The strain Sphingomonas sp. WG (CCTCC No. M2013161) was maintained on Luria-Bertani (LB) agar slants at 4°C and was activated in seed medium (10 g/L glucose, 1 g/L yeast extract (YE), 5 g/L tryptone, 2 g/L KH₂PO₄, and 0.1 g/L MgSO₄) at 28°C. The activated strain was sequentially transferred to a 250 mL flask containing 50 mL fermentation medium (67 g/L glucose, 3.4 g/L YE, 3 g/L K₂HPO₄, 0.1 g/L MgSO₄, pH 7.0; Zhou et al., 2017), and incubated at 32°C, 150 rpm for 72 h to produce WL gum. To amplify the recombinant plasmid DNA, Escherichia coli DH5α was chosen as the host strain, and related recombinant E. coli DH5α strains were the donor strains in bacterial conjugation. E. coli DH5α was cultivated at 37°C in LB medium supplemented with tetracycline or gentamicin when necessary. The plasmid pJQ200SK containing the sacB gene as the counter-selectable marker (Pelicic et al., 1996) was applied to construct the welA deletion strain △welA. The plasmid pBBR1MCS-3 (Kovach et al., 1995) was used to introduce welA gene into the △welA to construct the complemented strain △welA/pBBR1MCS-3-welA. The plasmid pET32a⁽⁺⁾ and E. coli BL21(DE3) were used to express the possible diguanylate phosphodiesterase encoded by Gene 1094. Both plasmids pJQ200SK and pBBR1MCS-3 were obtained from Wuhan Miaoling Bioscience and Technology Co., Ltd. Enzymes used in molecular cloning, including restriction enzymes, LA Taq DNA polymerase, and the DNA Ligation Kit were obtained from TaKaRa Biotechnology (Dalian, China). KOD FX DNA Polymerase was purchased from TOYOBO (shanghai) Biotech Co., Ltd. Primers were synthesized by Shanghai Sangon Biotechnology (Shanghai, China). All antibiotics used were obtained from Sigma Chemical Co. (St. Louis, MO, United States). Other reagents were obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China) and Shanghai Macklin Biochemical Co., Ltd.

Based on the amino acid similarity to GelA in gellan biosynthesis pathway (Harding et al., 2004), the gene welA (Gene 3707) was annotated on the whole genome of Sphingomonas sp. WG (GenBank no. LNOS00000000) (Li et al., 2016a) and its deduced protein WelA was obtained and analyzed by multiple bioinformatical tools. Firstly, the physical and chemical properties, hydrophilicity of WelA were predicted using ProtParam¹ and ProtScale² on the ExPASy Server (Wilkins et al., 1999). Furthermore, the related amino acid sequences were retrieved in the GenBank database using BlastP³ (Altschul et al., 1990) program, and the phylogenetic tree was constructed using ClustalX version 2.0 (Larkin et al., 2007) and the MEGA7 program (Kumar et al., 2016). The secondary structure prediction and domain architecture analysis were performed by SOPMA⁴ (Geourjon and Deleage, 1995) and SMART⁵ (Letunic et al., 2015), respectively.

welA gene was knocked out from the chromosome of Sphingomonas sp. WG by double homologous recombination methods. Firstly, the upstream and downstream fragments (approximately 1,000 bp) of welA were amplified by PCR with primers welA5flFor/welA5flRev and welA3flFor/welA3flRev containing the tails with identity to the ends of each other (underlined) (Supplementary Table 1), respectively. Secondly, the obtained fragments were gel purified and fused into one DNA fragment with the help of their tails by fusing PCR reaction using “nested” primers welAdelFor/welAdelRev containing SacI and XbaI recognition sites (underlined). Thirdly, the resulting fusion product was inserted into the SacI-XbaI digested pJQ200SK to obtain the recombinant plasmid pJQ200SK-△welA. Finally, the pJQ200SK-△welA was transferred into Sphingomonas sp. WG from E. coli DH5α by triparental conjugal mating with the help of E. coli pRK2013. The single-crossover recombinants were screened by streptomycin and gentamicin. The deletion strains that undergo double-crossover recombination were screened on LB plates supplemented with 5% sucrose since single-crossover recombinants bearing the sacB gene are lethal in the presence of sucrose. The single-crossover recombinants were identified by PCR using two pairs of primers welA5flFor/welAinRev and welAinFor/welA3flRev, and the deletion strain △welA was confirmed by PCR using welAinFor/welAinRev as primers.

A welA fragment was inserted into the plasmid pBBR1MCS-3 to complement the welA gene. The ORF of welA was obtained by PCR using the Sphingomonas sp. WG genomic DNA as the template and welAexFor/welAexRev (The underlined parts were XhoI and XbaI recognition sties) as primers. Subsequentially, the gel-purified PCR product was digested by XhoI and XbaI and ligated into the linearized pBBR1MCS-3. The recombinant plasmid pBBR1MCS-3-welA was confirmed by colony PCR and DNA sequencing. Finally, the complemented strain of △welA with pBBR1MCS-3-welA was obtained by triparental conjugal mating and screened on LB supplemented with streptomycin and tetracycline. The complemented strain designated as △welA/pBBR1MCS-3-welA was further identified by colony PCR using the specific primers welAinFor/welAinRev for partial welA gene. The wild type containing the empty vector pBBR1MCS-3 named as WT/pBBR1MCS-3 was constructed through triparental conjugal mating using the method as described in our previous work (Li et al., 2021).

To elucidate the possible role of WelA in the EPS production, four different strains, wild-type of Sphingomonas sp. WG (WT), △welA,△welA/pBBR1MCS-3-welA and WT/pBBR1MCS-3 were cultivated in 50 mL fermentation medium in 250 mL shake flasks at 32.5°C, 200 rpm for different time. The fermentation parameters such as the biomass, the WL gum production, and the fermentation broth viscosity were measured as previously described (Li et al., 2018). The specific growth rate μ and time for the bacterial population to double (tD) were calculated according to the growth curve. The EPS production was determined with the phenol-sulfuric acid colorimetric method described previously using glucose as the standard (Zhou et al., 2017). The viscosity of fermentation broth was measured on a Brookfield Viscometer DV-III equipped with an ultra-low adaptor (Brookfield Engineering Laboratories) using the rotor spindle LV3 at 5 rpm at 25°C.

The WT, △welA, and △welA/pBBR1MCS-3-welA strains were cultured in the LB medium at 30°C until OD₆₀₀ was 0.6. The motility was tested on SIM media or fermentation medium supplemented with 0.3% (for swimming motility) and 0.5% (for swarming motility) of agar where 5 μL of bacterial suspension was dropped in each plate center and inoculated for 72 h (for swimming motility) and for 168 h (for swarming motility) at 28°C. The radius of circles formed by bacterial migration was measured to express the motility.

Cells of WT and △welA at the exponential phase (about 16 h of fermentation) were collected and harvested by centrifugation, respectively. Total RNA was extracted using RNAiso Plus (TaKaRa Biotechnology Company, China), and RNA integrity was assessed by an Agilent 2100 Bioanalyzer system (Agilent Technologies, CA, United States). RNA with a RIN value higher than 8.0 was chosen to produce a transcriptome library. According to the manufacturer’s instructions, cDNA libraries were generated using a NEBNext Ultra Directional RNA library prep kit for Illumina (NEB, United States). The cDNA library was sequenced on an Illumina sequencing platform (Illumina HiSeq™ 4000 platform) by the Shanghai Personal Biotechnology Co., Ltd. Library quality was assessed on an Agilent Bioanalyzer 2100 system.

Raw data (raw reads) in fastq format were filtered by removing low-quality reads, reads containing adapters, and ploy-N to obtain high-quality clean data used in all downstream analyses. The high-quality clean reads were mapped to the reference genome of Sphingomonas sp. WG (GenBank no. LNOS00000000) using Bowtie2-2.2.3 (Langmead and Salzberg, 2012). The level of gene expression was estimated by FPKM (Fragments Per Kilo bases per Million fragment). Differential expression analysis of the two groups (WT as the control and △welA as the experiment group) was performed using the DESeq (version 1.18.0) (Anders and Huber, 2010). Genes with a minimal twofold difference in expression (| log2 fold change (FC)| ≥ 1) and p-value < 0.05 were identified as differentially expressed genes (DEGs). Thus, the number of up-regulated and down-regulated genes was calculated based on the log₂FC value.

DEGs with similar expression patterns were clustered by the Pheatmap R package using the Complete Linkage method. Furthermore, functional enrichment analysis was performed by Gene Ontology (GO) and KEGG enrichment analysis. For GO enrichment analysis, each DEG was mapped to GO terms in the Gene Ontology database⁶ by topGO (Alexa et al., 2006). The DEGs were also mapped to the KEGG orthology terms in the KEGG pathway database⁷ using the KOBAS software (Bu et al., 2021).

The expressional levels of 10 typical DEGs identified by RNA-Seq were examined using qRT-PCR. Total RNA of WT and △welA cultured for 16 h were reverse-transcribed by a PrimeScript^® RT Reagent Kit with gDNA Eraser (Perfect Real Time) (TaKaRa, Dalian, China) according to the manufacturer’s instructions. qRT-PCR was performed using the primers listed in Supplementary Table 2 with a TB Green^®Premix Ex Taq™ II (Tli RNaseH Plus) kit (TaKaRa, Dalian, China) on ABI 7500 instrument (Applied Biosystems). The qRT-PCR reaction system and conditions were the same as described in our previous work (Li et al., 2018). The relative expression of each gene was calculated according to the 2^{–ΔΔ Ct} formula, and the 16S rRNA gene served as an internal standard to standardize results. The expression of each gene was repeatedly detected at least three times. The data are expressed as arithmetic means ± the standard deviation.

Firstly, WT, △welA, and △welA/pBBR1MCS-welA were cultured in fermentation media for about 12 h and harvested by centrifugation, respectively. Secondly, nucleotide extraction was carried out as previously reported (Newell et al., 2011) and was resuspended in 500 μL water under vigorous vortexing. Finally, the concentration of c-di-GMP was analyzed by reversed phase-coupled HPLC-MS/MS on a Waters Xevo TQ-S Triple Quad Mass Spectrometer using the method reported by Spangler et al. (2010).

Heterologous expression of possible diguanylate phosphodiesterase encoded by Gene 1094 in E. coli BL21 (DE3) and its activity assay.

Firstly, the recombinant expression plasmid was constructed. the ORF of Gene 1094 was amplified by PCR using the Sphingomonas sp. WG genomic DNA as the template and gene1094exFor/gene1094exRev containing BamHI and HindIII recognition sties as primers (Supplementary Table 1). Sequentially, PCR product was gel-purified, digested by BamHI and HindIII and ligated into the linearized pET32a⁽⁺⁾. The recombinant plasmid pET32a-1094 was confirmed by PCR using gene1094exFor/gene1094exRev as primers and DNA sequencing. Secondly, the recombinant plasmid was transformed into the E. coli BL21 (DE3) host to obtain the expression strain. Thirdly, the protein expression was induced by 0.8 mM isopropyl-D-thiogalactopyranoside (IPTG) at 28°C for 16 h in accordance with the pET system manual (Novagen, Germany). The cells untreated with IPTG was used as the control group. Then the crude enzyme samples were obtained by ultrasonic disruption of the cells, centrifugation and ultrafiltration. The samples were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) to confirm the expression of the target protein.

Finally, the hydrolysis reaction of c-di-GMP was carried out in the mixture as follows: 50 mM Tris-HCl buffer (pH 9.35) containing 50 mM NaCl, 5 mM MgCl₂ and 0.5 mM EDTA, 100 μM c-di-GMP as the substrate and the crude enzyme. The reaction was performed at 37°C for 2 h and was terminated by heating the samples at 100°C for 3 min. The precipitated proteins were removed by centrifugation at 16,000 × g and the obtained supernatant was filtered using a 0.22 μm membrane and analyzed by an HPLC instrument (Agilent 1260 Infinity II) equipped with a ZORBAX SB-C18 column (15 × 4.6 mm) and detected at 254 nm. The mobile phase A was 100 mM KH₂PO₄ containing 4 mM tetrabutylammonium hydrogen sulfate, and mobile phase B contained 75% phase A and 25% methanol. The phase B content in mobile phase was gradient from 0 to 30% from 2.5 to 5 min; 30 to 60% from 5 to 10 min; 60 to 100% from 10 to 14 min and at 100% for 7 min, then from 100 to 50% from 21 to 22 min and finally to 0 at 23 min. c-di-GMP was used as the standard. The enzyme activity (1 U) was defined as the amounts of enzyme required to hydrolyze 1 μM c-di-GMP per minute in 1 ml reaction mixture.

One putative hybrid sensor histidine kinase (HK)/response regulator (RR) encoding gene welA was annotated in the genome of Sphingomonas sp. WG (Li et al., 2016a). The length of the open reading frame of welA was 2,385 bp, and therefore, the deduced protein WelA contained 794 amino acids. According to the predicted results of ProtParam, the calculated molecular mass and pI of WelA were 84.5 kDa and 5.14, respectively. The secondary structure prediction results by SOMPA showed the percentages of α-helix, extended strand, β turn, and random coil predicted were about 42.57, 18.14, 6.30, and 33.00. Structural domain analysis by SMART showed that this protein contains two PAS domains (residue 76–145 and residue 293–358), a HisKA domain (residue 418–484), a HATPase_c domain (residue 525–653), and a REC domain (residue 676–787), suggesting that this protein had both HK and RR domains and might be a hybrid sensor HK/RR. Besides, it also had two transmembrane regions (residue 15–34 and 41–63), indicating that it might be associated with the cell membrane.

Subsequently, proteins homologous to WelA were retrieved from the Swissprot database by the BLASTp program. A phylogenetic tree was constructed by ClustalX multiple sequence alignment program version 2.0 (Larkin et al., 2007) and MEGA 7 program (Kumar et al., 2016; Figure 1A). WelA from Sphingomonas sp. WG showed 50.12% identity with the hybrid sensor HK CckA (Genbank no. P0DOA0.1) (Supplementary Figure 1) which initiated the conserved CckA–ChpT–CtrA–CpdR regulatory system controlling cell division, replication, and virulence of Brucella abortus (Willett et al., 2015). It also showed similarity to other sensor kinases such as DivL and FixL in varying degrees.

To investigate if WelA is involved in the regulation of WL gum biosynthesis, welA was deleted from the genome of Sphingomonas sp. WG by double homologous recombination using the method reported in our previous work (Li et al., 2021). The mutant strain ΔwelA was verified by PCR with primers welAinFor and welAinRev (Figure 1B), and the specific product about 530 bp was lost. Besides, the welA coding sequence was introduced into ΔwelA to obtain the complemented strain ΔwelA/pBBR1MCS-3-welA, which was also elucidated by amplifying the partial welA coding sequence using welAinFor and welAinRev as primers (Figure 1C).

Subsequently, the fermentation of the WT, ΔwelA, and ΔwelA/pBBR1MCS-3-welA was performed (Figure 2). As the fermentation proceeded, the WT strain and the ΔwelA strain entered the log phase or exponential phase very quickly (within 6 h), and their biomass increased significantly. Compared with WT strain, the deletion of welA did not reduce the cell growth rate, in contrast, it led to the enhanced growth rate. The μ_max for WT strain and ΔwelA in the log phase was about 0.17 and 0.25 h^–1, respectively. ΔwelA/pBBR1MCS-3-welA showed a different growth curve in which it had a long lag phase for about 21 h and then entered the log phase with a μ about 0.12 h^–1. The tD time for WT, ΔwelA, and ΔwelA/pBBR1MCS-3-welA was 4.26, 3.23, and 5.85 h, respectively. The slow specific growth rate of the complementary strain might be related to the replication burden of the plasmid pBBR1MCS-3-welA or the dose-dependent effect of the welA overexpression. Therefore, the growth curve of the other strain WT/pBBR1MCS-3 was determined. The empty vector had a very slight impact on the growth of the bacteria. Therefore, the slow growth rate of the complemented strain might be caused by a dose-dependent effect of the welA overexpression. Due to the influence of the metabolites such as organic acids, the pH of the fermentation broth changed. The pH value of WT was decreased and was the lowest among the three strains. The pH of the ΔwelA/pBBR1MCS-3-welA was the highest, which might be related to its lowest biomass. Furthermore, the viscosity of the fermentation broth was detected as the accumulation of the EPS will enhance the viscosity. The viscosity of ΔwelA fermentation broth was only 0.6% of WT, while the broth viscosity of the complemented strain could restore to 40% and 44% when cultured for 72 and 84 h, respectively. Similar to viscosity, welA regulated the WL gum biosynthesis, and the deletion of welA resulted in a remarkable decrease in WL gum production. The WL gum yield of ΔwelA was only 44% and 49% of WT when cultured for 72 and 84 h. The gene complementation could restore the WL gum production to 73% and 74% of WT at 72 and 84 h. All these phenomena suggested that WelA plays important regulatory role in the WL gum biosynthesis process.

Transcriptome analysis of WT (the control group) and ΔwelA (the experiment group) was conducted to understand the genes differentially regulated by WelA. An average of 32,410,509 and 36,273,418 raw sequencing reads in the control group and experimental group was generated by Illumina sequencing, respectively. After the trimming process, about 93% and 90% of the libraries remained clean reads, respectively (Supplementary Table 3). The correlation of gene expression level between samples was checked to test the sequencing reliability of the experiment and the rationality of sample selection, and the results were shown in Supplementary Figure 2. The square of Pearson correlation coefficient (R²) between biological repeats was higher than 0.92 and suggested that the difference between individuals was very small and the sequence results were credible. Moreover, the six samples could be clustered into two groups, the WT and ΔwelA group, which indicated that the deletion of welA resulted in a significant change of gene expression (Supplementary Figure 2). Therefore, differential expression analysis was performed. A total of 386 DEGs with 187 up-regulated genes and 199 down-regulated genes were detected (Supplementary Figure 3). To validate the accuracy of RNA-seq results, the expression of 10 typical DEGs that involved in signaling transduction pathway, sugar metabolism and WL gum biosynthesis, and c-di-GMP biosynthesis or hydrolysis was further confirmed by qRT-PCR. The results were shown in Figure 3. According to correlation analysis, the expression results of qRT-PCR were strongly positively consistent with those from RNA-Seq (R² = 0.7642), which suggested that the data from the transcriptome were reliable.

GO enrichment analysis was performed to gain the information of DEGs functional categories. The results (Figure 4) showed that the “bacterial-type flagellum (GO:0009288)” and “cell projection (GO:0042995)” were the top GO terms in the “cellular component” group; The “methyltransferase activity (GO:0008168)” and “transferase activity, transferring one-carbon groups (GO:0016741)” were significantly enriched in “molecular function” group. The “secretion (GO:0046903),” “secretion by cell (GO:0032940),” “protein secretion (GO:0009306),” and “peptide secretion (GO:0002790)” were significantly more clustered in the “biological process category.” Besides, the DEGs were also mapped to different KEGG categories, including cellular processes, environmental information processing, genetic information processing, human diseases, and metabolism. The top 20 enriched pathways were also shown in Figure 4. The most DEGs were enriched in “Flagellar assembly (ko02040),” “Porphyrin and chlorophyll metabolism (ko00860),” “Valine, leucine, and isoleucine biosynthesis (ko 00290),” “C5-Branched dibasic acid metabolism (ko00660).” Since we focused on the role of WelA on the biosynthesis of WL gum, the regulation pathway of WelA was mainly analyzed in the following step.

WelA was mapped to the two-component signal transduction systems (ko02020) due to its identity to CckA, and its deletion resulted in the up-regulated expression of chpT (Gene 3285) and ctrA (Gene 1411) (Table 1). It has been reported that CtrA is an active transcriptional regulator controlling genes involved in cell cycle progression and flagellar motility in Caulobacter crescentus (Laub et al., 2000). Besides CtrA, CpdR (Gene 1015) is another RR that can also be phosphorylated by ChpT. Normally, CpdR inhibits CtrA, and this inhibition effect could be relieved by phosphorylation. However, the expression level of CpdR was not significantly changed after welA deletion. In general, genes controlled by CtrA are enriched in certain functional categories, i.e., cell motility, signal transduction, and cell wall/membrane envelope biogenesis (Brilli et al., 2010). In C. crescentus, the architecture of the CtrA promoters has been analyzed, and two classes of CtrA binding motifs: a full site represented by the sequence TTAA (N₇) TTAA, and a half site (TTAA) were revealed (Zhou et al., 2015). It has been reported that the expression of the fla2 genes involved in the biosynthesis of polar flagella of Rhodobacter sphaeroides is dependent on the two-component system, including CckA-ChpT-CtrA (Vega-Baray et al., 2015). Based on the motif sequence scanning, CtrA might regulate the expression of many genes. The expression of two special genes involved in c-di-GMP hydrolysis and flagellar assembly was up-regulated by CtrA. The first gene was Gene1094 encoding the diguanylate phosphodiesterase that might be related to hydrolysis of c-di-GMP. The heterologous expression of Gene 1094 in E. coli BL21 (DE3) showed that the crude enzyme had weak activity to catalyze the hydrolysis of c-di-GMP (Supplementary Figure 4). Its highly expression level indicated that c-di-GMP level might decreased and was involved in the regulation of WL gum biosynthesis. Therefore, the intracellular c-di-GMP was extracted and determined. As expected, the intracellular c-di-GMP concentration was significantly decreased in ΔwelA (Supplementary Figure 5). Generally, reduced c-di-GMP level might increase flagellar motility and decrease the biofilm formation (Floyd et al., 2020). The second one is FlgN (Gene 283), which is the flagellar synthesis and assembly chaperone, and therefore, the motility might be changed.

Besides the CckA-ChpT-CtrA pathway, the other three genes were mapped to the two-component signal transduction systems (ko02020), including the up-regulated DctP (Gene1533) in the C4-Dicarboxyate transport, the down-regulated CheR (Gene 3280) in flageller assembly, and Cyt C (Gene 371) in electron transfer system in Oxidative phosphorylation process.

Many genes were involved in the biosynthesis of WL gum. However, most genes related to the biosynthesis of nucleotide-sugar precursors such as pgm, ugp, ugd, and rlmA to rmlD, the tetrasaccharide repeating unit assembly such as welB, welK, welL, and welQ, polymerization of the assembled repeating units such as welS, welG, welE, welC, etc., were not identified as DEGs (Table 1). The expressional level of gene encoding PMM (Gene 592) catalyzing the synthesis of α-D-mannose 1-phosphate, the precursor of GDP-mannose, was down-regulated (the log2FC was -1.19). Besides, two genes (Gene 3600 and 3601) encoding AtrD and AtrB in the wel cluster were identified as up-regulated DEGs.

According to the GO enrichment results (Figure 4), the DEGs under the category of biological process related to processes of flagellum-dependent cell motility (GO:0001539) and under the categories of the cellular component involved in bacterial-type flagellum (GO:0009288) were highly expressed. KEGG pathway analysis revealed that 14 DEGs were involved in flagellar biosynthesis, including 12 highly and 2 lowly expressed genes. Taken together, WelA might play a regulatory role in flagellar biosynthesis of Sphingomonas sp. WG. Therefore, the organization of the genes for flagellar biosynthesis was analyzed. Most of the genes were clustered together in the smaller cluster I and the larger cluster II within the chromosome, as shown in Figure 5 and Supplementary Table 4. Most of the highly expressed genes in ΔwelA encode essential components for flagellar biosynthesis, including flagellin, flagellar basal body protein, flagellar hook-associated protein, and export apparatus for flagellar components. Furthermore, three genes that encode transcriptional regulatory proteins FlbD, FlaEY, and FlaF were also highly expressed. However, some genes in the chemosensory system that also influence motility, such as CheR, McpA, and twitching motility protein PilT, were down-regulated.

Therefore, it is inferred that the flagellar formation and the motility of the mutant strain might change. Therefore, the swimming and swarming motilities of WT and welA mutant strain were measured. ΔwelA showed higher swimming motility than WT on SIM medium (Figure 5C); however, when cultured on solid fermentation medium, its swimming motility was much lower (Figure 5D). Its swarming motility independent of the presence of a flagellum was lost on both media (Figure 5E).