Opening the black box: explainable deep-learning classification of wood microscopic image of endangered tree species.

BACKGROUND: Traditional method of wood species identification involves the use of hand lens by wood anatomists, which is a time-consuming method that usually identifies only at the genetic level. Computer vision method can achieve "species" level identification but cannot provide an explanation on what features are used for the identification. Thus, in this study, we used computer vision methods coupled with deep learning to reveal interspecific differences between closely related tree species. RESULT: A total of 850 images were collected from the cross and tangential sections of 15 wood species. These images were used to construct a deep-learning model to discriminate wood species, and a classification accuracy of 99.3% was obtained. The key features between species in machine identification were targeted by feature visualization methods, mainly the axial parenchyma arrangements and vessel in cross section and the wood ray in tangential section. Moreover, the degree of importance of the vessels of different tree species in the cross-section images was determined by the manual feature labeling method. The results showed that vessels play an important role in the identification of Dalbergia, Pterocarpus, Swartzia, Carapa, and Cedrela, but exhibited limited resolutions on discriminating Swietenia species. CONCLUSION: The research results provide a computer-assisted tool for identifying endangered tree species in laboratory scenarios, which can be used to combat illegal logging and related trade and contribute to the implementation of CITES convention and the conservation of global biodiversity.

Analyses of high spatial resolution datasets identify genes associated with multi-layered secondary cell wall thickening in Pinus bungeana.

BACKGROUND AND AIMS: Secondary cell wall (SCW) thickening is a major cellular developmental stage determining wood structure and properties. Although the molecular regulation of cell wall deposition during tracheary element differentiation has been well established in primary growth systems, less is known about the gene regulatory processes involved in the multi-layered SCW thickening of mature trees. METHODS: Using third-generation [long-read single-molecule real-time (SMRT)] and second-generation [short-read sequencing by synthesis (SBS)] sequencing methods, we established a Pinus bungeana transcriptome resource with comprehensive functional and structural annotation for the first time. Using these approaches, we generated high spatial resolution datasets for the vascular cambium, xylem expansion regions, early SCW thickening, late SCW thickening and mature xylem tissues of 71-year-old Pinus bungeana trees. KEY RESULTS: A total of 79 390 non-redundant transcripts, 31 808 long non-coding RNAs and 5147 transcription factors were annotated and quantified in different xylem tissues at all growth and differentiation stages. Furthermore, using this high spatial resolution dataset, we established a comprehensive transcriptomic profile and found that members of the NAC, WRKY, SUS, CESA and LAC gene families are major players in early SCW formation in tracheids, whereas members of the MYB and LBD transcription factor families are highly expressed during late SCW thickening. CONCLUSIONS: Our results provide new molecular insights into the regulation of multi-layered SCW thickening in conifers. The high spatial resolution datasets provided can serve as important gene resources for improving softwoods.

Retrieving complete plastid genomes of endangered Guibourtia timber using hybridization capture for forensic identification and phylogenetic analysis.

The high economic value and increased demand for timber have led to illegal logging and overexploitation, threatening wild populations. In this context, there is an urgent need to develop effective and accurate forensic tools for identifying endangered Guibourtia timber species to protect forest ecosystem resources and regulate their trade. In this study, a hybridization capture method was developed and applied to explore the feasibility of retrieving complete plastid genomes from Guibourtia sapwood and heartwood specimens stored in a xylarium (wood collection). We then carried out forensic identification and phylogenetic analyses of Guibourtia within the subfamily Detarioideae. This study is the first to successfully retrieve high-quality plastid genomes from xylarium specimens, with 76.95-99.97% coverage. The enrichment efficiency, sequence depth, and coverage of plastid genomes from sapwood were 16.73 times, 70.47 times and 1.14 times higher, respectively, than those from heartwood. Although the DNA capture efficiency of heartwood was lower than that of sapwood, the hybridization capture method used in this study is still suitable for heartwood DNA analysis. Based on the complete plastid genome, we identified six endangered or commonly traded Guibourtia woods at the species level. This technique also has the potential for geographical traceability, especially for Guibourtia demeusei and Guibourtia ehie. Meanwhile, Bayesian phylogenetic analysis suggested that these six Guibourtia species diverged from closely related species within the subfamily Detarioideae ca. 18 Ma during the Miocene. The DNA reference database established based on the xylarium specimens provides admissible evidence for diversity conservation and evolutionary analyses of endangered Guibourtia species.

Spatial organization and connectivity of wood rays in Pinus massoniana xylem based on high-resolution µCT-assisted network analysis.

MAIN CONCLUSION: Spatial organization and connectivity of wood rays in Pinus massoniana was comprehensively viewed and regarded as anatomical adaptions to ensure the properties of rays in xylem. Spatial organization and connectivity of wood rays are essential for understanding the wood hierarchical architecture, but the spatial information is ambiguous due to small cell size. Herein, 3D visualization of rays in Pinus massoniana was performed using high-resolution µCT. We found brick-shaped rays were 6.5% in volume fractions, nearly twice the area fractions estimated by 2D levels. Uniseriate rays became taller and wider during the transition from earlywood to latewood, which was mainly contributed from the height increment of ray tracheids and widened ray parenchyma cells. Furthermore, both volume and surface area of ray parenchyma cells were larger than ray tracheids, so ray parenchyma took a higher proportion in rays. Moreover, three different types of pits for connectivity were segmented and revealed. Pits in both axial tracheids and ray tracheids were bordered, but the pit volume and pit aperture of earlywood axial tracheids were almost tenfold and over fourfold larger than ray tracheids. Contrarily, cross-field pits between ray parenchyma and axial tracheids were window-like with the principal axis of 31.0 µm, but its pit volume was approximately one-third of axial tracheids. Additionally, spatial organization of rays and axial resin canal was analyzed by a curved surface reformation tool, providing the first evidence of rays close to epithelial cells inward through the resin canal. Epithelial cells had various morphologies and large variations in cell size. Our results give new insights into the organization of radial system of xylem, especially the connectivity of rays with adjacent cells.

Anatomical adaptions of pits in two types of ray parenchyma cells in Populus tomentosa during the xylem differentiation.

Pits in ray parenchyma cells are important to understand the functional anatomy of the ray parenchyma network in the xylem but have been less studied. Herein, pits in two types of ray parenchyma cells, contact cells and isolation cells, across different developmental stages were qualitatively studied using 48-year-old Populus tomentosa trees. The timing of differentiation and death was determined by histochemical staining and polarized light microscopy. The dimension, shape and density of pits as well as cell wall thickness were measured using SEM and optical microscopy images of semi-thin radial sections and macerated ray parenchyma cells, and analyzed by multi-factor analyses of variance. Results showed that secondary wall thickening and lignification of contact cells begun near the cambium, contrarily those of isolation cells have started until the transition zone. But even in the sapwood, contact cell walls were still much thinner than isolation cell walls. Moreover, district anatomical adaptions of pits during the xylem differentiation were present between horizontal walls and tangential walls, between contact cells and isolation cells. Ray pits were simple to slightly bordered, whereas sieve-like pits were only shown on tangential walls of isolation cells. Pit density of horizontal walls was similar between contact cells and isolation cells, nevertheless greater pits were present on tangential walls, especially for isolation cells. In addition, pits of ray parenchyma cells in the heartwood were smaller and more bordered than those in the sapwood, particularly on the horizontal walls. Moreover, isolation cells had pits with the smaller dimensions, greater pits on the tangential walls, more bordered pits on horizontal walls, as well as longer and narrower cell morphology with much thicker cell walls than contact cells. To a certain extent, all these anatomical adaptations were developed to ensure distinct functions of the two types of ray parenchyma cells in the xylem and finally to support tree growth in demand.

Metabolome and transcriptome analyses identify the plant immunity systems that facilitate sesquiterpene and lignan biosynthesis in Syringa pinnatifolia Hemsl.

BACKGROUND: Syringa pinnatifolia Hemsl. is a shrub belonging to the Oleaceae family. The peeled woody stems and roots of S. pinnatifolia are used in Chinese traditional medicine. This plant has been used for centuries, and modern pharmacological research has revealed its medicinal value. However, the wild populations of S. pinnatifolia have been decreasing, and it has been listed as an endangered plant in China. To elucidate the molecular mechanism leading to the synthesis of the major components of S. pinnatifolia for its further development and sustainable use, this study compared peeled stems and twigs at the metabolic and molecular levels. RESULTS: Peeled stems with the purple substance visible (SSP) and peeled twigs without the purple substance (TSP) were compared at different levels. Microscopic observation showed resin-like fillers in SSP and wood fiber cell walls approximately 1.0 µm thicker than those in TSP (wood fiber cell thickness approximately 2.7 µm). In addition, 104 volatile organic compounds and 870 non-volatile metabolites were detected in the non-targeted and widely-targeted metabolome analyses, respectively. Among the 76 differentially accumulated metabolites (DAMs) detected, 62 were up-accumulated in SSP. Most of these DAMs were terpenes, of which 90% were identified as sesquiterpenes in the volatile organic compound analysis. In the analysis of the non-volatile metabolites, 21 differentially accumulated lignans were identified, of which 18, including five subtypes, were accumulated in SSP. RNA sequencing revealed 4,421 upregulated differentially expressed genes (DEGs) and 5,522 downregulated DEGs in SSP compared with TSP, as well as 33,452 genes that were not differentially expressed. Analysis of the DEGs suggested that sesquiterpenes and lignans were mostly biosynthesized via the mevalonate and phenylpropanoid pathways, respectively. Additionally, in SSP, the enriched Gene Ontology terms included response to biotic stimulus and defense response, while the enriched Kyoto Encyclopedia of Genes and Genomes pathways included plant-pathogen interaction and many other pathways related to plant immunity. CONCLUSIONS: This study provides metabolome and transcriptome information for S. pinnatifolia, suggesting that biotic stimuli, including pathogens, are potential and valuable approaches to promoting the biosynthesis of the metabolites linked to the medicinal properties of this plant.

[Microscopic histochemical comparison during mountain-agarwood formation].

Mountain-agarwood plays an important role in ethnic medicine in China for its pharmaceutical value. Modern pharmacological researches demonstrated that mountain-agarwood was effective for its anti-myocardial ischemia, antibacterial, anti-inflammatory, antitumor and analgesic effects. Mountain-agarwood derives from the peeled roots, stems or twigs of Syringa pinnatifolia which belongs to Syringa genus. It often depends on the purple substance and fragrance to estimate the formation of mountain-agarwood. However, the mechanism of mountain-agarwood formation has not been reported. To observe the microcosmic change in the process during the formation of mountain-agarwood, this study described the microscopic and histochemical characteristics of mountain-agarwood formation through histochemical staining. Our results showed that a significant difference of the distribution of tyloses existed during mountain-agarwood formation. It was observed that inchoate mountain-agarwood had more starch granules and viable cells than mountain-agarwood formed with high level or low level. The amount of polysaccharide and degree of lignification were increased during the mountain-agarwood formation. The results indicated that the mountain-agarwood, which meets the quality requirements for pharmaceutical use, contained the following characteristics: a large amount of purple tyloses in heartwood; yellow-brown tyloses distributing in heartwood and sapwood which were less in the latter; lignification with high level; a few viable cells; lots of polysaccharide and few starch granules in xylem rays cell. This study is aimed to reveal the change of histochemical characteristics during mountain-agarwood formation, and lay the foundation for exploring the mechanism of mountain-agarwood formation.

A strategy for developing high-resolution DNA barcodes for species discrimination of wood specimens using the complete chloroplast genome of three Pterocarpus species.

MAIN CONCLUSION: A method for extraction of wood DNA and a strategy for designing high-resolution barcodes for wood were developed. Ycf1b was the prioritized barcode to resolve the Pterocarpus wood species studied. DNA barcoding, an effective tool for wood species identification, mainly focuses on universal barcodes and often lacks high resolution to differentiate species, especially for closely related taxa within the same genus. Therefore, more highly informative DNA barcodes need to be identified. This study is the first to report a strategy for developing specific DNA barcodes of wood tissues. The complete chloroplast genomes of leaf samples of three Pterocarpus species, i.e., P. indicus, P. santalinus, and P. tinctorius, were sequenced, and thereafter, the most variable DNA regions were identified on the scale of the complete chloroplast genomes. Finally, wood DNA was extracted from 30 wood specimens of the three Pterocarpus species, and DNA recovery rates of the selected regions were tested for applicability to verification on the wood specimens studied. The seven regions with the most variation (rpl32-ccsA, rpl20-clpP, trnC-rpoB, ycf1b, accD-ycf4, ycf1a, and psbK-accD) were identified from the chloroplast genome by quantifying nucleotide diversity (Pi > 0.02), which was remarkably higher than that of the plant universal barcodes (rbcL, matK, and trnH-psbA) and the previously reported barcodes (ndhF-rpl32 and trnL-F) used for phylogenetic analysis in Pterocarpus. After comprehensive evaluation of species discrimination ability and applicability, the ycf1b region performed well in terms of the recovery success rate (76.7%) and species identification (100%) for wood specimens of the three Pterocarpus species, and was identified as the preferred high-resolution chloroplast barcode for selected Pterocarpus species. It will offer technical support for curbing illegal timber harvesting activities and for conserving endangered and valuable wood species.

Machine learning approaches outperform distance- and tree-based methods for DNA barcoding of Pterocarpus wood.

MAIN CONCLUSION: Machine-learning approaches (MLAs) for DNA barcoding outperform distance- and tree-based methods on identification accuracy and cost-effectiveness to arrive at species-level identification of wood. DNA barcoding is a promising tool to combat illegal logging and associated trade, and the development of reliable and efficient analytical methods is essential for its extensive application in the trade of wood and in the forensics of natural materials more broadly. In this study, 120 DNA sequences of four barcodes (ITS2, matK, ndhF-rpl32, and rbcL) generated in our previous study and 85 downloaded from National Center for Biotechnology Information (NCBI) were collected to establish a reference data set for six commercial Pterocarpus woods. MLAs (BLOG, BP-neural network, SMO and J48) were compared with distance- (TaxonDNA) and tree-based (NJ tree) methods based on identification accuracy and cost-effectiveness across these six species, and also were applied to discriminate the CITES-listed species Pterocarpus santalinus from its anatomically similar species P. tinctorius for forensic identification. MLAs provided higher identification accuracy (30.8-100%) than distance- (15.1-97.4%) and tree-based methods (11.1-87.5%), with SMO performing the best among the machine learning classifiers. The two-locus combination ITS2 + matK when using SMO classifier exhibited the highest resolution (100%) with the fewest barcodes for discriminating the six Pterocarpus species. The CITES-listed species P. santalinus was discriminated successfully from P. tinctorius using MLAs with a single barcode, ndhF-rpl32. This study shows that MLAs provided higher identification accuracy and cost-effectiveness for forensic application over other analytical methods in DNA barcoding of Pterocarpus wood.

DNA Barcode Authentication and Library Development for the Wood of Six Commercial Pterocarpus Species: the Critical Role of Xylarium Specimens.

DNA barcoding has been proposed as a useful tool for forensic wood identification and development of a reliable DNA reference library is an essential first step. Xylaria (wood collections) are potentially enormous data repositories if DNA information could be extracted from wood specimens. In this study, 31 xylarium wood specimens and 8 leaf specimens of six important commercial species of Pterocarpus were selected to investigate the reliability of DNA barcodes for authentication at the species level and to determine the feasibility of building wood DNA barcode reference libraries from xylarium specimens. Four DNA barcodes (ITS2, matK, ndhF-rpl32 and rbcL) and their combination were tested to evaluate their discrimination ability for Pterocarpus species with both TaxonDNA and tree-based analytical methods. The results indicated that the combination barcode of matK + ndhF-rpl32 + ITS2 yielded the best discrimination for the Pterocarpus species studied. The mini-barcode ndhF-rpl32 (167-173 bps) performed well distinguishing P. santalinus from its wood anatomically inseparable species P. tinctorius. Results from this study verified not only the feasibility of building DNA barcode libraries using xylarium wood specimens, but the importance of using wood rather than leaves as the source tissue, when wood is the botanical material to be identified.

DNA barcoding of vouchered xylarium wood specimens of nine endangered Dalbergia species.

MAIN CONCLUSION: ITS2+ trnH - psbA was the best combination of DNA barcode to resolve the Dalbergia wood species studied. We demonstrate the feasibility of building a DNA barcode reference database using xylarium wood specimens. The increase in illegal logging and timber trade of CITES-listed tropical species necessitates the development of unambiguous identification methods at the species level. For these methods to be fully functional and deployable for law enforcement, they must work using wood or wood products. DNA barcoding of wood has been promoted as a promising tool for species identification; however, the main barrier to extensive application of DNA barcoding to wood is the lack of a comprehensive and reliable DNA reference library of barcodes from wood. In this study, xylarium wood specimens of nine Dalbergia species were selected from the Wood Collection of the Chinese Academy of Forestry and DNA was then extracted from them for further PCR amplification of eight potential DNA barcode sequences (ITS2, matK, trnL, trnH-psbA, trnV-trnM1, trnV-trnM2, trnC-petN, and trnS-trnG). The barcodes were tested singly and in combination for species-level discrimination ability by tree-based [neighbor-joining (NJ)] and distance-based (TaxonDNA) methods. We found that the discrimination ability of DNA barcodes in combination was higher than any single DNA marker among the Dalbergia species studied, with the best two-marker combination of ITS2+trnH-psbA analyzed with NJ trees performing the best (100% accuracy). These barcodes are relatively short regions (<350 bp) and amplification reactions were performed with high success (≥90%) using wood as the source material, a necessary factor to apply DNA barcoding to timber trade. The present results demonstrate the feasibility of using vouchered xylarium specimens to build DNA barcoding reference databases.