Food security is challenged by increasing geopolitical tensions and rapid climate change. About 50% and 25-40% of losses in crop yield world-wide are attributed to abiotic and biotic stressors, respectively (Kumar et al., 2022). Early and rapid diagnosis of the responsible stressor may allow timely intervention and implementation of remedial measures in the field. By enabling timely mitigation, guiding resource deployment, and informing predictive models, it minimizes crop losses and enhances resilience against future challenges (Ristaino et al., 2021).

Incorporating advanced technology into agricultural practices is essential for safeguarding food security. Traditional plant disease detection methods, including visual scouting and laboratory analysis, provide essential confirmation of disease presence and pathogen identification. Serological assays, like ELISA, offer high-throughput capabilities, while qPCR confirmation analysis provides sensitive quantification of pathogens. However, these methods rely on visible symptoms or require sample processing, delaying detection until later stages of disease propagation. Innovative methods, such as volatile compound analysis and remote sensing, offer non-invasive characterization of plant health and can detect subtle changes before symptoms appear, allowing for earlier intervention. This timing is crucial in managing polycyclic diseases and preventing widespread crop loss. Furthermore, techniques such as bio-photonics and advanced imaging provide real-time, in situ analysis, minimizing disturbance to the plant and offering more comprehensive spatial-temporal data, complementing the accuracy of traditional methods with improved speed and preventative capabilities (Martinelli et al., 2015).

Recent research has underscored the effectiveness of Raman spectroscopy (RS) in early diagnosis of both biotic and abiotic stresses in crops. This non-invasive laser technique enables rapid analysis without requiring chemical labelling or compromising the integrity of the sample, making it a cost-effective solution for breeders (Saletnik et al., 2024).

RS analyzes light scattering from molecules. When a photon interacts with a molecule, it can scatter elastically (no energy change) or inelastically (Raman scattering, with an energy shift). This energy shift, or Raman shift is specific to the molecule’s vibrational states (Langer et al., 2019). The resulting Raman spectrum, a plot of intensity versus Raman shift, provides a unique fingerprint for each molecule and its functional groups whereas peak heights within the spectrum correlate with the concentrations of specific molecules in the sample. The integration of qualitative and quantitative data from Raman spectrum allows profiling the vibrational states of biological molecules (Saletnik et al., 2021). The ability to acquire spectra representing detailed molecular information from intact tissues, with minimal interference from vibrational bands of water, renders this technology a powerful analytical tool for studying biological samples (Gierlinger and Schwanninger, 2007; Orkoula and Kontoyannis, 2014).

RS has emerged as a useful tool for diagnosing plant diseases by identifying pathogen-induced changes in plant metabolite profiles. This technology has been applied effectively to detect fungal infections in important crops like sorghum, wheat, and corn. RS coupled with chemometrics has been used to identify the infection of four fungal pathogens on maize kernels or fungal infected sorghum grains. Surface-enhanced Raman spectroscopy (SERS) has enabled the detection of Fusarium wilt of banana (FWB) by fingerprinting, an assay based on the database that includes the spectral characteristics of the pathogen. In combination with multivariate statistical analysis, RS has also been used for identifying a bacterial disease, Huanglongbing, in citrus trees. Candidatus Liberibacter solanacearum (Lso) infection on tomato can also be detected 3 weeks following infection. The distinct spectral features produced by different viruses enable RS to differentiate infections caused by various viral combinations or individual viruses on rose, wheat or tomatoes. The method can also identify grapevine infections attributable to grapevine fanleaf virus (GFLV) and grapevine stem pitting-associated virus (GRSPaV) with 100% and 80% accuracy, respectively, before appearance of phenotypic symptoms (Weng et al., 2021).

In most of these studies Raman spectra can detect metabolic changes in plants during pathogen interactions. Metabolites with resonance structures contribute significantly to the spectral features observed. For instance, bands associated with carotenoids appear in the ranges of 1001–1151 cm^-1 and 1520–1550 cm^-1. The resonance structures of these metabolites play crucial roles in their biological functions including light absorption in photosynthetic systems and antioxidant activity under stress conditions (Chung et al., 2021).

To sense biotic stresses from the environment, plants have developed a system that recognizes pathogen-associated molecular patterns (PAMPs), which are composed of conserved immunogenic structures derived from pathogens, including bacteria, fungi, and viruses. PAMPs can be detected by Pattern recognition receptors (PRRs) locate on cell-surface. These receptors are receptor-like proteins (RLPs) or receptor-like kinases (RLKs) which transmit information about the extracellular stimuli to the cytosol to render a series of corresponding PTI as the first layer of plant defense system (Van Der Burgh and Joosten, 2019).

Flagellin Sensing 2 (FLS2) and the elongation factor Tu Receptor (EF-Tu receptor, EFR) are two well studied RLKs in Arabidopsis thaliana. FLS2 specifically detects the flagellin-derived immunogenic peptide flg22, whereas EFR recognizes elf18, a peptide from the bacterial EF-Tu (Gómez-Gómez and Boller, 2000; Zipfel et al., 2006). Binding of flg22 strengthens the dimerization between FLS2 and another RLK, Brassinosteroid insensitive 1-associated receptor kinase (BAK1) resulting in mutual phosphorylation thus triggering the cytosolic immune response (Chinchilla et al., 2007). Besides the similar heterodimerization, EFR has also been proposed to activate BAK1 via an allosteric regulation (Mühlenbeck et al., 2024).

In contrast to bacteria-derived immunogenic peptides, defence response against fungal infection is triggered by chitin, a key component of fungal cell walls, which serves as a PAMP for fungi. Chitin elicitor receptor kinase 1 (CERK1, LysM receptor-like kinase 1, LYK1) is the first identified chitin receptor that physically interacts with chitin to initiate the MAPK signaling cascade. Mutations in CERK1 greatly reduce ROS production which is part of plant immunity response against fungi (Miya et al., 2007). Two other LysM receptor-like kinases, LYK4 and LYK5 have been identified as the core components of the chitin receptor complex. Besides their redundant role in chitin-responsive MAPK activation, LYK5 could also induce phosphorylation of CERK1 and LYK4 and is indispensable for ROS production in response to chitin. Mutations in either LYK4 or LYK5 renders plants partially insensitive to chitin treatment whereas a lyk4/5 double mutant shows little or no chitin response (Cao et al., 2014; Wan et al., 2008).

Upon recognizing PAMPs, PRRs activate various downstream signaling pathways including the Mitogen-Activated Protein Kinase (MAPK) pathways. These pathways, which are highly conserved across eukaryotes, serve as essential signaling networks that regulate a wide range of physiological processes in plants and they effectively transducing external stimuli into appropriate cellular responses. Particularly important for mediating responses to both biotic and abiotic stresses, MAPK pathways consist of 3 sequentially activated protein kinases: MAPK, MAPK kinase (MAPKK), and MAPKK kinase (MAPKKK). This tiered structure facilitates signal amplification and specificity (Sun and Zhang, 2022).

The activation of MAPK pathways by PAMPs and other stress signals induces several downstream effects, including the production of reactive oxygen species (ROS), activation of defence-related genes, and the establishment of systemic acquired resistance (SAR) (Genot et al., 2017; Mersmann et al., 2010). In response to increased ROS levels, plants may enhance synthesis of protective pigments. For instance, carotenoids and flavonoids, known for their antioxidant properties, often increase in abundance to scavenge ROS and mitigate oxidative damage, helping to stabilize the photosynthetic apparatus. Similarly, flavonoids such as anthocyanins can be upregulated in response to various stresses, providing photoprotection by absorbing excess light and shielding tissues from injury. Conversely, chlorophyll content may decline under stressful conditions leading to chlorosis and consequently reduced photosynthetic efficiency (Han et al., 2012; Xu and Rothstein, 2018).

As fungal infestations pose significant challenges to many agricultural systems, implementing effective early detection strategies is crucial to prevent crop loss. Timely intervention facilitated by early detection technology can help minimize chemical use, enhance food security, and promote economic sustainability in farming practices. In previous studies, we have demonstrated the potential of RS as an early diagnosis system based on detecting metabolites changes mediated by innate immunity against bacteria (Chung et al., 2021). Here, we extend and broaden the application of this early diagnosis system to detect fungal infections in plants.

Our previous study has demonstrated the application of RS for detecting early bacterial infections in plants (Chung et al., 2021). We introduced the Elicitor Response Index (ERI) and Infection Response Index (IRI) to quantify significant increase in spectral intensity and evaluate PTI response in inoculated leaves. In this previous study, both indexes competently identified PTI response at 24 hpi following elicitor treatment or pathogen introduction, however, the decrease in spectral intensity that often followed the appearance of the ERI or IRI peak was not considered. Our results here, along with previous observations, suggest that the negative difference should be also considered as a feature in the responses to pathogens. Therefore, in this study, we employed a modified ERI/IRI which could represent both significantly positive or negative differences in Raman spectra. Our results showed that following chitin treatment, WT Arabidopsis showed a positive ERI that peaked at 9 hours post-inoculation (hpi) and subsequently decreased from 12 hpi onward. ( Figure 1 , Supplementary Figures 2 , 3 ). To verify the response specificity, we performed similar assays with chitin receptor mutants, cerk1 and lyk4/5 which displayed ERI of 0 or close to 0 when treated with chitin or inoculated with fungi ( Figure 1 , Supplementary Figure 2 ). The mutants cerk1 or lyk4/5 were previously reported to show increased susceptibility to Alternaria brassicicola, Erysiphe cichoracearum and Fusarium oxysporum f. sp. cubense 4 (Huaping et al., 2017; Miya et al., 2007; Wan et al., 2008) and similar results were observed following inoculation with C. higginsianum ( Figure 2B ). Infected leaves of cerk1 and lyk4/5 showed lesions of smaller sizes but more severe water-soaked leaves as compared to infected WT leaves. Notably, chitin treatment on the knock-out (KO) mutants of bacterial pattern recognition receptors (PRRs), fls2 resulted in 0 to negative ERI at 6 and 9 hpi, rather than the anticipated positive ERI observed in WT. Further analyses revealed a positive spectral fluctuation at 3 hpi ( Supplementary Figure 4 ), Consistent with these findings, fls2 demonstrated weaker yet a more prolonged MAPK activation when compared to WT ( Supplementary Figure 5 ).

Chitin and its derivative, chitosan are known for their ability to boost plant immune responses across a wide range of plant species. These elicitors serve as a signal to activate plant immune systems. The recognition of chitin by its receptors prompts a series of defensive responses, including the production of reactive oxygen species, the activation of signaling pathways, and the expression of defence-related genes. These responses not only provide direct antimicrobial effects but also endow the treated plants with defenses against future pathogen attacks. The ability of chitin to universally trigger such immune responses renders it a crucial component in the enhancement of plant resilience to pathogens like bacteria, fungi, viruses, and nematodes (Saberi Riseh et al., 2024).

In addition to its direct antimicrobial effects that inhibit the proliferation of various microorganisms, an increasing number of studies have also shed light on the crosstalk between bacteria- and fungi-induced immunity. A recent study has indicated that the common regulatory kinase BAK1 can phosphorylate the juxtamembrane (JM) region of CERK1 in response to treatment with pathogenic bacteria or bacteria-derived PAMPs, such as flg22 and elf18. While this bacteria-mediated phosphorylation does not fully phosphorylate CERK1 as observed with chitin treatment, the findings suggest that it primes CERK1 to enhance its signaling capabilities in response to chitin. Consequently, this priming leads to increased resistance against potential fungal infections (Gong et al., 2019). On the other hand, chitin soil amendment was reported to trigger systemic resistance by enhancing PTI against P. syringae infection in lettuce, tomato and Arabidopsis (Makechemu et al., 2024). Micrografting experiments confirmed that chitin perception in roots strengthens the expression of key PTI components in distal leaves, including BIK1 and RBOHD. Chitin addition also renders plants more responsive to subsequent treatment of flg22 or elf18 resulting in a stronger ROS burst and potentiated calcium influx (Makechemu et al., 2024). Despite these differences, the significant crosstalk between the fungal and bacterial MAPK pathways leads to a more robust and coordinated defence response. Common components within the MAPK signaling cascades could integrate signals from both types of pathogens, allowing plants to generate a typical defence mechanism via a potential mutual regulation.

The absence of FLS2 during chitin treatment may potentially disrupt the homeostasis maintained by the intrinsic crosstalk between the two PTI systems. The positive spectral difference observed as early as 3 hours post-inoculation (hpi) suggests a hyper-responsive state that precedes the expected recession seen in WT plants after 12 hpi ( Supplementary Figures 3 , 4 ). This finding may also indicate the possibility of an ultrasensitive response caused by multisite phosphorylation or potential allosteric regulation from FLS2-mediated PTI (Ferrell and Ha, 2014a, b). It is conceivable that a faster, albeit partial, activation of PTI could be triggered due to a lowered response threshold to the elicitor or pathogen ( Supplementary Figure 5 ). Such dynamics invite further exploration into the nuances of plant immune responses and their implications for understanding plant-pathogen interactions.

Changes in carotenoids peak intensity are usually one of the most significant features to identify pathogen infection or disease severity (1151 cm^-1 (Gupta et al., 2020), 1180 cm^-1 (Perez-Guaita et al., 2017) and 1521 cm^-1 (Movasaghi et al., 2007)). Carotenoids have been shown to serve not only as scavengers of reactive oxygen species (ROS) during PTI (Han et al., 2012) but also as important precursors of stress hormones, such as abscisic acid (ABA). These hormones play a crucial role in regulating the expression of downstream genes in response to both biotic and abiotic stresses (Felemban et al., 2019; Jia et al., 2022). Moreover, Raman shift can be affected by the fluctuation of metabolites content as a result of pathogenic activity, such as decomposition or fermentation (Weng et al., 2021).

The resulting differential Raman peaks could assist in distinguishing between treatments with different elicitors or infections by various pathogens. For instance, the band representing nucleic acid at 789.7 cm^-1 increased intensity during P. syringae DC3000 infection and flg22 treatment (Movasaghi et al., 2007), suggesting a differential expression of defence genes. Additionally, the intensity of peaks at 482 cm^-1, 542 cm^-1, and the region between 742 and 1046 cm^-1 were generally higher in comparison to both chitin treatment and fungal infection. The peaks at around 542 cm^-1 and 742.3 cm^-1 indicates the accumulation of flavonoids and pectin, respectively (Moreno et al., 2022; Gupta et al., 2020). These observations suggest a differential defence mechanism mediated by bacteria or their PAMPs.

A Raman band at 1578 cm^-1 was only seen in P. syringae DC3000 inoculation, which may indicate increased levels of quinonoid compounds (Aymen et al., 2013), potentially derived from flavonoid metabolism. Quercetin, for example, is known to enhance resistance to Pseudomonas syringae pv. tomato DC3000 by modulating salicylic acid biosynthesis in Arabidopsis and this compound can be converted to various quinones via ROS-mediated pathways (An et al., 2023).

The nitrate peak at 1046 cm^-1 generally increases in all pathogen infection scenarios. In Arabidopsis, the high-affinity nitrate transporters NRT2.1 and NRT2.2 are upregulated in response to nitrogen deficiency. Double mutants lacking both NRT2.1 and NRT2.2 exhibit significantly reduced nitrate uptake compared to WT plants, regardless of nitrogen availability. A Raman peak at 1046 cm^-1 has been shown to decrease in intensity in nrt2.1/nrt2.2 double mutants (Huang et al., 2020). Furthermore, NRT2.1 appears to repress responses to biotrophic pathogens, potentially prioritizing resources for abiotic stress resistance over immune responses. Whereas the role of NRT2.6 in nitrate uptake is less clearly defined, its involvement in plant responses to necrotrophic pathogens seems to be related to reactive oxygen species (ROS) production and metabolic shifts, rather than a direct regulation of nitrate transport during infection. These studies emphasize the interplay between nitrate transport systems and defence signaling pathways (Camañes et al., 2012b, a; Dechorgnat et al., 2012).

Nitrate content could also be affected due to impaired nitrogen assimilation, altered nutrient allocation or compromised transport during pathogen infection. For example, Fusarium pseudograminearum infection was reported to reduce the ability of wheat plants to transfer nitrogen from roots to shoots (stem nitrogen transfer efficiency, or sNTE). This leads to a less efficient nitrogen use and reduced grain yield and quality (Buster et al., 2023). Infection of soybeans by Phytophthora sojae induces stress, causing infected plants to reallocate resources from growth towards defence mechanisms. This shift of nitrogen allocation away from growth and towards the synthesis of defence-related compounds, results in a state of apparent nitrogen limitation (Li et al., 2013). Soil-borne bacteria, fungi, and oomycetes cause vascular wilt diseases by colonizing xylem vessel. As a defence strategy against these pathogens, plants may alter xylem morphology to restrict pathogen spread but this strategy could also limit nutrient transport (Yadeta and Thomma, 2013). Collectively, the nitrate peak at 1046 cm^-1 may serve as a marker of responses triggered by PAMPs or pathogen infection. These findings also suggest its potential application for discriminating between bacterial and fungal spectra.

In this study, we have also demonstrated the effectiveness of this RS for detecting fungal infections in both Arabidopsis and Brassica vegetables. A positive IRI was observed within 48 hpi in most experimental trials. However, Choy-Sum infected by C. higginsianum did not exhibit any positive IRI within the same detection window. Interestingly, the symptoms in infected Choy-Sum were more severe than those in infected Pak-Choy, and positive differences were generally noted across all three trials. A similar trend was observed in C. higginsianum-inoculated cerk1 and lyk4/5 plants ( Figures 1B, C , Supplementary Figure 2 ), suggesting a higher susceptibility of Choy-Sum to C. higginsianum.

The IRI could successfully predict fungal infections in randomized controlled trials with an accuracy rate of 76.2% for Arabidopsis and 72.5% for Pak-Choy ( Figure 4 ). The inoculated plants exhibited a false-negative rate of 22.7% to 23.8%, with 5% of the inoculated samples showing a negative IRI. There was no correlation between symptom severity and IRI value, indicating that the zero IRI observed should not be attributed to differential susceptibility among individuals. Considering the dynamic nature of the IRI as observed in the analyses of Raman spectra during fungal infection ( Figure 4 ), our results suggest that the IRI of certain individual plants may increase at different time points. This hypothesis is supported by a parallel trial involving 30 Arabidopsis plants, which achieved an accuracy rate of 76.7% without any negative IRI ( Supplementary Figure 12 ). Additionally, the false positive rate ranged from 25% to 31.6%, with instances showing no observable symptoms compared to the inoculated leaves. In summary, our results indicate that variations in metabolite content among individuals can result in differing IRI outcomes. Implementing real-time detection to monitor IRI dynamics may help mitigate this issue and enhance overall accuracy.

While Raman spectroscopy holds promise, a significant limitation is the difficulty in differentiating between closely related bacterial or fungal species by current biomarkers, such as pigments, which are often insufficient for species-level identification. Environmental factors and plant stress responses can also alter Raman spectra, potentially confounding the detection of specific pathogens (Payne and Kurouski, 2021). To better integrate data from various sources for diagnostic purposes, more work should be done in connecting Raman Spectroscopy with the knowledge graphs that model plant response to pathogens.

As highlighted by Murray et al., a systems-level approach integrating Raman spectroscopy with genomics, metabolomics, and remote sensing data is crucial for addressing the limitation (Murray et al., 2025). Plant Reactome knowledge graphs may provide insights into specific metabolite changes that are indicative of specific pathogens, enabling more targeted Raman-based detection strategies. The knowledge gaps regarding plant-pathogen interactions and the need for well-curated OMICs datasets underscore the importance of future research in this area. Large-scale data collection efforts, coupled with expert biocuration, are needed to develop robust spectral signatures for disease diagnosis.

Future research should focus on developing more specific biomarkers, integrating Raman spectroscopy data with other data types (genomics, metabolomics, remote sensing), and creating well-curated OMICs datasets to improve the accuracy and reliability of disease diagnosis. An integrative approach is essential for building robust and scalable disease surveillance systems.

Here, our study advances the assessment of plant defence responses against fungal infections by utilizing the Elicitor Response Index (ERI) and the Infection Response Index (IRI). Elicitor treatments and biological assays demonstrated the capability of this method to quantify metabolic changes rapidly, within hours, and to identify early responses that occur before the appearance of visible symptoms. The randomized controlled assays confirmed its applicability in predicting infections in Arabidopsis and Pak-Choy. Furthermore, the acquired spectral data show-cased the potential to distinguish plant responses to various pathogens, thereby expanding the role of RS in pathogen detection. By harnessing RS, researchers can gain valuable insights into the complex interactions between plant pigments and their functions, which enhances our understanding of plant biology and improves agricultural practices. Overall, our findings underscore the promise of RS as a non-invasive monitoring tool that can strengthen crop health management and support sustainable agricultural practices.