The effects of glyphosate, glufosinate, paraquat and paraquat-diquat on soil microbial activity and bacterial, archaeal and nematode diversity

In this study, we investigated the effects of one-off applications of glyphosate, glufosinate, paraquat, and paraquat-diquat on soil microbial diversity and function. All herbicides were added to soil as pure compounds at recommended dose and were incubated under laboratory conditions for 60 days. High-throughput phylogenetic marker gene sequencing revealed that none of the herbicides significantly influenced the richness, evenness and composition of bacterial and archaeal communities. Likewise, the diversity, composition and size of nematode communities were not significantly influenced by any of the herbicides. From a functional perspective, herbicides did not significantly affect fluorescein diacetate hydrolysis (FDA) and beta-glucosidase activities. Furthermore, the ability of soil organisms to utilise 15 substrates was generally unaffected by herbicide application. The only exception to this was a temporary impairment in the ability of soil organisms to utilise three organic acids and an amino acid. Given the global and frequent use of these herbicides, it is important that future studies evaluate their potential impacts on microbial communities in a wider-range of soils and environmental conditions.

Upper limit of recommended application rate for each commercial herbicide (L or kg/ha) Concentration of herbicide active applied to soil (ppm)

Results and Discussion
Bacterial and archaeal diversity. Soil bacterial communities were dominated by members of the Proteobacteria, Firmicutes and Bacteroidetes, while archaeal communities were dominated by representatives of the Chrenarchaeota (Fig. 1). Members of the Acidobacteria, Actinobacteria and Nitrospirae also constituted a significant fraction of the community in each soil (Fig. 1).
Relative to the controls, none of the herbicides led to significant changes in the richness (observed OTUs and Chao1 richness estimates) or evenness (Simpson's Diversity Index) throughout the course of the experiment (Fig. 2). These findings are consistent with previous phylogenetic marker gene sequencing studies that found negligible effects on the alpha diversity of rhizosphere bacterial communities in response to: 1) glyphosate, applied at recommended dose in the form of Roundup PowerMax 29 or Roundup Plus 10,12 , and 2) glufosinate, applied at recommended dose in the form of Basta 30 . We are not aware of any studies that used phylogenetic marker gene sequencing to characterise the alpha diversity of soil microbial communities in response to paraquat and paraquat-diquat, or their commercial formulations. In addition, our study is the first, to our knowledge, to investigate the effects of glyphosate, glufosinate, paraquat, and paraquat-diquat on the alpha diversity of soil microbial communities as pure compounds.
Our results indicate that, relative to the controls, the composition of bacterial and archaeal communities was not significantly affected by any of the herbicides. Similarly, Newman et al. 29 did not observe significant changes in the relative abundances of Proteobacteria (P = 0.096) and Acidobacteria (P = 0.445) in response to glyphosate applied as Roundup PowerMax at recommended dose. Likewise, using 16S rDNA pyrosequencing, Barriuso et al. 10 did not detect a significant change in community composition in response to glyphosate applied at recommended dose in the form of Roundup Plus. A range of other studies, using microbial fatty acid analysis (e.g. PLFA) or DNA fingerprinting (e.g. DGGE and T-RFLP), demonstrate that when applied at or below the recommended field-rate, the effects of pure glyphosate on microbial community structure are negligible 14 , while those of Roundup range from negligible 5-10 to minor [11][12][13] .
Information concerning the effects of glufosinate on microbial community composition is limited. As seen in our study, recommended doses of glufosinate in the form of Basta and Liberty have been shown to have negligible effects on microbial community composition as represented by PLFA 18 and DNA fingerprinting 19 , respectively. For paraquat and paraquat-diquat, information concerning their effects on soil microbial community composition is even more limited. In fact, we are not aware of any studies that investigated their effects on microbial community composition using culture independent methods. Nematode community structure. Nematode communities were dominated by representatives of the families: Monochidae, Pangrolaimidae, Tripilidae, Tylenchidae, Cephalobidae and Dorylaimidae (Fig. 3). Relative to the controls, none of the herbicides led to a significant change in the total numbers of nematodes, or to the diversity (Simpson's Diversity Index) and composition of nematode communities. Similarly, relative to the control, none of the herbicides led to significant changes in nematode trophic groups or guilds. Our findings are consistent with previous studies that found negligible effects of glyphosate 38 , glufosinate as Basta 39 , paraquat 40 , and paraquat-diquat as Sprayseed 250 47 on nematode community composition, at recommended application rates.
Microbial activity. From a functional perspective, our results indicate that: 1) none of the herbicides influenced total microbial enzyme activity (FDA) or beta-glucosidase activity; and that 2) while herbicide addition temporarily impaired the ability of soil organisms to utilise three organic acids and an amino acid (Fig. 4), their effects were otherwise negligible.
Our findings are in agreement with previous studies that found no changes in FDA 11 and beta-glucosidase 48 activity, as well as substrate utilisation patterns 5,14 in response to recommended application rates of glyphosate. In contrast, Araújo et al. 49 and Panettieri et al. 50 found that recommended rates of glyphosate and GLYFOS Ultra (36% glyphosate) stimulated FDA and beta-glucosidase activity, respectively. This suggests that the effects of glyphosate on soil microbial activity can differ between soils.
For glufosinate, our findings are consistent with those of Zablotowicz et al. 51 , who observed no effects of one, two and four-times the recommended dose of glufosinate (i.e. 10, 20 and 40 ppm) on soil FDA activity, after two,  four and seven days exposure. After one day, however, they observed a slight reduction in FDA activity at recommended dose and larger reductions at higher doses 51 . We are not aware of other studies that have characterised FDA and beta-glucosidase activities in response to paraquat and paraquat-diquat application. Likewise, we are not aware of previous studies that have investigated the effects of glufosinate, paraquat, and paraquat-diquat on the induced respiratory responses of soil microbial communities to multiple substrates.
The herbicide treatments significantly influenced community level physiological profiles on Day 7 (P = 0.03). This shift in response to herbicide addition was related to a reduction in the ability of soil organisms to utilise at least two organic acids (Fig. 4). Organic acids released by roots have been shown to play an important role in the ability of plant growth-promoting rhizobacteria to colonise banana roots 52 . Hence, it is possible that glyphosate, glufosinate, paraquat, and paraquat-diquat may reduce colonisation of roots by beneficial soil organisms. Nonetheless, our results indicate that organic acid utilisation was impaired only temporarily, hence limiting the likelihood of potential alterations in rhizosphere recruitment.

Conclusions.
In response to recommended application rates of glyphosate, glufosinate, paraquat, and paraquat-diquat as pure compounds we observed no significant effects on: 1) the richness, evenness and composition of soil bacterial and archaeal communities; 2) the composition of nematode communities; 3) soil FDA and beta-glucosidase activities; and 4) the induced respiratory responses of soil microbial communities to multiple substrates, except for three organic acids and an amino acid at Day 7 only. Within the context of our short-term, laboratory-based experiment, these findings indicate that single applications of these herbicide actives at recommended dose pose little threat to soil biodiversity and function. In practice, however, herbicides are applied multiple times per year on farms with different soils and environmental conditions, and commercial formulations contain a range of additional compounds that are not disclosed. Indeed the soil used in this study was collected from a banana plantation, where the herbicides tested in this study are in regular use. It is possible, therefore, that herbicide-intolerant organisms were already depleted. For this reason, despite being largely consistent with previous studies, our results must be interpreted with caution. Further work is essential to more fully understand the potential impacts of herbicides on soil microbial communities.

Materials and Methods
Experimental design and soil sampling. Soil was collected using a 5 cm diameter corer to a depth of 10 cm from a banana (Musa acuminata AAA var. Williams) paddock in East Palmerston, Queensland, Australia (S17° 35′32″ E 145° 49′ 58″), then transported to the Centre for Wet Tropics Agriculture, South Johnstone, Australia for further processing. The site is typical of the hydrological catchments that drain to the Great Barrier Reef lagoon. The soil had a clay loam texture (38% sand, 30% silt, 33% clay) and a pH in water of 6.7. The soil was passed through a 4 mm stainless steel sieve, adjusted to 50% water holding capacity and then 1.6 kg of soil was loaded into each of 44, 2 L plastic containers with lids that allowed gas exchange. These minicosms were incubated at 27 °C for 14 days prior to the application of treatments to allow the soil to equilibrate following physical disturbance. To maintain soils at 50% water holding capacity, the containers were weighed on a weekly basis, to calculate evaporative losses, which were corrected for by addition of water.
A total of five treatments were applied (four herbicides and a control). Each treatment was replicated four times and was arranged in a randomised block design. Our treatments included glyphosate, glufosinate, paraquat and an equimolar concentration of paraquat:diquat, which were solubilised in 10 ml water. For the control we included a 10 ml water-only treatment. Each herbicide active was applied at a dose corresponding to the upper recommended rates for the commercial herbicides that contain them (Table 1). To achieve this, the quantity of active compound that would be applied per cm 2 of soil at the upper recommended rate was first multiplied by the . Inhibitory effects of herbicides on the respiratory responses of soil microbial communities to added substrates after seven days exposure. Asterisks indicate significant differences between control and herbicide treated soils (P < 0.10 • , P < 0.05*, P < 0.05**, Tukey's HSD). Error bars represent standard errors of the means (n = 4). exposed surface area of soil in each container (408 cm 2 ). This number was then multiplied by four, as the depth of soil in each container was 4 cm and the bulk density of the soil was c. 1 g/cm 3 . All treatments were applied to the surface of the soil using a fine mist sprayer. Soils were incubated at 27 °C for 60 days.
For Microresp and characterisation of bacterial and archaeal diversity sub-samples of soil were collected from each container after 1, 3, 7, 14, 30 and 60 days incubation as soil cores using sterile 50 ml plastic tubes. Samples for DNA extraction were transferred to −20 °C storage and those for Microresp were maintained at 4 °C for 2-4 days until further processing. For nematodes, 200 g of fresh soil was collected from each container after 3, 30 and 60 days and extraction was performed immediately. Samples for enzyme assays were collected after 3, 7, 14 and 30 days and were maintained at 4 °C for 2-4 days until further processing. Processing of sequence data. Primer sequences were removed from each fastq file using the QIIME 54 v1.9.1 script multiple_extract_barcodes.py. The header line of each sequence was then modified to contain a sample ID using a custom bash script and each file was quality filtered using the QIIME script multiple_split_libraries.py with the homopolymer filter deactivated. The forward reads from each sample were then concatenated into a single file and checked for chimeras against the October 2013 release of the GreenGenes database 55 using UCHIME ver. 3.0.617 56 . Sequences were clustered at 97% similarity using UCLUST v. 1.2.22 57 . GreenGenes (October 2013 release) taxonomy was then assigned to the representative OTU sequences using BLAST+ v. 2.2.30. Data were rarefied to 350 sequences per sample for all comparisons of diversity. The mean numbers of observed OTUs, the estimated total OTUs (Chao 1) as well as Simpson's Diversity Index values were calculated using QIIME.

Characterisation of bacterial
Characterisation of nematode communities. Nematodes were extracted using a modified Baermann funnel technique 58 , in which mesh baskets, each containing a single sheet of tissue paper and 200 g fresh soil, were placed in metal trays with 250 ml of deionised water. After 48 hours the soil was discarded and the solution was passed through a 25 µm sieve and backwashed twice with 10 ml of deionised water to collect the nematodes. From each 20 ml soil extract a 1 ml sub-sample was used to determine the total nematode abundance at low magnification, and then at higher magnification, 100 nematodes were identified to family-level 59,60 . Nematode abundances were expressed as numbers per 100 g of fresh soil. Nematodes were also categorized into feeding groups (i.e. fungivores, bacterivores, plant parasites, predators and ominvores) and functional guilds, which delineate feeding groups on the colonizer-persister (cp) scale according to their r and K characteristics (where cp = 1-5 and larger numbers represent more K dominated taxa) 35,61 . The groupings were as follows: Tylenchidae (fungivore, Fu2), Aphelenchidae (fungivore, Fu2), Rhabditidae (bacterivore, Ba1), Pangrolaimidae (bacterivore, Ba1), Cephalobidae (bacterivore, Ba2), Acrobelinae (bacterivore), Monhysteridae (bacterivore, Ba3), Tripilidae Enzyme assays. Fluorescein diacetate (FDA) hydrolysis assays were used to provide a measure of total microbial enzyme activity and were performed using a modified version of the method initially proposed by Schnürer and Rosswall 44 . Briefly, 2 ml deionized water was added to 5 g fresh soil in 50 mL centrifuge tubes and incubated at 27 °C for seven days. Post-incubation, soils were shaken for 30 min in 20 ml 60 mM potassium phosphate buffer and 200 μl fluorescein diacetate solution (2000 μg/ml). Reactions were terminated by addition of 20 ml acetone.
Beta-glucosidase assays were performed as a measure of organic matter degradation potential using a modified method of Eivazi and Tabatabai 45  L-arginine), carbohydrates (L-arabinose, D-fructose, D-galactose, D-glucose) and a phenolic compound (protocatechuic acid ethyl ester). Sterile distilled water was added to controls. Statistical analyses. Univariate responses were analysed using linear mixed-effect models that were implemented using the R packages lme4 62 and lmerTest 63 . These variables included: 1) the observed (Sobs) numbers of bacterial and archaeal OTUs, 2) the predicted (Chao1) numbers of bacterial and archaeal OTUs, 3) bacterial and archaeal OTU-level diversity as represented by the Simpson's Diversity Index, 4) total nematode abundance, 5) nematode family-level diversity as represented by the Simpson's Diversity Index, 6) the abundances of nematode trophic groups and guilds, 7) FDA activity, and 8) beta-glucosidase activity. For these analyses, measurements of the response variable y itc (for treatment i (i.e. herbicide), day t (days since treatment application), and container c) were assumed to follow the model: The terms α i , β t and γ it are fixed effects for treatment i, day t and their interaction, respectively. The random effect a c accounts for the container of soil indexed by c from which repeated measurements were taken, which we assumed was independent (between containers) and normal with mean zero and variance σ c 2 . The residuals were also assumed independent and normal with variance σ 2 . The significance of treatment effects was assessed using F-tests.
Multivariate responses were analysed using permutational multivariate analysis of variance (PERMANOVA 64 ) implemented using the R package vegan 65 . Multivariate analyses focussed on determining whether the treatments influenced: 1) the composition of bacterial and archaeal communities as represented by Hellinger transformed (square root of the relative abundances) OTU relative abundances 66 , 2) the composition of nematode communities as represented by the Hellinger transformed abundances of 14 nematode families, and 3) the substrate utilisation potential of microbial communities as represented by their induced respiratory responses to 15 substrates. All analyses were performed using R 3.2.3.