.5-1 cm and length between 3-5 cm increased the likelihood of the 3D reconstruction software to identify the center of rotation.
After machine optimization, the samples remained blurry, possibly due to auto digestion and interior movement through liquid evaporation (13).Cellular degradation on a micro level during the duration of the scan is reasonable (i.e.evaporation), therefore, it was necessary to x the tissues to preserve the morphological and molecular features of the tissue.We tested xatives to assist in dehydrating the stems prior to imaging.In addition, the ability for a contrast agent to penetrate the cellular structures is dependent on proper xation of the plant tissues (10,14).Fixative selection is important, as there can be an interaction due to crosslinking between the xatives and contrast agents (15).This interaction can cause the clumping of contrast in the tissues.The optimal combination of xing and contrast agent to use are ones that have similar pH values (16).Not one individual xative was recommended or preferred over another in the literature; therefore, we compared the images of samples prepared using four acidic xatives.After the stems were cut to the appropriate size, they were immediately placed into 1.5-mL micro-centrifuge tubes, lled with one of the following four xing agents: 70% ethanol (ETOH), formaldehyde alcohol acetic acid (FAA), Farmer, and Carnoy's (Supplementary Table 1).All samples used fresh xatives and no vacuum was applied during xation.The samples were then stored at 4°C in xative solution overnight and scanned the following day.
The addition of contrast improved the x-ray attenuation.Capillary action in a desiccator did absorb some of the contrast but it was insu cient to penetrate all the stem tissue.Different tissue reactions materialized according to the combination of xative and contrast agent and the density of the tissues; one-day soak was su cient for absorption of contrast after xation to increase the x-ray attenuation into the internode whereas a two-day soak in contrast was needed for the thicker node material (Fig. 3 & 4).
This difference is due to changes in the cross-linking of the xative-contrast combinations within the organic tissue.These compounds bind to a variety of chemical groups in the tissues, often affecting the charge at the site of attachment (15).The internodal tissues xed with FAA and Carnoys with contrast (Osmium Tetroxide or Phosphotungstate) were comparable (Fig. 3 & 4).When two combinations give comparable results, it is important to use the least toxic combination.The combination of FAA and phosphotunstate resulted in better images for the node sections while Carnoy's and phosphotunstate worked better for the internode sections.FAA was chosen for its ability to absorb into the high-density cells (nodes) allowing the contrast to penetrate and distribute evenly.Even though FAA and Carnoy's were comparable for the internodes, Carnoy's seemed to have better distribution of the contrast within the internodal tissues.There was no difference in x-ray attenuation between the two contrast agents; therefore, we chose to use phosphotunstate over osmium tetroxide because it was less toxic.
To further ensure the removal of all water out of the cells, the tissue was chemically dried with hexamethyldisilane (HMDS) (Fig. 3).Plant stems required a longer dry time then previously recorded ( 16).
This phenomena may be because the structure of plant cell walls impedes water loss.Therefore, allowing the chemical drying agent to fully evaporate from the plant under the hood was required.The introduction of contrasting agents and chemical drying increased the signal to noise ratio of the grey values to above 100 with the peak just under 300 in both the node and internode (Fig. 5E & 6E).This process allowed the plant tissue to stop the auto digestion, absorb the contrast agent uniformly and dry the tissue.Internal structures such as the vascular tissue could now be differentiated.

Validation
The optimized micro-CT images were compared with environmental SEM (TM-3000, Hitachi, High-Technologies Corporation, Tokyo, Japan) images of paired stem sections.Similar levels of detail can be seen (Fig. 5 & 6) between the micro-CT and SEM 2D images.Both methods could differentiate between cell primordia within the parenchyma tissues.The successful separation of the cellular boundaries within the micro-CT images permits the study of the spatial organization of the tissues and allows for detailed 3D micro-CT images to be created.The resulting 3D images were used to develop a sophisticated 3D numerical model to predict structural behaviors (2).

Analysis of plant tissue
We believe this is the rst time cereal nodal structures have been 3D imaged in detail.Niklas (17) discussed the physics for nodes in detail, but only described them as septa or diaphragm like.The introduction of transverse septa or nodes to the hollow tube increases the strength and exibility of the tube-shaped stem (Fig. 1).These struts reduce the tube length, thereby, reducing regionally increasing stiffness of the tube.Nodes, solid amorphous tissue connecting long stem segments, can increase stiffness as much as 16 -20%, even though they only contribute about 2% of the total stem weight (Niklas 1989).The internode, long segments of the stem, has the characteristic distribution of the strengthened tissues like vascular bundles equally distributed as far from the stem axis as possible with a hollow pneumatic center (18).
The micro-CT images allowed for clear visualization of cellular boundaries.The mechanical, parenchyma and lumen areas have distinct densities that are visualized by the micro-CT (Fig. 7).We did not nd a statistical difference between the overall oat and wheat stem diameter (p=0.18).However, the thickness of the mechanical layer changed in size between species.This result has been seen between various wheat genotypes (19).Wheat clearly had larger vascular bundles composed of sclerenchyma in the mechanical layer, which could contribute to increased internodal strength compared to oat (20).Grasses, like cereals, use cylindrical geometry to maximize their strength and exibility with the least amount of cellular tissue.A cylinder and to a greater extent, a hollow tube, is the most e cient geometric form to resist bending, elastic buckling, and torsion (21).Therefore, it is not surprising that plants use variations of this geometry in their growth.
In both cereals, two distinct rings separate from each other during scan preparation (Fig. 7D & H).The outer ring is not continuous, unlike the inner ring; an area can be clearly seen where the outer ring begins wrapping around the inner ring.This phenomenon is not noted in studies that used fresh material (22)(23)(24)(25).The ring alignment overlap can be seen in the 2D cross sections of the micro-CT scans (Fig. 7D & H).This overlap could be an artifact from the direction in which the cells form the apical meristem align during development.A spiral pattern can be seen clearly during growth at the seedling stages, where oat at leaf arrangement is clockwise while wheat's is counterclockwise.
Lateral and 3D images show various specialize structures with-in the nodal tissue that vary between species.First, oat nodes were found to be dense throughout the entirety of the nodal structure (Fig. 7A-C) while wheat contained large air pockets surrounding a center pith (Fig. 7E).These air pockets can also be seen on the lateral view (Fig. 7F & G).Secondly, the inner and outer thickness of nodal diaphragm or septa differed between oat and wheat (p= 0.031).The thinnest section of the nodal diaphragms is signi cantly greater in oat (1.94 ± 0.1 mm) (Fig. 7B & C) than wheat (1.58 ± 0.12mm) (Fig. 6F & G).The outer thickness of the node is also greater in the oat (2.94 ± 0.2 mm) than wheat (2.35 ± 0.3 mm).The nodes act as spring-like joints (26) where the spring constants of the nodes are directly related to the natural frequencies of vibration of the stems when set in motion (21,26,27).The differences in the thickness of the nodal diaphragms results in different spring constants.Wheat contains a lighter "spring" that is exible, while the oat "spring" is thick and stiff.Wheat also contains equally spaced air spaces resembling at discs that compress without damaging cellular tissue to add more exing capabilities within the node (Fig. 7E-G).These multiple pneumatic air spaces and mechanical structures add additional strength (18).
Within a exed grass stem, total strain energy is divided between the nodal diaphragms and the internodal walls where both structural elements operate as a single mechanical system (28).The internodal walls provide rigidity and strength, while the nodes contribute exibility.The cellular makeup and distribution pattern of those cells is different for each structure.

Conclusions
This method has been developed to visualize in 3D, low-density organic matter using micro-CT.Effective xative and contrast combinations are dependent on tissue cell density and chemical crosslinking with reactive cell components.However, when various xative and contrast mixes result in equal x-ray attenuation, a less toxic choice should be made.The complete drying of the cellular tissues is required to prevent evaporation during the scan that can cause internal movement and results in unde ned cellular details.We found for the determination of cell wall boundaries, this method is as accurate as SEM and produces 3D images over the length of stems.For the rst time the detailed 3D rendering of nodal tissue, with the aid of micro-CT imaging, allowed researchers to quantitatively study the interaction between the node and internode tissue.

Plant Material
Greenhouse grown oat (Avena sativia) "Gopher" plants were used for optimizing the micro-CT scanning parameters for stem segments, while eld grown oat and wheat (Triticum aestivum) "Linkert'' used for structural comparisons.All plants were scanned one week after owering.Well grown main stems were selected from each plant for all experiments.Optimization of parameters were replicated four times.Four plants per species were used for comparison of plant structural morphology.All stems were carefully cut into 3-5cm lengths using a new razor blade to prevent crushing of the stem edges sub-setting segments into nodes and internodes for imaging (Fig. 1).

Optimization of Sample Preparation
Straight stem sections, both nodes and internodes, of diameter 0.5 -1 cm were selected from the main stem of a well grown plant.The large sections were carefully cut into sections of 2-3 cm long and then immediately placed into 15-mL micro-centrifuge tubes.Node and internode sections were submerged in four different xatives at a concentration of 1:20 (wt./v tissue to xative) and stored overnight at 4°C (Sup Table 1) Then samples were transferred to a new 1.5-mL centrifuge tube lled with 1 ml of phosphotungstate contrast agent and stored in the refrigerator at 4°C for an addition 1-2 days depending on the experiment.After removing samples from the contrast agent, the stem sections were rinsed with 25mM sodium phosphate solution (Na2HPO4 (99+%, Acros Organics, New Jersey, USA) three times.The stem segments were then dried using the optimized drying method.

Micro-CT Parametrization
Greenhouse grown oat segments were secured onto a Te on stage holder using Cyanoacrylate (Scotch Super Glue, 3M Company, St. Paul, MN) and Zapit Accelerator (Dental Ventures of America, Inc., Corona, CA) to minimize vibration during scanning.Micro-CT machine (XT H 225, Nikon Metrology Inc., Brighton, MI, USA) with the following settings was used to collect images: exposure time (354 ms), tube acceleration voltage (100 kV), and tube current (40 µA).The samples were brought as close to the target as possible to maximize image resolution.Straight stem sections were used to allow the center of rotation to be identi ed readily in 3D reconstruction.The images were reconstructed using 3D CT Pro 3.3 (Nikon Metrology, Inc., Brighton, MI, USA) and visualized using VG Studio MAX 3.2 (Volume Graphics GmbH, Heidelberg, Germany).The same magni cation was used for all samples Scanning Electron Microscopy (SEM) Environmental A table-top environmental scanning electron microscope (TM-3000, Hitachi, High-Technologies Corporation, Tokyo, Japan) was used to obtain scans for complementary fresh stem segments to determine how detailed the micro-CT images could be.The SEM images were used to compare CT images for a more detailed understanding of tissue microstructures.Equivalent stem segments to those imaged by micro-CT of nodes and internodes were selected to obtain SEM images of their transverse sections.The tissues were secured onto an SEM holder using SEM mounting carbon conductive tabs (Ted Pella Inc. Redding, CA USA).The holder containing the sample was placed on a stage in the SEM machine and scanning was performed under compo mode, the mode that could differentiate materials with different densities, with an accelerating voltage of 15 kV.

Trait Measurements and Statistics
Structural measurements of four oat and four wheat samples were extracted from their micro-CT images using VG Studio MAX 3.2 (Volume Graphics GmbH, Heidelberg, Germany).Internode cross sections were measured for stem wall thickness, and inner and outer shell thicknesses (Fig S1).The ratio between the inner and outer shell thicknesses were calculated.Node longitudinal sections were measured for total internal node height and height of the nodal diaphragm (Fig S1).After scanning, a subset of stems were sliced and internal measurements were veri ed using a light microscope (DP71, Olympus America Inc, PA).
ANOVA and Tukey HSD were conducted between oat and wheat for all traits using JMP Pro 14.2.0 software (SAS Institute Inc., Cary, NC, USA, 1989-2019).D) The node without using a contrast agent has grey values under 100.

Declarations
E) The node immersed in Phosphotungstate for 2 days and dried in HMDS shows an increase of grey value distributed throughout the node from below 100 to a peak of 300.D) The node without using a contrast agent the grey values are under 100.
E) The node immersed in Phosphotungstate for 2 days and dried in HMDS shows an increase of grey value distributed throughout the node from below 100 to a peak of 300.

Figures Figure 1
Figures

Figure 3 Fixed
Figure 3

Figure 4 Visual
Figure 4