Soil samples were collected from six different farms of Portuguese DWR in September and October 2012, i.e. Arnozelo, Aciprestes, Carvalhas, Cidrô, Granja and S. Luiz. Details of geographic coordinates and altitudes of these farms are given in Fig. 1A. The sampling strategy was adapted from Klingen et al. (2002) and Goble (2010) and presented in Fig. 1B and the authors find it quite similar to that undertaken by Clifton et al. (2015). In brief, at each site, the surface litter was removed and the soil was dug to a depth of 20 cm with a soil core borer (width = 20 mm) at five places within 0.25 m² area. All five sub-samples from one site were put in the same polyethylene bag and sealed with a rubber band. This mix of five subsamples was considered as one soil sample from a site. The next sampling site was at 20 m away and the soil borer was washed with 5% sodium hypochlorite (NaOCl) between the sites. In total, 183 soil samples were collected, out of which 155 were from vineyards and 28 were from adjacent hedgerows. Hedgerows were mainly composed of oaks (Quercus spp. L., Fagaceae) and pine (Pinus spp. L., Pinaceae) trees. Soil samples were brought inside the laboratory and were spread on a tray and left overnight for the moisture to be equilibrated with the room temperature. This was done to avoid infestation with entomopathogenic nematodes (EPN), if any, as suggested by Quesada-Moraga et al. (2007). Soil samples were always processed within 24 hours of spreading on to the trays. The number of soil samples collected from each farm is provided in Table 1.

Figure 1. 

Geographic coordinates and altitudes of the farms and details of the soil sampling strategy adopted. a Details of the six farms of the Douro Wine Region, Portugal, which were considered in this study b Details of the soil sampling strategy from vineyards and adjacent hedgerows.

Two hundred and fifty grams (g) of sieved soil was put in a plastic bowl with small holes on the cap for ventilation. A total of 183 soil samples were used to compare the effect of habitat-type on fungal isolations. For each soil sampling site, four such bowls, i.e. 1 kg of the soil was analysed in total and four late instar T. molitor larvae were put in each of these bowls, i.e. the total number of larvae used (n) = 16. To study the effect of insect baiting, 81 of the total 183 soil samples were baited with late instar larvae of G. mellonella (n = 8) and T. molitor (n = 8) similarly, such that the total number of larvae, irrespective of the bait-insect type, remained same, i.e. n = 16. These 81 soil samples were from the three farms with a relatively diverse landscape, i.e. S. Luis, Carvalhas and Granja, as reported by Carlos et al. (2013). Hence, these farms were chosen to enhance the fungal diversity, in theory. This would facilitate studying the effect of insect baiting on a rather diverse group of EPF. Galleria mellonella was given heat shock by immersing in 56 °C water prior to baiting, to reduce the tendency of silk web formation within soil samples as suggested by Meyling and Eilenberg (2006). Bowls were kept in an environmental chamber (Panasonic MLR-352H-PE) at a temperature of 22 °C and relative humidity of 85%, in the dark. Bowls were frequently inverted, shaken gently and kept upside down for the total incubation period of three weeks as per Meyling and Eilenberg (2006).

The presence of insect cadavers was observed every day for the first week and every second day for the remaining two weeks. Everyday monitoring was necessary for the first week as death by EPN, if any, generally was caused within the first three days of larvae incubation in soils, although slightly delayed infection cannot be neglected. The schedules were monitored rigorously and the insect cadavers were observed quite carefully. Any cadavers with a foul smell were constantly discarded. Obtained cadavers were washed with 1% NaOCl for three minutes, followed by three distinct washes of 100 ml sterilised distilled water for three minutes each. It was done to isolate only the fungi which have penetrated the insect cuticles and proliferated within the insect haemocoel or have been ingested into the haemocoel. The cadavers were subsequently cultured on to potato dextrose agar (PDA) (Liofilchem) plates supplemented with 0.1 g/l streptomycin (Acros) and 0.05 g/l tetracycline (Acros). In cases of mixed infections or inhibited fungal growth, cadavers were cultured on to oatmeal agar (OA) supplemented with 0.5 g/l chloramphenicol (Acros) and 0.6 g/l cetyl trimethyl ammonium bromide (CTAB) (Sigma) as described in Posadas et al. (2012). Repeated culturing on OA or/and Sabouraud dextrose agar (SDA) (Prolabo) was undertaken until the pure culture of fungus was obtained. Plates were repeatedly observed through a low magnifying stereomicroscope (Olympus SZX9, 40X magnification) and, if any emergence of nematodes were observed, they were discarded no matter if a fungal growth was present or absent. Any possibility of cross-contamination or external contamination was carefully monitored as described by Steinwender et al. (2014). No colony forming units (CFUs) were observed in any of the tests for contaminations. To confirm Koch’s postulates, all the obtained fungi were tested using bioassays for pathogenicity against the larvae from which they were isolated. The method was initially described by Ali-Shtayeh et al. (2003), however, a modified protocol was used as described in Sun and Liu (2008) and Goble et al. (2010). The fungi found pathogenic to insect larvae were considered further in this study.

The appearance on the infected larvae and morphological characteristics were used as the preliminary identification of fungi. Morphological characteristics that were used for identification are described in a taxonomic key (Domsch et al. 2007). For molecular identification, DNA was extracted from fungal mycelium as described earlier by Möller et al. (1992). Moreover, the protocol was optimised for hard-to-crush mycelium and spores as in Sharma et al (2018). The fungal internal transcribed spacer (ITS) region was amplified using the forward primer ITS1-F (5’-CTTGGTCATTTAGAGGAAGTAA-3’) and reverse primer ITS4 (5’-TCCTCCGCTTATTGATATGC-3’) (Gardes and Bruns 1993). The PCR reaction was performed as described in Yurkov et al. (2015). Primers used for PCR reactions were also used for amplicon sequencing. Sequences were edited using BioEdit 7.2.1 (Hall 1999) and further aligned using MAFFT version 7 (Katoh and Standley 2013) to validate polymorphisms amongst sequences. Obtained ITS sequences from EPF were aligned with those from the respective type strain sequences using BLASTn and the identity results are shown in Suppl. material 1: Table S4. Newly generated sequences were submitted to EMBL nucleotide sequence database and the accession numbers are provided in the Suppl. material 1: Table S4.

Fungal species richness (S) was compared in terms of habitat-types and bait-insects used for isolation. Jaccard’s similarity coefficients (J) for fungal species shared between different habitats and bait-insects were measured as described in Garrido-Jurado et al. (2015). J = a/(a+b+c), where “a” represents the number of species occurring in both variables, “b” represents the number of species occurring only in variable 1 and “c” represents the number of species occurring only in variable 2. J can range between 0 (no shared species) to 1 (all shared species). Software IBM SPSS Statistics 22 was used to perform statistical data processing. Infections were counted qualitatively per site, i.e. whether a particular fungus infected one or several insect larvae of the same bait-insect, it was registered as one infection for that fungal species, as described in Klingen et al. (2002) and Goble et al. (2010). Therefore, effects of soil habitat-types and bait-insects are counted in accordance with the number of soil samples found harbouring a fungal species as in Klingen et al. (2002), Goble et al. (2010) and Clifton et al. (2015). Data were treated using Fisher’s exact test as it gives the exact P value for a 2×2 contingency table (https://www.graphpad.com/). Besides, farm type variations could only be analysed using the χ² (chi-square) test and Monte Carlo simulations were used in case the cells have the expected count of less than 5. Data used for different analyses, i.e. (1) effect of bait-insect type on the occurrence of EPF; (2) effect of habitat-type (hedgerows vs. vineyards) on EPF occurrence; and (3) effect of farm type on EPF occurrence, are provided in detail within the Suppl. material 1: Tables S1, S2 and S3, respectively. To compare possible factors which may influence fungal recoveries, a principal component analysis (PCA) was performed. The PCA was conducted on the mean-centred and scaled data in order to investigate the discriminations of the obtained fungal species. For the PCA plots, only those soils samples were considered where both the bait-insects, i.e. T. molitor and G. mellonella were used, i.e. soils from the farms S. Luis, Carvalhas and Granja (Suppl. material 1: Table S1). Fungi with isolation frequencies of <10% from either vineyards or hedgerows were considered as rare EPF. Hierarchical clustering was then employed to investigate the degree of similarities of fungal isolations based on their ecological proximities, i.e. in terms of habitat-type and bait-insect type. The resulting dendrogram was obtained based on the Euclidean distance and Ward aggregation method as in Sharma et al. (2018). Software R 3.4.2 was used to generate PCA plots and hierarchical clustering.

The total numbers of soil samples used were 183 and the number of soil samples found positive (N) with any EPF were 81, i.e. 44.26% ± 3.67% soils. A total of 12 different species were observed (Table 1). Clonostachys rosea f. rosea (Link) Schroers, Samuels, Seifert & Gams was found in the maximum number of soil samples i.e. 30.05% ± 3.38% (N = 55), followed by B. bassiana (12.57% ± 2.37% (N = 23)), Purpureocillium lilacinum (Thom) Luangsa-ard, Houbraken, Hywel-Jones & Samson (9.29% ± 2.14% (N = 17)) and Metarhizium robertsii Bischoff, Rehner & Humber (6.01% ± 1.75% (N = 11)).

Isolations of Beauveria pseudobassiana Rehner & Humber (3.38% ± 1.31% (N = 6)), Cordyceps sp. Fries (1.64% ± 0.94% (N = 3)), Lecanicillium dimorphum (Chen) Zare & Gams (1.64% ± 0.94% (N = 3)), Cordyceps cicadae (Miq.) Massee (1.10% ± 0.77% (N = 2)), Lecanicillium aphanocladii Zare & Gams (1.10% ± 0.77% (N = 2)), Metarhizium guizhouense Chen & Guo (1.10% ± 0.77% (N = 2)), Beauveria varroae Rehner & Humber (0.55% ± 0.54% (N = 1)) and Purpureocillium lavendulum Perdomo, García, Gené, Cano & Guarro (0.55% ± 0.54% (N = 1)) were also observed (Table 1). The fungal occurrence was the highest in the farm Granja, i.e. 61.54% ± 9.54% (N = 16), followed by Carvalhas (59.09% ± 7.4% (N = 26)), Arnozelo (45% ± 11.12% (N = 9)), S. Luiz (37.25% ± 6.77% (N = 19)), Aciprestes (30% ± 10.24% (N = 6)) and Cidrô (22.73% ± 8.93% (N = 6)) (Table 1).

To test the effect of insect baiting on EPF recoveries, bait-insects G. mellonella (n = 8) and T. molitor (n = 8) were employed on 81 soil samples from the three farms which had quite diverse landscapes, i.e. S Luiz, Carvalhas and Granja. Hence, in total, 16 larvae from two different bait-insects were used. Eleven EPF species were observed amongst the three farms and a few significant differences were detected within fungal recoveries (Fig. 2A, Suppl. material 1: Table S1). Significantly more soil samples were found positive for B. bassiana when G. mellonella was used as a bait-insect, i.e. 15 isolates (18.52% ± 4.31%) than T. molitor, i.e. 4 isolates (4.94% ± 2.4%) (P = 0.038). On the contrary, isolation of M. robertsii was increased significantly by T. molitor, i.e. 10 isolates (12.35% ± 3.65%) compared to G. mellonella, i.e. 2 isolates (2.47% ± 1.72%) (P = 0.003).

Clonostachys rosea f. rosea was isolated more often by T. molitor, i.e. (14.81% ± 3.94% (N = 12)) than by G. mellonella, i.e. (11.11% ± 3.49% (N = 9)). Moreover, T. molitor specific isolations were noticed for M. guizhouense, i.e. 2.47% ± 1.72% (N = 2). However, G. mellonella recovered more C. cicadae and L. dimorphum, i.e. 2.47% ± 1.72% (N = 2) than 1.23% ± 1.22% (N = 1) by T. molitor, in cases of both the fungi. Galleria mellonella specific isolations for Cordyceps sp. (3.79% ± 2.09% (N = 3)), L. dimorphum (2.47% ± 1.72% (N = 2)) and P. lavendulum (1.23% ± 1.22% (N = 1)) were also recorded (Fig. 2A, Suppl. material 1: Table S1). Overall, using G. mellonella yielded slightly more fungal species (i.e. S = 10) than T. molitor (i.e. S = 7) (Table 2).

Figure 2. 

Effect of insect baiting and habitat-type on the isolation of the entomopathogenic fungi. a Occurrence (% of soil samples ± SE) of entomopathogenic fungi when different bait-insects were incorporated b Occurrence (% of soil samples ± SE) of entomopathogenic fungi when soils were collected from different habitat-types. Bars with asterisk (*) show significant isolations, i.e. (P<0.05).

To study the habitat type variation, 183 soil samples from all the six farms were considered, i.e. 155 from vineyards and 28 from hedgerows. As two different bait-insects, G. mellonella and T. molitor, were used in the three farms, i.e. S. Luiz, Carvalhas and Granja and only one bait-insect T. molitor was used in the other farms, i.e. Aciprestes, Arnozelo and Cidrô, the numbers of bait-insects larvae used to study the habitat-type variations in each farm were kept constant, i.e. n = 16.

Out of 155 soil samples from vineyards, a total of nine EPF species were observed in 61 vineyards’ soils, i.e. 39.35% ± 3.81% soils were found harbouring at least one EPF. Six fungal species were observed solely from vineyards, i.e. Cordyceps sp. (1.94% ± 1.1% (N = 3)), L. aphanocladii (1.29% ± 0.9% (N = 2)), L. dimorphum (1.94% ± 1.1% (N = 3)), M. robertsii (7.10% ± 2.06% (N = 11)), M. guizhouense (1.29% ± 0.9% (N = 2)) and P. lavendulum (0.65% ± 0.64% (N = 1)). Although M. robertsii was isolated only from vineyards, however, recoveries were not significant (P = 0.220). Three species, i.e. P. lilacinum, C. rosea f. rosea and B. bassiana were shared amongst both habitat-types. Purpureocillium lilacinum was isolated more frequently from vineyard soils i.e. 16 isolates (10.32% ± 2.44%) than hedgerows, i.e. 1 isolate (3.57% ± 3.50%), however, non-significantly (P = 0.228) (Fig. 2B, Table 1).

Beauveria bassiana was slightly more abundant in hedgerows, i.e. 4 isolates in 28 samples (14.29% ± 6.61%) than in vineyards, i.e. 19 isolates in 155 samples (12.26% ± 2.63%), although differences were not significant (P = 0.759) (Table 1), (Fig. 2B). Clonostachys rosea f. rosea was also more frequent in hedgerows, i.e. in 15 of the 28 samples (53.57% ± 9.42%) than in vineyards i.e. 40 of the 155 samples (25.81% ± 3.51%) (P = 0.006). Moreover, B. pseudobassiana only occurred in hedgerows, i.e. 6 isolates (21.43% ± 7.75%) (P<0.001). Beauveria varroae (3.57% ± 3.50% (N = 1)) and C. cicadae (7.14% ± 4.86% (N = 2)) (P = 0.023) were also noticed in hedgerows’ soils only (Fig. 2B). Overall, significantly higher number of soil samples were found positive for EPF in hedgerows, i.e. 20 isolates in 28 samples (71.43% ± 8.53%), than in vineyards, i.e. 61 isolates in 155 samples (39.35% ± 3.92%) (P<0.001) (Table 1). However, fungal species richness (S) was higher in soils from vineyards, i.e. S = 9 than from hedgerows, i.e. S = 6 (Table 2). Additional information on the habitat-types variations is shown in Suppl. material 1: Table S2.

Those EPF which were recovered from all six farms using T. molitor larvae (n = 16) only, were considered to study the farm type variations. This was done to avoid any bias as T. molitor was the bait-insect used in all six farms. Nine EPF species were recovered and C. rosea f. rosea was isolated significantly more from Carvalhas, i.e. from 18 of the total of 48 soil samples collected from the respective farm (N = 18/48), (37.5% ± 6.98%) (χ² = 12.981, df = 5, P = 0.0024). Metarhizium robertsii was isolated more frequently from Granja (N = 8/11) (72.72% ± 13.4%) (χ² = 33.657, df = 5, P<0.001). Beauveria bassiana was found distributed throughout all farms, i.e. Aciprestes (N = 3/20) (15% ± 7.98%); Arnozelo (N = 2/20) (10% ± 6.7%), S. Luiz (N = 3/51) (5.88% ± 3.29%), Carvalhas (N = 2/44) (4.55% ± 3.14%), Cidrô (N = 1/22) (4.55% ± 4.44%) and Granja (N = 1/26) (3.85% ± 3.77%). Purpureocillium lilacinum was found in four of the six farms, i.e. Arnozelo (N = 2/20) (10% ± 6.7%), Carvalhas (N = 2/44) (4.55% ± 3.14%), S. Luiz (N = 2/51) (3.92% ± 2.71%) and Granja (N = 1/26) (3.85% ± 3.77%). More details about other fungi are in the supplementary information (Suppl. material 1: Table S3).

A PCA was performed on the EPF recovery data from the 81 soils of the three farms, i.e. S. Luis, Carvalhas and Granja, where both habitat-types and bait-insects were incorporated. This kind of analysis was done to understand which element(s), i.e. bait-insect(s) and/or habitat-type(s), governs the recovery of the EPF. Using PCA, 89.9% of the variance among fungal recoveries could be described by the three principal components, i.e. PC1 (55%), PC2 (21.7%) and PC3 (13.2%) (Fig. 3A, B, C). Second principal component (PC2) was slightly dominated by the type of bait-insect used (Fig. 3A, C). The occurrences of B. bassiana and P. lilacinum were slightly and marginally governed by insect baiting using G. mellonella, respectively. However, the isolations of C. rosea f. rosea and M. robertsii were slightly and mainly governed by baiting using T. molitor, respectively (Fig. 3A–D). Third principal component (PC3) could distinctly separate the two habitat-types (Fig. 3B, C). The isolations of C. rosea f. rosea were mostly governed by semi-natural habitats. However, M. robertsii and P. lilacinum were highly and slightly influenced also by cultivated habitats, respectively. Codyceps cicadae recovery was governed only by hedgerows (Fig. 3A–D). Hierarchical clustering dendrogram of the ecological proximities of fungi, after profiling their recoveries from bait-insects and habitat-types, placed B. bassiana and P. lilacinum closer, while C. rosea f. rosea and M. robertsii were quite different and distinct (Fig. 3E). Moreover, the dendrogram also separated rare EPF, i.e. those with an isolation frequency of <10% from either of the habitat-types (cluster 1), from relatively more frequent EPF (cluster 2) (Fig. 3E).

Figure 3. 

Principal component analysis (PCA) and hierarchical clustering of the observations based on the fungal isolations. aPC1 vs. PC2. bPC1 vs. PC3. cPC2 vs. PC3. d PCA 3D plot e Hierarchical clustering dendrogram to access the ecological proximities of obtained fungi based on their respective isolation profiles. Software R 4.3.2 was used to obtain the PCA plots and the hierarchical clustering. There was no fungal isolation from hedgerows from the farm Granja when bait-insect T. molitor was used and hence, it could not be included in any of the analysis which relies on proportions, i.e. PCA plots, hierarchical clustering. To reduce any bias, the authors also discarded the soil samples (N=1) which yielded the fungal isolations, when G. mellonella was used, from the hedgerows of the farm Granja. The blue balls represent relatively more frequent EPF, i.e. Beauveria bassiana, Beauveria pseudobassiana, Clonostachys rosea f. rosea, Cordyceps cicadae, Purpureocillium lilacinum and Metarhizium robertsii. The red balls represent other fungi such as Cordyceps sp., Lecanicillium aphanocladii, Lecanicillium dimorphum, Metarhizium guizhouense and Purpureocillium lavendulum. Hierarchical clustering based dendrogram classified isolated EPF into two clusters, i.e. rarely occurring EPF (cluster 1) and relatively more frequent EPF (cluster 2). Abbreviations used are: Beauveria bassiana (B.b), Beauveria pseudobassiana (B.p), Cordyceps cicadae (C.c), Cordyceps sp. (C.sp), Lecanicillium aphanocladii (L.a), Lecanicillium dimorphum (L.d), Metarhizium guizhouense (M.g), Purpureocillium lavendulum (P.la), Purpureocillium lilacinum (P.l), Clonostachys rosea f. rosea (C.rr) and Metarhizium robertsii (M.r).

Considering the number of soil samples and the objectives, this study was comparable with others on EPF occurrence and diversity (Tarasco et al. 1997, Klingen et al. 2002, Ali-Shtayeh et al. 2003, Quesada-Moraga et al. 2007, Sun et al. 2008, Imoulan et al. 2011, Schneider et al. 2012). The ‘Galleria-bait method’, i.e. using G. mellonella for EPF recovery from soils, was described by Zimmermann in the year 1986 (Zimmermann 1986). Since then it has been used quite often in numerous studies as the only method for EPF isolations, in the past three decades (Chandler et al. 1997, Bidochka et al. 1998, Ali-Shtayeh et al. 2003, Meyling and Eilenberg 2006, Quesada-Moraga et al. 2007, Sun and Liu 2008, Sun et al. 2008, Sevim et al. 2009, Fisher et al. 2011, Muñiz-Reyes et al. 2014, Pérez-González et al. 2014, Fernández-Salas et al. 2017, Gan and Wickings 2017, Kirubakaran et al. 2018). Similarly, in few other studies, insect baiting using T. molitor is the only method used for the EPF recovery (Sánchez-Peña et al. 2011, Steinwender et al. 2014).

Beauveria bassiana was isolated significantly more from G. mellonella (P = 0.038) (Fig. 2A) as in South Africa by Goble et al. (2010). Klingen et al. (2002) found insect-specific isolations of B. bassiana by G. mellonella in Norway. Studies in Iceland and Greenland also concluded that B. bassiana was isolated more often by G. mellonella (Oddsdottir et al. 2010, Meyling et al. 2012). Many previous reports are available on the recovery of different fungi from G. mellonella, for example, C. cicadae (Barker and Barker 1998), P. lilacinum (Imoulan et al. 2011), Lecanicillium spp. (Hypocreales: Cordycipitaceae) (Asensio et al. 2003, Meyling and Eilenberg 2006), as in the present study. To our knowledge, this study reports the first isolation of P. lavendulum from an insect.

In the present study, insect-specific isolation of M. guizhouense and significant isolation of M. robertsii was reported from T. molitor (P = 0.003) (Fig. 2A) (Suppl. material 1: Table S1). Comparing G. mellonella and T. molitor, insect-specific isolation of Metarhizium has been reported using the latter (Oddsdottir et al. 2010). Hughes et al. (2004) found that, out of the 20 soils sampled, 15 harboured Metarhizium when T. molitor was used as bait-insect, compared with just four when G. mellonella was used. Metarhizium was found to be the most abundant EPF in the soils from the tropical forests of Panama, although the soils were collected within 5 m from the nest of leaf-cutting ants (insect host) which possibly increased EPF recovery. Nonetheless, the major drawback of the study was that a very limited number of soil samples were used and the results were not analysed statistically (Hughes et al. 2004). In the present study, 81 soil samples were used to study the effect of insect baiting on EPF recovery. Moreover, a random selection of soil samples was promoted to reduce any bias for an enhanced EPF recovery and to maintain a practical scenario where no prior information on the presence of insect-host is necessary.

To our knowledge, this is the first report on the significantly higher recovery of M. robertsii by T. molitor when compared with that from G. mellonella. Galleria-bait is still a widely used method to isolate EPF from soils. Even the most recent reports, i.e. those reported in the past few months, overlook the use of T. molitor while studying with ecologies of EPF such as Metarhizium (Fernández-Salas et al. 2017, Gan and Wickings 2017, Hernández-Domínguez and Guzmán-Franco 2017, Kirubakaran et al. 2018). This study signifies that the use of both of the bait-insects is more important than considered before and T. molitor should always be used along with G. mellonella, especially when Metarhizium is being isolated from soils. Enhanced recovery of Metarhizium from T. molitor could be due to the higher sensitivity of the insect towards this fungus. Vänninen et al. (2000) found that even after three years post application, M. anisopliae could kill over 80% of the T. molitor baited in soils from different places.

In this study, 15.3% of the total soil samples were from hedgerows, which were comparable with 20.5% of the soil samples from hedgerows examined by Meyling and Eilenberg (2006). Beauveria bassiana was slightly more abundant in hedgerows than in vineyards (Table 1), (Fig. 2B). Some previous studies also did not report any significant habitat preference for B. bassiana (Klingen et al. 2002, Quesada-Moraga et al. 2007). Only the soils from hedgerows could lead to the isolation of B. pseudobassiana and it was significant (P<0.001) (Fig. 2B). This finding agreed with Meyling and Eilenberg (2007), who found B. pseudobassiana only in hedgerows. Cordyceps cicadae was also isolated in significant amounts from hedgerows (P = 0.023) (Fig. 2B). Barker and Barker (1998) reported that C. cicadae isolations were restricted to forest soils (i.e. less disturbed soils). To our knowledge, this is the first report on the isolation of C. cicadae from Mediterranean soils. Clonostachys rosea f. rosea was isolated more from less disturbed (i.e. orchard) soils than intensively disturbed (i.e. field crops) soils in this study as in Sun et al. (2008).

A possible reason of higher occurrence of B. bassiana and the habitat-specific occurrence of B. pseudobassiana and B. varroae in hedgerows could be the relatively higher dependence of Beauveria on secondary infections on insect hosts, as hedgerows are expected to host rather diverse insect communities (Goble et al. 2010). Besides, factors such as reduced ultra-violet radiation and temperatures, increased humidity and long-term environmental stability could also lead to an increased viability of these fungal spores (Meyling et al. 2009). Mycoparasitism, a characteristic of B. bassiana (Vega et al. 2009) and C. rosea (Keyser et al. 2016), could provide dominance amongst opportunistic saprophytes in hedgerows.

Although Purpureocillium lilacinum and M. robertsii were isolated more from vineyards’ soils, the results were, however, non-significant, i.e. P = 0.228 and P = 0.220 (Fig. 2B). Moreover, two strains of M. guizhouense were also isolated only from vineyards (Table 1). Purpureocillium lilacinum could tolerate a wide range of temperatures, from 8 °C to 38 °C and pH (Roumpos 2005). As these properties provide robustness against agricultural disturbances, according to Wei et al. (2009), P. lilacinum is the most widely tested fungus under field conditions. Higher isolations of Metarhizium spp. from crop cultivated lands in Spain and Mexico have been reported (Quesada-Moraga et al. 2007, Sánchez-Peña et al. 2011). Tillage seemed to distribute MetarhiziumCFUs evenly throughout the field which subsequently increases chances of fungal recovery from different sites (Kepler et al. 2015).

Fungal species richness (S) was higher in soils from vineyards, i.e. S = 9 than hedgerows, i.e. S = 6 (Table 2). Few genera mentioned in Table 1 were previously reported to be isolated more often from relatively more disturbed soils, for example, Lecanicillium (Meyling and Eilenberg 2006). Moreover, Sun et al. (2008) found higher species richness in soils of crop fields than from orchards soils (i.e. less disturbed soils), as in the present study.

More diverse fungal species in cultivated soils is not surprising. Practices such as ploughing, reseeding and fertilising increase environmental patches and niche availability for EPF and subsequently increase fungal diversity (Sun et al. 2008). The higher organic matter also increases biological activity in the soil which positively affects the presence of saprophytic fungi which lead to lesser organic resources for EPF and therefore, reduced survivability (Goble et al. 2010).

Studies on the EPF ecology in soils consider either different bait-insects or habitat-types or both, as discussed earlier. Principal component analysis was done to understand the most important factor, if any, that governs the recoveries of EPF. It was found that isolations of B. bassiana were slightly governed by baiting with G. mellonella, irrespective of the habitat-type incorporated (Fig. 3A, C, D). However, the isolations of M. robertsii were influenced both by the cultivated habitat-type as well as by baiting with T. molitor (Fig. 3A–D). The ecological proximities of B. bassiana and P. lilacinum could be explained as P. lilacinum was isolated more frequently from vineyard soils than from hedgerows and B. bassiana isolations were almost equal from vineyards to those from hedgerows (Figs 2B, 3D, E). Moreover, the bait-insect G. mellonella favoured P. lilacinum and B. bassiana isolations (Fig. 2A). Distinct profiles of C. rosea f. rosea and M. robertsii suggest their unique ecologies in terms of habitat-type and bait-insect preferences (Fig. 3D, E). The main advantage of fungal profiling by hierarchical clustering based dendrogram is that those EPF which were not isolated in this study can also be investigated for their roles in the biological control of interest pests in agroecosystems, if they exhibit similar ecological profiles (Sharma et al. 2018).

Entomopathogenic fungi was observed in 44.26% ± 3.67% of the soil samples and it was comparable to previous studies in Finland (38.6%) (Vänninen 1996), Palestine (33.6%) (Ali-Shtayeh et al. 2003), Alicante province, Spain (32.8%) (Asensio et al. 2003), South Africa (21.53%) (Goble et al. 2010), UK (17.6%) (Chandler et al. 1997) and southern Italy (14.9%) (Tarasco et al. 1997). More diverse fungal species were found in the present study when compared with the other studies in Mediterranean regions, for example, in Italy (Tarasco et al. 1997), Spain (Asensio et al. 2003, Quesada-Moraga et al. 2007, Garrido-Jurado et al. 2015), Turkey (Sevim et al. 2009) and Morocco (Imoulan et al. 2011). Different studies suggest that Metarhizium spp. are either absent (Ali-Shtayeh et al. 2003, Oliveira et al. 2012) or less prevalent in the Mediterranean region (Tarasco et al. 1997, Asensio et al. 2003, Quesada-Moraga et al. 2007, Garrido-Jurado et al. 2015). Surprisingly, Garrido-Jurado et al. (2015) reported just four isolates of M. robertsii from 270 soil samples in Spain which was quite a small number compared with the 11 isolates from 183 soil samples found in the present study. Occasional isolations of many species were noticed in the present study and, according to our knowledge, this is the first isolation of entomopathogenic strains of B. varroae, L. aphanocladii, L. dimorphum, M. robertsii and M. guizhouense in Portugal.