International standard strain C. sakazakii BAA-894 (ATCC, United States) was used in this study. Strain was stored in Luria-Bertani (LB) media (Oxoid, United Kingdom) containing 15% glycerol (Biosharp, China) at −80°C. To initiate all experiments, one loop of strain was innoculated in LB and cultured overnight. And then bacterium was diluted in LB media containing different concentrations of pCA to 10⁶ CFU/mL; the growth or inactivation of bacteria was monitored by serial dilution of bacteria and plating on LB agar. The plates were incubated at 37°C for 24–48 h, and then colony forming unit (CFU) was numbered.

Bacterial cells were washed and resuspended in phosphate buffered saline (PBS) (Vazyme, China). The bacterial suspension (5 × 10⁹ CFU) were administered by gavage to 3-day-old female Sprague Dawley rats (6/group). Where appropriate, pCA was administered at a dose of 1 mg/g weight 2 h after bacteria challenging, while an equal volume of pCA was administered without bacterial inoculation as the negative infection control and bacterial inoculation but without pCA treatment as positive infection control. Rats were maintained in their home cages in the animal house at 24 ± 1°C and 55 ± 5% humidity, with a 12-h light-dark cycle (light on at 8:00 and off at 20:00) To analyze the colonization of bacteria in blood and organs, the rats were sacrificed after 24 h of injection; organs were homogenized in ice-cold PBS and serially diluted. Bacterial load was determined by plating the diluents on LB agar.

The bacterial suspension from in vitro experiment was centrifuged at 4°C for 2 min at 12,000g and washed twice with PBS. A protease inhibitor cocktail (Beyong, China) was added, and the bacteria were lysed by ultrasound (200 W, working 3 s and pausing 3 s) for 10 min. The protein extractions were centrifuged to obtain the supernatant (21,000g, 15 min at 4°C). Clean up the protein using a microporous filter (0.22 μm), and the extracted protein was stored at –80°C for subsequent MS analysis.

The protein samples for LC-MS/MS analysis were prepared according to the protocols described previously with minor modifications (Xue et al., 2022). Specifically, the protein sample (50 μg, qualified by optical density at 260 nm) was denatured by adding one-fourth 8M urea (Aladdin, China). Add 50 mM ammonium bicarbonate (Aladdin, China) to make the total volume reach 100 μL. Incubate the mixture at 37°C for 30 min. Subsequently, the protein sample was reducted by reacting with 1 μL 200 mM Dithiothreitol (DTT) (Aladdin, China) and incubated for 30 min at 60°C. Cooldown to room temperature for 5 min. The protein was alkylated by adding 1 μL 500 mM iodoacetamide (Aladdin, China). Keep shaking for 30 min without light. Then quench iodoacetamide by adding DTT (1 μL, 200 mM) and vortex for 10 min. The samples were subsequently digested by trypsin (Thermo, United States) (4 μL, 0.25 μg/μL) at 37°C overnight. Terminate digestion by adding 10% trifluoroacetic acid (Aladdin, China) to the final concentration of 0.4%. At last, the samples were desalted with a C18 SPE column (Millipore, United States) and dried under vacuum. The vacuum-dried samples were submitted for MS analysis.

The digested peptide samples were analyzed using a Q Exactive Plus mass spectrometer and an EASY nano Liquid chromatography (EASY nLC 1200, Thermo Scientific) with an EASY nanoelectrospray interface according to the methods described before (Xue et al., 2022). The nano liquid chromatography system was equipped with a Thermo Scientific Acclaim Pepmap nano-trap column (C18, 5 μm, 100 Å, 100 μm × 2 cm) and a Thermo Scientific EASY-Spray column (Pepmap RSLC, C18, 2 μm, 100 Å, 50 μm × 15 cm). The nano liquid chromatography used solvent A (0.1% formic acid) and solvent B (80% CH₃CN/0.1% formic acid), whose gradients were as follows: 0–8% B for 3 min, 8–28% B for 42 min, 28–38% B for 5 min, 38–100% B for 10 min. The mass spectra were searched against the UniProt database, and the MS raw data for each sample were searched using Maxquant (Version 2.0.3.1). Related parameters and instructions were as follows: samples with carbamidomethylation of cysteine set as a fixed modification. Oxidation (M) is set as the variable modifications. Searches were performed with trypsin cleavage specificity allowing two miscleavage events. The precursor mass tolerance was set to 10 parts-per-million (ppm) and a fragment mass tolerance of 0.02 Da. A maximum false discovery rate (FDR) of 1.0% was set for protein and peptide identifications. Protein identification was based on at least one unique peptide identification. Protein quantification was calculated as the median of unique peptides of the protein.

The crystal structures of all the proteins used in the docking were predicted based on Alphafold2 (Jumper et al., 2021); the 3D structures of the small molecules were downloaded from the PubChem database (Kim et al., 2016), and the Chem3D V20 was used to minimize the energy in the MMFF94 field (Hajos, 2002). AutoDock Vina 1.1.2 was used to perform molecular docking (Gaillard, 2018). Before docking, PyMOL 2.5 was used to treat all receptor proteins, including removing water molecules, salt ions, and other small molecules (Skern, 2018). Then set up the docking box using the PyMOL plug and define the center of mass of the active site residue as the center of the docking box with the side length set to 22.5 angstroms. In addition, use ADFRsuite 1.0 to convert all processed small molecules and receptor proteins into the PDBQT format required for AutoDock Vina 1.1.2 docking. When docking, the global search verbosity is set to 20, leaving the rest of the parameters set by default. The output of the highest-scoring docking conformations is considered to be the binding conformations, and finally, PyMol is used for visual analysis of docking results.

All experiments were conducted in triplicate, as independent experiments. Gene Ontology (GO) and KEGG Enrichment Analysis (KEGG) analysis was performed on DAVID’s online website¹ and enrichment scores were calculated automatically (Dennis et al., 2003). Fuzzy C-means (FCM) algorithm was conducted with R (Bezdek et al., 1984; Ihaka and Gentleman, 1996). Statistical analyses were performed with GraphPad Prism software or R as described in Experimental Procedures for individual analysis. For all, p < 0.05 was considered statistically significant.

The effect of pCA on the growth of C. sakazakii is shown in Figure 1A. The bacteria in the control group began to grow exponentially after 20 mins and reached stationary phase at 6 h. Cronobacter sakazakii was inhibited at the concentration of 300 μg/mL pCA, and reached the stationary phase after 9 h. When the concentration of pCA reached 500 μg/mL, the growth of bacteria was inhibited entirely. These results indicated that pCA could inhibit the growth of C. sakazakii in vitro. We also studied the effect of pCA in vivo (Figure 1B). Rats were used as the infection model. In the pCA treatment group, the rats received gavage treatment of pCA (1 mg/g weight) 2 h after bacterial infection, while an equal volume of saline was administered as the control. The results showed that pCA could significantly improve the survival rate of infected rats. Moreover, pCA treatment significantly reduced the bacterial density of C. sakazakii in the blood, liver, and spleen of the rat sacrificed 1 day after infection (Figure 1C). Taken together, these results suggest that pCA can effectively inhibit bacterial growth in vivo and in vitro.

Using label-free relative quantitative mass spectrometry, we systematically monitored the protein expression profiles of C. sakazakii at 2 mins, 30 mins, 1 h, and 2 h after pCA exposure (Figure 2A). C. sakazakii not exposed to pCA at the initial stage and after 2 h of normal growth were set as controls. We obtained 15,621 unique peptides that matched 1,553 proteins. Among them, 905 proteins met the quantitative requirements (Figure 3A). T-test (p < 0.05) was used to determine the up-and down-regulated peptides according to the criterion that the folding change of differential expressed proteins was more than two times (Figure 2B and Supplementary Table 1). Compared with the control group, 2 mins of pCA exposure could significantly change 39 protein abundances, and abundance changes increased over prolonged exposure time to pCA (Figure 2C). After 2 h of pCA treatment, there were 90 up-regulated proteins and 113 down-regulated proteins. Surprisingly, however, the bacterial proteome showed even more changes after 2 h in standard growth control. It is suggested that prolonged time and drug treatment can cause changes in the proteome.

To distinguish the pCA-induced proteome changes from time-induced proteome changes, fuzzy C-means (FCM) algorithm was used to cluster the protein expression profiles at different time stages (Figure 3 and Supplementary Table 2). In general, we observed eight different time-pattern clusters representing different regulatory proteins, indicating different expression dynamics. Clusters 2 and 4 represent down-regulated proteins, clusters 6 and 7 represent up-regulated proteins, clusters 1, 3, and 5 represent proteins showing bimodal expression patterns, and cluster 8 characterizes proteins which expression does not change much. When time interference was excluded by superimposing the protein expression 2 h after normal growth, some proteins were continuously upregulated or downregulated, clustered in clusters 7 and 8, respectively. The up- and down-regulated proteins accounted for 11 and 17% of the total 905 proteins. Among the core-altered proteins (membership > 0.58), 40% were downregulated, 3.3 times as many as the upregulated. Overall, we found that pCA caused protein expression changes and more downregulated proteins than upregulated proteins, consistent with the observed inhibition of bacterial growth exposed to pCA.

GO analysis revealed that the altered proteins were enriched in the molecular function (Figure 4 and Supplementary Table 3). The NADP binding and glycine decarboxylation via glycine cleavage system were significantly enriched among molecular function clusters. The upregulated protein was significantly enriched in NADP binding. NADP-binding proteins are associated with the metabolic status of bacteria, so we hypothesize that pCA causes changes in the metabolic status of bacteria. In addition, pCA-induced up-regulated proteins were enriched in the biological process of the Protoporphyrinogen IX biosynthesis, and Protoporphyrinogen IX was reported to play an essential role in the biosynthesis of heme, suggesting that the bacteria were in an iron-deficient state. It is well documented that polyphenols can chelate iron ions, and perhaps pCA can competitively chelate environmental iron ions, making it difficult for bacteria to obtain iron. Additionally, pCA-induced down-regulated proteins were enriched in seven biological-function-related clusters and one cellular component-related cluster. However, pCA-up-regulated proteins were enriched in only two clusters. Both up-regulated and down-regulated proteins were found in three clusters: the plasma membrane, ATP binding, four iron, and four sulfur cluster binding. Still, the number of down-regulated proteins was more than up-regulated in the same cluster. Collectively, GO analysis infers that pCA leads to alterations in the bacterial metabolic state and an overall metabolically repressed state.

By performing KEGG pathway enrichment analysis, the down-regulated proteins were significantly enriched in seven pathways (p < 0.05), and the up-regulated proteins were enriched in only one pathway (Figure 5 and Supplementary Table 3). pCA distorted L-phenylalanine, tyrosine, and tryptophan biosynthesis. Of the seven pathways down-regulated, “metabolism” was the most common, including several subclasses: “glyoxylic and dicarboxylic acid metabolism,” “fructose and mannose metabolism,” “arginine and proline metabolism,” “glycine, serine and threonine metabolism.” In addition, some important categories were also significantly enriched (e.g., “phosphotransferase system,” “oxidative phosphorylation”). These results illustrated that pCA might inhibit cellular energy metabolism while the balance of amino acid metabolism might be disturbed.

Three proteins were identified that were affected by pCA to increase deamidation significantly. These include: A7MKI5, which has 64 amino acid L-asparagine deamidation, A7MGU5: 43 amino acid L-asparagine deamidation, A7MKM7:379 amino acid L-asparagine deamidation. Notably, all three proteins have important effects on the physiological functions of bacteria. A7MKI5 is a crucial protein of bacteria, termed Elongation factor Tu (EF-TU), a G protein with the intrinsic ability to hydrolyze GTP into GDP, contributes to the overall translation fidelity, and plays a vital role in the elongation process of extension of protein biosynthesis. And A7MGU5 is an ancillary SecYEG translocon subunit, which is thought to act as a molecular chaperone, helps maintain the typical conformation of proteins in bacteria. A7MKM7 is a Peptidase M3 domain-containing protein with zinc ion metal-binding ability and metallopeptidase activity. In summary, we found that the effect of pCA on posttranslational modifications mainly affected the deamidation of bacteria, including three critical proteins.

Molecular docking was performed to help understand the structure-based correlation between the deamidation of proteins and the structure of pCA (Figure 6). The structures of the three proteins affected by deamidation were predicted using alpha fold, and then molecular docking results were obtained using autodock Vina. We obtained the binding affinities of individual combinations. A negative binding affinity indicates the presence of binding. The binding energy of A7MKM7/pCA was −4.9 kcal/mol, followed by A7MKI5/pCA with a binding energy of −4.8 kcal/mol and A7MGU5/pCA complex, the binding energy is −3.5 kcal/mol. As shown in Figure 6A, in the A7MGU5/pCA complex, pCA was in contact with Gln-46, Trp-42, Asn-43, and Trp-39 on a long helix, where phenolic hydroxyl forms a hydrogen bond with Gln-46 The aromatic ring forms a pipi-stack action with the Trp directly below. In the A7MKI5/pCA complex (Figure 6B), pCA was contact with Phe-219, Glu-216, Asp-217, Arg-289, Asn-91, Asn-64, and Thr-65 on the A7MKI5 protein, in which the carboxyl group forms a hydrogen bond with the amide bond on the backbone of asn-64, which is consistent with the deamidation site analyzed by the proteomic data. It also includes hydrophobic interaction with Phe-219 and Glu-216. In the A7MKM7/pCA complex (Figure 6C), pCA came into contact with Glu-383, Lys-386, Lys-394, Glu-395, Arg-407, Phe-409, Ala-393, and Phe-382 on the A7MKM7 protein, the phenolic hydroxyl group formed a hydrogen bond with Glu-383, and the carboxyl group formed a hydrogen bond with amide bond on the main chain of GLU-395. Hydrophobic interaction formed with Phe-409, Ala-393, and Phe-382.