AgNP 50–80 nm nanoplates (stored as 450 mg/L Ag dispersions in polyvinylpyrrolidone) were obtained from Moscow State University (Moscow, Russia). A dispersion of AgNP 15 nm nanospheres [particle size 15 nm, 10.16% Ag (w/w) in a stabilizing vehicle material consisting of polyoxyethylene glycerol trioleate (4%; w/w) and polyoxyethylene (20) sorbitan mono-laurate (Tween 20; 4%; w/w; manufacturer’s information) (Van der Ploeg et al., 2014)] was obtained from a commercial source (Nanostructured & Amorphous Materials, Houston, TX, USA); AgNP 20–40 nm nanospheres (particle size range 20–40 nm, powder) were manufactured by a patented vapor condensation process from QSI-Nano^® (Santa Ana, CA, USA); AgNP 50 nm nanorods (with reported width of 50 nm and length of 5–10 μm, dispersion of 8.6% Ag (w/w) in an aqueous solution containing polyvinylpyrrolidone (<1% w/w), acrylic/acrylate copolymer (<2% w/w) and polycarboxylate ether (<2% w/w)) (Hund-Rinke and Schlich, 2014) were supplied by the Fraunhofer Institute (Schmallenberg, Germany). AgNO₃ was purchased from Sigma–Aldrich (Zwijndrecht, Netherlands).

Physio-chemical characterizations of AgNPs were done following the risk assessment of engineered nanoparticles (ENPRA) nanomaterial dispersion protocol for toxicological studies (Kermanizadeh et al., 2012). The stock suspension was sonicated for 8 min (in 4°C water, twice) at 38 ± 10 KHz, and diluted in BIS–TRIS buffer. During this dilution process the primary AgNPs dispersions were continuously inverted and vortexed to prevent the particles from sedimentation. Characterization of particle size and morphology of the four kinds of AgNPs were analyzed using transmission electron microscopy (TEM; JEOL 1010, IEOL Ltd., Japan). Size distribution of AgNPs suspensions were analyzed at 1, 24, and 96 h after incubation in the exposure medium by dynamic light scattering (DLS, Malvern, Instruments Ltd., UK). Concentrations of AgNPs suspensions (designated as AgNPs_(total) hereafter) in the exposure media were determined using Atomic Absorption Spectroscopy (AAS; Perkin Elmer 1100B) at 1, 24, 48, 72, and 96 h. The Ag-ion concentrations released from the AgNPs_(total) at 1 mg/L in test media were also measured at the same times.

Soil samples were collected from a site dominated by deciduous trees (52°07’06.7′′N 5°11′23.1′′E, Bilthoven, Netherlands). Moist soil samples were selected and sieved with an 8-mm sieve and kept at 4°C for the duration of the experiments. The soil moisture content was maintained at the initial content which was 18.4% of the dry soil weight by regularly supplying distilled water. Subsequently, 60 g soil subsamples were placed in glass jars and incubated for 120 h. Ten grams of this soil sample was added to 90 mL of 10 mM (2.09 g/L) BIS–TRIS (Sigma–Aldrich B9754). This buffer was adjusted to pH 7 using 1 M HNO₃. The buffer with soil added was then centrifuged for 10 min at 1500 rpm, after which the supernatant was filtered (to remove the soil particles) and diluted five times with the same buffer (soil microbial community extract).

A 15 mL soil extract of 1% w/v soil microbial community extract was added to the different silver suspensions treatments. Exposure concentrations were selected to range from 0.070 to 0.678 mg/L for the 50–80 nm nanoplates, 0.072 to 0.708 mg/L for the 15 nm nanospheres, from 0.108 to 0.814 mg/L for the 20–40 nm nanospheres and from 0.141 to 1.529 mg/L for the 50 nm nanorods based on range finding experiments. The effect of AgNO₃ was tested in the range from 0.003 to 0.05 mg/L as a positive reference to assess the toxicity of Ag⁺. Considering the rapid aggregation process of nanoparticles, the soil microbial community extracts were exposed to AgNPs for less than 2 h. To obtain a solution of the ion species of AgNPs, the AgNPs suspensions were centrifuged at 30,392g for 30 min (Sorvall RC5Bplus centrifuge, Fiber lite F21-8 × 50 year rotor) and filtered by a filter with 0.02 μm pore diameter. This step yields supernatants with negligible amounts of particles (Hua et al., 2014), which were used as Ag-ion suspension. AgNP_(ion) was operationally defined as the Ag measured in the supernatant. Accordingly, AgNP_(particle) is the difference between the Ag measured in the AgNP suspension and AgNP_(ion).

Community metabolic diversity was determined by evaluating the carbon substrate utilization patterns using commercial Ecoplates (Biolog, Hayward, CA, USA; Garland and Mills, 1991) containing 3 times 31 carbon sources (Supplementary Table S1). Biolog Ecoplates are comprised of ecologically relevant, structurally diverse compounds, yet do not include e.g., recalcitrant substrates nor specific substrates typical of the soils used in this study. It is therefore impossible to directly relate substrate utilization profiles to the actual functioning of the soil microbial communities. Nonetheless, the number of substrates used can serve as a proxy of the metabolic diversity of the microbial community (Garland, 1999; Krause et al., 2014). Moreover, differences in utilization profiles indicate that functionally distinct microbial communities can develop depending on treatment (e.g., Hunting et al., 2013a,b, 2015; Wang et al., 2015). Each well of the Biolog Ecoplates was inoculated with 100 μl of supernatant extracted from different AgNPs exposed soil samples as described above, and incubated for 96 h at 20°C, to enhance microbial growth. Optical density (OD) was measured at 590 nm (OD₅₉₀) immediately after inoculation of the plates and at 24, 48, 72, and 96 h to ensure that a saturation of the utilization rate in all samples was reached.

The average OD₅₉₀ of each carbon source (three replicates) was calculated for each measurement. Average well color development (AWCD) was used to represent the average microbial metabolic activity. AWCD was calculated for the 31 substrates as described by (Garland and Mills, 1991) and transformed (Weber et al., 2007).

The transformed values of AWCD of soil extracts exposed to AgNPs were used as the total effect (E_(total)). The effect of AgNO₃ was also tested in order to allow to calculate the effect of ions released from the corresponding AgNPs (E_(ion)). The response addition model (RA) was used to calculate the relative contribution to toxicity of AgNPs_(particle) and AgNPs_(ion) because different mechanistic pathways of NP_(particle) and NP_(ion) were assumed, according the assumptions of Hua et al. (2014):

Substrate utilization was chosen as the transformed OD₅₉₀ value greater than 0.25. Shannon indices (Shannon, 1948) were also calculated for each microbial community exposed to differently AgNPs as a measure of the diversity of the microbial communities active during incubation.

The exposure concentrations for each AgNPs are subject to exponential decay as they decrease along with time. This dissociation model heads downhill gradually reaching a plateau (P). The dynamics of the exposure concentrations were accounted using the time weighted average (TWA) concentrations for AgNPs_(total) and AgNPs_(particle) (1–96 h) depending on the nonlinear fit of ion release profiles:

where C₀ is the C (concentration, mg/L) value when T (time, h) is zero. P is the C value at infinite times. K is the rate constant. Statistics of the nonlinear fit of ion release profiles for each AgNPs_(total) and AgNPs_(particle) are shown in Supplementary Table S2.

The EC_x values were based on TWA concentrations for AgNPs_(total) and AgNPs_(particle) and calculated by the dose–response-inhibition model in GraphPad Prism 6.0. Statistically significant differences between different soil extracts exposed to four AgNPs were determined by a two-way ANOVA and Turkey’s honestly significant difference tests (with significance level set as p < 0.05). To relate the microbial functional composition under four AgNPs exposure levels, the 31 carbon substrates utilization were analyzed using a Euclidean-based cluster analysis and a two-way analysis of similarities (ANOSIM) in PAST 3.0. (Hammer et al., 2001; Villéger et al., 2008).

Transmission electron microscopy images reflecting the different shapes and primary sizes of the four AgNPs are shown in Supplementary Figure S2. Data on size distributions after 1, 24, 48, and 96 h are given in Table 1 It can be seen that AgNPs aggregated directly after submerging into the exposure medium. The particle sizes in the exposure medium changed from 191, 196, 187, and 288 nm after 1 h of incubation to 368, 385, 373, and 496 nm after 96 h of incubation, respectively. The 96 h ion release profiles of AgNPs are shown in Figure 1 It can be seen that the percentage of AgNP_(ion) in the AgNPs suspension at 1 mg shifted from 10, 8, 12, and 4% after 1 h to 13, 10, 14, and 6% after 96 h of incubation in the buffer, respectively.

The dose–response curves of AgNPs_(total) and AgNO₃ based on the TWA concentrations are provided in Figure 2A As described above, toxicity was expressed as the values of AWCD. It can be seen that all exposures of the microbial community to the nanoparticle suspensions induced significant toxicity. The 50–80 nm nanoplates_(total), 15 nm nanospheres_(total) and 20–40 nm nanospheres_(total) showed much lower 96 h EC₅₀ values compared to that of 50 nm nanorods_(total). The EC₅₀ value derived from the AgNO₃ curve was 0.064 mg/L, which implied that the Ag⁺ was the most toxic among all the different silver species tested.

The dose–response curves of AgNPs_(particle) based on the TWA concentrations are provided in Figure 2B These curves could be derived after using the RA model. The 50–80 nm nanoplates_(particle), 15 nm nanospheres_(particle) and 20–40 nm nanospheres_(particle) showed much lower 96 h EC₅₀ values of 0.201 mg/L, 0.212 mg/L and 0.222 mg/L, respectively, compared to the 50 nm nanorods_(particle) (0.342 mg/L), which indicated that the 50–80 nm nanoplates, 15 nm nanospheres and 20–40 nm nanospheres were much more toxic to soil microbes, compared to the 50 nm nanorods.

The dose–response curves based on initial concentrations of the differently sized and shaped AgNPs on the catabolic capabilities of a microbial community are shown in Supplementary Figure S3. The EC₅₀ values of AgNPs_(total) and AgNPs_(particle) using TWA versus initial concentrations are summarized in Table 2

The relative contributions of AgNPs_(particle) and AgNPs_(ion) to the overall toxicity of AgNPs to a microbial community at different effect levels is given in Table 3 At the EC₂₀ level, the contributions of 50–80 nm nanoplates_(particle), 15 nm nanospheres_(particle), 20–40 nm nanospheres_(particle), and 50 nm nanorods_(particle) to the overall toxicity were 32–43%. The contribution to toxicity of the particles increased with increasing exposure concentrations. The 50–80 nm nanoplates_(particle), 15 nm nanospheres_(particle), 20–40 nm nanospheres_(particle), and 50 nm nanorods_(particle) accounted for about 60–68% of the relative contribution to toxicity at the EC₉₀ level of the suspensions.

Toxic effects of AgNPs to the soil microbial community were mainly reflected in a change of the community metabolic diversity. Thus, the functional diversity of microbial communities under the four different AgNPs treatments needed to be further explored. Figure 3 presents the Shannon index of microbial communities under the four different AgNPs treatments at the EC₂₀, EC₆₀, and EC₉₀ levels. The values of Shannon index decreased with increasing exposure concentrations. No significant difference was observed between the 20–40 nm and 15 nm nanospheres exposure. In contrast, 50–80 nm nanoplates and 50 nm nanorods exposure significantly reduced the community metabolic diversity (p < 0.05).

Patterns in carbon substrate utilization reflected differences in the functional diversity of the microbial communities among different AgNPs treatments (Supplementary Figure S2 and Figure 4). The microbial communities in the control soil extracts were in general able to metabolize 20 of the 31 carbon sources, and the utilization of most of the substrates decreased with the increasing AgNPs concentrations (Supplementary Figure S2). At the EC₂₀ level, the utilization of Glycogen was inhibited in all AgNPs exposed microbial communities, however, the ability of microbial communities to utilize Glycyl-L-glutamic acid was stimulated by the AgNPs exposure. At the EC₆₀ level, the similar pattern of carbon utilization was only obtained in microbial communities under treatment of 15 and 20–40 nm nanoshperes, which suggested that microbial efficiency in exposure of spherical AgNPs was similar. At the EC₉₀ level, exposure to AgNPs significantly inhibited utilization of all substrates (p < 0.05).

Figure 4 shows the cluster analysis of the carbon source metabolisms of the microbial communities under the four AgNPs treatments at the EC₂₀, EC₆₀, and EC₉₀ levels. Microbial metabolic diversity differed significantly between treatments depending on both EC levels and shapes (two-way ANOSIM: EC level R = 0.953, p < 0.05; shape R = 0.876, p < 0.05), showing that the developed microbial communities were functionally distinct.