Experimental Determination of Quantum Dot Size Distributions, Ligand

Kim E. Sapsford , Katherine M. Tyner , Benita J. Dair , Jeffrey R. Deschamps , and ... Jennifer Amey , Nathaniel J. Miska , Lisa Pang , Corey R. Hine ...
0 downloads 0 Views 1MB Size
NANO LETTERS

Experimental Determination of Quantum Dot Size Distributions, Ligand Packing Densities, and Bioconjugation Using Analytical Ultracentrifugation

2008 Vol. 8, No. 9 2883-2890

Emma E. Lees,†,‡ Menachem J. Gunzburg,† Tich-Lam Nguyen,† Geoffrey J. Howlett,§ Julie Rothacker,‡ Edouard C. Nice,‡ Andrew H. A. Clayton,‡ and Paul Mulvaney*,† School of Chemistry and Bio21 Institute, UniVersity of Melbourne, ParkVille, Victoria 3010, Australia, The Department of Biochemistry and Molecular Biology and Bio21 Institute, UniVersity of Melbourne, ParkVille, Victoria 3010, Australia, and Ludwig Institute for Cancer Research, PO Box 2008, Royal Melbourne Hospital, ParkVille, Victoria 3050, Australia Received June 7, 2008

ABSTRACT Analytical ultracentrifugation (AUC) was used to characterize the size distribution and surface chemistry of quantum dots (QDs). AUC was found to be highly sensitive to nanocrystal size, resolving nanocrystal sizes that differ by a single lattice plane. Sedimentation velocity data were used to calculate the ligand packing density at the crystal surface for different sized nanocrystals. Dihydrolipoic acid poly(ethylene glycol) was found to bind between 66 and 60% of the surface cadmium atoms for CdSe nanocrystals in the 1.54-2.59 nm radius size regime. The surface ligand chemistry was found to affect QD sedimentation, with larger ligands decreasing the sedimentation rate through an increase in particle volume and increase in frictional coefficient. Finally, AUC was used to detect and analyze protein association to QDs. Addition of bovine serum albumin (BSA) to the QD sample resulted in a reduced sedimentation rate, which may be attributed to an associated frictional drag. We calculated that one to two BSA molecules bind per QD with an associated frictional ratio of 1.2.

1. Introduction. Semiconductor nanocrystals were originally proposed as substitutes for dyes in fluorescent biolabeling some 10 years ago.1,2 Since the original experiments, there has been a plethora of papers describing methods for conjugating nanocrystals to biological samples including peptides,3-6 proteins,7-9 antibodies,10-12 and oligonucleotides.13,14 The success of QDs as biological probes requires careful control over their physical properties: their size, shape, composition, and surface chemistry. A key parameter is particle size and size distribution since this determines the spectral position and purity of photoluminescence. Transmission electron microscopy (TEM) is routinely used to size QD samples, providing high-resolution detail of QD shape and structure. However TEM provides only limited, qualitative information on any surface-bound organic mate* Corresponding author. Telephone: (61)3-8344-2420. Fax: (61)3-93481595. E-mail: [email protected]. † School of Chemistry and Bio21 Institute, University of Melbourne. ‡ Ludwig Institute for Cancer Research, Royal Melbourne Hospital. § The Department of Biochemistry and Molecular Biology and Bio21 Institute, University of Melbourne. 10.1021/nl801629f CCC: $40.75 Published on Web 07/30/2008

 2008 American Chemical Society

rial. Surface chemistry dictates the solubility of QDs in various solvents, governs the method of biofunctionalization, and also has a direct impact on quantum yield15 and blinking properties.16 The surface chemistry also contributes to the final hydrodynamic diameter, which is a critical parameter for the use of QDs as potential diagnostic and therapeutic agents. Recent in vivo studies show that QD biodistribution and renal clearance are highly sensitive to hydrodynamic diameter.17 To date a number of different techniques have been employed to characterize QD surface chemistry, notably X-ray photoelectron spectroscopy,18 nuclear magnetic resonance spectroscopy,19 and Rutherford backscattering.20 Here, we use analytical ultracentrifugation (AUC) to characterize the nanocrystal size distribution, composition, and surface chemistry. AUC has previously been used to probe the properties of various nanoparticle systems: the hydrodynamic radii of FePt nanoparticles;21 size-dependent sedimentation of CdSe, Fe3O4, and gold nanocrystals;22 pH-dependent aggregation of TiO2 particles;23 and conjugation of the DNA

Table 1. Radii of Core CdSe Samples Obtained from TEM and AUC Analysisa sample

λ (nm)

rTEM (nm)

rAUC (nm)

reffAUC (nm)

S

547 1.49 ( 0.11 1.54 ( 0.13 3.07 ( 0.19 16.8 576 1.85 ( 0.14 1.90 ( 0.18 3.57 ( 0.24 25.9 608 2.48 ( 0.18 2.43 ( 0.17 4.26 ( 0.21 43.3 617 2.71 ( 0.20 2.59 ( 0.10 4.47 ( 0.13 49.6 a λ is the first exciton peak position, r TEM is the CdSe radius determined from Yu et al.’s calibration curve28 (which is based on TEM data), rAUC is the CdSe radius calculated from AUC analysis, reffAUC is the CdSe/ligand radius calculated from AUC analysis, and S is the average sedimentation coefficient. Standard deviations are shown for each data set. TEM data are shown with a 7.5% standard deviation, which is within the reported range of 5-10%. 1 2 3 4

binding protein LacI to gold nanocrystals.24 More recently, Colvin’s group demonstrated the dramatic effect that the surface coating has on the sedimentation rate of nanocrystals.22 By varying the polymer chain length of polystyrene stabilized gold nanocrystals, they found that longer polymer chain lengths resulted in slower sedimentation rates, which was attributed to a reduction in the overall particle density. Here, we systematically explore the effect of nanocrystal size, composition, and surface chemistry on sedimentation rate. By studying a range of different systems that isolates each of these variables, we have been able to model their individual contribution to the particle’s sedimentation rate. Importantly, this analysis provides quantitative estimates of the number of ligands bound to the crystal surface. This is particularly useful for characterizing QD-bioconjugates, namely, quantifying the number of biomolecules bound per particle. Other techniques used to verify the conjugation of biomolecules to QDs have difficulty in analyzing the individual and the collective components of the bioconjugate. For example, UV absorbance is routinely used to measure protein concentration; however, the large extinction coefficients of QD materials at wavelengths below 280 nm directly interferes with protein absorption. As mentioned previously, TEM techniques provide only limited, qualitative information on any surface-bound material so it cannot be used effectively to verify conjugation. Agarose gel electrophoresis has been successfully used to detect changes in surface chemistry elicited by surface molecules25,26 and biomaterials.27 However this analysis is often complicated by the strong inherent charge of the QDs. As AUC is sensitive to the mass, density, and shape of particles in solution, it is a powerful tool to characterize the physical properties of QDs and their complexes. In addition to characterizing the physical properties of QDs, we also show the suitability of AUC as a technique for characterizing QD-protein bioconjugates. 2. Results and Discussion. 2.1. Core CdSe QDs. 2.1.1. SensitiWity of AUC to Nanocrystal Size. To validate AUC as a technique for sizing nanocrystals, we compared the size of core CdSe QDs to those obtained from sedimentation velocity analysis. The core CdSe sizes were deduced from the first exciton peak position using the calibration curve of Yu et al.28 (Table 1). Sedimentation velocity profiles for four different sized core CdSe QD samples are shown in Figure 1A-D. Analysis of the sedimentation velocity profiles using a c(s) continuous size distribution model gave excellent fits 2884

Figure 1. Sedimentation velocity data for four different sized core CdSe QD samples (A-D). Open circles represent the data; solid lines represent the fit to data. Every twelfth (A, B) or sixth (C, D) scan is shown for clarity. Sedimentation coefficient distributions of the four core CdSe QD samples (E-H). Open circles represent the data, solid lines represent the fit of the data to eq 1 using a ligand footprint of 30 Å2, and dashed lines show sedimentation coefficient distributions obtained by solving eq 1 for particle sizes obtained from Yu et al.’s calibration curve,28 assuming a standard deviation of 7.5%. The first exciton peak positions for the four CdSe samples are (A, E) 547 nm, (B, F) 576 nm, (C, G) 608 nm, and (D, H) 617 nm.

to the data as exhibited by the solid lines in Figure 1A-D. The resulting sedimentation coefficient distributions are shown in Figure 1E-H. In order to convert the sedimentation coefficient distributions to radii, to compare the radii obtained from TEM to those obtained from AUC, the sedimentation coefficient distributions were fit to eq 129 which describes the theoretical sedimentation coefficient for a spherical particle Nano Lett., Vol. 8, No. 9, 2008

S)

MWt(1 - νFs) 6NAπreffηsf

(1)

where MWt is the total particle molecular weight (core + ligand), νj is the particle’s partial specific volume, Fs is the solvent density, NA is Avogadro’s number, reff is the effective radius of the particle, ηs is the solvent viscosity, and f is the friction factor (where f ) 1 for a sphere). Equation 1 includes terms that are dependent on the surface ligand chemistry, namely MWt, νj, and reff. Therefore, in order to assign values to MWt, νj, and reff and fit the data to eq 1, it is necessary to approximate the number of ligands bound to the crystal surface. The ligand packing density, LPD, is the fraction of the surface occupied by the ligands and is given by LPD )

4πrc2 LF × SA

(2)

where rc is the core CdSe radius, LF is the ligand footprint, that is the effective surface area of the ligand, and SA is the number of surface atoms. SA was calculated from the outermost CdSe monolayer,18,20,33 its derivation is included in the Appendix. LF was sequentially varied, with the best fit of eq 1 to each of the experimentally observed sedimentation coefficient distributions giving the most appropriate value of LF. Once the optimal value for LF was obtained, it was then applied to every sample capped with that particular ligand. For DHLA-PEG capped CdSe nanocrystals, LF was determined to be 30 Å2. Once the ligand footprint has been determined for a given system, MWt, νj, and reff may be calculated. These derivations are included in the Appendix. By fitting the data to eq 1 using a standard deviation, we were able to extract the size distribution and average nanocrystal radius for each CdSe core sample. As outlined in Table 1, these values are in strong agreement with the sizes determined from Yu et al.’s calibration curve, which is based on TEM data. Sizing nanocrystals from TEM micrographs has an error typically on the order of a single lattice plane, (0.175 nm in the case of CdSe. Here the AUC data yield particle sizes that are within 0.12 nm of those determined from TEM micrographs (Table 1), well within the error of sizing particles from TEM images. The nanocrystal radii determined from AUC data are, at most, only 5% different to the predicted sizes from Yu et al.’s calibration curve (Table 1). Figure 1H highlights that a small difference in size is magnified when expressed as a sedimentation coefficient. From Yu et al.’s sizing curve, a CdSe sample with an absorption maxima of 617 nm is predicted to have an average radius of 2.71 nm, which corresponds to an average sedimentation coefficient of 54 S (dashed line). Experimentally, the average sedimentation coefficient was found to be 49 S (open circles) which yields an average radius of 2.59 nm. This shows that only a 5% difference in nanocrystal radius (0.12 nm) culminates in a 10% difference in sedimentation coefficient, highlighting the sensitivity of AUC to nanocrystal size. The sensitivity of AUC to nanocrystal size is also evident when comparing CdSe samples with absorption maxima of 576 and 608 nm (Figure 1, panels F and G, respectively). These samples display distinguishable sedimentation coefNano Lett., Vol. 8, No. 9, 2008

ficient distributions centered around 26 and 43 S, respectively. Solving for eq 1 yields average nanocrystal radii of 1.90 and 2.43 nm, respectively. Significant differences in sedimentation coefficients were detected for QD samples that differ in radius by only 0.16 nm. This result demonstrates the acute sensitivity of AUC to changes in nanocrystal size and shows that AUC is capable of resolving nanocrystal sizes that differ by a single lattice plane. Included in the Supporting Information is a plot of eq 1 for CdSe nanocrystals up to 4.5 nm in radius (bulk CdSe) which shows the dependence of sedimentation coefficient on nanocrystal size (Figure S-1A). Nanocrystals ranging from 1 to 4 nm in radius exhibit sedimentation coefficients from 7 to 125 S. This relationship accounts for the sensitivity of AUC to nanocrystal size and the observation from the experimental data that small changes in nanocrystal size elicit large changes in sedimentation coefficient. Another advantage of AUC is that the sedimentation velocity data intrinsically reveals the size distribution of the sample. The width of the sedimentation coefficient distribution reflects the nanocrystal size distribution (Figure 1E-H). By fitting the sedimentation coefficient distribution to eq 1 the size distribution of the sample can be deduced. The sedimentation coefficient distributions shown for the nanocrystal sizes obtained from TEM were fit to eq 1 assuming a standard deviation of 7.5%, which is within the reported range of 5-10%.28 The experimentally determined size distributions obtained from the AUC data represent a 4-9% standard deviation about the average radius (Table 1) which is in good agreement with typical size distributions reported from TEM analysis. 2.1.2. Ligand Packing Density. As mentioned previously, the ligand packing density at the crystal surface is a critical parameter for analyzing the experimentally observed sedimentation coefficient distributions. By analyzing a range of different sized CdSe nanocrystal samples capped with the same ligand, DHLA-PEG, we calculated the ligand footprint of DHLA-PEG to be 30 Å2. Having determined LF, the LPD for the different sized nanocrystals may be calculated. For nanocrystals within the size regime studied here, that is between 1.54 and 2.59 nm radius, we calculate packing densities of 33 and 30% respectively (Figure S-1B). As DHLA-PEG binds only to Cd2+ sites on the nanocrystal surface, 66% and 60% of the surface cadmium atoms are bound to DHLA-PEG in the case of 1.54 and 2.59 nm nanocrystals, respectively. The moiety of DHLA-PEG is a terminal dithiol, which is capable of binding two Cd2+ sites. As we observe packing densities greater than 50% of the surface Cd2+ sites, we postulate that a proportion of the DHLA-PEG ligands must bind through only one of the thiol groups. For example, for the largest nanocrystal size of 2.59 nm we calculate that at least 9.5% of the DHLA-PEG ligands bind through a single thiol group, more may do so but this would result in vacant Cd2+ sites. 2.2. Core/Shell CdSe/ZnS QDs. Thus far we have examined core CdSe QDs to demonstrate the sensitivity of AUC analysis to nanocrystal size and to calculate the ligand packing density at the crystal surface. For biological ap2885

Figure 2. Normalized absorbance (solid line) and photoluminescence (dashed line) spectra, representative TEM images (scale bar ) 10 nm), and corresponding radius frequency histograms of five different sized core/shell CdSe/ZnS QD samples. 2886

Nano Lett., Vol. 8, No. 9, 2008

Table 2. Radii of Core/Shell CdSe/ZnS DHLA-PEG Samples Obtained from TEM and AUC Analysisa sample

λ (nm)

rc (nm)

rcsTEM (nm)

rcsAUC (nm)

S

562 1.42 2.62 ( 0.37 2.52 ( 0.25 35.3 579 1.42 3.26 ( 0.49 3.23 ( 0.18 56.6 595 1.85 3.08 ( 0.50 3.05 ( 0.16 53.3 598 1.53 3.60 ( 0.37 3.69 ( 0.33 74.6 608 2.41 3.46 ( 0.34 3.11 ( 0.13 60.9 a λ is the first exciton peak position of the CdSe/ZnS particles, r is the c core CdSe radius, rcsTEM is the CdSe/ZnS radius determined from TEM, rcsAUC is the CdSe/ZnS radius determined from AUC, and S is the average sedimentation coefficient. Standard deviations are shown for both data sets. 1 2 3 4 5

Figure 3. Sedimentation coefficient distributions of five different sized core/shell CdSe/ZnS QD samples capped with either PsAA or DHLA-PEG. The first exciton peak positions for the five CdSe/ ZnS samples are (A) 562 nm, (B) 579 nm, (C) 595 nm, (D) 598 nm, and (E) 608 nm. Open circles represent the data, solid lines represent the fits of the data to eq 1. PsAA capped particles are shown in blue (LF ) 50 Å2), DHLA-PEG capped particles are shown in red (LF ) 30 Å2).

plications however, it is advantageous to use core/shell QD samples for improved photostability. We next compared the size distributions of core/shell CdSe/ZnS DHLA-PEG QDs obtained from TEM data to those obtained from sedimentation velocity analysis. Representative TEM micrographs and corresponding radius/frequency histograms of five different sized core/shell CdSe/ZnS samples are shown in Figure 2. The average particle radii are included in Table 2. The core/ shell CdSe/ZnS sedimentation coefficient distributions shown in Figure 3 (red circles) were fit to eq 1 in the same manner as for the core CdSe data. The core CdSe radius was deduced from the first exciton peak position using the calibration Nano Lett., Vol. 8, No. 9, 2008

curve of Yu et al.28 The ligand footprint for DHLA-PEG was maintained at 30 Å2. The ZnS shell thickness was sequentially varied to obtain the best fit of eq 1 to the data. The core/shell CdSe/ZnS radii obtained from AUC analysis were in good agreement with those determined from TEM images, deviating by at most 0.35 nm (Table 2). This error is reasonable considering the compound errors associated with the core and ligand calculations. 2.2.1. Surface Ligand Chemistry. For each of the five different sized core/shell CdSe/ZnS samples, two different surface ligand coatings were employed. This enabled us to study the effect of surface ligand chemistry on QD sedimentation in systems directly relevant to biological applications. Sedimentation velocity experiments were performed on CdSe/ZnS QDs coated with either phosphonoacetic acid (PsAA), a short chain ligand with a low MW and high density (PsAA 5.3 Å, 140.03 g/mol, 1.508 g/cm3), or DHLAPEG, a longer molecule with a high MW and low density (DHLA-PEG 54.5-56.7 Å, 790 g/mol, 1.225 g/cm3). Samples capped with PsAA were found to sediment more quickly, i.e., have a larger sedimentation coefficient, than those capped with DHLA-PEG (Figure 3). To highlight the effect of surface ligand chemistry on QD sedimentation, eq 1 has been plotted as a function of nanocrystal size for the case of PsAA and DHLA-PEG (Figure S-2). The contribution of the ligands to MWt, νj, reff, and S are shown. Although DHLA-PEG increases MWt to a greater extent than PsAA, this is countered by its contribution to reff and νj, which results in larger, less dense particles. The overall effect is that DHLA-PEG capped particles have a reduced sedimentation rate compared to those capped with PsAA (Figure S-2D). This result is in agreement with Jamison et al.’s finding for polystyrene capped gold nanocrystals.22 Larger surface ligands have a greater capacity to reduce the particle density thereby decreasing the sedimentation rate. The sedimentation coefficient distributions in Figure 3 show some interesting features. In Figure 3B both the PsAA and DHLA-PEG QD distributions are bimodal. The average sedimentation coefficients are 81 and 45 S for PsAA capped particles. These values shift to 56 and 29 S, respectively, when capped with DHLA-PEG. The low abundant, smaller species may represent homogeneous ZnS nuclei formed during the shelling process which would account for their presence in both samples. Also apparent in the PsAA sedimentation coefficient distributions shown in panels B and E of Figure 3 are low abundant species with high sedimentation coefficients. These represent small QD aggregates. The appearance of these features in the sedimentation data allows for accurate characterization and quality control of QD samples. 2.2.2. QD Protein Association. Given the sensitivity of AUC to changes in QD size and surface chemistry, we next examined the effect of protein association on QD sedimentation. The sedimentation velocity profiles of 3.05 nm CdSe/ ZnS capped with DHLA-PEG in the presence and absence of BSA were compared (Figure 4). The average sedimentation coefficient decreased from 55 to 45 S upon addition of 2887

Figure 4. Sedimentation coefficient distributions of 3.05 nm radius CdSe/ZnS particles capped with DHLA-PEG in the presence (green circles) and absence (red circles) of BSA. Solid lines represent the fits of the data to eq 1. The inset shows the change in average sedimentation coefficient with increasing numbers of bound BSA for different frictional ratio values.

BSA. In order to quantify the QD:protein stoichiometry it is necessary to consider the effect that the frictional ratio has on the data analysis. Thus far we have analyzed the QD samples using a frictional ratio of 1.0, in keeping with their near spherical morphology. However, globular proteins typically have a frictional ratio of 1.2 which is likely to affect the frictional ratio of the QD:protein complex. The inset of Figure 4 shows the effect of increasing the number of associated BSA molecules on QD sedimentation coefficient for frictional ratios ranging from 1.0 to 1.4. For the case of a particle with no associated BSA and a frictional ratio of 1.0, the expected sedimentation coefficient is 53 S, which is what we observe for the DHLA-PEG capped particles in the absence of BSA (Figure 4). Addition of BSA, with no associated change in frictional ratio, is expected to increase the sedimentation coefficient (red solid circles). However, this was not observed experimentally, rather the QD sedimentation rate decreased to 45 S in the presence of BSA (Figure 4). This indicates that the association of BSA to the QD increases the frictional ratio, thereby reducing the sedimentation rate. As shown in the inset of Figure 4, a sedimentation coefficient of 45 S may be obtained in one of two ways: (i) assuming a frictional ratio of 1.2 with one to two BSA bound per QD (blue solid squares), or (ii) assuming a frictional ratio of 1.3 with five to seven BSA bound per QD (orange open squares). Given that the size of BSA is on the same order of magnitude as the QDs, five or more BSA per QD would be approaching steric limitations. Therefore, we postulate that one to two BSA molecules are associated per QD which results in a frictional ratio of 1.2. This result demonstrates the applicability of AUC in detecting and quantifying conjugation of biomolecules to QDs. 3. Conclusion. We have shown that AUC analysis of core CdSe and core/shell CdSe/ZnS QDs yields size distributions that are in excellent agreement with those obtained from 2888

TEM analysis. Furthermore, AUC analysis achieves monolayer resolution of nanocrystal sizes due to the high sensitivity of the sedimentation coefficient to particle size. From AUC analysis we were able to extract the size of both the nanocrystal and the nanocrystal/ligand complex. The choice of surface ligand coating was found to affect the rate of QD sedimentation. This was attributed to the contribution of the surface ligands to QD size and density, with larger ligands reducing QD sedimentation to the greatest extent. Finally, we have demonstrated that QD protein interactions can be monitored by AUC analysis as addition of protein induces a frictional drag which slows the QD sedimentation rate. 4. Experimental Section. 4.1. Materials. Phosphonoacetic acid(PsAA,98%),tetramethylammoniumhydroxide(TMAOH, 25 wt % in methanol), and bovine serum albumin (BSA, 98%) were purchased from Sigma-Aldrich. Analytical grade (AR) chloroform (CHCl3) was obtained from Merck. AR grade hexane was purchased from Univar, and AR grade methanol, ethanol, and acetone were purchased from Chem Supply. All chemicals and solvents were used as received. Milli-Q grade (R > 18 MΩ cm) water was used throughout. 4.2. Sample Preparation. CdSe core30,31 and CdSe/ZnS core/shell32,33 QDs were synthesized according to previously reported methods. The QDs were transferred into aqueous media via ligand exchange employing either (i) PsAA or (ii) DHLA-PEG (PEG, n ) 12), the synthesis of which is described by Uyeda et al.25 For PsAA capped QDs ∼ 7.5 nmol QDs in CHCl3 were precipitated with methanol/acetone and redispersed in a 0.1 M PsAA/methanol solution where the pH had been adjusted to 8 using TMAOH. The QD/ PsAA solution was heated to 60 °C overnight, precipitated with hexane/ethanol and redispersed in water. To obtain DHLA-PEG capped QDs, the QDs were first transferred into water using PsAA as outlined above, after which a 10000 molar excess of DHLA-PEG was added to the QD/PsAA solution and heated to 60 °C overnight. The QD/DHLAPEG solution was buffer exchanged into phosphate buffered saline (PBS) using an ultracentrifugal filtration device (Vivaspin, 50 kDa MWCO) and purified using three cycles of concentration/dilution. The QD/BSA sample was prepared by adding a 10 times molar excess of BSA to the QD/DHLAPEG solution in PBS followed by overnight incubation at room temperature. For each of the QD samples in aqueous media the optical density at the first absorption maximum was recorded (Nanodrop spectrophotometer ND-1000) and the volume was adjusted (assuming the same extinction coefficient of the sample in CHCl3) such that the final QD concentration was 1.5 µM. 4.3. Measurements. Absorption spectra were recorded on a Cary 5 UV-vis-NIR spectrophotometer operated in dual beam reference mode. Steady-state photoluminescence spectra were recorded on a Fluorolog-3 spectrofluorimeter (Horiba Jobin Yvon) with slit widths of 1 nm and an integration time of 1 s. Concentrations were adjusted such that the QD samples had optical densities of