Introduction

The nanoparticles (NPs) have been increasingly manufactured in industry because of the well-known characteristics such as high reactivity, electromagnetic properties and high antibacterial property [1]. Zinc oxide nanoparticle (nZnO), one of the most popular manufactured metal oxide nanomaterials, has unique properties (i.e., surface area and reactive sites) due to the extremely small size and is increasingly used in a range of products, such as sunscreens, cosmetics and antibacterial ointments [2]. The wide applications have caused a rapid increase in the production of nZnO, with 30,000 tons worldwide in 2010 [3]. As a result, the amount of nZnO entering the environment is increasing and the occurrences of nZnO, in the range 1–10 μg/L or higher, have been commonly reported in natural water and sediments [4, 5]. Consequently, there is a critical need to investigate the toxic effect and the potential health risk of nZnO [6].

To date, the studies of toxic effects for nZnO were mostly focused in the field of individual pollution, and the results demonstrated that nZnO can produce toxic effect upon bacterial, crustaceans, earthworms and mammalian cells [7, 8]. Adams et al. [9] determined toxicity of nZnO, nano-titanium dioxide (nTiO2) and nano-silicon dioxide (nSiO2) upon the Escherichia coli and found that nZnO was the most toxic nanoparticle. Furthermore, some studies were also performed for the purpose of understanding the toxicity mechanism of nZnO [10]. nZnO can cause damage to the organ and change osmoregulatory of Oreochromis niloticus [11]; the phosphodiester bond of L-R-phosphatidylethanolamine in Escherichia coli can be broken by nZnO [12]. Moreover, a variety of studies proved that the toxicity of nZnO is related to the dissolution of Zn2+ [13]. For instance, nZnO caused the cytotoxicity by means of interfering with the homeostasis of Zn2+ [14]; Zhang et al. [15] proved the toxic difference of various nZnO particles mainly depended on their dissolution.

Organisms are typically exposed to multiple mixtures of pollutants rather than single chemicals [16]. In the process of transportation and disposal, it is conceivable that NPs are able to form nanoparticle–toxin complexes due to their high surface area and large aggregates [17]; thus, there are ongoing concerns on evaluating the environmental risk for the mixtures containing NPs. Recently, the toxic effects of nZnO combined with other chemicals were investigated in few studies [18]. The joint effects of the nZnO and surfactants, for example, were investigated at equitoxic mixtures in acute toxicity test, which showed that the joint effects can be explained by the interactions between the Zn2+ and the surfactants [19]. In the field of toxicology, chemicals are classified as narcotic or reactive compounds based on their mode of action (MOA) for a better mechanistic understanding of interactions in the mixture toxicity [20]. A variety of studies proved that mixtures of compounds exerting only one MOA (narcotic or reactive) can be assumed as additive behavior, whereas the interactions of differently acting compounds tend to yield a less or more mixture toxicity [21]. Unfortunately, the toxic effects of nZnO combined with other chemicals were rarely investigated, leading to the fact that in the mixture pollution, it is still unclear whether the dissolved Zn2+ also mainly accounts for the nZnO toxicity in the mixtures of nZnO–chemicals.

In the field of mixture toxicology, the interactions of chemicals always cause the changes in the different joint effects, including synergism, antagonism and additivity [22]. For instance, the joint effects of the nZnO and pollutants were investigated in acute toxicity test, and the results showed antagonism [23]; the additive effect between (nZnO) with nano-copper oxide (nCuO) was identified in the mixture toxic effects upon Scenedesmus obliquus [24]; the antagonistic effect between nZnO with Pb was observed in the mixture toxic effects upon Leucaena leucocephala seedling [25]. However, the joint effects between nZnO and chemicals with different MOA have been rarely investigated, and the predictive powers of the concentration addition (CA) and independent action (IA) models have not been verified, although CA and IA models were extensively employed to predict the toxic effects of mixtures [26].

Vibrio fischeri, the marine bacterium, has been widely used as the test organism for investigating the toxicity of pollutants, including nZnO [27], antibiotics [26, 28], and heavy metals [29]. In recent years, reactive compounds (i.e., antibiotics) and narcotic compounds (i.e., lignin phenols) were frequently detected in Yangtze River Basin and reported in many previous studies [30, 31]. In addition, researches have suggested that organisms are exposed to the metal NPs and metal ions in the Yangtze River [32, 33]. Therefore, the purpose of this study is to (1) explore the role of Zn2+ in nZnO toxicity of the binary mixtures containing nZnO and chemicals with different MOA, (2) evaluate the joint effects of nZnO and tested chemicals, and (3) investigate the predictive powers of CA and IA models for the mixture toxicities of nZnO and tested chemicals.

Materials and methods

Test materials

The freeze-dried marine bacterium, V. fischeri, was supplied by the Institute of Soil Science, Chinese Academy of Sciences, Nanjing PRC. nZnO (30 ± 10 nm), ZnSO4 (Zn2+), four narcotic compounds (aniline (AL), 2-nitroaniline (NAL), p-toluidine (TD) and hydroquinone (HQ)) and five reactive compounds (sulfamethoxazole (SMZ), sulfapyridine (SPY), sulfadiazine (SD), tetracycline hydrochloride (TTC) and oxytetracycline hydrochloride (OTC)) were purchased from Aladdin Reagent Company (Shanghai, China, www.aladdin-e.com) and were used without further purification. The detailed information of 9 organic chemicals is listed in Additional file 1: Table S1.

Toxicity test

Single toxicity test

The single toxicity test was performed following our previous methods [28]. That is, a 3% NaCl solution was used as the diluent and the bioluminescence of V. fischeri was recorded over a range of chemical concentrations by a SpectraMax M5 plate reader (Molecular Devices, Sunnyvale, California). The exposure time of tested chemicals with V. fischeri for acute and chronic toxicity test was 15 min and 12 h, respectively. The inhibition of the tested chemicals toward bioluminescence was calculated as Eq. 1. Based on the decrease in light emission, the obtained concentration relationship data were fitted using dose–response model (Eq. 2) [34] and reported in unit of mg/L. The detailed information about single toxicity test is presented in Additional file 1: Fig. S1.

$$Y = \frac{{I_{{\text{control}}} - I_{s} }}{{I_{{\text{control}}} }} \times 100\%$$
(1)

where Y is the inhibition ratio or response, Icontrol and Is are the average relative light units of V. fischeri exposed to the controls and test chemicals, respectively.

$$Y = A_{2} + \frac{{A_{2} - A_{1} }}{{1 + 10^{{(\log C_{0} - C) \times P}} }}$$
(2)

where A1 and A2 are bottom inhibition and top inhibition, respectively. C is the concentration of the tested chemical, C0 is the value of C at 50% of the inhibition ratio, and P is the parameter of slope for the concentration–response relationship curve.

Mixture toxicity test

The binary mixtures, including mixtures at equitoxic ratios and the mixtures at non-equitoxic ratios based on the single toxicity results (EC50), were prepared in ratios (1:102.5, 1:102, 1:101.5, 1:10, 1:100.5, 1:1, 100.5:1, 10:1, 101.5:1, 102:1, 102.5:1) of the individual concentration (n (nZnO or Zn2+):m (chemicals), mg/L). The binary mixture toxicity tests were conducted in a same method as the analysis of individual toxicity test. Mixture toxicity data were fitted and described as \(\text{EC}_{{_{{\text{i,M}}} }}\) (Eq. 3) [35]. The joint effects of the mixtures were represented as the sum of toxic units (TU) [36], as shown in Eq. 4.

$$EC_{\text{i,M}} = \frac{{C_{\text{A}} + C_{\text{B}} }}{{\frac{{C_{\text{A}} }}{{{\text{EC}}_{\text{i,A}} }} + \frac{{C_{\text{B}} }}{{{\text{EC}}_{\text{i,B}} }}}}$$
(3)
$${\text{TU}} = \frac{{C_{\text{A}} }}{{{\text{EC}}_{{ 5 0 {\text{A}}}} }} + \frac{{C_{\text{B}} }}{{{\text{EC}}_{{ 5 0 {\text{B}}}} }}$$
(4)

where EC50A and EC50B are median effective inhibition concentrations of components A and B, respectively. ECi, M is the effective concentration of the mixtures. CA and CB are the concentrations of the individual chemical in mixtures at median inhibition. Simple additivity is characterized by 1.2 > TU > 0.8, while TU > 1.2 represents antagonism and TU < 0.8 indicates synergism [37].

Toxicity prediction

Concentration addition (CA) and independent action (IA) are two classical models for mixture toxicity prediction and are widely used to predict the joint effect of mixtures [38]. CA and IA models are expressed mathematically as Eqs. (5) and  (6), respectively:

$${\text{EC}}_{\text{x,m}} { = }\left( {\frac{{P_{\text{A}} }}{{{\text{EC}}_{\text{x,A}} }} + \frac{{P_{\text{B}} }}{{{\text{EC}}_{\text{x,B}} }}} \right)^{{^{ - 1} }} ,$$
(5)

where ECx,m is the concentration of the mixture eliciting X% effect, PA and PB are the concentration ratios of A and B components in the mixture, ECx,A and ECx,B denote the concentrations of the A and B components that elicit an X% effect.

$${\text{E}}({\text{c}}_{\text{m}} ) = 1 - \left( {1 - E\left( {{\text{c}}_{\text{A}} } \right)} \right) \times \left( {1 - E\left( {{\text{c}}_{\text{B}} } \right)} \right),$$
(6)

where \(E\left( {C_{m} } \right)\) is the toxic effect of mixture, \(E\left( {C_{A} } \right)\) and \(E\left( {C_{B} } \right)\) are the effect from analyte A and B if applied singly at an exposure concentration of A and B, respectively.

Statistics

SPSS 25.0 software (SPSS Inc.) was used to test the significant difference of the results and P < 0.05 was considered to be statistically significant. The statistic quality of linear models was evaluated by determination coefficients (R2), the formula as shown Eq. 7 [39]. The parameters of mean absolute percentage error (MAPE) (Eq. 8) and root mean square error (RMSE) (Eq. 9) were applied to measure the prediction accuracy of CA and IA models [40]. These indices were obtained by the following equations:

$${\text{R}}^{2} { = }\left( {\frac{{\sum\nolimits_{i = 1}^{n} {(y_{i} - \bar{y})(\hat{y}_{i} - } \bar{\hat{y}})}}{{\sqrt {\sum\nolimits_{1 = 1}^{n} {(y_{i} - \bar{y})^{2} \sum\nolimits_{i = 1}^{n} {(\hat{y}_{i} - } } \bar{\hat{y}})^{2} } }}} \right)^{2} ,$$
(7)

where \(\bar{\hat{y}}\) is the mean predicted responses.

$${\text{MAPE = }}\frac{1}{M}\sum\limits_{{{\text{k}} = 1}}^{M} {\left| {\frac{{\hat{p}_{ettc} (k) - p_{ettc} (k)}}{{p_{ettc} (k)}}} \right|} \times 100\% ,$$
(8)
$${\text{RMSE = }}\sqrt {\frac{1}{M}\sum\limits_{\text{k = 1}}^{M} {\left[ {\hat{p}_{ettc} (k) - p_{ettc} (k)} \right]^{2} } } ,$$
(9)

where M is the number of sample intervals.

Results and discussions

The single toxicity of tested chemicals

To investigate the mixtures’ toxicity of nZnO and chemicals with different MOA, the single toxicity of nZnO, Zn2+ and other 9 chemicals to V. fischeri was determined in acute and chronic toxicity test. The data were fitted by the model of dose–response and obtained curves are presented in Fig. 1. The values of R2 suggested a good fitting (0.977–0.999). In the case of acute toxicity (Fig. 1a), nZnO and Zn2+ presented higher toxic effects than other tested chemicals, EC50 values for tested chemicals were ranging from 1.17 mg/L to 319.24 mg/L, and the order of acute toxicity was as follows: nZnO > Zn2+> HQ > SMZ > NAL > TD > AL > OTC > TTC > SD > SPY. As far as chronic toxicity test (Fig. 1b), results showed that OTC presented higher toxic effects than other tested chemicals, EC50 values for tested chemicals were ranging from 1.17E−2 mg/L to 100.64 mg/L, and the order of chronic toxicity was as follows: OTC > TTC > HQ > SMZ > NAL > SD > TD > nZnO > Zn2+> SPY > AL.

Fig. 1
figure 1

The single toxic effects of chemicals upon V. fischeri and the fitted dose response curves. a The acute toxicity test results, b the chronic toxicity test results

Full size image

Mixture toxicity of nZnO and chemicals with different MOA

Acute toxicity test

Based on the results of the single acute toxicity test, the mixture toxicity of nZnO/Zn2+ and 9 tested chemicals with different MOA was determined at their equitoxic ratios (Additional file 1: Fig. S2). As shown in Additional file 1: Fig. S2, the relationship between the luminescent inhibition ratio and the concentration of these mixtures was good, with R2 ranging from 0.978 to 0.999 for nZnO–chemicals and from 0.982 to 0.993 for Zn2+–chemicals, respectively. The difference between \(EC_{{50{\text{M}}}}^{15\text{min} }\) for binary mixture containing nZnO and \(EC_{{50{\text{M}}}}^{15\text{min} }\) for binary mixture of Zn2+–chemicals is presented as Fig. 2a. As shown, the difference was not significant for tested chemicals (P > 0.05). To further verify the results, the acute mixture toxicity of nZnO/Zn2+ combined with SMZ (a reactive compound) and AL (a narcotic compound) was subsequently determined at non-equitoxic ratios (Additional file 1: Fig. S3). Figure 2b indicates that in acute toxicity test, the difference between \(EC_{{50{\text{M}}}}^{15\text{min} }\) for nZnO–SMZ and \(EC_{{50{\text{M}}}}^{15\text{min} }\) for Zn2+–SMZ was still not significant (P > 0.05). Furthermore, the same conclusion can be obtained for nZnO–AL and Zn2+–AL at their non-equitoxic ratios (P > 0.05, Fig. 2c).

Fig. 2
figure 2

Comparison of mixture toxicity between Zn2+–chemicals with nZnO–chemicals. a Zn2+–chemicals and nZnO–chemicals at equitoxic ratios. B in acute test, b Zn2+–SMZ and nZnO–SMZ at non-equitoxic ratios in acute test, c Zn2+–AL and nZnO–AL at non-equitoxic ratios in acute test, d Zn2+–chemicals and nZnO–chemicals at equitoxic ratios in chronic test, e Zn2+–SMZ and nZnO–SMZ at non-equitoxic ratios in chronic test, f Zn2+–AL and nZnO–AL at non-equitoxic ratios in chronic test

Full size image

Chronic toxicity test

Based on the results of the single chronic toxicity, the mixture toxicity of nZnO/Zn2+ and these chemicals was determined at their equitoxic ratios (Additional file 1: Fig. S4). As shown in Additional file 1: Fig. S4, the relationship between the luminescent inhibition ratio and the concentration of these mixtures was good, with R2 ranging from 0.966 to 0.994 for nZnO–chemicals and from 0.956 to 0.991 for Zn2+–chemicals. In the case of equitoxic ratios for chronic toxicity test, the difference between \(EC_{{50{\text{M}}}}^{{ 1 2 {\text{h}}}}\) for binary mixtures containing nZnO and \(EC_{{50{\text{M}}}}^{{ 1 2 {\text{h}}}}\) for binary mixtures containing Zn2+ is presented as Fig. 2d. It can be observed that the difference was still not significant for tested mixtures (P > 0.05). To further verify above results, the chronic mixture toxicity of nZnO/Zn2+ combined with SMZ (a reactive compound) and AL (a narcotic compound) was determined at their non-equitoxic ratios (Additional file 1: Fig. S5). The results of Fig. 2e, f consistently indicated that in chronic toxicity test, the difference between \(EC_{{50{\text{M}}}}^{{1 2 {\text{h}}}}\) for binary mixtures of nZnO–SMZ/AL with \(EC_{{50{\text{M}}}}^{{1 2 {\text{h}}}}\) for binary mixtures of Zn2+–SMZ/AL at their non-equitoxic ratios was still not significant (P > 0.05).

Consequently, dissolved Zn2+ mainly accounted for the nZnO toxicity in the mixtures of nZnO-reactive chemicals and in the mixtures of nZnO-narcotic chemicals, not only in acute toxicity test but also in chronic toxicity test.

Joint effects of nZnO and chemicals with different MOA

Based on the mixture toxicity results, the joint effects of mixtures of nZnO and tested chemicals were analyzed according to Eq. 3. In the case of acute toxicity test, it can be observed from Fig. 3a that TU values for the mixtures of nZnO–chemicals at equitoxic ratios ranged from 0.75 to 1.77. Figure 3b shows that TU values for nZnO–SMZ and nZnO–AL at their non-equitoxic ratios were ranging from 0.93 to 1.25 and from 0.99 to 1.88, respectively. According to the study of Broderius et al. [37], those results indicated that in acute toxicity test, (1) the joint effects of nZnO and chemicals with different MOA were mainly additivity or antagonism, but rarely synergism. For example, the TU value lower than 0.80 can only be obtained in the mixture of nZnO–NAL and the joint effect was viewed as synergism; and (2) the joint effects for nZnO–SMZ and nZnO–AL were consistent additivity in the acute test at non-equitoxic ratios. In the case of the mixture of nZnO–SMZ, for example, the TU values were 0.86–1.15.

Fig. 3
figure 3

Joint effects of nZnO–chemicals at equitoxic ratios (a) and at non-equitoxic ratios (b)

Full size image

As for chronic toxicity test, it can be observed from Fig. 3a that TU values for nZnO and tested chemicals at equitoxic ratios ranged from 0.47 to 2.45, indicating the joint effects of nZnO and chemicals with different MOA were additivity or antagonism or synergism. The synergism can only be obtained in one mixture (nZnO–HQ), and the joint effects for other mixtures were mainly additivity or antagonism. Furthermore, Fig. 3b suggests that for the mixtures of nZnO–SMZ and nZnO–AL, the joint effects were additivity in the mixtures where the difference between the concentrations of the components is large (e.g., lg (n/m) = − 2.5, 2, 2.5), whereas the joint effects were antagonism in the mixtures where the difference between the concentrations of the components is small. In the case of the mixture of nZnO–AL, for example, the TU values was 1.57, the corresponding lg n/m was 0.

Consequently, it can be concluded that for both acute toxicity test and chronic toxicity test, the joint effects of nZnO and chemicals with different MOA were mainly additivity or antagonism. Similar results were obtained for the joint effects of nZnO combined with propiconazole by Hackenberger et al. [41]. Zhang et al. [42] also found that the binary joint effects of Zn2+ and 11 nitro-substituted benzenes to Photobacterium phosphoreum were mainly antagonism.

Mixture toxicity of nZnO–chemicals predicted by CA and IA models

Based on the mixture toxicity results of nZnO and chemicals, the validity and applicability of CA and IA models were further verified (Fig. 4). Results indicated that (1) the prediction accuracy of CA and IA models was satisfied in the mixtures when the difference between the concentrations of the components was large (Fig. 4b, c). When lg n/m was − 2.5 or 2.5 for all test mixtures, for example, the values of MAPE and RMSE were mostly ranging from 0.13 to 0.24, 0.00302 to 0.00319 for CA model and were from 0.17 to 0.30, 0.00307 to 0.00340 for IA model, respectively; (2) the prediction accuracy of CA and IA models was poor in the equitoxic mixtures when the joint effects were antagonism or synergism (Fig. 4a). For example, the MAPE values were 1.26 for CA model and 1.97 for IA model in the mixture of nZnO–AL in chronic test; and (3) overall, the prediction accuracy of IA model was better than that of CA model, not only in the mixtures at equitoxic ratios but also in the mixtures at non-equitoxic ratios, as proved by the MAPE values of 0.105 to 2.506 and 0.108 to 2.242 for CA and IA model, respectively. It is well known that CA and IA models were used to predict the toxicity of mixture based on the theoretical assumption that chemicals in the mixture do not interact with each other, therefore both models may underestimate or overestimate the joint effects of binary mixtures [43]. CA model was used by Azevedo et al. [44] to predict the mixture toxicity of nZnO and nano-silver (nAg), the antagonism effect was observed and the mixture toxicity was overestimated. Wang et al. [45] reported the CA model was unsuitable to predict the mixture toxicity of Zn2+–sodium dodecyl benzene sulfonate at equivalent-effect concentration ratio on Vibrio qinghaiensis sp. Q67.

Fig. 4
figure 4

The results of MAPE and RMSE values for CA and IA models from binary mixtures. anZnO–9 chemicals at equitoxic ratios, bnZnO–SMZ at non-equitoxic ratios, cnZnO–AL at non-equitoxic ratios

Full size image

The mixture toxicity mechanism of nZnO chemicals

By now, the mixture toxicity of engineered nanomaterials (ENMs) with chemicals is of great interest in the field of toxicology. The mechanisms for the mixture toxicity can be mainly classified into the following types (Fig. 5): (1) ENMs effectively affect bioavailability of pollutants either positively or negatively by adsorption, complexation and degradation [46]; (2) the toxicokinetics of pollutants, including the process of uptake, biotransformation, distribution and elimination of pollutants in test organism, can be affected by ENMs by modifying the structure and function of cellular membrane, changing the metabolism pathways and altering the chemical species of pollutants [47]; and (3) ENMs influence the toxicodynamics of pollutants by interfering with the interactions of a toxicant with a biological target and its biological effects [48]. For example, ENMs may ease the entering and transport of pollutants in organisms via “Trojan horse effect” [49], because of the high surface to volume ratio [50]; Carbon nanotubes enter cells through damaging the cell membrane, which subsequently facilitated the entry of pollutants and induced a synergistic toxicity [51]; De La Torre-Roche et al. [52] indicated the dissolved Ag+ from silver nanoparticles (AgNPs) inhibited the activity of aquaporin and decreased the uptake of p,p′-DDE, and therefore, reducing the bioaccumulation of p,p′-DDE in test organisms.

Fig. 5
figure 5

The mechanisms underlying the toxicities of NMs and pollutant

Full size image

In the case of single toxicity, studies proved that the released Zn2+ mainly accounted for the nZnO toxicity upon V. fischeri and Escherichia coli, respectively [27, 53]. In the field of mixture toxicity, the role of released Zn2+ in nZnO toxicity remains controversial. The work of Yi et al. [54], for example, indicated no significant difference between the toxicity of nZnO–triphenyltin and Zn2+–triphenyltin to Tigriopus japonicas. While Lakshmi Prasanna et al. [55] reported that the surface defects of nZnO induced antibacterial activity via reactive oxygen species generation rather than by dissolved Zn2+. The possible reason for the above difference can be concluded as following: the main toxicity mechanism of nZnO may be different for the divergence of exposure condition, because the species of dissolved zinc ion could be changed by the components in the nature [56].

In the present study, the difference of the joint effects between nZnO–chemicals and Zn2+–chemicals was not significant (P > 0.05), not only in acute toxicity test but also in chronic toxicity test, suggesting that the toxicity of nZnO on V. fischeri was due mainly to the dissolved Zn2+. Thus, the joint effects of the nZnO and tested chemicals can be explained by the interactions between the dissolved Zn2+ and these chemicals. The joint effects of tested mixtures were mainly antagonism and additivity (Fig. 3a). The possible reasons for the antagonistic effect can be explained as follows: first, metal ions interact with organic compounds which reduce the effective dose of the pollutants in organism. In case of antibiotics, interactions between metal ions and the functional groups of antibiotics are the main mechanism for decreasing the mixture toxicity [57]. The work of Kim et al. [58], for example, revealed that the complexation reactions between Cu2+ and the phenolic compounds (narcotic compounds) played an important role in reduction of the Cu2+ concentrations and therefore decreased toxicity of the binary mixtures. Second, the components of a binary mixture may compete for binding sites [59]. Hackenberger et al. [41] found that two compounds in the treatment canceled the effect of one another in the mixtures of nZnO–propiconazole and Zn2+–propiconazole. The only synergism effect on V. fischeri occurred in the nZnO–HQ binary experiment. It was well known that phenols can cause damage to cell membrane [60]. Consequently, we speculated that the HQ disrupted the cell membrane integrity of V. fischeri, which facilitated the entry of released Zn2+ and increased the mixture toxicity.

Conclusions

Our results indicated that no significant difference of the toxicity between nZnO and Zn2+ was observed not only in single-component system but also in mixture systems of nZnO/Zn2+ and chemicals with different MOA, suggesting that nZnO toxicity was mainly caused by released Zn2+. Furthermore, the joint effects of nZnO and chemicals at equitoxic ratios were mainly antagonism and additivity, while the joint effects of nZnO–SMZ or nZnO–AL were additivity at non-equitoxic ratios. Moreover, the prediction accuracy of CA and IA models was not very well in binary mixtures at equitoxic ratios.