**1. Introduction**

Microbial pesticide, and especially mycoinsecticides, products based on living fungi to control arthropod pests, were given valuable research efforts in the past decades [1–5]. *Metarhizium* strains are soil-dwelling organisms detected extensively all over the world, regardless of climatic and soil limitations [6,7]. Members of the genus are facultative saprophytes and may either live freely within the topsoil or in the presence of a suitable arthropod host act as parasites [3,8]. The number and scope

of research on *Metarhizium* species sugges<sup>t</sup> that strains and isolates of *M. anisopliae* have been given the highest scientific attention within the genus, and also, they are the most widely used organisms in microbial pest control [9,10]. The first scientific recognition of *Metarhizium anisopliae* dates to Russia in 1879, when E. Metchniko ff discovered a fungus that not only covered the cadaver of a chafer, but was evidently the cause of death of the arthropod [6]. It was then named *Entomophthora anisopliae*, referring to the chafer, *Anisoplia austriaca*. Later, N.V. Sorokin repositioned this species to the genus *Metarhizium* [9,11]. When the species finds an arthropod to parasite, its conidial growth is predominantly green, giving the reason for the original name of "green muscardine" to the condition induced by the fungus [7,12].

Strains and isolates of *M. anisopliae* have long been recognized as entomopathogens, with a wide range of targeted (host) arthropods including mites, ticks, and members of the following insect orders: Diptera, Coleoptera, Hemiptera, Lepidoptera, Isoptera, Orthoptera, Thysanoptera, Homoptera, Sternorrhyncha, Heteroptera. Ongoing research of the past two decades, however, has shown that the position and e ffect of *M. anisopliae* is more complex. The fungus was found to colonize plants within rhizosphere, have a symbiotic relationship with plants, promote plant growth, and may act as an antagonist to plant diseases [13–18]. Commercialized products based on strains and isolates of *M. anisopliae* dominate the selection of mycoinsecticides worldwide. Formulation, application methods, targeted environment (arable or protected production), targeted crops, targeted pests, and strategies of use (inundative and non-inundative, or conservative way) are varying [4,5,10,19]. The potential of *M. anisopliae* isolates against pests of sweet potato (*Ipomoea batatas*) has been tested for more than three decades. One of the earliest virulence tests was performed in 1984, where the e fficacy of three *M. anisopliae* strains were investigated in laboratory conditions against adult individuals of the sweet potato weevil (*Cylas formicarius*) [20]. A subsequent study compared 12 isolates of three fungal entomopathogens including *M. anisopliae*, also on adults of the same pest [21]. This resulted in one of the *M. anisopliae* isolates giving the lowest LD50 values. In another laboratory experiment, *M. anisopliae* isolates were found not only to infect and destroy coleopterans, but to have an e ffect on the feeding and reproduction characteristics of *Cylas puncticollis* as well [20]. It was only in 2014, when the pathogenicity of *M. anisopliae* against *C. formicarius* was evaluated not only as a standalone treatment, but in a combination with *Beauveria bassiana* [22]. The possible ways of transmitting the fungal disease in sweet potato beetle was investigated when fecundity, expressed in the number of eggs and the rate of viable eggs was significantly hindered even when the eggs themselves had no contact with the fungus. It appeared that the presence of *M. anisopliae* altered the behavior of the pest, resulting in less eggs being positioned appropriately [23].

One of the earliest accounts of testing the e fficacy of the fungus in field conditions dates to 1998, when damage by the Banded Cucumber Beetle (*Diabrotica balteata*) and White grub (larvae of *Phyllophaga* spp.) were evaluated using *M. anisopliae* [24]. Although a single application before planting was found to have promising results against *D. balteata*, the e ffects on the other pest (i.e., Melolontha larvae) were uncertain, which may sugges<sup>t</sup> that more research should be focusing on finding the conditions to enhance the e fficacy of *M. anisopliae* on *M. melolontha* larvae [24].

Laboratory essays and open field experiments together suggests that, there are many abiotic and biotic factors contributing to the success and failure of using *M. anisopliae* in pest control. Among them several factors need further attention, such us soil chemical composition, soil microbiota, and biological activity [3,25]. Soil properties are governed by a complexity of factors, so in order to obtain helpful suggestions that can be used in the practice of Integrated Pest Management (IPM) or organic production, complex studies are required, with a set-up of complex models, and their viability must be trialed in realistic situations as well [26]. Altogether, more information is needed on what mechanisms endophytes establish and interact within a plant, the *M. anisopliae* on circumstances that favor the establishment of endophytism, so as to utilize its benefits [18,27,28]. Since the e ffect of *M. anisopliae* against *M. melolontha* larvae in sweet potato has not been a widely researched topic, we set up the present study to find answers to the following questions. (1) Can the fungal entomopathogen *M. anisopliae* strain NCAIM 362 (commercialized against coleopteran larvae) serve as an effective biological control agen<sup>t</sup> against *M. melolontha* larvae in sweet potato?; (2) In sweet potato production, which soil parameters can significantly influence the efficacy of *M. anisopliae*?; (3) Is *M. anisopliae* more effective in sweet potato than the chemical insecticide?

#### **2. Material and Methods**

#### *2.1. Experimental Set-up under Open Field and Greenhouse Conditions*

Open field experiments were conducted between 2018 and 2019. Sweet potato plants Beauregard variety were obtained in 4-leaf stage from the Lajosmizse Sweet Potato Company (Lajosmizse, Hungary), and planted in eight rows/block, each row containing 22 plants. The soil was chernozem (6.5 pH). The field was chosen for our experiment because the soil inhabiting pests was dominated by *M. melolontha* larvae. This was determined before the experiment, with an average of one 3rd instar larvae/m<sup>2</sup> detected. *M. melolontha* larvae infection was also influenced by the nearby (200 m distance) oak forests and orchards (100 m distance, mostly apple, pear, and plum trees at a 1.4 ha area). Since open field sweet potato production in the temperate zone usually involves the application of compost and soil cover systems (using agro-foil or textile), we followed and tested this routine. The eight rows and 22 plants within each row were also treated or not with compost and covered by agro-foil or textile (Figure 1, Figures S1 and S2). From each row, half of the plants were treated with *M. anisopliae* strain NCAIM 362 and the other half served as control (no *M. anisopliae*). There were 4 replications to each treatment, resulting in a total of eight replicates for each type. The presence of compost was marked K<sup>+</sup> or K<sup>−</sup>; the presence of agro-foil and textile; and the presence or absence of *M. anisopliae* (M+ or M−) (Figure 1A, Figures S1 and S2). The whole system was set up at the end of May, 2018, connected to automatic irrigation system (Irritrol junior max, placed below the soil cover systems, so each plant go<sup>t</sup> the same amount of water), while the *M. anisopliae* treatment in Wettable Powder (WP) formulation (as it was commercially recommended) was added on 27 June, after all plants were carefully checked. No plant pathogen symptoms or pest damages were detected on plants, and all plants were in the phenophase when the fungal entomopathogen treatment was added. This was done by preparing a 10% fungal solution (1400 g *M. anisopliae* to 12.6 L of water) transferred to all 700 plants. Treatment was added to each plant separately using a 20-mL syringe. The whole system was daily controlled until harvest. Crop harvest started on 1 October, with leaves and stems harvested first. Next, all soil covers were removed, and tubers were mechanically harvested. Each tuber from each treatment and cover system were separately collected, and tuber weights for each plant were measured and assigned to cover systems and treatments (M+ or M−). Because synthetic pesticides (i.e., α-cypermethrin) against soil inhabiting insects' larvae are not allowed in open field sweet potato control in Europe, this treatment was only used under well controlled conditions in a greenhouse experiment. Next, the damage made by soil inhabiting insects' larvae were evaluated using the following classification system: 0—no damage, 1—superficial damage, found only on the epidermal surface of tubers, 2—deep damage, found in deeper tissues (Figure 2). As no severe damages were detected, there was no reason to set up more levels in our classification system. The weight of missing tuber parts at level 2 damages were assessed by the following method: using gelatinized plastic with the same density as that of sweet potato tubers. Each hole was filled with this plastic. After drying (24 h), the plastic was removed and its weight (g) was measured (Figure 2). Yield was also measured at the end of the experiment by measuring every tuber under each plant. Weight results were averaged per compost use, soil cover systems, treatments and blocks. The whole experiment was replicated again in the next year, using the same cover systems, treatments, and methods.

Experimental set-up under greenhouse conditions was conducted in 2019, starting from May, parallel with the second-year field experiments. Soil properties, its chemical and microbial compositions and biological activities, were monitored under standardized and controlled conditions. The same sweet potato variety was obtained and used from the same company. For one experimental plot, there were

210 plants in three treatments (control 70 plants, fungal treatment 70 plants, and α-cypermethrin 70 plants); each divided in two sections (35 plants for each treatment) with (P+) and without (P−) *M. melolontha* larvae, all treatments replicated seven times again. Plants first were potted in 30 L plastic containers using 2:1 universal substrate/pea<sup>t</sup> ensuring the same soil pH as under open field conditions. Pots were then organized in rows (Figure 1B, Figures S3 and S4). The whole system was connected to an automatic irrigation system, controlled by Irritrol junior max. Temperature inside the greenhouses were controlled and kept around 35 ◦C during the vegetation period. Micro and macro elements were added twice, first after potting and later, in mid-July, to each plant using automatized Dosatron® systems. During the course of the whole experiment soil moisture, pH and EC were tested every three days. *M. melolontha* larvae were collected from natural environment (forest soil) from about 100 km from the experimental site and placed into the sweet potato containers when tubers were already developed, on 2 September. Two third-instar larvae were placed into each *M. melolontha*-treated container. The soil insecticide α-cypermethrin and *M. anisopliae* in a same WP formulation were added on 13 September. The insecticide was added in a concentration of 10 mL/10 L to each treated plant. The fungus was applied the same way and in the same concentration as described for the open field experiment. Tuber damage and yield weight were evaluated, as described above, too. The ratio of survived, dead, and infected *M. melolontha* larvae were counted at the end of the experiment by manually searching for larvae from containers after the plants were removed during harvest.

For soil chemical analyses, microbial assay and biological activity measurements from soil samples were collected twice: one month after planting (first week in June) and again, a month later. The same soil samples were divided and used for chemical assay, microbial analyses, and biological activity. From the soil of each treated and control plants (6 plants soil sample/treatments) 100 g soil was put into sterile pots and deposited at −70 ◦C until analyses. Damages on sweet potato tubers were assessed in the same way as under open field conditions.

#### *2.2. Chemical Composition Assay of Sweet Potato Soil*

EDX measurements were used to identify the elemental composition of the soil. Soil samples were dried in a drying cabinet at 80 ◦C to constant weight. Dried samples were powdered in a mortar using an electric grinder and were stored in airtight boxes. Samples were evaluated in homogenized powder form using a JEOL (Peabody, MA, USA) JSM 5510 LV scanning electron microscopy at various magnifications. The same samples were further analyzed with Scanning Jeol JEM 5510 JV and Oxford Instruments EDS Analysis System Inca 300 (UK) to determine the elemental composition of samples (*W*t%). Values are the means of five measurements from each soil samples and replicates [29,30].

#### *2.3. 16S rRNA Gene Amplicon Sequencing of Soil Bacterial Community and Biological Activity Assay*

The soil bacterial community analysis was performed based on amplicon sequencing of the 16S rRNA gene as in our previous work [31]. Briefly, total genomic DNA was extracted using the DNeasy PowerSoil Kit (Qiagen), a part of the 16S rRNA gene was amplified with primers containing the Bacteria-specific sequences Bakt\_341F (5-CCTACGGGNGGCWGCAG-3; [32]) and Bakt\_805NR (5-GACTACNVGGGTATCTAATCC-3; [33]), and DNA sequencing was performed on an Illumina MiSeq platform using MiSeq standard v2 chemistry as a service provided by the Genomics Core Facility RTSF, Michigan State University (East Lansing, MI, USA). There, Illumine-compatible, dual indexed adapters were added by PCR with primers targeting the CS1 and CS2 sites. PCR products were then batch normalized using SequalPrep DNA Normalization plates and all product recovered from the normalization plate was pooled. Subsequently, a clean-up of this pool was performed with Agencourt AMPure XP magnetic beads. Quality control and quantification was carried out using a combination of Qubit dsDNA HS (Thermo Fisher Scientific, Waltham, Massachusetts, USA), Fragment Analyzer High Sensitivity DNA (Advanced Analytical) and Kapa Illumina Library Quantification qPCR (Kapa Biosystems, Wilmington, MA, USA) assays. The pool was then loaded onto a standard MiSeq v2 flow cell Illumina. Sequencing was performed in a 2 × 250 bp paired end format using a v2, 500 cycle

MiSeq reagen<sup>t</sup> cartridge. Custom sequencing and index primers complementary to the CS1/CS2 oligomers were added to appropriate wells of the reagen<sup>t</sup> cartridge. Base calling was done by Illumina Real Time Analysis (RTA) v1.18.54 and output of RTA was demultiplexed and converted to FastQ format with Illumina Bcl2fastq v2.19.1. Raw sequence data were submitted to NCBI under BioProject ID PRJNA632727.

**Figure 1.** (**A,B**). Field (**A**) and greenhouse (**B**) experiment. Field experiment was replicated 4 times, having eight replicates for each cover. Greenhouse experiment was replicated seven times. Abbreviations: *M. anisopliae* present (M+) or absent (M−). Blue represents control, red represents insecticide, white represents fungal treatment.

For biological activity assays homogenized soil samples, sieved through a 1.6 mm sieve to remove stones and plant debris were used. For the FDA hydrolysis, 1 g of soil was measured, placed in a 500-mL conical flask, 50 mL of 100 mM potassium phosphate buffer (pH 7.6) and 0.l5 mL 12.01 μM FDA was added to start the reaction. Blank was prepared without the FDA substrate along with a control probe without soil sample. Time was monitored, and the hydrolysis took place at 37 ◦C for 1 h, hand-stirring every 5 min. After the hydrolysis, 2 mL of acetone were added to each probe to stop the reaction. Then the probes were centrifuged on 4000 rpm for 10 min and sieved through Whatman nr. 1 filter papers. Fluorescein concentrations were determined with spectrophotometer (PG Instruments T60 UV/VIS Spectrophotometer) on 490 nm. The obtained absorbance values were placed in the equation of calibration graph obtained by 0.03–10 μg/mL fluorescein standards, from where we obtained the

FDA enzyme activities of the soil probes in μg/g soil/h. Determination was replicated three times for each sample and treatment.

**Figure 2.** Sweet potato deep damage (defined in gram/tuber) on tubers with compost and fungal treatment and different soil cover systems. Analysis of variance (ANOVA) was used, followed by Tukey's HSD test to compare the effect of *M. anisopliae* on tuber damages using average damage/tuber/plant/compos<sup>t</sup> application/soil cover/block (*n* = 22). Grey bars represent *M. anisopliae*, blue bars represent control (no fungus). Bars represent standard errors. Upper figure presents damage assessment using gelatinized plastic. Different letters (a, b) means statistical significant differences.

## *2.4. Data Analyses*

Sweet potato damage data from the field experiment were first tested for the normality of errors and homogeneity of variances. Because data were normally distributed, analysis of variance (ANOVA) was used, followed by Tukey's HSD (Honestly Significant Difference) test to compare the effect of *M. anisopliae* on tuber damages (deep damages only) using average damage/tuber/plant/compos<sup>t</sup> application/soil cover/block/treatment (*n* = 11). Data were first compared between years, then block and side effects were tested using multivariate ANOVA; MANOVA). Because no significant differences were detected between years, and no blocks and side effects detected, pooled and averaged data between years were used for further analyses. Next, crop yield (average tuber weight/plant/compos<sup>t</sup> application/soil cover/fungal treatment/block (*n* = 11) were compared between control and fungal treatment using the same method (data were normally distributed).

Data from greenhouse experiment were again tested for the normality of errors and homogeneity of variances. Here, only the crop weight data was normally distributed, therefore analysis of variance (ANOVA) was used, followed by Tukey's HSD test to compare the effect of treatments (fungal, insecticidal treatment and control) using average tuber weight (g)/plant/treatment/block (*n* = 35). Tuber damage data and *M. melolontha* larval survival and infection data did not meet the assumption of normality, therefore the nonparametric Kruskal-Wallis test was used, followed by a Mann-Whitney U test to compare damages (averaged on plants/treatments/blocks (*n* = 35)) and average survived and dead larvae (average number/plant/treatment/block (*n* = 35)). All analyses were made using R version 3.0.1 [34] and values below *p* ≤ 0.01 were considered as statistically significantly different.

Chemical composition values of the soil were compared between collection dates and between treatments using analysis of variance (ANOVA), followed by Tukey's HSD test (data of five measurements/treatments and control).

Statistical analyses of soil bacterial communities were described in Benedek et al. [31], the differences were that the resulting sequence reads were processed using the mothur v1.41 software ([35]; based on the MiSeq standard operating procedure, downloaded on 03/04/2020) and the removal of chimeric sequences was performed using VSEARCH [36]. OTUs (operational taxonomic units) were defined at a 97% nucleotide sequence similarity level. For the statistical analysis of amplicon sequencing data, the subsampling of reads was performed to the read number of the smallest dataset (*n* = 19,791). Microbial α diversity (estimated using the Shannon-Wiener and Inverse Simpsons's (1/D) diversity indices) and species richness values (using the Chao1 and the ACE richness metrics) were calculated using mother v1.38.1. Linear regression was used to assess the variation in total bacterial diversity indices (Shannon and Simpson) under different treatments and control, *R*<sup>2</sup> values computed using PAST. Variation in bacterial community composition was also compared between genera for each treatment and control with ANOVA followed by Welch F test using mean percentages of DNA from total samples.

Data of soil biological activity was again normally distributed, thus, analysis of variance (ANOVA) was used, followed by Welch F test to compare the biological activity under different treatments and control using average data/plant/block (*n* = 6). Analyses were made using R version 3.0.1 [34].
