Salt Elicitation of Centella asiatica Enhances Antimicrobial and Keratinocyte Wound-Healing Activity: From Germplasm Screening to Nicotinamide Accumulation

1. Introduction

Centella asiatica (L.) Urb. (Apiaceae), known in Thailand as Bua Bok, is one of the most widely used medicinal herbs of tropical Asia and a common ingredient of skin-care products. Its reputation rests largely on the pentacyclic triterpenoids asiaticoside, madecassoside, asiatic acid and madecassic acid, which stimulate collagen synthesis, keratinocyte migration and wound closure in cellular and animal models [1,2]. These properties underlie the wide use of the plant in dermatology and cosmetology [3]. Extracts of the plant also inhibit a range of bacteria and fungi, an activity attributed to triterpenoid acids and phenolic constituents [4,5]. Because wound infection and delayed re-epithelialisation are linked problems, an extract that combines antimicrobial and keratinocyte-stimulating properties would be particularly useful for topical applications.

The chemical composition of a medicinal plant is dynamic and can be substantially influenced by environmental conditions. Abiotic stress redirects carbon and nitrogen into defensive and protective metabolites, and mild salinity is one of the cheapest stresses that can be applied under cultivation [6,7,8]. In C. asiatica, NaCl increases centelloside and phenolic contents and antioxidant capacity while the plant tolerates concentrations up to 100 mM [9,10]. We recently showed, in the same accession panel used here, that a short 100 mM NaCl treatment for 6 days enhances phytochemical accumulation and antioxidant activity in a genotype-dependent manner [11]. Salt stress also triggers the accumulation of compatible solutes such as glycine betaine and proline and reorganises NAD and amino acid metabolism, changes that are usually studied as markers of tolerance rather than as sources of bioactivity [12,13].

Thai C. asiatica germplasm is chemically and morphologically diverse, and accessions differ several-fold in centelloside content and biomass [14,15]. This diversity implies that the response to an elicitor will also vary, so that a single accession cannot represent the species. A screen across many accessions is needed to identify genotypes in which salt elicitation actually improves the desired bioactivity, before investing in metabolomic and functional characterisation of a selected line.

Three research gaps still remained. The effect of salt elicitation on the antimicrobial activity of C. asiatica extracts, and the consistency of that effect across germplasm, had not been examined. The metabolites induced in a responsive accession, and their contribution to any gain in activity, were unknown. It was also unclear whether an elicited extract retains the keratinocyte-stimulating properties expected of C. asiatica or whether the stress-altered chemistry compromises them. Previous studies of salt-treated C. asiatica measured bulk phytochemical indices rather than bioactivity, and none linked a germplasm screen to metabolomics and functional assays.

Here we addressed these questions in four linked steps. We screened extracts of 39 Thai accessions grown with or without NaCl against E. coli, S. aureus and C. albicans, using a growth ratio that expresses the effect of elicitation within each accession. We then profiled the metabolome of the most responsive accession, PHE, by untargeted LC-HRMS, and examined the changed metabolites both as fold changes and as shares of the total extract. We tested the hypothesis emerging from the metabolomics, namely that nicotinamide contributes to the enhanced activity, by comparing the NaCl extract, the untreated extract and pure nicotinamide in 24 h growth kinetics. Finally, we evaluated cytotoxicity, proliferation and scratch-wound closure in HaCaT keratinocytes. The results show that salt elicitation adds two distinct bioactivities to C. asiatica that are carried by different components of the altered metabolome.

2. Results

2.1. NaCl Elicitation Shifts Extracts from Growth Promotion Toward Inhibition

In the 4 h antimicrobial screening, extracts of untreated plants tended to promote rather than inhibit microbial growth, and elicitation removed this effect (Figure 1). The pattern was clearest for the two bacteria. Untreated extracts stimulated E. coli (Figure 1a) and S. aureus (Figure 1b) by a mean of 14 to 15%, whereas NaCl extracts of the same accessions were neutral, and the number of accessions with net inhibition roughly tripled for E. coli (7 to 22) and nearly doubled for S. aureus (11 to 19). For C. albicans (Figure 1c) the responses of individual accessions were much more scattered in both directions, and the condition means did not differ.

To express the effect of elicitation within each accession we used the growth ratio (GR), the growth of the microorganism with the NaCl extract relative to that with the untreated extract of the same plant, so that GR < 1 indicates stronger suppression after elicitation (Figure 1d). Confidence intervals (CI) and false discovery rates (FDR) were used throughout to summarise uncertainty. NaCl extracts reduced bacterial growth to about nine tenths of the untreated level, with a large effect size for E. coli (geometric mean GR 0.88, Hedges’ g = −0.84) and a moderate one for S. aureus (0.90, g = −0.56), and most accessions fell below unity (31 and 26 of 39). For C. albicans the mean shift was small and not significant (0.92, 22 of 39 below unity) because accessions responded in opposite directions. These conclusions held for every statistical approach applied (Table S2) and were unaffected by whether outliers were removed (Figure S2, Table S6).

2.2. The Response to Elicitation Is Accession Dependent and Identifies PHE as the Most Responsive Line

Behind the overall shift lay a wide spread of individual responses (Figure 2). For E. coli (Figure 2a) more than one third of the accessions responded significantly, all in the direction of stronger suppression (GR 0.66 to 0.92), and a single accession (TW) responded in the opposite direction. For S. aureus (Figure 2b) only PLK and NBP responded significantly and TW again reversed, which marks TW as a genotype in which elicitation weakens antibacterial activity. For C. albicans (Figure 2c) no accession reached significance after correction, yet the growth ratios spanned a six-fold range (0.37 in PHE to 2.18 in LPG), which shows that the absence of a mean effect reflects opposing responses rather than insensitivity.

Clustering of the three-organism profiles (Figure 3a) separated accessions that responded mainly for E. coli and C. albicans from those that responded for S. aureus, in which the yeast response was often reversed. Averaging the responses across organisms (Figure 3b) placed PHE clearly first (mean log₂GR −0.72, equivalent to 0.61-fold growth), followed by SPK, NPT, PLK, CMI and NPT2. Ranking by activity under NaCl alone (Figure 3c) also placed PHE first, followed by PLK, NPTT and SPK. The two rankings differed for accessions such as NPT and CMI, whose high ratio-based scores arose because their untreated extracts were strongly growth-promoting while their NaCl extracts were merely neutral. PHE was the only accession ranked first by both criteria, and its extract inhibited all three organisms after elicitation (66% inhibition of C. albicans and about 20% of each bacterium), so it was selected for the subsequent work.

2.3. Gram-Negative and Gram-Positive Responses Are Inversely Related

The three organisms did not respond to elicitation in parallel (Figure 4). Under untreated conditions (Figure 4a) accessions active against E. coli were also active against C. albicans (r = 0.55), whereas under NaCl (Figure 4b) the strongest association was between S. aureus and C. albicans (r = 0.62). The growth ratios (Figure 4c) showed the underlying structure. The E. coli and S. aureus responses were negatively correlated (Figure 4d, r = −0.42), the E. coli and C. albicans responses were positively correlated (Figure 4e, r = 0.44), and the S. aureus and C. albicans responses were again negatively correlated (Figure 4f, r = −0.38). Accessions whose elicited extracts gained activity against the Gram-negative bacterium therefore tended to lose activity against the Gram-positive one, which points to different induced constituents acting on the two bacteria.

2.4. Untargeted Metabolomics of Accession PHE: Nicotinamide Is the Most Strongly Induced Metabolite

Leaf extracts of untreated and NaCl-treated PHE plants (n = 3) were profiled by LC-HRMS in positive ionisation mode. A total of 783 features were annotated. A plausibility filter first removed annotations attributable to microbial, synthetic or animal sources (Table S13), and a stricter filter then retained features annotated as known plant or primary metabolites or matching a plant flavonoid database, leaving 126 plant-derived metabolites after removal of duplicate annotations. All results below refer to this set after probabilistic quotient normalisation (PQN).

Elicitation reorganised the metabolome extensively. The two conditions separated completely along the first principal component, which captured three quarters of the variance (Figure 5a), and about 30% of the metabolites changed by at least two-fold (37 of 126, 14 increased and 23 decreased, Figure 5b and Table S7). Nicotinamide, whose identity was confirmed with an authentic standard, showed by far the largest increase (9.3-fold), followed by the osmoprotectant glycine betaine (5.1-fold). Smaller increases occurred among phenolics, including the flavonol glucuronide miquelianin, a dicaffeoylquinic acid isomer and a pentahydroxyflavone isomer (2 to 3-fold each). The decreased group was dominated by nitrogen-containing primary metabolites and their derivatives, including pipecolic acid, N-acetylornithine, the dipeptide alanyl-proline and, with lower annotation confidence, aspartate and adenosine (0.16 to 0.47-fold), together with abscisic acid and raffinose. The heat map (Figure 5c) shows that these changes were consistent across replicates, and two thirds of them were also significant on raw, un-normalised peak areas. Several further features annotated as alkaloids of other plant families also changed, but because their spectral matches were weak or the compounds are not known from C. asiatica they are listed in Table S7 without interpretation.

2.5. Nicotinamide Becomes One of the Most Abundant Constituents of the Elicited Extract

Because a large fold change in a trace compound has little practical meaning, each metabolite was also expressed as its share of the total plant-derived signal (Figure 6). Most differential metabolites were minor constituents, but nicotinamide, glycine betaine, pipecolic acid and adenosine were exceptions, being both abundant and strongly changed (Figure 6a). At the class level (Figure 6b), elicitation increased the share of alkaloid and other nitrogen-containing metabolites (46% to 54% of the signal) at the expense of sugars, flavonoids and nucleoside-like compounds, and this class was also over-represented among the differential metabolites (odds ratio 3.4, Figure S4 and Table S8). Among the ten most abundant metabolites (Figure 6c), nicotinamide rose from a minor to a major constituent (1.4% to 10.3% of the signal) and glycine betaine became the single most abundant compound (3.8% to 14.9%), whereas pipecolic acid, adenosine and γ-aminobutyric acid, which had dominated the untreated extract, fell to a fraction of their former share. Elicitation thus concentrated the extract onto a few dominant compounds, and the ten most abundant metabolites accounted for a larger share of the signal after treatment (62% versus 57%).

2.6. Pathway View: NAD Salvage and Glycine Betaine Synthesis

Placing the changed metabolites on their pathways gave a coherent picture (Figure 7). In the NAD pathway (Figure 7a), the de novo precursor aspartate declined while the salvage-cycle intermediates nicotinamide and, to a lesser and non-significant extent, nicotinic acid rose, the pattern expected when NAD⁺ turnover accelerates and nicotinamide is released faster than it is recycled. In the glycine betaine pathway (Figure 7b), choline was unchanged while glycine betaine accumulated, the pattern produced by increased flux through choline monooxygenase and betaine aldehyde dehydrogenase. Phosphorylated intermediates and NAD⁺ itself are poorly retained under the reversed-phase conditions used and were not detected.

2.7. Kinetic Validation: The NaCl Extract Is a Stronger Bacterial Growth Inhibitor than the Untreated Extract or Nicotinamide Alone

Because the 4 h screen captures only an early effect on a dense inoculum, the PHE extracts and pure nicotinamide were compared over a full 24 h growth cycle at 10 mg/mL by optical density at 600 nm (OD₆₀₀), with growth summarised as the area under the curve (AUC) (Figure 8).

Against E. coli (Figure 8a,d) the three treatments produced markedly different growth profiles. The NaCl extract maintained E. coli growth close to baseline throughout the 24 h period, whereas the untreated extract permitted approximately half of the growth observed in the negative control. In contrast, nicotinamide exhibited a growth profile closely comparable to that of the negative control. Based on AUC analysis, the NaCl extract produced 88% growth inhibition, compared with 36% inhibition for the untreated extract, while nicotinamide showed essentially no inhibitory activity. Amoxicillin failed to inhibit the E. coli strain used in this study because the strain carries amoxicillin resistance. Therefore, amoxicillin should not be considered a valid positive control for this organism and is presented only for completeness.

Against S. aureus (Figure 8b,e), the NaCl extract again showed stronger antibacterial activity than the untreated extract, with 76% versus 59% inhibition based on AUC. Nicotinamide showed a distinct time-dependent effect. Growth initially tracked the negative control but plateaued after approximately 12 h, whereas the control continued to increase. Thus, nicotinamide produced only 11% inhibition based on AUC but reached 45% inhibition at 24 h. Amoxicillin completely suppressed S. aureus growth, confirming its effectiveness as a positive control. Against C. albicans (Figure 8c,f), both extracts almost completely suppressed growth, with activity comparable to ketoconazole. In contrast, nicotinamide caused OD₆₀₀ to fall below the initial baseline from approximately 12 h onward, suggesting loss of viable yeast biomass through cell lysis rather than simple growth inhibition.

2.8. The NaCl Extract and Nicotinamide Are Non-Cytotoxic, Stimulate HaCaT Proliferation and Accelerate Wound Closure

None of the treatments was cytotoxic to HaCaT keratinocytes at any concentration up to 500 µg/mL in the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Figure 9a). Instead, the NaCl extract stimulated the cells in a bell-shaped manner, with viability rising to a plateau of about 180% of untreated cells at 100 to 250 µg/mL before declining toward baseline at the highest dose. The untreated extract showed the same shape at a lower amplitude (about 140%) that did not remain significant after correction, and nicotinamide had no effect at any concentration. At the peak concentrations the NaCl extract exceeded both nicotinamide and the untreated extract (Table S11).

In the scratch assay, treatment with the NaCl extract at 100 µg/mL, the concentration showing the strongest effect, significantly improved wound closure. Wound closure reached 84% of the positive control, compared with 55% in the negative control, and was not significantly different from the positive control (Figure 9b). The untreated extract showed only a small and non-significant improvement, reaching 67%. Nicotinamide showed a similar effect to the NaCl extract, with 83% wound closure, and both treatments were more effective than the untreated extract. The representative images (Figure 9c) show a smaller remaining gap after 24 h in the NaCl extract and nicotinamide groups, compared with the wider gaps observed in the negative control and untreated extract groups.

3. Discussion

3.1. Salt Elicitation Converts Growth-Promoting Extracts into Inhibitory Ones in a Genotype-Dependent Way

The clearest effect of elicitation across the germplasm was not the appearance of a strong new antimicrobial activity but the disappearance of an old one, the tendency of untreated extracts to feed microbial growth. Methanol extracts of unstressed leaves carry a large load of sugars, free amino acids and nucleosides. When such an extract is added to a rich medium, it acts as a supplement rather than an inhibitor, and a dense inoculum grows faster than the vehicle control. Salt stress reversed this balance. Nutrient-like metabolites were drawn down, nitrogen-containing secondary metabolites accumulated, and the extract shifted from promoting to restraining bacterial growth. The average shift was modest, about one tenth of the untreated growth level, but in the most responsive accessions it reached one third, and the effect for E. coli was large by conventional standards and independent of how the data were treated.

The screening also makes clear why one accession cannot stand for the species. Responses ranged from strong suppression to outright reversal, and in one accession elicitation weakened activity against both bacteria. For C. albicans the responsive and non-responsive genotypes cancelled each other, so that no overall effect was visible even though a strong anti-Candida response emerged in the most responsive line. This picture matches the genotype-dependent phytochemical response of the same panel to short-term salinity [11] and the chemotype diversity of Thai C. asiatica [15]. Elicitation protocols for medicinal plants should therefore be developed on selected genotypes rather than on the species as a whole.

An unexpected feature of the screening was that the E. coli and S. aureus responses moved in opposite directions. Gram-negative and Gram-positive bacteria differ in their outer permeability barrier and in the classes of plant compounds that reach their targets [16]. A shift in the balance of induced metabolites within an accession could therefore favour activity against one group at the expense of the other. This pattern argues against a single broad-spectrum active principle and points instead to a mixture of constituents whose proportions vary among genotypes. The kinetic experiments with PHE support this interpretation.

3.2. Nicotinamide Dominates a Nitrogen-Centred Metabolic Response

The metabolome of the PHE accession responded to NaCl in the way expected of a salt-stressed glycophyte. Glycine betaine accumulated while its precursor choline did not, the signature of increased flux through choline monooxygenase and betaine aldehyde dehydrogenase [12,17]. Glycine betaine stabilises proteins and membranes and supports osmotic adjustment [18], and its accumulation is associated with salinity resistance in other species [19]. Its role is therefore protective for the plant rather than antimicrobial. Bacteria themselves take up glycine betaine as an osmoprotectant [20], so it was not pursued as a candidate for the bioactivities described here.

Nicotinamide, whose identity was confirmed against an authentic standard, was the most strongly induced metabolite, and its rise was mirrored by a fall in aspartate, the starting substrate of de novo NAD synthesis in plants. Together these changes point to accelerated NAD⁺ turnover. Stress-activated NAD⁺-consuming enzymes such as poly(ADP-ribose) polymerases and sirtuins release nicotinamide, which is recycled through the salvage pathway or methylated to trigonelline [21,22]. Nicotinamide has long been regarded as an early signal of oxidative stress in plant cells and accumulates under UV-B and oxidative stress [23,24], and exogenous nicotinamide exerts antioxidative and DNA-hypomethylating effects in plant cells [25]. The parallel loss of adenosine, pipecolic acid, N-acetylornithine and dipeptides fits the remobilisation of nitrogen from amino acid and nucleoside pools that accompanies salt acclimatisation [13]. The elicited metabolome is thus one in which nitrogen has been redirected from storage and transport forms toward a small number of protective compounds.

Viewing the same changes as shares of the extract, rather than as fold changes, gave them a practical meaning. Fold changes identify regulated metabolites, but an ingredient for a topical product is defined by its composition. After elicitation, nicotinamide alone accounted for about one tenth of the plant-derived signal and, together with glycine betaine, for about one quarter. The extract had become dominated by a few compounds. The enrichment of a well-characterised, water-soluble and pharmaceutically accepted molecule such as nicotinamide is a valuable outcome in itself, independent of the more exotic secondary metabolites. Two points temper this conclusion. Peak area is not molar abundance, because ionisation efficiency differs between compounds, and the total plant-derived signal fell under NaCl before normalisation, which may reflect a lower extractable metabolite content per unit dry mass or a matrix effect. The shares therefore describe relative composition rather than absolute yield, and targeted quantification of nicotinamide will be needed before the extract can be standardised.

3.3. Nicotinamide Contributes to the Antibacterial Gain, and Other Induced Constituents Reinforce It

The kinetic experiments answered the question raised by the metabolomics, namely how much of the enhanced activity nicotinamide alone can explain. Nicotinamide is an antimicrobial agent in its own right. It arrests the growth of E. coli, S. aureus, Pseudomonas aeruginosa and C. albicans by binding DNA and blocking cell division [26]. The concentration needed for complete inhibition is high, however, and it differs between species, being 30 mg/mL for E. coli and 25 mg/mL for S. aureus. Against C. albicans the compound is far more potent in other test systems, with a half-maximal inhibitory concentration of about 2.4 mg/mL. Its antifungal action involves cell wall remodelling and inhibition of the NAD⁺-dependent histone deacetylase Hst3, and it is active against fluconazole-resistant isolates and in a mouse model of candidiasis [27,28].

The concentration used here, 10 mg/mL, therefore lay below the threshold for complete bacterial inhibition but several-fold above the anti-Candida threshold, and the results followed these thresholds closely. Nicotinamide drove the yeast OD below its starting value, which indicates lysis rather than growth arrest. Against the bacteria its effect was partial and delayed. S. aureus grew normally at first and then levelled off in the second half of the growth cycle, whereas E. coli was essentially unaffected. Nicotinamide therefore does restrain the bacteria, but only late and only partly, exactly as expected of a sub-inhibitory dose.

The NaCl extract behaved very differently. It contains nicotinamide at only a fraction of the concentration applied as pure compound, yet it suppressed the growth of both bacteria throughout the growth cycle, and it did so far more strongly than the untreated extract. The elicited extract must therefore contain additional antibacterial constituents that act alongside nicotinamide, and their combined effect exceeds that of nicotinamide alone. Such reinforcement is common in plant extracts, in which different constituents act on different targets, improve one another’s uptake, or block efflux and resistance mechanisms [29]. Alkaloid-type compounds were enriched among the induced metabolites, and this class is a well-documented source of antibacterial, antibiotic-enhancing and antivirulence activity [30]. The flavonol glucuronide and the dicaffeoylquinic acid that increased are also antibacterial in other systems [31]. C. asiatica extracts already inhibit Gram-negative and Gram-positive bacteria through triterpenoid acids and phenolics [5,32], and elicitation appears to add nicotinamide and a set of nitrogen-containing metabolites to this existing repertoire. We refrain from naming specific alkaloids as active principles, because their annotations rest on spectral similarity to compounds not known from C. asiatica. Bioassay-guided fractionation of the elicited extract, and tests of defined mixtures of nicotinamide with the induced phenolics, are the logical next steps.

The two antimicrobial assays also illuminate each other. In the 24 h assay the untreated PHE extract was clearly active, whereas in the 4 h screening the same extract appeared neutral. The screening measures an early effect on a dense inoculum and is designed to rank many genotypes quickly, while the kinetic assay uses a lower inoculum and follows a complete growth cycle, so it captures delayed and cumulative effects. The two are therefore complementary rather than contradictory. The growth of E. coli in the presence of amoxicillin reflects the resistance of the strain used and does not affect the comparison between extracts.

3.4. The Elicited Extract Stimulates Keratinocytes, and Nicotinamide Accounts for Most of the Wound-Healing Effect

Neither extract nor nicotinamide was cytotoxic to HaCaT cells at any concentration tested, which places the elicited extract comfortably within the safety range required for topical use. Instead of reducing viability, the NaCl extract raised MTT reduction well above the untreated level in a bell-shaped manner, with a peak at intermediate concentrations. MTT reduction reflects mitochondrial and cytosolic reductase activity as well as cell number [33], so part of this signal may represent metabolic activation rather than proliferation, and a direct cell count or DNA-based assay would separate the two. Even so, a stimulation of this size in serum-free medium indicates a strong trophic effect. Nicotinamide alone did not increase viability, so this trophic effect comes from other constituents of the elicited extract. Asiaticoside and madecassoside, the classical wound-healing triterpenoids of C. asiatica, stimulate keratinocyte and fibroblast growth and migration [34,35] and are plausible contributors, although they were not among the plant-derived features that changed here and their levels in this accession remain to be quantified.

In the scratch assay the elicited extract accelerated wound closure to a level that no longer differed from the positive control, whereas the untreated extract produced only a small improvement. Nicotinamide alone matched the elicited extract almost exactly (83% versus 84% of the positive control). This close correspondence identifies nicotinamide as the main driver of the enhanced wound closure. Nicotinamide is an established dermatological active. It improves barrier function by increasing ceramide and other stratum corneum lipid synthesis in keratinocytes [36], improves the appearance of ageing skin in clinical use [37], promotes fibroblast migration, proliferation and collagen synthesis in vitro [38], and accelerates epithelialisation and tissue regeneration in full-thickness wounds [39]. It also has a favourable safety profile [40]. Because the elicited extract contains nicotinamide at only a fraction of the concentration applied as pure compound, other constituents must reinforce its effect, and the untreated extract, which lacks the nicotinamide enrichment, was correspondingly weaker. Glycine betaine is a humectant used in skin care and may add to the cosmetic value of the extract, but it was not tested separately here.

Taken together, the two functional assays show that nicotinamide contributes to both gains produced by elicitation, but to different extents. The wound-healing gain is largely reproduced by nicotinamide alone, whereas the antibacterial gain depends on nicotinamide acting together with other induced constituents. This distinction matters for product development. The elicited extract cannot be replaced by nicotinamide alone without losing part of the antibacterial component, and the two activities could be optimised somewhat independently, for example by combining elicitation with the selection of accessions that respond strongly in the antibacterial screening.

3.5. Practical Implications

A 10-day treatment with 100 mM NaCl in a sub-irrigation system is simple and inexpensive, and it is compatible with the hydroponic and soil systems already used for C. asiatica [14]. The present data suggest that such a treatment can be used deliberately to raise the nicotinamide content and the antibacterial and keratinocyte-stimulating activity of the harvested leaves. Two cautions apply. Salt reduces biomass in this species even at tolerable concentrations [9], so the gain in activity per gram must be weighed against the loss of yield, and residual sodium in the raw material may need to be controlled for some applications. Genotype selection is the second lever. Because responses ranged from strong suppression to reversal, an elicitation programme should begin with a screening of the available germplasm, as done here. PHE is the primary candidate from this panel, and SPK and PLK, which ranked next by both criteria, are the natural accessions to examine next.

3.6. Limitations and Outlook

Some aspects of this work define the scope of the conclusions and point to useful next steps. The identity of nicotinamide was confirmed with an authentic standard, whereas the other metabolites discussed are putatively annotated by spectral and formula matching, which is the usual level of confidence in untargeted profiling [41,42]. Targeted quantification of nicotinamide in the extracts would allow the elicited raw material to be standardised. The metabolomic comparison was made on one accession selected for its strong response, so the metabolic signature of less responsive accessions remains to be described, and the relative composition reported here complements rather than replaces absolute yields. The 4 h screening was designed for throughput across 39 accessions and was complemented by the 24 h kinetic assay for the selected accession, and minimum inhibitory concentrations of the elicited extract are a natural next measurement. The functional assays used a single keratinocyte line and one concentration for the scratch assay, and confirmation in primary keratinocytes and in vivo wound models would strengthen the translational case. Finally, the contribution of individual induced metabolites was tested with one pure compound, and bioassay-guided fractionation will show which additional constituents act with nicotinamide.

4. Materials and Methods

4.1. Plant Material and Cultivation

Thirty-nine accessions of C. asiatica (L.) Urb. collected from different regions of Thailand and maintained as living germplasm at Phichit Agricultural Research and Development Center [11] were propagated from stolon explants in pots of homogeneously mixed commercial potting soil, with the stolon placed at three quarters of the pot height (Table S1). Pots were grown in a modified sub-irrigation system in which each pot stood in a container of half-strength Hoagland solution kept at about one quarter of the pot height, and plants were raised for 40 days under greenhouse conditions before treatment. Plants of uniform size and developmental stage were then selected to reduce physiological variation among replicates.

4.2. NaCl Treatment

The selected plants were transferred to a controlled-environment growth chamber (30 °C, 50 to 70% relative humidity, 12 h light and 12 h dark, 300 µmol m⁻² s⁻¹ photosynthetic photon flux density from plant-growing LED lamps). NaCl-treated plants received nutrient solution containing 100 mM NaCl for 10 days, and untreated plants received nutrient solution without added salt. The pH of the nutrient solution was adjusted to 5.8 daily. The experiment followed a completely randomised design with five biological replicates per accession and treatment, and treated and untreated plants were harvested at the same time.

4.3. Extraction

The youngest fully expanded leaves were dried in a hot-air oven at 40 °C for 24 h and ground to a fine powder. Powder (50 mg) was extracted with 1 mL methanol for 48 h, and the extract was used for the antimicrobial screen and for metabolomics. For the kinetic, cell viability and wound-healing assays, extracts of accession PHE were dried using a rotary evaporator and redissolved to the required concentrations in DMSO.

4.4. Antimicrobial Screen at 4 h

The microbial strains Escherichia coli, Staphylococcus aureus, and Candida albicans were obtained from the Department of Microbiology, Faculty of Medical Science, Naresuan University which sourced the strains from the National Institute of Health of Thailand. A 100 µL portion of each methanol extract (equivalent to 5 mg leaf powder) was transferred to a well of a 96-well plate and the solvent was evaporated under a stream of nitrogen. Overnight cultures in nutrient broth were adjusted to OD₆₀₀ = 0.1, and 200 µL of the suspension was added to each well and mixed with the extract residue. Plates were incubated at 37 °C and OD₆₀₀ was read at 0 and 4 h. Growth inhibition (%) was calculated as 100 × (1 − ΔOD_extract / ΔOD_control), where ΔOD is the change in OD₆₀₀ between 0 and 4 h and the control contained microorganisms without extract, so that negative values indicate growth stimulation. Each accession and treatment was tested with five biological replicates per organism.

4.5. Growth Ratio and Statistics for the Screen

Within each accession, treatment and organism, replicate values were screened for single outliers with Grubbs’ test (α = 0.05, at most two iterations) [43], which removed one value in 19 of 234 groups. Relative growth was defined as G = 1 − inhibition / 100, with a floor of 0.02 to avoid zero or negative values, and the growth ratio as GR = mean G(NaCl) / mean G(untreated) for each accession and organism. Effects across accessions were tested on log₂GR (n = 39) with a one-sample t-test and a Wilcoxon signed-rank test, summarised as the geometric mean GR with 95% CI and Hedges’ g [44], and confirmed with a linear mixed model of replicate-level log₂G with treatment as fixed effect and accession as random effect. For each accession, a 95% CI of GR was obtained by bootstrap (2000 resamples) [45] and the treatment effect was tested with Welch’s t-test on log₂G [46] with Benjamini and Hochberg correction within organism [47]. Condition means were compared with zero by one-sample t-tests and between conditions by paired t-tests on accession means. The ratio-based index was the mean log₂GR over the three organisms and the NaCl-condition index the mean z-score of % inhibition under NaCl. Correlations were Pearson’s r. The influence of replicate handling was assessed by repeating the analysis on all five replicates (Table S6).

4.6. Untargeted LC-HRMS Metabolomics

Methanol extracts of untreated and NaCl-treated PHE leaves (three biological replicates per condition, prepared as in 4.3) were analysed on an UltiMate 3000 UHPLC system coupled to a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific). Separation used a Hypersil GOLD aQ C18 column (2.1 × 100 mm, 1.9 µm) held at 60 °C with 0.1% formic acid in water (A) and in acetonitrile (B) at 0.325 mL/min. The gradient held 5% B for 2 min, rose linearly to 95% B over 25 min, held for 2 min and re-equilibrated, for a total run of 30 min. The injection volume was 2 µL and the autosampler was kept at 8 °C. Blank injections of 0.1% formic acid in water followed each sample to limit carry-over. Electrospray ionisation was operated in positive mode at 3.5 kV with the capillary at 325 °C. Acquisition alternated full-scan MS (m/z 75 to 800, resolution 180,000) with data-dependent MS/MS (resolution 22,500) using dynamic exclusion, under Xcalibur 3.1.

Raw files were processed in Compound Discoverer 3.3 (Thermo Fisher Scientific). Retention times were aligned with the ChromAlign algorithm [48] with a tolerance of 0.5 min and a mass tolerance of 2 ppm. Features were extracted with predefined adduct rules, a mass deviation of 2 ppm, 30% intensity tolerance and a signal-to-noise threshold of 2, and quantified by peak area. Molecular formulas were predicted from accurate mass, and compounds were annotated by MS/MS spectral matching against mzCloud and by mass and formula matching against ChemSpider, KEGG and in-house lists. Compounds were retained if their original peak rating was at least 2 in at least three samples. Gap filling searched for missing features within 5 ppm and kept centroid peaks with signal-to-noise ≥ 2. Instrumental drift was corrected with the SERRF procedure using a random-forest model with 200 trees [49]. The export used for statistics contained 783 annotated features with name, formula, calculated mass, m/z, retention time, mzCloud score and peak areas for the six samples.

4.7. Metabolomics Data Processing and Statistics

Annotations were screened for biological plausibility before analysis. Features annotated as microbial natural products, synthetic drugs, industrial chemicals, animal-specific metabolites or catalogue identifiers without a plant identity were removed, with the reason recorded for each feature (Table S13), leaving 421 plant-plausible features. A stricter set of 129 features annotated as known plant or primary metabolites or matching the Arita laboratory flavonoid database [50] was then selected, and three pairs of features with identical peak areas across all samples (isobaric duplicate annotations) were merged, keeping the annotation with the higher mzCloud score, to give 126 metabolites. Peak areas were normalised by probabilistic quotient normalisation [51] using the median sample as reference and log₂ transformed. Differences between conditions were tested with Welch’s t-test and corrected by the Benjamini and Hochberg procedure, and metabolites with |log₂FC| ≥ 1 and FDR < 0.05 were considered differential. The same tests were repeated on raw peak areas, and metabolites significant in both analyses were flagged as robust. PCA was performed on autoscaled log₂ intensities. Metabolites were assigned to chemical classes from their formulas and names, and class enrichment among differential metabolites was tested with one-sided Fisher’s exact tests with FDR correction. The share of each metabolite was calculated from raw peak areas as a percentage of the summed area of the 126 metabolites in each sample. The identity of nicotinamide was confirmed by comparison of retention time, accurate mass and MS/MS spectrum with an authentic standard (Tokyo Chemical Industry Co., Ltd., HPLC grade). Other annotations correspond to confidence levels 2 and 3 [41,42].

4.8. Kinetic Antimicrobial Assay

Each of the three microorganisms was pre-cultured for 24 h under its respective growth conditions. Subsequently, 1% (v/v) of each stock culture was inoculated into freshly prepared nutrient broth. For the assay, 990 µL of the inoculated medium was dispensed into sterile microtubes and supplemented with 10 µL of either the extract or nicotinamide stock solution, resulting in a final concentration of 10 mg/mL. The negative control consisted of inoculated nutrient broth without treatment, while the blank contained the corresponding test substance in the absence of microorganisms. Amoxicillin (20 µg/mL) and ketoconazole (50 µg/mL) were included as positive controls for the bacterial and yeast strains, respectively. Mixtures (200 µL per well) were transferred to 96-well plates and OD₆₀₀ was recorded hourly from 0 to 24 h in a multi-detection microplate reader at 37 °C. Three replicates were run per treatment. Growth was expressed as % change in blank-corrected OD₆₀₀ from 0 h, the AUC from 0 to 24 h was calculated by the trapezoidal rule [52], and inhibition was calculated as 100 × (1 − AUC_treatment / AUC_negative control). Treatments were compared by one-way ANOVA with Tukey’s HSD test [53] and pairwise with Welch’s t-test.

4.9. HaCaT Culture and MTT Assay

Epidermal keratinocyte (HaCaT) cells were cultured in DMEM/F-12 medium supplemented with 10% FBS, penicillin 100 U/mL and streptomycin 0.1 mg/mL at 37 °C in 5% CO₂ incubator. Cells were seeded at 10,000 per well in 96-well plates and incubated for 24 h. The medium was then replaced with serum-free DMEM/F-12 containing the extracts or nicotinamide at final concentrations of 0, 10, 25, 50, 100, 250 and 500 µg/mL with the DMSO concentration fixed at 0.5% in all wells, including the untreated control. After 24 h, 10 µL of 5 mg/mL MTT was added for the final 2 h of incubation, the formazan was dissolved in DMSO (200 µL) and absorbance was read at 540 nm [33]. Viability was expressed as % of the untreated control, and a value below 70% was taken as cytotoxic [54]. Three independent experiments were performed. Viability at each concentration was compared with 100% by one-sample t-tests with FDR correction within each agent, and agents were compared at each concentration by ANOVA with Tukey’s HSD.

4.10. Scratch Wound-Healing Assay

HaCaT cells were seeded at a density of 150,000 cells per well in 24-well plates and cultured for 24 h until a confluent monolayer. A linear scratch was then generated across the cell monolayer using a sterile P200 pipette tip. Detached cells and debris were removed by washing the wells twice with 1 mL PBS. Cells were subsequently cultured in serum-free medium supplemented with the respective extracts or nicotinamide at 100 µg/mL. Medium containing 2% fetal bovine serum (FBS) was used as the positive control, whereas serum-free medium alone served as the negative control. Images of the scratched areas were captured immediately after scratching (0 h) and following 24 h of treatment using an inverted phase-contrast microscope. The cell-free area was quantified using ImageJ software [55]. Wound closure was calculated as the percentage reduction of the cell-free area from 0 to 24 h and expressed relative to the positive control (set to 100%). Six replicates were analysed per group. Groups were compared by one-way ANOVA with Tukey’s HSD and by Welch’s t-test against the negative control.

4.11. Software, Generative AI Use and Data Availability

Analyses were performed in Python 3.11 with pandas 2.2, NumPy 1.26, SciPy 1.13, statsmodels 0.14 and scikit-learn 1.4, and figures were drawn with Matplotlib 3.9. Compound Discoverer 3.3 (Thermo Fisher Scientific, Waltham, MA, USA) and ImageJ version 1.54m (National Institutes of Health, Bethesda, MD, USA) were used as described above. During the preparation of this study, the authors used Claude (Anthropic, San Francisco, CA, USA) to generate code for data visualizations and to improve the clarity of the writing. The authors have reviewed and edited all output and take full responsibility for the content of this publication.

5. Conclusions

A 10-day treatment with 100 mM NaCl changed the bioactivity of C. asiatica leaf extracts in a way that was consistent in direction but strongly genotype dependent in magnitude. Across 39 Thai accessions the elicited extracts suppressed bacterial growth more than the untreated extracts, and the most responsive accession, PHE, gained activity against all three test organisms. In this accession, elicitation raised nicotinamide to one tenth of the plant-derived metabolome, together with the osmoprotectant glycine betaine and a broader reorganisation of nitrogen and NAD metabolism. Functional tests showed that nicotinamide contributes to both bioactivities of the elicited extract but to different extents. Its enhanced keratinocyte-stimulating and wound-closing activity was matched by nicotinamide alone. Its enhanced antibacterial activity exceeded what nicotinamide produced at the same concentration, which lies below the species-specific inhibitory thresholds of the compound, and therefore reflects nicotinamide acting together with other induced constituents. Salt elicitation combined with accession selection is thus a practical route to C. asiatica raw material with dual antimicrobial and wound-healing properties, and confirmation of the active constituents by targeted quantification and fractionation is the next step.