The experiment was conducted at the Horticulture Research Centre, University of South Africa, located in Florida, South Africa. The plants were grown outdoors under summer conditions, with daytime temperatures ranging from 18°C to 28°C and nighttime temperatures between 15°C and 18°C. All plants were placed in mesh cages measuring 60 × 60 × 90 cm. Each mesh cage housed eight plants, which were replicates. Manual irrigation was applied as necessary throughout the study.

Two commercially relevant potato cultivars, namely, Sifra and Valor, with contrasting responses to P. syringae, as identified in preliminary screenings, were selected. Certified disease-free tubers were directly sown into 20-cm pots filled with sterilised potting soil. Thirty-five days after sowing, when the plants had developed four fully expanded compound leaves, they were inoculated with a cell suspension of the bacteria. Inoculation was done following the method outlined by Djami-Tchatchou et al. (2019a). In brief, the cell suspension was pressure-infiltrated into the abaxial surface of the leaves using a needleless syringe, with infiltration conducted at three sites per leaf across three leaves per plant. Only lower leaves were selected for inoculation, and control plants were mock-inoculated with the buffer. Successful inoculation was marked by an extensive permeation of the inoculum beneath the leaf epidermis.

The experiment included a total of 32 plants, with 16 plants assigned to each cultivar. Each cultivar group was further divided into control and treatment groups, resulting in four groups: control Sifra, treated Sifra, control Valor, and treated Valor. Each group consisted of eight plants. This design enabled a comparison of the differential responses of Sifra and Valor under pathogen stress conditions Figure 1.

Leaf samples were collected at 24, 48, and 72 hours after inoculation from three randomly selected plants per treatment group. The collected samples were immediately frozen at -80°C and kept in the freezer until use. The frozen leaf tissue was ground in liquid nitrogen using a mortar and pestle, and secondary metabolites were extracted with methanol.

Metabolite extraction was performed according to a modified method from Madala et al. (2014). 1g of each leaf sample was ground into powder using liquid nitrogen in a mortar and pestle. The powdered samples were then weighed into 2 mL centrifuge tubes, to which 1.5 mL of 80% ice-cold methanol was added. The samples were vortexed for 30 seconds and subsequently sonicated for 30 minutes to enhance cell disruption. After sonication, the samples were centrifuged at 13, 950 rpm for 5 minutes at 4°C. The resulting supernatant was filtered through 0.22 µm nylon filters into glass vials with 500 µL inserts (Agela Technologies, Tianjin, China). Additionally, the extract was transferred into new 2 mL Eppendorf tubes and stored at 4°C until analysis. Three biological replicates were prepared for each group to ensure reproducibility.

Metabolite profiling was performed using a Shimadzu 9030 liquid chromatography–quadrupole time-of-flight mass spectrometer (LC-MS qTOF). Data from LC-MS included retention times, which indicate the duration each metabolite took to pass through the chromatography column; mass spectra, which provide information on molecular weight and structure; and quantitative data on metabolite abundance, derived from peak intensities. These data were exported in mzML format and pre-processed using XCMS Online (version 2.7.2), which incorporates XCMS (version 1.47.3) and CAMERA (version 1.34.0). The pre-processing employed UPLC-qTOF parameters with the centWave feature detection method. The maximum allowed m/z deviation was set at 15 ppm, the signal-to-noise ratio at 6, the peak width range at 5–20 s, the m/z difference at 0.01, and the pre-filters for peak number and intensity at 3 and 100. Individual samples yielded between 969 and 9, 749 features per file across six genotypes under control and bacterial treatments. Retention time correction was performed using the obiwarp method with a profStep of 1, while peak alignment and grouping were conducted using the density method, with a bandwidth of 5, an m/z width of 0.015, a minimum fraction of 0.5 across all samples, and a minimum sample requirement of 1. Missing peak intensities were filled using the Fill Peaks function. After these pre-processing steps, a Kruskal–Wallis non-parametric statistical test was performed, accompanied by post-hoc analysis, with a significance threshold of p ≤ 0.01. Following statistical filtering, 603 features were retained. From the filtered features, 15–16 metabolites were confidently annotated per time point (24, 48, and 72 hours post-inoculation). The retained features were further annotated and identified using, GNPS, and Sirius 6.0.3, referencing databases such as HMDB, MassBank, and PubChem. Metabolites were identified based on accurate mass, fragmentation patterns, and elemental composition. The data were then normalized, log-transformed, and Pareto-scaled, with missing values estimated before multivariate analysis. Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA), and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) were conducted using SIMCA^® 17 software, with metabolites exhibiting Variable Importance in Projection (VIP) scores ≥1 considered significant. Additional data exploration included heatmap generation and pathway analysis using MetaboAnalyst, and the identified metabolites were mapped to biological functions using the Kyoto Encyclopedia of Genes and Genomes (KEGG).

Principal component analysis (PCA) was employed to investigate patterns in metabolite profiles of cultivars Sifra and Valor across various time points (24, 48, and 72 hours). PCA plots illustrate how samples cluster based on their overall metabolic composition, and in the present study, they emphasized variation attributed to cultivar differences and treatment duration.

Focusing only on the duration of treatment, the PCA plots revealed clear distinctions between the sampling times with 24, 48 and 72 hours samples forming three distinct clusters (Figures 2A, B). On close examination, it was evident that the metabolite profile captured at 48 hours was transitional between 24 hours and 72 hours since it was between these two end points. The separation is clearer on the three-dimensional PCA plot (Figure 2B) as each time point occupied a unique region in the principal component space.

On cultivar-specific metabolite profiles (Figures 2C, D), the PCA plots showed a chromatographic separation between the two genotypes across all time points. The two cultivars formed clusters, however, with overlap, indicating that genetic background is a significant factor influencing metabolic composition, albeit without a strong expression characterised by non-overlapping PCA clusters. The tight clustering within each cultivar group reflected consistent metabolic profiles within genotypes.

This consistent separation implies that each cultivar possesses a unique baseline metabolic profile and potentially distinct responses to treatment. These findings suggest that genotype-specific regulation of metabolic pathways is crucial in shaping the overall metabolic landscape.

The PCA results indicated that both treatment time and plant genotype have a significant impact on metabolic profiles. The principal components capture the main sources of metabolic variation, with both genotype and time playing a significant role. Time influences the progression of the metabolic response, while genotype defines both the baseline and potentially the specific response mechanisms. This analysis underscores the importance of considering both time-related and genetic factors when interpreting differences and changes in plant metabolism.

PLS-DA was further employed to investigate the metabolic response to treatment in the two potato genotypes, Valor and Sifra, at three time points, 24, 48, and 72 hours. This supervised method offers clearer group separation than PCA by incorporating information about treatment groups, allowing for a more focused view of how metabolite profiles differ in response to treatment over time (Figures 3A–C).

In both cultivars, P. syringae-treated samples and the control exhibited varied levels of separation from 24 hours through 72 hours. At 24 hours, the groups were already and most clearly distinct. The separation was especially greater in Valor, which indicates a strong and progressive shift in Valor’s metabolic profile following treatment. The results suggest that Valor undergoes extensive metabolic reprogramming in response to inoculation with P. syringae, reflecting a strongly pronounced metabolomic response within the first 24 hours of inoculation (Figure 3A).

In contrast, Sifra had less distinction between the P. syringae-treated and control groups, especially at 24 hours (Figure 3A). This limited separation suggests that the P. syringae had a much weaker impact on Sifra’s metabolite profile which carried over to the 48 and 72 hour time points (Figures 3B, C). The overall stability of Sifra’s metabolic profile across the time points may indicate that it is less responsive to the inoculation or that it has a more stable metabolic architecture under the tested conditions. Interestingly, the Valor samples are also less resolved at the 48 and 72 hour time points which may mean lessening of effects after the 24 hour time point.

Together, these results highlighted a clear genotype-dependent difference in metabolic responsiveness. Valor demonstrated a strong, time-dependent response to treatment, while Sifra exhibited a more stable and less altered profile. This suggests that Sifra may possess greater metabolic stability or tolerance to the treatment, potentially reflecting underlying genetic or physiological differences in stress perception or adaptation.

To enhance the distinction between treatment groups and eliminate unrelated orthogonal variation, OPLS-DA was applied at each time point. This method separates predictive variation linked to treatment from systematic variation caused by baseline cultivar differences, thereby facilitating the identification of metabolites responsive to pathogens. The OPLS-DA score plots (Supplementary Figure 1) showed minimal separation between control and pathogen-inoculated samples at 24 hours post-inoculation for Sifra, consistent with the limited early metabolic response noted in the PLS-DA analysis. At 48 hours, however, clear separation emerged in both cultivars, with Valor showing the most significant distinction between treatment groups along the predictive component. Sifra also exhibited separation at this time point, though it was less pronounced than Valor’s. The predictive component captured pathogen-induced metabolic variation, while the orthogonal component filtered out systematic cultivar-specific baseline differences. By 72 hours, the trend reversed, with Sifra demonstrating greater separation between control and pathogen-stressed samples, while Valor’s discrimination weakened significantly compared to the 48-hour mark, reflecting a temporal shift in defense strategies between the two cultivars.

Heatmap analysis was employed to examine metabolite changes in Sifra and Valor potato cultivars at 24, 48, and 72 hours post-inoculation to investigate how the different potato varieties respond to pathogen infection. The heatmap displayed metabolite levels on a colour scale ranging from 1 (yellow, downregulation) to +1 (navy black, upregulation). This analysis facilitated identifying how metabolite patterns shift in response to pathogen attack (Figures 4A–C) (Supplementary Tables 1–3). The heatmap differentiation was based on a few metabolites, 14 for 24 hours, 16 for 48 hours and 17 for the 72 hour interval. The metabolites selected contribute towards the heatmap had the highest VIP scores (Supplementary Tables 4A–C). Overall, and based on the selected metabolites, the P. syringae-induced differentiation was only evident at 24 and 48 hours post-inoculation.

To better understand the observed metabolite changes, pathway enrichment and topology analyses were conducted using MetaboAnalyst’s MetPA tool across all three time points (24 h, 48 h, and 72 h post-inoculation). This approach helped in placing the differentially accumulated metabolites (VIP > 1) within broader biological pathways, as defined in the KEGG database. The resulting enrichment scores and pathway impact values gave an insight into the timing and nature of the metabolic responses in each cultivar (Supplementary Figures 2 A–C, Supplementary Table 4A–C).

Sifra exhibits higher levels of alpha-solanine, quercetin 3-alloside, and lysophospholipids in both control and treated plants at 24 hours post-inoculation (hpi), in contrast to Valor’s lower baseline. This observation may suggest an evolutionary strategy where the plant remains in a constant state of readiness, rather than waiting for a specific trigger to initiate its defence (Züst and Agrawal, 2017). This idea is consistent with the concept of systemic acquired resistance described by Fu and Dong (2013), which posits that signals from a local infection can prepare the entire plant for potential future attacks. Additionally, previous research has shown that plants significantly reprogram their metabolism in response to pathogens (Bednarek, 2012).

Our metabolomics data illustrate that the duration and timing of this metabolic reprogramming are as crucial as its intensity for effective defence. As shown in our PCA and PLS-DA analyses (Figures 2, 3), these timing differences create distinct metabolic signatures that differentiate the two cultivars across all time points. In support of this, Kopczewski et al. (2022) found that cucumber plants infected with P. syringae exhibited metabolic changes not only at the infection site but throughout the entire plant, highlighting the importance of systemic metabolic responses in defence. Studies also emphasise that the timing of pathway induction is vital in determining resistance, particularly in long-term defence mechanisms involving phenylpropanoid and glycoalkaloid biosynthesis (Dong and Lin, 2021; Yadav et al., 2020; Liu et al., 2024).

By 24 hours post-inoculation (hpi), Sifra and Valor exhibit fundamentally different metabolic baselines. As clearly shown in the heatmap analysis (Figure 4A), Sifra displays constitutively elevated levels of defence metabolites, akin to wild crop relatives that maintain ongoing baseline defences (Züst and Agrawal, 2017), even in the absence of infection. In contrast, Valor shows minimal accumulation of pre-induction metabolites, indicating a reliance on pathogen-induced activation of its defence mechanisms (Ahuja et al., 2012; Balmer et al., 2013). The early activation of phenylalanine metabolism in Sifra (pathway impact: 0.423) is noteworthy. The pathway enrichment analysis (Supplementary Figure 2A) confirms that this pathway produces various defence compounds, including flavonoids, lignin, and antimicrobial phenolics (Dixon and Paiva, 1995; Maeda and Dudareva, 2012; Vogt, 2010). Recent advances in synthetic biology have shown that genetic engineering can enhance these pathways, suggesting that continuous activation of these defences could improve crop resistance. In experiments involving rice and potato-virus interactions, early activation of the phenylpropanoid pathway has been linked to increased disease resistance (Das et al., 2025; Kogovšek et al., 2016). Additionally, the sustained activity of ether lipid metabolism in Sifra at this early stage supports findings that changes in membrane composition can act as an early defence barrier (Hou et al., 2016). This may explain why some cultivars exhibit resistance even without visible signs of induced defences, as they are already primed (Conrath et al., 2015).

At 48 hpi, Sifra and Valor exhibit contrasting metabolic strategies (Balmer et al., 2013; Bigeard et al., 2015). The dramatic difference between cultivars is most apparent in (Figure 4B), where Valor undergoes early metabolic reprogramming, marked by a significant increase in glycerophospholipid metabolism (pathway impact: 0.246). The accumulation of phosphatidic acid and phosphatidylglycerol indicates active membrane remodeling and signalling (Testerink and Munnik, 2011; Ruelland et al., 2015; Hou et al., 2016). This spike likely reflects the activation of pattern-triggered and effector-triggered immune responses (Jones and Dangl, 2006; Bigeard et al., 2015). However, such a robust response may be energetically expensive, leading to an inability to maintain defence, as evidenced by a decline at 72 hpi (Berger et al., 2007; Walters et al., 2013). A similar decline in metabolic defence was observed in Valor, mirroring trends seen in other potato-virus pathosystems (Manasseh et al., 2023; Zheng et al., 2012). The increase in glycerophospholipids observed in Valor aligns with previous studies indicating that lipid-based signalling is crucial for regulating plant immune responses (Hou et al., 2016; Cacas et al., 2016). Phosphatidic acid, as a signalling molecule, is involved in activating defence gene expression and inducing structural changes in membranes during infection (Testerink and Munnik, 2011; Hong et al., 2010).

By 72 hpi, the ability to maintain defence over time is vital. (Figure 4C) demonstrates that Sifra continues to show elevated levels of solanidine and lysophosphatidylglycerol, indicating an ongoing defence response. In contrast, Valor shows a reduction in metabolic activity, which may stem from energy depletion or pathogen suppression (Berger et al., 2007; Bostock et al., 2014). Lee et al. (2013) demonstrated that extracellular metabolites produced by P. syringae can actively suppress plant defence responses. This finding may help explain the decline in Valor’s metabolic defences during the later stages of infection. Our research highlights that not only the presence but also the timing of these accumulations is critical for effective defence (Liu et al., 2024). The ongoing metabolism in Sifra at 72 hpi stands in contrast to what is typically observed in susceptible plants, where defence often fails due to pathogen interference or energy depletion (Dodds and Rathjen, 2010; Doehlemann et al., 2017). This ability to maintain a sustained response may be crucial for achieving durable resistance (Mauch-Mani et al., 2017; Conrath et al., 2015).

Our VIP analysis (Figures 5A–B) identified several metabolite classes that serve as molecular signatures of the different defence strategies:

The early and consistent presence of quercetin 3-alloside (VIP score: 3.4) in Sifra suggests a novel form of defence priming (Agati et al., 2012; Falcone Ferreyra et al., 2012). Quercetin compounds not only fight pathogens directly but also help activate other defence mechanisms (Treutter, 2006). The glycosylation of quercetin into quercetin 3-alloside may enhance its stability and facilitate storage within the cell (Vogt and Jones, 2000; Bowles et al., 2005). This finding is particularly relevant given recent advancements in our understanding of flavonoid biosynthesis and regulation (Nabavi et al., 2020; Tohge et al., 2017). These attributes position quercetin 3-alloside and other early markers, such as alpha-solanine, as promising candidates for resistance screening and breeding (Sakurai, 2022; Liu et al., 2024).

We observed consistent trends in glycoalkaloid accumulation: alpha-solanine at 24 hours hpi, solanine and solasonine at 48 hpi, and solanidine at 72 hpi. This temporal progression is visible across the VIP plots in (Figures 5A–C). This pattern indicates a strict regulation of glycoalkaloid metabolism over time (Friedman, 2006; Zhao et al., 2021). Each of these compounds may influence different aspects of the pathogen’s biology (Liu et al., 2024). The shift from glycosylated molecules like alpha-solanine to the aglycone solanidine suggests that enzymatic modifications play a role in modulating the activity and distribution of these molecules during the defence response (Itkin et al., 2013). Shenkarev et al. (2007) explained these key enzyme steps in potato glycoalkaloid production, showing how the plant regulates its defence chemicals. Glycosylated molecules may also function as stable reservoirs, readily activated when needed (Cardenas et al., 2015). Research has shown that glycoalkaloid biosynthesis is regulated at both the gene expression and post-transcriptional levels, with specific transcription factors directly involved in activating the key biosynthetic pathway (Liu et al., 2024; Itkin et al., 2013). Multiple genetic controls, including networks of co-expressed genes, regulate the metabolism of steroidal glycoalkaloids in potato tubers. This supports the notion that the differences between cultivars, such as Sifra and Valor, stem from the genetic regulation of these pathways (Shen et al., 2021). The consistent activity of this pathway in Sifra may stem from variations in the regulation of these genes.

Dastmalchi et al. (2019) showed that glycoalkaloid content infection in potato varies among different cultivars. Notably, cultivars with higher glycoalkaloid levels also demonstrated greater antibacterial activity, suggesting a direct relationship between metabolite accumulation and the intensity of the plant’s defence. Similarly, Dahlin et al. (2017) showed that glycoalkaloids, such as α-solanine and solanidine, can destabilise cell membranes and internal structures of Phytophthora infestans, exhibiting broad antimicrobial properties. Together, these findings support the idea that Sifra’s early and sustained glycoalkaloid accumulation may provide a stronger and more durable biochemical defence against pathogens like P. syringae.

Our findings reveal high concentrations of lysophospholipids and phosphatidic acid derivatives, underscoring the importance of lipid signalling in plant defence (Hou et al., 2016; Ruelland et al., 2015). Sifra consistently exhibits elevated levels of lysophosphatidylcholine and lysophosphatidylethanolamine, both of which function as membrane components and signalling molecules (Cacas et al., 2016; Hou et al., 2016). Their persistent presence in Sifra suggests that lipid signalling is continuously active, potentially through phospholipase A activity, which may prepare cell membranes to transmit signals swiftly upon pathogen detection (Hong et al., 2010). Pathway impact and topology analysis indicate that glycerophospholipid metabolism occupies a central position within the defence-related metabolic network, functioning as a regulatory hub integrating membrane dynamics with downstream defence responses. This systems-level role of lipid pathways mirrors the network-based interpretation proposed by Risoli et al. (2025).

The contrasting strategies of Sifra (constitutive defence) and Valor (induced defence) illustrate evolutionary trade-offs in resource allocation (Züst and Agrawal, 2017). While constant defences can deplete energy resources for growth and reproduction (Stamp, 2003; Koricheva, 2002), pathogen-induced responses conserve resources but may result in delayed protection (Karban and Baldwin, 1997; Agrawal, 1999). In high-disease environments, the reliability of Sifra’s constitutive defences is crucial (Walters et al., 2013; Mauch-Mani et al., 2017), especially as climate change intensifies disease pressures (Bebber et al., 2013; Ghini et al., 2016; Velásquez et al., 2018).

The key pathways and metabolites identified in our study offer promising targets for engineering disease resistance (Pollastri and Tattini, 2011). Enhancing phenylalanine metabolism, flavonoid biosynthesis, and glycoalkaloid biosynthesis could bolster resistance in susceptible cultivars. Recent advancements in synthetic biology show that it is possible to engineer complex defence pathways in plants (Nielsen and Keasling, 2016). Engineering transcription factors may allow us to replicate the coordinated defences observed in Sifra (Chattopadhyay et al., 2019). The time-course metabolite patterns identified in this study could serve as biomarkers for resistance breeding (Ghatak et al., 2018). Early markers, such as quercetin 3-alloside and alpha-solanine, could facilitate the rapid identification of resistant lines, while sustained responses at 72 hpi may indicate durable resistance (Sakurai, 2022; Mur et al., 2017). This strategy aligns with recent metabolomics-driven approaches in cereals and Solanaceae crops (Effah et al., 2023; Jose et al., 2023).