Disruption of dnaK results in transcriptional reprogramming of biofilm-associated pathways in Clostridioides difficile

Abstract

Introduction. Clostridioides difficile is an opportunistic enteric pathogen posing significant challenges in human and animal healthcare due to the recurrent nature of infections and associated financial burden. The molecular chaperone protein, DnaK (Hsp70), encoded by dnaK, plays a central role in the heat stress response and protein homeostasis. Disruption of dnaK in C. difficile 630Δerm has previously been associated with increased biofilm formation and multiple phenotypic alterations.

Gap Statement. Although disruption of dnaK has been shown to alter biofilm formation and multiple phenotypes in C. difficile, the global transcriptional mechanisms linking DnaK to biofilm development and stress adaptation remain poorly understood.

Aim. This study aimed to characterize the global transcriptional response associated with dnaK disruption during planktonic and biofilm growth to identify pathways linking the DnaK-mediated stress response with biofilm formation.

Methods. This study investigated the global transcriptional changes associated with dnaK disruption during planktonic and colony biofilm culture. Colony biofilms of C. difficile strains 630, 630Δerm and 630Δerm:dnaK were cultured on mixed nitrocellulose ester semi-permeable membranes and examined structurally using low-vacuum scanning electron microscopy. Total RNA was extracted from both planktonic cultures and 24 h colony biofilms, followed by Illumina RNA sequencing. Differential gene expression analysis was performed using DESeq (P_adj<0.05). Gene set enrichment analysis (Kyoto Encyclopedia of Genes and Genomes [KEGG] and Gene Ontology) was conducted to identify significantly enriched metabolic pathways and biological processes.

Results. Disruption of dnaK resulted in global transcriptomic remodelling, particularly during biofilm culture, where 72.4% of genes were significantly differentially expressed relative to planktonic culture. A core number of pathways were involved in biofilm formation in all strains, including increased sporulation with reduced translation, amino acid biosynthesis and flagellar assembly. The dnaK mutant exhibited a unique pattern of enrichment of pathways associated with cell division, peptidoglycan synthesis, carbon metabolism and amino acid biosynthesis, alongside a coordinated reduction of motility and chemotaxis pathways.

Conclusion. Disruption of the dnaK gene appears to link the stress response to biofilm formation through a number of pathways associated with biofilm development. Understanding this link may identify systems relevant to persistent and recurrent infections.

Data Summary

RNA sequencing data generated in this work have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) repository, accession number GSE336654. The authors confirm all supporting data, code and protocols have been provided within the article or through supplementary data files.

Introduction

The opportunistic enteric pathogen Clostridioides difficile causes significant healthcare problems due to the recurrent nature of infections, becoming increasingly difficult to treat with each recurrence [1, 2]. The limited number of approved treatments for C. difficile infection (CDI) are threatened by the rise in antimicrobial resistance [3]. New treatment options are emerging, such as spore therapy, bezlotoxumab monoclonal antibody therapy and a glycopeptide antibiotic EVG7, which is a semisynthetic vancomycin derivative [4–6]. However, no single treatment has been successful in inhibiting disease recurrence in all cases; therefore, further understanding of the mechanism of recurrence is still necessary. Recurrence of infection has been attributed to the ability of C. difficile to form biofilms and spores within the gut [1, 2, 7]. Spores are important for the zoonotic and faecal–oral transmission and survival of C. difficile when exposed to oxygen, heat or disinfecting agents, allowing them to persist on hospital equipment [8–11]. Sporulation has been strongly associated with biofilm formation in C. difficile through the master regulator Spo0A [8]. However, the regulatory mechanisms of both processes are multifactorial and, particularly with respect to biofilm formation, not yet fully characterized. Multiple processes have been associated with biofilm formation in C. difficile in a complex web of interrelated pathways, which differ between strains [1, 2, 12]. The main systems that have been linked to biofilm formation in C. difficile include motility, sporulation and signalling by a second messenger molecule, c-di-GMP [1, 2, 12].

Initiation of biofilm formation in C. difficile has also been linked to stressors such as the presence of antimicrobials, including antibiotics such as vancomycin and metronidazole as well as heat stress [13]. A physiologically relevant increase in temperature from 37 to 41 °C, as would be notable in CDI, caused increased expression of Class I heat shock genes in C. difficile 630, which are involved in the stress response in C. difficile [14, 15]. Included in this response was increased expression of the dnaK gene (CD2461), found on the dnaJK operon, encoding the highly conserved DnaK molecular chaperone [14–16]. Also referred to as Hsp70 (heat shock protein 70), DnaK is responsible for protein folding, particularly under environmental conditions that cause stress to the cells. Disruption of the dnaK gene in C. difficile 630Δerm using ClosTron mutagenesis led to various phenotypic changes in the mutant 630Δerm:dnaK cells, including increased cell length by ~50%, loss of motility and increased aggregate biofilm production [17]. This identified the dnaK gene as a potential link between the bacterial stress response under physiologically relevant heat stress and biofilm formation, which is associated with increased disease severity. This work investigates transcriptomic changes in dnaK-deficient C. difficile in both planktonic and colony biofilms to identify pathways that may bridge the gap between heat stress and increased biofilm formation in C. difficile. We hypothesized that disruption of the dnaK gene would result in coordinated transcriptional changes involving pathways previously associated with biofilm formation, including sporulation, motility, cell wall remodelling and the stress response.

Methods

C. difficile strains and culture conditions

C. difficile strains 630 (ATTC: BAA-1382), 630Δerm (originally gifted from Professor Minton’s Lab, University of Nottingham) and 630Δerm:dnaK [17] were obtained from Ulster University frozen stocks maintained at −80 °C on TSC ‘Protect’ Anaerobe Cryobeads (Gem Scientific, UK). All strains were cultured on brain heart infusion (BHI) (37 g l⁻¹) (OXOID, UK) agar (1.2% wt/v) supplemented with 5 g l⁻¹ of yeast extract and 1 g l⁻¹ of l-cysteine (BHIS) under anaerobic conditions (80% nitrogen, 10% hydrogen and 10% carbon dioxide) at 37 °C in a Don Whitley Scientific MACS500 anaerobic workstation. Overnight cultures were set up in triplicate for all experiments from a single colony in 10 ml of BHIS broth, preconditioned to anaerobic conditions and cultured for 16–18 h under anaerobic conditions.

Colony biofilm growth on semi-permeable membranes

Colony biofilms of all strains 630, 630Δerm and 630Δerm:dnaK were cultured on white mixed nitrocellulose ester membranes (MF-Millipore, Germany) of pore size 0.22 µm (MF-Millipore, Germany) in a method adapted from James et al. (2018) [18]. The membranes were placed on BHIS agar, and a 30 µl volume of a 1% overnight bacterial inoculum in BHIS broth was carefully spotted to the centre of each membrane. All membrane biofilms were cultured for 24 h at 37 °C under anaerobic conditions. All assays were set up with three biological replicates, each with two technical replicates, for each strain, with three replicates of blank controls that were incubated only with sterile BHIS broth spotted onto the membranes.

Low vacuum scanning electron microscopy of colony biofilms

Colony biofilms were removed from the anaerobic cabinet and fixed in 1 ml 2.5% glutaraldehyde in PBS (pH 7.4) for 1 h at room temperature. The biofilms were then rinsed by carefully submerging each in a separate 1 ml aliquot of 1× PBS (pH 7.4) in a 24-well microtitre plate. The membranes were returned to the original BHIS agar plate to reduce over-drying, while wet samples were being imaged. Samples were imaged under low vacuum using an FEI QuantaTM 200 (FEI Company, Eindhoven, The Netherlands) environmental scanning electron microscope (ESEM) equipped with a large field detector (LFD). All samples were processed using an accelerating voltage of 30 kV at spot size 4.0 and a working distance of 7.6–7.8 mm and a scan speed of 30 µs. Images were captured at 200×, 800× and 12,000× magnifications.

RNA extraction

C. difficile colony biofilms grown on mixed nitrocellulose ester membranes (13 mm ø) for 24 h were used for total RNA extraction by the RNeasy PowerBiofilm kit (Qiagen, Germany). For each extraction, 6 biofilms were harvested by physically scraping them off the membrane using sterile plastic loops and into a solution of 350 µl MBL (RNeasy PowerBiofilm reagent) and 5 µl ß-mercaptoethanol (Sigma-Aldrich, US). The biofilm was homogenized in the solution by repeated pipetting up and down before being transferred into the bead tube. For planktonic cell harvest, a 10% inoculum from an overnight culture in 20 ml BHIS was cultured anaerobically for 6 h at 37 °C before 5.5 ml was centrifuged at 14,000 ×g for 5 min. The supernatant was removed, and the cell pellets resuspended as per the biofilm samples in 350 µl solution containing 345 µl of solution MBL and 5 µl of ß-mercaptoethanol before being transferred to the bead tube. A 100 µl aliquot of FB solution from the RNeasy PowerBiofilm kit was also added to each bead tube before briefly vortexing and incubating at 65 °C for 5 min. Cell numbers were not quantified prior to RNA extraction due to the extraction kit specifying an input of ~50 mg biomass. Preliminary sacrificial biofilms and cultures were used to determine reproducible harvesting conditions that consistently yielded the required biomass (data not shown). The tubes were then vortexed horizontally using a vortex genie (Scientific Industries Inc., USA) with a horizontal adapter on full speed for 10 min. The bead tubes were then centrifuged for 1 min at 13,000 g and the supernatant transferred into a clean microtube. To the supernatant, 100 µl of ice-cold IRS solution was added and mixed by inverting several times before being placed on ice for 5 min at 4 °C. The microtubes were centrifuged as before for 1 min at 13,000 ×g at room temperature, and the supernatant was transferred into a new microtube. To the supernatant, 450 µl of PB solution was added as well as 450 µl of ethanol. The solution was mixed by pipetting up and down until uniformly mixed before being transferred into spin columns in 650 µl portions and centrifuged as before until all the solution had been used and the flow-through discarded. The spin columns were washed with 650 µl of well-shaken PW solution centrifuged through the column at 13,000 ×g for 1 min, and the flow-through discarded. The columns were then centrifuged again as before with no additional reagent before 50 µl of DNase and RDD buffer supplied in the kit were added and incubated for 20 min at room temperature. To the columns containing DNase solution, 400 µl of WB buffer was added and the columns centrifuged as before and the flow-through discarded. The columns were then washed once more as before with 650 µl of PW solution. Aliquots of 650 µl of ethanol were then added to the column, centrifuged as before, and the flow-through discarded. To remove residual reagent and dry the columns, they were centrifuged again for 1 min at 13,000 ×g and the catch tube discarded and replaced with a clean microtube. To remove the RNA bound to the column, 50 µl of RNase-free water was added directly to the centre of each column and incubated at room temperature for 5 min before being centrifuged at 13,000 ×g for 1 min. The flow-through was placed on ice and the column transferred to a clean microtube. A fresh 50 µl of RNase-free water was added to the column and incubated at room temperature for 5 min before the elution was repeated. The RNA quantity and purity in the elutions were assessed using a Nanodrop ND1000 spectrophotometer. First and second elutions of acceptable purity and quantity from each sample were combined and treated with an in-solution DNase digest step using combined 80 µl of elution, 10 µl of DNase and 10 µl of buffer RDD from the Qiagen RNeasy PowerBiofilm kit. This was incubated at room temperature for 20 min before being re-bound to the spin column using the PB solution and ethanol and subsequently washed with WB and PB solutions as described above. The final purified RNA was eluted in 50 µl of RNase-free water. The quantity and purity of the elution was tested via Nanodrop.

RNA sequencing

RNA sequencing was carried out on the Illumina platform at Deepseq (University of Nottingham, UK). Dry ice was used for shipment of the RNA samples to Deepseq, where the sequencing facility reported that the samples were quality-checked again using an Agilent TapeStation 4200 and the Agilent RNA ScreenTape Assay Kit (Agilent, USA). Following this, a Qubit RNA BR assay kit from Life Technologies was used to quantify the total concentration of the RNA samples so that an aliquot of 1 µg of total RNA could be used for rRNA depletion using the QIAseq FastSelect –5S/16S/23S kit (Quiagen, Germany). A NEBNext Ultra II Directional RNA library prep kit for Illumina (NEB, USA) was then used with NEBNext^® Multiplex Oligos for Illumina^® (96 Unique Dual Index Primer Pairs) to create the whole transcriptome indexed libraries. After second-strand synthesis using eight cycles of PCR, the libraries were quantified using a Qubit Fluorometer with a Qubit dsDNA HS Kit (Thermo Fisher Scientific, USA). Library fragment-length distributions were analysed using the Agilent TapeStation 4200 and the Agilent High Sensitivity D1000 ScreenTape Assay (Agilent). Libraries were pooled in equimolar amounts and quantified using the KAPA Library Quantification kit for Illumina (Roche, Germany). The pooled library was then loaded onto the Illumina NextSeq 500, on a NextSeq 500 Mid Output v2.5 150 cycle kit (Illumina, USA), to generate ~10 million 75 bp paired-end reads per sample. RNA sequencing was performed on three biological replicates for each strain in biofilm and planktonic mode of growth, allowing DESeq to estimate gene-wise variability across replicates and improve modelling of expression counts for differential expression analysis.

Differential gene expression analysis

Raw sequencing reads were trimmed to remove Illumina adapters and low-quality nts (Q<20) using Trim Galore (v0.6.7), and reads shorter than 20 bp were discarded. Trimmed reads were aligned to the appropriate Clostridium difficile reference genome using HISAT2 (v2.2.0). StringTie (v2.2.1) was used for transcript assembly and gene-level quantification. Raw gene-level read counts were supplied to DESeq2 (v1.34.0) for differential expression analysis. DESeq2 normalization was performed internally by estimating sample-specific size factors using the median-ratio method, followed by dispersion estimation and statistical testing using the default settings. Resulting P values were corrected for multiple testing using the Benjamini–Hochberg method, and genes with an adjusted P value<0.05 were considered differentially expressed. No minimum log_₂ fold-change (log₂FC) threshold was applied. Seven DESeq2 gene expression contrasts were performed to identify the differences in gene expression between each strain grown as a colony biofilm compared to planktonic growth and between mutant and parent strains. Data generated from RNA sequencing including HTSeq counts are available in the NCBI GEO repository, accession no. GSE336654. Full gene lists with log₂FCs per contrast are available as supplementary data (Table S3).

Gene set enrichment analysis by KEGG and biological processes

Gene set enrichment analysis (GSEA) was used to identify KEGG pathways and biological processes (BPs) from Gene Ontology (GO) that were significantly enriched in each dataset. KEGG pathway gene sets for C. difficile 630 (organism code: cdf) and C. difficile 630Δerm (organism code: pdf) were retrieved in R using KEGGREST, and the associated genes were matched to those present in the expression datasets for each strain. GO annotations were downloaded from QuickGO for C. difficile 630 (Taxon: 272563) and mapped to C. difficile 630Δerm using the old locus-tag gene identifiers shared between the two strains. Genes were ranked by log₂FC prior to enrichment analysis. No minimum log₂FC threshold was applied.

Quantitative Real-Time PCR (qPCR) validation

Primers used in this study were selected for genes involving key virulence factors in C. difficile, including motility, toxin production and sporulation, with reference genes adk, rpsJ and gyrA (Table S2). The primer efficiencies were tested using cleaned-up PCR products generated from each primer pair. A fivefold dilution series was created from PCR products in nuclease-free water. A master mix of Sybr Green Master Mix 10× (Roche, Germany) and nuclease-free water was aliquoted into microfuge tubes for each primer pair being tested. The primers were added from a working stock containing 10 mM forward and 10 mM reverse primers to a final concentration of 1 mM. The master mix for each primer set was aliquoted into PCR tubes, one for each dilution being tested and one for each of the no-template controls containing nuclease-free water. To each tube, the template or nuclease-free water control was added in aliquots of 3 µl before being vortexed briefly to mix and centrifuged gently for 30 s at 5,000 ×g to gather all liquid to the bottoms of the tubes. The mixture was aliquoted into three technical replicates per dilution onto the qPCR plate.

qPCR was carried out on a Lightcycler^® 480 (Roche, Germany) with an initial denaturation step of 95 °C for 5 min, followed by 50 cycles of denaturation at 95 °C for 10 s, primer annealing at 57 °C for 10 s and then an extension step of 72 °C for 30 s. A melting curve was performed with all runs from 65 °C to 97 °C to assess the quality of each run.

To calculate the percentage primer efficiency, the technical replicates for each dilution were averaged and the log value of sample quantity calculated from the dilution. The slope of log sample quantity against average threshold cycle (Ct) value was calculated using the formula below in Microsoft Excel.

=SLOPE(Average Ct value range, log quantity range)

The percentage primer efficiency was then derived from the calculated ‘slope’ using the following formula:

(10^(−1/Slope)−1)×100

Reverse transcription of extracted RNA was performed by using 500 ng of RNA and 50 ng of random primers (Promega, UK) made up to 12 µl using nuclease-free water. This was incubated for 10 min at 70 °C. After incubation, 4 µl of 5× first strand buffer, 2 µl of 0.1 M DTT mix and 1 µl of 10 mM dNTPs mix containing (dATP, dTTP, dCTP and dGTP) to a final concentration of 0.2 mM were added with 1 µl of RNaseOUT (Invitrogen, USA). A further incubation for 2 min at 42 °C was performed before 200 U of reverse transcriptase was added and incubated first for 50 min at 42 °C, then a further 15 min at 70 °C. All reactions were performed alongside negative controls for each RNA sample, which contained no reverse transcriptase enzyme. End-point PCR was performed on all reverse transcription reactions and controls with adk reference gene primers (Table S1) and visualized using gel electrophoresis to validate that the reverse transcription reaction was efficient.

qPCR was performed on all RNA samples diluted 1 : 10 in RNase-free water using the same cycle conditions and master mix as previously described. A separate qPCR plate was used for each of the validated qPCR primers.

Standard curves were generated for each primer set using the same qPCR cycle conditions as described above. For each primer set, a pooled cDNA sample from the experimental condition with the lowest average Ct values was selected. This ensured that the lowest possible Ct values were included in the standard curves, and the dynamic range of Ct values included was wide enough to cover all samples. The pooled cDNA was first diluted 1 : 10 before being diluted fivefold and added to the qPCR master mix as previously described.

The percentage primer efficiencies were calculated as previously described before being converted to a different scale using the formula below for calculation of relative gene expression:

Converted primer efficiency=[primer efficiency (%)/100]+1

The Pfaffle method [19] was used to calculate relative gene expression for each sample with multiple housekeeping genes, using the formula below, where E is the converted primer efficiency, GOI refers to ‘gene of interest’ and REF refers to ‘reference gene’:

$${\displaystyle {\displaystyle \mathrm{R}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{G}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}=\frac{({\mathrm{E}}_{\mathrm{G}\mathrm{O}\mathrm{I}}{)}^{\Delta \mathrm{C}\mathrm{t}\mathrm{G}\mathrm{O}\mathrm{I}}}{\mathrm{G}\mathrm{e}\mathrm{o}\mathrm{M}\mathrm{e}\mathrm{a}\mathrm{n}[({\mathrm{E}}_{\mathrm{R}\mathrm{E}\mathrm{F}}{)}^{\Delta \mathrm{C}\mathrm{t}\mathrm{R}\mathrm{E}\mathrm{F}}]}}}$$

Results

C. difficile colony biofilms

Visible differences in colony biofilm structure could be observed by eye for the 630Δerm:dnaK strain, which appeared shrivelled and had reduced extracellular polymeric substance compared to the parent and wild-type strains. To assess a hypothesized structural difference in more detail, colony biofilms of strain 630, 630Δerm and 630Δerm:dnaK were grown on mixed nitrocellulose ester membranes for 24 h prior to being imaged by low vacuum scanning electron microscopy (SEM) with an LFD detector. The structure of the 630 and 630Δerm 24 h colony biofilms was similar at all magnifications (Fig. 1). Many tears in the relatively smooth surface of the biofilm revealed a complex biofilm matrix within. At a higher magnification (12,000×), the cells on the surface of the 630 and 630Δerm colony biofilms appeared very tightly packed with minimal gaps between them. However, the structure of the 630Δerm:dnaK colony biofilm did not exhibit the same pattern of tearing on the biofilm surface. With the mutant, the tears were much longer and apparently connected but were relatively thin in comparison to the ‘tears’ of the parent and wild-type biofilms. At 12,000× magnification, the 630Δerm:dnaK cells were faintly visible as an apparently more irregular network under the surface of the biofilm, with some small gaps forming between the cells. Overall, the dnaK mutant biofilms appeared denser, with the lengthened cells tangled and shrouded in a thicker matrix.

Fig. 1. C. difficile 630, 630Δerm and 630Δerm:dnaK 24 h colony biofilms cultured on semi-permeable 13 mm ᴓ mixed nitrocellulose ester membranes with pore size 0.2 µm and imaged by low-vacuum SEM equipped with an LFD at 200×, 800× and 12,000× magnification.

Overall differential gene expression per contrast

Seven differential gene expression comparisons were performed between mutant and parent strain under the same growth conditions and between growth conditions within the same strains. RNA sequencing data were confirmed by qPCR validation (Table S3). C. difficile 630Δerm had only 6.5% of genes statistically significantly differentially expressed under planktonic conditions compared to the wild-type 630 (Fig. 2). Whereas when grown in a biofilm, the same strains differed by 40.8% of the transcriptome. Disruption of the dnaK gene in 630Δerm:dnaK resulted in a 52.8% change in the transcriptome compared to 630Δerm under planktonic growth, whereas this increased to a 64.0% difference when the strains were grown as colony biofilms.

Fig. 2. Differential gene expression between C. difficile strains under planktonic and biofilm growth conditions. Differential expression was compared between strains 630 (630), 630Δerm (erm), and 630Δerm:dnaK (dnak), cultured planktonically (p) or as 24 h colony biofilms (b). Diagonal boxes identify the strain and growth condition included in each comparison. Lower-left panels show volcano plots of DESeq2-estimated log₂FC (horizontal-axis) against −log₁₀ Benjamini–Hochberg-adjusted P value (vertical-axis). For contrasts labelled A_vs_B, positive log₂FC values indicate higher expression in A relative to B, whereas negative values indicate lower expression in A relative to B. Genes with an adjusted P value<0.05 are shown in red when upregulated in A and blue when downregulated in A; genes not meeting this criterion are shown in grey. No minimum log₂FC threshold was applied. Upper-right panels report the number and percentage of genes with an adjusted P value<0.05 in each comparison. Percentages were calculated relative to the total number of genes tested. Grey panels indicate comparisons that were not performed.

When comparing biofilm to planktonic growth for each of the strains, 630 and 630Δerm showed similar changes in transcriptome with 63.2 and 66.0%, respectively. 630Δerm:dnaK, when grown in a colony biofilm compared to planktonic growth showed the biggest change in transcriptome with 72.4% of genes statistically significantly differentially expressed (P_adj<0.05).

Across the three strains, the biofilm-to-planktonic contrasts revealed a strong shared transcriptional response, with a core set of 634 genes consistently significantly increasing in expression and 625 consistently significantly decreasing in expression (Fig. 3a). However, the 630Δerm:dnaK biofilm-to-planktonic contrast has a large number of differentially expressed genes falling outside this shared core. There were 301 genes that were uniquely significantly increased when compared to the other biofilm-to-planktonic contrasts and 260 genes that were uniquely significantly decreased. Genes belonging to each group can be identified in the gene-membership tables (Tables S4–6). Also notably, 183 genes are significantly decreased in a 630Δerm:dnaK biofilm contrast, where they are increased in 630 and 630Δerm, and oppositely, 153 are increased in a 630Δerm:dnaK biofilm when decreased in 630 and 630Δerm.

Fig. 3. Overlap of significantly differentially expressed genes between C. difficile strains under planktonic and biofilm growth conditions. UpSet plots showing the overlap and uniqueness of significantly upregulated and downregulated genes across three sets of comparisons: (a) biofilm versus planktonic growth in the 630, 630Δerm and 630Δerm:dnaK strains; (b) 630Δerm:dnaK versus 630Δerm under biofilm and planktonic conditions; and (c) 630Δerm versus 630 under biofilm and planktonic conditions.

When comparing 630Δerm:dnaK to 630Δerm under either planktonic or biofilm conditions, a large portion of genes were significantly differentially expressed only in a biofilm (547 increased and 517 decreased), and a large number (352 decreased and 306 increased) were significantly differentially expressed regardless of the conditions (Fig. 3b). When considering the 630Δerm compared to 630, relatively few genes were significantly differentially expressed in planktonic conditions; however, a large number were differentially expressed (771 increased and 657 decreased) when forming a biofilm (Fig. 3c).

GSEA

GSEA of all contrasts were compared not only to identify patterns of changing pathways when forming a biofilm but also to identify differences in gene expression between strains under the same conditions. GSEA identifies clusters of coordinated gene expression where genes belonging to a specific pathway are significantly concentrated in a positive or negative direction without applying a differential expression significance cutoff. As a baseline, comparing gene expression in C. difficile 630 colony biofilms to planktonically cultured cells revealed 15 KEGG pathways and 7 BPs that were enriched in this contrast (Fig. 4). The only category that was significantly increased in expression was ‘sporulation resulting in formation of a cellular spore’. The categories that were decreased in expression centred around metabolic processes. When gene expression in C. difficile 630Δerm colony biofilm was compared to planktonically cultured cells, a very similar pattern emerged with focus on the same categories combined with the addition of a decrease in bacterial-type flagellum-dependent cell motility and bacterial-type flagellum organization, as well as decreases in some additional metabolic processes including selenocompound metabolism, pyrimidine metabolism and porphyrin metabolism (Fig. 4). In contrast, when comparing gene expression in C. difficile 630Δerm:dnaK colony biofilms to planktonically cultured cells, there were double the number of KEGG pathways (30 in total) that were enriched, with over half of them being unique to C. difficile 630Δerm:dnaK biofilm formation when compared to the parent and wild-type strains. There were nine core pathways enriched in all strains forming a colony biofilm and were involved with metabolic processes and flagellar assembly. The additional KEGG pathways that are enriched in 630Δerm:dnaK are heavily involved with carbon fixation, amino acid biosynthesis and remodelling of the cell wall. Peptidoglycan synthesis was increased in expression in 630Δerm:dnaK colony biofilms, where no other KEGG pathways were positively enriched in the parent or wild-type strains forming a biofilm.

Fig. 4. GSEA of C. difficile strains under planktonic and biofilm growth conditions. GSEA was performed on transcriptomic comparisons of C. difficile strains 630, 630Δerm (erm) and 630Δerm:dnaK (dnaK) cultured planktonically or as a 24 h colony biofilm to identify enriched KEGG pathways and GO BPs terms associated with growth condition and strain differences. Fractions notated beside each data point represent the rank of KEGG pathway or GO by NES (normalized enrichment score), showing from 1 /N, most enriched pathway to N/N. N=number of pathways enriched.

When comparing 630Δerm:dnaK colony biofilms to parent strain 630Δerm colony biofilms, there were eight KEGG pathways, which were all positively enriched except for flagellar assembly, which was negatively enriched. The eight pathways that were positively enriched included biosynthesis of amino acids, specifically phenylalanine, tyrosine and tryptophan, as well as fatty acids and secondary metabolites, alongside oxidative phosphorylation and both 2-oxocarboxylic acid and starch and sucrose metabolism. There were also 12 BPs enriched, which were mostly positively enriched, with 3 being negatively enriched. The three negatively enriched BPs were all involved with motility, namely chemotaxis, bacterial-type flagellum-dependent cell motility and bacterial-type flagellum organization. The largest enriched BP in this comparison was cell division, followed by translation, then equally multiple metabolic processes including lipid metabolic process, amino acid biosynthetic process, fatty acid biosynthetic process and aromatic amino acid family biosynthetic process. The differences in transcription profiles of 630Δerm:dnaK colony biofilms compared to parent strain 630Δerm colony biofilms were not reflected when the same strains were compared in planktonic growth. More metabolic pathways were differentially expressed in the planktonically grown culture comparison, with only three KEGG pathways showing similar enrichment patterns to the biofilm comparison, being a negative enrichment in flagellar assembly and a positive enrichment in both 2-oxocarboxylic acid metabolism and oxidative phosphorylation. Carbon metabolism was positively enriched, when planktonic 630Δerm:dnaK were compared to 630Δerm under the same conditions. However, this pathway is not enriched in 630Δerm forming a biofilm and is strongly decreased in expression in 630Δerm:dnaK forming a biofilm. This same pattern was reflected in five other KEGG pathways, namely, TCA cycle, arginine and proline metabolism, butanoate metabolism, other carbon fixation pathways and microbial metabolism in diverse environments. Five KEGG pathways were enriched exclusively in the 630Δerm:dnaK to 630Δerm planktonic comparison, one of which, the two-component system, was decreased in expression, whereas lipoic acid metabolism, one-carbon pool by folate, glyoxylate and dicarboxylate metabolism and pyruvate metabolism were all increased in expression.

The remaining two comparisons, between 630 and 630Δerm when grown in a biofilm and when grown planktonically, had the least number of enriched categories. When grown in a biofilm, five KEGG pathways and five BPs were enriched in the comparison between 630 and 630Δerm. The five KEGG pathways included positive enrichment of histidine metabolism, fructose and mannose metabolism, microbial metabolism in diverse environments and flagellar assembly with only ribosome KEGG pathway negatively enriched. The five BPs reflected this with a positive enrichment of l-histidine biosynthesis and bacterial type flagellar dependent cell motility, with a negative enrichment in sporulation, fatty acid biosynthesis and translation. When grown planktonically, five KEGG pathways and three BPs were enriched in the comparison between 630 and 630Δerm. Notably, the three BPs that were enriched in 630Δerm planktonic cells, when compared to 630, were all motility-related and all positively enriched. This is also reflected in the increased expression of the flagellar assembly and bacterial chemotaxis KEGG pathways in 630Δerm planktonic cells when compared to 630.

Discussion

C. difficile biofilms are increasingly recognized as spore reservoirs contributing to persistence and recurrence of CDI [1, 2, 7]. Understanding the regulatory networks that promote biofilm development may help to identify pathways that could be targeted to reduce this disease persistence. The phenotypic effects of disruption to the dnaK gene in C. difficile have been extensively reported to demonstrate a link between the stress response and biofilm formation in C. difficile [17, 20, 21]. The transcriptomic analysis in this paper reveals the specific pathways that are affected by disruption of the dnaK gene and closely align with known pathways of biofilm formation.

Initial colony biofilm investigations found that disruption of the dnaK gene resulted in structural differences in colony biofilm morphology at 24 h (Fig. 1). The 24 h 630Δerm:dnaK colony biofilm structure had longer and apparently connected tears in the surface, which were relatively thin in comparison to the parent and wild-type colony biofilms at 24 h. Similar structural differences were previously observed when C. difficile 630, 630Δerm and 630Δerm:dnaK were grown as aggregate biofilms on glass coverslips and imaged by high-vacuum SEM [20]. These findings reveal that disruption of the dnaK gene in C. difficile 630Δerm causes similar biofilm phenotypic traits regardless of the method of biofilm culture.

Transcriptomic analysis showed that disruption of the dnaK gene caused widespread metabolic and transcriptional reprogramming in comparison to the parent strain 630Δerm. A coordinated upregulation of metabolic processes of the dnaK mutant in planktonic culture was seen increased expression in 13 KEGG pathways, including metabolism of pyruvate, d-amino acids, lipoic acids, butanoate and carbon. This higher rate of metabolic processes in 630Δerm:dnaK aligns with increased metabolism due to a dysregulated stress response and/or a delay in achieving stationary phase [17, 22]. Molecular chaperone proteins DnaJK are both transcriptionally and proteomically increased in expression in C. difficile 630Δerm when in stationary phase [22]. Therefore, we propose that 630Δerm:dnaK is delayed in metabolically achieving stationary phase due to the lack of efficient chaperone proteins. Alternatively, it is physiologically possible that the cells are both delayed in growth phase alongside experiencing a stress response. This is supported by the reduced planktonic growth of 630Δerm:dnaK at 37 °C previously reported, which was originally hypothesized to be a result of a dysregulated heat stress response in the cells [17].

After 6 h of planktonic growth, cultured 630Δerm:dnaK cells expressed increased oxidative phosphorylation as well as sporulation as a BP, in contrast to 630Δerm under the same conditions. Spore formation is a known stress response as a way of survival and is also strongly associated with biofilm formation [8, 9]. Sporulation is an energy intensive process where phosphorylation of key regulatory factors, such as Spo0A, plays a critical role in initiation [10, 23]. This premature increased expression of genes associated with the sporulation pathway in planktonic 630Δerm:dnaK suggests earlier initiation of spore formation, which may result from the heightened stress response in this strain, leading to earlier biofilm initiation. Of note, functional spores have not been recoverable from the DnaK-deficient strain [20, 24]. Therefore, although the genetic pathway for sporulation has been increased in expression, physical spores have not been observed, indicating a problem with production of the spore proteins or spore assembly. These cells are either failing to sporulate post-transcriptionally or, as is seen in Bacillus subtilis, the alternative phosphorylation of Spo0A could have redirected cellular pathways towards early biofilm formation [23].

Of particular interest in the transcriptome of 630Δerm:dnaK colony biofilm cells, compared to 630Δerm, was a significantly increased enrichment of cell division, which was not observed in any other contrast. This, combined with the increased translation, tRNA modification and other biosynthetic and metabolic processes in the 630Δerm:dnaK biofilm cells, could indicate that the dnaK mutation has resulted in a cellular attempt to increase growth and proliferation in a biofilm compared to 630Δerm under the same conditions. Rapidly proliferating cells in the 630Δerm:dnaK biofilms could have explained the increased biofilm; however, previous studies found no significant increase in cell number in 630Δerm:dnaK aggregate biofilms compared to the parent strain [20]. Therefore, the question remains whether the transcriptionally increased cell division pathways are phenotypically successful in the dnaK mutant. A post-transcriptionally failed cell division process could be attributed to the 50% longer cell phenotype observed in 630Δerm:dnaK in comparison to the parent and wild-type strain. However, since this phenotype is also observed in planktonic 630Δerm:dnaK cells [17], it would have been expected to see an increase in cell division in the 630Δerm:dnaK to 630Δerm planktonic contrast.

A strong coordinated reduction in motility genes in 630Δerm:dnak biofilm cells compared to 630Δerm was noted by significant enrichment in flagellar assembly, chemotaxis, bacterial-type flagellum-dependent cell motility and bacterial-type flagellum organization. This lack of motility aligns with phenotypical observations of the dnaK mutant [17, 20] as well as enhanced biofilm formation, as it is well established that genes associated with flagella are decreased in C. difficile biofilm formation [25, 26]. In C. difficile mutant cells where either flagellar assembly or post-translational modifications have been altered, biofilm formation has been affected [13, 27]. For example, multiple different modifications in flagellar assembly of C. difficile R20291 led to increased cell hydrophobicity, which was concluded to be responsible for increased biofilm formation via increased aggregation [27]. Similarly, a restructuring of the cell surface suggested to cause increased hydrophobicity in 630Δerm:dnaK cells [21], aligns with significantly increased enrichment in peptidoglycan synthesis in 630Δerm:dnaK biofilms compared to planktonic cells.

The 630Δerm planktonic cells also showed a significant increase in five pathways relating to motility and chemotaxis compared to 630, which has not been phenotypically observed in previous work [28, 29]. Conversely, four out of five of these motility-related pathways were significantly decreased in DnaK-deficient planktonic cells compared to 630Δerm. Taken together, these observations strongly suggest that disruption of the dnaK gene led to a transcriptional shift in motility towards a sessile cellular state seen in biofilm initiation.

Overall, loss of erythromycin resistance in 630Δerm when cultured planktonically resulted in five KEGG pathways and three BPs that were significantly enriched compared to the wild-type 630. Previous genomic comparison of 630Δerm and wild-type 630 found multiple genomic variations in addition to the loss of erythromycin resistance [29]. None of these variations were directly associated with motility or chemotaxis and therefore cannot easily explain the transcriptomic changes identified in these pathways in this work [29]. However, due to the high quantity of mobile genetic elements in C. difficile 630 [30], it is entirely possible that subculturing of C. difficile 630Δerm by different labs may have led to variations in different lab strains. Importantly, the 630Δerm strain used in this study was the parent strain used to derive the 630Δerm:dnaK. Therefore, we can determine that the transcriptional differences in 630Δerm:dnaK should not be due to other genomic differences in 630Δerm [17].

Conclusion

This work generated the first global transcriptional description of a DnaK-deficient C. difficile strain grown as a colony biofilm and identified several pathways linking the stress response to increased biofilm formation. A core number of KEGG pathways did remain enriched in all strains, including a decrease in flagellar assembly, as well as decreases in biosynthesis of amino acids, aminoacyl-tRNA, secondary metabolites, cofactors and ribosomes and a reduction in glycolysis/gluconeogenesis. This is reflected in the BPs, where tRNA aminoacylation for protein translation and translation were reduced in all strains forming a biofilm. This paints a very clear picture that reduction in protein metabolism in biofilms is a core process as cells enter a sessile state. Perhaps by disruption of the dnaK gene, producing a key molecular chaperone for protein folding, this has promoted an active stress response together with reduced translation, leading to earlier and enhanced biofilm formation [17, 20]; via eradication of motility processes, forcing cells into a sessile state alongside premature upregulation of sporulation-associated genes at 6 h planktonic growth. Further proteomic analysis would generate a fuller picture of how disruption of the dnaK gene affects translation and protein folding. However, these findings support a role for dnaK in coordinating stress adaptation and biofilm-associated physiology and provide a framework for future mechanistic studies investigating persistence-related phenotypes in C. difficile.

Abbreviations

BHI

brain heart infusion

BHIS

brain heart infusion-supplemented

BP

biological process

CDI

C. difficile infection

ESEM

environmental scanning electron microscope

GEO

Gene Expression Omnibus

GO

Gene Ontology

GSEA

gene set enrichment analysis

KEGG

Kyoto Encyclopedia of Genes and Genomes

LFD

large field detector

log₂FC

log_₂ fold-change

NCBI

National Center for Biotechnology Information

SEM

scanning electron microscopy