Detection and Genomic Characterisation of Clostridioides difficile from Spinach Fields

Despite an increased incidence of Clostridioides difficile infections, data on the reservoirs and dissemination routes of this bacterium are limited. This study examined the prevalence and characteristics of C. difficile isolates in spinach fields. C. difficile was detected in 2/60 (3.3%) of spinach and 6/60 (10%) of soil samples using culture-based techniques. Whole genome sequencing (WGS) analysis identified the spinach isolates as belonging to the hypervirulent clade 5, sequence type (ST) 11, ribotypes (RT) 078 and 126 and carried the genes encoding toxins A, B and CDT. The soil isolates belonged to clade 1 with different toxigenic ST/RT (ST19/RT614, ST12/RT003, ST46/RT087, ST16/RT050, ST49/RT014/0) strains and one non-toxigenic ST79/RT511 strain. Antimicrobial resistance to erythromycin (one spinach isolate), rifampicin (two soil isolates), clindamycin (one soil isolate), both moxifloxacin and rifampicin (one soil isolate), and multi-drug resistance to erythromycin, vancomycin and rifampicin (two soil isolates) were observed using the E test, although a broader range of resistance genes were detected using WGS. Although the sample size was limited, our results demonstrate the presence of C. difficile in horticulture and provide further evidence that there are multiple sources and dissemination routes for these bacteria.


Introduction
Clostridioides difficile is an enteric spore-forming and toxigenic pathogen that was historically classified as a hospital-acquired infection [1]. Common symptoms range from watery non-bloody diarrhoea with abdominal pain to life-threatening fulminant colitis [2]. However, the steady increase in community acquired C. difficile infection(s) (CDI) in recent years has motivated investigative research on other sources and dissemination routes. As a result, specific C. difficile ribotypes, such as RT078, are now being classified as communityacquired [3,4].
Virulence in C. difficile is mainly determined by the production of toxin A (tcdA) and B (tcdB), which alter the host's gut by causing damage to the epithelial barrier, leading to the translocation of commensal bacteria and cell death [5][6][7]. In addition, accessory genes tcdR and tcdC are part of the pathogenicity locus (PaLoc) and play a key role in regulating toxin production [8,9]. Apart from toxins A and B, it is estimated that up to 30% of C. difficile strains can produce the transferase C. difficile binary toxin (CDT), which belongs to the binary toxin family and is generally associated with hypervirulent strains [10][11][12].
Epidemiological investigations of C. difficile cases often use ribotyping that differentiates strains based on polymorphisms in the 16S-23S rRNA intergenic spacer region [13]. Certain ribotypes are more often linked with a higher occurrence and severity of disease in humans and some of these have been isolated from the food chain. Multi-drug resistant RT027, for example, is often isolated from clinical samples and is frequently associated with more severe illnesses and has been isolated from different food animals [14][15][16]. Moreover, RT014/020 and RT001/072 are confirmed endemic ribotypes in Europe [17].

C. difficile Isolation and Confirmation
Soil and spinach samples were tested for C. difficile by adding 90 mL of maximum recovery diluent (MRD) to 10 g of each sample in a stomacher bag and blending for 1 min in a Star Blender LB 400 Stomacher (VWR, Lutterworth, Leicestershire, UK). To ensure the selection of C. difficile spores, vegetative cells were eliminated by heating at 60 °C for 25 min. To encourage spore germination, 10 mL of the heat-treated mixture was added to 90 mL of brain heart infusion broth (BHI) (Oxoid, Basingstoke, Hampshire, UK (CM1135B)) with 0.1% sodium taurocholate (Sigma-Aldrich, Gillingham, UK (86339-25G)) and selective C. difficile supplements (containing 8 mg/l of cefoxitin and 250 mg/l of D-cycloserine) (Oxoid, Basingstoke, Hampshire, UK (SR0096E)). Samples were incubated at 37°C for 3-5 days under anaerobic conditions in an A35 Anaerobic workstation (Don Whitley, Victoria Works, West Yorkshire, UK) [38].
DNA was extracted from isolated colonies with the typical C. difficile morphology (irregular ground-glass colonies) using the DNeasy blood & tissue kit (QIAGEN GmbH, Hilden, Germany (69504)) and was confirmed as C. difficile by testing for the tpi gene using PCR [39]. Briefly, 12.5 µL of Mastermix (QIAGEN Ltd., Manchester, UK (206143)), 0.625 µL of each primer (0.5 µM), 9.25 µL of nuclease-free water (Invitrogen, Biosciences Ltd., Dún Laoghaire, Dublin, Ireland (LSKNF0500)) and 2 µL of template were mixed together and amplified using a Veriti 96-Well Thermal Cycler (Applied Biosystems, Warrington, Cheshire, UK). The PCR cycle started with denaturation for 3 min at 95 • C, followed by 40 cycles of 30 s at 95 • C, 30 s at 55 • C and 30 s at 72 • C, and a final extension of 30 s at 72 • C. Electrophoresis in a 2% agarose gel stained with SYBR™ Safe DNA Gel Stain (Biosciences Ltd., Dún Laoghaire, Dublin, Ireland (S33102)) was then used to confirm the tpi PCR fragment generated (230-bp) under UV light. C. difficile RT078, obtained from a clinical isolate at St. James's Hospital (Dublin, Ireland) and supplied by Prof. Thomas Rogers from Trinity College Dublin, and C. sporogenes, available in the Teagasc Ashtown Food Research Centre collection, were used as the positive and negative controls, respectively.

Characterization of Toxin and Accessory Genes
The confirmed C. difficile isolates obtained were further characterised for toxin (tcdA, tcdB, cdtA and cdtB) and accessory genes (tcdC and tcdR) by conventional PCR. Primer sequences, concentration and amplification protocols for toxin A (tcdA), toxin B (tcdB), toxin CDT (cdtA and cdtB) and accessory genes (tcdC and tcdR) were as described by Marcos et al. [24]. The total PCR mixture for every reaction had a final volume of 25 µL with the corresponding volume for each primer depending on its concentration, 12.5 µL of Mastermix (QIAGEN), 2 µL of template and nuclease-free water to reach the total volume. As previously described, a 2% agarose gel stained with SYBR™ Safe DNA Gel Stain (Biosciences) was carried out to separate the PCR fragments: 369-bp for tcdA, 160-bp for tcdB, 375-bp for cdtA, 510-bp for cdtB, 718 bp for tcdC and 300 bp for tcdR. C. difficile RT078 and C. sporogenes were the corresponding positive and negative controls, respectively.

Antimicrobial Susceptibility Testing
The susceptibility of isolates to a range of relevant antibiotics (vancomycin, erythromycin, metronidazole, clindamycin, moxifloxacin and rifampicin) was tested using the E-test strips (bioMérieux, Marcy-l'Étoile, France), following the protocol described by the manufacturer. The aforementioned antibiotics were selected for testing due to their common use in CDI treatment [40][41][42][43][44].
Isolates were cultured in Mueller-Hinton broth (Oxoid, Basingstoke, Hampshire, UK (CM0405)) at 37 • C for 24 h or until an OD 600 nm = 0.5 (10 8 CFU/mL) was achieved, measured using a DeNovix DS-C spectrophotometer (DeNovix Inc., Wilmington, NC, USA). A sterile swab (Sparks Lab Supplies, Dublin, Ireland (SW001)) was dipped into the broth solution and then spread on Brucella agar plates with vitamin K and haemin (Sigma-Aldrich, Gillingham, UK (B2926-500G)) and 5% defibrinated horse blood (TCS Biosciences Limited, Botolph Claydon, Buckingham, UK (HB034)). Plates were allowed to dry for 15 min before the E-test strip was placed on top of the inoculated agar with a sterile forceps and incubated for 48 h at 37 • C in anaerobic conditions as previously described. After incubation, MIC values were read using the scale (µg/mL) provided by the manufacturer. Values obtained for each antibiotic were compared to the breakpoint values described in EUCAST [45] and classified as susceptible or resistant according to the criteria.

PCR-Ribotyping
Ribotype testing of all isolates obtained was carried out following the PCR ribotyping protocol by ECDC [46], with minor changes. Amplification of the 16S and 23S rRNA genes was undertaken to identify the specific intergenic spacer region (ISR). Primers for the amplification of the C. difficile 16S (forward primer) and 23S (reverse primer) rRNA genes were as described by Bidet et al. [47]  After the PCR, denaturation of the generated fragments was necessary before analysis in an automated sequence and fragment analysis system was undertaken. Exactly 2 µL of the PCR product was added to 9.5 µL of Highly Deionized (Hi-Di) Formamide (Applied Biosystems, Warrington, Cheshire, UK) and 0.5 µL of GeneScan 1200 LIZ Size Standard (Applied Biosystems, Warrington, Cheshire, UK). The mix was denatured for 2 min at 95 • C in a thermal cycler, followed by cooling of the plate for 10 min in a fridge before the PCR products were analysed on an ABI 3500 Genetic Analyzer (Applied Biosystems, Warrington, Cheshire, UK) with default settings for POP7 and 50 cm capillary length. The raw data files (*fsa files) from the ABI 3500 Genetic Analyzer were uploaded to the free-touse website WEBRIBO (https://webribo.ages.at/; accessed on 7 July 2022) to compare our isolates with existing ribotypes stored on the database [13].

Whole Genome Sequencing (WGS)
The C. difficile isolates (n = 8) obtained from 2 spinach leaves and 6 soil samples were selected for WGS in order to identify genetic determinants of virulence and resistance. WGS was performed in Teagasc Food Research Centre Moorepark by Dr. Fiona Crispie and Dr. Gaston Allende. DNA libraries were prepared with the Illumina DNA preparation kit as per the manufacturer's instructions. This was followed by 2 × 150 bp sequencing on the NextSeq 2000 with P2 reagents. Nucleotide sequence data reported are available in the GenBank database under the accession number PRJNA884639.

Characterisation of the C. difficile Isolates Using Conventional PCR, E-Test and PCR-Ribotyping
A total of 8 (two spinach and six soil) out of 120 samples analysed (6.7%) were confirmed as C. difficile positive ( Figure 2). In Field 1, two spinach and one soil samples were positive, while one positive soil sample was obtained in Field 2 ( Figure 2). Half of the positive samples (four soils) were obtained in Field 3.

Characterisation of the C. difficile Isolates Using Conventional PCR, E-test and PCR-Ribotyping
A total of 8 (two spinach and six soil) out of 120 samples analysed (6.7%) were confirmed as C. difficile positive ( Figure 2). In Field 1, two spinach and one soil samples were positive, while one positive soil sample was obtained in Field 2 ( Figure 2). Half of the positive samples (four soils) were obtained in Field 3. The presence or absence of toxin and accessory genes in the C. difficile isolates is shown in Table 1. Toxin-encoding genes were detected by PCR in seven of the eight isolates. The tcdA and tcdB genes were detected in five out of the six soil isolates, while tcdB, cdtA and cdtB were present in the two spinach isolates. Accessory genes (tcdC and tcdR) were detected in all isolates except in isolate 7, which also lacked the PaLoc.
The resistance or susceptibility of the C. difficile isolates to the six antibiotics tested and their corresponding ribotype is also shown in Table 1. All isolates, except for one spinach isolate (ribotype 078) from Field 1, were resistant to at least one antibiotic. Resistance to rifampicin was observed in 5/8 of the isolates, followed by erythromycin (3/8), vancomycin (2/8), clindamycin and moxifloxacin (1/8 each). All isolates were susceptible to metronidazole. Two soil isolates from Fields 2 (RT003) and 3 (RT050) were resistant to the same three antibiotics (erythromycin, vancomycin and rifampicin). The rest of the isolates presented varied resistance and ribotype patterns. The presence or absence of toxin and accessory genes in the C. difficile isolates is shown in Table 1. Toxin-encoding genes were detected by PCR in seven of the eight isolates. The tcdA and tcdB genes were detected in five out of the six soil isolates, while tcdB, cdtA and cdtB were present in the two spinach isolates. Accessory genes (tcdC and tcdR) were detected in all isolates except in isolate 7, which also lacked the PaLoc.
The presence of toxin-encoding genes, sequence type (ST) and clade of the spinach (isolates 2 and 3) and soil (isolates 1,4,5,6,7,8) isolates are shown in Table 2. The detection of toxin-encoding genes by WGS analysis was in broad agreement with the results obtained by PCR, with the exception of tcdA in the spinach isolates. Isolates 2 and 3 (both spinach isolates from Field 1) belonged to clade 5 and ST11. The rest of the isolates (obtained from soil from Fields 1, 2 and 3) belonged to clade 1 and had different STs (19, 12, 46, 16, 79 and 49). Antibiotic resistance genes detected in the C. difficile isolates via ResFinder, Prokka and CARD software tools are shown in Table 3. ResFinder and Prokka analysis detected that isolates 2 and 3 (spinach from Field 1) carried resistance genes for aminoglycosides (D19aph(3')-III_1, ant(6)-Ia_1, ant(6)-Ia_2, ant(6)-Ia_3) and tetracyclines (tet(M)_10, tet(40)_1, tet(M)_4) and all soil isolates, except for 7 and 8, had the vancomycin resistance gene (vanB). Genes encoding antibiotic resistance were not detected in isolates 7 and 8 (soil from Field 3) using ABRicate and Prokka. However, according to the analysis of the isolates using Resistance Gene Identifier (RGI) and CARD, all isolates encoded CDD-1 or CCD-2 (carbapenem resistance), qacG (disinfecting agents and antiseptics resistance) and a mutation in the 23S rRNA that confers the resistance to erythromycin and clindamycin. A spinach isolate (isolate 3) had resistance to nucleoside antibiotics (SAT-4), and soil isolates (1, 4, 5, 6, 7 and 8) carried resistance to glycopeptide antibiotics (vanXY and vanR in the vanG cluster). Isolate 5 (soil from Field 3) also encoded the cdeA gene, which confers resistance against fluoroquinolones.
A further comparison between the antimicrobial susceptibility profiles found by E-test and CARD is presented in Table 4.
The pangenome analysis carried out for the 8 C. difficile isolates is shown in Figure 3. The core genome consisted of 3010 gene clusters and 25,718 genes. In the total pan-genome comparison, 6381 gene clusters were detected with 33,243 genes identified. Two distinct clusters were observed: the first cluster included isolates 2 and 3 (spinach), while the second cluster included the soil isolates (1,(4)(5)(6)(7)(8).  In addition, PlasmidFinder detected the genes of the replicon repUS43_1_CDS12738(DOp1) in the spinach isolates 2 and 3 (99.83% and 100% identity), respectively.
The pangenome analysis carried out for the 8 C. difficile isolates is shown in Figure 3. The core genome consisted of 3,010 gene clusters and 25,718 genes. In the total pan-genome comparison, 6381 gene clusters were detected with 33,243 genes identified. Two distinct clusters were observed: the first cluster included isolates 2 and 3 (spinach), while the second cluster included the soil isolates (1, 4-8). In addition, Figure 4 shows the phylogenetic relationship and distance between the C. difficile isolates. Isolates 2 and 3 (spinach isolates) had the closest distance (19), followed by soil isolates 4 and 6 (9431), and 4 and 7 (9516). Spinach isolates (2 and 3) presented a high distance from all of the soil isolates (>96,008). In addition, Figure 4 shows the phylogenetic relationship and distance between the C. difficile isolates. Isolates 2 and 3 (spinach isolates) had the closest distance (19), followed by soil isolates 4 and 6 (9431), and 4 and 7 (9516). Spinach isolates (2 and 3) presented a high distance from all of the soil isolates (>96,008).

Ribotypes of C. difficile
In our study, ribotypes 078 and 126 were detected on spinach leaves while RT614, 003, 087, 050, 511 and 014/0 were found in the soil. The majority of these ribotypes have been previously linked to animal, food and human sources.
Ribotype 078 is frequently reported in farm animals such as cattle [76][77][78], poultry [79,80] and pigs [81][82][83], with no distinguishable differences between human and porcine RT078 isolates [84,85]. It has also been found in food products derived from animals purchased at retail such as ground pork, turkey [86] and beef [87]. In vegetables, it was detected in ginger and carrots from Canada [33] and in potatoes from Ireland and Italy [28]. In humans, hypervirulent C. difficile RT078 accounted for 17% of ribotyped isolates in 2021 in Ireland [88], with similar data reported for Germany (16.7%) by Rabold et al. [89]. Interestingly, the RT078 strain isolated in our study was identified as similar to C. difficile strain M120 by WGS, a strain that has been associated with clinical infection in the UK and Ireland [19], as well as in calves in Canada [90] and the pig farm environment in Spain [91].  (1,(4)(5)(6)(7)(8) with the distances between isolates presented as a distance matrix.

Ribotypes of C. difficile
In our study, ribotypes 078 and 126 were detected on spinach leaves while RT614, 003, 087, 050, 511 and 014/0 were found in the soil. The majority of these ribotypes have been previously linked to animal, food and human sources.
Ribotype 078 is frequently reported in farm animals such as cattle [76][77][78], poultry [79,80] and pigs [81][82][83], with no distinguishable differences between human and porcine RT078 isolates [84,85]. It has also been found in food products derived from animals purchased at retail such as ground pork, turkey [86] and beef [87]. In vegetables, it was detected in ginger and carrots from Canada [33] and in potatoes from Ireland and Italy [28]. In humans, hypervirulent C. difficile RT078 accounted for 17% of ribotyped isolates in 2021 in Ireland [88], with similar data reported for Germany (16.7%) by Rabold et al. [89]. Interestingly, the RT078 strain isolated in our study was identified as similar to C. difficile strain M120 by WGS, a strain that has been associated with clinical infection in the UK and Ireland [19], as well as in calves in Canada [90] and the pig farm environment in Spain [91]. Ribotype 126 has been described in cattle [76,92,93] and pigs [94,95]. It may not be a coincidence that both ribotypes RT078 and RT126, which are frequently reported in cattle, were detected in the spinach isolates from Field 1 that was located adjacent to fields with bovine animals. Apart from farm animals, Agnoletti et al. [96] and Troiano et al. [97] isolated RT126 from shellfish. Tkalec et al. [28] confirmed the presence of ribotype 126 in potatoes from France, Spain, Austria and Romania, and Primavilla et al. [34] found this ribotype in lettuce served in Italian hospitals. In Spain, RT126 was one the most common ribotypes among clinical isolates [98], while in Portugal, it accounted for 3.8% of the CDI cases [99]. Recently, Azimirad et al. [100] identified RT126 as the predominant ribotype (11.2%) in hospital patients in Tehran. Hypervirulent RT126 was observed in clinical isolates in Taiwan by Hung et al. [101], in Kuwait by Jamal and Rotimi [102] and in Australia by Knight et al. [82]. The ribotype 126 detected in spinach leaves in our study was similar to C. difficile strain DSM 29020, which has been reported in hospital patients in Korea and China [103,104].
Of our soil isolates, ribotype 003 was previously reported in poultry faeces (1%) [105] and meat (28.6%) [106], calves [76] and horses (5.6%) [107]. Indeed, the close proximity of the field where it was isolated to a stud farm might highlight the potential for its zoonotic transfer. Regarding its link to humans, in Germany, RT003 accounted for 11.1% of CDI cases in 2012-2013 [89], and strain W0023a, which is similar to our isolate by WGS, has been detected in clinical isolates from Korea and the USA [103,108].
RT087, the ribotype of another of our soil isolates, was identified in poultry meat by de Boer et al. [106] and Tkalec et al. [109] (14.3 and 20%, respectively). The latter authors also reported that ribotype 087 was more common in clinical isolates than RT078 in a Slovenian study.
To the best of our knowledge, this is the first time RT050 has been reported in soil samples, although it has been reported in pigs (12%) [110] and was recently associated with hospitalized adult patients in Mexico [111] and in a CDI outbreak in the intensive care unit of the Amsterdam University Medical Centre [112]. Furthermore, both RT087 and RT050 were similar to C. difficile CFSAN096664, which was previously isolated from hospital patients in the USA [113].
Ribotype 014, previously described in retail ground beef (8.3%) [114], was found in soil in Field 3 that was located next to a dairy farm. A variant of this ribotype, RT014/020, was the predominant ribotype among soil isolates obtained from domestic gardens in Australia by Shivaperumal et al. [26]. In humans, a study from Germany noted RT014/0 as the most prevalent ribotype among CDI patients (22.2%) [89], and our WGS analysis suggested that this isolate was similar to C. difficile DSM 105001, which has been reported before in US hospital patients [115].
Isolates of RT614 have not been previously reported in soil but have been found in clinical samples in Portugal [116]. This isolate was similar to DSM 27639, which has been reported in German hospital patients [117].
Our remaining soil isolate was ribotype 511 and has not been associated with animal, food or human sources. However, due to the ribotype's lack of toxin genes, it may be carried asymptomatically [118] or be associated with hospital-acquired diarrhoea [119]. WGS analysis suggested that RT511 was most closely related to strain C. difficile W0003a, previously observed in clinical isolates in Korea [103].

Toxigenicity of the C. difficile Isolates
The C. difficile population consists of six distinct phylogenetic clades labelled 1, 2, 3, 4, 5, and C-I [120,121]. Most genotypes within clades 1-3 produce both toxins A and B and, in addition to these, strains from clades 2, 3, and 5 have genes encoding the binary toxin, CDT. Clade 5 includes ribotype 078, which is typically associated with community-acquired cases and has a higher rate of complications and mortality in humans [3,110,122]. In the present study, spinach isolates (isolates 2 and 3) both belonged to this highly toxigenic clade 5 (ST11) and contained the binary toxin CDT, characteristic of hypervirulent strains [10], apart from toxins A and B. Interestingly, while conventional PCR yielded a negative result for the tcdA gene in the spinach isolates, this gene was detected using WGS. Thus, SNP analysis was carried out to search for possible mutations in the primer region used to amplify tcdA that could have affected the PCR analysis and might explain the negative result. However, no SNPs were detected in the primer or promoter region. The use of a positive tcdA control (C. difficile RT078) and a negative control belonging to C. sporogenes in the conventional PCR analysis ruled out issues with the PCR reaction (data not shown), and thus, further research is required to explain this apparent anomaly.
Pangenome and phylogenetic analysis showed the existence of two clusters among our samples. The first cluster included the spinach samples belonging to clade 5 and the second cluster comprised the soil samples that belonged to clade 1. Thus, the soil samples were closely related to each other but were phylogenetically distinct from the spinach isolates (who were, in turn, closely related). However, due to the limited sample size of this study, further studies would need to investigate this potential relationship considering a larger number of isolates.

Antibiotic Susceptibility of the C. difficile Isolates
An interesting finding in our study was the resistance differences observed in antibiotics tested phenotypically and genotypically, supporting previous similar observations in C. difficile by Muñoz et al. [123].
In spinach isolates, similar susceptibilities to those observed in this study in RT078 against all antibiotics tested were described previously in isolates from humans by Álvarez-Pérez et al. [98], rabbits by Drigo et al. [124] and pigs by Zhang et al. [83]. Freeman et al. [14] observed vancomycin resistance in RT078 isolates while we observed susceptibility, although the lack of vancomycin resistance genes we reported in this ribotype would explain our results. In addition, Zhang et al. [83] found resistance to tetracycline in RT078 isolates from pigs in China.
In RT126, phenotypic resistance to erythromycin had been previously reported by Álvarez-Pérez et al. [98] and can be explained by the existence of a mutation in the target site of C. difficile 23S rRNA that confers resistance against the antibiotic. Furthermore, the gene SAT-4, which encodes nucleoside antibiotic resistance and was identified in RT126 in our study, was detected in Campylobacter coli and Enterococcus faecium by other authors [125,126].
In all soil isolates (RT614, RT003, RT087, RT050, RT511, RT014/0), vancomycin resistance genes (vanXY and vanR) were detected, even though phenotypic resistance was only observed in isolates of RT003 and RT050, contrary to the susceptibility reported by Kecerova et al. [107] in RT003. These results support the findings of Suzuki et al. [128] in C. difficile RT027 strains that carried vanRG and vanG genes but were phenotypically susceptible to vancomycin. Therefore, the presence of vancomycin resistance genes in C. difficile does not always result in their expression in vitro.
Phenotypic resistance to clindamycin in RT014/0, previously reported in human and pig isolates in Australia [129,130] and to erythromycin in RT003 and RT050, is probably due to the mutation detected in the antibiotic target site of C. difficile 23S rRNA, which confers resistance to both antibiotics. The resistance of C. difficile strains to erythromycin and clindamycin has been widely reported in human isolates and treatment with clindamycin is a risk factor for developing CDI [18,131].
Additionally, the in vitro resistance to moxifloxacin in RT087, in contrast to the results reported by Freeman et al. [14] for this ribotype, could be explained by the presence of the cdeA gene. This gene confers resistance against fluoroquinolones in E. coli and has been previously detected in C. difficile [132].
To the best of our knowledge, no previous authors have described antimicrobial resistance patterns for RT614 and RT511, possibly due to a lower frequency of reports in humans, even though they both showed phenotypic resistance to rifampicin in our study and have been associated with an increased severity of CDI [133].
The CDD-1 and CDD-2 genes, detected in all of the isolates analysed in this study, confer a high resistance against a broad range of beta-lactams [134]. While beta-lactam antibiotics were not tested by E-test in this study, these enzymes previously reported in C. difficile allow the bacteria to have intrinsic resistance to antibiotics such as penicillin, monobactams and cephalosporins [135].
Interestingly, the qacG gene present in every isolate was identified previously in staphylococci isolated from the food industry, conferring resistance to benzalkonium chloride that is a frequently used quaternary ammonium disinfectant [136]. Resistance to quaternary ammonium compounds is plasmid-borne in Gram-positive bacteria, linked to the Small Multi-drug Resistance (SMR) family transporters (qacC/D and qacE∆1, qacG, qacH and qacJ) [137]. While the qacG gene has not been reported in C. difficile before, in 2013, it was detected in a S. aureus MRSA strain isolated from pork [138]. However, Seier-Petersen et al. [139] described no reduced susceptibility in vitro to biocides in S. aureus MRSA strains where the qacG gene was present. More recently, the gene was found in the Staphylococcus species in Turkey [140] and carbapenem-resistant A. baumannii in China [141].

Conclusions
It was concluded that the prevalence of C. difficile was 10% in soil and 3.3% in spinach, both of which were lower than the results reported in other similar studies. Both spinach isolates (toxigenic RT078 and RT126) carried virulence genes (toxins A, B and CDT) and RT126 was phenotypically resistant to erythromycin. Soil isolates included toxigenic ribotypes (003, 014/0, 050, 087 and RT614) and a novel non-toxigenic ribotype (RT511). RT003 and RT050 were resistant to erythromycin, vancomycin and rifampicin, RT087 to moxifloxacin and rifampicin, and RT014/0 to clindamycin. WGS suggested there were inconsistencies between AMR phenotypes and genotypes. Moreover, the soil isolates were closely related but genetically distinct from the spinach strains.