Pesticide soil microbial toxicity: setting the scene for a new pesticide risk assessment for soil microorganisms (IUPAC Technical Report)

3.1 Pesticide exposure estimation

Determination of the level and the duration of the exposure should be an indispensable part of any toxicity assay. Temporal measurement of the dissipation of pesticides during any assay would define the level and the duration of exposure of the soil microbial community to the pesticide in question. Furthermore, detection and quantitation of pesticide TPs formed in soil would provide information on the potential involvement of TPs on toxicity effects observed. This is already done in the current regulatory framework, although the selection of TPs, relevant for inclusion in risk assessment, relies solely on the quantity of the TP formed (more than 10 % of the parent), excluding TPs that do not meet the 10 % rule but show high potency due to their chemical structure or mode of action. Correlation testing between the measured soil concentrations of pesticides and their TPs with temporal microbial measurements would point to the causal agent of the potential toxicity on the soil microbial community (parent vs TPs or a combination of both). Such an example is offered by the study of Vasileiadis et al. [31], who identified via correlation testing and a series of soil and in vitro tests, 3,5-dichloroaniline, the main TP of the fungicide iprodione (3-(3,5-dichlorophenyl)-N-(propan-2-yl)-2,4-dioxoimidazolidine-1-carboxamide) in soil, as the main driver of the toxicity observed on a range of functional microbial attributes.

3.2 Standardization of methods in soil microbial ecology

The recent methodological advances in soil microbial ecology and microbial ecotoxicology revealed a previously unprecedented microbial complexity and diversity and highlighted the multifaceted role of soil microorganisms in ecosystem functioning [35]. The methods available could be classified into: (a) functional methods, that determine the activity of enzymes, rates of microbial processes or the activity of microbial groups controlling key functions in terrestrial ecosystems; (b) diversity methods, which determine the composition of the soil microbiota operating at varying resolution from phylogenetically distinct (e.g. bacteria, archaea, fungi) to functionally distinct microbial groups (i.e. AOM, AMF, methane-oxidizing microorganisms, nitrogen-fixing bacteria, denitrifying bacteria). A key step for the implementation of these established and advanced methods in environmental risk assessment is their standardization. Standardization of functional methods is well advanced compared to the diversity methods.

Functional methods include measurements of (i) the activity of soil microbial exoenzymes involved in carbon, nitrogen, phosphorus and sulfur cycling, and (ii) the rates of microbial processes like soil microbial respiration, potential nitrification or denitrification. These methods have been used in several studies for assessing the toxicity of pesticides on soil microorganisms [25, 33, 34, 38, 39]. For these methods, relevant ISO standards are available (Table 1), while at the same time high throughput approaches facilitating their rapid regulatory uptake have been developed for measuring soil microbial exoenzymes [40] and soil microbial respiration through MicroResp^® [41]. However, the frequent lack of consistency in the response of these methods to pesticide exposure, appearing as effects deviating from the expected dose-dependent pattern, might be a plausible explanation for their exclusion from environmental risk assessment [42, 43].

Nucleic acid-based methods determining the abundance (DNA) or the activity (RNA) of key functional microbial groups via quantitative Polymerase Chain Reaction (q-PCR) and Reverse Transcription-q-PCR (RT-q-PCR), respectively, could be also considered as tools for measuring microbial functions, although they were not considered in environmental risk assessment due to lack of standardization [43]. Relatively recent advances with the development of ISO standards for direct soil DNA extraction (ISO11063-2012 replaced by ISO11063-2020) [44] and determination of the abundance of soil microbial groups via q-PCR (ISO17601) [45] addressed these concerns and paved the way for their implementation in pesticide environmental risk assessment. Several soil studies have determined effects of pesticides on the abundance of phylogenetically distinct microbial groups like bacteria, fungi or archaea or the abundance of functional microbial groups like AOM (ammonia-oxidizing microorganisms) [31, 46, 47], sulfur-oxidizing bacteria [34] and degraders of biogenic aromatic compounds [48] using q-PCR standardized protocols. Besides mere determination of microbial abundance in soils, RT-q-PCR approaches based on soil RNA offer a more sensitive measurement of the potential effects of pesticides on microbial activity, by measuring the number of mRNA transcripts of microbial genes involved in key soil microbial functions. However, only a few studies have used RT-q-PCR approaches to determine pesticide effects on soil microbial activity, mostly targeting AOM [42]. This most probably reflects the sensitivity of soil RNA to lysis during laboratory handling which often results in low recovery of gene transcripts.

Diversity methods include Phospholipid Fatty Acids Analysis (PLFAs), molecular PCR-based fingerprinting like Denaturating Gradient Gel Electrophoresis (DGGE) and Terminal Restriction Fragment Length Polymorphism (TRFLP), and the more recent and powerful amplicon sequencing approaches. Phospholipid Fatty Acids Analysis is a well standardized method (ISO/TS29843-1 and -2) [49, 50], which provides information about the overall size and the composition of the living soil microbial community, however at a rather low phylogenetic resolution [51]. Conversely, PCR-based molecular fingerprinting or sequencing methods could provide robust phylogenetic information at family, genus or even species level, but are faced with limited standardization. The depth of phylogenetic analysis of the soil microbial community offered by these methods has increased from the lower resolution of earlier fingerprinting methods like TRFLP and DGGE [52, 53], to the higher resolution amplicon sequencing approaches or so called meta-taxonomics [54]. Denaturating Gradient Gel Electrophoresis and TRFLP, either as stand-alone approaches or in combination with clone libraries, have been heavily used in the period of 2000–2015 to determine effects of pesticides on the diversity of phylogenetically and functionally distinct microbial groups [55, 56]. However, they could not provide quantitative information for less abundant members of the soil microbiota. Since 2015, several studies have used meta-taxonomic approaches to identify effects of pesticides on the diversity of bacteria, fungi [57, 58] and functional microbial groups like AOM [31]. Benchmarking protocols for the preparation and setup of meta-taxonomic analysis of the soil bacterial (https://earthmicrobiome.org/protocols-and-standards/16s/) and fungal diversity (https://earthmicrobiome.org/protocols-and-standards/its) were developed by the Earth Microbiome Project and have been largely adopted by most recent pesticide soil ecotoxicity studies making a major step towards their formal standardization. However, what we are still missing is the standardization of the bioinformatic pipeline for the analysis of amplicon sequencing data and the way to make use of these multivariate data in a regulatory context. These issues and ways to resolve them are further discussed in Sections 4.5 and 5.1 respectively.

Thiele-Bruhn et al. [59] classified the different methods available in soil microbial ecology according to the Ecosystem Services Approach, in line with the recommendation of EFSA [19], that functional endpoints are more attainable and ready-to-implement compared to diversity endpoints. In this approach, methods, traditional and new ones, were classified according to their relevance for a list of eight ecosystem services, namely: (a) biodiversity, genetic resources, cultural services; (b) food web support; (c) biodegradation of pollutants; (d) nutrient cycling; (e) pest control and plant growth promotion; (f) carbon cycling; (g) greenhouse gas emissions; (h) soil structure formation. Most of the above-mentioned services are emanated from the same basic functions. The authors highlighted the potential of q-PCR-based methods to define, in a standardized way, pesticide effects on microbial groups performing key soil functions involved in soil fertility (nitrogen and phosphorus transformation, plant growth promotion) and environmental quality (greenhouse gas emissions, natural attenuation of pollutants). All these point towards the use of endpoints which represent a microbial function until we will be able to make use of the wealth of data derived from meta-taxonomic analysis in a standardized way and in a regulatory context.

3.3 Soil microbial indicators

A first assessment of the potential toxicity of pesticides on aquatic and terrestrial macro-organisms is performed through tests on specific organisms from different trophic levels identified as bioindicators (single species tests). A cornerstone organism for pesticide ecotoxicology is Daphnia magna, used as an indicator of aquatic invertebrates, while Onchorhynchus mykiss, Eisenia foetida, Selenastrum capricornutum and Lemna minor are used as toxicity indicators for fishes, earthworms, algae and aquatic plants, respectively [19, 60]. These bioindicators were selected based on the following criteria: (i) they have a key ecological role; (ii) they are characterized by high sensitivity to pesticides, compared to other species in the same group of organisms; (iii) they show a clear ecotoxicological response to pesticides; (iv) we have a very good knowledge of their life cycle; and (v) there are standardized tests available for the determination of their response to pesticides.

If the same logical path is to be followed for the establishment of a new framework to assess the toxicity of pesticides on soil microorganisms, new microbial groups which fulfil the above criteria should be identified to serve as candidate bioindicators. Previous studies have pointed to AOM [61], AMF [19, 33], nitrogen-fixing bacteria [62], protists [63], microalgae [64] and cyanobacteria [65], all of which are involved in key soil microbial functions as shown in Fig. 2. In the following sections we will try to describe the pros and cons of these potential bioindicator microbial groups with relevant examples from the literature.

Fig. 2:

The different functional microbial groups that could be used as indicators of the soil microbial toxicity of pesticides and their ecological role in soil. Upon their application, either on the foliage or directly in soil, pesticides come into contact with different microbial groups performing significant soil functions like (a) Arbuscular mycorrhizal fungi (AMF) that contribute to enhanced PO₄³⁻ uptake by plants; (b) nitrogen-fixing bacteria (NIF), which fix atmospheric N₂ to NH₄⁺, operating as free-living or as symbionts in the roots of leguminous plants; (c) ammonia-oxidizing bacteria (AOB), ammonia-oxidizing archaea (AOA), comammox bacteria and nitrite-oxidizing bacteria (NOB) collectively transforming NH₄⁺ to NO₃⁻ (d) denitrifying bacteria that transform NO₃⁻ to N₂ through the intermediate production of deleterious gases like NO and N₂O (e) cyanobacteria on surface soil that photosynthesize and release O₂ in soil (f) heterotrophic bacteria and fungi that decompose soil organic matter and release organic nitrogen that could be further hydrolyzed to NH₄⁺ feeding into the nitrification process (g) protists that server a range of soil functions acting as saprotrophs, phototrophs and phagotrophs feeding on bacteria, fungi and small eukaryotes.

3.4 A new risk assessment scheme and tests required for the determination of the toxicity of pesticides on soil microorganisms

3.4.1 A new risk assessment scheme proposed

Environmental risk assessment of pesticides relies on a tiered system composed of tiers of increasing complexity and realism. Tier I is based on the assessment of the toxicity using highly conservative assays. If risk assessment indicates potential undesirable effects at this step of the process, then assessment is performed at Tiers II and III, which are characterized by less conservative and more realistic assays at both exposure and toxicity level. This tiered scheme is the backbone of the current environmental risk assessment of pesticides, but this is not yet applicable to soil microorganisms. Earlier studies in the late 70s provided the first experimental guidelines on how to determine the toxicity of pesticides on soil microorganisms [25, 138]. They suggested that if significant inhibitory effects are observed at lab scale, the toxicity of pesticides on soil microorganisms should be further explored at field scale. Following the same philosophy, Karpouzas et al. [139] proposed a similar tiered system for the assessment of the soil microbial toxicity of pesticides. Their approach was composed of three tiers of increasing complexity based on the ecotoxicological response of key soil functional groups like AOM and AMF or others which remain to be defined (Fig. 3): (i) a Tier I in vitro screening of pesticides against a set of soil-derived AOM and AMF strains that cover the different ecophysiological and phylogenetic variants of these microbial groups; (ii) a Tier II toxicity assessment in lab soil microcosms (or planted pot studies when AMF are considered) against natural assemblages of AOM and AMF; and (iii) a Tier III toxicity assessment at field scale (under cropping conditions) against natural assemblages of AOM and AMF. In cases where an unacceptable risk for soil microorganisms is still evident, refinement of exposure could be an option to minimize risk.

Fig. 3:

A three-tier risk assessment scheme proposed for assessing the risk associated with the use of pesticides for soil microorganisms. AOM: Ammonia-oxidizing microorganisms; AMF: Arbuscular mycorrhizal fungi (adopted upon permission by Karpouzas et al. [129]).

4.1 Low-risk pesticides

The growing public concern about the effects of synthetic pesticides on environmental quality and soil health, has shifted attention to low-risk pesticides. However, their introduction in pesticide market has been hampered by the stringency of the current EU regulatory framework which treats low-risk pesticides in the same way as synthetic ones [152, 153]. This is reflected in the much lower number of low-risk pesticides reaching the market in Europe compared to USA, China, Brazil and India [154]. Low-risk pesticides is a broad group which encompasses (a) microbials, where the active agent is a microorganism that protects crops from fungal and insect infestations; (b) natural products or biochemicals or botanicals and peptides, that are biogenic compounds, products of the secondary metabolism of plants and microorganisms, with strong biocidal activity; (c) semiochemicals, including pheromones and allelochemicals, that are emitted by animals, plants and other organisms and impose a physiological or behavioral response in individuals of the same species or other species respectively [155]; and (d) double stranded RNA (dsRNA)-based pesticides that work through the natural gene silencing RNA interference (RNAi) mechanism of several living organisms [156]. Their mode of action is based on gene-specific small interfering RNA (siRNA) or microRNA (miRNA) molecules that bind on mRNA with complementary nucleotide sequences and trigger their lysis, hence preventing protein production [157]. Prevention of the biosynthesis of proteins involved in the pathogenesis of pests or pathogens could lead to plant protection.

As their name indicates, low-risk pesticides are characterized as safe and environmentally friendly solutions, mostly due to their biological origin. However, hard evidence to verify their low-risk character is scarce. Amongst natural products, azadirachtin is the hallmark compound of this group. It is one of the most studied low-risk pesticide regarding its off-target toxicity to soil microorganisms and the results obtained to date do not support a low-risk profile. Azadirachtin showed low in vitro toxicity towards AMF, compared to synthetic pesticides [76]. This was not verified in soil studies, where the recommended dose of azadirachtin negatively affected the abundance and the transcriptional activity of AOM, nitrogen-fixing and denitrifying bacteria [98, 158] and reduced the diversity of bacteria, fungi and AMF [56, 99]. In the same study, azadirachtin showed equivalent or stronger inhibitory effects than the synthetic pesticides tested. When looking at semiochemicals, little is known regarding the off-target toxicity of pheromones to the soil microbiota, as expected due to their volatile nature that precludes exposure of soil microorganisms. On the other hand, there are a few studies looking at the effects of allelochemicals acting as herbicides on the soil microbiota. Hence, Romdhane et al. [29] compared the effects of the allelochemical β-triketone herbicide leptospermone (2,2,4,4-tetramethyl-6-(3-methylbutanoyl) cyclohexane-1,3,5-trione) and its synthetic derivative sulcotrione (2-[2-chloro-4-(methanesulfonyl)benzoyl]cyclohexane-1,3,5-trione) on the soil microbial community. In line with the results of azadirachtin, leptospermone induced stronger changes in the soil bacterial community composition than sulcotrione. These studies certainly challenge the general perception that natural products are characterized by lower off-target toxicity compared to synthetic pesticides.

In accordance with the other low-risk pesticide groups, little is known regarding the potential effects of microbial pesticides on the soil microbiota. Potential toxicity effects of microbials on the soil microbial community largely depend on the mode of action of the microbial pesticide itself. Hence microbial pesticides based on microorganisms which do not act through the production of biocidal compounds are not generally expected to affect soil microorganisms. This was clearly the case for the nematode parasitic fungi Paracoccus lilacinus strain PL251, which did not have a direct inhibitory effect on AOM [101]. In contrast Yu et al. [159], using a different P. lilacinus strain PL1210, showed strong inhibitory effects on nitrification and AOM abundance, which were attributed to antimicrobial metabolites that the tested strain produces. Other relevant studies also suggested that microbial pesticides based on microorganisms acting through parasitism or antagonism (i.e. Metarhizium brunneum, Fusarium oxysporum f.sp. stringae, Bacillus amyloliquefaciens) did not appear to induce strong and persistent effects on the soil microbial community [160], [161], [162]. Still the potential of a microbial pesticide to adversely affect the soil microbiota should be determined at the strain level taking into consideration its capacity to produce bioactive metabolites [163] or its cargo of transferable antibiotic resistance genes [164]. Such information could be initially deduced by the genomic data of the microbial strain and further verified experimentally.

The dsRNA-based pesticides have not reached the market yet, but they are expected to become available in the forthcoming years. Although studies regarding their efficacy [165], mode of application [166], persistence [167] and toxicity on non-target organisms [168] are available, we still have no idea about their potential effects on the soil microbiota. The highly specific mode of action of the RNAi mechanism on target mRNA suggests limited off-target toxicity effects. Bioinformatic analysis focusing on the design of dsRNA molecules with minimum homology to mRNA of non-target organisms, coupled with biological experimentation starting with phylogenetically close organisms should be considered in the assessment of the soil microbial ecotoxicity of dsRNA pesticides [169]. However, differences in the RNAi machinery of different organisms and variable uptake and internal distribution of dsRNA in different organisms are factors that might strongly influence the efficiency but also the ecotoxicity of ds-RNA pesticides [170], including effects on the soil microbiota.

According to the registration framework in Europe, low-risk pesticides undergo the same registration process as synthetic pesticides. We argue that the risk assessment procedure for the different types of low-risk pesticides should be adjusted to account for the characteristics, properties and needs of these products. This should certainly include the implementation of tools for monitoring the environmental fate and persistence of low-risk pesticides in soil.

4.2 Nano-pesticides

Nano-pesticides are plant protection products which are composed of nanoparticles, particulate substances which have one dimension less than 100 nm, that are used to increase the efficacy and reduce the environmental footprint of pesticide active ingredients [171]. They could be classified into two broad categories: (i) organic pesticide active ingredients that are complexed onto nanocarriers (soft nanoparticles like polymers, solid lipid or hard nanoparticles like silica or carbon nanotubes), (ii) inorganic active ingredients without a nanocarrier that are used directly (e.g. CuO, Cu(OH)₂) [172, 173].

Nano-pesticides of the first group (organic pesticides complexed onto nanocarriers) show increasing efficacy (by 10 to 20 %), and persistence, whereas we know very little about their ecotoxicity compared to their corresponding conventional formulation and the pure active ingredients. For now, there are only a few studies looking at the off-target toxicity, including the soil microbiota, of organic nano-pesticides, their conventional formulations and the active ingredients. For example, Maruyama et al. [174] assessed the impact of the herbicides imazapic (rac-2-[4-methyl‐5-oxo‐4-(propan‐2-yl)-4,5-dihydro-1H-imidazol-2-yl]-5-methylpyridine-3-carboxylic acid) and imazapyr (rac-2‐[4-methyl‐5‐oxo-4-(propan-2-yl)-4,5-dihydro-1H-imidazol-2-yl] pyridine-3-carboxylic acid) encapsulated in alginate/chitosan and chitosan/tripolyphosphate nanoparticles on soil microorganisms involved in nitrogen cycling, but they did not include in their study any conventional formulation of the two herbicides. They found that encapsulated herbicides were less toxic compared to the non-encapsulated compounds.

Regarding nano-pesticides of the second group (inorganic active ingredients without a nanocarrier), we have got several studies that have investigated their effects on the soil microbiota. In a few studies, the authors examined the toxicity of nano-pesticide formulations comparatively to the inorganic active ingredient to be able to distinguish the contribution of the two components, inorganics or nanocarriers, in the toxicity observed [175, 176]. In the most comprehensive of these studies, Swart et al. [175] analyzed separately the effects of the formulated silver and CuO nanoparticles, their ionic counterparts CuCl₂ and AgNO₃, and their inert carriers on the soil microbiota at different concentration levels. They noted a dose-dependent effect on the soil bacterial community by the formulated products and their ionic controls but not by the nanocarriers, suggesting that the inorganics are the main drivers of the effects on soil microbiota. In support of this, Zheng et al. [177] demonstrated, via in vitro testing, that the toxic effect of silver nanoparticles on Paracoccus denitrificans, a model denitrifying bacterium, was mostly driven by silver ions decreasing the expression of denitrification and glucose assimilation enzymes and increasing polyhydroxybutyrate biosynthesis. Several other studies have reported the effects of inorganic nano-pesticides on the soil microbiota with the measured endpoints varying. Sillen et al. [178] studied the impact of silver nanoparticles on the activity of the rhizosphere microbiota via transcriptomic analysis. They noted an increased transcription of genes involved in copper resistance (e.g. efflux pumps of Ag/Cu) and a decreased transcription of ammonia monooxygenase and urease of AOA but not of AOB. In a similar study, Simonin et al. [179] showed that Cu(OH)₂ nano-pesticides, applied to soil under three different fertilization regimes (ambient, low, high), did not induce any significant effect on AMF and nitrogen-fixing bacteria. Similarly, Xu et al. [180] tested the effects of TiO₂ and CuO nano-pesticides on the soil microbiota and showed that the latter exhibited higher toxicity, as denoted by the significant reduction of microbial biomass carbon, total PLFAs, urease, phosphatase and dehydrogenase activity.

Overall, the small size and high surface/volume ratio of nano-pesticides coupled with the high toxicity of their inorganic active ingredients, favor their strong interaction with soil microorganisms and increase the possibility of effects on the soil microbiota. Future studies should focus on the assessment of the effects of organic nano-pesticides on the soil microbiota, comparatively to their conventional formulation and active ingredients.

4.3 Pesticide mixtures (and co-formulants)

Formulated pesticides are composed of the active ingredients and several other co-formulants that aim to maximize active ingredients efficiency on the target. The identity of these co-formulants in pesticide commercial formulations is not disclosed due to intellectual property rights. However, they could be responsible for or enhance the toxicity (e.g. by increasing the soil bioavailability) of pesticide compounds on soil microorganisms. Ecotoxicological studies exploring the toxicity of active ingredients and the corresponding commercial formulation should be employed to distinguish the contribution of the different components of pesticide formulations on the soil microbial toxicity. In such a study, Crouzet et al. [136] showed that the commercial formulation of the herbicide mesotrione, when applied at 10 times the recommended dose, induced stronger effects on the structure of the soil cyanobacteria community compared to the same levels of the pure pesticide compound. In a similar study, Rousidou et al. [101] showed that co-formulants of the formulation of the bionematicide BIOACT^® (glucose or skimmed milk), which contains spores of the nematode parasitic fungus P. lilacinus, were responsible for the transient inhibition of AOM upon soil application of BIOACT^®.

Besides additives, several of the commercial pesticide formulations contain more than a single active substance [181], hence releasing mixtures of pesticides in the environment. This might trigger effects deviating from those seen when applied individually. To date, most studies available on the toxicity of pesticide mixtures have focused on aquatic toxicity endpoints [182]. They suggested that, for mixtures composed of active ingredients with the same or different mode of action, the concentration addition (CA) or the independent action (IA) models, respectively, could predict toxicity [183]. On the other hand, we still know little about the potential toxicity of pesticide mixtures on soil biota and microbiota. Evaluating the applicability of CA and IA models for assessing the toxicity of pesticides on the soil microbiota or designing new models, more relevant for soil ecosystems, could be a new frontier in pesticide soil microbial ecotoxicology.

4.4 Transformation products (TPs)

Environmental risk assessment should go beyond parent compounds and address also the potential toxicity of pertinent pesticide TPs. In a benchmarking study, Sinclair and Boxall et al. [184] explored via meta-analysis the toxicity of 89 transformation products (TPs) obtained from 37 parent compounds. They showed that 70 % of the pesticide TPs have either similar or lower toxicity to aquatic organisms compared to their parent compounds. The remaining 30 % of the TPs tested were more toxic than the parent compound. This list included TPs that (i) carried the pesticide toxicophore moiety (i.e. oxime carbamate compounds) (ii) were the active compounds formed from a pro-pesticide (i.e. organophosphate compounds) or (iii) had a more potent mode of action compared to the parent compound (i.e. carbamate and organotic compounds). The EU regulatory framework still uses an exposure-driven assessment approach whereby TPs that are formed at levels exceeding 10 % of the applied parent compound are considered as relevant for environmental risk assessment [185]. This approach has often been criticized as it excludes from risk assessment a large number of TPs formed at lower amounts than the cutoff value of 10 %: the rational for defining a relevant TP should rather be based on their toxicophore moieties or mode of action [186, 187]. A reflection of these decisions was the late (20 or 30 years after their market introduction) identification of toxicologically relevant TPs like the carcinogenic N-nitroso-dimethylamine (N,N-dimethylnitrous amide), a TP of the fungicide tolylfluanid (N-[dichloro(fluoro)methanesulfanyl]‐Nʹ,Nʹ-dimethyl-N-(4-methylphenyl)sulfuric diamide) [188]. The pitfalls of this exposure-driven a priori toxicological characterization of pesticide TPs were highlighted by Storck et al. [189] who proposed the implementation of a posteriori characterization of the toxicological relevance of TPs.

To date, there are several examples in the literature where TPs rather than the parent compounds are incriminated for toxicity on the soil microbiota. Papadopoulou et al. [42] identified quinone imine, a TP of the fruit preservative ethoxyquin, as the main driver of the significant reduction in the abundance and diversity of AOM in soils treated with ethoxyquin. More recently Vasileiadis et al. [31] showed that 3,5-dichloroaniline, a major TP of iprodione, was responsible for the strong inhibition in the abundance and activity of AOM in soils treated with iprodione. In both these studies the toxicity of TPs was tested at levels relevant to their rate of formation by the parent compound, an issue that should be always considered in relevant soil studies. One of the better illustrated examples of toxicity of pesticide TPs comes from chlorothalonil. Using a conservative exposure scenario, the repeated application (4 times) of chlorothalonil in soil at dose rates 5 times the recommended led to the production of levels of 2,4,5-trichloro-6-hydroxybenzene-1,3-dicarbonitrile which induced strong inhibitory effects on the soil microbiota [190, 191].

In silico screening of the potential toxicity of pesticides on the soil microbiota could be an invaluable tool for avoiding the market introduction of compounds whose application could lead to devastating effects on the soil microbiota. A database and prediction tool for the biotransformation of organic contaminants (EnviPath) [192], has been updated with all freely accessible EU regulatory data on pesticide degradation in lab soil studies with the aim to develop more accurate prediction for pesticide biotransformation pathways [193]. Complementary tools (i.e., QSAR) enabling the prediction of the soil microbial toxicity of TPs, such as TyPoL, could allow for a targeted investigation of the TPs toxicity. Further validation of these in silico prediction tools with in vitro and in soil experimental data will enable their use in the prediction of soil microbial ecotoxicity of pesticides in the frame of environmental risk assessment.

4.5 Diversity analysis

The introduction of amplicon sequencing approaches has revolutionized our view of microbial diversity across all environmental compartments [194], including soil, which is considered the most complex environmental matrix [195]. Amplicon sequencing could monitor, at high resolution level, the composition of bacterial, archaeal, fungal and eukaryotic soil communities [196]. In addition, specific protocols allow the analysis of the diversity of key functional microbial groups like AOA, AOB [31, 197], and AMF [198], although the depth and quality of the analysis of the functional microbial diversity could be compromised by the lack of high quality or updated databases (i.e. Abell et al. [87], for AOB). Sequencing data generation is followed by in silico analysis relying on online databases (Silva, UNITE, etc.) and the use of freeware bioinformatic tools. As all DNA-based PCR methods, amplicon sequencing analysis could be affected by the presence of relic DNA from dead soil microorganisms complicating the interpretation of the data [199]. DNA relics can be removed from soil samples through treatment with DNAses [200]. Alternatively, soil RNA (and thereof cDNA) being unstable and leaving no relics upon cell lysis, could be used as a target matrix for amplicon sequencing providing insights into the active fraction of the soil microbiome [201].

The different steps of the process (PCR amplification, Sequencing, Bioinformatic Analysis) show variable levels of standardization with PCR amplification steps being more standardized than data analysis, where several issues remain open [202]. The Earth Microbiome Project, a concerted action started in 2010 [203], provided standardized amplicon sequencing protocols for studying soil microbial diversity using validated barcoded primer sets and thermocycling conditions for the amplification of the 16S rRNA gene of bacteria and archaea [204], the ITS region of fungi [205] and the 18S rRNA gene of soil eukaryotes [206] (more details available at https://earthmicrobiome.org/). These protocols have been acknowledged by the scientific community and peers which encourage all groups working on soil microbiomes to utilize the Earth Microbiome Project protocols in relevant studies to ensure comparability and high quality of the data produced [207].

The analysis of amplicon sequencing data involves several steps like: (i) merging of amplicons, removal of primers and barcodes, quality control (ii) development of the data matrix by clustering into Operational Taxonomic Units (OTUs) or using Amplicon Sequence Variants (ASVs) (iii) production of the taxonomic table which will be used for further analysis [208]. In all these steps decisions made and parameters selected are known to strongly affect the outcome of the analysis. For example, the use of ASVs instead of OTUs have often a marked effect on the output of community diversity analysis, with the ASV method outperforming the OTU method when the sequencing depth is adequate [202]. The latter reflects the sequencing effort efficiency to capture the existing microbial diversity and could be determined via rarefaction analysis [209]. In a first effort to contribute to a standardized platform of analysis, the Earth Microbiome Project has proposed QIIME as a tool for bioinformatic processing of the sequencing data [210], while other software like DADA2 [211], and USEARCH [212] could be used for the same purpose. It is beyond the purpose of this paper to provide a detailed technical guide on amplicon sequencing data analysis. The readers could refer to excellent reviews [213], [214], [215], [216] and new perspectives in this topic [217]. Overall, the lack of standardization in amplicon sequencing data processing and analysis hampers their use in a regulatory context. We believe that the on-going effort by the scientific community to set such standards will pave the way for their inclusion in the regulatory scheme in a near future.

Today, several studies have used amplicon sequencing to study the effects of pesticides on the soil microbial diversity [31, 57, 110, 218]. Most of these studies were aligned with the amplicon sequencing protocols of the Earth Microbiome Project, enabling the meta-analysis of the global dataset. In this context, Rocca et al. [219] presented a first meta-analysis platform called the Microbiome Stress Project, which was based on amplicon sequencing datasets obtained from experiments looking on the effects of a range of stressors on microbial diversity. This platform could be used as a base for a future systematic meta-analysis of the effects of pesticides on the soil microbiota.

Besides the lack of standardization, soil microbial ecotoxicology lacks a conceptual framework on how to translate the enormous volume of information generated by amplicon sequencing to a realistic toxicity assessment for soil microorganisms. It is now rather well-documented that, despite functional redundancy, soil microbial diversity losses often lead to reciprocal negative effects on soil microbial functioning [150, 220]. However, we still need to define the normal operating range for the different microbial endpoints as determined by amplicon sequencing. This will enable us to establish threshold values beyond which observed microbial diversity losses will have critical ecotoxicological effects. For example, we still have no answer to the question “What is the level of soil microbial diversity reduction that we could consider as imposing no unacceptable risk for the soil microbiota?”. Aquatic ecotoxicologists have identified Hazard Concentration 5 % (HC5) as a threshold value for environmental concentrations of pesticides at which, if exceeded, an unacceptable risk to the aquatic biota occurs [221]. Forthcoming research efforts should define normal operating range values per microbial endpoint that can be used by soil microbial ecotoxicologists to define relevant environmental thresholds that will preserve soil microbial diversity and functions.

5.1 Introducing advanced statistical approaches in soil microbial ecotoxicology

Aquatic ecotoxicology was a pioneer in the development and implementation of advanced statistical approaches like species sensitivity distribution (SSDs) [222] or principal response curves (PRCs) [223]. The former has been used as a refinement tool to determine threshold environmental pesticide concentrations that have no unacceptable effects on 95 % of the population of a trophic level, relying on data from single species tests [224]. On the other hand, PRC is a multivariate ordination method used for the analysis of data derived from complex aquatic mesocosm studies. It provides an overview of the response of the community of a given group of organisms (i.e. arthropods) on a range of toxicant concentrations, while at the same time it gives information of the response of the individual species to the toxicant [225]. Webster et al. [226] first suggested the potential use of SSDs in soil microbial ecotoxicology. The construction of SSDs could be based on EC₅₀ values calculated from amplicon sequencing data for ASVs, whose abundance show a dose-dependent response (Fig. 4a). Species Sensitivity Distributions could be constructed for a whole microbial domain (i.e. bacteria, fungi, protists), distinct bacterial or fungal phyla or classes (i.e. Proteobacteria, Actinobacteria, etc.), but also for functionally distinct microbial groups like AOB, where phylogeny could provide direct functional assignments. Regarding functional microbial groups, in vitro toxicity tests using representative strains (equivalent to single species tests used in aquatic ecotoxicology) could be used to calculate EC₅₀ used as input data for the construction of SSDs per microbial function. These functional endpoints could be used for the construction of Functional Species Distribution providing pesticide environmental concentration levels that could have no unacceptable effects to 90 or 95 % of the key soil microbial functions (Fig. 4a). Similarly, amplicon sequencing data derived from soil microcosm or mesocosm studies could be used for the construction of PRCs for the whole prokaryotic, fungal or protist community providing a domain-specific assessment of their response to pesticide exposure while also pointing to specific microorganisms that exhibit differential response to the general pattern (Fig. 4a).

Fig. 4:

A glance in novel tools that could be used to advance the assessment of the toxicity of pesticides on soil microorganisms. (a) The construction of species sensitivity distributions (SSDs), Function Sensitivity distributions (adapted by Webster et al. [202]) and principal response curves (PRCs) using data from in vitro assays or amplicon sequencing datasets from soil studies; (b) synthetic microbial communities performing a collaborative function like nitrification (left bottle) composed of an ureolytic bacterium, an ammonia-oxidizing bacterium (AOB), an ammonia-oxidizing archaeum (AOA) and a nitrite-oxidizing bacterium (NOB) or simulating soil food-web (right bottle) composed of phagotrophic protists and bacteria or fungi whose function can be monitored like benzoate degraders, AOA and AOB.

The concept of SSDs was only recently employed to assess the toxicity of nano-pesticides on the soil microbial community. Hund-Rinke et al. [102] determined the effects of silver nanoparticles applied at five dose rates on the soil bacterial community using amplicon sequencing. They calculated EC₅₀ values for different bacterial orders and verified the sensitivity of Nitrosomonadales to silver nanoparticles, as derived from potential ammonia oxidation studies. More recently, Swart et al. [175] tested the toxicity of silver and copper nanoparticles on the soil and gut microbiota of earthworms. Based on bacterial amplicon sequencing data, they calculated ASV specific EC₅₀ values which were then used for the construction of SSDs and the calculation of HC₅, which indicated high risk for the soil microbiota by CuO nanoparticles.

The use of SSDs but also PRCs could make feasible the implementation of amplicon sequencing data in pesticide risk assessment. Alternatively, SSDs derived from (i) meta-analysis of already available amplicon sequencing datasets from studies that meet certain experimental criteria of quality and (ii) novel soil microcosm studies (including control and at least four dose rates) testing the toxicity of a range of pesticides on soil microorganisms using amplicon sequencing measurements, could be used to identify responsive and sensitive microbial groups. These could be further promoted as microbial indicators in soil microbial ecotoxicity testing.

5.2 Ecosystem level assessment

Most soil studies to date have assessed the toxicity of pesticides on individual taxa or functional groups separately [62] or in the context of a biochemical pathway (e.g. denitrification), ignoring the ecological dimension of the effects observed [106]. In nature, microorganisms are assembled into complex and diverse communities where subpopulations are intertwined by metabolic links or other types of interactions [227], potentially important for ecosystem functioning. The resilience and robustness of microbial consortia to external perturbations has been attributed either to microbial diversity, which enables tolerant members to fill functional voids left by intolerant species [228], or to functional complementarity of network members, resulting in better exploitation of resources and elevated resistance to stress [229]. Ammonia-oxidizing microorganisms and NOB constitute one of the most well studied functional microbial networks whereby the nitrite produced by AOM through ammonia oxidation is used as a substrate by NOB to produce nitrate [83]. Recent studies suggested that the interactions between AOM and NOB could be even more complex where the latter hydrolyze urea to NH₃, which is then taken up by urease-negative AOB that, in turn, produce the NO₂ that is consumed by NOB [230]. These functional microbial networks operating within nitrogen cycling could be the focus of future research on the effects of pesticides. This is possible due to (i) our deep knowledge of the physiology, diversity and functioning of the microbial groups participating in these networks, (ii) the availability of relevant monitoring methods and (iii) the tight association of microbial groups participating in the different steps of the process. Studies exploring the potential resistance and resilience of such microbial networks to pesticide exposure will provide an ecosystem level look at the impact of pesticides on soil microbial functioning.

Microorganisms also interact with other organisms within the soil-food web. Predator–prey relations are particularly important for soil ecosystem functioning with protists-bacteria being the best studied model [231]. Predation by protists influences bacterial diversity and productivity with consequences on the flux of organic nutrients into biomass at higher trophic levels [232]. The diversity of both protists and bacteria interactively determines the performance of the predator [233]. External perturbations (i.e. pesticides) could affect diversity at both trophic levels with possible effects on ecosystem functioning [234], an aspect overlooked so far in pesticide research. Considering the power and the resolution of amplicon sequencing tools, we posit that soil studies should study bacterial, fungal and protistan communities and identify, via appropriate co-occurrence network analysis, pesticide-driven interactions between protists and bacteria/fungi.

Synthetic microbial communities are currently at the forefront of microbial ecology and biotechnology with applications in the biosynthesis of molecules of industrial interest or in the biodegradation of pollutants [235]. Recently, Karkaria et al. [236] proposed the use of computational methods for building stable synthetic microbial communities. The concept of synthetic microbial communities could be used to study the effects of pesticides on nitrifying microbial networks composed of ureolytic bacteria, AOB, AOA and NOB [237]. Multitrophic synthetic microbial communities composed of microorganisms from different trophic levels exhibiting predator–prey interdependencies could be also used in soil microbial ecotoxicology to verify pesticide-driven effects on protists and their prey (bacteria or fungi) (Fig. 4b). This approach could be utilized as a refinement option when Tier I in vitro tests suggest unacceptable risks for the soil microbiota.

5.3 Toxicity of pesticides on the soil interactome (plant microbiome and gut microbiome of terrestrial organisms)

Beyond soil, pesticides could directly or indirectly affect the composition of the microbial communities colonizing internal (endophytic) or external (epiphytic like in phyllosphere or carposphere) plant tissues composing all together the plant microbiome whose importance for agricultural production and plant communities is immense [238]. Recent evidence suggests that pesticide effects go beyond plants in agricultural systems and reach the symbiome of insects and earthworms with reciprocal implications for ecosystem functioning [239].

Pesticide applied directly on plant foliage could directly affect epiphytic microbial communities. On the other hand, pesticides applied via soil drenching, besides affecting the seed and rhizosphere microbiome, could be taken up by plants and could impose strong effects on the endophytic microbiome. Schaeffer et al. [240] showed, using amplicon sequencing, that fungicides could impose detrimental effects on yeasts inhabiting the nectar of almond trees like Metschnikowia which are important players in pollination. Katsoula et al. [58] showed that repeated applications of iprodione on the foliage and in the rhizosphere of pepper plants induced significant alterations in the epiphytic and soil microbiota and led to an adaptation of both microbial communities, rhizospheric and epiphytic, to enhanced biodegradation of iprodione. In contrast, other studies have suggested limited effects of pesticides on the epiphytic microbial communities [241, 242].

The effects of pesticides extend to the symbiome of earthworms and arthropods exposed to pesticides through soil or through feeding on plants. Recent studies investigated the effects of nano-pesticides on the gut microbiome of earthworms and showed contrasting results [175, 243]. Interestingly Swart et al. [175] reported that CuO nanoparticles negatively affected bacteria considered as common residents of the gut of earthworms like Candidatus Lumbricincola, which are expected to have serious implications for host vigor. On the other hand, the regular exposure of insects to pesticides like atrazine (6-chloro-N²-ethyl-N⁴-(propan-2-yl)-1,3,5-triazine-2,4-diamine), trichlorfon (dimethyl rac-2,2,2-trichloro-1-hydroxyethyl phosphonate), or fenitrothion (O,O-dimethyl O-(4-nitro-3-methylphenyl) phosphorothioate) could select in the insect symbiome for microorganisms that carry catabolic genes like atz which encode for enzyme catalyzing the degradation and detoxification of atrazine [244], [245], [246].

The soil microbiome is the main reservoir for plant and insect microbiomes [247]. Considering the effect of pesticides on the soil microbiota and the direct and indirect exposure of plants and insects to pesticides, we posit that effects of pesticides might be better examined at the ecosystem level where the microbiome of all biotic interactors will be studied in parallel.

5.4 Interactive effects of pesticides with other pollutants

Pesticides often encounter other organic pollutants in agricultural soils like veterinary drugs (antibiotics and anthelmintics), which end up in agricultural soil mainly through the application of manure from livestock units [248, 249]. Kurenbach et al. [250] first showed that treating bacterial pathogenic strains with herbicides like glyphosate ([(phosphonomethyl)amino]acetic acid) and dicamba (3,6-dichloro-2-methoxybenzoic acid) could induce a multiple antibiotic resistance phenotype associated with increasing upregulation of efflux pumps. This was further verified in soil where the application of glyphosate, glufosinate (ammonium rac-2-amino-4-[hydroxy(methylphosphonoyl)] butanoate), and dicamba increase the prevalence of antibiotic resistance genes and mobile genetic elements in soil microbiomes [251]. These findings suggest that the application of herbicides in agricultural soils might exert a selection pressure favoring the co-selection of antibiotic resistance thereby contributing to the global antimicrobial resistance problem. However, the full extent of this phenomenon at microbiome level is not known (i.e. this is common among all pesticide groups), while the combined effects of veterinary drugs and pesticides, beyond antibiotic resistance gene dispersal, should be investigated in an ecotoxicological context.

Another group of emerging pollutants in agricultural soils are microplastics. They are plastic fragments of size smaller than 5 μm formed by the disintegration of plastic fragments released in soils and their levels in certain agricultural soils could reach up to 6.7 % [252]. A few studies in recent years have highlighted the potential adverse effects of microplastics on the soil microbiota [253, 254]. Microplastics are known to interact with pesticides in agricultural soils offering sites for their adsorption [255, 256]. Although this might suggest that microplastics might act as sinks for pesticides limiting their bioavailability and hence reducing their adverse effects on the soil microbiota, we still lack data to support this hypothesis.

Another aspect which was only recently acknowledged is the ecological role of plastisphere, the interface between soil and plastic surfaces. The composition of the microbial communities colonizing the plastisphere is only just starting to be explored compared to aquatic ecosystems [257]. Recently, Zhu et al. [258] showed that the plastisphere selects for microbial communities that are involved in diverse metabolic pathways including an enrichment in antibiotic resistance genes and bacterial pathogens compared to the surrounding soil. In this framework, the adsorption of pesticides on microplastic surfaces could favor the selection of microorganisms with tolerance or increased biodegradation capacities against pesticides. Studies trying to explore these interactions will strongly promote our knowledge on the interactive effects of pesticides with microplastics on the soil microbiota and will explore microbial evolutionary aspects that have not yet been investigated before.

5.5 Beyond amplicon sequencing: looking at pesticide effects using shotgun meta-omic approaches

Amplicon sequencing analysis could provide insights into the phylogenetic identity of the soil microbial responders to pesticides, but it falls short in providing solid functional assignments to the members of soil microbial communities. Such information could be derived through the application of shotgun meta-genomic and meta-transcriptomic approaches. To date the application of these meta-omic tools in soil microbial ecotoxicology is rare. Wu et al. [259] used a combination of amplicon sequencing and shotgun meta-genomics to look at the effects of thiamethoxam ((EZ)-{3-[(2-chloro-1,3-thiazol-5-yl)methyl]-5-methyl-1,3,5-oxadiazinan-4-ylidene}nitramide) on the soil microbiota. Beyond effects on bacterial diversity and abundance which were temporal and recovery was observed, they noted negative effects of thiamethoxam on (i) the relative abundance of nitrogen cycling genes (ii) Carbon metabolism genes involved in glycolysis, and Krebs cycle (iii) genes involved in signal regulating pathways like ABC transporters. The use of shotgun meta-genomic and meta-transcriptomic analysis in soil microbial ecotoxicology would go beyond descriptive phylogenetic information on pesticide microbial responders and look at effects at the metabolic potential and functional level while at the same time acquiring strong phylogenetic signals.

Abbreviations

AMPA

2-Amino-3-(5-methyl-3-oxo-2,3-dihydro-1,2-oxazol-4-yl)propanoic acid

AOA

Ammonia-oxidizing archaea

AOB

Ammonia-oxidizing bacteria

AOM

Ammonia-oxidizing microorganisms

ASV

Amplicon Sequence Variant

AMF

Arbuscular mycorrhizal fungi

CA

Concentration addition

DGGE

Denaturating Gradient Gel Electrophoresis

dsRNA

Double Stranded RNA

EC₅₀

Effective Concentration 50 %

ERA

Environmental Risk Assessment

EFSA

European Food Safety Authority

EU

European Union

HC5

Hazard Concentration 5 %

IA

Independent action

IC₅₀

Inhibitory Concentration 50 %

ISO

International Organization for Standardization

LC₁₀

Lethal Concentration 10 %

LC₅₀

Lethal Concentration 50 %

miRNA

MicroRNA

NOB

Nitrite-oxidizing bacteria

OECD

Organization for Economic Co-operation and Development

OTU

Operational Taxonomic Unit

PLFAs

Phospholipid Fatty Acids Analysis

PRCs

Principal Response Curves

q-PCR

quantitative Polymerase Chain Reaction

QSAR

quantitative structure-activity relationship

RT-q-PCR

Reverse Transcription q-PCR

siRNA

Small Interfering RNA

SSDs

Species Sensitivity Distribution

TRFLP

Terminal Restriction Fragment Length Polymorphism

TPs

Transformation Products