The biofouled RO module was harvested from a full-scale desalination plant located on the Red Sea coast in the Kingdom of Saudi Arabia (22.299815 N, 39.116812 E). The water quality parameters and the operational conditions of the plant have been discussed elsewhere (Belila et al., 2016). To isolate the bacteria, the biomass was harvested from the biofouled RO module using a sterile spatula and suspended it in sterile seawater. Serial dilutions of the suspended biomass were made and then 100 μl was plated on Zobell Marine Agar (HiMedia, India, 55.25 g powder/L of Milli-Q) plates, followed by incubation at 30°C for 72 h. The Zobell Marine medium mimics seawater composition and allows bacteria in the marine environment to grow abundantly. Three phenotypically different colonies were selected to be streaked on fresh agar plates (Supplementary Figure 1). This procedure was repeated three times to obtain pure cultures of the selected bacteria. The isolates were then grown in Marine Broth (HiMedia, India), and 500 μl of bacterial culture was mixed with 500 μl of 50% glycerol solution and stored at −80°C for future use.

Pure cultures of bacteria were streaked on Zobell Marine Agar plates and incubated at 30°C overnight. A single colony from these plates was inoculated in Marine Broth and incubated overnight at 30°C with shaking at 120 rpm. These bacterial cultures were diluted to OD₆₀₀ of 0.01 in fresh Marine Broth and incubated at 30°C with shaking at 120 rpm. The OD₆₀₀ of these bacterial samples was measured using a Spectronic 200 (Thermo Fisher Scientific, United States) every hour until the cells reached the stationary phase.

We performed a biofilm formation assay for the isolates in 96-well tissue culture-treated polystyrene plates (Greiner Bio-One, Germany), as described previously (Barbara and Coffey, 2014). Briefly, the overnight grown cells were diluted to an OD₆₀₀ of 0.05, and 100 μl of the diluted cultures was added to the wells of polystyrene plates under sterile conditions. The plates were incubated for 24 or 48 h and culture media were discarded at the end of each incubation. The plates were rinsed in water and stained with 0.1% of crystal violet (CV) for 10 min. The unbound CV was discarded, and plates were rinsed twice in freshwater and dried overnight. The bound CV was dissolved in 30% acetic acid and the absorption was measured in a microplate reader at OD₅₉₀ using SpectraMax M3 (Molecular Devices, United States).

A single colony of bacterial isolates was inoculated in Marine Broth using a sterile inoculation loop. The bacterial colony was mixed with vortexing and a volume of 150 μl of this bacterial suspension was spotted on fresh Marine Agar plates and spreaded on an agar surface using a sterile L-shaped spreader. These plates were incubated upside down at 30°C for 72 h. The agar surface was used for EPS extraction as this method mimics biofilm growth and yields the maximum amount of EPS (Rehman et al., 2013). The biomass was scraped from these plates using a sterile spatula and suspended in 1x PBS. The EPS was extracted using formaldehyde and NaOH, as described previously (Liu and Fang, 2002). Briefly, 60 μl of formaldehyde (36.5%) (Thermo Fisher Scientific, United States) was added to 10 ml of EPS solution and incubated at 4°C for 1 h. To this mixture, 4 ml of NaOH (1N) was added and incubated at 4°C for another 3 h. The mixture was centrifuged at 16,000 × g at 4°C for 20 min followed by filtration of supernatant through 0.2 μm syringe filter. The filtrate was dialyzed using 3.5 KDa dialysis membranes (Spectra/Por^®, United States) at 4°C for 24 h. Finally, the solution was lyophilized at −50°C for 48 hrs.

The carbohydrate content of EPS was measured using a modified phenol-sulfuric acid method and glucose standards (Dubois et al., 1951). Briefly, the lyophilized EPS was suspended in Milli-Q water at a concentration of 1 mg/ml. We mixed 1 ml of this suspension (and glucose standards separately) with 0.5 ml of phenol, followed by the addition of 2.5 ml of concentrated sulfuric acid. The solution was mixed and allowed it to cool at room temperature for 1 h. The absorbance was measured at OD₄₈₅ using a spectrophotometer (Spectronic 200, Thermo Fisher Scientific, United States).

The protein and DNA content of EPS was measured using a Qubit^TM Protein Assay kit and Qubit^TM dsDNA BR Assay kit, respectively (Thermo Fisher Scientific, United States). The proteins and polysaccharides were quantified in 1 mg of freeze-dried EPS (total). The amount of protein and polysaccharide reported here represents per mg dry-weight of total EPS.

We determined the molecular weight of EPS by high-performance gel permeation chromatography (HP-GPC; Agilent Technologies 1260 Infinity) equipped with columns PL-aquagel-OH 60 and PL-aquagel-OH 30 connected in series (length 300 mm and particle size 15 and 8 μm, respectively) (Agilent, United States). The HP-GPC was equipped with a differential refractometer and a UV detector, which was set at 280 nm. For the elution, 0.1 M of NaNO₃ and NaN₃ (0.02% w/v) (Honeywell Fluka^TM, United States) were used. The eluent was degassed and filtered through 0.1 μm pore size membrane filters (MF-Millipore^TM, United States). The EPS samples were dissolved in eluent at 1 mg/ml concentration and filtered through 0.2 μm filters. A 100 μl of EPS solution was injected into the HP-GPC, flow rate was adjusted to 0.5 ml/min, and the temperature of the columns was maintained at 30°C. The polyethylene glycol standards were used to calibrate the system (EasiVial PEG/PEO, pre-weighted calibration kit, 2 ml, Agilent, United States). The HP-GPC analysis was performed in duplicate.

We analyzed the lyophilized EPS samples by FTIR spectroscopy in attenuated total reflection (ATR) mode using a Nicolet^TM iS10 FTIR spectrometer (Thermo Fisher Scientific, United States) equipped with DTGS detector and SMART iTR^TM sampling accessory (Thermo Fisher Scientific, United States). The spectrum was recorded in transmittance mode over a range of 3,900 to 600 cm^–1 at a resolution of 4 cm^–1.

We performed NMR spectrometry (both ¹H and ¹³C) to determine the chemical composition of the EPS produced by the RO membrane isolates, as described recently (Khan et al., 2013; Rehman et al., 2017). Briefly, for ¹H NMR, 2–3 mg of lyophilized EPS was dissolved in 650 μl of deuterated water D₂O and transferred the solution to 5 mm NMR tubes. The ¹H NMR spectra was acquired using a Bruker 700 MHz AVANCE III spectrometer equipped with a Bruker CPTCI multinuclear CryoProbe (Bruker BioSpin, Germany). The signal-to-noise ratio was improved by acquiring 1K scans with a recycle delay time of 3 sec. In order to suppress the water peak, each spectrum was induced using an excitation sculpting pulse sequence through a standard program (zgesgp) from the Bruker pulse library. Solid-state ¹³C NMR spectra of EPS were obtained using 900 MHz AVANCE III NMR spectrometer (Bruker BioSpin, Germany) using a triple-resonance 3.2 mm Bruker MAS probe. The temperature was maintained at 25°C. Bruker Topspin 3.5p17 software (Bruker BioSpin, Germany) was used for data collection and processing of spectra.

The total area under each spectrum was calculated for complete ¹H NMR spectra by trapezoid numerical integration using RStudio (1.2.5001) and trapz function in R-package (caTools) (Kincaid and Cheney, 1991). The relative abundances of the major ¹H resonances were measured by integrating the area in chemical shift ranges (as previously described by Hertkorn et al. (2013), which were normalized by total chart area.

The genome sequencing of isolates was done as described previously (Rehman et al., 2020b). Briefly, all the strains were grown in 20 ml of marine broth, centrifuged at 6,000 g for 10 min, and the resulting cell pellet was used for DNA extraction using the DNeasy PowerWater kit (Qiagen, Germany). DNA concentration was measured using the Qubit dsDNA HS/BR Assay kit (Thermo Fisher Scientific, United States). The sequencing library was prepared using NEB Next Ultra II DNA library prep kit for Illumina (New England Biolabs, United States) as per manufacturer instructions. Paired-end sequencing (2 × 301 bp) of the samples was done on Miseq (Illumina, United States) using MiSeq Reagent Kit v3, 600 cycles (Illumina, United States).

The sequencing reads were trimmed using Cutadapt v. 1.16 (Martin, 2011), and assembled using Megahit v. 1.1.3 (Li et al., 2016). The genome coverage was calculated by mapping reads back to assembly using minimap2 v. 2.12-r827 (Li, 2018). The genomes were annotated using PROKKA v. 1.14-dev (Seemann, 2014), and completeness was calculated using CheckM (Parks et al., 2015). Identification of carbohydrate-active enzymes (CAZymes) was made online using predicted amino acid sequences¹ (Zhang et al., 2018). The dbCAN2 database was searched for homologous sequences using HMMER with default values (E < 1e-15, coverage > 0.35). Comparative genome analysis of RO isolates was done using BLAST Ring Image Generator (BRIG) (Alikhan et al., 2011). For these analyses, the RO2 genome was used as a reference, while RO1 and RO3 genomes were used as a query.

The blastp was used to identify various proteins involved in the production of EPS.

The genome sequences reported in this article were deposited to DDBJ/ENA/GenBank under BioProject ID PRJNA616073 and accession number JAAZWB000000000 (RO1), JAAXCU000000000 (RO2), JAAXCV000000000 (RO3). The version described in this paper is version JAAZWB010000000, JAAXCU010000000, and JAAXCV010000000.

The basic statistics of the assembled genomes of RO1, RO2 and RO3 are given in Supplementary Table 1 and in Rehman et al. (2020b). Briefly, the sizes of genomes were between 4.1 and 4.2 Mbp, with completeness more than 98%. The BLASTn analysis of 16S rRNA genes of the isolates was conducted against 16S rRNA sequences (bacteria and archaea) database (NCBI, March 29, 2020) (Altschul et al., 1997). These analyses showed that the best match for RO1 (92.3% identity, 98% coverage) and RO2 (99.9% identity, 100% coverage) was Bacillus horikoshii (NR_119070.1), whereas the best match for RO3 (99.4% identity, 97% coverage) was Bacillus aquimaris (NR_025241.1).

Taxonomic analysis using a single gene such as 16S rRNA gene fails to provide sufficient resolution at the species level. Consequently, Average Nucleotide Identity (ANI) is fast becoming a preferred method to classify the newly sequenced genomes. The bacteria belonging to the same species typically showed >=95% ANI values (Jain et al., 2018). We used Genome taxonomy Database Toolkit (GTDB-Tk v1.0.2) to classify each genome by placing it in a reference phylogenetic tree and calculating ANI values against the reference genomes (Chaumeil et al., 2019). The GTDB-Tk analysis showed that both RO1 (91% ANI) and RO2 (92.2% ANI) are closely related to Bacillus sp. CHD6a (GCF_001293645.1), while RO3 (83.1% ANI) is related of B. aquimaris (GCF_000935355.1) (Supplementary Table 2). The ANI values, less than 95%, suggest that the RO isolates may represent novel species belonging to the genus Bacillus (Rashid et al., 2015). Furthermore, it appears that RO1 and RO2 represent two distinct species, as ANI between them was 92.74%, whereas RO1 and RO3 showed 77.95% ANI, and RO2 and RO3 showed 78.1% ANI values.

To identify proteins involved in carbohydrate metabolism a search against the dbCAN database was carried out, which showed that RO1, RO2, and RO3 carry 79, 84, and 76 carbohydrate-active enzymes (CAZymes), respectively (Supplementary Table 3). The detected CAZymes include Glycoside Hydrolases (GH), Carbohydrate Esterases (CE), Glycosyltransferases (GTs), as well as proteins with Carbohydrate-Binding Modules (CBM) (Supplementary Table 3). Most of the identified CAZymes families were shared between the RO isolates except GH35 and GH94, which were found only in RO2 (Supplementary Table 3 and Figure 1). The GH35 family includes B-galactosidases, exo-B-glucosaminidases, exo-B-1,4-galactanases, and B-1,3-galactosidases, whereas GH94 mainly act as phosphorylases² (Lombard et al., 2014). It is interesting to note that RO2, whose EPS is smaller molecular weight compared to RO1 and RO3 (Table 1), contains two extra GHs. It remains to be determined if these GHs have any role in degradation and production of low molecular weight EPS by RO2.

All the isolates contain YwqC, YwqD, and YwqE proteins encoded by ywqCDEF operon, which is implicated in the synthesis of capsular polysaccharides (Supplementary Table 4). The protein YwqD is a tyrosine kinase that phosphorylates YwqF and TuaD, and UDP-glucose dehydrogenases, which are the key enzymes required for capsular polysaccharide biosynthesis (Mijakovic et al., 2003). Studies show that the eps operon (epsABCDEFGHIJKLMNO) of B. subtilis synthesize major components of biofilm EPS (Vlamakis et al., 2013). EpsE, the best-studied protein encoded by eps operon is bifunctional; acting as glycosyltransferase and a molecular clutch inhibiting flagellar rotation (Blair et al., 2008). All the RO isolates possess proteins that show significant similarity to EpsE protein (Supplementary Table 4). These results show that RO isolates have the genetic potential to synthesize EPS components required for biofilm formation by the genus Bacillus, and that the polysaccharide component of EPS may be similar in composition to B. subtilis given the genetic similarities. However, further research is needed to confirm this suggestion.

The proteins encoded by tapA-sipW-tasA operon produce a major protein component of B. subtilis biofilms (Branda et al., 2006). The TasA forms amyloid fibers anchored to cell wall through TapA (Romero et al., 2010). SipW is a type I signal peptidase needed to process and release TasA and TapA to the extracellular milieu (Terra et al., 2012). Genome analyses reveal that only RO3 contain the genes that encode proteins similar to TasA, TapA, and SipW (Supplementary Table 4). The absence of TasA, TapA, and SipW similar proteins in RO1 and RO2 could result from incomplete genome sequences, as the genome sequences of RO1 and RO2 are 98.56% complete, whereas RO3 is 98.85% complete (Rehman et al., 2020b). Another possibility is that RO1 and RO2 produce functionally similar proteins with different primary sequences that could not be identified by blastp analysis.

Genome analysis showed that all the RO Bacillus strains carry genes, albeit different in number, involved in the biosynthesis of flagella (Figure 1). We detected 18 flagellar biosynthesis genes in the RO2 genome, and their organization in the genome is shown in Figure 1. Different flagellar biosynthesis genes found in the genomes of RO1, RO2, and RO3 are given in Supplementary Table 5. Bacterial flagella not only play a role in cellular motility but also take part in surface sensing and formation of biofilm (Belas, 2014). The RO membranes have different surface properties, such as hydrophobicity, surface roughness, and charge (Nguyen et al., 2012). Flagellar motion allows bacteria to overcome the hydrodynamic conditions and the repulsive forces due to surface chemistry (Lemon et al., 2007). Once attached, the Bacillus strains can produce EPS and activate other genes involved in the formation and further development of biofilm (Lee et al., 2010).

Previously, we used high-throughput sequencing to identify the composition and functional potential of the microbial community on the biofouled RO membranes (Rehman et al., 2019). We found that Firmicutes (to which genus Bacillus belongs) constitute the fourth most abundant Phyla detected on biofouled RO membranes. However, the relative abundance of Bacillus in the context of biofouling may be irrelevant. A previous study on the biofouling of the RO membrane using pure cultures of Bacillus showed that it is not the cell number of Bacillus, but their EPS and activation of the specific genes that determine the extent of biofouling (Lee et al., 2010).

After 24 h of incubation, biofilm formation was not detected for any isolate on polystyrene plates. It is possible that the RO isolates (Gram-positive Bacillus) were slow biofilm formers compared to biofilm forming Pseudomonas aeruginosa studied in our lab (Rehman and Leiknes, 2018), or that the laboratory conditions and culture media were not suitable for biofilm formation by these bacteria. Furthermore, the negative charges on Bacillus spp. (Malanovic and Lohner, 2016) and polystyrene surfaces (tissue culture treatment) may have prevented the initial attachment of bacteria to the surface. Indeed, Pang et al. (2005) have demonstrated that bacteria with higher negatively charged surfaces form less biofilm (Pang et al., 2005). Nonetheless, after 48 h of incubation, we observed biofilm formation for all the isolates, albeit in low amounts (Figure 2A). The formation of a surface-conditioning film most likely allowed the bacteria to overcome the electrostatic repulsion and attach to the surface (Lorite et al., 2011). We observed a slightly higher biofilm formation by these isolates on glass surface compared to polystyrene after 48 h (Supplementary Figure 3), which can be attributed to the different surface properties of glass and polystyrene. Similar to our results, Pang et al. (2005) have shown that some of their RO isolates yielded low biomass in a biofilm formation assay (Pang et al., 2005). There was no statistically significant difference in biofilm formed by the three isolates (p > 0.05).

Note that the bacteria isolated in this study represent only a fraction of the bacterial diversity on the RO membrane (Rehman et al., 2019). The microbes in RO elements are exposed to harsh conditions, such as high pressure, salinity, and shear force (Nagaraj et al., 2017a), and such conditions may be selective for microbes that have superior attachment and ability to survive under these conditions. It is not easier to replicate these harsh conditions in the lab, making it challenging to capture bacterial diversity through cultivation-based methods. Therefore, it is important to study the phenotypic and chemical characteristics of cultivable bacteria frequently detected in the RO system (Belila et al., 2016; Nagaraj et al., 2017b). Furthermore, Bacillus are chlorine resistant, which allows them to survive the pretreatment and deposit on the membrane, contributing to biofouling and flux decline (Bhojani et al., 2018).

In this study we extracted and analyzed the secreted fraction of EPS, which has a higher binding capacity (Kim et al., 2006). The amount of proteins and polysaccharides produced by all three isolates growing in biofilm mode was very similar (Figure 2B). The DNA was below the detection limit in the extracted EPS. Our results are consistent with those reported for acid mine drainage biofilms and EPS isolated from biofilm-forming marine bacteria (Jiao et al., 2010; Siddik and Satheesh, 2019), where a higher carbohydrate-to-protein ratio was reported. Furthermore, we have previously shown that the carbohydrate-to-protein ratio increases as the biofilm grows older (Matar et al., 2016). Jiao et al. (2010) suggested that the low protein content in the EPS of biofilms could be attributed to protein degradation by extracellular proteases and by harsh conditions. Further, studies have shown that marine bacterial isolates can produce extracellular proteases (Fulzele et al., 2011), which can potentially degrade extracellular proteins.

The molecular weight distribution of the EPS extracted from the three isolates was determined using size exclusion chromatography. The refractive index (RI) and UV chromatogram of the EPS samples were similar to those recently reported (Liu, 2018; Supplementary Figure 4). The RI chromatogram shows that the EPS from RO1, RO2, and RO3 were eluted as two size groups by the HPSEC column (Table 1 and Supplementary Figure 4). The first group (peak 1) of RO1 was eluted in the range of 12.4–13.7 min, and the second group (peak 2 and 3) was eluted between 14.2 and 18.0 min (Table 1). For RO2, the first group (peak 1) of the EPS components was eluted in the range 13.3–14.7 min, and the second group (peak 2 and 3) was eluted between 14.7 and 18.0 min (Table 1). For RO3, the first peak was eluted between 12.1 and 13.9 min and the second peak between 14.1 and 17.9 min (Table 1).

The size exclusion chromatography (SEC) column separates molecules based on their hydrodynamic volumes with larger molecules elute faster than smaller molecules. The first fraction (peak 1) of the EPS, which eluted faster, showed no UV absorbance in all the samples (Supplementary Figure 4). We identified this higher molecular weight non-UV absorbing fraction of EPS as polysaccharides (Alasonat and Slaveykova, 2012). The second group (peak 2 and 3) of the EPS components, which absorbed UV, was most likely proteinaceous substances as no DNA was detected in the EPS samples. The peaks represent refractive index (RI), which depends on the molar concentration and properties of the solute in a multi-component solution (Zhu et al., 2016). Therefore, the peak size may not reflect the amount of polysaccharide and protein. The peaks that were outside the calibration range of the column were not analyzed (Supplementary Figure 4).

The average molecular weights of extracellular polymers produced by RO1, RO2, and RO3 were 201.3, 62.4, and 213.1 KDa, respectively. The molecular weight of polysaccharides from strain RO2 was smaller than RO1 and RO3 (Table 1). However, the molecular weight of the EPS components from all the isolates was within the range described in previous studies (Brian-Jaisson et al., 2016; Liu, 2018). We observed that the polydispersity index (PD) of the proteinaceous substances was slightly higher than the PD of the polysaccharides, possibly because formaldehyde and NaOH can destroy disulfide bonds in glycoproteins, which could lead to higher polydispersity (Alasonat and Slaveykova, 2012).

The high molecular weight exopolysaccharides produced by RO isolates may serve as a scaffold for the attachment of proteins, lipids, and nucleic acids, which may cumulatively form biofilm matrix (Bales et al., 2013). Studies have shown that high molecular weight organic matter is responsible for fouling of NF membranes (Nilson and DiGiano, 1996). The physical properties of EPS produced by the RO isolates suggest that they can attach to and condition the surfaces for subsequent biofilm formation.

FTIR analysis was performed to identify the functional groups associated with the EPS polymers. The peaks were assigned to functional groups based on previous studies (Jiao et al., 2010; Brian-Jaisson et al., 2016; Maddela et al., 2018). Overall, the FTIR spectra of the isolates were very similar, indicating that the chemical composition of the EPS produced by these isolates was also very similar.

The FTIR spectra in the range 900–1,300 cm^–1 are characteristic of functional groups in polysaccharides and nucleic acids (Brian-Jaisson et al., 2016). However, DNA was not detected in the EPS samples; thus, spectra in this range most probably corresponded to polysaccharides (Figure 3). FTIR analysis of the EPS from all the isolates showed the presence of α-1,4 glycosidic linkage (peak 1 and 2), uronic acids (peak 3), O-acetyl group (peak 4), and carboxylic group (peak 5) (Figure 3; Tang et al., 2017; Maddela et al., 2018). Jiao et al. (2010) previously suggested that the proteins show FTIR spectra in the range of 1,500–1,700 cm^–1 (Jiao et al., 2010). In this range, peak 6 corresponds to stretching vibrations of C = O and C-N [amide (II)] and peak 7 corresponds to C = O and C-N [amide (I)] groups commonly found in the peptide bonds of proteins (Alasonat and Slaveykova, 2012; Tang et al., 2017). Lastly, peak 8 represents C-H vibrational stretching of aliphatic hydrocarbons, which is indicative of lipids (Figure 3; Jiao et al., 2010; Tang et al., 2017).

In a recent study, peak 1, 2, and 6 were correlated with the fouling potential of pure cultures of bacteria (Maddela et al., 2018). Furthermore, the authors suggested that EPS of high fouling bacterial strains contained uronic acids (peak 3) and O-acetyl groups (peak 4), which imparted gelling properties to EPS and enhanced attachment of bacteria to the membrane.

Overall, the FTIR spectra confirmed the presence of proteins, polysaccharides, and lipids in the EPS produced by the RO isolates. The results show that the Bacillus are capable of producing diverse biological molecules with varying chemical nature and charge. These molecules may facilitate biofilm formation by complex communities of microbes on the RO membrane.

¹H NMR analysis was carried out to further explore the chemical composition of EPS obtained from the RO isolates (Figure 4A). The three EPS samples showed well-resolved peaks. The relative abundance of the five ¹H chemical shifts regions was calculated, which were assigned based on (Table 2; Hertkorn et al., 2013). These results reveal small differences in the relative abundance of different structural categories of organic H between EPS of the RO isolates. However, these differences may not be significant as suggested by previous studies (Hertkorn et al., 2013; Fox et al., 2018).

Results based on the five ¹H chemical shifts showed that EPS from the RO isolates was similar to EPS of dissolved organic matter obtained from various marine sources (Hertkorn et al., 2013; Fox et al., 2018). Our EPS samples showed a higher abundance of aliphatics (Table 2), similar to what has been reported for dissolved organic matter from reverse osmosis-electrodialysis system (RO-ED) (Table 2; Koprivnjak et al., 2009). The aliphatics were the most abundant structural category of organic H identified for RO1 and RO2 (Table 2). These results suggest that hydrophobic compounds constitute a significant proportion of organic material deposited on RO membranes. Also, the abundance of HC-O (singly-oxygenated hydrogen and carbon) chemical shifts (ascribed to carbohydrates) in the EPS of RO isolates and organic matter from RO-ED (Koprivnjak et al., 2009) was similarly high. However, the abundance of HC-O was found to be lower in organic matter from other marine sources (Fox et al., 2018). The higher abundance of HC-O in RO-ED and also in our samples indicates that carbohydrates have a higher binding capacity to the RO membranes. However, experiments using purified polysaccharides and their ability to bind to RO membrane are needed to confirm this.

Furthermore, we identified proton chemical shifts indicating CH₃ groups (Peak A, 0.85 ppm), which were previously reported for exopolymers obtained from bacteria but not from algae (Figure 4A; Li et al., 2015). Similarly, CH₂ groups (Peak B, 1.25 ppm) associated with cutins, lipids, and waxes were also detected (Figure 3A; Simpson et al., 2007). We also observed proton resonances (Peak C, 2.1 ppm) indicating methyl groups of N-acetyl amino sugars (Figure 4A; Fox et al., 2018). Resonances corresponding to N-acetylated amino sugars have been reported for high molecular weight dissolved organic matter in surface seawater and semi-labile polysaccharides (Hertkorn et al., 2006; Abdulla et al., 2013). The acylated polysaccharides detected in the saline environment are more stable than their non-acetylated counterparts (Abdulla et al., 2010; Fox et al., 2018). Lastly, we detected chemical shifts (Peak D, 4.1 ppm) indicating fucosylated compounds in the EPS (Figure 4A). These compounds are produced by marine bacteria and play a role in the attachment of barnacles on different surfaces (Siddik and Satheesh, 2019), suggesting their role in surface conditioning.

Furthermore, the peaks in the aromatic area at 7.1 and 8.2 ppm corresponded to the aromatic amino acid side chain and amides in peptides, respectively, indicating presence of proteins and its hydrophobic nature. Proteins constitute an important component of transparent exopolymers (TEP), which are significant contributors to surface biofouling (Villacorte et al., 2013; Li et al., 2015). Based on the ¹H NMR analyses, it appears that the EPS produced by the RO isolates has chemical features similar to that of organic matter found in marine environments. Furthermore, it appears that EPS is structurally stable and suited for harsh environments.

Solid-state ¹³C NMR spectroscopy was conducted to identify the different functional groups associated with EPS. The ¹³C carbon atoms resonate at different frequencies depending on the electronic structure of the immediate environment. Aliphatic carbons resonate in the frequency range of 0–40 ppm, while secondary alcohols of carbohydrates and carbon atoms of glycoside bonds resonate in the frequency range of 60–90 and 95–106 ppm, respectively (Jiao et al., 2010). The resonance frequency of glycosidic carbon is also indicative of their linkage, i.e., alpha or beta oxygen linkages. Carbon atoms forming linkages to alpha oxygen resonate in the range 95–103 ppm, while linkages to beta oxygen resonate in the range 103–106 ppm.

As shown in Figure 4B, the peak observed at 20 ppm indicates methyl carbon (Khan et al., 2014), while the peak at 30 ppm indicates methylene groups in long-chain hydrocarbons (Metzger et al., 2009; Jiao et al., 2010), suggesting the presence of hydrophobic compounds in EPS. These hydrophobic compounds may represent lipids that originate from phospholipids and lipopolysaccharides of the bacterial cell membrane (Metzger et al., 2009). Recently, it has been shown that lipids constitute a considerable fraction of the organic matter deposited on the RO membrane treating seawater (Khan et al., 2014). In line with previous studies such as (Metzger et al., 2009), it appears that EPS produced by the bacterial isolates in this study contain large amounts of lipids. The presence of lipids in the EPS samples was also confirmed by ¹H NMR analysis (Table 2).

The NMR spectra confirmed the presence of secondary alcohols and revealed the type of glycoside bonds linking the monomers in the bacterial isolates. We detected a peak at 72 ppm, indicative of secondary alcohol carbon, in all the EPS samples (Figure 4B), albeit much lower in RO1 (Khan et al., 2013). The peaks at 90 and 102 ppm indicated that carbohydrates in all the samples contain α-glycoside bonds (Figure 4B). However, the peak at 102 ppm is small in RO2 and RO3 compared to RO1. It is not clear if the small peak for α-glycosidic bonds in RO2 can account for low molecular weight EPS produced by this bacteria (Table 1). FTIR analysis also confirmed the presence of α-1,4 glycoside linkage in all the EPS samples (Figure 4). The peaks at 130 and 158 ppm were indicative of proteins containing hydroxy aromatic amino acids, such as tyrosine (Figure 4B; Khan et al., 2013). These peaks were detected in all the EPS samples from RO isolates. A sharp peak at 173 ppm indicated the presence of an amide carbon, suggesting the presence of proteins in the sample (Figure 4B; Jiao et al., 2010).

The ¹³C NMR spectra of EPS harvested from the RO membrane bacterial isolates show many resonances that have been previously detected in foulants isolated from RO membranes used for treating seawater (Khan et al., 2014, 2015). Based on NMR spectra, Khan et al. (2013) argued that the foulants on RO membranes mainly originated from bacterial cells. Our results support this claim, where the ¹³C NMR spectra of EPS produced by bacteria isolated from the biofouled RO membrane in this work bear similarity to foulants on the RO membranes (Figure 4B). Therefore, EPS produced by our isolates can be used in studies investigating the role of EPS in membrane conditioning, biofilm initiation and flux decline of RO membranes.