Introduction

‘Query fever’ (Q fever) is a highly infectious zoonotic disease first identified in 1937 [1,2,3,4,5]. The causative agent, the obligate intracellular bacterium Coxiella burnetii, infects a variety of vertebrate species, including humans. Sporadic cases in humans occur annually worldwide, but occasional outbreaks are also common [1,2,3,4]. For example, in the Netherlands more than 4,000 human cases were reported between 2007 and 2010 [6]. While most human cases are self-limiting with fever and fatigue, acute forms range from mild flu-like symptoms to pneumonia or hepatitis. The disease can also become chronic (mainly endocarditis), and, though rarely fatal, remains highly debilitating even when treated with antibiotics [1,2]. Most human cases are linked to contact with infected livestock, especially goats and sheep, which suffer abortion and reproductive disorders. Infection usually occurs by the inhalation of aerosolized resistant small cell variants that are present in the excretions of infected animals. Other modes of transmission including ingestion of unpasteurized milk or dairy products; human-to-human contact is also possible but considered rare [1,2,4,5]. One of the most virulent reference strains of C. burnetii (strain RSA 493 / Nine Mile I [7]) was isolated from a guinea pig on which field-collected Rocky Mountain wood ticks Dermacentor andersoni had fed, suggesting that transmission through tick bites may also occur [8]. The small cell variants of the bacterium can survive and remain highly infectious for long periods in the environment, leading to the classification of C. burnetii as potential bioterrorism agent [9].

The evolutionary origin of Q fever is unclear since the C. burnetii ancestor and its primary lifestyle remain entirely unknown. Historically, C. burnetii was assigned to the taxonomic order Rickettsiales (Alphaproteobacteria), but it has been recently considered more closely related to the Legionellales order (Gammaproteobacteria) because of its genetic proximity to the Legionnaires' disease agent, Legionella pneumophila [10]. The Legionellales order includes many other intracellular bacteria infecting non-vertebrate species, such as, for instance, Rickettsiella species that are both widespread and biologically diverse in arthropods [11,12,13]. Within the Coxiella genus, the only known relative of C. burnetii which has been formally identified is C. cheraxi, a pathogen of crayfishes [14]. Many past descriptions of Coxiella were likely biased toward the detection of pathogenic strains since most C. burnetii isolates were collected from humans or domestic ruminants during Q fever outbreaks [1,5,15]. However, the advent of 16S rRNA gene sequencing as a universal DNA barcoding marker in bacteria has led to the description of a few novel Coxiella-like organisms in non-vertebrate species (listed in [16]), and particularly in ticks [17,18,19,20,21,22,23,24,25]. All these Coxiella-like organisms are closely related, but genetically distinct to C. burnetii, suggesting that some diversity exists within the Coxiella genus. The highly conserved nature of the 16S rRNA gene sequences has prevented researchers from establishing the exact relationship between C. burnetii and Coxiella-like organisms, and a sister clade relationship is commonly assumed [16,18,21,22,23].

The Coxiella-like organisms differ from C. burnetii in their biological traits and some may behave as subtle symbionts engaged in intricate interactions with ticks. In ticks belonging to Ornithodoros, Amblyomma and Rhipicephalus genera, Coxiella-like organisms were found to massively infect ovaries and to be maternally inherited through the egg cytoplasm [18,20,21,26]. In these tick species, the presence of the bacteria in the Malpighian tubules further suggests a possible role in nutrition by potentially provisioning their hosts with essential nutriments [20,21,26]. Indeed, the elimination of these bacteria with an antibiotic treatment was shown to negatively impact the fitness of the lone star tick A. americanum [27]. Accordingly, when the Coxiella-like bacterium found in A. americanum was recently sequenced [16], no recognizable virulence genes were found, indicating that this bacterium is likely non-pathogenic. In contrast, its genome encodes major vitamin and cofactor biosynthesis pathways, suggesting that it may be a vitamin-provisioning endosymbiont. This interaction exhibits the typical hallmarks of maternally-inherited symbionts with essential roles in arthropod biology [28,29]. Such patterns have been found in other exclusive blood-feeding species like bedbugs [30] and tsetse flies [31], two insect groups which rely on a single food source throughout their developmental cycle and harbor beneficial microbes that provide nutrients absent from their restricted diets. The Coxiella-like organisms of ticks share obvious similarities with these beneficial endosymbionts.

Here, we examine the origin of the Q fever pathogen, C. burnetii, by inferring the evolutionary processes that have shaped diversity within the entire Coxiella genus. To this aim, we first sampled an extensive range of ticks, with 58 tick species examined, and developed a sensitive detection method that reveals a wider Coxiella diversity than recognized in past studies. Second, instead of relying solely on the 16S rRNA gene, a molecular marker that is notoriously inadequate for inferring reliable fine-scale phylogenies [32], we used a novel multilocus typing method, allied to Whole Genome Sequencing (WGS) data, and conducted phylogenetic analyses on a large amount of DNA sequence data. Third, we examined two major ecological features of Coxiella-like organisms, i.e. their ability to be maternally-inherited through the tick egg cytoplasm and to grow in a vertebrate cell environment suitable to C. burnetii. Altogether, this corpus of data has led to the characterization of a large genetic and ecological diversity within the Coxiella genus, far beyond the C. burnetii type species. The Coxiella-like organisms of ticks form an ancient lineage of maternally-inherited tick endosymbionts that do not lie as a sister-clade to C. burnetii but rather form a basal lineage illustrative of the ancestral Coxiella life style.

Results

Ticks commonly harbor Coxiella-like organisms

We performed an extensive screening for the presence of Coxiella in 916 tick specimens from 58 species belonging to the two main tick families, Ixodidae (hard ticks, 36 species) and Argasidae (soft ticks, 22 species) (Fig 1 and Table 1). Except for 37 specimens (6 species) derived from laboratory colonies, all other tick specimens were sampled from natural populations in Europe, Americas, Africa, Oceania and Asia (n = 112 localities). In these populations, ticks were collected either in the host habitat or directly on hosts (Table 1). To detect C. burnetii and its relatives, we developed a detection method based on a nested polymerase chain reaction (PCR) using total tick DNA extracts to amplify a 539–542 base-pair (bp) fragment of the Coxiella rpoB gene (Table A in S1 Text).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Geographic origin of the sampled ticks and distribution of Coxiella and Rickettsiella infections.

Square size indicates the number of populations sampled per geographic area. Numbers refer to the tick species whereas letters discriminate the different populations screened within a species; this nomenclature is detailed in Table 1. The colors within squares indicate the infection status of populations. Two tick species were infected by both Coxiella and Rickettsiella at the species level, but not at individual and population levels.

https://doi.org/10.1371/journal.ppat.1004892.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of tick species and populations included in the analysis, with details on their origin, the population sample size, and the prevalence of Coxiella spp. and Rickettsiella spp.

https://doi.org/10.1371/journal.ppat.1004892.t001

Using this procedure, all the tick-borne bacteria we detected belong to the Legionellales order and can be unambiguously assigned either to Coxiella or to its sister genus, Rickettsiella. Whole tick DNA extracts from more than two thirds of the specimens (637 out of 916, 69.6%) and the species (40 out of 58, 70.0%) were found to be positive for Coxiella (Fig 1 and Table 1). Coxiella was found in most tested genera of hard ticks (Rhipicephalus, Ixodes, Amblyomma, Dermacentor, Haemaphysalis) and soft ticks (Ornithodoros, Argas). In almost all infected species, Coxiella was detected in >90% of the examined specimens, indicating high Coxiella prevalence in diverse tick species. For example, infection was apparently fixed in populations of most Rhipicephalus and Ornithodoros species (Table 1). In contrast, Coxiella was frequently absent in Ixodes species and displayed highly variable prevalence in the five infected species (out of 12 screened).

Other Legionellales bacteria of the genus Rickettsiella were found in 52 specimens (5.7%) from six Ornithodoros species and three Ixodes species (Fig 1 and Table 1). In two of the three Rickettsiella-infected Ixodes species, i.e. I. ricinus and I. uriae, Coxiella was also found, but in different individuals and in distinct populations (i.e., no co-infection by Coxiella and Rickettsiella occurred at individual and population levels; Table 1). Adding the Rickettsiella-positive samples, we found that 689 of the 916 examined tick specimens (75.2%) and 44 of 58 screened species (76%) harbored either Coxiella or one of its relatives.

High genetic diversity among tick-borne Coxiella

To characterize Coxiella genetic diversity, we developed a multi-locus typing method based on five conserved bacterial genes including rpoB and four other housekeeping genes: 16S rRNA, 23S rRNA, GroEL and dnaK (Table A and Fig A in S1 Text). Multi-locus sequences were obtained from a subsample of 85 Coxiella- and 12 Rickettsiella-positive tick specimens (one to four specimens per infected species were examined). All five bacterial genes were successfully amplified from 71 Coxiella- and 12- Rickettsiella positive specimens representing 35 Coxiella- and six Rickettsiella-infected tick species. For five other Coxiella-infected species (i.e., 14 individual ticks), only three to four bacterial genes were successfully amplified. The sequences were easily readable without double peaks, indicating that there was no coinfection of Coxiella/Rickettsiella strains in any specimen.

The overall dataset included 33 to 40 alleles per bacterial gene (Table 2) and 51 new multi-locus genotypes (43 in Coxiella and eight in Rickettsiella). Within the Coxiella genus, all pairs of 16S rRNA gene sequences are at least 93% identical (Table 2) and range in threshold values typically used to delineate other Legionellales genera such as Legionella [33] and Rickettsiella [34]. Each of the infected tick species harbored a specific bacterial genotype or a set of closely related genotypes. None of the Coxiella multi-locus genotypes identified in ticks was identical to those of the 15 C. burnetii reference strains (Table B in S1 Text), although some showed moderate levels of nucleotide identity: pairwise identity between the two groups ranged from 77.8% to 97.7%. For each bacterial gene, the genetic diversity was significantly higher in the Coxiella strains of ticks than in C. burnetii as illustrated by the metrics on their respective genetic diversity (Table 2, paired t test, all P < 0.02).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Genetic estimates for 85 Coxiella-like strains and for 15 Coxiella burnetii reference strains.

https://doi.org/10.1371/journal.ppat.1004892.t002

Coxiella burnetii originated from a tick-borne Coxiella ancestor

We constructed a multi-gene phylogeny of the entire Coxiella genus using a dataset that included the Coxiella and Rickettsiella sequences from ticks, the 15 C. burnetii reference genomes, as well as sequences from Legionella spp. and more distant outgroups that were available in GenBank (Table B in S1 Text). The concatenated sequences included 3009 unambiguously aligned base pairs (bp). Prior recombination tests showed that Coxiella and Rickettsiella strains did not exhibit a strictly clonal structure, but rather experienced significant genetic exchanges. We thus applied a sequence-based network approach that does not force relationships to be tree-like but rather incorporates recombination into the phylogenetic reconstruction. The network results (Fig 2), as well as the results from the Maximum Likelihood (ML) tree-based analysis (Fig B in S1 Text), consistently showed that the Coxiella genus can be split into four main clades (labeled A-to-D hereafter) with each clade clustering the Coxiella genotypes found in five to 15 tick species. The phylogenetic analyses also highlight that all C. burnetii isolates cluster into a unique subclade embedded within the A clade (Fig B and C in S1 Text and Fig 2). Notably, the closest relatives of C. burnetii are the Coxiella strains from soft ticks of the Ornithodoros and Argas genera, suggesting that the common ancestor of C. burnetii originated from a Coxiella hosted by soft ticks.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Phylogenetic network with concatenated 16S rRNA, 23S rRNA, GroEL, rpoB and dnaK sequences (3009 unambiguously aligned bp), including 71 Coxiella-like strains of ticks, 15 C. burnetii reference strains, and bacterial outgroups.

The four Coxiella clades are labeled A to D. A zoom on the A clade which contains C. burnetii isolates is shown in Supplementary Fig C in S1 Text. Each number corresponds to one tick species as detailed in Table 1. Blue—Coxiella-like organisms; red—C. burnetii; green—Rickettsiella; black- other bacteria. All multi-locus typing of Coxiella and Rickettsiella of ticks are new sequences from this study. The scale bar is in units of substitution/site.

https://doi.org/10.1371/journal.ppat.1004892.g002

The partitioning of Coxiella diversity among tick species revealed a complex structure, indicating a role for both co-divergence and horizontal transfer events in the evolution of this bacterial group. Closely related Coxiella-like organisms were frequently found in closely related tick species, a pattern suggestive of co-divergence between Coxiella and ticks (Fig B in S1 Text and Fig 2). For instance, all the Coxiella-like organisms found in the 12 examined Rhipicephalus tick species cluster together within the C clade, whereas all the Coxiella-like organisms in the Ixodes species cluster within the B clade (Fig B in S1 Text and Fig 2). Conversely, some Coxiella-like organisms found in related tick species are only distantly related and do not cluster together (e.g., the Coxiella-like organisms of Ornithodoros soft ticks are scattered among the A, B and C clades), a pattern suggestive of horizontal transfers among tick species.

Further analyses were conducted by examining public repositories of DNA sequencing data generated by the whole genome sequencing (WGS) projects of the cattle tick R. microplus and the deer tick I. scapularis. Using the 1,995,281 bp C. burnetii (str. Nine Mile I RSA 493) genome as a probe, we found clear evidence of Coxiella infections in R. microplus, but not in I. scapularis. A total of 31 contigs (514–2,349 bp, totaling 34,990 bp) from R. microplus sequencing were uniquely attributable to Coxiella. They matched 50 genes of C. burnetii with 68-to-100% nucleotide identity (Fig A and Table C in S1 Text and Fig 3A). Alignment of the 31 Coxiella contigs to other bacterial genomes, including the 15 C. burnetii reference genomes (19,304 unambiguously aligned bp), corroborates the finding of our prior five loci-based analyses: the Coxiella strain identified in R. microplus is evolutionarily related, but distinct, to C. burnetii (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Characterization of new Coxiella strains derived from whole-genome sequencing (WGS) of the cattle tick Rhipicephalus microplus.

(a) Percent identity of 50 genes uniquely attributable to Coxiella from R. microplus WGS versus 15 sequenced C. burnetii genomes and other reference genomes. (b) Bacterial phylogeny, comprising the Coxiella found in the R. microplus WGS data, reconstructed from the concatenated sequences of Fig 3A (19,304 unambiguously aligned bp) using maximum-likelihood (ML). Branch numbers indicate percent bootstrap support for major branches (1,000 replicates; only bootstrap values >90% are shown). The scale bar is in units of substitution/site.

https://doi.org/10.1371/journal.ppat.1004892.g003

Discussion

Since its original description, C. burnetii infections have been characterized in a wide variety of hosts. While only two species have been formally identified within the Coxiella genus, we show here that a far greater diversity of Coxiella exists in ticks. We detect the presence of Coxiella-like organisms in many more tick species than previously known [17,18,19,20,21,22,23,24,25] and describe a far wider genetic diversity among these bacteria than previously suspected. The incidence of Coxiella, as well as of its sister genus Rickettsiella, in ticks is exceptionally high, with approximately three quarters of tick species infected. Although possible tick-borne transmission of C. burnetii has been reported [1,2,8], none of the 43 new Coxiella genotypes identified here are identical to C. burnetii. We also demonstrate for genetically divergent Coxiella strains (i.e., members of the A, C and D clades) found in four tick species that infection is primarily transmitted maternally via the egg cytoplasm. These results converge to support the hypothesis that these Coxiella-like organisms are specific endosymbionts of ticks. Phylogenetic evidence further shows that one of the Coxiella-like organisms belonging to the A clade and primarily hosted by soft ticks has served as the progenitor of C. burnetii.

Three complementary lines of argument indicate a much longer evolutionary history for Coxiella-tick associations than for vertebrate-Coxiella associations. The first lies in the broad distribution of Coxiella and Rickettsiella bacteria across tick species, genera and families. The second concerns the extensive genetic diversity found in tick-borne Coxiella strains compared to C. burnetii strains, as illustrated by the clear subdivision of this genus into four highly divergent clades (A-D). Finally, the clustering of all C. burnetii strains within one of the clades of tick-borne Coxiella shows that the ancestor of C. burnetii was a tick-associated bacterium which succeeded in infecting vertebrate cells. The remarkably low genetic diversity of C. burnetii, previously noted in other studies [36,37], indicates a unique and recent emergence of this highly infectious vertebrate pathogen. Interestingly, this hypothesis was initially raised a decade ago from observations of the profound differences in genome architecture of C. burnetii relative to other pathogenic intracellular bacteria [38]. It was again emphasized from the genome sequencing of new C. burnetii strains [39]. Our data brings further support to this hypothesis by demonstrating that C. burnetii roots within the Coxiella phylogeny. Comparative genome sequences of C. burnetii [38, 39] and of the Coxiella-like organism from A. americanum [16] also suggest that Coxiella bacteria differ substantially in terms of genome size and gene content. The C. burnetii genome (A clade) has a size of ca 2Mb [38, 39], whereas the genome of the Coxiella-like organism isolated in A. americanum (D clade) is only about a 1/3 of this size (ie. 0.66 Mb) with a large percentage of missing genes [16]. This reduction in genome size may limit the transition to pathogenicity, and suggests that some Coxiella-like organisms may have evolved towards exclusive and irreversibly specialized interactions with their tick hosts. Overall, the diversity of genome sizes emphasizes that members of the different Coxiella clades may have retained a variety of evolutionary strategies to favour their spread and persistence in their hosts.

We identified Coxiella as a major emerging clade of bacterial endosymbionts allied to ticks. Coxiella-like organisms are maternally-transmitted through the egg cytoplasm at high frequency with 98–100% mother-to-offspring transmission, a pattern also reported in previous studies [20,21,26]. This transmission pattern is the rule for a variety of bacterial endosymbionts that live exclusively within arthropod cells [28,29,40]. While some, like Wolbachia, are globally common symbionts estimated to infect ca. 40% of insect species [41,42], others are globally rare, but common and important in particular arthropod groups [28]. This is precisely the case for Coxiella-like organisms; although they have not been found in other arthropod species, they are commonly associated with ticks. This leads to the obvious question of the phenotypic consequences of Coxiella-tick interactions. In some cases, Coxiella-like organisms of ticks likely act as obligate mutualistic symbionts required to support normal tick development, potentially provisioning their hosts with essential nutriments absent in vertebrate blood [16,20,21,26,27]. The ubiquity of Coxiella in some tick groups-such as in the Rhipicephalus genus in which infection is at fixation- corroborates the hypothesis of an obligate endosymbiont. This is not, however, the case for all tick species since some, such as I. ricinus and I. uriae, harbour Coxiella-like organisms at much lower frequencies. In these tick species, Coxiella is more likely to behave as a conditional mutualist-i.e., that confers advantages under certain environmental conditions- or as a reproductive parasite-i.e., that manipulates host reproduction toward the production of daughters (the transmitting sex), as commonly observed in arthropods with a variety of facultative symbionts [28,40]. It should also be noted that other endosymbionts also occur in ticks and may have evolved under complex multispecific interactions [17,24]. For instance, whereas the soft tick O. moubata was not found to be infected by a Coxiella-like organism in the present study, this tick species has been found to be infected by an endosymbiont belonging to the Francisella genus [17]. Endosymbionts other than Coxiella may thus interact with ticks, a pattern suggesting that endosymbiotic systems can be dynamic across tick lineages. These different hypotheses will now require specific testing.

Another question remains concerning the degree of vertebrate infection risk by the Coxiella-like organisms of ticks. Ticks are found worldwide and blood-feed on many different hosts; a combination of traits that may facilitate tick-to-vertebrate transfers of Coxiella. However, the bacteria observed in this study seem confined to ticks and, to our knowledge, none have ever been isolated from a vertebrate or associated with clinical symptoms. This suggests that these tick-associated bacteria currently pose a much lower infection risk to vertebrates than C. burnetii. As discussed above, the genome reduction of the Coxiella-like organism isolated in A. americanum, with the lack of nearly all the genes associated with pathogenicity [16], corroborates this view. Moreover, the inability to grow tick-borne bacteria in vertebrate cells highlights the significant barrier that must be overcome by the bacteria to successively achieve tick-to-vertebrate transmission. This type of transmission may, nonetheless, occasionally occur; an avian Coxiella-like organism was recently reported to induce fatal systematic infections in domestic birds [43,44,45]. A very similar infection pattern was found for another maternally inherited endosymbiont, Arsenophonus, a widespread bacterium in different insect groups [41,46,47,48,49]. In particular, some Arsenophonus strains were detected in the phloem of plants fed on by infected phytophagous insects and were assumed to be opportunistic plant pathogens [50]. In such cases, the plant host may act as an ecological arenas for the global exchange of endosymbionts like Arsenophonus, serving as a possible intermediate host for the horizontal transfer of bacteria among insect species [48]. In the case of Coxiella-like organisms, the extent of exchange between different tick species via the vertebrate host is yet to be established, but could be favoured by tick co-feeding (ticks feeding in close proximity on the host). The genetic similarity between Coxiella-like organisms found in unrelated tick species highlights the capacity to shift tick host species. Future research is now needed to assess the potential of different Coxiella-like organisms to infect vertebrates.

The reasons why C. burnetii is a highly virulent pathogen of vertebrates, but not Coxiella-like organisms (especially those from the A clade) remain unknown. As an intracellular pathogen with airborne transmission, C. burnetii has evolved specific mechanisms to survive in the abiotic environment, as well as to infect and exploit vertebrate cells [15]. Several evolutionary pathways may explain the acquisition of the genetic material necessary for this major lifestyle transition; this includes spontaneous genetic mutations in the genome of a tick-Coxiella ancestor, or the more likely transfer and integration of virulence genes from a co-infecting pathogen. The opportunity of gene transfer among bacteria, irrespective of their pathogenic or symbiotic properties, relies on their frequent co-occurrence within the same tick host [25,51,52]. The Coxiella-like organisms of the A clade may have dynamic genomes as observed in many arthropod symbionts: although they reside in confined intracellular environments, arthropod symbionts commonly experience variable degrees of recombination and gene transfer [53,54,55,56,57]. These gene transfers have served as immediate and powerful mechanisms of rapid adaptation in many endosymbionts, such as Wolbachia [56] and Hamiltonella [55,57]. This mechanism may explain the evolutionary transition from a Coxiella tick-symbiont of the A clade to the vertebrate pathogen C. burnetii. Other genetic connections are also possible; several C. burnetii genes that may contribute to major virulence traits, such as tissue tropism, are similar to eukaryotic genes and may have been acquired through lateral gene transfers from eukaryotes [38,39]. Detailed studies of virulence genes in C. burnetii and their homology with Coxiella-like organisms of the A clade will now be necessary to understand the remarkable emergence of the Q fever agent.

The evolutionary transition observed within the Coxiella genus is one of the rare cases reported to date of an arthropod-inherited symbiont evolving metabolic adaptations leading to the emergence of a vertebrate infectious disease. Another such transition occurred in the Rickettsia genus. The best-known members of this genus are transmitted by blood-feeding arthropods and are pathogenic in the vertebrate host. However, in recent years, many maternally-inherited Rickettsia endosymbionts found exclusively in arthropods have been discovered [58,59]. The examination of the evolutionary history of the Rickettsia genus revealed that this bacterium originated from endosymbionts of invertebrates and only secondarily became vertebrate pathogens [58,59]. Like Coxiella, some Rickettsia species of blood-feeding hosts have underwent a horizontal transmission through a vertebrate host, leading to pathogen emergence. Other bacteria, such as Arsenophonus [47,48] and Sodalis [60] may have had similar life cycle transitions, but the case of Coxiella is unique in that the arthropod host is no longer required to complete its life cycle.

In conclusion, we show that C. burnetii arose from a rare and recent event: the evolutionary transformation of a maternally inherited endosymbiont of ticks into a specialized and virulent pathogen of vertebrates. This raises a series of exciting questions related to both how Coxiella endosymbionts made the major evolutionary transition leading to the emergence of Q fever and their role in the population dynamics of ticks. Identifying the evolutionary processes that transform symbiotic bacteria into emerging pathogens will require further exploration into the biology of the entire Coxiella genus.

Methods

Supporting Information

Acknowledgments

We are grateful to X. Bailly, A. Barbour, L. Beati, T. Boulinier, J.L. Chapuis, F. Dantas-Torres, A. Estrada-Peña, M. Gauthier-Clerc, L. Gern, D. González Acuña, J. Gonzalez-Solis, U. Guidon, A. Guglielmone, K. Hansford, M. Hipfner, A. Jaeger, D. Jenny, M. Labruna, R. Lane, M.S. Latrofa, C. Lebarbenchon, J. Liu, J. Medlock, S. Nava, A.C. Norte, U.E. Schneppat, D. Spratt, P. Tortosa and M. Vittecoq for help at different stages of this work, and to three anonymous reviewers for helpful comments and suggestions.