Separation of Stabilized MOPS Gold Nanostars by Density Gradient Centrifugation.

This article describes the stabilization and postsynthetic separation of gold nanostars (AuNS) synthesized with a morpholine-based Good's buffer, 3-(N-morpholino)propanesulfonic acid. Resuspension of AuNS in ultrapure water improved the shape stability of the particles over 30 days. We demonstrated the sorting of nanostars via rate-zonal centrifugation through a linear sucrose gradient based on branch length and number. We determined that one round of centrifugation was sufficient for separation. Also, we improved the structural homogeneity and stability of the nanoparticles through the optimization of the storage conditions and established a robust method to sort AuNS based on size and shape.

Introduction

The physical and chemical properties of anisotropic gold nanoparticles are controlled by their dimension and shape.[1] Gold nanostars (AuNS) are high-performing substrate materials for surface-enhanced Raman spectroscopy,[2−6] photothermal therapy,[7] and photoacoustic imaging[8] because they exhibit strong electric field enhancements on multiple sharp tips.[9] AuNS are of interest among anisotropic nanoparticles because minor shape modifications, such as aspect ratio and branch length,[10,11] enable tunability of the localized surface plasmon (LSP) resonances.[2,8,12,13] Accessing specific structural features of AuNS has driven synthetic techniques that can control the chemical environment by introducing different reactants.[14−16] Typically, AuNS are prepared by seed-mediated syntheses;[17,18] however, these methods require strongly bound, cytotoxic surfactants that limit the biological applications. A commonly used seedless synthesis method for AuNS that avoids additional surfactants only involves two precursors: a Au salt (HAuCl4) and a biocompatible Good’s buffer, usually 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES).[10,13,19−21] Good’s buffer acts as both a reducing and shape-directing agent, with the ethane sulfonate group of HEPES binding to the Au surface and the hydroxyl group facilitating the assembly and bilayer formation to stabilize the AuNS.[21] Recently, we reported the preparation of AuNS with disparate branch lengths, branch tips with a small radius of curvature (3–5 nm), and multiple LSP resonances in the near-infrared (NIR) and second NIR window using a Good’s buffer with a morpholine ring, 3-(N-morpholino)propanesulfonic acid (MOPS).[11]

To exploit the unique structural properties of MOPS AuNS, challenges specific to the buffer must be overcome: (1) the morpholine moiety degrades once exposed to HAuCl4, which causes colloidal instability;[22] and (2) MOPS cannot form a stable bilayer without a hydroxyl group, which leads to heterogeneous solutions.[21] Adjusting the synthetic conditions is a common strategy to obtain stable and homogeneous Au nanoparticle solutions, but exquisite control over room-temperature reaction conditions is limited;[23] hence, postsynthetic separation offers an approach to obtain purer solutions of anisotropic particles. Centrifugation,[24] size-exclusion chromatography,[25] filtration or diafiltration,[26,27] and electrophoresis[28] have been used to produce metal nanoparticle solutions with narrow shape and size distributions. In particular, density gradient centrifugation (DGC) is a postsynthetic sorting technique that can refine the structural distributions of metal nanoparticles by rate-zonal separation.[29] In this method, the sorting of particles primarily depends on the sedimentation rate, which is a function of their size and shape. Standard gradient media are deleterious to anisotropic metallic nanoparticles with weakly bound surface ligands;[30] however, sucrose shows excellent compatibility.[31,32]

Here, we describe a strategy to obtain highly enriched populations of AuNS by combining the stabilization and sorting processes. We improved the homogeneity of MOPS AuNS and reduced the percentage of unwanted spherical byproducts by adjusting the storage conditions. By centrifuging the as-synthesized, unsorted mixtures through a sucrose density gradient, we isolated the solutions of AuNS based on branch length and number. We demonstrated that only a single round of DGC was needed to sort AuNS solutions by size and shape; an additional round did not significantly improve separation of the sorted samples.

Results and Discussion

Establishing MOPS AuNS Stability

Scheme summarizes the stabilization and separation of heterogeneous MOPS AuNS populations. AuNS solutions stored in ultrapure, deionized water maintained the shape integrity compared with those stored in the MOPS buffer. To overcome particle heterogeneity, we sorted the AuNS first by layering a concentrated solution on top of a sucrose density gradient and then performing DGC. In the subsequent rounds of DGC, the original (O) fractions were concentrated and centrifuged through a shallow linear gradient.

Scheme 1

Stabilization and Separation of MOPS AuNS

AuNS were synthesized using a modified version of our previously established procedure[11] with 150 mM of the Good’s buffer MOPS and 0.2 nM HAuCl4 (Methods). We selected these conditions because this MOPS concentration produced AuNS with two distinct LSP peaks (Figure ). The as-synthesized solutions were characterized by UV–vis spectroscopy to measure the bulk optical properties, and transmission electron microscopy (TEM) was used to visualize individual AuNS. The 0 min time point indicates an unstirred solution after the addition of HAuCl4. After 1 min of vortexing, one LSP peak at shorter wavelengths, λ1, formed between 700 and 750 nm (Figure a). Throughout the reaction, a LSP peak at longer wavelengths, λ2, formed at 800 nm and red shifted to ca. 1100 nm. After 24 h, both LSP peaks blue shifted and decreased in absorbance as the 520 nm LSP peak intensity increased (Figures b and S1), possibly due to a decrease in the particle size and the average branch length, as well as an increase in the number of spherical particles (Table S1). We hypothesize that the MOPS buffer was displaced from the surface of the particles, which resulted in a change in the shape.

Figure 1

Peak absorbance of MOPS AuNS in growth solution blue shifted over 30 days. (a) The absorbance was measured 24 h (red line) after the addition of HAuCl4 up to 30 days (black line) in 5 day increments. (b) TEM images of MOPS AuNS after 1, 15, or 30 days after the addition of HAuCl4.

Exchanging Solvent Storage Conditions

In an earlier work, we observed that the AuNS synthesized with MOPS transformed into spheres over 12 days when stored in the buffer growth solution.[11] Here, we resuspended the as-synthesized solution in ultrapure water either one (1× solution) or two (2× solution) times (Figure ). We tracked the changes in bulk optical properties at the two LSP peaks to determine whether AuNS’s shape changed; within the first 5 days, λ1 and λ2 blue shifted by 15 nm for 1× and 2× solutions. After 10 days, the LSP resonances of the 2× solution did not change, and the branch lengths decreased only minimally (by 5 ± 3 nm). Larger AuNS (>80 nm) remained in the solution over 30 days compared with the as-synthesized particles. The LSP peaks of the 1× solution, however, shifted to shorter wavelengths by an additional 10 and 20 nm, and the particle size decreased. The percentage of spherical particles increased for the 1× solution from 35 to 47, whereas the percentage remained the same for the 2× solution. We hypothesize that exchanging buffered growth solutions for ultrapure water stabilized AuNS, in contrast to similar systems that are stabilized by the excess buffer solution because excess MOPS degrades over time in the presence of Au salt.[33]

Figure 2

Shape stability of AuNS increased due to water resuspensions. (a) Plot of the LSP resonance shift over time for solutions at one (black) and two (red) resuspensions. Filled markers correspond to the λ1 peak and open markers correspond to the λ2 peak. (b) TEM images of AuNS 30 days after one or two resuspensions in water.

Comparing Linear and Step Gradients

Even with improved storage conditions, MOPS AuNS solutions still contained ca. 37% undesirable spherical particles. Because the as-synthesized AuNS showed a broad distribution of shapes (0–8 branches) and sizes (20–80 nm),[11] we used DGC to sort branched nanoparticles.[30] We compared two different types of gradient densities: linear and step (Figure ). A linear gradient contains a single continuous zone, whereas a step gradient has multiple density zones. Both gradients can enrich the solutions by sorting the nanoparticles into distinct bands.[29] Changing the gradient composition and centrifugation time can improve both dispersion through the gradient and the nanoparticle yield. Because sorting of particles with similar sizes but different shapes requires a shallow gradient (10%),[30] we started with an 50–60% (w/v) gradient of sucrose.

Figure 3

Linear gradient allows for a greater separation than step gradient. The control (no AuNS) was used to establish the linear and step gradients and contained the coloring agent Trypan blue. Once the gradients were established, AuNS were centrifuged through a gradient without Trypan blue (sorted AuNS). A linear gradient with an increasing viscosity and density produced two bands: a red one and a blue one. The particles had a greater spatial separation over a step gradient. The step gradient produced two bands, but the particles were not further separated within the bands.

To monitor the gradients, we added a standard coloring agent, Trypan blue, to the denser 60% (w/v) solution. For the linear gradient, the dense, blue solution was layered below the clear 50% (w/v) solution. Using a custom mixing program (Methods), the two sucrose solutions were mixed at an angle without AuNS to create a continuous linear gradient (Figure , control (no AuNS)). We determined the linearity of the sucrose gradients by fractionating the control gradient and calculated the concentration of sucrose in each fraction using the following equation[34]where c is the concentration of sucrose at point i, x is the concentration of the heavier solution (60%), y is the concentration of the lighter solution (50%), and Afn is the absorbance of fraction n, and A60% is the absorbance of 60% sucrose dye. The concentration of sucrose over the gradient was linear and contained an r2 value of 0.9998. For the step gradient, we layered five different density sucrose solutions: 50, 52.5, 55, 57.5, and 60% (w/v).

We created linear and step gradients without Trypan blue and then layered 500 μL of AuNS solutions (8–10 nM) and centrifuged samples at 4400g (Figure , sorted AuNS). DGC is limited intrinsically by the volume of the sample that can be layered on top of the gradient;[35] however, we found that the starting concentration of the layered sample had no effect on the degree of separation (Figure S2). In the linear gradient, two distinct bands were formed: (1) a thin red band of spherical nanoparticles and (2) a long blue/gray band of sorted AuNS. TEM analysis verified that the red band contained 80–85% spherical nanoparticles and the blue/gray band enriched the populations of AuNS. For the step gradient, centrifugation of AuNS produced bands in the top three zones and resulted in the reduced spreading of the particle distributions compared with the linear gradient. Moreover, each step gradient band contained nanoparticles with mixed particle shapes and size. Hence, the linear density gradients can separate similarly sized nanoparticles with small structural differences (i.e., AuNS with varying branch length and number), and step gradients are more useful for sorting a few, distinct populations of nanoparticles.

Characterizing Separation of AuNS in a Linear Gradient

After centrifuging AuNS through the linear gradient, the samples were fractionated at intervals of 4 mm from the meniscus 30 times (Methods). The corresponding fractions were collected in different tubes and labeled F1–F30. F1 indicates the top-most fraction and F30 is the bottom-most fraction. We found that the best separation of particles without forming a pellet at the bottom occurred for the centrifugation times of 1.5 h at 4400g (Figure a). To characterize sorted AuNS fractions (F1–F30), we measured the absorbance spectra and correlated the optical properties with TEM images (Figure b–c). Fractions F1–F4 produced a peak only at 520 nm, which corresponded with spherical nanoparticles. Fractions F5–F19 contained AuNS with a LSP at λ1 between 700 and 720 nm with a high-intensity absorbance (>0.3), which was the minimum absorbance needed to image the particles and the bulk optical properties. The second LSP at λ2 shifted from 800 to 1150 nm in fractions F5–F19; this shift indicated an increase in the branch length. In fractions >F19, the absorbance decreased by half, which indicated a low concentration of particles. Although particles with the same mass or shape may have comparable sedimentation rates, sorted AuNS showed improved particle homogeneity in each fraction.[30]

Figure 4

Separation of AuNS through a linear density gradient by size and shape. (a) Photographs of F5–F23 of the AuNS (top) layered on top of the gradient and (bottom) centrifuged through the linear sucrose gradient. (b) Absorbance spectra for every odd-numbered fraction from F5 to F23. (c) Representative TEM images from F5, F13, and F17 show branched particles in all of the different fractions.

Sorting AuNS by Branch Length and Number

To quantify particle size, branch length, and branch number of AuNS, more than 500 branches of every other fraction from F5 to F19 were analyzed manually by ImageJ (Figures a and S3). (We omitted F1–F4 because they contained mostly spherical particles.) For fractions F5–F19, the average tip-to-tip particle size increased from 21 ± 5 to 100 ± 17 nm, which also corresponded with an increase in the branch length from 18 ± 10 to 32 ± 18 nm (Figure b). We observed two populations of AuNS: shorter (<30 nm) and longer (>30 nm) branched particles. In all of the fractions, most branches were less than 30 nm. With an increase in the fraction number, the number of short branches decreased as the number of long branches increased. In fractions F5–F11, the population consisted of particles that had lower masses and hence remained closer to the top of the density gradient. In higher fractions F13–F19, AuNS had longer branches with larger particle sizes and greater mass per particle.

Figure 5

Branch length increase with fraction number. (a) TEM images for odd-numbered fractions in F5–F19. (b) The number of branches that are short (<30 nm) or long (>30 nm) for each fraction.

Rate-zonal separation efficiency depends on the particle shape or the number of branches (0–8) per particle (Figure ). We categorized the fractions into branch number populations: spheres (0), low (1–2), medium (3–4), and high (>5). We observed a decrease in the percentage of spheres from 20 to <1% from F5 to F13. This low percentage of spheres (<1% of the population) remained in fractions F13–F19, which created highly enriched populations of branched AuNS. Each fraction showed greater particle homogeneity than the as-synthesized solutions. From fractions F5 to F19, high branch number populations increased, whereas low branch numbers decreased. Particles with greater number of branches, longer branches, and larger sizes tended to sediment to the bottom of the gradient. These characteristic differences between spherical nanoparticles and heterogeneous AuNS allowed for sorting based on nanoscale structural features.

Figure 6

Analyzing the Au nanoparticle and AuNS branch number population for each fraction. (a) TEM images of high, medium, and low number of branches. The dimensions of all of the images are 150 nm × 200 nm. (b) The percentage of the AuNS population that have low (1–2), medium (3–4), or high (>5) number of branches.

Separating AuNS through a Second Round of Centrifugation and Fractionation

To achieve more refined populations of AuNS, we processed the original (O) fractions by DGC for a second time (Figure a). During round 1 of DGC, the AuNS dispersed throughout the top half of the tube; therefore, for round 2, we created a second linear 50–55% (w/v) sucrose gradient. AuNS in the top (T), middle (M), and bottom (B) regions of the gradient were analyzed for the number of branches per particle, but no differences were found between the three fractions compared with the original population (Figure b). We found that a single round of rate-zonal DGC efficiently enriched AuNS populations based on the differences in size and shape, and additional rounds did not improve the particle homogeneity.

Figure 7

Round 2 of DGC fractions did not enrich branch populations. (a) Odd-numbered original fractions were layered on top of a 50–55 wt % sucrose gradient and centrifuged for 45 min at 4400g. Top (T), middle (M), and bottom (B) fractions of the each of the round 2 fractionated samples were collected. (b) The branch number distribution for each of the T, M, and B samples was analyzed manually. The population distribution was similar among the original fractions (O) and the refractionated samples.

Conclusions

In summary, we established techniques to stabilize and separate the heterogeneous solutions of AuNS synthesized with a morpholine-based Good’s buffer. The modified storage conditions can be generalized to other metal nanoparticles that change shape in solution over time. We found that linear gradients are better suited to sort similarly sized particles with small differences, such as branch length and number. The robust techniques described herein can produce highly enriched populations of anisoptropic nanoparticles that can be exploited for different shape-dependent applications, such as surface-enhanced spectroscopies, sensing of chemical and biological analytes, bioimaging, and therapeutics.

Methods

Gold Nanostar Synthesis

The Good’s buffer used was 3-(N-morpholino)propanesulfonic acid (MOPS buffer, Sigma Aldrich). The 1 M stock MOPS solution was made by dissolving the buffer salt in Millipore water (18.2 MΩ cm) using a medium-sized stir bar to ensure thorough mixing. The pH of the MOPS solution was measured using a Thermo Scientific pH meter and was adjusted using the concentrated solutions of NaOH. For fine pH adjustments, HCl was added dropwise.

AuNS were synthesized by adding 0.2 mM (final concentration) gold(III) chloride trihydrate (HAuCl4; Sigma Aldrich) to 150 mM of MOPS buffer. Each solution was vortexed in a 50 mL Falcon tube for 1 min before the addition of HAuCl4 and for 1–5 min afterward. After vortexing, the growth solution was left undisturbed at room temperature for 24 h. The 0 min time point indicates an unstirred growth solution after the addition of HAuCl4.

Particle Characterization Techniques

AuNS solution (3 mL) was placed in a 1 cm plastic Brookhaven cuvette, and the absorbance spectra were measured from 400 to 1400 nm using a Cary 5000 UV–vis–NIR spectrophotometer (Agilent Technologies).

For TEM grid preparation, carbon Type B, 300 mesh copper grids (Ted Pella) were treated with 0.1% (w/v) poly-l-lysine (Sigma Aldrich) for 5 min. Then, 40 μL of 10× concentrated Au nanoparticle solution was left to rest on the treated grids for 30–60 s and then wicked away with filter paper. A JEOL 1230 TEM was used for imaging the particles. The representative images were collected from different areas of the grid. Structural features, such as circularity and Feret diameter, were characterized using the Analyze Particles plugin on ImageJ for at least 500 particles per sample. A circularity threshold of 0.9 was used to define spherical particles, and the Feret diameter corresponds to the largest tip-to-tip distance on a particle. The branch length was measured manually from the tip to the base of the branch.

Density Gradient Centrifugation

Linear sucrose density gradients were formed using a gradient maker (BioComp Instruments) with 9 mL starting solutions of 50 and 60% w/v sucrose in water. A custom mixing program alternated five times between the following two steps: (1) time: 5 s, angle: 76, speed: 30 rpm; (2) time: 15 s, angle: 76, speed: 0 rpm. Step gradients were layered by hand using solutions of 50, 52.5, 55, 57.5, and 60% w/v sucrose in water. For the first round of DGC, 500 μL of a concentrated solution of bare AuNS (8–10 nM) was layered on top of the density gradient in an Ultra-Clear SW28 centrifuge tube (Beckman Coulter) and then centrifuged at 4400g for 90 min using a Thermo Fisher Scientific Sorvall Legend XT 120 v Benchtop centrifuge. The samples were fractionated at intervals of 4 mm from the meniscus (BioComp Instruments). Each fraction was dialyzed in Thermo Fisher 20K Slide-A-Lyzer Dialysis Cassettes for 24 h to remove sucrose from the solution.

For round 2 of DGC, we repeated round 1 DGC 20 times and the odd-numbered fractions were combined to form 400 μL of 8–10 nM solution. We collected each fraction over the 20 individual rounds of DGC and analyzed the final distributions of branch number and length of the combined fractions. Each of the combined fractions was layered on top of a linear 50–55% (w/v) sucrose gradient and the centrifugation time was set at 45 min at 4400g. We note that the reproducibility and scalability of this technique are demonstrated by the uniformity of the fractions that were combined from 20 rounds of DGC.