Zwitterionic compounds are less ecotoxic than their analogous ionic liquids

Zwitterionic compounds (ZIs) have been attracting much attention due to their properties. Unlike ionic liquids – constituted by separated ions – the ZIs structure – with the cation and anion covalently bonded – results on an added complexity and diversity, suggesting that their ecotoxicological behavior should probably not be extrapolated from those of ionic liquids. This study addresses the aquatic toxicity of ZIs towards the bacterium Allivibrio fischeri and the microalga Raphidocelis subcapitata . Sixteen ZIs, comprising five different cationic groups − ammonium, imidazolium, pyridinium, pyrrolidinium and piperidinium − and two anionic groups − sulfonate and carboxylate − were studied, and the relationships between their structure and toxicity are reported. All studied ZIs are harmless or practically harmless for both microalga and bacterium (median effective concentration, EC 50 >100 mg·L -1 ), presenting a significantly lower hazardous potential to aquatic species than their ionic liquids counterparts. The results also show that the increased hydrophobicity of ZIs, promoted by the increase of cation alkyl chain or spacer size, has a significant influence on EC 50 values for microalga, reflected on a higher toxicity. However, no significant differences were observed when considering the various cationic groups of the ZIs studied, unlike what is known for the ionic liquids ecotoxicity. Also, no relationships were found between the chemical structure of ZIs and EC 50 values estimated for the bacterium A. fischeri . The structural differences between ZIs and ionic liquids results into different mechanisms of interaction with microalgae and bacteria membranes, which may explain why the ecotoxicity heuristic rules previously reported for ionic liquids do not seem to apply to ZIs.


Introduction
Solvents are essential in many chemical processes, influencing the environmental performance of chemical industries.Despite their wide application, conventional organic solvents employed in chemical industries present several drawbacks, such as high volatility, flammability, and toxicity, which represents a clear handicap for the development of sustainable processes in line with the Green Chemistry principles.In this context, the search for novel and alternative environmentally benign solvents has gained increased attention. 15][6] One such effort was successfully achieved in 2001 when Ohno and co-workers 7 reported the synthesis and application of zwitterionic compounds (ZIs) derived from ionic liquids as alternative electrolyte materials.Zwitterions are compounds characterized by the simultaneous presence of both positively and negatively charged groups in the same chemical structure, bearing a total zero net charge.The covalently bonded anionic and cationic groups, which are linked by an alkyl or other type of spacer, confer to the ZIs both a global neutral charge and an extremely high polarity.Zwitterions are widely spread in nature, amino acids being probably the better known example of this type of molecules.The large number of anionic and cationic groups available in the chemical inventory allows the synthesis of a wide range of zwitterionic structures.Most systems studied use, as cationic groups, quaternary ammonium or imidazolium, whereas the most commonly used anionic groups are sulfonate and carboxylate. 8,9However, the number of possible structural variations by the modification of headgroups, spacers and hydrophobic tails is theoretically endless, which supports their tunable character.4][15] Recently, the use of aqueous solutions of ZIs derived from ionic liquids as alternative solvents in extraction and separation systems was demonstrated.Ferreira et al. 13,14 studied the ability of ammonium-based ZIs (also known as sulfobetaines) to induce the formation of aqueous biphasic systems when mixed with polymers or salts.These systems presented a high performance as selective separation platforms 13 and allowed the development of integrated bioreaction-separation processes. 14 Kuroda et al. 11 also demonstrated how the tunable character of ZIs can be implemented, by designing a specific solvent with a high capacity to dissolve cellulose, while keeping its biocompatible character towards the cell walls.The increased interest in the application of alternative solvents, like ionic liquids and ZIs, and their consequent industrialization, will result in the release of their residues to aquatic systems, potentially affecting the aquatic biota.As commonly referred, legislation concerning these scenarios is nowadays stricter in Europe as REACH (Registration, Evaluation, Authorization and Restriction of Chemicals) 16 requires ecotoxicological data for all chemicals produced or imported into the European Union above 1 ton/year.Therefore, the optimization of the technical performance must run in parallel with the minimization of hazardous potential, to simultaneously improve the chemical's economic viability and reduce their environmental impact.Nevertheless, and although we recognize the importance of creating ecotoxicological databases to characterize the various classes of compounds allowing their licensing, in this work, we intend to go further and provide experimental evidence on the toxicology of ZIs derived from ionic liquids.In literature, only a few studies dedicated to the environmental behavior of ZIs [17][18][19][20][21][22][23] can be found, and without consistency to be used for the understanding of their ecotoxicity and allowing a comparison with ionic liquids congeners.This work addresses thus the study of the aquatic ecotoxicity of a set of sixteen ZIs, towards the marine bacterium Allivibrio fischeri and the freshwater microalga Raphidocelis subcapitata.Included in the set of ZIs under study are some of the most common charged groups (cationic groups: ammonium, imidazolium, pyridinium, pyrrolidinium and piperidinium; anionic groups: sulfonate and carboxylate).The toxicity of these ZIs and its relation to their chemical structure are discussed, and further compared with the ecotoxicity data available in literature or determined in this work for their ionic liquid congeners.

Toxicity to Raphidocelis subcapitata
Bulk cultures of the freshwater microalga R. subcapitata were maintained in batch cultures of Woods Hole MBL medium at (20 ± 2)°C, with continuous aeration, under a 16h:8h (light:dark) photoperiod cycle.Four days before starting the tests, an inoculum was harvested from the bulk microalgae culture and incubated at (23 ± 1)°C under continuous illumination (coolwhite fluorescent light, intensity about 7000 lux).This inoculum was used to start the test, which initiated with 1.0x10 4 cells•mL - 1 of R. subcapitata in the log exponential growth phase, following the OECD guideline N201 26 adapted to the use of 24- well microplates. 27Microalgae were exposed to a geometric range of concentrations of each chemical, which was obtained by dilution of a stock solution.Both the stock solution and the dilutions were prepared using Woods Hole MBL medium.Each treatment consisted of a blank control (test solution with no algae), an algae control (no chemical) and three replicates of the tested concentration.The test conditions were the same as mentioned above for the inoculum.After 96 h of exposure, the absorbance at 440 nm was measured (Shimadzu UV-1800) and used to estimate the microalgae densities using a previously established calibration curve. 28To exclude any interference in the absorbance, each respective blank control was used as the blank in the absorbance measurement of each replicate.Since the ZI C2PipC4S, betaine and the ionic liquid [C2C1Im][C1CO2] interfere with the absorbance measurements at 440 nm, microalgae densities for these chemicals were determined by counting under a microscope in a Neubauer chamber.Cell densities were used for yield calculations.

Toxicity to Allivibrio fischeri
Acute toxicity to the Gram-negative bacteria A. fischeri was assessed using the Microtox® Toxicity Test (Microtox® M500 analyzer, Modern Water, UK), by determining the luminescence inhibition after 30 min of exposure to each compound.Bacteria, purchased freeze-dried, were activated by rehydration, and exposed to a range of diluted aqueous solutions (from 0 to 81.9 % in volume) of each compound.Dilutions were made following the manufacturer protocol with the supplied diluent reagent (2 wt% NaCl solution in water), thus ensuring optimal osmotic conditions for the bacteria.The light emission of the bacteria was measured and compared to that of the blank control (bacteria in blank diluent), to determine the light emission relative to control.

Statistical analysis
The obtained data (light emission relative to control for A. fischeri, and yield for R. subcapitata) were used to estimate the effective concentrations (EC50 and EC20 values, i.e. concentrations eliciting 50% and 20% of effect, respectively) and the corresponding 95% confidence intervals.This estimative was achieved by fitting the data to the logistic equation, using the least-squares method (non-linear regression) in the STATISTICA software package (version 8, StatSoft Inc., USA).Correlations were assessed by Pearson's correlation (Statistica, StatSoft).An alpha of 0.05 was used in all analyses.

Results and discussion
A total of sixteen ZIs constituted by five different cationic (ammonium, imidazolium, pyridinium, pyrrolidinium and piperidinium) and two anionic (sulfonate and carboxylate) groups were studied on this work (see Table 1 and Figure 1).The EC50 values of ZIs for R. subcapitata (96 h of exposure) and A. fischeri (30 min of exposure) are given in Table 2.For a more detailed view on the ecotoxicological profile of these ZIs, EC20 values are additionally provided in Table S1 of the ESI.All experimental data are in mass units (g•L -1 ) to be in line with the toxicological categories adopted by the European Commission 29 and to facilitate the comparison with data previously reported in literature, which is commonly presented in weight per volume units.
A global evaluation of the EC50 values (Table 2) suggests that none of these chemicals could be considered hazardous to the aquatic environment according to the United Nations Globally Harmonized System of Classification and Labelling Chemicals (GHS), 30 as their EC50 values are above 100 mg•L -1 .Such a low toxicity of the tested compounds agrees with the low toxicity commonly attributed to ZIs. 9,10,31 This is particularly valid for ZIs  with short alkyl chains, as the elongation of alkyl chains seems to promote an increase in toxicity, as previously observed for sulfobetaines 18 -ZIs composed of ammonium-based cationic groups and sulfonate anionic groups -and alkyl dimethyl amine oxides. 17The increased toxicity with the elongation of the alkyl chains reveals an effect of the ZIs chemical structure on their toxicity.To better understand these effects, a detailed analysis of the relationship between the structural features of the ZIs chemical (cationic and anionic groups, and spacer length) and EC50 values determined for both R. subcapitata and A. fischeri will be carried out in the following sections.

Toxicity to R. subcapitata
The median effective concentrations of the tested ZIs to the microalga R. subcapitata are presented in Figure 2.For comparison, in this figure are also presented the results obtained for betaine -a well-known naturally occurring, nontoxic and biodegradable zwitterionic compound -and ionic liquids of similar chemical structure 27,[32][33][34] and some organic solvents. 32More details about name, acronym and chemical structure of ionic liquids used in this work for comparison purposes can be found in Table S2 and Figure S1.
According to the Global Harmonized System (GHS) of classification and labelling of chemicals by the United Nations UN 30 all studied ZIs are considered not hazardous to the aquatic environment, concerning short-term (acute) aquatic hazard, as the EC50-96h values are above 100 mg•L -1 .Furthermore, the results presented in Figure 2 show that ZIs exhibit higher EC50 values than the correspondent ionic liquids, which are considered practically harmless or hardly toxic:
structure and thus chemical properties, on the environmental profile of the tested ZIs.Indeed, despite the harmless character of ZIs for the microalga, it is possible to identify some dependency of the EC50 values on their chemical structures -cf. Figure 2. Considering the sulfobetaines (NiiiC3S and NiiiC4S), it is clear that the length of the alkyl chains in the cationic group has a pronounced effect on the EC50 values.The increase of alkyl chain from 1 (N111C3S) to 5 carbons (N555C3S) lead to a decrease of the EC50 by 2 orders of magnitude (346-fold).9][40] It is true that the increase of the alkyl chain length in ZIs leads to an increased hydrophobicity, as given by the logarithmic function of the octanol-water partition coefficient (log(KOW)) values presented in Table 1 (for a hydrophobic compound log(KOW) > 0).Indeed, a similar trend was previously observed for the ecotoxicity of ionic liquids for the same aquatic species. 41This relationship is supported by the linear correlation found between the logarithm function of EC50 and the number of carbons in alkyl chains of the sulfobetaines family (NiiiC3S) depicted in Figure 3 (Pearson correlation, ρ = -0.976,p = 0.005, n = 5).However, for ZIs based in the imidazolium cations and carboxylate anions, the elongation of alkyl chain from 1 (C1ImC3C) to 4 carbons (C4ImC3C) induces only a slight decrease in the EC50 values (1.2-fold), despite the pronounced hydrophobicity augment (cf.Table 1), suggesting that hydrophobicity changes do not always affect in the same extent the (eco)toxicity of a compound.Concerning the size of the spacer between the cationic and anionic groups (cf. Figure 2) -N111C3S vs N111C4S, N222C3S vs N222C4S and C1ImC3C vs C1ImC5C -our results suggest that the increase in the spacer length translates into a decreased on the EC50 values, except for the pair N111C3S vs N111C4S.Probably, this trend is also related with the hydrophobicity of ZIs, in accordance with data previously reported by Davies et al. 18 concerning the toxicity of sulfobetaines to Daphnia magna.Indeed, the small difference in hydrophobicity between N111C3S and N111C4S (cf.Table 1) should explain the negligible effect of the spacer length in the (eco)toxicity of these ZIs to microalga.Different cationic groups were studied; among them ammonium, imidazolium, pyrrolidinium, pyridinium and piperidinium.However, it was not possible to identify any specific trend regarding the impact of this structural feature on the EC50 values (cf. Figure 2).The cation confers to the ZIs an aromatic or non-aromatic character.ZIs based on imidazolium and pyridinium families are aromatic, whereas ZIs derived from ammonium, pyrrolidinium and piperidinium families are nonaromatic (cf. Figure 1).The role of ZIs hydrophobicity on the EC50 values for the whole EC50-96h dataset is depicted in Figure 4, distinguishing both aromatic and non-aromatic ZIs.For nonaromatic ZIs, the correlation between hydrophobicity (expressed as log (KOW)), and toxicity (as log (EC50)), is strong and significant (Pearson correlation: ρ = -0.9814,p < 0.001, n = 10) but weak and non-significant for the aromatic ones (p = 0.121, n = 6).For the latter, a pronounced decrease of log (KOW) does not translate into an increase of EC50 values.
A linear relationship between hydrophobicity and toxicity of non-aromatic ZIs was reported in a previous study, which proposed to use log(KOW) as a single physicochemical property for defining sulfobetaines toxicity to the freshwater planktonic crustacean Daphnia magna. 35However, this study was carried out for quaternary alkylammonium sulfobetaines with longer radicals than those used here.In the same study, it was proposed that ZIs with lower molecular weight can cross the cellular membrane directly or be carried by an ion channel, which may outweigh the dependence of the (eco)toxicity on hydrophobicity. 35This might explain why ZIs with similar log(KOW), but structurally different, can exhibit significantly different toxicities, as is the case of N111C4S and C4ImC3C (log (KOW) = -2.35 and -2.37, and EC50 = 48.85 and 16.44 g•L -1 , respectively).It is interesting to note that the studied aromatic mechanism driving the organisms' response may be taking place here.Nevertheless, in order to clarify whether the relationship between hydrophobicity and (eco)toxicity is significant for aromatic ZIs, further ecotoxicological tests should be performed.Interestingly, the opposite trend was observed for ionic liquids, with aromatic compounds (e.g.[C4C1Im]Br) exhibiting higher toxicity than the non-aromatic ones (e.g. [C4C1Pyr]Br).This was observed for the R. subcapitata microalga 42,43 and other freshwater species, [42][43][44][45] and it was explained by the increased water solubility of ionic liquid-based aromatic cations. 43he effect of ZIs anionic group (sulfonate vs carboxylate) can be addressed by comparing the toxicity of N222C3C and N222C3Scf.Figure 2. Following their differences in hydrophobicity (log (KOW) = -2.13 and -1.80, respectively), a higher EC50 value was expected for N222C3C than for N222C3S, which was not observed (EC50 = 22.39 and 41.26 g•L -1 , respectively).Nevertheless, due to the higher hydrophilic character of the carboxylate group, carboxybetaines tend to present larger solubility in water and lower self-association effects, 9 which may result in a higher interaction of ZIs with the microalga membrane.

Toxicity to A. fischeri
The median effective concentrations estimated for the ZIs following exposure of the bacterium A. fischeri are presented in Figure 5.The results obtained for betaine and previously reported for structurally similar ionic liquids 33,44,[46][47][48][49] and organic solvents 50,51 are also presented for comparison.The EC50 values of studied ZIs varied by 344-fold, with N333C3S presenting the lowest (EC50 = 0.6714 g•L -1 ) and C1PyrC4S the highest value (EC50 = 231.1 g•L -1 ), thus a range similar to that observed for the microalga.Following the hazard ranking from Passino & Smith. 52all ZIs are harmless or practically harmless also for the bacterium.Furthermore, the equivalent ionic liquids used for comparison (cf.Table S2 of ESI) exhibit significantly higher toxicity than the studied ZIs: for example, C1PyrC4S (231. ), again showing the hazardous behavior of ionic liquids towards the ZIs.The determined EC50-30 min of betaine to A. fischeri (141.6 mg•L -1 ) is lower than the value reported in a previous study (242 mg•L -1 ); 53 but the fact that the exposure period in that work was not clarified prevents further comparison.The reduced antimicrobial activity of ZIs, in particular of sulfobetaines, is in agreement with the results of a previous study, in which sulfobetaines were shown to exhibit a lower toxicity to bacterium than quaternary ammonium salts. 54nlike to what was observed for the median effective concentration of ZIs to the freshwater microalga, it is difficult to identify any trend regarding the toxicity dependency for the bacterium on the ZI chemical structures.The ZIs harmless character may explain the absence of clear trends in the data obtained, although the quantification of the EC50 values was well established.Moreover, previous works concerning the ecotoxicity of other type of solvents, which also record very high EC50 values for bacterium, were able to observe a relationship between the toxicity and the chemical structure of the compounds under study.However, the trends reported are variable according to the compounds and biological species used.For instance, A. fischeri exposed to amine oxides showed increased EC50 values with increased alkyl chain length from 12 to 14 carbons, which is the opposite trend from that reported for freshwater organisms, such as Daphnia magna and R. subcapitata. 20,21On the other hand, for several Gram-positive and Gram-negative bacteria, increased toxicity of sulfobetaines was caused by increasing chain length from 10 to 16 carbons. 55n particular, sulfobetaines with 10 carbon atoms did not exhibit antimicrobial activity to the tested bacteria. 55The same trend was observed for bacteria exposed to ZIs composed of both a pyridinium and a sulfonate charged groups. 56Here, the following trends were observed: EC50 of N444C3S > N555C3S ≥ N111C3S ≥ N222C3S > N333C3S and C4ImC3C > C1ImC3C.Applying the rationale used before there is an apparent inconsistency on the trend observed.For the sulfobetaines, there is a decrease of EC50 values from N111C3S to N333C3S but, comparing the EC50 values of N333C3S and N444C3S, there is a significant increase of EC50 with the alkyl chain length.Some authors suggested that this behavior may be related with self-aggregation effects that may become more pronounced as the size of cationic group alkyl chain increases. 9Considering the reported data for the ecotoxicity of tetraalkylammonium chloride ionic liquids ([Niiii]Cl) to the bacteria A. fischeri, 46 where alkyl chains varying from 1 to 3 carbons were tested, and the experimental value determined in this work for [N4444]Cl (cf.Table S1), it is possible to conclude that the EC50 of the ionic liquids decreases continually with the increase of cation alkyl chain length from 1 to 4 carbons, presenting a distinct behavior to the one here presented for ZIs.Nevertheless, similar trend shifts in ZIs properties/behavior were previously reported. 13oncerning imidazolium-based ZIs, a trend for decreased toxicity with increasing hydrophobicity can be identified through a strong correlation between the two variables, but this correlation was not found to be significant, possibly due to the low sample size (ρ = 0.844, p = 0.072, n = 5).When comparing the anionic groups, carboxylate and sulfonate, is the former that results in higher EC50 values, probably due to the higher hydrophilicity of carboxylated ZIs.Interestingly, both trends are the opposite to that observed for the microalga.While further studies are needed to better understand how these compounds interact with this type of microorganism, the role of the salinity )

HARMLESS PRACTICALLY HARMLESS HARDLY TOXIC
of the test medium in modulating the observed trends should not be ruled out.The presence of electrolytes/salts may interfere with the complex balance of forces governing the structure and the solution properties of ZIs, 9,10 and consequently the resultant toxicity.However, the exposure to the ZIs was carried out under the same conditions of salinity, which allows comparing the ZIs toxicity data.][57][58] Their mechanism of toxic action against Gram-negative bacteria is assumed to be bacteriostasis, indicating that ZIs may interfere with the bacterial metabolism or ability to reproduce, but do not destroy their cell membrane. 57Concerning their mechanism of toxic action to aquatic species in general, it is believed to occur through narcosis, a nonspecific disturbance of the membrane integrity and functioning, resulting from the partition of pollutants into biological membranes. 59ZIs are suggested to act as polar narcotics, 35 also known as "baseline toxicants".When considering sulfobetaines, the toxicity mechanism suggested by Davies et al. 35 is supported by other studies concerning pharmaceutical zwitterionic compounds. 60,61he differences in the toxicity of different chemicals to Gramnegative bacteria like A. fischeri can be attributed to differences in the molecular interactions between the chemical and key cellular events, namely (i) the mechanisms involved in the uptake, i.e., concerning composition of the cell wall, transport and diffusion across the membrane, (ii) changes induced by the toxic chemicals in metabolic pathways which supply cell energy, and (iii) interactions with the luciferase complex which is responsible for the luminescence. 62The cell wall limits the uptake and further toxic effects of chemicals.In the Gramnegative bacteria, the cell wall is characterized by the presence of an external membrane, exhibiting negatively charge lipopolysaccharides in the surface.This external membrane, which is not present in Gram-positive bacteria, has been assumed as the main responsible for the decreased sensitivity of Gram-negative bacteria to chemicals, compared to Grampositive bacteria. 55,63he complex interactions of ZIs with the components of the cell wall allied to chemical interactions with the medium 10 might explain the lack of obvious trends in the toxicity of ZIs to A. fischeri.The non-hazardous character of the tested ZIs to the bacterium and the macroalga are quite promising in supporting of their environmentally friendliness, which allied to their interesting properties, contributes to the attractiveness of these compounds.However, to address the limited knowledge on their ecotoxicological assessment, future studies should address their chronic toxicity, biodegradability and potential for bioaccumulation.

Conclusions
The ecotoxicity of sixteen hydrophilic ZIs composed of cationic and anionic groups commonly found in ionic liquids, was here determined considering the microalga R. subcapitata and the bacterium A. fischeri.The studied ZIs were shown to present a harmless character to both tested aquatic species, with EC50 values above 100 mg•L -1 .Their EC50 values to the freshwater microalgae R. subcapitata could be related to the chemical structure through the featured hydrophobicity (most pronounced for non-aromatic ZIs), following the general heuristic rule of increased toxicity with increased hydrophobicity.However, no general consistent relationships were found for the marine bacteria A. fischeri, suggesting that the chemical structure may have no relevant toxicity repercussions depending on the test model.Furthermore, despite the structural similarity between the studied ZIs and some ionic liquids, the heuristic rules commonly used to explain the ecotoxicological behavior of most ionic liquids, are not able to explain the behavior of all ZIs.When compared with their ionic liquid counterparts, the ZIs studied present a more benign character, supporting the idea that these compounds may become attractive in several chemical and engineering applications.Nevertheless, a deeper understanding of the ZIs toxic behavior and underlying mechanisms is required for their more comprehensive ecotoxicological profiling.This effort will certainly allow the establishment of a more robust basis for the rational design of sustainable ZIs.

Fig. 2 .
Fig.2.Median effective concentration (EC50), in g•L -1 , obtained for the microalga R. subcapitata after 96 h of exposure to several zwitterionic compounds (blue bars), ordered by increasing toxicity.The results for betaine (determined in this work), and some ionic liquids of similar structure27,[32][33][34] (green bars) (cf.ESI) and organic solvents 32 (orange bars) are also represented.The error bars represent standard error.

i 7 Fig. 4 .
Fig. 4.Relationship between the logarithmic function of the octanol-water partition coefficient of ZIs (expressed as log (KOW)) and the corresponding toxicity (expressed as log (EC50-96h), g•L -1 ) for R. subcapitata.Aromatic ZIs correspond to the orange symbols and non-aromatic ZIs correspond to the blue symbols.The correlation represented by the black straight line (statistical summary presented below) refers only to the latter.

10 Fig. 5 .
Fig.5.Median effective concentration (EC50), in g•L -1 , obtained after 30 min of exposure of the bacteria A. fischeri to each ZI tested, ordered by the level of toxicity.The results for betaine (determined in this work), some ionic liquids of similar structure33,44,[46][47][48][49] (cf.ESI) and organic solvents50,51 -some considering 15 min of exposure -are also presented.The error bars represent standard error.

Table 1 .
Name, acronym, and properties − molecular weight (Mw) and logarithmic function of the octanol-water partition coefficient (KOW) − of the studied ZIs.