The aim of the current study was to analyse two steps in the life cycle of a bacterial parasite, characterize the specificity of the interaction, with regard to genetic and environmental factors, and relate these findings to what is known about host-parasite coevolution in this system. We focused on the activation of the parasite's resting stages and on the attachment of the activated spores to the host tissue where it enters the host. Our study revealed that P. ramosa spores captured by the filter feeding Daphnia are indiscriminately activated by every Daphnia clone and Daphnia species tested (Tables 1 and 3). Furthermore, activation was not only found to be independent of the host genotype or species and host gender but also of the environmental conditions (namely, density, temperature and food conditions). The following step of the infection process, however, the attachment of the activated spore to the oesophagus wall of the host, depended strongly on the combination of the D. magna and parasite genotype, but not on the host's gender, nor the environmental conditions in which they were kept (Tables 1, 2, 3).
Previous studies with the Daphnia-Pasteuria system were not able to disentangle the activation, attachment and proliferation steps. Thus, variation in infection success as reported in earlier studies [19, 32, 35–40] may be explained by the combined effects of these steps. However, the binary polymorphism found in infection trials with high doses of single parasite clone  correlates perfectly with the results of our attachment-test (Table 1). This suggests that only Pasteuria clones able to attach to the oesophagus are able to infect the host. Ben-Ami et al.  proposed that D. magna might be either completely resistant or susceptible to P. ramosa depending on the genotype-genotype interaction. They called this the 'binary infection hypothesis'. Our data are consistent with this hypothesis and further pinpoint which specific step of the infection process is responsible for the high degree of specificity. For a given combination of host and parasite genotypes, the activated spores are either able to attach and then infect or they do not attach and do not infect. We did not see any evidence for a graded (quantitative) form of interaction.
Spore attachment is a key step in Daphnia-Pasteuria coevolution
The Daphnia-Pasteuria system has become one of the prime examples of antagonistic coevolution. Host and parasites show strong genetic effects for resistance, virulence and infectivity; genotype × genotype interactions have been reported within and across populations and selection acts rapidly in natural populations [18, 19, 40, 41]. Our study suggests that the parasite-dependent  host population structure and the coevolution  described for this system are mainly driven by the properties of a unique step, the attachment step. First, this step revealed very strong host genotype by parasite genotype interactions (Table 1). Second, the attachment step is independent of the environmental conditions. Third, a recent study of D. magna - P. ramosa coevolution using resurrected host and parasite isolates from lake sediments showed a signal of fluctuating selection only for infectivity but not for parasite virulence . Virulence (the parasite's effect on infected hosts) was also observed to evolve but at a slower rate . The authors proposed that the difference between the evolution of virulence and infectivity resulted from different genes contributing to these traits. Here we give a mechanistic explanation for this finding. Infectivity depends on the attachment and, most likely, on the ligands present on the host and on the parasite. On the other hand, expression of virulence may depend on the host's immune response during the within-host proliferation step. It is likely that these processes are determined by different sets of genes.
The identification of the attachment step as the key step in the coevolutionary dynamics in this system will allow us to improve our understanding of the patterns of antagonistic coevolution. For example, evolutionary models studying the coevolution of the infectivity and the virulence steps  can fit our system in relation to the coevolution of the attachment and the proliferation steps. Those models typically characterize infection outcomes as binary (Yes/No), while empirical data suggest they are more quantitative [15–17]. We show that we can observe a binary outcome when individual steps of the infection process are considered. Furthermore, our method provides a fast and reliable way to test individuals and populations for their susceptibility to Pasteuria. Ongoing research in our group showed that up to 400 Daphnia individuals can be tested in a day (P Luijckx, in preparation). The assay we developed makes it possible to test for susceptibility without the potentially confounding effect of the within-host proliferation step in the infection trials.
From the environment to the host body cavity
The resting endospores of P. ramosa can remain dormant for decades under harsh environmental conditions [24, 29]. Before attachment to the host, the spores need to be activated (Figure 1d). The filter-feeding Daphnia capture particles, including parasites, from the water and transport them on a mucus-layered pathway from the phyllopods to the mouth. During this process, the parasite's exosporium opens by an unknown trigger, releasing the activated spore form within less than 10 min (Figure 1). Despite the fact that spore activation is a necessary step for the infection, this step is entirely unspecific with regard to Daphnia species and clone, host gender and the environmental conditions (Tables 1, 2, 3). The signal that triggers spore activation may be related to chemical substances in the mucus of the filtering apparatus, but other factors (for example, mechanical) cannot be excluded.
Once the activated spore enters the oesophagus, it will attach to the oesophagus wall if the host and the parasite genotype are compatible. There it presumably penetrates the gut wall and enters the host's body cavity. A similar attachment process on the cuticula is also known from P. penetrans but, in this case, the parasite seems to be able to attach to any area of the nematode's body surface . It has been proposed that the lower part of P. nishizawae attaches to the host because this part is densely covered by microfibres . In contrast, in P. ramosa it is the upper part of the peripheral fibres (Figure 1e) that are most densely covered with a layer of microfibres. These fibres may be involved in the attachment (Figure 1f).
An endospore adhesin epitope, situated on the exosporium of P. ramosa, has been identified and it has been suggested that it may be a ligand responsible for the recognition and the binding onto the host . However, according to our results, it is unlikely that this epitope is involved in the attachment because the exosporium of P. ramosa is removed during the activation step. A later study, analysing surface proteins of P. ramosa spores by two-dimensional gel electrophoresis, proposed that a collagen-like protein may be responsible for the binding onto the host but might suffer the same problem of the previous study . We propose that later studies on candidate proteins responsible for the specific attachment to the host in this system should investigate the spores once activated.
The development of Pasteuria, from the moment they attach to the oesophagus until the vegetative stage can be detected in the hemolymph (about 8 days at 20°C ), is unknown. Also, the penetration mechanism is poorly described. Sayre and Wergin  show a transmission electron micrograph of P. penetrans with a structure they call a 'germ tube' crossing the host epithelium. Our hypothesis is that the endospore makes a hole across the host epithelium and injects its cortex into the host. As one response of Daphnia to wounding is an increase of Phenoloxidase (PO) activity , one might expect the penetration process to trigger an immune response but this remains an open question. However, resolving the infection process will allow the study of the immune response during the proliferation step without the confounding effect of genetic variation in the attachment step.
Environment effects and the proliferation step
We found that environmental effects do not influence the activation and attachment step (Table 2). Excluding these steps, we suggest that the proliferation step is the one responsible for the reported sensitivity of the overall infection process for environment effects [32, 34]. The activation and the attachment steps seem independent of the host's immune system (defined as a system that is potentially able to kill parasites), while the proliferation step is likely to be governed by the host's immune system. The immune system may lead to variation between and within those Daphnia clones that allow Pasteuria attachment (and, thus, enable the parasite to enter the host), thereby contributing to local and temporal adaptation, maternal effects and induced resistance [29, 34, 48]. We suggest that future studies on host immunity should use only Pasteuria clones that can attach to a given clone of Daphnia so that all the variation observed is likely to originate from variation during the proliferation step. These factors highlight the importance of controlling the host and parasite genotypes and breaking down the infection process in order to understand the respective role of each step in host-parasite interactions.
Resolving the infection process leads to a better understanding of host-parasite interactions
Resolving the infection process in its sequential steps has been proposed in a number of theoretical models [10, 11] but experimental data are scarce. Our approach is transferable to other host-parasite systems and our results suggest that this can provide important new insights into host-parasite interactions and their evolution. Increasing the degree of the resolution of the infection processes highlights a universe of possibilities of the different levels at which host and parasites interact. The different steps might differ in how they are influenced by the environment. They might also differ in which sets of genes regulate them. As it is probably the case for our study system, different steps of the infection process might follow distinct evolutionary dynamics and be explained by different model (for example, balancing selection, directional selection) [10, 11]. However, because of the sequentiality of the steps, it is possible that the selection on one might depend on the selection on other steps. We propose that analysing infection as a succession of well characterized steps will help to reconcile the empirical data with predictions based on alternative coevolutionary models (for example, Red Queen and Selective Sweep models).
Spores of all P. ramosa clones tested, and which were isolated from natural D. magna populations, were activated by all D. magna clones as well as by six other Daphnia species (Table 3) and even a Cladoceran from a different genus, Simocephalus vetulus. Also, apart from the natural host, D. magna, D. dolichocephala, also became infected following attachment of the activated spores to the host oesophagus. This suggests that the triggers for spore activation and, to a lesser extent, for attachment are phylogenetically conserved. This may facilitate the host range evolution of the parasite. Indeed, despite its high specificity on the level of the host clone, P. ramosa infections have been reported in several species within the family Daphniidae . It will be necessary to test more clones of different Daphnia species in order to determine their pattern of susceptibility and resistance to the parasite. Importantly, phylogenetically conserved steps of the infection process can be ruled out as major factors in coevolution, but are, perhaps, the most appropriate targets for vaccine and drug development. In fact, the genes involved in some infections steps have been worked out for some systems [50, 51] and can be of use in biomedicine for diseases control [52, 53].