A correlation analysis of Light Microscopy and X-ray MicroCT imaging methods applied to archaeological plant remains’ morphological attributes visualization

In this work, several attributes of the internal morphology of drupaceous fruits found in the archaeological site Monte Castelo (Rondonia, Brazil) are analyzed by means of two different imaging methods. The aim is to explore similarities and differences in the visualization and analytical properties of the images obtained via High Resolution Light Microscopy and X-ray micro-computed tomography (X-ray MicroCT) methods. Both provide data about the three-layered pericarp (exo-, meso- and endocarp) of the studied exemplars, defined by cell differentiation, vascularisation, cellular contents, presence of sclerenchyma cells and secretory cavities. However, it is possible to identify a series of differences between the information that can be obtained through each of the methods. These variations are related to the definition of contours and fine details of some characteristics, their spatial distribution, size attributes, optical properties and material preservation. The results obtained from both imaging methods are complementary, contributing to a more exhaustive morphological study of the plant remains. X-ray MicroCT in phase-contrast mode represents a suitable non-destructive analytic technique when sample preservation is required.

in this work, several attributes of the internal morphology of drupaceous fruits found in the archaeological site Monte Castelo (Rondonia, Brazil) are analyzed by means of two different imaging methods. The aim is to explore similarities and differences in the visualization and analytical properties of the images obtained via High Resolution Light Microscopy and X-ray micro-computed tomography (X-ray Microct) methods. Both provide data about the three-layered pericarp (exo-, meso-and endocarp) of the studied exemplars, defined by cell differentiation, vascularisation, cellular contents, presence of sclerenchyma cells and secretory cavities. However, it is possible to identify a series of differences between the information that can be obtained through each of the methods. These variations are related to the definition of contours and fine details of some characteristics, their spatial distribution, size attributes, optical properties and material preservation. the results obtained from both imaging methods are complementary, contributing to a more exhaustive morphological study of the plant remains. X-ray Microct in phase-contrast mode represents a suitable nondestructive analytic technique when sample preservation is required.
The identification of the taxonomic provenance and preservation conditions of plant remains recovered from archaeological deposits is at the basis of archaeobotanical research. The whole procedure involves a detailed description of observable characteristics of the morphological phenotype of the sample, together with a comparative analysis of modern and ancient exemplars of related taxa.
On the other hand, plant remains from archaeological sites are not always recognizable through their external morphological characteristics, for reasons that include the presence of similar traits between different species, morphological changes induced by taphonomic processes, and other issues which hinder the identification of specimens. In these cases, the analysis of the internal characteristics of the specimen can increase the amount of morphological data used to determine the anatomic and taxonomic provenance of ancient plant remains.
In general, Archaeobotany has a wide palette of microscopy techniques that allow for various descriptions of both external and internal morphological features of archaeological plant samples 1,2 . In fact, it is possible to obtain imaging methods. The methodology is focused on the application of two different imaging methods X-ray Computed Microtomography (X-ray MicroCT) and High Resolution Light Microscopy (HRLM), whose performance for visualization and analysis are compared on the basis of a specific set of drupe morphological characteristics (see "Referenced internal morphological attributes of drupes" section).
For the first imaging method, tomographic 3D images at a micrometer scale of samples SMC_01 and SMC_10 were reconstructed based on the X-ray MicroCT technique in propagated-based phase-contrast mode at the IMX Beamline of the Brazilian Synchrotron Light Laboratory (LNLS-CNPEM). The technical features of the IMX Beamline include the detector PCO pco.2000 area camera, 0.82 μm effective pixel size, 2048 × 2048 pixel, 14-bit CCD cooled camera.
For both experiments an X-ray energy range (pink-beam) 4 to 14 keV was selected. The exposure time/ projection varied from 0.4 s in SMC_01 to 0.5 s in SMC_10. The sample-to-detector distance and the angular range were established at 260 mm and 360°, respectively. The effective pixel size was 0.82 μm for a total of 2048 projections on each microtomographied sample. A 350 μm Si filter was used in both cases. These parameters were set based on a previous test imaging experiment, performed on the same exemplar SMC_01 at the IMX Beamline, and another 10 specimens imaged using industrial microtomographers ( www.nature.com/scientificreports/ selected in order to optimize resolution and contrast for the minimum necessary scanned area of the sample to reveal internal specific characteristics and the global time requirements to complete the experiment (delimited by the synchrotron experiments dynamic). X-ray MicroCT images result from the application of mathematical algorithms 22 to X-ray experimental data, obtaining a sequence of 2D images (slices) and stacked together to reconstruct a 3D map of the object's X-ray attenuation coefficient (transmission or absorption tomography). If certain experimental conditions are fulfilled, then phase shifts of the transmitted X-rays can also be detected. Absorption and phase modulation are effects that occur on X-rays crossing any kind of materials. The use of phase sensitive imaging techniques has the advantage of enhancing the visibility of low absorption materials as well as accentuating the edges between materials with small differences in refraction index. This requires a highly spatially-coherent X-ray beam provided mostly by synchrotron light sources 5,23,24 , but also by microfocus X-ray sources 25 . This study used the X-ray propagation based-phase contrast MicroCT via synchrotron radiation 26,27 .
The in-house software Raft, developed at LNLS-CNPEM [28][29][30] , was applied to reconstruct data acquired from the IMX Beamline. The Paganin filter (sometimes considered a PCI projection restoration) 31 , regularly implemented in the IMX Beamline reconstruction software for edge enhancement of low contrast samples, were applied to projections acquired in this study. Visualization, segmentation and 3D morphological analysis of microtomographic images were achieved using CT Analyzer 1.15.4.0, developed by Bruker 32 and the FIJI distribution version 1.52p of the image processing software ImageJ [33][34][35] with the plug-ins MorphoLibJ 36 , 3DViewer 37 and Volume Viewer 38 . All figures were edited using GNU Image Manipulation Program (GIMP) in the version 2.10.2 (https ://www.gimp.org/).
For the HRLM imaging method, 2D photomicrography of the internal morphological characteristics of the sample SMC_11 were examined and captured using two different microscopes. The first of them was an Olympus System Microscope model BX51 coupled to an Olympus Microscope Digital Camera Model DP71 in the Plant Anatomy Laboratory of the Biology Institute, University of Campinas. The second was an Erns Leitz GmbH microscope model Ortoplan-Pol coupled to a Kodak Digital Camera model DC4800 in the Laboratory of Complex Fluids of the Institute of Physics, University of Sao Paulo.
As non-destructiveness (including damage or modification) is considered a relevant issue of the imaging method for plant remains, only one specimen was subjected to the HRLM method of sample preparation in the Plant Anatomy Laboratory of the Biology Institute, University of Campinas. The specimen SMC_11 was previously fixed in FAA (formaldehyde, acetic acid, 50% ethanol) for 24 h 39 . Then, it was dehydrated in an ethanol www.nature.com/scientificreports/ series and embedded in hydroxyethyl methacrylate resin (Historesin Leica) 40 . A total of 96 longitudinal slices of 8.0 µm thick were obtained from one of the halves of the lengthwise divided sample, using a Microm HM340E rotary microtome. The slices were stained with 0.05% Toluidine Blue in sodium acetate buffer pH 4.7 41 and fixed to microscope slides. All slices were mounted with water for examination.

Referenced internal morphological attributes of drupes
The internal characteristics considered here for the analysis of the fruit remains comprises the three-layered pericarp (endocarp, mesocarp and exocarp) defined by cell differentiation, vascularisation, presence of cellular contents, sclerenchyma cells, cavities and ducts. This set of descriptive variables is based on specific studies on fruit anatomy and classification [42][43][44] . In general, drupes can be distinguished from other fruits by their fleshy mesocarp and hard endocarp tissues. The peach could be considered a typical example of a drupe 42,43 . The exocarp (strictu sensu) is usually uniseriate and, in many fruits, represents the main protective layer of the pericarp. Notwithstanding, multiseriate exocarps are very seldom found in several fruits. This layer of tissue may be composed of regular epidermal cells of isodiametric or elongated shapes in surface view, but palisade cells could appear in some drupes. The outer tangential walls of the epidermal cells are generally thick and covered by a well-developed cuticle. Stomata, lenticels and trichomes are frequently present. Anthocyanins are regularly reported as contents in the cell sap of epidermal cells, producing the red, purple, blue and black colors of the fruit. Calcium oxalate crystals, raphides, siliceous impregnations, cutines, suberines and waxes also occur in specialized epidermis cells 42,45 .
The endocarp (strictu sensu) is often composed of smaller cells than the exocarp, especially in fleshy fruits, and they are frequently elongated. Drupes develop a distinctive hard multilayered endocarp formed by sclerenchymatic tissue. Differently from parenchymatous endocarps, its main function is to serve as supporting tissue and to protect the seed from injuries. The sclerenchyma is formed by thick secondary walls and strongly lignified cells denominated sclereids (short cells), fibers (long cells) and fiber-sclereids. Several types of sclereids could be present in the endocarp structure. The most common are brachysclereids, or stone cells, roughly isodiametric or somewhat elongated cells; macrosclereids, represented by elongated columnar cells; and osteosclereids or bone-cells, also columnar but with enlarged ends 42,45 .
As crystals are often found in the vicinity of sclerenchymatic cells, a crystal layer frequently accompanies the endocarp in drupes 42 . Inorganic deposits on plants consist mostly of calcium salts and anhydrides of silica which occurs in many plant families. Although calcium carbonate crystals are rarely present, the crystalline form of calcium oxalate is very common. They appear as prismatic, raphide, druse and styloid crystal morphologies. In some tissues, calcium oxalate crystals arise in cells that resemble adjacent, crystal-free cells. In others, the crystals Besides crystals, idioblasts can also produce oils, mucilage and tannins 42 and can be distributed all over the pericarp. Other secretory structures in drupes are the cavities and ducts, formed by several specialized cells which release those secondary metabolites to the intercellular space. This characteristic is very conspicuous in the mesocarp of drupes of the Anacardiceae family i.e. [46][47][48][49][50][51][52] , also referred to as lacunar mesocarp 44 .
The parenchyma plays an essential part in the formation of the fleshy mesocarp of drupes. It generally consists of large rounded or elongated (in a radial direction, generally the inner mesocarp) cells with thin walls and large vacuole. In ripe drupes, these are usually distinguished by their juiciness. The cell sap of parenchyma may be enriched with anthocyanin and sugar, together with tannins, acids and other contents. Crystals of different shapes may also occur and appear in specially preformed layers. Small intercellular spaces may eventually develop between the cells. As the parenchyma represents the edible part of drupes, it functions as a storage tissue in maturing fruits; it may contain starch, sugar, oil, fats, etheric oil (deposited in idioblasts, special glands or oil ducts) and latex in laticifers. Collenchymatic tissue is also found frequently in the fleshy mesocarp of drupes.

Results
pericarp. Both HRLM and X-ray MicroCT longitudinal sections show a three-layered internal structure marked by straight differences between cellular components (Fig. 3). In HLRM images, the exocarp corresponds to a multiseriate layer stained in brown-green. Cells are small and slightly regular with prevailing isodiametric shapes and some contents. No palisade cells were observed. A fine cuticle layer is present on the outer cells in some regions of the exocarp. The mesocarp appears as the most voluminous tissue, formed by large, irregular parenchymatic cells. Primary cell walls are well preserved (Fig. 4a). The endocarp is a multiseriate layer formed by mostly regular, small and highly lignified sclerenchymatic cells (brachysclereids) stained in green-blue (Fig. 4b).
No structure related to seed-embryo tissues was observed in HRLM images from the studied exemplar. In X-ray MicroCT images, virtual longitudinal and transversal sections allow one to distinguish the threelayered pericarp in terms of the differences between the morphologies of their composing cells (Fig. 3b,c), very similar to those observed in HRLM images. In sample SMC_01, the limits between the layers are barely defined, resulting in a gradate transition from one cell morphology to another. Any relevant contrast level differences based on variations of the density values of layers are evident (Fig. 4c-f). This character makes the segmentation of images less evident, even when non-algorithmic procedures are used. A membranous tissue remnant of the seed-embryo structure is clearly observable inside the central cavity. It appears weakly attached to the inner side of the endocarp (Fig. 5) (Supplementary Video S1 online).

presence of cellular contents. Most of the exocarp cells show the presence of brown-reddish stained
contents in the HRLM images. They could be associated with phenolic contents and/or pigments, commonly present in the skin of fruits (Fig. 4a). However, this characteristic is not observed in the X-ray MicroCT images, at least for the experimental parameters set for this study.
HRLM also shows crystal contents in the cells of the endocarp. Some of these cells contain calcium oxalate crystals which exhibit the typical extinction cross under polarized light (Fig. 6a,b). These particles were also evidenced in X-ray MicroCT images, which exhibit several white and compacted particles in the inner layer of some cells. Their high X-ray attenuation coefficient makes them easily distinguishable from tissues and other particles in the samples because they are comparatively denser and their composition is very different from surrounding material (Fig. 6c-e). A total of 19 particles were segmented from a random section of the endocarp of sample SMC_01, most of them measuring between 1.50E+07 μm 3 and 6.40 E+07 μm 3 (Fig. 7a,b). The www.nature.com/scientificreports/ reconstructed volumes of some of them suggest crystal-like multi-faceted structures defined by edges and angles (Fig. 6f). Notwithstanding, the resolution of the X-ray microtomographies in this study is not high enough to confirm this morphology. X-ray MicroCT images also show a second type of dense particle present in the mesocarp and inside the cavity, next to the open pointy region of the samples. They do not seem to be directly associated with intracellular contents and their 3D morphological analysis shows a total of 111 particles which are bigger than the crystal ones. The volume of most of these particles is between 6.6 E+05 μm 3 and 5.4 E+09 μm 3 , while others can reach 1.2 E+11 μm 3 (Fig. 7c,d). Their Sauter diameter (the diameter of the sphere with the same volume/area ratio of a given particle) ranges between 5.6 E+05 μm 3 and 2.9 E+06 μm 3 . They moreover lack the multi-faceted morphology and might be better described as amorphous concretion-like structures (Fig. 8). The complete list of volume Vascularisation. Several segments of the vascular system of the fruit can be observed in HRLM slices as vascular elements or fragments of bundles showing spiral thickened cell walls (Fig. 9a-c). They are present abundantly in the fleshy mesocarp layer, and are also morphologically distinguishable in X-ray MicroCT longitudinal virtual slices (Fig. 9d,e). The volumetric reconstruction of the more conspicuous bundles through the mesocarp (at least 12) reveals their distribution in the region near to the endocarp (Fig. 9f-i).
cavities and ducts. Some microtomed longitudinal sections show several cavities with sub-circular contours which is compatible to those previously described for several species of the Anacardiaceae. The diameter of these structures ranges from 100 µm to 250 µm and some of them still obtrude with contents ( Fig. 10a-d).
They were observed less frequently in the distal half of the sample. Nonetheless, the microtomographied exemplars showed some of them in this position in both longitudinal and transversal virtual sections of the mesocarp (Fig. 10e-j). Contents are not evident in 2D virtual slices. Their size ranges from 80 µm to 100 µm in transversal and longitudinal diameter, according to their 3D sub-spherical/drop morphology (Figs. 10k and 11a). Another type of structure with contents was also observed in the Micro-CT images, appearing as open spaces along the mesocarp, where some substances persist, probably due to their secretory function (Fig. 11b).

conclusions
The high resolution images obtained by both techniques were able to evidence each of the studied morphological traits in the archaeobotanical samples from Monte Castelo. HRLM and X-ray MicroCT are able to recognize the three differentiated layers of tissues composing the pericarp of the fruit and some of its most relevant histological characteristics: the presence of vascular tissues in the mesocarp; the existence of mineral particles within and between the tissues; and the occurrence of secretory structures. However, it is possible to point out some differences between the information obtained through each method.  www.nature.com/scientificreports/ Similarly, the use of polarized light coupled to HRLM enhances the analysis of mineral particles, giving a clear idea about their crystalline nature. This was not possible using microtomography. Despite this, the reconstructed volumes of the particles show differences between faceted and non-faceted morphologies, suggesting that a morphological analysis of crystalline minerals could be achieved focusing on appropriate X-ray MicroCT experimental settings.
Indeed, the possibility of performing 3D reconstructions of samples and visualizing their internal structure in a three-dimensional space allows the user to reveal the position and distribution of vascular bundles, mineral particles and secretory structures. Moreover, dynamic and non-destructive virtual slicing addresses the full range of slicing planes describing characteristics in a way that is not available to HRLM static slicing.
The non-destructive character of X-ray MicroCT helped to describe the presence of remnant tissues inside the central cavity of the samples which are absent in the microscopy images and probably represent the seed-embryo structures. As shown by microtomography, these structures are weakly tied to the rest of the fruit and could www.nature.com/scientificreports/ be detached and lost from the microtomed-for-microscopy sample. Likewise, some differences in the average size of the secretory structures were observed when measured by X-ray MicroCT or HRLM images. This effect could be linked to distortions originating in sample preparation procedures, including sample dehydration and resin embedding, meaning all quantitative comparisons between specimens should be made under the same imaging method 53 . Results obtained from both imaging methods are complementary in terms of a more exhaustive analysis of morphological characteristics of plant remains. However, X-ray MicroCT represents a suitable alternative analytic technique when sample preservation is required, for example when it is a rare, fragile specimen, or it will later be intended for further analysis. Based on the results of this study, the application of this analytical technique presents two types of challenge.
The first is of a technical nature, and has to do with the limitations of the technique in defining contours and subtle definition of details when applied to biological materials. Beyond the loss of information that this may represent for the study of plant remains, its resolution depends fundamentally on the technological development Figure 10. Sub-circular cavities observed in the microtomed longitudinal section from SMC_11 (a-d); Virtual X-ray MicroCT sections of cavities from sample SMC_01, suggesting their sub-circular/drop morphology (e-j); Volume reconstruction of one of these X-ray MicroCT imaged cavities (k) (CT Analyzer; ImageJ-FIJI, plug-in: 3D Viewer).
The second challenge is a methodological one, linked to some basic procedures of archaeobotanical studies. A good part of Archaeobotany research is focused on taxonomic identification of archaeological plant remains. This task is based primarily on morphological comparative analysis between old specimens and contemporary plant reference material. These references may be previous bibliographical studies, modern described plant exemplars itself or, as is generally the case, a combination of both. However, the analytical parameters used by morphology and plant anatomy to describe different plant taxa have been constructed based on two-dimensional, light or SEM microscopy images.
This means that a large number of the variables used to describe and compare plant material, relies on images of a different nature than three-dimensional X-ray MicroCT. It is therefore necessary to consider providing three-dimensional descriptions of reference plant material to archaeobotanical studies using this technique. Nonetheless, it is worth mentioning that virtual slicing would offer two-dimensional images equivalent to Light Microscopy, but properly related to volume, which could pave the way for new taxonomical standards.
The present study has shed light on some differences observed between analytic images based on the variations of the light reflected by the object (HRLM) from those based on the variations of the X-ray attenuation coefficient (and phase shifts) of the irradiated object (X-ray phase-contrast MicroCT). A more detailed analysis of these differences, their impact on the recognition and comparability of specific morphological characteristics, and the production of three-dimensional reference material, will be valuable in assisting the implementation of X-ray MicroCT to the study of archaeological plant remains.