Arabidopsis thaliana (Col-0) wild type (WT) and mutants including fls2 (salk_062054), efr-2 (salk_068675), bak1 (salk_116202), bik1 (salk_005291), rbohd (salk_083046), pub12/13 (Zhou et al., 2018), and pub25/26 (Wang et al., 2018) were used. Ten-day-old seedlings grown on soil were transferred to a growth room at 70% humidity and 21°C with a 10-h-light/14-h-dark photoperiod (short-day condition). Light intensity was 100 μmol/m²s⁻¹. Five- to six-week-old plants with about 17–18 leaves were used. Flg22 contains the 22-amino acid sequence QRLSTGSRINSAKDDAAGLQIA corresponding to a conserved motif of the amino terminus of flagellin from Pseudomonas species. Elf18 contains the sequence AcSKEKFERTKPHVNVGTIG corresponding to the conserved N-terminal 18-amino acid peptide of the bacterial protein EF-Tu. These peptides (Axil Scientific Pte Ltd., Singapore Science Park II, Singapore) were dissolved in sterile H₂O to a concentration of 5 mM, and aliquots of 100 μM stock solution were stored at −20°C. 3-hydroxydecanoic acid (Sigma, St. Louis, MO, USA, H3648) (Kutschera et al., 2019) was dissolved in ethanol to a concentration of 5 mM, and Chitin, GlcNacβ1-4[GlcNacβ1-4GlcNAc]5GlcNAc (ELICITYL, Crolles, France, GLU437) was dissolved in water to a concentration of 5 mM. Aliquots of 100 μM stock solutions were stored at −20°C.

An overnight culture of Pseudomonas syringae pv. Tomato DC3000 (Pst DC3000) (Rufián et al., 2019) was suspended in 10 mM MgCl₂ to approximately 5 × 10⁵ cfu/ml (OD₆₀₀ = 0.001). About 50 μl of either the bacterial suspension or 10 mM MgCl₂ (mock) was infiltrated on the abaxial side of Arabidopsis leaf by a needleless syringe. Commercial Choy Sum (B. rapa v. parachinensis) seeds were grown in growth chambers at 25°C under a 16-h-ligh/8-h-dark photoperiod with 75% humidity. Two-week-old Choy Sum plants were inoculated with the Xanthomonas campestris py. campestris (Xcc) isolate (ATCC33913). The Xcc strain was grown in yeast extract, glucose, and calcium carbonate (YGC) medium at 28°C. Harvested bacterial cells were resuspended in YGC medium to 1 × 10⁸ cfu/ml. Using a needleless syringe, the abaxial side of the fourth leaf of each plant was inoculated with 200 μl of Xcc suspension on both sides of the central leaf vein near the leaf margins. The leaves of separate plants inoculated with 200 μl of sterile media served as the mock control.

The Raman spectroscopy system with an 830 nm excitation wavelength was previously described (Huang et al., 2020). The Raman shift axis was calibrated using a polystyrene sample (Creely et al., 2005), and the Raman spectra were recorded in the range of 400–1,700 cm⁻¹. The 830 nm excitation wavelength has the advantage of deeper penetration depth and generating a comparatively lower fluorescence background signal from the leaf tissue, resulting in spectra with a higher signal-to-noise ratio (SNR). The preprocessing of the Raman spectra consisted of cosmic ray removal and smoothing with the Savitzky–Golay filter function (MATLAB, Inc., Natick, MA, USA) with the fifth-order polynomial and frame length of 11. The Raman spectra presented in this study were obtained after removing any residual fluorescence background using a polynomial subtraction method with nonnegative constraints (Lieber and Mahadevan-Jansen, 2003).

For Arabidopsis, two plants from each independent biological sample were investigated (Huang et al., 2020). The seventh or eighth leaf of 5–6-week-old plants was used for infiltrations (Figure 1). At the designated time after infiltrations, leaf discs (6 mm diameter) were excised with a puncher from the regions proximal to the infiltrated sites (Figure 2A). Two leaf discs were obtained from each elicitor-treated leaf, one from the left and the other from the right side of the central leaf vein. A total of four leaf discs were analyzed for each genotype per experiment. The leaf discs were placed with a drop of water between a fused silica slide of thickness 0.5 mm and a fused silica coverslip, which was 100 μm thick. Three locations were measured in each leaf disc, with five Raman spectra acquired per location and an exposure time of 10 s per spectrum. Thus, an average Raman spectrum of an independent biological sample was obtained from the mean of a total of 60 Raman spectra. The experiments were repeated with three to four independent biological samples.

For Raman spectral acquisitions in Choy Sum, the two leaf discs from a single leaf were excised from the proximal regions (PRs) of the infiltrated site at 24 and 48 h after infiltration. Three locations were measured in each leaf disc, with five Raman spectra acquired per location and an exposure time of 10 s per spectrum. Thus, an average Raman spectrum of one independent biological sample for each time point was obtained from the mean of a total of 30 Raman spectra. Three biological replicates for each time point were performed for both Xcc and mock infiltration.

After mock or elicitor treatments, the leaf samples were collected at time periods indicated in each experiment (0.5, 1, 3, 6, and 24 h) and immediately frozen in liquid nitrogen. Total RNA was extracted from the entire infiltrated leaf using the RNeasy plant mini kit (74104; Qiagen, Hilden, Germany) with DNase I (79254; Qiagen, Hilden, Germany) treatment. About 1 μg of RNA was used for complementary DNA (cDNA) synthesis, and reverse transcription-quantitative PCRs (RT-qPCRs) were performed as described (Park et al., 2018) on the following genes. Immune response genes—WRKY53 (At4g23810), early senescence WRKY networking regulator; WRKY29 (At4g23550); WRKY30 (At5g24110); ZAT12 (At5g59820), ZINC FINGER OF ARABIDOPSIS THALIANA12; PR1 (At2g14610), PATHOGENESIS-RELATED GENE1; PDF1.2 (At5g44420), PLANT DEFENSIN; MPK3 (At3g45640), Mitogen-activated Protein Kinase3; SAND (At2g28390), MONENSIN SENSITIVITY1 (MON1) a zinc finger protein, PMR4 (At4g03550), PAD3 (At3g26830), MPK3 (At3g45640), and MYB51 (At1g18570). Tryptophan-metabolism related genes. MYB122 (At1g74080), MYB51 (At1g18570), MYB34 (At5g60890), CYP79B2 (At4g39950), CYP79B3 (At2g22330), CYP83B1 (At4g31500), CYP71A13 (At2g30770), CYP71B15/pad3 (At3g26830) Carotenoid biosynthetic genes-PSY, phytoene synthase (At5g17230); PDS, phytoene desaturase (At4g14210); Z-ISO, z-carotene isomerase (At1g10830); ZDS, z-carotene desaturase (At3g04870); CRTISO, carotenoid isomerase (At1g06820); εLCY, Lycopene ε-cyclase (LUT2, At5g57030); βLCY, Lycopene β-cyclase (At3g10230); CYP97A3, (LUT5, At1g31800); CYP97C1, (LUT1, At3g53130); CYP97B3, (At4g15110); BCH1, β-carotene hydroxylase 1 (At4g25700); BCH2, β-carotene hydroxylase 2 (At5g52570); CCS, COPPER CHAPERONE FOR SOD1(At1g12520); ZEP, Zeaxanthin epoxidase (ABA1, At5g67030); NPQ, NON-PHOTOCHEMICAL QUENCHING1 (At1g08550), NXS, neoxanthin synthase (ABA4, At1g67080); NCED3, 9-cis-epoxycarotenoid dioxygenase (At3g14440); ABA2, ABA DEFICIENT 2 (At1g52340); AAO3, ABSCISIC ALDEHYDE OXIDASE3 (At2g27150); CCD1, carotenoid cleavage dioxygenase1 (At3g63520); CCD4, At4g19170; CCD7, At2g44990; CCD8, At4g32810; D27, β-carotene isomerase (At1g64680); CYP711A, (At2g26170). The expression of the Arabidopsis Act2 (At3g18780) was used as an internal control. Primer pairs for RT-qPCR were designed using Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi) or manually by eye. All RT-qPCR primers are listed in Supplementary Table 1.

The callose assay was performed as described by Ton et al. (2005). Callose-mediated fluorescence was visualized using a DAPI filter set (an excitation filter 365 nm; a dichroic mirror 395 nm; and an emission filter 397 nm) of an Axio Imager M2 microscope (Zeiss, Oberkochen, Germany). Images collected with a photometric CoolSNAP HQ2 camera system (Roper Scientific Germany, Planegg, Germany) were acquired by an x100 lens using the MetaMorph software (Version 7.7.0.0; Molecular Devices, Sunnyvale, CA, USA). Callose foci within the frame of a single image (magnification ×100; 5,655 μm² per region) were quantified using the ImageJ software (http://rsb.info.nih.gov/ij/, Schneider et al., 2012). For each experiment, the number of callose deposits was measured in two plants with about four to five leaf regions surveyed per plant. This provided about 8–10 data points per experiment. The experiment was repeated with three independent biological samples. For Choy Sum, callose deposition was visualized in three independent biological samples at 24 and 48 hpi.

Total carotenoids were extracted and analyzed as described with minor modifications (Sumanta et al., 2014). Briefly, 30 mg of the fresh leaf sample was homogenized in a tissue homogenizer with 600 μl of methanol solvent. After centrifugation at 10,000 × g for 15 min at 4°C, 0.1 ml of the supernatant was mixed with 0.9 ml of methanol. The mixture was analyzed for photosynthetic pigment content by a spectrophotometer (TECAN, Männedorf, Switzerland, The spark multimode microplate reader) according to Sumanta et al. (2014). For each experiment, the quantification of total carotenoids was measured in three to five infiltrated plants per experiment. The quantification of carotenoids was expressed as the mean value of 15–20 different leaf extracts in four replicated experiments.

We used flg22 and elf18 to induce plant innate immunity responses without related pathogenic symptoms (Bektas and Eulgem, 2015). To reduce variability between biological samples, Arabidopsis WT (Col-0) and mutant plants were grown under short-day conditions for about 5–6 weeks with about 17–18 leaves (Farmer et al., 2013; Kurenda et al., 2019) (Figure 1a). The abaxial side of leaf seventh or eighth was infiltrated with a needleless syringe containing (50 μl) water, 1.0 μM flg22, or 1.0 μM elf18 (Figure 1b). At 24 hours post-infiltration (hpi), no pathological symptoms were seen at the infiltration sites in response to the PAMP nor the slight wounding caused by the syringe infiltration (Figure 1c). However, compared to mock infiltration, the infiltration of either induced the expression of PTI-specific marker genes but with different kinetics (Supplementary Figure 1). Some early responsive genes (WRKY30, MPK3, MYB51, CYP79B2, etc.) showed increased expression as early as 3 hpi, whereas the expression of other genes (PR1, PDF1.2, ZAT12, and MYB122) increased only at 6 hpi. The expression of most activated genes returned to basal levels by 24 h (Supplementary Figure 1) although several genes (PR1, WRYK29, and WRKY53) still maintained high expression levels at this time point. One day after infiltration, callose deposition in the elicitor-infiltrated leaf was more significantly induced than that in the mock-infiltrated control leaf (Figure 1d), which was consistent with the increased expression of genes related to callose synthesis (Supplementary Figure 1). The gene expression results and the callose deposition data together confirmed that, under our experimental conditions, both flg22 and elf18 induced a rapid and robust innate immunity response in the infiltrated leaves.

We used a custom-built near IR excitation Raman instrument (Figure 1e) to measure the Raman spectra from mock- or elicitor-infiltrated leaf samples (Figure 2). To quantify the PTI response using Raman spectroscopy, we developed an ERI as described as follows (Supplementary Figure 2):

Because a robust induction of PTI marker genes was seen with 1 μM flg22 or elf18 (Seo et al., 2019), we use this elicitor concentration to perform a time-course experiment on leaf regions PR or distal (DR) from the infiltrated sites (Figures 2A–C). For flg22, the highest Raman shift spectrum and a high ERI value were obtained at PR 24 hpi. The ERI value decreased to almost zero at 48 and 72 hpi (Figures 2B,D). Low ERI values were obtained in DR, which gradually increased with time at 48–72 hpi indicating signal transmission from PR to DR. For elf18, Raman shift signals were also detected at 24 hpi in PR with high ERI values, which decreased to almost zero at 48–72 hpi. A lower ERI value was seen in DR (Figures 2C,D). Based on these results, for subsequent experiments, we chose to analyze the Raman spectra at PR 24 hpi with an elicitor concentration of 1.0 μM.

To confirm that the Raman spectra changes obtained by the elicitors were indeed caused by PTI signaling, we first analyzed the Raman spectra of a well-characterized mutant of the PTI pathway. Arabidopsis plants possess two PAMP receptors. The FLS2 receptor specifically recognizes flg22, whereas the EFR receptor is specific for elf18 (Chinchilla et al., 2006; Zipfel et al., 2006). Figure 3 shows the response of WT plants to both flg22 and elf18 with similar ERI values. By contrast, the fls2 mutant was refractory to flg22 but remained responsive to elf18, and similarly, the efr-2 mutant responded to flg22 but not to elf18 (Figure 3). These results confirmed that the immune response is specific to the elicitor used. We also measured the Raman spectra of mutants, including bak1, bik1, and rbohd, that are blocked in the PAMP signal transduction downstream of receptor activation (Figure 3). Previous publications have reported that the following mutants of the PAMP-signaling pathway, fls2, efr-2, bak1, bik1, and rbohd, have significantly reduced callose deposition after elicitor treatment (Lu et al., 2009; Zhang et al., 2010; Jaillais et al., 2011; Belkhadir et al., 2012; Daudi et al., 2012; Chaudhary et al., 2014; Smith et al., 2014). Callose deposition patterns and expression changes of marker genes of these mutants confirmed these previous results and verified their positive regulatory role in PTI response (Supplementary Figures 3, 4A). All these mutants did not show any significant changes in Raman spectra after elicitor treatment and hence had low ERI values (< 1 × 10⁴) (Figure 3).

Following the activation of the PAMP receptor, the PTI-signaling pathway is fine-tuned by downregulation to prevent excessive or prolonged activations that would be deleterious to the host plants. One common mechanism of signaling desensitization is by the polyubiquitin-mediated proteolysis of signaling components, and several such components have been identified. The FLS2 receptor itself is known to be poly-ubiquitinated by PUB12/13 (Lu et al., 2011; Zhou et al., 2015), and the coactivator BIK1 is similarly modified by PUB25/26 (Wang et al., 2018). Therefore, both PUB12/13 and PUB25/26 are the negative regulators of the PTI response, and the mutants that are deficient in these E3 ligases should display hyperresponsiveness to elicitors. In a time-course experiment, we found that the pub mutants (pub12/13 and pub25/26) indeed showed increases in gene expression levels as early as 0.5 or 6 h after 1 μM elicitor treatment when little response was detected with WT (Supplementary Figure 4B). Based on the earlier gene expression kinetics, we decided to measure changes in Raman spectra at 10 hpi. Figure 4 shows that at 10 h after flg22 treatment, the WT ERI value was only 1.12 × 10⁴ as compared to an ERI value of 9.44 × 10⁴ at 24 hpi (Figure 3A). Similarly, in the case of elf18, the ERI value at 10 hpi was 0.63 × 10⁴ compared to 8.02 × 10⁴ at 24 hpi (Figures 4A,B for 10 h and Figures 3A,B for 24 h). These results suggest that in WT plants the PAMP signaling pathway was only slightly activated at 10 h, and full activation would require 24 hpi. By contrast, at 10 hpi of either elicitor pub12/13 displayed an ERI value about 13–21 times higher than WT values, and for pub25/26, about 6.4–13.3 times higher (Figure 4B). These results showed that the hyper-responsive nature of the pub mutants to elicitors can be confirmed by Raman spectroscopy. Interestingly, higher ERI values were found in pub12/13 targeting an FLS2 receptor compared to those of pub25/26 targeting a downstream regulator, BIK1, in the flg22/elf18 signaling pathway (Figure 4B).

The ERI values were evaluated based on the Raman spectrum analysis at 24 spots in 4 individual plants (2 leaf discs per plant) with 4 independent experiments giving a range of numbers between the high (>5 × 10⁴) and low (0.0–1.0 × 10⁴) level according to the PTI response (Figure 4B). These results confirmed the accuracy of the ERI indexing based on Raman spectral assay for determining PTI responses. Consistent with these analyses, the pub12/13 and pub25/26 mutant plants showed more callose deposits than WT and the two control mutants, fls2 and bik1 (Figures 4C,D).

Elicitor Response Index presents a quantitative and holistic measure of changes in the Raman spectra in elicitor-treated samples relative to mock control samples. In this case, we consider changes in the peak intensities for individual vibrational bands and consider the metabolites that give rise to these changes. For WT plants at 24 hpi, the majority of the Raman peaks that exhibit changes are common to both elicitor responses, flg22 and elf18. The difference spectra with annotated peaks are shown in Supplementary Figure 5. Table 1 summarizes the relevant vibrational bands (p < 0.05) for various elicitors, and the peak assignments are based on previously published reports. An increase in peak intensity at vibrational bands originating from carotenoids, pectin, phenylpropanoids, cellulose, lignin, and proteins is observed, which contributes toward a high ERI value seen at 24 hpi. Among these, the most prominent increase in peak intensity was associated with carotenoids. WT plants showed a significant increase in peak intensities at around 1,001, 1,151, and 1,521 cm⁻¹ at 24 hpi for both the elicitors (Figure 5A). Peak intensities at 1,151 and 1,521 cm⁻¹ were also increased in the case of receptor mutants treated with a heterologous elicitor (Figure 5A). These strong Raman peaks are attributed to carotenoids, and it is these molecules that exhibit the largest change in concentration. Previous reports have attributed peaks in the 1,100–1,200 and 1,400–1,600 cm⁻¹ Raman shift regions to carotenoids (Payne and Kurouski, 2021). These peaks are due to in-phase carbon-carbon stretching vibrations of the main polyene chain with the exact Raman shift locations of these peaks being dependent on the polyene chain length. These results suggest that an early event in plant PTI response is likely an increase in carotenoid concentration, which contributes significantly toward a high ERI value. To confirm the Raman spectral results with regard to an increase in carotenoid concentration, we extracted the photosynthetic pigments (chlorophyll a, b, and carotenoids) from the treated leaves and measured changes in their content directly. The treatment of WT with elicitors produced a quantitative increase of more than 5% of total carotenoids. In the case of fls2 and efr-2 mutants, the carotenoid content was increased by 10% when challenged with heterologous elicitors (Figure 5B). The results are supported by increased expression of genes related to carotenoid biosynthesis at early time points of elicitor treatment (Supplementary Figure 6).

Specific peaks that exhibit statistically significant changes for a particular elicitor were also observed. Only flg22 treatment in WT showed a statistically significant change in peak intensity at 789.7 cm⁻¹. In the literature, this vibrational band region is typically associated with chlorophyll a and phosphodiester bonds (C5′-O-P-O-C3′) in DNA. elf18 treatment specifically resulted in changes for the Raman peak at 1,065 cm⁻¹. This band is attributed to skeletal –C–C– stretching of lipids or C–O stretch of cell wall polysaccharides (Movasaghi et al., 2007).

The spread in Raman peak intensities that exhibited changes was further analyzed using histogram plots. It provides a better classification inference with high statistical significance. We found that the Raman shift at 1,521 cm⁻¹ and the region between 1,549 and 1,552 cm⁻¹ represent carotenoids and proteins (amino acid-tyrosine and tryptophan), respectively, and provide the best discriminating classification between mock and treated plants (Supplementary Figure 7). As the other two carotenoid Raman peaks (1,001 and 1,151 cm⁻¹) are present in lower wavenumber regions where there is a significant spectral overlap from other biomolecules such as lipids and proteins, clear differentiation was not observed in these Raman peaks. The changes observed at 1,521 and between 1,549 and 1,552 cm⁻¹ are the potential Raman spectral biomarkers, indicating a dynamic early PTI response in plants.

Pseudomonas syringae causes bacterial specks on Arabidopsis and induces chlorosis at later stages of infection (Whalen et al., 1991; Uppalapati et al., 2010; Luo et al., 2017). To see if Raman spectroscopy can be used to detect early Pseudomonas infection, we performed similar experiments in Arabidopsis infected with Pst DC3000. In these experiments, we estimated the Infection Response Index (IRI) using the same method as the ERI value. Similar to the elicitor study, the difference Raman spectrum between the infected and mock sample at 24 hpi exhibited high IRI values. The IRI value decreased to almost zero at 48 and 72 hpi (Figures 6A,C).

To further validate the results observed in Arabidopsis in a crop plant, Xcc infection in Choy Sum was analyzed. Xcc causes Black rot disease, which is a serious problem in all Brassica vegetables resulting in significant yield losses. Typical external visible symptoms of Xcc infections are chlorosis and V-shaped lesions in the leaves (Meenu et al., 2013; Nuñez et al., 2018). Under our growth conditions, the characteristic visual symptoms of Xcc infection are seen within 4–5 days. Raman spectra were measured at 24 and 48 hpi from mock and Xcc infiltrated plants to detect spectral changes before the appearance of symptoms. Like Arabidopsis, the difference Raman spectrum between the infected and mock sample at 24 hpi exhibited high IRI values, which decline at 48 hpi (Figures 6B,D). Similar to the Arabidopsis elicitor experiments, callose deposition was also observed in the Xcc-infected leaves at 24 hpi (Supplementary Figure 8A).

The vibrational bands contributing toward a high IRI value in Arabidopsis and Choy Sum were largely similar and resembled the ERI spectral data (Table 2; Supplementary Figure 9). Like the elicitor treatment, the most dominant changes upon bacterial infection in Arabidopsis and Choy sum were observed around 1,001, 1,151, and 1,521 cm⁻¹, which are attributed to carotenoids (Supplementary Figure 8B). Apart from carotenoids, increase in peak intensities at vibrational bands originating from pectin, phenylpropanoids, cellulose, nitrates, nucleic acids and proteins were the dominant contributors towards the high IRI value. A significant change at 1,253 cm⁻¹ peak intensity was seen only in P. syringae-infected Arabidopsis, which is associated with proteins and lipids (C–N in-plane stretching), whereas Xcc-infected Choy sum showed changes at 791.8 cm⁻¹ attributed to nucleic acid phosphodiester bond, which were not observed in Arabidopsis infection. As observed with the elicitor study, the Raman shift at 1,521 cm⁻¹ and the region between 1,549 and 1,552 cm⁻¹ provide the best classification between mock and infected plants.