Identification of Soil Bacteria Susceptible to TiO₂ and ZnO Nanoparticles

Abstract

Because soil is expected to be a major sink for engineered nanoparticles (ENPs) released to the environment, the effects of ENPs on soil processes and the organisms that carry them out should be understood. DNA-based fingerprinting analyses have shown that ENPs alter soil bacterial communities, but specific taxon changes remain unknown. We used bar-coded pyrosequencing to explore the responses of diverse bacterial taxa to two widely used ENPs, nano-TiO₂ and nano-ZnO, at various doses (0, 0.5, 1.0, and 2.0 mg g⁻¹ soil for TiO₂; 0.05, 0.1, and 0.5 mg g⁻¹ soil for ZnO) in incubated soil microcosms. These ENPs significantly altered the bacterial communities in a dose-dependent manner, with some taxa increasing as a proportion of the community, but more taxa decreasing, indicating that effects mostly reduced diversity. Some of the declining taxa are known to be associated with nitrogen fixation (Rhizobiales, Bradyrhizobiaceae, and Bradyrhizobium) and methane oxidation (Methylobacteriaceae), while some positively impacted taxa are known to be associated with the decomposition of recalcitrant organic pollutants (Sphingomonadaceae) and biopolymers including protein (Streptomycetaceae and Streptomyces), indicating potential consequences to ecosystem-scale processes. The latter was suggested by a positive correlation between protease activity and the relative abundance of Streptomycetaceae (R = 0.49, P = 0.000) and Streptomyces (R = 0.47, P = 0.000). Our results demonstrate that some metal oxide nanoparticles could affect soil bacterial communities and associated processes through effects on susceptible, narrow-function bacterial taxa.

INTRODUCTION

Engineered nanoparticles (ENPs) are finding increasing industrial application (4, 40) and will inevitably enter natural ecosystems, with soils predicted to be a substantial sink (22). The potential of ENPs to be toxic in these environments has been suggested by studies evaluating impacts on bacterial cultures (42, 44), protozoa grazing on ENP-exposed bacteria (52), and hydroponically grown plants (38); thus, ENPs may affect ecosystems through both population- and community-level effects. However, when ENPs are released into natural media, they can be transformed by biotic (29) and abiotic (33) processes in ways that could change their effects on natural communities from what is observed in single populations. Understanding the effects of ENPs in the environment requires studying natural materials with intact chemical, physiological, and community factors that can mediate the effects. However, studies of ENP impacts on soil bacteria and the processes they carry out remain few and inconclusive, with responses ranging from none (50), to subtle (31), to substantial (21, 35).

Two ENP types are used in such large amounts that they are particularly likely to accumulate in the environment: nanoparticulate titanium dioxide (nano-TiO₂) and zinc oxide (nano-ZnO). These ENPs are used in applications as diverse as sunscreens, cosmetics, and coatings, and therefore are increasingly being introduced into the environment (22, 46). Both of these ENPs can alter the composition of the soil bacterial community, and the effects increase with dose (21). However, prior work did not reveal which taxa were susceptible. We hypothesized that community shifts would be dominated by changes in a few taxa that respond strongly, with most taxa showing only modest changes. We used a pyrosequencing approach to identify susceptible taxa. The number of bacterial taxa with decreased relative abundance was greater than the number of taxa that increased. Among the taxa with decreased relative abundance were bacteria potentially involved in N₂ fixation and methane oxidation, while among those that increased were bacteria potentially involved in refractory organic compound mineralization.

MATERIALS AND METHODS

Experimental design.

The microcosm design was described in detail in Ge et al. (21). Annual grassland soil (0 to 10 cm) was collected from the University of California Sedgwick Reserve (34°40′32″N, 120°2′27″W) which has a Mediterranean climate with hot, dry summers and cool, wet winters. The soil is a weakly acidic Pachic Argiustoll (pH 6.0); surface soil has a loam texture and contains 2.2% C and 0.21% N (54). Ten 2-kg soil samples were randomly collected and mixed to form one composite sample. Soil was placed in a sterile plastic bag, sealed, and transported to the laboratory on ice. Visible rocks, roots, and fresh litter were removed, and the soil was sieved to 2 mm and then stored at 4°C. The nano-TiO₂ was semispherical, 15- to 20-nm diameter, and reported as 81% anatase and 19% rutile (Evonik Degussa, Parsippany, NJ); the nano-ZnO was spheroid, 20- to 30-nm diameter, and was 100% zincite (Meliorum, Rochester, NY). Detailed characterizations of the nano-TiO₂ and nano-ZnO were published previously (33). Three sets each of nano-TiO₂ stock dispersions (5, 10, and 20 mg ml⁻¹) and of nano-ZnO stock dispersions (0.5, 1.0, and 5.0 mg ml⁻¹) were produced by the addition of nanoparticles to Nanopure water, with vigorous stirring for 10 min, followed by sonication in a Bransonic model 2510 bath (output power, 100 W; frequency, 40 kHz) (Branson, Danbury, CT) for 30 min.

Each microcosm consisted of 74 g of soil in a 250-ml sterile plastic bottle (Fisher, Pittsburgh, PA) (21). The exposure doses of nano-TiO₂ (0.5, 1.0, and 2.0 mg g⁻¹ soil) and nano-ZnO (0.05, 0.1, and 0.5 mg g⁻¹ soil) were chosen based on pure culture studies to represent three scenarios: low, medium, and high concentrations (1, 5, 28, 55). Microcosms without ENPs were used as controls. The soil water content was adjusted to 0.18 (ca. 22.5% of water-holding capacity) to maximize microbial activity. The nano-TiO₂ and nano-ZnO stock dispersions were added to each microcosm using a previously described method (21) to achieve the target exposure doses and soil water content. Microcosms were sealed with plastic caps, incubated at 20°C in the dark, and vented by opening the caps in a fume hood for around 40 min every 3 to 4 days. Soil water content was maintained by weighing the bottles at 2-week intervals and adding sterile Nanopure water to replace any lost water. A total of 56 microcosms were prepared to achieve four replicates per exposure dose (control, three doses of nano-TiO₂, and three doses of nano-ZnO) and sampling time (15 and 60 days). Before microcosm construction, four additional soil samples were stored at −80°C for characterization of the baseline soil conditions.

Soil DNA extraction, quantification, and PCR.

Soil DNA was extracted from 0.3 g of soil using a Powersoil DNA isolation kit (MoBio, Carlsbad, CA) according to the manufacturer's instructions. DNA was quantified using a Quant-iT DNA assay kit, high sensitivity (Invitrogen, Eugene, OR) which is a fluorescence-based method showing high sensitivity and selectivity for double-stranded DNA (dsDNA) (21). Fluorescence intensity was measured using a Synergy multi-mode microplate reader (Biotek, Winooski, VT) with excitation and emission wavelengths of 485 and 528 nm, respectively.

Genes encoding 16S rRNA were PCR-amplified using the extracted DNA as the template and a primer set well suited for bar-coded pyrosequencing analysis (47). The forward primer contained the 454 primer B (CTA TGC GCC TTG CCA GCC CGC TCA G), a 2-base linker (TC), and the bacterial primer 27F (AGA GTT TGA TCC TGG CTC AG); the reverse primer contained the 454 primer A (CGT ATC GCC TCC CTC GCG CCA TCA G), a unique 12-base error-correcting bar code, a 2-base linker (CA), and the bacterial primer 338R (TGC TGC CTC CCG TAG GAG T) (18, 27). The unique bar codes were used to assign sequences to samples. The PCR was run on a PTC-100 thermal cycler (MJ Research, Watertown, MA) in 0.2-ml tubes; the final volume of the reaction mixtures was 25 μl, containing 1× PCR buffer, 2 mM MgCl₂, 0.2 mM each deoxynucleoside triphosphate (dNTP), 0.4 μM each primer, 2.5 U of GoTaq DNA polymerase (Promega, Madison, WI), 0.2 μg μl⁻¹ bovine serum albumin (BSA), and 10 ng of template DNA. The thermal cycling scheme was an initial denaturation at 95°C for 3 min, followed by 30 cycles of denaturation at 95°C for 30 s, annealing at 72°C for 30 s, and extension at 72°C for 60 s, with a final extension step at 72°C for 7 min. Negative controls containing all the components except DNA templates were included to test for contamination. For each sample, triplicate PCR runs were pooled to reduce random PCR bias. After size and quality verification by 1% agarose-gel electrophoresis, PCR products were purified to remove unincorporated primers and nucleotides using a QIAquick PCR purification kit (Qiagen, Valencia, CA).

Amplicon library preparation and pyrosequencing.

The bar-coding amplicon library used for pyrosequencing was prepared using a previously described process (18) but with some modifications. Briefly, the purified PCR products from each sample were quantified, equally pooled by amount, and concentrated to a final volume of 60 μl. The DNA concentration of this amplicon library was determined as 56.4 ng μl⁻¹, using a Quant-iT DNA assay kit, high sensitivity; the DNA quality was assessed based on the absorbance ratios of 260/280 nm as 1.84, using a ND-1000 spectrophotometer (Nanodrop, Wilmington, DE). Pyrosequencing was performed on a 454 Genome Sequencer FLX platform using titanium chemistry (Roche, Branford, CT).

Sequence preprocessing.

All sequences that passed the quality controls of the Genome Sequencer FLX software were further screened to remove low-quality sequences that had an average quality score of less than 25 and contained any ambiguous characters (27). Remaining sequences with the correctable bar codes were assigned to samples by examining the 12-bp unique bar codes, and the primers were trimmed using the Pyrosequencing Pipeline program in the Ribosomal Database Project (http://pyro.cme.msu.edu/init/form.spr). After initial trimming and screening, qualified sequences were processed similarly to others (11) using the QIIME program (7). Briefly, similar sequences were clustered into operational taxonomic units (OTUs) based on a cutoff dissimilarity of 0.03 using the UCLUST method (17). The number of sequences that clustered into the same OTU was counted as the abundance (also called frequency) of that specific OTU. The resulting sample-OTU matrix was used for OTU-based community analysis. A representative sequence was chosen from each OTU by selecting the longest sequence that was most similar to other sequences in the OTU. Representative sequences were aligned against a Greengenes template alignment (http://greengenes.lbl.gov/Download/Sequence_Data/Fasta_data_files/core_set_aligned.fasta.imputed) with a minimum alignment length of 150 bp and a minimum identity of 75% using the nearest alignment space termination (NAST) algorithm (14). After the aligned sequences were filtered by removing columns comprised of only gaps and positions known to be hypervariable using a lane mask (http://greengenes.lbl.gov/Download/Sequence_Data/lanemask_in_1s_and_0s), an approximately maximum-likelihood phylogenetic tree was inferred using the FastTree method (43). The resulting phylogenetic tree was used for the downstream phylogenetic-based community analysis.

Protease activity.

Protease activity was measured after 15 and 60 days to evaluate ENP effects on a specific soil microbial function that is critical to maintaining soil N supply (48). Several groups of bacteria are also known to be prodigious producers of extracellular proteases (45), and so protease has the potential to be sensitive to shifts in specific bacterial taxa, notably the family Streptomycetaceae and the genus Streptomyces (8). Protease activity was assayed using a previously described method (36). Briefly, 1 g of soil was incubated in a shaking water bath with 2.5 ml of 0.2 M Tris buffer (pH 8.1) and 2.5 ml of 20 g liter⁻¹ sodium caseinate solution at 50°C for 2 h. After incubation, the enzymatic reaction was stopped by the addition of 5 ml of 15% trichloroacetic acid (TCA), and then the reaction mixture was centrifuged at 4,000 × g for 20 min. After centrifugation, 0.5 ml of the supernatant was mixed with 0.75 ml of 1.4 M Na₂CO₃ and 0.25 ml of 3-fold diluted Folin-Ciocalteu's phenol reagent, and the mixture was incubated at room temperature for 1 h. The tyrosine concentration was determined spectrophotometrically at 700 nm. Tyrosine standards were treated in the same way as the samples.

Statistical analysis.

To test whether bacterial communities differed among treatments, the community dissimilarities were illustrated through two complementary multivariate techniques, principal coordinate analysis (PCoA) and multivariate regression tree (MRT) analysis, and were statistically tested by nonparametric multivariate analysis of variance (NP-MANOVA) with 9,999 permutations (2). PCoA is a widely used unconstrained ordination approach, while MRT analysis is a constrained classification approach that is well suited for complex ecological data sets and has been successfully used for exploring key environmental variables that influence microbial communities (19, 20). Both methods were performed for confirmation of the community shift patterns. PCoA and NP-MANOVA were conducted based on two widely used community distances, OTU-based Bray-Curtis and phylogenetically based weighted UniFrac (39), while MRT analysis was conducted based on the total sum of squares of the different OTUs (13). Both Bray-Curtis distance and weighted UniFrac distance were calculated accounting for both abundances (magnitudes of membership) and membership (OTU or branch of phylogenetic tree).

To test whether bacterial communities systematically changed with ENP dose, the community dissimilarities (Bray-Curtis and weighted UniFrac distances) were exponentially or linearly regressed against the ENP concentrations in the soils. A Mantel test (9,999 permutations) was also conducted to confirm the regression results, by examining the Spearman rank correlation between the environmental distance (Euclidean distance) of the dose and the Bray-Curtis or weighted UniFrac distance. Because pyrosequencing depends on a random sequencing strategy (57), the number of final qualified sequences varied over samples, which may cause bias when communities are directly compared at different sequencing (sampling) efforts. For example, if one sample yielded 5,000 sequence counts and another yielded 10,000, the community comparison results could be more affected by sequencing effort than underlying community dissimilarities. Therefore, to increase the reliability of community comparison among samples, we rarefied the sample-OTU matrix through a random subsampling process to make the sequence counts of all samples the same (5,741, the smallest sequence count among samples) and conducted the above-mentioned analyses using the rarefied sample-OTU matrix.

To explore individual taxon variations with metal oxide ENP exposure, the qualified sequences were first assigned to a set of hierarchical taxa (phylum, class, order, family, and genus) using the program Classifier in the Ribosomal Database Project (http://rdp.cme.msu.edu/classifier/). The number of sequences assigned to a specific taxon was counted as the abundance of that taxon, and the abundance proportion of a specific taxon in the whole community was calculated as the relative abundance of that taxon. Then, the relationship between the relative abundance of each taxon and the ENP exposure was explored through linear regression analysis; if the statistical test on the regression equation was significant (P < 0.05), either positively or negatively, at both sampling times, the taxon was identified as a sensitive soil bacterium. To minimize the effects of random sequencing, only those taxa that existed in all samples were used in regression analysis.

One-way analysis of variance (ANOVA) was performed to test the global effect of exposure dose on protease activity at each sampling time. Where the global effect was significant (P < 0.05), a Tukey test was performed to test the significance (α = 0.05) between the control and each ENP treatment. A two-tailed Pearson correlation analysis was performed to test for correlation between protease activity and the relative abundance of selected bacterial taxa.

Analyses were conducted using QIIME (http://qiime.sourceforge.net/) (7), R (http://www.r-project.org/), or SigmaPlot (Systat Software, San Jose, CA).

RESULTS

Effects of nano-TiO₂ and nano-ZnO on overall bacterial communities.

In total, 531,408 qualified sequences (average, 8,857 ± 1,283 sequences per soil sample) were obtained after initial trimming and screening. These sequences yielded 37,208 OTUs (average, 3,415 ± 352 OTUs per soil sample) at a dissimilarity of 0.03. The Chao1 method and the abundance-based coverage estimator (ACE) (30) method yielded estimates of 50,435 (95% confidence interval, 49,848 to 51,049) and 50,553 (95% confidence interval, 50,083 to 51,040) total species, respectively. The rank-abundance plot demonstrated that many rare and a few common taxa were encountered, with 27,867 OTUs (75%) detected less than five times and with 657 OTUs (2%) detected more than 100 times.

Both OTU-based and phylogenetic-based PCoA illustrated differences among bacterial communities across treatments. Communities exposed to low and medium doses of nano-TiO₂ (0.5 and 1.0 mg g⁻¹ soil) or nano-ZnO (0.05 and 0.1 mg g⁻¹ soil) overlapped the control after 15 days, while communities exposed to high doses of nano-TiO₂ (2.0 mg g⁻¹ soil) and nano-ZnO (0.5 mg g⁻¹ soil) separated from the control after both 15 and 60 days (Fig. 1). The PCoA graphs also revealed a time-dependent effect of ENP exposure; communities exposed to low and medium doses of both ENPs were more similar to the control after 15 days, while they became more similar to the high-dose treatments after 60 days (Fig. 1). The stronger effects of nano-ZnO were reflected by greater separation from the control with nano-ZnO than with nano-TiO₂ at the same exposure dose (0.5 mg g⁻¹ soil).

Fig 1.

Principal coordinates analysis (PCoA) to illustrate the shifts of soil bacterial communities exposed to nano-TiO₂ (a and b) and nano-ZnO (c and d) based on Bray-Curtis distance and weighted UniFrac distance. Con, control; TL, TM, and TH, low (0.5 mg g⁻¹ soil), medium (1.0 mg g⁻¹ soil), and high (2.0 mg g⁻¹ soil) doses of nano-TiO₂, respectively; ZL, ZM, and ZH, low (0.05 mg g⁻¹ soil), medium (0.1 mg g⁻¹ soil), and high (0.5 mg g⁻¹ soil) doses of nano-ZnO, respectively. Exposure time is indicated by the numerical suffix; e.g., Con15 represents the control at day 15.

MRT analysis revealed a nested structure, with communities splitting first by exposure time, then by dose of either nano-TiO₂ or nano-ZnO (Fig. 2). Exposure time explained 26.6% and exposure dose explained 24.7% of the community variations.

Fig 2.

Multivariate regression tree (MRT) to reveal the hierarchical environmental determinants on soil bacterial communities associated with exposure dose (0, 0.5, 1.0, and 2.0 mg g⁻¹ soil for TiO₂; 0.05, 0.1, and 0.5 mg g⁻¹ soil for ZnO) and exposure time (0, 15, and 60 days). In an MRT, each split is represented graphically as a branch that is labeled with the levels of the classification variable; bar plots show the multivariate means of OTUs at each branch; the numbers of samples included in the splits are shown under the bar plots. The heat map shows the normalized abundances of OTUs that occurred in more than 50 samples. Con, control; TL, TM, and TH, low (0.5 mg g⁻¹ soil), medium (1.0 mg g⁻¹ soil), and high (2.0 mg g⁻¹ soil) doses of nano-TiO₂; ZL, ZM, and ZH, low (0.05 mg g⁻¹ soil), medium (0.1 mg g⁻¹ soil), and high (0.5 mg g⁻¹ soil) dose of nano-ZnO. Exposure time is indicated by the numerical suffix (see the legend of Fig. 1).

The results of NP-MANOVA further confirmed the results of PCoA and MRT analysis and showed a time-dependent dose effect of nano-TiO₂ and nano-ZnO on bacterial community structure (Table 1). After 15 days of exposure, communities exposed to high doses of either nano-TiO₂ or nano-ZnO were significantly different from all other treatments; after 60-days, compared to 15 days of exposure, the community dissimilarities between high and low or medium treatments tended to decrease, while the dissimilarities between control and low or medium treatments tended to increase, as reflected by the variation of P values.

Table 1.

Significance test of dissimilarity of soil bacterial communities using NP-MANOVA

ENP and effect or test paira	Significance test of dissimilarity by exposure time and distance estimationb
	15 days		60 days
	Bray-Curtis	Weighted UniFrac	Bray-Curtis	Weighted UniFrac
Nano-TiO2
Global	0.001*	0.006*	0.000*	0.000*
Con, TL	0.367	0.252	0.031*	0.031*
Con, TM	0.169	0.317	0.027*	0.027*
Con, TH	0.029*	0.029*	0.029*	0.029*
TL, TM	0.313	0.338	0.028*	0.057
TL, TH	0.030*	0.030*	0.030*	0.030*
TM, TH	0.029*	0.029*	0.280	0.168
Nano-ZnO
Global	0.003*	0.004*	0.000*	0.002*
Con, ZL	0.319	0.433	0.061	0.086
Con, ZM	0.111	0.087	0.056	0.027*
Con, ZH	0.029*	0.029*	0.029*	0.029*
ZL, ZM	0.515	0.399	0.543	0.578
ZL, ZH	0.030*	0.030*	0.030*	0.058
ZM, ZH	0.029*	0.029*	0.029*	0.085

^aCon, control; TL, TM, and TH, low (0.5 mg g⁻¹ soil), medium (1.0 mg g⁻¹ soil), and high (2.0 mg g⁻¹ soil) doses of nano-TiO₂; ZL, ZM, and ZH, low (0.05 mg g⁻¹ soil), medium (0.1 mg g⁻¹ soil), and high (0.5 mg g⁻¹ soil) doses of nano-ZnO.

^b*, significant dissimilarities at a significance level of 0.05.

We next explored the relationship between community shift and exposure dose. The community dissimilarity estimated by both Bray-Curtis and weighted UniFrac distances increased linearly in response to increasing nano-TiO₂ and nano-ZnO after 15 days of exposure and exponentially in response to increasing both ENPs after 60 days (Fig. 3). The Mantel test of the Spearman rank correlation with 9,999 permutations also revealed a significant positive correlation between the ecological distances of bacterial communities and the environmental distance of the exposure doses at both sampling times (Table 2).

Fig 3.

Regression analysis to illustrate the systematic shifts of soil bacterial communities along dose gradients of nano-TiO₂ (a, b, e, and f) and nano-ZnO (c, d, g, and h) after 15-day and 60-day exposures. The boundary of the box closest to zero indicates the 25th percentile, a line within the box marks the median, and the boundary of the box farthest from zero indicates the 75th percentile. Error bars above and below the box indicate the 90th and 10th percentiles, respectively.

Table 2.

Mantel test of the Spearman rank correlation between the ecological distances (Bray-Curtis distance and weighted UniFrac distance) of bacterial communities and the environmental distance (Euclidean distance) of exposure doses at a significance level of 0.05.

ENP	Mantel test result by exposure time and distance estimation
	15 days				60 days
	Bray-Curtis		Weighted UniFrac		Bray-Curtis		Weighted UniFrac
	r	P	r	P	r	P	r	P
Nano-TiO2	0.52	0.000	0.43	0.001	0.52	0.000	0.45	0.002
Nano-ZnO	0.60	0.000	0.49	0.000	0.52	0.000	0.30	0.008

Responses of individual taxa to nano-TiO₂ and nano-ZnO.

The program Classifier in the Ribosomal Database Project was used to identify the taxonomic affiliations of the 531,408 sequences at hierarchical taxonomic levels (see Table S1 in the supplemental material). Regardless of the treatments, Actinobacteria (relative abundance of 43.8%) was the dominant taxonomic group at the phylum level, followed by Proteobacteria (27.2%), Acidobacteria (8.1%), Bacteroidetes (5.0%), Gemmatimonadetes (1.5%), Firmicutes (1.4%), and another 12 bacterial phyla with the relative abundance of less than 0.3%. Actinobacteria (43.8%), Alphaproteobacteria (15.4%), and Sphingobacteriia (4.8%) were the most abundant groups at the class level; Actinomycetales (35.2%), Rhizobiales (9.8%), Sphingobacteriales (4.8%), and Solirubrobacterales (4.5%) were the most abundant groups at the order level; Bradyrhizobiaceae (5.1%), Nocardioidaceae (4.9%), Streptomycetaceae (4.8%), Micromonosporaceae (4.7%), and Chitinophagaceae (4.1%) were the most abundant groups at the family level; and Streptomyces (4.2%) and Gp6 (3.4%) were the most abundant groups at the genus level.

To explore the effects of nano-TiO₂ and nano-ZnO on individual taxa, we defined sensitive soil bacteria as those whose relative abundance showed a significant (P < 0.05) slope (either positive or negative) with the concentration of either ENP at both sampling times. Of the identified taxa that exist in all samples, 9, 10, and 3 taxa were positively correlated with nano-TiO₂, nano-ZnO, and both nanoparticles, respectively, while 25, 15, and 12 taxa were negatively correlated with nano-TiO₂, nano-ZnO, and both nanoparticles, respectively (Table 3).

Table 3.

The number of taxa showing positive, negative, or no response to nano-TiO₂, nano-ZnO, and both nanoparticles at hierarchical taxonomic levels

Taxon	No. of taxa by ENP(s) and response type
	Nano-TiO2			Nano-ZnO			Both ENPs
	Positive	Negative	None	Positive	Negative	None	Positive	Negative	None
Phylum	1	1	9	0	0	10	0	0	10
Class	1	1	18	0	1	18	0	1	18
Order	0	4	13	1	1	13	0	1	14
Family	4	8	37	4	7	40	2	5	41
Genus	3	11	58	5	6	64	1	5	62
Total	9	25	135	10	15	145	3	12	145

The taxa with significant responses (positive or negative) to both metal oxide ENPs were selected to illustrate the effects of nano-TiO₂ and nano-ZnO on individual taxa (Fig. 4). There were significant positive linear relationships of families Sphingomonadaceae and Streptomycetaceae and of the genus Streptomyces with the exposure doses of both nano-TiO₂ and nano-ZnO at both sampling times; significant negative linear relationships of the class Alphaproteobacteria, the order Rhizobiales, families Bradyrhizobiaceae, Geodermatophilaceae, Methylobacteriaceae, Micromonosporaceae, and Rhodospirillaceae, and the genera Actinoplanes, Balneimonas, Blastococcus, Bradyrhizobium, and Skermanella occurred with both nanoparticles.

Fig 4.

Soil bacteria susceptible to both nano-TiO₂ and nano-ZnO at both sampling times. Error bars indicate the standard error of the mean (n = 4).

The slopes of the dose-response curves varied among taxa within a range of 0.05 to 8.15 (Fig. 4; see also Table S2 in the supplemental material). When we compared paired slopes of dose-response curves for the same taxon at the same sampling time but for different metal oxide ENPs, the slopes of nano-ZnO were always higher (2 to 6 times) than that of nano-TiO₂ (Fig. 4, note the different x axes), indicating the stronger effects of nano-ZnO on these bacteria.

Protease activity.

Protease activity was significantly increased by both ENPs (Fig. 5). This increase paralleled the responses of the family Streptomycetaceae and the genus Streptomyces (Fig. 4b and c) but not of other taxa that are well-known for extracellular protease production (i.e., Bacillus, Lactococcus, Serratia, Pseudomonas, Aeromonas, and Vibrio). As a result, there was a positive correlation between protease activity and the relative abundance of Streptomycetaceae (R = 0.49, P = 0.000) and Streptomyces (R = 0.47, P = 0.000) (see Fig. S1 in the supplemental material).

Fig 5.

Effects of nano-TiO₂ (a) and nano-ZnO (b) on protease activity. Error bars indicate the standard error of the mean (n = 4). Bars within the same bar cluster labeled by the same letter do not differ at a P value of <0.05.

DISCUSSION

Do ENPs affect soil bacterial communities?

Debate remains about how significantly ENPs released into soil will affect soil microbial communities (21, 31, 35, 50). Pure culture studies regularly report both acute toxicity and sublethal effects of ENPs on bacteria through mechanisms including membrane disorganization, DNA damage, surface-coating-related photocatalytic oxidation, and reactive oxygen species (ROS) production (5, 10, 23, 44). This suggests that ENPs could be toxins when released into soil, e.g., harming or killing microorganisms. However, while some studies evaluating ENP effects on soil communities reported shifts in bacterial community composition and reduced biomass and enzyme activities (16, 21, 35), others found subtle (31) or no (50) effects of ENPs on soil microbial communities. The apparent contradiction of soil microbial responses to ENPs could be partially caused by the inherent toxicity differences among ENPs (e.g., metal/metal oxides versus fullerenes) (1, 26, 28) and also possibly by differences among soils used in experiments (49) or experimental conditions. ENPs, depending on their core and surface chemistries, may undergo agglomeration, sorption, desorption, dissolution, and migration differently in different soils that have different textures, pHs, ionic strengths, and organic matter contents; ENPs may thereby vary in bioavailability and so alter the soil microorganisms' exposure to the ENPs (34).

Our pyrosequencing data demonstrate that both nano-TiO₂ and nano-ZnO induced soil bacterial community shifts through either direct toxicity or indirect effects on some sensitive taxa (Fig. 1, 2, and 4; Table 1), showing that these ENPs can reduce some specific soil bacterial populations. Since these data were counts determined from a constrained number of sequences, for one species to decrease in this data set, others would necessarily increase. However, as there were many taxa that could respond, a significant response at the taxon level would suggest a strong enough specific effect that it would likely have been due to a distinct target mechanism. As the overall biomass and activity of this community declined as a result of the ENPs (21), significant negative responses are most likely due to toxicity. Increases could result from a specific enhancement due to either the ENP or a release from competition from taxa that were repressed. Previous studies also demonstrated distinct effects of nano-TiO₂ on phyllosphere microbial communities (51), activated sludge bacterial communities (56), and microbial biofilms in stream microcosms (3), which partially support our observations in the terrestrial system studied here. The potential mechanism could be the direct toxicity of ENPs on soil bacteria through the release of metal ions or attachment-related cell damage (5, 10, 23, 28, 44). ENPs may also indirectly affect soil bacteria by changing nutrient availability or the bioavailability of cooccurring contaminants and by changing physical properties of the soil due to their large surface area and high reactivity (4, 25). Further research is needed to partition the relative importance of different factors in influencing soil microbial communities.

Both nano-TiO₂ and nano-ZnO altered soil bacterial community diversity, including abundances of specific functional groups, but it is possible that over the long term such community changes could reverse, for example, if the pressures driving the community shift are removed. However, in a somewhat comparable study, microbial communities in soils experimentally polluted with various heavy metals were different from those in uncontaminated soil 34 months after the exposures (24). Another study showed that microbial communities did not recover even 12 months after the metal stress was removed (15). Both studies suggested that some ecological niches might have been taken over by tolerant species and that pollution-induced community shifts could be preserved for a long time. Since most ENPs are chronically released into the environment and are not biodegradable, such materials are likely to accumulate in soil and thus could cause long-term effects.

The effects of nano-TiO₂ and nano-ZnO on soil bacterial communities are dose dependent.

Once substantial effects of ENPs on soil bacterial communities were observed, the next questions were whether and how the bacterial community varied as a function of exposure dose. It is necessary to address this question to quantitatively predict potential impacts of ENPs on soil microbial communities. Our results showed that, despite the high complexity and diversity of soil bacterial communities, ENP-induced communities changed with the dose of both nano-TiO₂ and nano-ZnO (Fig. 3), consistent with our previous results characterized through terminal restriction fragment length polymorphism (T-RFLP) analysis (21).

For both metal oxide ENPs examined in this study, the dose-response curves changed over time (Fig. 3). The different dose-response curves revealed temporal shifts of soil bacterial communities under ENP exposure, which coincide well with the results of PCoA (Fig. 1) and NP-MANOVA (Table 1). The shifting curve shapes over 60 days imply a saturation process whereby high doses elicit maximum responses over a short time frame, while lower doses cause protracted responses. Such dose dependencies in the rate of community change could occur if nanoparticles are diffusing slowly to receptors or if receptors are highly sensitive to the local dose.

Because the slope of the dose-response curve reflects how strongly the community responds to increasing nano-TiO₂ or nano-ZnO exposure, a higher slope implies a greater impact of specific nanoparticles on soil bacterial communities. We can compare the relative effects of nano-TiO₂ and nano-ZnO on soil bacterial communities by comparing the slopes of the dose-response curves. Our results show stronger effects of nano-ZnO on the community (Fig. 3). This is consistent with pure-culture studies showing higher bacterial toxicity from nano-ZnO than from nano-TiO₂ (1, 28).

Soil bacteria that are susceptible to nano-TiO₂ or nano-ZnO exposure.

In addition to an examination of the effects of nano-TiO₂ or nano-ZnO on the overall structure of soil bacterial communities, pyrosequencing allows the examination of the responses of hundreds of individual taxa simultaneously. Through the identification of sensitive taxa, it becomes possible to envision how ENPs could affect ecosystems. Further, a major endeavor in nano-toxicology is to hasten ENP toxicity assessments by using high-throughput screening (HTS) (12). The environmental utility of HTS would benefit from the identification of sensitive and environmentally important bacteria as screening targets.

To our knowledge, this is the first study to explore the taxon-specific responses of soil bacteria in a complex community to ENPs. We used a conservative standard to identify the sensitive taxa, defined as those whose relative abundance significantly (P < 0.05) responded to either nano-TiO₂ or nano-ZnO exposure at both sampling times; the additional taxa that showed significant responses to certain nanoparticles at only one sampling time (15 or 60 days) were not classed as sensitive. To increase the reliability of the comparison among communities, we explored the response patterns of individual taxa using their relative abundances. Previous studies have found significant correlations between quantitative PCR and pyrosequencing-based abundance estimates (6), and the relative abundance has been used in characterizing the relationship between environmental gradients and bacterial responses (11, 32, 37, 47).

A number of taxa susceptible to nano-TiO₂ or nano-ZnO exposure were identified, and the slopes of dose-response curves varied among taxa (Fig. 4; see also Table S2 in the supplemental material). Considering that both nano-TiO₂ and nano-ZnO reduced total soil microbial biomass (21), taxa that declined in relative abundance almost certainly declined in absolute abundance, but it is hard to evaluate the changes in absolute abundance of those that increased their relative abundance. They may have actually increased in absolute terms, remained unchanged while other organisms declined, or even declined but to a lesser degree than other taxa, thereby increasing as a proportion of the community.

Notably, some of the taxa sensitive to metal oxide ENP exposure are known to be functionally significant in organic matter decomposition, N₂ fixation, and methane oxidation, indicating that specific ecosystem processes carried out by such bacteria may have changed. For example, the relative abundance of the family Sphingomonadaceae, known as a decomposer of recalcitrant organic pollutants (53), increased in the presence of both nanoparticles (Fig. 4a). It has been reported that Sphingomonas paucimobilis maintained a higher population density in hexachlorocyclohexane-contaminated soil than in uncontaminated soil (41), which may indicate the relative success of this genus in contaminated environments. The family Streptomycetaceae and the genus Streptomyces also responded positively to both nanoparticles (Fig. 4b and c). Bacteria in these taxa metabolize biopolymers including protein, cellulose, chitin, and lignocellulose (8). The order Rhizobiales, the family Bradyrhizobiaceae, and the genus Bradyrhizobium, which contain symbiotic N₂-fixing bacteria, declined in response to both nanoparticles (Fig. 4e, f, and n). Thus, these ENPs could interfere with symbiotic N₂ fixation in exposed legume crops, such as soybean (an important food crop). The family Methylobacteriaceae, which contains methanotrophs and other taxa that use one-carbon compounds as their sole source of carbon and energy (9), decreased in response to both nanoparticles (Fig. 4h). Methanotrophs are vital for methane oxidation to CO₂ and hence contribute to reducing methane emissions from terrestrial ecosystems.

Although it is difficult to ascribe specific functions to particular groups of organisms, it is notable that protease increased in response to both nano-TiO₂ and nano-ZnO dose (Fig. 5) and correlated with the relative increase of both Streptomycetaceae and Streptomyces (Fig. 4b and c; see also Fig. S1 in the supplemental material), organisms known to produce substantial amounts of extracellular proteases (8). The increase in protease occurred although overall both microbial activity and biomass declined with ENP dose (21). Although this correlation does not provide conclusive evidence that the increase in Streptomycetaceae was directly responsible for the increase in protease, it is suggestive. Although many microorganisms produce extracellular proteases, the only other bacterial group that increased with both ENPs was the Sphingomonadaceae (Fig. 4), while other known producers of extracellular proteases, e.g., Bacillus, Lactococcus, Serratia, Pseudomonas, Aeromonas, and Vibrio (45), did not change with ENP dose. The Streptomycetaceae were also quite abundant (4.8% of total sequences), and most of these organisms fell within the Streptomyces(87.5% of Streptomycetaceae); so it would be easy to imagine that changes in the abundance of the group could cause measurable changes in the activity of a single enzyme system. Thus, we hypothesize that the increase in protease activity was driven by the relative increase in Streptomyces in response to ENP additions.

This study expanded from prior work (21) to identify specific bacterial populations that are sensitive to ENPs in soil. The applied ENPs affected communities and so appear to be bioavailable, but effects on individual taxa could result either from direct toxicity or indirect abiotic effects such as changes to water or nutrient availability or from biotic interactions. As some of the sensitive taxa are known to carry out important and defined roles in nitrogen and carbon cycling, the accumulation of ENPs in soil could affect critical, ecosystem-level, biogeochemical processes.