1. Introduction
Arcobacter butzleri (homotypic synonym “
Aliarcobacter butzleri”), formerly addressed as
Campylobacter butzleri and first described in 1991, is a facultatively pathogenic bacterium from the order Campylobacterales and the family Arcobacteraceae [
1,
2,
3,
4,
5]. Moreover, in 1991, the genus name
Arcobacter was proposed [
6].
Arcobacter spp. are phylogenetically closely related to
Campylobacter spp. [
7]. Nevertheless, the phenotypic characteristics of the genera are partly different [
8]. In human patients,
A. butzleri has been associated with partly severe infections comprising diarrhea including travelers’ diarrhea, enteritis, gangrenous appendicitis, peritonitis, endocarditis and bacteremia and accordingly, it has been regarded as an emerging pathogen since 2002 [
9,
10,
11,
12,
13]. Nosocomial transmission has been reported [
14].
Next to
A. butzleri, the likely importance of
A. cryaerophilus,
A. skirrowii and other species for human enteric disease has been suggested, mostly due to epidemiological associations or based on case reports [
12,
15,
16,
17,
18]. Other species, such as, e.g.,
A. lanthieri, have been primarily associated with livestock [
19], although virulence factors with likely relevance for human disease have been found in their genomes as well [
20,
21,
22]. Altogether, the etiological relevance of
A. butzleri is considered to be the best established [
15,
16]. Cell invasion, induction of immune responses and toxin production have been associated with its pathogenic potential [
16,
18,
23]. As expected, due to phylogenetic relatedness, various virulence genes are homologous to those in
Campylobacter spp. [
24]. In particular, invasive and adhesive properties have been demonstrated [
25]. Selection of the microorganism under antibiotic pressure can be facilitated by its pronounced resistance to antimicrobial drugs or even multidrug-resistance [
16,
26]. In a recent meta-analysis on
Arcobacter spp. [
27], high minimum inhibitory concentrations, suggestive of a lack of antimicrobial susceptibility, were recorded, particularly for beta-lactams with up-to 99.2% and 97.4% for penicillin derivates and cephalosporines, respectively, followed by macrolides with up-to 39.8%, fluoroquinolones with up-to 14%, aminoglycosides with up-to 12.9% and tetracyclines with up-to 7.1%. Further, sequencing approaches have shown a high diversity of sequence types, mostly without clear-cut associations to specific hosts or geographic regions [
24].
Livestock, including poultry and pigs, has been identified as a reservoir of
A. butzleri, and raw meat products, as well as contaminated water, are of relevance for the pathogen’s transmission [
15,
24,
28,
29,
30,
31]. Accordingly,
A. butzleri-associated disease is considered zoonotic [
32]. Meat contamination is assumed to be caused by spillage of gastrointestinal fluids from animals during the slaughtering process [
12]. Adverse environmental conditions such as food processing and storage can be survived by the microorganism, supporting its spread to human individuals [
16].
Culture-based isolation and differentiation of
A. butzleri is still not a standardized routine procedure in diagnostic laboratories assessing human sample materials [
15,
32]. The lack of standardized diagnosis has also been blamed for the lack of reports of
A. butzleri-associated disease in regions where the microorganisms are known to be highly prevalent like, e.g., in Nigeria [
33]. Though cultural growth of
Arcobacter spp. in microaerobic or aerobic atmosphere is feasible, and—other than
Campylobacter spp.—the microorganisms even grow at low temperature of 15 °C, common growth protocols suggest enrichment steps, i.e., use of selective broths and agars to suppress concomitant bacterial flora, as well as incubation times of 4–5 days [
13,
23,
34].
A. butzleri also grows on standard agars such as blood agar, chocolate agar and MacConkey agar under standard conditions such as a temperature of 37 °C and an atmosphere enriched with 5% CO
2, with colonies showing positive results in cytochrome oxidase and motility testing [
13]. Biochemical differentiation is challenging [
12,
13], and the reliability of matrix-assisted laser-desorption-ionization time-of-flight mass spectrometry (MALDI) largely depends on the quality of the underlying database as reported elsewhere [
18]. Accordingly,
A. butzleri-associated gastroenteritis is assumed to frequently go undetected, making an estimation of the role of this microorganism in infectious gastroenteritis challenging [
12]. In line with this, a previously reported association of
Arcobacter spp.-isolations with 0.11% to 1.25% of diarrhea cases might considerably underestimate the microorganisms’ true prevalence [
18].
Considering the challenges of traditional culture-based
Arcobacter spp.-detection and identification in the diagnostic routine setting [
12,
13,
15,
18,
23,
32,
33,
34], molecular diagnostic approaches based on PCR or real-time PCR with or without subsequent sequence analysis were established [
13,
35]. Of note, however, early
Arcobacter spp.-specific PCR assays targeting genes such as
gyrA or the 16S and 23S ribosomal RNA genes showed restricted diagnostic accuracy including cross-reactions with non-target species [
35]. For various
Arcobacter spp., including
A. butzleri, fluorescence resonance energy transfer (FRET)-based and hybridization probe-based real-time PCR assays have been introduced, established and applied for human diagnostic, agricultural and environmental use [
17,
36,
37,
38,
39,
40,
41,
42,
43,
44,
45,
46,
47]; however, the validation studies were usually based on quite limited sample counts.
In the study presented here, the aim was to contribute to available evidence on the diagnostic accuracy of selected
A. butzleri-specific real-time PCR assays [
36,
37,
38] based on a test comparison without a reference standard [
48], applying latent class analysis [
49] and using a sample collection with unknown
A. butzleri prevalence but expected high pretest probability due to a known high local abundance of this microorganism [
26].
4. Discussion
Considering the yet-poor standardization of the diagnosis of Arcobacter butzleri in human stool samples in the routine diagnostic setting, the study was conducted to comparatively assess the diagnostic accuracy of three published real-time PCR assays with different target genes in a latent class analysis (LCA)-based test comparison without a reference standard using a sample collection with high pretest probability. The assessment led to a number of results.
First, the initial assessment with stool samples spiked at high titers with either
A. butzleri or the phylogenetically closely related
A. cryaerophilus indicated that all assessed real-time PCR assays are not perfectly specific for
A. butzleri but have the potential of cross-reaction with phylogenetically related
Arcobacter species. The lack of positive real-time PCR signals with samples spiked with 12
A. lanthieri strains at least indicates that such cross-reactions do not necessarily occur with all representatives of the genus. The observed cross-reaction of the
rpoB-assay with the single “outstander” sample, which was high titer-spiked with
Campylobacter coli, was considered as hardly relevant for the diagnostic situation due to the extraordinarily large associated Ct value shift. While the reaction with
A. cryaerophilus was expected for the
gyrA-assay due to its design [
36] and could be easily discriminated from reactions with
A. butzleri due to distinct melting temperatures, such a discrimination was unfeasible for the hybridization probe-based
rpoB/C-assay and
hsp60-assay. However, Ct-values shifts of 14.9 when comparing
A. butzleri-spiked samples and
A. cryaerophilus-spiked samples in the
hsp60-PCR, and of 4.5 when comparing them in the
rpoB/C-PCR, indicated base-mismatching-associated reduced likeliness of cross-reaction with
A. cryaerophilus. This might result in lower susceptibility of the hybridization probe assays to non-specific reactions in case of use with clinical samples without spiking-associated exorbitantly high concentrations of pathogen DNA. Nevertheless, the results confirm the previously described challenges regarding the design of PCR assays with reliable selectivity for
A. butzleri [
35].
Second, LCA-based prevalence estimation of 14.7%
A. butzleri-positive cases among the assessed Ghanaian stool samples confirmed the study’s assumption that the chosen specimen collection was associated with high pre-test probability. This finding matches previous reports on
A. butzleri prevalence rates in Ghanian livestock being even higher than
Campylobacter spp.-prevalence rates [
56] next to reports on high enteric infection and colonization rates with bacterial pathogens in Ghanaian individuals [
57,
58] and on high
A. butzleri-prevalence rates in Western African Nigeria [
33]. In line with the ethical clearance of the here-presented study, the residual volumes of the samples were completely anonymously assessed; thus, no statements on epidemiological features such as age, sex and symptom associations can be provided. As stated above, this design is an admitted deviation from STARD criteria [
52]. However, it is nevertheless in line with diagnostic routine conditions, despite the long storage of extracted DNA, because as-good-as-possible diagnostic accuracy for the detection of the target pathogen is required irrespective of the epidemiological background.
Third, and in line with previous findings [
35], the LCA-assessment indicated considerable difference regarding the diagnostic accuracy of the assessed three published real-time PCR assays for the detection of
A. butzleri [
36,
37,
38]. The FRET-assay targeting the
gyrA gene [
36] showed the best specificity of close to 100%, albeit for the price of a very low sensitivity of less than 15%. Considering the high Ct values measured for the
gyrA-PCR-positive Ghanian sample materials, it is likely that target DNA quantities close to the PCR’s detection threshold have been the reason for the poor sensitivity result. A 10-fold higher detection threshold observed with the dilution series of the positive control plasmid compared to the
hsp60-assay and the
rpoB/C-assay as described above also speaks in favor of this conclusion. The hybridization probe-based
hsp60-assay showed intermediate sensitivity of only slightly less than 95% but the comparably worst specificity of less than 97%. The latter is also the reason for the calculated sensitivity being lower than the calculated sensitivity of the
rpoB/C-assay, although more positive results were obtained with the
hsp60-assay from the sample collection. Focusing on the observed higher likeliness of the
rpoB/C-assay to cross-react with
A. cryaerophilus-spiked samples compared to the
hsp60-assay, the latter’s worse specificity might be considered surprising. However, the melting curve analysis of the non-
A. butzleri-associated positive results in the
gyrA-assay indicated a very low prevalence of only six samples positive for
A. cryaerophilus, of which four showed negative reactions in the
hsp60-assay. Due to the quantitatively low relevance of
A. cryaerophilus as a potentially cross-reacting agent in the Ghanaian sample collection, its effect on the diagnostic accuracy estimations has most certainly been negligible. The
rpoB/C-assay was the only assay with a calculated sensitivity of more than 95%, at least if the uncertainty due to the broad 95%-confidence interval is accepted, associated with a still-acceptable specificity of slightly more than 98%. In spite of the large confidence interval of the
rpoB/C-PCR’s estimated sensitivity, the true sensitivity of the assay is nevertheless likely to be actually high due to the observed combination of (a) a high absolute number of positive PCR signals; (b) the almost perfect agreement of its results with the results of the
hsp60-assay, which showed sensitivity only slightly lower than 95%; and (c) the quite-acceptable specificity. The
rpoB/C-assay’s relatively good specificity may seem surprising considering the high rate of observed matching between positive
rpoB/C-PCR results and positive
gyrA-PCR results, with melting curves indicative of non-
A. butzleri DNA. However, the observed non-
A. butzleri-specific melting curves may have been due to co-colonization of the stool donators’ gut with both
A. butzleri and microorganisms with DNA more readily reacting with the
gyrA-assay and so, it cannot be assumed for certain that the associated
rpoB/C-PCR results must have been false positive in all these instances. The same applies to positive
hsp60-PCR results in concordance with non-
A. butzleri-specific positive
gyrA-PCR signals. This co-colonization hypothesis is supported by the finding of quite-similar mean Ct values in
hsp60-PCR and
rpoB/C-PCR, respectively, irrespective of the observation of
A. butzleri-specific or non-
A. butzleri-specific melting temperatures in the
gyrA-RCR. In contrast, the broad spectrum of observed melting temperatures in the case of non-
A. butzleri-specific positive
gyrA-PCR signals indicates that several phylogenetically related non-target microorganisms potentially associated with cross-reactivity were abundant in the sample collection.
This study has a number of limitations. First, a culture-based reference standard for the test comparison was not available. Respective attempts were unfeasible due to the retrospective design of the study. However, due to the lack of standardization regarding the cultural growth-based diagnosis of A. butzleri, uncertainties regarding the sensitivity and specificity of such an approach, even from fresh sample materials in the case of a prospective study design, would have limited its value as a reliable reference standard for the test comparison. Second, the choice of microorganisms used for the initial spiking experiments was restricted by the availability within the research group. However, the spiking approach was considered as no more than an initial proof-of-principle, while the main study was based upon the LCA approach. Third, lacking funding of this investigator-initiated investigation made sequencing-based confirmatory testing from all obtained PCR amplicons unfeasible; thus, LCA was applied for the diagnostic accuracy estimation. Due to this lack of a sequencing option, unfortunately, gyrA PCR signals with melting temperatures different from the expected values for A. butzleri and A. cryaerophilus could not be resolved. Fourth, and as repeatedly stated above, the ethical requirement of thorough anonymization for the test comparison made any comparison of A. butzleri detections, both qualitatively and with focus on recorded Ct values, with clinical symptoms of the assessed individuals unfeasible. Future studies should be conducted in order to address this highly relevant issue and to estimate the clinical relevance of such real-time PCR findings. Fifth, and as an intrinsic limitation of the chosen mathematic approach, the principle of LCA implies that the calculated diagnostic accuracy does not necessarily refer specifically to A. butzleri but to a meta-structure sharing genetic elements detected by the three compared assays. This might as well mean a combination of A. butzleri and other phylogenetically closely related microorganisms abundant in the assessed Ghanaian stool samples. The observation of potential cross-reactivity within the Arcobacter/“Aliarcobacter” genus described in this study makes this option rather likely. Accordingly, LCA can help to estimate prevalence rates on a population level but cannot decide on the correctness of an individual PCR result.