Molecules of Congo red caught hopping between insulin fibrils: a chiroptical probe of the dye–amyloid binding dynamics

Robert Deca, Viktoria Babenkoab and Wojciech Dzwolak*a
aDepartment of Chemistry, Biological and Chemical Research Centre, University of Warsaw, Pasteura 1, 02-093 Warsaw, Poland. E-mail: wdzwolak@chem.uw.edu.pl; Fax: +48 22 822 5996; Tel: +48 22 552 6567
bDepartment of Medicine, Imperial College London, London, UK

Received 2nd September 2016 , Accepted 5th October 2016

First published on 5th October 2016


Abstract

Misfolded protein molecules tend to self-assemble into highly-ordered aggregates, the so-called amyloid fibrils whose presence in vivo is often symptomatic of certain neurodegenerative disorders. Amyloid fibrils from different proteins share the capacity to bind a number of small planar molecules such as Congo Red (CR). The underlying docking interactions between fibrils and ligands remain poorly understood despite their importance for research on amyloid-inhibiting drugs. Here, we describe a novel approach to study binding of small ligands to amyloid fibrils. We probe dynamics of CR–amyloid interactions using two structural variants of insulin fibrils with strong quasi-opposite chiroptical properties and the ability to induce circular dichroism in bound CR molecules. We demonstrate that in mixed non-equilibrium complexes of CR with both chiral variants of fibrils, dissociation of CR–amyloid complex is the rate-determining step of the overall equilibration of the system. Kinetics of entropy-driven “hopping” of CR molecules between saturated and unoccupied binding moieties on amyloid fibrils follow exponential trajectories enabling calculation of dissociation rates of CR–amyloid complex. Temperature dependencies of CR-exchange rates reveal a heterogeneous distribution of CR–amyloid binding energies in the range of 38–53 kJ mol−1. Our results have been discussed in the context of mechanisms of interactions between amyloid fibrils and amyloid-specific stains.


Introduction

Amyloid fibrils are highly-ordered thermodynamically stable β-sheet-rich aggregates of misfolded protein molecules.1,2 In vivo, occurrence of amyloid deposits has been long recognized as a histopathological hallmark of certain neurodegenerative maladies such as Alzheimer's disease.3 More recent studies have found examples of biologically-functional amyloid4–6 and these findings are paralleled by attempts to employ amyloid fibrils in material science and nanotechnology.7,8 One of the characteristic features of amyloid fibrils distinguishing them from native and fully unfolded proteins is the capacity to bind a number of organic stains such as Thioflavin T9 and Congo red.10 In fact, an ongoing research effort is looking at similar molecules (in terms of size, planarity, and aromaticity) as possible candidates for therapeutic inhibitors of amyloid growth.11–13 It was demonstrated earlier that amyloid fibrils have a non-zero net chiral bias of stain-binding surface moieties which results in electronic transitions of these achiral dyes becoming active in corresponding circular dichroism (CD) spectra upon binding to amyloid.14–16 We have shown that under certain conditions, insulin forms amyloid fibrils with strong chiroptical properties.17–19 Depending on the self-assembly conditions, these fibrils may constitute quasi-enantiomeric chiral superstructures capable of inducing either positive or negative extrinsic Cotton effects in bound Thioflavin T, Congo red18 and other small, planar and aromatic molecules such as 2-anthracenecarboxylate.20 While both the chiral variants (termed +ICD and −ICD) are assembled from identical L-only insulin molecules, structural determinants of the chiral phenotype appear to be first established on the level of β-sheet twist (as captured by vibrational CD spectroscopy21) and are subsequently transferred to dye-binding moieties.

Because optical signals of dye complexes with +ICD and −ICD fibrils are additive and the system obeys the Lambert–Beer law, it is possible to design a CD-based experiment allowing one to probe reversibility of the dye–amyloid binding and dynamics of “hopping” of dye molecules between binding sites on amyloid surfaces. Furthermore, analysis of kinetics of the hopping process at different temperatures could, in principle, provide insights into murky energetics of dye–amyloid interactions. In this proof of concept study, we have used the two chiral isomers of insulin amyloid obtained under conditions described earlier17–19 and CR as a model ligand (Fig. 1).


image file: c6ra22067c-f1.tif
Fig. 1 (A) Atomic force microscopy images of two insulin amyloid variants with quasi-opposite chiroptical properties. (B) Induced circular dichroism spectra of separated and mixed +ICD and −ICD insulin fibrils in the presence of 2 μM CR in phosphate buffer pH 7.4. (C) Schematic explanation of CD-detected hopping of CR molecules in non-equilibrium mixtures of dye's complexes with optically active insulin fibrils.

Results and discussion

While CR added to either type of aggregate produces strong positive or negative ICD spectra with main peaks around 542.5 nm, adding the dye to approximately equimolar (i.e. “racemic” in terms of optical activity) mixture of both types of fibrils produces flat CD signal (Fig. 1). However, when CR is pre-incubated with an excess of one of the chiral variants (e.g. +ICD) with the equimolar portion of the other variant (−ICD) added afterwards, the transition to the flat-CD-equilibrium state is not instantaneous, as it requires initial dissociation of +ICD·CR complex with rate k1:
image file: c6ra22067c-t1.tif
which is followed by hopping of released CR molecules on −ICD fibrils and formation of −ICD·CR complex (Fig. 1):
image file: c6ra22067c-t2.tif
where k2 and k3 are association rates for CR complexes with the respective fibrils. Hence, +ICD·CR complex is a slowly releasing reservoir of CR molecules while −ICD amyloid plays the role of a molecular sponge absorbing the released dye. As Fig. 2A shows, the initial spontaneous complexation of CR by +ICD is very fast, but the subsequent process of racemic equilibration after addition of the −ICD portion is significantly decelerated implying that dissociation of amyloid–CR complex is the rate determining step for the overall transition. The same conclusions are reached when −ICD·CR complex is formed first and +ICD fibrils are added afterward (data not shown). Time-lapse ICD spectra were collected for non-equilibrium samples containing mixtures of pre-incubated complexes of CR with one type of chiral fibrils to which the opposite “naked” chiral fibrils were added at t = 0. The rate of ICD signal decay is strongly temperature-dependent as the data obtained at 15 and 35 °C show (Fig. 2B). The mixing stoichiometry was such that the amounts of +ICD and −ICD fibrils used would yield approximately (within 3% standard deviation) the same absolute CD intensity at 542.5 nm after staining with identical amounts of CR. We note the lack of full kinetic and thermodynamic equivalence of CR binding in −ICD·CR and +ICD·CR. There are visible differences in kinetics depending on the order of mixing (i.e. whether dissociation with k1 or k4 constant is the rate determining step). Also, the asymptotic values of equilibrium ICD plateaus are consistently positive which could imply presence of larger population of strongly CR-binding moieties on +ICD fibrils – either due to distinct distributions of such sites on +ICD and −ICD fibrils per mass unit or due to the particular stoichiometry of mixing. Possible differences in rotational strength (i.e. induced CD signal intensity per amyloid-bound CR molecule) of +ICD·CR and −ICD·CR complexes could also contribute to this effect. The latter inequivalence arises from the diastereomeric relationship between +ICD and −ICD aggregates and its certain aspects have been observed and discussed before.17,18 Importantly, the primary way the increasing temperature affects the system is by overcoming energy barrier of dissociation of amyloid·CR complex rather than by shifting thermodynamic equilibrium between +ICD·CR and −ICD·CR as illustrated in Fig. 2C: in a fresh quasi-racemic mixture of +ICD·CR and −ICD·CR, positive T-jumps cause an irreversible shift.

image file: c6ra22067c-f2.tif
Fig. 2 (A) Kinetic trajectories plotted according to CD signal at 542.5 nm after a +ICD sample is added to neat CR solution (red line), followed by addition of chiroptically equivalent portion of −ICD fibrils in 20th minute (green line). (B) Stacked time-lapse CD spectra of pre-incubated and equilibrated +ICD·CR complexes mixed with stoichiometric portions of −ICD (top row) and vice versa (bottom row) collected over the period of 100 min at 15 and 35 °C. (C) Drift of CD signal at 542.5 nm of quasi-racemic mixture of separately prepared +ICD·CR and −ICD·CR complexes (blue points) subjected to a temperature ramp (the superimposed red line, right axis).

Assuming reversibility of CR-fibril docking, stationary concentration of free CR during the equilibration phase, and excess of unoccupied binding moieties on amyloid fibrils, the kinetics of CR hopping can be described as:

ΔICD(t) = C1[thin space (1/6-em)]exp(−k1t) + C2[thin space (1/6-em)]exp(−k4t) + C3
where ΔICD(t) is the measured signal, C1, C2, C3 are constants (defined in Experimental section) and k1 and k4 are dissociation rates of +ICD·CR and −ICD·CR, respectively. Kinetic trajectories of CR-exchange between +ICD and −ICD fibrils obtained at different temperatures are shown in Fig. 3. As the experimental data fit well to this equation, calculation of k1 and k4 at different temperatures allowed us to evaluate energy barriers of CR–amyloid dissociation from the corresponding Arrhenius plots (Fig. 3, Table 1).


image file: c6ra22067c-f3.tif
Fig. 3 Temperature-dependent normalized kinetic trajectories for two ways of mixing substrates: −ICD·CR with +ICD (A), and +ICD·CR with −ICD (B). Inset shows Arrhenius plots calculated from data in panel (B).
Table 1 Binding energy of CR–insulin amyloid [kJ mol−1]
Complex Mixing of −ICD·CR with +ICD Mixing of +ICD·CR with −ICD
−ICD·CR 52.65 ± 1.93 38.51 ± 2.44
+ICD·CR 52.77 ± 2.56 40.84 ± 2.47


In our case, the thus determined energy barriers have the significance of effective CR–insulin fibrils binding energies. The values juxtaposed in Table 1 are in the approximate range between 38 and 53 kJ mol−1. These remarkably low energy barriers indicate that complexes of CR and insulin amyloid fibrils are, in fact, at least under the conditions examined in this study, very dynamic entities. There are several important aspects that must be considered here. First, it is implausible that amyloid fibrils (other than self-assembled from homopolypeptide chains) would have only one type of CR-binding moiety with the same binding energy. Typically, amyloid fibrils, such as studied here, are more likely to have different types of binding sites with distinct affinities to CR. Therefore the data obtained in our study correspond to averaged values. Certainly, this also implies that the experimental kinetic trajectories fitting the double exponential equation would be more accurately described by complex multi-exponential functions. Second, the two types of amyloid fibrils used in this study as CR-reservoir and CR-absorbing sponge are not exact enantiomers which creates additional asymmetry in energy barriers for the alternative mixing regimes (explained in Fig. 4). This effect is likely to contribute to the observed variations in CR binding energy for different orders of mixing the substrates reported in Table 1 (e.g. 52.77 ± 2.56 kJ mol−1 compared with 40.84 ± 2.47 kJ mol−1).


image file: c6ra22067c-f4.tif
Fig. 4 Various CR hopping scenarios between fibrils with different distributions of CR-binding energies. (A) Naïve case of a homopolypeptide amyloid (and its exact enantiomer) with a single type of CR-binding site, hopping is entropy-driven. (B) Hopping between exact amyloid enantiomers featuring populations of binding sites with different affinities to CR (possible enthalpic contribution). (C and D) Hopping between fibrils with asymmetric populations of binding sites. Cases (C) and (D) correspond to alternative regimes of substrate mixing and highlight origins of asymmetric activation barriers.

Several factors including complexity of these systems and heterogeneity of dye-binding sites hampers studies on energetics and dynamics of interactions between amyloid fibrils and small molecules of potential biomedical interest. Our approach constitutes a complimentary approach to the existing calorimetric22 and in silico23,24 methods. This proof-of-concept study utilizes a particular pair of amyloid forms which are accessible to us. The fact that both +ICD and −ICD fibrils give strong and opposite in sign CD signals after binding to CR is advantageous since it doubles the intensity of the measured signal. However, the here presented approach is also adaptable to a situation where the hopping of ligand molecules occurs between labelled fibrils (reservoir) and non-proteinaceous25 or even achiral sponge. Should the dye–sponge binding be effectively irreversible this would also reduce the number of fitted parameters. Despite medical importance the ongoing search for inhibitors of amyloid growth is complicated due to number of reasons including limited number of adequate physicochemical tools. With this work we are showing that chiroptical spectroscopy may offer an interesting and insightful way to characterize certain poorly understood aspects of amyloid–stain interactions.

Conclusions

In summary, we have demonstrated a new experimental approach to the obscure problem of amyloid–stain binding dynamics by probing time-dependent induced circular dichroism of Congo red in non-equilibrium complexes with two chiral variant of insulin fibrils. Our results show that the binding of CR to insulin amyloid fibrils is a structurally-inhomogeneous process. CR–amyloid docking is not only reversible, but the corresponding energy barriers of dissociation are remarkably low. We argue that a similar approach may be taken to study binding interactions between clinically-important amyloid fibrils and small ligands while screening them for possible applications as amyloid-targeting drugs.

Experimental

Preparation of +ICD and −ICD insulin amyloid samples

The two types of insulin amyloid fibrils characterized by opposite extrinsic Cotton effects upon staining with amyloid-specific dyes such as Congo Red (CR) or Thioflavin T (ThT) were obtained through the process of so-called ‘vortex-induced chiral bifurcation’ described in our earlier papers.17–19 Typically, 10 Eppendorf tubes, each containing 0.6 mL of freshly prepared solution of 0.75 wt% bovine insulin (from Sigma, USA) in 0.1 M NaCl, D2O, pD 1.9 (uncorrected pH-meter readout), were vortexed in TS-100C Thermo-Shaker accessory (BioSan, Latvia) at 1400 rpm, 50 °C for 72 hours. As Fourier transform infrared (FT-IR) spectroscopy was used as an independent way of monitoring insulin aggregation in the conformation-sensitive amide I region, H2O, whose bending vibrations overlap amide I/I′ band, was replaced with D2O. After 72 hours of agitation, insulin samples randomly converted (chiral bifurcation17,18) either to +ICD or −ICD types of insulin amyloid which was determined by induced circular dichroism after staining the fibrils with ThT according to previously described protocol.18

Fresh phosphate buffer was prepared through adjusting pH of 0.05 M Na3PO4 solution containing also 0.1 M NaCl with H3PO4 to final pH of 7.4. Subsequently, acidic +ICD and −ICD samples with the chiral phenotypes already determined using the ThT assay were washed with the phosphate buffer. The procedure consisted in centrifugation of amyloid samples (5 minutes at 13[thin space (1/6-em)]400 rpm), followed by replacing upper 200 μL of supernatant with a 200 μL portion of phosphate buffer, vortexing the sample, and centrifugation. The procedure was carried out 3 times. Final pH of +ICD/−ICD samples suspended in the phosphate buffer was approximately 7.3. Insulin fibrils of either type remained stable upon the buffer-exchange: no decrease in terms of β-sheet content, as probed by attenuated total reflectance Fourier transform infrared (ATR FT-IR) spectroscopy nor changes in chiroptical characteristics were observed.

CD/ICD measurements

Prior to ICD measurements, fresh 1 mM stock solution of CR in H2O was prepared. Sample for CD measurements was prepared by diluting a 4 μL portion of either +ICD or −ICD amyloid (already suspended in phosphate buffer as described above) with larger 2 mL volume of the same buffer. This was followed addition of 4 μL of the CR stock solution and brief vortexing of sample. We have checked that induced CD signal of amyloid-bound CR follows Lambert–Beer law in respect to CR concentration. In the case of kinetic measurements, a 4 μL portion of +ICD fibrils (or 3.5 μL portion of −ICD fibrils) was diluted with 2 mL of the phosphate buffer and the resulting solution was placed in 1 cm quartz cuvette along with a magnetic bar. The slightly different concentrations of +ICD and −ICD fibrils were chosen as CR-complexes of the two chiral variants of insulin fibrils have different induced CD intensities. The cuvette was subsequently placed in a J-815 spectropolarimeter (Jasco, Japan) equipped with a temperature-controlling Peltier accessory and a magnetic stirrer. After adding 4 μL portion of CR stock solution, CD measurements were carried out continuously at the desired temperature and maximum stirring rate for 20 minutes. Subsequently a portion of alternative chiral variant (3.5 μL of −ICD or 4 μL of +ICD fibrils) was added and CD measurements followed for another 100 minutes. The selected ratios of amyloid[thin space (1/6-em)]:[thin space (1/6-em)]CR mixing correspond to following molar ratios of insulin monomers to CR: 1.29 for +ICD[thin space (1/6-em)]:[thin space (1/6-em)]CR, 1.13 for −ICD[thin space (1/6-em)]:[thin space (1/6-em)]CR, and – consequently – 2.42 for mixtures of +ICD and −ICD in the case of kinetic measurements. It is important to stress that such CR-fibrils mixing stoichiometry enabled keeping most of remaining CR-binding sites unoccupied. This was verified through CD-monitored titration in which increasing amounts of CR were added to chiral insulin fibrils at the fixed concentration used in the kinetic experiments. A linear dependence of induced CD signal at 542 nm was observed for gradually added CR at least up to its final concentration 2.5 times above that used in the kinetic experiments.

While for each stationary CD spectrum 5 individual scans were accumulated, in the case of kinetic experiments, each spectrum consisted of a single scan. Spectra were baseline corrected through subtraction of buffer spectra.

Atomic force microscopy (AFM)

For AFM imaging, fresh samples of insulin fibrils (before the elution with phosphate buffer) were diluted 100-times with acidified D2O (pD 1.9). Thus obtained samples were deposited on cleaved mica surfaces. The samples were left to dry up overnight. Tapping-mode AFM measurements were carried out using Nanoscope III AFM from Veeco, USA, and TAP300-Al sensors (res. frequency 300 kHz) from BudgetSensors, Bulgaria. Other details of AFM-measurements were described earlier.18

Kinetic model

For reversible binding of insulin fibrils to CR and the simultaneous formation of +ICD·CR and −ICD·CR complexes:
image file: c6ra22067c-t3.tif
kinetics of formation of respective complexes would be described by the following:
image file: c6ra22067c-t4.tif

Given the excess of dye-binding sites over total number of CR molecules, it can be assumed that:

image file: c6ra22067c-t5.tif

Furthermore, we assume that during the equilibration (hoping of CR between reservoir and sponge) the stationary state condition is achieved (constant concentration of free dye):

image file: c6ra22067c-t6.tif
thus:
image file: c6ra22067c-t7.tif
where:
image file: c6ra22067c-t8.tif
Hence:
image file: c6ra22067c-t9.tif

While for the opposite order of mixing, one obtains:

image file: c6ra22067c-t10.tif

The experimentally measured signal (ΔICD) is a transient difference between positive ellipticity of +ICD·CR and negative ellipticity of −ICD·CR. ΔICD signals were normalized so that ΔICDN is equal to “1” at t = 0:

image file: c6ra22067c-t11.tif
becomes:
image file: c6ra22067c-t12.tif

Assuming that:

image file: c6ra22067c-t13.tif
one obtains:
ΔICDN = C1ek1t + C2ek4t + C3

And for the opposite order of mixing:

image file: c6ra22067c-t14.tif

Hence:

ΔICDN = C4ek4t + C5ek1t + C6
where:
image file: c6ra22067c-t15.tif

The optimization was carried out using a weighted least-square method. Values of k1 and k4 obtained from fitting experimental data to these equations are shown in Table 2.

Table 2 Values of k1 and k4 dissociation constants at different temperatures
T [°C] Mixing of −ICD·CR with +ICD Mixing of +ICD·CR with −ICD
k1 [min−1] k4 [min−1] k1 [min−1] k4 [min−1]
15 0.0138 ± 0.0026 0.0657 ± 0.010 0.0166 ± 0.0024 0.0647 ± 0.015
20 0.0161 ± 0.0014 0.0922 ± 0.0080 0.0115 ± 0.0025 0.0666 ± 0.0051
25 0.0201 ± 0.0014 0.100 ± 0.0060 0.0243 ± 0.0010 0.119 ± 0.0065
30 0.0263 ± 0.0013 0.128 ± 0.0065 0.0377 ± 0.00087 0.181 ± 0.0078
35 0.0312 ± 0.00081 0.195 ± 0.0071 0.0367 ± 0.0018 0.176 ± 0.0090
40 0.0438 ± 0.0013 0.233 ± 0.0086 0.0728 ± 0.0017 0.329 ± 0.014


Acknowledgements

This work was in part supported by the National Science Centre of Poland, grant no. 2011/03/N/ST4/00736 (to V. B.) and in part by University of Warsaw (project 501/64-BST-176309 to W. D.). The study was carried out at the Biological and Chemical Research Centre, University of Warsaw, established within the project co-financed by EU from the European Regional Development Fund under the Operational Programme Innovative Economy, 2007–2013, and with the use of CePT infrastructure financed by the same EU programme.

References

  1. F. Chiti and C. M. Dobson, Annu. Rev. Biochem., 2006, 75, 333–366 CrossRef CAS PubMed .
  2. V. N. Uversky and A. L. Fink, Biochim. Biophys. Acta, 2004, 1698, 131–153 CrossRef CAS PubMed .
  3. J. Hardy and D. J. Selkoe, Science, 2002, 297, 353–356 CrossRef CAS PubMed .
  4. D. M. Fowler, A. V. Koulov, W. E. Balch and J. W. Kelly, Trends Biochem. Sci., 2007, 32, 217–224 CrossRef CAS PubMed .
  5. J. Li, T. McQuade, A. B. Siemer, J. Napetschnig, K. Moriwaki, Y. S. Hsiao, E. Damko, D. Moquin, T. Walz, A. McDermott, F. K. Chann and H. Wu, Cell, 2012, 150, 339–350 CrossRef CAS PubMed .
  6. M. S. Dueholm, S. V. Petersen, M. Sønderkær, P. Larsen, G. Christiansen, K. L. Hein, J. J. Enqhild, J. L. Nielsen, K. L. Nielsen, P. H. Nielsen and D. E. Otzen, Mol. Microbiol., 2010, 77, 1009–1020 CAS .
  7. T. Scheibel, R. Parthasarathy, G. Sawicki, X. M. Lin, H. Jaeger and S. L. Lindquist, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 4527–4532 CrossRef CAS PubMed .
  8. C. Li, S. Bolisetty and R. Mezzenga, Adv. Mater., 2013, 25, 3694–3700 CrossRef CAS PubMed .
  9. H. LeVine, Methods Enzymol., 1999, 309, 274–284 CAS .
  10. W. E. Klunk, J. W. Pettegrew and D. J. Abraham, J. Histochem. Cytochem., 1989, 37, 1273–1281 CrossRef CAS PubMed .
  11. W. M. Berhanu and A. E. Masunov, J. Biomol. Struct. Dyn., 2015, 33, 1399–1411 CAS .
  12. R. Malisauskas, A. Botyriute, J. G. Cannon and V. Smirnovas, PLoS One, 2015, 10, e0121231 Search PubMed .
  13. R. Mishra, D. Sellin, D. Radovan, A. Gohlke and R. Winter, ChemBioChem, 2009, 10, 445–449 CrossRef CAS PubMed .
  14. E. P. Benditt, N. Eriksen and C. Berglund, Proc. Natl. Acad. Sci. U. S. A., 1970, 66, 1044–1051 CrossRef CAS .
  15. R. Khurana, V. N. Uversky, L. Nielsen and L. Fink, J. Biol. Chem., 2001, 276, 22715–22721 CrossRef CAS PubMed .
  16. W. Dzwolak and M. Pecul, FEBS Lett., 2005, 579, 6601–6603 CrossRef CAS PubMed .
  17. W. Dzwolak, A. Loksztejn, A. Galinska-Rakoczy, R. Adachi, Y. Goto and L. Rupnicki, J. Am. Chem. Soc., 2007, 129, 7517–7522 CrossRef CAS PubMed .
  18. A. Loksztejn and W. Dzwolak, J. Mol. Biol., 2008, 379, 9–16 CrossRef CAS PubMed .
  19. A. Loksztejn and W. Dzwolak, J. Mol. Biol., 2010, 395, 643–655 CrossRef CAS PubMed .
  20. M. Nishijima, H. Tanaka, C. Yang, G. Fukuhara, T. Mori, V. Babenko, W. Dzwolak and Y. Inoue, Chem. Commun., 2013, 49, 8916–8918 RSC .
  21. G. Zhang, V. Babenko, W. Dzwolak and T. A. Keiderling, Biochemistry, 2015, 54, 7193–7202 CrossRef CAS PubMed .
  22. R. E. McKnight, D. R. Jackson and K. Yokoyama, Eur. Biophys. J., 2013, 42, 495–501 CrossRef CAS PubMed .
  23. C. Wu, Z. Wang, H. Lei, W. Zhang and Y. Duan, J. Am. Chem. Soc., 2007, 129, 1225–1232 CrossRef CAS PubMed .
  24. C. Wu, J. Scott and J. E. Shea, Biophys. J., 2012, 103, 550–557 CrossRef CAS PubMed .
  25. E. Y. Ozmen and M. Yilmaz, J. Hazard. Mater., 2007, 148, 303–310 CrossRef CAS PubMed .

This journal is © The Royal Society of Chemistry 2016
Click here to see how this site uses Cookies. View our privacy policy here.