Inhibition of a conserved bacterial dual-specificity phosphatase confers plant tolerance to Candidatus Liberibacter spp

Summary

“Candidatus Liberibacter spp.” are insect-vectored, fastidious, and vascular-limited phytopathogens. They are the presumptive causal agents of potato zebra chip, tomato vein clearing, and the devastating citrus greening disease worldwide. There is an urgent need to develop new strategies to control them. In this study, we characterized a dual-specificity serine/tyrosine phosphatase (STP) that is well conserved among thirty-three geographically diverse “Candidatus Liberibacter spp.” and strains that infect multiple Solanaceaea and citrus spp. The STP is expressed in infected plant tissues, localized at the plant cytosol and plasma membrane, and interferes with plant cell death responses. We employed an in silico target-based molecular modeling and ligand screen to identify two small molecules with high binding affinity to STP. Efficacy studies demonstrated that the two molecules can inhibit “Candidatus Liberibacter spp.” but not unrelated pathogens and confer plant disease tolerance. The inhibitors and strategies are promising means to control “Candidatus Liberibacter spp.”

Graphical abstract

Highlights

• •A conserved protein phosphatase of “Candidatus Liberibacter spp.” was characterized
• •The phosphatase interferes with plant cell death responses
• •Ligand screening identified two small molecule inhibitors of the phosphatase
• •The inhibitors conferred plant tolerance to “Candidatus Liberibacter spp.”

Microbiology; Plant biology; Interaction of plants with organisms

Introduction

Several devastating plant diseases are caused by fastidious (unculturable) “Candidatus Liberibacter spp.” For instance, potato zebra chip disease, caused by “Candidatus Liberibacter solanacearum” (CLso), leads to millions of dollars in losses to the potato industry.¹ Similarly, citrus greening or Huanglongbing (HLB) disease, caused by “Candidatus Liberibacter asiaticus” (CLas), is the most devastating disease of citrus today and threatens citrus production worldwide. HLB is transmitted by an insect vector, the Asian citrus psyllid (Diaphorina citri). Since its inception in 2006, in Florida alone, HLB caused a ≥50% reduction in citrus acreage and economic losses upwards of US $8 billion.^2,3

The citrus industry is evaluating multiple antibiotic-based therapies to control “Candidatus Liberibacter spp.” Between the 1970s and 1980s, tetracycline, penicillin, and other antibiotics were tested for HLB suppression in citrus.^4,5,6 Trunk applications of penicillin and oxytetracycline in symptomatic sweet orange trees provided tolerance, tree health, and yield improvements.^4,7,8 In a drastic measure to control HLB, the Florida citrus industry has obtained a Section 18 emergency exemption from the Environmental Protection Agency (EPA) to use commercial antibiotic/bactericide products. Some products (e.g., ArborOTC, Firewall^TM, and FireLine) have been commercially used in Florida over the last few years.^7,8,9 Although promising in the short term, these bactericides are also being used in animal and human medicine,¹⁰ and their extensive deployment in agriculture in the long term poses threats to the emergence of antibiotic resistance.¹¹ Thus, there is a need to explore alternative therapies for disease management by advancing knowledge of the “Candidatus Liberibacter spp.” pathosystem and employing interdisciplinary approaches.

Target-based molecular modeling and ligand screening are commonly employed in biomedical and pharmaceutical research to discover new drugs and inhibitors.¹² Fewer studies have used this approach in plant-microbial interactions and are limited to the studies of plant proteins.^13,14 Here, we characterize a “Candidatus Liberibacter spp.” dual-specificity serine/tyrosine phosphatase (STP), conserved among 33 “Candidatus Liberibacter spp.” strains, and functions to interfere with host cell death responses. Using in silico molecular modeling and ligand screening approaches, we further identify two pharmacophore-based inhibitors of STP that demonstrated significant antimicrobial activity against “Candidatus Liberibacter spp.” in plant tissues.

Results and discussion

STP is a conserved dual-specificity phosphatase encoded by diverse strains of Candidatus Liberibacter spp

A prior proteomics analysis annotated a unique dual-specificity phosphatase (CLIBASIA_03975, hereafter referred to as serine/tyrosine phosphatase, STP) and classified it as a putative virulence protein based on the absence of detectable homologs in other organisms, which is indicative of rapid evolution.¹⁵ Phylogenetic analysis of 88 protein-coding genes from 33 bacterial genome datasets suggests that CLso and CLas genomes are evolutionarily closer to each other than other bacterial species.^16,17 We further compared the genomes of multiple CLso and CLas strains¹⁷ to analyze their STP sequences. STP was present and conserved among geographically diverse CLso and CLas strains (Figures 1A and 1B), sharing ≥69% amino acid similarity (55% identity) between CLso and CLas isolates (Figure 1A). Phylogenetic analyses also reveal a high degree of conservation of STP genes within CLso and CLas isolates (Figure 1B). For instance, the STP protein shares ≥95% amino acid similarity within CLso isolates (Figure S1A) and ≥99% amino acid similarity within CLas isolates (Figure S1B). Expression analysis of infected plants confirmed STP presence and expression in C. Liberibacter-infected potato and citrus shoots (Figure 1C). Next, we evaluated whether CLso and CLas STP possess phosphatase activity. The respective coding sequences were cloned into binary vectors and expressed in Arabidopsis thaliana protoplasts as HA-tag fusions, followed by immunopurification using anti-HA beads. Enzymatic activity of the immunopurified CLso and CLas STP and a non-phosphatase protein control (GFP) control were determined using a p-nitrophenyl phosphate (pNPP)-based substrate assay.¹⁸ Both CLso and CLas STPs demonstrated robust phosphatase activity (Figure 1D). Together, these results indicate that STP is a conserved and active phosphatase expressed in C. Liberibacter-infected plant tissues.

Figure 1.

A conserved serine/tyrosine phosphatase (STP) encoded by multiple “Candidatus Liberibacter spp.”

(A) Multiple sequence alignment and (B) phylogenetic relationships of the putative conserved serine tyrosine phosphatase (STP) encoded by “Candidatus Liberibacter solanacearum” (CLso), asiaticus (CLas), europaeus (CLeu), and americanus (CLam) strains isolated and sequenced from diverse geographical locations. Identical amino acids (100% conserved sequence) across all strains are shaded in black, while similar residues are shaded in gray. Residues that lack similarity across 1 or more isolates are shaded white. Pharmacophore/binding hotspot regions are shown in the sequence logo. Enlarged sequence alignments are presented in Figure S1. The maximum likelihood method and 1,000 bootstrap replicates were employed to construct the unrooted phylogenetic tree.

(C) Expression of STP in leaves of healthy potatoes or challenged with psyllids carrying CLso+ or not (CLso-).

(D) Expression of STP in leaves of healthy citrus or citrus infected with CLas+. Lanes “-” and “+” are negative (no template) and positive controls, respectively. Uncropped raw agarose gel images used to prepare e and f are presented in Figure S9.

(E) Phosphatase activity assays of CLso and CLas STPs. The amount (pmols) of substrate metabolized by CLso STP and CLas STP, compared to the GFP protein control, over 45 min of reaction time. Statistically significant differences were determined with one-way ANOVA followed by Tukey’s honestly significant difference (HSD) test. Different lowercase letters represent significant differences among the group means (p ≤ 0.05). p value of ANOVA is indicated. Data are represented as mean ± SEM.

STP localizes to the host cytosol and plasma membrane and attenuates Prf^D1416V-mediated plant cell death response

C. Liberibacter spp. is an obligate and intracellular pathogen. We next sought to determine the sub-cellular localization of STP in plant cells. For this, CLas STP was selected and tagged with a green fluorescent protein at the C-terminus (STP::GFP) and transiently overexpressed in grapefruit protoplasts. Fluorescence imaging showed accumulation of STP::GFP primarily in the cytosol and plasma membrane (Figure 2A). STP::GFP was also localized to a few unidentified cytosolic structures, possibly mitochondria or vesicles. These data support a model wherein STP could interfere with host cellular proteins and processes.

Figure 2.

STP accumulates in plant cytosol and plasma membrane and suppresses Prf^D1416V-induced plant cell death

(A) Subcellular localization of STP. Grapefruit protoplasts were transfected with full-length CLas STP::GFP encoding binary vector. A plasma membrane (PM) marker gene tagged with red fluorescent protein (RFP, PM-RFP) was co-transfected to indicate PM. After 24 h, protoplasts were observed using a confocal laser-scanning microscope. STP::GFP and PM localization is indicated by green and red fluorescence signals, respectively. Overlapping green and red signals are shown in the third panel. Arrowheads point to the STP localization in the cytosol, plasma membrane, and a few unidentified organelles, possibly mitochondria.

(B and C) Transient expression of the CLso and CLas STP peptides suppresses Prf^D1416V-induced cell death in N. benthamiana. Agrobacterium tumefaciens carrying binary expression vectors with either CLso or CLas STP were co-infiltrated with Prf^D1416V containing plasmid in N. benthamiana leaves, along with Prf^D1416Valone, empty vector control (EV) or mock-infiltrated (Mock) controls. Prf^D1416V-induced cell lesions phenotype was assessed at 15 dpi. Representative images are shown in B and C (left panels). Quantification of the data shown in the graphs (right panels) was done by scoring the cell lesion phenotypes of five independent biological replicates as follows: 0 = no lesions, 0.5 = moderate lesions, and 1 = strong lesions. Data are represented as mean ± SEM (n = 5). Statistically significant differences in b were determined by the Kruskal-Wallis test followed by the Mann-Whitney U test for comparing treatments with the mock. p values of Mann-Whitney U tests are indicated. Statistically significant differences in c were determined using one-way ANOVA followed by Tukey’s HSD test. Unlike lowercase letters represent significant differences among the group means (p ≤ 0.05). F and p values of ANOVA are indicated. Data are represented as mean ± SEM.

Little is known about the functions of bacterial-encoded dual-specificity phosphatases. In humans, multiple dual-specificity phosphatases have demonstrated activity toward MAP-kinases (MAPKs) and mediate positive and negative cellular immune and inflammation responses.¹⁹ The dual-specificity phosphatases were also active targets in drug design for fighting diseases, cancer, and inflammation therapies.¹⁹ Given the significance of MAPK activations, phosphorylation, and dephosphorylation events in general during plant-pathogen interactions and signal transduction, we evaluated whether STP could interfere with plant immune responses. We utilized a well-established transient Prf-mediated cell death genetic assay.^20,21 Prf is a tomato NBARC-LRR resistance (R) protein that confers resistance to Pseudomonas syringae pv. tomato (Pst). Prf^D1416V is an autoactivated kinase mutant of Prf with single amino acid substitution at D1416 in the isoleucine-histidine-aspartic acid (IHD) motif that triggers a hypersensitive response (HR)-like programmed cell death (PCD) or lesions in plants in a pathogen or effector-independent manner.^20,21 The PCD phenotype was further quantitated by a scoring scale of 0 (no lesions), 0.5 (moderate lesions), and 1 (strong lesions) using five independent biological replicates.²⁰ Potential effectors or inhibitors of cell death can be examined by screening for suppressing Prf^D1416V -induced cell death. STPs encoded in a binary expression vector were co-infiltrated with Prf^D1416V in N. benthamiana (Figures 2B and 2C). Leaves infiltrated with Prf^D1416V alone, Prf^D1416V + empty vector (EV), and mock infiltrations were used as controls. As expected, Prf^D1416Valone and EV infiltration induced prominent lesions by 15 dpi (Figures 2B and 2C). In contrast, leaves co-infiltrated with Prf^D1416V + CLas or CLso STP showed a significant reduction of lesions (Figures 2B and 2C), similar to mock-infiltrated leaves (Figures 2B and 2C).

Molecular dynamic simulation identifies binding hotspots on STP tertiary structure

Genetic manipulation of C. Liberibacter spp. and its encoded proteins is not feasible due to their unculturable and obligatory lifestyle. Hence, we explored alternative approaches to block STP activity in planta. Target-based molecular modeling and ligand screening are commonly employed in biomedical and pharmaceutical research for drug discovery.¹² Fewer studies have employed this approach in plant-microbial interactions and are limited to studies of plant proteins.^13,14 To use these tools for identifying C. Liberibacter spp. inhibitors, we first generated homology models of CLso and CLas STPs, using experimentally determined structures of a protein template in the protein database bank.²² The homology model was built based on the template protein (PDB: 4RKK²³) using SWISS-MODEL via the HHblits protein sequence search package while enabling the highest sequence coverage and identity.²⁴ This selected template was a phosphatase with a carbohydrate-binding domain and a dual-specificity phosphatase domain. Glucose-6-phosphate (G6P) was used as a reference substrate to represent the phosphate ions and α-D-glucose native substrates. Both CLso and CLas STP structures overlapped on residue 40 to 191 with a 0.72 coverage rate and 0.29 sequence identity rate (Figure 3A). Visual molecular dynamics (VMD)²⁵ was used to depict the structural differences between the homology model CLso STP and CLas STP (Figure 3). Sequence alignment showed a homology value (QH) of 0.9563, root-mean-square deviation (RMSD) of 0.6389 Å, and percent identity of 62.5 which suggested that the two proteins had a high degree of structural similarity, in line with the multiple sequence alignment results (Figure 1). The two homology models had a good overlap on various substrate-binding domains with the template. Next, we identified binding hotspots (pharmacophores) using the CLso STP protein model. Identification of primary and allosteric (if present) substrate-binding site(s) of STP was initially approximated via Autodock Vina,²⁶ followed by in-depth analyses via nanoscale molecular dynamics (NAMD)²⁷ simulations and Drugui.²⁸ The goal was to accurately identify binding features in an environment resembling what the enzyme would encounter in vivo. NAMD combined Drugui was used to determine the binding features by sending multiple types and quantities of selected ligands to the dynamically changing receptor. To mimic the cytosolic environment in silico, the protein was immersed in a water sphere in the presence of different concentrations of the ligand(s). Each system was minimized and simulated to discern the hotspots where ligand(s) preferentially bound while the protein was still undergoing structural motions. Once the hotspots were identified, the binding features and locations (in 3D space) were recorded as pharmacophore points. From this, we identified at least sixty-seven potential binding hotspots (Figure 3B), with binding energies varying between −1.01 and −2.62 kcal/mol. The acetate drug-like probe had the highest binding affinity of −2.62 kcal/mol, and the site where acetate was bound was considered an active site. The interaction points that participate in substrate binding were further isolated by identifying clusters where several high-affinity hotspots lie closer, suggesting an active site (Figure 3C) (Table S3). The CLas STP pharmacophore clusters, the G6P substrate, and the template/homology model (PDB: 4RKK²²) were overlapped to determine whether the pharmacophores generated were in the active site. The protein-ligand structures after alignment showed that the selected CLas STP active site coincided with the dual specificity phosphatase (DSP) domain on laforin (Figure 3D).

Figure 3.

Homology modeling of CLso and CLas STP and identification of binding hotspots

(A) The homology model of CLso STP (gray) and CLas STP (cyan) is overlaid with the template (PDB ID: 4RKK in blue).

(B) Hotspot distribution when NAMD simulations were performed with select probes on the CLso STP enzyme. The red spot depicts a binding hotspot with the highest affinity. The area demarcated by the red box depicts pharmacophore distribution at the primary site.

(C) Multiple pharmacophores were clustered around the CLso STP enzyme, implying potential binding site(s).

(D) The selected pharmacophores (green) occupied a binding pocket in the CLas STP (gray mesh), consistent with the experimental substrate analog alpha-maltohexaose (red).

In silico ligand screening identified competitive inhibitors of STP

Ligand screening was performed in silico by pooling the features of both the ligand (via Autodock) and the receptor pocket (NAMD/Drugui) using ELIXIR-A.^29,30,31 This hybrid approach allowed the screening of a pool of pharmacophore features orders of magnitude larger than those that would have been possible when using either the ligand or pocket alone. The previous approach would allow for deciphering more effective and site-specific inhibitors. The zincpharmer algorithm was used to sift through a library of compounds for possible hits that match the identified pharmacophore with potential inhibitors.³² Lipinski’s rule of five and high pharmacophore scores in zincpharmer were used to narrow down the hits.³³ The screening efforts isolated two potential candidates with higher affinities (G6P3510, −5.38 ± 0.379 kcal/mol and G6P6373, −5.41 ± 0.438 kcal/mol) to the active binding site on CLso STP than a native substrate, glucose-6-phosphate (G6P, −4.74 ± 0.196 kcal/mol) (Table 1) (Figure 4A1–3; Figure S2A1–3). G6P is a primary metabolite in the cell that is fundamental to numerous biochemical processes and was used as a native substrate for the phosphatases. Both the inhibitors conformed with Lipinski’s rule of five, which defines properties necessary for good permeation³⁴ (Table 1). Based on the octanol-water partition coefficient (logP) analysis, G6P was likely to be hydrophilic, while G6P3510 and G6P6373 were hydrophobic due to aromatic groups. Furthermore, the G6P substrate appeared to have a higher bioactivity score (1.23) when compared to the two screened compounds (−0.19 and −0.54) (Table 1). Interestingly, the two compounds had distinct structures and functional groups, although they were predicted based on similar pharmacophores compared to the original substrate G6P. This underscores the value of large-scale screening of in silco chemical databases to identify unique chemistries or interactions.

Table 1.

Structural and functional information of potential inhibitor ligands with G6P control

Compound name	Glucose 6-phosphate	G6P3510	G6P6373
InChI Key	NBSCHQHZLSJFNQ-DVKNGEFBSA-N	FNDOMYALHUMMDS-SSDVNMTOSA-N	GKLIPLBJOGUXHG-QGMBQPNBSA-N
Octanol-water partition coefficient logP	−3.31	3.90	−0.87
molecular weight (Daltons)	260.13	420.85	373.31
number of hydrogen bond acceptors	9	8	10
number of hydrogen bond donors	6	2	1
“Rule of five” violations	1	0	0
Number of Rotatable Bonds	7	8	5
Enzyme inhibitor scorea	1.23	−0.19	−0.54
Average docking affinity on CLso STP, kcal/mol ± std (n = 10)	−4.74 ± 0.196	−5.38 ± 0.379	−5.41 ± 0.438
Average docking affinity on CLas STP, kcal/mol ± std (n = 10)	−4.38 ± 0.379	−5.30 ± 0.283	−5.27 ± 0.320

^aThe enzyme inhibitor score is one of the drug-likeness & bioactivity score provided by Molinspiration.

Figure 4.

Identification of competitive inhibitors (G6P3510 and G6P6373) that bind to CLas STP

(A) structural information of the screened compounds; (1) a native enzyme substrate G6P; (2) Screened inhibitor G6P6373; (3) Screened inhibitor G6P3510.

(B) The interactions of STP with G6P, G6P6373, and G6P3510, respectively.

(C) G6P (red) occupies the active site, and an isolated ligand (blue) occupies the same site as G6P on CLas STP (cyan).

For any ligand to be an effective drug (i.e., an inhibitor), it is essential that, after binding, the ligand blocks any possibilities for the enzyme-substrate to bind to the enzyme. This could happen only if the ligand binds more strongly to the pocket than the substrate and if the ligand occupies a vast majority of possible binding sites that would otherwise be available for the substrate to bind. Thus, it is critically important that the substrate-binding site(s) and inhibitor-binding site(s) overlap. Accordingly, the nature of interactions was deciphered via Autodock Vina docking simulation, this time in the presence of substrate(s) and potential inhibitor(s) on respective enzymes. Autodock Vina binding confirmations showed that the substrate and inhibitors had similar contact residue groups on CLso STP (Figure S2B). G6P formed van der Waals interactions with residues Asn 43 while forming hydrogen bonding with Tyr 160 and Thr 41 (with a distance of 1 Å). G6P3510 formed van der Waals interactions with residues Tyr 40, Asn 43, and Gln 57 while forming hydrogen bonding with Thr 41 (with a distance of 1 Å). G6P6373 formed van der Waals interactions with residue Thr 59 while forming hydrogen bonding with Tyr 40 and Thr 41 (with a distance of 1 Å). In a manner similar to the CLso STP interactions, the two inhibitors, G6P3510 and G6P6373, also displayed high binding affinities of −5.30 ± 0.283 kcal/mol and −5.27 ± 0.320 kcal/mol, respectively, vs. −4.38 ± 0.379 kcal/mol for G6P to the active binding site on CLas STP (Table 1) (Figure 4B1–3). The substrate and inhibitors showed similar contact residue groups on CLas STP (Figure 4C). G6P formed van der Waals interactions with residues Tyr 169 and Phe 40 while forming hydrogen bonding with Thr 41, Asn 43, and Asn 59 (with a distance of 1 Å). G6P3510 formed van der Waals interactions with residues Lys136, Asn 43, Glu 84, and Ser 85 while forming hydrogen bonding with Asn 59 (with a distance of 1 Å). G6P6373 formed van der Waals interactions with residue Phe 40, Gln 57, Asn 59 Pro 83, Ser 85, and Lys 136 while forming hydrogen bonding with Asn 43 and Thr 41 (with a distance of 1 Å). In summary, the in silico simulations identified multiple potential interactions between inhibitors and the STP proteins.

G6P3510 and G6P6373 interact with STPs via multiple intermolecular hydrogen bonds

MD simulations were performed to validate the performance and stability of the two small molecules in the binding pocket. The G6P/CLas STP, G6P3510/CLas STP, and G6P6373/CLas STP complexes were simulated for 100 ns to find the conformational changes within the active site. RMSD and root-mean-square fluctuations (RMSF) of the structures were used to determine whether the system had reached equilibrium during the simulation. The RMSD changes of G6P/CLas STP, G6P3510/CLas STP, and G6P6373/CLas STP complexes were analyzed to evaluate the stability of each complex during 100 ns MD simulations (Figure S3A). All three complexes plateaued after 20 ns of simulation and remained in the range of 2.5 Å to 4 Å. The final RMSDs were 3.082 Å, 3.683 Å, and 2.972 Å for G6P/CLas STP, G6P3510/CLas STP, and G6P6373/CLas STP, respectively, which reached equilibrium. The RMSF variances of the α-carbon residue of each residue in three complexes were calculated to find the flexibility of the protein structures. A higher RMSF value represents a much more flexible protein structure. Overall, the RMSF values of the three complexes ranged from 0.46 Å to 7.7 Å (Figure S3B). The outliers were from the tails. Although the N-terminal peptide covered the tails, they were still more flexible than the rest of the structure. No protein-ligand interactions were found at the last two amino acid positions, so the last two amino acids were excluded in subsequent RMSF analyses. After excluding these outliers, the RMSF ranged from 0.46 Å to 4.1 Å. Then, the average RMSF values were 1.33 Å, 1.32 Å, and 1.22 Å for G6P/CLas STP, G6P3510/CLas STP, and G6P6373/CLas STP, respectively. The two molecules screened, G6P3510 and G6P6373, caused lower RMSF values than G6P, indicating interactions between compounds and the protein (e.g., hydrogen bonding, hydrophobic interactions, ionic, and water bridge) had some impact on stabilizing the protein structure.

Intermolecular interaction analysis helps reveal the ligands’ binding properties at the active site. For instance, in the G6P/CLas STP complex, the protein-ligand interaction diagram indicated the formation of three strong hydrogen bonds during a 100ns simulation (Figure S4A). The residues Lys 73, Lys 129, and Lys 67 formed hydrogen bonds with G6P, which lasted 110.5%, 62.1%, and 61.9% of the simulation time, respectively. A ratio greater than 100% indicates that a single residue simultaneously forms multiple hydrogen bonds with the ligand. Meanwhile, water bridge interactions were also detected on these three residues. In the last 1 ns frame of the simulation, G6P also formed three salt bridges with the residue Lys 67 and Lys 73 at the active site, which increased the affinity for binding (Figure S5A). In the complex of G6P3510/CLas STP, multiple strong hydrogen bonds on residue Phe40, Thr41, Gln57, Asn 59, Lys 136, and Ser 137 were detected during the simulation, accounting for 71.9%, 124.8%, 15.1%, 2%, 72.2% and 20.6% of binding contribution during the MD simulation, respectively (Figures S4B and S5B). Due to the presence of an aromatic ring, strong hydrophobic interactions were detected at residues Phe 40, Ala 56, Tyr 169, Phe 172, and Pro 173, accounting for 8.1%, 7.6%, 14.9%, 46%, and 14.0%, respectively. Even though the small molecule remained in a fixed active site, stronger bonds were formed during the MD simulation. Also, the constant presence of hydrogen bonds indicated that the molecule bound stably at the active target site. In G6P6373/CLas STP complex, five central hydrogen bonds were detected at residue Gln 57 (32.6%), Gly 60 (28%), Lys 136 (70.4%), Thr 188 (21.2%), and Gln 189 (25.2%) (Figures S4C and S5C). Additionally, all these hydrogen bonds could switch to water-mediated ones during the simulation. Similar to the G6P3510 complex, this complex also detected more non-covalent bond formation during the simulation. All three molecules formed at least one or more stable non-covalent contacts with the protein throughout the simulations. In summary, the two screened compounds had better binding properties than G6P due to the observation of more protein-ligand contacts, and the simulation results confirmed the stability of the protein-ligand complexes.

G6P3510 and G6P6373 competitively bind to STP and inhibit phosphatase activity

To validate binding and determine the binding kinetics of the potential inhibitors of STPs, we performed ligand-binding assays using Bio-layer interferometry (BLI).³⁵ Briefly, recombinant CLso and CLas STPs were produced in Escherichia coli with C-terminal 6xHis tag and purified by one-step purification using Ni column (Figure S6). Based on BLI binding kinetics data, the two test molecules, G6P3510 and G6P6373 along with the positive control G6P interacted on both STPs of CLso and CLas generating association and dissociation signals (Figures 5A and 5B; Table 2). For CLso STP, the affinity constants, K_D, for G6P3510 and G6P6373 were 1.381 × 10⁻⁵ M and 1.42 × 10⁻² M, respectively (Table 2). The K_D for the native substrate G6P was 6.688 × 10⁻⁶ M. The K_D values indicated that G6P3510 had an affinity closer to that of G6P (Table 2). A molecule with an affinity closer to but higher than the native substrate suggests competitive inhibition. Compared to G6P, G6P6373 dissociated faster on CLso STP, leading to higher KD values and lower affinity. Compared to G6P, G6P3510 associated and dissociated faster; however, the compound’s affinity was slightly lower than that of G6P. For CLas STP, the K_D for G6P3510 and G6P6373 was 8.144 × 10⁻⁵ M and 2.401 × 10⁻⁵ M, respectively, while it was 2.361 × 10⁻⁵ M for the native substrate G6P, indicating that both compounds had a comparable affinity to that of G6P (Table 2). It was evident that both the compounds had a comparable association rate to that of G6P; however, G6P3510 dissociated faster than the other two. Overall, the BLI kinetics data indicate that G6P6373 had a higher affinity to the active site than G6P3510, yet a lower affinity than the G6P native substrate, suggesting competitive inhibition.

Figure 5.

G6P3510 and G6P6373 bind to STP and inhibit phosphatase activity in vitro

The Biolayer interferometry (BLI) sensorgram shows the binding of CLso STP (A) and CLas STP (B) with G6P3510 and G6P6373. Glucose-6-phosphate (G6P), a known phosphatase substrate, was used as a BLI assay reference control. All experiments were conducted in three replicates; each of the curves was globally fitted using three replicates. Phosphatase activity of CLso and CLas STP in the presence of 125 μM (C) or 250 μM (D) of G6P3510 or G6P6373. Assays with GFP were used as a negative control. Statistically significant differences were determined by the Kruskal-Wallis test, followed by the Dunn's test for multiple pairwise comparisons. The p values were corrected for type I error using the Benjamini-Hochberg multiple testing correction. Different lowercase letters represent significant differences among the group means (p ≤ 0.05). Data are represented as mean ± SEM.

Table 2.

The kinetic energy analysis of the protein-ligand interactions

Ligand	Targeted protein	KD (M)	Ka (1/Ms)	Kd (1/s)	R2
G6P	CLso STP	6.688 × 10−6	1.758×101	1.176 × 10−4	0.4832
G6P3510	CLso STP	1.381 × 10−5	1.064×102	1.469 × 10−3	0.8156
G6P6373	CLso STP	1.42 × 10−2	2.691×101	3.82 × 10−1	0.8876
G6P	CLas STP	2.361 × 10−5	2.166×103	5.114 × 10−2	0.7641
G6P3510	CLas STP	8.144 × 10−5	1.285×103	1.047 × 10−1	0.5428
G6P6373	CLas STP	2.401 × 10−5	1.924×103	4.618 × 10−2	0.6449

A known substrate, G6P was used as the positive control in this assay.

Next, we tested if G6P3510 and G6P6373 can inhibit the STP phosphatase activity in vitro. For this, CLso STP and CLas STP were expressed in Arabidopsis protoplasts as C-terminal HA-tag fusions and immunopurified using anti-HA beads. A GFP HA-tagged protein was simultaneously expressed and purified as a non-phosphatase control. The immunopurified STP and GFP proteins were treated with 125 μM or 250 μM G6P3510 (dissolved in 0.1% DMSO, along with mock (0.1% DMSO), followed by phosphatase assays. The results indicated that 125 μM and 250 μM G6P3510 and G6P6373 significantly inhibited CLas/CLso STP phosphatase activity compared with the mock treatments (Figures 5C and 5D). Collectively, the results show that G6P3510 and G6P6373 can bind to and inhibit STP phosphatase activity in vitro.

G6P3510 and G6P6373 inhibit STP virulence function in planta

Since STP interfered with Prf^D1416V-mediated cell death^20,21 (Figure 2), we next tested if the G6P3510 and G6P6373 can block STP virulence function in planta. STP was co-infiltrated with Prf^D1416V in N. benthamiana (Figure 6). G6P3510 (50 μM, dissolved in 0.1% DMSO) or G6P6373 (50 μM, dissolved in 0.1% DMSO) were co-infiltrated with Prf^D1416V and CLas STP (Prf^D1416V+STP+G6P3510 or G6P6373), along with mock (DMSO 0.1%), Prf^D1416V alone, Prf^D1416V + STP, Prf^D1416V + EV, Prf^D1416V + EV + G6P3510 or G6P6373, and Prf^D1416V + G6P3510 or G6P6373. Leaves infiltrated with Prf^D1416V alone, Prf^D1416V + EV, and mock infiltrations served as controls. Five biological replicates were performed for each treatment (Figure 6). As expected, Prf^D1416Valone and EV infiltration resulted in the induction of prominent lesions by 15 dpi (Figure 6). In contrast, leaves co-infiltrated with Prf^D1416V + CLas or CLso STP showed a significant reduction of lesions (Figure 6), similar to mock-infiltrated leaves (Figure 6). However, leaves co-infiltrated with G6P3510 or G6P6373 and Prf^D1416V + STP, the cell death lesions were strong and comparable to Prf^D1416V alone or other Prf^D1416V + EV controls (Figure 6). Collectively, these results support that G6P3510 and G6P6373 can inhibit STP activity in plant tissues.

Figure 6.

G6P3510 and G6P6373 inhibit STP activity in planta

A. tumefaciens co-infiltrations were performed by mixing 50 μM of G6P3510 (A and B) or G6P6373 (C and D) with Prf^D1416V, CLas STP (Prf^D1416V+STP+G6P3510 or G6P6373), along with mock (DMSO 0.1%), Prf^D1416V alone, Prf^D1416V + STP, Prf^D1416V + +empty vector (EV), Prf^D1416V + EV + G6P3510 or G6P6373, and Prf^D1416V + G6P3510 or G6P6373. Prf^D1416V induced cell lesions were assessed at 15 dpi. Representative images are shown in (A) and (C). Quantification of the data shown in (B) and (D) was done by scoring the cell lesion phenotypes of five independent biological replicates as follows: 0 = no lesions, 0.5 = moderate lesions, 1 = strong lesions. Data are represented as mean ± SEM (n = 5). Statistically significant differences were determined by one-way ANOVA followed by Tukey’s HSD test. Different lowercase letters represent significant differences among the group means (p ≤ 0.05). F and p values of ANOVA are indicated in (D). Data are represented as mean ± SE (n = 5). Statistically significant differences were determined by the Kruskal-Wallis test followed by the Dunn test for multiple pairwise comparisons. The p values were corrected for type I error using the Benjamini-Hochberg multiple testing correction. Different lowercase letters represent significant differences among the group means (p ≤ 0.05). p values of the Kruskal-Wallis test are indicated in (D). Data are represented as mean ± SEM.

G6P3510 and G6P6373 inhibit Candidatus Liberibacter spp. in plant tissues

Since STP interferes with plant immune responses (Figure 6) and is expressed in infected plant tissues (Figure 1), we next evaluated the effects of G6P3510 and G6P6373 on “Candidatus Liberibacter spp.” accumulation. A major bottleneck to evaluating chemistries or therapies for “Candidatus Liberibacter spp.” has been the fastidious or unculturable nature of the pathogens. Conventional assays rely on foliar sprays, stem injections, and grafting using whole plants, which can be inherently variable due to the challenges associated with the delivery of the inhibitors to the bacteria that reside unevenly within the mature plant vasculature.³⁶ We have recently demonstrated that plant hairy root matrices allow ex vivo propagation of “Candidatus Liberibacter spp.” and can be used for faster antimicrobial assays and citrus genetic engineering.^37,38,39 Because hairy roots possess intact phloem and xylem that are anatomically and metabolically similar to normal plant vasculature,^40,41,42 they would also provide an in planta environment that is similar to the natural host environment for CLso and CLas when compared to cell-free systems and using surrogate bacteria.^43,44,45,46

The efficacy of the two potential inhibitors (G6P3510 and G6P6373) was first tested using the CLso-potato and CLas-citrus hairy root cultures. Briefly, CLso/CLas containing hairy roots were treated with G6P3510 and G6P6373 (5, 10, 20, and 50 μM, dissolved in 0.1% DMSO) for 72 h. Untreated and DMSO (0.1%) samples were used as controls. After treatment, all tissue samples were subjected to molecular diagnostics, and relative titers of the CLso and CLas were estimated by qPCR analysis. Both G6P3510 and G6P6373 inhibited CLso (FDR corrected p ≤ 0.01) at all the dosages tested (5, 10, 20, and 50 μM) when compared to untreated controls (Figures 7C and 7D; Table S4). G6P3510 and G6P6373 also showed inhibitory activity against CLas at most tested dosages (FDR corrected p ≤ 0.05 to 0.01) (Figures 7E and 7F; Table S5).

Figure 7.

G6P3510 and G6P6373 inhibit ‘Candidatus Liberibacter spp.’ in potato and citrus hairy root tissues

(A and B) Potato and citrus hairy roots (HR) infected with CLso and CLas were treated with mock (DMSO) or 5, 10, 20, and 50 μM of G6P3510 (C1) and G6P6373 (C2) for 72 h. Relative titers of CLso (C and D) and CLas (E and F) were estimated after 72 h of treatment by respective compounds, followed by qPCR analysis, along with untreated (UT) and DMSO (0.1%) treated samples. The bacterial titers were estimated relative to untreated samples set to 100%. Data are represented as mean ± SEM (n = 5 for [C and D] and n = 3 for [E and F]), and p values were estimated using one-way ANOVA followed by a post-hoc LSD test. The p values were corrected for multiple comparisons using the Benjamini-Hochberg (BH) FDR method. The experiment was independently repeated twice, and all replication attempts were successful.

It is possible that G6P3510 and G6P6373 may inhibit unrelated plant- or other pathogen-encoded serine/tyrosine phosphatases. To investigate this, first, we performed BLASTP analysis of CLas/CLso STPs and the predicted proteomes of potato, citrus, and Pseudomonas syringae. We did not find any significant matches to citrus proteins, while one significant (E-value <0.006) match was recovered in the potato (XP_006352385.1) and P. syringae (WP_052963584.1) proteomes. Both these matches, however, had poor coverage (∼40–60%) and poor sequence conservation (∼49–55% similarity) when compared to the high degree of conservation (>95% similarity) among the Ca. Liberibacter STPs (Figure 1A), suggesting that non-target effects on unrelated STPs are limited. Second, we tested the activity of G6P3510 and G6P6373 directly on P. syringae pv. tomato DC3000’ (PstDC3000) growth in vitro. The results showed that neither of the molecules was inhibitory to PstDC3000 growth (Figure S7), suggesting that non-target effects on unrelated microbes may be limited and underscoring the robustness of the target-based approach for finding specific inhibitors.

For preliminary proof of activity in planta, we focused on “Candidatus Liberibacter spp.”-infected potato and citrus plants maintained in the greenhouse. The leaf tissues of these plants are generally amenable to penetration by either foliar spraying³⁷ or syringe infiltration.⁴⁷ For potato treatments, four-week-old plants were challenged with psyllids carrying CLso, followed by foliar spraying G6P3510 and G6P6373 (10 and 50 μM, dissolved in 0.1% DMSO). The timeline, dosages, and frequency of applications were according to previously optimized methodologies.³⁷ Briefly, eight sprays were performed bi-weekly after the psyllid challenge. Chlorosis, wilting, and stunting disease symptoms on aerial shoots were monitored weekly, and qPCR-based molecular diagnostics were performed at 28 days post-infection (dpi). Characteristic symptoms of CLso-infected potatoes, such as leaf curling and wilting, appeared in the control/untreated plants as early as 21 dpi and were exacerbated by 28 dpi (Figure 8A). Most plants died by 5–6 weeks. In contrast, the disease symptoms of plants treated with G6P3510 and G6P6373 (10 and 50 μM, dissolved in 0.1% DMSO) were visually less prominent compared to the untreated controls at 28 dpi (Figures 8A and 8B). PCR diagnostics revealed a statistically significant (Tukey’s HSD p ≤ 0.01 to 0.005) decrease in CLso accumulation in plants treated with G6P3510 and G6P6373 when compared to untreated controls (Figure 8C; Table S6), which correlated with the attenuated disease symptoms on the plants. Neither of the compounds exhibited phytotoxicity symptoms (e.g., chlorosis, necrosis, or stunting) on healthy plants throughout the developmental stages (Figure S8).

Figure 8.

G6P3510 and G6P6373 cf. tolerance to “Candidatus Liberibacter spp.”

(A and B) Visual symptoms of healthy and CLso-infected potato plants sprayed with mock (DMSO), 10 μM, and 50 μM solutions of G6P6373 (C1) and G6P3510 (C2). Red arrows indicate characteristic disease symptoms of CLso-infected potatoes, such as leaf curling and wilting.

(C) Relative titers of CLso in the mock, C1, and C2 treated plants at 28 days post-treatment.

(D) Leaves of CLas-infected citrus plants were infiltrated with mock (0.1% DMSO) or 100 μM solutions of G6P6373 (C1) and G6P3510 (C2).

(E) Relative titers of CLas in the mock, C1, and C2 treated plants at 28 days post-treatment. The relative bacterial titers were plotted relative to mock samples set to 100%. Data are represented as mean ± SEM (n = 3–5), and p values were estimated using one-way ANOVA followed by post hoc Tukey’s HSD test.

For citrus in planta treatments, two fully expanded leaves of four to five-month-old CLas-positive citron rooted plants (n = 3 to 5 plants per treatment) were syringe-infiltrated⁴⁷ with a 0.1% DMSO (mock) solution or 100 μM solution of C1 and C2 (dissolved in 0.1% DMSO). The plants were subsequently maintained in the greenhouse, and leaves were sampled after 28 days to determine the relative CLas titers. The qPCR molecular diagnostics results showed significantly (p ≤ 0.01) reduced CLas titers in C1 and C2 treated plants compared to mock plants (Figures 8D and 8E; Table S6). Combined with the hairy root bioassay results (Figure 7), the greenhouse plant trials sufficiently demonstrate the antimicrobial activity of the two molecules against Candidatus Liberibacter spp. in planta.

Limitations of the study

This study successfully employed an interdisciplinary approach of target-based molecular modeling and ligand screening to identify two pharmacophore-based inhibitors with demonstrated antimicrobial activity against “Candidatus Liberibacter spp.” in hairy root assays and greenhouse plant experiments. Additional research would be needed to advance these antimicrobials as commercial products for practical application to manage HLB. For instance, unlike herbaceous crops like potatoes or greenhouse citrus plants, there are challenges associated with delivering active ingredients (AIs) into mature citrus trees in the field that are large, woody, and perennial. The waxy cuticle of the leaves is a barrier to the penetration of AIs by foliar application. Alternatively, trunk injections could deliver AIs into the citrus trees. However, only a few commercial products, mainly those based on oxytetracycline, were successfully delivered by trunk injections to manage HLB in citrus trees.⁸

In addition to possessing antimicrobial activity, the success of any new therapy in the field conditions is predicated on the physicochemical properties (e.g., solubility, pH, stability) of the AI that allow it to penetrate the plant cells and translocate systemically. In the current study, we used DMSO (0.1%), an EPA-approved inert ingredient (solvent or cosolvent) for pesticide formulations/products (https://www.federalregister.gov/d/2022-22129) to solubilize the AIs for foliar application and leaf infiltration studies. Further translocation and tracing studies in planta are needed to determine the systemicity of the new AIs. This will also aid in the selection of optimal carriers and formulation. Notably, multiple EPA-approved commercially available surfactants or carriers can be evaluated to enhance the absorption and subsequent translocation of AIs in planta.

Candidatus Liberibacter spp. accumulates considerably high levels in sink tissues such as young leaves and distal roots. Furthermore, in the case of HLB, citrus roots may act as a reservoir for CLas between periods of foliar flushing and seasonal fluctuations.^48,49 Hence, understanding the systemic movement and translocation of the AIs in planta can also help determine the impacts of the AIs on the bacterial communities in the root systems.

Another limitation to consider is the production scale and availability of the new AIs. Depending on the age and size of the citrus tree, the volume to treat could be much larger than the plants in the greenhouse, hence requiring significantly higher amounts of AIs. Because the new AIs are experimental compounds, the current synthesis scales are smaller (Table S7). The synthesis needs to be scaled up for large-scale field tree studies and applications to produce sufficient quantities. This will also help lower the costs of the interventions and make them profitable (Tables S7 and S8).

Lastly, as with any new AIs entering commercialization, they must undergo rigorous human and environmental safety assessments and obtain federal (e.g., EPA, FDA) regulatory approvals, likely in collaboration with the industry, before they can be safely deployed as commercial pest control products.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-HA peroxidase monoclonal antibody	Roche	Cat# 12013819001; RRID: AB_390917
Mouse Anti-His monoclonal antibody	Genscript	Cat# A00186
Bacterial and virus strains
Candidatus Liberibacter spp.	This study	N/A
Chemicals, peptides, and recombinant proteins
CLso STP recombinant protein	This study	N/A
CLas STP recombinant protein	This study	N/A
G6P3510 chemical	This study	MCULE ID-8749857699-0
G6P6373 chemical	This study	MCULE ID-4680703503
Experimental models: Organisms/strains
Solanum tuberosum L. var. Atlantic	This study	N/A
Bactericera cockerelli (Šulc)	This study	N/A
Oligonucleotides
Primers used for qPCR and RT-qPCR	This study	Table S2
Recombinant DNA
pHBT-binary vector containing CLas or CLso STP-HA	This study	N/A
PrfD1416V plasmid	Du et al., 2012	N/A
Software and algorithms
ELIXIR-A	GitHub	https://github.com/sfernando-BAEN/ELIXIR-A
	GitHub	https://github.com/sfernando-BAEN/ELIXIR-A-Vina-Batch-Screening-Module
Bioinfokit	GitHub	https://github.com/mandadi-lab/bioinfokit
Other
Genbank accession numbers for ‘Candidatus Liberibacter spp.’ STP proteins.	This study	Table S1
BLItz/Octet N1 system with Anti-Penta-HIS (HIS1K) Biosensors	Sartorius	https://www.sartorius.com/en/products/protein-analysis/octet-bli-detection/octet-label-free-detection-systems/octet-n1

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kranthi Mandadi (kkmandadi@tamu.edu).

Materials availability

There are restrictions on the availability of some materials due to the limited amount of the original stocks and/or quarantines associated with regulated pathogens. Direct inquiries regarding the materials generated as part of this study should be directed to the lead contact, Kranthi Mandadi (kkmandadi@tamu.edu).

Data and code availability

• •All data is made available in the manuscript and related supplementary files. Accession numbers are listed in the supplementary tables and key resources table. The Python bioinfokit package used for data analysis and visualization was deposited on GitHub [https://github.com/mandadi-lab/bioinfokit].
• •The source code for algorithms used for pharmacophore mapping was deposited in GitHub [https://github.com/sfernando-BAEN/ELIXIR-A] and [https://github.com/sfernando-BAEN/ELIXIR-A-Vina-Batch-Screening-Module]. All code is publicly available and was last accessed on 29 Jan 2024.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Plant and insect propagation

Potato (Solanum tuberosum L. var. Atlantic) plants were grown in Metro-Mix® 360 (Sungro Horticulture, MA, USA) in a growth chamber under the following conditions: 21-22°C, a 14-h/10-h light/dark photoperiod, and 50% relative humidity. Bactericera cockerelli (Šulc) psyllids, both CLso-positive (haplotype B) and CLso-negative, were collected in the Rio Grande Valley in Texas,⁵⁰ and maintained in the laboratory in nylon cages (Bugdorm, Taiwan). The presence and haplotype of ‘Candidatus Liberibacter solanacearum’ (CLso) in psyllid colonies were periodically confirmed by PCR using primers specific for the single sequence repeat (SSR) loci of CLso (SSR-F and SSR-R, Table S2).⁵¹

Methods detail

STP comparative genomics and phylogenetics and expression analysis

C. Liberibacter spp. STP sequences and NCBI accession numbers used for multiple sequence alignment and phylogenetic analyses are provided in Table S1. Multiple alignments of protein sequences were performed with prank PRANK v.170427.⁵² Phylogenetic analyses were performed using a maximum likelihood approach in RAxML 8.2 with 1,000 bootstrap replicates.⁵³ The resulting phylogeny was visualized using FigTree 1.4.3.⁵⁴ Expression of STP in CLso- and CLas-infected potato and citrus tissues was confirmed using reverse-transcription (RT)-PCR. Briefly, total plant RNA was isolated from 50 mg of symptomatic leaf tissue using the Direct-zol RNA Miniprep kit (Zymo Research, Irvine, CA), following the manufacturer’s instructions. The RNA quality and concentration were evaluated by gel electrophoresis and NanoDrop1000 spectrophotometer, respectively (Thermo Fisher Scientific, Waltham, MA). One μg of RNA was used to prepare complementary DNA using Invitrogen Superscript IV Reverse Transcriptase following the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA), followed by PCR amplification using STP gene-specific primers (Table S2).

Homology modeling, pharmacophore identification, and ligand screening

The inhibitor screening was performed in several steps, beginning with developing the STP homology model, identifying active sites and pharmacophores, and then ligand/inhibitor screening. Simulation platforms, including Autodock Vina,²⁶ Visual Molecular Dynamics (VMD),²⁵ Drugui,²⁸ Enhanced Ligand Exploration and Interaction Recognition Algorithm (ELIXIR-A),^29,30,31 Nanoscale Molecular Dynamics (NAMD)²⁷ were used during in silico analysis. ELIXIR-A, the algorithm created for pharmacophore mapping, was deposited in GitHub [https://github.com/sfernando-BAEN/ELIXIR-A] and [https://github.com/sfernando-BAEN/ELIXIR-A-Vina-Batch-Screening-Module]. NAMD simulations were performed in the Texas A&M High Performance Research Computing. To perform MD simulations using NAMD, four separate files were used: A Protein Data Bank (PDB or .pdb) file that stores atomic coordinates and velocities for the system; A Protein Structure File (PSF or .psf) that holds structural information of the protein, such as various types of bonding interactions; A force field parameter file that contains the mathematical expressions of the potentials experienced by the atoms in the system; and a configuration file in which the user specifies all the options that NAMD should use in running a simulation. These files were created by substrate ligands and probe molecules (small molecule alcohols, amines, amides, and esters) that were used to interrogate the enzymes’ active sites. Small probes were selected so they could diffuse rapidly and explore small and even transient pockets in the simulations. This property also helps sample many binding events while enabling equilibrium in reasonable timeframes. The binding hotspots (pharmacophores) were isolated by selecting high-affinity probe binding spots, clustering them, and merging them to assess high-affinity sites.²⁸ All the interaction diagrams were developed using Schrodinger Maestro software. ELIXIR-A^29,30,31 and the zincpharmer algorithm were used to screen and match the identified pharmacophores with potential inhibitors.³² The top hits were selected based on the following criteria: i) a combination of the location of the functional groups (e.g., proton donor/acceptor, hydrophobic groups, positive/negative ion, and exclusion spheres); ii) stabilization of the most effective conformation; iii) Lipinski’s rule of five that defines properties necessary for good permeation³⁴ (i.e., the molecule having less than five proton-donators, the molecular weight is less than 500 Daltons, log P smaller than 5, the molecule having < 10 acceptors, and the molecule using biological transporters so that the ligand does not attach too strong); iv) Molinspiration drug-likeness & bioactivity score (www.molinspiration.com), and v) having at least three pharmacophoric points.

Molecular dynamics simulation

A 100 ns MD simulation was performed using the Schrödinger-Desmond platform to analyze STP proteins' conformational and interaction stability in complex with screened ligands.⁵⁵ The CLas STP protein was prepared by Maestro’s Protein Preparation Wizard.^56,57 C and N termini were capped to reduce the flexibility of the termini. For the screened ligands, Schrödinger's Glide extra-precision mode was used to reproduce the docking poses from the previous Vina docking studies.⁵⁸ The ligand-protein complexes were minimized under the OPLS3e force field.⁵⁹ The system was solvated using an SPC solvent model with orthogonal boundary conditions maintaining a buffer distance of 5 Å. The system charge was neutralized using chloride or sodium ions. Each system was minimized under the OPLS3e force field using Desmond's default relaxation protocol. The system was simulated under isothermal–isobaric conditions at 300 K and 1.01325 bar pressure. A thousand frames were recorded at every 100 ps following the initial frame. The Simulation Interaction Diagram (SID) module was used to assess the quality of the entire simulation over time, including protein-ligand interactions, complex root means square deviations (RMSD), and complex root means square fluctuations (RMSF).

STP expression, protoplast transfection, immunopurification, and phosphatase activity assays

The endogenous expression of STP in CLso and CLas infected plant tissues (potato and citrus) was confirmed by Reverse Transcription Polymerase Chain Reaction (RT-PCR) analysis. Briefly, total RNA was extracted from healthy or infected shoot tissues (50 mg) using a Direct-zol RNA Miniprep kit (Zymo Research, Irvine, CA), following the manufacturer’s instructions. The RNA concentration, integrity, and quality were checked using agarose gel electrophoresis and NanoDrop1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA). Using a random hexamer primer, one microgram of total RNA was converted to complementary DNA (cDNA) using a Superscript IV first-strand cDNA synthesis kit (Invitrogen). The cDNA was subsequently used for PCR amplification using CLso and CLas STP-specific primers (Table S2).

For STP expression in Arabidopsis protoplasts and immuno-purification, the pHBT-binary vector-based CLas/CLso STP-HA tagged constructs were made and used for transfection. Briefly, the Arabidopsis leaf strips were excised from the 4-week-old plants. They were lysed in an enzyme solution (1% cellulase R10, 0.2% macerozyme R10, 0.4 M mannitol, 20 mM KCl, 20 mM MES pH 5.7) at room temperature for 3-4 hours. The protoplasts were re-suspended in 0.5 mL W5 (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, 2 mM MES pH 5.7) solution by gentle shaking. Before PEG-Ca²⁺ transfection, the W5 protoplast solution was resuspended in MMg solution (0.4 M mannitol, 15 mM MgCl₂, 4 mM MES pH 5.7) at a density of 5 x 10⁵-5 x 10⁷ cells /mL. This density of protoplasts provided an excellent range to achieve measurable phosphatase activity. Next, 40% (w/v) PEG solution with 0.2 M mannitol and 100 mM CaCl₂ was used for PEG-Ca²⁺ transfection. After transfection, the protoplasts were resuspended in 50 μL WI (0.5 M mannitol, 20 mM KCl, 4 mM MES pH 5.7). Lastly, protoplasts were aliquoted into two wells of a 24-well tissue culture plate with 250 μL WI, and proteins were allowed to express by overnight incubation at room temperature. For immunopurification, the protoplasts were lysed by vertexing in an extraction buffer (100 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 7.5, 2 mM NaF, 2 mM Na₃VO₄, 1 mM dithiothreitol (DTT), 0.5% Triton X-100, 10% glycerol and 1×protease inhibitor). After centrifugation at 14,000 g at 4°C for 20 min, 250 μL of extraction buffer was added to resuspend the pellet and incubated with anti-HA affinity beads (Sigma) at 4°C for 2h with gentle shaking. Beads were collected and washed three times with washing buffer (20 mM Tris-HCl, pH7.5, 100 mM NaCl, 1 mM EDTA, 1% Triton X-100) and once with 50 mM Tris-HCl, pH7.5. The beads were then resuspended in 100 μL 1 X phosphatase assay buffer (25 mM HEPES, pH 7.5, 0.2 mM EDTA, 10 mM DTT, 200 μg/mL BSA), and stored on ice until initiation of the phosphatase assay. An aliquot of the immunoprecipitated protein beads was analyzed by SDS-PAGE and immunoblotting with anti-HA peroxidase antibodies (Roche) to confirm the presence of the proteins before the phosphatase activity assays.

The CLso/CLas STP phosphatase activity was measured using p-Nitrophenyl Phosphate (pNPP) chromogenic substrate. In this assay, dephosphorylation of pNPP can be measured at absorbance (OD) 405 nm and used to calculate the enzymatic activity of the phosphatase. Briefly, 10 μL pNPP (Thermofisher, 250 mM in 1 x phosphatase assay buffer) was mixed with 40 μL immunoprecipitated STP protein beads in 1 x phosphatase assay buffer and incubated at 30°C for 30-45 minutes. The reaction was stopped by adding 1 mL of 1 N NaOH, and absorbance (OD) was read at 405 nm using the microplate reader (Agilent BioTek Cytation 7 Multimode Reader by Agilent Technologies). The enzyme activity was estimated using the following formula:

Enzyme activity [μmoles/min μg] = 50 [vol] x OD405 x 1/time [min] x enzyme [μg] x 1/ 18,000 [molar extinction coefficient].

G6P3510 and G6P6373 molecules were custom synthesized and purchased from Mcule. Chemical specifications (Formula, InChIKey, Smiles) and Mcule IDs of G6P3510 and G6P6373 are provided in Table S7. Next, the effects of G6P3510 and G6P6373 on CLso/CLas phosphatase activity were determined by pre-treating immunopurified proteins with G6P3510 and G6P6373 (125 μM or 250 μM, along with DMSO as mock, for around 15-30 min followed by addition of the pNPP to start the reaction at room temperature (2). After reaction completion, the OD405 absorbance was measured using the microplate reader. The enzyme activity was calculated using the equation above.

Bio-layer interferometry (BLI) and binding kinetics assays

Bio-layer interferometry (BLI) experiments were conducted using the BLItz/Octet N1 system with Anti-Penta-HIS (HIS1K) Biosensors (Sartorius, Goettingen, Germany). BLI is an optical detection technique that can determine the affinities of protein-ligand interactions on the sensor by reflecting a light beam.⁶⁰ STP protein with the hexahistidine (His₆) tags was attached to the fiber optic sensor tip with a selective tag immobilization method. The sensor tip was dipped into the small-molecule solution medium with a known concentration to load the small molecule ligands onto the sensor. The association of small molecules to the sensor represented by the association constant, k_a, causes the white light to shift in the interference pattern. The unbinding phenomenon, represented by the dissociation constant, k_d, will cause the white light to shift in the opposite direction on the sensor. The dissociation/association ratio defines the affinity constant for the binding interactions (K_D) and is represented by the equation as follows:

$$K_D=\frac{\left[A\right]·\left[B\right]}{\left[AB\right]}=\frac{k_d}{k_a}$$

Where K_D represents the affinity constants (M, or mol·L^-1), k_d represents the dissociation rate (s^-1), and k_a represents the association rate (M·s^-1). Here, [A], [B], and [AB] represent the concentrations (M) of the unbound STP protein, small molecule ligands, and protein-ligand complex, respectively. The affinity constants will reveal the overall binding kinetics of the protein-ligand complex, which helps to evaluate the performance of the screened molecules.

The binding kinetics of CLso and CLas STP and the screened molecules G6P6373 and G6P3510 were performed using recombinant STP proteins purified from E. coli and stored in 50 mM Tris-HCl, 500 mM NaCl, 10% Glycerol, 2 mM DTT, pH 8.0 buffer. The HIS1K sensors were hydrated in a 250 μL assay buffer for 30 minutes. The assay buffer was prepared with 9.95% 10X Pall Forte Bio kinetics buffer (Sartorius, Goettingen, Germany ) and 89.55% PBS (Thermo Fisher Scientific, Waltham, MA), and 0.5% Dimethyl sulfoxide (DMSO) (Sigma-Aldrich, St. Louis, MO). The BLI test used a cycle of 30s assay buffer baseline, 300s STP protein loading, 120s baseline, 120s analyte association, and 120s analyte dissociation. The assay buffer was used in the two baseline and last dissociation phases. The concentration of protein and analytes was optimized for significant and measurable binding signals. For the CLas STP protein assay, 4 μL of CLas STP protein with 0.34 mg/ml concentration was loaded, and the testing analyte 1 μM G6P, 50 μM G6P6373, or 50 μM G6P3510 was associated in a 250 μL assay buffer, respectively. For the CLso STP protein assay, 4 μL of CLas STP protein with 0.30 mg/ml concentration was loaded, and the testing analyte G6P, G6P6373, or G6P3510 was associated with 1 μM in a 250 μL assay buffer, respectively. All the BLItz experiments were conducted at room temperature. The Kinetics energies were analyzed in BLItz Pro 1.3.0.5 software (Sartorius, Goettingen, Germany) per manufacturer recommendations, and the interaction curves were exported from BLItz Pro and replotted with R studio.

In vitro and in planta efficacy trials

For the hairy root bioassays, CLso-potato and CLas-citrus hairy roots were generated using CLso- and CLas-infected plant tissues.³⁷ The presence of CLso and CLas in the hairy roots was confirmed by qPCR amplification of the 16s rDNA⁶¹ and Ribonucleotide reductase β-subunit gene, nrdB (RNR),⁶² respectively (Table S2). The validated hairy roots were transferred to a multi-well titer plate containing Gamborg’s B5 media. G6P3510 and G6P6373 inhibitors at 0, 5, 10, 20, and 50 μM were added, vacuum-infiltrated, and incubated for 72 hrs. Untreated and DMSO (0.1%) were used as negative controls, respectively. Three to five biological replicates were used for each treatment. After the treatments, tissue samples were treated with PMAxx (propidium monoazide) dye (Biotium, Fremont, CA) to inactivate dead bacterial DNA. Subsequently, total DNA was extracted, and quantitative PCR (qPCR) analysis was carried out to quantify the viable bacterial titers and plotted relative to the untreated controls (set to 100%) using the ^ΔΔCt method.⁶³ After normalization of target Ct with an endogenous reference gene (Ct') to correct for DNA template concentration differences among the samples, it was plotted relative to untreated controls. The untreated was set to 100% for ease of visual representation.³⁷ Statistically significant differences among the controls and treatments were estimated using one-way ANOVA followed by a post-hoc LSD test (Figure 5). The p values were further corrected for multiple comparisons using the Benjamini-Hochberg (BH) FDR method.⁶⁴

Non-target effects of G6P3510 and G6P6373 on unrelated bacteria were evaluated with Pseudomonas syringae pv. tomato DC3000 (PstDC3000). DC3000 was cultured in a modified Luria-Bertani (LM) liquid medium either in the absence or presence of 5, 10, and 20 μM of G6P3510 and G6P6373 at 28 °C. Growth was recorded by measuring optical densities (OD 600 nm) at 0, 2, 4, 6, 8, 10, and 24 h and was converted to colony forming units (CFU)/80 μL.

For the in-planta trials, one-month-old potato plants (var. Atlantic) grown in a growth chamber at 21°C were challenged with 20 psyllids carrying CLso. The challenged plants were maintained in secured cages to confine the insects to the plants. The psyllids were also restricted to two lower branches of the plant and contained using an organza drawstring bag tied around the branch. After one week, the bags containing psyllids were clipped off the plant to prevent the escape of the newly emerging psyllids and nymphs. All the challenged plants were sprayed with either 0, 10, and 50 μM of G6P3510 and G6P6373 (3-4 biological replicate plants/treatment) twice a week for up to four weeks.³⁷ The visual disease symptoms on the plants were monitored weekly, and tissue samples were collected at 28 days post-challenge for molecular diagnostics. Phytotoxicity assessments were performed similarly by treating healthy potato plants with 50 μM of G6P3510 and G6P6373 and monitoring visual symptoms of chlorosis, necrosis, or stunting.

For citrus treatments, four to five-month-old CLas-infected citron rooted plants maintained in the greenhouse were used for leaf infiltration. All the leaves of the plants were syringe-infiltrated⁴⁷ with either 0.1% DMSO (0.1%) solution or 100 μM solution of G6P3510 and G6P6373 (3 to 5 biological replicate plants/treatment) (Figure 8D). After 28 days, leaf tissues were sampled and subjected to genomic DNA isolation and qPCR molecular diagnostics.

DNA isolation and qPCR diagnostics

Total DNA was isolated from ∼100 mg of tissue homogenized using Precellys 24 homogenizer (MO BIO Laboratories, Carlsbad, CA, USA).⁶⁵ The quality and quantity of DNA were determined using a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA) and by agarose gel electrophoresis. Approximately 50 ng of total DNA was used for qPCR analysis, performed on a CFX384™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc., Hercules, CA, USA) using iTaq™ universal SYBR® Green supermix (Bio-Rad Laboratories, Inc.), following the manufacturer’s instructions. The 16s rDNA Lso-F and HLB-R primers were utilized for CLso detection,⁶¹ while RNR primers were used for CLas detection⁶² (Table S2). A dissociation melting curve analysis showed no non-specific amplification or primer dimers. Relative CLso and CLas titers were estimated using the ^ΔΔCt method.⁶³ The CLso and CLas Ct values were normalized to RIBOSOMAL PROTEIN L2 (StRPL2) and the citrus GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE C2 (GAPC2) to correct for DNA template concentration differences among the samples. They were plotted relative to the untreated controls, which were set to 100%.

Quantification and statistical analysis

All the data analysis, statistics, and graphs were displayed using Microsoft Excel software (version 2009) and the Python bioinfokit package (https://github.com/mandadi-lab/bioinfokit). The graphical abstract was created with Biorender (https://BioRender.com).

Published: February 15, 2024