pubs.acs.org/NanoLett
Liposome-Based Chemical Barcodes for Single Molecule DNA Detection Using Imaging Mass Spectrometry Anders Gunnarsson,† Peter Sjo¨vall,†,‡,* and Fredrik Ho¨o¨k†,* †
Department of Applied Physics, Division of Biological Physics, Chalmers University of Technology, Fysikgra¨nd 3, SE-412 96 Go¨teborg, Sweden, and ‡ Department of Chemistry and Materials Technology, SP Technical Research Institute of Sweden, P.O. Box 857, SE-501 15 Borås, Sweden ABSTRACT We report on a mass-spectrometry (time-of-flight secondary ion mass spectrometry, TOF-SIMS) based method for multiplexed DNA detection utilizing a random array, where the lipid composition of small unilamellar liposomes act as chemical barcodes to identify unique DNA target sequences down to the single molecule level. In a sandwich format, suspended target-DNA to be detected mediates the binding of capture-DNA modified liposomes to surface-immobilized probe-DNA. With the lipid composition of each liposome encoding a unique target-DNA sequence, TOF-SIMS analysis was used to determine the chemical fingerprint of the bound liposomes. Using high-resolution TOF-SIMS imaging, providing sub-200 nm spatial resolution, single DNA targets could be detected and identified via the chemical fingerprint of individual liposomes. The results also demonstrate the capability of TOF-SIMS to provide multiplexed detection of DNA targets on substrate areas in the micrometer range. Together with a high multiplexing capacity, this makes the concept an interesting alternative to existing barcode concepts based on fluorescence, Raman, or graphical codes for small-scale bioanalysis. KEYWORDS TOF-SIMS, liposome, DNA, barcode, single molecule detection
T
he capability to identify and quantify genetic and proteomic biomarkers has improved rapidly during the last decades.1 In particular, novel concepts for DNA detection have provided new possibilities to perform gene-expression profiling or single polynucleotide polymorphism analysis with high multiplexing capability.2 Currently, the main approach relies on DNA microarray technology, which enables simultaneous detection of thousands of different targets on a single chip.3 However, multiplexing relies in this case on photolithography or robotic spotting, which makes the production labor intensive and often difficult to accomplish with high reproducibility. To circumvent some of the drawbacks associated with such ordered arrays, various detection schemes based on target-specific barcodes have increased in popularity.4 These are either surfacebased, so-called random arrays,5 or solution-based,6 in which case the probe molecules are either distributed randomly on a planar surface or immobilized on suspended microspheres, respectively. In both cases, multiplexing is accomplished through encoding of target-specific barcodes associated with each type of probe molecule, making it possible to decipher the identity of the bound target. The potential of multiplexed encoding using fluorescence has been extensively investigated using, for example, micro-
spheres containing precisely controlled ratios and concentrations of different organic dyes7,8 or quantum dots.9,10 Furthermore, Mirkin’s group has demonstrated that Ramanactive dyes in close proximity to gold nanoparticles may provide more narrow-band spectroscopic decoding than that of conventional luminescence-based methods, thus potentially improving the multiplexing capacity.11 The same group also demonstrated both DNA12 and protein13 detection using DNA sequences as unique barcodes, providing an essentially limitless multiplexing capacity. However, this method requires multiple preparation steps, including target separation using magnetic beads, barcode identification using conventional DNA array technology and, for operation in high sensitive mode, signal amplification via silver deposition utilizing gold nanoparticles as catalyzing seeds.14 Mass spectrometry (MS) techniques could potentially offer attractive readout alternatives due to their inherent multiplexed detection capabilities. Direct detection of the target molecules by MS would naturally eliminate the need for a separate barcode component with the ultimate multiplexing capability limited by mass interferences between the target molecules (and molecular fragments) in the recorded mass spectra. However, due to the relatively poor detection sensitivity of MS, direct multiplexed detection of DNA targets require a high density of surface-immobilized probes.15,16 In fact, despite high surface densities and the use of matrix enhancement, spot sizes in the millimeter range are needed for limit of detections (LODs) in the high nanomolar regime.16
* To whom correspondence should be addressed. E-mail: (F.H.) fredrik.hook@ chalmers.se. Phone: (+)46 (0)31-772 6130. Fax: (+) 46 (0)31-772 3134. (P.S.) E-mail: peter.sjo¨
[email protected]. Phone: (+)46 10 516 5299. Fax: +46 33 10 3388. Received for review: 12/20/2009 Published on Web: 01/19/2010 © 2010 American Chemical Society
732
DOI: 10.1021/nl904208y | Nano Lett. 2010, 10, 732-737
To enhance the sensitivity of MS detection, a barcode approach may be applied in which direct target detection is replaced by detection of a target-specific barcode compound existing in multiple copies per target molecule, as recently demonstrated by Qiu et al.17 Here, the binding of target DNA to DNA-modified gold nanoparticles was detected at a LOD in the sub-nanomolar regime while simultaneously identified by laser desorption/ionization time-of-flight (LDI-TOF) MS analysis of nanoparticle-attached poly(ethylene)glycol chains of different molecular weights, acting as barcodes. However, the identification of different PEG-chains was made from a homogeneous mixture of encoded gold particles and not on an individual particle basis, in contrast to the case in conventional suspension- or random arrays. This work nicely illustrates the potential strength of MS-based decoding of random arrays. However, considering the limited spatial resolution of LDI-TOF or matrix assisted LDI-TOF, which at best can reach 25-50 µm,18 there is significant room for further improvement with respect to LOD, sample consumption, as well as barcode capacity. As an imaging mass spectrometry technique with high spatial resolution, molecular specificity, and detection sensitivity, time-of-flight secondary ion mass spectrometry (TOF-SIMS)19 is emerging as a promising tool for chemical microanalysis of biological cells, tissues, and model systems. By detection of molecular and characteristic fragment ions, the spatial distribution of specific biomolecules can be mapped in biological samples at submicrometer resolution without preselection or labeling of the probed molecule and without the need for matrix deposition. For example, the capability to detect and identify individual 300 nm liposomes with different lipid composition on a silicon dioxide substrate was recently demonstrated using TOF-SIMS (Gunnarsson et al. submitted). In one of the few attempts to demonstrate the potential of TOF-SIMS as a bioanalytical sensor, Brandt et al. used peptide nucleic acid (PNA) on a microarray chip (spot size ∼100 µm) to capture unlabeled DNA targets. The captured DNA strands could then be subsequently identified through their phosphate content, which is completely missing in PNA. However, the multiplexing component of this method relies, in analogy with conventional DNA arrays, on spatial arrangement of probePNA modified spots.20 In this work, we investigate the potential of TOF-SIMS for multiplexed DNA detection on a random DNA array. By letting DNA-modified liposomes with different lipid composition act as chemical barcodes, the identification of different target-DNA sequences was accomplished. The three-component sandwich assay used for this purpose (see schematic illustration in Figure 1) is composed of (i) surface-bound probe-DNA strands, (ii) unlabeled target-DNA strands, and (iii) capture-DNA strands conjugated to barcoding liposomes. The feasibility of using TOF-SIMS to detect and identify target-DNA-mediated liposome binding was verified using two unique target-DNA sequences. Different analysis modes, © 2010 American Chemical Society
FIGURE 1. Schematic illustration of the sensing template based on the self-assembly of biotinylated copolymer (PLL-g-PEG/PLL-g-PEGbiotin), Neutravidin, and biotinylated probe-DNA on SiO2. The surrounding gold surface is functionalized with thiol-PEG to avoid unspecific binding. After exposing the sensor surface to the sample solution, target-DNA detection is accomplished by introducing a mixture of liposomes consisting of POPC or D13-DPPC, exposing 15 free-hanging single-stranded DNA bases (A′ or B′). The 30 bases single-stranded target (A or B) hybridizes in a sandwich format with both the surface-immobilized probe-DNA (15 bases) and the captureDNA conjugated to the liposome (A′ or B′) thereby immobilizing liposomes with a specific composition to the surface. After subsequent rinsing of unbound liposomes, the sample is freeze-dried and the lipid composition of the bound liposomes is analyzed by TOF-SIMS.
optimizing either mass or image resolution, were evaluated with respect to LOD and multiplexing capacity. In order to facilitate the preparation for vacuum analysis,21,22 the sensor substrate (SiO2) was surrounded by a gold grid (producing a SiO2 stripe of dimension 0.1 × 2 mm2). By selectively modifying the gold with thiol-PEG, a hydrophilic surface was produced which enabled the adjacent sensor surface to be covered by a thin water film during the rinsing and plunge freezing steps preceding freeze-drying. In addition, the chemical contrast generated by the thiolPEG-modified Au surface serves as a control during TOFSIMS analysis to confirm that the surface chemistry was preserved after freeze-drying. The sensor substrate was functionalized as previously described23 with a few modifications (see Supporting Information for details). In brief, the gold grid was first coated with a self-assembled monolayer of thiol-PEG, making it inert toward unspecific binding of DNA and liposomes.24 Since the thiol-PEG binds only to Au, the SiO2 surface could subsequently be coated with biotin-modified PLL-g-PEG (PLLg-PEG/PLL-g-PEGbiotin, molar ratio 100:1). The PLL-g-PEG coating has previously been shown to effectively minimize unspecific binding of DNA and lipid vesicles23 while the small fraction of PLL-g-PEGbiotin provides specific attachment of biotinylated probe-DNA strands using Neutravidin as a linker 733
DOI: 10.1021/nl904208y | Nano Lett. 2010, 10, 732-737
identical experiment with the difference that the suspension only contained target B, the corresponding spectrum yields a strong signal for D13-DPPC (m/z 179.16, 197.19). Similarly, peaks from both types of lipids appear from a sample that contained both target-DNA A and B. These spectra demonstrate the feasibility of the assay, that is, the differently labeled liposomes bind to the surface only if the specific corresponding target-DNA is present. The ion images in Figure 2B shows the signal intensity distributions of the two lipid-specific fragments across the 100 µm stripe, obtained from the analysis of the sensor substrate incubated only with target-DNA A. The strong signal from m/z 184.11 indicates a large concentration of POPC liposomes within the sensor area, while the weak signal at m/z 197.19 demonstrate a much lower surface concentration of D13DPPC lipids. It is clear from Figure 2B that there is a small, nonzero background signal from the D13-DPPC fragment (m/z 197.19), although the corresponding target-DNA B was not present in the suspension. This background signal can only be explained by the presence of D13-DPPC on the sensor surface; because of the high mass resolution (half width of the m/z 197 peak is ∼0.075 Da) and the unique masses of the fragment ions used to detect D13-DPPC, there are no mass interferences with other ions produced in the sample. Whether the presence of D13-DPPC on the surface originates from unspecific binding of liposomes or preparational artifacts, such as lipid deposition from solution during freeze-drying or weak exchange of the capture-DNA between different liposomes, cannot be deduced from these results alone. The slightly higher background signal (m/z 197.19) at the SiO2 area compared to the Au (Figure 2B) may suggest a weak exchange of capture-DNA (specific binding to the SiO2 area) but could also be attributed to a lower ionization yield of the lipid fragment ion on the Au surface. In fact, high image resolution measurements suggest that lipid deposition is the main cause of background signal (see further below). To determine the limit of detection of the current protocol, the analysis was performed at different target concentrations. Figure 3A shows the normalized signal intensity from a substrate containing probe-DNA sequences A′ and B′ exposed to capture-DNA modified POPC and D13-DPPC liposomes, at reducing concentrations of target-DNA A (in the absence of target-DNA B). The results demonstrate a limit of detection below 100 pM, at which concentration targetDNA was still detected with a signal-to-background ratio of around 4:1. The results in Figure 3A further enable us to determine the affinity constant (Kd) of the DNA-target hybridization. Assuming that the magnitude of the lipid signal is proportional to the liposome coverage, the lipid signal is proportional to the target-DNA molecule coverage as long as the coverage of surface immobilized probes (controlled by the ratio of nonbiotinlyated to biotinylated polymer) is kept equal or below the jamming limit of the liposomes (69 µm-2). Indeed, from Figure 3A, the Kd of the DNA interaction can be estimated to approximately 0.5 nM, which is in very
FIGURE 2. (A) Positive ion spectra for three different samples exposed to different target-DNA (A, B, or both) and two reference spectra for the different lipids. Important mass peaks are color coded to simplify interpretation, where red and green peaks are characteristic for POPC and D13-DPPC, respectively. The reference spectra are taken from pure D13-DPPC liposomes and a POPC bilayer, dried on a SiO2 surface. (B) Positive ion images from sample incubated with 100 nM DNA A for the POPC-specific fragment, m/z 184.11 and the D13-DPPC specific fragment, m/z 197.19. The surrounding gold region shows no signal. Scale bar: 50 µm.
molecule. The ratio of biotinylated to nonbiotinylated polymer was adjusted to around one probe-DNA per projected liposome area.25 This, in combination with a target-DNA concentration below the value of the dissociation constant (Kd) for the target hybridization, ensures that a majority of the liposomes are bound to the surface with one targetDNA.26 Liposomes were modified with capture-DNA by incubating the liposome suspension with hybridized pairs of cholesterol-terminated DNA strands.27 Figure 2A shows TOF-SIMS spectra recorded on substrates incubated with two different target-DNA sequences (see Supporting Information for details). The top spectrum, which was recorded on a sensor substrate exposed to targetDNA A (100 nM), and subsequently incubated with a mixture of POPC and D13-DPPC liposomes, each labeled with captureDNA A′ and B′, respectively, shows strong signal only for ion fragments specific to POPC (m/z 166.08, 184.11). In an © 2010 American Chemical Society
734
DOI: 10.1021/nl904208y | Nano Lett. 2010, 10, 732-737
FIGURE 3. (A) Normalized signal intensity from POPC and D13-DPPC as a function of reducing concentrations of target-DNA A. The POPC and D13-DPPC signals are sums of characteristic peaks for POPC (m/z 58 + 86 + 104 + 166 + 184 + 224) and D13-DPPC (m/z 66 + 98 + 117 + 179 + 197 + 237), in both cases normalized to the added signal intensities from the unspecific organic hydrocarbon fragments C2H3 and C2H5. The average of three measurements and the standard deviations are shown for each concentration. (B) Signal intensity (counts per pixel, 1.18 × 1.18 µm2) from POPC (m/z 184) and D13DPPC (m/z 197) at decreasing size of the analysis area, extracted from the data for the 100 nM target-DNA A sample. The average and standard deviations from nine separate spectra, obtained from three separate areas in each of the three analyzed total areas, are shown for each analysis area.
FIGURE 4. (A-D) Positive ion images of individual 300 nm liposomes for two substrates incubated with 100 pM DNA A (A,C) or DNA B (B,D), respectively. The top and bottom rows show ion images for m/z 184 (specific for POPC) and m/z 197 (specific for D13-DPPC), respectively. Small, high intensity spots [inset] represent individual liposomes (some which are marked with red [POPC] and green [D13DPPC] arrows to simplified interpretation). Field of view is 50 × 50 µm. (E) Spectra from (i) a single POPC or (iii) D13-DPPC liposome and spectra from the total area in (ii) panel A and (iv) panel D. The POPC (m/z 184) and D13-DPPC (m/z 197) specific peaks are colored in red and green, respectively.
good agreement with previous reports for the same DNA sequence measured with alternative techniques.26 Furthermore, the smallest area that could be probed while still obtaining a significant signal level above the background was as small as 1 µm2 at 100 nM, demonstrating that the sensor concept is compatible with small scale sensors and minute sample volumes (see Figure 3B). For this mode of analysis, the nonzero background signal from D13-DPPC is currently setting the limit of detection in terms of concentration and its reduction would directly improve the sensitivity of the method. In fact, utilizing the high lateral resolution and the chemical specificity of TOFSIMS, capable of spatially resolving individual nanometersized liposomes (Gunnarsson et al. submitted), we demonstrate here that under conditions at which each individual liposome is mediated by a single DNA target26 these liposomes can act as chemical barcodes for individual DNA targets in which case only the binding kinetics sets the © 2010 American Chemical Society
ultimate limit of detection. Figure 4A-D shows ion images specific for POPC (m/z 184, upper row) and D13-DPPC (m/z 197, lower row) obtained from TOF-SIMS analysis of substrates incubated with 100 pM target-DNA A (left column) or DNA B (right column). Individual liposomes are clearly visible in the POPC image from the DNA A sample, while virtually no liposomes are present in the D13-DPPC image, indicating specifically bound POPC liposomes on the surface, each mediated by a single DNA target. Similarly, the images for the DNA B sample show a large number of specifically bound D13-DPPC liposomes and only a few nonspecifically bound POPC liposomes. The observation of individual liposomes in these images were made possible by optimizing the instrument for high lateral resolution (
