Insights into the utilisation of 1,2-propanediol and interactions with the cell envelope of Clostridium perfringens

C. perfringens is a frequent inhabitant of the infant gut especially during the breastfeeding phase [11], yet few isolates have been characterised in detail. C. perfringens FMT 1006 has been isolated from infant feces [17]. To determine the genetic potential of FMT 1006, we isolated DNA and performed genome analysis combining nanopore (Oxford Nanopore Technologies) and MiSeq sequencing (Illumina). The genome was compared to other strains with closed genomes that were obtained from animals, food and the environment (Table 1). Genomes with ANI values above 95% are considered as the same species [18]. Based on our analysis using EDGAR 3.0 [19], the C. perfringens strains used in this study were closely related (ANI 96–99%) (Fig. S1). The pangenome encompassed 4117 genes, while the core genome consisted of 2033 genes. There were 982 singletons, and 1100 dispensable genes present in a subset of the strains.

The genome of FMT 1006 had a size of 3.2 Mbp and a GC content of 28.5% similar to the other strains (3.0-3.6 Mbp) (Table 1). FMT 1006 harboured 2774 coding DNA sequences (CDS) and a total of 124 RNA components (Table 1). FMT 1006 possessed a total of 100 unique CDS compared to the other six strains (Fig S1A). KEGG identified 1544 proteins with e-value cut-off of 0.01. All genomes possessed the plc gene encoding the alpha toxin Plc (phospholipase C).

Annotations of all genomes were performed with Prokka and refined with InterProScan [20, 21], which confirmed the presence of pduCDE genes (also sometimes annotated as dhaB) in C. perfringens FMT 1006 and the other strains encoding the three subunits of glycerol/diol dehydratases (Fig. 1). pduCDE (T_01016-T_01018) and dhaT (T_01024), encoding the 1,3-propanediol oxidoreductase (DhaT), which has been shown previously to perform alcohol dehydrogenase (PduQ) activity [22], were located in cluster with genes related to cobalamin turnover pduGH (T_01019-T_01020) and pduO (T_01021) (cluster_pdu). Other potential pdu related genes such as pduL (T_00814), pduU (T_00802) and pduV (T_00803) were found within a gene cluster that mainly contained genes related to ethanolamine utilisation (Fig. 1) (cluster_eut). The gene encoding protein AdhE_1 (T_00801) that was identified as PduQ by InterProScan, was in proximity to pduU and pduV (Fig. 1). A candidate gene pduW (T_01799) was located at a different location on the genome (cluster_pduW). When characterised with additional BLASTp analysis, PduW was 100% identical to PduW of other C. perfringens strains, and had 71% identity with PduW from Clostridium baratii (Suppl. Table 1).

We conducted additional BLASTp (v. 2.11.0+) analysis using previously functionally characterised PduP of Salmonella (AAD39015.1) [23] to identify possible homologues of PduP. No homologue of PduP was present in the genome of C. perfringens, the closest hits were AdhE_2 (T_00811) with 29% identity and another acetaldehyde dehydrogenase (T_02661) with 31% identity. Taken together, theoretical homologues of enzymes related to 1,2PD utilisation were identified with identity values above > 40% except for PduP (Suppl. Table 1).

Similar to FMT 1006, the genomes of the other assessed C. perfringens strains possessed most pdu genes except pduP conferring the overall potential of utilizing 1,2PD (Fig. 1). Genes were located on two clusters with the exception of pduW (T_01799), and gene syntety of pdu and eut was conserved except for SM101, which lacked the eut cluster (Fig. 1). In Limosilactobacillus reuteri, Salmonella or Listeria, the pdu genes were encoded in one operon that also included genes related to cobalamin synthesis and to ethanolamine utilisation [24‐26] (Fig. 1). The pdu operon of Propionibacterium freudenreichii DSM 20271 was organised similar to C. perfringens with pdu present in two different loci [7] indicating that organisation of pdu differs between species.

As C. perfringens FMT 1006 possessed the general ability for 1,2PD utilisation based on its genome sequence but lacked a gene encoding PduP, which catalyses the formation of propionate, we evaluated the growth and fermentative activity in anaerobically prepared yeast extract casitone (YC) medium supplied with 50 mM glucose (YC-G50), 50 mM of 1,2PD (YC-PD50) or without additional carbon source (YC) at 37 °C for 24 h in serum flasks. The concentration of 1,2PD was chosen to reflect concentration ranges in feces [2, 6]. We measured optical density (OD₆₀₀) and determined substrate utilisation and metabolite formation with high performance liquid chromatography with refractive index detector (HPLC-RI).

In YC-G50, C. perfringens reached highest density (1.98 OD₆₀₀) at 24 h (Fig. 2A). At 6 h, C. perfringens had a growth rate of 0.28 OD₆₀₀/h and a doubling time of 2.44 h in YC-G50. C. perfringens reached similar OD₆₀₀ values at 24 h when incubated in YC and in YC-PD50 (0.30 and 0.25 OD₆₀₀, respectively) (Fig. 2A). Low optical density was also observed when A. hallii, B. obeum, L. reuteri or R. gnavus were grown in a similar medium with 1,2PD as added carbon source [10] suggesting that 1,2PD metabolism did not contribute to biomass formation.

Fermentative activity of C. perfringens was the highest in YC-G50. After 24 h, 42.5 ± 5.4 mM lactate was formed from 24 ± 0.9 mM glucose, followed by 6.8 ± 1.2 mM butyrate and 3.3 ± 1.6 mM acetate (Fig. 2B, D, E and F). In YC-PD50, 33.3 ± 5.8 mM 1,2PD (78.6%) were consumed and converted to 7.3 ± 0.2 mM 1-propanol and 1.8 ± 0.2 mM propanal without detectable levels of propionate (Fig. 2C, G, H and I). Propanal is a toxic compound and to prevent intracellular damage, BMC are formed during the utilisation of 1,2PD [27]. Acetate, lactate, butyrate, were present in low amounts (below 2 mM) (Fig. 2D, E and F). 1,2PD, 1-propanol, propanal or propionate were not detected when C. perfringens was incubated in YC (Fig. 2C, G, H and I). When A. hallii was tested at similar conditions, the carbon provided in form of 1,2PD was fully recovered as propanol, propanal and propionate (C_{1,2PD used}=C_propanol+C_propanal+C_propionate) [28], while in this study, the proportion was only approx. 34% (C_{1,2PD used} > C_propanol+C_propanal).

As we could not recover the expected metabolites from the liquid phase, we examined levels of 1-propanol, propanal and other potential metabolites in the headspace using proton-transfer-reaction mass spectrometry (PTR-TOF-MS) and estimated the concentrations in liquid media according to Henry’s law assuming air-water equilibrium. The detection of 1-propanol and propanal was confirmed by gas chromatography-MS (GC-MS). Smaller amounts of 2-ethyl-2-butenal and 2-methyl-pentanal were detected by both PTR-MS and GC/MS. Propanal levels were estimated 4.4-times higher based on headspace measurement than levels determined with HPLC-RI, and estimates for 1-propanol were 1.6-fold higher (Suppl. Table S2). The carbon C_1,2PD that could be recovered as C_propanol+C_propanal was approx. 60%. As we observed a peak in HPLC-RI profiles at the retention time of 1,3-propanediol (1,3PD), we suggest that additional aldehyde hydration events occur in the liquid medium, which might account for the discrepancy between measured and estimated levels.

Taken together, our data show that C. perfringens metabolised 1,2PD to 1-propanol with a C_1,2PD recovery of 33–60% depending on analysis of compounds present in the liquid broth or the headspace.

Based on cultivation studies and genome characterisation, C. perfringens FMT 1006 did not form propionate when incubated in YC-PD50 and did not possess any genes related to propionate formation via PduP. We next determined gene expression profiles of C. perfringens when grown in YC-G50, YC-PD50 and YC. RNA was collected during the exponential phase of growth in YC-G50, while samples for RNA profiling of YC and YC-PD50 were obtained when the OD₆₀₀ increased after 3 and 4 h incubation, respectively. Custom gene expression microarrays were synthesised using maskless DNA photolithography [29‐31] and were hybridised using RNA transcribed to fluorescently labelled cDNA.

In YC-PD50, genes of the pdu cluster, i.e. pduCDE, pduGH, pduO, dhaG and dhaT had higher fluorescence signal (p < 0.05) compared with YC-G50 or YC (Table 2). At the same time, genes of the eut cluster_eut (eutA, eutB, eutM and pduL) had a lower signal (p < 0.05) in YC-PD50 compared to YC or YC-G50 (Table 2). pduW was not differently expressed. In YC-G50, genes related to glucose utilisation (acetyl-coA transferase thlA, L-lactate dehydrogenase ldh) or electron transfer (e.g. flavodoxin T_00556 or NADH oxidoreductase T_00907) were significantly (p < 0.05) upregulated compared to YC-PD50 (Table 2).

We conducted additional quantitative PCR (qPCR) analysis to determine expression levels of pduC, pduD, dhaT and pduL relative to housekeeping genes ftsZ, recA and gyrA. In agreement with microarray analysis, pduC, pduD and dhaT were 0.06–0.09 fold downregulated in YC relative to YC-G50, while the same genes were upregulated 2.7–5.6 fold in YC-PD50 relative to YC-G50 (Table 3). pduL was 2.6-fold higher expressed in YC relatively to YC-G50 (Table 3).

As we observed that the addition of 1,2PD did not support growth of FMT 1006 and that the intermediate propanal was detected after 24 h of incubation, we determined if glucose addition improved biomass formation and the metabolism of 1,2PD when supplied at different concentrations. We combined glucose and 1,2PD at equimolar concentrations (YC-PD10-G10 and YC-PD50-G50) and compared to YC-PD50. FMT 1006 was cultured for 24 h at 37 °C, and turbidity was determined using a McFarland densitometer. HPLC-RI was used to determine substrate utilisation and metabolite formation.

Final turbidity was low in YC-PD10-G10 (1.9 ± 1.4 McFarland units, MF), YC (1 ± 0.4 MF), YC-PD10 (0.7 ± 0.2 MF) and YC-PD50 (1.4 ± 1 MF). When glucose was supplemented to YC-PD50 (YC-PD50-G50, 9 ± 1 MF), turbidity was similar to YC-G50 (10.7 ± 0.6 MF) (Fig. 3A). Compared to cultivation in YC-G50, the presence of 50 mM 1,2PD (YC-PD50-G50) shifted the ratio of acetate:lactate from 1.0:9.0 to 1.0:2.6 (Fig. 3B). Low concentrations of 1,2PD (YC-PD10) led to higher yields of propanal from 1,2PD (60.2% mM_propanal/mM_1,2PD) compared to YC-PD50 (8.3% mM_propanal/mM_1,2PD). In contrast, the proportion of 1-propanol was similar in YC-PD10 and YC-PD50 (21.0% mM_propanol/mM_1,2PD) (Fig. 3C). When glucose was present, the proportion of propanal and 1-propanol was 12% and 88% and 27.1 and 70.1% in YC-PD10-G10 and YC-PD50-G50, respectively (Fig. 3C).

Bild vergrößern

Taken together, we observed that propanal production did not impact final turbidity (p > 0.05) when glucose was provided together with 1,2PD even at 50 mM, suggesting that the low biomass observed in the absence of glucose may not be due to propanal production. Within the Pdu pathway, conversion of propanal by PduQ and PduP requires NAD⁺/NADH as co-factors (Fig. 4). PduQ uses NADH as electron donor, while PduP employs NAD⁺ as electron acceptor [32, 33]. In previous studies, Salmonella formed 1-propanol and propionate in a ratio of 1:1 [34, 35]. In contrast, C. perfringens converted 1,2-PD to 1-propanol without concomitant production of propionate. In our genome-based pathway model, C. perfringens only possessed DhaT producing NAD⁺ and lacked NADH-generating PduP (Fig. 4). Lactate production was highest during growth in YC-G50, and lower lactate production in the presence of both glucose and 1,2PD may be associated with DhaT activity, which used NADH as electron donor instead of lactate dehydrogenase Ldh (Fig. 4). Similarly A. hallii produced more 1-propanol (and not propionate) when grown in the presence of 1,2PD and glucose, compared to 1,2PD alone [28] suggesting that A. hallii also regenerated NAD⁺ pool through activity of PduQ.

To investigate further whether 1-propanol formation depended on the concurrent formation of acetate and/or lactate, we evaluated the impact of addition of 1,2PD (50 mM) to cultures grown in YC-G50 at early exponential (3 h), late exponential (6 h) and during stationary growth phase (12 h). After 24 h growth, highest turbidity was achieved when 1,2PD was added at 3 h (10.6 ± 0.6 MF), followed by 12 h (9.4 ± 0.2 MF) and 6 h (7.5 ± 0.2 MF) (Fig. 3D). Turbidity recorded at 3 h before injection (3.4 ± 0.0 MF) was lower (p < 0.05) compared to 24 h (Fig. 3D). Consumption of glucose (19.6–23.6 mM) and 1,2PD (48.4–63.2 mM) was similar at all three conditions, suggesting that substrate utilisation was not affected by the concurrent presence of glucose and 1,2PD (Fig. 3E and F). These observations are in contrast to Clostridium beijerinckii or C. butylicum, whose Pdu metabolism was repressed when glucose was provided [36, 37].

When 1,2PD was added after 3 or 6 h, levels of acetate and lactate were similar, and the ratio of acetate: lactate was 0.6:1.0 compared to 2.9:1.0 at 12 h (Fig. 3E). When 1,2PD was added at 3 h, the recovery of C_1,2PD used as C_propanol+C_propanal was 87% with 35.9 ± 3.0 mM 1-propanol and 19.3 ± 4.7 mM propanal produced (Fig. 3F). At 6 and 12 h, the recovery of C_{1,2PD used} as C_propanol+C_propanal was lower (57% and 29%, respectively) (Fig. 3F). Conversion to 1-propanol was higher (p < 0.05) when 1,2PD was added during exponential growth compared to stationary phase.

These differences in recovery of 1,2PD metabolites again suggest that the conversion of pyruvate to acetate or lactate plays a role in the production of 1-propanol. In the presence of glucose, acetate formation can deliver ATP (Fig. 4), and when pyruvate is transformed to acetate, there is less lactate formed, and NADH consumed which can provide NADH to PduQ (Fig. 4). In infant feces, C. perfringens likely has access to both glucose and 1,2PD. In agreement, 1-propanol was recovered in 54-63% of fecal samples of breastfed infants that were collected between 2 weeks and 2 months of life [38, 39]. 1-propanol was detected in feces of breastfed infants as early as eight days after birth [39], in agreement with early colonisation by Clostridium spp. [40]. While C. perfringens might not be able to contribute to the intestinal propionate pool through 1,2PD cross-feeding, it can provide 1-propanol to the gut intestinal environment.

At the provided cultivation conditions, C. perfringens FMT 1006 growth was not supported by the presence of 1,2PD. As an alcohol, 1,2PD could increase in membrane fluidity, leading to leakage of cofactors and even a collapse of metabolic activity [41]. Ethanol and butanol have been shown to decrease proton flux across the membrane, to alter ATPase activity and to reduce energy production [41]. Therefore, we investigated the effect of 1,2PD on the state of the membrane using the stain Laurdan (1-(6-(dimethylamino)naphthalen-2-yl)dodecan-1-one). Due to its apolarity, Laurdan can dissolve within the lipidic membrane, shifting emission spectra from 440 nm (gel-phase) to 490 nm (liquid-phase). We used phenylethanol (PE) (10 and 50 mM) as positive control as PE increases fluidity [42], and tested 1,2PD (10 and 50 mM) and the pathway intermediate propanal (20 and 50 mM).

Addition of PE significantly increased (p < 0.05) membrane fluidity in a dose-dependent manner compared to cells that were not treated as indicated by a lower Generalised Polarisation (GP) value. In comparison, 1,2PD increased membrane fluidity less than 0.9-fold when added at 50 mM, and reduced membrane fluidity by ∆GP 1.05 at 10 mM (Fig. 5A). The aldehyde propanal did not affect overall membrane fluidity. Huffer et al. [41] assessed the effect of different alcohols on bacterial membranes, and concluded that strains whose membrane fluidity did not change significantly were more tolerant towards alcohol exposure. These tolerant strains included C. beijerinckii [41]. As there was only a slight effect on membrane fluidity even at a concentration of 50 mM, our results point out that the C. perfringens FMT 1006 membrane was tolerant towards exposure to 1,2PD at levels that might be similar as observed in vivo [2, 6].

Exposure to alcohols might not only affect membrane fluidity but can also cause alterations in lipid composition. We therefore evaluated membrane composition of C. perfringens FMT 1006 after 24 h incubation in YC-PD50, YC and YC-G50. We extracted the lipidic fraction using the SIMPLEX protocol [43] and determined lipid composition with liquid chromatography–tandem mass spectrometry (LC-MS/MS). For data extraction, peak identification and normalisation we used MS-DIAL [44] and further analysed the results with RStudio.

Membrane lipids were mainly composed of diacylglycerols (DAG), monoacylglycerols (MAG), triacyclglycerols (TAG), and glycerophosphoethanolamines (GPE) (Fig. 5B). During growth in YC-G50, the most abundant lipids were DAG (33%), TAG (36%) and GPE (30%). When cultured in YC and YC-PD50, the proportion of GPE (10% and 8%) was lower compared to YC-G50, while the proportions of TAG (54% and 38%) and DAG (36% and 52%) were higher (Fig. 5B). There was little difference in the composition of the MAG and DAG fractions when C. perfringens was incubated in YC-G50 or YC-PD50 (data not shown). In contrast, the composition of GPE lipids, which are major components of the lipidic membranes of gram positive bacteria [45], differed (Fig. 5C).

There were little difference in the composition of the MAG and DAG fractions when incubated in YC-G50 and YC-PD50 (data not shown) while the composition of GPE lipids shifted (Fig. 5C). The membrane of C. perfringens FMT 1006 had a higher abundance of shorter GPE when cultured in YC-G50 compared to YC-PD50 (Fig. 5C) suggesting a more pronounced membrane rigidification due to the presence of 1,2PD (Fig. 5C) [45].

As we observed changes in membrane lipid composition, we investigated expression of genes of the fatty acid biosynthesis cluster fab, which has been linked to membrane lipid modification during stress conditions in Bacillus and Clostridium sp [46, 47]. Based on microarray data, expression levels of fab related genes did not significantly differ between growth in YC-PD50 compared to YC or YC-G50 (Suppl. Table S3).

Materials were purchased from Merck unless stated otherwise. C. perfringens FMT 1006 was obtained from the strain collection of the group of Functional Microbe Technology (FMT) (Aarhus University, Denmark) and was originally isolated from infant feces [17]. FMT 1006 was routinely cultivated in Wilkins-Chalgren medium supplemented with 5 g/L soya peptone (WCSP, Thermo Fisher), 1 g/L Tween 80, and 0.5 g/L L-cysteine-HCl. The pH was adjusted to 7.1 before boiling, and L-cysteine-HCl was added before degassing for a final pH of 6.5 after autoclaving. Samples were dispersed in serum flasks (30 mL) or Hungate tubes (10 mL) under constant CO₂ flow, closed and autoclaved at 120 °C for 20 min. For initiation of working cultures, cells were streaked from frozen stocks on anaerobically prepared WCSP agar plates (WSCP with the addition of 15 g/L agar (VWR)) in an anaerobic chamber (10% CO₂, 5% H₂, 85% N₂, Baker Ruskinn) and were incubated at 37 °C for 24 h.

Colonies were transferred to a 10 mL Hungate tube containing anaerobically prepared, modified yeast extract-casitone (YC) medium as described by Huertas-Díaz et al. [2]. supplied with 50 mM glucose (YC-G50). After two transfers in YC-G50, cultures were used in the assays. Strain purity was routinely checked using bright field microscopy.

To determine growth behaviour and to collect samples for RNA isolation and analysis of membrane lipids, working cultures were inoculated (2%) in YC in serum flasks in YC, YC-G50 and, YC-PD50. Growth was monitored as optical density at 600 nm (OD₆₀₀) using a cell densitometer (Ultrospec 10, Biochrom). When incubated with glucose, C. perfringens showed regular growth behaviour with a lag phase, exponential and stationary growth phase, while during incubation in YC and YC-PD50, there was only a minor increase in turbidity at around 3–4 h. We therefore collected cell pellets of cultures grown in YC-G50 at early-stage growth phase (0.2 OD₆₀₀), for isolation of RNA. Samples for RNA profiling of YC and YC-PD50 were obtained when the OD₆₀₀ increased from 0.14 to 0.20 and from 0.16 to 0.30 after 3 and 4 h incubation, respectively. Cell pellets were snap frozen in liquid nitrogen and placed at − 80 °C until further processing. All experimental work was performed in independent biological replicates (n = 2–7). Cell pellets were collected from the remaining biomass after 24 h incubation for determination of the composition of membrane lipids and stored at − 80 °C.

To test for the impact of glucose on 1,2PD metabolism, working cultures were inoculated (2%) in Hungate tubes containing YC-G50, YC, YC-PD10, YC-PD10-G10, YC-PD50, YC-PD50-G50. Additionally, we investigated how 1,2PD addition affected the fermentation activity of C. perfringens grown in YC-G50 and added 1,2PD after 3, 6 and 12 h of incubation. Turbidity was measured prior addition of 1,2PD and after 24 h incubation. Growth rate in YC-G50 was calculated at 6 h of growth as 1/(duration of incubation * ln [2]) / ln (OD₆₀₀/initial OD₆₀₀).

DNA was extracted from working cultures following the instructions from manufacturer (Genomic DNA purification kit GeneJET, Thermo Scientific for Gram-positive bacteria). Purified DNA was suspended in 40 µL of TE buffer. Libraries were prepared and sequenced using MiSeq (Department of Biology, Aarhus University) and Oxford Nanopore Technologies (Plasmidosaurus) and standard library preparation protocols. For quality filtering and assembly, the following process was used: FitLong (v0.2.1) was used to remove 5% of the lowest quality reads and to reduce the reads to 250 Mb [48]. A rough assembly was created with Miniasm (v0.3) which reduced to 100× coverage [49]. Flye (v2.9.1) assembly was run with parameters selected for high quality ONT reads [50] and Medaka (v1.8.0) was used to refine the Flye assembly [51]. Contig analysis was performed using Bandage (v0.8.1), quality check and contamination was assessed by CheckM (v1.2.2) [52, 53]. The assembly was refined using Polypolish (v0.5.0) with the Illumina reads previously trimmed with Trimmomatic (v0.39) and quality checked with FastQC (v0.11.9) [54‐56]. Annotations were done with Prokka (v1.14.5) [20] and InterProScan (v5.60-92.0) was used for annotation refinements [21]. Metabolic pathways were assessed with KofamKOALA (v2024.02.01) [57]. Annotations were evaluated using Artemis software (v18.2.0) [58]. Additional BLASTp was performed with selected Pdu proteins of FMT 1006 against the NCBI using the database (non-redundant protein sequences). For comparative genome analysis of C. perfringens FMT 1006 and the complete genomes of six other strains of C. perfringens (Table 1), EDGAR3.0 was used [19] to determine the core genome ANI values based on all-against-all comparisons [18, 19].

To isolated RNA, we followed TRizol procedure. TRizol (1 mL) was added to the pellets and incubated at room temperature for 5 min. Chloroform (0.2 mL) was added, mixed by shaking, and the sample was incubated for 3 min, followed by centrifugation for 15 min at 12 000 × g at 4 °C. The aqueous phase was transferred into a new 1.5 mL tube and mixed with an equal volume of 70% ethanol. The sample was transferred to a RNeasy spin column and processed according to the manufacturer’s instruction (RNeasy Mini Kit, Qiagen). Chromosomal DNA was removed using the TURBO DNA-free™ Kit (Thermo Fisher) and the absence of DNA was confirmed using quantitative PCR (qPCR) with primers targeting the 16 S rRNA gene as described [59]. A total of 10 µg of RNA were simultaneously reverse transcribed and labelled using 7 µg of 5′ Cy3 random nonamer (Biomers) as previously described [30]. Briefly, the mixture was diluted in nuclease free water and incubated for 70 °C for 5 min and placed immediately on ice. For reverse transcription, a mixture of 6 µL of 5X first strand buffer, 1.5 µL of 0.1 M DTT, 1 µL of 10 mM dNTP, 4 µL of Superscript 3 (200 U/µL, Thermo Fisher) and 1 µL of RNase out (40 U/µL, Thermo Fisher) was added to the previous mixture and incubated for 5 mi at 25 °C followed by 3 h at 42 °C. For hydrolysis, a mixture of 16 µL of 200 mM NaOH and 16 µL of 20 mM of EDTA were added to the mixture and incubated for 10 min at 65 °C. For the neutralisation step, 64 µL of 1 M HEPES (pH 7) was added to the previous mixture. For purification, QiaQuick PCR purification kit was used according to instructions.

Probes for the C. perfringens were designed using OligoPicker software [60] with default settings. For microarray design a total of 2618 unique oligo sets out of total 2771 genes. Each gene was represented for an average of 2.5 different probes of 70mer length. A 4-part chamber array design was used with 35.000 spots from which each probe would be represented approximately 5 times and each gene 12.5 times. 60 probes were added with ‘QC 25mer’ as an internal quality control and 10 empty sequences for background subtraction. For gene expression analysis, 0.125 µg of Cy3-labelled cDNA were used per chamber in a 4-part array. Data extraction was performed using NimbleScan 2.1 (Roche-NimbleGen), employing robust multichip analysis for inter-array normalization purposes. All experiments were performed in biological and technical duplicates. Data normalisation was performed using robust multichip analysis (RMA) method followed by quality control of QC25 k-mers [61]. Background probes were added to all parts of microarray and background was subtracted after normalisation. For final data comparison, DESeq2 analysis was used [62]. Further information is provided in supplementary methods. qPCR was performed to determine relative expression levels of major genes of pdu as described in the supplementary methods [59].

To investigate changes in the fluidity of the cell membrane, a Laurdan staining method was adapted taking into account the requirements for anaerobic conditions of FMT 1006. All procedures were performed in an anaerobic chamber and all chemicals and materials were pre-heated to 37 °C to avoid membrane-associated thermal changes. Laurdan stock (1 mM) in acetone was kept at -20 °C.

Samples (0.5 mL) were withdrawn from a 24 h culture in YC-G50 and mixed with PBS buffer (pH 7.4) to a final density of 0.5 OD₆₀₀. The cell pellet was washed with phosphate buffered saline (PBS) twice at 10,000 × g for 30 s. Samples were resuspended in PBS containing 50 mM 1,2PD (PD50), 10 mM 1,2PD (PD10), 50 mM PE, 50PE, 10 mM PE (10PE), 50 mM propanal, 20 mM propanal, or PBS only as negative control together with Laurdan (10 µM). Samples were incubated for 30 min at 37 °C.

After incubation, cells were washed twice with PBS buffer and pellets were resuspended in the same treatment solutions without Laurdan. Cell suspensions (200 µL) were added to a 96-well black clear bottom plate and fluorescent intensity was determined using a fluorescence microplate reader (CLARIOstart Plus, BMG LABTECH) at excitation 350 nm and emission 380 nm to 600 nm at 37 °C. The experiment was repeated three times, and each sample was analysed in technical duplicates.

The Generalised Polarisation (GP) values were calculated using fluorescent emission intensity values at 440 nm and 490 nm as follows: GP=(I₄₄₀-I₄₉₀)/(I₄₄₀ + I₄₉₀).

To extract membrane lipids, we followed the SIMPLEX method for lipidomics as described [43] followed by LC-MS/MS for identification. Eluent A was prepared with MeOH/ACN/H2O (1:1:3), 5 mM ammonium acetate, 10 nM EDTA. Eluent B contained isopropanol with ammonium acetate (5 mM) and EDTA (10 nM). A 2 µL injection under a flow rate of 200 µL/min. Using a column Kinetex 1.7 μm C18 100 Å size 100 × 2.1 mm. Instrument control and data processing software was performed by Analyst (TF 1.7.1). Data processing was performed using MS Dial for lipidomics. Only proton adduct was analysed by MSDial [44]. Data was extracted after normalisation and analyzed by DESeq2 package in R. Analysis performed from biological duplicates. Further information in Supplementary materials and methods.

Substrate levels and fermentation metabolites present in supernatants were analysed by high performance liquid chromatography coupled to a refractive index detector (HPLC-RI) using a 1260 Infinity II LC System (all Agilent). Compounds were separated using a Hi-Plex H column (300 × 7.7 mm) attached to a guard (50 × 7.7 mm) column. The samples (10 µL injection volume) were eluted with 5 mM H₂SO₄ at a flow rate of 0.6 mL/min at 40 °C. Fermentation metabolites were quantified using external standards.

PTR-TOF-MS measurements were performed as previously reported [63]. The system was operated with H₃O⁺ as the primary ion using the proton affinity to detect compounds by direct time-of-flight mass spectrometry [63]. Settings were 2.4 mbar, voltage 600 V and a temperature 80 C for an E/N of 140 Td The can mode ranged from m/z 5 to 200 and the airflow was set to 400 mL/min After calculating of concentrations within the headspace, Henry’s law was applied to calculate equilibrium liquid concentrations [63]. In addition, headspace gas phase sampling was done on adsorbent tubes coupled to thermal-desorption gas chromatography with mass spectrometric detection (TDGCMS) to verify identification of unknowns. The method has been described in more detail before [64].

DESeq2 analysis was used both for microarray analysis and for lipidomics comparison [62]. Significance was calculated by Wald test. For statistical analysis of turbidity, we employed Kruskal-Wallis test with posthoc Dunn’s test adjusted by Holm using the package ‘rstatix’ v.0.7.2. Figures were made with RStudio and posterior modifications by Adobe Acrobat Pro, Excel or PowerPoint.