DBALM: A novel method for identifying ornamental flowering plants based on DNA barcodes–leaf morphology

Abstract Whereas the presence of flowers on ornamental flowering plants is essential for their identification via traditional methods, ornamental flowering plants cannot be reliably identified in non‐flowering stages likewise. Here, DBALM (DNA Barcodes–Leaf Morphology), a new approach that combines DNA barcoding data with micromorphological features of the leaf epidermis and that is not limited by the flowering stage, was used to identify 16 evergreen rhododendron cultivars. First, the sequences of DNA barcodes, ITS, matK, psbA‐trnH, and rbcL, were obtained from the DNA of leaves. Phylogenetic analysis was conducted to clarify the groupings among all the samples based on the four markers. Then, microscopic features of the leaf epidermis were used to further distinguish individuals from the same clade. DNA barcoding permitted the 16 cultivars to be divided into eight groups. The microscopic features of the leaf epidermis permitted cultivars within the same clade to be distinguished. The matK + psbA‐trnH combination was the most effective barcode combination in this study. In addition, the new primer matK‐Rh_R was designed, and it increased the amplification rate of evergreen rhododendron cultivars to 100%. In sum, DBALM was capable of accurately identifying the 16 evergreen rhododendron cultivars using data collected from a single leaf in the vegetative growth stage. This method can greatly facilitate the identification and breeding of ornamental flowering plants.


| INTRODUC TI ON
The diversity of economically important ornamental flowering plants is astounding. One of the major challenges in both horticultural research and ornamental plant industry is accurately identifying cultivars. Indeed, the misidentification of cultivars is the cause of major losses to the ornamental plant industry and has major implications for horticultural research. The accurate identification of ornamental flowering plants is thus essential for cultivation and breeding of horticultural plants, including international trade. Although experts in ornamental flowering plants have often been employed for identification, there are several drawbacks associated with this approach (Eagles et al., 2001;Li et al., 2020;Nadeem et al., 2017).  (Arif et al., 2010;Azizi et al., 2021;Lu et al., 2019;Sharma et al., 2020;Wang et al., 2018;Yali, 2022). Some other molecular marker techniques are costly, but the amplification rate and stability of the results of these methods are low (Arif et al., 2010;Azizi et al., 2021;Nadeem et al., 2017;Shi et al., 2016;Yali, 2022). DNA barcoding has been widely used for classification and identification of plants because of its simplicity and high repeatability (Liu et al., 2021;Vere et al., 2015). However, the resolution of DNA barcoding is not sufficiently high to permit closely related cultivars to be distinguished, as cultivars vary in ploidy and bud mutations (Azizi et al., 2021;Shi et al., 2016;Yan et al., 2015). Morphological features of vegetative organs, such as micromorphology of the leaf epidermis, ultrastructural features, and structure of leaf veins, have frequently been used to classify and identify plants (Cristina et al., 2008;Franklin, 1945;Keating, 1984;Payne, 1979;Silva et al., 2016;Wang et al., 2006).
Observations of micromorphological features of the leaf epidermis can be easily made. The micromorphological features of bud mutants and polyploid cultivars differ from those of stock plants (Chen et al., 2009(Chen et al., , 2010Yi et al., 2016)

| Identification based on DNA barcodes
Molecular samples from fresh material were obtained via the drying treatment method at 40°C to minimize the degradation of DNA (Shen et al., 2017(Shen et al., , 2022. The mCTAB method was used to extract total DNA (Li et al., 2013). Four markers used in previous studies of members of the genus Rhododendron (Fu et al., 2022;Liu, 2011;Liu et al., 2012;Ma et al., 2013;Yan et al., 2015), ITS, rbcL, matK, and psbA-trnH, were used in our DNA barcoding analysis. Primers and references for the PCR protocol are provided in Table 3. The amplification success rate of matK was low (Liu, 2011;Liu et al., 2012;Ma et al., 2013;Yan et al., 2015). We could not obtain a single positive TA B L E 1 Information on the experimental materials used for the collection of DNA barcoding data and micromorphological data of the leaf epidermis of Rhododendron.  (Nguyen et al., 2015), respectively. We conducted two independent simultaneous runs (three hot chains and one cold chain for each run), and sampling was conducted every 1000 generations for 1,000,000 generations. The average standard deviation of the split frequencies was less than 0.01. ML analysis was conducted by using 10,000 ultrafast bootstraps (Minh et al., 2013). The BI and ML models (Kalyaanamoorthy et al., 2017;Lanfear et al., 2017) are shown in Table A1 in Appendix 1. The tree was plotted via using iTOL (https://itol.embl.de/).

| Micromorphological trait-based identification
Water-mounted slides of leaf epidermis were prepared following the method of Wang et al. (2007). Four 1 × 1-cm pieces of the leaves of each individual were collected and used to make micromorphological observations. A Nikon E800 optical microscope, NIS-Elements F 3.0 software, and ImageJ v1.8.0 software (http://rsbweb.nih.gov/ ij/) were used to characterize the size of cells, anticlinal wall pattern, height and width of anticlinal wall waves, stomata density and the presence of hairs. In this study, a measurement ( Figure 1) and data analysis was conducted on the height and width ratio of the anticlinal wall waves. Variation in character among cultivars is intermittent when there is no overlap in the boxplots, and the character can be used to distinguish cultivars. The terminology of the micromorphological traits of the leaf epidermis followed that of Dilcher (1974).

| Efficacy of DBALM
The clades generated via the DNA barcoding data were divided into  into different subgroups. The resolution provided by the matK + psbA-trnH combination for the identification of evergreen rhododendron cultivars was the same as that provided by the ITS + matK + psbA-trnH + rbcL combination. The highest resolution for the identification of the evergreen rhododendron cultivars is provided by the four-marker combination, according to previous studies (Yan et al., 2015). Thus, the matK + psbA-trnH combination is sufficient for distinguishing between evergreen rhododendron cultivars via DNA barcoding.
Recent studies (Fu et al., 2022) have shown that DNA barcoding has a 35% success rate for identifying wild species in the genus Rhododendron. But when the chloroplast genome combined with nuclear genes was used for identification, success rate was increased to 55%. In contrast, the identification of cultivars of evergreen rhododendrons might be more complicated than that of wild species because of the close relationships among cultivars. In fact, success rate of the identification of the cultivars of evergreen rhododendrons using the same DNA barcodes (ITS + matK + psbA-trnH + rbcL) in this study was 10% lower than that of wild species in the genus Rhododendron (Fu et al., 2022). However, when DBALM was used to identify the cultivars of evergreen rhododendrons, success rate was 100%. Therefore, we speculate that DBALM has a higher success F I G U R E 3 Phylogenetic tree built via the Bayesian inference and maximum likelihood methods using the ITS + matK + psbA-trnH + rbcL combination of markers. Note: Bayesian inference (BI) and ML bootstrap supports are shown at the nodes (BI/ML). rate than the use of chloroplast genomes and nuclear genes for the identification of cultivars in the genus Rhododendron and that it has broad application prospects. DBALM enhanced the efficiency and accuracy of identification of evergreen rhododendron cultivars.
DNA barcoding and leaf epidermis microscopic characteristics are widely used in the classification of horticultural cultivars, such as banana and orange (Yi et al., 2016;Zulkifli, 2013). Therefore, although this study only used 16 evergreen rhododendrons as an example, we speculate that this new method could be effective for the identification of a variety of cultivated evergreen plants. The advantages of DBALM are manifold. First, cultivars can be identified at any growth stage, including non-flowering stages. Second, few materials are needed for identification using this method that minimizes damage to the plant. Third, cultivars can be performed by those lacking expertise in particular plant groups. Fourth, the results of the identification procedure are repeatable and objective.

F I G U R E 4
Phylogenetic tree built via the Bayesian inference (BI) and maximum likelihood (ML) methods using the matK + psbA-trnH combination of markers. Note: Bayesian inference (BI) and ML bootstrap supports are shown at the nodes (BI/ML). The groups are indicated in different colors. The inner rings with cultivar names indicate the clades identified using DNA barcoding; the outer rings with letters indicate the groups identified using leaf epidermal micromorphological characteristics.
Fifth, the procedure can be performed relatively rapidly and is less costly. However, the accuracy of DBALM identification is highly dependent on the comprehensiveness and accuracy of the database. Although genomics is currently a heated topic in the fields of identification, taxonomy, phylogeny, evolution, and biodiversity conservation, it might not be the most effective approach in many cases, especially for cultivar identification. The potential value of morphological characteristics, including micromorphological characteristics, should not be ignored.

ACK N OWLED G M ENTS
We thank Dr. Dong Zhang of Lanzhou University for his help with the phylogenetic analysis. We thank Ling Guo of China National Botanical Garden for her guidance in identifying horticulture plants.
We thank Yanlei Liu of Hebei University of Engineering for his help in writing. We thank Shuangling Li of Hebei Normal University for her help in writing.

DATA AVA I L A B I L I T Y S TAT E M E N T
The alignments of the ITS, matK, psbA-trnH, and rbcL sequences of  F I G U R E A 5 Phylogenetic trees built via the BI and ML methods using the sequences of four makers. (i) psbA-trnH + rbcL; (j) ITS + matK + psbA-trnH.
(i) (j) F I G U R E A 6 Phylogenetic trees built via the BI and ML methods using the sequences of four makers. (k) ITS + matK + rbcL; (l) ITS + psbA-trnH + rbcL.
(k) (l) F I G U R E A 7 Phylogenetic trees built via the BI and ML methods using the sequences of four makers. (m) matK + psbA-trnH + rbcL.

(m)
A PPEN D I X 4 F I G U R E A 8 Adaxial epidermal micromorphological characteristics of evergreen rhododendron cultivars. Note: The lower right corner of the small figure represents the specimen collection numbers.