Fidaxomicin resistance in Clostridioides difficile: a systematic review and predictive modeling with RNA polymerase binding sites

ABSTRACT

Fidaxomicin (FDX), an RNA polymerase (RNAP) inhibitor antibiotic, is a guideline-recommended therapy for Clostridioides difficile infection. Mutations associated with reduced FDX minimum inhibitory concentrations (MICs) have been identified. However, the molecular characterization of these mutations on FDX binding and the development of FDX resistance have not been studied. The purpose of this systematic review was to identify FDX resistance in C. difficile isolates and determine whether single nucleotide polymorphisms associated with increased FDX MIC aligned with the RNAP binding pocket interacting residues. A systematic literature search was done in PubMed (1991–2023) with identified articles and their bibliographies searched for papers that included C. difficile genetic mutations and increased FDX MIC. Visualization of FDX-RNAP interactions was performed on Schrödinger Maestro using the publicly available C. difficile RNAP with fidaxomicin sequence (code 7L7B) on the Protein Data Bank. Seven articles were identified after applying inclusion and exclusion criteria. The most common mutation in clinical and laboratory isolates was at position V1143 of the β subunit, which accounted for approximately 50% of the identified mutations. Most other mutations occurred within the β′ subunit of RNAP. Approximately one-third of the identified mutation aligned directly with FDX interacting residues with C. difficile RNAP (7/20) with most of the remainder occurring within 5 Å of the binding residues. C. difficile strains with elevated FDX MIC align closely with the known RNAP binding residues. These data demonstrate the potential to identify genomic methods to identify emerging FDX resistance.

INTRODUCTION

Clostridioides difficile infection (CDI) is a common healthcare-associated infection that accounts for an estimated 462,100 cases in the United States, of which 235,700 cases were healthcare-associated CDI and 226,400 cases were community-associated CDI in 2017 (1). The US Centers for Disease Control and Prevention designates C. difficile as an urgent threat due to a high burden of disease and increased number of CDI caused by hypervirulent strains. CDI treatment relies heavily on two antibiotics, of which fidaxomicin (FDX) is recommended due to its high efficacy, including a low rate of recurrences and microbiome-sparing activity. Recent updates in the Infectious Diseases Society of America and Society for Healthcare Epidemiology of America guidelines have placed FDX as the first-line therapy for CDI, resulting in selection pressure for resistant development and emergence of FDX-resistant isolates (2). However, antibiotic susceptibility testing for C. difficile is not performed routinely in clinical settings due to its costly, laborious, and difficult process.

FDX, an RNA polymerase (RNAP) inhibitor antibiotic, exerts its bactericidal activity on C. difficile by inhibiting the initiation of transcription. Studies using Mycobacterium tuberculosis showed that the active component of fidaxomicin (lipiarmycin A3) binds at the base of the RNAP clamp and through interactions with RNAP switch regions (SW) 2, SW3, and SW4, trapping RNAP in an open-clamp state. The open-clamp state as a result prevents the σ module, σR2 and σR4, from engaging with promoter -10 and -35 elements, respectively (3). The mechanism of FDX’s narrow spectrum of activity has recently been elucidated via cryogenic electron microscopy. Six structural components of the RNAP were critical to FDX binding, including the β clamp, the β′ SW2, β SW3, β SW4, β′ zinc-binding domain (ZBD), and β′ lid. Specifically, FDX formed hydrogen bonds or salt bridges with key residues of the β and β′ subunits, including β R1121, β′ K84, β′ K86, β′ R326, and β′ D237 and H294 of the RNAP σ subunit. The interaction between FDX-RNAP is further stabilized by the cation-ϖ interaction between FDX and R89 of the β′ subunit. Other C. difficile RNAP residues, including β Q1074 of the β switch region 3 and β S1140 of the β clamp, interact with FDX via van der Waals or hydrophobic interactions (4). The known molecular mechanism of resistance to FDX is primarily based on single point mutations in either rpoB or rpoC of the C. difficile RNAP that results in a nonsynonymous substitution. If genomic targets of FDX resistance could be identified, genomic methods to identify emerging resistance could be developed. Therefore, the purpose of this systematic review was to identify genomic changes associated with elevated FDX and explore how often those single nucleotide polymorphisms (SNPs) aligned at or close to the key residues of the FDX-RNAP binding site.

RESULTS

The flow diagram of the study selection process is shown in Fig. 1. A total of 125 studies were initially identified, of which 115 studies were from the PubMed database search and 10 studies were from bibliography screening. After abstracts and titles were screened, 24 studies were retrieved, of which 17 were assessed for eligibility. A total of seven studies that met the inclusion and exclusion criteria were included in the final analysis (Table 1) (5–11). A total of 18 isolates were identified across seven studies, of which 10 isolates (56%) are laboratory strains and 8 isolates (44%) are clinical strains.

Fig 1.

Preferred Reporting Items for Systematic Review and Meta-Analyses flow chart.

TABLE 1.

Summary of fidaxomicin non-susceptible isolates and association with fidaxomicin binding site^c

Country, reference	Study type	Strain type	Reference strain for WGS	Gene name	Nucleotide change	Amino acid change	Parent strain FDX MIC (mg/L)	Mutant strain FDX MIC (mg/L)	MIC method	FDX binding site	Mutation within 5 Å of FDX binding site
USA (9)	SP	Lab	CD630	marR/CD22120	ΔT349	Frameshift after AA117	0.25	1	AM	N	N
USA (9)	SP	Lab	CD630	rpoB	A3221G	Gln1074Arg	0.25	4	AM	Y	Y
UK (8)	MC	Lab	R20291	rpoB	T3428A	Val1143Asp	0.25	>32	AM	N	Y
UK (8)	MC	Lab	R20291	rpoB	T3428G	Val1143Gly	0.25	8	AM	N	Y
UK (8)	MC	Lab	R20291	rpoB	G3427T	Val1143Phe	0.25	2	AM	N	Y
USA (7)	SP	Lab	ATCC 9689a	rpoC	G709T	Asp237Tyr	0.125	2–4	BM	Y	Y
USA (7)	SP	Lab	ATCC 9689a	rpoB	G3427T	Val1143Phe	0.125	2–4	BM	N	Y
USA (7)	SP	Lab	ATCC 9689a	rpoB	C3220A	Gln1074Lys	0.125	2–4	BM	Y	Y
USA (6)	SP	Lab	ATCC 43255	rpoC	C2341A/ A2342G and C3381A	Gln781Arg and Asp1127Glu	0.016– 0.25	2–4	BM	N	–b
USA (6)	SP	Lab	ATCC9689a	rpoB	C3220A	Gln1074Lys	0.016	1	BM	Y	Y
France (10)	CR	Clinical	CD630	rpoB	T3428G	Val1143Gly	0.063	16	AM	N	Y
Germany (11)	CR	Clinical	CD630	rpoB	T3428A	Val1143Asp	NA	>64	AM	N	Y
Japan (5)	FDX CS	Clinical	CD630	rpoB	G3427C	Val1143Leu	0.12	8	AM	N	Y
Japan (5)	FDX CS	Clinical	CD630	rpoC	A265G	Arg89Gly	0.12	4	AM	Y	Y
Japan (5)	FDX CS	Clinical	CD630	rpoB	T3428G	Val1143Gly	0.03	16	AM	N	Y
Japan (5)	FDX CS	Clinical	CD630	rpoB	T3428A	Val1143Asp	0.06	>64	AM	N	Y
Japan (5)	FDX CS	Clinical	CD630	rpoC	A260G	Arg89Gly	0.03	64	AM	Y	Y
Japan (5)	FDX CS	Clinical	CD630	rpoB	A3446C C976T	Gln1149Pro Arg326Cys	0.03	0.25	AM	N	Y

^a Corrected from Q1073R in the original paper.

^b –, R1121 is a binding residue that is six amino acid away from Asp1127.

^c WGS, whole-genome sequencing; SP, serial passage experiments; MC, mutant creation experiments; CR, case report; FDXCS, FDX clinical study; AA, amino acid; AM, agar dilution method; BM, broth microdilution method; Y, yes; and N, no.

Laboratory-generated FDX mutants

Laboratory-derived mutants were reported in multiple studies from 2011 to 2018 and were created by either a single step or serial passages on Brucella agar under FDX selection or introduction of desired single nucleotide changes into C. difficile genome. Mutants were derived from different strains including CD630, R20291, ATCC9689, and ATCC43255.

Clinical isolates with elevated FDX MIC

The emergence of clinical C. difficile isolate with reduced FDX was first reported in a phase III clinical trial of FDX against CDI in 2011 in which a single strain with elevated FDX MIC of 16 mg/L was isolated from a patient with recurrent CDI (12). More recently, a clinical C. difficile isolate with markedly reduced FDX susceptibility (MIC > 64 mg/L) was identified out of a study of 50 clinical isolates of patients with severe CDI conducted in 2018 in Germany (11). Ninety-four percentage of isolates (17/18) identified across seven studies exhibited nonsynonymous mutations in rpoB or rpoC. Within these isolates, 15 isolates conferred mutations associated with elevated FDX MIC (MIC > 1 mg/L) compared to the wild type. The most common SNP identified was a substitution of thymine (T) at position 3428 of rpoB (33%) to adenine (A) or guanine (G), which resulted in the nonsynonymous substitution of valine at position 1143 to aspartic acid or glycine, respectively. Mutation at V1143 was the most identified mutation in both laboratory and clinical settings (50%), followed by mutation at position Q1074 (17%), one of the key FDX-interacting residues in C. difficile RNAP. Different mutations occurred at V1143 in which nonsynonymous substitutions at this position conferred a wide range of FDX MIC: V1143F (MIC = 2–4 mg/L), V1143G (MIC = 16 mg/L), V1143L (MIC = 8 mg/L), and V1143D (MIC ≥ 64 mg/L). Approximately one-third of the identified mutations (35%) aligned exactly with FDX-RNAP-interacting residues, including β R326, β Q1074, β′ R89, and β′ D237. Two independent reports of FDX MIC were obtained for isolates that harbor mutation at β′ R89 in which nucleotide change at A265G was reported with FDX MIC of 4 mg/L and A260G with MIC of 64 mg/L.

The structural details of the FDX-RNAP complex were visualized for wild type and mutated strains. Some residues were hidden to gain better visualization on the interactions of the residues of interest. Figure 2 shows that β′ R89 formed a hydrogen bond with β′ K86 and salt bridge interaction with two neighboring residues β E1139 and β′ D237. All three residues were FDX-interacting residues in C. difficile RNAP in which they participated in either salt bridge interaction or hydrogen bond. However, when the residue was mutated to glycine, which was a mutation previously observed in a clinical study (5), the salt bridge interaction with β E1139 and β′ D237 was both lost. A similar scenario was observed for V1143. Valine at this position maintained good contact with FDX and formed hydrogen bonds with neighboring residues, including β E1139, β S1140, and β E1147. Both β E1139 and β S1140 belonged to the clamp region of C. difficile RNAP and interacted with FDX via van der Waals/hydrophobic interactions. Mutating valine to phenylalanine induced steric clashes with FDX and the interacting residue R326 of the β′ SW2. On the other hand, the substitution of valine with aspartic acid at position 1143 resulted in the formation of salt bridge interaction between mutated residue D1143 with β′ R326.

Fig 2.

Interaction between FDX and RNAP residues. (A) Comparison between FDX-RNAP interaction when ARG89GLY mutation occurred in C. difficile RNAP. ARG89 formed a hydrogen bond (yellow dash) with LYS86 and salt bridge interaction (pink dash) with GLU1139 and ASP237. Mutation to GLY89 resulted in the loss of salt bridge interaction with GLU1139 and ASP237. (B) Comparison between FDX-RNAP interactions when mutation occurred at position VAL1143. At this position, valine formed hydrogen bonds with GLU1139, SER1140, and GLU1147. Mutation to PHE1143 resulted in steric clashes and bad/ugly contacts (orange/red dash) with FDX and ARG326. Mutation to ASP1143 resulted in the formation of salt bridge interaction between the oxygen of ASP1143 and nitrogen of ARG326.

The total predicted changes in protein stability upon mutations (∆∆G) in rpoB and rpoC at the key FDX-RNAP binding residues and common identified mutation were quantified using the MAESTRO webserver. The mutation sensitivity profile was created for all possible substitutions at each interacting residue and common mutated residue location (Fig. 3; Table S1). As a result, the box plot depicted the distribution of predicted changes in protein stability values when substituted with each amino acid, and the impact of all possible mutations at each position was visualized. The results showed mutations at β Q1074K (∆∆G = 0.085 kcal/mol, C_pred = 0.86), β′ R89G (∆∆G = 0.03 kcal/mol, C_pred = 0.86), and β′ D237Y (∆∆G = 0.05 kcal/mol, C_pred = 0.86) were destabilizing mutations (∆∆G_pred > 0), while β V1143G (∆∆G_pred = −0.08 kcal/mol, C_pred = 0.86), β V1143D (∆∆G_pred = −0.11 kcal/mol, C_pred = 0.85), β V1143F (∆∆G_pred = −0.24 kcal/mol, C_pred = 0.83), and β′ R326C (∆∆G_pred = −0.05 kcal/mol, C_pred = 0.86) were stabilizing mutations (∆∆G_pred < 0).

Fig 3.

Total predicted change in stability upon mutation in rpoB (ΔΔG_pred). (A) Scanning for (de)stabilizing mutations at FDX-RNAP key binding residues in rpoB (L1071, V1072, T1073, Q1074, D1114, V1116, V1117, V1120, R1121, E1139, and S1140) and commonly identified mutation V1143. (B) Scanning for de(stabilizing) mutations at FDX-RNAP key binding residues in rpoC (K84, S85, K86, R89, D237, L238, P240, S252, K314, M319, and R326). ΔΔG_pred < 0.0 indicated a stabilizing mutation.

DISCUSSION

FDX use has increased due to changes in guideline recommendation placing FDX as first-line therapy. This has resulted in antimicrobial selection pressure increasing the likelihood of the emergence of FDX non-susceptible isolates. C. difficile strains with elevated FDX MIC have been reported in both laboratory and clinical settings; however, the mechanism of FDX resistance and implications for patient outcomes are still under investigation. A recent study elucidated the mechanism of action of FDX and identified the key residues that are critical to FDX-RNAP binding. In this review, we demonstrated that most isolates with FDX elevated MIC harbored mutations in positions at or within 5 Å of the FDX-RNAP binding site, including β R1121G, β V1143D, β V1143F, β V1143G, β V1143L, β′ D237Y, and β Q1074K. The most common mutation identified was V1143 in which the amino acid valine was susceptible to nonsynonymous substitutions with either aspartic acid, glycine, phenylalanine, and leucine that resulted in changes to the FDX MIC. The structural details of the FDX-RNAP complex provided insights into the binding pocket of RNAP at this position, which revealed the formation of salt bridge interaction between the mutated residue and neighboring residue β′ R326 or the induction of steric clashes when substituted with phenylalanine. In the case of β′ R89, mutation to glycine resulted in the loss of two salt bridge interactions, which originally arose from the interaction between the cationic ammonium of arginine and the anionic carboxylate of aspartic acid and glutamic acid.

Point mutations have been studied extensively due to their contribution to the fitness of the proteins in terms of protein folding and folding kinetics. Substitution of valine at position 1143 in rpoB, which is the most common mutation observed in both clinical and laboratory settings, resulted in a wide range of FDX MIC (2 to >64 mg/L). V1143 locates only three amino acids away from the key FDX interacting residues of the RNAP clamp β S1140, and the substitution of V1143D led to the formation of a salt bridge between β D1143 and β′ R326, a key binding residue of FDX and C. difficile RNAP. This observation may explain the substantial effect that this mutation had on FDX MIC, as salt bridge interaction has been known to participate in molecular recognition, protein interaction, protein folding, and protein stability (13, 14). The distance between the negatively charged oxygen atom of D1143 and the positively charged nitrogen atom of R326 was 3.47 Å, which did not exceed the permissible distance of 4 Å for salt bridge length, indicating a stable salt bridge (13, 14).

The overview on Gibbs free energy prediction provided some insights into the destabilizing effects of mutations at each interacting and common mutated residue in rpoB and rpoC. Interestingly, C. difficile RNAP residues known to confer FDX resistance when mutated, such as β V1143 and β′ R326, did not result in a defect in the protein stability. Wild-type residue was often predicted to be more stable than the mutated residues because the new amino acid side chain might not pack as well as the wild-type residues. However, the analyses of protein stability are greatly complicated due to the protein sequence-structure-energy relationship and many other biological factors, such as pH of the microenvironment, or technical factors, such as the accuracy of the computational tools. This observation came solely from one prediction tool. Other studies using the MAESTROweb server have demonstrated a decrease in protein stability at a pH of 8–8.5. Therefore, further work is necessary to elucidate the effects of disruption of interatomic interaction due to single point mutation on the enzyme activity as the mechanism appeared to be multifactorial.

One identified mutation located in the marR/CD22120 gene caused a frameshift after amino acid 117 and resulted in a twofold increase in FDX MIC when compared with the parent strain. Resistance to FDX also arose in a laboratory-generated mutant due to a frameshift mutation occurring after amino acid 117 in a homolog of the multidrug resistance-associated transcriptional regulator MarR (CD22120). MarR functions as a transcriptional repressor of genes encoding efflux pumps, and its inactivation results in the upregulation of efflux mechanisms of antimicrobial resistance. The deletion identified in the laboratory mutant has also been detected in the genomes of eight clinical isolates from five different countries. This observation suggests a potential link between the loss of MarR, and the development of low-level resistance to fidaxomicin requires further study.

In this review, only nine different mutations were identified across 18 strains (including double mutations). Point mutations with elevated FDX MIC have also been explored in a recent study by Kolte et al. (15) on 26,557 C. difficile genome sequences on uncharacterized strains using EnteroBase. Eight and four genomes were detected to confer V1143D and V1143G mutations in rpoB, respectively, while only one genome presented mutation at position R89G in rpoC. 3 × 10⁻⁴ sequences were associated with ∆T349 in marR, which was a mutation detected earlier from mutants created by serial passages. The slow emergence of those mutations may be due to slow uptake of the drug clinically or there may be a bacterial fitness cost associated with isolates harboring reduced FDX susceptibility. A study on C. difficile isolates with mutation at V1143 demonstrated a fitness cost associated with bacterial growth, sporulation, and toxin production capacities. Mutated isolates seemed to decline more rapidly and were more likely to be outcompeted by the parent strains, which may explain the reason why resistant isolates are less likely to be seen in clinical settings. However, this review highlights that C. difficile strains with reduced FDX susceptibility can be isolated from CDI patients. This is best highlighted by the phase III FDX clinical study in Japan that showed FDX-resistant isolates emerged after patients had been exposed to FDX in which six isolates demonstrated a 30- to 2,000-fold reduction in FDX susceptibility. This observation confirmed that resistance to FDX is multifactorial, and further studies are needed to explore it.

In this systematic review, mutation at or close to the FDX binding site influenced FDX MIC to C. difficile phenotype and provides evidence that genotyping models may be able to predict FDX resistance. Several computational tools have been developed to predict AMR phenotypes from whole-genome sequencing (WGS) data, including XGboost, a machine learning-based MIC prediction model originally used for Salmonella. Since most strains with elevated FDX MIC conferred a mutation either at the key FDX interacting residues or located within 5 Å of the interacting residues, this observation highlights the potential to develop a similar model for C. difficile and FDX susceptibility. The ability to predict the emergence of resistant strains would be greatly beneficial for antimicrobial stewardship decisions and public health surveillance.

MATERIALS AND METHODS

A systematic review literature search was done in PubMed (1991–2023) with identified articles and their bibliographies searched for papers that included C. difficile genetic mutations and increased FDX MIC.

The systematic literature search was conducted in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses. A comprehensive search was performed using the following keywords in PubMed: (“Clostridioides difficile” OR “Clostridium difficile”) AND “fidaxomicin” AND “antimicrobial resistance.” Filters for full text and English-language studies were applied. Published abstracts were excluded if the full article was also published to avoid redundancy. The bibliographies of identified articles were searched for additional research articles. Included articles studied C. difficile isolates and contained information on SNPs that were associated with a change in FDX minimum inhibitory concentration. Mutations were further categorized as derived from laboratory mutants or clinical isolates. Data collected included geographic location, year of publication, types of studies, strain types and reference strains for WGS, number of isolates tested or obtained, nucleotide changes, amino acid change, susceptibility testing methodology, and FDX MIC of the parent strains and the mutated strains.

Schrödinger Maestro version 14.0 was utilized to visualize and gain novel molecular insights on protein-ligand interactions, using the deposited Clostridioides difficile RNAP with FDX structure on Protein Data Bank (Code 7L7B). The structure was imported and prepared using the Schrödinger Maestro Protein Preparation Workflow to fill in missing side chains, optimize H-bond assignments, and perform restrained minimization. The preprocess steps fixed any detected structural defects and added missing information. Hydrogen assignments and bonding network were then optimized to ensure the correct orientations of side chains (Asn, Gln, His) at the specified pH value. Finally, the protein structure underwent restrained minimization to fix clashes that occurred when hydrogen bonds or missing sidechains were added. Several valence errors were found due to missing H or an incorrect number of bonds, which were corrected during preprocess steps. The structural visualization was generated with a pH value of 8 in accordance with pH value that was used to prepare the crystal structure assay, and only water molecules within 5 Å of the ligands were included in the analysis.

Predicted Gibbs free energy change (∆∆G_pred in kcal/mol) values and the corresponding prediction confidence estimation (C_pred) were calculated using a Multi AgEnt Stability pRedictiOn tool (MAESTROweb) (https://pbwww.services.came.sbg.ac.at/maestro/web). A positive value of ∆∆G (∆∆G > 0) depicted a destabilizing mutation, while a negative value (∆∆G_pred < 0) indicated the protein was stable despite the mutated residue. C_pred was provided as confidence estimation for ∆∆G prediction only. The confidence estimation was confined to a value between 0.0 and 1.0, where 0.0 indicated an unreliable prediction and 1.0 corresponded to a highly reliable prediction. Nucleotide changes were obtained from the reference sequence NC_013316 on the National Center for Biotechnology Information.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.01206-24.

Table S1. aac.01206-24-s0001.docx.

Mutation sensitivity profile for possible substitutions at each interacting residue and common mutated residue location.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.