Highly specific determination of gentamicin by induced collapse of Au–lipid capsules

Abstract

Residues of gentamicin in food pose a threat to human health. Determining gentamicin residues relies largely on adequate analytical techniques. Herein, a simple and highly specific colorimetric method for effective detection of this aminoglycoside antibiotic in milk based on gentamicin-induced collapse of an Au–lipid capsule is first proposed. The strong interaction between gentamicin and phosphatidylcholine resulted in the collapse of Au–lipid capsule and consequently, the color of AuNPs changed from wine red to blue. The concentration of gentamicin could be determined with the naked eye or a UV-vis spectrometer. The results showed that the absorption ratio (A₆₆₄/A₅₃₁) was liner with the gentamicin concentration in the 0–0.2 μM range with a linear 0.99 correlation coefficient. The detection limit was 7.4 nM. The coexisting substances, including L-arginine, guanidine hydrochloride, Tween-20, ammonium hydroxide, sodium chloride, potassium chloride, calcium chloride, glucose, and other common antibiotics such as streptomycin, amikacin, kanamycin, chloramphenicol, tetracycline, ampicillin and carbenicillin, did not interfere with the gentamicin determination with this method. Furthermore, the established method was successfully applied for the qualitative and quantitative analysis of gentamicin in pretreated milk products.

---

1. Introduction

Gentamicin is an aminoglycoside antibiotic, which is used to treat many types of bacterial infections, particularly those caused by Gram-negative organisms.^1,2 This antibiotic has been widely used not only as an antibacterial drug in human therapy, but also as a veterinary drug in animal husbandry and a crop-protection agent in agriculture.^3,4 However, gentamicin shows a comparatively narrow safety margin and may cause many side effects such as loss of hearing, toxicity to kidneys, and allergic reactions to drugs.⁵ Moreover, the residual amount of gentamicin found in the environment may also lead to antibiotic resistance from the pathogenic bacterial strains, which poses a serious threat to human health.⁶ It is of great importance to establish efficient, accurate and economical methods for the detection of gentamicin residue in environmental media.

Several methods have been designed for determining residual antibiotics, including gentamicin in the environment. Various immunoassays, such as the enzyme linked immunosorbent assay (ELISA), fluorescence immunoassay (FIA), radioimmunoassay (RIA) and immunochromatographic assay (ICA), have been employed for detecting antibiotic residues.^7–12 However, due to the cross-reactions with complicated compounds in food, immunoassays are susceptible to interference in real sample analysis.¹³ High-performance liquid chromatography (HPLC) is another high-sensitive method that can provide reliable results. However, due to the lack of a chromophore group, post-column derivatization and fluorescence detection are required for trace level gentamicin detection.^14–16 Liquid chromatography-mass spectrometry (LC-MS) was also employed for detecting gentamicin and other antibiotics with excellent performance,^17,18 but complicated sample preparation and high cost restrict its applications.

In recent years, AuNPs-based colorimetric sensors have been proven as a versatile analytical tool with high sensitivity, due to their unique properties such as color, biocompatibility, stability and distance-dependent surface plasmon resonance (SPR) absorption.^19,20 Liposome is an artificially-prepared spherical vesicle composed of a lamellar phase lipid bilayer and this unique structure inherently provides liposome with a powerful capability for modified with AuNPs on its surface. In this study, we found that the collapse of Au–lipid capsule could be specifically induced by gentamicin. Subsequently, the color of AuNPs changed from wine red to blue. We demonstrated that this phenomenon was applied successfully to effectively detect trace amounts of gentamicin residue in milk samples.

2. Experimental

2.1. Chemicals and materials

Chloroauric acid (HAuCl₄·3H₂O) was purchased from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). Sodium citrate tribasic dihydrate was from Bodi Chemical Factory of Tianjin (Tianjin, China). Phosphatidylcholine (soybean, >98%) containing alkyl chains of 16 carbon atoms was purchased from Aladdin, which is from soybean. Streptomycin sulfate, amikacin, chloramphenicol, tetracycline hydrochloride, kanamycin sulfate, carbenicillin disodium salt, ampicillin sodium salt and gentamicin sulfate were all USP grade and purchased from Solarbio (Beijing, China). All other chemicals were of analytical grade and used as purchased. Ultrapure water (18.25 MΩ cm), obtained from a water purification system, was used in the whole study. All the glassware was cleaned with aqua regia and thoroughly rinsed with ultrapure water before use.

2.2. Instrumentation

The absorption spectra were obtained on an evolution 300 UV-visible spectrophotometer (Thermo, USA) at room temperature (25 °C). Scanning electron microscopy (SEM) measurements were performed on an S-4800 (Hitachi, Japan) at 10 kV. Transmission electron microscopy (TEM) measurements were performed on a HT7700 (Hitachi, Japan) at 80 kV.

2.3. Nanoparticle synthesis

Au particles were prepared by citrate reduction of HAuCl₄ according to a previous report with necessary modifications.²¹ Typically, 200 μL 1% HAuCl₄ was added to 20 mL ultrapure water (18.25 MΩ cm) that was brought to a boil with vigorous stirring in a round-bottomed flask fitted with a reflux condenser. 400 μL of 1% trisodium citrate was then added rapidly to the solution, and the mixture was heated under reflux for another 30 min. The solution was cooled to room temperature and stored at 4 °C until it was used.

2.4. Preparation of Au–lipid capsule

The phosphatidylcholine liposome was fabricated according to a previous report with necessary modifications.²² The purchased phosphatidylcholine (soybean, >98%) (15 mg) was added into 10 mL ultrapure water (18.25 MΩ cm). Then, it was vortexed vigorously to make a phosphatidylcholine suspension. The phosphatidylcholine suspension with white color was heated at 60 °C for 30 minutes to exceed the phase-transition temperature of the used phosphatidylcholine molecules and then was sonicated at 25 °C for 30 minutes to form the phosphatidylcholine liposomes. The AuNPs colloidal solution and the phosphatidylcholine liposomes aqueous solution were mixed at a 1 : 1 volume ratio and then mixed immediately by pipetting to make the Au–lipid capsule.

2.5. Characterization of AuNPs and Au–lipid capsule

The size and morphology of AuNPs and Au–lipid capsule were characterized by a Hitachi S-4800 field emission scanning electron microscope and a Hitachi H-7700 field transmission electron microscope. The SEM images were acquired by operating at an accelerating voltage of 10 kV. To obtain high resolution images from the SEM analysis, all samples were deposited on a silicon wafer and allowed to dry for 30 min. Then, the excess liquid was absorbed from the edges with a filter paper to prevent adhesion. In particular, the collapse of Au–lipid capsules and aggregate of the AuNPs made by treating the solid supported drying the Au–lipid capsule solution treated by gentamicin.

The TEM samples were prepared by placing a drop of the samples onto a Formvar-coated copper grid. The grid was then stained by placing a drop of 1% phosphotungstic acid on its coated-side for 20 s. Excess stain on the grid was soaked away by touching a filter paper strip. The grid was then dried under a stream of nitrogen gas. All TEM images were taken under an 80 kV electron accelerating voltage.

2.6. Sample preparation

The liquid milk bought from local supermarket was pretreated to remove protein and fat.²³ Typically, 1.2 mL of 300 g L⁻¹ trichloroacetic acid was added into 3.0 mL of the spiked milk samples in a centrifuge tube. After thorough vortexing, the mixtures were centrifuged at 10 000 rpm for 10 min, and the supernatant was adjusted to the original volume with trichloroacetic acid again. Finally, the solution was filtered using a syringe and a 0.22 μm filter and was then used for gentamicin determination.

3. Results and discussion

3.1. The mechanism of the sensing system

To better understand the sensing strategy employed in this study, a schematic for detecting gentamicin by induced collapse of Au–lipid capsule is outlined in Scheme 1. The AuNPs existed on the outside of the liposome membrane to form an Au–lipid capsule. This was because the amine head groups at the outer layer of the phosphatidylcholine liposome could be capped with citrate-stabilized gold nanoparticles through electrostatic interactions.²⁴ Liposome contributed to stabilize the AuNPs dispersion, preventing their aggregation. Gentamicin, which carries five amino groups, could establish electrostatic interactions with the phosphate group on the surface of the phosphatidylcholine liposome. Due to the fluidity of the liposome surface, it was observed that there was a high-affinity interaction between gentamicin and the phosphatidylcholine molecules on the liposomal surface and subsequently resulted in liposome collapse and aggregation.²⁵ As shown in Scheme 1, when the concentration of gentamicin was less than 0.2 μM, the Au–lipid capsules slightly collapsed. When the gentamicin was above 0.2 μM, the Au–lipid capsules thoroughly collapsed. The color of the suspension was accordingly changed to violet blue. These changes were easily readable by the naked eye. Subtle differences could be measured by a spectrophotometer. Given that the strong binding affinity between phosphatidylcholine and gentamicin makes this method highly specific, this study essentially offers a simple but specific and rapid method for gentamicin detection.

Scheme 1 Schematic of the Au–lipid capsule sensing system for gentamicin detection.

3.2. Characterization of AuNPs and Au–lipid capsule

Fig. 1A showed the surface plasma resonance of AuNPs Fig. 1A(a) and Au–lipid capsule Fig. 1A(b). In the absence of liposome, AuNPs were wine red in color and displayed an intense surface plasmon band at about 531 nm. While in the presence of liposome, the absorbance of AuNPs at 531 nm slightly decreased, indicating the presence of AuNPs on the outer membrane of liposome surfaces without disturbing the spherical topography. The SEM in Fig. 1B, Fig. S3† and TEM image in Fig. S1† confirmed that AuNPs were present on the outer membrane of liposome surfaces, which was consistent with the previous report.²⁶

Fig. 1 (A) UV-vis absorption spectra of AuNPs; (a) Au–lipid capsule (b); (B) SEM image of AuNPs (a) and Au–lipid capsule (b).

Moreover, the stability of the Au–lipid capsule was tested. As key factors for most electrostatic reactions, the influence caused by pH and ionic strength of the Au–lipid capsule suspension was tested. The Au–lipid capsule suspension was treated with varied pH or different NaCl concentrations for one hour, and then the absorption ratio A₆₆₄/A₅₃₁ was observed. In this test, the absorption ratio of A₆₆₄/A₅₃₁ was monitored to represent the aggregation level. As shown in Fig. S2A,† the A₆₆₄/A₅₃₁ absorption ratio was the highest at pH 2, indicating the maximum aggregation level of AuNPs. In general, AuNPs prepared using tri-sodium citrate carry negative charges. The lower pH would weaken the electrostatic interaction between the AuNPs and amine head groups at the outer layer of the liposome and lead to AuNPs aggregation. Moreover, the absorption ratio of A₆₆₄/A₅₃₁ was found to increase with increasing ionic strength, as shown in Fig. S2B.†

3.3. Optimization of experimental conditions

3.3.1. The effect of media pH. Fig. 2A illustrates the influence of media pH on the absorption ratio (A₆₆₄/A₅₃₁) in the presence of 0.6 μM gentamicin. As shown, at the media pH under weakly basic conditions (pH = 7–9) or acidic conditions (pH = 5–6), the probe did not show a good response. The absorption ratio (A₆₆₄/A₅₃₁) was the highest at pH 5, indicating the AuNPs maximum aggregation level, which coincided with the solution color change (Fig. 2B). Therefore, pH 5 was set as the operational pH for subsequent experiments.

Fig. 2 (A) Effect of pH value on the absorption ratio (A₆₆₄/A₅₃₁) of Au–lipid capsule with the addition of 0.6 μM gentamicin; (B) image of the Au–lipid capsule with the addition of 0.6 μM gentamicin under different pH conditions (pH = 5–9). Data are from three separate experiments. The data are expressed as means ± SD. Error bars represent the standard deviation.

3.3.2. The effect of temperature. Fig. 3 illustrates the influence of temperature on the absorption ratio (A₆₆₄/A₅₃₁) in the presence of 0.6 μM gentamicin. It was observed that the temperature had little effect on the Au–lipid capsule response to gentamicin. This might be because the temperature influenced the mobility of the individual lipid molecules. For convenience, all the absorption measurements were performed for subsequent experiments at room temperature (25 °C).

Fig. 3 Absorption ratio, A₆₆₄/A₅₃₁, of the Au–lipid capsule (red line) and Au–lipid capsule with 0.6 μM gentamicin (blue line) at different temperatures. These experiments were performed three times with similar results each time. The data are expressed ±SD. Error bars represent the standard deviation.

3.4. Colorimetric sensing of gentamicin

To demonstrate the performance of the Au–lipid capsule probe, different concentrations of gentamicin ranging from 0 to 0.8 μM were added to aqueous solutions of the Au–lipid capsule. Upon addition of increasing concentrations of gentamicin, the color of the Au–lipid capsule gradually changed from initially wine red to purple and finally to blue (Fig. 4A). These changes are related to a plasmon resonance coupling effect of AuNPs: the reduction of the distance between AuNPs particles due to aggregation led to a strong enhancement of the localized electric field and increased refractive indices.²⁷ Addition of gentamicin induced a serious collapse of the Au–lipid capsule, leading to an increase of AuNPs aggregation. The aggregation of AuNPs was observed by UV-vis spectroscopy, which is shown in Fig. 4B. As expected, with the increase of gentamicin concentration, the surface plasmon resonance at 531 nm decreased, while at the same time, a new absorption band around 664 nm appeared and gradually increased. The corresponding effect was evaluated by comparing the A₆₆₄/A₅₃₁ values in the presence of different concentrations of gentamicin for quantitative analysis (Fig. 4C). Consistently, the A₆₆₄/A₅₃₁ increase was significantly observed in the concentration ranging from 0 to 0.4 μM and a slight increase was observed in the 0.4–0.8 μM concentration range.

Fig. 4 (A) Visual colorimetric change of the Au–lipid capsule solution upon gentamicin addition at different concentrations; (B) UV-vis absorption spectra of the Au–lipid capsule upon gentamicin addition with different concentrations (0, 0.05, 0.1, 0.2, 0.4, 0.6, and 0.8 μM). (C) The plot of ratio A₆₆₄/A₅₃₁ versus gentamicin concentration. All experiments were performed in triplicate. The data are expressed as means ± SD. Error bars represent the standard deviation.

The gentamicin-induced collapse of the Au–lipid capsule and the aggregation of AuNPs were further confirmed by SEM analysis (Fig. 5). First, the initial AuNPs were well dispersed on the liposome surface to form an Au–lipid capsule (Fig. 5a). However, after adding 0.02 μM of gentamicin, the slight collapse of Au–lipid capsule and the random agglomerate of AuNPs, driven by attraction between the negative charges on the surface of phosphatidylcholine liposome and positive charges on the gentamicin molecules, was observed (Fig. 5b). When the concentration of gentamicin was increased up to 0.1 or 0.2 μM, the Au–lipid capsule seriously collapsed and large numbers of AuNPs accumulated (Fig. 5c and d).

Fig. 5 SEM characterized Au–lipid capsule aggregation upon gentamicin addition for concentrations up to 0 μM (a), 0.02 μM (b), 0.1 μM (c), 0.2 μM (d).

A good linear correlation existed between the absorption ratio A₆₆₄/A₅₃₁ of Au–lipid capsule and the concentration of gentamicin in the 0–0.2 μM range with a 0.99 correlation coefficient (Fig. 6). The detection limit of the proposed method was 7.4 nM, which was calculated as LOD = 3 × (SD/S), where SD is the standard deviation of the response and S is the slope of the calibration curve.

Fig. 6 Standard calibration curves of A₆₆₄/A₅₃₁ against the gentamicin concentration from 0 to 0.2 μM. All experiments were performed in triplicate. The data are expressed as means ± SD. Error bars represent the standard deviation.

3.5. Specificity of the assay

Specificity is an important aspect to evaluate the performance of a new proposed assay. Thus, it is necessary to explore the selectivity of the proposed assay. The selectivity of the probe for gentamicin was evaluated by monitoring the absorption ratio (A₆₆₄/A₅₃₁) in the presence of various bioanalytes compared with a blank test (Fig. 7A). First, the responses of the Au–lipid capsule to gentamicin and to other antibiotic molecules were compared. The absorption ratio (A₆₆₄/A₅₃₁) showed little change in the presence of 0.8 μM of other antibiotics, including streptomycin, chloramphenicol, tetracycline, amikacin, kanamycin, ampicillin and carbenicillin. This result revealed that the Au–lipid capsule showed no cross-reactivity with the above mentioned antibiotics. It was demonstrated that instead of other aminoglycoside antibiotics, there was a strongest interaction between gentamicin and phosphatidylcholine, due to the positive charge, special conformation and the amphiphilic properties of gentamicin.²⁸ Second, we evaluated Au–lipid capsule response to molecules carrying positively charged groups such as L-arginine, guanidine hydrochloride, ammonium hydroxide. Au–lipid capsule showed no response to these molecules. In addition, we also monitored the Au–lipid capsule response to KCl, CaCl₂, NaCl, glucose and Tween-20, which might coexist with gentamicin in the environment. The UV-vis spectra revealed that these molecules did not interfere in gentamicin detection. However, the control results of AuNPs did not show specificity towards gentamicin (Fig. 7B). All the results showed that the probe of Au–lipid capsule could detect gentamicin with high selectivity.

Fig. 7 (A) Absorption ratio A₆₆₄/A₅₃₁ of Au–lipid capsule in the presence of different analytes compared to the Au–lipid capsule solution; (B) absorption ratio A₆₆₄/A₅₃₁ of AuNPs in the presence of different analytes compared to the AuNPs solution. Error bars show the standard deviations of three independent measurement.

3.6. Analysis of gentamicin in real samples

To evaluate the practical application of the proposed colorimetric method, the detection of milk sample was carried out by the well-known standard addition method.^29–31 The Maximum Residue Limits (MRLs) of some aminoglycoside antibiotics in milk are between 0.14 μM and 1.0 μM,³² and we chose 0.05 μM and 0.1 μM to study the recoveries of gentamicin. As showed in Table 1, the recoveries of gentamicin were 88.9% and 108.6% with the coefficient of variation less than 10% (n = 6), indicating the promising feasibility of this colorimetry for gentamicin quantification. Furthermore, a red-to-blue color change could also be observed upon addition of the milk sample with naked eye (Fig. 8). Therefore, the proposed method could be employed to analyze the antibiotics in pretreated milk samples.

Table 1 Detection of gentamicin levels in spiked milk

Samples	Added concentration (μM)	Measured concentration (μM)	Recovery (%)	CV (%)
Milk 1	0.05	0.04445	88.9	3.8485
Milk 2	0.1	0.1086	108.6	4.6905

---

Fig. 8 Visual colorimetric change of the optimized Au–lipid capsule probe: (a) with the addition of the extract from a blank milk sample; (b) with the addition of the extract containing 0.02 μM gentamicin; (c) with the addition of the extract containing 0.05 μM gentamicin; (d) with the addition of the extract containing 0.1 μM gentamicin.

4. Conclusion

In this study, a novel colorimetric sensor was proposed for the highly sensitive and selective detection of gentamicin. The strong electrostatic interaction between gentamicin and phosphatidylcholine rapidly induced the collapse of the Au–lipid capsule and consequently, the AuNPs aggregated. As a result, the color of Au–lipid capsule solution changed from red to blue, which could be determined with the naked eye or a UV-vis spectrometer. Parameters that affect the sensitivity and the possible interfering substances were investigated. Compared to other traditional detection method for gentamicin (Table S1†), the proposed approach presented a satisfactory linear range, low detection limit, short detection time and good accuracy and specificity for the convenient detection of gentamicin. In addition, we found that the Au–lipid capsule was suitable to monitor gentamicin in milk samples efficiently, which could be applied as a promising candidate for on-site detection of this commonly used antibiotic.

Acknowledgements

This project was supported by the National Natural Science Foundation of China (NSFC) [Grant 31270860], the Program for New Century Excellent Talents in University (NCET-13-0480) and the Yangling Agricultural Hi-tech Industries Demonstration Zone (2014NY-35). We are particularly grateful to Jianlong Wang for his good suggestions.

References

1. M. J. Weinstein, G. M. Luedemann, E. M. Oden and G. H. Wagman, Antimicrob. Agents Chemother., 1963, 161, 1.
2. R. F. Moulds and M. S. Jeyasingham, Australian Prescriber, 2010, 33, 134–135.
3. S. A. McEwen and P. J. Fedorka-Cray, Clin. Infect. Dis., 2002, 34, S93–S106.
4. A. K. Vidaver, Clin. Infect. Dis., 2002, 34, S107–S110.
5. R. Oertel, V. Neumeister and W. Kirch, J. Chromatogr. A, 2004, 1058, 197–201.
6. D. A. Tadesse, S. Zhao, E. Tong, S. Ayers, A. Singh, M. J. Bartholomew and P. F. McDermott, Emerging Infect. Dis., 2012, 18, 741–749.
7. N. Pastor-Navarro, Á. Maquieira and R. Puchades, Anal. Bioanal. Chem., 2009, 395, 907–920.
8. B. G. Knecht, A. Strasser, R. Dietrich, E. Märtlbauer, R. Niessner and M. G. Weller, Anal. Chem., 2004, 76, 646–654.
9. S. Rebe Raz, M. G. Bremer, W. Haasnoot and W. Norde, Anal. Chem., 2009, 81, 7743–7749.
10. N. Byzova, E. Zvereva, A. Zherdev, S. Eremin, P. Sveshnikov and B. Dzantiev, Anal. Chim. Acta, 2011, 701, 209–217.
11. Y. Bang-Ce, L. Songyang, Z. Peng and L. Xiao-Hong, Food Chem., 2008, 106, 797–803.
12. Y. Chen, Y. Shang, X. Li, X. Wu and X. Xiao, Food Chem., 2008, 108, 304–309.
13. S. Jongberg, V. Orlien, L. H. Skibsted and S. Weigel, Eur. Food Res. Technol., 2009, 228, 659–664.
14. P. Viñas, N. Balsalobre and M. Hernández-Córdoba, Talanta, 2007, 72, 808–812.
15. P. Edder, A. Cominoli and C. Corvi, J. Chromatogr. A, 1999, 830, 345–351.
16. A. Posyniak, J. Zmudzki and J. Niedzielska, J. Chromatogr. A, 2001, 914, 59–66.
17. K. Vučićević-Prčetić, R. Cservenak and N. Radulović, J. Pharm. Biomed. Anal., 2011, 56, 736–742.
18. S. B. Turnipseed, S. B. Clark, C. M. Karbiwnyk, W. C. Andersen, K. E. Miller and M. R. Madson, J. Chromatogr. B, 2009, 877, 1487–1493.
19. S. K. Ghosh and T. Pal, Chem. Rev., 2007, 107, 4797–4862.
20. R. A. Sperling, P. R. Gil, F. Zhang, M. Zanella and W. J. Parak, Chem. Soc. Rev., 2008, 37, 1896–1908.
21. C.-C. Huang and W.-L. Tseng, Analyst, 2009, 134, 1699–1705.
22. C. Tsukada, T. Tsuji, K. Matsuo, H. Nameki, T. Yoshida and S. Yagi, J. Surf. Anal., 2014, 20, 230–233.
23. Y. Wang, D. Zhang, W. Liu, X. Zhang, S. Yu, T. Liu, W. Zhang, W. Zhu and J. Wang, Biosens. Bioelectron., 2014, 55, 242–248.
24. S. Chatterjee, M. Krikorian, H. D. Gafney and B. Gersten, 2012, arXiv preprint arXiv:1207.3474.
25. F. Van Bambeke, P. M. Tulkens, R. Brasseur and M.-P. Mingeot-Leclercq, Eur. J. Pharmacol., Mol. Pharmacol. Sect., 1995, 289, 321–333.
26. M. Bhuvana and V. Dharuman, Analyst, 2014, 139, 2467–2475.
27. M. E. Stewart, C. R. Anderton, L. B. Thompson, J. Maria, S. K. Gray, J. A. Rogers and R. G. Nuzzo, Chem. Rev., 2008, 108, 494–521.
28. R. Brasseur, G. Laurent, J. M. Ruysschaert and P. Tulkens, Biochem. Pharmacol., 1984, 33, 629–637.
29. X. Zhang, H. Zhao, Y. Xue, Z. Wu, Y. Zhang, Y. He, X. Li and Z. Yuan, Biosens. Bioelectron., 2012, 34, 112–117.
30. Z. Wu, H. Zhao, Y. Xue, Q. Cao, J. Yang, Y. He, X. Li and Z. Yuan, Biosens. Bioelectron., 2011, 26, 2574–2578.
31. X. Zhang, Y. Zhang, H. Zhao, Y. He, X. Li and Z. Yuan, Anal. Chim. Acta, 2013, 778, 63–69.
32. E. E. Loomans, J. van Wiltenburg, M. Koets and A. van Amerongen, J. Agric. Food Chem., 2003, 51, 587–593.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra21752k

‡ These authors contributed equally to this work.

---