Introduction
Conyza Less. (Asteraceae: Astereae) is a genus that comprises approximately 150 species worldwide (TICA 2023), including troublesome weedy species with accessions resistant to as many as six site-of-action herbicides (Heap Reference Heap2022). Broadly, Conyza weeds are closely related species because of their shared traits, environments, and niches and often overlap in cropping regions across the globe (Thébaud and Abbott Reference Thébaud and Abbott1995). For example, hairy fleabane [Conyza bonariensis (L.) Cronquist] and horseweed [Conyza canadensis (L.) Cronquist] co-occur in several field crops in at least 36 countries (Bajwa et al. Reference Bajwa, Sadia, Ali, Jabran, Peerzada and Chauhan2016). In addition, varieties of C. bonariensis, C. canadensis, and Sumatran fleabane [Conyza sumatrensis (Retz.) E. Walker] are often reported in floristic analyses of the genus (Sancho Reference Sancho2014). For instance, Conyza sumatrensis (Retz.) E. Walker var. leiotheca (S.F. Blake) Pruski & G. Sancho is a glabrous variety restricted to the Americas, while Conyza sumatrensis (Retz.) E. Walker var. sumatrensis is hirsute and globally dispersed (Pruski and Sancho Reference Pruski and Sancho2006). The presence of species and varieties may create complex infestations of Conyza spp., including weeds with differential responses to management practices, mainly chemical control.
As with any pest, proper identification of Conyza spp. is critical for the early development of effective strategies; however, these weeds have often been identified only to the genus level in South America (Mendes et al. Reference Mendes, Takano, Netto, Picoli, Cavenaghi, Silva, Nicolai, Christoffoleti, Oliveira and De Melo2021). In fact, Conyza spp. at the seedling and rosette growth stages are hardly distinguished due to the paucity of morphological traits useful for identifying them under field conditions. Although several dichotomous keys are available for the identification of specimens of Conyza spp., they require flowering plants and standard morphotypes for comparison (Wang et al. Reference Wang, Wu, Zhu and Lin2018). However, the taxonomy is not resolute due to overlapping traits among species, the existence of varieties within species, and interspecific hybrids (Thébaud and Abbott Reference Thébaud and Abbott1995). Due to obstacles to morphological classification and the relevance of these taxa as weeds of several crops, Conyza spp. are prime candidates for the use of molecular tools to support their taxonomic resolution.
By the 1980s, many molecular tools based on frequency data from markers were developed to support the identification and characterization of plant materials (Grover and Sharma Reference Grover and Sharma2016). These tools include random amplified polymorphic DNA, amplified fragment-length polymorphisms, microsatellites, and single-nucleotide polymorphisms (SNPs). However, although these markers have made valuable contributions to resolving phylogenetic issues, they could be problematic and even misleading for taxonomy (Arif et al. Reference Arif, Bakir, Khan, Al Farhan, Al Homaidan, Bahkali, Sadoon and Shobrak2010). In this context, novel molecular techniques based on gene sequencing such as DNA barcoding and even whole-genome sequencing have emerged (Hebert et al. Reference Hebert, Cywinska, Ball and DeWaard2003). The introduction of gene sequencing approaches featuring lower-error databases was instrumental for easy assignment of unknown plant samples to appropriate species.
DNA barcoding is a method that uses universally amplified, short, and polymorphic DNA markers (DNA barcodes) for genetically identifying organisms in terms of taxonomy at the species level (Hebert et al. Reference Hebert, Cywinska, Ball and DeWaard2003). The DNA barcodes most reported for plants include the internal transcribed spacer (ITS) region of the nuclear ribosomal cistron; the plastid intergenic spacers trnH-psbA, atpF-atpH, and psbK-psbI; the plastid intron within trnL; and the plastid genes matK, rbcL, rpoC1, and rpoB (Li et al. Reference Li, Yang, Henry, Rossetto, Wang and Chen2015; Yang et al. Reference Yang, Lv and Zhang2020). For example, five of eight Conyza spp. in Australia were genetically identified by the combination of ITS and rbcL gene regions as an adequate two-locus DNA barcode (Alpen et al. Reference Alpen, Gopurenko, Wu, Lepschi and Weston2014). In another study, the chloroplast genome sequencing of C. bonariensis revealed the plastid intergenic region rps16-trnQ as a barcode region that could be used to separate three Conyza spp. from Australia (Wang et al. Reference Wang, Wu, Zhu and Lin2018). The use of the entire chloroplast genome as a unique “super DNA barcode” for species of the Asteraceae family has also been reported (Chen et al. Reference Chen, Zhou, Cui, Wang, Duan and Yao2018; Gichira et al. Reference Gichira, Avoga, Li, Hu, Wang and Chen2019; Wang et al. Reference Wang, Wu, Zhu and Lin2018).
Although DNA barcodes resolved taxonomic issues for several closely related species, species delimitation using only molecular data is still a major challenge in some genera (Guo et al. Reference Guo, Huang, Liu and Wang2016; Starr et al. Reference Starr, Naczi and Chouinard2009). In addition, plastid gene regions are mostly inherited uniparentally and cannot reliably distinguish interspecific hybrids from the parent species (Daniell et al. Reference Daniell, Lin, Yu and Chang2016; Park et al. Reference Park, Lee, Lee, Lee, Joh, Park, Kim, Song and Yang2021). Thus, morphological markers have been studied in association with molecular markers as a suitable method of species classification (Han et al. Reference Han, Sebastin, Wang, Lee, Cho, Hyun and Chung2021; Ochoa et al. Reference Ochoa, Catan, Targa, Fraño, Caro and Chaila2023; Yang et al. Reference Yang, Zhang, Yang, Wang, Orr, Wang and Zhang2022). Combining a variety of morphological markers across the weed life cycle with both plastid and nuclear gene regions may generate an accurate platform to distinguish Conyza spp. This strategy would allow us to identify Conyza spp. at the field level and at any time during the season, mainly early in the season, for the development of effective strategies against these weeds.
In recent years, soybean [Glycine max (L.) Merr.] has been cultivated in approximately 45 million ha in Brazil (USDA 2022) across five macroregions (MRSs) that differ in climate and management (Kaster and Farias Reference Kaster and Farias2012). Field surveys have revealed infestations of Conyza spp. in almost half of the Brazilian soybean area, with higher frequencies in the southern region of the country (Da Silva et al. Reference Da Silva, Karam, Vargas, Adegas, Gazziero, Ikeda, Cavalieri, Costa and Perina2021; Lucio et al. Reference Lucio, Kalsing, Adegas, Rossi, Correia, Gazziero and Da Silva2019). Thirteen species of the genus Conyza are found in croplands, urban areas, and natural lands in Brazil, among which four are reported as weeds of field crops (Flora e Funga do Brasil 2023). These weedy species include fleabane [Conyza blakei (Cabrera) Cabrera], C. bonariensis, C. canadensis, and C. sumatrensis (Piasecki et al. Reference Piasecki, Mazon, Agostinetto and Vargas2019; Santos et al. Reference Santos, Oliveira, Constantin, Francischini, Machado, Mangolin and Nakajima2014; Vidal et al. Reference Vidal, Kalsing, Goulart, Lamego and Christoffoleti2007). Most of the information about Conyza spp. identification in Brazil is based only on few morphological or low discriminant molecular markers. There is no information available on the distribution of each Conyza sp. across Brazil, and the frequency of species overlaps in soybean fields across MRSs is unclear.
In this study, we identified Conyza spp. by combining two approaches and surveyed their spatiotemporal dispersion in soybean fields. We hypothesized that (1) Conyza spp. can be suitably identified by associating morphological and molecular markers and (2) dispersion and species overlaps of the Conyza spp. depend on geography. Thus, the aim of the present study was to identify Conyza spp. through DNA barcodes and morphological markers and survey their geographic dispersion across soybean-cropping MRSs and seasons in Brazil.
Materials and Methods
Plant Material
A total of 394 soybean fields were surveyed throughout the five MRSs of Brazil in 2019, 2020, and 2021, as follows: 130 in MRS 1, 111 in MRS 2, 98 in MRS 3, 54 in MRS 4, and 1 in MRS 5 (Figure 1). The fields were selected based on the representativeness of the cropping area and management system as well as the occurrence of mature plants that have escaped control by herbicide programs. Each field was farmed by a different grower and was treated as an accession of Conyza spp., and sampling was not repeated on the same field to explore the diversity of accessions (at least 50 km apart) within MRSs. Two distinct sampling strategies were employed to address the objectives of the current study. In 20 out of 394 fields, 20 plants per accession were harvested separately (single-plant sample) to obtain plant material for the development of morphological and molecular markers. In 374 out of 394 fields, 10 plants per accession were bulked and combined into a sample (10-plant sample) to obtain plant material for the survey on the geographic dispersion of Conyza spp. After sampling, samples were identified and stored as described in Burgos et al. (Reference Burgos, Tranel, Streibig, Davis, Norsworthy and Ritz2013).
Sampling sites of Conyza spp. in 394 soybean fields throughout five cropping macroregions (MRS 1–MRS 5) and three growing seasons (2019–2021) in Brazil. 1, 2, 3, 4, and 5 refer to MRS 1, MRS 2, MRS 3, MRS 4 and MRS 5, respectively. Accessions of Conyza spp. totaled 177, 124, and 93 in 2019, 2020, and 2021, respectively.
Growing Conditions and Species Identification
Whole-plant assays were carried out from May to August each year (late fall to winter in the Southern Hemisphere), when Conyza spp. mainly germinate and are established in crop fields (Vidal et al. Reference Vidal, Kalsing, Goulart, Lamego and Christoffoleti2007). Seeds from single-plant samples were germinated in petri dishes in growth chambers at 22 ± 1 C and with 10-h light, and then one seedling was transplanted to a 1-dm−3 pot filled with soil potting mix. Seeds from 10-plant samples were germinated in 1 by 1 cm cell trays in a greenhouse, and then four seedlings were transplanted into pots 1.2 dm−3 in size filled with the same substrate. The potting mix was composed of 50% soil, 25% rice bark, 25% peat, and traces of N-P-K, with a 162 kg m−3 density, 52% water retention, 5.5 pH, and 0.6 ± 0.3 mS cm−1 electrical conductivity. Plants were grown in open greenhouses with natural temperature (23.3 ± 9.3 C), humidity (65.1 ± 13.2%), and photoperiod (11.2 ± 0.4 h) conditions and received 2 mm of sprinkler irrigation four times a day. The accession species were classified using the morphological keys of Pruski and Sancho (Reference Pruski and Sancho2006) and Sancho (Reference Sancho2014), in which the most useful traits were involucre shape, capitulescence type, and inflorescence type. Twelve voucher specimens comprising the species and varieties found in this study were deposited in the Irina Delanova Gemtchújnicov Herbarium/BOTU of State University of São Paulo (BOTU 34833-34844).
Phylogenetic Analysis Based on Morphological Traits
In this assay, 314 plants of Conyza spp. that were established from the 400 single-plant samples (20 accessions × 20 plants) were evaluated for 32 morphological traits throughout five weed development stages (Table 1). Qualitative traits were evaluated as described by Pruski and Sancho (Reference Pruski and Sancho2006) and Sancho (Reference Sancho2014), and quantitative traits were measured with a Vernier caliper (Insize, WorldTools, Joinville, BR). After that, qualitative trait data were converted to numerical scores according to Table 1, and data of all variables were normalized on a 0 to 1 scale to perform clustering analysis. Principal component analysis (PCA) was performed to highlight the morphological markers among the 32 distinct traits using the adegenet package in the R environment (Jombart Reference Jombart2008). The data were used to build dendrograms to cluster Conyza spp. based on Euclidean distances using the software PAUP (set as default; Wilgenbusch and Swofford Reference Wilgenbusch and Swofford2003).
List of morphological traits and their scores.
a Quantitative traits.
b Qualitative traits.
DNA Isolation, PCR Amplification, and Sequencing
Leaf tissue (100 mg) from the fourth leaf was freeze-dried and ground into a powder to extract genomic DNA using a DNeasy® Plant Mini Kit (Qiagen, Hilden, DE, USA) following the manufacturer’s protocol. DNA quantity and quality were assessed by a spectrophotometer (NanoDrop®, Thermo Scientific, Waltham, MA, USA), and samples were considered suitable when the absorbance ratio at 260/280 nm was >1.8. DNA samples were resuspended in 100 µl of deionized water, and then dilutions were made up to 10 ng µl−1, after which DNA samples were stored at −20 C until gene amplification.
Five distinct DNA barcodes were used in the current study based on nuclear ribosomal or chloroplast gene regions: ITS, matK, rbcL, rps16-trnQ, and trnF-trnL (Table 2). Primer sequences for rbcL and trnF-trnL were obtained from previous research on universal DNA barcodes for plant species, while primer sequences for ITS, matK, and rps16-trnQ were designed. In such cases, DNA coding sequences annotated in the NCBI database for Conyza spp. were aligned, and homologous regions were used for primer design in Primer3Plus (Untergasser et al. Reference Untergasser, Cutcutache, Koressaar, Ye, Faircloth, Remm and Rozen2012). Primers were designed based on three published sequences of C. bonariensis, C. canadensis, and C. sumatrensis from the NCBI database, according to Supplementary Table S1. The polymerase chain reactions (PCR) contained Taq High Fidelity Pol Master Mix 2x (Red, Cellco Biotec, São Carlos, BR) and the optimized conditions were 95 C for 2 min, 40 cycles of 94 C for 30 s, 58 C for 30 s, and 72 C for 1 min, followed by an extension stage at 72 C for 5 min. PCR products were visualized for quality and size using a UV transilluminator (iBright® CL1500 Imaging System, Thermo Scientific) after electrophoresis through a 1.5% agarose gel.
Primer sequences and characteristics of the barcoding gene regions. a
a Abbreviations: AEf, amplification efficiency (number of successful PCR amplifications/number of DNA samples extracted); AS, amplicon size; SEf, sequencing efficiency (number of successful sequences analyzed/number of successful PCR amplifications); TM, melting temperature.
b Internal transcribed spacer region of the nuclear ribosomal cistron (18S-5.8S-26S).
c Plastid gene regions: matK, maturase K; rbcl, ribulose 1,5-biphosphate carboxylase.
d Intergenic spacer regions (between rps16 and trnQ and between trnF and trnL gene regions).
e Primer sequences redesigned in the present study based on external DNA sequences of Conyza spp.
f Primer sequences obtained from Kress and Erickson (Reference Kress and Erickson2007), Fazekas et al. (Reference Fazekas, Burgess, Kesanakurti, Graham, Newmaster, Husband, Percy, Hajibabaei and Barrett2008), and Taberlet et al. (Reference Taberlet, Gielly, Pautou and Bouvet1991).
High-quality amplicons were purified using the QIAquick® PCR Purification Kit (Qiagen) and then sequenced in both directions by BPI (Biotecnologia, Pesquisa e Inovação, Botucatu, BR). The sequences were proofread using Chromas Lite v. 2.1.1 (Technelysium, Brisbane, Australia), and consensus sequences of both directions were built in BioEdit v. 7.0.5 (Hall Reference Hall1999). A total of 208 DNA annotations representing combinations of the DNA barcodes, species, and varieties were deposited in the NCBI under the accession numbers provided in Supplementary Table S2.
Phylogenetic Analysis Based on Barcoding Gene Regions
The barcoding gene regions were examined by sequencing gene targets in 89 out of 314 plants from the taxonomic assay. These 89 plants were randomly selected within each species and the variety was identified in the taxonomic assay. First, an exploratory assay was carried out employing three plants of each Conyza sp. to assess the discrimination ability of each barcoding gene region; this was repeated twice. Consensus sequences were analyzed, filtered (removal of low-scoring sequences or parts of sequences including sequence ends, low-scoring base pairs, and unreliable SNPs), and aligned with Guidance2 (Sela et al. Reference Sela, Ashkenazy, Katoh and Pupko2015), and nucleotide diversity, Tajima’s D, and genetic distances were calculated in DnaSP v. 6.3.3 (Rozas et al. Reference Rozas, Ferrer-Mata, Sánchez-DelBarrio, Guirao-Rico, Librado, Ramos-Onsins and Sánchez-Gracia2017). Then, the gene regions with this discrimination ability were amplified as a proof of concept in the 89 selected plants, and external sequences from the NCBI were added as outgroups for comparison (Supplementary Table S1). The sequences were filtered and aligned as described in the exploratory assay and employed to construct dendrograms based on neighbor-joining (NJ) distance in PAUP (set as default; Wilgenbusch and Swofford Reference Wilgenbusch and Swofford2003). Node support within the NJ trees was assessed by a 10,000-replicate bootstrap test. Given that gene regions with discrimination ability were sequenced for all plants, a multilocus combination was made, and an additional NJ tree was constructed for the obtained concatenated sequence. In the case of multilocus analysis, outgroup sequences were not added, because they contained only a single locus.
Phylogenetic Analysis Combining Morphological and Molecular Approaches
The morphological and molecular data of the 89 plants described in the previous section were combined in a two-approach analysis to associate the discrimination ability of each method. Polymorphic sites of the barcoding gene sequences were converted to binary data (0 or 1), and then the molecular data set was positioned in tandem with the matrix of morphological data. The combined data set was used to construct dendrograms to cluster Conyza spp. based on Euclidean distances as previously described for the morphological assay of this study.
Dispersion of Conyza Weeds across Soybean Fields in Brazil
Accessions of Conyza spp. (n = 374, 10-plant samples) from soybean fields across MRSs and seasons of Brazil (Supplementary Table S3) were morphologically classified as described in the morphological assay, using 16 plants per accession. Spatial maps were generated to illustrate the spatiotemporal dispersion of Conyza spp. and their varieties across the respective geographic origins using the ggplot2 package in R (Ginestet Reference Ginestet2011). A color-coded classification scheme was used to identify each species and variety, and maps were plotted by species to visualize the regions in which species overlaps occurred.
Results and Discussion
Species Identification through Morphology
Out of 314 plants of Conyza spp., 42 plants were identified by taxonomy as C. bonariensis and 272 as C. sumatrensis, following the dichotomous keys of Pruski and Sancho (Reference Pruski and Sancho2006) and Sancho (Reference Sancho2014). Conyza bonariensis mainly varied from C. sumatrensis by its reddish phyllaries (vs. green), disk-like involucre (vs. bell-shaped), and corymbiform capitulescence (vs. thyrsoid-paniculate) (Figure 2). Two varieties were observed within C. bonariensis (var. angustifolia, n = 8; var. bonariensis, n = 34) as well as within C. sumatrensis (var. leiotheca, n = 28; var. sumatrensis, n = 244). Conyza bonariensis (L.) Cronquist var. angustifolia (Cabrera) Cabrera differed from Conyza bonariensis (L.) Cronquist var. bonariensis in its paniculiform, pyramidal capitulescence (vs. corymbiform, frequently flat-topped) (Figure 3). Conyza sumatrensis var. leiotheca differed from C. sumatrensis var. sumatrensis by its leaves, stems, and involucres being glabrous or subglabrous (vs. moderately to largely pilose) (Figure 4).
Details of phyllary color (A), involucre shape (B), and capitulescence and inflorescence types (C) between Conyza bonariensis (ERIBO) and Conyza sumatrensis (ERISU) from soybean fields in Brazil.
Details of capitulescence and inflorescence types between Conyza bonariensis var. angustifolia (ERIBOvarAn) and Conyza bonariensis var. bonariensis (ERIBOvarBo) from soybean fields in Brazil.
Details of involucre and leaf indumentum (pilosity) between Conyza sumatrensis var. leiotheca (ERISUvarLe) and Conyza sumatrensis var. sumatrensis (ERISUvarSu) from soybean fields in Brazil.
We did not find C. blakei and C. canadensis among the 314 plants from 20 accessions (fields) across Brazil, although they were reported in field crops in the country (Piasecki et al. Reference Piasecki, Mazon, Agostinetto and Vargas2019; Vidal et al. Reference Vidal, Kalsing, Goulart, Lamego and Christoffoleti2007). Conyza blakei differs from C. bonariensis and C. sumatrensis by its pinnatisect lower leaves (vs. linear or obovate) and narrow capitulescence (vs. moderately elongated) (Sancho Reference Sancho2014). Conyza canadensis differs from C. bonariensis and C. sumatrensis by its subradiate capitulum (vs. disciform) and long-haired leaf margins (vs. subglabrous) (Pruski and Sancho Reference Pruski and Sancho2006). Thus, given the distinctive morphological traits, both C. blakei and C. canadensis would probably be easily distinguished from C. bonariensis and C. sumatrensis regardless of the species variety. Although C. sumatrensis var. leiotheca is often misidentified as C. canadensis (Pruski and Sancho Reference Pruski and Sancho2006), the capitulum type (disciform) allowed us to effectively identify the plants evaluated in our study.
Principal component 1 (PC1) 1 of the PCA was mainly correlated (≥0.70) with leaf angle in the rosette state and phyllary color, involucre shape, and capitulescence and inflorescence type when plants were flowering (Figure 5). PC2 was mostly correlated (≤ −0.70) with the number of leaves and plant height in the rosette stage and the number of leaves when plants were undergoing stem elongation. The first two PCs of the PCA accounted for 37.5% of the total variance in the morphological data and indicated that these eight traits could serve as markers to distinguish between Conyza spp. (Figure 5). Euclidean trees based on morphology separated C. bonariensis from C. sumatrensis, illustrating monophyletic clades for each species, except for C. bonariensis var. angustifolia (Figure 6A). In fact, the plants of this species variety clustered with the C. sumatrensis clade because of many morphological overlaps, such as in the number of leaves, leaf angle, and inflorescence type.
Principal component analysis (PCA) of 32 morphological traits assessed across five developmental stages of Conyza bonariensis and Conyza sumatrensis. The analysis included 314 plants (single-plant samples) from soybean fields in Brazil. The length and direction of each vector indicate the strength and type (positive or negative) of the correlation between morphological traits and one of the principal components (PCs); percentages correspond to the proportion of the total variability accounted for by each PC. S1, S2, S3, S4, S5, and S6 refer to the seedling, rosette, stem elongation, beginning of flowering, main stem capitulum and beginning of senescence development stages, respectively.
Euclidean dendrograms of Conyza bonariensis (var. angustifolia and var. bonariensis) and Conyza sumatrensis (var. leiotheca and var. sumatrensis) based on morphological traits (A), DNA barcoding gene regions (B), and their combination (C). The analysis included 89 plants (single-plant samples) from soybean fields in Brazil. Dendrogram based on morphology (A) including 32 morphological traits (see Table 1). Dendrogram based on DNA barcodes (B) including all polymorphic sites of gene regions. Dendrogram based on both approaches (C) including the four morphological markers (phyllary color, involucre shape, capitulescence type and inflorescence type) and all polymorphic sites of the ITS and rps16-trnQ regions.
Overall, the traits related to reproductive structures were the most stable variables in our study and were very consistent with those in other studies (De Ulzurrun et al. Reference De Ulzurrun, Acedo, Garavano, Gianelli and Ispizúa2018; Hao et al. Reference Hao, Qiang, Liu and Cao2009; Thébaud and Abbott Reference Thébaud and Abbott1995). However, quantitative traits such as plant height, number of leaves, and leaf angle were quite variable, even though they showed large contributions to the total variance in the PCA (Figure 5). Conyza spp. commonly show extensive phenotypic plasticity for vegetative traits that are evidenced by observations of variable forms in the same field (Thébaud and Abbott Reference Thébaud and Abbott1995). Leaf color, cotyledon width, and number of teeth were not stable traits in our study but were recognized as key traits to distinguish Conyza spp. in Argentina (De Ulzurrun et al. Reference De Ulzurrun, Acedo, Garavano, Gianelli and Ispizúa2018). Thus, we adopted a more conservative approach by selecting only phyllary color, involucre shape, and capitulescence and inflorescence type as reliable markers to be combined with molecular data.
Characteristics of Barcoding Gene Regions
In the exploratory assay, high amplification success rates were observed for the matK, rbcL, and trnL-trnF regions, while for the ITS and rps16-trnQ regions, the rates ranged from 56% to 67% (Table 2). All five DNA barcodes had a high sequencing success rate that ranged from 90% to 100% of PCR products, which shows that overall, the regions were more easily sequenced than amplified. Nucleotide diversity calculated by locus was 2.4% and 4.9% for ITS and rps16-trnQ, respectively, while no SNPs were found in the other three barcoding gene regions (Table 3). Intraspecific distance was 0.0450 and 0.0500 for ITS and rps16-trnQ, respectively, while interspecific distances were slightly higher for each gene region (0.0597 and 0.0647 for ITS and rps16-trnQ, respectively).
Diversity and variation of the barcoding gene regions.
a Not calculated due the absence of diversity.
In addition to simple DNA sequence variability, crucial characteristics for barcoding loci also include primer universality and easy amplification and sequencing (Fazekas et al. Reference Fazekas, Burgess, Kesanakurti, Graham, Newmaster, Husband, Percy, Hajibabaei and Barrett2008; Kress and Erickson Reference Kress and Erickson2007). In this study, ITS and rps16-trnQ were the more variable gene regions and the most difficult to amplify, even with adjustments in PCR protocols and the design of new primer sequences. Even so, these DNA barcodes were chosen to identify the 89 plants selected in the current study, as the other gene regions showed no ability to separate Conyza spp. collected from Brazil. In another study, the ITS region had the highest genetic variability among Conyza spp. (0.3 to 15%), while rbcL showed interspecific distances ranging from 0% to 0.8% (Alpen et al. Reference Alpen, Gopurenko, Wu, Lepschi and Weston2014). Therefore, the data from our and other studies indicated that some universal DNA plant barcodes, such as rbcL, are likely not capable of resolving the taxonomic obstacles presented by Conyza spp.
Species Identification through Barcoding Gene Regions
Phylogenetic analysis based on the ITS, rps16-trnQ, and ITS + rps16-trnQ regions was used to distinguish 89 plants comprising the species and varieties identified with morphological data (Figure 7). The origins of accessions of C. bonariensis (var. angustifolia, n = 8; var. bonariensis, n = 30) and C. sumatrensis (var. leiotheca, n = 28; var. sumatrensis, n = 23) are detailed in Supplementary Table S2. The ITS region exhibited the highest separation of C. bonariensis from C. sumatrensis, supporting the monophyly of all plants of C. bonariensis var. angustifolia (Figure 7A). The rps16-trnQ gene region did not support monophyletic clusters for each species; hence, this proposed DNA barcode was unable to discriminate between the species (Figure 7B). The concatenation of the two regions failed to improve the ability of ITS individually, and even smaller bootstrap support was observed for the clade of C. bonariensis (Figure 7C).
Neighbor-joining dendrograms of Conyza bonariensis (var. angustifolia and var. bonariensis) and Conyza sumatrensis (var. leiotheca and var. sumatrensis) based on the nucleotide sequences of ITS (A), rps16-trnQ (B), and both gene regions (C). The analysis included 89 plants (single-plant samples) from soybean fields in Brazil. Colors represent each species and its varieties. Bootstrap values are given above branches.
NJ analysis of the ITS gene region supported clear monophyly for five of the eight Conyza spp. in Australia, including plants of C. bonariensis and C. sumatrensis (Alpen et al. Reference Alpen, Gopurenko, Wu, Lepschi and Weston2014). A similar analysis based on the rps16-trnQ gene region separated 13 specimens of C. bonariensis, C. canadensis, and C. sumatrensis from Australia into monophyletic clades (Wang et al. Reference Wang, Wu, Zhu and Lin2018). In our study, these two gene regions were not capable of distinguishing the species, as not all plants formed a single cluster in the tree, as expected for a universal DNA plant barcode. Curiously, only the plants featuring the morphotype of C. bonariensis var. angustifolia were spread across the clades, even in the analysis based on the ITS or ITS + rps16-trnQ regions. We hypothesized at least three different reasons for these findings in our study: misidentification of species, incomplete lineage sorting, and interspecific hybridization.
Incorrect C. bonariensis var. angustifolia identification is unlikely, because plants differ taxonomically as they mature, showing well-defined plant morphotypes, identified according to Sancho (Reference Sancho2014). In addition, we submitted three of these plants as voucher specimens to the herbarium where the plant species identification was confirmed, and vouchers were annotated appropriately. Incomplete lineage sorting is more likely, because Conyza comprises paraphyletic species that have recently speciated from a common ancestor (Alpen et al. Reference Alpen, Gopurenko, Wu, Lepschi and Weston2014; Marochio et al. Reference Marochio, Bevilaqua, Takano, Mangolim, Oliveira and Machado2017). Thus, DNA barcodes may not delimit species if the mutation rate at the target regions is insufficient to allow novel diversity to emerge (Simeone et al. Reference Simeone, Piredda, Papini, Vessella and Schirone2013; Van Velzen et al. Reference Van Velzen, Weitschek, Felici and Bakker2012). Interspecific hybridization is equally likely, as there is strong evidence that Conyza hybrids can be generated from plants that can interact freely in crop fields (Zelaya et al. Reference Zelaya, Owen and Van Gessel2007).
After alignment of the rps16-trnQ gene sequences, an insertion or deletion of six nucleotides (TGAAAT) was evident in our plant material in the gene region around 500 base pairs (Supplementary Table S4). The presence of these nucleotides appeared randomly in both species sampled from soybean fields, with the insertion predominantly present in C. bonariensis, while the deletion was present in C. sumatrensis. Interestingly, this difference in the gene sequences was not found in the external sequences of C. sumatrensis downloaded from the NCBI as outgroups for comparison (Supplementary Table S1). This finding reinforces the possibility of hybridization between different Conyza spp. from Brazil and raises the question of whether hybrids of C. bonariensis and C. sumatrensis occur spontaneously in soybean fields. Although evidence of hybridization was shown for diploid species of Conyza (Thébaud and Abbott Reference Thébaud and Abbott1995; Zelaya et al. Reference Zelaya, Owen and Van Gessel2007), little is known about interspecific gene flow among hexaploid species. To our knowledge, there was a unique report of hybrids between C. bonariensis and C. sumatrensis, namely Conyza ×daveauana Sennen in France, with unknown fertility (McClintock and Marshall Reference McClintock and Marshall1988). The proper confirmation of interspecific hybridization between C. bonariensis and C. sumatrensis from Brazil warrants further investigation.
Species Identification through Molecular and Morphological Approaches
The ITS and ITS+rps16-trnQ regions were able to differentiate C. bonariensis var. bonariensis from C. sumatrensis while failing to separate C. bonariensis var. angustifolia from C. sumatrensis (Figure 7). This result persisted when only polymorphic sites of both barcoding gene sequences were selected, converted to binary data, and analyzed through Euclidean distances (Figure 6B; Supplementary Table S4). To efficiently differentiate this unclassified variety of the species C. bonariensis, we assessed 32 traits and chose 4 of them as morphological markers (Figure 5). The combined, two-approach analysis considering the polymorphic sites of each region and the four morphological markers did not differentiate C. bonariensis var. angustifolia (Figure 6C). In fact, this variety of C. bonariensis remained spread among clusters and thus unclassified, which did not confirm our hypothesis about the combination of morphological and molecular data. However, this joint analysis was capable of discriminating 81 of 89 plants (91%) and demonstrated the potential of the combination of DNA barcodes and morphology to identify closely related weedy species.
No studies have combined DNA barcodes and morphological traits to classify Conyza spp., despite the relevance of these widespread weedy species worldwide. In other closely related weeds, such as Citrullus spp., Echinochloa spp., and Gallium spp., this approach supported taxonomic resolution (Deroo et al. Reference Deroo, Eckstein, Benaragama, Beattie and Willenborg2019; Shaik et al. Reference Shaik, Lepschi, Gopurenko, Urwin, Burrows and Weston2016; Tabacchi et al. Reference Tabacchi, Mantegazza, Spada and Ferrero2006). Thus, this study revealed a potential tool to differentiate species of Conyza in Brazil and worldwide, because the molecular and morphological markers can be applied to any species of the genus. For example, we believe that the identification of other weeds such as C. canadensis may be easier, as this species appears to be less related to the genus than other weeds in this genus (Thébaud and Abbott Reference Thébaud and Abbott1995). This study could also be a valuable tool not only for identifying Conyza spp. in field crops but also for ecological and conservation purposes due to the size (around 150 species) of this genus (TICA 2023).
Dispersion of Conyza Weeds across Soybean Fields in Brazil
Conyza sumatrensis was the prevailing Conyza species found in the field survey across five MRSs and three seasons in Brazil, with a high frequency and dispersion regardless of the MRS and season (Figure 8). This weed was detected in 166 of 177 fields in 2019, 99 of 104 fields in 2020, and 88 of 93 fields in 2021, and C. sumatrensis var. sumatrensis was the most common variety of the species. Conyza bonariensis was the other Conyza spp. found throughout the three seasons in the country but was only sparsely dispersed in MRS 1 and in a few sites in MRS 2, MRS 3, and MRS 4. In this case, the weed occurred in 14 of 177 fields in 2019, 6 of 104 fields in 2020, and 6 of 93 fields in 2021, in which the most-noted variety was C. bonariensis var. bonariensis. The overlaps between these species were relatively rare and were detected only in 3 of 177 fields in 2019, 1 of 104 fields in 2020, and 1 of 93 fields in 2021 (data not shown on the map).
Occurrence of Conyza bonariensis (ERIBO) and Conyza sumatrensis (ERISU) across soybean-cropping macroregions (MRS 1–MRS 5) and seasons (2019–2021) in Brazil. The analysis included 374 fields where 16 plants were identified using 10-plant samples. Light blue and blue dots refer to Conyza bonariensis var. angustifolia and bonariensis, respectively. Light red and red dots refer to Conyza sumatrensis var. leiotheca and sumatrensis, respectively. 1, 2, 3, 4, and 5 refer to MRS 1, MRS 2, MRS 3, MRS 4, and MRS 5, respectively.
While the dispersion of C. bonariensis in Brazil was dependent on geography, the dispersion of C. sumatrensis was not, which partially confirms our hypothesis about the overlaps between these species. Accessions of C. sumatrensis were detected over the seasons under high frequency and dispersion across Brazil; therefore, there were no large differences among MRSs (Figure 8). If the environment has had slight or even no influence on the proportion of Conyza spp., we can assume that multiple resistance to herbicides is the main factor explaining these findings. In fact, accessions of C. sumatrensis featuring resistance to herbicides with different sites of action have often been documented in Brazil (Albrecht et al. Reference Albrecht, Pereira, De Souza, Zobiole, Albrecht and Adegas2020; De Pinho et al. Reference De Pinho, Leal, Dos Santos Souza, De Oliveira, De Oliveira, Langaro, Machado, Christoffoleti and Zobiole2019). While accessions of glyphosate-resistant C. bonariensis are being well controlled by alternative herbicides in soybean fields, multiple-resistant C. sumatrensis probably survives a variety of herbicide programs.
As previously shown in the morphological study, we also did not identify plants of C. blakei and C. canadensis in the 374 accessions (fields) across different cropping MRSs and seasons (Figure 8). C. canadensis was reported in at least 16 studies carried out in Brazil in the Web of Science database, including cases of characterization of accessions resistant to glyphosate. Despite the possibility of occurrence at very low frequencies and eventual species shifts over time, it is likely that misidentification occurred in these studies from Brazil. In fact, a variety of C. sumatrensis generally has glabrous or subglabrous involucres, is restricted to the Americas, and is often misidentified as C. canadensis (Pruski and Sancho Reference Pruski and Sancho2006). It is necessary to improve the scientific rigor applied to species identification in studies with Conyza spp. to avoid producing further inconsistent information, as this might affect management actions.
In conclusion, the identification of Conyza spp. is a fundamental step in developing effective strategies against these weeds, and species misidentification may result in poor management efficacy. Combining ITS or ITS+rps16-trnQ regions and the four morphological markers discriminated 81 of 89 plants (91%) of both Conyza spp. (except C. bonariensis var. angustifolia). Conyza sumatrensis was detected in 353 of 374 (94%) soybean fields across MRSs and seasons, while C. bonariensis was sparsely dispersed, mainly in the southern part of the country (MRS 1). These results support the discrimination between C. bonariensis, C. sumatrensis, and other closely related weed species in soybean fields in Brazil and other cropping systems worldwide. Our study provides relevant information to support species identification and management to minimize the dispersion of herbicide resistance in Conyza spp.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/wsc.2024.3
Acknowledgments
The authors are thankful for the support from field agronomists, field scientists, and technical assistants of Corteva Agriscience who collected and submitted seed samples for the identification of Conyza spp. We are also grateful for the skilled research assistance in the laboratory and technical assistance from Caroline Andreato, Felipe Nunes, Lucas De Marco, Igor Araujo, and Matheus Moro. Special thanks are due to Gisela Sancho, Gustavo Heiden, and Karen Alpen for sharing key materials as well as for providing helpful comments on the development of different parts of the study. The present study received partial funding from the Corteva Agriscience and the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior–Brasil (CAPES) Finance Code 001. The authors have no conflicts of interest to declare and note that the funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
