A retrospective study of community-acquired Clostridium difficile infection in southwest China

To identify the prevalence and characteristics of community-acquired Clostridium difficile infection (CA-CDI) in southwest China, we conducted a cross-sectional study. 978 diarrhea patients were enrolled and stool specimens’ DNA was screened for virulence genes. Bacterial culture was performed and isolates were characterized by PCR ribotyping and multilocus sequence typing. Toxin genes tcdA and/or tcdB were found in 138/978 (14.11%) cases for fecal samples. A total of 55 C. difficile strains were isolated (5.62%). The positive rate of toxin genes and isolation results had no statistical significance between children and adults groups. However, some clinical features, such as fecal property, diarrhea times before hospital treatment shown difference between two groups. The watery stool was more likely found in children, while the blood stool for adults; most of children cases diarrhea ≤3 times before hospital treatment, and adults diarrhea >3 times. Independent risk factor associated with CA-CDI was patients with fever. ST35/RT046 (18.18%), ST54/RT012 (14.55%), ST3/RT001 (14.55%) and ST3/RT009 (12.73%) were the most distributed genotype profiles. ST35/RT046, ST3/RT001 and ST3/RT009 were the commonly found in children patients but ST54/RT012 for adults. The prevalence of CA-CDI in Yunnan province was relatively high, and isolates displayed heterogeneity between children and adults groups.


Results
Clinical features. Among the 978 community-acquired diarrhea patients, 712 cases (72.80%) were children, 266 (27.2%) were adults. The average age was 1.35 ± 2.24 years for children and 48.39 ± 17.72 for adults. The clinical characteristics of two patient groups were different, as shown in Table 1. Except for use antibiotics before hospitals treatment, duration of vomiting and vomit frequency, all the clinical features of two groups were statistical significance (P < 0.05). In children group, the proportion of male, fevers and vomit, the watery stool of fecal property and diarrhea days ≤3 before hospitals treatments were higher than adults; while, the diarrhea days >3 before hospitals treatments and the mucoid, blood stool were more commonly found in adults diarrhea cases. Children patients were enrolled from all the four sentinel hospitals and adults' cases were collected only from hospital A and D.
Fecal samples detection. Samples were collected from 2013-2016, and details of the distribution in each year were shown in Table 2. Toxin genes tcdA and/or tcdB were found in 138/978 cases (Supplementary file 1, S1), and the total positive rate was 14.11% for fecal samples virulence genes ( Table 2). Among them, 118 cases (85.51%) were tcdA + /tcdB + , 7 cases (5.07%) were tcdA + /tcdB − , and 13 cases (9.42%) were tcdA − /tcdB + (S1). In addition, three cdtA + /cdtB + positive cases from hospital D were found in this study and the rest were negative for binary toxins genes (S1). For 138 positive cases, 91 was tcdC + , 124 were tcdR + , and 122 were tcdE + . Among the 138 positive virulence gene samples, 102 cases (73.91%) were from children, only 36 (26.09%) cases from adults. However, the positive samples of children was 14 no statistical significance (χ 2 = 0.100, P = 0.752). As showed in Table 2, the clinical features of children group were quite similar, such as 46.08% cases had fever and vomit, 83.33% patients had watery stool, 65.69% of cases had diarrhea over 3 days, 93.48% of children patients had vomit more than 3 days, and 76.09% cases with vomit frequency less than 5 times. Furthermore, all these clinical features of children group showed no statistical difference (P > 0.05). While, for adults, the distributions of sentinel hospitals (χ 2 = 6.250, P = 0.012), fecal property (χ 2 = 11.721, P = 0.008), duration of diarrhea (χ 2 = 6.702, P = 0.010) had statistical significance. Compared with children cases, most of the positive cases of adults were from hospital D (86.11%), and the blood stool in adult cases (27.78%) had a relative high proportion. Around 61.11% patients had diarrhea over 3 days. Logistic analysis showed that fever (OR = 2.776, 95%CI = 1.342-5.742, P = 0.006) was the risk factor for positive virulence genes of fecal samples. Other clinical characteristics showed no statistical significance (P > 0.05) ( Table 3).

Co-infection with other enteric pathogens.
In addition to confirming C. difficile infection, we also screened the stool samples for other intestinal pathogens [15][16][17] . A diverse collection of pathogens were detected including diarrheagenic E. coli, Salmonella, Shigella, Yersinia, rotavirus, calicivirus, astrovirus and adenovirus. Finally, 16 stool samples with positive toxin genes were also found co-exist with one/two other pathogens mentioned above. Details were summarized in supplementary file 1. All 16 samples were positive for tcdA and tcdB, and negative for cdtA/B, except for sample YNCD 654 and 659, which were tcdA negative but tcdB positive. Sample YNCD550 was detected positive of both diarrheagenic E. coli and rotavirus (S1). The age of sample YNCD306 was 45-year-old, and the age of the rest 15 samples were no more than 2-year-old. Three cases (YNCD550, 822, and 831) had co-infection with rotavirus, 13 cases had co-infection with diarrheagenic E. coli, and one case (YNCD601) had co-infection with non-typhoidal Salmonella (S1). According to the bacterial culture result, 6 strains (patients) were found co-infected with diarrheagenic E. coli, non-typhoidal Salmonella and rotavirus. Among them, one case (YNCD831) had co-infection with rotavirus; four cases had co-infection with diarrheagenic E. coli and one case (YNCD601) had co-infection with non-typhoidal Salmonella (S1). The age of these 6 patients were no more than 2-year-old. Furthermore, two C. difficile strains (YNCD399 and YNCD591) were non-toxigenic, others were toxigenic isolates.
Analyzed with the co-infections, 16 cases of positive virulence genes were excluded, 122 cases (12.47%) were CA-CDI for virulence genes detection. For bacterial cultures, 6 cases were excluded for co-infection, and 5 non-toxigenic strains were also excluded, only 44 cases (4.50%) were CA-CDI for isolations.

Discussion
To our knowledge, this was the first cross-sectional study of CA-CDI in southwest China to address both clinical features and molecular characteristics of C. difficile. It is known that C. difficile is found in 30-35% newborns and in 10-15% infants up to 2 years as a commensal bacterium [18][19][20] , but is potentially harmful for children above two years 5 . Although C. difficile isolated from an infant under 2 years is normally considered as colonization, we still pay attention to pediatric CDI in this age group if clinical symptoms are typical, or other common intestinal pathogens or physiological diarrhea are excluded. Therefore, we detected toxin genes in all these age groups, and found similar positive isolated rates in both stool samples (14.33% for children, and 13.53% for adults) and isolates (6.04% for children, and 4.51% for adults) between children and adults. The presence of bloody diarrhea was considered as severe CDI for both children and adults 21 . In contrast to studies reported [21][22][23][24] , in which a higher proportion of severe CDI in children than that in adult was found, adult patients with bloody diarrhea (38/44 = 86.36%) were much larger than pediatric patients (6/44 = 13.64%) in this study. This might be attributed to lacking of mature toxin binding receptors on epithelial cells, or in complete cellular uptake of the toxins and/or protecting microbiome composition of infants, which leading to insensitive to free C. difficile toxins in the intestinal tract 22,25 . Antibiotics are usually considered as a risk factor, but not as a prerequisite for developing CDI. A previous study showed that nearly 30% of patients with CA-CDI did not receive antibiotics in the 12 weeks prior to infection 11 . Here, only 2.7% of patients received antibiotics before admission treatment in hospitals.
According to PCR detection of tcdA/tcdB of stool samples, 138 samples were positive. Furthermore, 55 isolates were successfully cultured, including 7 non-toxigenic strains which might be missed if only tcdA/tcdB positive stool samples were cultured. The average isolation rate (5.62%), and the positive rate (14.11%) for fecal virulence genes in this study was similar with previous studies. A study in United Kingdom illustrated the incidence of CA-CDI increased from 0 to 18 cases per 100,000 persons per year between 1994 and 2004 26 . Data from North America and Europe suggest that 20-27% of all CDI cases are community-associated, with an incidence of 20-30 per 100,000 population 27 . A large study in the Netherlands displayed that the proportion of positive toxin fecal samples (1.5% of 2,443 samples) was remarkably similar to that seen in the UK study by examining all fecal samples using a toxin EIA 28 .
In China, reports on community-acquired CDI were extremely rare. There is only one case of community-acquired CDI, due to a moxifloxacin susceptible RT 027 strain documented in China 29 . Although RT078 was recognized as key pathogen causing CA-CDI, especially in younger adults without hospital contact, but no RT 078 nor 027 were identified here. It was reported that susceptible populations and molecular types of hospital acquired CDI and CA-CDI were distinct. A Europe-wide survey of CDI in 2008 showed that ribotypes 014/020 (15%), 001 (10%), and 078 (8%) were most prevalent; the prevalence of ribotype 027 was 5% 30 . While in another study in North China, ST54 was the dominant type and accounted for 29.2% (66/226) and the other most frequent types were ST3 (25.7%), ST2 (9.7%), ST35 (10.6%), and ST37 (8.4%), but they displayed distinct proportions in diarrhea adults, healthy infant and healthy adults 31 . Similarly, ST3, ST35 and ST54 were the major ST types in our study with different composition rate among diarrhea infants (32.56%, 20.93% and 11.63%) and diarrhea adults (8.33%, 8.33% and 33.33%). However, ST37, previously reported dominant types in China from some studies [32][33][34] was not found in this study, which may due to divergence of geography, population, habits and customs. A recent study of hospitalized C. difficile infection conducted in Eastern China 13 showed that ST37/RT017 (16.5%) was the most dominant genotype. RT001 (14.4%), RT012 (13.4%), RT017 (16.6%) and ST2 (11.2%), ST3 (16.3%), ST37 (16.6%), ST54 (12.9%) were predominant. Interestingly, the genotypes profiles in our study were a little bit distinct from that report. ST35/RT046, ST3/RT009 and ST3/RT001 were most common genotypes for pediatric patients, while the most genotype of adult patients was ST54/RT012 in our study.
According to the MLST scheme, almost all isolates clustered in clade 1, and only 3 isolates (ST39) were in clade 4 in this study. The C. difficile population structure consists of six distinct phylogenetic clades designated 1-5 and C-I 35,36 , and the latter one was highly divergent, entirely nontoxigenic, and potentially a new species or subspecies of C. difficile 37 . Clade 1 is a highly heterogeneous cluster with toxigenic and nontoxigenic STs and RTs, in North American and Europe 35 . Interestingly, the ST /RT types mentioned above were not the dominant ones in our study, and ST35/RT046, ST3/RT001, ST3/RT009 and ST54/RT012 took highest proportion here. The most notable types in clade 4 is ST 37 (RT017), which has been associated with outbreaks in Europe 38,39 , North America 40 , and Argentina 41 , and is responsible for much of the CDI burden in Asia 42 . However, no ST 37 was identified, leading to focus on clade 4 (containing ST 39 in this study) in China in following research. This further emphasized the importance of national wide surveillance of CDI in China.
This study was the first systemic surveillance for CA-CDI, included children and adults cases in southwest China, in which the prevalence and risk factors were determined. Detailed clinical information and laboratory tests were collected and performed to identify the molecular and epidemiological characteristics of CA-CDI in this area. This study gave us an eye on CA-CDI and caused attention focused on CDI surveillance in China. However, it was indentified several limitations in this study. Firstly, the study was conducted in an urban region that probably showed a poor representation of the potential CA-CDI. Secondly, the diarrhea cases were selected from outpatient cases. But the patients who did not to seek medical advice were not recruited. Thirdly, some of the specific information of C. difficile infection probably missed when data collections, such as co-morbidities, exposure to proton-pump inhibitors and histamine-2 antagonists etc. Therefore, further research involved diarrhea cases from urban, rural, outpatient from hospital, and patients from other medical facilities might be done to evaluate the prevalence, clinical features and burdens of C. difficile infection. . Hospital B to D were tertiary care academic medical center in southwest China, and hospital A was a primary health service center. These four hospitals covered the entire Kunming area. The majority of diarrhea patients came from 12 districts in Kunming, and other cities in Yunnan province, while the rest of patients were from Sichuan or Guizhou provinces in southwest China. The stool specimens from diarrhea patients during the study period were collected and transported to the Yunnan Provincial Center for Disease Control and Prevention. To avoid overrepresentation, only the first stool specimen from each patient was included.

Methods
Diarrhea was defined as more than 3 times/per day, accompanied by changes in fecal traits (loose, watery or unformed stool) 43 . Clinical information of patients were collected when they came to hospitals, including gender, age, manifestations et al. Patients <18 years was considered as children, ≥18 years were adults. All these cases were defined as community-acquired diarrhea, indicated that the onset of diarrhea symptoms occurring while the patient was outside a healthcare facility, or the patient had no prior stay in a healthcare facility within the 12 weeks prior to symptom onset 44,45 . Therefore, all the C. difficile detection positive cases (stool specimens of virulence genes and bacterial isolations) in this study were defined as community-acquired C. difficile infection (CA-CDI).
Previous to C. difficile detection, the intestinal pathogen spectrums were conducted for all the stool specimens, including diarrheagenic E. coli, Salmonella, Shigella, Yersinia, rotavirus, calicivirus, astrovirus and adenovirus [15][16][17] . We analyzed the co-infection results between C. difficile and other enteric pathogens mentioned above.
Detection of virulence genes in fecal samples. DNA extraction was performed on all fecal samples using a fecal sample DNA extraction kit (Tiangen, Beijing) based on the manufacturers' instruction. The DNA samples were stored at −20 °C and tested for the presence of virulence genes tcdA, tcdB, tcdC, tcdR, tcdE, cdtA and cdtB, using primers shown in Table 4. PCR amplification was performed using 20 μl system, 10 μl of Taq premix, 8 μl of water, 0.5 μl of upstream and downstream primers (10 μmol), and 1 μl of template DNA. The amplification reactions were: 94 °C for 5 min; 94 °C 15 sec, 55 °C 30 sec, 72 °C 30 sec, 30 cycles, and finally at 72 °C for 10 min. The products were observed by 1.5% agarose gel electrophoresis and gel imager (Bio-Rad, GelDoc).
Bacterial culture. Stool specimens from diarrhea patients were collected using Transwabs (MW&E Ltd., Wiltshire, England), and stored in brain heart infusion (BHI, Oxoid, UK) with 15% glycerol in −80 °C. All faces specimens were inoculated on selective cycloserine-cefoxitin-fructose agar plates (CCFA, Oxoid, UK) with 5% egg yolk after ethanol shock treatment and incubated in an anaerobic jar (80% nitrogen, 10% hydrogen and 10% carbon dioxide) (Mart, NL) at 37 °C for 48 h. C. difficile colonies were identified on the basis of their typical morphology and Gram stain as well as the characteristic odour. Suspected colonies were further confirmed by API 20 A (BioMerieux, France) on their biochemical characteristics and 16 SrRNA gene with primer shown in Table 4.
Bacterial DNA isolation, PCR-ribotyping and toxin gene profiling. Crude bacterial template DNA for toxin profiling was prepared by resuspension of cells in a 5% (wt/vol) solution of Chelex-100 resin (Bio-Rad). All isolates were screened by PCR for the presence of the toxin A (tcdA) and toxin B (tcdB) genes and the binary toxin (cdtA and cdtB) genes 46,47 and the regulating genes of tcdC, tcdR, and tcdE. PCR ribotyping was performed as previously described 48 . PCR ribotyping reaction products were concentrated using a Qiagen Min-Elute PCR purification kit (QIAGEN) and run on a QIAxcel capillary electrophoresis platform (QIAGEN). Visualization of PCR products was performed with QIAxcel ScreenGel software (v1.3.0; QIAGEN). PCR ribotyping banding patterns were identified by comparison of banding patterns with a reference library consisting of a collection of 24 reference strains from the European Centre for Disease Prevention and Control (ECDC), and a collection of 30 isolates from American Type Culture Collection (ATCC). Interpretation of the capillary electrophoresis data (PCR ribotyping banding patterns) was performed using the BioNumerics software package v.7.6 (Applied Maths, Saint-Martens-Latem, Belgium). Isolates that could not be identified with the available reference library were designated with internal nomenclature.

Multilocus sequence typing.
MLST was performed on all recovered isolates using the primers and methods developed by Griffiths et al. 49 . Seven housekeeping genes (adk, atpA, dxr, glyA, recA, sodA and tpi) were amplified and products were sent for bidirectional sequencing to TaKaRa, Japan. The complete allele sequences were analyzed using DNAStar and MEGA4 software and allele and ST assignments were performed using the C. difficile database at pubMLST (https://pubmlst.org/cdifficile). A minimum spanning tree was created using BioNumerics version 7.6. During the experiment, adk gene for seven strains couldn't amplified, and we re-designed the primer (Table 4) based on the Clone Manager Professional Suite 8 software.
Statistical analysis. Data analysis was performed by IBM SPSS software (version 19.0 for Windows, Armonk, NY). Univariate analysis was performed using the χ 2 test or Fisher's exact tests where appropriate. Logistic regression analysis was used to identify independent risk factors. Kruskal-Wallis and χ 2 test were used to analyze correlation among RTs, STs and clinical features of strains. Odds ratios (ORs), 95% confidence intervals (95%CI), and P values were calculated to assess the differences between groups. Two-sided significance was assessed for all variables, applying a cut-off value of P < 0.05. Ethics statements. The human sample collection and detection protocols were carried out in accordance with relevant guidelines and regulations. All experimental procedures were approved by the Ethics Review Committee [Institutional Review Board (IRB)] of National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention. All adult subjects provided informed consent, and a parent or guardian of any child participant provided informed consent on their behalf. Data availability. All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).