In this study, the strains used were A. hydrophila ATCC 7966 (wild type strain, WT), E. coli MC1061, E. coli S17 λpir, suicide vector pRE112 and shuttle vector pBBRMCS1, which were all kept in our laboratory ( Supplementary Table S1 ). Both E. coli and A. hydrophila were cultured in fresh LB liquid medium, and incubated at 200 rpm 37°C and 30°C, respectively. When needed, chloramphenicol (Cm) and ampicillin (Amp) were added to the LB medium.

The deletion of the target gene was performed using homologous recombination, as previously described (Wang et al., 2019). Briefly, about 500 bp upstream and 500 bp downstream sequences of the target gene was amplified and then fused in pRE112 suicide vector. The recombinant plasmid was transformed into E. coli MC1061 competent cells, and the plasmid of the positive clone was extracted and transformed into E. coli S17 competent cells. The E. coli S17 with recombinant plasmid was conjugated with WT at a ratio of 1:4 (v/v), and then screened in LB agar plate with 30 μg/mL Cm and 100 μg/mL Amp. The selected colonies were further screened and inoculated into the LB agar plate containing 20% sucrose or 30 μg/mL Cm, respectively. Finally, the positive colonies were picked up to verify by PCR and DNA sequencing, and then stored at −80 °C before use.

The complemented ΔuidR strain was constructed using ligase (Clon Express II One Step Cloning Kit, Vazyme) to ligate the gene sequence of with the digested pBBRMCS1 plasmid (HindIII and BamHI), producing a recombinant plasmid. Finally, the recombinant plasmid was transformed into ΔuidR strain by electroporation to generate the complemented strain of ΔuidR, which was finally confirmed by PCR and DNA sequencing. The pair primers for gene deletion and the construction of the complemented strain used in this study were listed in Table 1 .

The overnight bacteria were separately diluted 1:100 in fresh LB medium and thoroughly mixed, then 300 μl bacteria solution were added into the holes of the growth curve plate and cultured overnight for 16 h at 30°C. Bacterial growth was measured using a growth curve analysis system (Bioscreen C, BioRad) at 600nm. There were three independent replicates of this experiment.

The biofilm formation of bacterial strain was analyzed by crystal violet staining method, as previously described (Wang et al., 2019). In brief, the overnight bacterial strains were transferred to fresh LB medium at the ratio of 1:100 and incubated until the logarithmic growth phase (OD600nm ~1.0) at 200 rpm, 30 °C. The culture bacterial solution was then diluted at 1:100 and transferred to a 96-well microplate, and incubated at 30 °C for 24h stationary incubation. Then, the wells were washed slowly with double-distilled water for three times and then dried out in air. Subsequently, 250 μL of 0.1%crystal violet solution (w/v) was added in each well, and then stained at room temperature for 30 min. After that, the wells were rinsed slowly with double-distilled water for three times and then dried again. Finally, 300 μL of 95% ethanol solution (v/v) was added to each well to dissolve the dyed crystal violet, and the absorbance at 595 nm was detected using a SpectraMax^® i3 multifunctional microplate reader (Molecular Devices Corporation, California, USA). Each sample in the experiment was repeated in eight wells for technical replicates, and performed three independent times for biological replicates.

The preparation of protein in biofilm was performed as previously described (Li et al., 2016). Briefly, the overnight bacterial strain was transferred to fresh 5 ml LB medium at the ratio of 1:100, and then incubated at 30°C to an OD₆₀₀ of 1.0. The cultured bacterial solution was diluted at 1:200 in a polystyrene Petri dish (Fisher Scientific, Franklin, MA) and incubated at 30°C for 24h stationary incubation. After being washed carefully three times with PBS, the mature biofilm protein was obtained by leaching twice with PBS buffer. The bacterial cells were then washed three times with PBS and centrifuged at 12,000 rpm, 30 min, 4°C, and resuspended in lysis buffer (6 M urea, 2 M thiourea, 0.1 M Tris-HCl (pH 8.5), protease inhibitor). The bacterial suspensions were then ruptured by sonication at 4°C for 15 mins and centrifuged (12,000 rpm, 30min, 4°C) to collect the supernatant. The supernatant was transferred to a new tube, and the protein concentration was determined by Bradford assay. Approximately, 50 μg extracted protein from each group was reduced with 100 mM dithiothreitol (DTT) for 1 hour at room temperature, and alkylated with 50 mM iodacetamide (IAA) in the dark for 30 min in UT buffer (8 M urea, 0.1 M Tris, pH 8.5). The proteins were then digested by trypsin (Promega Corporation, Madison, USA) at 1:50 ratio by a FASP (Filter aided sample preparation) method, as previously descried (Wisniewski, 2016). The digested peptides were desalted by a C18 column (Waters, Massachusetts, USA) and then dried down by a CentriVap concentrator (Labconco Corporation, Missouri, USA).

The digested peptides were dissolved in 0.1% Formic acid (FA) and separated on an EASY-nLCTM 1200 UHPLC system (Thermo Fisher Scientific, Massachusetts, USA) and analyzed with an Orbitrap Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific, Massachusetts, USA) as previously described (Zhang et al., 2020). The scanning mode of mass spectrometry is data-dependent acquisition (DDA), with an electrospray voltage of 2.3 kV, a capillary temperature was 320 °C, and a full MS scan with a scan range of 350 to 1500 m/z and automatic gain control (AGC), a maximum ion implantation time of 45 ms, a normalized collision energy of 27%, an intensity threshold of 8.3×10³, and a dynamic exclusion parameter set to 60 s. Each sample in this experiment was repeated in three for technical replicates, and performed three independent times for biological replicates.

The raw MS data were searched using Maxquant 1.6.17.0 against A. hydrophila ATCC 7966 Uniport database. Proteins with a protein abundance ratio difference greater than 1.5 or less than 0.667 and a P value <0.05 were considered as differentially expressed proteins (DEPs) and submitted for subsequent bioinformatics analysis.

Gene Ontology (GO) annotations were classified and enriched according to GO terms of biological processes (BP), cell components (CC), and molecular functions (MF). Briefly, DAVID online software (https://david.ncifcrf.gov/) was used to perform GO enrichment analysis of DEPs, and then the GO plot package in the R language software was performed to visualize (Damian et al., 2015; Li et al., 2020). The KEGG pathways in differential expressed proteins were enriched and analyzed by the online tool OmicsBean (http://www.omicsbean.cn/) and then visualized with the GO plot package in the R language software. The prediction of protein-protein interaction (PPI) of DEPs network was analyzed by STRING version 11.5 (https://string-db.org/) and visualized using Cytoscape version 3.8.2 (Shannon et al., 2003; Damian et al., 2018).

The qPCR assay of biofilm was performed as previously described (Cai et al., 2019). Briefly, the total RNA of each bacterial sample in biofilm state was extracted according to the TRIzol-chloroform (Takara standard Co. LTD., Japan) extraction method. Then, the total RNA was reverse-transcribed into cDNA according to the operating steps of the Prime Script RT reagent Kit (TransGen Biotech, Beijing, China)instructions. Finally, the real-time fluorescent quantitative PCR was used to detect the transcription regulation-related genes using the Real-Time PCR Detection System (Bio-Rad, California, USA) instrument with the pair primers listed in Supplementary Table S2 . Each sample in this experiment was repeated in eight wells for technical replicates, and performed three independent times for biological replicates.

The ΔuidR strain was successfully constructed using a two-step homologous recombination method. As shown in Figure 1A , when compared to the wild-type (WT) strain, the DNA fragment of uidR was absent in ΔuidR strain using the PCR pair primers (P5P6) for target gene amplification. Additionally, the band of PCR product of ΔuidR strain was significantly lower than WT strain when using PCR pair primers (P7P8) to amplify the upstream and downstream range of the target gene, indicating that the knocked-out strains were successfully constructed. The positive clone was selected and further confirmed by DNA sequencing.

In order to investigate the effect of the uidR deletion in A. hydrophila, the bacteria growth and the biofilm formation were measured. As shown in Figure 1B , the growth between the ΔuidR and WT strain had no significant difference. Meanwhile, the crystal violet staining analysis showed that the ΔuidR strain significantly increased the ability of biofilm formation when compared to WT strain (P<0.001) ( Figures 1C, D ). This was further confirmed by fluorescence microscope using acridine orange staining method, which showed that the fluorescent intensity of ΔuidR in biofilm was significantly stronger than WT ( Figures 1E, F ).

Moreover, the complemented strain was validated by PCR ( Supplementary Figure S1 ) and DNA sequencing, and the property of biofilm formation in the complemented strain was restored to the WT level ( Figure 2 ). These data indicated that the deletion of uidR negatively affected the A. hydrophila biofilm formation, and inferred that the TetR family transcriptional regulator may play an important role in the regulation of the biofilm formation in A. hydrophila.

To further investigate the role of uidR as a transcriptional regulator, we used a label-free quantitative proteomics method to compare the DEPs of the ΔuidR and WT strain. As shown in Figure 3A , a total of 1496 proteins were identified by mass spectrometry, with 220 proteins being differentially expressed between the ΔuidR and WT strain, including 120 up-regulated and 100 down-regulated proteins in the ΔuidR strain. Moreover, the correlation analysis of protein intensity among three biological repeats in each group showed that these samples had a high reproducibility with a correlation coefficient >0.96, indicating the stability and reliability of the proteomics data ( Figure 3B ).

To further understand the functions of these identified proteins, we classified the 220 DEPs by Gene Ontology (GO) categories, including biological process (BP), cell component (CC), and molecular functioning (MF). As showed in Figure 4A , the DEPs participated in diverse biological functions in the BP classification. Among DEPs, 55 proteins were involved in a single-organism metabolic or catabolic process, followed by 47 DEPs that were involved in small molecule metabolic process. According to the CCs enrichment, most of the DEPs were distributed in the cytoplasm, some were distributed in the oxidoreductase complex and the outer membrane ( Figure 4B ). As shown in the MF classification ( Figure 4C ), the increasing abundance proteins were mostly enriched in the oxidoreductase activity, cofactor binding, CH-OH group acting on the donor in the activity, and the glycerol-3-phosphate dehydrogenase activity and ubiquinone-8 oxidoreductase activity. The decreasing abundance proteins were preferring to NAD or NADP as the acceptor, and CH-OH group as the donor processes, while anion binding process presented unbiased.

We further performed KEGG enrichment analysis on DEPs, and a total of 63 proteins were enriched, which participated in a total seven metabolic pathways. As shown in Figure 4D , 25% of DEPs was involved in the metabolic pathways of microorganisms in diverse environments, 22% for the carbon metabolite pathway, 13% for the propionate metabolic pathway, 13% for the pyrimidine metabolic pathway, and 11% for glyoxylate and dicarboxylate metabolism. Since previous research had reported that the proteins related to glyoxylate and dicarboxylate metabolism pathway were involved in bacterial biofilm formation in Marinobacter hydrocarbonoclasticus (Vaysse et al., 2009), we were then interested in the role of these proteins on uidR mediated biofilm formation in this study.

PPI relationships among the DEPs between the ΔuidR and WT strain in biofilm were established based on the STRING online software. As shown in Figure 5 , nine DEPs (cysk, hisC, thrC, AHA_2966, AHA_4139, ilvE, AHA_1270, kbl, AHA_0245) were mainly enriched in pyridoxal phosphate (FDR=0.0451), eight DEPs (AHA_0245, mmsA-1, AHA_2266, pflB, ackA-2, prpB, AHA_4081, gldA) were mostly enriched in glycerolipid metabolism (FDR=0.0059), five DEPs (AHA_2091, treC, AHA_1466, psuG, AHA_0115) were enriched in glycosidase (FDR=0.0451) and eight DEPs (gldA, AHA_2461, AHA_2460, glpC, AHA_1652, glpK, AHA_4006, dhaK) were enriched in propanoate metabolism (FDR=0.0355). Interestingly, there were still a series of proteins with similar functions showing significant differences in some pathways which had no different significance in the FDR value, such as nine DEPs (AHA_3367, AHA_0417, cheW-1, cheB-1, AHA_1033, AHA_3970, AHA_0422, AHA_1038, AHA_3469) in chemotaxis and six DEPs (clpA, groS, AHA_0007, AHA_0861, cspD, AHA_0008) in stress response.

Furthermore, PPI analysis also showed that seven altered proteins including two down-regulated proteins (AHA_1951 and AHA_4140), and five up-regulated proteins (AHA_3062, AHA_3063, aceA, aceB and gltA) were involved in glyoxylate and dicarboxylate metabolism pathway (FDR=0.0355) ( Figure 6A ). The results showed that most of DEPs in this pathway interacted with each other, and most of them were up-regulated in uidR deletion strain, indicating uidR may negatively regulate the expression of these proteins during biofilm formation in A. hydrophila.

To further validate the accuracy of the proteomics data, six genes involved in the glyoxylate and dicarboxylate metabolism pathway (AHA_1951, AHA_4140, AHA_3062, aceA, aceB and gltA) were selected to compare their mRNA levels between the ΔuidR and WT strain in biofilm by qPCR method. As shown in Figure 6B , when compared to WT strain, the transcription of AHA_3062, aceA, aceB and gltA were significantly increased, while AHA_4140 and AHA_1951 were significantly decreased in biofilm formation, being consistent with the proteomics results. And the results indicated the reliability of the quantitative proteomics data.

In order to further verify the effects of proteins related to the glyoxylic acid and dicarboxylic acid metabolic pathway on the biofilm formation of A. hydrophila, four genes (AHA_3062, AHA_3063, AHA_4140 and aceB) involved in this pathway were deleted to evaluate their biofilm formation capabilities by crystal violet staining method. As shown in Figure 6C , the deletion of these four genes significantly reduced the biofilm formation capability in A. hydrophila when compared with WT stain, respectively. These results further validated the role of the genes related to toglyoxylate and dicarboxylate metabolism in biofilm formation. Moreover, since the proteins encoded by these genes were up-regulated in ΔuidR strain in proteomics results, suggesting that the TetR family transcription factor may negatively regulate glyoxylate and dicarboxylate metabolism related genes and then affect the biofilm formation in A. hydrophila.