Foliar Glyphosate Treatment Alters Transcript and Hormone Profiles in Crown Buds of Leafy Spurge and Induces Dwarfed and Bushy Phenotypes throughout its Perennial Lifecycle

Abstract

Leafy spurge (Euphorbia esula L.) is an invasive weed of North America and its perennial nature attributed to underground adventitious buds (UABs) that undergo seasonal cycles of para‐, endo‐, and ecodormancy. Recommended rates of glyphosate (∼1 kg ha^–1) destroy aboveground shoots but plants still regenerate vegetatively; therefore, it is considered glyphosate‐tolerant. However, foliar application of glyphosate at higher rates (2.2–6.7 kg ha^–1) causes sublethal effects that induce UABs to produce stunted, bushy phenotypes. We investigated the effects of glyphosate treatment (±2.24 kg ha^–1) on vegetative growth, phytohormone, and transcript profiles in UABs under controlled environments during one simulated seasonal cycle. Because shoots derived from UABs of foliar glyphosate‐treated plants produced stunted, bushy phenotypes, we could not directly determine if these UABs transitioned through seasonally induced endo‐ and ecodormancy. However, transcript abundance for leafy spurge dormancy marker genes and principal component analyses suggested that UABs of foliar glyphosate‐treated plants transitioned through endo‐ and ecodormancy. Glyphosate treatment increased shikimate abundance in UABs 7 d after treatment; however, the abundance of shikimate gradually decreased as UABs transitioned through endo‐ and ecodormancy. The dissipation of shikimate over time suggests that glyphosate's target site was no longer affected, but these changes did not reverse the altered phenotypes observed from UABs of foliar glyphosate‐treated leafy spurge. Transcript profiles further indicated that foliar glyphosate treatment significantly affected phytohormone biosynthesis and signaling, particularly auxin transport; gibberellic acid, abscisic acid and jasmonic acid biosynthesis; ethylene responses; and detoxification and cell cycle processes in UABs. These results correlated well with the available phytohormone profiles and altered phenotypes.

Abbreviations
• ABA
• abscisic acid
• ABC
• ATP-BINDING CASSETTE
• EPSPS
• 5-enolpyruvylshikimate-3-phosphate synthase
• GA
• gibberellic acid
• JA
• jasmonic acid
• IAA
• indole acetic acid
• P450
• cytochrome P450 monooxygenase
• qRT-PCR
• quantitative real-time polymerase chain reaction
• RD
• ramp down in temperature and photoperiod
• RD + V
• ramp down in temperature and photoperiod plus an extended vernalizing cold treatment
• TPM
• transcripts per kilobase million
• UABs
• underground adventitious buds
• UGT
• UDP-glucosyl/glycosyl transferase
• ZOG
• zeatin O-glucoside

Core Ideas

• Continuous increases in glyphosate-resistant weeds mandates a better understanding of glyphosate's effects on plant physiology.
• Higher rates (2.24–6.74 kg ha^–1) of foliar-applied glyphosate can cause a sublethal effect in the underground portion of leafy spurge and induce altered vegetative growth patterns from underground buds in the following shoot generations.
• We investigated the effects of glyphosate treatment (±2.24 kg ha^–1) on vegetative growth and metabolite and transcript profiles in underground buds of leafy spurge under controlled environments during its perennial lifecycle.

Glyphosate [N-(phosphonomethyl)glycine] is by far the most common broad-spectrum herbicide used in agricultural systems. However, continuous increases in glyphosate-resistant weeds mandates the need for new weed management strategies (Norsworthy et al., 2012), as well as a better understanding of glyphosate's effects on plant physiology (Vencill et al., 2012). Control of invasive weeds in noncultivated ecosystems of North America, including leafy spurge, generally requires long-term management programs integrating multiple methods including the use of herbicides such as glyphosate (Lym, 2000). Leafy spurge is a herbaceous perennial weed that reproduces by both seeds and vegetative propagules, but its persistence is mainly attributed to vegetative growth from an abundance of UABs. The seasonal cycles of para-, endo-, and ecodormancy occurring in UABs of leafy spurge during the summer, fall, and winter, respectively (Anderson et al., 2005, 2010), allow escape from conventional control measures, often requiring follow-up application with herbicides for long-term management. Although recommended field rates of glyphosate (∼1 kg ha^–1) are sufficient to destroy aboveground shoots of leafy spurge, these rates cause little or no damage to UABs. As a result, leafy spurge regenerates vegetatively from UABs and has been considered glyphosate-tolerant (Gottrup et. al., 1976).

Following foliar application, glyphosate rapidly translocates to the apical meristems, root meristems, and underground reproductive organs of perennial plants (Shaner, 2009), and inhibits nuclear encoded and chloroplast localized (Della-Cioppa et al., 1986) 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) of the shikimate–chorismate biosynthetic pathway (Amrhein et al., 1980; Steinrucken and Amrhein, 1980). Inhibition of EPSPS leads to increased accumulation of shikimate (Boocock and Coggins, 1983; Gruys et al., 1992; Franz et al., 1997; Herrmann and Weaver, 1999), which affects critical steps in the production of chorismate as a precursor for aromatic amino acids, auxin, and many other secondary products essential for plant growth and development.

Studies have demonstrated that application of glyphosate at rates (2.2–6.7 kg ha^–1) greater than recommended field rates causes sublethal effects in the underground portion of leafy spurge and induced new vegetative shoot growth from UABs that exhibits variegated leaves and stunted and bushy phenotypes (Maxwell et al., 1987; Doğramaci et al., 2014a, 2015a). Under greenhouse conditions, these foliar glyphosate-induced effects on subsequent vegetative growth from UABs were persistent in at least four vegetative growth cycles (removal of aerial shoots and 6 wk of new shoot growth from UABs constitutes one growth cycle) and also affected numerous molecular processes, including biosynthesis or signaling of phytohormones and cell cycle processes (Doğramaci et al., 2015a). An expanded investigation into glyphosate-induced altered phenotypes under field conditions has also been completed (Doğramaci et al., 2016). When field plots were treated with high rates of glyphosate (i.e., 3.4 and 6.7 kg ha^–1), the number of shoots derived from UABs of foliar glyphosate-treated plants was significantly increased in subsequent years after the application, and new vegetative shoots also had stunted and bushy phenotypes, as previously observed for foliar glyphosate-treated plants under greenhouse conditions. Further, quantification of the abundance of a selected set of transcripts in UABs of treated versus nontreated field grown plants (6.7 vs. 0 kg ha^–1) via quantitative real-time polymerase chain reaction (qRT-PCR) indicated that foliar glyphosate treatment impacted molecular processes in UABs involved in phytohormone biosynthesis and signaling and cell cycle processes. These glyphosate-induced effects on vegetative growth and transcript abundance persisted in the field for at least two seasonal cycles after the initial treatment (Doğramaci et al., 2016). Overall, these results provided a snapshot of molecular events in response to foliar glyphosate treatment in leafy spurge under field conditions and confirmed that these altered vegetative growth patterns in UABs were consistent both under greenhouse and field conditions. However, in both of these studies (Doğramaci et al., 2014a, 2016), the molecular data were limited to a selected set of genes for the qRT-PCR analysis and, in the case of field study, it was also complicated by the uncontrolled environmental conditions.

Therefore, to gain a better understanding of the effects that foliar glyphosate treatment has on vegetative growth and physiology of leafy spurge UABs during seasonal cycles, this study was performed under controlled environmental conditions. The objectives of this study were to (i) evaluate the effects of foliar glyphosate treatment on vegetative growth from UABs of leafy spurge after simulated seasonal conditions that induce paradormant UABs into endo- and ecodormancy; (ii) identify changes in transcript abundance, shikimate, and phytohormone profiles in UABs of control and foliar glyphosate-treated leafy spurge plants after the simulated environmental treatments; and (iii) identify the molecular mechanisms associated with glyphosate-induced phenotypes.

Materials and Methods Plant Material, Glyphosate Treatments, and Vegetative Growth

Leafy spurge plants were propagated from a uniform biotype (‘1984-ND001’) in cone-tainers and maintained in a greenhouse as described by Anderson and Davis (2004). Foliar tissues of 4-mo-old greenhouse-grown leafy spurge plants were treated with 0 or 2.24 kg ha^–1 glyphosate plus 0.25% surfactant as described by Doğramaci et al. (2014a). Each experiment included four biological replicates and each replicate consisted of 46 plants; Fig. 1 outlines the flow of experiments. Plants were maintained under growth-conducive conditions for 7 d after treatment, and UABs were collected from a subset (a third of the total treated plants) as the baseline paradormant samples. The remaining plants were exposed to a ramp down in temperature and photoperiod (RD) for 12 wk to induce endodormancy in UABs; a subset of these plants were further exposed to an extended vernalizing cold treatment (RD + V) for 8 wk to induce ecodormancy and flowering competence under controlled conditions (Foley et al., 2009). At the end of each environmental treatment, UABs were collected from control and glyphosate-treated plants to obtain transcriptome, shikimate, and phytohormone profiles. Additionally, to confirm the dormancy status and vegetative growth patterns in UABs of control and glyphosate-treated plants, aboveground shoots of six plants per replicate were decapitated for each treatment, and vegetative growth rate was measured as described by Foley et al. (2009). After 6 wk of vegetative growth from UABs of foliar glyphosate-treated and control plants, the height of the tallest shoot from each cone-tainer was measured and analyzed using the generalized linear mixed model (PROC GLIMMIX) procedure of SAS version 9.2 (SAS Institute Inc., Cary, NC).

Quantification of Shikimate Levels

Quantification of shikimate in crown buds of the 24 samples [three environmental treatments (baseline paradormant plants, RD to induce endodormancy, and RD + V to induce ecodormancy) × two glyphosate treatment (0 or 2.24 kg ha^–1) × four replicates] were accomplished spectrophotometrically (Cromartie and Polge, 2000; Shaner et al., 2005). In brief, 0.5 g of frozen tissue was extracted in 2 mL 0.25 N HCl at room temperature and 250 μL of this extract was added into 1 mL 0.25% (w/v) periodic acid and 0.25% (w/v) meta-periodate. After incubation at 37°C for 15 min, 1 mL of 0.6 M NaOH/0.22 M Na₂SO₃ was added into the sample and read with a spectrophotometer (DU 7400, Beckman, Indianapolis, IN); 510 µM shikimate (Sigma, St. Louis, MO) was used as the standard. These assays included four biological and two technical replicates; technical replicates were averaged prior to calculating the mean of the biological replicates and the SE.

Hormone Profiling

Quantification of phytohormones [abscisic acid (ABA), auxins, cytokinins, and gibberellic acid (GA)] in the 24 samples was conducted at the National Research Council of Canada, Plant Biotechnology Institute (Saskatoon, SK, Canada) via an ultra-performance liquid chromatography–electrospray ionization–tandem mass spectrometry system. Deuterated forms of each of the hormones were used as internal standards (Abrams et al., 2003; Zaharia et al., 2005). Preparation of the samples and quantification of the phytohormones was performed as described by Chiwocha et al. (2003, 2005). Profiling data obtained from four biological replicates were used to calculate the mean and SE.

RNA Extraction, cDNA Library Preparation, and Transcript Analyses by RNA-Seq

Total RNA was extracted from crown buds of the 24 samples according to the pine tree RNA extraction protocol (Chang et al., 1993). The RNA was treated with amplification grade DNase1 (Invitrogen, Carlsbad, CA), quantified with the Qubit 2.0 Fluorometer (Invitrogen), and quality was confirmed by agarose gel electrophoresis. The RNA-seq libraries were prepared with the TruSeq Stranded mRNA LT Sample Preparation Kit (Cat. No. RS–122–2101, Illumina, San Diego, CA) starting from 1 µg of total RNA. The pool of barcoded RNA-seq libraries was quantified via qRT-PCR using the Library Quantification kit (Cat. No. KK4824, Kapa Biosystems, Wilmington, MA). The size range of the final cDNA libraries was determined on an Agilent bioanalyzer DNA7500 DNA chip (Agilent Technologies, Santa Clara, CA). The cDNA libraries were sequenced on two lanes for 151 cycles from each end of the cDNA fragments on a HiSeq2500 using a TruSeq SBS sequencing kit version 1 (Illumina). The sequence images were transformed with the Real Time Analysis version 1.17.21.2 (Illumina) software to bcl files, which were demultiplexed to fastq files with CASAVA version 1.8.2 (Invitrogen). The quality-scores line in fastq files processed with Casava version 1.8.2 use an ASCII offset of 33, known as Sanger scores.

Transcriptome Assembly, Annotation, and Expression Analyses

Initial read quality was assessed by using the ‘FastQC (multi file)’ program under the ‘Public Apps → NGS → QC and Processing’ in the iPlant Discovery Environment (Oliver et al., 2013). Next, the program ‘Sickle-Quality-Base-Trimming’ (Joshi and Fass, 2011) was used to trim reads for quality and length using the parameters ‘quality format of sanger’, ‘quality threshold of 20’, and ‘minimum read length of 70 bases’ in the iPlant Discovery Environment. The number of raw fragments and trimmed fragments are provided in Supplemental File S1. To ensure the most complete transcriptome was assembled for use as a reference database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE71317, accessed 14 Aug. 2017), trimmed reads from these 24 samples along with samples from several other RNA-seq studies on leafy spurge (Doğramaci et al., 2015a; Chao et al., 2016) were combined into two files (one for each paired end) by using the ‘Concatenate Multiple Files’ program and the reads were kmer normalized by using the program ‘Trinity Normalize By K-mer Coverage’ (Grabherr et al., 2011) in the iPlant Discovery Environment with the default parameters of no more than 30× coverage for a given kmer. The program ‘Trinity’ (Robertson et al., 2011) was then used to assemble the resulting paired end read files. This combined assembly was annotated by ‘BlastX’ (McGinnis and Madden, 2004) against the nonredundant database with a minimum E-value cutoff of 10^–5. BlastX analysis against the TAIR10 protein sequence database was also used to identify the most similar Arabidopsis thaliana L. genes with a similar E-value cutoff. This assembly was used to map fragments and quantify sequences in the RSEM program suite (Li and Dewey, 2011). The embedded program suite EBseq (Leng et al., 2013) was used to identify the probability that any given sequence was differentially expressed between any sample groupings. The annotated assembled transcriptome and expression data as transcripts per kilobase million (TPM) value with differentially expressed genes noted by the posterior probability of the false discovery rate is provided in Supplemental File S1. In this manuscript, only component-based gene expression analyses were considered; however, contig-level expression analyses were also performed (data not shown, but are available through the Gene Expression Omnibus; http://www.ncbi.nlm.nih.gov/geo, accessed 14 Aug. 2017). For all subsequent analyses, only components with >10 TPM in all replicates from at least one treatment were considered to be expressed. Although the abundance of selected transcripts presented in the tables represent fold changes (log₂), the mean TPM values (log₂) are available in Supplemental File S1, Page 1). The expression data are deposited at the Gene Expression Omnibus as GEO dataset query GSE71321 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE71321, accessed 14 July 2017) and GSE71406 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE71406, accessed 14 July 2017).

Quantitative Real-Time Polymerase Chain Reaction Analyses

To confirm transcript profiles obtained from RNA-seq for selected genes, qRT-PCR was used. Sequences from a leafy spurge expressed sequence tag database (Anderson et al., 2007) or obtained by RNA-seq technology were used for designing primer pairs with the Primer-Select program of Lasergene 8 (DNASTAR, Inc., Madison, WI). These primer pairs were used to quantify the abundance of selected transcripts (see Supplemental File S2). Total RNA samples were reverse transcribed into cDNA as described in Doğramaci et al. (2013). In brief, 5 μg of total RNA was treated with DNase1 amplification grade (Invitrogen), and reverse transcribed using Super Script First-Strand Synthesis Kit3 (Invitrogen) in a 20-μL volume according to the manufacturer's instructions. After cDNA synthesis, each reaction was diluted to 800 μL total volume and stored at −80°C. For qRT-PCR reactions, 1 μL total cDNA was added to a 10-μL PCR reaction mixture containing 5 μL of LightCycler 480 SYBR Green I Master and 0.5 μL of each primer-set. Transcript abundance was measured from three technical and four biological replicates using a LightCycler 480 II (Roche, Indianapolis, IN). All transcript values were normalized using the reference gene SAND, which has been identified as one of the best internal control genes for various tissues and treatments, including crown bud growth and development for leafy spurge (Chao et al., 2012). QbasePLUS version 2.4 software (Biogazelle, Ghent, Belgium) was used to normalize expression values. Pearson's and Spearman's tests were performed to obtain correlations between the RNA-seq and qRT-PCR data (Supplemental File S2). The gene abbreviations and descriptions of all putative homologous leafy spurge genes included throughout this report were obtained from an A, thaliana website (www.Arabidopsis.org, accessed 14 July 2017) and are presented in Supplemental File S2.

Gene Set Enrichment Analyses

The transcriptome dataset with > 10 TPM in all replicates for at least one treatment (i.e., 13,378 and 12,809 components for control and glyphosate-treated samples, respectively) was analyzed using Ariadne Pathway Studio 9.0 Software-Resnet Plant version 2.1 (Ariadne Genomics Inc., Rockville, MD) to obtain gene set enrichment analyses by using Arabidopsis Gene Identifier designations for A. thaliana genes (Mueller et al., 2003; Subramanian et al., 2005). The Mann–Whitney U-test algorithm with a 0.05 enrichment p-value cutoff was used to identify predefined sets of genes over-represented between treatments on the basis of their AraCyc metabolic pathways; their A. thaliana signaling pathways (http://pmn.plantcyc.org/ARA/class-tree?object=Pathways, accessed 14 Aug. 201); and gene ontology (biological processes, cellular component, molecular function; http://www.geneontology.org/, accessed 14 July 2017).

Results Vegetative Growth and Phenotype Changes in Response to Glyphosate

The RD treatment induced a transition from para- to endodormancy in UABs, and an additional extended cold treatment (RD + V) induced a transition from endo- to ecodormancy and flowering competence in leafy spurge control plants (Fig. 2), as previously established for leafy spurge (Foley et al., 2009; Doğramaci et al., 2010). However, new shoots derived from UABs of foliar glyphosate-treated plants displayed minimal growth (Fig. 2), and dwarfed and bushy phenotypes (Fig. 3) before and after RD and RD + V treatments; after the RD + V treatment, new shoots derived from foliar glyphosate-treated plants were not flowering competent (Fig. 3). On the basis of the vegetative growth results, we could not directly determine if these UABs of foliar glyphosate-treated plants also went through similar phases of dormancy, as was observed for the UABs of control plants. Thus a transcriptional approach was used to look for the hallmarks of dormancy transitions on the basis of the gene expression differences.

Transcriptome Profiling

Transcriptome analyses for UABs of leafy spurge treated with or without glyphosate under simulated seasonal conditions identified 223,732 assembled components, with between 40,972 (endodormant Replicate 1, control) and 48,645 (endodormant Replicate 1, glyphosate-treated) components represented among the 24 libraries (Supplemental File S1 Page 2). Overall, the transcriptome analyses identified different abundance among 19,127 transcripts (a false discovery rate of ≤0.05); of these, 14,613 had >10 TPM in UABs with or without glyphosate treatment (Supplemental File S1 Page 2).

Among the control samples (0 kg ha^–1 glyphosate), 18,189 components had >10 TPM for at least one of the dormancy phases (para-, endo-, ecodormancy) for all the replicates (Supplemental File S1 Page 2). Of these components, 9037 were differentially expressed (Table 1), and 6776 had similarity to genes in the A. thaliana database (www.Arabidopsis.org, accessed 14 July 2017). Therefore, the function of proteins encoded by leafy spurge transcripts discussed throughout this manuscript are based on the most homologous match to A. thaliana (provided in Supplemental File S1 to Supplemental File S3). The para- to endodormancy transition identified differential expression of 6070 transcripts with 4583 A. thaliana hits; endo- to ecodormancy transitions identified the differential expression of 6507 transcripts with 4840 A. thaliana hits (Table 1). However, among the glyphosate-treated samples (2.24 kg ha^–1 glyphosate), 17,746 components had >10 TPM for at least one of the dormancy phases for all the replicates (Supplemental File S1 Page 2); of these components, 10,014 were differentially expressed and 7529 had similarity to genes in the A. thaliana database (Table 1). After RD treatment, 7159 transcripts were differentially expressed (with 5415 A. thaliana hits) in endodormant UABs; after extended cold treatment (RD + V), 5449 transcripts were differentially expressed (with 4131 A. thaliana hits) in ecodormant UABs (Table 1). Further, a direct comparison of transcriptome profiles obtained from UABs of glyphosate-treated plants (2.24 kg ha^–1) to UABs of controls (0 kg ha^–1) identified 4993, 2311, and 1301 differentially expressed transcripts in the para-, endo-, and ecodormant phases, respectively (Table 1). Reliability of the RNA-seq data was validated via qRT-PCR analysis of 44 selected genes (Supplemental File S2), which revealed a similar trend in expression patterns, with the exception of a few genes. Additionally, Pearson's or Spearman's correlation coefficient tests confirmed a significant correlation between the RNA-seq and qRT-PCR data and validated use of the global transcript data (Supplemental File S2).

Table 1 Number of differentially expressed genes (DEGs) with a false discovery rate of ≤0.05, DEGs with similarity to Arabidopsis thaliana (AT hit), and the distribution of increased and decreased transcripts in underground buds of glyphosate-treated vs. control (2.24 vs. 0 kg ha^–1) plants at specific dormancy phases [paradormancy (Para), endodormancy (Endo), ecodormancy (Eco)], or interphase changes (‘Endo vs. Para’ or ‘Eco vs. Endo’) resulting from 0 or 2.24 kg ha^–1 glyphosate in underground adventitious buds of leafy spurge.

Treatment	Comparison	DEGs	DEGs with AT hits	DEGs decreased	DEGs increased
0 kg ha–1 glyphosate	Endo vs. Para	6,070	4583	2044	2539
	Eco vs. Endo	6,507	4840	2992	1848
	Para–Endo–Eco†	9,037	6776	na	na
2.24 kg ha–1 glyphosate	Endo vs. Para	7,159	5415	2791	2624
	Eco vs. Endo	5,449	4131	2297	1834
	Para–Endo–Eco†	10,014	7529	na‡	na
Para	2.24 vs. 0 kg ha–1	4,993	3950	1699	2251
Endo	2.24 vs. 0 kg ha–1	2,311	1834	1300	534
Eco	2.24 vs. 0 kg ha–1	1,301	1031	426	605

†

three-way comparison.

‡

na, not applicable.

Principal component analyses of the differentially expressed transcripts (19,127) revealed similarities and differences for the environmental treatment effects with or without glyphosate treatment (Fig. 4). The X-component of the principal component analyses explained 41.9% of the variance and clearly distinguished each phase of dormancy with or without glyphosate treatment. The Y- and Z-components of the principal component analyses explained 28.5% and 20.7% of the variance, respectively. The variance level in the Z-component suggests that foliar glyphosate treatment appears to have the most impact on UABs at the paradormant phase. Clustering of the differentially expressed transcripts suggest that UABs of foliar glyphosate-treated plants transitioned through similar phases of endo- and ecodormancy after the RD and RD + V treatments, respectively. The transcript abundance of several marker genes (e.g., HY5, MAF3-like, RD22, and RVE1-like) for endo- and ecodormancy (Doğramaci et al., 2014b, 2015b) also supports this hypothesis. As seen in Table 2, the abundance of these transcripts had similar changes during dormancy transitions with or without foliar glyphosate treatment. Thus, we refer to UABs as para-, endo-, and ecodormant in control or foliar glyphosate-treated plants in the remainder of the manuscript, and mainly focus on the transcriptional changes caused by foliar glyphosate treatment at each of these dormancy phases.

Table 2 Fold changes in transcript abundance according to RNA-seq and quantitative real time–polymerase chain reaction (qRT-PCR) for selected genes used as markers during transitions from paradormancy (Para) to endodormancy (Endo) and endo- to ecodormancy (Eco) in underground buds of leafy spurge for control (0 kg ha^–1) and glyphosate-treated (2.24 kg ha^–1) plants.

Gene ID†	TAIR ID	RNAseq				qRT-PCR
		Endo vs. Para		Eco vs. Endo		Endo vs. Para		Eco vs. Endo
		0 kg ha–1	2.24 kg ha–1	0 kg ha–1	2.24 kg ha–1	0 kg ha–1	2.24 kg ha–1	0 kg ha–1	2.24 kg ha–1
COP1	AT2G32950	−0.02	0.01	−0.08	−0.23	0.26	−0.03	1.57	1.05
HY5	AT5G11260	0.39	1.74	0.97	0.41	0.40	0.82	1.06	0.77
RD22	AT5G25610	0.95	1.31	−0.89	0.07	0.89	1.25	−0.88	−0.10
RVE1-like	AT5G17300	1.61	2.43	1.29	0.79	1.47	2.28	1.40	0.75
MAF3-like	AT5G65060	1.23	2.79	8.27	6.89	1.21	1.70	7.97	7.97

†

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) and gene IDs.

Shikimate Abundance

The shikimate pathway leads to production of chorismate, which is a precursor of aromatic amino acids (e.g., tyrosine, phenylalanine, tryptophan), auxin, and many other secondary products essential for plant growth and development. In this study, shikimate abundance was significantly increased in paradormant UABs 7 d after foliar glyphosate treatment (Fig. 5), indicating that the target site (EPSPS) of glyphosate was inhibited. However, shikimate abundance was significantly decreased in endodormant UABs of glyphosate-treated plants but it was still higher than the endodormant control samples. Shikimate abundance was further decreased in ecodormant UABs of foliar glyphosate-treated plants, at which stage, it was similar to levels in UABs of control plants. These results suggest that simulated seasonal conditions allowed for dissipation of shikimate in UABs, or that the target site is no longer inhibited, and shikimate was used for downstream processes. Further, the abundance of transcripts (e.g., EMB1144, EPSPS, SK1) involved in various stages of the shikimate pathway had changes in UABs of foliar glyphosate-treated plants at all phases of dormancy but the amplitude of these changes was minor (Table 3; Supplemental File S1 Page 1). The results for the abundance of shikimate and the transcripts involved in the shikimate pathway did not explain the altered phenotypes derived from UABs of foliar glyphosate-treated plants after transitioning to endo- and ecodormancy (Fig. 2, Fig. 3). Therefore, to gain additional insights into the involvement of growth regulators and pathways that might explain these altered phenotypes, we obtained phytohormone profiles to correlate with the associated transcript profiles.

Table 3 Fold changes (log₂) in the abundance of transcripts involved in shikimate, amino acid, and auxin biosynthesis, and auxin metabolism, transport, and signaling caused by glyphosate treatment (2.24 vs. 0 kg ha^–1) at specific dormancy phases [paradormancy (Para), endodormancy (Endo), and ecodormancy (Eco)] in underground adventitious buds of leafy spurge.

[Table omitted. See PDF]

†

Differentially expressed genes (false discovery rate ≤ 0.05) with >10 transcripts per kilobase million for all replicates for at least one of the treatments.

‡

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) to obtain primary gene IDs and abbreviations.

Phytohormone Profiling and the Abundance of Associated Transcripts

Significant changes were detected for phytohormone profiles in UABs before and after environmental treatments (RD and RD + V) with or without foliar glyphosate treatment. The list of phytohormone profiles quantified (auxin, ABA, cytokinins, and GA) is available in Supplemental File S2. The phytohormone profiling data are incomplete for some samples because some of the signals were below the limit of either quantification or detection. Even though statistical analyses could not be performed because of the incomplete data for some samples, the available data were included in the graphs as mean values (Fig. 6 to Fig. 9).

Auxin Abundance and Associated Transcripts

Auxins were represented by the biologically active indole acetic acid (IAA) and its conjugates with aspartic acid and glutamic acid. Bioactive IAA levels were significantly increased in endodormant UABs of foliar glyphosate-treated plants compared with the controls: although IAA levels were slightly decreased in para- and eco-dormant buds in response to the foliar glyphosate treatment, these changes were not significant (Fig. 6). Indole acetic acid conjugated with aspartic acid had mainly increased abundance in UABs of foliar glyphosate-treated plants compared with the controls at each dormancy phase; however, these changes were significant only for paradormant UABs. Changes in IAA conjugated with glutamic acid were not significant.

Although the abundance of one transcript involved in the tryptophan and auxin biosynthetic pathway (TRP3) was significantly increased in paradormant UABs 7 d after glyphosate treatment, the overall abundance of transcripts involved in these pathways (TRP1, TRP3) had only minor changes in endo- and ecodormant UABs as a result of foliar glyphosate treatment (Table 3). Transcripts involved in auxin catabolism mainly had increased abundance [three components of UDP-GLUCOSYL/-GLYCOSYL TRANSFERASES (UGT)74E2] in paradormant UABs 7 d after treatment (Table 3). However, only one of the components (comp116697_c3) had a significant increase in abundance in UABs after transitioning into endo- or ecodormancy (Table 3). Further, the abundance of several transcripts known to be involved in auxin transport, response, or signaling [ATP-BINDING CASSETTE (ABC)B1, ABCB4, LAX3, PID, PILS7, PIN3] were also altered in para-, endo-, and ecodormant UABs as a result of foliar glyphosate treatment. The transcript abundance of some auxin transporters (ABCB4, PILS7, PIN3) were significantly increased in paradormant UABs of foliar glyphosate-treated plants 7 d after treatment (Table 3). However, the abundance of several auxin transporters (ABCB1, LAX3, PID, PIN3) decreased significantly in endodormant UABs as a result of foliar glyphosate treatment; only ABCB1 and PIN3 had a significant decrease caused by foliar glyphosate treatment in ecodormant UABs (Table 3).

Cytokinins and Associated Transcripts

Low levels of bioactive free base cytokinins (t-Z, c-Z, dhZ and iP) were present in some samples but they were not reliably quantified for all samples (Fig. 7). Higher levels of the biosynthetic precursor zeatin riboside (ZR) (trans- and cis-isomers, dhZR) were also quantified. Levels of t-ZR were significantly increased in UABs of foliar glyphosate-treated plants compared with the controls at each dormancy phase, whereas changes in c-ZR levels were only significant in endodormant UABs of foliar glyphosate-treated plants compared with endodormant UABs of non-glyphosate-treated plants. dhZR levels were reliably quantified only in UABs of glyphosate-treated plants and were highest in endodormant UABs. Moreover, quantification of t- zeatin O-glucoside (ZOG) and c-ZOG (catabolism products of zeatin) suggests that high levels of zeatin must have been produced. t-ZOG levels were increased in UABs of foliar glyphosate-treated plants; however, these changes were significant only in ecodormant UABs (Fig. 7). c-ZOG levels were significantly increased in para- and endodormant UABs of foliar glyphosate-treated plants compared with non-glyphosate-treated plants; the level of c-ZOG in ecodormant UABs of foliar glyphosate-treated plants was not significant.

The abundance of several transcripts involved in cytokinin biosynthesis (Table 4; Supplemental File S1 Page 1) decreased significantly (ADK2, LOG5) or increased (IPT5) 7 d after foliar glyphosate treatment in paradormant UABs. Significant changes and similar trends in transcript abundance for IPT5 and LOG5 were also observed in endodormant UABs as a result of foliar glyphosate treatment, but only IPT5 was significantly impacted by glyphosate in ecodormant UABs. Although the abundance of some transcripts involved in cytokinin catabolism (CKX1, CKX3, UGT73C1) were significantly increased in paradormant UABs of foliar glyphosate-treated plants, only UGT73C1 was significantly increased in response to glyphosate treatment in endo- and ecodormant UABs (Table 4). Additionally, transcripts involved in cell cycle processes (e.g., CDC2A, CDKB1;2, CDKB2;2, CYC1B, CYCD1;1, CYCD3;1, CYCD3;2) decreased significantly in abundance in paradormant UABs 7 d after glyphosate treatment (Table 4). Several of these transcripts (CDKB1;2, CDKB2;2, CYCD1;1, CYCD3;1, CYCD3;2) also decreased significantly in abundance in endodormant UABs as a result of foliar glyphosate treatment but, interestingly, no significant changes were observed for these transcripts in ecodormant UABs in response to foliar glyphosate treatment.

Table 4 Fold changes (log₂) in the abundance of transcripts involved in cytokinin biosynthesis, metabolism, signaling, and cell cycle processes caused by glyphosate treatment (2.24 vs. 0 kg ha^–1) at specific dormancy phases [paradormancy (Para), endodormancy (Endo), and ecodormancy (Eco)] in underground adventitious buds of leafy spurge.

[Table omitted. See PDF]

†

Differentially expressed genes (false discovery rate ≤ 0.05) with >10 transcripts per kilobase million for all replicates for at least one of the treatments.

‡

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) to obtain primary gene IDs and abbreviations.

Abscisic Acid Abundance and Associated Transcripts

Abscisic acid profiles indicated that foliar glyphosate treatment mainly led to decreased abundance of ABA and its metabolites compared with the controls at all dormancy phases in UABs (Fig. 8). For example, bioactive ABA levels decreased significantly in para- and ecodormant UABs of foliar glyphosate-treated plants compared with the controls; bioactive ABA levels also decreased in endodormant UABs compared with the controls, but these changes were not significant. The main ABA metabolism pathway appears to work through 8’-hydroxylation to produce phaseic acid, which is reduced to dihydrophaseic acid. In UABs of glyphosate-treated plants, phaseic acid levels decreased significantly in UABs of foliar glyphosate-treated plants compared with the controls at each phase of dormancy. However, dihydrophaseic acid levels decreased compared with the controls at each phase of dormancy, but these changes were only significant for para- and endodormant UABs. The changes in the levels of ABA glucose ester and 7'OH-ABA were not significant; the quantification of neophaseic acid and t-ABA did not produce enough data to obtain statistical analyses.

Several transcripts involved in ABA biosynthesis (AAO4, ABA2) were positively affected by foliar glyphosate treatment in all phases of dormancy (Table 5; Supplemental File S1 Page 1). Although foliar glyphosate treatment had a significant and negative effect on the abundance of NCED3 in paradormant UABs 7 d after treatment, no significant changes were observed in endo- and ecodormant UABs of foliar glyphosate-treated plants (Table 5). Additionally, the transcripts involved in ABA catabolism were either negatively (CYP707A) or positively (CYP707A3) affected in paradormant UABs 7 d after glyphosate treatment. Additionally, in endo- and ecodormant UABs of foliar glyphosate-treated plants, CYP707A3 also increased significantly. The transcripts involved in response to ABA signaling also had either decreased (ABI1, COR6.6) or increased (RCAR10) abundance in paradormant UABs of foliar glyphosate-treated plants; only RCAR10 had a significant change in abundance in endo- and ecodormant UABs of glyphosate-treated plants (Table 5). Because foliar glyphosate treatment appears to reduce the abundance of bioactive ABA or its metabolites in UABs, the decrease in NCED3 (an upstream committed step in ABA biosynthesis) observed in para- and endodormant UABs of foliar glyphosate-treated plants could be a limiting factor.

Table 5 Fold changes (log₂) in the abundance of transcripts involved in abscisic acid (ABA) biosynthesis, catabolism, and signaling caused by glyphosate treatment (2.24 vs. 0 kg ha^–1) at specific dormancy phases [paradormancy (Para), endodormancy (Endo), and ecodormancy (Eco)] in underground adventitious buds of leafy spurge.

[Table omitted. See PDF]

†

Differentially expressed genes (false discovery rate ≤ 0.05) with >10 transcripts per kilobase million for all replicates for at least one of the treatments.

‡

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) to obtain primary gene IDs and abbreviations.

Gibberellic Acid Abundance and Associated Transcripts

The GA profiles in UABs were detectable for only a few samples from both control and foliar glyphosate-treated plants, thus producing limited results (Fig. 9). Regardless, the available data indicate that GAs detected (i.e., GA8, GA19, GA24, and GA53) belong to either the early 13-hydroxylation pathway or to the nonhydroxylation pathway (Doğramaci et al., 2015a). The increased 13-hydroxylation pathway products G19 and GA53 (precursors of biologically active GA1) in UABs of foliar glyphosate-treated plants suggests that at least some modules of the bioactive GA biosynthesis pathway are stimulated after foliar glyphosate treatment; however, on the basis of these incomplete results, it is difficult to glean accurate conclusions about the abundance of GA profiles.

Although the transcripts involved in initial steps of GA biosynthesis (CPS1/GA1, KS/GA2, KO/GA3) increased significantly in paradormant UABs 7 d after glyphosate treatment, the downstream transcripts involved in this pathway (KAO1, GA3OX1/GA4) had only minor or insignificant changes (Table 6; Supplemental File S1/Page 1). The foliar glyphosate treatment led to an increase in the abundance of CPS1/GA1 and KS/GA2 in endo- and ecodormant UABs; however, the changes were only significant in endodormant UABs. Some transcripts involved in downstream GA biosynthesis (KAO1) or catabolism (GA2OX8) had decreased abundance that was significant in endo- (KAO1, GA2OX8) or ecodormant (GA2OX8) UABs of foliar glyphosate-treated plants. These results appear to indicate that some components of the GA biosynthesis pathway were significantly affected by foliar glyphosate treatment and some downstream transcripts involved in GA catabolism were negatively impacted in endo- and ecodormant UABs of foliar glyphosate-treated plants. Further, significant increases (GID1B, SCL14) were observed for several transcripts involved in GA signaling in all phases of dormancy in UABs of foliar glyphosate-treated plants, whereas several others had a significant decrease in para- (SHR) or endo-dormancy (SCR). Thus it is difficult to correlate transcript data with the limited GA profiles obtained from this study.

Table 6 Fold changes (log₂) in the abundance of transcripts involved in gibberellic acid (GA) biosynthesis, catabolism, and signaling caused by glyphosate treatment (2.24 vs. 0 kg ha^–1) at specific dormancy phases [paradormancy (Para), endodormancy (Endo), and ecodormancy (Eco)] in underground adventitious buds of leafy spurge.

[Table omitted. See PDF]

†

Differentially expressed genes (false discovery rate ≤ 0.05) with >10 transcripts per kilobase million for all replicates for at least one of the treatments.

‡

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) to obtain primary gene IDs and abbreviations.

Transcripts Involved in Other Growth Related Processes

Although the abundance of ethylene was not quantified in this study, foliar glyphosate treatment positively affected the transcripts involved in ethylene biosynthesis but these changes were not significant at any of the dormancy phases. Significant increases were observed for the transcripts involved in ethylene signaling (EIL1, EIN3, EIN4, ERS1, ETR2) in paradormant UABs 7 d after treatment. Although the abundance of these transcripts remained increased in endo- and ecodormant UABs of foliar glyphosate-treated plants, only a few of these changes remained significant, probably because of increases in ethylene signaling associated with the transition from para- to endodormancy (Table 7; Supplemental File S1 Page 1). Additionally, some members of the AP2/ERF transcription factors known to be responsive to ethylene had significantly increased (DREB26, ERF1, ERF59, ERF105, RAP2.12) or decreased (TINY) abundance, but only in paradormant UABs of foliar glyphosate-treated plants 7 d after treatment (Table 7).

Table 7 Fold changes (log₂) in the abundance of transcripts involved in ethylene biosynthesis, response, and signaling caused by glyphosate treatment (2.24 vs. 0 kg ha^–1) at specific dormancy phases [paradormancy (Para), endodormancy (Endo), and ecodormancy (Eco)] in underground adventitious buds of leafy spurge.

[Table omitted. See PDF]

†

Differentially expressed genes [false discovery rate (FDR) ≤ 0.05] with >10 transcripts per kilobase million (TPM) for all replicates for at least one of the treatments.

‡

Differentially expressed genes (FDR ≤ 0.05) with <10 TPM for all replicates for at least one of the treatments.

§

The Arabidopsis Information Resource (TAIR) was used to annotate homologs of leafy spurge transcripts (TAIR ID) to obtain primary gene IDs and abbreviations.

Significant changes in the abundance of transcripts involved in biosynthesis or signaling associated with brassinosteroid biosynthesis and catabolism, jasmonic acid (JA) biosynthesis and signaling, and salicylic acid signaling were also observed (Table 7). Significant changes in brassinosteroid biosynthesis (BR6ox1) or catabolism (BEN1) increased in paradormant UABs of foliar glyphosate-treated plants, whereas the abundance of transcripts involved in several steps of JA biosynthesis and signaling were either negatively (AOS, LOX2, WRKY70) or positively affected (LOX2, LOX4) by foliar glyphosate treatment in paradormant UABs 7 d after treatment. Several of these transcripts also had significant increases in abundance (LOX2, LOX4) in endo- and ecodormant UABs. Because chloroplast-localized lipoxygenases are required for wound-induced JA accumulation in A. thaliana (Bell et al., 1995), these results suggest that some of the glyphosate-induced stress responses persist in endo- and ecodormant UABs of foliar glyphosate-treated plants.

Four large gene families associated with detoxification processes [ABC TRANSPORTERS, CYTOCHROME P450 MONOOXYGENASES (P450s), GLUTATHIONE S-TRANSFERASES and UGTs] had differential expression between UABs of control and foliar glyphosate-treated plants (Supplemental File S1). Overall, UABs of foliar glyphosate-treated plants had 169 (113 increased), 82 (63 increased), and 52 (47 increased) members of these gene families that were differentially expressed during the para-, endo-, and ecodormant phases, respectively (Supplemental File S1 Pages 3–5).

Gene Set Enrichment Analyses

Identification of overrepresented pathways and gene sets via gene set enrichment analyses correlate well with the transcript and phytohormone profiles mentioned above (Supplemental File S3). Among these gene sets, several pathways were uniquely enriched with increased or decreased transcripts in UABs of foliar glyphosate-treated plants. For example, endodormant UABs of foliar glyphosate-treated plants had uniquely enriched pathways with increased transcript abundance for AraCyc pathways, such as brassinosteroid biosynthesis II; A. thaliana signaling pathways such as guard cell ABA signaling and stress ABA signaling; and numerous biological processes including auxin polar transport, cellular amino acid biosynthetic process, cellular response to ethylene stimulus, regulation of the G2–M transition of mitotic cell cycle, response to ABA stimulus, and response to GA stimulus (Table 8; Supplemental File S3). Likewise, after the para- to endodormancy transition, the uniquely enriched pathways caused by foliar glyphosate treatment with decreased transcript abundance included numerous AraCyc pathways such as ABA glucose ester biosynthesis, ethylene biosynthesis from methionine, the methionine salvage pathway, IAA degradation IV, monoterpene biosynthesis, the superpathway of phenylalanine and tyrosine biosynthesis, and the cytokinin biosynthesis pathways. Enriched gene sets with decreased transcript abundance also included several A. thaliana signaling pathways such as ethylene signaling; JA, ethylene, and salicylic acid crosstalk signaling;and biological processes such as the GA-mediated signaling pathway, the IAA biosynthetic process, auxin efflux and influx transmembrane transporter activity, negative regulation of the ethylene-mediated signaling pathway, the tryptophan catabolic process, the endoplasmic reticulum unfolded protein response, and glutathione transferase activity.

Table 8 Uniquely enriched (P < 0.05) pathways and biological processes with increased or decreased transcript abundance in response to glyphosate treatment (2.24 vs. 0 kg ha^–1) during para- to endo-, and endo- to ecodormancy transitions identified via gene set enrichment analyses.

Paradormancy to Endodormancy	Endodormancy to Ecodormancy
Increased	Increased
Arabidopsis signaling pathways	Arabidopsis signaling pathways
Guard cell ABA† signaling	Gibberellin signaling
Stress ABA signaling	SCARECROW signaling
AraCyc pathways
Brassinosteroid biosynthesis II
Biological processes	Biological processes
Auxin polar transport	Auxin mediated signaling pathway
Cellular amino acid biosynthetic process	Auxin efflux transmembrane transporter activity
Cellular response to ethylene stimulus	Ethylene biosynthetic process
Regulation of G2–M transition of mitotic cell cycle	Long-day photoperiodism, flowering
Response to ABA stimulus	Regulation of circadian rhythm
Response to gibberellin stimulus	Response to ethylene stimulus
	Terpenoid biosynthetic process
Decreased	Decreased
Arabidopsis signaling pathways	Arabidopsis signaling pathways
Ethylene signaling	Plant growth auxin signaling
Jasmonic acid, ethylene, and salicylic acid crosstalk signaling	Senescence auxin signaling
AraCyc pathways	AraCyc pathways
ABA glucose ester biosynthesis	Glycolysis II (plant plastids)
Cytokinin glucoside biosynthesis	Superpathway of carotenoid biosynthesis
Ethylene biosynthesis from methionine	Superpathway of phenylalanine, tyrosine and tryptophan bio.
IAA degradation IV
Methionine salvage pathway
Monoterpene biosynthesis
Superpathway of phenylalanine and tyrosine biosynthesis
Biological processes	Biological processes
Auxin efflux	Amino acid transport
Auxin efflux and influx transmembrane transporter activity	Aromatic amino acid family metabolic process
Endoplasmic reticulum unfolded protein response	Carotenoid biosynthetic process
GA-mediated signaling pathway	Cellular amino acid biosynthetic process
Glutathione transferase activity	Flower morphogenesis
IAA biosynthetic process	Jasmonic acid mediated signaling pathway
Negative regulation of ethylene mediated signaling pathway	Plant-type cell wall biogenesis
Tryptophan catabolic process	Response to ABA stimulus
	Response to auxin stimulus
	UDP-glucosyltransferase activity

†

ABA, abscisic acid; IAA, indole acetic acid; GA, gibberellic acid.

In ecodormant UABs of foliar glyphosate-treated plants, uniquely enriched pathways with increased transcript abundance included A. thaliana signaling pathways such as gibberellin signaling, SCARECROW signaling, and numerous biological processes such as the auxin-mediated signaling pathway, auxin efflux transmembrane transporter activity, ethylene biosynthetic process, response to ethylene stimulus, long-day photoperiodism, flowering, regulation of circadian rhythm, and the terpenoid biosynthetic process (Table 8; Supplemental File S3). Ecodormant UABs containing uniquely enriched pathways with decreased transcript abundance included numerous AraCyc pathways such as glycolysis II (plant plastids), the superpathway of carotenoid biosynthesis, and the superpathway of phenylalanine, tyrosine, and tryptophan biosynthesis. Enriched gene sets with decreased transcript abundance included several A. thaliana signaling pathways such as plant growth auxin signaling, senescence auxin signaling, and numerous biological processes such as the JA-mediated signaling pathway, response to ABA stimulus, response to auxin stimulus, the aromatic amino acid family metabolic process, the carotenoid biosynthetic process, the cellular amino acid biosynthetic process, amino acid transport, flower morphogenesis, plant-type cell wall biogenesis, and UDP-glucosyltransferase activity.

Discussion

Vegetative growth rates are generally a good indicator of the dormancy status in UABs of leafy spurge. Because foliar glyphosate treatment (2.24 kg ha^–1) altered vegetative growth from paradormant UABs and induced stunted and bushy phenotypes, vegetative growth measurements were not an efficient method of directly determining transitions from para- to endo- and ecodormancy in UABs of foliar glyphosate-treated plants. However, the transcript abundance of several marker genes (e.g., COP1, HY5, MAF3-like, RVE1-like) and the results of the principal component analyses revealed that UABs of foliar glyphosate-treated plants subjected to simulated seasonal conditions (i.e., RD and RD + V) transitioned through endo- and ecodormancy (Fig. 2). These results confirm that foliar glyphosate treatment does not inhibit UABs of leafy spurge from transitioning through normal seasonally-induced dormancy phases. However, the vegetative growth patterns induced by foliar glyphosate treatment (i.e., stunted and bushy phenotypes) were persistent after going through phases of endo- or ecodormancy (Fig. 2 and Fig. 3).

Several reports indicate that glyphosate can be metabolized slowly in plants (Duke, 2010; Sandberg et al., 1980). For example, ¹⁴C-glyphosate levels remained similar in leaves and roots in several weedy species [Convolvulus arvensis L., Cirsium arvense (L.) Scop., Ipomoea purpurea (L.) Roth, Apocynum cannabinum L., and Asclepias syriaca L.] 3 to 4 wk after treatment, with the main glyphosate metabolite (aminomethylphosphonic acid) occurring only to a limited extent (Wyrill and Burnside, 1976; Sandberg et al., 1980). However, in Cyperus rotundus L., foliar-applied glyphosate was rapidly absorbed and converted to aminomethylphosphonic acid, and bud elongation from underground tubers of these plants was inhibited (Wang, 2001). Based on the stunted and bushy phenotypes observed in UABs of foliar glyphosate-treated plants, our results suggest that leafy spurge rapidly absorbs glyphosate. Indeed, studies indicated that 0.5 and 1% of ¹⁴C-labeled glyphosate translocated into the root system of leafy spurge within 12 h (Gottrup et al., 1976). Enzymes involved in metabolizing glyphosate (e.g., glyphosate oxidoreductase-like) have been speculated to exist in plants but no corresponding enzymes or genes have been identified yet (Duke, 2010). Therefore, we are not able to determine the extent of glyphosate metabolism based on transcript abundance alone. However, vacuolar sequestration of glyphosate has been proposed to result in lower shikimate levels, even in the presence of high cellular glyphosate levels (Ge et al., 2010).

Foliar Glyphosate Treatment Has Different Long-Term Effects on Shikimate Abundance

Although shikimate abundance initially increased in paradormant UABs 7 d after foliar glyphosate treatment, it decreased as UABs transitioned into endo- and ecodormancy. These results suggest that either seasonal conditions (i.e., a reduction in temperature and photoperiod) or a time lapse (i.e., 12–20 wk) allowed for dissipation of shikimate, or EPSPS was no longer inhibited and shikimate was used for downstream processes. These results are similar to those obtained in previous studies where foliar glyphosate treatment initially increased shikimate abundance in treated aerial tissues and UABs of leafy spurge plants 7 d after glyphosate treatment, whereas shikimate abundance was reduced in new shoots derived from UABs of foliar glyphosate-treated plants (Doğramaci et al., 2015a). Other studies also demonstrated that changes in shikimate abundance following glyphosate treatments are dynamic (Mueller et al., 2008). For example, in Trifolium repens L. and Conyza canadensis (L.) Cronquist, increasing shikimate levels were observed up to 6 d after treatment; in Chenopodium album L., increasing shikimate concentrations leveled off in ∼3 d; and in Amaranthus palmeri S.Watson and Polygonum pensylvanicum L., shikimate was accumulated at lower levels (compared with the other weeds) and then started to decrease within 6 d. Collectively, these results suggest that metabolism of glyphosate appears to be species-specific.

In this study, the minimal changes observed in abundance of transcripts involved in shikimate biosynthesis following foliar glyphosate treatment (Table 3) were similar to those previously observed in leafy spurge (Doğramaci et al., 2014a, 2015a) or other weedy species such as A. palmeri (Gaines et al., 2010) and C. canadensis (Nol et al., 2012). The enrichment of AraCyc pathways downstream of the shikimate pathway (e.g., the superpathway of phenylalanine, tyrosine, and tryptophan biosynthesis), which contained transcripts with decreased abundance, correlates well with the decreased shikimate levels during transitions from para- to endo- and ecodormancy (Table 8).

Despite the decreased shikimate levels, the persistent stunted and bushy phenotypes observed for new aerial shoot growth from UABs of foliar glyphosate-treated plants over time raises questions about what processes might be involved. Two different hypotheses have been proposed (Doğramaci et al., 2015a). First, foliar glyphosate-induced vegetative growth from UABs following decapitation of aboveground shoots causes root-sequestered glyphosate or its metabolites to be remobilized in UABs, and remobilization of these substances continually causes subsequent generations to be exposed to residual toxicity and impact shoot apical meristem maintenance. Alternatively, foliar glyphosate treatment causes irreversible effects in UABs after the treatment that persistently induces stunted and bushy phenotypes from these UABs. In this study, the results for vegetative growth and shikimate abundance may support the latter hypothesis, because the effects of foliar glyphosate treatment on vegetative growth remained persistent after going through seasonal phases of endo- and ecodormancy, whereas the abundance of shikimate gradually returned to the levels observed in the nontreated controls (Fig. 5). If glyphosate was sequestered in the root system and remobilized after growth induction, as mentioned in the first hypothesis, one would expect shikimate levels to increase in new shoots derived from endo- and ecodormant buds as well. Glyphosate's metabolites could also be sequestered in the root system and may affect the vegetative growth, but there is no indication in the literature that these metabolites cause stunted or bushy phenotypes in following generations in perennials. Although exogenous application of the main glyphosate metabolite (i.e., aminomethylphosphonic acid) did not affect shikimate levels in glyphosate-resistant soybean [Glycine max (L.) merr.], it did induce phytotoxic effects and some injuries (Reddy et al., 2004) but it did not induce stunted or bushy phenotypes, as observed in leafy spurge. Regardless, the reduction in shikimate levels and minor changes in the related transcripts did not completely explain the altered vegetative growth of UABs of foliar glyphosate-treated plants. Thus we examined the hormone and transcript profiles to determine the impact that foliar glyphosate treatment has on the pathways involved in growth and development.

Foliar Glyphosate Treatment Affected Auxin Transport and Signaling

An increased abundance of shikimate is a marker for inhibition of EPSPS (a critical step for the production of tryptophan to auxin biosynthesis pathways). Thus increased shikimate (Fig. 5) and decreased auxin levels (Fig. 6) in paradormant UABs 7 d after treatment, along with the decreased shikimate levels and increased auxin levels observed in endodormant UABs, correlate well. These results are also in line with previous studies, where foliar glyphosate-treatment led to increased auxin levels in subsequent generations of aerial shoots and decreased shikimate abundance (Doğramaci et al., 2015a) or where glyphosate treatment altered endogenous level of free IAA in shoots, hypocotylar regions, and roots (Westwood and Biesboer, 1985).

Although UABs of foliar glyphosate-treated plants had minimal changes in the abundance of transcripts involved in auxin biosynthesis in all phases of dormancy, glyphosate treatment did have major effects on the transcripts involved in auxin catabolism, response and signaling, and transport (Table 3). The results for the auxin levels in para- and endodormant UABs (Fig. 6) correlated well with several transcripts that are responsive to auxin (ARF5, IAA4, IAA9, IAA29, LRP1; Table 3). Because auxin is known to inhibit IAA's ability to repress some auxin response factor (Liscum and Reed, 2002), the foliar glyphosate-induced increase in the abundance of auxin in endodormant UABs (Fig. 6) correlates well with the positive (ARF5) or negative (IAA4, IAA29) impact on abundance of transcripts involved in auxin response and signaling; a similar trend was also observed in ecodormant UABs. Because ARF5 is involved in positive auxin signaling and IAA4 and IAA29 are involved in negative auxin signaling, it appears that foliar glyphosate treatment has a downstream impact that enhances positive auxin signaling in endo- and ecodormant UABs.

Further, the abundance of transcripts involved in auxin transport could also play a role in auxin signaling in UABs of foliar glyphosate-treated plants. For example, PILS7 had increased transcript abundance in paradormant UABs 7 d after treatment (Table 3); these results are in agreement with our previous reports, where the abundance of PILS7 increased significantly in UABs of leafy spurge 3 to 7 d after foliar glyphosate treatment (Doğramaci et al., 2014a). PILS are a family of auxin transport facilitator genes that are required for auxin-dependent regulation of plant growth and organ formation by determining the cellular sensitivity to auxin (Barbez et al., 2012; Feraru et al., 2012). PILS proteins facilitate intracellular auxin accumulation at the endoplasmic reticulum membrane, which regulates intracellular auxin accumulation and ultimately decreases nuclear auxin signaling (Barbez et al., 2012; Peer, 2013). Studies have demonstrated (Barbez et al., 2012) that ectopic expression of some PILS resulted in various abnormalities, including dwarfed a bushy A. thaliana plants. Thus significant increases in the abundance of PILS7 in paradormant UABs 7 d after treatment may play some role in dwarfed and bushy growth in the UABs observed in this study. However, the increased abundance of PILS7 was not significant in endo- and ecodormant UABs, and we hypothesize that environmental factors leading to endo- and ecodormancy override the initial impact that foliar glyphosate treatment has on the transcript abundance of PILS7 in paradormant UABs (Table 3; Supplemental File S1 Page 1).

Additionally, other transcripts involved in auxin transport also increased significantly (ABCB4) in paradormant UABs of foliar glyphosate-treated plants or decreased (e.g., ABCB1, PID, PIN3) in endodormant UABs. Significant changes in the abundance of some of these transcripts, such as PINs, have also been linked to altered phenotypes. PINs encode auxin efflux carriers that mediate tissue-specific cell-to-cell polar auxin transport (Peer, 2013) that is critical for maintenance of embryonic auxin gradients, developmental patterning, and differential growth responses (Blilou et al., 2005). These results are consistent with our previous studies that also highlighted the transcripts involved in auxin transport (e.g., PILS, PINs, ABC TRANSPORTERS), that were significantly affected by foliar glyphosate treatment under greenhouse and field conditions (Doğramaci et al., 2014a, 2015a, 2016). Indeed, enrichment of genes associated with the ‘auxin polar transport process’ and ‘auxin efflux and influx transmembrane transporter activity process’ (Table 8) supports our hypothesis that foliar glyphosate treatment is likely to impact polar auxin transport in UABs of leafy spurge. Additionally, enrichment of genes associated with IAA biosynthetic processes containing transcripts with decreased abundance may be caused by feedback mechanisms in response to increased levels of auxin (Spiess et al., 2014) in endodormant UABs. Although some biological processes associated with auxin (e.g., the auxin-mediated signaling pathway and auxin efflux transmembrane transporter activity) were enriched with transcripts showing increased abundance in ecodormant UABs of foliar glyphosate-treated plants, some A. thaliana signaling pathways (plant auxin signaling pathways, senescence auxin signaling) and biological processes (response to auxin stimulus) were enriched with transcripts showing decreased abundance. Collectively, these results suggest that foliar glyphosate treatment leads to long-term effects on auxin biosynthesis, transport, signaling, and response in UABs of leafy spurge. Because exogenous application of auxin represses stem elongation in shoots derived from UABs of leafy spurge (Chao et al., 2006; Horvath, 1998), the significant changes in the abundance of transcripts involved in auxin transport and bioactive IAA levels (Fig. 6) in UABs of foliar glyphosate-treated plants may have had some influence on the stunted and bushy phenotypes observed in this study.

Effects of Foliar Glyphosate Treatment on Cytokinins and Cell Cycle Processes are Likely to Contribute to the Altered Phenotypes

Cytokinins are known to play a positive role in cell division in A. thaliana (Werner et al., 2001; Jones et al., 2010). Although our data for free-base cytokinins was limited (Fig. 7), the results suggest that foliar glyphosate treatment led to increased abundance of free-base cytokinins in para- and endodormant UABs, which at first appears to contradict the decreased abundance of transcripts involved in cell cycle processes. However, because cytokinin catabolites (t-ZOG and c-ZOG) also increased in UABs of foliar glyphosate-treated plants (Fig. 7), our results are consistent with the increased abundance of transcripts involved in catabolism at all phases of dormancy (Table 4), suggesting that free-base cytokinins are likely to be catabolized. Because cytokinins also play positive roles in shoot branching (Vanstraelen and Benková, 2012) and regulation of auxin synthesis in A. thaliana (Jones et al., 2010), the changes in the cytokinin profiles observed in response to foliar glyphosate treatment could play some role in the stunted and bushy phenotypes observed in this study (Fig. 3). Indeed, changes in the cytokinins are known to impact polar auxin transport through modulation of auxin efflux carrier activity (Jones et al., 2010). Thus the results for cytokinin and auxin profiles (Fig. 6 and Fig. 7) and the transcript abundance of several auxin transporters (Table 3) suggest that foliar glyphosate treatment could impact auxin–cytokinin interactions in UABs, which might also play some role in the stunted and bushy phenotypes observed in this and other studies.

Impact of Foliar Glyphosate Treatment on GA Biosynthesis in UABs May Result from Disruption of Cellular Organelles

Even though the GA profiles were incomplete, foliar glyphosate treatment appeared to result in increased abundance of precursors of bioactive GA in UABs (Fig. 9). Additionally, transcripts involved in the initial steps of GA biosynthesis (CPS1/GA1, KS/GA2, KO/GA3) increased significantly as a result of foliar glyphosate treatment (Table 6) but the transcripts involved in downstream GA biosynthesis (KAO, GA3OX1) had only minor changes in abundance in paradormant UABs. These results are in line with our previous reports (Doğramaci et al., 2014a), where the most striking change in abundance was observed for CPS1/GA1 3 to 7 d after glyphosate treatment and only slight changes in abundance were observed for downstream transcripts. Although the significant changes in abundance were observed for some of these transcripts in endodormant UABs as well, the residual effects of foliar glyphosate treatment on these transcripts were minimized in ecodormant UABs.

These results are intriguing because GA biosynthesis is known to occur in different compartments in the cell, where CPS1/GA1, KS/GA2, and KO/GA3 catalyze the first steps of the GA biosynthetic pathway in the plastids (Railton et al., 1984) to produce ent-kaurenoic acid, which is then converted into GA12 by ent-kaurenoic acid oxidase in the endoplasmic reticulum. Conversion of GA12 to bioactive forms of GA occurs in the cytosol, and GA12 is a branch point for a spectrum of bioactive gibberellins catalyzed by 2-oxoglutarate–dependent dioxygenases (e.g., GA20OX and GA3OX). Because ent-kaurenoic acid oxidase is the rate-limiting step for production of GA intermediates (Fleet et al., 2003; Hedden and Phillips, 2000; Sponsel and Hedden, 2004), the insignificant change in abundance of this transcript in paradormant UABs 7 d after foliar glyphosate treatment suggests that bioactive GA biosynthesis was probably impaired. Indeed, glyphosate is known to rapidly disrupt proplastids, chloroplasts, and the endoplasmic reticulum in other species (Campbell et al., 1976; Mollenhauer et al., 1987). The results presented in this study support our hypothesis (Doğramaci et al., 2014a) that metabolite flux through subcellular compartments could be impacted in paradormant UABs of leafy spurge after foliar glyphosate treatment. More specifically, increased transcript abundance for the initial steps of GA biosynthesis in plastids but decreased abundance of downstream GA biosynthesis occurring in the endoplasmic reticulum and in the cytoplasm could suggest that foliar glyphosate treatment disrupts metabolite flux between these subcellular compartments, which also could play some role in the stunted phenotypes observed (Fig. 2, Fig. 3). Indeed, this hypothesis is supported by the fact that mutants of CPS1/GA1 and KS/GA2 produce dwarfed phenotypes in A. thaliana (Yamaguchi et al., 1998), and loss of function of CPS1/GA1 and KS/GA2 also has been shown to cause a dwarfed phenotype in rice (Oryza sativa L.) (Sakamoto et al., 2004). Additionally, loss of KO/GA3 or KAO gene function causes dwarfed phenotypes in pea (Pisum sativum L.) (Davidson et al., 2003), and some chemicals [e.g., phosphon D (tributyl[2,4-dichlorobenzyl]phosphonium chloride), cycocel (2-chloro-N,N,N,-trimethylethanaminium chloride), ancymidol (cyclopropyl[4-methoxyphenyl]5-pyrimidinylmethanol), and paclobutrazol ([2S,3S]-1-[4-chlorophenyl]-4,4-dimethyl-2-[1H-1,2,4-triazol-1-yl]-3-pentanol)] that are known to block synthesis of GA also induce stunted growth (Sponsel and Hedden, 2004).

Changes in ABA May Result from the Impact Foliar Glyphosate Treatment has on Plastids in UABs

Foliar glyphosate treatment induced a decrease in the abundance of bioactive ABA and its metabolites in UABs compared with the controls at all dormancy phases (Fig. 8). Transcripts (NCED3, ABA4) in the upstream ABA biosynthesis pathway in plastids had decreased abundance, whereas some transcripts (AAO4, ABA2) involved in downstream ABA biosynthesis in the cytosol had increased abundance in paradormant UABs of foliar glyphosate-treated plants 7 d after treatment (Table 5). The significant decrease in abundance of these upstream genes (NCED3, ABA4) could be caused by early disruption of plastids in paradormant UABs after foliar glyphosate treatment. Because glyphosate's target site (EPSPS) is a chloroplast-localized enzyme (Della-Cioppa et al., 1986) of the shikimate biosynthetic pathway, it is not surprising that many processes, such as the initial steps of ABA or GA biosynthesis that occur in the plastids, are significantly affected.

Foliar Glyphosate-Induced Dwarfed Phenotypes Could Involve Ethylene or GA Signaling Crosstalk

Ethylene is involved in numerous processes throughout the plant's lifecycle including growth, senescence, and stress responses (Bleecker and Kende, 2000). Although the abundance of several transcripts involved in ethylene signaling and response (e.g., AP2/ERF TFs) in paradormant UABs of foliar glyphosate-treated plants were affected 7 d after treatment, the environmental conditions that induced endo- or ecodormancy appeared to lessen these effects (Table 7). In this study, foliar glyphosate treatment affected ethylene signaling pathways in paradormant UABs, as observed by the increased transcript abundance for DREB26 and ERF1 in this and previous studies (Doğramaci et al., 2014a, 2016). Indeed, exogenous application of the ethylene precursor (1-aminocyclopropane-1-carboxylic acid) caused increases in the transcript abundance of DREB26 and ERF1 in leafy spurge crown buds that caused stunted vegetative growth (Doğramaci et al., 2013). Further, overexpression of several AP2/ERFs, including DREB26, has been shown to result in dwarfed phenotypes in other species (Krishnaswamy et al., 2011; Mizoi et al., 2012), and overexpression of some members of the DREB1 class of AP2/ERF TFs are known to induce dwarfism through regulating GA metabolism (Magome et al., 2008; Tong et al., 2009), which would correlate well with the GA results observed in this study.

Changes in Xenobiotic Gene Families Suggest That Foliar Glyphosate Treatment Induces Long-Term Phytotoxic Effects beyond Inhibition of EPSPS

Members of large gene families known to be associated with detoxification processes (ABC TRANSPORTERS, GLUTATHIONE S-TRANSFERASES, P450s, and UGTs) had differential expression in UABs of foliar glyphosate-treated plants at all phases of dormancy (Supplemental File S1 Page 3–5). Previous reports indicate that P450s, together with glutathione S-transferases and UGTs, are involved in herbicide biochemical modification through metabolism, whereas ABC transporters are involved in compartmentalizing herbicides and their metabolites in other plant species, and this process may be linked to herbicide resistance in some weeds (Werck-Reichhart et al., 2000; Brazier et al., 2002; Peng et al., 2010; Busi et al., 2011; Cummins et al., 2013). In this study, it is possible that members of these gene families could also play some role in reducing the residual phytotoxic effects that foliar glyphosate treatment appears to have on leafy spurge UABs. If we assume that these leafy spurge homologs perform similar functions as they do in A. thaliana, significant changes in the abundance of these gene families suggest that detoxification processes could play some direct or indirect role in reducing the foliar glyphosate-induced phytotoxic effects in UABs; however, they do not prevent the stunted and bushy phenotypes from being expressed in subsequent generations of vegetative tissue derived from UABs.

Conclusions

Because nonuniform emergence of vegetative shoots allows weedy plants to escape conventional control measures, determining how key processes in molecular networks regulate bud dormancy and vegetative reproduction could identify critical new targets for manipulating vegetative reproduction. In this study, we used 2.24 kg ha^–1 of foliar applied glyphosate to determine its short- and long-term effects on vegetative reproduction of leafy spurge under controlled environments, aiming to cover gaps or questions raised from previous studies related to the effects of foliar glyphosate treatment on vegetative reproduction from UABs under controlled environments (Doğramaci et al., 2014a, 2015a) or under field conditions (Doğramaci et al., 2016).

Foliar glyphosate treatment of leafy spurge inhibited EPSPS, as determined by increased shikimate abundance in paradormant UABs; however, the abundance of shikimate dissipated over time as UABs transitioned from para- to endo- and ecodormancy. Overall, these results suggest that the effects of foliar glyphosate treatment on target site inhibition in UABs is short-lived. Indeed, the most significant changes in transcript abundance often occurred in paradormant UABs within 7 d after the glyphosate treatment; however, the abundance of transcripts involved in JA biosynthesis suggest that foliar glyphosate treatment probably caused some long-term stress responses in endo- and ecodormant UABs. Although abundance of shikimate returned to normal levels in endo- and eco-dormant UABs of foliar glyphosate-treated plants, these changes were not sufficient to prevent the dwarfed and bushy phenotypes resulting from UABs of foliar glyphosate-treated plants. Collectively, these results suggest that UABs of leafy spurge were irreversible affected within 7 d after glyphosate treatment, but the cause of these long-term phytotoxic effects will require further studies.

It should not be surprising that foliar glyphosate treatment induces phytohormonal disturbance and affected plant growth and development in UABs of leafy spurge, since studies indicate that glyphosate treatment mainly affects shoot and root apical meristems, which are the production sites of phytohormones (Cakmak et al., 2009; Gomes et al., 2014). Regardless, most of the significant processes identified in these studies are associated with the plastids or the endoplasmic reticulum, which would be in agreement with glyphosate's known effects on disruption of these organelles in other species (Campbell et al., 1976; Mollenhauer et al., 1987). In this study, significant changes were observed in the abundance of transcripts involved in the initial steps of GA biosynthesis, which occur in plastids, but insignificant changes were observed for downstream transcripts involved in GA biosynthetic processes occurring in the endoplasmic reticulum and cytoplasm. These results suggest that metabolite flux through these compartments were likely to be affected in UABs of glyphosate-treated plants. Further, the increase in the transcript abundance of PILS7 is also a noteworthy outcome and suggests that the impact of foliar glyphosate treatment on PILS7 in UABs could affect intracellular auxin accumulation at the endoplasmic reticulum membrane and in downstream auxin signaling. Likewise, changes in the transcript abundance of PINs and members of ABC TRANSPORTERS in UABs of foliar glyphosate-treated leafy spurge also suggest that auxin homeostasis was affected. Indeed, significant changes in the abundance of these transcripts are linked to dwarfed and bushy phenotypes in other species. Further research on critical targets of the pathways identified in this study (e.g., ABC TRANSPORTERS, GA1/CPS1, PILS7, PINs) will be needed to determine how altering these potential targets could serve for manipulating plant growth and development in perennials.

Supplemental Information

Supplemental File S1. Excel spreadsheets for RNA-seq results summary; transcript data used for Tables 3–7; differentially expressed transcripts identified by pairwise comparisons for glyphosate-treated vs. controls (2.24 vs. 0 kg ha^–1) at para-, endo-, and ecodormant phases.

Supplemental File S2. Excel spreadsheets containing a list of selected primer pairs used for qRT-PCR studies, RNA-seq and qRT-PCR results for selected genes, Pearson and Spearman's rank correlation results, and a list of the phytohormones quantified.

Supplemental File S3. Excel spreadsheets containing the results of the gene set enrichment analysis.

Conflict of Interest Disclosure

The authors declare that there is no conflict of interest.

Acknowledgments

The authors thank Wayne A. Sargent, Leonard W. Cook, Cheryl A. Huckle, Brant B. Bigger, and Laura C. Nessa for their technical assistance. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.