The collection of wasps and the behavioral experiments comply with the current laws in Italy. No specific permits are required for collection of wasps and colonies. The species used in the experiments is not endangerd or protected in Italy.

In this system, parasite and host meet when they are both immature organisms (Fig 1). Tiny triungulins, the first infective stage of X. vesparum, enter into all larval stages of P. dominula [18, 19], while they are incapable to penetrate into adult wasps: hence the parasite cannot be transmitted horizontally among adults. The abdomen of a parasitized P. dominula wasp shows a typical irregular shape, due to the extrusion of the parasite between abdominal segments (see Fig 1). In summer, the winged male parasite emerges from his puparium, that protrudes from the host’s abdomen, and wanders in search of a receptive female to inseminate, dying shortly after mating [20]. The neotenic female parasite instead is permanently confined within the hemocoel of her host and may overwinter inside the hosts. In the spring, P. dominula wasps infected by X. vesparum females are the vector of the parasite transmission: mature triungulins leave their mother’s cephalothorax and are deposited directly on a wasp nest, where they can infect new larval hosts, or on flowers, where foraging non-parasitized wasps may accidentally pick them up and carry them via phoresy to their own nest [18, 21, 22].

Our main sampling sites were four bushes in the plain of Sesto Fiorentino (Florence, Italy: 43° 50’ 7” N, 11° 11’ 46” E): two bushes of C. radicans (height: 2.5 m; length: 14 and 12 m, hereafter called A and B), one of Rosa L., cultivar (height: 2.5 m; length: 10, hereafter R) and one of Jasminum officinale L. (height: 2.5 m; length: 11, hereafter J). These 4 bushes, prominent landmarks in the flat landscape, all located within a range of 40 m from each other at the corners of a square fence, were intensively patrolled by P. dominula and other insects (pers.obs.). Wasps were collected across 7 years (2009–2015), froze and dissected to evaluate the proportion of parasitized and non-parasitized workers (N = 523, 15-min sampling per bush from 10.00 to 13.00). In July 2015 (1^st week of July, 15-min collection from 10.00 to 13.00), we collected, froze and dissected P. dominula wasps from 13 C. radicans bushes in 4 different sites in Tuscany where we previously found parasitized nests, to control if the preference for this plant was a local phenomenon. These bushes were prominent sunlit landmarks, similar in size to bushes A and B, far apart 10–40 m from each other. These locations are: Fornello (Pontassieve); Chiesina Uzzanese (Pistoia); Pieve a Nievole (Pistoia); Chiocchio (Greve in Chianti).

To evaluate site-fidelity to trumpet creepers, 658 P. dominula wasps (333 parasitized workers, 121 non-parasitized workers, 204 non-parasitized males) were collected from bush A, more attractive than bush B for wasps due to its rich blooming, marked on the thorax with fast-enamel paint and released near the same bush (2^nd and 3^rd weeks of July). At mid-July parasitized wasps were easily recognizable without dissection because the parasite was fully extruded from the wasp abdomen. This mark-recapture study was repeated across 3 years (2011–2013). We carried out 30-min daily censuses on bush A from 10.00 to 13.00 over 6 different days. The wasps that were recaptured at least three times after the first marking were considered as “resident”.

Focal wasps were individually marked and video-recorded at 1 m distance, to avoid any interference (2^nd week of July 2015, 10-min scan periods, from 10.00 to 13.00, bush A). Wasps were collected, froze and dissected at the end of observations to verify parasite’s occurrence, ovary development and the amount of fat bodies. We selected 10 workers parasitized by one X. vesparum (6 with one male and 4 with one female) and 9 non-parasitized workers as control (small body size, no signs of wearing on wings, undeveloped ovaries, scant fat bodies). It was not possible to perform blind observations because the protrusion of the parasite from host’s abdomen is evident. We evaluated the frequency (N events/10 min) of the following behavioral categories: a) flight; b) bud patrol, i.e. feeding on buds and calices, touching the substrate with antennae and mandibles, further split into long bud patrol: 30–60 sec or more; and short bud patrol: < 30 sec; c) leaf patrol, i.e. inspection on leaves, < 20 sec; d) flower patrol, i.e. inspection of the outer corollas, < 20 sec; e) entry into flower; f) aggressive interaction with other wasps/insects approaching the focal wasp, usually chased away after a quick antennation.

Twelve wasps, collected on bush A and parasitized by one or two X. vesparum females, were individually enclosed into vials covered with a mesh large enough for a male parasite, but not for an infected wasp, to pass through, and hung on branches of bush A. After three hours we verified the occurrence of male parasites inside the vials as a proxy to quantify successful mating (1^st week of July 2010, from 10.00 to 13.00).

To assess the occurrence of mature triungulins in the brood canal of female parasites, we dissected under a stereomicroscope (Nikon 645 c-dss) X. vesparum females infecting two sets of workers. 1) Lab workers which emerged from nests collected in Sesto Fiorentino and were housed under controlled laboratory conditions (at 26 ± 2 C° and at a natural photoperiod, water and sugar cubes ad libitum). Workers harbouring parasites of both sexes were caged together until the end of July, when female parasites were analysed (2010, N = 30; 2011, N = 24). 2) Field parasitized workers (bushes A and B, mid-July, 2011–2014), from which 115 X. vesparum females were isolated and analysed. Moreover, we checked under a stereomicroscope the outer and inner corollas of 30 flowers for the presence of triungulins (bushes A and B, mid-July, 2011–2012).

Chi-square test was carried out to compare: the proportion of parasitized and non-parasitized wasps across 7 years among 4 bushes; the site-fidelity among males, parasitized and non-parasitized females across 3 years. Seven spatial behaviors were used as dependent variables into Generalized Linear Model (GLM) to compare parasitized (10) and non-parasitized (9) wasps. We computed a matrix of distances among behavioral frequencies by means of Bray-Curtis dissimilarities, using the vegdist function of the Vegan package for R, and then a principal coordinates analysis (PCoA) with cmdscale R function. We extracted n-1 dimension. A scatterplot of the first two dimensions visualized the dissimilarity pattern of samples. The probability to attribute the sample to its group (parasitized/non-parasitized) was evaluated with a Jack-knife procedure. All the statistical analyses were conducted using SPSS 20.0 and R.

In this long-term field study (2009–2015) we sampled wasps from four flowering bushes, similar in size and sun exposure. Wasps were collected between late June and early July, i.e. during the peak of the worker phase in the colony life cycle. The number of P. dominula ranged between 10 and 30 at each sweep netting (Fig 2). Two thirds of P. dominula workers collected on trumpet creepers (bushes A and B) in June-July were parasitized (i.e. 175 parasitized and 84 non-parasitized wasps), while both rose and jasmine bushes (R and J) were visited mainly by non-parasitized wasps (95%, 248 non-parasitized and 16 parasitized wasps). This asymmetrical distribution of parasitized wasps, more numerous on trumpet creepers, was consistent with no significant difference across the 7 years of the study (chi-square-test, df = 6. A, χ² = 2.33, P = 0.88; B, χ² = 1.69, P = 0.94; R, χ² = 1.85, P = 0.93; J, χ² = 3.22, P = 0.78). The percentage of parasitized wasps collected on two trumpet creepers (A+B, clumped together because not significantly different from each other, χ² = 0.26, P = 0.60) was 66–69%; this rate is significantly higher (χ² = 213.35, P < 0.001) than the 6% of parasitized wasps collected on rose and jasmine bushes clumped together (not significantly different from each other, R+J, χ² = 0.04, P = 0.94).

In August, during 3 years (2011–2013), sporadic checks on bushes A and B revealed the presence of P. dominula males and few aggregations of 2–7 parasitized wasps inside corollas, resting or seeking floral nectar (see August clusters, Fig 1F). We discovered 10 small clusters inside flowers of bush A (mean wasp number = 3.9) and 7 clusters on bush B (mean wasp number = 3.6). This late-summer phenomenon was never observed on rose and jasmine bushes.

To exclude that the attraction towards C. radicans was a local phenomenon, on July 2015 we sampled wasps from 13 flowering bushes of the same species far 30–40 km from our main sampling site. In these sweep-netting collections, parasitized wasps ranged between 22% and 77% (Table 1). A further mid-August screening showed that 8 bushes, first attractive for parasitized females, were then intensively patrolled by P. dominula males. Moreover, 7 small aggregations (mean wasp number = 2.9) were discovered inside the flowers of 3 bushes.

We evaluated site-fidelity (Fig 3) in relation to both wasp parasitism status and sex by means of a repeated mark-recapture study focused on bush A, the most attractive for parasitized wasps (see Fig 2). We recaptured a fifth of parasitized workers (N = 59 resident out of 333, 18%) on bush A for 3–5 days after their marking, and non-parasitized male wasps in a similar percentage (N = 45 resident out of 145, 22%; parasitized male wasps appear only later in the season because of the parasite bivoltine cycle, see below), while non-parasitized workers were observed much less frequently (N = 5 resident out of 121, 4% only, Fig 3). These rates were consistent across 3 years (chi-square-test, df = 2; Parasitized workers: χ² = 1.04, P = 0.59; non-parasitized male wasps: χ² = 0.08, P = 0.95; non-parasitized workers: χ² = 0.08, P = 0.96 respectively) and significantly different between parasitized and non-parasitized workers (χ² = 10.8, P = 0.001). Site-fidelity was not significantly different between parasitized wasps and males (χ² = 1.02, P = 0.31).

Parasitized wasps spend more time (30–60 sec or more) than non-parasitized workers in long patrol of closed and open buds (Wald χ² = 13.52, P < 0.01, Fig 5). Focal parasitized workers on buds chased out an approaching wasp or other insect more frequently than non-parasitized workers (aggressive interactions, Wald χ² = 9.09, P < 0.01). Non-parasitized workers landed on these inflorescences briefly, flying quickly among buds, leaves and flowers; in comparison to parasitized wasps they showed short bud patrols (Wald χ² = 9.08, P < 0.01), leaf and flower patrols (lasting 10–20 seconds, Wald χ² = 6.91, P < 0.01 and Wald χ² = 5.55, P < 0.01 respectively). Entry into flowers to search for ovarian nectary was rare and quick overall in comparison to what was observed for bees and other pollinators (pers. obs). Early in the morning, when wasps were inactive, we observed single wasps sheltering inside the tubular corollas, usually facing outwards, mainly during windy and rainy days. Three checks on 20 flowers, containing one wasp each, showed that the majority was parasitized (13, 17 and 18 parasitized out of 20 wasps, respectively; none was marked). No aggregations inside flowers were detected until August.

The two-dimensional plot obtained after PCoA (Fig 6) showed a good separation between parasitized and non-parasitized groups on the first axis. The jack-knife procedure correctly attributed 17 on 19 specimens (89.5%). The high percentage of correct attribution of individuals to their groups supports the different behavioral frequencies between the two groups.

Four out of 12 vials on bush A, enclosing wasps parasitized by X. vesparum females, harboured one dead X. vesparum male, while one vial contained two males, all dead at the time of our survey. Dissections of X. vesparum females, infecting lab workers which were housed under controlled laboratory conditions, showed groups of mature mobile triungulins inside the brood canal at the end of July, i.e. after 6–8 weeks from workers’ emergence (2010: 17 out of 30 X. vesparum females released triungulins, 57%; 2011: 24 out of 24 females, 100%). At mid-July, dissections of X. vesparum females infecting workers collected from bushes A and B confirmed the development of triungulins also in the field (2011: 32 out of 34, 94%; 2012: 26 out of 37, 70%; 2013: 2 out of 13, 15%; 2014: 13 out of 31, 42%). A screening of 30 C. radicans flowers collected from bushes A and B also showed a few mobile triungulins on the outer and inner corollas, 1–2 per flower (mid-July, 2011: 22 triungulins; 2012: 6 triungulins).