In Vitro and In Vivo Infectious Potential of Coxiella burnetii: A Study on Belgian Livestock Isolates

Q-fever is a zoonosis caused by the gram-negative obligate intracellular pathogen Coxiella burnetii. Since its discovery, and particularly following the recent outbreaks in the Netherlands, C. burnetii appeared as a clear public health concern. In the present study, the infectious potential displayed by goat and cattle isolates of C. burnetii was compared to a reference strain (Nine Mile) using both in vitro (human HeLa and bovine macrophage cells) and in vivo (BALB/c mice) models. The isolates had distant genomic profiles with one - the goat isolate - being identical to the predominant strain circulating in the Netherlands during the 2007–2010 outbreaks. Infective doses were established with ethidium monoazide-PCR for the first time here applied to C. burnetii. This method allowed for the preparation of reproducible and characterized inocula thanks to its capacity to discriminate between live and dead cells. Globally, the proliferative capacity of the Nine Mile strain in cell lines and mice was higher compared to the newly isolated field strains. In vitro, the bovine C. burnetii isolate multiplied faster in a bovine macrophage cell line, an observation tentatively explained by the preferential specificity of this strain for allogeneic host cells. In the BALB/c mouse model, however, the goat and bovine isolates multiplied at about the same rate indicating no peculiar hypervirulent behavior in this animal model.


Introduction
Q-fever is a zoonosis caused by Coxiella burnetii, an intracellular gram-negative bacterium. Q-fever manifests in humans as acute or chronic illness. Acute Q-fever ranges from an asymptomatic state to an abrupt, flu-like illness that can be accompanied with high fever, general malaise, pneumonia, myalgia and hepatitis. Chronic Q-fever occurs in individuals in approximately 1-5% of cases, and typically manifests as endocarditis [1,2]. Humans contract the disease mainly by inhalation of airborne particles contaminated with C. burnetii (feces, milk or birth products -placenta and amniotic fluid-) of infected animals [3,4]. During abortion, up to 10 9 C. burnetii cells per gram of placenta can be excreted [5]. Considering the infective dose of this bacterium has been reported to be close to one [6], these products are obviously hazardous to humans.
Attention to Q-fever as a potential public health threat increased following the recent outbreaks in the Netherlands. On this occasion, the number of cases notified by the Dutch authorities rose to more than 4,000 between 2007 and 2010, representing a quite unique phenomenon [7]. The outbreak was retrospectively associated with an increase of dairy goat farm density in the affected area. However, because association with farming or other occupational activities could not be documented for all human Q fever cases reported during the outbreaks [8], intrinsic factors (e.g. the presence of unusual virulence factors) associated with the circulating isolates have been questioned [7,9,10,11,12,13,14]. Scientific evidence however is missing to clarify this point. Extensive genotypic characterization of human and animal C. burnetii isolates demonstrated the polyclonal nature of the Dutch outbreaks [9,15,16,17,18]. Genetic diversity was limited though, and all isolates could be rooted to one founder genotype named CbNL01 [17] belonging to group G in 6-locus Multiple Loci Variable Number of Tandem Repeats Analysis (MLVA) typing [9], type-1 in single nucleotide polymorphism (SNP) analysis [15] and MST33 in multispacer sequence typing (MST) [18]. In Belgium, investigations were subsequently conducted in order to assess whether this particular strain was disseminating from the neighboring country, the Netherlands, and to evaluate its potential threat for the human population and the agricultural sector.
C. burnetii is a small pleomorphic c-proteobacterium with an intracellular life style. Following internalization by a host cell, C. burnetii replicates in a large acidophilic parasitophorous vacuole [19]. Upon changes in the surrounding conditions, the organism undergoes peculiar developmental cycles resulting in a differentiated stage called the large cell variant, which starts replicating exponentially. The replicating phase is followed by a maturation step that leads to a stable C. burnetii form, the small cell variant that is somewhat similar to an endospore in its ability to resist harsh environmental conditions [19].
The infectious potential of C. burnetii is also associated with the structure of its lipopolysaccharide (LPS), which consists in either the virulent phase I (LPS I) or the avirulent phase II (LPS II) [20]. Bacteria in phase I are characterized by a full-length LPS and cause Q-fever in humans or a similar disease in experimental animal models. Upon serial passages in cell cultures or repeated injections in embryonated eggs, virulent C. burnetii undergo a shift to the phase II form characterized by a non-reversible shortening of the LPS molecule [21]. This transition proceeds through gradual reduction of the polysaccharidic part of the LPS, which in turn is accompanied by large genomic deletions [20]. Phase II bacteria are avirulent and do not replicate in immuno-competent hosts.
Because of its intracellular life-cycle, considerable constraints are associated with the laboratory handling of C. burnetii and to the establishment of accurately documented in vitro and in vivo infection protocols. The infectious potential of C. burnetii has been investigated with cell lines [22,23] and animal models, such as guinea pigs [24,25,26,27] and mice [28]. While guinea pigs seem particularly susceptible to C. burnetii infection in aerosol models, the use of mice provides researchers with the benefits of the numerous genetic and biological resources engineered for this model animal to study candidate virulence factors expressed by a given pathogen or to assess the efficacy of candidate vaccines. In many studies, however, repeatability and strain to strain reproducibility are often questionable due to the methodology used to characterize the bacterial inocula. Often the methodology fails in taking into consideration the proportion of live/dead cells (e.g. quantitative PCR-qPCR-based methods) or lacks sensitivity (e.g. microscopic counting of infection foci in cell monolayers).
In this study, we applied in vitro and in vivo infection models to compare the proliferative capacity of the reference Nine Mile strain in phase I (NMI) with that exhibited by two C. burnetii field strains isolated from infected Belgian animals. The latter consisted in one strain isolated from a cow and another one isolated from a goat imported from the Netherlands and displaying a genotypic profile identical to the main C. burnetii strain circulating during the 2007-2010 Dutch outbreaks. We used ethidium monoazide (EMA)-PCR to determine the exact amount of living organisms in the inocula used to infect cultured cell lines or BALB/c mice. The monitored parameters were the proliferative capacity measured by real-time PCR quantification of the number of C. burnetii genome-equivalents over time, the organs' weight and the serological response.

Cell Lines and Culture Media
Human carcinoma HeLa cells were cultivated in Dulbecco modified Eagle medium (DMEM) supplemented with 10% foetal calf serum (FCS) and 0.1 mM of nonessential amino acids solution (all from Life Technologies, Carlsbad, CA). SV40-transformed bovine macrophage cells (SV40 macrophages) were maintained in RPMI1640 medium supplemented with 10% FCS and 2 mM Lglutamine (Life Technologies).

Animals
Specific-pathogen-free (SPF) 6-week old female OF1 and BALB/c mice were purchased from Charles River experimental animal facility (Charles Rivers, Wilmington, MA). Mice were housed in SPF isolated cages (Tecniplast, Buguggiate, Italy) at the L3 animal facility in CODA-CERVA under controlled conditions. Experiments conducted in this study were all approved by the Animal Ethical Committee of CODA-CERVA (Project Approval No. RF 10/6228) and performed in L3 containment.

Isolation of C. burnetii Strains from Infected Animal Samples
C. burnetii strains from animals suspected of infection by the Q fever agent were isolated by a two-step protocol including: (i) an in vivo isolation procedure in OF1 mouse strain (Charles Rivers), in which 200 ml of homogenized abortion material or bulk milk were injected intraperitoneally (i.p.) in 7 week-old female mice at day 1 (first challenge) and day 21 (second challenge). Spleens were harvested at day 32 and analyzed for the presence of C. burnetii by real-time qPCR as previously described [29]. Only those samples displaying a Ct value ,25 were kept for processing in the second step, (ii) which consisted in C. burnetii amplification in seven-day old embryonated chicken eggs obtained by injection in the yolk sac of 100 ml of a 1:10 dilution of the spleen homogenate in phosphate buffered saline (PBS) from the previous step. Egg embryos dying before day 5 were discarded. Depending on egg lifespan and not later than one week after injection, yolk sacs were collected, washed twice in physiological water (Biorad, La Jolla, CA) and homogenated mechanically using the ULTRA-TURRAX Tube Drive system (IKA, Staufen, Germany). DNA was extracted from 200 ml of this homogenate and 1/50 th of the DNA preparation was analyzed for the presence of C. burnetii by real-time qPCR [29]. Only those samples displaying a Ct value ,20 were used to infect cell lines and BALB/c mice.

Ethidium Monoazide (EMA)-PCR and Real Time qPCR
Yolk sacs from C. burnetii-infected embryonated egg were homogenized in PBS buffer before analysis by EMA-PCR. Working dilutions were those described in the results section. EMA (Geniul, Spain) was added to a final concentration of 100 mM. Samples were then incubated for 30 minutes at 4uC in the dark and vortexed regularly. Subsequently, they were exposed to visible light for 30 minutes in an appropriate instrument (PhAST blue system, Geniul). DNA was isolated from 175 ml of EMA-treated samples with MagMax beads (Life Technologies) as described by the manufacturer. One tenth of the recovered DNA sample was used in qPCR analysis according to a real-time PCR procedure targeting com1 and published by Kersh et al. [32]. For quantitative measurements (qPCR), a calibration curve was built up by determining the threshold cycle (Ct) for each serial dilution of a reference sample of our own consisting of a highly concentrated C. burnetii suspension. When dilutions were plotted on a graph in logarithmic scale as a function of Ct values, a calibration curve was obtained. This curve was transformed into an absolute quantification curve by adjusting the curve parameters with the Ct determined on an external reference (LSI) containing precisely 40 genome equivalents per ml.

In vitro Infection Protocol
Cells were suspended in infection medium (DMEM for HeLa cells or RPMI 1640 for SV40 macrophages, both supplemented with 5% FCS) and infected separately with ca. 10 6 'living' C. burnetii organisms by adding adjusted dilutions of the embryonated egg homogenates (Multiplicity of Infection = 100). After 24 hours incubation (37uC, 5% CO 2 ), infected cells were washed three times in 200 ml culture medium. After the third washing, 200 ml fresh medium supplemented with 50 mg/ml Gentamicin (Life Technologies) was added to the wells. Cells were harvested from four individual wells for each time point by adding distilled water (100 ml/well), transferred into 1.5-ml screw capped tubes and heated for 30 min at 82uC. The number of C. burnetii genome equivalents was established with the real-time PCR targeting com1. Uninfected control cells were kept spatially close to infected cells and were always found negative in qPCR, ruling out possible cross-contamination. Growth kinetics of C. burnetii were established on 1 6 10 4 HeLa cells or SV40 macrophages cultivated in 96-well tissue culture plates.

In vivo Infection Protocol
Six-week old female BALB/c mice (Charles River) were injected i.p. with 10 4 C. burnetii cells (200 ml) categorized as 'living' by EMA-PCR conducted on the same day. Dilutions of C. burnetii from stock suspensions were obtained in physiological water (Biorad). Control mice were injected with physiological water only. A group of five mice was sacrificed at each time point for each strain. Serum, spleen and lungs were collected for further analysis (Fig.S1). Overall, 20 mice were enrolled for each strain study.

ELISA Quantification of mice anti-C. burnetii IgM and IgG
Mice IgM and IgG antibody levels directed against C. burnetii were quantified following adaptation of a commercially available ELISA kit (Ruminant Q Fever -Serum/Milk ELISA kit, LSI, Lisieux, France). Analyses were conducted according to the manufacturer's instructions with the following modifications. One hundred ml of diluted mouse serum were deposited on plates supplied with the kit, coated with C. burnetii whole-cell antigens, and incubated for 1 hour at 37uC. Optimal mice serum dilution and anti-mouse conjugate dilution were determined following twoway titrations conducted on both negative samples originating from uninfected animals and positive (C. burnetii infected) samples (Fig. S2). Maximal absorbance within the linear range occurred when mouse serum was diluted 1:20 and all subsequent analyses were performed using serum samples diluted 1:20. For mouse IgM quantification, peroxidase-conjugated goat polyclonal anti-mouse IgM (AbD Serotec, Oxford, UK) was found to best perform at 1:500 dilution. For mouse IgG quantification, the protein-G based kit's conjugate was used at 1:100 dilution. Optical densities were read out at 490 nm on a microplate reader (Perkin Elmer Wallac Victor, Santa Clara, MA). The robustness of this mouse-adapted commercial ELISA was thoroughly validated through repeated measurements taken in independent assays.

Bioinformatics and Statistical Analyses
Genetic relatedness of the MLVA profiles was analyzed by Minimum spanning tree using the Bionumerics software package (Applied Maths, Belgium). GraphPad Prism version 6 (GraphPad Software, San Diego, CA) was used to perform two-way analysis of variance test (ANOVA) with post-test in order to validate strain to strain and time-point variations determined with a significance level set to p,0.05.

Isolation and Genotypic Characterization of Two C. burnetii Field Strains
Following large screening programs involving goat and dairy cattle in Belgian farms in 2009-2011, a number of field samples suspected of C. burnetii infection were collected. A total of 11 goat and 20 bovine farms were found positive and therefore analyzed. These samples contained sufficient bacterial loads for partial genetic analysis by 10-locus MLVA. Minimum spanning tree analysis of the genetic profiles obtained for the bovine samples ( Fig. 1) showed consistent genetic homogeneity, with only one marker varying among those successfully amplified (average of 9.4/1060.7). In contrast, C. burnetii strains from goat samples were of greater variability. Genetic profiles associated with C. burnetii infecting goats were not found in cattle, and vice-versa, providing indications for a discrete, possibly species-specific lineage within cattle herds. Among goat samples, we found one (named CAP3 in Fig. 1) displaying a genetic profile identical to the main strain of the Dutch 2007-2010 outbreaks (named CbNL01). This sample was collected in a farm close to the Netherlands' border in goats imported from this country (as attested by the sanitary label found on the animal ear). From the 31 positive samples, two (CAP3 and BOV1) were kept for further bacterial amplification and characterization (Fig. 1). Isolation was first obtained in OF1 mice as described in Materials and Methods. The complete genotype of these strains was obtained by both MLVA and a single nucleotide polymorphism typing scheme [9,30,31] and were compared to that of the reference strain Nine Mile phase I (NMI). Results confirmed the genetic diversity observed on raw clinical samples and the high level of identity between the CAP3 goat sample and CbNL01 (Table 1). These strains, from here on named CbBEC1 (the goat strain) and CbBEB1 (the bovine strain), were studied for their infectious potential using in vitro cell cultures and a mouse infection model.
Enumeration of Viable C. burnetii Cells by EMA-PCR Because C. burnetti is an obligate intracellular pathogen, inocula can only be characterized by indirect enumeration methods. We therefore applied ethidium monoazide (EMA) qPCR to define the ratio of live/dead C. burnetii in the inocula used for in vitro and in vivo infections. The principle of the test is briefly outlined in Fig2a. Calibration curves were established on serial dilutions of a C. burnetii DNA stock solution by amplifying the single-copy gene com1 (see Materials & Methods). To assess the potential of EMA-PCR in discriminating live and dead cells, we inactivated C. burnetii-containing samples diluted 1:10 5 (corresponding roughly to 10 6 cells/ml) by heating at 72uC for 30 minutes. After this treatment, PCR amplification of C. burnetii DNA was undetectable within 45 PCR cycles, demonstrating template DNA inactivation by EMA and confirming the capacity of the technique to differentiate dead from living cells (Fig. 2b). Reproducibility and repeatability of the technique were evaluated by testing each of three different strains five times by two different operators (Fig. 2c). Overall, results demonstrated the robustness of the EMA-PCR approach for characterizing C. burnetii inocula both quantitatively and qualitatively. To ensure that repeated freeze/thaw cycles did not affect the viability of C. burnetii stock samples, EMA-PCR was conducted on each inoculum prior to infection on the day of the experiment.

Infectious Potential of Three C. burnetii Strains in BALB/c Mice
Because in vitro infection models unlikely account for the complexity of the mechanisms associated with C. burnetii pathogenesis, we also analyzed the infectious potential of NMI, CbBEC1 and CbBEB1 in vivo in a mouse model. Groups of five female BALB/c mice were injected i.p. with 10 4 C. burnetii cells. Spleen, lung and serum were collected 1, 2, 4 and 8 weeks p.i. and bacterial loads, organ weights and antibody responses were analyzed. Bacterial colonization of the spleen was transiently higher for the NMI and the CbBEB1 strains as compared to CbBEC1. This difference was mostly significant 1 week p.i. and characterized by a 3610 4 fold and a 8610 3 fold increase in C. burnetii counts for NMI and CbBEB1, respectively (Fig. 4 upper-left  panel). However, bacterial loads in the spleen decreased with time and both NMI and CbBEB1 were completely cleared from the spleen 8 weeks p.i. (Fig. 4 upper-left panel). An increase in spleen weight was noticed in mice 1 week p.i. with NMI and CbBEB1. Spleens from animals infected with the latter two strains were   Table  S1). Colonization of the spleen by the CbBEC1 strain was lower but constant in time. The latter strain was not completely cleared from the spleen 8 weeks p.i. (Fig. 4 upper-left panel). Colonization of the lung was minimal for the three strains, and no significant difference was observed in terms of bacterial load or organs weight ( Fig. 4 lower panels). At the serological level, IgG and IgM titers were similar for all three strains (Fig. 5). Differences were observed in the associated kinetics, though. IgM titers peaked at 2 weeks p.i. for NMI, 2-4 weeks p.i. for CbBEB1, 4 weeks p.i. for CbBEC1and decreased afterwards (Fig. 5 left panel). IgG responses were maximal at 4 weeks p.i for NMI and 8 weeks p.i. for CbBEB1 and CbBEC1 (Fig. 5 right panel). Altogether, NMI was thus the most actively replicating strain, followed by CbBEB1 and CbBEC1. Serologically, the three strains elicited similar antibody responses with different kinetics. Globally, the field-derived strains showed similar infective behavior in BALB/c mice.

Discussion
The Dutch outbreaks that took place between 2007 and 2010 were unprecedented in the history of Q fever in three aspects. First, these outbreaks were the largest ever recorded with more than 4,000 human cases [33]. Second, related to that, international scientific awareness (and funding) has been brought back to this neglected zoonosis. Third, unprecedented amounts of prospective data were generated at human and veterinary levels to shed light on the etiology, pathology and biology of C. burnetii [33]. Several genotypes of C. burnetii could be identified during the Dutch outbreaks, but genetic diversity was limited and could be rooted back to a major clone termed CbNL01. Circulation of highly infectious C. burnetii strains has been suggested but not demonstrated so far [17].
In this work, we studied the infectious potential of two Belgian field isolates by using in vitro and in vivo models of infection and comparing it to that of the NMI reference strain. One strain included in our study displayed a genetic profile undistinguishable from that of the Dutch CbNL01 strain (as defined by MLVA and SNP typing). Our study also reports the use of EMA-PCR applied for the first time to characterize the live/dead ratios of C. burnetii cells in the inocula used for experimental infection models.
In the absence of an experimental procedure adequately modeling a human Q fever infection in the laboratory [27], cell culture infection and mouse infection by the intra-peritoneal route are provisional options offering plasticity and flexibility to study several aspects of C. burnetii pathogenic potential. Before starting the work with infection models, we questioned current methodologies aimed at determining infective C. burnetii doses. We first looked for a protocol able to discriminate between live and dead bacteria. Quantitation of total C. burnetii loads in infectious inocula can be obtained by quantitative PCR through real-time amplification of the single-copy gene com1 [32,34]. This measurement is appropriate when one single strain is used in the study [35]. However, this method does not provide any information regarding the viability of the detected bacteria, a critical parameter that is traditionally measured through cell culture infections of serially diluted samples [28]. The latter method is tedious and lacks sensitivity. Combining qPCR and EMA dye staining, such as reported here, turns out to be an effective means to achieve viability characterization. EMA selectively penetrates damaged bacterial cells and, upon photolysis, inhibits the PCR amplification of genomic DNA from dead cells, proving direct measurement of intact (living) cells. EMA-PCR is widely used as viability test for various pathogens [36,37,38,39] and it was here successfully applied to C. burnetii. In our hands, EMA-PCR was found robust and reproducible when performed within well-defined dilution limits (from 1:100 in our setting). Non-specific chicken DNA originating from embryonated eggs competes with bacterial DNA in binding EMA dye and is likely to explain the failure of the method to discriminate between live and dead C. burnetii cells at higher sample concentration. The developed procedure specifically applies to the embryonated egg matrix, and needs to be adapted for use on other matrices such as animal tissues or biological fluids.
The NMI reference strain replicated more efficiently than the field strains in both our in vitro and in vivo infection models, an observation that can be explained either by a higher infectious potential of NMI per se [6,41], or a higher adaptation of this strain to in vitro cultures as compared to field strains. In order to exclude differences associated with alterations induced by in vitro culturing, the field strains were processed the same way (one passage on mice and twice on embryonated eggs). Because in vitro adaptation in non immuno-competent hosts might lead to irreversible changes in the LPS structure shifting from the virulent phase I to the avirulent phase II [20], the number of culture passages in embryonated eggs was kept to a minimum.
The Belgian bovine isolate replicated faster than the goat isolate in SV40-transformed bovine macrophages, although both grew slower that the NMI reference in this cell line. The difference was more significantly marked at later time points, ca. 5 days post infection as the probable consequence of the quite long doubling  time of fresh C. burnetii field strains in cell cultures [40]. This observation, together with the apparent genetic homogeneity observed for bovine C. burnetii isolates, suggests that host-specificity might exist within Coxiella strains, as it does for other intracellular pathogens [42]. It has been shown recently that bovine milk products derived from various European countries contained C. burnetii DNA displaying highly similar genetic profiles as assessed by 6-locus MLVA [43]. This homogeneity in profiles also resembled that observed in the bovine isolates reported in the present work. Another study, using whole-genome microarrays, compared the genome content of 52 C. burnetii isolates retrieved from hard ticks, mammals and humans, demonstrated the existence of a single conserved genomotype for hard tick isolates. [44].
The in vivo behavior of C. burnetii has been studied in a variety of animal models. We chose to use BALB/c mice because of the intermediate level of sensitivity of this mouse strain to the Q fever agent and its successful use by several research groups in the past [45]. In our intra-peritoneal infection model, colonization of BALB/c mice by C. burnetii was observed in all tested organs, albeit at low level in lungs. Colonization of the spleen was a good indicator of infectivity and persistence over time, allowing individual C. burnetii strains to be compared to the NMI reference. In contrast with previous works [34,45], correlation between splenic bacterial loads and spleen weight could not be observed by us at individual mouse level. NMI proliferated faster than the two field strains. Clearance from the spleen was observed for NMI and CbBEB1 but not for CbBEC1, which persisted 8 weeks post infection. The latter strain also elicited delayed IgM and IgG responses compared to the two other strains. Our in vitro experiments together with the BALB/c infection model showed that CbBEB1 and CbBEC1 proliferative capacity in the considered hosts are similar. In spite of the fact that one of the used field strain was highly similar -if not identical -to the Dutch outbreak strain CbNL01, no hypervirulent behavior could be demonstrated in mice. Figure S1 Schematic representation of the protocol used for in vivo infection in BALB/c mice. 6-week old female BALB/c mice were injected i.p. with 10 4 C. burnetii living bacterial cells as defined by EMA-PCR. Mice were housed in independent SPF cages for the uninfected control and the infected animals. Groups of five mice were sacrificed 1, 2, 4 and 8 weeks post infection for each strain and for uninfected controls. Spleen, lungs and blood were collected for further analyses. (TIFF) Figure S2 Adaptation of a commercially available ELISA kit for mouse anti-C. burnetii IgM and IgG quantification. Optimal serum dilution and secondary antibody complex/ conjugate concentrations was investigated by two-way titration. Titrations were conducted on negative samples (NC-dashed lines) derived from uninfected animals or from positive samples (PC-full lines) originating from C. burnetii infected mice. The applied conditions were as follows: serum dilution 1:20, anti-mouse IgM conjugate 1:500, kit's anti-IgG conjugate 1:1000. (TIFF)

Supporting Information
Table S1 Data on spleen weight (mg) and associated bacterial load (genome equivalent) obtained from individual mice for each of C. burnetii isolates used in this study. (XLSX)