Wood specimens of P. santalinus and P. tinctorius heartwood were supplied from Wood Collections of Chinese Academy of Forestry (WOODPEDIA). Indium-tin oxide (ITO) coated glass slides and double-sided conductive tape were purchased from China’s Sinopharm. ACN in chromatographic grade were purchased from Merck (Darmstadt, Germany). TFA were purchased from Sigma-Aldrich (Taufkirchen, Germany). ddH₂O was purified by filtration system (Sartorius arium-comfort II, Germany). 2-MBT were supplied by Bruker Beijing laboratory (Bruker, Germany).

Matrix coating was carried out with an ImagePrep electronic matrix sprayer (Bruker Daltonics, Germany). The spray of matrix was performed according to the spray method provided by the instrument. The MALDI matrix for each wood section detection was 6 mL 2-MBT, which was prepared by ACN (12mg/mL), TFA and double distilled water (80:0.2:19.8, v/v/v). In previous researches, 2-MBT, DMCA, 2,5-dihydroxybenzoic acid, α- cyano-4-hydroxycnmic acid, graphene oxide, and silver nanoparticles have been used as matrices. It was found that 2-MBT had better ionization efficiency, more effective signal peaks and better signal intensity for wood slices (He et al., 2019; Liu et al., 2019b).

Small wood blocks of about 1 cm³ were cut from each specimen, without any chemical or boiling treatment. Transverse sections of 20 µm in thickness were cut with a sliding microtome (Leica SM2010R, Germany). All sections were held flatly by slides for optical imaging and matrix coating. Before entering the mass spectrometer, the wood slices were sticked on indium-tin oxide (ITO) coated glass slides with double-sided conductive tape by using clean tweezers. 6 mL of 2-MBT matrix solution was uniformly sprayed onto the surface of the wood tissue by ImagePrep electronic matrix sprayer in two times (3 mL each time). After 2-3 seconds of spray each time, the slices were then slightly dried for 60-120 seconds. This process was repeated 50 times. The whole spraying process took about 40 minutes. After being sprayed, the wood slices were put into a vacuum tank for 10 minutes drying. During spraying, be sure to keep the spraying environment wet enough. The coated glass slides were installed on the imaging target plate, and then put it into the mass spectrometer.

All of the profiling and imaging experiments were carried out using a Bruker Autoflex Speed MALDI-time-of-flight (TOF)/TOF mass spectrometer (Bruker Daltonics, Germany), equipped with a 2000 Hz solid-state, Smartbeam Nd: YAG laser (355 nm). Sample was analyzed in the positive-ion and reflection mode with broadband detection. A mass range of m/z 400 to 1300 were performed. 50 μm laser raster step-sizes was used for the in situ detection of low-molecular weight compounds in wood tissue sections. When collecting profiling data, evenly print 30 points each time, 100 shots for each point (if the ideal peak cannot be found, then evenly print 30 points each time, 200 shots for each point). A total of 3000 shots are stacked as one time, and each wood slice was printed 3 times, and 3 profiling data were saved respectively. All wood samples were tested under the same conditions (including laser energy, detector gain, number of shots, etc.).

For data collected from MALDI-TOF-MSI, the Bruker’s FlexAnalysis 3.4 software was performed for the analysis of the preliminary mass spectra. According to the botanical structure characteristics of National standard of the People’s Republic of China: Hongmu (General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China, 2017), the fiber, vessel, axial parenchyma, wood ray and other tissues in the measurement area of each sample were defined. In each sample, these four tissue regions of wood structure were also areas of interest for the distribution analysis of metabolic compounds. The Bruker FlexImaging 4.1 software was used for the reconstruction of the ion map of low molecular weight compounds.

The representative heartwood samples were selected from the herbarium, and the chemical fingerprints of two species have been obtained through the detection of MALDI-FOF-MS. The flow chart of analytical methods was shown in Figure S1 . The overall average mass spectra of P. santalinus and P. tinctorius gained by MALDI-TOF-MSI under positive ionization mode were showed in Figure 2 . In the range of 400-1300 m/z, chemical signals were detected in the mass spectra of P. santalinus and P. tinctorius, and the P. santalinus samples showed more high abundant ions than P. tinctorius. However, when the wood chip samples of P. santalinus and P. tinctorius detected by DART-FTICR-MS, the peak range was 200-800 m/z (Zhang et al., 2019a). There was a significant difference between the number of peaks detected in the mass spectra of P. santalinus and P. tinctorius ( Figure S2 ). 202 peaks were detected in P. santalinus and 123 peaks were detected in P. tinctorius (Table S1). The P. santalinus sample has about 64% more peaks than P. tinctorius. The wavelength and shape of peaks in the spectra of P. santalinus and P. tinctorius were very similar. There were 96 common peaks, such as 421.66, 465.62, 497.58, 582.76, 596.77, 747.66 and 886.04 m/z, but the abundance of ions were more or less different. Several significantly different peaks can be observed from the spectra, such as 465.54, 529.47, 885.97 and 1202.48 m/z, which hardly existed in the spectra of P. tinctorius, and 497.49 and 1052.91 m/z, which could be found in the spectra of P. santalinus. These peaks with significant differences are benefit the differentiation of the two wood species.

Figure 3 showed the tissue heterogeneity distribution of some representative molecules. Ion images directly showed the existence of significantly different metabolites in the two kinds of wood slices. Such as m/z 504.7, 529.5, 616.7, 879.6, 895.6, 111.6, 1216.5, 1218.5, 1230.5, 1236.4 ( Figure 3B5, B15, B20, B21, B24, B29, B31, B32, B33 ), among these peaks, the ion abundance of P. tinctorius was very small or absent, while at the m/z 458.6, 534.8, 582.7, 854.7, 926.6, 1052.9, 1140.6, 1272.9 ( Figure 3A3, A6, A11, A19, A22, A23, A25, A34 ) etc., the ion abundance of P. santalinus was very small or absent. This result is consistent with the specific peaks obtained from the mass spectrum, showed in Table S1. Ion image showed that some metabolites with low relative content (in blue) can also be observed and detected by MALDI-TOF-MSI. Common LC-MS will homogenize plant tissue, and metabolites with low content cannot be detected, but MALDI-TOF-MS can detect metabolites with low content at a specific location. MALDI-TOF-MS imaging makes it easy to intuitively lock the category of differential compounds of two wood species, and facilitate further exploration through data analysis. As showed in Figure 4 , volcano plot displayed two important indicators (Fold change/p-Value). With more than 1.5 times of change and P<0.05 as the screening condition, the differential metabolites that can serve as marker compounds between the two wood species can be screened intuitively and reasonably. The differential metabolites screened from the volcano plot include all the significant differential metabolites shown in the ion image. It fully shows that mass spectrometry imaging has the ability to screen differential expression factors among samples, and is more intuitive than the results of mathematical statistics.

The ion images showed the heterogeneous distribution and relative abundance of different metabolites in wood xylem tissues, which were obviously related to the botanical structure of P. santalinus and P. tinctorius. It mainly involves four types of tissues, namely fiber, axial parenchyma cell, wood ray and vessel ( Figure 3A1, B1 ). The location of metabolites in wood tissue structure can be preliminarily determined by corresponding optical images of tissue section. The anatomical characteristics of the parenchyma cells of P. santalinus and P. tinctorius are tangential banded, aliform or confluent, with obvious structural characteristics, so it is better to locate and compare the distribution of metabolites in the ion images. Some metabolites are obviously distributed in parenchyma cells, such as m/z 566.7, 596.7, 610.7, 747.6 and 749.6, the relative content of metabolites in the parenchyma cells of P. santalinus was about 50-75%. While in the parenchyma cells of P. tinctorius, the relative content of metabolites m/z 610.7, 747.6 and 749.6 can also reach about 50%. Wood fiber is the cell type with the largest tissue proportion in the Pterocarpus species. The high abundant metabolites are distributed in the wood fiber cells of the two species, with the widest relative content range. In P. santalinus, m/z 567.7, 583.7, 596.7, 610.7, 763.6, 879.6, 1111.6, 1202.5, 1203.5, 1217.5 and 1236.4 etc. were involved, and m/z 596.7, 763.6, 854.7 and 926.6 etc. were involved in P. tinctorius. There were also many metabolites with high content in the wood ray cells. Because both P. santalinus and P. tinctorius are all rays storied, and the rays are thin and densely arranged, the ion image on the wood transverse section showed a clear distribution.

The distribution quantity of metabolites in P. santalinus is obviously more than that in P. tinctorius ( Figure 5 ). The ion image showed that there were few high abundance metabolites in the vessel, which may also be related to the fine wood structure of the two species, namely, the vessel diameter was small, the vessel arranged not too tightly (vessels exclusively solitary or few in radial multiples of 2-3 cells, scattered in 2-5 cells/mm²) and the distribution density was also small, so the vessel tissue structure was not fully displayed. In general, there were more metabolites in fibers, parenchyma cells and rays, and less metabolites in vessel were observed by ion imaging, which has a certain relationship with tissue ratio and resolution of ion imaging. The localization of all metabolites revealed by ion imaging combined the chemical information and histological characteristics of plants, which is helpful to understand the coevolution and provenance relationship between two species of Pterocarpus wood. MALDI-TOF-MSI analysis made up the gap between morphology and chemistry in complex biological samples.

PCA (Principal Component Analysis) is a commonly used multivariate statistical method to investigate the correlation between multiple variables. Through linear dimensionality reduction of data, it can generate features that are easy to be found and more easily understood by humans (Rasmus and Age 2014; Hur et al., 2010). The PCA score plot of the MALDI-TOF-MS spectra of P. santalinus and P. tinctorius were shown in Figure S3 . The results showed clustering behavior related to the wood species. All samples were divided into two independent groups. The separation trend between the two groups was good, and there was no abnormal point. PCA can effectively process high-dimensional data, but is not sensitive to variables with small correlation. PLS-DA (Partial least squares Discriminant Analysis) integrates multiple linear regression analysis, canonical correlation analysis and principal factor analysis, which can avoid potential problems such as data non normal distribution, factor indeterminacy and model failure to identify. Variable importance on projection (VIP) is the variable weight value of the PLS-DA model variable, which is used to measure the influence intensity and interpretation ability of the accumulation difference of various metabolites on the classification and discrimination of each group of samples. VIP value larger than 1 is a common screening standard for different metabolites (Ahn et al., 2020). The variables are sorted by combining the VIP value and the p value in the t-test, and the selected highly correlated variables can be considered as candidate variables for species differences (Rajalahti et al., 2009). Due to the high dimensional and massive characteristics of mass spectrometry data, the principle of VIP value larger than 1.5 and P value less than 0.05 was used to screen potential markers of interspecific differences by combining single-dimensional and multi-dimensional statistical analysis methods. Using this screening principle, fifteen variables were considered as markers ( Figure S4 and Table 1 ). The fifteen variables were all in the significant screening results of interspecific differences in ion images obtained by MALDI-TOF-MSI ( Figure 3 ). Each variable has high VIP value and strong species difference expression. The results of heat map ( Figure S5 ) and correlation analysis ( FigureS6 ) also confirm this result. It is indicated that the fifteen variables can be used as chemical markers for samples classification of P. santalinus and P. tinctorius, which can effectively reduce the time and workload of locating compounds and massive data evaluation.

For dry wood, especially the identification of chemical components of wood specimens that have been dried and preserved for many years or even decades, it has always been a difficult point in the field of wood identification, whether it is the use of DNA extraction technology, or the use of chemical component chromatography/mass spectrometry technology for species level identification (Jiao et al., 2018b; Zhang et al., 2019b). Compared with fresh plant materials, such as leaves, roots, stems, or xylem of fresh wood, secondary mass spectrometry can be obtained to identify compounds by MALDI-TOF-MS/MS (Wang et al., 2016; Kuo et al., 2019; Li et al., 2020; Shen et al., 2022). However, for dry wood, compared with fresh tissue, it is difficult to ionize. We have not succeeded in obtaining secondary mass spectrometry using MALDI-TOF-MS/MS in this study. If the secondary mass spectrometry of dried wood samples can be obtained, combined with the ion image of MALDI-TOF-MSI, it can improve the understanding of the unique functions of individual species and their biological functions in plant growth and development, as well as the impact on the macro and micro properties of wood.

MALDI-TOF-MSI is a powerful, cost-effective, fast and robust species classification technology, which has been successfully applied to the classification and identification of microorganisms, herbal active substances and proteins. However, the wood MALDI-TOF-MSI database is not yet available, which does not meet the needs of wood identification and classification. In order to further distinguish species that are closely related and difficult to classify, it is increasingly important to establish, categorize and expand databases.