The bacterial strains used in this study are listed in Supplementary Table 1. All bacterial strains were stored at −80°C, and the frozen stocks were streaked onto fresh agar plates before experiments were performed. Escherichia coli WM3064 were grown in LB supplemented with 0.3 mM DAP (2,6-diamino-pimelic acid) at 37°C. Pseudoalteromonas lipolytica strains were grown in SWLB (seawater LB, 1% tryptone and 0.5% yeast extract dissolved in seawater) or 2216E at 25°C. Kanamycin (50 μg/mL) and erythromycin (25 μg/mL) were used to maintain the gene knockout vector pK18mobsacB-ery in E. coli and P. lipolytica, respectively. Chloramphenicol (30 μg/mL) was used to maintain the expression vector pBBR1MCS-cm in both E. coli and P. lipolytica.

The isolation of colony morphology variants from biofilms was conducted as we previously described (Zeng et al., 2015). Briefly, pellicle biofilms of P. lipolytica wild-type and EPS+ strains were grown in SW-LB medium in test tubes without shaking for the indicated times at 25°C. The harvested biofilm cells were then diluted in seawater using 10-fold serial dilutions and subsequently plated on fresh SWLB agar plates to obtain 30–300 colonies per plate. The wrinkled or smooth colonies were examined after 3 days of incubation at 25°C. The randomly selected variant strains were inoculated into fresh liquid medium or onto agar plates for three rounds of overnight passaging to ensure the stable variation. The AT00_08765 gene was sequenced in 24 wrinkled variants using the sequencing primers listed in Supplementary Table 2.

A swimming motility assay was conducted on semisolid agar plates. Overnight cultures of the indicated strains were inoculated onto 2216E plates containing 0.25% agar (w/v) and incubated for the indicated times at room temperature. The diameter of the swimming motility was also measured after incubation. The assays were performed with three independent colonies for each strain.

Congo red (CR) binding assay was conducted on agar plates or liquid cultures containing different concentrations of CR. Bacterial strains producing cellulose lead to the pink or red phenotype on Congo red agar plates. Briefly, fresh colonies were streaked onto 2216E agar plates containing 100 μg/mL CR and incubated at 25°C for 3 days. An increasing depth of color indicated high cellulose matrix production levels. To quantify cellulose production, 20 μg/mL CR was added to 1 ml of overnight culture and continued to culture with shaking condition for 3 h. The cultures were then centrifuged at 15,000 rpm for 5 min. The cell pellet was examined and the supernatant was collected. The amount of CR bound to the cells was determined by measuring the absorbance of the supernatant at 490 nm as previously described (Sajadi et al., 2019). Each sample was prepared and assayed in triplicate.

Pseudoalteromonas lipolyticaΔflhA mutant and the EPS+ strains were grown on 2216E agar medium at 25°C. The fresh grown colonies were inoculated in 2216E medium with an adjustment OD₆₀₀ between 0.5 and 1.0. The bacterial suspension was transferred to a formvar-coated copper mesh membrane for 2 min and then covered with 30 g/l phosphotungstic acid at pH 7.0 for another 2 min. After air drying the membrane, the cells were observed and imaged using a Hitachi H-7650 microscope.

Gene deletion mutants were constructed using our previously published method (Wang et al., 2015). Primers used in this study are listed in Supplementary Table 2. Briefly, the upstream and downstream regions of the target gene open reading frame were PCR-amplified from wild-type genomic DNA. The two fragments were then ligated into the suicide plasmid pk18mobsacB-ery to create the deletion vector in a host of E. coli WM3064, and then integrated into the chromosome of the P. lipolytica strain by conjugation. The P. lipolytica deletion mutants were generated by screening for erythromycin resistance followed by sacB-based counterselection. The vector pBBR1MCS-Cm was used for gene expression in P. lipolytica. The target gene with its native promoter was PCR amplified and ligated into the vector pBBR1MCS-Cm in a host of E. coli WM3064. The plasmid was transferred to P. lipolytica by conjugation as described above. The resulting deletion mutants and expression plasmids were further confirmed by sequencing using the primer sets pK18-f/pK18-r and pBBR1MCS-f/pBBR1MCS-r, respectively, listed in Supplementary Table 2.

The genomes of four biofilm variant strains (V3, V4, SV1 and SV2) were sequenced using the whole-genome shotgun method by GENEWIZ Co., Ltd. (Suzhou, Jiangsu, China). The raw sequencing data were processed after filtering, and the statistics of the sequencing data are listed in Supplementary Table 3. In total, the average clean sequencing data for each variant strain was greater than 3000 Mb, and the average genome sequencing depth exceeded 500. SNPs and InDels were detected based on the aligned result of the assembly sequence and the P. lipolytica reference. The mutation results revealed by whole-genome re-sequencing are listed in Tables 2, 3. Mutations in flhA, dgcB wspC and bcsC in the corresponding variant strains were also further confirmed by PCR-amplification and sequenced using the primer sets listed in Supplementary Table 2. The whole-genome sequencing project has been submitted to the SRA database under the accession numbers PRJNA756106, PRJNA756166, PRJNA755839 and PRJNA755904 for V3, V4, SV1, and SV2, respectively.

Previously, we found that Pseudoalteromonas lipolytica produced a number of wrinkled colony morphology variants during the development of biofilms (Zeng et al., 2015). In this study, we further characterized 24 wrinkled variant strains that were randomly isolated from P. lipolytica biofilms. As shown in Figure 1A, all 24 isolated variant strains, designated V1 to V24, exhibited wrinkled colony morphology, which shared a similar phenotype to the EPS+ wrinkled variant, as we have previously reported (Liu et al., 2018). Swimming motility assays showed that all of the variant strains dramatically decreased swimming motility by approximately 2- to 10-fold compared to that of the wild-type strain after 20 h of incubation (Figure 1B). Among them, the V3 strain was completely defective in swimming motility on semisolid agar plates, whereas V4, V5, V8, V20, and V21 exhibited approximately twofold reduced swimming motility compared to that of the wild-type strain. In addition, the wrinkled variant strains showed a darker red color on Congo red plates than the wild-type strain, suggesting that the variant strains produced more cellulose polysaccharide than the wild-type strain (Figure 1C and Supplementary Figure 1). As shown in Figure 1C, the cellulose production of the variant strains increased approximately twofold to threefold compared to that of the wild-type strain. Thus, the self-generated wrinkled colony morphology variant strains in P. lipolytica biofilms showed reduced swimming motility and increased cellulose production.

We previously identified that the colony morphology of the wrinkled variant was associated with mutation in the AT00_08765 gene (Zeng et al., 2015). Thus, the total genomic DNA of these twenty-four wrinkled colony morphology variant strains was extracted and used as templates to amplify the AT00_08765 genes for sequencing. The results showed that nineteen variant strains harbored mutations at different positions within the AT00_08765 gene, except for the V3, V4, V5, V20, and V21 variant strains (Table 1). The mutation types include base deletion, base insertion or base substitution, which result in frameshift mutation, nonsense mutation or non-synonymous mutation in the AT00_08765 gene, leading to a defective gene product. To further confirm the physiological function of AT00_08765, the AT00_08765 gene from the wild-type strain was cloned via the broad-host-range vector pBBR1MCS and then expressed in three wrinkled variant strains (V1 to V3). The results showed that expression of the wild-type AT00_08765 significantly restored swimming motility for the variant V1 and V2 strains that harbored mutation in AT00_08765, but not for the V3 variant strain that did not harbor mutation in AT00_08765 (Figure 2A). Moreover, the expression of AT00_08765 also restored cellulose production to the wild-type level for the V1 and V2 strains, whereas cellulose production for the V3 variant strain remained unchanged (Figure 2B). Taken together, we demonstrated that the phenotypes of the majority of wrinkled variants were mainly caused by spontaneous mutations in the AT00_08765 gene.

To explore the molecular basis of the wrinkled variant without harboring a mutation in the AT00_08765 gene, we performed a whole-genome re-sequencing approach based on the genome sequence of wild-type P. lipolytica to identify the genomic change in the V3 and V4 wrinkled variants, which exhibited relatively lower and higher activity in swimming motility, respectively, compared to other wrinkled variants (Figure 1). Statistically, both of the genomes of the V3 and V4 strains showed 99.83% coverage to the reference wild-type genome, and the sequencing depths both exceeded 500× (Supplementary Table 3). By aligning the sequences to that of the reference wild-type genome, sixteen sites of mutations (designated M1 to M16) were found in these two variants. Of these sixteen mutations, fourteen were point mutations, including five synonymous mutations and eight non-synonymous mutations located in the coding region, and one mutation located in the intergenic region. Besides, two insertion mutations were also found located in the coding region (Table 2). Given that the V3 and V4 variants exhibited different swimming motility activities, the mutated genes that determine the alteration of swimming motility should be different from each other. Thus, three mutations of M4, M15, and M16 that were only found in the V3 variant strain and one mutation of M6 that was only found in the V4 variant strain might be critical in determining the corresponding phenotype for the V3 and V4 variants. Among them, M15 was located at AT00_RS08430, which encodes a flagellar biosynthesis protein FlhA that is required for formation of the rod structure of the flagellar type III protein export apparatus (Terahara et al., 2018). To verify that flhA mutation in P. lipolytica causes defects in swimming motility, the intact flhA gene from the wild-type strain and the variant flhA gene from the V3 variant strain were each cloned via the broad-host-range vector pBBR1MCS, and then expressed in the V3 variant strain, respectively. The results showed that the expression of wild-type flhA to the V3 variant successfully restored swimming motility, whereas the expression of the variant flhA failed to do so (Figure 3A). To further investigate the physiological function of FlhA, an in-frame deletion of the flhA gene was constructed in P. lipolytica (Figure 3B). As expected, the deletion mutant strain ΔflhA completely lost swimming motility as the V3 variant strain, and complementation of flhA to the ΔflhA mutant strain also restored swimming motility (Figure 3C). To verify FlhA is required for the flagellar assembly, we examined the cell morphology of the ΔflhA mutant strain using transmission electron microscopy. As expected, flhA mutation causes defects in flagellar assembly (Figure 3D). In contrast, P. lipolytica wild-type strain harbor a single polar flagellar (Zeng et al., 2015). Moreover, the ΔflhA mutant strain displayed wrinkled colony morphology on SW-LB agar medium, and complementation of flhA to the ΔflhA mutant strain also restored smooth colony morphology, which was similar to that of the wild-type strain (Figure 3E). Importantly, the ΔflhA mutant strain also increased cellulose production approximately twofold, which is similar to that noted for the V3 variant strain (Figure 3F), whereas the expression of flhA in the ΔflhA mutant strain reduced cellulose production to the wild-type level (Figure 3G). Taken together, these results demonstrated that the wrinkled variant of P. lipolytica harboring a mutation in flhA gene cause defects in swimming motility and an overproduction of cellulose.

Based on the whole-genome re-sequencing results, the M6 mutation might cause phenotypic changes in the V4 variant strain. M6 was located at AT00_RS12195, which encodes a diguanylate cyclase DgcB that mediates the synthesis of a c-di-GMP molecule. To investigate the physiological function of the variant dgcB, the dgcB gene from the wild-type strain and from the V4 variant strain were each cloned and expressed in the wild-type strain. The results showed that overproduction of wild-type DgcB dramatically decreased swimming motility while increasing cellulose production; however, overproduction of the variant DgcB from the V4 variant failed to do so (Figures 4A,B). To further test the physiological function of wild-type DgcB, an in-frame deletion of the dgcB gene was constructed in P. lipolytica (Figure 4C). However, the results showed that the colony morphology, swimming motility and cellulose production of the ΔdgcB mutant strain displayed similar characteristics to those of the wild-type strain (Figure 4D). More than forty genes encode a diguanylate cyclase in the genome of P. lipolytica, and twelve of them were annotated as dgcB (data not shown). This result indicates that deletion of one dgcB gene (AT00_RS12195) in wild-type P. lipolytica might have a minor impact on physiological performance. We next sequenced the AT00_RS12195 gene in other biofilm wrinkled variants that have a higher swimming motility as shown Figure 1B. The results showed that biofilm variants V20 and V21 also harbor the same point mutation in AT00_RS12195 (data not shown). Thus, the increasing swimming motility of these wrinkled variants might be correlated with the AT00_RS12195 mutation. Therefore, we deleted AT00_RS12195 in the P. lipolytica EPS+ host, a wrinkled variant, to investigate the corresponding phenotype. Although deletion of AT00_RS12195 in the EPS+ host induced minimal effects on the colony morphology and cellulose production level, the EPS+ ΔAT00_RS12195 strain displayed higher swimming motility than the EPS+ variant strain after 20 h of incubation on swimming agar medium (Figure 4E). Taken together, these results showed that the AT00_RS12195 mutation increased swimming motility in wrinkled variants of P. lipolytica, and overproduction of AT00_RS12195 led to cellulose overproduction and decreased swimming motility.

To further identify the genetic factor associated with the wrinkled colony morphology, we screened suppressor variant strains that switched the phenotype from wrinkled to smooth within a 5-day-old biofilm of P. lipolytica EPS+, a wrinkled variant strain with an AT00_08765 gene mutation. As shown in Figure 5A, two suppressor variant strains with stable smooth colony morphology, designated SV1 and SV2, were isolated from the ancestral strain EPS+. We next performed a whole-genome re-sequencing approach based on the genome sequence of the P. lipolytica EPS+ strain to identify the genomic change that occurred in these two variants. Statistically, both of the genomes of SV1 and SV2 showed 99.83% coverage compared to the genome of the EPS+ strain, accompanied by a sequencing depth greater than 500× (Supplementary Table 3). By aligning the genome sequences to that of the reference genome, eight sites of mutations were found in these two variants, in which the point mutation in AT00_08765 originally existed in the EPS+ strain was also listed in Table 3. Notably, the genome of SV1 harbors a non-synonymous mutation in AT00_09010 (or AT00_RS08640), which encodes a methyltransferase CheR that mediates chemosensory signaling in conjunction with the methylesterase CheB in the chemosensory pathway. The point mutation in cheR of SV1 was also confirmed by sequencing the amplified PCR product using the primers listed in Supplementary Table 2. Furthermore, we deleted the cheR gene in the EPS+ host to verify the physiological functions of the cheR gene product (Figure 5B). As expected, a null cheR and AT00_08765 mutant in which the chemosensory system was ineffective resulted in the smooth colony morphology (Figure 5A). For the SV2 variant strain, a single nucleotide deletion mutation in bcsC (AT00_RS02190) was located within the bcs (bacterial cellulose synthesis) operon, ranging from AT00_RS02190 to AT00_RS02225 as we have previously reported (Zeng et al., 2015). Similarly, the deletion mutation in bcsC of the SV2 variant strain was further confirmed by sequencing the PCR-amplification product. Intact bcsC is required for the expression of cellulose as reported in many microorganisms (Zogaj et al., 2001; Solano et al., 2002). Thus, it is hypothesized that bcsC mutation in the EPS+ strain results in the loss of cellulose production, leading to smooth colony morphology. Importantly, we previously demonstrated that deletion of the bcs gene within the cellulose operon in the P. lipolytica ΔAT00_08765 host results in a morphological switch from wrinkled to smooth, suggesting that the cellulose operon and its product are required for the generation of wrinkled colony morphology (Zeng et al., 2015). Moreover, deletion of cheR or bcs genes in EPS+ host significantly reduced the cellulose production, and formed relatively less or rarely pellicle (Figure 5A). Since EPS+ derived smooth variant exhibited wild-type smooth colony morphology, we originally hypothesized that the smooth variants may recover the wild-type motility activity. However, results showed loss of cheR or bcs in EPS+ host only slightly increased swimming motility compared to that of the EPS+ strain (Figure 5C), suggesting AT00_08765 and cheR involved in motility are independently from their role in controlling cellulose production. Moreover, the EPS+ strain harbors an intact single polar flagellar (Figure 5D), indicating that the reduced swimming motility of the EPS+ strain compared to that of the wild-type strain is independently from the biosynthesis of flagellar. Next, we deleted cheR in ΔflhA host to check whether cheR involved in the flhA-mediated wrinkled colony morphology. As we observed that ΔflhAΔcheR double mutant strain developed a relative reduced wrinkled colony phenotype compared to that of the ΔflhA strain (Figure 5E), suggesting the AT00_08765-CheR signaling pathway may converge with the flhA dependent pathway in terms of the emergence of rugosity. Taken together, we demonstrated that the chemosensory pathway is crucial for the determination of wrinkled colony morphology by controlling the production of cellulose in P. lipolytica.

To gain a better insight into the evolution of genetic variants in P. lipolytica biofilms, we summarized the emergence of different types of genetic variants during the biofilm growth of P. lipolytica, each of which exhibited a specific phenotype (Figure 6). As noted thus far, the biofilm growth of P. lipolytica produced various phenotypic variants by introducing genome mutations in specific targets, such as wspF-like or flhA in wrinkled variants (EPS+) with higher cellulose production, bcs in smooth variants (EPS−) with lower cellulose production, in translucent variants (CPS−) with lower capsular polysaccharide production (Zeng et al., 2019) and hmgA in pigment production variants with higher pyomelanin production (Zeng et al., 2017). Although all of these different genetic variants can be produced during biofilm formation, it is noted that the emergence frequency of those different variants with specific phenotype mainly depends on the culture conditions (i.e., colony biofilm or pellicle biofilm, 2216E medium or HSLB medium, 25°C or 30°C). In general, the EPS+ and CPS− variants were generated at a relatively high frequency in P. lipolytica biofilm under laboratory culture condition. In all, the study of the molecular basis of these biofilm variants provides insights into the evolution within biofilms of P. lipolytica.