Fructan utilization by members of marine Gammaproteobacteria involves SusC/D-like proteins

TL;DR

This study shows that some marine bacteria use SusC/D-like proteins to break down fructans, a type of sugar found in the ocean.

Contribution

The study reveals that SusC/D-like proteins, previously thought unique to Bacteroidota, are also used by marine Gammaproteobacteria for fructan utilization.

Findings

P. distincta uses a SusD-like protein and SusC-like transporter to import inulin-type fructans.

A periplasmic exo-active GH32 enzyme degrades imported fructans in P. distincta.

Comparative genomics shows SusC/D-like proteins are common in marine Gammaproteobacteria and often co-occur with GH32s.

Abstract

Fructans are ubiquitous in terrestrial ecosystems, however, these glycans are underexplored in the marine environment. We have discovered that the Antarctic gammaproteobacterium Pseudoalteromonas distincta is highly adapted to the degradation of fructose-containing substrates. This is enabled by proteins encoded in several genomic regions, including a fructan polysaccharide utilization locus (PUL). In addition to a glycoside hydrolase from family 32 (GH32), the fructan PUL encodes two proteins that have been described as specific for the phylum Bacteroidota and were previously unknown for the class Gammaproteobacteria (phylum Pseudomonadota): a glycan-binding SusD-like protein and a SusC-like TonB-dependent transporter (TBDT), which work as a complex in glycan import in Bacteroidota. Proteome, biochemical, sequence, and structural analyses indicate that the SusD-like protein and SusC-like TBDT of P. distincta mediate the uptake of inulin-type fructans, followed by degradation by a periplasmic exo-active GH32. In contrast, P. distincta likely degrades levan-type fructans via an extracellular endo-acting GH32 that is not encoded in the fructan PUL. Comparative genomics identified further SusD-like proteins and SusC-like TBDTs in Gammaproteobacteria, most of which are co-encoded with GH32s, indicative of fructan PULs, and are frequently associated with the marine habitat. Our study thus suggests that SusC/D-like complexes are not exclusive to the phylum Bacteroidota. It further shows that fructans contribute to the marine glycan pool and are targeted by specialized marine communities.

Figures

The gammaproteobacterium P. distincta is specialized in the degradation of fructose-containing substrates. (a) The genomic context of two GH32-encoding genes, one in a fructan PUL (PdGH32PUL) and one distal to the PUL (PdGH32DIS), is shown. In addition, P. distincta contains a genomic region likely specific for sucrose (GH13_18) and β-glucans (GH16_3). The cluster might be even longer (Fig. 2). Fructan PUL: EGI74656-61, distal fructan-related CAZymes: EGI72340/41, sucrose/β-glucan: EGI73127-36 (Fig. 2). (b) Growth of P. distincta in MPM supplemented with 0.2% of fructans, FOSINU, or sucrose. Experiments were conducted in triplicate. There is a different time scale for bacterial levan and the control (no substrate: P. distincta in MPM without carbon source). P. distincta was also grown on fructose (Fig. S1) and substrate controls were performed (Fig. S2). Inu: inulin; LevPL: plant-derived levan; LevBAC: bacterial levan; FOSINU: fructo-oligosaccharides from inulin.

The abundance of fructan PUL-encoded proteins is increased in inulin-grown and levan-grown cells. Relative protein abundance of selected proteins of P. distincta grown on fructose-containing substrates and α-glucans (control), indicated above the plot. Relative protein abundance values (%riBAQ, relative intensity-based absolute quantification, giving the relative protein abundance within the sample, n = 3) were log-normalized and z-scored. Protein accessions were abbreviated, e.g. 74657 refers to EGI74657.1. Inu: inulin; LevPL: plant-derived levan; LevBAC: bacterial levan; FOSINU: fructo-oligosaccharides from inulin.

A SusC/D-like complex in P. distincta. Superposition of the AF2-predicted 3D structures of PdSusC/DPUL, colored using the pLDDT confidence scores, with the corresponding levan-binding complex from Bacteroides thetaiotaomicron VPI-5482T (PDB ID 6ZAZ, [64]) in gray. Boxed region “I” highlights “hinge” loops responsible for the interaction between the glycan-binding SusD-like protein and SusC-like TBDT. Boxed region “II” shows residues of the SusD-like protein from B. thetaiotaomicron binding FOS from levan and their superposition with residues of the P. distincta PdSusDPUL (colored with confidence scores), when similar or identical, shown in top view (see also Fig. S10). A levan-interacting residue of the SusC-like protein in B. thetaiotaomicron [75]—also shown in box “I”—is marked with an arrow head. pLDDT confidence scores: “very high” (pLDDT > 90) dark blue, “high” (90 > pLDDT>70) cyan, “low” (70 > pLDDT > 50) yellow and “very low” (pLDDT < 50) orange.

The PUL-encoded SusD-like protein binds FOSINU, whereas the PUL-encoded GH32 cleaves inulin, levan, and sucrose. (a) Binding of PdSusDPUL to FOSINU. ITC raw titration data (upper panel) and integrated heat peaks (lower panel) as a function of the molar ratio of ligand per receptor (a representative experiment is shown, n = 3, Fig. S12). (b) Activity of PdGH32PUL on fructans and sucrose, detected by RSA. Bars represent mean ± standard deviation (n = 3). Values were corrected based on controls with the respective glycan, but without the addition of CAZyme. A non-active CAZyme is shown for reference (Fig. S13a). Inu: inulin; FOSINU: fructo-oligosaccharides from inulin; LevPL: plant-derived levan; LevBAC: bacterial levan.

SusD-like proteins co-occur with SusC-like TBDTs and GH32s in Gammaproteobacteria. The GEM and UGCMP databases were searched for gammaproteobacterial SusD-like proteins (4681 representative genomes). Shown is their genomic context as well as the presence of additional GH68/GH32 pairs or GH32s.

Fructan-degrading activity is relevant in marine surface waters around the globe. Global abundance (shown as % of mapped reads) of the (a and b) GH32 compared to the abundance of the (c and d) GH16_3 using the ocean gene atlas v2.0 metagenome datasets (a and c) and metatranscriptome datasets (b and d), Tara oceans microbiomes reference catalog v2 + metaG Arctic inside (prokaryotes) and Tara oceans microbiomes reference catalog v2 + metaT Arctic inside (prokaryotes), respectively.

Proteins that are encoded in different genomic regions mediate the utilization of inulin, levan, or sucrose, but could be linked during downstream processing. Our model for the utilization of fructose-containing substrates in P. distincta is presented. Left: Fructan PUL-encoded proteins enable inulin use. Presumably, the small-sized inulin is shuttled directly into the periplasm by the SusC/D-like transport complex, where it is degraded by the exo-active CAZyme PdGH32PUL, as previously described for human gastrointestinal Bacteroidota [17]. Top center: A distal PdGH68DIS could produce levan as EPS, this levan could be degraded/remodeled by the endo-levanase PdGH32DIS. Fragments released from PdGH32DIS might be further degraded in the periplasm by the fructan PUL-encoded PdGH32PUL, which accepts both linkages. The transporter importing these oligosaccharides is unknown. Right: Proteins encoded in a sucrose PUL may metabolize fructose and sucrose, released upon degradation of inulin and levan. In inulin and levan, fructosyl units are β(2,1)- or β(2,6)-linked to sucrose, respectively. P: phosphate.

Full text

Introduction

We are only beginning to understand the diversity of glycan structures that marine algae—but also fungi and bacteria—produce and how such glycans contribute to carbon flux in the ocean. Members of the phylum Bacteroidota and the class Gammaproteobacteria (phylum Pseudomonadota) are specialists in the degradation of such oceanic glycans and have evolved unique glycan utilization strategies.

In the Bacteroidota phylum, gene clusters encoding glycan-disassembling carbohydrate-active enzymes (CAZymes), such as glycoside hydrolases (GHs), and glycan-transporting SusD-like proteins and SusC-like TonB-dependent receptors or transducers (referred herein as transporters, TBDTs) are referred to as polysaccharide utilization loci (PULs), where “Sus” refers to homologs of the starch utilization system of the human intestinal Bacteroides thetaiotaomicron [1, 2]. The SusD-like protein and the SusC-like TBDT are structurally designed to function as a complex (referred herein also as SusC/D-like complex or SusC/D-like pair), in which the outer membrane-tethered SusD-like protein, a lipoprotein, binds and transmits the oligosaccharides released by extracellular CAZymes to the SusC-like TBDT using the “pedal-bin” mechanism [3]. Compared to “classical” TBDTs, SusC-like TBDTs are larger to accommodate the SusD-like protein [4]. In contrast, carbohydrate utilization clusters of the class Gammaproteobacteria, often referred to as PULs as well, lack SusD-like proteins and the “SusC-like” specific characteristics of TBDTs. In this study, we have discovered a predicted SusC-like TBDT/SusD-like protein pair encoded in a PUL of the Antarctic gammaproteobacterium Pseudoalteromonas distincta ANT/505. In addition to these Bacteroidota-specific proteins, the PUL encodes a GH32 which suggests fructan utilization [5], a substrate which is largely unexplored in the marine ecosystem. Fructans are polyfructose chains, in which fructosyl units are β(2,1)- or β(2,6)-linked to sucrose, referred to as inulin or levan, respectively. In terrestrial ecosystems, where fructan production is widespread [6], fructans serve as storage compounds or osmolytes in plants and are frequently found as exopolysaccharides (EPS) in Gram-positive and Gram-negative bacteria. In bacteria, levan and inulin are produced by GH68s [6]. In contrast, little is known about marine sources of fructans and their relevance in marine ecosystems. Although there are some reports of marine bacteria that produce fructans as EPS [7, 8], fructose-containing EPS [9], or that encode a GH68 [10], little is known about marine algae that produce fructose-containing EPS, such as the microalgae Dunaliella salina [11]. However, plants in ice-free coastal Antarctic regions contain fructans [12].

The ability to degrade fructans is widespread among terrestrial bacteria [13] and has been thoroughly investigated in Bacteroidota of the human intestine [14–17]. For the marine counterpart, inulinase activity of marine fungi and bacteria [18–21] or the presence of GH32-containing PULs in metagenome-assembled genomes of marine bacteria [22, 23] have been reported, but in-depth biochemical and structural analyses of the corresponding CAZymes, as in the case of the Thermotoga maritima GH32 [21, 24], are rare. Moreover, little is known about fructan utilization pathways in general.

In this study, we propose a marine utilization pathway for fructans in the Antarctic gammaproteobacterium P. distincta, which was isolated from sea-ice covered surface water [25]. Our results suggest that specialized marine Gammaproteobacteria degrade fructans using PULs that encode SusC/D-like transport complexes and GH32s.

Materials and methods

Carbohydrates

Plant-derived levan (Timothy grass) and fructo-oligosaccharides (FOS_INU_, DP2–8) from inulin (chicory) were purchased from Megazyme/NEOGEN, inulin (chicory) from Alfa Aesar (VWR) and bacterial levan (Erwinia herbicola), pullulan and maltotriose from Sigma Aldrich (Merck).

Cultivation of Pseudoalteromonas distincta

Cells of overnight cultures in marine broth 2216 (16°C and 200 rpm) were pelleted and washed with an artificial seawater medium (MPM) [26]. This suspension was then used to inoculate main cultures of P. distincta in MPM supplemented with 0.2% of the respective substrates to an OD_600nm_ of 0.1. Inoculated MPM medium without any carbon source served as a control. Cells were cultivated at 10°C and 150 rpm. The same conditions were used for Pseudoalteromonas arctica A 37–1–2 (DSM-18437) [27] and Alteromonas stellipolaris LMG 21856 (DSM-15672) [28]. Colony forming units (CFUs) of P. distincta were determined on marine broth 2216 agar plates, which were incubated at 10°C.

Proteome analysis

Cells were harvested by centrifugation (15 min, 9500 × g, 4°C) in the late exponential growth phase (three biological replicates). For whole-cell proteomes, cell pellets were resuspended in TE-buffer (10 mM Tris, 10 mM EDTA) and cells were lysed via sonication (4 cycles of 25 s, 30% power, pulse 5, Bandelin Sonopuls HD 2070 ultrasonic homogenizer). Cell debris was removed by centrifugation. Intracellular soluble fractions, outer membranes, and outer membrane vesicles were prepared as described before [29–31]. Proteins were extracted from the supernatant using StrataClean beads according to published protocols [32, 33], referred to as the extracellular proteome. Here, 20 μL of bead solution was used to extract 20–25 μg of protein. Otherwise, the Pierce BCA assay determined protein concentration and 25 μg of protein was separated by 1D-SDS-PAGE. Lanes were cut into 10 equal pieces and proteins were digested with trypsin as described before [33]. Peptides were subjected to reversed phase chromatography and analyzed using an online-coupled LTQ Orbitrap (whole-cell proteomes) or LTQ XL Orbitrap (subcellular fractions) mass spectrometer equipped with a nanoelectrospray ion source (Thermo Fisher Scientific Inc., Waltham, MA, USA). Data analysis was performed using MaxQuant v. 1.6.0.16 with the integrated Andromeda search engine [34]. The false discovery rate (FDR) was set to 1% on peptide and protein level, and match between runs was enabled (for subcellular protein fractions only between replicates). %riBAQ values (relative intensity-based absolute quantification giving the relative protein abundance in the sample) were calculated to determine protein abundance per sample. Proteins that were quantified in at least two out of three replicates were considered for calculation of mean values.

For statistical analysis in the case of whole-cell proteomes, the data set was filtered for at least one valid value per protein and missing values were imputed with a small constant. Values were then log-transformed and a row-wise z-score transformation was applied. Significant differences were determined by Welch’s t-test (FDR 5%) in Perseus v. 2.1.0.0 [35, 36] using the filtered, imputed, and log-transformed abundance values of proteins. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [37, 38] partner repository, see Data Availability Statement. PSORTb v. 3.0 [39], DeepLocPro v. 1.0 [40], SignalP v. 6.0 [41], and DeepSig [42] were used to evaluate suggested subcellular localization of proteins.

Cloning

The Gibson assembly strategy [43] was used to clone genes into the pET28a vector with primers listed in Table S1. Primers were designed to produce P. distincta proteins (PdSusD_PUL_: EGI74657.1, PdGH32_PUL_: EGI74659.1, PdGH68_DIS_: EGI72340.1, and PdGH32_DIS_: EGI72341.1) as recombinant N-terminal His_6_-tagged constructs, removing predicted signal peptides from sequences [41, 42]. The NEBuilder HiFi DNA Assembly Master Mix (NEB) was used for assembly following the manufacturer’s protocol. Sequencing (Eurofins) confirmed sequence identity. For protein production, plasmids were transformed into competent Escherichia coli BL21(DE3) cells. Since PdGH32_DIS_ was not active using the predicted open reading frame (ORF), the corresponding gene was re-cloned using an alternative ORF (Table S2) resulting into active CAZyme.

Protein production and purification

E. coli BL21(DE3) clones, harboring the corresponding plasmids, were grown in lysogeny broth (LB) containing 30 μg mL^−1^ kanamycin at 37°C and 120 rpm. Medium was cooled down before isopropyl β-D-1-thiogalactopyranoside (IPTG) induction (0.3 or 1 mM) and cells were grown overnight at 20°C and 120 rpm. In the case of CAZyme activity analysis, proteins were produced using the ZYP5052 autoinduction medium [44], except for the re-cloned PdGH32_DIS_, where the Enpresso B medium (Enpresso GmbH, Berlin, Germany) was used according to the manufacturer’s instructions. Cells were harvested by centrifugation (e.g. 1 L culture: 35 min, 13 881 × g, 4°C) and pellets were stored at −20°C until further treatment. A chemical lysis protocol [45] was used to disrupt the cells (resuspension buffer: 50 mM Tris–HCl pH 8.0, 25% sucrose, ~10 mg lysozyme; lysis buffer: 20 mM Tris–HCl pH 7.5, 100 mM NaCl, 1% sodium deoxycholate, 1% Triton-X). For CAZyme activity tests, cells were lysed via ultrasonication on ice (Bandelin Sonoplus HD2200, KE76 sonotrode; 2 × 3 min, 1 min pause, 50% power, 0.5 pulse) in buffer A (300 mM NaCl, 20 mM Tris–HCl pH 8.0, 20 mM imidazole).

For protein purification, 1 mL or 5 mL HisTrap HP columns were used, equilibrated in buffer A (300–500 mM NaCl, 20 mM Tris–HCl pH 8.0, 20 mM imidazole) and charged with 0.1 M NiSO_4_. The proteins were eluted using a linear gradient of 0%–100% buffer B (300–500 mM NaCl, 20 mM Tris–HCl pH 8.0, 500 mM imidazole) within 20–60 min. If necessary, the purity of the sample was increased by size-exclusion chromatography (SEC) using a Superdex 200 16/60 column (Cytiva) in buffer C (CAZyme activity: 25 mM Tris–HCl pH 8.0, 50 mM NaCl, otherwise: 20 mM Tris–HCl pH 8.0, 200–250 mM NaCl). Otherwise, samples were desalted using a HiPrep 26/10 column and buffer C. Protein concentration was determined using a Nanodrop Spectrophotometer using the molecular weight and calculated extinction coefficient [46]. If necessary, proteins were concentrated at 1890 × g and 4°C using Amicon Ultra centrifugal filter units (Merck) or using a stirred cell ultrafiltration device (10 kDa cut-off).

Reducing sugar assay

The degradation of different oligo- and polysaccharides by CAZymes was carried out using final concentrations of 2 mg mL^−1^ substrate and 30 μg mL^−1^ of CAZyme in 1 × PBS, pH 6.0 (120 μg mL^−1^ in the case of the re-cloned PdGH32_DIS_). The enzymatic reactions were incubated for 24 h at room temperature. If not otherwise stated, glycans without the addition of CAZyme served as controls. For the quantification of reducing ends released upon degradation, 20 μL of the reaction solutions were mixed with 20 μL DNS-reagent (0.4 M NaOH, 30% potassium sodium tartrate, 1% 3,5-dinitrosalicylic acid; DNS) and incubated at 100°C for 15 min. dH_2_O (180 μL) was added and 200 μL of the mixtures were transferred to a microtiter plate. The absorption at 540 nm was measured using a Tecan infinite 200 pro.

Fluorophore-assisted carbohydrate electrophoresis

The enzymatic reactions (10 μL) were lyophilized, resuspended in ANTS (0.1 M in DMSO, 4 μL, Invitrogen) and NaCNBH_3_ (1 M in DMSO, 4 μL), and incubated at 37°C overnight. The labeled samples were mixed with loading dye (62 mM Tris pH 6.8, 0.014% bromophenol blue, 10% glycerol, 8 μL) and separated at 400 V on a 30% polyacrylamide running gel with a 4% stacking gel for about 1 h under consistent cooling. Imaging was performed using a 312 nm UV table.

Isothermal titration calorimetry

PdSusD_PUL_ was dialyzed against 50 mM Tris pH 8.0, 100 mM NaCl at 4°C for at least 48 h. Isothermal titration calorimetry (ITC) analysis was performed using a MicroCal ITC 200 machine. The dialyses buffer was used to dissolve the substrates and to wash the ITC sample cell. All samples were centrifuged before analysis. Injections with 20 mM of inulin (calculated as DP17), 5 mM of plant-derived levan, 7.5 μm of bacterial levan, and 20 mM of sucrose or fructose detected no signal. Since little heat was released during titrations at a concentration of 31.7 mM FOS_INU_ (calculated as DP5) and due to good solubility at high concentrations, we tested even higher concentrations of FOS_INU_: 63.5 mM of FOS_INU_ were injected into 160.97 μm of PdSusD_PUL_, which was performed in triplicate. The following settings were selected: cell temperature 20°C (293.15 K), reference power 10 μCal/s, stirring speed 500 rpm, filter period 1 sec, injection spacing 200 sec, 2 μL injection volume (first injection 0.3 μL), 18 injections in total. As a control, FOS_INU_ was titrated into buffer and buffer into buffer. Due to the fact that the low affinity resulted in a C-value of 0.01 [47] and to the best of our knowledge SusD-like proteins bind in a 1:1 ratio, the n-value was fixed to 1. MICROCAL ORIGIN v.7 was used to analyze the data applying a single-site binding model.

Protein annotation and structure prediction

SusD-like proteins, TBDTs, and SusC-like TBDTs of P. distincta, P. arctica, and A. stellipolaris, as well as SusD-like proteins of the phylogenetic analysis, were predicted with hmmscan (e-value <1E-15 and coverage >0.35). Sequences were screened for SusD-like protein homologs using the PFAM models PF07980 (SusD_RagB), PF12741 (SusD-like), PF12771 (SusD-like_2), and PF14322 (SusD-like_3) [48]. To annotate TBDTs, sequences were screened against PF13715 (CarboxypepD_reg-like domain), PF07715 (Plug), and PF00593 (TonB_dep_Rec). TBDTs also meeting the cut-offs for the TIGRFAM profile TIGR04056 (OMP_RagA_SusC) [49] were classified as SusC-like TBDTs. CAZyme families were assigned with hmmscan vs. dbCAN-HMMdb-V13 using the same threshold.

AlphaFold2 (AF2)-predicted structures were downloaded from the AlphaFold Protein Structure DB [50], which is integrated in the Uniprot knowledgebase [51] and visualized using PyMOL Molecular Graphics System v2.5.4 (Schrödinger, LLC.).

Co-occurrence analysis

Gammaproteobacterial genomes of the unified genome catalog of marine prokaryotes (UGCMP) [52] and the Genomes from Earth’s Microbiomes (GEM) catalog [53] were dereplicated using dRep (v3.4.2, flags: -l 0 -comp 50 -con 10 -pa 0.95 --checkM_method taxonomy_wf) [54]. Resulting 4681 genomes were reannotated using prokka (v1.14.6) [55] and screened for SusD-like homologs using hmmsearch (v3.3.2, gathering threshold) [56] with PFAM models PF07980 (SusD_RagB), PF12741 (SusD-like), PF12771 (SusD-like_2), and PF14322 (SusD-like_3) [48].

In order to examine polysaccharide utilization-related functions around susD-like genes, relevant functions were predicted using a five-gene sliding window. For TBDT annotation, sequences were compared to PF13715 (CarboxypepD_reg-like domain), PF07715 (Plug), and PF00593 (TonB_dep_Rec) using hmmsearch (gathering threshold). Sequences containing PF07715 and PF00593 were kept for further analysis. TBDTs also fitting the gathering threshold for TIGRFAM profile TIGR04056 (OMP_RagA_SusC) [49] were classified as SusC-like proteins. CAZyme families were assigned with hmmscan vs. dbCAN-HMMdb-V12 and dbCAN-sub as well as diamond blastp (v2.1.1.155, e-value cut-off 1E-102) against CAZyDB.07262023, all provided by dbCAN [57]. Results were then filtered using the hmmscan-parser.sh script with an e-value cut-off of 1E-15 and a minimum coverage of 0.35. CAZyme families predicted by at least two tools were kept for further analysis, not considering GT and AA assignments. Results were visualized with the ComplexUpset R package (v1.3.5) [58].

The 4681 representative gammaproteobacterial genomes were searched for GH32 family encoding genes with hmmsearch (v3.3.2, option settings: -E 1 --domE 1 --incE 0.01 --incdomE 0.03) using GH32 HMM profiles from dbCAN-HMMdb-V12 and dbCAN-sub. Hits were further assigned to a final CAZyme family as described above. Their genetic neighborhood 5 genes up- and downstream was reconstructed similar to the predictions mentioned above, including hmmscan (gathering threshold) against the full Pfam-A database (as of 20240508).

Global abundance of fructan-utilization activity

The abundance of genes and transcripts was analyzed using the Ocean Gene Atlas v2.0 (https://tara-oceans.mio.osupytheas.fr/ocean-gene-atlas/) [59, 60] with the Tara Oceans Microbiomes reference catalog v2 + metaG Arctic inside (prokaryotes) and Tara Oceans Microbiomes reference catalog v2 + metaT Arctic inside (prokaryotes) as subjects, and the GH32 and GH16_3 HMM models as queries. The website’s integrated hmmsearch tool was used with default parameters, applying a threshold of 1E-15 and calculating the abundance as % of mapped reads. Retrieved hits were further curated using hmmscan v3.3.2 against the dbCAN-HMMdb-V13 database. Results were filtered with the hmmscan-parser.sh script from dbCAN, using an e-value cut-off of 1E-15 and a minimum coverage of 0.35. In addition, the P. distincta fructan PUL (EGI74656.1-EGI74659.1; ADOP01000005.1:65326-73395) was used as query for a BLASTN search (default parameters) against the Tara Oceans Microbiomes reference catalog v2 + metaG Arctic inside (prokaryotes). Abundance values (% of mapped reads) from surface water samples (0.22–3 μm) were visualized in R using the ggplot2 and rnaturalearth packages.

Results

Pseudoalteromonas distincta contains several genomic regions to utilize fructose-containing substrates

The P. distincta fructan PUL classifies as a Bacteroidota PUL. Upstream of a GH32 (PdGH32_PUL_), the PUL encodes a SusC-like TBDT (PdSusC_PUL_) and a SusD-like protein (PdSusD_PUL_) (Fig. 1a). In addition to the receptor/β-barrel domain PF00593 and plug domain PF07715 that identify “classical” TBDTs, the SusC-like subclade SusC/RagA model TIGR04056 was identified for the transport protein PdSusC_PUL_ (SusC/D-like protein homologs are named RagA/B in the human oral Porphyromonas gingivalis and are specific for oligopeptides [61]). In line with this, the Pfam family for SusD/RagB homologs PF07980 was identified for the co-encoded putative glycan-binding protein PdSusD_PUL_, which contains a lipoprotein signal peptide. Apart from PdSusC/D_PUL_, no other SusC-like TBDTs and SusD-like proteins were identified in P. distincta.

The gammaproteobacterium P. distincta is specialized in the degradation of fructose-containing substrates. (a) The genomic context of two GH32-encoding genes, one in a fructan PUL (PdGH32PUL) and one distal to the PUL (PdGH32DIS), is shown. In addition, P. distincta contains a genomic region likely specific for sucrose (GH13_18) and β-glucans (GH16_3). The cluster might be even longer (Fig. 2). Fructan PUL: EGI74656-61, distal fructan-related CAZymes: EGI72340/41, sucrose/β-glucan: EGI73127-36 (Fig. 2). (b) Growth of P. distincta in MPM supplemented with 0.2% of fructans, FOSINU, or sucrose. Experiments were conducted in triplicate. There is a different time scale for bacterial levan and the control (no substrate: P. distincta in MPM without carbon source). P. distincta was also grown on fructose (Fig. S1) and substrate controls were performed (Fig. S2). Inu: inulin; LevPL: plant-derived levan; LevBAC: bacterial levan; FOSINU: fructo-oligosaccharides from inulin.

Additional genomic regions are presumably related to the utilization of fructans or fructose-containing substrates by P. distincta: a distal GH32 (PdGH32_DIS_) and GH68 (PdGH68_DIS_), as well as a PUL that encodes proteins to cleave sucrose (GH13 subfamily 18, GH13_18) and to metabolize fructose (fructokinase) (Fig. 1a). However, the PUL also encodes a GH16_3, which could indicate β-glucan use [62]. It is unknown whether this genetic region is actually one PUL dedicated to a common substrate or whether it is two PULs.

Using fructans of terrestrial origin, we confirmed that P. distincta grows on inulin, oligosaccharides from inulin (FOS_INU_), sucrose, and fructose (Fig. 1b, Fig. S1). In comparison to inulin, P. distincta reaches lower optical densities on plant-derived levan and hardly grows on bacterial levan (Fig. 1b). This was further verified by CFU counting (Fig. S2) and by demonstrating that fluorescently labeled inulin and plant-derived levan are internalized by P. distincta (Fig. S3, Supplementary Methods).

Inulin and levan stimulate the production of proteins encoded in the fructan polysaccharide utilization locus

The relative abundance of proteins (%riBAQ) encoded in the genomic regions associated with the use of fructans was increased in cells grown on inulin, FOS_INU_, and both levans compared to cells grown on fructose and α-glucans (Fig. 2). The fructan PUL-encoded proteins PdSusC_PUL_, PdSusD_PUL_, and PdGH32_PUL_ were abundant in whole cell proteomes of cells grown on inulin, FOS_INU_, and both levans (Fig. S4, Supplementary Dataset 1), and were significantly more abundant compared to pullulan-grown cells (Fig. S5, Supplementary Dataset 1). In comparison, the relative protein abundance of the distal PdGH68_DIS_ and PdGH32_DIS_ was very low in fructan-grown cells (%riBAQ <0.002, Fig. S6). However, PdGH68_DIS_ was the most abundant protein in the extracellular protein fraction of inulin-grown cells, and the third most abundant in cells grown on levan derived from plants (Fig. S7, Supplementary Dataset 2).

The abundance of fructan PUL-encoded proteins is increased in inulin-grown and levan-grown cells. Relative protein abundance of selected proteins of P. distincta grown on fructose-containing substrates and α-glucans (control), indicated above the plot. Relative protein abundance values (%riBAQ, relative intensity-based absolute quantification, giving the relative protein abundance within the sample, n = 3) were log-normalized and z-scored. Protein accessions were abbreviated, e.g. 74657 refers to EGI74657.1. Inu: inulin; LevPL: plant-derived levan; LevBAC: bacterial levan; FOSINU: fructo-oligosaccharides from inulin.

Moreover, relative abundance of proteins encoded in the sucrose-related part of the sucrose/β-glucan genetic region was often higher in cells grown on fructose-containing substrates (DP ≥ 2) compared to α-glucan conditions (Fig. 2, Fig. S8). Therefore, proteins encoded in the β-glucan part (GH16_3 and co-encoded TBDT) are likely regulated independently from the sucrose PUL. In addition, relative abundance of proteins related to the utilization of α-glucans was increased in cells grown on bacterial levan (Fig. 2), but %riBAQ values were much lower compared to α-glucan-grown cells (Fig. S9). P. distincta hardly grew on bacterial levan. The increased protein abundance of α-glucan-related proteins might indicate that P. distincta uses an α-glucan-type storage polysaccharide, like glycogen [63], when cultured with bacterial levan.

Inulin specificity of the SusC/D-like complex

Structural and sequence comparisons showed that the P. distincta SusD-like protein and the SusC-like TBDT (PdSusC/D_PUL_) are indeed very similar to Bacteroidota fructan-specific homologs. For comparisons, we used proteins from human intestinal Bacteroidota, where fructan utilization is well-studied and 3D crystal structures of levan-specific proteins are available.

The P. distincta SusD-like protein and the SusC-like TBDT have a high sequence similarity to the Bacteroidota homologs known or suggested to bind and transport levan or inulin (PdSusD_PUL_: 53.5%–61.5%, PdSusC_PUL_: 56.6%–61.7%, Table S3, Fig. S10). Furthermore, the AF2-predicted 3D structures of the P. distincta proteins superimpose very well with the corresponding levan-specific complex from B. thetaiotaomicron VPI-5482^T^ (PDB ID 6ZAZ, [64]) (Fig. 3). However, some levan-binding residues of B. thetaiotaomicron VPI-5482^T^ are missing in P. distincta (Fig. 3, Fig. S10) and in a large-scale phylogenetic analysis, PdSusD_PUL_ did not cluster with these selected Bacteroidota homologs (Fig. S11, Supplementary Methods). We therefore investigated ligand binding by PdSusD_PUL_. Affinity gel electrophoresis (1% of glycan used) and surface plasmon resonance did not detect binding of PdSusD_PUL_ to levan or inulin (data not shown). Similarly, isothermal titration calorimetry (ITC) did not show binding to inulin, levan, sucrose, or fructose at the concentrations used (data not shown, see Materials and Methods). However, PdSusD_PUL_ did show low affinity to FOS_INU_ (K_a_ 43 ± 4 M^−1^, Fig. 4a, Fig. S12a). The interaction was enthalpically favorable and entropically unfavorable (Fig. S12b). Attempts to obtain a 3D crystal structure for PdSusD_PUL_ to gain further insight into ligand binding were not successful (Supplementary Methods).

A SusC/D-like complex in P. distincta. Superposition of the AF2-predicted 3D structures of PdSusC/DPUL, colored using the pLDDT confidence scores, with the corresponding levan-binding complex from Bacteroides thetaiotaomicron VPI-5482T (PDB ID 6ZAZ, [64]) in gray. Boxed region “I” highlights “hinge” loops responsible for the interaction between the glycan-binding SusD-like protein and SusC-like TBDT. Boxed region “II” shows residues of the SusD-like protein from B. thetaiotaomicron binding FOS from levan and their superposition with residues of the P. distincta PdSusDPUL (colored with confidence scores), when similar or identical, shown in top view (see also Fig. S10). A levan-interacting residue of the SusC-like protein in B. thetaiotaomicron [75]—also shown in box “I”—is marked with an arrow head. pLDDT confidence scores: “very high” (pLDDT > 90) dark blue, “high” (90 > pLDDT>70) cyan, “low” (70 > pLDDT > 50) yellow and “very low” (pLDDT < 50) orange.

The PUL-encoded SusD-like protein binds FOSINU, whereas the PUL-encoded GH32 cleaves inulin, levan, and sucrose. (a) Binding of PdSusDPUL to FOSINU. ITC raw titration data (upper panel) and integrated heat peaks (lower panel) as a function of the molar ratio of ligand per receptor (a representative experiment is shown, n = 3, Fig. S12). (b) Activity of PdGH32PUL on fructans and sucrose, detected by RSA. Bars represent mean ± standard deviation (n = 3). Values were corrected based on controls with the respective glycan, but without the addition of CAZyme. A non-active CAZyme is shown for reference (Fig. S13a). Inu: inulin; FOSINU: fructo-oligosaccharides from inulin; LevPL: plant-derived levan; LevBAC: bacterial levan.

Hydrolysis of inulin and levan by the fructan PUL-encoded glycoside hydrolase from family 32

The PUL-encoded CAZyme PdGH32_PUL_ hydrolyzed inulin, FOS_INU_, plant-derived levan, and sucrose, but also showed low activity on bacterial levan, detected using reducing sugar assay (RSA) and fluorophore-assisted carbohydrate electrophoresis (FACE) (Fig. 4b, Fig. S13a–d). Results, based on end-point measurements, indicated exo-activity of PdGH32_PUL_, because sucrose was still hydrolyzed (Fig. 4b, Fig. S13d). The suspected exo-activity was further confirmed by a time-course experiment using inulin and plant-derived levan (Fig. S14). In line with this, AF2 predictions indicate a “funnel”-like binding pocket of PdGH32_PUL_ (Fig. S15).

A special feature of the PUL-encoded PdGH32_PUL_ is an additional carbohydrate-binding module (CBM38), which protrudes between blade IV and V of the five-bladed β-propeller catalytic module (Fig. S16). Attempts to obtain 3D crystal structures of PdGH32_PUL_ were not successful (Supplementary Methods).

In contrast to the PUL-encoded PdGH32_PUL_, the distal PdGH32_DIS_ of P. distincta is levan-specific (Fig. S17) and endo-active (Fig. S18). In agreement with this, the predicted catalytic groove of the distal PdGH32_DIS_ is larger compared to the PUL-encoded PdGH32_PUL_ (Fig. S15).

Moreover, we detected high sucrose-hydrolyzing activity for PdGH68_DIS_ (Fig. S13a and d), but no production of higher molecular weight compounds.

SusC/D-like complexes of Gammaproteobacteria are mostly associated with strains of marine origin and fructan utilization

The phylogenetic analysis identified 13 additional gammaproteobacterial SusD-like proteins within a huge number of Bacteroidota homologs (1005 in a total of 1055 protein sequences; Fig. S11). Gammaproteobacterial sequences are of sediment/soil, aquatic, and marine origin (Table S3) and are encoded in GH32-containing PULs (data not shown). Two selected gammaproteobacterial strains, P. arctica A 37–1–2 and A. stellipolaris LMG 21856, which harbor fructan PULs encoding a SusC-like TBDT and a SusD-like protein, were shown to grow on inulin. P. arctica was also able to grow on plant-derived levan (Fig. S19).

To further investigate the distribution of SusC/D-like complexes in Gammaproteobacteria, we searched the GEM and UGCMP databases for gammaproteobacterial SusD-like proteins. In a total of 4681 representative Gammaproteobacteria genomes, 32 clusters (31 genomes) encoded SusD-like proteins, out of which 26 were of marine origin (Fig. 5). Most SusD-like protein-coding sequences occurred as SusC/D-like pairs (91%). Out of these SusC/D-like pairs, 83% co-occurred with GH32s, indicating fructan PULs (Fig. 5). The number of PULs containing SusD-like proteins and SusC-like TBDTs might be even higher due to incompleteness of clusters (occasionally fragmented genomes). With respect to the marine catalog UGCMP, apart from Gammaproteobacteria, one PUL was also identified in Marinisomatia (Fig. S20).

SusD-like proteins co-occur with SusC-like TBDTs and GH32s in Gammaproteobacteria. The GEM and UGCMP databases were searched for gammaproteobacterial SusD-like proteins (4681 representative genomes). Shown is their genomic context as well as the presence of additional GH68/GH32 pairs or GH32s.

A total of 631 GH32s were identified in 574 gammaproteobacterial clusters (426 genomes) using the GEM and UGCMP catalogs. 4% of the identified GH32s are encoded in the above-mentioned PULs containing SusD-like proteins and SusC-like TBDTs. 7% co-occurred with TBDTs instead, indicating classical Gammaproteobacteria PULs. 89% did not contain SusD-like proteins, SusC-like TBDTs, or TBDTs in their genomic neighborhood suggesting that many GH32 are not encoded in PULs. A global analysis using the Ocean Gene Atlas v2.0 indicates that the GH32 (96% of bacterial origin) is relevant in oceanic surface waters (Fig. 6a and b), even when compared to the GH16_3 (Fig. 6c and d), which is active on β-glucans like laminarin, a polysaccharide found in high concentrations in marine waters [65]. Using the P. distincta fructan PUL as query for a BLASTN against the same metagenomes supports that marine gammaproteobacterial GH32s co-occur with SusD-like proteins and SusC-like TBDTs (Fig. S21).

Fructan-degrading activity is relevant in marine surface waters around the globe. Global abundance (shown as % of mapped reads) of the (a and b) GH32 compared to the abundance of the (c and d) GH16_3 using the ocean gene atlas v2.0 metagenome datasets (a and c) and metatranscriptome datasets (b and d), Tara oceans microbiomes reference catalog v2 + metaG Arctic inside (prokaryotes) and Tara oceans microbiomes reference catalog v2 + metaT Arctic inside (prokaryotes), respectively.

Proteins that are encoded in different genomic regions mediate the utilization of inulin, levan, or sucrose, but could be linked during downstream processing. Our model for the utilization of fructose-containing substrates in P. distincta is presented. Left: Fructan PUL-encoded proteins enable inulin use. Presumably, the small-sized inulin is shuttled directly into the periplasm by the SusC/D-like transport complex, where it is degraded by the exo-active CAZyme PdGH32PUL, as previously described for human gastrointestinal Bacteroidota [17]. Top center: A distal PdGH68DIS could produce levan as EPS, this levan could be degraded/remodeled by the endo-levanase PdGH32DIS. Fragments released from PdGH32DIS might be further degraded in the periplasm by the fructan PUL-encoded PdGH32PUL, which accepts both linkages. The transporter importing these oligosaccharides is unknown. Right: Proteins encoded in a sucrose PUL may metabolize fructose and sucrose, released upon degradation of inulin and levan. In inulin and levan, fructosyl units are β(2,1)- or β(2,6)-linked to sucrose, respectively. P: phosphate.

Discussion

Here we show that marine Gammaproteobacteria such as P. distincta degrade fructans. Our results further indicate that a SusC/D-like outer membrane transport complex, encoded in a GH32-containing PUL of P. distincta, is involved in fructan uptake. The presence of a glycan-binding SusD-like protein and a SusC-like TBDT in P. distincta is in contrast to the current state of research that, to the best of our knowledge, such transport complexes are restricted to the Bacteroidota phylum.

Like SusD-like proteins in Bacteroidota, the fructan PUL-encoded SusD-like protein (PdSusD_PUL_) of P. distincta contains a lipoprotein signal peptide that directs export and membrane anchoring. In the gammaproteobacterial PdSusD_PUL_, the conserved cysteine is followed by acidic residues (Fig. S22), which signal lipoprotein export to the bacterial surface in Bacteroidota [66]. In addition, the P. distincta SusC/D-like complex is structurally predicted to be highly similar to the levan-binding SusC/D-like complex from B. thetaiotaomicron VPI-5482^T^ [3, 64]. This similarity includes specific loops at the top of the SusC-like TBDT β-barrel and close to the fructan binding site of the SusD-like protein that mediate interaction between the two proteins (Fig. 3). In comparison, the gammaproteobacterial SusC-like TBDTs identified in this study lack the N-terminal carboxypeptidase D regulatory-like domain. The exact function of the carboxypeptidase D regulatory-like domain preceding the plug domain is unknown, but many SusC-like TBDTs of B. thetaiotaomicron VPI-5482^T^ possess it [67]. This includes the levan-specific protein (BT_1763) where it was shown that the deletion of the carboxypeptidase D regulatory-like domain causes a growth defect [64]. It was speculated that it might mediate specificity for the various TonB orthologs. With one exception (EGI71291), all the other TBDTs of P. distincta also lack the carboxypeptidase D regulatory-like domain, suggesting a functional complex despite the missing domain.

The phylogenetic analysis of the SusD-like protein (Fig. S11) indicates the possibility of horizontal gene transfer. The transfer of genes or operons related to polysaccharide use has been described before, from marine Bacteroidota to Bacteroidota of the human intestine [68, 69], but also from a marine flavobacterium (Bacteroidota) to Pseudomonadota, including Gammaproteobacteria [70]. However, in the latter case, the transfer of Bacteroidota alginolytic operons to Gammaproteobacteria did not include the SusC/D-like complex. It is therefore intriguing how the fructan-related complex was able to establish in Gammaproteobacteria, whereas SusC/D-like complexes with other specificities did not – or not yet.

The detected affinity of the P. distincta PdSusD_PUL_ (~K_a_ 43 M^−1^) to the inulin-derived oligosaccharide FOS_INU_ was much lower compared to the affinity of the SusD-like protein of B. thetaiotaomicron strain Bt-8736 (the K_d_ was reported as ~0.5 mM, which corresponds to K_a_ 2 × 10^3^ M^−1^) to inulin [17]. A higher affinity binding, which in the case of PdSusD_PUL_ may also demonstrate binding to inulin, may require a functional complex with the SusC-like TBDT, as already speculated for a non-binding SusD-like protein encoded in a xylo-oligosaccharide PUL [71].

In human gastrointestinal Bacteroidota, it has been shown that the specificity toward levan or inulin is mediated by SusC/D-like complexes and by the subcellular localization of GH32s [17].

Like PdSusD_PUL_, the P. distincta fructan PUL-encoded CAZyme PdGH32_PUL_ contains a lipoprotein signal peptide. However, the conserved cysteine in PdGH32_PUL_ is not followed by acidic residues (Fig. S22), which could indicate that the CAZyme is not exposed to the surface. Together with its high relative protein abundance in the outer membrane and outer membrane vesicles (Fig. S7, Supplementary Dataset 2), this suggests that PdGH32_PUL_ is anchored to the outer membrane, with a possible orientation toward the periplasm. In this scenario, the small-sized inulin might enter via the SusC/D-like complex and be degraded by PdGH32_PUL_ (Fig. 7). Similarly, inulin-degrading Bacteroidota from the human intestine lack extracellular GH32s or extracellular CAZymes are not necessary for growth [17, 72]. In contrast, growth of P. distincta on levans may be enabled by the extracellular levan-specific PdGH32_DIS_, which is not encoded in a PUL. Bacterial levan has a high molecular weight (>1000 kDa [73]) and is branched [74] compared to plant-derived levan (~12.5 kDa), which could explain differences seen in growth on levans. Genetic manipulation of P. distincta to knock out the SusD-like protein/SusC-like TBDT and the GH32s, or cryo-electron microscopy as performed for the levan-specific SusC/D-like complex of B. thetaiotaomicron [75], could further confirm the suspected model for the utilization of inulin and levan by P. distincta.

In addition to fructan degradation, the GH68s of P. distincta, Pseudoalteromonas carrageenovora 9^T^ [10], and other marine Gammaproteobacteria (Fig. 5, Fig. S20) indicate fructan production [6]. Although we could not confirm that PdGH68_DIS_ produces higher molecular weight FOS or fructans, GH68s with transfructosylation activities [76, 77] were identified as closest PdGH68_DIS_ homologs using DALI [78]: GH68 of Microbacterium saccharophilum K-1 (formerly Arthrobacter, z-score 66.2, identity 77%, PDB ID 3VSS [79]) and GH68 of Gluconacetobacter diazotrophicus SRT4 (formerly Acetobacter diazotrophicus, z-score 61.3, identity 56%, PDB ID 1W18 [80]). The production of fructans and fructose-containing EPS by marine bacteria has been reported before [7, 8], which indicates fructose-based EPS as one source for marine fructans.

The distribution and abundance of GH32s suggests that fructans are important polysaccharides in marine ecosystems (Fig. 6a and b), although the frequency of fructan PULs detected in our analysis of gammaproteobacterial genomes and in particle-associated bacterial metagenome-assembled genomes from phytoplankton blooms [22] is rather low. However, GH32s often occur without a PUL context.

GH32-containing PULs may have the advantage that the expression is fine-tuned by regulatory elements. Overall, it remains speculative whether the only benefit of acquiring a fructan PUL from a putative Bacteroidota ancestor was an extended substrate repertoire or whether the SusC/D-like tandem came with advantageous aspects. For instance, for P. gingivalis, a Bacteroidota pathogen associated with human periodontitis, it was suggested that the peptide-binding RagB-like protein (corresponding to SusD-like protein) mediates specificity [61]. In addition, a SusD-like protein might be beneficial when endo-active extracellular inulinases and CBM-containing non-catalytic proteins are missing, as in P. distincta.

Our findings thus show that fructans contribute to the marine glycan universe targeted by specialized Gammaproteobacteria, such as P. distincta, which have PULs encoding GH32s, SusD-like proteins and SusC-like TBDTs.

Supplementary Material

Supplementary_materials_wrag030