The leaf samples used in the study were collected from 10 citrus orchards located in 10 towns in Renhua County (24°56’ ~ 25°27’N, 113°30’ ~ 114°02’E) in Shaoguan City, Guangdong Province, China, which is one of the main production area of ‘Gonggan’ but suffered from HLB recently. From each orchard, 4~8 citrus plants with no HLB symptoms or HLB symptoms to different degree were selected randomly, and leaves distributing uniformly on the outside of canopy were sampled from each plant. The collected leaves were carefully placed in sterile bags. Finally, a total of 61 leaf samples were obtained for this study. All the samples were placed in an insulated ice box immediately after sampling and transported to the laboratory, where they were stored at a temperature of 4 °C until further processing.

The leaf samples were sterilized using a rigorous protocol. Firstly, they were washed with 75% alcohol for 30 seconds in an ultra-clean workbench. Following this, the samples were washed with sterile water 3~5 times to eliminate any residual alcohol. Then, the samples underwent further sterilization with 2% sodium hypochlorite for 1 min and were subsequently washed again with sterile water 3~5 times. To ensure complete disinfection, the final elution solution from the cleaning process was collected. A 100 µL of this solution was plated on LB agar plates for microbial culturing. The absence of microbial growth after incubation confirmed successful complete disinfection. Next, the midrib (approximately 2 mm in width) was excised using a sterile surgical blade, and then cut into small pieces (approximately 5 mm in length) and placed in a sterile 2 mL centrifuge tube. The samples were ground into a fine powder using a homogenizer, and DNA extraction was performed using the Tiangen DP305-02 plant kit (Faddetta et al., 2021).

The quality of extracted DNA was assessed using a Nanodrop spectrophotometer and gel electrophoresis. For the PCR amplification of the 16S rRNA gene specific to CLas, the OI1/OI2c primers were used (Jagoueix et al., 1996). The PCR reaction was set up as follows: a 25 µL system containing 9.5 µL ddH₂O, 12.5 µL 2×Taq PCR Master Mix, 1 µL 10 μM forward primer, 1 µL 10 μM reverse primer, and 1 µL DNA template. The PCR program consisted of an initial denaturation step at 95 °C for 3 min, followed by 34 cycles of denaturation at 95 °C for 30 s, annealing at 64 °C for 35 s, extension at 72 °C for 80 s, and a final extension step at 72 °C for 7 min. For the quantitative detection of CLas using qPCR, specific primers CLas-4G and HLBr were used, along with the HLBp probe (Bao et al., 2020). The qPCR reaction was set up in a 20 µL system, including 10 µL SuperMix, 0.4 µL 10 μM forward primer, 0.4 µL 10 μM reverse primer, 0.4 µL 10 μM probe, 0.4 µL Passive Dye, 1 µL DNA template, and 7.4 µL nuclease-free water. The qPCR program included an initial incubation step at 50 °C for 2 min, followed by denaturation at 95 °C for 10 min, and 40 cycles of denaturation at 95 °C for 15 s and annealing/extension at 60 °C for 1 min.

The detection methods commonly used for HLB include PCR and qPCR, which are well-established for their accuracy and widespread use in HLB detection (Ángel et al., 2014). However, there is currently no consensus on the threshold for distinguishing between HLB-positive and -negative samples. Various studies have reported different threshold values, such as CT values of 32 (Mira et al., 2019), 33 (Paula et al., 2018), 36 (Turechek et al., 2012), or 40 (Bai et al., 2013). In this experiment, based on the PCR and qPCR results, a threshold for defining a healthy strain was set at a CT value greater than 34, while a CT value less than 34 was considered indicative of an HLB-positive strain. With this criteria, all the 61 midrib samples were divided into infected group (36 samples) and healthy group (25 samples) ( Supplementary Table 1 ).

The 61 DNA samples were sent to Novogene for amplicon sequencing of the 16S rRNA gene. The sequencing was performed using the Illumina NovaSeq platform with a Paired_End approach. The V3~V4 region (341F: CCTAYGGGRBGCASCAG; 806R: GGACTACNNGGGTATCTAA T) was targeted for sequencing (SRR25558690-SRR25558750, https://submit.ncbi.nlm.nih.gov/subs/). After obtaining the raw data, a series of data processing steps were carried out. The barcode and primers were removed, followed by merging, filtering, and decontamination of the paired-end sequences using QIIME2. Taxonomic classification was performed using the SILVA database with a similarity threshold of 97% (Almeida et al., 2018). The resulting high-quality data was then analyzed using R version 4.0.2, utilizing packages such as ‘barplot’ and ‘vegan’ for creating stacked histograms. TBtools was used for constructing Venn diagrams (Chen et al., 2020), and ImageGP was used for producing PCoA diagrams. Function prediction was done using the ‘PICRUSt2’ package in R (Douglas et al., 2020), and importance ranking was determined using random forest analysis with the ‘RandomForest’ package in R (Shibahara et al., 2017), which mainly contains a random forest classification and selection of important variables contributing to the difference between groups. After removing low-quality and chimeric sequences, a total of 4,138,080 sequences were obtained for the midrib bacterial community. These sequences were further decontaminated to eliminate any host contamination. The data saturation was evaluated using a dilution curve, as shown in Supplementary Figure 1 . Based on the curve, it can be concluded that the decontaminated data was suitable for subsequent analysis.

After removing contamination and other undesired sequences, sequencing data was obtained from midrib samples, which revealed a total of 281 genera at the genus level. The top 30 genera accounted for 98.52% of the total abundance ( Figure 1A ; Supplementary Table 2 ). To analyze the community composition, we focused on these top 30 genera. Significant differences (P < 0.001) were observed in the bacterial community between healthy and infected plants ( Figure 1B ). Among the healthy plants, the most abundant genera were Pseudomonas (53.27%), Erwinia (37.36%), Escherichia-Shigella (4.90%), Pantoea (4.05%), and Bacillus (0.23%). In contrast, infected plants had the highest abundance of Pantoea (66.01%), followed by Pseudomonas (13.38%), Erwinia (7.06%), Corynebacterium (0.90%), Streptomyces (0.86%), Romboutsia (0.72%), and Bacillus (0.70%) ( Figure 1A ; Supplementary Table 3 ). Among the 281 genera, 23 genera showed significant differences in abundance between the two groups ( Figure 1C ). Additionally, the abundance of Pseudomonas and Erwinia significantly decreased in infected citrus compared to healthy ones. Conversely, the abundance of Pantoea, Bacillus, Vulcaniibacterium, Meiothermus, Streptomyces, pathogen CLas and etc. significantly increased in infected samples ( Figure 1C ). Furthermore, compared to healthy samples, the α diversity index showed a significant increase (Chao1), but a significant decrease (Good’s coverage) in infected samples. The overall diversity (Shannon and Simpson indices) also tended to decrease in infected samples ( Table 1 ). These results suggest that the bacterial communities in the leaf midribs of citrus plants underwent significant changes due to HLB infection.

In total, 62 genera were found to be present in both healthy and infected citrus midribs ( Figure 1D ; Supplementary Table 4 ), which were considered hereafter as shared taxa between healthy and infected leaf midribs. These shared taxa mainly included Pseudomonas, Erwinia, Pantoea, Bacillus, and others. Additionally, the infected group exhibited 216 unique genera, which was much higher in number than the healthy group (3 unique genera). This suggests an increased bacterial richness in infected midribs compared to healthy ones ( Figure 1D ; Table 1 ).

To analyze the differences in shared taxa between healthy and infected midribs and identify which taxa contribute to these differences, a random forest importance ranking was conducted using the shared bacterial community. The results showed that the model had the lowest error ratio when considering the top 18 genera ( Figure 2A ). The importance values, known as Mean Decrease Accuracy, were then used to rank the genera. The analysis revealed that 6 genera were of significant importance: Escherichia-Shigella, Streptomyces, Erwinia, Pseudomonas, Xanthomonas, and Dokdonella ( Figure 2B ). Among these, 4 genera (Escherichia-Shigella, Streptomyces, Erwinia, and Pseudomonas) showed significant differences between healthy and infected midribs. Streptomyces was found to be significantly enriched in infected midribs, while the other three genera were significantly enriched in the healthy samples ( Figure 2C ). This result suggests that these 4 genera were closely associated with the onset of HLB in citrus plants.

Additionally, among the 62 genera, 21 genera exhibited significant differences in abundance between healthy group and infected group ( Figure 2C ), indicating that approximately one-third of the shared taxa underwent changes in abundance in association with HLB. In detail, Pseudomonas and Erwinia were significantly enriched in healthy midribs, while Pantoea, Bacillus, and the pathogenic bacteria Candidatus liberibacter were significantly enriched in infected samples. These findings suggest that Pantoea and Bacillus might be considered as risky species or pathogen helper while Pseudomonas and Erwinia might play important roles in the growth and development of citrus plants, or even possess antagonistic properties against CLas. However, the healthy group and the infected group could not be well distinguished according to the shared taxa ( Figure 2D ).

In order to better understand the functional implications of the differences in shared taxa between the healthy and infected groups, we focused on four genera (Erwinia, Pseudomonas, Streptomyces, Scherichia-Shigella) with significant contribution to the observed differences, and identified their proteins in association with disease resistance.

The results indicate that the enzyme glutathione S-transferase (K00799) was present in genera Erwinia (OTU3) and Pseudomonas (OTU4). This enzyme is known to be involved in detoxification processes and also plays a role in gibberellic acid (GA) biosynthesis. In the genus Streptomyces (OTU26), several proteins of interest were identified, including endo-1,4-β-xylanase (K01181), α-L-rhamnosidase (K05989), and 3-oxoacyl-[acyl carrier protein] reductase (K00059). Similar to the other genera, these proteins were enriched in the infected samples ( Table 2 ; Supplementary Tables 9–12 ), suggesting that they might play a role in disease progression or the biochemical response of leaf midribs to infection by CLas.

Similarly, the gluconic acid 2-dehydrogenase (EC:1.1.99.3) were observed in Erwinia, while 3-oxoacyl-[acyl-carrier-protein] reductase (EC:1.1.1.100) was observed in Pseudomonas, in addition to glutathione S-transferase (EC:2.5.1.18). Furthermore, several enzymes, such as cellulases (EC:3.2.1.4), α-L-rhamnosidase (EC:3.2.1.40), endo-1,4-β-xylanase (EC:3.2.1.8), and 3-oxoacyl-[acyl carrier protein] reductase (EC:1.1.1.100) were observed in Streptomyces. Notably, all of these enzymes were found to be enriched in the diseased samples ( Table 3 ; Supplementary Tables S13–16 ), suggesting their potential involvement in disease development or the response of infected midribs to pathogens.

Our study demonstrates that HLB infection decreased the diversity but increased the richness of bacterial community in the leaf midribs of citrus plants. Previous studies revealed that the diversity and richness of bacteria in infected midribs increased, while specific key groups in leaves and roots decreased in the early stages of HLB infection, followed by an enrichment of beneficial taxa (Blaustein et al., 2017; Ginnan et al., 2020). Additionally, it has been found that when Citrus reticulata cv. Shatangju was invaded by CLas, the alpha diversity of its bacterial community initially decreased and then increased, but not reaching a significant level (Yan et al., 2021). Overall, the invasion of pathogenic bacteria is a dynamic process (Tang et al., 2020), so the diversity of plant microbiome may fluctuate depending on the period of invasion and sampling used.

In this study, the shared taxa in leaf midribs consisted of 62 genera, including Pseudomonas, Erwinia, Pantoea, and Bacillus, etc. These shared taxa have been commonly found in various crops such as barley, rice, sugarcane, grapes, soybeans, Arabidopsis, and in citrus plants as well. For instance, Streptomyces, Sphingomonas, Kaitobacter, and Bacillus have been identified as shared taxa in citrus rhizosphere by Penyalver et al. (2022). Similarly, Pseudomonas, Sphingomonas, and Streptomyces were identified as shared taxa in the rhizosphere, root endosphere, flower and flush of citrus in California by Xi et al. (2023). Additionally, according to a large number of literature, Pseudomonas, Sphingobium, Chittinopaga, and Agrobacterium were identified as shared taxa in citrus (see the review by Srivastava et al., 2022). These reports suggest that there was some variation in the composition of shared taxa among different citrus varieties, and this variation may be attributed to the variations in the surrounding environments (Trivedi et al., 2020) or the genetic makeup of the host. Although these reported shared taxa are derived from multiple niches other than leaf midribs, the bacteria can move between different niches (Xiong et al., 2021). Therefore, the shared taxa of a specific niche might be the shared taxa of another niche as well.

It is worth mentioning that, in this study, these shared taxa were found in both the healthy and the infected midribs. Therefore, it is likely that these taxa in the healthy midribs are involved in the plant growth and development while these taxa in the infected midribs are involved in disease resistance. For example, the shared taxa in maize plants are active nitrogen fixers or contribute to biological nitrogen fixation (Zhang et al., 2022), which demonstrates that the core taxa shared in all xylem sap samples were involved in plant growth and development. The shared core microbiota of soil microbiomes plays a crucial role in maintaining the functional stability of soil microbiota in afforestation ecosystems (Jiao et al., 2022). Additionally, the rotation between chili/eggplant and banana crops generates nine unique antagonistic shared core species, predominantly including Bacillus and Pseudomonas, which have significant impacts on the reduction in the incidence rate of banana wilt (Hong et al., 2023). These antagonistic shared core species persisted in high abundance in the rhizosphere soils of both chili/eggplant and banana in the rotation system compared to the banana monoculture system. Therefore, the shared taxa may play important roles in plant health, nitrogen fixation, ecosystem stability, and disease management. On the other hand, microbial functions highly depend on environments and a given microbial taxa can function differently in different environments (Djemiel et al., 2022). In this study, we identified the shared taxa in leaf midribs of citrus, but did not isolate them. It is necessary to perform the isolation and to test their potential functions in alleviating HLB in the future.

Our study identified 62 genera as shared taxa, accounting for 99.36% of all taxa in terms of abundance (infected group 98.27%, healthy group 99.94%). This suggests that the shared taxa were abundant in citrus midribs, with the unique taxa only accounting for 0.64% of all taxa (infected group 1.73%, healthy group 0.06%). Although the abundance of unique taxa was very low, they might play a crucial role in the health of citrus and thus deserve further investigation, especially considering that some of them were recruited by the infection by CLas. Interestingly, rare or key taxa can play a pivotal role in determining the structure and function of microbiomes (Banerjee et al., 2018). Rare taxa exhibit a wide range of environmental adaptation and have less functional redundancy compared to abundant taxa (He et al., 2022b), contribute significantly to the health and functioning of ecosystems (Lyons et al., 2005). Thus, it is crucial to consider the role of rare taxa in maintaining the overall health and functioning of the citrus microbiome. Further investigation on the specific functions and interactions of rare taxa with CLas might provide valuable insights into the mechanisms underlying HLB and the solutions to mitigate its impact.

Our study identified three metabolic pathways with significant importance to the shared taxa, i.e., 1,4-dihydroxy 6-naphthoate biosynthesis I, categol degradation III ortho cleavage pathway, and L-isoleucine biosynthesis II. These pathways were were also enriched in healthy midribs. Interestingly, 1,4-dihydroxy 6-naphthoate, as the precursor of menaquinone, is related to the biosynthesis of menadione (MK), which is a potential target for evaluating antibiotics in Gram-positive bacteria (Hiratsuka et al., 2008; Choi et al., 2017). This suggests that the enrichment of 1,4-dihydroxy 6-naphthoate and other related metabolites in the healthy midribs could be attributed to antagonistic taxa contributing to antibacterial processes and maintaining plant health. Additionally, the categol degradation III ortho clearance pathway is involved in the degradation of aromatic hydrocarbon pollutants and environmental remediation (Heinaru et al., 2000; Basu et al., 2003; Murakami et al., 2003; Ontanon et al., 2015). L-methionine, another significantly important metabolic pathway, has antioxidant and free radical scavenging functions (He et al., 2022a). Meanwhile the effector protein encoded by CLas prophage has been reported to target ROS clearance-related proteins in citrus, inhibiting the accumulation of reactive oxygen species in plants and promoting the infection of pathogenic bacteria (Du et al., 2023). These information likely suggest that L-methionine may promote the invasion of CLas by clearing ROS.

Our comprehensive analysis of proteins revealed that the Spermidine/Putrescine transport system ATP-binding protein was not only a significantly important proteins, but also showed significant intergroup differences. Spermidine and putrescine are well recognized for their importance as polyamines in various cellular processes, including gene expression, cell growth, survival, stress response, and proliferation (Shah and Swiatlo, 2008; Gevrekci, 2017). Moreover, spermidine can promote the activity of the gene colibactin, and its intracellular pool in bacteria is tightly regulated through de novo synthesis and transport (Nanduri and Swiatlo, 2021). Therefore, maintaining polyamine homeostasis is crucial for proper bacterial physiology and has a significant impact on bacterial pathogenesis. Therefore, Yatin (2002) concludes that inhibitors of polyamine production shows promise in disease prevention and treatment. In these scenarios, it can be deduced that the Spermidine/Putrescine transport system ATP-binding protein identified in this study may play a vital role in the proliferation of CLas and the production of virulence factors. Further study is needed to understand the specific mechanisms involved.

In this study, it is clear that the metabolic pathways, proteins, and enzymes associated with significantly different shared taxa could distinguish between the healthy and infected groups, while the significantly different shared taxa could not. Therefore, it is reasonable to conclude that the functions (i.e. metabolic processes) but not the species of the shared taxa in midribs contributed to the onset of HLB. Microbes are normally metabolically versatile for better adaptation to changing environments, thus express only partial functional genes in a particular environment (Zhou et al., 2014; Koskella, 2020). Furthermore, the functionality of microbial community is more determined by environments than by its composition (Zhou et al., 2019). In summary, the functional differences in the significantly different shared taxa were instrumental in the symptoms of host HLB. Understanding these functional differences can provide insights into potential therapeutic targets and strategies for managing HLB.

In this study, the functional characterization of the shared taxa revealed the presence of several proteins and enzymes significantly associated with the disease. One such protein was glutathione S-transferase, which acted as an antioxidant enzyme in oxidative stress and counteracted external stressors (Jin et al., 2022; Tang et al., 2023). The downregulation of SymE protein expression in infected citrus suggested its crucial role in restraining potential pathogens in healthy plants. Enzymes such as endo-1,4-β-xylanase and cellulases are well known for their ability to degrade cellulose and hemicellulose (Ariaeenejad et al., 2019; Thornbury et al., 2019; Thapa, 2020; Li et al., 2021), both of which are crucial components of plant cell walls. Therefore, the secretion of these enzymes by microorganisms is critical for their colonization inside the plant tissues and the interactions in host-pathogen-microbe relationships. The enzyme α-L-rhamnosidase efficiently hydrolyzed natural active substances like rutin (Bang et al., 2015; Yadav et al., 2017; Kalinova et al., 2018). Rutin, a flavonoid, functions in the growth and resistance of plants against both biotic and abiotic stress. Hence, α-L-rhamnosidase likely plays a crucial role in the citrus defense mechanisms against stress. Another enzyme, 3-oxoacyl-[acyl carrier protein] reductase, participates in the biosynthesis of fatty acids (Hu et al., 2018; Guo et al., 2019; Cross et al., 2021). Its overexpression can lead to an increase in the accumulation of fatty acids, thereby enhancing anti-oxidative activity and non-biotic stress resistance (Ye et al., 2022). Therefore, 3-oxoacyl-[acyl carrier protein] reductase may contribute to the defense against pathogenic bacteria in the citrus leaf midrib bacterial community. Gluconic acid 2-dehydrogenase is involved in the synthesis of gluconic acid, which can dissolve mineralized elements and promote their absorption in the host (de Werra et al., 2009; Yu et al., 2019; Jiménez-Gómez et al., 2020; Uroz et al., 2022). This may lead to an improvement of plant growth condition.

Thus, some bacterial taxa in the leaf midribs of citrus plants, such as Erwinia, Pseudomonas, and etc., may engage in host defense against CLas, while Streptomyces may be involved in either the defense against CLas or the pathogenicity of CLas. Overall, the presence and activity of these specific proteins and enzymes in the midribs provide insights into the complexity of defense mechanisms and interactions between the host and its bacterial community in the context of disease. However, conflicting results have also been reported in other studies (Girard et al., 2020; Saldierna Guzmán et al., 2021; Yuan et al., 2021; Shen et al., 2022), which suggests that Erwinia and Pseudomonas may be opportunistic pathogens or isolates from Erwinia and Pseudomonas may be highly divergent in pathogenesis or beneficial nature (Girard et al., 2020). There are some isolates from these two genera being reported as plant growth promoting rhizobacteria (e.g., Erwinia gerundensis, Pseudomonas fluorescens) or pathogen (e.g., Erwinia amylovora, Pseudomonas syringae) so far.