The suitability of native flowers as pollen sources for Chrysoperla lucasina (Neuroptera: Chrysopidae)

Green lacewings (Neuroptera: Chrysopidae) are key biological control agents found in a broad range of crops. Given the importance of enhancing their presence and conservation, in this study, we aim to identify and to determine the relative importance of the pollen consumed by Chrysoperla lucasina (Lacroix, 1936) from 29 pollen types offered by 51 native plant species sown in an experimental farm in Villarrubia in the south of Spain. For the purposes of this study, C. lucasina specimens were captured in the late spring of 2016 and 2017. The pollen types and other components in the alimentary canal of C. lucasina were microscopically identified using the transparency method, which is a novel technique applied to green lacewings captured in the field. The results show that (i) C. lucasina feeds on over half of the pollen types offered by the sown plant species, with no differences in behaviour by sex or year; (ii) Capsella bursa-pastoris was the most frequently identified pollen type in the alimentary canal; (iii) the majority of pollen types identified correspond to sown native plant species and not to surrounding plant species; and that (iv) most of the adults studied also consumed honeydew. Our feeding study has important implications for the selection of plant mixtures for ground cover restoration and flower vegetation strips in Mediterranean agroecosystems, which complements our previous findings on how C. lucasina use native plant species as host and reproduction sites. The plant species Capsella bursa-pastoris and Biscutella auriculata, which are best suited to provide pollen, host and reproduction sites for C. lucasina in late spring, should consequently be included in the proposed plant mixtures for Mediterranean agroecosystems.


Introduction
Agricultural intensification has increase global agricultural production, while also expanding agricultural land area and the use of external inputs such as pesticides, fertilizers and irrigation [1,2], as observed in Mediterranean vineyards, olive groves and fruit orchards [3][4][5]. However, these agricultural practices have had adverse environmental effects such as loss of natural habitats and biodiversity, leading to a decline in ecosystem services, particularly pest control and set of available pollen resources for chrysopids [56,57]. Previous studies have reported that chrysopids, which consume pollen where and when available, mainly feed on herbaceous plant species belonging to the families Apiaceae, Asteraceae, Brassicaceae, Chenopodiaceae-Amaranthaceae, Fabaceae, Liliaceae, Plantaginaceae, as well as tree and shrub species from the families Ericaceae, Cistaceae, Oleaceae and Pinaceae [38,58]. Hence, knowledge about how insects in general, and more specifically chrysopids, use floral resources, particularly in SNH, is critical for successful habitat management strategies aimed at improving natural biological control [29,38,59,60].
The principal objective of this study is to identify and to determine the relative importance of the pollen consumed by C. lucasina adults by observing the alimentary canal of individuals collected from sown native plant species. Specifically, we were interested in: 1. whether C. lucasina is a generalist or specialist pollen feeder, 2. the principal sown native species exploited, 3. the importance of surrounding natural plants species in relation to the pollen consumed, 4. whether the pollen consumed by both sexes show a similar pattern, 5. and the possibility of identifying other components (fungal spore, honeydew) in the alimentary canal.
This should enable us to advance our knowledge of the feeding behaviour and ecology of C. lucasina and to suggest the most effective plant species to attract and conserve this key predator in Mediterranean agroecosystems.

Study area
The study was carried out on an experimental farm in the village of Villarrubia (37˚49 0 49@N, 4˚54 0 20@W) in the southern Spanish province of Cordoba. The farm is bordered by a commercial orange orchard, an olive orchard and various irrigated crops, as well as, to the south, by the Guadalquivir river, with its riverbank vegetation. In addition, Pinus halepensis, Pinus pinaster and Pinus pinea trees can be found in the Sierra Morena mountain range, five kilometres to the north of the farm [61]. No specific authorizations were required for the field study, which did not involve endangered or protected species.
In the experimental farm, in order to test the way in which chrysopid species exploited the plant resources, a set of 51 plant species were planted (S1 Table) which need to [62]: (i) be native angiosperms, (ii) be self-sowing winter annuals which do not compete for water with orchards during the summer season and (iii) have a bloom period prior to olive blooming in order to attract natural enemies to the agroecosystem. The potential use of these plants as ground cover and for seed production [25], as well as their attractiveness to on chrysopid species, were tested [28] in previous studies. The farm was tilled in late November 2015 and 2016, and the seeds were planted on consecutive days. The plots were irrigated once during germination and several times during plant development when necessary.
We sampled three 3x3 and 1x9 metre sampling areas (a total of 9 m 2 ) in 2016 and 2017, respectively, for every plant sown in each sampling event. In 2016, 40 plant species belonging to 13 families (S1 Fig and S2 Table) were randomly planted in three blocks of 40 plots. Each plant species was sown in three 3x3 m plots (a total of 120 samples). In 2017, although 35 plant species were sown in a single randomized plot configuration with different areas, we sampled only 20 plant species based on the chrysopid abundance results for 2016 sampling (S1 Fig and  S2 Table). The three samples (9 m 2 ) were located equidistantly from the centre of each plot to avoid border effect (60 samples in total). Due to their poor development, five plant species sampled in 2016 (Anarrhinum bellidifolium, Helianthemum ledifolium, Tuberaria guttata, Aegilops geniculata and Aegilops triuncialis) and two species sampled in 2017 (Medicago orbicularis and Medicago polymorpha) were omitted from the analyses.

Insect collection
Insects were collected only on emerged and well-developed blooming plant species (S2 Table). Three samples (9 m 2 ) per plant species were vacuumed twice in May 2016 and 2017 between 9 a.m and 2 p.m for 40 seconds using an InsectaZooka aspirator (BioQuip 1 Products Inc., Rancho Dominguez, CA, US). Samples were stored at -20˚C and cleaned in the laboratory. Chrysopid adults were identified under a stereomicroscope up to species level and were sexed according to the Iberian chrysopid key [63].

Alimentary canal content analysis
The alimentary canal contents of all chrysopid adults collected were analysed by slightly modifying the transparency method described by Bello et al. [55]. Though previously used in studies of aquatic insect feeding and benthic fauna [64][65][66][67], this was the first time that this method was used to analyse chrysopid specimens collected from the field. The adult chrysopids were defrosted at room temperature, washed three times with distilled water and then vortexed to remove external pollen on the insect's surface. After removing the wings, legs and antennae, each individual was placed in a vial, covered with Hertwig´s solution and heated in an oven at 60˚C for a period of 24 hours. Two specimens, which were lost during the transparency manipulation process, were excluded from the pollen identification analysis. Subsequently, the specimens were mounted on a slide for microscopic examination under a DMI600B inverted microscope (Leica, Wetzlar, Germany) at 1000x magnification to identify the pollen. In addition, photomicrographs of the most frequently occurring pollen types were taken using a C-1 confocal microscope system (Nikon, Japan) at 1000x magnification to obtain a 3D picture of the shape of the pollen grain and to observe the autofluorescence of the exine according to the method described by Castro et al. [68]. Before pollen was identified, we divided the alimentary canal into five segments: I-head, II-thorax, III-initial abdomen, IV-medium abdomen and V-final abdomen. This was done to show the percentage of the alimentary canal occupied by the pollen in each segment, which was classified into three different categories: absent (0% grains), medium (under 50%) and high (50-100% occupation of the alimentary canal) at 40x magnification. The presence of pollen in each segment provided a rough approximation of the intensity of consumption, of when the pollen was consumed and of any differences between sexes with regard to the amount of pollen consumed.
Given that other components, such as nectar, honeydew, fungal spores, other fungi and/or arthropod exuviae, in the chrysopids alimentary canal have been previously reported [38,40,69,70], we visually checked their presence under an inverted microscope according to the method described by Lacey et al. [71]. Based on the findings of Villenave et al. [38,72], honeydew consumption was indirectly identified by the presence of fungal spores Cladosporium sp. and Alternaria sp.

Pollen identification
We examined the pollen according to their morphological traits (polar and equatorial axes, shape, apertures and exine ornamentation), up to type (plant species and/or genus assemblages), family, genus and/or species level. We identified the pollen using the method described by Valdés [73] and our reference pollen collection. This collection was created by growing the native plant species sown in the field under controlled greenhouse conditions: 25˚C, 50-60% humidity and a photoperiod of 16:8h (light:dark). The collection was deposited at the Department of Environmental Protection in the Estación Experimental del Zaidín, a Spanish Scientific Research Council (CSIC) centre in Granada.

Statistical analyses
All analyses were carried out using the R software packages agricolae, FactoMineR, Factoextra and vegan [74][75][76][77]. As the data did not follow a normal distribution, the differences in the number of pollen types identified and the percentage of the alimentary canal containing pollen by sex were determined using the Kruskal-Wallis test with Bonferroni adjustment. Multiple correspondence analysis (MCA) was used to summarize and visualize the presence or absence of each pollen type in each adult chrysopid. MCA was applied to the different types of pollen recorded more than four times in the adult chrysopids examined in order to minimize the effect of pollen types occasionally consumed. As the sown native plant species differed slightly each year, we carried out MCA annually. A permutational multivariate analysis of variance (PERMANOVA), with Jaccard distance and 999 permutations, was then performed each year to determine whether pollen type composition differed significantly by sex in each year. Differences between sexes in relation to each fungal spore identified were also determined using the Fisher test in each year.  (Table 1). C. lucasina consumed 17 of the 29 pollen types, with the following families being the most notable among the adults examined (Table 1): types Anthemis arvensis, Calendula arvensis, and Crepis capillaris on Asteraceae; types Borago officinalis and Echium plantagineum on Boraginaceae; types Capsella bursa-pastoris and Biscutella auriculata on Brassicaceae; and types Silene latifolia, Silene vulgaris and Vaccaria hispanica on Caryophyllaceae. Five pollen types from the following surrounding natural plants were also identified: Apiaceae, Ericaceae, Trifolium arvense, Castanea sativa and Pinus pinea.

Results
Type C. bursa-pastoris was the pollen type most frequently identified in the alimentary canal (Table 1 and Fig 2). Despite some difficulty, we managed to differentiate between B. auriculata and C. bursa-pastoris pollen up to plant species according to grain pollen size (S2 Fig). We found that the most abundant pollen in the alimentary canal was from the species C. bursa-pastoris, which was thus identified in 86  These results are particularly noteworthy when we take into account that both plant species share some months during their bloom period (S2 Table); however, a slight difference in the phenology of both species was observed in 2016, with C. bursa-pastoris showing signs of senescence, while B. auriculata was blooming at the time of sampling. This explains why the plant species C. bursa-pastoris was not sown in the experimental farm in 2017.
We recorded similar pollen types in the alimentary canal of C. lucasina in both years studied; only seven pollen types, C. arvensis, C. capillaris, Ericaceae, Mentha aquatica, Rhamnaceae, Salvia verbenaca and Scabiosa atropurpurea, were absent from one of the sampling years, while twelve pollen types were not recorded in either of the sampling years ( Table 1). The consumption of each pollen type by C. lucasina individuals showed different patterns of association, with no significant differences by sex in either year (Fig 3); PERMANOVA 2016 R 2 = 0.038, d.f. = 1, p = 0.15; 2017 R 2 = 0.024, d.f. = 1, p = 0.173. In 2016, the pollen consumed by C. lucasina were grouped in the following three pollen assemblages: E. plantagineum, V. hispanica and S. vulgaris; C. bursa-pastoris and Festuca arundinacea; and P. pinea, A. arvensis and S. latifolia isolates (Fig 3A). Meanwhile, in 2017, there were four groups: C. capillaris, S. vulgaris, Apiaceae, F. arundinacea and V. hispanica; P. pinea, E. plantagineum and A. arvensis; Reseda luteola and S. verbenaca; S. latifolia, C. bursa-pastoris and T. arvensis isolates (Fig 3B).
In addition to pollen, we identified fungal spores in the alimentary canal contents of 98 of the 107 C. lucasina adults examined. Alternaria sp. and Cladosporium sp. were the most  (Fig 4).

Discussion
In this study, the alimentary canal contents of C. lucasina provided great insight into their late spring feeding behaviour. The transparency method also enabled us to show that C. lucasina had mainly fed on the plant species sown and to investigate their previous visits to plants.

PLOS ONE
The suitability of native flowers as pollen sources for Chrysoperla lucasina (Neuroptera: Chrysopidae)

PLOS ONE
The suitability of native flowers as pollen sources for Chrysoperla lucasina (Neuroptera: Chrysopidae) We show that 17 of the 29 pollen types on the 51 native plants sown were consumed by C. lucasina, which have a rich polyphagous diet without any differences observed by sex. Furthermore, multiple correspondence analyses found no stable pattern of association in either year among the pollen types consumed, thus confirming that C. lucasina is not a specialist pollen feeder. We also observed that the principal pollen types in the alimentary canal of C. lucasina were type C. bursa-pastoris (including B. auriculata and C. bursa-pastoris plant species) belonging to the Brassicaceae family, as well as type A. arvensis from the Asteraceae family which, according to previous studies, are important sources of pollen for C. carnea s.l. especially in spring [38,58]. Most of the pollen types identified have been reported in previous studies of the feeding patterns of chrysopids, whose feeding habits were shown to be opportunistic and eclectic, although, in some cases, their diet was found to specialize in particular pollen types [29,36,38,57,58,78]. These feeding differences in chrysopids could be explained by the variety of plant species in certain habitats and their availability for chrysopids, but might also be due to different chrysopid species, with probably slightly different feeding behaviours, present in the C. carnea complex [29, 57,58]. It might be necessary to carry out further research into the 12 pollen types in our study that did not appear in the chrysopid alimentary canal in order to determine feeding preferences when type C. bursa-pastoris is unavailable in the landscape.
C. lucasina emerge in spring and is one of the most frequently identified and abundant chrysopid species during the blooming stage of herbaceous vegetation [36,79]. Green lacewing populations have different patterns of activity and levels abundance depending on the species and stage of development and habitat, as observed in vineyards, with fewer captures observed during the month of May [17]. Thus, the provision of sown plants which bloom when chrysopid populations decreasing could contribute to increasing predator/prey ratios at key moments in order to improve biological control.
In a previous study, we selected 42 native plant species to evaluate their attractiveness to chrysopids based on chrysopid captures from sown plant species and on their use as

PLOS ONE
The suitability of native flowers as pollen sources for Chrysoperla lucasina (Neuroptera: Chrysopidae) reproduction sites [28]. By linking these previously published findings to the current results, we observed that one of the most attractive plant species for chrysopids is B. auriculata, whose pollen is the second more abundant in the alimentary canal of C. lucasina. Immature and adult stage C. lucasina were found on B. auriculata, which was visited for feeding, reproductive and refuge purposes. On the one hand, the lack of captured adult or immature stage chrysopids in 2016 from C. bursa-pastoris can be explained by signs of plant senescence at the moment of sampling. Although C. bursa-pastoris was not on the sown plant list for 2017 [28], its pollen was consumed by 81% of the chrysopid adults examined as compared to 40% from B. auriculata in that year. We therefore concluded that C. bursa-pastoris selectively attracts C. lucasina. We also observed that the plant species such as Nigella damascena, Moricandia moricandioides and Silene arvensis [28], most visited by chrysopids are not used as pollen sources by C. lucasina. However, other plant species, such as B. officinalis, E. plantagineum and P. rhoeas were abundantly consumed in certain chrysopid specimens despite the presence of bees and bumblebees in these plant species at the time of sampling. The present study demonstrates that, given the feeding preferences of C. lucasina, the plant species C. bursa-pastoris and B. auriculata should be included in conservation biological control programs.
Our findings could also facilitate the selection of suitable plant species in the context of a probable shift from woody pollen sources in spring to herbaceous pollen sources in summer according to the study by Bertrand et al. of C. carnea [29] given that the blooming period of the sown plant species in our study coincides with C. lucasina activity in Mediterranean agroecosystems [80]. Villa et al. [58] found that C. carnea s.l. mainly feeds on Olea europaea (woody strata) in all seasons and to a lesser extent on herbaceous plants in spring and summer. As suggested by Villenave et al. [38], this highlights the importance of providing plant species with different blooming periods that correspond to seasonal chrysopid flight activity and, as recommended by Bertrand et al.
[29] that of increasing plant and vegetation diversity to attract chrysopid populations to agroecosystems. Despite the availability of several flowering plant species over time, Villenave et al. [38] have shown that C. lucasina is attracted to large patches of flowering plant species. In our study, although the pollen type C. bursa-pastoris was consumed, the plant species was sown in small 9 m 2 patches and even in surrounding non-sown plants.
We also found that the alimentary canal of C. lucasina females contained significantly more pollen than that of males. The number of pollen grains observed by Villenave et al. [36] in the diverticulum, an alimentary canal structure of adult chrysopids where the pollen is stored before being digested, is similar to that reported in our study. In our view, the differences between sexes could be related to the higher feeding requirements of female green lacewings as compared to those of males. A previous study found that protein titre levels significantly affect the reproductive physiology and fecundity of C. carnea [81]. Females reach protein titre levels during egg laying on roughly the fourth or fifth day following emergence, while males reach this point at the end of the second day [81]. Thus, the proteinaceous food provided by flowers is essential for the reproduction of adult lacewings [40,[43][44][45][46][47]. However, the floral architecture and nectaries need to be accessible to ensure that these food resources can be reached [49]. Furthermore, as found for bees by Liolios et al. [82], the nutritional requirements of chrysopids are probably satisfied by a small number of abundant plants with accessible pollen present in the landscape and blooming for long periods of time in a manner unrelated to pollen protein content. This could be related to the attraction of green lacewings to large patches of flowering plants [38], the reported consumption by C. carnea s.l. of mostly O. europaea pollen in olive agroecosystems [58] or our results regarding its feeding on C. bursa-pastoris and B. auriculata. Nevertheless, Liolios et al. [82] have observed that the protein content of O. europaea pollen is lower than that of other common plant species surrounding olive orchards, such as Reseda sp., Papaver rhoeas, Trifolium sp. and S. verbenaca, which were also sown for our study.
Given the high capacity of chrysopid movements [83][84][85], we also identified mixtures of entomophilous and/or anemophilous pollen types from surrounding trees and herbaceous vegetation such as Ericaceae, C. sativa, P. pinea, Foeniculum vulgare and Rhamnaceae which were also reported as chrysopid resources in previous studies [36,58]. The presence of these pollen types was low in the C. lucasina alimentary canal as compared to the pollen types offered by the sown plant species. Anemophilous plant species, such as type P. pinea, whose pollen may have been consumed on vegetation surfaces, were at the blooming stage in our study area and at the time of sampling [86]. Furthermore, as C. lucasina had already been known to use P. halepensis for reproductive purposes [80], its use of this tree stratum for feeding and refuge cannot be ruled out.
With respect to honeydew, Villenave et al. [38,72] managed to indirectly identify its consumption through the presence of fungal spores, arthropod exuviae and pollen from anemophilous plant species in the alimentary canal of chrysopids. We found that most chrysopid adults have fungal spores in their alimentary canal, while previous studies have reported that Alternaria sp. and Cladosporium sp. fungal spores reaching peak abundance during the months of May and June in our study area [87,88]. We therefore concluded that a large proportion of C. lucasina adults feed on honeydew in late spring given the presence of fungal spores and anemophilous pollen in the alimentary canal of chrysopids.
Finally, in our view, although flower visitation studies are an important aid to selecting suitable plant species as refuges and reproduction sites, they need to be supplemented by flower foraging analyses. Only then can a full picture be obtained of how chrysopids exploited plant species and of how to guarantee their presence in the required habitat and time.

Conclusions
This study shows that the C. lucasina captured feed mainly on pollen from sown native species (>77%), with no differences observed by sex or year, and that C. bursa-pastoris was the most abundant pollen type consumed by C. lucasina. The identification of anemophilous pollen and fungal spores in its alimentary canal, confirms that C. lucasina feeds on honeydew on vegetation surfaces. Our recommendation at the conclusion of this study is to include C. bursa-pastoris and B. auriculata in the list of the most attractive plant species for C. lucasina. These native plant species are associated with the most frequently consumed pollen types identified in its alimentary canal, are able to attract and maintain C. lucasina populations and also promote a more heterogeneous landscape in Mediterranean agroecosystems. Finally, analysis of floral visits by chrysopids should be complemented with feeding studies in order to provide a more complete picture of plant resources use by chrysopids.  Biscutella auriculata (A and B) and Capsella bursa-pastoris (C and D) flowers; photomicrographs of B. auriculata (E) and C. bursa-pastoris (F) pollen grains taken by a confocal microscope at 1000x magnification. Both images (E and F) show exine autofluorescence after merging multiple optical sections; microscopic images of B. auriculata (G) and C. bursa-pastoris (H) pollen grains at 1000x magnification. (PDF) S1