Patterns of ea infestation in rodents and insectivores from intensied agro-ecosystems, NW Spain

Background Fleas frequently infest small mammals and play important vectoring roles in the epidemiology of (re)emerging zoonotic disease. Rodent outbreaks in intensied agro-ecosystems of NW Spain have been recently linked to periodic zoonotic disease spillover to local human populations. Obtaining qualitative and quantitative information about the composition and structure of the whole ea and small mammal host coexisting communities is paramount to understand disease transmission cycles and to dilucidate the disease-vectoring role of ea species. The aims of this research were to: i) characterise and quantify the ea community parasiting a small mammal guild in intensive farmlands in NW Spain, and ii) determine and evaluate patterns of co-infection and the variables that may inuence parasitological parameters. Methods We conducted a large scale survey stratied by season and habitat of eas parasitizing the small mammal host guild. We report on the prevalence, mean intensity, and mean abundance of ea species parasitizing Microtus arvalis, Apodemus sylvaticus, Mus spretus and Crocidura russula. We also report on aggregation patterns (variance-to-mean ratio and Discrepancy index), co-infection by different ea species in hosts (Fager Index), and used Generalized Linear Mixed Models (GLMM) to study ea parameter variation according to season, habitat and host sex. Results Three ea species dominated the system (99.4%), namely: Ctenophthalmus apertus gilcolladoi, Leptopsylla taschenbergi and Nosopsyllus fasciatus. Results showed a high aggregation pattern of eas in all hosts. All host species in the guild shared C. a. gilcolladoi and N. fasciatus, but L. taschenbergi mainly parasitized mice (M. spretus and A. sylvaticus). We found signicant male-biased infestation patterns in mice, seasonal variations in ea abundances for all rodent hosts (M. arvalis, M. spretus and A. sylvaticus), and relatively lower infestation values for voles inhabiting alfalfas. Simultaneous infections by two or three ea species occurred in 36.8% of all hosts, and N. fasciatus was the commonest ea co-infecting small mammal hosts. Conclusions The generalist N. fasciatus and C. a. gilcolladoi dominated the ea community, and a high percentage of co-infections with both species occurred within the small mammal guild. Nosopsyllus fasciatus may show higher competence of inter-specic transmission, and future research should unravel its role in the circulation of rodent-borne zoonoses.

host sex, age, condition, immune function), the effect of environmental habitat and conditions (temperature, humidity) and co-infection, are all relevant aspects that need to be quanti ed to build any baseline knowledge required to understand ea life cycles and their relative ecological and epidemiological roles [5,6].
Many eas parasitize rodents [1], which account for 25% of all living mammals and act as the main host type for >80% of all known ea species [5]. Rodents are a key mammal group in terms of public health, as they are involved in the ampli cation and spillover of many zoonoses affecting humans globally [7], and their eas often play signi cant vectoring roles in the transmission cycles of disease [7][8][9]. In NW Spain, common vole (Microtus arvalis) populations massively invaded lowland agricultural landscapes between 1970-1990s, putatively colonizing newly irrigated fodder crops from natural peripheral mountainous habitats [10][11][12]. In recently colonized farmland, common vole populations are cyclic [13], and periodically become a crop pest when overabundant, causing serious public health impacts due to the ampli cation and spillover of zoonotic diseases like tularemia [10,14,15]. In these intensively farmed landscapes, common voles coexist in the same microhabitats with other sympatric rodents and insectivores (mainly, wood mouse Apodemus sylvaticus, Algerian mouse Mus spretus, and greater whitetoothed shrew Crocidura russula) [16]. Characteristics of the host (morphological, physiological, immunological, behavioural and phylogenetic traits) and their shelters (i.e., burrows, nests) are critical to ea lifecycles [1,5]. In other study systems, ea speci city is an important trait in uencing the ea community [17,18]. Fleas can infest hosts phylogenetically close [18], switching between coexisting species within guilds [19]. Host density is also a relevant factor to consider since it involves variations in ea richness [20,21]. Furthermore, climatic conditions, local factors and speci c host features shape infestation patterns at the local level [18,[22][23][24]. Due to all these sources of variation, general patterns should not be inferred but studied in detail in local ea communities.
In the Palearctic region, up to six different taxonomic families of eas can infest small mammals [1], and 68 ea species have been identi ed to datef in the Iberian peninsula [25]. Three of these ea families occur in NW Spain: Ceratophyllidae, Ctenophthalmidae and Pulicidae [26][27][28][29]. Recent surveys reported that the main ea species infesting common voles in intensive farmland of NW Spain were Ctenophthalmus apertus, Nosopsyllus fasciatus and Leptopsylla taschenbergi [30].
Fleas parasitizing common voles in NW Spain are known to harbour zoonotic bacteria such as Francisella tularensis (i.e., the etiological agent of tularemia) and several Bartonella species (agent of bartonelloses) [30]. Yet, nothing is known about ea distribution, relative abundance and co-infection patterns within and between the small mammal host guild. Improving our basic knowledge on how eas interact with their local hosts in farming landscapes will aid in the understanding of disease circulation in landscapes frequently scourged by rodent-driven zoonoses. Here we report on the patterns of ea infestation in the small mammal community inhabiting the intensively farmed landscapes in NW Spain dominated by colonizing common voles. Speci cally, we document and quantify ea-host speci city and describe patterns of abundance, prevalence, intensity and aggregation of each ea species on each of the main small mammal hosts. We also evaluated patterns of ea co-infection in hosts and studied relative abundance variation according to season, habitat (i.e., crop type) and host sex.

Study area
The study was conducted in intensively farmed landscapes in Castilla-y-León region, NW Spain. These landscapes are steppe-like crop mosaics dominated by cereal elds (mainly wheat and barley) with scattered irrigated crops (e.g. sweet beet, sun ower, corn and alfalfa) and interspersed by remnant seminatural vegetation (fallows or set-aside, eld margins, grassy road verges and wild ower strips; [16]). Climate is continental-Mediterranean, characterized by wide seasonal temperature oscillations: long, cold and humid winters with frequent frost events, followed by dry and hot summers with variable but persistent drought periods; precipitation is mostly concentrated during spring and autumn [32].

Small mammal trappings
Fieldwork consisted of seasonal live trappings (March, July and November) conducted at three independent localities (>60 km apart) in the provinces of Palencia (42°01´N, 4°42´W), Valladolid (41°34´N, 5°14´W) and Zamora (41°50´N, 5°36´W), all within the Castilla-y-León region. Between July 2009 and July 2015, we monitored 6 study areas (two replicates per locality, each replicated consisted of an area of ca. 40 km 2 ). In each study area, we sampled the three most relevant habitats: cereals (most abundant crop type), alfalfas (most favourable crop type in terms of cover and food availability for voles) and fallows (see [16] for more details on trapping procedures and habitat use by voles). In brief, we randomly selected 12 elds (4 cereals, 4 alfalfas and 4 fallows) amongst those available in a given area, avoiding sampling the same locations during consecutive seasonal trappings, and set 25 traps in each eld (8 cm × 9 cm × 23 cm; LFAHD Sherman©) interspaced 2 m, and deployed in a T-line shape from eld margins towards the inside of crop elds [16]. Traps were baited with carrot and apple and set during 24h. Each trapped animal was provided with a unique code (we noted the date, site and crop eld where it was trapped). For this study, we captured and sampled 2,254 small mammals belonging to the species M. arvalis (61.2%), A. sylvaticus (23.1%), M. spretus (13.5%), C. russula (1.9%) and other species (0.3%).

Flea collection from trapped animals and identi cation
Captured rodents and shrews were individually transferred to lab-animal plastic cages (29 x 22 x 14 cm; Panlab®). In order to follow the ethical legislation about welfare of animals used in research [37] and to maximise the number of animals that arrived alive to the laboratory, we provided captured animals with food, water and bedding material until the euthanasia procedure. In the lab, each animal was sexed, weighed and euthanized with CO 2 , following a humane protocol approved by the University of Valladolid Ethical committee in animal research (code: 4801646). Immediately after death, eas were carefully collected from each animal by blowing the fur and combing it with a lice comb while holding the animal over a white plastic pan (520 x 420 x 95 mm) half-lled with water. Fleas from each individual were counted, collected from the water surface using a pair of tweezers and stored in individually labelled tubes lled with 70% ethanol [38,39]. We ensured that no eas were missed from each individual by placing the animal carcasses in sealed plastic bags and leaving them for one hour in the fridge before checking again for eas. Fleas were subsequently studied with a 10x and 40x optical microscope (Nikon Optiphot-2) and identi ed at the species level using dichotomous keys [40]. We collected a total of 4,715 eas from 1,239 small mammal hosts: 3,900 eas from M. arvalis (n= 941), 698 from A. sylvaticus (n = 238), 87 from M. spretus (n = 49), 14 from C. russula (n = 6), and 16 from other small mammals (n = 5). A total of 4,266 individual eas were identi ed.

Data analysis
For each host species, we obtained information on the prevalence, mean abundance and mean intensity of each ea species (as de ned in Bush et al. [41]). Data were summarized as prevalence ± 95% con dence intervals (CI; traditional Clopper-Pearson con dence limits), and mean intensity or abundance ± standard error. We also quanti ed the level of skewness of the ea distribution on hosts (a measure of the asymmetry of the probability distribution of a real-valued random variable about its mean) using two complementary indices: (i) the variance-to-mean ratio (VMR), and (ii) the Discrepancy index (D) following Poulin [42]. Co-infection was also quanti ed (hosts infested by more than one ea species). These descriptive statistics were obtained using the Quantitative Parasitology (QPweb) software version 1.0.14 [43]. We used the Fager Index [44] to determine the degree of co-occurrence of ea species, regardless of abundance variations. This index ranges between 0 (species never infest simultaneously) and 1 (species always co-occur) and it is calculated as follows: Co-infection differences according to host sex were tested using Pearson's chi-square test or G-test chisquare test, depending on minimum sample sizes. Relating to rodent hosts, we studied ea prevalence, mean intensity and mean abundance variation according to sampling month, crop type (except for M. spretus because of the small sample size), and host sex using Generalized Linear Mixed-effects Models (GLMM), with a negative binomial distribution for mean intensity and mean abundance. Models included the factors year (2009-2015) and trapping site (Palencia/Valladolid/Zamora) as random terms (to account for possible temporal or spatial variations) whenever possible, and the factors habitat (crop) type (alfalfa/cereal/fallow), host sex (male/female) and season/month (November/March/July) as explanatory variables. Because of sample size limitations, some mixed models did not converge and we then had to include site or site and year as xed effects instead of random effects. The model selection followed a backwards-selection procedure (using the "drop1" function in R), removing non-signi cant terms (we report both signi cant, P = 0.05 level, and marginally signi cant effects, P = 0.10 level, effects). Differences between levels of the categorical factors (crop type and month) were tested using post-hoc Tukey tests. Statistical models were carried out using the "lme4" [45] and "R2admb" [46] packages, and Gtest chi-square test using "RVAideMemoire" [47] of the R3.6.1 software [48].

Flea community
The ea community included the following species: Ctenophthalmus apertus apertus (n = 2), C. a. gilcolladoi (n = 1879), C. baeticus baeticus (n = 11), Leptopsylla taschenbergi amitina (n = 460), Nosopsyllus fasciatus (n = 1903) and Rhadinopsylla beillardae (n = 8); in addition, two specimens were identi ed at genus level only (Ctenophthalmus spp.). Two species dominated the small mammal ea community: N. fasciatus and C. a. gilcolladoi (frequency = 44.6% and 44.1%, respectively), followed by L. taschenbergi (10.8%). Patterns of ea infestation differed between small mammal host species (Table 1 and Figure 1; complementary information in the additional Figure S1). The most abundant ea species infesting common voles were N. fasciatus and C. a. gilcolladoi accounted for 98.5% of all eas identi ed (48.5% and 50.0%, respectively). By contrast, L. taschenbergi was the most abundant ea infesting mice (56% and 48% of all eas for A. sylvaticus and M. spretus, respectively). Regarding shrews, we found that individuals were only infested by C. a. gilcolladoi and N. fasciatus. Other Ctenophthalmus spp. different from C. a. gilcolladoi (C. a. apertus, C. baeticus and Ctenophthalmus spp.) were seldom found in M. arvalis (in one, eight and two animals respectively). The ea R. beillardae was occasionally identi ed in the most abundant rodent species (two eas in tow M. arvalis, ve eas in two A. sylvaticus, and one ea in one M. spretus).
Regarding L. taschenbergi, we found that this ea was more abundant on males than on females prevalence: Χ 2 = 7.15, P = 0.028; intensity: NA; abundance: Χ 2 = 9.01, P = 0.011). The infestation was more prevalent and severe in July and among males than females in both host species. Furthermore, crop type explained variation in intensity and abundance A. sylvaticus (intensity: Χ 2 = 7.86, P = 0.020; abundance: Χ 2 = 6.15, P < 0.046), with a greater intensity those from cereal compared with fallows.

Co-infections
The majority of hosts were infested with one or two ea species (63.2% and 34.5% respectively).

Discussion
The ea community parasitizing the small mammal guild studied here was mainly (99.4%) composed by N. fasciatus, C. a. gilcolladoi and L. taschenbergi, and showed species-speci city and marked aggregation patterns. We found strong sex-biased differences in mice host, lower ea infestation in voles captured from alfalfas, and seasonal variations differing between host and ea species. Interspeci c coinfections were frequent in some hosts, with up three different ea species.

Flea community
The northern rat ea (N. fasciatus) was the commonest and most abundant ea species in the studied small mammal community, parasitizing both rodents and insectivores. This ea mainly parasitizes rodents but can facultatively infest a wide range of mammalian hosts [49], which can explain their overall high abundance and prevalence rates. A pattern of generalist eas reaching heavier infestations have been also determined in other systems [50]. Ctenophthalmus eas are a generalist group that diverges on their distribution range owing to geographical speci city [51]. Although all Ctenophthalmus species identi ed here are endemisms of the Western Mediterranean area, C. a. gilcolladoi is the typical ea of the open habitats of this region [51,52]. L. taschenbergi was the most abundant ea found on A. sylvaticus and M. spretus, consistent with its well-known mouse-speci city [52] and low infestation rate on common voles.
Our results show the typical aggregation pattern of eas on rodents found elsewhere [53]. Poulin D-Index values for the most abundant eas in M. arvalis (N. fasciatus and C. a. gilcolladoi) are lower than in mice. Behavioural traits could explain these differences [19], since the social behaviour of common voles could facilitate the switching of eas between hosts, reducing ea aggregation. Regarding variance-to-mean ratios, we found differences between N. fasciatus and C. a. gilcolladoi despite having similar values of abundance. The host-generalist strategy of N. fasciatus [5] could explain the reduced aggregation level. Generalist parasites with broad habitat requirements, like N. fasciatus, would have access to more hosts, having more resources available and reduced intraspeci c competition. Another important aspect to consider is the local asynchrony in abundance uctuations of the rodent species (summer peak in voles versus autumn peaks in mice; [13,54]). Thereby, a high number of potential hosts are accessible to generalist eas through different periods of the year, potentially causing a dilution effect for N. fasciatus and lower aggregation levels. Conversely, highly specialised eas (such as L. taschenbergi) would tend to parasitize in the fewer suitable hosts that are available (i.e. A. sylvaticus and M. spretus).
Variations of ea parasitological parameters variation according to season, crop type and host sex M. arvalis harboured the highest ea burden and infestation rate, followed by A. sylvaticus. Larger body size and greater complexity of the burrow system can lead to heavier ea infestations [55], which could explain the lower values in M. spretus and C. russula (although the sample size of the later is too small to draw solid conclusion). Moreover, the fossorial life of M. arvalis could increase the probability of getting infested by naïve eas because burrows are used by preimaginal stages to develop until their rst blood feeding [5].
We found seasonal variations in the main ea species whatever the rodent host. In general terms, prevalence models could be easily understood from a ea phenology point of view, since maximum values occurred during the season with the greatest ea activity season: C. a. gilcolladoi showed higher presence in March/November, with a signi cant drop in July, and vice versa for N. fasciatus and L. taschenbergi, which in in accordance with previous studies [5,49,52]. At a global scale, N. fasciatus has no marked seasonality in Atlantic conditions [5]. However, N. fasciatus eggs require at least 23ºC during four days or a peak of 30ºC for three hours to hatch [56,57]. The extreme cold conditions during the autumn-winter months in the study region (including November and March) may lead this species to shorten its reproductive period or modify its activity pattern, recovering in spring as the weather conditions become more suitable.
Seasonality in ea patterns is common, as well as variations owing to local climatic conditions and hosts traits [18,[22][23][24]. Assuming the same phenology in individuals belonging to a certain species living under the same climatic conditions, differences in infestation parameters can be explained by host characteristics. The patterns obtained in this study also seemed to be in uenced by particular hosts traits. Prevalence models followed the same general rule according to ea phenology, but many seasonal peculiarities arose considering variations in intensity regarding one host that did not show the others. Overall maximum infestation rate occurred during summer, when small mammal species are more active and reproduce [36], potentially increasing the exposure of hosts through contacts with other infested animals [19]. However, we found that maximum infestation rate and maximum intensity occurred simultaneously during this favourable season in mice, but not in voles. The low ea abundance in voles during this favourable season for transmission, despite a high percentage of infected hosts, could be explained by a diluting effect owing to the increase in the population density on voles compared to colder months (see [13] for more details about host density dynamics). Similarly, intensity of N. fasciatus also differed between voles and mice during July. Since the intensity and prevalence raised signi cantly in voles during the summer, the reproductive period of this ea [52], infestation burden remained low in mice until March. If burrows and mice themselves were as suitable as voles, no differences in intensity should exist between both. Nevertheless, intensity remained low in mice during the most suitable season for N. fasciatus reproduction. A possible reason could be that wood mouse nests or their own body are less suitable for this ea compared with voles. This type of differences at host species level has been described in other small mammal communities [58]. Therefore, the maximum intensity of N. fasciatus in mice is reached at minimum mice population (after the winter mortality and before the spring reproductive recruitment) [13], which may facilitate the ea aggregation in the surviving animals. Furthermore, A. sylvaticus have a more individualistic behaviour and spent more time inside the burrows during the cold months [34], which could reduce the probability of ea transmission to other hosts, preventing a decrease in ea intensity [19].
A similar pattern may be occurring for C. a. gilcolladoi on wood mouse in March, when prevalence patterns differed from voles. A noticeable seasonality was found in the later, following the typical ea phenology with peaks during temperate months [5,52]. We found lower infestation values and no differences in prevalence throughout the year in mice. These results suggest that C. a. gilcolladoi may prefer infesting common voles rather than mice in this ecosystem, although no investigation has been yet done to determine the relationship between C. a. gilcolladoi and their hosts in this system. This possible preference was not expected since M. arvalis is a host species than colonized the study region during the 1970s, invading from mountain habitats characterized by Atlantic weather conditions [10]. Ctenophthalmus apertus is endemic to Mediterranean open habitats, and absent from Atlantic climatic areas [51] and thus, does not occur within the original distribution range of M. arvalis. A host shift could have possibly occurred, if this new host species offered better conditions than other hosts [58]. The colonization by a new species, the common vole, may therefore have altered the host-parasite system. If either of them acts as a reservoir or vector of any pathogen, the consequences and implications of this alteration could be unexpected [59,60]. Future investigation could elucidate the consequences of such shifts in terms of pathogen transmission risk.
We also found differences in ea parameters depending on the habitats used by hosts. Common voles inhabiting alfalfas had low ea prevalence and abundance than those from other crop types. L. taschenbergi infestation in voles also appeared associated with cereal habitats. Wood mouse inhabiting cereals had heavier ea burdens compared with other habitats. Some authors have linked animal condition and abundance with the quality and quantity of food supply [61]. The most favourable crop type for voles are alfalfas, where they reach highest densities and better body condition [61]. Wood mouse has a strong preference for woody habitats [62], avoiding habitat without shrub [63] and with insu cient cover to protect them from predators [64] such as cereals (especially recently sown crops in November and stubbles in July). Animals inhabiting sub-optimal habitats may have greater ea infestation because they are in worse condition, favouring the egg production and survival of ea larvae [65].
Lastly, we found sexual differences in ea infestation in both mice species, with greater prevalence and severity in male than in female hosts. Such patterns were not found in M. arvalis, except with the less frequent ea species, L. taschenbergi. This male-biased difference is usually linked to sexual size dimorphism, immunosuppression by sexual hormones or behavioural differences that facilitates ea encounter and horizontal transmission [66]. Such patterns have been already reported in several mice species [22,67,68]. However, the absence of sex-bias in our voles suggests that differences at host scale may cause these dissimilarities. This may be due to the colonial lifestyle [69] and the aggressive behaviour of female voles [70] that could increase the horizontal transmission among females, balancing the ea burden between both sexes. Male voles are more mobile than females [69] and also more active [71] and therefore more prone to ea encounters [5], potentially explaining why the less frequent ea species (L. taschenbergi) occurred more often in male than in female voles.

Co-infections and implications
In the Mediterranean agricultural landscapes studied here, two dominant ea species (N. fasciatus and C. a. gilcolladoi) were shared by 40% of the hosts of the small mammal guild. Flea co-occurrence is common in small mammals [72], and may be due to apparent facilitation via suppression of the host immune system [73]. It is known that different eas can be infected by the same pathogen [30], while several pathogens can be harboured by a shared host [7]. Since eas are important vectors for many pathogens [1], coinfections could have important consequences in terms of disease dispersal and zoonotic transmission risk [74,75], increasing the circulation of pathogens through a shared host population. Important not forgetting generalist and abundant eas that could infest closely humanrelated animals and, eventually, bite humans. The generalist ea N. fasciatus was found in the most abundant co-infections of the four small mammals, having an apparently high tolerance to cohabit with other ea species. In fact, we identi ed this ea and C. a. gilcolladoi (the two more abundant eas) in 82.7% of all the co-infections analysed. Noteworthy, previous work [30] detected a 65% Bartonella spp. prevalence in N. fasciatus collected from our focal common vole population, and 33% in C. a. gilcolladoi. Additional investigation should be carried out to know the potential of these two eas in the transmission of bartonelloses. Moreover, N. fasciatus is a typical parasite of Rattus spp. [1], and both eas are known to parasite Mus domesticus in Spain [76]. Rattus and Mus spp commensal rodents are widely spread in rural areas, especially linked to the presence of domestic livestock, and they can be infected by Bartonella sp. [77,78]. Density can in uence the ea-host system [20,21], and uctuations in wild rodent densities can facilitate the encountering with commensal rodents, eas and humans, increasing the possibility of disease transmission in the rural population inhabiting this agricultural area. The other ea species (other Ctenophthalmus spp. and R. beillardae) were more rarely found. Therefore, their possible role in the circulation cycle of zoonoses among this small mammal host guild would probably be less relevant.

Conclusion
Six different ea species parasitized the studied small mammal guild inhabiting continental Mediterranean farmland, although C. a. gilcolladoi and the generalist N. fasciatus were found to dominate the ea community. Co-infections with both ea species, which often harbour zoonotic pathogens, frequently occurred within the focal host guild. The role that ea species might play in zoonotic transmission should be elucidated, considering also seasonal patterns and sex-biased differences. Moreover, the most abundant host (i.e. M. arvalis) is a recent colonizer and, unlike other small mammal hosts, is characterized by large uctuations in abundance. The consequences of uctuating M. arvalis abundances for the transmission cycles of ea-vectored diseases should also be investigated.

Declarations
Ethics approval and consent to participate The trapping methods applied in this study were approved by our institution ethics committee (CEEBA, Universidad de Valladolid; authorization code: 4801646) and we counted with the o cial trapping permits from DGMN (Junta de Castilla-y-León), as well as compulsory national certi cates (B and C categories) to manipulate living animals for research.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the principal investigator JJ Luque-Larena (j.luque@agro.uva.es) on reasonable request.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions JJLL and FM obtained all the funding and designed the monitoring. JJLL, RRP and MFF collected the data. SHC, MFF and FM performed the statistical analysis. SHC and MFF drafted the manuscript. FM, JJLL and RRP critically revised the paper. All authors read and approved the nal manuscript.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.