Fumigant methyl iodide can methylate inorganic mercury species in natural waters

Abstract

Methyl iodide or iodomethane (CH₃ I) has recently been registered as a fumigant in many countries, although its environmental impacts are not well understood. Here we report the results of a study on the methylation of mercury by CH₃ I in natural water by incubation experiments using both Hg (¹⁹⁹ HgCl₂ and )- and hydrogen (CD₃ I)-stable isotope addition techniques. We find that methylation of Hg⁰ , and Hg²⁺ by CH₃ I can occur in natural water under sunlight, while only Hg⁰ and can be methylated in deionized water. We propose that the methylation of Hg by CH₃ I in natural waters is mediated by sunlight and involves two steps, the reduction of Hg²⁺ to Hg⁰ / and the subsequent methylation of Hg⁰ / by CH₃ I. Further quantitative assessment suggests that CH₃ I-involved methylation of inorganic Hg could be an important source of CH₃ Hg⁺ in an environment where CH₃ I has been used in large amounts as a fumigant.

ARTICLE

Received 16 Mar 2014 | Accepted 9 Jul 2014 | Published 19 Aug 2014

DOI: 10.1038/ncomms5633

Fumigant methyl iodide can methylate inorganic mercury species in natural waters

Yongguang Yin1,2,3, Yanbin Li4, Chao Tai2,5, Yong Cai2,3 & Guibin Jiang1

Methyl iodide or iodomethane (CH3I) has recently been registered as a fumigant in many countries, although its environmental impacts are not well understood. Here we report the results of a study on the methylation of mercury by CH3I in natural water by incubation experiments using both Hg (199HgCl2 and CH3 201Hg )- and hydrogen (CD3I)-stable isotope addition techniques. We nd that methylation of Hg0, Hg22 and Hg2 by CH3I can occur in natural water under sunlight, while only Hg0 and Hg22 can be methylated in deionized water.

We propose that the methylation of Hg by CH3I in natural waters is mediated by sunlight and involves two steps, the reduction of Hg2 to Hg0/Hg22 and the subsequent methylation of

Hg0/Hg22 by CH3I. Further quantitative assessment suggests that CH3I-involved methylation of inorganic Hg could be an important source of CH3Hg in an environment where CH3I has been used in large amounts as a fumigant.

1 State Key Laboratory of Environmental Chemistry and Ecotoxicology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, No. 18, Shuangqing Road, Haidian, Beijing 100085, China. 2 Department of Chemistry and Biochemistry, Florida International University, University Park, Miami, Florida 33199, USA. 3 Southeast Environmental Research Center, Florida International University, University Park, Miami, Florida 33199, USA. 4 Key Laboratory of Marine Chemistry Theory and Technology, Ministry of Education, Ocean University of China, Qingdao 266100, China. 5 Institute of Resources and Environment, Henan Polytechnic University, Jiaozuo 454000, China. Correspondence and requests for materials should be addressed to Y.C.(email: mailto:[email protected]

Web End [email protected] ) or to G.J. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633

Methyl iodide or iodomethane (CH3I) is one of the most abundant organoiodine compounds in natural environments1. Anthropogenic input of CH3I to the

environment has been rapidly increasing since the 1990s, after it was suggested as a replacement for the ozone-depleting fumigant methyl bromide (CH3Br). CH3I is preferred over CH3Br because it is similar to, or more effective than, CH3Br in controlling a wide variety of soil pests and weeds2, and it has less impact on the ozone layer owing to its rapid photodegradation in the troposphere2. The U.S. EPA approved the use of CH3I as a fumigant in 2008 (ref. 3). A commercially available CH3I

formulation was then registered in many countries, such as the USA, Japan, and New Zealand.

The potentially negative impact of using CH3I has not been adequately addressed4, despite its increasing use on farmland. Compared with CH3Br, CH3I degrades more slowly in soil, hence increasing its chance of being transported to an aquatic environment via runoff5,6. CH3I is known to have the capability of methylating several metals and metalloids, including Sn2,

As3, Pb2 and Hg0, following an oxidative addition mechanism710. Among these metals/metalloids, mercury deserves special attention because its methylation product, methylmercury (CH3Hg), is among the most widespread of all highly toxic contaminants. Previous work has demonstrated that CH3I can be used as a methylation reagent for synthesizing

CH3Hg from Hg0 in a laboratory setting10. However, such a methylation reaction has not been conrmed in natural waters with low concentrations of Hg0 and CH3I (ref. 11). The undetectable CH3Hg in the previous study could be attributed to the possible degradation of CH3Hg, which was not monitored during the experiment11. It is understandable that direct methylation of Hg2 by CH3I has not been observed1214, as this reaction is theoretically impossible according to the mechanism of the oxidative addition reaction. However, recent studies have indicated that a rapid conversion of Hg2 to Hg0 can occur in natural waters15,16, suggesting that Hg2 could also play an important role in the CH3I-involved methylation of inorganic mercury in the environment.

In this study, we show that CH3I has the potential of methylating inorganic mercury in aquatic environments under sunlight. We used both mercury (199HgCl2 and CH3 201Hg )-and hydrogen (CD3I)-stable isotopes to examine the possible methylation of inorganic mercury species by CH3I in natural water. Various concentrations of 199HgCl2, Hg2(NO3)2, Hg0 and CH3I were added to deionized water or natural water and incubated with or without sunlight. CH3 201Hg , spiked in the

incubation media, was used as a control to correct for the degradation of produced CH3Hg. CD3I was used to validate the source of the methyl group in CH3Hg. Our results demonstrate that methylation of inorganic mercury by CH3I

could be an important source of CH3Hg in CH3I-contaminated surface water.

ResultsMethylation of inorganic mercury by CH3I in natural water.

Natural water samples were collected from a pond located at the Florida International University (FIU). Methylation experiments were carried out in a closed 0.5-l Teon bottle under natural sunlight or in darkness. The methylation of inorganic mercury species (Hg2, Hg22 and Hg0) by CH3I was measured (Fig. 1).

Signicant production of CH3Hg was observed in waters exposed to sunlight for all tested mercury species. For samples tested in the dark, no signicant formation of CH3Hg occurred for Hg2 and Hg0 (P40.1) (Fig. 1a,c), and a 90% reduction in the formation of CH3Hg was observed for Hg22 (Po0.05)

when compared with that under sunlight (Fig. 1b). Most previous studies also showed that sunlight is necessary for the methylation of Hg0 by CH3I10,11,17, although contrast results were also observed18. In sunlight, the methylation rates of these three mercury species (r, equation (9), see Methods) were in the order of Hg0 (1.670.09 pmol l 1 E 1 m2)4Hg22 (0.424

0.025 pmol l 1 E 1 m2)4Hg2 (0.1250.010 pmol l 1 E 1 m2). These results suggest that methylation of mercury by CH3I can

occur in natural water and this process is driven by sunlight. The 1-day methylation yield of Hg0 by CH3I was calculated to be0.6%, comparable with the results of Celo et al.18,19 and Hall et al.11 (1.1 and 0.29%, respectively). Although methylation rates of Hg0 and Hg22 were B13- and 4-fold higher than that of

Hg2, methylation of Hg2 by CH3I should be more important as Hg2 is the main inorganic mercury species in natural water. The formation of CH3 199Hg with the addition of 199Hg2 demonstrated that the generated CH3Hg was from the methylation of spiked 199Hg2 (Supplementary Fig. 1a). Further experiments were conducted to examine the origin of the methyl group of CH3Hg using a CD3I tracer method. When natural water containing Hg2 was spiked with CD3I and incubated under sunlight for 4 days, formation of CD3Hg was detected using gas chromatographymass chromatography (GCMS) (Supplementary Fig. 1b) (detailed description is given in Supplementary Note 1). Based on the oxidative addition theory, Hg2 cannot be directly methylated by CH3I. It is therefore

8

15.0

100

1 )

Hg2+

1 )

Hg22+

1 )

Hg0

CH 3Hg+ concentration (pmol l

CH 3Hg+ concentration (pmol l

12.5

80

CH 3Hg+ concentration (pmol l

6

10.0

60

4

7.5

40

5.0

2

20

2.5

0

0.0

0 1 2

Time (day)

3 4

0

0 1 2

Time (day)

3 4

0 1 2

Time (day)

3 4

Figure 1 | Methylation of inorganic Hg by CH3I in pond water. (a) Hg2, (b) Hg22 and (c), Hg0. Solid line and dashed line represent the methylation of inorganic Hg under sunlight and in the dark, respectively. CH3I was spiked into the pond water to form a nal concentration of 1 mmol l 1. The nal concentrations of Hg2, Hg22 and Hg0 were 4.99, 4.99 and 4.24 nmol l 1 (as Hg), respectively. The methylation rates for Hg2, Hg22 and Hg0were calculated to be 0.1250.010, 0.4240.025 and 1.670.09 pmol l 1 E 1 m2, respectively. All the tests were conducted in triplicate, and the error bars indicate the s.d. of three repeated measurements.

2 NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633 ARTICLE

proposed that the observed methylation of Hg2 in natural water may proceed through a two-step reaction mechanism, the reduction of Hg2 to Hg0 or Hg22 followed by the methylation of Hg0 or Hg22 by CH3I (Equations (15)).

Considering the electronegativity of the iodine atom, CH3I is expected to dissociate to CH3./I. or CH3/I , but not CH3 /I (ref. 20). CH3. and CH3 can then methylate Hg22 or Hg0.

According to the density functional theory, the methylation reaction of Hg0 and CH3I is thermodynamically favourable17. The requirement of sunlight for this reaction indicates the existence of a high activation barrier17. This proposed mechanism of Hg2 methylation is supported by the facts that 95.2 pmol l 1 Hg0 was detected in the pond water spiked with 4.99 nmol l 1 Hg2 (Supplementary Fig. 2) and that the methylation rate of Hg2 was only 7.5% and 29.5% compared with that of Hg0 and Hg22 (Fig. 1). The formation of Hg22 during Hg2 photoreduction in the presence of dissolved organic matter (DOM, Suwannee River humic acid) was also conrmed using Raman spectrometry (Supplementary Fig. 3). These results provide strong evidences to support the proposed two-step reaction mechanism of CH3I-involved methylation of mercury in natural water.

CH3I !

hv CH3 I 1

CH3I !

hv CH3 I 2

Hg2 natural water !

hv Hg0 or 12 Hg22 3

CH3

1

2 Hg22 ! CH3Hg 4

CH3 Hg0 ! CH3Hg 5 2I ! I2 6

4H 4I O2 ! 2I2 2H2O 7 To evaluate the importance of CH3Hg formed through CH3I-involved process in aquatic environments, additional experiments were performed in natural waters from the Florida Everglades. In these experiments, the methylation of Hg2 by CH3I was also observed (Supplementary Fig. 4), indicating that this process may widely occur in aquatic environments. Methylation rates (r) were calculated to be 0.0250.001 pmol l 1 E 1 m2 and

0.1250.010 pmol l 1 E 1 m2 for surface water of the Florida Everglades and FIU pond water, respectively.

Methylation of inorganic mercury by CH3I in deionized water. Experiments were conducted in deionized water, where factors (for example, DOM, Fe2) possibly facilitating the reduction of Hg2 are absent, to validate the aforementioned mechanism of photoinduced methylation of Hg2 by CH3I. The presence of

CH3Hg under sunlight was observed after 1 day of incubation when Hg22 or Hg0 was used as a precursor, while the methylation of Hg2 was not detected (Supplementary Fig. 5). The formation of CH3Hg was not observed for all three inorganic mercury precursors in the dark. These results demonstrate that Hg2 cannot be directly methylated by CH3I.

In deionized water, the experimentally generated CH3Hg became undetectable after 3 days (Supplementary Fig. 5). This loss of CH3Hg is not probably the results of photodegradation as previous studies showed that CH3Hg present in deionized water cannot be photodemethylated21,22. During the course of the experiments, rapid degradation of CH3Hg was always accompanied with a change in solution colour from clear to

brown. The brown substance was identied to be I2 based on the results of the following two experiments. A purple colour appeared in the organic phase when an extraction was performed with CH2Cl2 (ref. 23), indicating this substance could be I2 (Supplementary Fig. 6). The brown solution turned blue when starch was added, indicating the formation of I3 amylose complex. Furthermore, ultraviolet-visible spectra analysis showed that the brown substance in water and the purple substance in CH2Cl2 gave a maximum absorbance similar to that of I3 in water and I2 in CH2Cl2, conrming the formation of I2 (Supplementary Fig. 6). I2 may be produced from the combination of I or through the oxidation of I by dissolved

oxygen (equations 6 and 7)5. Thus, it was assumed that I2, originating from CH3I, caused the degradation of CH3Hg in deionized water. To test this hypothesis, I2 or I solution was added to deionized water containing CH3 201Hg . For trials with I2, immediate analysis of the solution indicated that CH3 201Hg disappeared completely (Supplementary Fig. 7a). An increase in the amount of 201Hg0 and 201Hg2 was also observed, conrming

the degradation of CH3 201Hg . However, the concentration of CH3 201Hg showed no signicant change in the presence of I (Supplementary Fig. 7b), indicating that I could not degrade CH3 201Hg .

In contrast to deionized water, the degradation of CH3 201Hg in natural water showed no signicant difference (P40.1) in the presence or absence of CH3I (Fig. 2). In addition, the water colour did not change to brown during the 5 days of incubation (Supplementary Fig. 6). It is proposed that any I2 formed would be quickly reduced to I in the presence of humic substances24, thus inhibiting the degradation of CH3Hg by I2. This assumption was partially supported by the fact that the rapid degradation of CH3 201Hg in the presence of CH3I was not observed when Suwannee River humic acid was added to deionized water (Supplementary Fig. 8). The degradation of CH3 201Hg in Suwannee River humic acid solution correlated with cumulative photosynthetic active radiation (PAR) photon ux (Supplementary Fig. 8), indicating that sunlight was the

CH 3201Hg+concentration (pmol l

8

7

1 )

6

5

4

3

2

1

0

0

1 2 3 4

Time (day)

Figure 2 | Degradation of CH3 201Hg in pond water. Solid line and dashed line represent the degradation of CH3Hg in the presence(1 mmol l 1) or absence of CH3I, respectively. The degradation rate of CH3 201Hg was calculated to be 0.0150.001 m2 E 1 in the presence of

CH3I and 0.0130.002 m2 E 1 without CH3I. The degradation of CH3Hg showed no signicant difference in the presence or absence of CH3I

(P40.1). All the tests were conducted in triplicate, and the error bars indicate the s.d. of three repeated measurements.

NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633

driver for the degradation of CH3 201Hg . The rapid degradation of CH3Hg by I2 would explain why the methylation of Hg0 by CH3I was not detected under sunlight in deionized water in a previous study11.

Pathway of methylation of inorganic mercury by CH3I. Based on the results obtained in deionized and natural waters, a pathway of photoinduced methylation of inorganic mercury by CH3I

in natural water was proposed (Fig. 3). Under the inuence of solar irradiation, Hg2 can be reduced to Hg0 or Hg22 (probably facilitated by DOM) in natural waters25, while CH3I can be decomposed to CH3 /I or CH3/I . CH3 radicals and CH3

ions can then methylate Hg22 or Hg0 through oxidative addition reactions. DOM may play an important role in this process. On one hand, DOM can be an effective scavenger of I2, therefore inhibiting the degradation of CH3Hg induced by I2. On the other hand, DOM may facilitate the reduction of Hg2 to Hg0 or Hg22 under sunlight25, subsequently enhancing the oxidative methyl transfer.

Assessment of CH3I-involved CH3Hg formation in environments. After being applied in farmland, CH3I could be transported into surrounding ponds or lakes via runoff water. The levels of CH3I in runoff water were estimated by laboratory column-leaching experiments (Supplementary Fig. 9). CH3I

applied to soils can be quickly leached out with concentration reaching 0.202.22 mmol l 1 depending on the soil type used (Supplementary Fig. 9), which is two to ve order of magnitude higher than its concentration in coastal waters26. The reaction of Hg2 and CH3I in pond water can be described as a pseudo-second-order reaction (Fig. 4). The photomethylation rate constant of Hg2 by CH3I (kM, see equation (13)) was then calculated to be (1.510.19) 10 2 l mol 1 E 1 m2 in pond

water. By using the obtained methylation rate constant and assuming that the increased CH3I concentration in aquatic water was in the range of 07 mmol l 1 (comparable to its concentration in leachates obtained in column experiments), the differences in CH3Hg concentration at steady state in

natural water with or without the presence of CH3I were estimated (equation (17)). Figure 5 showed that the increased steady-state concentration of CH3Hg is highly dependent on the concentration of CH3I and Hg2. If we assume that the concentration of CH3I entering natural water is 1 mmol l 1 (comparable with that obtained in leachates) and Hg2

concentration is 6.13265.4 pmol l 1 (literature reported

15

r(1014 mol l1 E1 m 2 )

10

5

0

2 4 6 8 10[Hg2+][CH3I](1012 mol2 l2)

Figure 4 | Relationship between the formation rate of CH3Hgand the product of Hg2 and CH3I concentrations in a pond water. To calculate the relationship between the formation rate of CH3Hg (r) and concentrations of Hg2 and CH3I, various concentration of CH3I (0, 0.1,0.2, 0.5, 1.0 and 2.0 mmol l 1) and 199HgCl2 (0, 1.0, 2.0, 3.0, 4.0 and5.0 nmol l 1) were prepared using pond water containing 1 ng l 1(5.0 pmol l 1) CH3 201Hg . The formation rate was observed to be positively related to the product of Hg2 and CH3I concentrations. Photomethylation rate constant of Hg2 by CH3I (kM, see equation (13) in main text) was calculated to be (1.510.19) 10 2 l mol 1 E 1 m2. All the

tests were conducted in triplicate, and the error bars indicate the s.d. of three repeated measurements.

1

0

CH3I

31.90

hv

28.71

CH3+I

30

25.52

Hg22+

22.33

25

19.14

hv

hv

20

+ nmol l1

15.95

Hg2+

I2

12.76

CH3Hg+

15

CH 3Hg

9.570

6.380

10

Hg0

3.190

O 2

DOM

5

0.000

8

7

hv

CH3++I

CH3I

CH 4 3 3 4

3 I mmol l

6 5 5 0

1 2 1 0 0 1

2

Hg2+ nmol l1

Figure 3 | Proposed pathway of photoinduced methylation of inorganic Hg by CH3I in natural water. The methylation of Hg by CH3I is proposed to include two steps, the reduction of Hg2 to Hg22 or Hg0 and the subsequent methylation of Hg22 or Hg0 by CH3. or CH3 (solid line), which are photo-decomposition products of CH3I (dashed line). I2, which is produced from the combination of I or oxidation of I by dissolved

oxygen, can rapidly degrade the generated CH3Hg. However, this process is negligible in natural waters as I2 can be quickly reduced to I by DOM.

Figure 5 | The effects of CH3I and Hg2 on the increased steady-state concentration of CH3Hg. The obtained photomethylation rate constant of Hg2 by CH3I (kM 1.51 10 2 l mol 1 E 1 m2) and photo-

demethylation rate constant of CH3Hg (kD 0.015 m2 E 1) in the pond

water were used to calculate the increased steady-state concentration of CH3Hg (equation (17)). The concentrations of Hg2 and CH3I are assumed to be in the range of 05 nmol l 1 and 07 mmol l 1, respectively.

4 NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633 ARTICLE

mercury levels in natural waters)27,28, the steady-state concentration of CH3Hg in natural waters was estimated to increase by 0.0052.966 pmol l 1 through the CH3I-involved methylation of mercury (see Supplementary Table 1). This increase in CH3Hg is comparable to the concentration of

CH3Hg in these natural waters (o0.0255.085 pmol l 1)27,28. To evaluate the relative importance of CH3I-involved methylation versus biotic methylation in sediments, we further compared the formation rate of CH3Hg from photomethylation by CH3I in water body and the diffusion rate of CH3Hg from sediment (generated by biotic methylation) with water (see Supplementary Table 2). The results showed that the calculated formation rates of CH3Hg from photomethylation by CH3I are in the range of 0.53258.5 pmol m 2 d 1, comparable with the diffusion rates of CH3Hg from sediment ( 3.0327.0 pmol m 2 d 1)

(Supplementary Table 2), indicating that CH3I-involved formation of CH3Hg could be comparable with that diffused from sediment (mainly generated by biotic methylation). Both estimations suggest that photomethylation of mercury by CH3I

could be an important source of CH3Hg in areas with CH3I fumigant application.

DiscussionMethylation of mercury by sulphate-reducing29 or iron-reducing bacteria30 has been deemed to be the main source of CH3Hg in aquatic environments. However, this study shows that methylation of inorganic mercury by CH3I could be another important source of

CH3Hg. This would be particularly true in an environment where CH3I is used in large amounts as a fumigant. In addition to acting as a methylation agent, CH3I can also serve as a potential mobilizing agent for mercury sulde (HgS)12. It has been demonstrated that CH3I could methylate sulphide into volatile (CH3)2S and (CH3)2S2, and then liberate free Hg2 into water from insoluble HgS12. Considering HgS is an important inorganic mercury species in anoxic environments such as soil31, sediment32 and sulphidic water33, this dissolution process of HgS could increase the leaching and release of Hg2 from soil and sediment into surface water, and therefore enhance the potential formation of CH3Hg. As CH3Hg is a neurotoxin, the impact of

CH3I application to farmland on biogeochemical cycling of mercury should be carefully evaluated.

With respect to the production of CH3Hg, most previous studies focused on the methylation of Hg2 in the environment; however, there is a lack of knowledge on the methylation of Hg0.

Results of this study imply that methylation of Hg0 can occur in the environment (for example, by CH3I). Although Hg0 is present at low concentrations in water and sediment, Hg2, through photomediated reduction in natural waters, provides a sufcient and continuous source of Hg0. Additionally, Hg0 is the main Hg species in the atmosphere, and the co-occurrence of Hg0 and released CH3I from fumigated elds by volatilization, makes the photochemical methylation of Hg0 possible in the atmosphere. Therefore, the methylation of Hg0 by CH3I and its contribution to the CH3Hg pool needs to be considered.

Fumigants are widely applied in agricultural elds to control pests and weeds. With the phase out of CH3Br, the use of CH3I is expected to increase dramatically in the future. CH3I is a

pulmonary and dermal irritant26, and is also suspected to be carcinogenic34, neurotoxic18,26 and endocrine disrupting4. This study shows that CH3I could also indirectly threaten the health of humans and wildlife by forming a toxic chemical, CH3Hg, suggesting the necessity of a more comprehensive risk assessment of CH3I use as a fumigant.

In summary, we have demonstrated that the registered fumigant CH3I could methylate inorganic Hg species, including

Hg0, Hg22 and Hg2, into toxic CH3Hg in natural water under sunlight. The photoinduced CH3Hg formation mechanism by CH3I is proposed, which involves the reduction of Hg2 to Hg0/Hg22 and the subsequent methylation of

Hg0/Hg22 by CH3I, following an oxidative addition reaction. Quantitative assessment suggests that methylation of inorganic

Hg by CH3I could be an important source of CH3Hg in CH3I-contaminated surface water.

Methods

Reagents. CH3Hg standard was purchased from Ultra Scientic (North Kingstown, RI). Enriched 201HgO and 199HgO were purchased from Oak Ridge National Laboratory (Oak Ridge, TN). CH3 201Hg was synthesized from enriched

201HgO using methylcobalamin35. Enriched 199HgCl2 solution was prepared by dissolving 199HgO in 10% HCl. Elemental mercury (Hg0) and Hg2(NO3)2 were purchased from Fisher Chemicals (Pittsburgh, PA). Hg0 stock solution was freshly prepared following the method by Whalin and Mason36. One drop of Hg0 was rst transferred into a piece of silicon tubing. After capping at both ends with Teon plugs, the tubing was immersed into deionized water. Next, the gaseous Hg0 was equilibrated with water for over 48 h under anoxic conditions. The concentration of Hg0 in water was determined before the experiment was conducted. CH3I and

CD3I were purchased from Fisher Chemicals and Hengye Zhongyuan Chemicals (Beijing, China), respectively. Deionized water was prepared from a Barnstead NANOpure Diamond Life Science (UV/UF) ultrapure water system (Barnstead International, Dubuque, IA).

Methylation of inorganic mercury by CH3I. Water samples were collected from a pond in FIU (25450N, 80220W) and the Florida Everglades (25450N, 80440W), and stored in pre-cleaned 2-L Teon bottles using trace metal clean procedures. The samples were kept in a cooler and transported to the laboratory within 3 h. Methylation experiments were conducted immediately on arrival at the laboratory. The methylation experiments were carried out in a 0.5-l Teon bottle under natural sunlight or in darkness (wrapping bottles with aluminum foil). 199HgCl2,

Hg2(NO3)2 or Hg0 was spiked into each bottle to form a nal concentration of about 1,000 ng l 1 (4.99 nmol l 1) as mercury, respectively. CH3I and CH3 201Hg were then added to each bottle to form a nal concentration of 141.9 mg l 1

(1 mmol l 1) and 1 ng l 1 (4.975 pmol l 1), separately. For the experiments conducted in deionized water, 199HgCl2, Hg2(NO3)2 or Hg0 was spiked to each bottle to form a nal concentration of about 1,000 ng l 1 (4.99 nmol l 1) as mercury, respectively. CH3I were then added to each bottle to form a nal concentration of 141.9 mg l 1 (1 mmol l 1). To investigate the effects of Hg2 and CH3I concentrations on the methylation rate, various concentration of CH3I (0, 0.1, 0.2, 0.5,1.0 and 2.0 mmol l 1) and 199HgCl2 (0, 1.0, 2.0, 3.0, 4.0 and 5.0 nmol l 1) were spiked into pond water containing 4.99 pmol l 1 CH3 201Hg . Bottles were incu

bated under ambient temperature and light conditions for 4 days. During the experiment, the intensity of ambient PAR was measured at 15min intervals routinely using LI-192 Quantum Sensor (LI-COR Biosciences, Lincoln, NE). Triplicates (three separate bottles) were employed for each trial. CH3 202Hg,

CH3 201Hg and CH3 199Hg in the incubated samples were determined after 0,0.5, 1, 2, 3 and 4 days of incubation using the aqueous phenylation purge-and-trap followed by the GC inductively coupled plasmaMS (ICP-MS) detection. A CD3I

addition technique was adopted to investigate the origin of the methyl group in the generated CH3Hg. HgCl2 and CD3I were added to 2 l of environmental water to form a nal concentration of 9.97 mmol l 1 as Hg2 and 20 mmol l 1, respectively. The bottle was then incubated under ambient temperature and light conditions for 4 days. The formation of CD3Hg was veried by the aqueous phenylationsolid phase microextraction (SPME)GCMS.

Analysis of CH3Hg. Water samples (20 ml), L-ascorbic acid (100 mmol l 1,1 ml), concentrated HCl (1 ml) and CH2Cl2 (5 ml) were added into a 50-ml polypropylene centrifuge tube (Corning Inc, Lowell, MA) sequentially. After shaking for 30 min, the samples were left to stand for 5 min to let CH2Cl2 separate from water. CH2Cl2 (2.5 ml) phase was then transferred to a 50-ml polypropylene centrifuge tube containing 35 ml 1% (v/v) HCl. The tube was purged with N2 at

150 ml min 1 to completely volatilize CH2Cl2 and release CH3Hg into aqueous phase. CH3Hg in the aqueous phase was analysed by aqueous phenylation and purge-and-trap pre-concentration followed by GCICPMS37. The articial formation of CH3Hg using this method has been demonstrated to be negligible, attributing to the omission of the distillation procedure and the replacement of the derivatization reagents (from sodium tetraethylborate or sodium (n-propyl)borate to sodium phenylborate (NaBPh4))37.

Aqueous phenylation reactions were performed in 300-ml glass bubblers from Exeter Scientic Glass Co. (Birdsboro, PA). The prepared sample (35 ml) was added to a 100 ml solution containing 30% (w/v) NaCl and 0.17% (w/v) NaOH. Two millilitres of citric buffer (pH 5.0, 2 mol l 1) and 1 ml of 1% (w/v) NaBPh4 (Fisher Chemicals, Fairlawn, NJ) were added. After 15 min of reaction, the phenylation products were purged from the bubbler at 45 C for 45 min with

NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633

200 ml min 1 of N2, and trapped on a Tenax trap. The trap was dried for 5 min by 200 ml min 1 of N2. Finally, the trapped mercury species were thermally desorbed and separated using a fused silica capillary column (15 m 0.53 mm i.d., coated

with DB-1 lm of 1.5-mm thickness) and then measured with an Elan DRC-e ICPMS (PerkinElmer Sciex, Waltham, MA).

PhenylationSPMEGCMS38 was used to validate the origin of the methyl group in produced CH3Hg. Natural waters (2,000 ml) were transferred to 2-l

Teon bottles and spiked with 4 mg Hg2 and 40 mmol CD3I. The solution was incubated under ambient temperature and sunlight for 4 days. The produced CD3Hg was extracted from each 40-ml water sample after adding 1 ml concentrated HCl and 5 ml CH2Cl2. The CH2Cl2 phase (4 ml) was transferred into a 100-ml polypropylene centrifuge tube containing 35 ml 1% HCl and purged with N2 at 150 ml min 1 to remove CH2Cl2. Then, pH was adjusted to 7 using 0.17%

(w/v) NaOH. Sample solution (35 ml), 2 mol l 1 HAc-NaAc buffer (5 ml, pH 4.9) and 15 g NaCl were added to a 100-ml glass vial sealed with a polytetrauoroethylene-coated septum. One millilitre 1% (w/v) NaBPh4 solution was injected into the vial using a syringe. The vial was stirred by a magnetic stirring bar and maintained at 55 C for 30 min in an oven. Next, headspace SPME extraction was carried out at 55 C for 10 min. After extraction, the bre was introduced into the GC injection port (200 C) for thermal desorption and GCMS analysis in splitless mode. The GC temperature programming is as following: 80 C for l min, increased to 280 C at a rate of 8 C min 1 and held there for 10 min.

Analysis of Hg0. Hg0 in water was purged and trapped onto a home-madegold trap and detected by atomic uorescence spectrometer. Hg2 solution was used as the calibration standard. Hg2 standard and 0.5 ml SnCl2 solution(10% (w/v)) were added to 100 ml 1% (v/v) HCl. The mixture was left to react for 15 min and then purged for 20 min with 400 ml min 1 of N2 at room temperature to let Hg0 completely volatilize and adsorb on the gold trap. The trap was dried for 5 min by 400 ml min 1 of N2. Finally, the trapped Hg0 was thermally desorbed and determined by a PS Analytical mercury speciation system (Orpington, Kent, UK). The analytical procedure for purgeable Hg0 in water samples was similar to that for the Hg2 standards, except that SnCl2 and HCl were not added.

Characterization of Hg22 using Raman spectrometry. Formation of Hg22 during the photoreduction of Hg2 was investigated using surface-enhanced

Raman spectrometry. Suwannee River humic acid and Hg2 were added into deionized water to form a nal concentration of 40 mg l 1 dissolved organic carbon and 0.5 mmol l 1 Hg2, respectively. After incubation under sunlight for 3 days, Raman-scattering spectra of Hg22 in the solution with and without the addition of KI (1.0 mol l 1) were collected using the following method. Au nanoparticle (AuNPs) solution was prepared by heating the solution containing1 mol l 1 AuCl4 and 5 mg l 1 dissolved organic matter (pH 6.8) at 80 C for24 h39. AuNP-coated Si wafer was then made by dropping AuNPs on Si wafer and then dried in a vacuum desiccator at room temperature. Next, water samples containing mercury species were dropped on the wafer and dried in a vacuum desiccator at room temperature. The AuNPs-enhanced Raman-scattering spectra of samples were obtained by using a Leica microscope in a confocal Raman spectroscopy system (Renishaw InVia Raman microscope, Britain). Spectra collections were carried out using a 50 objective lens and 60-s scan between

155.48 and 1,132.18 cm 1 with a 532-nm laser (5 mW) and 2,400 lines per mm grating.

Column experiments for evaluating the leachability of CH3I. The leachability of CH3I was investigated using columns (16 cm length 5 cm i.d.) containing

vegetable soil, sandy soil or quartz sand. To minimize the losses of soil during the experiment, a well-permeable quartz plate was placed at the bottom of the column and the top 3 cm of the columns was lled with quartz sand. The columns were pre-saturated with deionized water and then 27.36 mg CH3I (corresponding to an application rate of 125 lb ac 1) was added into the columns. The columns were then leached with deionized water at a rate of 4.1 ml min 1 for 3 h. Leachates (50 ml) were continuously collected at an interval of B12 min.

Leachates (10 ml) in 15-ml polypropylene centrifuge tube were extracted with CH2Cl2 (2 ml) (Corning Inc, Lowell, MA). After shaking for 30 min, the samples were left to stand for 5 min to let CH2Cl2 separate from water. One ml CH2Cl2 phase was collected for GCMS analysis.

The determination of CH3I was performed using an Agilent 7,890 gas chromatograph system, coupled to a quadruple Agilent 5,975 electron ionization mass spectrometric detector (Agilent Technologies, Palo Alto, CA, USA) equipped with a DB-624 fused silica capillary column (30 m 0.25 mm i.d., 1.4-mm lm

thickness). The gas chromatograph system was equipped with a split/splitless injection port operating in split mode (50:1) at 120 C. The injection volume was 2 ml. The oven temperature was programmed from 40 C (2 min) to 90 C at5 C min 1, and then to 240 C at 15 C min 1. The carrier gas was helium with a constant ow of 1 ml min 1. The mass spectrometer was operated in SIM mode (m/z 127 and 142). Solvent delay was set at 2 min. Quantication was conducted

using an external standard curve method.

Data analysis. The methylation rates of Hg0, Hg22 and Hg2 (r) were calculated by nonlinear regression of CH3Hg concentration CCH Hg t against t using

equation (9) (Origin, Version 6.0 for Windows, OriginLab Corp., Northampton, MA). The net production rate of CH3Hg can be described as the difference of the photomethylation rate of inorganic mercury by CH3I (r PAR) and the photo-

degradation rate of CH3Hg (kD CCH Hg PAR) at t time (equation (8)).

Next, the function of CCH Hg at t time could be obtained (equation (9)) by integrating equation (8). Degradation of CH3Hg can occur in natural waters and this process was deemed to be driven by sunlight21,40. In this study, the spiked

CH3 201Hg was also observed to be rapidly degraded under sunlight in both tested waters, while degradation of CH3 201Hg was not observed in the dark. The effect of CH3Hg photodegradation was taken into account in the calculation, represented by the photodegradation rate constant kD. A model based on the rst-order chemical kinetics was used to describe the photodegradation of CH3 201Hg

in water (equation (10))21,40. The rate constant of CH3 201Hg photodegradation, kD, was calculated by linear regression of lnCCH Hg t against t.

dCCH Hg

dt r PAR kD CCH Hg PAR 8

CCH Hg t

r r kD CCH Hg 0 e k J

kD kD40

CCH Hg t r J CCH Hg 0kD 0

dCCH Hg

dt kD CCH Hg PAR 10

lnCCH Hg t lnCCH Hg 0 kD J 11

where CCH Hg 0 and CCH Hg t are the concentrations of CH3Hg at the

beginning and t time, respectively (pmol l 1), kD is the rate constant of CH3Hg photodegradation (m2 E 1), PAR is the photosynthetically active radiation(E m 2 d 1), J is the cumulative PAR photon ux (E m 2) and r is the PAR normalized formation rate of CH3Hg (pmol l 1 E 1 m2).

The differences in CH3Hg concentrations at steady state with or without the presence of CH3I were calculated and used to evaluate the inuence of CH3I on CH3Hg production. In natural waters, the concentration of CH3Hg was assumed to be determined by several key processes, including photodegradation of CH3Hg, input of CH3Hg from sediment and runoff, methylation in water body and methylation by CH3I. Next, the changing rate of CH3Hg could be calculated by the photodegradation rate kD C PAR, the CH3Hg input rate from

sediment and runoff (R0), the methylation rate (R0) and the rate of photomethylation by CH3I (r PAR). As r was found to be positively related to the

product of CHg and CCH I (equation (13), Fig. 4) and equation (14) could be obtained by combining equation (13) with equation (12).

dC

dt R0 R0 r PAR kD C PAR 12

r kM CHg CCH

I

13

dC

dt R0 R0 kM CHg CCH

PAR kD C PAR 14 where R0 is the rate of CH3Hg input from sediment and runoff (pmol l 1 d 1),

R0 is the non-CH3I involved methylation rate in water body (pmol l 1 d 1), kM is the PAR-normalized formation rate constant of CH3Hg (l pmol 1 E 1 m2),CHg is the concentration of Hg2 (pmol l 1), CCH Iis the concentration of CH3I (pmol l 1). The predicted concentration of CH3Hg at steady state in the presence of CH3I (C CH H ) can be calculated by making the right side of equation (14) equal to zero (equation (15)).

C CH H

R0 R0 kM CHg CCH

I

PARkD PAR

15

The predicted concentration of CH3Hg at steady state in the absence of CH3I (CFCH H ) can be calculated by making CCH I of equation (15) equal to zero (equation (16)).

CFCH H

R0 R0kD PAR

16

Next, the differences in CH3Hg concentrations at steady state with or without the presence of CH3I (DC ) can be calculated by equation (17).

DC C CH H CFCH H

kM CHg CCH

I

I

kD 17

The parameters for calculating DC are given in Supplementary Table 1.

To evaluate the relative importance of CH3I-involved methylation versus biotic methylation, the CH3Hg photomethylation rate within the entire water column was calculated as follows. The sunlight intensity at Z depth (m) is calculated from the light intensity in the surface water (PAR (0)) by exponentially decreasing it with depth according to the Beer-Lambert equation. Combined with the calculation of the photomethylation rate of CH3Hg in the surface water, the photomethylation

6 NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5633 ARTICLE

rate at Z depth can be described as equation (18). Next, mercury methylation is integrated with respect to water column depth to obtain the methylation rate within the entire water column (equation (19)).

dC Zdt kM CHg CCH

I

PAR0 e kZ 18

where C Z is the concentration of CH3Hg at depth Z (pmol l 1), PAR(0)

is the PAR above the surface of the water (E m 2 d 1) and k is the light attenuation coefcient of sunlight (m 1).

Then,

Z

dC Zdt dZ Z kM CHg CCH

I

PAR0 e kZdZ 19

The parameters for calculating

R

dC Z

dt

dZ are given in Supplementary

Table 2.

References

1. Yokouchi, Y. et al. Global distribution and seasonal concentration change of methyl iodide in the atmosphere. J. Geophys. Res. Atmos. 113, D18311 (2008).

2. Ohr, H. D., Sims, J. J., Grech, N. M., Becker, J. O. & McGiffen, M. E. Methyl iodide, an ozone-safe alternative to methyl bromide as a soil fumigant. Plant Dis. 80, 731735 (1996).

3. Pelley, J. Methyl iodide, a fumigant under re. Environ. Sci. Technol. 43, 6898 (2009).

4. Mileson, B. E., Sweeney, L. M., Gargas, M. L. & Kinzell, J. Iodomethane human health risk characterization. Inhal. Toxicol. 21, 583605 (2009).

5. Gan, J. Y. & Yates, S. R. Degradation and phase partition of methyl iodide in soil. J. Agr. Food Chem. 44, 40014008 (1996).

6. Guo, M. X., Zheng, W., Papiernik, S. K. & Yates, S. R. Distribution and leaching of methyl iodide in soil following emulated shank and drip application.J. Environ. Qual. 33, 21492156 (2004).7. Dai, S., Huang, G. & Cai, Y. A study of methylation of inorganic tin by iodomethane in an aquatic environment with 13C carbon isotope tracer technique. Appl. Organomet. Chem. 3, 115120 (1989).

8. Chen, B. W. et al. Simulate methylation reaction of arsenic(III) with methyl iodide in an aquatic system. Appl. Organomet. Chem. 20, 747753 (2006).

9. Ahmad, I., Chau, Y. K., Wong, P. T. S., Carty, A. J. & Taylor, L. Chemical alkylation of lead (II) salts to tetraalkyllead (IV) in aqueous solution. Nature 287, 716717 (1980).

10. Baldoni, A. A. & Miyashiro, J. J. Preparation of organo mercuric compounds. United States Patent 2914451 (1959).

11. Hall, B., Bloom, N. S. & Munthe, J. An experimental study of two potential methylation agents of mercury in the atmosphere: CH3I and DMS. Water Air

Soil Poll. 80, 337341 (1995).12. Thayer, J. S., Olson, G. J. & Brinckman, F. E. Iodomethane as a potential metal mobilizing agent in nature. Environ. Sci. Technol. 18, 726729 (1984).13. Minganti, V., Capelli, R., Drava, G. & De Pellegrini, R. Solubilization and methylation of HgS, PbS, and SnS by iodomethane, a model experiment for the aquatic environment. Chemosphere 67, 10181024 (2007).

14. Celo, V. & Scott, S. L. Kinetics and mechanism of the mercury(II)-assisted hydrolysis of methyl iodide. Inorg. Chem. 44, 25072512 (2005).

15. Amyot, M., Mierle, G., Lean, D. R. S. & McQueen, D. J. Sunlight-induced formation of dissolved gaseous Hg in lake waters. Environ. Sci. Technol. 28, 23662371 (1994).

16. ODriscoll, N. J., Siciliano, S. D., Lean, D. R. S. & Amyot, M. Gross photoreduction kinetics of mercury in temperate freshwater lakes and rivers: Application to a general model of DGM dynamics. Environ. Sci. Technol. 40, 837843 (2006).

17. Castro, L., Dommergue, A., Larose, C., Ferrari, C. & Maron, L. A theoretical study of abiotic methylation reactions of gaseous elemental mercury by halogen-containing molecules. J. Phys. Chem. A 115, 56025608 (2011).

18. Algren, D. A., Schwartz, M. D. & Schier, J. G. Acute neurologic toxicity associated with inhalational exposure to methyl iodide and iodine-131. Clin. Toxicol. 44, 722723 (2006).

19. Celo, V., Lean, D. R. S. & Scott, S. L. Abiotic methylation of mercury in the aquatic environment. Sci. Total Environ. 368, 126137 (2006).

20. Craig, P. J. & Rapsomanikis, S. Methylation of tin and lead in the environment: Oxidative methyl transfer as a model for environmental reactions. Environ. Sci. Technol. 19, 726730 (1985).

21. Zhang, T. & Hsu-Kim, H. Photolytic degradation of methylmercury enhanced by binding to natural organic ligands. Nat. Geosci. 3, 473476 (2010).22. Hammerschmidt, C. R. & Fitzgerald, W. F. Iron-mediated photochemical decomposition of methylmercury in an Arctic Alaskan lake. Environ. Sci. Technol. 44, 61386143 (2010).

23. Lyday, P. A. in Ullmanns Encyclopedia of Industrial Chemistry (Wiley-VCH, 2000).

24. Skogerboe, R. K. & Wilson, S. A. Reduction of ionic species by fulvic acid. Anal. Chem. 53, 228232 (1981).

25. Nriagu, J. O. Mechanistic steps in the photoreduction of mercury in natural waters. Sci. Total Environ. 154, 18 (1994).

26. Schwartz, M. D., Obamwonyi, A. O., Thomas, J. D., Moorhead, J. F. & Morgan, B. W. Acute methyl iodide exposure with delayed neuropsychiatric sequelae: Report of a case. Am. J. Ind. Med. 47, 550556 (2005).

27. Horvat, M. et al. Total mercury, methylmercury and selenium in mercury polluted areas in the province Guizhou, China. Sci. Total Environ. 304, 231256 (2003).

28. Barringer, J. L. et al. Mercury and methylmercury dynamics in a coastal plain watershed, New Jersey, USA. Water Air Soil Poll. 212, 251273 (2010).29. Gilmour, C. C., Henry, E. A. & Mitchell, R. Sulfate stimulation of mercury methylation in freshwater sediments. Environ. Sci. Technol. 26, 22812287 (1992).

30. Fleming, E. J., Mack, E. E., Green, P. G. & Nelson, D. C. Mercury methylation from unexpected sources: Molybdate-inhibited freshwater sediments and an iron-reducing bacterium. Appl. Environ. Microb. 72, 457464 (2006).

31. Barnett, M. O. et al. Formation of mercuric sulde in soil. Environ. Sci. Technol. 31, 30373043 (1997).

32. Jonsson, S. et al. Mercury methylation rates for geochemically relevant Hg-II species in sediments. Environ. Sci. Technol. 46, 1165311659 (2012).33. Graham, A. M., Aiken, G. R. & Gilmour, C. C. Dissolved organic matter enhances microbial mercury methylation under suldic conditions. Environ. Sci. Technol. 46, 27152723 (2012).

34. Moser, F., Horwitz, A., Chen, J., Lim, W. A. & Voigt, C. A. Genetic sensor for strong methylating compounds. ACS Synth. Biol. 2, 614624 (2013).

35. Bancon-Montigny, C., Yang, L., Sturgeon, R. E., Colombini, V. & Mester, Z. High-yield synthesis of milligram amounts of isotopically enriched methylmercury (CH3 198HgCl). Appl. Organomet. Chem. 18, 5764 (2004).

36. Whalin, L. M. & Mason, R. P. A new method for the investigation of mercury redox chemistry in natural waters utilizing deatable Teon bags and additions of isotopically labeled mercury. Anal. Chim. Acta 558, 211221 (2006).

37. Mao, Y. X., Liu, G. L., Meichel, G., Cai, Y. & Jiang, G. B. Simultaneous speciation of monomethylmercury and monoethylmercury by aqueous phenylation and purge-and-trap preconcentration followed by atomic spectrometry detection. Anal. Chem. 80, 71637168 (2008).

38. Yin, Y. G. et al. Possible alkylation of inorganic Hg(II) by photochemical processes in the environment. Chemosphere 88, 816 (2012).

39. Alvarez-Puebla, R. A., dos Santos, D. S. J. & Aroca, R. F. SERS detection of environmental pollutants in humic acid-gold nanoparticle composite materials. Analyst 132, 12101214 (2007).

40. Lehnherr, I. & Louis, V. L. S. Importance of ultraviolet radiation in the photodemethylation of methylmercury in freshwater ecosystems. Environ. Sci. Technol. 43, 56925698 (2009).

Acknowledgements

We thank Leo Scinto for providing the instrument for PAR measurement, and Bruce McCord and Watson Lees for commenting on an earlier version of the manuscript. This research was supported by the National Natural Science Foundation of China (21120102040, 20937002), National Basic Research Program of China (2013CB430002), the U.S. Department of Energy (DE-FG01-05EW07033) and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB01020300).

Author contributions

Y.Y., Y.L., Y.C. and G.J. designed research; Y.Y., Y.L. and C.T. performed research; Y.Y., Y.L., Y.C. and G.J. wrote the paper.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Yin, Y. et al. Fumigant methyl iodide can methylate inorganic mercury species in natural waters. Nat. Commun. 5:4633 doi: 10.1038/ncomms5633 (2014).

NATURE COMMUNICATIONS | 5:4633 | DOI: 10.1038/ncomms5633 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2014 Macmillan Publishers Limited. All rights reserved.