Compound-Specific Chlorine Isotope Fractionation in Biodegradation of Atrazine †

two Figures and one Table illustrating the GC-qMS method optimization for chlorine analysis, one Table illustrating the method comparison of the GC-qMS for chlorine analysis, one Figure and one Table considering H-abstraction during chlorine CSIA, two Figures illustrating the results of HPLC concentration analysis. ABSTRACT Atrazine is a frequently detected groundwater contaminant. It can be microbially degraded by oxidative dealkylation or by hydrolytic dechlorination. Compound-specific isotope analysis is a powerful tool to assess its transformation. In previous work, carbon and nitrogen isotope effects were found to reflect these different transformation pathways. However, chlorine isotope fractionation could be a particularly sensitive indicator of natural transformation since chlorine isotope effects are fully represented in the molecular average while carbon and nitrogen isotope effects are diluted by non-reacting atoms. Therefore, this study explored chlorine isotope effects during atrazine hydrolysis with Arthrobacter aurescens TC1 and oxidative dealkylation with Rhodococcus sp. NI86/21. Dual element isotope slopes of chlorine vs. carbon isotope fractionation (Λ Arthro Cl/C = 1.7 ± 0.9 vs. Λ Rhodo Cl/C = 0.6 ± 0.1) and chlorine vs. nitrogen isotope fractionation (Λ Arthro Cl/N = -1.2 ± 0.7 vs. Λ Rhodo Cl/N = 0.4 ± 0.2) provided reliable indicators of different pathways. Observed chlorine isotope effects in oxidative dealkylation (ɛ Cl = -4.3 ± 1.8 ‰) were surprisingly large, whereas in hydrolysis (ɛ Cl = -1.4 ± 0.6 ‰) they were small, indicating that C-Cl bond cleavage was not the rate-determining step. This demonstrates the importance of constraining expected isotope effects of new elements


INTRODUCTION
The herbicide atrazine has been used in agriculture to inhibit growth of broadleaf and grassy weeds 1 . In the U.S. atrazine was the second most commonly used herbicide in 2012 and is still in use today 2 . In the European Union atrazine was banned in 2004 3 , but together with its metabolites it is still frequently detected at high concentrations in groundwater 4,5 . The massive and widespread use has led to a wide-ranging presence of atrazine in the environment, which can have harmful effects on living organisms and humans 6 . Therefore, the environmental fate of atrazine is of significant concern and much attention has been directed at detecting and enhancing its natural biodegradation. However, assessing microbial degradation of atrazine in the environment is challenging with conventional methods like concentration analysis. Sorption and remobilization of the parent compound and its metabolites, as well as further transformation of the metabolites inevitably lead to fluctuations in concentrations 7-10 , which make it difficult to assess the net extent of atrazine degradation in the field.
In recent years, compound-specific isotope analysis (CSIA) has been proposed as an alternative approach to detect and quantify the degradation of atrazine [11][12][13] .
In contrast to, and complementary to traditional methods, CSIA informs about transformation without the need to detect metabolites. The reason is that during (bio)chemical transformations molecules with heavy isotopes are typically enriched in the remaining substrate since their bonds are more stable and, therefore, usually react slower than molecules containing light isotopes (normal kinetic isotope effect). The ratios of heavy to light isotopes (e.g. 13 C/ 12 C for carbon) in the remaining substrate, therefore, change during transformation. Observing such changes can be used as direct (and concentration-independent) indicator of degradation 14,15 .
Isotope values are reported in the δ-notation relative to an international reference material, e.g.
for carbon 14,15 : δ 13 C = [( 13 C/ 12 C)Sample -( 13 C/ 12 C)Reference] / ( 13 C/ 12 C) Reference (1) The magnitude of the degradation-induced isotope fractionation depends on different factors, which can make isotope ratios of specific elements particularly attractive to observe degradationinduced isotope fractionation. To this end, first, an element needs to be directly involved in the (bio)chemical reaction. For example, a carbon isotope effect would be quite generally expected in organic molecules, whereas a chlorine isotope effect would be primarily expected if a C-Cl bond is cleaved. Second, isotope fractionation depends on the underlying kinetic isotope effect (see above), but also on the extent to which this effect is represented in the molecular average isotope fractionation described by the enrichment factor ɛ (see below). Atrazine, for example, contains only one chlorine atom but eight carbon and five nitrogen atoms. Hence, chlorine isotope effects at the reacting position are fully represented in the molecular average, whereas position-specific carbon and nitrogen isotope effects are diluted by non-reacting atoms 14, 15 .
Most of the publications studying the chemical mechanisms of abiotic and microbial atrazine degradation have focused on the analysis of carbon ( 13 C/ 12 C) and nitrogen ( 15 N/ 14 N) isotope fractionation. Thereby, ɛ-values in the range of -5.4 ‰ to -1.8 ‰ for carbon and -1.9 ‰ to 3.3 ‰ for nitrogen were observed 9, 10, 16, 17 . Chlorine isotope effects for microbial atrazine degradation were so far not reported due to analytical challenges 18 : Until recently 19, 20 , suitable methods were not available for chlorine isotope analysis of atrazine. However, from the magnitude of chlorine isotope effects observed for dechlorination of trichloroethenes (-5.7 ‰ to -3.3 ‰, where intrinsic isotope effects are diluted by a factor of three 21 ), very large chlorine enrichment factors ɛCl (-8 ‰ to -10 ‰ or even larger) could potentially occur for a C-Cl bond cleavage in atrazine. For example, enzymatic hydrolysis of the structural homologue ametryn (atrazine structure with a -SCH3 instead of a -Cl group) yielded a sulfur isotope enrichment factor ɛS of -14.7 ‰ ± 1.0 ‰ 17 .
If the cleavage of carbon-chlorine bonds is involved in the rate-determining step of a (bio)transformation, chlorine isotope effects could, therefore, enable a particularly sensitive detection of natural transformation processes by compound-specific (i.e., molecular average) isotope analysis. NI86/21 (see Figure 1). A. aurescens TC1 was directly isolated from an atrazine-contaminated soil 24 . By expressing the enzyme TrzN, it is capable of performing hydrolysis of atrazine 24, 25 .
Rhodococcus sp. NI86/21 uses a cytochrome P450 system for catalyzing oxidative dealkylation of atrazine 26 . For these two pathways, carbon isotope fractionation was very similar, but significant differences were observed in nitrogen isotope effects 9, 10, 16, 17 . Plotting the changes of isotope ratios of these two elements relative to each other results in the regression slope Λ for carbon and nitrogen 27,28 ΛC/N = Δδ 15 N / Δδ 13 C ≈ ɛN / ɛC Hence, dual element (C, N) isotope trends for oxidative dealkylation of atrazine with Rhodococcus sp. NI86/21 (Λ Rhodo C/N = 0.4 ± 0.1) 16 were significantly different compared to hydrolysis with A. aurescens TC1 (Λ Arthro C/N = -0.6 ± 0.1) 9 offering an opportunity to distinguish atrazine degradation pathways in the field. However, in environmental assessments it is advantageous to have isotopic information of as many elements as possible in order to distinguish degradation pathways and sources at the same time 29-31 . Therefore, information from a third element, chlorine, would be highly valuable. Also on the mechanistic end, information gained from a change in the chlorine isotope value could lead to a more reliable differentiation of transformation pathways and contribute to a better mechanistic understanding of the underlying chemical reaction 31 . Along these lines, triple element (3D) isotope analysis was already accomplished for chlorinated alkanes 31, 32 and alkenes 33,34 .
Until now, however, compound-specific chlorine isotope analysis has not been accessible so that chlorine isotope ratio changes for hydrolysis of atrazine have only been analyzed in abiotic systems or via computational calculations 35,36 . For oxidative dealkylation, chlorine isotope effects have, so far, not been studied. Recently a GC-qMS method for chlorine isotope analysis of atrazine has been brought forward 20 which offers the opportunity to enable deeper mechanistic insights into its transformation processes. Therefore, our objective was to analyze carbon, nitrogen and chlorine isotope effects associated with the biodegradation of atrazine via hydrolysis with A. aurescens TC1 and via oxidative dealkylation with Rhodococcus sp. NI86/21. In addition, we computationally predicted the chlorine isotope effect associated with hydrolysis and oxidative dealkylation for comparison. Further, we evaluated whether the additional information from chlorine isotope fractionation is a particularly sensitive indicator for transformation processes and whether it can confirm previously proposed mechanisms of different pathways. With this study, we bring forward information about degradation-induced chlorine isotope fractionation of atrazine as a basis to apply triple element (3D) isotope analysis in environmental assessments.

MATERIAL & METHODS
Bacterial strains and cultivation. A. aurescens strain TC1 was grown in mineral salt medium supplemented with approx. 20 mg/L of atrazine according to the protocol of Meyer et al. 9 Likewise, Rhodococcus sp. strain NI86/21 was cultivated in autoclaved nutrient broth (8 g/L, Difco TM ) with approx. 20 mg/L of atrazine according to the protocol of Meyer et al. 16 . In the lateexponential growth phase the strains were harvested via centrifugation (4000 rpm, 15 min). For the start of the degradation experiments, cell pellets of each strain were transferred to 400 mL fresh media and atrazine was added to achieve a starting concentration of 20 mg/L. All experiments were performed in triplicate at 21 °C on a shaker at 150 rpm. Control experiments, which were performed without the bacterial strains, did not show any degradation of atrazine.
Concentration measurements via HPLC. The process of atrazine degradation was monitored by concentration measurements. For analysis, 1 mL samples were taken and filtered through a 0.22 µm filter. Atrazine and its degradation products were directly analyzed using a Shimadzu UHPLC-20A system, which was equipped with an ODS column 30 (Ultracarb 5 μM, 150 × 4.6 mm, Phenomenex). After sample injection (10 µL) an adequate gradient program (see SI) was used for compound separation. The oven temperature was set to 45 °C and the compounds were detected by their UV absorbance at 222 nm. Quantitation was performed by the software "Lab Solutions" based on internal calibration curves.
Preparation of samples for isotope analysis. According to the protocol of Meyer et al. 9 between 10 and 260 mL of sample were taken for isotope analysis of atrazine at every sampling event. After centrifugation (15 min, 4000 rpm) the supernatant was collected in a new vial.
Subsequently, samples were extracted by adding dichloromethane (5-130 mL) and shaking the vial for at least 20 min. The sample extracts were dried at room temperature under the fume hood.
Afterwards, the dried extracts were dissolved in ethyl acetate to a final atrazine concentration of approx. 200 mg/L.
Isotope analysis of carbon and nitrogen. The protocol for isotope analysis of carbon and nitrogen was adapted from the studies of Meyer et al. 9,16 . A TRACE GC Ultra gas chromatograph hyphenated with a GC-III combustion interface and coupled to a Finnigan MAT253 isotope ratio mass spectrometer (GC-C-IRMS, all Thermo Fisher Scientific) was used. Each sample was analyzed in triplicate. Sample injection (2-3 µL) was performed by a Combi-PAL autosampler (CTC Analysis). The injector had a constant temperature of 220 °C, was equipped with an "A" type packed liner for large volume injections (GL Sciences) and was operated for 1 min in splitless and then in split mode (split ratio 1:10) with a flow rate of 1.4 mL/min. For peak separation, the GC oven was equipped with a DB-5 MS column (30 m × 0.25 mm, 1 µm film thickness, Agilent).
The temperature program of the oven started at 65 °C (held for 1 min), ramped at 20 °C/min to 180 °C (held for 10 min) and ramped again at 15 °C/min to 230 °C (held for 8 min). In the combustion interface, a GC Isolink II reactor (Thermo Fisher Scientific) was installed, which was operated at a temperature of 1000 °C. After combustion of the analytes to CO2 and subsequent reduction of any nitrogen oxides, the compounds were analyzed as CO2 for carbon and N2 for nitrogen isotope measurements. Three pulse of CO2 or N2, respectively, were introduced at the beginning and at the end of each run as monitoring gas. Beforehand, these monitoring gases were calibrated against RM8563 (CO2) and NSVEC (N2), which were supplied by the International Here δ 37 Cl0 refers to the chlorine isotope value at the starting point (t = 0) of an experiment.
Regression slopes Λ of dual element isotope plots were obtained by plotting isotope ratios of two different elements against each other, e.g. carbon vs. nitrogen (see eq. 2). The uncertainties of the at 300 K. The tunneling contributions to the overall kinetic isotope effect were omitted.

RESULTS & DISCUSSION
Observation of normal chlorine isotope effects in biotic hydrolysis and oxidative dealkylation. Atrazine degradation by A. aurescens TC1 resulted in the metabolite 2hydroxyatrazine, whereas the metabolites DEA and DIA were observed for Rhodococcus sp.
NI86/21 (see Figure S4 and S5 in the SI). Detection of these expected degradation products ( Figure 1) demonstrates that hydrolysis and oxidative dealkylation were the underlying biochemical reactions during atrazine degradation, respectively. In both biodegradation experiments -biotic hydrolysis with A. aurescens TC1 and oxidative dealkylation with Rhodococcus sp. NI86/21 -normal chlorine isotope fractionation was observed (see Figure 2A).

Evaluation of δ 37 Cl values during biotic hydrolysis according to Equation 3 resulted in a small
normal isotope effect of ɛCl = -1.4 ± 0.6 ‰. In oxidative dealkylation with Rhodococcus sp.
NI86/21 approx. 90 % degradation was reached after approx. 186 h in all three replicates (see SI, Figure S5). Evaluation of changes in chlorine isotope ratios resulted in a surprisingly large normal isotope effect of ɛCl = -4.3 ± 1.8 ‰ considering that the C-Cl bond is not broken during the reaction (see Figure 1). In a next step, carbon and nitrogen isotope effects were therefore analyzed to confirm whether the same reactions mechanisms are at work as observed in previous studies 9, 16 . Observed carbon and nitrogen isotope fractionation is consistent with previous studies.

Carbon and nitrogen isotope fractionation for atrazine degradation by A. aurescens TC1 and
Rhodococcus sp. NI86/21 was consistent with previous studies: Both experiments showed significant changes in isotope ratios (see Figure 2B and C). For hydrolysis with A. aurescens TC1, an inverse nitrogen isotope effect (ɛN = 2.3 ± 0.3 ‰) and a normal carbon isotope effect (ɛC = -3.7 ± 0.4 ‰) were observed, which were slightly smaller compared to the results of a former publication of Meyer et al. (ɛN = 3.3 ± 0.4 ‰, ɛC = -5.4 ± 0.6 ‰) 9 , but gave the same dual element isotope plot (Λ Arthro C/N = -0.6 ± 0.1) confirming that the same mechanism was at work (see Figure 3A).  Multi-element isotope approach. Results of chlorine isotope analysis were combined with data for carbon and nitrogen isotope measurements in dual element isotope plots (see Figure 3B and C). For hydrolysis with A. aurescens TC1 regression slopes of Λ Arthro Cl/C = 1.7 ± 0.9 and Λ Arthro Cl/N = -1.2 ± 0.7 were obtained. Oxidative dealkylation by Rhodococcus sp. NI86/21 resulted in regression slopes of Λ Rhodo Cl/C = 0.6 ± 0.1 and Λ Rhodo Cl/N = 0.4 ± 0.2. Since the dual element isotope plots of chlorine and carbon and of chlorine and nitrogen provide significantly different regression slopes for the respective elements, they can provide an additional line of evidence to differentiate the two degradation mechanisms of atrazine from each other.

Surprising mechanistic evidence from chlorine isotope effects. For degradation with
A. aurescens TC1, rather small chlorine isotope fractionation was observed (ɛCl = -1.4 ± 0.6 ‰) despite the fact that the chlorine is cleaved off during hydrolysis (see Figure 1). For oxidative dealkylation with Rhodococcus sp. NI86/21, the chlorine is not cleaved off (see Figure 1), therefore, no or just a small chlorine isotope effect was expected. However, here more pronounced chlorine isotope fractionation was observed (ɛCl = -4.3 ± 1.8 ‰).
The corresponding apparent kinetic isotope effects (AKIECl, see Table 1)   For microbial hydrolysis of atrazine an experimental AKIE Arthro Cl value of 1.0014 ± 0.0006 was calculated (see Table 1). Dybala-Defratyka et al. 35 reported a more pronounced AKIE alk.hydrol. Cl value of 1.0069 ± 0.0005 (see Table 1). However, that study 35 was conducted in an abiotic alkaline solution at 21 °C so that another hydrolysis pathway was involved. Newer data reported a much smaller value of AKIE alk.hydrol. Cl = 1.0009 ± 0.0006 36 for the same alkaline hydrolysis at 50 °C.
Later on it was confirmed that abiotic alkaline hydrolysis performed earlier at 21 °C resembles rather neutral than alkaline conditions 36 . Table 2 Table 2). However, hydrolysis at neutral pH is too slow to be of relevance. Computational calculations taking into account the transition state structures at a molecular level predicted AKIECl values ranging from 0.9996 to 1.0014 for alkaline, acidic and enzymatic hydrolysis (see Table 1 and 2) 36, 44 . Hence, on the mechanistic level, the computational studies predict that the formation of a Meisenheimer complex rather than the subsequent C-Cl bond cleavage is the rate-determining step during the nucleophilic aromatic substitution reaction catalyzed by TrzN 36, 44 . In both abiotic pathways the C-Cl bond at the transition state of the rate determining step is almost intact giving rise to very small AKIECl (the computed bond orders for both alkaline and acidic hydrolysis are the same and equal to 1.03, see also Table 2). In this study, we therefore observed a similarly small AKIE Arthro Cl value for enzymatic hydrolysis in A. aurescens TC1 which resembles acid-catalyzed hydrolysis rather than alkaline hydrolysis 9 . Hence, a consistent picture emerges that different hydrolytic pathways give rise to experimental AKIECl values much lower than the Streitwieser Limit of 1.02 [41][42][43] indicating that the chlorine isotope effect is masked in all cases and that the C-Cl bond cleavage is not the rate-determining step. Interestingly, this is in contrast to ametryn hydrolysis where strong sulphur isotope effects were observed in enzymatic hydrolysis by TrzN 17 . In conclusion, since chlorine isotope effects were found to be masked, information from chlorine isotope analysis alone would not be enough to differentiate the different reaction mechanisms. This illustrates the importance of analyzing more than one element for mechanistic differentiation.  Meyer et al. 16 concluded that oxidative dealkylation of atrazine with 6 Rhodococcus sp. NI86/21 is initiated by hydrogen atom transfer based on the observed product 7 distribution and the carbon and nitrogen isotope effects. Hydrogen atom transfer leads directly to 8 a homolytic cleavage of the C-H bond adjacent to the nitrogen atom (α-position of the ethyl or 9 isopropyl group) producing a relative unstable 1,1-aminoalcohol which is then further transformed 10 to DEA or DIA 16 . In parallel, two additional products could be detected which were formed by 11 oxidation of the C-H bond in the β-position of the ethyl or isopropyl group. For this mechanistic 12 pathway, chlorine isotope effects would be expected to be rather small since the chlorine is not 13 involved in the reaction steps. The closed mass balance of the concentration analysis (see 14 Figure S5 in the SI) of this study and the results of product distribution of Meyer et al. 16 also 15 indicate that there is no C-Cl bond cleavage taking place since corresponding hydrolysis products 16 were not detected. Furthermore, our computations for hydrogen atom transfer at a catalytic center 17 mimicking cytochrome P450 predicted no chlorine isotope effect (AKIE hydro.atom trans. Cl = 0.9999, 18 see Table 1). Hydride transfer promoted by the hydronium ion also resulted in no chlorine isotope 19 effect (AKIE hydride trans. Cl = 0.9997, see Table 1). At previously located transition state structures 20 for these two reactions 16 the carbon-chlorine bond remains intact and no stretching of this bond is 21 involved in the reaction coordinate (hydrogen transfer) mode. The observed more pronounced 22 AKIE Rhodo Cl value of 1.0043 ±0.0018 in this study (see Table 1) could, therefore, be indicative of 23 isotope effects caused by enzymatic interactions. Meyer et al. 16 proposed that for oxidative 24 dealkylation no selectivity itself is observed, however, the preferred oxidation of the α-position 25 over the β-position could be explained by steric factors of the catalyzing enzyme which could have 26 an influence on the transformation pathway. Thus, the sensitive chlorine isotope effect, which is 27 observed even though the C-Cl bond is not cleaved during degradation, can be interpreted as an 28 indicator that non-covalent interactions between the cytochrome P450 complex and the chlorine 29 cause significant chlorine isotope fractionation 45 . 30

CONCLUSION 31
Since atrazine is frequently detected in groundwater systems, major efforts should be put into 32 understanding its environmental fate. We provide an approach to 3D-isotope (C, N, Cl) analysis 33 of atrazine and explored isotope fractionation in different transformation pathways. Together, this 34 provides the basis to more confidently assess sources and degradation of atrazine in the 35 environment. Specifically, we demonstrated that pronounced changes in chlorine isotope values 36 are not an indicator of microbial hydrolysis (as one might have expected without knowledge of 37 our experimental data), but -surprisingly -rather of oxidative dealkylation. Therefore, although 38 trends are different than expected, they can nonetheless be used for a more confident identification 39 of different sources and transformation pathways in field samples. Regarding the sensitivity of 40 chlorine isotope effects, our study demonstrates the importance of performing controlled 41 laboratory experiments before applying the approach in the field. Specifically, in other cases 42 chlorine isotope fractionation can be much more pronounced than observed for atrazine in this 43 study. Large chlorine isotope effects were observed in proof-of-principle experiments by Ponsin 44 et al. 20 studying hydrolytic dechlorination of S-metolachlor, an herbicide containing also only one 45 chlorine atom. Here preliminary data suggest a large chlorine isotope effect of ɛCl = -9.7 ± 2.9 ‰ 46 for abiotic alkaline hydrolysis. Therefore, in the case of other substances chlorine isotope effects 47 can be even more sensitive indicators of degradation provided that the C-Cl bond cleavage occurs 48 in the rate-determining step of a reaction. Further, gaining deeper insights into these chemical 49 processes is the basis for understanding the biotic catalysis of organic micropollutant degradation. 50 This, in turn, is essential for identifying and developing optimized strategies for micropollutant 51 removal in order to make successful bioremediation possible. 52

CONFLICT OF INTEREST 53
There are no conflicts to declare.