Anal. Chem. 2009, 81, 5655–5662
Near-Infrared Spectroscopic Study on Guest-Host Interactions Among G0-G7 Amine-Terminated Poly(amidoamine) Dendrimers and Porous Silica Materials for Simultaneously Determining the Molecular Weight and Particle Diameter by Multivariate Calibration Techniques N. Heigl,† S. Bachmann,† C. H. Petter,† M. Marchetti-Deschmann,‡ G. Allmaier,‡ G. K. Bonn,† and C. W. Huck*,† Institute of Analytical Chemistry and Radiochemistry, Leopold-Franzens University, Innrain 52a, 6020 Innsbruck, Austria and Institute of Chemical Technologies and Analytics, Vienna University of Technology, Getreidemarkt 9/164-IAC, 1060 Vienna, Austria The guest-host interactions of poly(amidoamine) (PAMAM) dendrimers and porous silica surfaces were investigated by near-infrared (NIR) diffuse reflection spectroscopy. G0-G7 of amine-terminated PAMAM (PAMAM-NH2) dendrimers were analyzed comprising early, mid, and late generations. For early stages, the adsorption process of the partly protonated dendrimers to the negatively charged silica surface strongly depends on the size/shape characteristics of the guest (PAMAM-NH2 dendrimers) and host (porous silica) materials. G0-G4 (15-45 Å) show smaller particle sizes than the pore diameter of the silica (60 Å) and thus have access to the interior surface of the host material. For mid and later stages (G5-G7; 54-81 Å) only low amounts of the dendrimers adsorb to the silica surface due to the inaccessibility to the interior surface. The loading capacity of the silica material with adsorbed PAMAM-NH2 was evaluated by means of capillary zone electrophoresis (CZE), whereas deviations from the theoretical to the effective particle size and molecular weight (MW) was determined by gasphase electrophoretic mobility molecular analysis (GEMMA) and matrix-assisted laser desorption/ionization linear time-of-flight mass spectrometry (MALDIlin TOF-MS). Deviations from the theoretical to the actual values showed a maximum of 13.8% and 28.0% for the particle size and MW, respectively. The NIR absorption spectra show a distinct band at 4932 cm-1 (νsym (NH) + amide II) due to the adsorbed dendrimers. It was found that the absorbance tends to increase with decreasing generation number. On this basis multivariate calibration was performed with the theoretical data and the data obtained by GEMMA and * To whom correspondence should be addressed. Phone: +43 512 507 5195. Fax: +43 512 507 2965. E-mail:
[email protected]. † Leopold-Franzens University. ‡ Vienna University of Technology. 10.1021/ac900375z CCC: $40.75 2009 American Chemical Society Published on Web 06/12/2009
MALDI-lin TOF-MS. All in all, the calculated partial least-squares regression (PLSR) model containing the GEMMA/MALDI-lin TOF-MS reference values showed better results than the models exclusively calculated from the theoretical values. This indicates that the theoretical values do not imply the structural imperfections arising during the synthesis that may be present in the PAMAM-NH2 dendrimers. Over the last 10 years dendrimers have attracted increasing attention in the fields of chemistry, physics, pharmacy, and clinical chemistry due to their assumed highly compact, well-defined globular shape.1-5 The branched molecular architecture, consisting of a core, an interior of shells (generations), and terminal functional groups, makes it easy to control the size, molecular weight, and chemical functionality of the dendrimers. Synthesis can be performed by either the divergent or the convergent approach, meaning the stepwise construction of iterative stages around a desired core to produce mathematically defined core-shell structures.6,7 About 20 years ago the first articles on poly(amidoamine) (PAMAM) dendrimers (Figure 1) were reported,8 representing the best characterized and best understood family of dendrimers: today commercially available and mainly used for biomedical applications like for magnetic resonance imaging (MRI) contrast agents,9 calibration standards,10 or drug delivery (1) Esfand, R.; Tomalia, D. A. DDT 2001, 6, 427–436. (2) Miller, L. L.; Duan, R. G.; Tully, D. C.; Tomalia, D. A. J. Am. Chem. Soc. 1997, 119, 1005–1010. (3) McKendry, R.; Huck, W. T. S.; Weeks, B.; Fiorini, M.; Abell, C.; Rayment, T. Nano Lett. 2002, 2, 713–716. (4) Demadis, K. D. J. Chem. Technol. Biot. 2005, 80, 630–640. (5) Larsen, G.; Lotero, E.; Marquez, M. J. Phys. Chem. B 2000, 104, 4840– 4843. (6) Tomalia, D. A. Aldrichimica Acta 2004, 37, 39–57. (7) Thornton, A.; Bloor, D.; Cross, G. H.; Szablewski, M. Macromolecules 1997, 30, 7600–7603. (8) Tomalia, D. A.; Baker, H.; Dewald, J.; Hall, M.; Kallos, G.; Martin, S.; Roeck, J.; Ryder, J.; Smith, P. Polym. J. (Tokyo) 1985, 17. (9) Fischer, M. V.; Voegtle, F. Angew. Chem. 1999, 111, 934–955.
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
5655
systems.11 It is a well-explored fact that even with a very precise procedure structural imperfections are inevitable during the synthesis of ethylenediamine (EDA)-core PAMAM dendrimers, especially of higher generations (>G6). Three structural imperfections are frequently reported, namely, (i) generational imperfections due to the presence of trailing generations and dimers, (ii) skeletal defects such as missing branches, intramolecular loops, etc., and (iii) substitutional deviations resulting from heterogeneity between molecules due to surface functionalization. To investigate the chemical and physical properties a variety of analytical techniques have been applied for characterizing dendrimers such as size exclusion chromatography (SEC),12 high-performance liquid chromatography (HPLC),13 infrared (IR) and Raman spectroscopy,14-17 fluorescence spectroscopy,18 capillary electrophoresis,19 electron paramagnetic resonance (EPR) spectroscopy,20 and various types of mass spectrometry (MS).21,22 Moreover, several strategies of the adsorption behavior of PAMAM dendrimers onto solid substrates have been developed resulting in potential applications that provide a deeper understanding of the interaction at interfaces.23-32 Parameters like pH value, concentration, dendrimer generation, and characteristics of chemical surfaces of both the solids and the dendrimers are considered to be the main driving forces. Porous silica can be implemented for the investigation of guest (dendrimers)-host (silica) interactions that can give insight on the accessibility of an adsorbed molecule to the host surface. This effect has already been studied by Ottaviani et al.33 by means of EPR spectroscopy. The fact that the external amino groups are partly protonated in aqueous solution leads to strong electrostatic interactions with the Si-OH (10) Mulholland, G. W.; Bauer, B. J. J. Nanopart. Res. 2000, 2, 5–15. (11) Twyman, L. J.; Beezer, A. E.; Esfand, R.; Hardy, M. J.; Mitchell, J. C. Tetrahedron Lett. 1999, 40, 1743–1746. (12) Nourse, A.; Millar, D. B.; Minton, A. P. Biopolymers 2000, 53, 316–328. (13) Islam, M. T.; Shi, X.; Balogh, L.; Baker, J. R., Jr. Anal. Chem. 2005, 77, 2063–2070. (14) Deutsch, D. S.; Siani, A.; Fanson, P. T.; Hirata, H.; Matsumoto, S.; Williams, C. T.; Amiridis, M. D. J. Phys. Chem. C 2007, 111, 4246–4255. (15) Popescu, M.-C.; Filip, D.; Vasile, C.; Cruz, C.; Rueff, J. M.; Marcos, M.; Serrano, J. L.; Singurel, G. J. Phys. Chem. B 2006, 110, 14198–14211. (16) Davis, A. P.; Ma, G.; Allen, H. C. Anal. Chim. Acta 2003, 496, 117–131. (17) Tabakovic, I.; Miller, L. L.; Duan, R. G.; Tully, D. C.; Tomalia, D. A. Chem. Mater. 1997, 9, 736–745. (18) Shcharbin, D.; Bryszewska, M. BBA-Gen. Subjects 2006, 1760, 1021–1026. (19) Peterson, J.; Ebber, A.; Allikmaa, V.; Lopp, M. Proc. Estonian Acad. Sci. Chem. 2001, 50, 156–166. (20) Han, H. J.; Sebby, K. B.; Singel, D. J.; Cloninger, M. J. Macromolecules 2007, 40, 3030–3033. (21) Kallos, G. J.; Tomalia, D. A.; Hedstrand, D. M.; Lewis, S.; Zhou, J. Rapid Commun. Mass Spectrom. 1991, 5. (22) Mu ¨ ller, R.; Laschober, C.; Szymanski, W. W.; Allmaier, G. Macromolecules 2007, 40, 5599–5605. (23) Liu, D.; Gao, J.; Murphy, C. J.; Williams, C. T. J. Phys. Chem. B 2004, 108, 12911–12916. (24) Esumi, K.; Fujimoto, N.; Torigoe, K. Langmuir 1999, 15, 4613–4616. (25) Tsukruk, V. V. Adv. Mater. 1998, 10, 253–257. (26) Tully, D. C.; Frechet, J. M. J. Chem. Commun. 2001, 1229–1239. (27) Pericet-Camara, R.; Papastavrou, G.; Borkovec, G. M. Langmuir 2004, 20, 3264–3270. (28) Rahman, K. M. A.; Durning, C. J.; Turro, N. J.; Tomalia, D. A. Langmuir 2000, 16, 10154–10160. (29) Goino, M.; Esumi, K. J. Colloid Interface Sci. 1998, 203, 214–217. (30) Esumi, K.; Goino, M. Langmuir 1998, 14, 4466–4470. (31) Esumi, K.; Saika, R.; Miyazaki, M.; Torigoe, K.; Koide, Y. Colloid Surf. A 2000, 166, 115–121. (32) Imae, M. U. T. J. Colloid Interface Sci. 2006, 293, 333–341. (33) Ottaviani, M. F.; Turro, N. J.; Jockusch, S.; Tomalia, D. A. J. Phys. Chem. B 2003, 107, 2046–2053.
5656
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
Figure 1. G3 EDA-core PAMAM-NH2 dendrimer.
(hydrophilic), Si-O-Si (hydrophobic), and Si-O- groups depending on the pH.34-36 The size and shape characteristics of the guest (particle size) and host material (pore size) are crucial for the adsorption of the dendrimers to the internal surface of the porous material, as can be gathered from Figure 2. Here we report an investigation on the adsorption and interaction of PAMAM-NH2 dendrimers (G0-G7) onto porous silica with NIR diffuse reflection spectroscopy37 in combination with multivariate data analysis (MVA). Furthermore, GEMMA and MALDI-lin TOF-MS were drawn as reference techniques for building a multivariate calibration model, whereas CZE was utilized to quantify the amount of adsorbed dendrimers onto the silica material. Generally, MVA-based calibration techniques are applied to combine the NIR spectral data with target parameters transferred from reference techniques or to expose similarities and hidden data structures in the spectra.38-41 NIR radiation impinging on a sample leads to overtone and combination vibrations showing lower absorptivities compared to MIR absorption bands. Reflection and scattering from the silica surface can lead to increasing penetration depths of the NIR radiation and thus providing important information about the adsorbed dendrimers to the interior silica surface, even at deeper sample layers. In contrast, MIR mainly interacts with the external sample layers, providing information primarily about the scanned surface.42 The fact that every material and surface structure has its unique physicochemical fingerprint in the NIR region makes near-infrared spectroscopy (NIRS), especially coupled to diffuse reflection devices and combined with MVA, a powerful tool in the field of material science and, in particular, for the simultaneous expression (34) Reynhardt, J. P. K.; Yang, Y.; Sayari, A.; Alper, H. Adv. Funct. Mater. 2005, 15, 1641–1646. (35) McCain, K. S.; Schluesche, P.; Harris, J. M. Anal. Chem. 2004, 76, 939– 946. (36) Cahill, B. P.; Papastavrou, G.; Koper, G. J. M.; Borkovec, M. Langmuir 2008, 24, 465–473. (37) Siesler, H. W.; Ozaki, Y.; Kawata, S.; Heise, H. M. Near-Infrared SpectroscopyPrinciples, Instruments, Applications; Wiley-VCH: Germany, 2002. (38) Esbensen, K. Multivariate Data Analysis-In Practice, 5th ed.; CAMO: Oslo, Norway, 2005. (39) Kowalski, B. R. Chemometrics-Mathematics and Statistics in Chemistry; D. Reidel: Dordrecht (Holland), 1984. (40) Martens, H.; Martens, M. Multivariate Analysis of Quality, Wiley & Sons Ltd.: New York, 2001. (41) Massart, D. L.; Vandeginste, B. G. M.; Deming, S. N.; Michotte, Y.; Kaufman, L. Chemometrics: a Textbook; Elsevier: New York, 1988. (42) Dahm, D. J.; Dahm, K. D. Interpreting Diffuse Reflectance and Transmittance: IM Publications: Chichester, 2007.
Figure 2. Idealized guest-host interaction of dendrimers with varying particle size at the silica surface having a constant pore size. Dendrimers showing smaller particle size than the pore size have access to the interior surface of the silica material. Larger sized molecules can only interact with the silica surface. Øp is the pore radius; Ød is the dendrimer radius.
of chemical and physical parameters within a short time.43 In summary, the main objective of this work was to set up a noninvasive and rapid method based on NIRS for simultaneously determining the particle size and molecular weight via multivariate calibration techniques based on guest-host interactions of PAMAMNH2 dendrimers with porous silica. EXPERIMENTAL SECTION All PAMAM-NH2 dendrimers with an EDA core (G0-G7) were purchased as 5-20% standard solutions in methanol from Sigma-Aldrich (St. Louis, MO). Thin-layer chromatography (TLC) plates (Polygram Sil G, specific surface area (BET) ) 500 m2/g, mean pore size ) 60 Å, specific pore volume ) 0.75 mL/g, particle size ) 5-17 µm) were purchased from Macherey-Nagl (Du ¨ ren, Germany). Methanol (purity g 99.9%) for washing the TLC plates was obtained by Aldrich (Milwaukee, WI). Capillary Zone Electrophoresis (CZE). Chemicals. Diethylenetriamine, ReagentPlus 99% (DIEN), and o-phosphoric acid (85%) were purchased from Sigma-Aldrich (St. Louis, MO). Polyimide-coated fused silica capillaries (75 µm i.d. and 360 µm o.d., TSP 0750375) were obtained from Polymicro Technologies (Phoenix, AZ). Standard and Capillary Zone Electrophoresis Background Electrolyte (CZE BGE) Solution. Stock solutions (5000 ppm) of low-, mid-, and high-generation dendrimers (G0, G4, G7) in MeOH were diluted with H2O to the desired concentrations for further incubation and calibration (1000-4000 ppm). The CZE BGE was prepared at a concentration of 40 mM DIEN. The pH (pHmeter type Multilab 540, WTW, Weilheim, Germany) was adjusted to 3.0 by adding phosphoric acid. Sample Preparation. One milligram of silica (scratched from the TLC plate) and 300 µL of the dendrimer solutions (G0, G4, G7 at varying concentrations) were mixed in a 2 mL vial. The mixture was incubated for 1 h (Thermomixer comfort, Eppendorf, Germany) at 25 °C and 1400 rpm. Subsequently, it was centrifuged for 5 min at 1300 rpm (Centrifuge, Eppendorf, Germany). Finally, the supernatant was analyzed by CZE. Instrumentation. All CZE experiments were performed on an Agilent HP 3D-CE (Waldbronn, Germany) instrument equipped (43) Heigl, N.; Petter, C. H.; Rainer, M.; Najam-ul-Haq, M.; Vallant, R. M.; Bakry, R.; Bonn, G. K.; Huck, C. W. J. Near Infrared Spectrosc. 2007, 15, 269– 282.
with a diode array detector (190-600 nm). Data processing of the UV spectra was done by Agilent ChemStation Rev A.06.01 software. Analyte signals were detected at 214 nm. A fused silica capillary with a total length of 40 cm and an effective length of 32.5 cm was used throughout the experiments. Sample injection was performed at 50 mbar for 4 s by pressure injection at 25 °C and +15 kV voltages. New capillaries were conditioned by flushing with 0.1 M NaOH (for 20 min) followed by water (for 15 min). Finally, the conditioning procedure was completed by flushing with the CZE BGE for 20 min. Prior to each run the capillary was purged with the CZE BGE for 3 min. Between days or after a change of separation electrolyte, the capillary was conditioned by rinsing successively with water and the separation electrolyte for 20 min, respectively. The capillary was stored overnight filled with water. Matrix-Assisted Laser Desorption/Ionization Linear Timeof-Flight Mass Spectrometry (MALDI-lin TOF-MS). Chemicals. Ammonium acetate (g99.0%) and trihydroxyacetophenone (THAP) were purchased from Fluka (Buchs, Switzerland). Cytochrome c (Cyt C) and bovine serum albumin (BSA) were obtained from Sigma-Aldrich (St. Louis, MO). Sample Preparation. Sample concentration calculations are based on the ideal molecular weights of the PAMAM-NH2 dendrimers (Table 1). Five micromolar solutions were prepared by diluting each stock solution (G0-G7) with 25 mM ammonium acetate buffer (pH 6.8). Ten milligrams of THAP/mL in MeOH was used as MALDI matrix solution. For sample preparation, 0.4 µL of 25 mM ammonium acetate buffer was deposited on a stainless steel target, and onto that droplet 0.6 µL of dendrimer sample (equivalent to 3 pmol) was pipetted. Directly into that droplet 0.8 µL of THAP solution was pipetted, causing the sample volume to spread over the sample spot area. The sample was allowed to dry at room temperature. Instrumentation. MALDI mass spectrometry was performed on an AXIMA CFRplus instrument (Shimadzu Biotech, Kratos Analytical, Manchester, U.K.) equipped with an ultra-high mass detector (HM1 High-Mass System, CovalX, Zu ¨ rich, Switzerland). Data are the average of 200-500 unselected single laser shots and were smoothed using the company-supplied SavitzkyGolay44 algorithm; mass calibration was performed externally using the singly protonated molecular ions of THAP, Cyt C, (44) Savitzky, M.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627–1639.
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
5657
0 0 0.6 4.2 6.7 11.2 15.7 28.0 a
0 1 2 3 4 5 6 7
n.a. ) not ascertainable with this method. n.m. ) not measured. b Theoretical values were obtained by the manufacturer (Dendritech).
0.52 1.43 3.24 6.63 13.32 25.92 50.16 90.97 0.52 1.43 3.26 6.91 14.21 28.82 58.05 116.49 n.a. n.a. 13.8 n.m 4.6 5.9 4.7 6.6 n.a. n.a. 33 n.m. 43 51 64 76 4 8 16 32 64 128 256 512
PAMAM-NH2 generation
15 22 29 36 45 54 67 81
MALDI-lin TOF-MS-derived molecular weight [kDa]
difference between theoretical and measured molecular weight [%] theoretical number of NH2 groups based on an ideal structureb
theoretical particle diameter based on an ideal structureb [Å]
GEMMA-derived particle diameter [Å]
difference between theoretical and measured particle diameter [%]
theoretical molecular weight based on an ideal structureb [kDa]
molecular weight particle diameter
Table 1. GEMMA- and MALDI-lin TOF-MS-Derived Data Compared to the Theoretical Valuesa 5658
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
and BSA and the singly charged dimer of BSA (m/z 200-130 000). Molecular weights are average data of 5 independently prepared sample spots. Gas-Phase Electrophoretic Mobility Molecular Analysis (GEMMA). Sample Preparation. Sample preparation was done according to MALDI-lin TOF-MS experiments resulting in 5 µM dendrimer solutions in 25 mM ammonium acetate buffer. Instrumentation. For all measurements, independent of the size of the dendrimers, the nanoelectrospray (nES) source settings and the nanodifferential mobility analyzer (nDMA) settings were identical. A pressure of 4 psi was applied on the sample solution for sample introduction (final flow rate ) 67 nL/min) through a fused silica capillary (25 µm i.d., 160 µm o.d., 25 cm length, uncoated, Polymicro Technologies, Phoenix, AZ). For optimum spray conditions, 2 kV spray voltage, 0.3 L CO2/min, and 1 L/min compressed synthetic air (both Air Liquide, Vienna, Austria) were applied. The GEMMA instrument consists of a nES source with a Po-210 radioactive source (NRD, Grand Island, NY), a nDMA, and an ultrafine condensation particle counter (uCPC) as the detector (all parts TSI Inc., Shoreview, MN). The multiply, positively charged ion produced in the nES source are charge reduced to singly charged ions by the Po210 source. The nDMA was operated in the negative polarity mode for the central electrode. Ten scans across the selected sizing range (3-10 nm) were averaged for a final GEMMA spectrum from which the average electrophoretic mobility diameter (EMD) of the particle was obtained. Using an external calibration curve, which is based on the correlation of the measured EMD of well-defined globular proteins and their respective molecular weights,45 the molecular mass of the dendrimers was calculated. Near-Infrared Spectroscopy (NIRS). Sample Preparation. First the TLC plates were cut into 20 × 10 mm pieces. Five microliters of PAMAM-NH2 solution (G0-G7, c ) 1% in H2O) was spotted onto the surface. The same procedure was repeated three times for each generation (24 plates). The loaded plates were allowed to dry for 15 min at room temperature. After that, every single plate was washed three times with fresh MeOH and brought into an oven at 105 °C to get rid of water adsorbed onto the surface. Before taking the NIR spectra the loaded and washed TLC plates were dried for 1 h at room temperature. Finally, five NIR spectra were recorded of each sample in diffuse reflection mode (120 spectra). The limit of detection (LOD) was found to be 0.125% in H2O for G0, 0.400% for G4, and 0.840% for G7, respectively. Instrumentation. NIR spectra were recorded with a scanning polarization interferometer Fourier-transform NIR spectrometer (FT-NIR) (NIRVis 1013; Buchi, Flawil, Switzerland) equipped with a tungsten-halogen lamp and a temperated lead-sulfide detector (30 °C). The instrument offers a resolution of 12 cm-1, an absolute wavelength accuracy of ±2 cm-1, and a relative wavelength reproducibility of 0.5 cm-1 between 4000 and 10 000 cm-1 (2500 and 1000 nm). For reflection spectra recording a horizontally coupled integrating sphere device were utilized. Chemometric software NIRCal 4.21 (Buchi) was used for creating the principal component analysis (PCA) and partial least-squares (45) Bacher, G.; Szymanski, W. W.; Kaufman, S. L.; Zo¨llner, P.; Blaas, D.; Allmaier, G. J. Mass Spectrom. 2001, 36, 1038–1052.
(PLS) regression models. The Unscrambler v9.6 (CAMO, Oslo, Norway) was drawn for comparison. For testing the models the collected spectra were divided into a learning set (c-Set, 67%) and a test set (v-Set, 33%), both consisting of independent samples. Measurements were carried out at room temperature (23 °C) from 4000 to 10 000 cm-1. Quantitative Analysis. The optimum number of factors for creating a PLSR model was determined by the predicted residual error sum of squares (PRESS) that shows the sum of squares of deviation between predicted and reference values.
∑ (x
PRESS )
n
- yn)2
(1)
Selection of the best quantitative regression model is based on the following calculated values. (1) BIAS, i.e., the average deviation between predicted values (yn) and actual values (xn), in the calibration set. Bias )
1 N
∑ (x
n
- yn)
(2)
(2) Standard error of estimation (SEE): the standard deviation of differences between reference values (obtained from another technique) and NIRS results in the calibration set.
SEE )
N - 1r - 1 ∑ (x
n
- yn - Bias)2
(3)
(3) Standard error of prediction (SEP), i.e., the standard deviation of differences between reference values and NIRS results in the validation set.
SEP )
N1 ∑ (x
n
- yn - Bias)2
(4)
(4) The quality of quantitative calibration data is described by the correlation coefficient (R) defined as
R)
∑ (x
n
- ¯x) - (yn - ¯y )
N
∑ (
N
(xn - ¯x)
2
∑ (y
(5) 2
n
- ¯y ) )
N
where xn represents the true (reference) values, yn the predicted values, ¯x and ¯y are the mean values, and N is the number of spectra in the calibration sets. RESULTS AND DISCUSSION Potential applications of dendrimers at surfaces become more and more important, e.g., for the modification of solid surfaces to create problem-specific composite materials that are compatible with different kinds of environments, such as biomedium, or used for nanopatterning or as surface-based sensors. As already described in the literature, PAMAM-NH2 dendrimers are partly protonated in aqueous solution (pH ≈ 8.5), which leads to chemical and electrostatic interactions of the positively charged dendrimers and the negatively charged silica surface.33 Basic surface conditions of the host material (porous silica) lead to
Table 2. Capacities of the Silica Material Loaded with Low-, Mid-, and High-Generation PAMAM-NH2 Dendrimers
PAMAM-NH2 generation G0 G4 G7
before incubation
after incubation
c [ppm]
PAMAM-NH2 c (supernatant) bound on SiO2 [ppm] [µmol/mg]
1000 2000 4000 1000 2000 4000 1000 2000 4000
689 1106 2990 731 1527 3034 909 1730 3460
510 1470 1660 20 30 70 0.8 2.3 4.6
an increased amount of adsorbed dendrimers and vice versa; around neutral conditions it is an intermediate case. Another reason for the strong interaction is the ratio between the pore size of the host material to the particle size of the guest particle. For this study, silica with an average pore diameter of 60 Å was taken to investigate the adsorption behavior of the PAMAM-NH2 dendrimers ranging from 15 to 81 Å (theoretical values), as can be gathered from Table 1. Theoretically, as long as the particle diameter is smaller than the pore diameter, the relatively flexible structure of low-generation dendrimers (up to G4) allows accessing the interior surface of the host material. Particles larger than the pores of the host only adsorb onto the silica surface. CZE. In order to evaluate the interactions between the dendrimers and the silica and, furthermore, confirm the above made presuppositions CZE was implemented. Therefore, representative samples for low (G0), mid (G4), and high (G7) generations were probed while the calibration plots of peak area versus concentration for at least five data points per concentration in a range from 1000-4000 ppm were taken into consideration. The supernatant solutions after the adsorption process were compared to the dendrimer solution before the adsorption process. The CZE calibration curves showed high linearity (R2G0 ) 0.998, R2G4 ) 0.995, R2G7 ) 0.992) over the concentration range. It was found, as expected, that the quantity of low-generation dendrimers (G0) adsorbed to the silica is higher than that for mid- and high-generation dendrimers (G4 and G7), as can be gathered from Table 2. Low-generation dendrimers (G0-G3) have access to the pores of the silica material or the interior surface, which shows a much higher surface area than the outer surface and hence adsorbed dendrimers to a greater extent. Mid generations (G4, G5) may have partial access to the pores. It is assumed that the flexibility of the particles plays a crucial role whether the interior surface is accessible or not. In the case of the higher generations (G7), the CZE experiments showed clearly that only a small fraction of the guest particles adsorb to the surface of the silica material, which may be due to the inaccessibility to the interior surface. MALDI-lin TOF-MS. It is known that structural defects may occur during synthesis of dendrimers, especially for higher generations, leading to discrepancies between theoretical values and the actual molecular weight and respective particle size. Using MALDI-lin TOF-MS, this fact was already observed for G0 where m/z values with multiple mass differences of ±114 Da were Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
5659
Figure 3. (a) NIR absorption spectra of the pure silica (dashed line) and dendrimer-loaded silica from 4000 to 7500 cm-1. (b) Second-derivative spectra (Savitzky-Golay, 9 points) showing increasing absorbance intensity at 4932 cm-1 (PAMAM-NH2) with decreasing particle size or generation. It has to be considered that absorbance signals in a show negative values in b.
Figure 4. (a) NIR absorbance intensities (Figure 3) visualizing the dendrimer/silica interaction at wavenumbers rich of physicochemical information and (b) showing the regression coefficients from 4000 to 10 000 cm-1, while the wavenumbers indicating significant linearities regarding particle size and MW are highlighted. Table 3. Characteristic Vibrations Observed in the NIR Absorption Spectra from 4000 to 7500 cm-1a wavenumber [cm-1] 4212 4404 4416 4536 4932 5136 5268 5640 7116 7308
molecular vibration 3 × δ (OH) ν (OH) + δ (OH) ν (CH) + ν (CdO) νsym (NH) + amide II ν (OH) + δ (OH) 2 × ν (CH) 2 × ν (OH) 2 × ν (CH) + δ (CH)
a “ν” is for stretching, “δ” is for bending mode, “2 ×” indicates the first overtone, 3 × indicates the second overtone.
observed, indicating incomplete (up to 2 missing branches) or unstopped (up to 7 additional branches) synthesis and also intramolecular ring formation (-60 Da). It is also a well-known analytical problem to prepare homogeneous dendrimer formulations after synthesis containing only a single component. This was also observed for higher generation dendrimers, always showing 5660
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
a noteworthy presence of the last two generations (with all their incomplete and unstopped byproduct). This easy to follow observation for smaller generations (G0-G3) can be one explanation for the increasing mass defect observed for higher generation dendrimers (G4-G7) (Table 1). During synthesis easily accessible amino groups, like those from smaller particles still present due to incomplete synthesis, are more easily linked to the next branch than others. Another reason may be the increasing surface density hampering complete conversion of one generation to the next as the increasing number of terminating amino groups also increases the number of possible charges and subsequently repulsive forces during synthesis. A linear decrease, starting with G3, of actual molecular masses compared to theoretical values was ascertained. GEMMA. A mass defect, explainable by the same considerations mentioned before, was also observed for the determined EMDs. Although the dendrimers’ diameters increased constantly with the increasing number of generation, the determined diameters differed significantly from the manufacturer’s data. For higher generations (G5-G7) also particles of smaller size were detected, which again could be correlated to the diameter of the
Figure 5. Three-dimensional factor plot representing PCA one, two, and three.
previous generation, corroborating the hypothesis of the present byproduct of incomplete synthesis or unremoved residues of former generations. For the GEMMA data the deviation of the experimental from theoretical data did not follow a linear relation. NIRS. Vibrational spectroscopic techniques such as Raman, IR, and NIR spectroscopies offer the noninvasive and fast analysis of solid samples, while foremost NIRS can provide information about deeper sample layers and allows for multivariate calibration. Vibrational frequencies can be very sensitive to small changes in the bonding and geometrical arrangements and provide lots of chemical information simply by visual spectra interpretation. Chemometric spectra transformations such as derivatives, smoothing, scatter corrections, and normalizations can indicate hidden data structures and correct for unwanted effects like scattering due to, for example, inconsistent particle size.38,40 For this study silica in the form of commercially available TLC plates was taken as a host material that provides a reproducible, uniform sample surface and thickness (0.25 mm), which is to prevent unwanted scatter effects in the spectra. The 24 PAMAM-NH2-loaded TLC plates were placed onto a horizontally coupled integrating sphere device, and 5 spectra, each representing the average of 10 scans, were consecutively recorded on each of the plates,
resulting in a total of 120 NIR spectra. Figure 3a shows the averaged, pretreated (multiplicative scatter correction (MSC)) absorption spectra, whereas Figure 3b depicts the secondderivative (Savitzky-Golay, 9 points) spectra. It can be clearly seen that the adsorption of different PAMAM-NH2 generations onto the silica produces distinct absorption bands in the nearinfrared. Moreover, the absorption intensity at specific wavenumbers exhibits a linear trend with the number of generation, which can be seen in Figure 4a. In more detail, the absorption band at 4932 cm-1, representing a νsym (NH) + amide II combination band, is aroused exclusively by the adsorbed dendrimers; presumably the terminal NH2 functionalities are responsible for that. Even offset effects (parallel shifts toward higher absorbances) from 5000 to 7400 cm-1 and shifts toward higher wavenumbers in the 4100-4600 cm-1 region occur due to dendrimer adsorption. The detailed list of absorption bands is gathered in Table 3. G0-G4 show the most intense absorbance at 4932 cm-1, indicating that larger amounts of lower generation dendrimers (G0-G4) adsorb to the silica than higher generations (G5-G7) do. As already mentioned, the small particles have access to the pores or, in other words, more dendrimers adsorb onto the silica surface. These assumptions correspond well to the obtained NIR spectra and absorption characteristics at specific wavenumbers, as can be seen in Figures 3 and 4. The calculated PCA brings out the correlations of special features within the individual spectra or samples. The 3-dimensional factor plot in Figure 5 represents principal components (PC) one, two, and three, whereas PC one and two describing >95% of the total variance. Spectra of G0-G4 are placed along the increasing values (from left to right) of PC 1, with growing number of generation and thus increasing particle size and molecular weight, respectively, which indicates a related property of G0, G1, G2, G3, and G4. The mentioned generations all have access to the interior surface of the host due to their small particle size (15-45 Å) compared to the pore size of the silica (60 Å). G5 and G6 are classified into a separate quadrant, toward the negative values (top down) of PC two. This may be an indicator for the decreasing possibility of the larger dendrimers to gain access to the pores. G7, which most likely only adsorbs onto the outer host surface and does not penetrate the pores, is positioned remote to all other samples and shows the most negative value on PC two. The cluster
Table 4. Calibration Parameters for the Calculated PLSR Models method
PLS
calibration wavenumbers data pretreatment calibration factors
4000-7608 cm-1 multiplicative scatter correction (MSC) 7
particle size [Å]
min/max SEE decline (SEE) [%] SEP decline (SEP) [%] R2 a
molecular weight [kDa]
manufacturer (Dendritech)
GEMMA (G0,G1,G3 ) theoretical valuea)
manufacturer (Dendritech)
MALDI-lin TOF-MS (G0 ) theoretical value)
15/81 4.3
15/76 3.6
0.52/116.49 11.4
0.52/90.97 8.5
4.1
11.7
19.4 5.1
24.4
34.1 0.9556
9.0 30.0
0.9629
0.9063
0.9157
Theoretical values were obtained by the manufacturer (Dendritech).
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
5661
model can be well interpreted in terms of the guest-host interactions and demonstrates the interrelationship between the dendrimer generations and the absorbance intensities. In summary, it can be stated that the NIR absorbance signal linearly increases/decreases, especially at 4932 cm-1, according to decreasing number of generation, particle size, or molecular weight. The dependence of absorbance intensities to the respective particle size and molecular weight of the dendrimers can be used to calculate a quantitative partial least-squares regression (PLSR) model,38-41 predicting the particle size and molecular weight of unknown samples. Here it is intended to simultaneously determine the mentioned sample properties; hence, the same PLS model has to be applied for predicting the particle size and MW. Table 4 highlights the calibration parameters, whereas particular attention has to be drawn to the standard error of prediction (SEP) that gives the standard deviation of the predicted (NIRS) to the reference values (MALDI-lin TOF MS, GEMMA). Two models were created for determining the particle size and molecular weight, respectively; one is exclusively calculated with the theoretical values obtained by the manufacturer and the other with additional values collected by MALDI-lin TOF-MS (molecular weight) and GEMMA (particle size). In the case where no reference values could be received, the theoretical values were used. The results showed a significant decrease of the SEE at 19.4% and the SEP at 24.4% for the particle size and 34.1% and 30.0% for the molecular weight, respectively. Moreover, an enhancement of all relevant calibration parameters, itemized in Table 4, for the model established with the MALDI-lin TOF-MS and GEMMA reference data was achieved. The theoretical values do not imply the structural deformations that may be present in the samples, which becomes evident by the SEP that could be decreased from 5.1 to 4.1 Å and from 11.7 to 9.0 kDa. It can be stated clearly that the conducted experiments give more precise reference values for calculating a NIR-spectra-based PLSR model than the theoretical data. NIRS can be counted, besides MALDIlin TOF and GEMMA, to the integral methods, meaning that
5662
Analytical Chemistry, Vol. 81, No. 14, July 15, 2009
single particles are not investigated but large representative particle groups, which provides realistic insight on the interaction of the guest-host interactions on a larger scale. CONCLUSION In the case of the molecular weight, relatively large deviations from the theoretical values to the measured values were found (up to 28.0%), whereas for the particle size deviations only up to 13.8% were detected. The fact that the mass defects steadily rise with increasing number of generation but the particle size defects remain constant leads to the presumption that the inner branches of mid- and high-generation PAMAM-NH2 dendrimers are not completely developed or missing. On the other hand, the dendrimer scaffold seems to be almost completely developed, as indicated by the small deviations from the measured to the theoretical values for the particle size, even for mid and higher generations. NIR diffuse reflection spectroscopy has turned out to be a highly suitable analytical tool for the investigation of the guest-host interactions of PAMAM-NH2 dendrimers and porous silica material. Furthermore, the results demonstrate the high efficiency of multivariate techniques to calibrate for changes that take place at the interface of porous silica with PAMAM-NH2 dendrimers. In the case studied, the established PLSR models give an accurate idea of the evolution of the guest-host system, which certainly leads to a high measuring expenditure, but as soon as the NIRS-based model is calibrated, it can be used to much several quantitative analytical information, such as particle size and molecular weight, simultaneously within a very short period of time. ACKNOWLEDGMENT The authors thank Leopold-Franzens University, Innsbruck, Austria, for financial support (Nachwuchsfo¨rderung 2008). Received for review February 18, 2009. Accepted May 29, 2009. AC900375Z
