Tomato (Moneymaker) and N. benthamiana were grown at 21 ± 3°C, 30-50% humidity, and a 16/8-hour light/dark photoperiod.

Previously created tomato single adc1 and adc2 and double adc1/adc2 mutants (Wu et al., 2019) and the new odc1 mutant were generated via CRISPR-Cas9 (Jacobs et al., 2015) using two gRNAs targeting each gene. These constructs were transformed into tomato (Wittmann et al., 2016) and genotyping was conducted as described by Wu et al., 2019 using primers presented in Supplementary Table 1 .

Germinated seedlings were genotyped, then transferred into 11 cm diameter pots at 18 days after germination (DAG), and phenotypes measured/counted every 3 days from 26-47 DAG. Plant height was measured to nearest 0.5cm from the cotyledon to shoot apical meristem. Only fully expanded adult leaves originated from the primary stem were counted. All buds and opened flowers were counted from each plant. Photos of every replicate were taken 37 DAG and leaf and flower photos taken at 47 DAG.

The youngest, fully expanded adult leaf was cut from each replicate (aged 4-5 weeks). The terminal and two primary leaflets were kept in water at ~6°C (cold treatment; CT) or room temperature (RT) for 24 hours (in darkness). Samples for LC-MS or RNA extraction were taken without allowing CT leaves to warm.

Tomato ADC1, ADC2, and ODC1 genes were cloned from WT tomato gDNA into pUC57 vectors as level I modules for further Golden Gate cloning (Binder et al., 2014). PCR-based site-directed mutagenesis of key SlADC catalytic residues was conducted in level I modules, replacing nucleotides to encode an arginine instead of the WT lysine or cystine. Constructs were cloned into level II T-DNA expression vectors (Binder et al., 2014) and transformed into A. tumefaciens GV3101 using electroporation.

A. tumefaciens strains containing T-DNA expression constructs and a p19 silencing suppressor plasmid-containing strain were grown in Yeast Extract Broth (YEB) media (28°C, 180rpm), centrifuged, and resuspended in infiltration buffer (10 mM MgCl₂, 5 mM MES, and 150µM acetosyringone) to an optical density at 600nm (OD₆₀₀) of ~0.5. After an additional incubation (~4 hours, 28°C, 180rpm), all expression construct-containing cultures were mixed 1:1 with the p19-containing strain to a final OD₆₀₀ of 0.4 and infiltrated into the abaxial surface of N. benthamiana leaves using a blunt end syringe.

50mg (exact weight recorded) of plant tissue was flash frozen, ground in pre-cooled adapters using a TissueLyser II (Qiagen ^®; 22 Hz, 30 sec, x2) and homogenized in ADC activity buffer (300µL of 5 mM Tris [pH 7.5], 0.75% polyvinylpolypyrrolidone [PVPP], 1 mM ascorbate, 50 µM pyridoxal 5 phosphate [P5P], 0.5 x protease inhibitor [Roche complete]; 2µM ¹³C₁ D₂ citrulline [Eurisotop^®] and 2.5µM D₅ tryptophan [Sigma-Aldrich^®] as internal standards). Samples were then centrifuged (10min, 4°C, 18000g) and the supernatant (SN) collected. 10 µL of each ADC and ODC enzyme substrates ¹³C₆ arginine [Carl Roth^®] and ¹³C₅ ¹⁵N₂ ornithine [Eurisotop^®] were then added to the SN of each sample (180 µl) to a final concentration of 50 or 500µM and incubated for 1 hour at 37°C before the assay was stopped by using 5µL of heptafluorobutyric acid (HFBA) and centrifuged (30 min, 4°C, 18000g). For the “whole extract” (WE) protocol, the initial centrifugation step and collection of SNs was omitted, with the heavy isotope enzyme substrates directly included in the ADC activity buffer. When testing for LDC activity, 500 µM ¹³C₆ ¹⁵N₂ lysine (LGC Standards^®) was included in the buffer.

For N. benthamiana samples, which require an additional extraction step, 80µL of the stopped WE assay samples was combined with 200µL methanol and 100µL chloroform, and incubated (10 minutes, 25°C, 950 rpm). 120µL of ultrapure MilliQ water was added, the incubation repeated, a final centrifugation step (18,000 g, 30 minutes, 4°C), and the upper phase used for profiling.

To determine if metabolite levels fell into the linear range of their respective calibration curves ( Supplementary Table 2 ), dilutions (in water, 0.1% FA, 0.1% HFBA) of representative samples were tested prior to running all samples. LC-MS profiling was performed using a Micro-LC M5 (Trap and Elute) and QTRAP6500+ (Sciex) system operated in Multiple Reaction Monitoring (MRM) mode. Chromatographic separation was achieved on a HaloFused C18 micro column (150 x 0.5 mm; 2.7 μm; AB Sciex) and a Luna C18 (2) micro trap column (5μm; 100 Å; 20×0.5mm; Phenomenex); injection volume was 50 μL. Analytes were ionized using an Optiflow Turbo V ion source equipped with a SteadySpray T micro electrode (10-50 μL min^-1) in positive ion mode. Supplementary Table 3 presents information concerning the chromatographic gradients and MS parameters. A shorter trap gradient was chosen for the assays without HFBA in the solvent to reduce the loss of the ion paring agent in the sample itself. A general loss of sample volume of about 10% (data not shown) using the reduced trapping time was deemed acceptable.

Vendors Sciex OS software was used to analyze acquired MRM data. The metabolite content in each replicate was calculated using commercial standards (external calibration in water, 0.1% FA, 0.1% HFBA), except for ¹³C₅ agmatine, ¹³C₄ putrescine, and ¹³C₄ ¹⁵N₂ putrescine, which were quantified based on their natural variants. Data were normalized against exact sample fresh weight (FW), and D₅ tryptophane or ¹³C₁D₂ citrulline.

Note: A small ¹³C₅ agmatine peak was observed in control assays with no plant extracts. As no enzyme was present in these samples, the production of ¹³C₅ agmatine can only be explained by chemical decomposition or insource fragmentation of ¹³C₆ arginine within the MS ion source. This is further confirmed by the retention time of this ¹³C₅ agmatine peak, which is not congruent to the natural agmatine but rather to arginine. All samples with low ADC activity (e.g., tomato adc1/adc2 double mutant extracts) were double-checked for accumulation of enzymatically produced ¹³C₅ agmatine (retention time of agmatine) instead of the chemically derived variant (retention time of arginine) ( Supplementary Figure 1 ).

RNA was extracted from flash frozen tomato samples using the Universal RNA Purification Kit (Roboklon GmbH); on- and off-column DNase (ThermoFischer Scientific) treatments were both conducted as tomato ADCs do not contain introns. RNA (850 ng) was converted to cDNA using the RevertAid cDNA Synthesis Kit (ThermoFischer Scientific) and diluted 1:10 in RNA-free water. RT-qPCRs were conducted as described by Wu et al., 2019, using the CFX96 system (BioRad), and 3 technical replicates per biological replicate. SlADC1 and SlADC2 transcript levels were calculated using 2^-ΔΔCT and normalized to SlTIP41 (Lacerda et al., 2015) expression as previously described by Wu et al., 2019. RT-qPCR primers used in this work are found in Supplementary Table 1 .

N. benthamiana leaf tissue was flash frozen and ground in pre-cooled adapters, then homogenized in extraction buffer (80 mM Tris [pH 7.5], 2.5% SDS, 50 mM EDTA, 2.5x strength protease inhibitor; 33 µL per leaf disk) and loading dye added (20 mM Tris-HCl pH 8.6, 40% [v/v] glycerol, 10% [v/v] 2-mercaptoethanol, and 0.1% [w/v] bromophenol blue; 13 µL per leaf disk). Samples were heated (90°C, 10 minutes), briefly centrifuged, and loaded onto a 10% SDS PAGE gel. Proteins were transferred to a PVDF membrane using Trans-Blot^® Turbo Transfer System (BioRad), blocked (5% milk powder in TBST [50 mM Tris-HCl, 150 mM NaCl, 0.05% Tween-20]), then incubated overnight (6°C, rocking) with the epitope-detecting antibody (i.e., monoclonal, conjugated anti-HA-HRP antibody [Roche] or monoclonal anti-c myc [raised in mouse; Sigma Aldrich]). Washed blots (3x 5 min in TBST) were either imaged directly or blocked, incubated with a conjugated anti-mouse-HRP antibody (Sigma Aldrich) for 2 hours, and washed again. ECL Select™ Western Blotting Detection Reagent (Amersham™, cytiva) was used for detection and blots were imaged using an Amersham™ Imager 600, GE Healthcare Life Sciences. Ponceau staining (0.1% [w/v] Ponceau S, and 5% [v/v] glacial acetic acid) of all blots was conducted to verify even protein loading.

Microsoft Excel and R were used for basic statistical analysis and generation of graphs and heatmap diagrams, which were then assembled using Adobe Illustrator. Code to generate metabolite heatmaps is available: https://github.com/TheDepe/Metabolite-Measurement.

We first sought to improve the sensitivity of our LC-MS-based method to quantify ADC activity and PA-related metabolites from plant extracts, even with low ADC expression. In the previously established protocol, ADC activity was quantified in the supernatant (SN) of cellular extracts ( Figure 2B ; Wu et al., 2019). We rationalized that SNs may lack organelles containing ADC proteins and metabolites (e.g., chloroplasts), consequently reducing ADC activity. Therefore, we compared ADC activity in SNs (i.e., previous protocol) to whole cellular extracts (termed here on out as ‘whole extract’ or WE) where the extraction buffer is applied to ground leaf tissue without an additional centrifugation step ( Figure 2B ). Furthermore, we considered whether a higher ADC substrate (i.e., ¹³C₆ arginine; Figure 2A ) concentration would elevate ADC activity. We therefore compared using ten times more ¹³C₆ arginine (i.e., 500 µM; “high”) to the original 50 µM (“low”) concentration ( Figure 2B ).

High versus low concentrations of the ADC substrate were added to the SN or WE of tomato leaves kept at either room temperature (RT) or cold-treated (CT); the latter is known to elevate SlADC transcript levels. Indeed, both SlADC1 and SlADC2 transcript levels were significantly induced by ~3.5- and 2.9-fold following CT relative to RT controls ( Supplementary Figure 2 ). When using the higher ¹³C₆ arginine concentration, we observed a dramatic increase in ADC activity ( Figure 2C ). While this was particularly evident in CT samples with induced SlADC levels (~11-fold for WE and ~7-fold for SN samples), a similar increase was observed in RT samples (i.e., ~11- and ~8-fold higher in WE and SN samples, respectively), although not statistically significant ( Figure 2C ). Interestingly, applying a higher ¹³C₆ arginine concentration enabled reliable detection of the heavy isotopic variant of putrescine produced from ¹³C₆ arginine (i.e., ¹³C₄ putrescine; Figure 2A ), which was not detectable with the low ¹³C₆ arginine concentration ( Figure 2D ). Overall, these results demonstrate that increasing the ¹³C₆ arginine substrate concentration elevated the sensitivity of our ADC activity assay, enabling reliable measurement of ¹³C₅ agmatine, and derived ¹³C₄ putrescine, even under ADC non-inducing conditions.

Unexpectedly, we observed no significant differences in levels of ¹³C₅ agmatine or ¹³C₄ putrescine between SN and WE samples ( Figures 2C, D ). However, the SN-based ADC activity protocol is more labor-intensive, time-consuming, and error-prone due to more handling steps, making the new WE protocol preferable.

Measuring multiple enzyme activities in parallel reduces experimental costs and workload. Therefore, we wanted to test if both ADC and ODC activities (i.e., decarboxylation of ornithine to putrescine) could be quantified in one LC-MS run, despite both pathways producing putrescine ( Figure 1 ). To achieve this, heavy isotopically labelled substrates of ADC and ODC were selected that would produce putrescine variants with sufficiently different molecular masses to be distinguishable by MS. For this, ¹³C₆ arginine (+6 mass) and ¹³C₅ ¹⁵N₂ ornithine (+7 mass) were selected as ADC and ODC substrates, which produce ¹³C₄ putrescine (+4) and ¹³C₄ ¹⁵N₂ putrescine (+6), respectively ( Figures 2A, E ).

We successfully detected ¹³C₄ ¹⁵N₂ putrescine in our tomato extracts and discriminated it from the ADC-derived ¹³C₄ putrescine, enabling simultaneous ADC and ODC activity quantification ( Figure 2F ). In contrast to ADC activity, ODC activity was not elevated by higher substrate levels and ODC activity was higher in all WE assays compared to the SN. Our results indicate that a substantial portion of ODC enzymes are localized in organelles and suggest that WEs are preferable to SNs for assaying enzyme activities.

We also considered the LC-MS gradient composition for optimization. Previously, 0.05% heptafluorobutyric acid (HFBA) was included in the gradient solvents as an ion pairing agent (Wu et al., 2019). However, due to its acidity, HFBA could cause corrosion of internal LC-MS components and also increase LC-MS contamination levels by adhering to the column and MS ion source. The assay from Figure 2 was analyzed twice; once with HFBA in the solvent, and once without, relying solely on the HFBA added to the sample (i.e., to stop enzyme activity assay; Figure 2B ) for ion pairing. In all cases, ¹³C₅ agmatine intensities showed no significant difference between the two gradient compositions ( Supplementary Figure 3 ). Therefore, HFBA can be omitted from the solvent without affecting activity quantification, whilst improving longevity and performance of the LC-MS system.

Monitoring changes across the wider PA network – particularly of anabolic compounds – would improve our understanding of PA metabolic fluxes during stress responses. Previously, we had only quantified the concentrations of agmatine, putrescine, spermidine, and spermine (Wu et al., 2019). Yet, to better comprehend the impact of abiotic stresses or extraction methods (e.g., SN and WE), we extended our protocol to include arginine, ornithine, citrulline, glutamine, N-acetylputrescine, cadaverine, and later, lysine. This was achieved by adjusting LC gradients to avoid co-elution of structural isomers such as acetyl-putrescine and agmatine, and glutamine and lysine ( Supplementary Tables 2 , 3 , Supplementary Figure 4 ).

When quantifying changes in the concentrations of numerous metabolites in response to external (e.g. biotic or abiotic stress) or internal (e.g. developmental program) stimuli, or due to differences in extraction protocols (e.g., SN versus WE), it is difficult to infer metabolic fluxes if the data are displayed using conventional bar charts, as such representations do not reveal the causal relationships between metabolites. Therefore, we presented our LC-MS profiling results as a heatmap-based pathway diagram to better visualize metabolic relationships of the PA network ( Figure 3 ). Following CT, agmatine, putrescine, ornithine, and citrulline levels increased, while arginine levels were reduced compared to the RT samples. This suggests that cold-induced SlADC transcript levels and ADC activity reduces the arginine pool to produce more agmatine and subsequently putrescine. Interestingly, spermidine and spermine levels are not influenced by CT, suggesting a regulatory mechanism that prevents the cold-induced increase in putrescine from causing an unregulated increase in these higher PAs.

Our results collectively show that a higher substrate concentration (i.e., 500 µM) increases assay sensitivity. Additionally, using WEs instead of SNs, now captures the activity of both cytoplasmic- and organelle-localized enzymes. We also found that omitting the ion pairing agent HFBA from the LC-MS gradient had no impact metabolite quantification but protects the LC-MS mechanical components. Finally, visualizing PA-related metabolite changes via a heatmap pathway diagram aids in understanding the interdependencies between metabolites.

Previously, we assumed a linear relationship between ADC protein abundance and its enzymatic activity; yet this assumption had not been experimentally validated (Wu et al., 2019; Gallas et al., 2024). To test for such a linear relationship, we expressed tomato ADC (SlADC) proteins in N. benthamiana leaves, where their expression levels can be easily modulated by inoculating varying amounts of T-DNA-containing A. tumefaciens strains. These T-DNA constructs encoded epitope-tagged SlADC1/2, driven by the constitutively active 35S promoter. They were transformed into A. tumefaciens and infiltrated into N. benthamiana leaves (termed agroinfiltration) with increasing bacterial loads (i.e., OD₆₀₀ 0.05, 0.1, 0.5, and 1). Two days post infiltration (dpi), samples were harvested for both immunoblotting and LC-MS profiling. Immunoblots showed increasing signal intensity of tagged SlADCs that mirrored the increasing bacterial loads ( Figure 4 ).

Previously, unprocessed N. benthamiana leaf samples rapidly clogged the micro-LC columns. Therefore, an additional liquid-liquid extraction step was applied to all N. benthamiana leaf samples following the assay. Here, a proportion of alkaloids would be separated away from the upper phase that is subsequently used for LC-MS profiling, reducing their interference with the micro-LC columns.

Comparable to the protein immunodetection analysis, rising bacterial load increased ADC activity ( Figure 4 ), confirming our hypothesis of a linear relationship between ADC abundance and activity. To confirm that this observation was not due to heightened biotic stress influencing ADC activity, the same experiment was conducted with epitope-tagged fluorophore controls. Again, increasing bacterial loads were associated with higher detectable epitope-tagged protein levels, but no mirrored increase in ADC activity was observed ( Supplementary Figure 5 ). Overall, these results confirm a linear relationship between SlADC1/2 protein abundance and ADC activity, demonstrating that ADC activity quantification can be used as a reliable proxy for ADC protein levels from plant extracts.

We next wanted to evaluate whether the enzymatic functionality of ADC mutant variants, recombinantly expressed in N. benthamiana leaves, can be assessed using our optimized ADC activity assay. We generated SlADC1 and SlADC2 mutants in the key catalytic lysine and cystine residues that are essential for enzymatic function ( Supplementary Figure 6 ). These residues were either individually mutated (i.e., SlADC1 K149A or C540A, and SlADC2 K156A or C548A) or both mutated to alanine (i.e., SlADC1 K149A C540A and SlADC2 K156A C548A). T-DNA constructs were generated encoding 35S-promoter driven, epitope-tagged versions of mutated SlADC1/2, transformed into A. tumefaciens, and agroinfiltrated into N. benthamiana leaves alongside complementary mCherry and GFP fluorophores, serving as negative controls. LC-MS analysis of wild-type (WT) SlADC1- and SlADC2-agroinfiltrated samples showed 72- and 85-fold higher ADC activity compared to the average ADC activity of the negative controls ( Figure 5 ). No significant difference between the negative controls and any single or double SlADC1 or SlADC2 mutant variant was found, despite all proteins being equally expressed ( Supplementary Figure 7 ). These results demonstrate that both lysine and cysteine residues are essential for SlADC1/2 catalytic activity.

As ADCs and ODCs are key enzymes in PA biosynthesis, we wanted to show that our LC-MS assay could quantify both activities following their transient expression in N. benthamiana. We generated 35S-promoter driven T-DNAs encoding, HA epitope-tagged SlADC1, SlADC2, SlODC1, and mCherry (negative control) and agroinfiltrated these into N. benthamiana leaves. At 2 dpi, samples were harvested for both LC-MS and immunoblot analysis. Immunoblot analysis showed that all proteins were detectable ( Supplementary Figure 8 ).

As anticipated, transient expression of SlADC1 and SlADC2 produced 83- and 58-fold higher ADC activity levels compared to the negative mCherry control, whilst SlODC1 expression produced no significant increase in ADC activity ( Figure 6A ). This pattern was mirrored in the accumulation of ¹³C₄ putrescine (generated from ¹³C₆ arginine), where SlADC1 and SlADC2 expression caused 42- and 43-fold higher levels compared to the control ( Figure 6B ). On the other hand, 20-fold higher ODC activity was observed upon transient SlODC1 expression, whilst no significant increase was observed following SlADC1 and SlADC2 expression ( Figure 6C ). Interestingly, expression of SlADC1 and SlADC2 enzymes, but not SlODC1, significantly increased putrescine compared to the mCherry control ( Figure 6D ). Also in this assay, we included a ¹³C₆ ¹⁵N₂ lysine substrate along with the isotopic variants of arginine and ornithine to determine substrate specificity of SlADC and SlODC. As expected, no LDC activity (i.e., production of ¹³C₅ ¹⁵N₂ cadaverine from ¹³C₆ ¹⁵N₂ lysine) was detected above background levels from any expressed enzyme.

Overall, these results demonstrate that tomato ADC and ODC enzymes have strict preference for their respective metabolic substrates and that their activities can be easily analyzed following transient expression in N. benthamiana leaves without the need for in vitro protein purification.

To gain deeper insight into the PA network, we analyzed tomato adc1, adc2, and adc1/adc2 mutants (Wu et al., 2019) using our optimized LC-MS profiling method. We also included an unpublished single odc1 tomato mutant, with the remaining two SlODC genes WT (Solyc03g098300.1 and Solyc03g098310.1). These tomato lines were all generated via CRISPR-Cas9 mutagenesis and carry deletions that lead to premature stop codons (Wu et al., 2019; Supplementary Figure 9 ). In contrast to A. thaliana where the loss of both ADC genes is embryonically lethal due to the absence of ODC, the tomato adc1/adc2 double mutant is viable, but cannot produce seeds. Since adc1/adc2 mutants produce no seeds, we identified these double mutants by genotyping the progeny of a parental line that was homozygous mutant for SlADC1 (sladc1/sladc1) and heterozygous for SlADC2 (SlADC2/sladc2) ( Supplementary Figure 9D ). Since adc1/adc2 mutants offer a unique opportunity to study the functional roles of ADCs, we analyzed phenotypic differences between these mutants and WT plants. Twenty genotyped tomato plants representing WT, adc1, adc2, adc1/adc2, and odc1 were examined every three days between 26 to 47 days after germination (DAG), with photos taken of representative plants at 37 DAG ( Figure 7A ). Leaves of adc1/adc2 mutants developed primary and secondary leaflets more slowly, often lacking intercalary leaflets entirely ( Figure 7B ). By 47 DAG, their leaves and leaflets were less flat and expanded compared to those of WT plants. Interestingly, no flowers, buds, or tissues indicating developing flower buds were ever observed on adc1/adc2 plants, demonstrating that ADC genes are indispensable for flower development ( Figure 7C, D ). By contrast, the single adc1 or adc2 mutants showed no obvious deficiency in flower development compared to WT. Additionally, the double adc1/adc2 mutant plants were shorter and had fewer adult leaves compared to WT or the other mutants ( Figures 7E, F ). Intriguingly, adc1 and odc1 were both significantly taller 47 DAG compared to the WT. Overall, this phenotypic analysis demonstrates the importance of ADCs for tomato growth and development of reproductive organs.

To determine the role each ADC enzyme plays in PA metabolism, transcript levels, ADC activity and PA-related metabolites were quantified in all five genotypes following an acute CT. SlADC1 and SlADC2 transcript levels were increased in all genotypes following CT, except for SlADC1 in the adc1 mutant ( Supplementary Figure 10 ). ADC activity was detectable for all genotypes following RT or CT, except for the adc1/adc2 double mutant ( Figure 8A ), consistent with agmatine being exclusively synthesized by ADCs. Following CT, ADC activity was significantly induced in the WT, adc2, and odc1 genotypes, but not adc1, despite significant transcriptional activation of its functional SlADC2 gene following CT ( Supplementary Figure 10B ). These results indicate that SlADC1 is the main contributor to ADC activity in adult tomato leaves following CT.

When quantifying PA metabolites, we again observed higher agmatine, putrescine, citrulline, and ornithine levels and lower arginine levels in WT tomato leaves following CT ( Figure 8B ). Contrastingly, the adc1/adc2 double mutant had no detectable agmatine above background levels and under both RT and CT conditions the adc1/adc2 mutant had higher arginine but lower putrescine levels. Interestingly, N-acetylputrescine levels were also lower in the adc1/adc2 mutant compared to WT, while spermidine and spermine levels remained unaffected. Finally, we noted that although lysine (the cadaverine precursor) was lower in all genotypes following CT, cadaverine was not increased and was only reliably detected in the adc1/adc2 mutant under both conditions.

In summary, we characterized the phenotypes, ADC activities, and PA network in the tomato adc1, adc2, adc1/adc2, and odc1 mutants, demonstrating the importance of ADC enzymes not only for PA synthesis, but also for plant development and generation of reproductive tissues.

In this work we optimized the sensitivity and workflow of our previously established LC-MS-based method for quantifying ADC activity and PAs from plant extracts. Moreover, we broadened the scope of this method, including ODC activity and 7 more PA-related metabolites (i.e., arginine, ornithine, citrulline, glutamine, N-acetylputrescine, lysine, and cadaverine) that were not covered by the previous method. Multiple aspects of our LC-MS approach were considered for optimization, including substrate concentration, extraction method, and presence of the ion-pairing agent HFBA in the solvent.

Using 10-fold higher substrate concentration increased ADC activity ~11-fold ( Figure 2C ), producing clear ¹³C₅ agmatine LC-MS peaks well above detection limits, even from samples with low levels of ADC proteins (e.g., RT single adc mutants; Figure 8 ). The ¹³C₆ arginine-derived putrescine variant (¹³C₄ putrescine) could only be detected with elevated substrate levels, but not at lower ¹³C₆ arginine concentrations ( Figure 2D ). Since at both substrate concentrations ¹³C₆ arginine levels were never depleted ( Supplementary Data File ), the higher concentration (i.e., 500 µM) likely increased ADC activity by almost reaching the reported SlADC1 Km of 600 µM (Guo et al., 2023), a situation representing 50% substrate saturation of the enzyme. A higher ¹³C₆ arginine concentration to achieve 100% enzyme substrate saturation was not considered as this could negatively affect the chromatographic separation and electrospray ionization process of coeluting metabolites (Jessome and Volmer, 2006). Taken together, our results suggest that 500 µM ¹³C₆ arginine increased assay sensitivity, facilitating detection of subtle fluctuations in ADC activity that occur in response to abiotic stresses.

Putrescine, the precursor for synthesizing higher plant PAs, is produced via the ADC- or ODC-dependent pathways. We aimed to establish an approach allowing quantification of both ADC and ODC enzymatic activities from the same tissue extract in one LC-MS sample. To quantify ODC activity, we used a stable ¹³C₅ ¹⁵N₂ ornithine isotope that, following decarboxylation by ODC, produces ¹³C₄ ¹⁵N₂ putrescine. This putrescine isotope differs in its molecular mass sufficiently from the ADC-derived ¹³C₄ putrescine (i.e., from ¹³C₆ arginine), enabling quantification of both ADC and ODC activities simultaneously ( Figure 2F ). In contrast to ADC activity, increasing ¹³C₅ ¹⁵N₂ ornithine substrate concentration did not elevate ODC activity. This is likely due to low SlODC protein levels as SlODC is not highly transcribed in adult tomato leaves (Kwak and Lee, 2001; Acosta et al., 2005; Sivakumar et al., 2022).

We compared ADC and ODC activities in WEs and SNs ( Figure 2B ) and consistently found higher ODC activity in WEs than SNs ( Figure 2F ). ODCs have been shown to localize to the endoplasmic reticulum (ER, Joshi et al., 2024) and predicted to localize to chloroplasts and mitochondria (Sivakumar et al., 2022). Centrifugation to obtain SNs removes these key organelles, thereby reducing ODC abundance and activity. In contrast, ADC activity did not differ between SNs and WEs ( Figure 2C ), indicating that SlADCs are predominantly localized in the cytoplasm of tomato leaf cells. Overall, this work demonstrates that analysis of WEs simplifies the workflow and is equally suitable for enzymes localized in the cytoplasm, ER or plastids. Comparative LC-MS analysis of SN and WE samples could also provide information regarding localization of natively expressed enzymes, thereby complementing classical subcellular localization studies.

Finally, we tested whether HFBA could be omitted from the LC-MS gradient whilst still enabling reliable ADC activity quantification. HFBA was used for two purposes in the previously published method: firstly, to stop proteinaceous reactions (i.e., enzymatic activity) at a defined timepoint ( Figure 2B ); and secondly, as an ion pairing agent, that improves resolution of charged metabolites. However, HFBA potentially reduces LC-MS platform longevity and integrity, as well as creating matrix suppression effects. Upon comparing metabolite peak areas of the same samples analyzed with or without HFBA in the LC-MS gradient, we observed no significant differences ( Supplementary Figure 3 ), indicating that sufficient ion-pairing occurs from the HFBA added directly into the samples. Therefore, HFBA can be omitted from the LC-MS gradient without compromising metabolite and enzyme quantification, improving LC-MS system longevity.

A gene’s transcript level is commonly used to estimate the amount and biological activity of the encoded protein, yet this approach is unsuitable for ADC and ODC, whose translation is also regulated post-transcriptionally (Kwak and Lee, 2001; Wu et al., 2019; Jiménez-Bremont et al., 2022). Here, we validated a linear relationship between ADC activity and ADC protein abundance ( Figure 4 ), showing that enzyme activity can be used as a proxy for protein abundance in planta. Therefore, with the simplicity and reliability of our LC-MS approach, it is now possible to efficiently compare ADC and ODC activities, and accordingly their protein levels, across many tissues and/or stimuli.

Furthermore, systematic comparison of ADC/ODC mRNA levels with enzyme activities quantified using our LC-MS approach could identify contexts where transcript levels and enzyme activities differ most. As such contexts with large differences would be indicative of post-transcriptional regulation, our approach has the potential to be used in systematic screens, uncovering tissue contexts or stimuli where ADC and ODC expression is regulated at the post-transcriptional level.

Agrobacterium-mediated transfer of T-DNAs encoding proteins of interest into N. benthamiana leaves has become a widely used method for functional analysis of plant genes. We therefore wanted to ensure that our LC-MS approach could be applied to ADC/ODC proteins ectopically expressed N. benthamiana leaves. Yet, when using the standard extraction protocol, these samples clogged our Micro-LC columns. This is likely due to high alkaloid levels in this species, making up to ~1% of its dry weight (Stephan et al., 2018), which would also be present in the WE samples. By including an additional extraction step, N. benthamiana samples no longer interfered with the columns, enabling key experiments to be conducted ( Figures 4 - 6 ). Using this system, we not only verified that ADC activity has a linear relationship to ADC protein abundance ( Figure 4 ) but also confirmed the functional relevance of key lysine and cysteine residues required for SlADC activity ( Figure 5 ). We were also able to study substrate specificities of key enzymes in the PA network by simply adding stable isotope variants of arginine, ornithine, and lysine to plant extracts. These assays showed that transiently expressed tomato ADCs and ODC1 enzymes had the expected substrate specificities ( Figure 6 ). This extraction enables analysis of key PA biosynthetic enzymes, ADC and ODC, by leveraging transient expression in N. benthamiana. This allows comparative studies of enzyme mutants or homologs, or treatment responses (e.g., inhibitor, hormone, or abiotic/biotic stress treatments) all within the same N. benthamiana leaf background, eliminating the need for time- and resource-intensive recombinant protein expression and purification, thereby streamlining and accelerating enzyme activity studies.

N. benthamiana is not only a commonly used system for in planta recombinant expression of proteins of interest but also serves as a host for the bacterial model pathogens Pseudomonas and Xanthomonas. The assays established here now enable the combination of transient expression of enzymes that alter PA homeostasis with infections by Pseudomonas or Xanthomonas, both of which have been shown to manipulate the PA pathway through injected toxins or effector proteins (Kim et al., 2013; Guo et al., 2023; Gerlin et al., 2021). Such assays, together with our MS-based platform for PA quantification, could provide a powerful tool to investigate how different PAs influence bacterial infection.

We also aimed to broaden our portfolio of PA-related metabolites that can be quantified by our LC-MS approach, particularly focusing on anabolic pathway components, many of which play important roles in plant stress responses (Joshi and Fernie, 2017; Hildebrandt, 2018; Batista-Silva et al., 2019; Moormann et al., 2022). We compared 11 PA network members in RT and CT tomato leaves, finding a CT-dependent reduction of arginine and lysine levels and increased agmatine, putrescine, citrulline, and ornithine levels ( Figures 2 , 8 ). The arginine pool was likely depleted due to increased SlADC1/2 expression and ADC activity following CT ( Supplementary Figures 2 , 10 , Figures 2C , 8A ), accumulating agmatine for putrescine synthesis, a metabolite known to mitigate cold stress (Cuevas et al., 2008; Chen et al., 2019; Upadhyay et al., 2020; González-Hernández et al., 2022). Interestingly, despite increased ADC activity, agmatine, and putrescine levels, we never observed CT-dependent changes in spermidine or spermine levels ( Figures 3 , 8 ). Similar results have been previously reported (Alcázar et al., 2005, 2006; Wu et al., 2019; Gallas et al., 2024), indicating that spermidine and spermine levels are strictly regulated, possibly by PA-dependent translational regulation (Jiménez-Bremont et al., 2022). Displaying the data as a pathway heatmap highlights metabolites whose homeostasis may be prioritized (i.e., spermidine and spermine) or diverted to other metabolic pathways during stress responses (e.g., lysine). Together, quantification and heatmap visualization of the PA network offer comprehensive insight into pathway fluxes, revealing genotype- or condition-specific changes and key regulatory nodes.

We also used our optimized LC-MS-based approach to study the role of tomato ADC and ODC enzymes on ADC activity and PA metabolism following an acute CT using single adc1 and adc2 and double adc1/adc2 tomato mutants (Wu et al., 2019) and a novel odc1 mutant ( Supplementary Figure 9 ). Importantly, we confirmed that SlADC1 and SlADC2 are the sole enzymes capable of synthesizing agmatine in tomato as no ADC activity was observed above background levels in the adc1/adc2 double mutant ( Figure 8A ). Accordingly, compared to WT, this adc1/adc2 mutant had higher arginine levels and lower agmatine and putrescine under both RT and CT conditions ( Figure 8B ). This demonstrates that in the absence of functional ADC enzymes, arginine builds up, and as agmatine cannot be produced, putrescine synthesis is reduced. As adc1/adc2 plants have a functional ODC pathway but still lower putrescine levels compared to WT, this indicates that SlODCs cannot compensate for the loss of SlADCs. Surprisingly, no difference in spermidine or spermine levels were observed in the adc1/adc2 mutant compared to WT in either RT or CT samples, despite reduced levels of their precursor putrescine ( Figure 8B ). This suggests that despite completely lacking the ADC-dependent pathway, spermidine and spermine levels are maintained, indicating a prioritization regulatory mechanism maintaining their levels at the expense of putrescine. Alternatively, PAs may be synthesized in other tissues with higher ODC expression, such as the roots (Kwak and Lee, 2001; Acosta et al., 2005) and transported through the plant. Grafting the arial part of adc1/adc2 plants to WT root stocks and analyzing ADC activity and PA levels in their leaves could give insight into long distance transport of such metabolites.

The adc1 single mutant also had severely reduced ADC activity compared to adc2 and WT in both RT and CT conditions ( Figure 8A ), indicating that ADC1 is the main contributor of ADC activity in tomato leaves. This is likely due to higher SlADC1 expression in tomato leaves ( Supplementary Figures 2 , 10 ), rather than SlADC1 being a more efficient enzyme, as we observed only slightly higher ADC activity levels when transiently overexpressing SlADC1 compared to SlADC2 in N. benthamiana ( Figures 4 - 6 ).

We also characterized the phenotypes of these single and double tomato mutants, as a double adc1/adc2 mutant in A. thaliana cannot be generated (Urano et al., 2005). Compared to WT, adc1/adc2 mutant plants were severely affected, being only two-thirds as tall as the WT at 47 DAG, with significantly fewer adult leaves and no flowers ( Figure 7 ). In fact, we never observed any tissue resembling developing floral buds on adc1/adc2 plants, even after months of growth. Previous studies have found higher PA levels in floral tissues compared to leaves in various plant species and that supplementation of PAs can promote floral development (Chen et al., 2019). Spermine is essential for A. thaliana flower development (Zierer et al., 2016), yet no significant differences in spermidine or spermine levels were observed in adc1/adc2 leaves compared to WT ( Figure 8 ). On the other hand, putrescine levels, which were significantly reduced in adc1/adc2 leaves, were reportedly highest in the early tomato flower stages (Antognoni et al., 2002). Overall, this work demonstrates that ADCs are essential for plant and flower development, even though the concentrations of key PAs, such as spermidine and spermine remain unchanged in the adc1/adc2 mutant.