Background
Infection with
Campylobacter spp. is a worldwide leading cause of diarrhoea, showing an increase in incidence over the last decade in North America, Europe and Australia [
1]. In Switzerland, as in other European countries, campylobacteriosis is currently the most commonly reported food-borne bacterial disease [
2] with up to 70% of all cases attributable to chicken consumption [
3], thus posing a relevant public health problem. In Switzerland, the incidence in 2017 was 85.4 cases per 100,000 people [
4] and the annual economic burden due to laboratory-confirmed campylobacteriosis has recently been estimated at around €8.3 million [
5]. It is assumed that a large number of additional cases go unreported [
5].
The clinical course of campylobacteriosis varies in symptoms and severity. Infection with
Campylobacter jejuni - the most frequently diagnosed species - is most often characterised by acute diarrhoea, fever, and abdominal pain [
6]. In most cases self-limiting, symptoms typically last from one day to a week, however, may persist for longer in 10–20% of patients and relapse may occur in another 5–10% [
6]. The disease is occasionally further complicated by bacteraemia and sepsis or by extra-gastrointestinal manifestations following enteric disease, such as reactive arthritis and Guillain-Barré syndrome [
6]. Complications and prolonged disease usually affect elderly or immunocompromised patients with comorbidities [
1].
The clinical manifestation of infection with
Campylobacter spp. is determined by host factors such as immune-competency, age, genetic background, and bacterial factors such as virulence of the pathogen, initial infectious dose, and initiated therapy [
6,
7]. Current research has identified novel mechanisms that may contribute to virulence and survival of
Campylobacter spp. in humans, such as the Type VI Secretion System (T6SS) [
8], one of several bacterial systems for protein secretion [
9]. With a structure similar to the tail of the contractile bacteriophages [
10], the T6SS allows the contact-dependent secretion of effectors into other bacterial or host cells [
11,
12]. The T6SS is encoded by a gene cluster of approx.17kb comprising 13 genes [
13]. One of the hallmark genes of the T6SS gene cluster is
hcp (encoding the secreted component hemolysin co-regulated protein) [
14], which has been defined as a marker of a complete T6SS [
15].
The expression of T6SS in
C. jejuni has been demonstrated to be of importance for host cell adhesion and invasion in vitro: mutants with a defective secretion system show an up to 50% reduced ability to adhere to and invade colonic epithelial cells and macrophages [
13]. Furthermore, in vivo studies show a higher persistence of
C. jejuni expressing T6SS, with the dynamics of T6SS-regulation favouring colonisation of the colon [
12,
13]. In polymicrobial environments such as the gastrointestinal tract, the T6SS also presents an advantage by allowing the pathogen to overcome the local microbiota [
16].
In view of the diverse functionality of T6SS and extensive in vitro/in vivo research, the secretion system may provide a survival benefit for
Campylobacter spp. contributing to successful and persistent infection [
17]. A possible connection between T6SS as a clinically relevant virulence factor in
Campylobacter spp. and manifestation of disease in humans has been insufficiently studied so far. Harrison et al. [
15] found a correlation between T6SS-positive
C. jejuni strains and bloody diarrhoea, and accordingly more severe disease, in a sample of 36 patients from Vietnam. A causal relationship could, however, not be proven [
15].
In this study, we aimed to compare the clinical course of Campylobacter infection caused by strains with and without T6SS in an observational case-control study, and to identify possible associations between this virulence factor and the clinical manifestations of disease.
Methods
Cases with detection of Campylobacter in any clinical isolate at the University Hospital Basel within the time period from April 2015 to January 2017 were eligible for this observational case-control study. A total of 119 consecutive C. jejuni and 15 C. coli bacteria were isolated. Coinfections were excluded. Rare Campylobacter species such as C. fetus (n = 2), C. upsaliensis (n = 2), and C. concisus (n = 1) were also detected, but subsequently excluded from the analysis due to the small sample size. Additionally, 22 Campylobacter isolates were detected at the University Children’s Hospital Basel during the same observation period. Due to small sample size, these data were only analysed descriptively and not included in the case-control study.
Patients and data collection
The study has been approved by the local ethical review board (EKNZ, approval number 2016–01183). The medical records of the included patients were retrospectively reviewed. Data extraction was performed blinded for T6SS status. Hospitalised cases included in this study were admitted via the emergency unit. Outpatients were most commonly treated in the emergency unit and in some cases monitored for a maximum of one night.
The following data were systematically collected: demographic characteristics, medical and travel history as pertaining to the present infection. In addition, clinical characteristics at presentation, characteristics of diarrhoeal disease, and laboratory results at presentation, as well as the requirement for antibiotic treatment, were documented. Underlying diseases were recorded according to the Charlson Comorbidity Index [
18‐
21]. Regarding treatment, fluid substitution and antibiotic therapy (including duration and application) were recorded. Information on the need for and duration of hospitalisation, need for ICU-stay and in-hospital mortality were extracted. Microbiological data included identification to the species level and the antibiotic susceptibilities to erythromycin and ciprofloxacin (μg/mL).
The following categorical factors were defined as primary clinical endpoints: need for hospital admission, antibiotic therapy, intensive care unit (ICU) stay, development of bacteraemia, and in-hospital mortality.
Isolate growth and identification
All routine clinical microbiology was performed in an ISO/IEC 17025 accredited laboratory. Stool samples were cultured in microaerobic conditions (5% O2, 10% CO2 and 85% N2) on Campylosel agar plates (bioMérieux, Marci I’Étoile, France) at 41 °C for 48 h. Bacterial isolates were identified using Matrix Assisted Laser Desorption Ionization – Time of Flight mass spectrometry using methods described by Bessède et al. [
22] on a Microflex LT mass spectrometer and MALDI Biotyper Compass software v. 4.1.70 (Bruker Daltonics, Germany). The value of “Score” ≥ 2.0 was accepted as a measure for reliable species identification.
Minimal inhibitory concentration (MIC) determination
MICs for erythromycin and ciprofloxacin of each isolate were determined by E-test (bioMérieux) and incubated under microaerobic conditions on MHF (Müller Hinton agar + 5% horse blood + 20 mg/l β-NAD) agar plates (bioMérieux) at 41 °C for 24 h. The interpretation of MICs was performed according to EUCAST guidelines (version 8.1;
www.eucast.org).
Type VI secretion system (T6SS) PCR detection
PCR detection of the
hcp gene as a locus representing the T6SS in
C. jejuni and
C. coli isolates was performed [
23]. We used
C. jejuni subsp.
jejuni strains from ATCC (Ref. 33560 and 43431) as positive controls. Briefly, the DNA of cultured isolates was extracted using the DNA Bacterial card on the EZ1 Advanced XL automated extraction system (Qiagen, Hombrechtikon) using the manufacturer’s protocol following a digestion step using proteinase K performed for 10 min at 56 °C and 10 min at 95 °C. The PCR used the following primers: Hcp_F 5′-CAAGCGGTGCATCTACTGAA-3′, Hcp_R 5′-TAAGCTTTGCCCTCTCTCCA-3′ and as a control the
gltA gene gltA_F 5′-GCCCAAAGCCCATCAAGCGGA-3′, gltA_R 5′-GCGCTTTGGGGTCATGCACA-3′. HotStarTaq (Qiagen) was used for the PCR with a melting temperature of 57 °C. PCR fragments (
hcp 463 bp and
gltA 142 bp) were detected on a 2200 TapeStation instrument (Agilent, Santa Clara, USA) using the D1000 ScreenTape assay (Agilent).
Whole genome sequencing (WGS)
Isolate details are given in Additional file
1: Table S1. DNA was extracted as above. Library generation with Nextera XT (Illumina, Cambridge, UK) was followed by sequencing 2 × 300 bp on an Illumina Miseq platform. Downstream analysis was performed in CLC Genomics Workbench v 10.1.1 with read trimming and QC, followed by coverage determination by mapping against the genome of NCTC11168 (accession number NC002163). All genomes gave a mean coverage over 62x, up to 488x. The presence of the T6SS was determined by mapping against the well characterised T6SS locus in strain 414 (accession number CM000855, locus tags C414_000040086- C414_000040100) [
13]. All isolates with reads mapping to under 25% of the gene cluster, also with mean coverage of the region lower than 1x, were considered as negative (
n = 103); those mapping to over 86% of the locus were considered to be positive (
n = 21); and those mapping to 26–41% of the locus considered as partial (n = 10). Read data generated for this study can be found in the European Nucleotide Archive (ENA,
http://www.ebi.ac.uk/ena/) under project number PRJEB30092.
Statistical methods
Descriptive analysis was performed with all data. For the analysis of the clinical impact of T6SS, only adult patients infected with C. jejuni were included into the statistical model to obtain a more homogenous cohort. Median and interquartile ranges are shown, unless otherwise indicated. Disease and host characteristics were compared in a univariate analysis between T6SS-positive (case) and negative strains (controls). For the calculation of adjusted odds ratios, a multivariable logistic regression model was constructed using variables that were statistically significantly associated with a positive T6SS status in the univariate analysis. Results with p-values < 0.05 (two-sided) were considered statistically significant.
Discussion
This study adds to the growing amount of research on the bacterial T6SS providing insight into its occurrence in clinical Campylobacter spp. and examining its role in human infection.
While 15.7% of the isolates from the adult population were positive for the complete T6SS locus, a proportion also carry partial T6SS gene clusters, as has previously been noted [
15,
24]. These were not identified by PCR primers targeting
hcp, meaning that the PCR of
hcp is a valuable marker for a functional T6SS [
15]. That mean coverage of the T6SS is lower than that across the genome may be a sequencing artefact relating to the relatively low %G + C content of this locus, or may indicate that the locus is lost from some bacteria during culture.
The clinical characteristics of infection with
Campylobacter spp. in the study population are similar to those previously described by various authors [
1,
6]. Infection with
C. jejuni and
C. coli were clinically indistinguishable: both manifested as acute gastroenteritis with a similar distribution in both species causing vomiting, abdominal pain, bloody diarrhoea and fever history. Complications arose in very few cases, with a single death attributable to invasive campylobacteriosis. In view of epidemiologic studies reporting a peak incidence of
Campylobacter infection in toddlers and young adults [
1,
25,
26], the median age of 55 in patients in the present study seems relatively high. However, Schmutz et al. [
27] describe a trend towards a higher median age in
Campylobacter infections in Switzerland. A possible selection bias towards an older and sicker population remains: 59.7% of all diagnosed patients required hospital admission and antibiotic therapy was prescribed in 61.2% of cases. Furthermore, 5.0% of patients infected with
C. jejuni had a positive blood culture, which is unusually high considering population-based studies from Denmark [
28] and the USA [
29]. These have described an incidence of
Campylobacter bacteraemia of less than 1% of all infections. Besides the aforementioned selection bias, this discrepancy may also reflect the higher number of automatic blood cultures performed in a hospital setting.
Various studies have investigated the epidemiology of T6SS in
C. jejuni. The T6SS seems to be more prevalent in isolates from Asia and Egypt than Europe, with 60.6% of human isolates T6SS-positive in Vietnam [
15], 33.3% in Thailand [
15] and 57.6% in a large Egyptian paediatric cohort [
30]. European studies have shown a prevalence of 2.6% in human isolates from the UK [
15] and 14% in chicken isolates from Spain [
24]. The results of the present study are in accordance with these studies in that they uphold the large difference between Asian and European strains. With a T6SS-prevalence of 16.8% in
C. jejuni the results are close to those from Spain [
24], although their strains were obtained from poultry and sewerage water. Bleumink-Pluym et al. also identified a T6SS-prevalence of about 10% in strains from various sources such as human, chicken, swine, and cattle [
14].
The reported prevalence of T6SS in
C. coli reaches from 18.0% in isolates from a paediatric cohort [
30] to 56.1% in isolates from retail chicken [
23]. Corcionivoschi et al. [
23] found a much higher percentage of T6SS-positive
C. coli in comparison with
C. jejuni. As they point out, only a small percentage of
Campylobacter infections in humans are caused by
C. coli, whereas the prevalence of
C. coli and
C. jejuni is almost equal in their study [
23]. The present study identified only one functional T6SS among 15
C. coli isolates. The low prevalence of 6.7% may be due to small sample size, however: Sainato et al. [
30] also recorded a relatively low frequency of T6SS in
C. coli (18%) in comparison to
C. jejuni (57.6%) in clinical isolates.
Our comparison between patients with T6SS-positive and -negative
C. jejuni isolates did not reveal a significant difference in the clinical manifestations and the course of disease. All variables defined as clinical endpoints (e.g. hospital admission, antibiotic therapy) showed a similar distribution in both groups. These findings agree with a recently published study examining 176 cases of infection with
C. jejuni and 126 cases of infection with
C. coli in a paediatric cohort from Egypt concerning the presence of the
hcp gene [
30]. The clinical symptoms did not differ in severity between T6SS-positive and –negative
Campylobacter strains [
30]. This stands in contrast to the findings of Harrison et al. [
15] of a positive association of T6SS-presence in
C. jejuni with bloody diarrhoea. Bleumink-Pluym et al. [
14] noted that 50% of their T6SS-positive strains were isolated from blood, also suggesting an association with more severe disease. The studies were performed with 36 and 8
C. jejuni strains, respectively, making it possible that the dissimilarity in findings is due to small sample size and potentially other virulence factors.
We found a surprisingly stable association (
p < 0.001) with T6SS-positive strains for a higher score in the Charlson Comorbidity Index, ascertained also in a multivariable model. Cancerous diseases and diabetes mellitus type 2 were identified as the leading comorbidities. These and other diseases assessed by the CCI show an immunosuppressing effect in patients. It seems that patients with a higher morbidity, but also with an immunocompromised status are more susceptible to
C. jejuni strains harbouring a T6SS. The association of T6SS-positive strains with allogenic haematopoietic stem cell transplantation (HSCT), solid organ transplant and immunosuppressant therapy (though not independently) supports this hypothesis. In a study of
Acinetobacter baummanii isolates, Kim et al. [
31] also found an association between infection with T6SS-positive strains and HSCT and immunosuppressant therapy. Our results suggest a possible contribution of the T6SS to infection of an immunocompromised and sickly host. The mechanism, however, is not clear and should be further investigated. The identified association may, however, be influenced by selection bias towards an older and more sickly population.
The rates of antibiotic resistance identified in the present study are similar to those reported by other authors [
6,
32] with macrolide resistance higher in
C. coli (20.0%) than in
C. jejuni (1.7%) and generally high resistance rates for ciprofloxacin: 61.2% of all
Campylobacter isolates. Analysis of T6SS-positive
C. jejuni strains showed a significantly higher resistance rate for ciprofloxacin. A possible explanation for this association may be found in a correlation with clonal complexes analysed by multi-locus sequencing typing. Kittl et al. [
3], Collado et al. [
33] and Kovac et al. [
34] examined the relationship between ciprofloxacin-resistance and sequence types in
Campylobacter jejuni showing a significant association with clonal complexes ST-21 [
3,
33,
34], ST-353 [
33,
34], ST-48 [
33] and ST-464 [
3]. Kittl et al. noted that the correlation was independent from the host, suggesting that certain STs may develop more easily quinolone resistance [
3]. Harrison et al. [
15] found a correlation between the T6SS and clonal complex ST-353. It seems that clonal complexes showing higher resistance for ciprofloxacin may also be those that are associated with carrying the T6SS.
Our study has some important limitations: the overall sample size is relatively small and for the statistical analysis we have only focused on C. jejuni isolates from adult patients (n = 119). Also, a relative high proportion of patients was hospitalised, which could reflect a more severely ill and immunosuppressed patient population at a tertiary hospital setting. Finally, we found only 20 C. jejuni isolates with a T6SS locus.