Metabolomic Analysis of Cassava Cultivars in Response to Sri Lankan Cassava Mosaic Virus Infection by Viruliferous Whiteflies

Abstract

Cassava mosaic disease, caused by Cassava mosaic begomoviruses in the family Geminiviridae, poses a major threat to cassava production, with Sri Lankan cassava mosaic virus (SLCMV) being the dominant strain in Southeast Asia. Transmitted via infected propagative stems and whiteflies (Bemisia tabaci), SLCMV’s impact on cassava metabolite dynamics remains poorly understood. This study investigated metabolite profile changes in resistant, tolerant, and susceptible cassava cultivars at 1, 3, and 7 days after inoculation by viruliferous whiteflies. Distinct metabolite patterns were observed among cultivars, with several pathways linked to plant defense identified, including flavonoid biosynthesis, phenylpropanoid biosynthesis, and purine metabolism. Secondary metabolite pathways, such as the energy-signaling SnRK1/AMPK-liked proteins, alpha-linolenic acid metabolism, and starch and sucrose metabolism, were also implicated. The results provide insights into metabolite-mediated defense mechanisms during early and late infection, enhancing understanding of cassava’s responses to SLCMV inoculation after exposure to viruliferous whitefly infestation. This study supports the development of SLCMV-resistant cassava cultivars.

Abiotic and biotic stresses induce significant changes in metabolic regulation and adaptations in plants. Changes in metabolic responses underlie changes in the levels of reactive oxygen species (ROS) and the release of chemicals such as phenolic and antioxidant compounds (Mittler, 2002; Mittler et al., 2004). Additionally, secondary metabolites (such as terpenoids and flavanols) may induce phytohormones during plant immune responses (Anjali et al., 2023). In particular, virus-infected infestation has been shown to alter plant metabolites, affecting virus infection of the plant by influencing the insect feeding behavior. Plant viruses can then modify host metabolites throughout insect feeding secretions (Fereres et al., 2016; Pan et al., 2021; Wu et al., 2019).

Plant viruses are primarily transmitted by insect vectors (including planthoppers, thrips, aphids, and whiteflies), accounting for over 80% of plant virus transmission (Lefeuvre et al., 2019; Roossinck, 2013). Viral outbreaks are closely linked to insect feeding behavior and insect migration between infected and healthy host plants. Controlling the spread of plant viruses requires strict management practices based on insect vector preferences and plant host susceptibility (Power, 2000). Infection and transmission involve a triad of interactions between viruses, insect vectors, and host plants, which influence virus spread, epidemics, inoculation, and insect population dynamics (Blanc and Michalakis, 2016; Eigenbrode et al., 2018; Mauck et al., 2012, 2018; Pan et al., 2021; Zhao et al., 2022). Therefore, studying and manipulating the interactions between viruses and insect vectors is key.

Begomoviruses, a type of DNA virus, are continually transmitted by insect vectors. They utilize insect vectors that enable them to pass from the insect gut into the circulatory system and reside in the salivary glands, thus leading to persistent transmission (Gilbertson et al., 2015; Power, 2000). Whiteflies (Hemiptera: Aleyrodidae), specifically the Bemisia tabaci (Gennadius) species, have been reported to be the exclusive vectors for begomovirus (Chikoti and Tembo, 2022; Fiallo-Olivé et al., 2020; Tembo et al., 2017). Following acquisition by whiteflies, begomovirus capsid proteins are continuously transferred to the anterior midgut and salivary glands, facilitating productive transmission (Ghanim et al., 2001; Ghosh et al., 2021; Gildow, 1993; Noris et al., 1998; Ohnishi et al., 2009; Uchibori et al., 2013; Wei et al., 2014). Whiteflies serve as primary vectors for various viral diseases such as cassava mosaic disease (CMD) and cassava brown streak disease (Maruthi et al., 2017; Minato et al., 2019; Saokham et al., 2021).

A major cause of CMD in Southeast Asia (SEA) is Sri Lankan cassava mosaic virus (SLCMV; genus Begomovirus, family Geminiviridae). Its spread across SEA’s cassava cultivation areas has led to significant yield losses. The dissemination of SLCMV is facilitated by the dissemination of B. tabaci and infected propagative stems (Duraisamy et al., 2012; Graziosi et al., 2016; Minato et al., 2019). In Asia, SLCMV transmission efficiency varies among three whitefly species (Asia I, Asia II-1, and Asia II-7), with Asia II-1 being the primary vector for SLCMV (Chaubey et al., 2015; Chi et al., 2020; Chittarath et al., 2021; Malik et al., 2022; Saokham et al., 2021; Wang et al., 2016).

The first SLCMV outbreak was reported in Kaun Moum, Cambodia, in 2015 (Wang et al., 2016). Subsequent sequencing studies revealed a high incidence of CMD in cassava production areas across Cambodia, Laos, and Vietnam, attributed to transmission by whiteflies (Chittarath et al., 2021; Uke et al., 2018; Wang et al., 2016). This led to a decline in cassava production in these regions (Pérez-Clemente et al., 2019). In Thailand, which is a leading cassava producer in SEA and globally, SLCMV was first detected in 2018 (Rakkrai et al., 2022). Recently, Saokham et al. (2021) found that 95% of SLCMV-infected cases in Thailand were transmitted by B. tabaci.

A significant challenge in combating SLCMV and ensuring cassava productivity is the lack of comprehensive understanding and characterization of SLCMV transmission by whiteflies to cassava plants, particularly in terms of plant metabolome annotation after SLCMV inoculation. Metabolomics presents a valuable tool for unraveling the intricacies of SLCMV transmission by whiteflies and elucidating broader implications for the spread of viruses, particularly those in the Begomovirus genus. In particular, metabolomics offers insights into how susceptible, tolerant, and resistant plants respond to viral infections (regarding plant metabolic processes), thereby altering the likelihood of virus infection. To further understand and characterize SLCMV transmission by whiteflies to cassava plants, the gap between the plant genome (a template that emerged via evolution) and metabolites (the resulting products) must be bridged in the context of the central dogma of molecular biology (i.e., genetic information flows only in one direction, from DNA, to RNA, to protein). In this context, metabolites are the end products of various biochemical pathways, with DNA encoding RNA, which encodes enzymes that catalyze reactions that ultimately lead to the production of specific metabolites. In summary, metabolomics will facilitate a deeper understanding of the chemical modifications in the host plant (Wintermantel, 2018).

This study examines the metabolite profiles of three cassava cultivars—TMEB 419 (resistant), Kasetsart 50 (KU 50) (tolerant), and Rayong 11 (R 11) (susceptible)—at 1, 3, and 7 days after inoculation (dai) with SLCMV following exposure to viruliferous whiteflies. Under field conditions, CMD symptoms typically appear after approximately one month of whitefly infestation, though asymptomatic plants may still harbor latent SLCMV infections, detectable only through molecular tools. The rate of symptom development varies among cassava landraces, highlighting the challenges of early detection and disease management. Our study emphasizes the importance of early molecular monitoring before visual symptom emergence, providing insights into the initial interactions between whiteflies, the viral pathogen, and cassava. Understanding these early-stage dynamics and their link to metabolite modifications after vector transmission of SLCMV may offer critical perspectives on cassava’s transcriptomic and proteomic responses to SLCMV infection.

Materials and Methods

Plant materials

Propagative materials of cassava landraces TMEB 419, KU 50, and R 11 were provided by the Thai Tapioca Development Institute (Thailand). They were planted under greenhouse conditions at the Department of Plant Pathology of the Faculty of Agriculture at Kasetsart University (Thailand). The greenhouse was maintained at 27–29°C, with a 14-h light period and 70–80% relative humidity.

Non-viruliferous whitefly propagation

Whitefly nymphs (B. tabaci) were collected from cassava fields in Sakaeo and Nakhon Ratchasima provinces and then transferred to green eggplants (Solanum melongena) for propagation. The whiteflies were reared in net cages at 27–29°C, with a 14-h light period and 70–80% relative humidity.

Fourth-generation adult whiteflies were confirmed as non-viruliferous (disease-free) using a modified protocol from Saokham et al. (2021) and Holterman et al. (2006) to verify the absence of SLCMV infection. First, the whiteflies underwent DNA extraction by grinding them in lysis buffer (200 mM NaCl, 200 mM Tris-HCl [pH 8.0], β-mercaptoethanol, and 10 mg/mL proteinase K), incubating the mixture at 65°C for 90 min, and centrifuging the mixture; the DNA was then collected. PCR was performed using the AV1 gene-specific primers (forward strand: 5′-GTT GAA GGT ACT TAT TCC C-3′; reverse strand: 5′-TAT TAA TAC GGT TGT AAA CGC-3′) described by Saokham et al. (2021). The PCR conditions involved initial denaturation at 94°C for 5 min; 35 cycles of 94°C for 40 s, 55°C for 40 s, and 72°C for 40 s; and a final elongation at 72°C for 5 min. The PCR products were visualized using gel electrophoresis with 1.2% agarose gel containing RedSafe Nucleic Acid Staining Solution (iNtRON Biotechnology, Seongnam, Korea) in 0.5× Tris-acetate-ethylenediaminetetraacetic acid (TAE) buffer.

SLCMV acquisition and inoculation by whiteflies

To enhance SLCMV acquisition efficiency, fourth-generation non-viruliferous adult whiteflies underwent a 1-h fasting period before a 48-h acquisition access period (AAP) on SLCMV-infected cassava plants in net cages. SLCMV infection in whiteflies was then confirmed via PCR, as described above.

Next, the viruliferous whiteflies (approximately 30 whiteflies per cage) were transferred to 9 net cages (three cultivars × three replicates, with each replicate in a separate cage), each containing a 1-month-old cassava plant, and they were then subjected to a 24-h inoculation access period (IAP) (Kaliappan et al., 2012). There was a 1-h fasting period before the IAP (Fig. 1).

Fig. 1.

Sri Lankan cassava mosaic virus (SLCMV) acquisition and inoculation by whiteflies. CMD, cassava mosaic disease.

At 1, 3, and 7 dai, leaves from the resultant SLCMV-infected cassava plants were collected and immediately frozen in liquid nitrogen at −20ºC until ultra-high-performance liquid chromatography high-resolution mass spectrometry (UHPLC-HRMS/MS) analysis.

Whitefly species identification

Whitefly species identification was confirmed by amplifying the mitochondrial cytochrome oxidase 1 (mtCO1) gene using primers C1-J-2195 (5′-TTG ATT TTT TGG TCA TCC AGA AGT-3′) and L2-N-3014 (5′-TCC AAT GCA CTA ATC TGC CAT ATT A-3′) (Frohlich et al., 1999; Saokham et al., 2021). The PCR amplification product size was approximately 1,258-bp. The PCR conditions involved initial denaturation at 94°C for 5 min; 35 cycles at 94°C for 40 s, 52°C for 40 s, and 72°C for 40 s; and a final elongation at 72°C for 5 min. PCR-positive samples were sequenced by Macrogen (Seoul, Korea) and subjected to a BLASTn analysis using the National Center for Biotechnology information (NCBI) database.

DNA isolation and viral load quantification

Genomic DNA was isolated from cassava leaves of three distinct cultivars at 1 and 7 dai using a modified cetyl trimethylammonium bromide technique (Doyle and Doyle, 1987). The resulting DNA pellet was dissolved in nuclease-free water with contained RNase (100 μg/mL; Thermo Fisher Scientific, Waltham, MA, USA). DNA quality and quantity were assessed by 1.5% agarose gel electrophoresis in 0.5× TAE buffer (containing RedSafe stain, iNtRON Biotechnology) and spectrophotometric analysis, respectively. Isolated DNA was stored at −20°C until further use.

Virus loads (titers) were quantified using quantitative polymerase chain reaction (qPCR) with SYBR Green dye on a CFX96 Connect Real-Time PCR System (Bio-Rad, Hercules, CA, USA). Amplification, targeting the targeting the AV1 gene, was performed in a total volume of 10 μL per reaction. Each reaction contained 0.5 μL of forward primer (5′-TAAGAGGTTTTGCGTTAAGTCCG-3′) and reverse primer (5′-TGTACAGCATCAATGCATTCTCG-3′). The reaction mixture also included 3 μL of nuclease-free water, 1 μL of DNA template (100 ng/mL), and 5 μL of 2× qPCRBIO SyGreen Mix Lo-ROX (Copenhagen Biotech Supply, Bronshoj, Denmark). The qPCR thermal cycling protocol began with initial denaturation at 95°C for 2 min, followed by 40 cycles consisting of denaturation at 95°C for 5 s, annealing at 60°C for 30 s, and extension at 65°C for 5 s. Fluorescence signals were recorded after each cycle to monitor viral amplification. A melting curve analysis was performed from 65°C, with a 10-s hold at every 1°C increment. All reactions were conducted in triplicate. A standard curve was generated in each run by amplifying known quantities of the SLCMV DNA template, derived from SLCMV-infected cassava, using a series of five 10-fold serial dilutions (from 10² to 10⁻²). Viral loads in samples collected at 1 and 7 dai post SLCMV infection were calculated by plotting their Ct values against the standard curve. Data visualization was performed using RStudio (Posit PBC, Boston, MA, USA).

Metabolite extraction and UHPLC-HRMS/MS profiling

Metabolites were extracted from powdered cassava leaves by shaking with 70% methanol. The clear solution was then analyzed by UHPLC-HRMS/MS, separating the metabolites using a UHPLC system (Vanquish, Thermo Fisher Scientific, Inc.) coupled with a Q-Exactive HF-X hybrid quadrupole-Orbitrap mass spectrometer system (Thermo Fisher Scientific, Inc.). The UHPLC system involved a Hypersil GOLD Vanquish C18 column (1.9 μm, 2.1 × 100 mm; Thermo Fisher Scientific, Inc.) with a guard column. The temperature was maintained at 40°C and the flow rate was 0.35 mL/min. The mobile phases A and B consisted of 0.1% (v/v) formic acid in water and 0.1% (v/v) formic acid in acetonitrile, respectively. The injection volume was 5 μL. The gradient elution was as follows: 0–4 min, 5% B; 4–14 min, 5–90% B; 14–18 min, 90% B; and 18–18.5 min, 90–5% B. The total run time was 25 min. The MS detection was performed using a Q-Exactive HF-X Orbitrap mass spectrometer with a heated electrospray ionization source (HESI) (Thermo Fisher Scientific, Inc.). Positive ions were detected using the full-scan MS1/data-dependent MS2 (dd-MS2) mode with the following parameters: full-scan MS1 resolution, 120,000; dd-MS2 resolution, 30,000; mass range, 150–2,000 m/z; auxiliary gas heater temperature, 400°C; capillary temperature, 320°C; maximum injection time, 30 ms; automatic gain control target, 3 × 10⁶; auxiliary gas, 10 arbitrary units (AU); sheath gas, 45 AU; sweep gas, 5 AU; and spray voltage, 3.5 kV (positive) or 2.5 kV (negative).

The raw MS data were collected using three technical replicates for each sample to ensure reproducibility, then processed using Compound Discover 3.3 (Thermo Fisher Scientific, Inc.). Normalization was carried out using quality control (QC) samples. The MS data were then used to search the plant metabolite databases of mzVault, ChemSpider, mzCloud, and Metabolika to identify metabolites.

Data analysis

Each identified metabolite was aligned based on its peak intensity (which represents the abundance of each metabolite individually). The raw data were normalized and exported as comma-separated values (.csv) files, with adjustments for column alignment. The data QC threshold was set to 25% standard deviation in MetaboAnalyst 6.0 (Pang et al., 2024).

The metabolomic profiles were analyzed using hierarchical clustering dendrograms, 2D principal component analysis (PCA) score plots, and heatmaps to illustrate the relationships among the three cultivars (TMEB 419, KU 50, and R 11) at each time point (1, 3, and 7 dai). Clustering was performed using Euclidean distance and the Ward clustering algorithm, visualized using a dendrogram. Volcano plots were generated using SRplot (http://www.bioinformatics.com.cn/srplot), with a significance threshold of P < 0.01. Log₂(fold change) for each metabolite was calculated as Δlog₂ [Δlog₂ (fold change of metabolite expression) = log₂ (initial period) – log₂ (late period)], and then the P-values, f.value and the false discovery rate were extracted from one-way ANOVA, which was conducted using R on the MetaboAnalyst platform.

Pathway enrichment analysis

Pathway enrichment analysis was conducted in MetaboAnalyst 6.0 by first annotating the metabolites using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (based on Arabidopsis thaliana), supplemented with data from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and Human Metabolome Database (HMDB). Next, the KEGG annotations were used to determine which functional pathways were affected, with significantly regulated metabolites being analyzed in gene-protein and protein-protein interaction networks based on the KEGG database (https://www.genome.jp/kegg/).

Results

SLCMV acquisition, whitefly species identification, and viral load quantification

After a 48-h AAP, the whiteflies were confirmed to have acquired SLCMV by PCR amplification of the AV1 gene. All the whiteflies tested positive for SLCMV, with the PCR product being associated with an approximately 928-bp band (Supplementary Fig. 1).

To confirm the whitefly species, the mtCO1 gene underwent PCR amplification, producing a 1,258-bp band, which was sequenced. BLASTn analysis indicated 99% similarity to the B. tabaci Asia II-1 nucleotide sequence.

Viral quantification by qPCR revealed distinct differences in SLCMV between 1 and 7 dai. A standard curve generated from five serial dilutions of SLCMV-infected samples exhibited a strong linear correlation, described by the equation y = −3.5333x + 51.183, with an R² value greater than 0.9997. At 1 dai, the SLCMV copy number (copies/ng) were comparable across all three cultivars: TMEB 419, KU 50, and R 11, all showing approximately 2.69E+05 copies/ng. However, by 7 dai, significant differences in viral accumulation were observed: R 11 exhibited the highest viral load at 1.58E+11 copies/ng, followed by KU 50 at 2.97E+09 copies/ng, while TMEB 419 maintained a relatively low viral load of 5.22E+05 copies/ng (Supplementary Table 1). The differential accumulation is further illustrated in Fig. 2, which present the viral copy number converted to a log10 fold scale, clearly showing the highest accumulation in R 11, a moderated increase in KU 50, and only minor changes in TMEB 419.

Fig. 2.

Investigation of comparative Sri Lankan cassava mosaic virus (SLCMV) concentrations (copies/ng) among TMEB 419, KU 50, and R 11 cultivars between 1 and 7 dai, by using the qPCR technique.

Metabolite clusters among resistant, tolerant, and susceptible cassava cultivars at 1, 3, and 7 dai

There were 67 metabolites that were first identified by untargeted metabolomics, using UHPLC-HRMS/MS analysis, and then found in at least three of the studied databases (mzCloud, ChemSpider, mzVault, and Metabolika). This yielded metabolite profiles for each of the three cultivars at 1, 3, and 7 dai (Supplementary Table 2). A hierarchical clustering dendrogram, 2D PCA score plot, and heatmap were generated using MetaboAnalyst to visualize the data.

The hierarchical clustering dendrogram revealed two main branches: (1) one with sub-branches grouping (i) R 11 at 1, 3, and 7 dai and (ii) KU 50 at 1 and 3 dai, and (2) a second with sub-branches grouping (i) TMEB 419 and KU 50 at 7 dai and (ii) TMEB 419 at 1 and 3 dai. Notably, KU 50 (tolerant) at 1 and 3 dai aligned more closely with R 11 (susceptible), while KU 50 (tolerant) at 7 dai clustered with TMEB 419 (resistant), suggesting more similar metabolite profiles in TMEB 419 and late-stage-infected KU 50 (Fig. 3).

Fig. 3.

Hierarchical clustering dendrogram of differential metabolite profiles in TMEB 419, KU 50, and R 11 at 1, 3, and 7 days after Sri Lankan cassava mosaic virus (SLCMV) inoculation (dai) via viruliferous whitefly infestation, which involved calculating the distance among populations using Euclidean distance and the Ward clustering algorithm.

The 2D PCA score plot demonstrated clear separation among the cultivars, reflecting distinct metabolite shifts after SLCMV inoculation via viruliferous whitefly infestation. Variation in PC2 (3.2%) and PC3 (0.7%) accounted for some of the variation in the cultivars, with distinct and overlapping regions in the 2D PCA score plot illustrating different metabolite response dynamics (Fig. 4).

Fig. 4.

2D principal component analysis score plot of relationships between TMEB 419, KU 50, and R 11 at 1, 3, and 7 dai. Each color represents a group of replicates.

Hierarchical clustering and PCA analyses revealed clear metabolic differentiation between R 11 (susceptible) and TMEB 419 (resistant) at all three time points based on metabolite expression peak intensity. Interestingly, KU 50 (tolerant) initially clustered with R 11 (susceptible) but later (at 7 dai) shifted toward TMEB 419 (resistant). This may be related to unique metabolic enrichment over time. These findings suggest that SLCMV inoculation after exposure to viruliferous whitefly infestation leads to distinct metabolic patterns in TMEB 419 (resistant) vs R 11 (susceptible), and similarity between KU 50 (tolerant) and TMEB 419 (resistant).

Metabolite functional annotation and pathway enrichment analysis

The functions of the metabolites and the results of the pathway enrichment analysis are presented in Table 1. The pathway enrichment analysis revealed that the identified metabolites were primarily associated with flavonoid biosynthesis (14.93%), flavone and flavanol biosynthesis (13.43%), phenylpropanoid biosynthesis (7.46%), phenylalanine, tyrosine, and tryptophan biosynthesis (5.97%), and purine biosynthesis (4.48%) (Fig. 5). Other identified pathways included pathways related to secondary metabolite biosynthesis, such as the energy-signaling SnRK1/AMPK liked, alpha-linolenic acid metabolism, indole alkaloid biosynthesis, starch and sucrose metabolism, and biosynthesis of unsaturated fatty acids.

Table 1.

Pathway functional annotation of identified metabolites

Metabolites	Pathway IDs			FDR	F-value
	HMDB ID	PubChem ID	KEGG ID
Flavonoid biosynthesis
Chlorogenic acid	HMDB0003164	1794427	C00852	2.80E-20	780.52
Kaempferol	HMDB0005801	5280863	C05903	1.04E-15	208.23
Naringenin	HMDB0002670	439246	C00509	1.08E-13	120.23
Taxifolin	METPA0191	NM	C01617	1.07E-13	120.74
Apigenin	HMDB0002124	5280443	C01477	9.53E-18	366.11
Quercetin	HMDB0005794	5280343	C00389	2.08E-19	587.7
(−)-Epigallocatechin	HMDB0038361	72277	C12136	1.35E-11	67.87
(−)-Epicatechin	HMDB0001871	72276	C09727	3.82E-11	59.97
Myricetin	HMDB0002755	5281672	C10107	5.47E-11	57.32
Flavone and flavonol biosynthesis
Kaempferol	HMDB0005801	5280863	C05903	1.04E-15	208.23
Apigenin	HMDB0002124	5280443	C01477	9.53E-18	366.11
Quercetin	HMDB0005794	5280343	C00389	2.08E-19	587.7
Quercitrin	HMDB0033751	5280459	C01750	6.42E-10	42.38
Chlorogenic acid	HMDB0003164	1794427	C00852	2.80E-20	780.52
Hyperoside	HMDB0030775	5281643	C10073	1.59E-13	114.62
Isorhamnetin	HMDB0002655	5281654	C10084	8.51E-16	214.07
Rutin	HMDB0003249	5280805	C05625	1.14E-20	895.49
Tangeritin	HMDB0030539	68077	C10190	9.37E-23	1,693.4
Phenylalanine, tyrosine and tryptophan biosynthesis
L-Tyrosine	HMDB0000158	6057	C00082	1.70E-13	113.36
L-Tryptophan	HMDB0000929	6305	C00078	1.02E-18	483.27
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
Xanthurenic acid	HMDB0000881	5699	C02470	6.08E-04	6.49
Phenylpropanoid biosynthesis
Chlorogenic acid	HMDB0003164	1794427	C00852	2.80E-20	780.52
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
p-coumaric acid	HMDB0002035	637542	C00811	1.33E-10	51.52
Esculetin	HMDB0030819	5281416	C09263	6.84E-15	166.98
Vanillin	HMDB0012308	1183	C00755	5.58E-06	13.28
Indole alkaloid biosynthesis
L-Tryptophan	HMDB0000929	6305	C00078	1.02E-18	483.27
Isoquinoline alkaloid biosynthesis
L-Tyrosine	HMDB0000158	6057	C00082	1.70E-13	113.36
Purine metabolism
Adenosine	HMDB0000050	60961	C00212	8.71E-10	40.76
Guanine	HMDB0000132	764	C00242	5.30E-18	393.53
Guanosine	HMDB0000133	6802	C00387	7.82E-16	217.16
Tropane, piperidine and pyridine alkaloid biosynthesis
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
Ubiquinone and other terpenoid-quinone biosynthesis
L-Tyrosine	HMDB0000158	6057	C00082	1.70E-13	113.36
p-coumaric acid	HMDB0002035	637542	C00811	1.33E-10	51.52
Phenylalanine metabolism
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
Pentose and glucuronate interconversions
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
Tyrosine metabolism
L-Tyrosine	HMDB0000158	6057	C00082	1.70E-13	113.36
Arginine biosynthesis
L-Arginine	HMDB0000517	6322	C00062	2.79E-09	35.26
Glucosinolate biosynthesis
L-Tryptophan	HMDB0000929	6305	C00078	1.02E-18	483.27
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
Glycerolipid metabolism
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
Biosynthesis of unsaturated fatty acids
Eicosapentaenoic acid	HMDB0001999	446284	C06428
Starch and sucrose metabolism
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
Sesquiterpenoid and triterpenoid biosynthesis
Lupeol	NM	259846	C08628	2.61E-03	4.98
Glycolysis/Gluconeogenesis
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
alpha-Linolenic acid metabolism
(−)-Jasmonic acid	HMDB0032797	5281166	C08491
Galactose metabolism
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
Cyanoamino acid metabolism
L-Phenylalanine	HMDB0000159	6140	C00079	5.99E-20	697.19
Biosynthesis of various plant secondary metabolites
Scopoletin	HMDB0034344	5280460	C01752	2.22E-23	2,147.1
Tryptophan metabolism
L-Tryptophan	HMDB0000929	6305	C00078	1.02E-18	483.27
Arginine and proline metabolism
L-Arginine	HMDB0000517	6322	C00062	2.79E-09	35.26
Glycine, serine and threonine metabolism
L-Tryptophan	HMDB0000929	6305	C00078	1.02E-18	483.27
Pyrimidine metabolism
Cytidine	HMDB0000089	6175	C00475	1.13E-19	635.28
Amino sugar and nucleotide sugar metabolism
Glucose 1-phosphate	HMDB0001586	65533	C00103	8.40E-09	30.69
The energy-signaling SnRK1/AMPK liked pathway
Epigallocatechin gallate	HMDB0003153	65064	C09731	1.09E-09	39.57
Isoflavonoid biosynthesis
Genistin	HMDB0033988	5284639	C09126	2.12E-10	48.59
Monoterpenoid biosynthesis
Kahweol	HMDB0035602	442495	C09893	4.86E-15	174.3
Aminobenzoate degradation
Vanillic acid	HMDB0000484	8468	C06672	1.36E-03	5.64
Vanillin	HMDB0012308	1183	C00755	5.58E-06	13.28
Others
Abscisic acid	HMDB0036093	5375200	NM	2.46E-13	108.29
Caffeic acid	HMDB0001964	689043	C01481	2.84E-16	245.85
Diosmetin	HMDB0029676	5281612	C10038	3.62E-04	7.1
Glycyl-L-leucine	HMDB0000759	92843	C02155	1.23E-16	271.54
Oleamide	HMDB0002117	5283387	C19670
Valylproline	HMDB0029135	9837272	NM	3.07E-05	10.39

HMDB, Human Metabolome Database; KEGG, Kyoto Encyclopedia of Genes and Genomes; FDR, false discovery rate; NM, not mentioned.

Fig. 5.

Overview of the pathway enrichment analysis of the identified metabolites showing the percentages for key metabolic pathways (directly illustrating the different functions of the metabolites). The x-axis shows the number of identified metabolites in each pathway.

Metabolite clusters after SLCMV inoculation by viruliferous whiteflies

Heatmap analysis of the 67 metabolites revealed six distinct metabolite clusters in the three cassava cultivars at 1, 3, and 7 dai (Fig. 6). Cluster 1 and 2 involved metabolites with high accumulation in all three cultivars (with preferential accumulation in KU 50). Cluster 3 involved metabolites with high accumulation in KU 50, but not in TMEB 419. Cluster 4 involved metabolites with high accumulation in R 11, but not in TMEB 419. Cluster 5 involved metabolites with high accumulation in both R 11 and TMEB 419. Lastly, cluster 6 involved metabolites with high accumulation specifically in TMEB 419.

Fig. 6.

Heatmap analysis of all 67 metabolites in three cassava cultivars (TMEB 419, KU 50, and R 11) at 1, 3, and 7 days after Sri Lankan cassava mosaic virus (SLCMV) inoculation (dai) via viruliferous whitefly infestation, identified using ultra-high-performance liquid chromatography high-resolution mass spectrometry analysis. The dashed borders indicate six clusters of metabolites with different accumulation patterns, based on peak intensity detection and t-tests/one-way ANOVA analyzed using MetaboAnalyst 6.0.

Despite some minor exceptions, these clusters clearly differentiated the metabolic responses among the cultivars. The functional annotations of these metabolites (Table 1) provide context for their accumulation patterns and offer insights into the plant metabolic responses to SLCMV inoculation after exposure to viruliferous whitefly infestation.

For example, in cluster 3, metabolites such as rutin, quercetin, quercitrin, hyperoside, tangeritin, and isorhamnetin (intermediates in flavone and flavonol biosynthesis) exhibited low accumulation in TMEB 419 and R 11, but high accumulation in KU 50 across all three time points; this was similar to glucose 1-phosphate, which was in cluster 1 and exhibited preferential accumulation in KU 50. Conversely, chlorogenic acid (cluster 1) exhibited low accumulation in R 11, while kaempferol (cluster 4), which is part of the flavone and flavonol biosynthesis pathway, exhibited low accumulation in TMEB 419.

Interestingly, in cluster 4, tryptophan (which is involved in phenylalanine, tyrosine, tryptophan, indole alkaloid, and glucosinolate biosynthesis, and glycine, serine, and threonine metabolism) mostly exhibited decreasing accumulation from 1 to 7 dai in KU 50 and TMEB 419, while it exhibited increasing accumulation from 1 to 7 dai in R 11. Distinct trends were observed for L-phenylalanine, which exhibited upregulation from 1 to 3 dai in KU 50. These metabolites likely reflect the differences in metabolic responses among the resistant, tolerant, and susceptible cultivars after SLCMV inoculation via viruliferous whitefly infestation.

In cluster 6, there was unique accumulation in TMEB 419 of 5 metabolites (kahweol, 14-deoxy-11,12-didehydroandrographolide, cytidine, lupeol, linolenic acid ethyl ester, and abscisic acid [also known as ABA]), which were solely upregulated in TMEB 419, while being downregulated in KU 50 and R 11.

Differential metabolites from 1 to 3 dai and from 3 to 7 dai

The volcano plot analysis of the 67 metabolites, based on log₂(fold change) and P-value, identified significantly differential metabolites from 1 to 3 dai and from 3 to 7 dai for each cultivar, as visualized in Venn diagrams (Table 2, Figs. 7 and 8, Supplementary Table 3).

Table 2.

Quantitative log2(fold change) and P-values in TMEB 419, KU 50, and R 11 from 1 to 3 dai and from 3 to 7 dai

Metabolites name	TMEB 419			KU 50			R 11
	Log2fold of 1 to 3 dai	log2fold of 3 to 7 dai	P-value	Log2fold of 1 to 3 dai	Log2fold of 3 to 7 dai	P-value	Log2fold of 1 to 3 dai	Log2fold of 3 to 7 dai	P-value
(−)-Caryophyllene oxide	−0.12	−0.33	0.0E+00	−1.02	−0.14	0.0E+00	0.36	−1.15	1.4E-03
(−)-Epigallocatechin	−0.75	0.36	1.6E-03	0.28	0.14	2.3E-02	0.35	−0.88	2.1E-05
(−)-Epigallocatechin gallate	0.11	−0.88	1.3E-04	0.45	−1.01	1.5E-03	−0.03	−1.20	2.1E-05
1-(1H-Benzo[d]imidazol-2-yl) ethan-1-ol	0.15	0.09	0.0E+00	−0.18	1.43	2.4E-02	0.45	0.06	2.7E-05
12-oxo Phytodienoic Acid	−0.32	0.68	3.9E-04	0.63	0.82	3.1E-07	0.76	−0.59	4.3E-06
14-Deoxy-11,12-didehydroandrographolide	0.65	1.12	9.3E-07	−1.71	0.48	0.0E+00	−0.22	−0.28	0.0E+00
19-Norandrostenedione	0.47	1.15	2.6E-03	−2.41	1.63	0.0E+00	−0.21	0.43	2.6E-02
2-(2,6-dimethoxyphenyl)-5,6-dimethoxy-4H-chromen-4-one	0.09	−0.12	0.0E+00	8.82	0.02	2.5E-08	0.03	−5.03	4.0E-10
2-(2-amino-3-methylbutanamido)-3-phenylpropanoic acid	0.07	0.80	4.8E-05	0.36	0.89	9.0E-06	0.63	−0.56	5.7E-07
2-Amino-1,3,4-octadecanetriol	−0.99	0.08	1.7E-04	−0.22	−1.37	1.6E-04	−0.30	0.56	1.0E-02
4-Chloro-L-Phenylalanine	0.14	−0.07	1.2E-02	0.21	−0.31	0.0E+00	−0.07	0.17	1.7E-03
7-Hydroxycoumarine	0.69	−0.42	3.6E-05	0.23	0.12	0.0E+00	−0.73	2.01	5.3E-07
7-{[(2S,3R,4S,5S,6R)-6-((2S,3R,4R)-3,4-dihydroxy-4-(hydroxymethyl) oxolan-2-yl]oxy}methyl)-3,4,5-trihydroxyoxan-2-yl]oxy}-5-hydroxy-3-(4-hydroxyphenyl)-6-methoxy-4H-chromen-4-one	0.02	0.06	0.0E+00	0.12	0.05	0.0E+00	0.53	−0.59	8.7E-05
8-{3-Oxo-2-[(2E)-2-penten-1-yl]-1-cyclopenten-1-yl} octanoic acid	−0.68	0.80	1.4E-04	−0.37	0.49	0.0E+00	−0.10	0.51	4.9E-03
9(Z),11(E),13(E)-Octadecatrienoic Acid methyl ester	0.66	0.12	6.8E-04	0.21	−0.47	2.2E-03	0.43	−0.33	1.6E-02
9-Oxo-10(E),12(E)-octadecadienoic acid	−0.49	0.56	6.9E-04	0.07	0.72	3.6E-05	0.60	−0.24	4.7E-04
9S,13R-12-Oxophytodienoic acid	−0.83	1.12	1.4E-04	−0.17	0.78	1.4E-04	−0.24	0.32	3.7E-02
Abscisic acid	0.06	0.46	1.0E-06	0.08	−0.11	0.0E+00	−0.27	0.34	5.8E-03
Adenosine	−0.17	−0.16	0.0E+00	−0.63	0.79	0.0E+00	−1.45	0.86	9.5E-05
All trans retinal	1.22	−0.96	0.0E+00	−3.13	−0.77	3.7E-06	3.97	−0.72	7.0E-08
Apigenin	0.33	−0.08	7.3E-05	−0.27	−0.15	0.0E+00	0.27	−0.13	1.9E-02
Apocynin	−0.18	0.40	0.0E+00	−0.70	0.10	0.0E+00	−0.22	0.46	0.0E+00
Caffeic acid	0.89	−0.48	3.6E-04	0.97	−0.41	6.5E-05	−0.84	2.18	1.5E-06
Chlorogenic acid	0.48	−0.62	6.3E-06	1.09	−1.39	3.5E-06	−0.16	0.21	0.0E+00
Cytidine	0.50	0.69	1.9E-07	−0.18	0.38	4.9E-03	0.42	−0.20	7.6E-05
Diosmetin	0.83	0.72	1.6E-03	0.27	0.31	0.0E+00	0.50	−0.15	0.0E+00
Eicosapentaenoic acid	−0.26	0.22	0.0E+00	−1.59	1.35	0.0E+00	−0.18	−0.02	0.0E+00
Epicatechin	−0.27	0.22	6.9E-03	0.31	−0.21	2.3E-02	1.24	−1.21	4.1E-06
Esculetin	0.53	0.00	6.5E-04	0.95	−0.38	1.3E-05	−0.24	0.46	2.2E-04
Genistin	−0.19	−0.30	3.2E-02	0.21	0.78	0.0E+00	1.09	−1.14	1.8E-06
Glucose 1-phosphate	−0.37	0.29	0.0E+00	0.19	−0.84	0.0E+00	−0.49	0.57	1.3E-02
Glycyl-L-leucine	−0.12	0.42	6.7E-04	0.49	0.68	2.0E-04	0.08	−0.05	2.1E-02
Guanine	0.02	0.20	1.2E-02	−0.03	0.77	6.1E-05	0.48	−0.09	5.3E-06
Guanosine	−0.03	0.22	3.3E-02	−0.05	0.75	4.2E-04	0.59	−0.23	3.4E-06
Hyperoside	−0.33	0.16	3.2E-04	0.31	0.43	1.3E-03	0.45	−1.06	3.0E-07
Isorhamnetin	0.00	0.21	8.7E-03	0.44	0.16	1.1E-03	0.52	−0.58	3.1E-06
Jasmonic acid	−0.23	0.38	0.0E+00	−1.56	1.50	0.0E+00	−0.16	0.18	3.4E-04
Kaempferol	−0.27	−0.07	2.6E-05	0.26	0.11	4.7E-03	0.53	−1.03	5.6E-08
Kaempferol-3-O-rutinoside (purifed)	−0.04	−0.18	4.9E-03	0.25	−0.09	1.1E-03	0.11	0.00	2.9E-02
Kahweol	0.70	1.21	7.1E-06	−1.18	1.28	0.0E+00	−0.41	−0.13	2.3E-02
L-Arginine	1.07	−0.12	2.1E-02	−0.21	1.41	6.6E-03	0.42	0.40	2.6E-05
L-Phenylalanine	1.09	−0.23	8.9E-09	0.82	0.37	7.1E-07	0.11	0.49	2.1E-08
L-Tyrosine	0.12	0.17	1.7E-03	0.05	0.42	5.2E-04	0.20	0.27	7.0E-05
Linolenic acid ethyl ester	−0.52	1.14	3.2E-03	−1.67	0.32	0.0E+00	−2.51	1.78	1.6E-03
Lupeol	0.26	−0.07	0.0E+00	0.23	−0.24	0.0E+00	−0.04	0.18	0.0E+00
Methyl 2-[(2-methoxy-2-oxoethyl) amino]acetate	0.04	0.23	6.0E-04	0.16	−0.38	4.2E-05	−0.60	0.64	2.9E-06
Methyl 3,5-di-tert-butyl-4-hydroxybenzoate	−1.17	0.20	3.8E-04	−0.71	1.08	0.0E+00	0.36	0.08	2.0E-04
Myricetin	−0.55	0.07	3.4E-05	−0.17	0.34	0.0E+00	0.05	−0.98	1.1E-07
Naringenin	0.17	−0.98	2.0E-07	0.52	−0.41	3.5E-03	1.13	−1.83	7.0E-07
Nootkatone	−0.31	0.36	0.0E+00	−1.57	1.32	0.0E+00	−0.20	0.17	0.0E+00
Oleamide	−0.05	0.25	0.0E+00	−1.54	1.38	0.0E+00	−0.16	0.07	0.0E+00
p-coumaric acid	0.30	−0.03	4.6E-04	−0.11	0.19	7.1E-03	−0.08	−0.78	4.8E-05
Polygodial	−0.29	0.39	0.0E+00	−0.73	1.09	0.0E+00	−0.17	0.15	7.6E-03
Quercetin	−0.28	0.22	0.0E+00	0.26	0.38	1.8E-04	0.27	−0.76	2.0E-05
Quercitrin	0.03	0.08	0.0E+00	−0.04	0.20	0.0E+00	0.20	−0.08	2.7E-02
Rutin	1.00	−0.58	2.7E-03	0.15	−0.01	0.0E+00	0.53	0.11	7.6E-06
Scopoletin	−0.99	1.89	5.5E-08	−2.40	1.13	1.7E-08	−2.54	3.11	2.2E-11
Tangeritin	0.08	−0.31	0.0E+00	9.48	−0.29	5.4E-09	0.39	−4.80	2.5E-09
Taxifolin	0.07	−0.32	0.0E+00	0.29	−0.33	0.0E+00	0.26	−0.58	1.5E-05
Tran-ferulic acid	0.32	0.26	2.1E-02	0.28	−0.57	2.0E-02	−0.28	0.91	4.8E-04
Tricin	−0.64	0.22	4.1E-05	−0.30	0.36	0.0E+00	0.54	−1.14	1.9E-06
Tryptophan	0.89	−0.48	9.0E-08	0.82	0.02	2.8E-05	0.37	−0.06	1.1E-06
Valylproline	−0.38	−0.14	0.0E+00	0.34	0.44	0.0E+00	−0.02	0.72	0.0E+00
Vanillic acid	0.16	−0.98	3.4E-04	−0.77	0.15	0.0E+00	0.77	−1.89	1.5E-04
Vanillin	−1.34	0.41	3.9E-03	−0.33	−0.51	8.8E-03	0.17	0.27	0.0E+00
Xanthurenic acid	0.13	0.02	1.4E-02	−0.27	0.18	2.7E-02	0.20	−0.35	0.0E+00
β-Asarone	−0.51	0.27	2.1E-03	−1.81	1.26	0.0E+00	−0.08	−0.70	6.8E-05

dai, days after inoculation.

Fig. 7.

Volcano plots of significantly differential metabolites in TMEB 419 from 1 to 3 (A) and 3 to 7 dai (B), KU 50 from 1 to 3 (C) and 3 to 7 (D) dai, and R 11 from 1 to 3 (E) and 3 to 7 (F) dai. The x-axis shows the log₂(fold change) of the metabolites, while the y-axis shows the P-value. Red points represent upregulated metabolites [log₂(fold change) ≥ 1, P > 2], while blue points represent downregulated metabolites [log₂(fold change) ≤ −1, P > 2].

Fig. 8.

Venn diagrams of upregulated (A) and downregulated (B) metabolites between TMEB 419, KU 50, and R 11 from 1 to 3 days after inoculation (dai) and from 3 to 7 dai.

Distinct upregulated metabolites from 1 to 3 dai were observed in each cultivar. In TMEB 419, two metabolites (L-arginine and L-phenylalanine) were upregulated. In KU 50, three metabolites [2-(2,6-dimethoxyphenyl)-5,6-dimethoxy-4H-chromen-4-one, chlorogenic acid, and tangeritin] were upregulated. In R 11, four metabolites (all-trans retinal, epicatechin, genistin, and naringenin) were upregulated (Figs. 7A, C, E, and 8A) (Table 2).

Additionally, distinct downregulated metabolites from 1 to 3 dai were observed in each cultivar. In TMEB 419, two metabolites (methyl 3,5-di-tert-butyl-4-hydroxybenzoate and vanillin) were downregulated. In KU 50, 12 metabolites [(−)-caryophyllene oxide, 14-deoxy-11,12-didehydroandrographolide, 19-norandrostenedione, eicosapentaenoic acid, jasmonic acid (JA), kahweol, linolenic acid ethyl ester, nootkatone, oleamide, scopoletin, and β-asarone] were downregulated. In R 11, three metabolites (adenosine, linolenic acid ethyl ester, and scopoletin) were downregulated (Table 2, Figs. 7A, C, E, and 8A).

Distinct upregulated metabolites from 3 to 7 dai were observed in each cultivar. In TMEB 419, six metabolites (14-deoxy-11,12-didehydroandrographolide, 19-norandrostenedione, 9S,13R-12-oxophytodienoic acid, kahweol, linolenic acid ethyl ester, and scopoletin) were upregulated. In KU 50, 11 metabolites (19-norandrostenedione, eicosapentaenoic acid, JA, kahweol, L-arginine, methyl 3,5-di-tert-butyl-4-hydroxybenzoate, nootkatone, oleamide, polygodial, scopoletin, and β-asarone) were upregulated. In R 11, four metabolites (7-hydroxycoumarin(e), caffeic acid, linolenic acid ethyl ester, and scopoletin) were upregulated (Table 2, Figs. 7A, C, E, and 8A).

Additionally, distinct downregulated metabolites from 3 to 7 dai were observed in KU 50 and R 11, though not in TMEB 419 (i.e., no metabolites were downregulated in TMEB 419 from 3 to 7 dai). In KU 50, three metabolites [(−)-epigallocatechin gallate, 2-amino-1,3,4-octadecanetriol, and chlorogenic acid] were downregulated. In R 11, 11 metabolites [(−)-caryophyllene oxide, (−)-epigallocatechin gallate, 2-(2,6-dimethoxyphenyl)-5,6-dimethoxy-4H-chromen-4-one, epicatechin, genistin, hyperoside, kaempferol, naringenin, tangeritin, tricin, and vanillic acid] were downregulated.

Changes in metabolite regulation between 1–3 and 3–7 dai

Several metabolites exhibited temporal changes in their regulation between early infection (1 to 3 dai) and late infection (3 to 7 dai).

Three metabolites (L-arginine, 19-norandrostenedione, and kahweol) were upregulated in the late stage in KU 50. Meanwhile only 19-norandrostenedione and kahweol were upregulated in the late stage in TMEB 419, while L-arginine was downregulated in late infection but upregulated in the early stage.

In KU 50 and R 11, four metabolites [(−)-caryophyllene oxide, linolenic acid ethyl ester, scopoletin, and (−)-epigallocatechin gallate] were downregulated in early infection but were absent or upregulated in late infection. Additionally, in KU 50, seven metabolites (19-norandrostenedione, eicosapentaenoic acid, JA, kahweol, nootkatone, oleamide, scopoletin, and β-asarone) were similarly downregulated in early infection but upregulated in late infection. Moreover, only one metabolite (chlorogenic acid) was upregulated in early infection but downregulated in late infection.

In R 11, two metabolites (linolenic acid ethyl ester and scopoletin) were upregulated in late infection but downregulated in early infection. Additionally, in R 11, three metabolites (epicatechin, genistin, and naringenin) were upregulated in early infection but downregulated in late infection.

Pathway enrichment analysis and biological implications

The changes in metabolites of cassava plants after SLCMV inoculation via viruliferous whitefly infestation were associated with several plant defense and secondary metabolite biosynthesis pathways (Table 1). The following insights were observed. First, scopoletin was consistently upregulated in all cultivars during late infection (from 3 to 7 dai), indicating its role in secondary metabolite biosynthesis across cultivars. Second, JA (involved in alpha-linolenic acid metabolism, which is critical for defense signaling) was upregulated in KU 50 (tolerant) during late infection (from 3 to 7 dai) (Table 1). Lastly, L-arginine (involved in the arginine and proline metabolism pathway) was upregulated in TMEB 419 (resistant) during early infection (from 1 to 3 dai) and in KU 50 (tolerant) during late infection (from 3 to 7 dai), suggesting its role in stress response and metabolic regulation in these cultivars.

Discussion

In this study, the associations of the pathogen–insect vector and the host plants were explored using metabolomics analysis and integration of biochemical networks. Our approach allowed us to identify distinct metabolite profiles associated with resistance, tolerance, and susceptibility in cassava cultivars under similar infection conditions. This comparison provided an opportunity to pinpoint metabolites/metabolic pathways linked to phenotypic characteristics related to differential plant responses to SLCMV inoculation.

The CMD2 resistance locus, located on chromosome 12, serves as a key genetic marker distinguishing the cassava genotypes among the selected cultivars. TMEB 419, developed by the International Institute of Tropical Agriculture (IITA) through Africa’s cassava breeding program, is recognized for its dominant CMD2-mediated resistance (Akano et al., 2002; Rabbi et al., 2014; Thuy et al., 2021; Wolfe et al., 2016). In contrast, KU 50, bred in Thailand and widely cultivated across SEA, is characterized by mild to minimal CMD symptoms in the primary growing season, along with high yields and starch content. Meanwhile, R 11, derived from a similar breeding background, demonstrates greater CMD susceptibility, showing severe symptoms and higher disease severity (Kongsil et al., 2024; Malik et al., 2022; Vannatim et al., 2025).

Our results highlight differences in viral accumulation that correlate with the observed phenotypic variations among cassava cultivars. The qPCR amplification clearly showed distinct viral load patterns between 1 to 7 dai for tolerant and susceptible cultivars, while the resistant cultivar, TMEB 419, maintained consistently low viral copy numbers throughout this period. In contrast, both KU 50 and R 11 exhibited increasing SLCMV titers, with R 11 showing the highest viral load increase. These differential viral populations are likely to the observed phenotypic responses. In the resistant cultivar, TMEB 419, a fundamental barrier or partial mechanism appears to effectively restrict viral replication, leading to an asymptomatic or only mildly symptomatic state. Conversely, while tolerant and susceptible cultivars permit viral replication, their distinct phenotypes (e.g., symptom reversion or recovery) may be governed by different post-infection mechanisms.

Seven days after SLCMV infection, in both tolerance and susceptible cassava cultivars, appears to represent a critical time point for viral replication and accumulation. This observation is supported by research on other geminiviruses. For instance, a study by Rodríguez-Negrete et al. (2014) on tomato yellow leaf curl virus (TYLCV) and tomato yellow leaf curl Sardinia virus (TYLCSV) in tomato plants noted that viral loads increased until at 15 dpi. Similarly, in Nicotiana benthamiana, TYLCV and TYLCSV reached initial maximum levels at 7 dpi before increasing again at 42 dpi. Further support comes from a study by Rentería-Canett et al. (2011), which investigated the co-infection of pepper huasteco yellow vein virus (PHYVV) and pepper golden mosaic virus (PepGMV) in pepper plants. These viruses, also transmitted by whiteflies, caused the first visible symptoms to emerge at 7 to 14 dpi. The research found that the viral particle, quantified by real-time PCR in leaf tissues, showed high viral loads at 7 dpi. Therefore, the 7-day post-inoculation time point is a crucial period for evaluating SLCMV viral load in cassava. This can serve as a foundational measurement for future studies, particularly those involving whitefly-mediated virus transmission in cassava.

This study used viruliferous whiteflies, i.e., B. tabaci Asia II-1 (confirmed by PCR-based detection of mtCO1), which is the same species reported to transmit SLCMV in Thailand (Saokham et al., 2021). This vector’s efficiency at transmitting SLCMV highlights its critical role in virus spread in cassava fields. Further studies are necessary to clarify the exact influence of B. tabaci Asia II-1 on SLCMV spread and severity, which may be useful for developing new insect management strategies in SEA.

The hierarchical clustering and 2D PCA score plot analyses provided insights into the metabolic changes in cassava cultivars with different resistance/tolerance/susceptibility statuses regarding SLCMV infection transmitted by viruliferous whiteflies. Specifically, the clustering results showed that the resistant and tolerant cultivars at 7 dai formed a distinct group, clearly separated from the susceptible and tolerant cultivars at earlier infection stages (1 and 3 dai). This clustering suggests that, by the late infection stage, the resistant and tolerant cultivars exhibit convergent metabolic profiles, potentially indicating a shared defense mechanism that becomes activated over time.

The biosynthesis of flavonoids, flavone, and flavanol was found to be a major pathway that the identified metabolites were involved in. Several metabolites exhibited fluctuating concentrations over time and upregulation and downregulation of the related pathways over time. Flavonoid production influences insect vectors’ feeding behavior, subsequently restricting their growth and development (Luan et al., 2013; Yao et al., 2019). Accordingly, our observation of high metabolite accumulation patterned with indicates inhibition of the viruliferous whiteflies’ feeding behavior on cassava plants. In addition, the upregulated biosynthesis of flavonoids, flavanols, and flavanones plays a crucial role in plant resistance to bacterial and fungal pathogens. This is achieved through ROS generation in response to pathogen infection, which decreases damage to the plant cells (Dai et al., 1996; Mierziak et al., 2014; Ramaroson et al., 2022). Consequently, the increased levels of the flavonoids, flavanols, and flavanones and the upregulation of the biosynthesis pathway observed in this study show a trend with contribute to both SLCMV resistance and plant responses against viruliferous whitefly infestation.

The heatmap analysis revealed that glucose 1-phosphate (a metabolite categorized under glycerolipid metabolism) exhibited uniquely high accumulation in KU 50 (tolerant) at 1, 3, and 7 dai. Changes in the glycerolipid pathway alters fatty acid dynamics, affecting both the composition and molar ratios of glycerolipid species (Shen et al., 2010). Farahbakhsh et al. (2019) found that lipid-related metabolite pathways can be utilized by plants to resist virus infection or, conversely, they can be exploited by viruses. This is supported by Zhang et al. (2016), who demonstrated an association between lipid enrichment and the recruitment of essential components for virus replication (including RNA-dependent RNA polymerase). Thus, we propose that, during SLCMV infection of KU 50 (tolerant), SLCMV perhaps correlative with glycerolipid metabolism in the host plant to facilitate viral replication and infection.

The plant hormones JA and ABA were both detected in this study. JA was uniquely highly accumulated in KU 50 (tolerant) at 3 dai, while ABA was particularly highly accumulated in TMEB 419 (resistant) at 1, 3, and 7 dai. These plant hormones are known to play significant roles in various plant defense pathways. JA contributes to host defense mechanisms by modulating plant secondary compounds and altering physical and chemical defenses, enhancing resistance to biotics stress. It also enhances plant resilience against herbivorous insects (Lou et al., 2005; Okada et al., 2015). JA facilitates defense against cucumber green mottle mosaic virus infection in bottle gourd by regulating alpha-linolenic acid metabolism (Li et al., 2023). Biomechanical defense mechanisms are also reflected in changes in alpha-linolenic acid metabolism, which directly depends on JA signaling. The bacterial pathogen Pseudomonas syringae secretes coronatine, a natural phytotoxin (Collemare et al., 2019; Win et al., 2012), which binds to the JA receptor known as coronatine-insensitive 1 (COI1), increasing the interaction of COI1 with numerous JASMONATE ZIM DOMAIN (JAZ) family proteins and ultimately disrupting plant metabolic homeostasis (Fig. 9) (Geng et al., 2012; Katsir et al., 2008; Nomura et al., 2005). ABA, which is produced after the cleavage of carotenoid precursors, regulates plant responses to both abiotic and biotic stresses, such as salt stress and biotic stress including root knot nematode (Meloidogyne graminicola) infection in rice (Oryza sativa) (Kyndt et al., 2017; Ruiz-Sola et al., 2014). Thus, this study elucidates the potential roles of JA and ABA in driving metabolic pathways associated with SLCMV resistance/tolerance. These findings align with the results of Malichan et al. (2023), which emphasized the involvement of JA and ABA in specific plant defense mechanisms.

Fig. 9.

Pathway analysis of significantly differential metabolites (both primary/intermediary metabolites and signaling hormones) implicated in the differential phenotypic cassava responses at 1, 3, and 7 days after Sri Lankan cassava mosaic virus (SLCMV) inoculation (dai) via viruliferous whitefly infestation. Different block colors indicate different metabolic pathways: red block represents tryptophan metabolism, green block represents phenylalanine metabolism (while green arrows indicate related genes and involved activities), blue block represents phenylpropanoid biosynthesis, yellow block represents flavonoid, flavone, and flavanol biosynthesis, separated block mentions alpha-linolenic acid metabolism (while green arrows indicate related genes and involved activities), and violet block and arrows represent the impact of each metabolic/biosynthesis pathway. This figure is modified from the Kyoto Encyclopedia of Genes and Genomes database-related figures in papers by Peluffo et al. (2010), Li et al. (2023), Mano and Nemoto (2012), Sasaki-Sekimoto et al. (2013), Sasaki-Sekimoto et al. (2013), Wagner et al. (2012), and Dehghan et al. (2014). HCT, p-hydroxycinnamoyl-CoA shikimate hydroxycinnamoyl transferase; C3H, p-coumarate 3-hydroxylase; CCoAOMT, caffeoyl-CoA O-methyl transferase; PEP, phosphoenolpyruvate; PAL, phenylalanine ammonia lyase; UGCT, UDP-glucose:glycoprotein glucosyltransferase; C4H, cinnamic acid 4-hydroxylase; 4CL, 4-coumarate; CHS, chalcone synthase; CHI, chalcone isomerase; FNSI, flavone synthase I; F3H, flavanone 3-hydroxylase; LAR, leucoanthocyanidin reductase; FLS, flavonol synthase; CPY450, cytochrome P450 monooxygenase; NPR1, NONEXPRESSOR OF PATHOGENESIS-RELATED GENES 1; PR, PATHOGENESIS-RELATED; COI1, coronatine-insensitive 1; JAZ, JASMONATE ZIM DOMAIN.

The JA pathway regulates the production of secondary metabolites and defensive proteins in plants (Pieterse and Dicke, 2007). The elevated JA levels in KU 50 at 3 dai, following whitefly infestation, suggest a role for JA in its tolerant phenotype. JA signaling is mediated by MYC2, which is suppressed by JAZ proteins, thereby inhibiting JA downstream responses (Song et al., 2022). Wang et al. (2012) highlighted the interplay between MYC2-JA signaling and whitefly infestation, revealing that whitefly salivary gland secretions modulate plant defenses. Transcriptome analysis identified highly expressed genes in the salivary glands, some linked to whitefly feeding behavior and plant resistance mechanisms. Their findings support a salicylic acid (SA)-JA crosstalk model, where early whitefly infestation activates SA signaling, enhancing whitefly performance while suppressing JA responses through MYC2 silencing (Li et al., 2014; Luan et al., 2014; Zhang et al., 2012, 2013).

Geminiviruses may exploit this mechanism by interfering with MYC2, thereby weakening JA-mediated defenses against whiteflies. The increased JA levels in KU 50 could contribute or link to its tolerance to SLCMV, whereas the reduced JA accumulation in R 11 may be accompanied by its susceptibility.

Certain metabolites, such as p-coumaric acid, have been found to influence cell wall components. This metabolite is associated with phenylpropanoid biosynthesis and ubiquinone and other terpenoid-quinone biosynthesis pathways, and it enhances lignin content in Arabidopsis seedlings (Buanafina and Fescemyer, 2012; Hatfield et al., 2016; Kemat et al., 2020). We propose that p-coumaric acid may contribute to cell wall reinforcement in TMEB 419 (resistant) and KU 50 (tolerant), as indicated by its high accumulation during early infection (1 and 3 dai). In contrast, in R 11 (susceptible), p-coumaric acid exhibited high accumulation only during late infection (7 dai). This pattern suggests that p-coumaric acid may play a critical role in the early defense mechanism against SLCMV infection after exposure to viruliferous whitefly infestation in TMEB 419 and KU 50. Its delayed accumulation in R 11 may reflect a weaker or slower defense response, underscoring the metabolic and phenotypic differences among these cultivars.

Tryptophan metabolism, represented by the metabolite L-tryptophan, exhibited distinct patterns in the three cultivars. In R 11, L-tryptophan was consistently highly accumulated at the three time points. In KU 50, it gradually declined over time, while in TMEB 419, it was entirely absent. L-tryptophan plays a key role in the biosynthesis of auxin (indole-3-acetic acid) and indoleamines (e.g., melatonin and serotonin), which are essential for plant development and defense against biotic and abiotic stressors (Erland and Saxena, 2019; Erland et al., 2019). The consistently high accumulation of L-tryptophan in R 11 (susceptible) indicates that it facilitates SLCMV infection. This may occur because SLCMV hijacks host resources to support its replication, systemic movement, and virion production. Auxin, known for promoting cell elongation and division, may inadvertently assist in these processes (Ljung, 2013). In contrast, the reduced L-tryptophan in KU 50 (tolerant) and its absence in TMEB 419 (resistant) suggest the correlation with defensive adaptations. These cultivars are consistent with suppressing tryptophan metabolism to limit the virus’s ability to exploit cellular processes. This highlights a potential regulatory mechanism in KU 50 (tolerant) that tends to help the plant adapt and tolerate the virus after exposure via viruliferous whitefly infestation.

Basal amino acids and their derivatives are associated with increased nucleotide production during SLCMV infection, suggesting heightened RNA synthesis. Relatedly, Wang et al. (2012) demonstrated the role of amino acids in whitefly-mediated transmission of Tomato yellow leaf curl China virus (TYLCCNV) in tobacco. Their study found that TYLCCNV-infected plants had lower amino acid levels compared to uninfected plants, allowing whiteflies to achieve more balanced nutrient absorption during feeding. Wei et al. (2019) and Zhang et al. (2022) studied Cucurbit chlorotic yellows virus (CCYV), a single-stranded RNA virus with a bipartite genome, and found increased nucleotide and lipid accumulation, which coincided with rising viral titers. This suggests that amino acids and related metabolic pathways are essential for CCYV assembly and replication, including facilitating CCYV transmission. In this study, amino acid-related pathways such as purine metabolism (e.g., adenosine, guanine, and guanosine) and arginine biosynthesis (L-arginine) exhibited high accumulation, particularly in KU 50. These metabolites showed a distinct trend: they were highly accumulated during early infection (1 dai) but either decreased or disappeared by late infection (7 dai). This pattern perhaps suggests that KU 50 (tolerant) undergoes a significant adaptive response over time, limiting the virus’s ability to exploit the amino acid-related pathways as the infection progresses. These findings articulate the temporal dynamics of tolerance mechanisms during SLCMV infection.

The changes in metabolites from 1 to 3 dai and from 3 to 7 dai, as visualized in the volcano plots, revealed significant differences among the cultivars. In TMEB 419 (resistant) during 1 to 3 dai, there was an increase in defense-related metabolites related to phenylalanine metabolism (e.g., L-phenylalanine) and arginine and proline metabolism (e.g., L-arginine), alongside a decrease in phenylpropanoid biosynthesis (e.g., vanillin) (Fig. 9). The contribution of phenylalanine, arginine, and proline metabolism to amino acid transduction indicates their importance in plant disease resistance. Specifically, phenylalanine metabolism is associated with the induction of SA, a hormone crucial for activating defense mechanisms through NONEXPRESSOR OF PATHOGENESIS-RELATED GENES 1 (NPR1) and PATHOGENESIS-RELATED (PR) genes (Backer et al., 2019; Li et al., 2023). Meanwhile, the contribution of arginine and proline metabolism to amino acid production has been linked to biotic and abiotic stresses (Mhlongo et al., 2020; Ting et al., 2020). Pérez-Clemente et al. (2019) observed elevated amino acids during citrus tristeza virus (CTV) infection. Similarly, Zhang et al. (2022) reported significant increases in amino acids in CCYV-infected cucumber plants.

Our heatmap analysis (Fig. 6) synthesized L-phenylalanine as a key metabolite in SA biosynthesis, activated via phenylalanine ammonia-lyase (PAL). PAL, a primary enzyme in phenylalanine metabolism (Fig. 9), showed distinct modifications in KU 50 (elevated at 1 dai) and R 11 (increased at 1 and 3 dai) after whitefly infestation. This metabolic shift promotes SA signaling, systemic resistance, and phytoalexin synthesis, essential for pathogen defense (Arbona and Gómez, 2016; Bolwell et al, 1986; Gurikar et al., 2022; Pedras et al., 2008).

Elevated SA levels may restrict SLCMV replication, reinforcing plant defense in TMEB 419 (resistant), KU 50 (tolerant), and R 11 (susceptible). Identifying these key secondary metabolites could enhance our understanding of begomovirus–whitefly–host interactions. Pathogen-infected plants often experience defense suppression, increasing their suitability for vector development (Zhang et al., 2012). This aligned with higher insect populations, greater pathogen acquisition, and enhanced transmission. Thus, SLCMV infection via whiteflies may influence vector multiplication and pathogen invasion dynamics, particularly in susceptible cultivars.

Metabolites that contribute to plant phenotypic differences can be controlled by multiple gene clusters. While these clusters may be dispersed across the genome, they are often coordinated through co-expression or regulation. The resulting divergence in metabolite profiles may involve post-transcriptional and translation regulation as well as post-translational modifications, including enzyme modulation. This contracts with gene transcript levels, which directly reflect transcriptional regulation or genes silencing.

For example, in the model plant A. thaliana (Marszalek-Zenczak et al., 2023), multiple gene clusters encode enzymes for structurally related triterpene metabolites. Some of these biosynthetic clusters are large and contain multiple enzyme-coding genes that are co-expressed in specialized metabolic pathways. This observation is reinforced by research on cassava (Drapal et al., 2019; Geng et al., 2017; Ma et al., 2023; Siriwan et al., 2023), which demonstrates that the abundance of functional enzymes, metabolites, and proteins does not solely rely on transcript-level expression data. These studies validated gene function using tools like qPCR, highlighting the importance of multiple regulatory layers. Similarly, a study on a yellow-green left mutant of Hami melon (Cucumic melo) by Han et al. (2023) showed that changes at the protein level matched shifts in metabolite profiles, which successfully explained the observed photosynthetic defects. In contrast, modifications at the transcript levels were insufficient to explain the phenotype. This underscores the crucial information that can be gained by integrating metabolite and gene analyses. We propose that metabolomics is a valuable tool for advocating and confirming the functional relevance of compounds and metabolites, which in turn contributes to pathway analysis. Our research could be expanded to include studies of the proteome and transcriptome, thereby providing valuable insights into the crosstalk among gene, protein, and metabolite regulation. Understanding these complex regulatory architectures can aid in synthetic biology, metabolic engineering, and crop improvement by targeting complex metabolite pathways.

This study identified distinct metabolite changes among resistant, tolerant, and susceptible cassava cultivars during different periods after SLCMV inoculation via viruliferous whitefly infestation. Some of these changes appear to enhance cassava fitness against whitefly infestation and delay SLCMV establishment in the resistant or tolerant cassava cultivars. Our findings provide valuable insights into cassava plants’ metabolic pathway responses to SLCMV during the early and late stages of infection. Importantly, this research contributes to the understanding of plant–virus interactions and offers critical information for breeding programs that aim to develop resistant cultivars and improve strategies for controlling plant viral diseases.

Electronic Supplementary Material

Supplementary materials are available at The Plant Pathology Journal website (http://www.ppjonline.org/).