Langmuir 2004, 20, 2837-2851
2837
Structural Characteristics of Activated Carbons and Ibuprofen Adsorption Affected by Bovine Serum Albumin M. Melillo,† V. M. Gun’ko,‡ S. R. Tennison,§ L. I. Mikhalovska,† G. J. Phillips,† J. G. Davies,† A. W. Lloyd,† O. P. Kozynchenko,§ D. J. Malik,| M. Streat,| and S. V. Mikhalovsky*,† School of Pharmacy and Biomolecular Sciences, University of Brighton, Lewes Road, Brighton BN2 4GJ, U.K., Institute of Surface Chemistry, 17 General Naumov Street, 03164 Kiev, Ukraine, MAST Carbon Ltd., Henley Park, Guildford, Surrey GU3 2AF, U.K., and Department of Chemical Engineering, Loughborough University, Loughborough, Leics LE11 3TU, U.K. Received October 31, 2003 Structural characteristics of a series of MAST carbons were studied using scanning electron microscopy images and the nitrogen adsorption isotherms analyzed with several models of pores and different adsorption equations. A developed model of pores as a mixture of gaps between spherical nanoparticles and slitlike pores was found appropriate for MAST carbons. Adsorption of ibuprofen [2-(4-isobutylphenyl)propionic acid] on activated carbons possessing different pore size distributions in protein-free and bovine serum albumin (BSA)-containing aqueous solutions reveals the importance of the contribution of mesopores to the total porosity of adsorbents. The influence of the mesoporosity increases when considering the removal of the drug from the protein-containing solution. Cellulose-coated microporous carbon Norit RBX adsorbs significantly smaller amounts of ibuprofen than uncoated micro/mesoporous MAST carbons whose adsorption capability increases with increasing mesoporosity and specific surface area, burnoff dependent variable. A similar effect of broad pores is observed on adsorption of fibrinogen on the same carbons. Analysis of the ibuprofen adsorption data using Langmuir and D’Arcy-Watt equations as the kernel of the Fredholm integral equation shows that the nonuniformity of ibuprofen adsorption complexes diminishes with the presence of BSA. This effect may be explained by a partial adsorption of ibuprofen onto protein molecules immobilized on carbon particles and blocking of a portion of narrow pores.
Introduction Ibuprofen (2-(4-isobutylphenyl)propionic acid) is a nonsteroidal anti-inflammatory drug. Its normal therapeutic concentration in blood is about 50 mg/L; however, it is toxic at the concentration of 250 mg/L or higher.1 In human plasma ibuprofen is 99% protein-bound to serum albumin.2 Removal of overdosed ibuprofen can be carried out using activated carbons or other adsorbents with a hemoperfusion column technique. Activated carbons are used as oral adsorbents or a constituent of hemoperfusion columns to treat cases of poisoning, as carbons can adsorb a variety of toxic organic and inorganic compounds. However, activated carbons produced from natural raw materials (fruit stones and shells, wood, etc.) have poor hemocompatibility. Therefore they have to be covered with a more hemocompatible semipermeable coating such as cellulose or poly(HEMA).3 However, there is an inherent disadvantage with coated carbons relating to significantly slower adsorption kinetics particularly toward high molecular solutes. The use of synthetic polymer-based carbons with controlled particle morphology and chemical composition allows the production of adsorbents, which can be tailored for target specific poisons and drugs without * To whom correspondence may be addressed. Fax: +44 (0) 1273 679 333. E-mail address:
[email protected]. † University of Brighton. ‡ Institute of Surface Chemistry. § MAST Carbon Ltd. | Loughborough University. (1) Flanagan, R. J. Ann. Clin. Biochem. 1998, 35, 261. (2) Halpern, S. M.; Fitzpatrick, R.; Volans, G. N. Adverse Drug React. Toxicol. Rev. 1993, 12, 107. (3) Cooney, D. O. Activated Charcoal in Medical Applications; Marcel Dekker: New York, 1995.
the need for polymer coatings. These carbons can be characterized by more appropriate adsorption kinetics than coated ones. The production of extracorporeal devices containing uncoated but biocompatible carbons may offer an alternative method for the rapid reduction of circulating titers of poisons in whole blood or plasma.4 However, when considering the potential use of adsorbents in hemoperfusion, it is necessary to take into account the role of proteins in the adsorption of target compounds.2,5-10 For most drugs there are two primary binding sites on each albumin molecule, as well as secondary, less specific, binding sites. Ibuprofen binds reversibly to albumin, and the main binding site has been localized on the sites IIIA.11 Ibuprofen is a high affinity selective binding ligand to human serum albumin (HSA) with dissociation constants in the order of 10-5 M.12 Epps et al.13 used the method of extrinsic fluorescence to determine the dissociation constant of the albumin-ibuprofen complex by competitive displacement of a fluorescent probe specific for the IIIA (4) Levy, H.; Dasgupta, A.; et al. J. Toxicol., Clin. Toxicol. 1995, 33, 457. (5) Gulyassy, P. F.; Depner, T. A. Am. J. Kidney Dis. 1983, 2, 578. (6) Routledge, P. A. Br. J. Pharmacol. 1986, 22, 499. (7) Chen, Y. Z.; Ung, C. Y. J. Mol. Graphics Modell. 2001, 20, 199. (8) Trappe, T. A.; White, F.; Lambert, C. P.; Cesar, D.; Hellerstein, M.; Evans, W. J. Am. J. Physiol. Endocrinol. Metab. 2002, 282, E551. (9) Haspel, H.; Gleich, L.; Shea, M.; Towle, T.; Schmid, N.; Ommert, S.; Carvalho, B.; Kellogg, G.; Bansal, P.; Contarino, M. Drug Serum Protein Binding on Tecan’s LabCD System; Tecan Boston: Medford. (10) Fowler, C. J.; Tiger, G.; Stenstrom, A. J. Pharmacol. Exp. Ther. 1997, 283, 729. (11) He, X. M.; Carter, D. C. Nature 1992, 358, 209. (12) Hage, D. S.; Noctor, T. A.; Wainer, I. W. J. Chromatogr., A 1995, 693, 23. (13) Epps, D. E.; Raub, T. J.; Kezdy, F. J. Anal. Biochem. 1995, 227, 342.
10.1021/la0360557 CCC: $27.50 © 2004 American Chemical Society Published on Web 02/28/2004
2838
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
Table 1. Structural Characteristics of Activated Carbonsa carbon MAST-37 MAST-48 MAST-58 MAST-65 MAST-75 Norit RBX carbon MAST-37 MAST-48 MAST-58 MAST-65 MAST-75 Norit RBX
SΣ, SDFT, Ssum,slit(2), Ssum,slit(1), SDS, Smic,slit, Smes,slit, Smic/ Smic,BR, Smes,BR, burnoff, SBET, SBET, m2/g m2/cm3 m2/g m2/g m2/g m2/g % m2/g m2/g m2/g Ssum ∆wslit m2/g m2/g 37 48 58 65 75 n/a
1150 1340 1820 1840 2330 1029
450 460 530 460 510 380
1187 1390 1846 1857 2448 948
1038 1117 1281 1490 1539
972 993 1221 1290 1505
1047 1169 1432 1458 1679 804
1268 1299 1770 1823 1762 1047
911 994 1167 1187 1243 756
137 175 265 271 435 48
0.87 0.85 0.82 0.81 0.74 0.94
0.099 0.181 0.285 0.262 0.328 0.279
990 1274 1514 1514 1710 1060
Vmic,slit, Vmes,slit, Vem,tot, Vem,gr, Vmic,slit/ Vmes,slit/ Vp, Vp, cm3/g cm3/cm3 cm3/g Vp Vp cm3/g cm3/g cm3/g
δDS, nm
xDS, nm
Vmic,BR, Vmes,BR, Fb, g/cm3 cm3/g cm3/g
1.15 1.32 1.72 1.73 2.12 0.51
0.15 0.18 0.21 0.21 0.51 0.16
0.39 0.48 0.58 0.60 0.98 0.47
0.39 0.34 0.29 0.25 0.22 0.42
0.45 0.45 0.50 0.43 0.47 0.21
0.50 0.56 0.73 0.73 0.80 0.41
0.65 0.73 0.99 0.98 1.30 0.08
1.87 2.25 2.76 3.31 3.86 1.69
1.10 1.32 1.62 1.94 2.26 0.99
0.44 0.42 0.42 0.42 0.38 0.81
0.56 0.56 0.57 0.57 0.61 0.16
0.48 0.58 0.73 0.73 0.82 0.43
0.67 0.75 0.99 1.00 1.30 0.08
161 178 325 327 571 51
∆wBR 0.0 -0.049 0.001 -0.001 -0.022 -0.074
DAJ
∆wDFT
2.810 2.792 2.714 2.703 2.576 2.928
0.136 0.125 0.049 0.155 0.033
a Parameters with DS subscript were calculated using Dubinin-Stoeckli equation; D AJ is the fractal dimension determined using the adsorption data at p/p0 < 0.85 with the Frenkel-Halsey-Hill equation; Vmic,slit and Vmes,slit, Smic,slit, and Smes,slit are the pore volumes and specific surface area of micro- and mesopores, respectively, calculated by integration for fV(x) and fS(x) over 0.2-1.0 nm and 1.0-25.0 nm ranges, respectively. Ssum,slit ) Smic,slit + Smes,slit + Smac,slit; ∆wslit ) SBET/Ssum,slit - 1. Vem ) 1/Fb - 1/F0 is the empty volume in grams of carbon; Vem,gr ) RpacVem is the empty volume inside granules, Rpac is the packing coefficient (assuming 0.586 for cubic lattice with spherical carbon granules). F0 ) 1.45 g/cm3. The values with the BR subscript were calculated with eqs 11-13. The Vmac and Smac values are not shown because they are very low. SDFT is the total specific surface area, Ssum,slit(2) was calculated using the same isotherms that were used on DFT calculations, ∆wDFT ) SDFT/Ssum,slit - 1 (Miromeritics ASAP 1010 isotherms).
site, and they found a dissociation constant of (2.7 ( 1.2) × 10-6 M. The interaction of drugs with high-affinity binding sites has been widely studied. Moreover recent studies are now focusing on the low-affinity binding sites of albumin.14-16 Many compounds are involved in the lowaffinity binding interaction, especially when ligand concentration is much higher than that of HSA in solution.14 The number of binding sites calculated for the HSAibuprofen complex, from self-diffusion coefficient NMR data, was approximately 38.16 However, many aspects of the interactions between albumin and low molecular weight organics (e.g., drugs or toxins) in the presence of a “third” component such as solid porous adsorbent are rather unclear. Therefore, the aim of this work is to study the influence of the textural characteristics of adsorbents on their potential to remove ibuprofen (as a model drug with a toxic effect on overdosing) from protein-containing and protein-free solutions. Experimental Section Materials. Activated carbons (MAST Carbon Ltd., Guildford, U.K.) produced from phenolic resin precursors differ in the degree of burnoff (shown in carbon’s code MAST-xx as percentage (xx) of burnoff in Table 1) which affects their textural characteristics. These carbons consist of spherical granules (diameter 0.2-0.5 mm slightly diminished with increasing burnoff (Figure 1)) composed of amorphous nanoparticles tightly attached one to another. According to Tennison,17 the latter have an average size of approximately 4 nm and a specific density of approximately 1.5 g/cm3. These parameter values provide an outer specific surface area of approximately 1000 m2/g. Consequently, for the carbon MAST-75, the size of nanoparticles should be smaller than 2 nm if these particles are nonporous.17 Commercial cellulose-coated activated carbon Norit RBX18 (utilized in Adsorba300C hemoperfusion column (Gambro)) was used as a control adsorbent in comparative investigations. Structural (14) Ji, Z.; Hanzhen, Y.; Liu, M.; Hu, J. J. Pharm. Biomed. Anal. 2002. (15) Sulkowska, A. Appl. Spectrosc. 1997, 51, 428. (16) Luo, R.-S.; Liu, M.-L.; Mao, X.-A. Spectrochim. Acta, Part A 1999, 55, 1897. (17) Tennison, S. R. Appl. Catal. 1998, 173, 289. (18) Mikhalovsky, S. V. Microparticles for hemoperfusion and extracorporeal therapy, Microspheres, Microcapsules and Liposomes. Medical & Biotechnology Applications; The MML Series; Arshady, R., Ed.; Citus Books: London, 1999; Vol. 2, Chapter 5, pp 133-169.
characteristics of the studied carbons are shown in Table 1. Certain parameters are expressed both per unit of mass and per unit of volume of the carbons. The latter is more appropriate considering the potential application of carbons in medical devices such as hemoperfusion columns. However, in other cases the carbon mass may be a more relevant characteristic than its volume. Bovine serum albumin (BSA) (99% purity, molecular weight (MW) ≈ 67 kDa) and ibuprofen used in the adsorption and fluorescence experiments were obtained from Sigma-Aldrich. Stock solutions of ibuprofen (0.05 M) and of BSA (3.00 g/L) were prepared in Tyrode buffer. The composition of the Tyrode buffer mimics the mineral composition and pH of blood, since it contains sodium chloride, sodium hydrogen carbonate, potassium chloride, sodium dihydrogen orthophosphate, magnesium chloride, calcium chloride, and glucose. Human blood plasma protein fibrinogen (Fg, MW ≈ 340 kDa, size ≈ 45 × 9 × 6 nm) spiked with radio-labeled 125I-fibrinogen (Amersham, UK) with the final activity of 6.25 × 105 CPM/mL of the bulk Fg solution was used to compare the adsorption capability of the carbon adsorbents with respect to large protein molecules. Nitrogen Adsorption. Low-temperature (77.4 K) nitrogen adsorption isotherms were recorded using Micromeritics Gemini II (Figure 2) and ASAP 2010 adsorption analyzers. The specific surface area (Table 1, SBET) was calculated according to the standard BET method19,20 both per gram and per cubic centimeter of carbons. It should be noted that application of the standard BET equation to microporous carbons (such as Norit RBX) characterized by the nitrogen adsorption isotherm of the type I is questionable; however, it is acceptable if cBET < 450.21 For MAST carbons, despite a significant contribution of mesopores to the total porosity (in contrast to Norit RBX), the cBET value criterion is satisfied (cBET < 450) for certain samples. Therefore, the SΣ ) Sext + SDA values are also shown in Table 1 to characterize the total specific surface area. As a whole, the SΣ and SBET values are closely related. Consequently, SBET (as a conventional parameter) may be utilized to characterize all the studied carbons. The total pore volume Vp was evaluated by converting the volume of nitrogen adsorbed at p/p0 ≈ 0.98 (p and p0 denote the equilibrium pressure and the saturation pressure of nitrogen at (19) Adamson, A. W.; Gast, A. P. Physical Chemistry of Surface, 6th ed.; Wiley: New York, 1997. (20) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area and Porosity, 2nd ed.; Academic Press: London, 1982. (21) Rouquerol, J.; Avnir, D.; Fairbridge, C. W.; Everett, D. H.; Haynes, J. M.; Pernicone, N.; Ramsay, J. D. F.; Sing, K. S. W.; Unger, K. K. Pure Appl. Chem. 1994, 66, 1739.
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2839
Figure 1. SEM (model JSM-6310, Japan Electron Optics Ltd.) images of (a, b, c) MAST-37 and (d, e, f) MAST-75 surfaces at different magnifications (bars are shown in figures). 77.4 K, respectively) to the volume of liquid nitrogen both per gram and per cubic centimeter of carbons. The RS plots19,20 for the studied carbons (Carbopack F graphitized carbon black was used as a reference adsorbent22) (parts b and c of Figure 2) were also used to calculate the external surface area Sext. The modified Dubinin-Stoeckli (DS) equation23,24 was utilized to estimate micropore contribution with correction relevant to adsorption in mesopores. The SDS values (Table 1) were calculated at x ) 0.2-1.0 nm corresponding to micropores. To calculate the specific surface area of micropores (SDA), the Dubinin-Astakhov (DA) equation25 was also applied. The pore size distributions (PSDs) fV(x) (differential PSD since fV(x) ∼ dVp/dx) of the studied carbons were calculated with overall
equation in the form proposed by Nguyen and Do (ND method)26
a)
∫
rk(p)
rmin
fV(x) dx +
∫
rmax
rk(p)
w t(p,x)fV(x) dx x - σsf/2
(1)
where rmin and rmax are the minimal and maximal half-widths of pores, respectively, w ) 1 for slitlike pores, rk(p) is determined by the modified Kelvin equation
rk(p) )
wγνm cos θ σsf + t(p,x) + 2 RgT ln(p0/p)
(2)
and t(p,x) can be computed using the modified BET equation (22) Kruk, M.; Li, Z.; Jaroniec, M.; Betz, W. R. Langmuir 1999, 15, 1435. (23) Dubinin, M. M.; Stoeckli, H. F. J. Colloid Interface Sci. 1980, 75, 34. (24) Fenelonov, V. B. Porous Carbon; Institute of Catalysis: Novosibirsk, 1995. (25) Dubinin, M. M. Adv. Colloid Interface Sci. 1968, 2, 217.
cz (1 - z) [1 + (nb/2 - n/2)zn-1 - (nb + 1)zn + (nb/2 + n/2)zn+1]
t(p,x) ) tm
[1 + (c - 1)z + (cb/2 - c/2)zn - (cb/2 + c/2)zn+1]
(3)
2840
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
where
tm ) am/SBET
(4)
b ) exp(∆/RgT)
(5)
c ) cs exp((Qp - Qs)/RgT)
(6)
cS ) γeE-QL/RgT
(7)
∆ is the excess of the evaporation heat due to the interference of the layering on the opposite pore wall (∆ ≈ 2.2 kJ/mol26), t(p,x) is the statistical thickness of the adsorbed layer, am is the BET monolayer capacity, cs is the BET adsorption coefficient on flat surface, QL is the liquefaction heat, E is the adsorption energy, γ is a constant, Qs and Qp are the adsorption heat on flat surface and in pores, respectively, z ) p/p0, n is the number (noninteger) of statistical monolayers of adsorbate molecules, and its maximal value for a given pore half-width x is equal to (x - σsf/2)/tm, and σsf ) (σs + σf)/2 is the average collision diameter of surface (carbon) and fluid (nitrogen) atoms. The Steele potential27 was used on calculations of Qs and Qp for nitrogen molecule in slitlike pores
U(x,y) ) φ(y) + φ(x - y)
(8)
with
[
( )
φ(y) ) 4πFsσsf2sf∆ 0.2
σsf y
10
- 0.5
( ) σsf y
4
-
]
σsf4
6∆(y + 0.61∆)3 (9)
∆ ) 0.3354 nm is the thickness of a nitrogen monolayer and y is the distance from the central plane of the outermost atom layer of one pore wall. The nitrogen adsorption data were utilized to compute fV(x) distributions with eq 1 using the modified regularization procedure CONTIN28 under non-negativity condition (fV(x) g 0 at any x) with a fixed regularization parameter R ) 0.01. For microporous Norit RBX, a model of slitlike pores was used. For MAST carbons a mixture of gaps between spherical particles11 with slitlike pores was used because of morphological features of these adsorbents (Figure 1). In the case of pores as gaps between spherical particles, eq 2 was used in the form20
[
p0 γvm 1 2 ln ) p RgT rk ((R + t′ + rk)2 - R2)1/2 - rk + R + t′
]
(10)
where R is the radius of nanoparticles, and t′ ) t + σsf/2. To consider two types of the porosity (slitlike pores and gaps between spherical nanoparticles), integral eq 1 can be rewritten as follows
aΣ )
∑ca ) i i
[ i
∑c ∫ i
i
rk,i(p)
rmin
fV,i(x) dx +
∫
rmax,i
rk,i(p)
w
]
ti(p,x)fV,i(x) dx
x - σsf/2
(11)
where ci ) cslit and csph are weight constants determining contributions of slitlike pores and gaps between spherical particles to the total adsorption (i.e., porosity), using the corresponding modified Kelvin equations (eqs 2 and 10, respectively). Equation 11 was solved using two approaches: (i) fV,slit(x) ) fV,sph(x) ) fV(x) (monoregularization with respect to overall fV(x) using modified ND/CONTIN (MNDC) method); and (ii) fV,slit(x) * fV,sph(x) with binary self-consistent (subsequent for fV,slit(x) and fV,sph(x)) regularization with respect to slitlike pores and gaps between spherical particles (initial fV(x) was calculated (26) Nguyen, C.; Do, D. D. Langmuir 1999, 15, 3608; 2000, 16, 7218. (27) Do, D. D.; Do, H. D. Appl. Surf. Sci. 2002, 196, 13. (28) Provencher, S. W. Comput. Phys. Commun. 1982, 27, 213-227; 229-242.
Figure 2. (a) Nitrogen adsorption isotherms and (b, c) the RS plots for MAST carbons and Norit RBX. with the monoregularization). The PSDs obtained with the binary regularization are shown only for MAST-75; however, related structural characteristics are given for all the carbons in Table 1. The fV(x) distributions determined with eqs 1 or 11 and linked to the pore volume can be transformed to the distributions fS(x) with respect to the specific surface area using the corresponding models of pores
fS(x) )
(
)
Vp w f (x) x V x
(12)
where w ) 1, 2, and 3 for slitlike, cylindrical, and spherical pores, respectively. However, the relationship for fS(x) and fV(x) is more
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2841
complex for pores as gaps between spherical particles, since the inner volume of aggregates of primary particle plays a role of pores but both outer and inner surfaces of these aggregates contribute the specific surface area. For a cubic lattice with spherical nanoparticles w ≈ 1.36; however, this value increases for a denser hexagonal lattice. For estimation of deviation of the pore shape from a slitlike one ∆wslit ) SBET/Ssum,slit - 1 (Table 1), eq 12 was used at w ) 1 on calculations of fS(x) for the model of slitlike pores with
Ssum,slit )
∫
xmax
xmin
fS(x) dx )
∫
xmax
xmin
(
)
Vp w f (x) dx (13) x V x
It should be noted that the PSDs of different adsorbents (activated carbons, silica gels, fumed oxides, etc.) calculated using the MNDC method are close to those calculated with Micromeritics DFT software.29-31 The pore size distribution functions were also shown as incremental PSDs (IPSDs)
fV,n(xi) ) 0.5(fV(xi) + fV(xi-1))(xi - xi-1)
(14)
for comparison with IPSDs calculated using the DFT technique. As a whole the IPSDs are inconvenient for comparison of one to another because the intensity in the ith point depends on the interval (xi - xi-1), i.e., the number of terms in the sum ∑i fV,n(xi) ) Vp. Therefore, both PSDs (differential fV(x) ∼ dVp/dx and incremental fV,n(x)) are used. Fractal dimension DAJ was determined using the adsorption data at p/p0 < 0.85 with the Frenkel-Halsey-Hill equation.32 The Fowler-Guggenheim (FG) equation was used to describe localized monolayer adsorption of nitrogen with lateral interactions
θl(p,E) )
Kp exp(zwΘ/kBT) 1 + Kp exp(zwΘ/kBT)
(15)
where K ) K0(T) exp(E/kBT) is the Langmuir constant for adsorption on energetically uniformed sites and the preexponential factor K0(T) is expressed in terms of the partition functions for an isolated gas and surface phases, z is the number of nearest neighbors of an adsorbate molecule (assuming z ) 4), w is the interaction energy between a pair of nearest neighbors, and kB is the Boltzmann constant, e.g., zw/kB ) 380 K for nitrogen.33 Equation 15 was used as a local isotherm θl in the overall adsorption isotherm in the form of the Fredholm integral equation of the first kind solved using the modified CONTIN procedure at a fixed regularization parameter (R ) 0.01). A maximum p/p0 value for an isotherm portion used to compute f(E) corresponded to nearly monolayer coverage Θ ) a/am ≈ 0.99. Detailed analysis of the structural characteristics of the MAST carbons using several adsorption equations with the DFT and MNDC methods was performed because of unusual texture of these adsorbents reflected, e.g., in the shape of nitrogen adsorption isotherms, the great granule strength,17 and features in the adsorption of complex organic compounds. Adsorption of Organics. Batch experiments were undertaken to study the adsorption of ibuprofen onto carbons in proteinfree solutions and in the presence of BSA. To determine the adsorption of the drug from protein-free solutions, samples of carbon (0.03 g) were mixed with solutions of ibuprofen in Tyrode buffer (10 cm3) at a range of concentrations (0.025-0.0025 M). The slurries were shaken at 298 K for 24 h. The residual (29) Gun’ko, V. M.; Do, D. D. Colloids Surf., A 2001, 193, 71. (30) Kowalczyk, P.; Gun’ko, V. M.; Terzyk, A. P.; Gauden, P. A.; Rong, H.; Ryu, Z.; Do, D. D. Appl. Surf. Sci. 2003, 206, 67. (31) Murphy, M. C.; Patel, S.; Phillips, G. J.; Davies, J. G.; Lloyd, A. W.; Gun’ko, V. M.; Mikhalovsky, S. V. In Characterisation of Porous Solids VI; Studies in Surface Science and Catalysis 144; RodriguezReinoso, F., McEnaney, B., Rouquerol, J., Unger, K., Eds.; Elsevier Science: Amsterdam, 2002; pp 515-520. (32) Avnir, D.; Jaroniec, M. Langmuir 1989, 5, 1431. (33) Jaroniec, M.; Madey, R. Physical Adsorption on Heterogeneous Solids; Elsevier: Amsterdam, 1988.
concentration of the solute in the solution was measured by UVvis spectrophotometry (Helios 2, UNICAM) at the wavelength λ ) 263 nm.34 The previously established linear Beer-Lambert relationship was used in the concentration analysis. Dilution was required to operate the analysis in the linear Beer-Lambert region for samples at high concentrations. The amount of ibuprofen adsorbed to carbon was determined as follows
qe )
V(Co - Ce) m
(16)
where qe is the amount of ibuprofen adsorbed on the carbon at equilibrium (mmol/g), V is the volume of solution (dm3), Co is the initial concentration of the solute (mmol/dm3), Ce is the residual (equilibrium) concentration (mmol/dm3), and m is the amount of carbon used (g). To study adsorption of ibuprofen in the presence of BSA, samples of carbon (0.03 g) were mixed with solutions of ibuprofen (10 cm3) at the concentrations of 0.025-0.0025 M in Tyrode buffer with the presence of BSA at a constant concentration of 1.0 g/L (0.1% w/v). At the same conditions, interaction between ibuprofen and BSA (without carbon adsorbents) was measured using a separation method with a Vivaspin 20 mL concentrator, 10000 MWCO PES (Sartorius), using 4 mL of each solution centrifuged at 3000 rpm for 20 min and the ibuprofen concentration was measured in the filtrate. The slurries were shaken for 24 h, and the final solutions were analyzed by UV spectrophotometry at λ ) 263 nm using the method described above. The absorbance of BSA at this wavelength was constant and additive to the ibuprofen absorbance (results are not shown here). A series of fibrinogen solutions (1000, 500, 200, 100, 50, 20, 10, and 5 µg/mL) was prepared from the bulk solution of 1000 µg/mL. Each sample of carbons (50 mg) was equilibrated with 0.2 mL of phosphate buffered saline (PBS) overnight, and then 0.8 mL of the Fg solution at a different concentration was added. Samples were incubated at 37 °C for 1 h followed by the extensive washing of carbons with PBS (five washes with 1 mL of PBS). Adsorbed protein was eluted with 2% SDS in 0.05 M Tris/HCl at pH 6.5 and 37 °C for 30 min. The radioactivity of the protein eluates and the carbon materials was measured using a WALLAC Wizard 1480 automatic gamma counter (LKBWALLAC, Sweden). The amount of adsorbed fibrinogen was estimated from the specific radioactivity of the bulk fibrinogen solution. Four samples of each material were tested. The ibuprofen adsorption equilibrium data were fitted with the Langmuir equation
qe )
q0aCe 1 + aCe
(17)
where a is a constant and q0 is the monolayer capacity (mmol/ g).19,35 To compute the distribution functions of changes in the Gibbs free energy (f(∆G)) on ibuprofen or fibrinogen adsorption, the Langmuir eq 16 was used as the kernel of the adsorption isotherm equation in the form of Fredholm integral equation of the first kind
θ)
z exp(-∆G/RgT)
∫ 1 + z exp(-∆G/R T) f(-∆G) d(-∆G)
(18)
g
where θ ) qe/q0 is the relative adsorption, z ) Ce is the equilibrium concentration, and Rg is the gas constant. The D’ArcyWatt equation36 with two terms37 was also applied as the kernel of the integral equation to estimate the distribution functions of (34) Terzyk, A. P.; Rychlicki, G. Colloids Surf., A 2000, 163, 135. (35) Khan, A. R.; Ataullah, R.; Al Haddad, A. J. Colloid Interface Sci. 1997, 194, 154. (36) D’Arcy, R. L.; Watt, I. C. Trans. Faraday Soc. 1970, 66, 1236. (37) Barton, S. S.; Evans, M. J. B.; MacDonald, J. A. F. Langmuir 1994, 10, 4250.
2842
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
changes in the Gibbs free energy on the ibuprofen adsorption
θ)
[
zK1 exp(-∆G/RgT)
∫ 1 + z exp(-∆G/R T) + g
zK2 exp(-∆G/RgT) 1 - z exp(-∆G/RgT)
Results and Discussion
]
f(-∆G) d(-∆G) (19)
where Ki is a constant which is a measure of ibuprofen attraction to the ith active site (on carbon surface and BSA). The K1 and K2 parameters were predetermined from condition of a minimum of a functional
Φ(K1,K2,β) )
∫ {spline(θ 1
0
exp(z))
-
}
zK2β zK1β 1 + zβ 1 - zβ
2
dz f min (20)
at ∂Φ/∂yi ) 0 (yi ) K1, K2, and β); where spline(θexp(z)) is a cubic spline of the adsorption data. Fluorescence Spectroscopy. Binding of certain ligands to BSA is accompanied by changes in intrinsic tryptophan fluorescence of albumin. Therefore fluorescence spectroscopy is widely used for monitoring the binding of small molecules (e.g., drugs) to serum albumin molecules because of its sensitivity, accuracy, and applicability.13,38,39 The fluorescence intensity was recorded after sequential addition of small amounts of ibuprofen into solutions of a known concentration of protein. Following the addition of ibuprofen, the solution was left to reach equilibrium at ambient temperature for 30 min. Solutions containing 0.23 g/dm3 of BSA were titrated with ibuprofen at the 1.5 × 10-3 to 1.0 × 10-4 M concentration range. The fluorescence intensity F was measured at an excitation wavelength of 284 nm and the emission wavelength of 340 nm. The following equation (eq 21 describing dependence of F on L) can be used to estimate the best-fit values of the F0, F∞, and Kd parameters38
F ) F0 -
(F0 - F∞)L Kd + L
(version 3.1 standard with HF/6-31G(d) and modified to use DFT with B3LYP/6-31G(d)) program package.42
Structural Characteristics of Carbons. The MAST carbons studied here possess atypical structural properties, which can be affected by several factors analyzed below. Detailed investigations of the textural characteristics of these carbons are of importance for a deeper understanding of their interaction with such complex organics as BSA, Fg, and ibuprofen. Enhancement of burnoff leads to an increase in the porosity of MAST carbons (Table 1, Vp, Vmic,BR, Vmes,BR, Vem,gr in cm3/g) and the specific surface area (SBET, SDFT, and SΣ in m2/g), but the fractality (DAJ) and the bulk density (Fb) decrease despite diminution of the particle size (see Figure 1). Therefore, changes in the porosity (Table 1, Vp in cm3/cm3) and specific surface area (SBET in m2/cm3) per cm3 of carbon are smaller than those per gram of carbon. Their maximal values are observed for MAST-58 with not maximal burnoff. Contributions of micropores (Table 1, SDS, Smic, Vmic per gram of carbon) and mesopores (Vmes, Smes,BR, and Rs plots at Rs > 1 in Figure 2) of MAST carbons increase (Vmes/Vp > 0.5) with burnoff as well as the xDS and δDS values (with simultaneous decrease in DAJ). Notice that SDS > Smic (since fDS(x) is overestimated at small x), but changes in both parameter values with burnoff are closely related. These results suggest that the morphology of nanoparticle of all the studied MAST carbons changes only slightly, but the density of particle packing reduces because of decomposition of a fraction of small nanoparticles. A simple equation of the pore volume in spherical carbon granules (Figure 1)
(21)
where F0 is the intensity of fluorescence emission of the albumin solution in the absence of the ligand, F∞ is the intensity of fluorescence emission of the albumin solution saturated with ligand, L is the drug concentration, and Kd the dissociation constant of the ligand-albumin complex. Quantum Chemical Calculations. The ibuprofen molecule HOOC(CH3)CHC6H4CH2CH(CH3)2 and its slightly modified hydrophilic and hydrophobic fragments CH3CH2COOH, CH3COOH, and C6H5CH2CH(CH3)2 were calculated by means of the GAMESS (PCGAMESS versions 6.2 and 6.3, and a current version (September 1, 2003) of GAMESS with DFT)40 and Gaussian 9441 program packages with the Hartree-Fock (HF) and density functional theory (DFT was used with the combined B3LYP exchange and correlation functional) methods using the 6-31G(d), 6-31G(d,p), and 6-311++G(3df,2p) basis sets. The free energy of solvation ∆Gs and other parameters of molecules solved in water (assuming infinite dilution) were calculated using a solvation model SM5.42/6-31G(d) with consideration for the geometry relaxation on solution by means of the GAMESOL (38) Epps, D. E.; Raub, T. J.; Caiolfa, V.; Chiari, A.; Zamai, M. J. Pharm. Pharm. 1999, 51, 41. (39) Parikh, H. H.; McElwain, K.; Balasubramanian, V.; et al. Pharm. Res. 2000, 17, 632. (40) (a) Granovsky, A. A. www http://classic.chem.msu.su/gran/ gamess/index.html. (b) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen, J. J.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S.; Windus, T. L.; Dupuis, M.; Montgomery, J. A. J. Comput. Chem. 1993, 14, 1347. (41) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Gill, P. M. W.; Johnson, B. G.; Robb, M. A.; Cheeseman, J. R.; Keith, T.; Petersson, G. A.; Montgomery, J. A.; Raghavachari, K.; Al-Laham, M. A.; Zakrzewski, V. G.; Ortiz, J. V.; Foresman, J. B.; Cioslowski, J.; Stefanov, B. B.; Nanayakkara, A.; Challacombe, M.; Peng, C. Y.; Ayala, P. Y.; Chen, W.; Wong, M. W.; Andres, J. L.; Replogle, E. S.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Binkley, J. S.; Defrees, D. J.; Baker, J.; Stewart, J. P.; Head-Gordon, M.; Gonzalez, C.; and Pople, J. A. Gaussian 94, Revision E.1; Gaussian, Inc.: Pittsburgh, PA, 1995.
Vem,gr ) Rpack
(
)
1 1 Fb F0
(22)
where Rpack is the coefficient dependent on the type of packing of carbon micro- and nanoparticles, gives values for a cubic lattice close to the Vp values (Table 1, Vp in cm3/g). This result may be considered as evidence of adequacy of the model used for the MAST carbons with spherical granules (Figure 1) consisting of roughly spherical nanoparticles. However, it does not exclude availability of slitlike micropores in these adsorbents. If the nanoparticle packing corresponds to a dense hexagonal lattice that Rpack decreases giving Vem,gr < Vp if these particles are nonporous. Similar carbons are characterized by the great strength17 possibly due to formation of strengthened tight contacts between nanoparticles (i.e., a granule may be considered as an individual particle with pores). Therefore, eq 22 may overestimate the empty volume between nanoparticles. Additionally, the formation of the mentioned dense contacts causes changes in contribution of narrow pores of different sizes. This effect is maximal for MAST-75 (Figures 3-5). These structural features of MAST carbons can be elucidated on analysis of the PSDs calculated using different structural models with the DFT and MNDC methods. Several models of pores are used on calculations of the PSDs: slitlike pores, pores as gaps between nonporous spherical particles and a mixture of slitlike pores and gaps between spherical particles. The model of pure slitlike (42) Xidos, J. D.; Li, J.; Zhu, T.; Hawkins, G. D.; Thompson, J. D.; Chuang, Y.-Y.; Fast, P. L.; Liotard, D. A.; Rinaldi, D.; Cramer, C. J.; Truhlar, D. G. GAMESOL-version 3.1, University of Minnesota, Minneapolis, 2002, based on the General Atomic and Molecular Electronic Structure System (GAMESS) as described in ref 40.
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2843
Figure 3. (a) Fitting of nitrogen adsorption isotherms for MAST-48 and MAST-75 using eq 11. (b) The corresponding pore size distributions calculated using a model of nanoparticles with interior slitlike pores (small values of error bars show good fitting result). PSDs for MAST-75 calculated with (c) the slitlike and slitlike/gap pore models at different average radius (R) of particles and (d) for slitlike pores and gaps between spherical particles calculated with self-consistent binary regularization.
pores shows deviation in the shape of pores (assuming SBET as the reference value) ∆wslit > 0 (Table 1) and this value increases with burnoff (because of diminution of a fraction of the smallest particles). Calculations by the DFT technique with the slitlike pore model also give SDFT < SBET; however 0 < ∆wDFT < ∆wslit with one exception of MAST-37 (Table 1). One may assume that is evidence of inadequacy of the BET equation for microporous carbons. However, deviation in the pore shape from slitlike may be a reason for smaller values of SDFT than SBET. Clearly a fraction of pores has a shape other than slitlike, and
more complex relief of the surfaces corresponds to larger surface area at the same pore volume. A model of pores as gaps between nonporous spherical particles strongly bound one to another cannot provide a great microporosity observed in the MAST carbons. Therefore, a mixture of two types of pores was also used. This model corresponds to deviation in the particle shape from spherical one and deviation in the pore shape from slitlike one. Application of this model with eq 11 to the MAST carbons gives good fitting of the isotherms (see fitting examples in Figure 3a) with small errors (error bars for fV(x) are shown in Figure
2844
Langmuir, Vol. 20, No. 7, 2004
Figure 4. Pore size distributions with respect to (a, b) the pore volume per (a) gram of carbon and (b) cubic centimeter of carbon and (c) the specific surface area per gram of carbon.
3b). The size of nanoparticles affects the PSDs (Figure 3c), and their changes are related to narrow pores at x ≈ 1 nm and mesopores. Gaps between clusters of nanoparticles (i.e., mesopores) may increase with growing particle size independently of the type of particle packing. Binary regularization, BR (average particle radius R ≈ 2.5 nm (MAST-75) and 4 nm (MAST-37)), gives the PSDs (BR-PSD) (Figure 3d) with the shape akin to that obtained with monoregularization with the same pore model. The BR-PSD shows that slitlike pores give a greater contribution to micropores than that of gaps between spherical particles; however, the latter are responsible
Melillo et al.
for mesopores of a large size at x > 12 nm. These PSDs overlap in other regions. A similar picture is observed for other MAST carbons (therefore their BR-PSDs are not shown here). The micropore peak of fV(x) (Figures 3e, 3f, and 4a) slightly shifts toward larger x values and becomes broader with increasing burnoff (compare MAST-37 and MAST1 75). A mesopore broad peak giving ∫25 1 fV(x) dx > ∫0.2fV(x) dx and Vmes/Vp > 0.5 depends on burnoff too. Thus, all the MAST carbons have micropores and are also characterized by broad distributions of mesopores (Figures 3-5) responsible for more than half of Vp. The contribution of narrow pores at x < 2 nm provides the majority of the specific surface area (Figure 4b, and compare SDS and SBET in Table 1) and a significant portion of the porosity. The PSD of mesopores depends on burnoff because of changes in interior pores of granules (Figure 1) or “exterior” pores as gaps between nanoparticles in granules due to faster burnoff of the smallest nanoparticles. Deviation from slitlike pore shape (Table 1, ∆wslit) shows that enhancement in burnoff leads to an increase in the contribution of pores with a nonslitlike shape that gives w > 1 (in eq 12), since ∆wslit > 0. The ∆wslit values can be used to estimate the values of the weight coefficients cslit and csph in eq 11 on calculations of the BR-PSDs and related parameters (Table 1, parameters with the BR subscript calculated using eq 12 modified akin to eq 11). The shape of MNDC fV,n(x) is akin to that of DFT IPSDs (Figures 3 and 5). The difference in their intensity is mainly caused by a different number of points in these distributions (see eq 14). A similar IPSDV in the mesopore range is observed for fumed oxides composed with nonporous spherical particles. The IPSDV calculated using a model of nonporous spherical nanoparticles for fumed silica A-380 (SBET ) 378 m2/g, Vp ) 0.78 cm3/g, and average nanoparticle diameter is 7.2 nm) is shown in Figure 5e as an example. However, this IPSDV demonstrates a very low microporosity of A-380 in contrast to the MAST carbons. Consequently, one can assume that a significant portion of the porosity (and the specific surface area) can be attributed to inner space of both microparticles (granules) and nanoparticles with complex shape. Calculation of the distribution functions of nitrogen adsorption energy f(E) (Figure 6) gives additional confirmation of the proposed model of pores in MAST carbons. The f(E) distributions reveal unusual changes of the highenergy peak. This peak for MAST-58 and -65 is higher than that for other samples, and this peak for MAST-37 is higher than that for MAST-48 and -75 despite smaller porosity and SBET of MAST-37. Diminution of the highenergy f(E) peak for MAST-75 is in agreement with a decrease in the fractal dimension DAJ and contribution of narrow pores (Figures 3-5). If all the pores of MAST-75 were of the slitlike shape, similar effects could not be observed. The nitrogen adsorption isotherms (Figure 2a), the Rs plots (Figures 2b and 2c), the pore size distributions (Figure 4), and all the calculated structural parameters (Table 1) show that the porosity of Norit RBX significantly differs from that of the MAST carbons. Norit RBX is characterized by a small contribution of mesopores, but deviation from the slitlike porosity is relative large (Table 1, ∆wslit). This result may be connected with contribution of pores of the nonslitlike shape formed because of the coating of Norit RBX particles with cellulose. One may assume that the observed structural differences of Norit RBX and MAST carbons is reflected in the adsorption of low molecular (ibuprofen) and high molecular (protein) organics.
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2845
Figure 5. Incremental PSDs of the MAST carbons calculated using Micromeritics DFT and MNDC (fV,n(x)) methods with respect to the pore volume: (a) MAST-37, (b) MAST-48, (c) MAST-58, (d) MAST-65, and (e) MAST-75. (f) The specific surface area (only DFT) with a model of slitlike pores for all samples and a model of pores as gaps between spherical nonporous particles only for MAST-75.
Adsorption of Organics. The ibuprofen molecule includes a polar COOH group, a lowly polar benzene ring, and an isobutyl “tail” (Figure 7). Calculations of the ibuprofen molecule and its fragments (for simplicity the corresponding molecules were calculated instead of the fragments) using the model SM5.42 with HF/6-31G(d) and B3LYP/6-31G(d) show that a polar “head” with COOH provides major contributions to change in free energy of solvation ∆Gs and to the dipole moment of the ibuprofen molecule in the aqueous solution (Table 2). Consequently, the structure of the adsorption complexes of ibuprofen
molecules may depend on the polarity of the carbon surfaces, e.g., amounts of oxygen-containing functionalities forming hydrogen bonds with adsorbate molecules. The presence of polar and charged protein molecules (as well as pH and the salinity of the aqueous solution) influence ibuprofen behavior due to formation of the corresponding complexes.2,7,9-16 On the other hand, the benzene ring in the ibuprofen molecules provides dispersion interaction with nonpolar basal planes of carbons. For instance, the orbital energy of the highest occupied MO (EHOMO) for the hydrophobic fragment is significantly
2846
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
Figure 6. Nitrogen adsorption energy distributions for the MAST carbons calculated using the FG equation: (1) MAST37, (2) MAST-48, (3) MAST-58, (4) MAST-65, and (5) MAST75.
smaller than that of the polar fragment (Table 2). Therefore the benzene ring has a great capability in dispersion interaction with nonpolar carbon (graphitelike) surface fragments. However, the ibuprofen molecule has a 3D (nonplanar) structure (Figure 7) and its dispersion interaction with basal planes of carbons may be smaller than that of individual benzene because of the steric effect. Solvation of the polar COOH group may reduce changes in the free energy of the ibuprofen adsorption (from the aqueous solution) in micropores with no formation of a complete solvate shell. However, this diminution may be smaller than that caused by changes in interaction of the hydrophobic fragment of ibuprofen molecule with water and carbon surface in micropores, i.e.
∆Gtotal ) (∆Gads - ∆Gs) < 0
(23)
where ∆Gads is change in the free energy due to ibuprofen adsorption in micropores (without formation of ‘half’ of the solvation shell) and ∆Gtotal is the total change in the Gibbs free energy on ibuprofen adsorption from the aqueous solution. In the case of ibuprofen adsorption in broader pores where a portion of the solvation shell remains
∆Gtotal ) (∆Gads - φ(x)∆Gs) < 0
(24)
where φ(x) < 1 is determined by a value of a molecule fragment interacting with the surface with removal of a portion of the solvation shell depending on the pore halfwidth x. Clearly charge-controlled adsorption of ibuprofen molecules (e.g., with formation of the hydrogen bonds) occurs through the COOH group, and the benzene ring is responsible for orbital-controlled adsorption to nonpolar basal planes. Protein molecules form a denser layer on hydrophobic surfaces than that on hydrophilic ones.43 Hence the use of polymer-coated microporous carbons may be ineffective for the adsorption of ibuprofen in the presence of proteins. This assumption is confirmed by a low adsorption of ibuprofen on Norit RBX, which significantly decreases (43) Vogler, E. A. Adv. Colloid Interface Sci. 1998, 74, 69.
Figure 7. (a) Charges of atoms of ibuprofen molecule (calculations with B3LYP/6-311++G(3df,2p)//6-31G(d,p)), localization of (b) the highest occupied molecular orbital (HOMO), and (c) the lowest unoccupied MO (LUMO) at the orbital energy EHOMO ) -6.80 eV and ELUMO ) -0.65 eV.
with the presence of BSA in contrast to uncoated MAST carbons (Figure 8). Figure 8 shows the adsorption isotherms of ibuprofen determined for protein-free and protein-containing aqueous solutions. For MAST carbons the amount of ibuprofen adsorbed per mass unit of adsorbent increases (Figure 8a) with increasing Vp and SBET per gram of carbon with parallel enhancement of the mesoporosity (Table 1). On the other hand, when adsorption results are expressed per volume unit of carbon, no large difference between MAST carbons is observed (Figure 8b), which is in agreement with the similarity in their PSDs (Figures 3-5) and morphology (Figure 1). The adsorption capacity of Norit RBX is significantly smaller in BSA-containing solution compared to the MAST carbons because of the difference in the PSDs of these carbons: Norit RBX is a microporous adsorbent with very low contribution of mesopores. Large protein molecules can easily block the entrances to micropores (Figure 9) thus inhibiting ibuprofen adsorption. It should be noted that adsorption of proteins (see fibrinogen adsorption graphs in Figure 10) or other biopolymers (e.g., lipopolysaccharide44) on microporous carbons is typically lower than that on carbons with a larger contribution of mesopores. The Fg adsorption onto (44) Gun’ko, V. M.; Betz, W. R.; Patel, S.; Murphy, M. C.; Mikhalovsky, S. V. Submitted for publication in Carbon.
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2847
Table 2. Parameters of Ibuprofen Molecule and Molecules Corresponding to Its Fragments Calculated Using Solvation Model SM5.42 with the HF/6-31G(d) and B3LYP/6-31G(d) Basis Setsc molecule
∆Gs, kJ/mol
µ, D
-EHOMO, eV
ELUMO,eV
(CH3)2CHCH2C6H4CH(CH3)COOHa (CH3)2CHCH2C6H4CH(CH3)COOHb (CH3)2CHC6H4CH(CH3)COOHa CH3CH2COOHa CH3CH2COOHb CH3COOHa C6H5CH2CH(CH3)2a C6H5CH2CH(CH3)2b
-24 -25 -25 -27 -30 -29 0.3 1.0
2.63 2.41 2.59 2.21 1.99 2.20 0.28 0.36
8.84 6.39 8.87 12.49 7.70 12.55 8.91 6.53
3.69 -0.11 3.72 5.27 0.66 5.20 3.94 0.14
a HF/6-31G(d) basis set. b B3LYP/6-31G(d) basis set. c Note. ∆G is the free energy of solvation, µ is the dipole moment, E s HOMO is the highest occupied molecular orbital, and ELUMO is lowest unoccupied MO.
Figure 8. Ibuprofen adsorption isotherms from (a, b) protein-free solution and (c, d) the solution with BSA on different carbons (each point represents a mean of three experiments). Amount of solute adsorbed per (a, c) gram of carbon and (b, d) cm3 of carbon is plotted against final (equilibrium) concentration. (c) Interaction between ibuprofen and BSA without carbon is shown.
microporous carbon Norit RBX is smaller than that on MAST carbons and characterized by small changes in the Gibbs free energy (Figure 10b). The difference in the Fg adsorption onto MAST-37 and -75 is akin to that observed on the ibuprofen adsorption; however, the isotherm shapes differ because of the scale factor for low (ibuprofen) and high (Fg) molecular compounds. Analysis of the polymer adsorption onto microporous carbons reveals that it occurs only on the outer surfaces of carbon granules.44 Consequently, a major portion of entrances to the inner space of microporous granules may be blocked, reducing the adsorption of low molecular weight compounds. This effect is well observed on the ibuprofen adsorption on microporous Norit RBX in the presence of BSA. A certain
adsorption of ibuprofen on this carbon in the presence of BSA may be connected with the adsorption of drug molecules onto immobilized protein macromolecules. The ibuprofen adsorption equilibrium data were fitted with the Langmuir and D’Arcy-Watt equations. It can be seen from the results presented in Table 3 and good fitting of the isotherms that the Langmuir equation adequately describes the ibuprofen adsorption isotherms on MAST carbons and Norit RBX, both in the presence and in the absence of BSA. The amount of ibuprofen required to form a monolayer on the adsorbent surface was determined with the Langmuir equation (Table 3, q0). This value is in good agreement with the assumption of monolayer adsorption of ibuprofen when compared with the mo-
2848
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
Figure 9. (a) Molecular model of human serum albumin (HSA) (XRD data from PDB (http://www.biochem.ucl.ac.uk/bsm/pdbsum) and a slitlike micropore (x ) 0.5 nm with consideration for σs of carbon atoms). (b) A model of a HSA molecule on a fragment of MAST-75 carbon surface.
lecular size of ibuprofen (estimated from crystallographic data13,38 or quantum chemical calculations (Figure 7)), taking in account the presence of a significant fraction of micropores in carbons. The titration of BSA with ibuprofen produced a concentration-dependent decrease in the tryptophan fluorescence of albumin due to the fluorescence quenching of the single tryptophan residue (Figure 11). The value of the dissociation constant of the ibuprofen-albumin complex was determined using eq 21 and showed a good correlation (R2 ) 0.994), with F0 ) 17.22 ( 0.22, F∞ ) 11.46 ( 0.16, and Kd ) 3.0 × 10-5 ( 7.8 × 10-6 M. Such a low Kd value indicates a strong interaction between ibuprofen and BSA, and it is in agreement with the data published elsewhere.12,13 This effect is also confirmed by the f(-∆G) distribution functions (∆G ) ∆Gtotal) (Figure 12) showing enhanced uniformity of the distribution functions on the ibuprofen adsorption in the presence of BSA (in comparison with the adsorption from proteinfree solution) independently of the type of the kernel (Langmuir or D’Arcy-Watt equations) of the integral equation. This result may be connected to a decrease in the influence of φ(x) on ∆Gtotal in eq 22 because of the BSA adsorption onto the carbon surfaces. Integral equations 18 and 19 give a better fitting than eq 20 (see sample fitting in Figure 12e). Equations 18 and 19 give slightly different distributions (Figure 12); however, a general tendency of BSA effect is the same: narrowing of the f(-∆G) distribution functions with remaining difference between Norit RBX and MAST carbons. The isotherm of “adsorption” of ibuprofen on BSA
(Figure 8c) shows that the interaction of ibuprofen with BSA is weaker than that with MAST carbons. The f(-∆G) distribution calculated using this isotherm and eq 18 (Figure 11b) is narrow and shifts toward low -∆G values. The surface area of BSA molecules (monomers) is higher than 1000 m2/g, and the BSA concentration is 30% of the carbon concentration. However, a number of sites in the BSA molecule, which can strongly bind the ibuprofen molecule, is low.14-16 These factors can be responsible for the observed shape of the isotherm and f(-∆G) on ibuprofen interaction with individual BSA. On the other hand, in the case of BSA immobilized on carbons the f(-∆G) peaks shift toward greater -∆G values. This effect can be considered as evidence of an important role of the carbon surface in absorption of ibuprofen and possibility of decomposition of ibuprofen-BSA complexes on the carbon surfaces. One can assume that relatively small ibuprofen molecules move from BSA sites due to diffusion into narrow pores of carbons characterized by the high adsorption potential. The Norit RBX is characterized by significantly smaller change in the Gibbs free energy of the ibuprofen adsorption, since the f(-∆G) peak shifts toward low -∆G values. This effect is may be caused by a low number of accessible pores appropriate for ibuprofen adsorption onto both initial and BSA-coated Norit RBX and the difference in the hydrophobic/hydrophilic properties of this cellulose-coated carbon and MAST carbons. Notice that in the presence of BSA, changes in f(-∆G) increase toward greater -∆G values even for Norit RBX. This effect suggests binding
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2849
Figure 11. Fluorescence titration of BSA solution, 0.23 g/dm3 with increasing ibuprofen concentration (each point represents a mean on three experiments). The solid line represents the fitting of the data points using eq 21.
Figure 10. (a) Adsorption isotherms of fibrinogen (Fg) on different carbons. (b) The corresponding distribution functions of changes in the Gibbs free energy on Fg adsorption calculated with eq 18. Table 3. Monolayer Capacity (q0) Calculated with Langmuir Equation and Correlation Coefficient (R2) carbon
q0 (mmol/g)
R2
MAST-37 MAST-48 MAST-58 MAST-65 MAST-75 Norit RBX MAST-37 + BSA MAST-48 + BSA MAST-58 + BSA MAST-65 + BSA MAST-75 + BSA Norit RBX + BSA
2.2 ( 0.1 3.1 ( 0.1 3.6 ( 0.1 4.4 ( 0.1 4.9 ( 0.1 7.1 ( 0.1 2.4 ( 0.2 3.1 ( 0.2 3.6 ( 0.2 4.4 ( 0.2 4.8 ( 0.2 1.5 ( 0.3
0.98 0.98 0.90 1.00 0.99 0.97 0.92 0.98 0.98 0.99 0.99 0.96
of a portion of ibuprofen to protein molecules immobilized onto the carbon surfaces. On comparison of the adsorption isotherms it is clear that MAST carbons can remove ibuprofen from both protein-free and protein-containing Tyrode buffer solution. The adsorption capacity of Norit RBX toward ibuprofen is significantly affected by the presence of albumin (Figure 8), while in protein-free solutions it performs comparably with MAST carbons (however, the isotherm shape shows a lower affinity of ibuprofen to Norit RBX surfaces). Similar results are observed on Fg adsorption (Figure 10). This
difference can be connected to the absence of mesopores in Norit RBX (since protein molecules can easily block micropores) in contrast to MAST carbons possessing mesopores with the size larger than the albumin molecule size (Figures 1, 3-5, and 7). The MAST carbons have both micropores (x < 1.0 nm) and mesopores (at x between 1 and 20 nm), and their pore size distribution is significantly broader than that of Norit RBX (Figure 4). Besides, MAST carbon particles are spongy (Figure 1) with large macropores (transport pores) providing penetration of a fraction of protein molecules into carbon granules (Figure 7b). BSA (MW ≈ 67 kDa) structure can be schematically represented by six helical subdomains assembled to form a heart-shaped molecule (HSA shown in Figure 7 is akin to BSA). The shape is asymmetric and can be approximated to an equilateral triangle with a side of ∼8 nm and average depth of ∼3 nm depending on the pH value.11,45,46 In solution the structure of albumin can be approximated to an asymmetric oblate ellipsoid with a diameter of ∼8 nm (which can increase to 14 nm at pH far from its isoelectric point).47 Therefore, no matter what exact structure albumin has in solution, its dimensions should fit well with the dimensions of the mesopores region (up to x ≈ 20 nm) of the MAST carbons. This could explain why these carbon adsorbents efficiently remove both protein bound and unbound ibuprofen from the solutions. In this study the mechanism of interaction between carbon, albumin, and ibuprofen has not been investigated. However, there is an indication that the ibuprofen-albumin complex may partially break down during interaction with the surface of activated carbons. Justification for this effect comes from comparison of q0 values for ibuprofen adsorption from the protein-free and albumin-containing solutions (Table 3). These values do not change in the presence of albumin (with one exception of Norit RBX) suggesting that the monolayer forms with free ibuprofen molecules rather than its albumin complex. Additionally, ibuprofen absorbs better on the MAST carbons than on BSA (Figure 8c). The (45) Carter, D. C.; He, XM.; Munson, S. H.; Twigg, P. D.; Gernert, K. M.; Beth Broom, M.; Miller, T. Y. Science 1989, 244, 1195. (46) Carter, D. C.; He, X. M. Science 1990, 249, 302. (47) Kiselev, M. A.; Gryzunov, I. A.; Dobretsov, G. E.; Komarova, M. N. Biofizika 2001, 46, 423.
2850
Langmuir, Vol. 20, No. 7, 2004
Melillo et al.
Figure 12. Distribution functions of changes in the Gibbs free energy on ibuprofen adsorption from (a, c) protein-free solution and (b, d) the solution with BSA calculated with (a, b) eq 18 (Langmuir isotherm) and (c, d) eq 19 (D’Arcy-Watt isotherm). (e) An example of fitting with eqs 19 and 20 for ibuprofen adsorption on MAST-48 carbon.
poor performance of Norit RBX in albumin solutions could be attributed to the lack of accessibility of its surface to the protein-drug complex. This may be due to at least two factors: (i) the protein-bound complex cannot diffuse easily through the cellulose membrane coating, and (ii) absence of mesopores in the Norit RBX structure and protein molecules adsorb mainly onto the outer surfaces of granules of this carbon. A complexation equilibrium constant for ibuprofen interacting with the carbon surfaces can be written as follows
Kib )
(
)
Cib,sCX,des f-ibγX ψib - ψX exp Cib,eqCs γ0γib kT
(25)
where Cib,s and Cib,eq is the concentration of adsorbed and nonbound ibuprofen, Cs is the concentration of active surface sites (assuming that ibuprofen interacts with both hydrophobic and hydrophilic sites), CX,des is the concentration of solvent molecules desorbed from active sites occupied by ibuprofen molecules, f-ib is the surface activity coefficient, γI is the mean activity coefficient of an i molecule in the solution, ψX and ψib is the mean potential
Ibuprofen Adsorption on Activated Carbons
Langmuir, Vol. 20, No. 7, 2004 2851
energy at the planes of adsorbed solvate or buffer molecules and ions (X) and ibuprofen (ib), respectively. Some of constants can be estimated from a difference in the free energy of ibuprofen in the adsorption and solution states (∆∆Gs ≈ -6 kJ/mol according to SM5.42/PM3 calculations with adsorption of ibuprofen in a slitlike pore shown in Figure 9a). Additionally, adsorbed ibuprofen (for maximal adsorption shown in Figure 8a) occupies approximately 40% of the surface area of MAST carbons and its molecule replaces approximately 10-12 H2O molecules from the surfaces. Assuming that the “specific surface area” of BSA (monomers) is approximately 1000 m2/g, it is possible to estimate
γ ) Kib,BSA/Kib,MAST75
(26)
for the ibuprofen interaction with BSA and, e.g., MAST75. Consideration for only two sites per BSA molecule for strong binding of ibuprofen molecule gives γ ) 1.8 × 10-4 for the maximal amount of adsorbed ibuprofen Cib,max ≈ 6.53 × 10-4 M/g BSA. This Cib,max value corresponds to approximately 44 ibuprofen molecules per BSA molecule, which is in agreement with the NMR investigations giving 38 ibuprofen molecules per HSA molecule.14 Assuming the site density on BSA akin to that on carbon surfaces, γ in eq 26 is equal to ≈3.5 × 10-3. A simple estimation of γ from the f(-∆G) peak values (Figure 12)
γ ) exp(-∆Gib,BSA)/exp(-∆Gib,MAST75)
(27)
gives a similar value γ ≈ 2.8 × 10-3. On the estimation of the γ values with eqs 26 and 27 the equilibrium Cib concentration with the presence of carbons was used. These low γ values (1.8 × 10-4 to 3.5 × 10-3) are in agreement with a weak influence of BSA on the ibuprofen adsorption onto MAST carbons (Figure 8). Apparent disagreement between the values of γ and Kd is caused by the concentration effect, since the strong binding of ibuprofen molecules to BSA occurs only on a minor portion of BSA sites and only for a small number of ibuprofen molecules. A strong structural effect in the case of the ibuprofen adsorption onto microporous carbon Norit RBX in the presence of BSA is caused by the steric factor. Conclusion Granules (0.2-0.5 mm) of the MAST carbons are composed with nanoparticles. Pores in these carbons can be attributed to several types: gaps between spherical particles tightly attached one to another at pore half-width between 0.2 and 25 nm and slitlike micropores in granules.
According to estimations of the specific surface area from the pore size distributions, the shape of narrow pores can deviate from that of slitlike pores. A developed model of a mixture of pores can describe deviation in the particle shape from the spherical one and deviation in the pore shape from the slitlike one. Micro/mesoporous MAST carbons are efficient adsorbents for ibuprofen removal from the aqueous solutions both in the presence and in the absence of serum albumin. The behavior of ibuprofen as a highly protein-bound drug is affected by BSA. Its adsorption onto all the studied adsorbents and especially on cellulose-coated microporous carbon Norit RBX depends on the presence of BSA. The absence of mesopores in the Norit RBX carbon and diffusion limitations imposed by the coating may be responsible for lowering its adsorption capacity toward ibuprofen in the presence of BSA molecules which can easily block the entrances to micropores. The adsorption of ibuprofen increases with increasing burnoff of the MAST carbons (since gaps between nanoparticles increase) when the adsorption values are expressed per gram of adsorbent since contributions of both micro- and mesopores increase per gram of carbon. No large difference in the removal of ibuprofen, however, is observed between these carbons if the adsorption capacity is expressed per volume of adsorbent, since the corresponding differences in their PSDs and the structural parameters per cubic centimeter of carbons are relatively small. The ibuprofen adsorption isotherms are adequately described by the Langmuir and D’Arcy-Watt equations in agreement with a model of monolayer adsorption of ibuprofen molecules in pores. Analysis of the distribution functions of changes in the Gibbs free energy of the ibuprofen adsorption from proteinfree and BSA-containing aqueous solutions shows that, despite low values of -∆G on interaction of ibuprofen with individual BSA, there is a BSA effect since diminution of the nonuniformity of the BSA-covered surfaces is observed. On the other hand, the effective interaction between adsorbent surfaces and ibuprofen remains in the presence of BSA. These results show the potential of polymer-based MAST carbons to remove ibuprofen (as a model of toxic organic compounds) from liquid media with the presence of proteins. Acknowledgment. This work was supported by EPSRC (U.K.), Grant GR/R05154. V.M.G. thanks the Royal Society for financial support of his visit to the University of Brighton (U.K.) and Dr. T. L. Petrenko (ISC, Kiev, Ukraine) for the use of the Gaussian 94 program package. LA0360557