Enhanced Single Molecule Mass Spectrometry via Charged Metallic

Oct 24, 2014 - Department of Physics, Virginia Commonwealth University, Richmond, ... For a more comprehensive list of citations to this article, user...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/ac

Enhanced Single Molecule Mass Spectrometry via Charged Metallic Clusters Christopher E. Angevine,† Amy E. Chavis,† Nuwan Kothalawala,‡ Amala Dass,‡ and Joseph E. Reiner*,† †

Department of Physics, Virginia Commonwealth University, Richmond, Virginia 23284, United States Department of Chemistry and Biochemistry, University of Mississippi, University, Mississippi 38677, United States



S Supporting Information *

ABSTRACT: Nanopore sensing is a label-free method for characterizing water-soluble molecules. The ability to accurately identify and characterize an analyte depends on the residence time of the molecule within the pore. It is shown here that when a Au25(SG)18 metallic cluster is bound to an α-hemolysin (αHL) nanopore, the mean residence time of polyethylene glycol (PEG) within the pore is increased by over 1 order of magnitude. This leads to an increase in the range of detectable PEG sizes and improves the peak resolution within the PEG-induced current blockade distribution. A model describing the relationship between the analyte residence time and the width of the peaks in the current blockade distribution is included. Finally, evidence is presented that shows the Coulombic interaction between the charged analyte and cluster plays an important role in the residence time enhancement, which suggests the cluster-based approach could be used to increase the residence time of a wide variety of charged analyte molecules.

R

application of the SMMS technique is the promise of measuring ion−analyte binding5−7 and differences in analyte conformation at the single molecule level.14 Given the role ion binding plays in protein folding15 and the importance of studying protein folding distributions with single molecule measurements,16 it appears SMMS could be coupled with standard mass spectrometry to extract critical information about physical and chemical properties of small protein molecules. This requires SMMS be demonstrated for other molecules besides PEG. The SMMS technique requires increasing analyte residence times to reduce the standard error of the estimated average from each blockade. For the case of PEG, it was found that increasing the ionic strength of the electrolyte increased the PEG residence time, which gave rise to the mass spectrum-like current blockade distributions seen in the initial SMMS results. However, increasing ionic strength will not increase analyte residence times for most analytes.17 Therefore; expanding the SMMS technique to molecules besides PEG requires the development of other methods for increasing residence times for a wide variety of molecules. One method for increasing analyte residence times is to modify physical characteristics of the solution (i.e., viscosity or temperature). This has been demonstrated with solid-state pores where reducing the solution temperature from T = 22 °C to T = 4 °C resulted in a ca. 1.7-fold increase in DNA residence times and increasing the solution viscosity 5-fold yielded a corresponding 5-fold increase in the residence time.17 In both

esistive-pulse nanopore sensing is a single molecule technique used to detect and analyze label-free analytes.1−3 The principle of operation is straightforward; a nanoscale channel is formed in a membrane that separates two chambers filled with electrolyte solution, and an applied transmembrane potential drives ionic current through the pore while molecules, commensurate in size with the pore, enter the channel and give rise to measurable current blockades. The magnitude of these blockades depends on the ratio of the molecular volume to the pore volume4 and in some cases the interaction between the molecule and the ions in solution.5−7 The duration of the blockades (i.e., the time the molecule spends inside the pore) depends on the electrophoretic drift through the pore, which can be affected by electroosmotic counterflow8 or entropic/enthalpic interactions between the analyte and pore.5−7 Water-soluble polymers played an important role in the development of nanopore sensing with early studies utilizing polyethylene glycol (PEG) of various molecular weights to characterize internal pore diameters9 and detailed pore geometries.10 It was later found that increasing the KCl concentration of the electrolyte increases the residence time of PEG in an α-hemolysin (αHL) pore, so that current blockades from individual molecules can be detected.4 Histograms of the averaged current from each blockade showed distinct peaks corresponding to PEG molecules with a specific number of monomer units.5,6,11−13 The similarity between these current blockade distributions and the corresponding mass spectra of PEG inspired the phrase “single molecule mass spectrometry” (SMMS).11 Though the promise of detecting mass differences between similarly sized molecules is intriguing, a more compelling © XXXX American Chemical Society

Received: April 29, 2014 Accepted: October 24, 2014

A

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

After the membrane is formed, a second pipette tip containing wild-type αHL (50 μg/mL) (List Biological, Campbell, CA) is positioned above the bilayer and a backing pressure (≈15 hPa) is applied for several seconds. A small trans-membrane voltage (typically 1−5 mV) is applied to verify pore insertion. When a sufficient number of pores (ca. 500) are in the membrane, the backing pressure is turned off and the protein tip is removed from solution. A quartz capillary (1 mm OD, 0.7 mm ID with filament, Sutter) is used to form a patch pipette tip with a laser-based pipette puller (P-2000, Sutter). This patch pipette contains a Ag/AgCl electrode. The patch pipette is brought down into contact with the membrane. If the patch membrane contains zero or more than one pore after membrane contact (as evidenced by comparing the measured current with the expected single channel conductance), then the tip is removed from the surface and a backing pressure is applied to remove the patch membrane. This process is repeated until a single αHL pore is localized within the patch area. The pure lipid composition facilitates the use of untreated glass for the patch pipettes and we typically observe leakage currents less than 1 pA throughout all measurements. Polydisperse PEG 1000 and PEG 1500 with mean molecular weights 1000 and 1500 Da, respectively, were purchased from Sigma-Aldrich (Sigma, St. Louis, MO). Monodisperse PEG28 (PEG with n = 28 monomer repeat units) and monodisperse PEG−diamine (PEG−diamine with n = 26 monomer repeat units) were purchased from Polypure (Polypure, Oslo, Norway). Au25(SG)18 Cluster Synthesis. One hundred milligrams of HAuCl4 (0.253 mmol/L) was dissolved in 50 mL of DI water, resulting in a yellow solution. Three hundred and seven milligrams of glutathione (1.00 mmol/L) was slowly added to the gold salt under slow stirring, while the yellow solution changed to a cloudy white suspension. Next the solution was cooled in an ice bath for 30 min. After 30 min, 94.6 mg of NaBH4 (2.5 mmol/L) was dissolved in 12.5 mL of ice cold DI water and added to the mixture all at once while stirring at 1000 rpm. The milky white color of the mixture rapidly turned black after the addition of NaBH4, indicating the formation of nanoparticles. After 1 h, the mixture was rotary evaporated until total volume was reduced to 5 mL, while the temperature was kept below 30 °C. Then 20 mL of methanol was added to the product mixture and centrifuged at 3800 rpm for 3 min. The resulting precipitate was washed three times with methanol. Au25(SG)18 was obtained using polyacrylamide gel electrophoresis (PAGE) as described previously.40 Data Collection and Analysis. Ionic currents were recorded using an amplifier head stage (Axopatch 200B, Molecular Devices, Carlsbad, CA), digitally sampled (Digidata 1440A, Molecular Devices) at 50 kHz and filtered with a fourpole 10 kHz low-pass Bessel filter. The data was recorded in axon binary files (.abf) with pClamp 10 software (Molecular Devices). Data analysis of the .abf files was performed with inhouse software written in Labview 8.5 (National Instruments, Austin, TX). A threshold algorithm, similar to one described previously,5 was used to calculate the average blockade depth and residence time for each event. The average blockade depth associated with the gold clusters from Figure 2 is about 25%. Therefore, to avoid confusing gold cluster and PEG-induced blockades, the current blockade threshold was set to detect events that remain 35% above the open-pore current for a minimum period of 140 μs. This allowed analysis of current blockades from both the open-nanopore state and the gold-

cases, the ionic conductivity and on-rate of analyte to the pore was reduced.18 This is unacceptable because both effects reduce the quality and sensitivity of SMMS analysis. To overcome these limitations, more sophisticated methods have been developed to increase analyte residence times. These methods include the application of external forces to the analyte,8,19−21 modification of the binding affinity of the analyte to the pore,22−28 and modification of the environment in or near the nanopore via polymers29 or nanoparticles.30 Although useful, most of these methods require complex methodologies (i.e., optical tweezers) that target a specific analyte (i.e., DNA). A need exists to develop simple methods that increase the residence time for a wide variety of different analyte molecules. This paper describes a method that utilizes charged metallic clusters to increase the nanopore residence time of oppositely charged analytes. Briefly, a negatively charged, glutathionethiolate capped, water-soluble Au25(SG)18 cluster31−37 is driven into the cis side vestibule of an αHL pore under an applied electric field. The cluster remains in the pore for extended periods (>10s), but only partially blocks the flow of ions (≈25% reduction) while PEG molecules enter the trans side of the pore and interact with the cluster−pore complex. This leads to an increase in the PEG residence time over a large range of PEG sizes and ionic strengths. It is shown both experimentally and with a simple model that increasing the PEG residence time yields higher resolution peaks in the SMMS current blockade distribution, which improves nanopore-based analysis of the polymer. Finally, we present evidence that the enhancement mechanism results from a Coulombic attraction between the oppositely charged analyte and cluster. This suggests the cluster-based approach should be applicable to other molecules besides PEG, which will enable the SMMS technique to work for a wide variety of charged analytes.



EXPERIMENTAL SECTION Membrane Formation and Patch Clamping. Experiments were carried out with patch pipette electrophysiology methods. Briefly, a 1 cm2 PTFE (Teflon) sheet with a ca. 100 μm hole in the center (Eastern Scientific, Rockville, MD) is fixed to a previously fabricated large holder with PDMS (KwikCast, WPI, Sarasota, FL).38 The holder positions the Teflon sheet ca. 300 μm from the top of a microscope coverslip mounted onto a homemade holder that sits on an inverted microscope (Axio Observer D, Zeiss, Germany). The microscope is bolted to an air-floated optical table within a Faraday cage enclosure (Kinetic Systems, Boston, MA). A prepaint mixture (2 μL of 0.5 mg/mL DPhyPC (1,2 diphytanolyl-snglycero-3-phosphatidylcholine; Avanti Polar Lipids, Alabaster, AL) dissolved in pentane) is applied to both sides of the 100 μm hole prior to attaching the Teflon sheet to the holder. Membranes are formed with a painting methodology closely related to the method of Muellar et al.39 A glass microcapillary (femtotip II, Eppendorf, NY) is positioned with a motorized manipulator (MPC-275, Sutter Instruments, Novato, CA) ca. 100 μm above the hole in the Teflon partition. Several picoliters of 10 mg/mL DPhyPC:hexadecane solution is ejected under pressure (Femtojet, Eppendorf) from the tip and adhered to the Teflon surface. A glass rod with a ball formed at the end is manipulated with a manual translation stage to wipe the lipid across the hole. The lipid/solvent mixture thins over a period of several seconds and a lipid bilayer membrane is formed. The membrane formation process is verified with brightfield optical microscopy. B

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

blocked state for the same pore. The average current for each blockade was normalized by the average open current 1 ms before and 1 ms after the blockade event. Blockade events were discarded if the averaged open-state current fell outside a narrowly defined window (fwhm of the open-pore current distribution and fwhm of the largest peak in the clusteroccupied current distribution). The n-values for each peak in the blockade distributions (see Figure 3a) were calibrated by assigning n = 28 to the largest peak.5,6,11 The residence time for each event was defined as the period 20 μs before the initial threshold crossing (initiate the blockade) to 20 μs before the second threshold crossing (completion of blockade). The residence time for a given blockade was assigned to an n-mer PEG molecule if the magnitude of the current blockade fell between the two minima surrounding the nth peak in the current blockade distribution. Ten-bin histograms (automatic bin widths) were calculated from the residence times for each n-sized PEG using the histogram analysis package in IGOR 6.22A (Wavemetrics Inc. Lake Oswego, OR). Single offset exponentials were used to perform weighted fits of each residence time distribution (see the Supporting Information, section 3). Unless otherwise noted, all reported error bars in graphs and uncertainties are ±1 standard deviation. The resolution for the nth peak in the current blockade distributions was calculated with Rn = 0.5 × (μn − μn+1)/(⟨δxn⟩ + ⟨δxn+1⟩) where μn and ⟨δxn⟩ correspond to the mean and standard deviation of the nth peak, respectively.41 Unless otherwise noted, μn and ⟨δxn⟩ were found from a multipeak Gaussian fit to each current blockade distribution using the Multipeak Fitting 2 package in IGOR.

Figure 1. Nanopore detection of gold clusters verifies the sign of the cluster charge and yields multiple blockade states. For presentation purposes, the data in this figure was filtered with a 100 Hz low-pass Hanning filter. (A) Clusters randomly enter and exit the pore from the cis side under an applied negative transmembrane potential (ground is fixed on the trans side). (B) Hanning-filtered, open-state current with no gold present is ⟨iopen⟩(−50 mV) = −(172.3 ± 0.6) pA. When a cluster enters the pore, the current is reduced as seen by the short and long-lived current blockades. The average on-rate for the clusters to the pore is kon (−50 mV) = (0.12 ± 0.01) s−1. (B, inset) No blockades are observed under a positive applied potential where the open-state current is ⟨iopen⟩(+50 mV) = (165.8 ± 1.0) pA. This verifies that the clusters are negatively charged. (C) An all-points histogram of the negative voltage current trace demonstrates the quantized nature of the current blockades. The large, off-scale peak, corresponds to the open-pore current and the smaller peaks between −150 and −130 pA correspond to the gold cluster blockade states. A least-squares fit to the largest blockade peak yields ⟨igold⟩ = −(133.7 ± 1.1) pA. Measurements were performed with an applied voltage of ±50 mV, [KCl] = 3.5 mol/L (pH = 7.2) and [Au25(SG)18] = 5 μmol/L.



RESULTS AND DISCUSSION Nanopore Current Blockades. Figure 1A shows an illustration of the experimental setup and the method of detection. Clusters diffuse in the cis side solution and randomly enter the pore under an applied voltage. Figure 1B shows shortand long-lived current blockades that result from gold clusters entering and exiting the pore under a negative applied voltage (ground is fixed on the trans side of the pore). The short-lived events could be used to characterize the gold clusters,1 but the focus here is on the longer-lived blockades (>1 s) that enable the increased PEG residence times shown in Figure 2. The diameter of the Au25(SG)18 clusters has not been characterized in solution, but the similarly sized cluster Au25(SCH2CH2Ph)18 has been crystallized, and found to have a diameter of 2.4 nm.31 This is small enough to allow entry into the cis side of the αHL pore (dopen = 2.6 nm), but too large to translocate across the narrow constriction ring separating the cis and trans sides of the pore (dring = 1.5 nm).42 This suggests that the long-lived blockades result from clusters being forced down against the constriction of the pore and held there under the applied transmembrane potential. The magnitude of a typical cluster-induced blockade is ⟨igold⟩/⟨iopen⟩ ≈ 0.75, where ⟨⟩ represents a time average, igold is the nanopore current with a gold cluster in the pore, and iopen is the current through an open pore. A similar behavior was previously observed for 3-mercapto-1-propanesulfonate (MPSA) coated gold nanoparticles interacting with wild-type αHL pores.43,44 In those reports, deeper blockades (⟨igold⟩/ ⟨iopen⟩ < 0.5) were observed and attributed to interactions between the nanoparticle ligands and the narrow constriction ring separating the cis-side vestibule from the trans-side

lumen.43 We rarely observe these deeper blockades in our experiments, which suggests that the Au25(SG)18 clusters do not strongly interfere with the salt bridges formed by the lysine and glutamate amino acids in the constriction ring. The cluster-induced blockades leave a sufficient dynamic range for characterizing other analytes that partition into the pore while a cluster is in the cis side of the pore. Control experiments verified that the blockades result from clusters and not channel gating or other spurious artifacts (Supporting Information, section 1). The inset to Figure 1B shows a blockade-free current trace resulting from a positive applied voltage. The absence of blockades verifies that the clusters are negatively charged. Figure 1C shows an all-points histogram of the current trace from Figure 1B, which shows five distinct blockade states. Electrospray ionization mass spectrometry suggests the cluster distribution is monodisperse (Supporting Information, section 2) so the multipeak distribution between |i| = (130−150) pA could result from several possibilities C

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 2. Current traces with and without a cluster in the pore demonstrate increased PEG residence times inside a gold-clusteroccupied αHL pore. (A) The empty pore exhibits an open-state current ⟨iopen⟩(−50 mV) = −(174.0 ± 2.7) pA. PEG molecules enter the pore at a rate of kon (−50 mV) = (11.5 ± 0.1) s−1 and yield shortlived current blockades. (B) When a cluster enters the cis side of the pore the current is reduced to ⟨igold⟩(−50 mV) = −(136.5 ± 2.6) pA. The negatively charged gold cluster in the pore yields longer-lived PEG blockades. Measurements were performed under an applied potential of −50 mV with [KCl] = 3.5 mol/L (pH = 7.2), [Au25SG18] = 5 μmol/L, [PEG1000] = [PEG1500] = 10 μmol/L and [PEG28] = 1 μmol/L.

including different charge states of the clusters, differently sized cluster isomers,45 the clusters residing in different parts of the pore,46 or cluster-induced modifications to the nanopore volume.47,48 From a cursory examination of time series data, it is clear that a cluster in the pore can increase the PEG residence time (Figure 2). Figure 2A shows a current trace resulting from PEG molecules entering and exiting the αHL pore from the trans side with no cluster present. The PEG blockades are deeper (⟨iPEG⟩/⟨iopen⟩ ≈ 0.35) and shorter-lived than the clusterinduced blockades seen in Figure 1B. Figure 2B shows a current trace after a cluster enters the same pore from the cis side. The resulting PEG-induced current blockades are still deep, yet they last for significantly longer periods of time. Molecular dynamics simulations have shown that PEG coordinates K+ cations to form crown-ether like structures in high ionic strength solutions.6 Our hypothesis is that the anionic cluster is forced into the pore under the applied potential and interacts with the oppositely charged, cation-like, PEG near the constriction ring of the pore. We justify this hypothesis later in the paper, but we start with a discussion of the improvements to nanopore analysis resulting from the cluster-induced residence time enhancement. Nanopore Blockade Analysis. Figure 3 verifies that the gold cluster improves the analysis of a polydisperse mixture of PEG in 3.5 mol/L KCl solution. Figure 3A shows histograms of the PEG-induced current blockades for the cluster-free and cluster-filled pore. Each current blockade is normalized with respect to the current through the open pore (black) or clusterfilled pore (red). The peaks in each distribution correspond to current blockades that arise from PEG molecules with a distinct number of n monomer repeat units.5,11 The blockade distribution for the gold-occupied pore (red) shows a greater range of resolvable PEG sizes (n = 13−39 vs n = 22−39) and narrower peaks than the cluster-free pore (black). The slight shift between the two distributions along the horizontal axis is

Figure 3. PEG-induced current blockade distributions and corresponding mean residence time distributions with (red) and without (black) a metallic cluster in the pore. (A) Distribution of normalized PEG current blockades yields peaks corresponding to different sized PEG molecules where larger PEG molecules yield deeper current blockades.5,11 The large peak in both distributions corresponds to the monodisperse PEG28 that calibrates the polymer repeat number for all peaks.11 The y-axis in both distributions corresponds to the normalized frequency such that the total area under each curve is one. The number and range of resolvable peaks appears greater in the gold trapped distribution (n = 13−39) than the open-pore distribution (n = 22−39). (B) Gold clusters in the pore increase the mean residence time by approximately 1 order of magnitude. This trend is observed over the entire range of measurable PEG sizes. The data was recorded with a −50 mV applied transmembrane potential (ground is fixed on the trans side) on a single pore fluctuating between the clusteroccupied and open-pore states. The cluster free and cluster-occupied distributions were calculated from 5443 and 5575 blockade events, respectively. Measurements were performed with [KCl] = 3.5 mol/L (pH = 7.2), [Au25SG18] = 5 μmol/L, [PEG1000] = [PEG1500] = 10 μmol/L and [PEG28] = 1 μmol/L.

expected because the gold cluster changes the effective size of the nanopore sensing volume.4,5,47,48 Even more compelling, Figure 3B shows that the mean residence time for the PEG molecules increases by about 1 order of magnitude over the entire range of measurable peaks when the cluster is inside the pore. Peak Narrowing with Cluster-based Detection. The initial SMMS results showed that increasing the PEG residence time made it possible to distinguish between current blockades that differ in size by a single monomer unit.11 Here we quantify the relationship between the residence time and the width of the peaks in the current blockade distribution for exponentially distributed residence times. This shows that increasing the PEG residence times with the gold clusters yields narrower peaks and more accurate characterization of the analyte molecules. D

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Longer lived blockades yield mean estimates with lower standard errors. We use this fact to derive an expression for the current blockade distribution with the goal of finding a relationship between the PEG residence times and the peak widths in the current blockade distribution. We start by assuming blockade currents corresponding to n-sized PEG are Gaussian distributed with mean μn and variance σ2 where the variance is independent of the PEG size. The duration of each blockade td for a given n-sized PEG molecule is random and exponentially distributed with mean residence time τ n (Supporting Information, section 3).5,6,11 The current is digitized at frequency fs so the standard error of the mean for each blockade can be approximated by σ2/sqrt( fstd) where fstd is the number of data points in the blockade. We assume that the current blockade distribution for n-sized PEG is a weighted sum over all possible blockade durations as follows ∞ ⎛ j ⎞ ⎟⎟Gn(j , x , μn , σ , α) Pn(x) = C ∑ exp⎜⎜ − ⎝ fs τn ⎠ j = j0

(1)

where x = ⟨iPEG⟩/⟨iopen⟩ or ⟨iPEG⟩/⟨igold⟩ depending on whether or not there is a gold cluster in the pore, C = [∑j ∞= j0exp(− j/ fsτn)]−1 is a normalization constant, j is an integer that parametrizes time with lower bound j0 = fstcutoff = 50 kHz × 140 μs = 7 for this paper, α is a constant that corrects for serial correlation in the current,49 and Gn is the normalized sampling distribution of the sampling mean, given by Gn(j , x , μn , σ , α) =

⎛ j(x − μ )2 ⎞ n ⎟ ⎜⎜ − exp ⎟ 2πασ 2 2ασ 2 ⎠ ⎝ j

(2)

From eqs 1 and 2, it can be shown that the mean of the nth peak in the distribution is ⟨xn⟩ = μn and the variance (⟨δxn2⟩ = ⟨xn2⟩ − ⟨xn⟩2) is given by j ⎛ ⎡ jo − 1 zn ⎤⎞ ⎜ 2⎢ ln(1 − zn) + ∑ j = 1 j ⎥⎟ 2 ⟨δxn ⟩ = ⎜ασ ⎢ ⎥⎟ znj0 ⎜ ⎢ ⎥⎦⎟⎠ ⎣ z 1 − ⎝ n

Figure 4. Increasing PEG residence times leads to narrower peaks in the current blockade distribution. (A) Weighted fits of the n = 28 peaks shown in Figure 3a for the open (black) and cluster-occupied (red) pores demonstrate peak narrowing with increased residence time. For comparison purposes, the peaks were rescaled with both means centered at zero and the peak heights set to unity. The ratio of the standard deviations of the open-pore and cluster-occupied peaks is (2.0 ± 0.3), which is in reasonable agreement with the ratio predicted by eq 3 (2.6 ± 0.1). Each histogram was constructed with 20 bins, and error bars shown on the graph correspond to ± N1/2 where N is the number of counts in each bin. Note that the fwhm for each peak is related to the standard deviation by fwhm ≈ 2.4 ⟨δxn⟩. (B) Relationship between the width of the peaks in the blockade distribution and the mean residence times is shown for all observed PEG sizes. The mean residence time and corresponding standard deviation of each peak in the cluster-occupied blockade distribution is plotted as open circles and fit with eq 3 (solid line). Each data point corresponds to a unique peak in the current blockade distribution. Data points corresponding to peaks n = 16, 20, and 28 are identified for clarity. The model is a weighted least-squares fit to the data using eq 3 with σ = 0.013, fs = 50 kHz, j0 = 7 and the freely adjustable parameter α = 2.2 ± 0.2. The reduced χ2 value for the fit is 1.3.

(3)

⟨δxn2⟩1/2

where zn = exp(−1/fsτn) and ⟨δxn⟩ = is the standard deviation of the nth peak in the current blockade distribution. Note that the standard deviation is related to the full-width at half-maximum of the peaks (fwhm ≈ 2.4 ⟨δxn⟩ for a single Gaussian peak). Given that a cluster in the pore increases the mean residence time of PEG in the pore, it is expected and shown in eq 3 and Figure 4A that the peak widths for the cluster-occupied pore are narrower than the peak widths for the open-pore case. This is explicitly shown in Figure 4A where the n = 28 peaks from Figure 3A are compared. The two peaks are rescaled to unit amplitude and centered at mean zero to make the comparison of peak widths more clear. Each peak is fit with a single Gaussian function and the results show that the cluster’s presence in the pore reduces the peak width by a factor of 2.0 ± 0.3, which is in reasonable agreement with the value predicted by eq 3 ⟨δx28,open (τ = 0.6 ms)⟩/⟨δx28,gold (τ = 11.5 ms)⟩ = 2.6 ± 0.1. The most important result from eq 3 is that increasing the mean residence time of the current blockades τn leads to smaller peak widths in the current blockade distribution ⟨δxn⟩. This in turn leads to more accurate estimates of the mean blockade current μn and a greater probability of associating each blockade with the correct analyte. The dependence between

⟨δxn⟩ and τn is shown explicitly in Figure 4B where the peak width for each n-sized PEG in the cluster-occupied pore (red data, Figure 3A) is plotted against the corresponding mean residence time in the pore. A least-squares fit using eq 3 shows excellent agreement with the data. The reduced peak widths in Figure 3A and modeled in Figure 4 lead to an increase in the peak resolution across all nsized PEG molecules. Figure 5 quantifies this enhancement by E

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

Figure 5. Increasing the residence time yields narrower peaks in the current blockade distributions with higher resolution (Rn = 0.5 × (μn − μn+1)/(⟨δxn⟩ + ⟨δxn+1⟩)). Higher values of R correspond to more easily resolvable peaks. The resolution for each peak in the current blockade distributions shown in Figure 3A is improved with a cluster in the pore. The red points correspond to the gold-occupied peaks and the black points correspond to the open-pore peaks. The multipeak Gaussian fitting algorithm did not converge for peaks corresponding to n < 23 in the open-pore configuration. Baseline resolution corresponds to R ≈ 1.5.41

comparing the resolution R (see Data Collection and Analysis section) of the peaks from both distributions shown in Figure 3A. Briefly, the resolution compares the width of adjacent peaks to the spacing between the peaks. Higher values of R correspond to peaks that are more easily resolved. Specifically, for the case of peaks with identical height, the lowest possible resolution for separating two adjacent peaks is 0.5 and baseline resolution is expected for R ≈ 1.5.41 Figure 6 shows that the cluster modified pore improves the peak resolution across the entire range of measurable peaks. In general, increasing peak resolution in the SMMS technique improves the analytical capabilities of the nanopore sensor because current blockades from differently sized analyte molecules can be more accurately distinguished. Residence Time Enhancement Mechanism. Full development of the cluster-based approach for improving SMMS requires an understanding of the residence time enhancement mechanism. Our hypothesis is that the cation-like PEG experiences a Coulombic attraction with the anionic cluster near the constriction region of the pore that confines the PEG for longer periods of time. We report a series of experiments that support this hypothesis. Nanopore sensing of PEG has demonstrated that K+ cations strongly bind to PEG and this causes PEG to behave like a polycation in high ionic strength solutions.5,11 Therefore, our hypothesis predicts that increasing the KCl concentration will increase the PEG charge and thus the strength of the Coulombic attraction between the PEG and cluster. This should lead to a correlation between the KCl concentration and the degree of residence time enhancement. Figure 6 confirms this prediction because it shows the residence time enhancement factor (ratio of cluster-occupied residence time to openpore residence time) grows with increasing KCl concentration. To further verify that the charged analyte plays an important role in the residence time enhancement, we performed a control experiment with PEG in LiCl. It was previously shown

Figure 6. Metallic clusters increase the PEG residence time over a wide range of KCl concentrations, which illustrates that the enhancement mechanism depends on the cation binding to the PEG. (A) Mean residence time for monodisperse PEG28 in the pore both with (red circles) and without (black circles) a gold cluster present shows that the enhancement is observed over a wide range of KCl concentrations. (B) PEG binds K+ cations,6 so if the enhancement mechanism depends on the Coulombic attraction between the PEG and cluster, then the residence time enhancement factor (tres(gold)PEG28/tres(open)PEG28) should grow with increasing KCl concentration. This behavior is observed over the entire range of KCl concentrations studied. For KCl concentrations of 1.5, 2.0, and 2.5 mol/L, the measurements were performed with a monodisperse PEG mixture ([PEG28] = 5 μmol/L). The 3.0 and 3.5 mol/L data was recorded using a polydisperse PEG mixture with [PEG1000] = [PEG1500] = 10 μmol/L and [PEG28] = 1 μmol/L. The applied transmembrane potential was −50 mV for all data shown. For each KCl concentration, data was recorded with the same pore fluctuating between the open and cluster-occupied states.

that PEG−Li+ cation binding is much weaker than PEG−K+ binding.7 Therefore, the PEG in LiCl is essentially neutral. Our hypothesis predicts the PEG residence time in LiCl should not be affected by a cluster in the pore. Figure 7 confirms this behavior, which shows a current trace of PEG blockades in LiCl both with and without a cluster in the pore. Unlike Figure 2B, the PEG blockades in LiCl do not appear dramatically affected by the cluster in the pore. However, a closer examination of the residence time distributions (Figure 7B) shows a ca. 2.5-fold increase in the PEG residence time with a cluster in the pore. It is worth noting that a similar enhancement factor is observed at the lower KCl concentrations ([KCl] < 2 mol/L) shown in Figure 6B. This is significantly less than the nearly 20-fold F

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

which have been shown to cause a structural change in PEG,6 increasing the entropic barrier for PEG translocation through the cluster-occupied pore. We rule out this possibility by measuring the current blockades from PEG−diamine in low ionic strength KCl solution. Low ionic strength here means cation binding to the PEG is negligible as evidenced by the fact that PEG blockades are not observed at the PEG concentrations used throughout the manuscript ([PEG28] ≈ 5 μmol/L). At [KCl] = 1 mol/L, the only source of charge on the PEG−diamine comes from the amine groups on each end of the PEG molecule. These amine groups are positively charged in the near pH neutral conditions used in our experiments. If the enhancement only results from a structural change of PEG in response to cation binding and not the charge on the PEG, then we expect to see no enhancement for PEG−diamine in low ionic strength solutions. Figure 8 shows

Figure 7. Limited residence time enhancement for PEG in LiCl supports the hypothesis that bound charge to the PEG is a dominant factor in the residence time enhancement mechanism. (A) Current trace shows current blockades in both the gold-occupied and open state that are essentially equivalent. Cation binding to PEG is much weaker for LiCl than KCl, which makes PEG essentially neutral in LiCl solutions.7 The neutral PEG does not interact strongly with the gold cluster and we see limited enhancement. (B) Residence time distributions for the open-pore (black circles) and gold-occupied (red circles) states. The solid lines are least-squares fits to the data with offset exponential functions. The mean residence time from both fits are tres(open)PEG28 = (31.8 ± 1.4) μs and tres(gold)PEG28 = (78.8 ± 5.0) μs for the open-pore and gold-occupied states, respectively. Lithium chloride yields a slight enhancement factor of (tres(gold)PEG28/ tres(open)PEG28 = 2.5). For comparison, the inset shows the residence time distributions for PEG28 in 3.5 mol/L KCl, which shows about a 1 order of magnitude greater enhancement than the LiCl (tres(gold)PEG28/tres(open)PEG28 = 17.1). This illustrates the important role the bound cations play in the residence time enhancement mechanism. Data was recorded in 3.5 mol/L LiCl (pH = 7.2) under a −50 mV applied transmembrane potential (ground is fixed on the trans side of the pore). The PEG28 concentration was increased 1000fold (5 mmol/L) because of the limited number of blockade events observed in LiCl.7

Figure 8. Current traces for PEG−diamine with and without a cluster in the pore to verify the role Coulombic attraction plays in the residence time enhancement. At low ionic strengths, we expect no cation binding with the PEG, so any enhancement results from the Coulombic interaction between the amine groups and the anionic cluster. (A) With no cluster in the pore, the PEG−diamine blockade events are very short-lived and undetectable. (B) When a cluster enters the pore, the current is reduced and the events become longer lived and detectable. The residence times in the gold-occupied state were exponentially distributed with a mean residence time of (100 ± 2) μs. Data was taken with [PEG−diamine] = 5 μmol/L, [KCl] = 1 mol/L (pH 7.2) and Vapp = −80 mV.

the current trace for PEG−diamine immersed in 1 mol/L KCl in which it is clear that the PEG−diamine is affected by the cluster in the pore. Without a gold cluster in the pore, the PEG−diamine blockades are too short-lived to be detected at the measurement bandwidth (B = 10 kHz). However, insertion of a cluster into the pore yields a significant increase in the number, magnitude, and duration of blockade events. The increased residence time for the PEG−diamine in 1 mol/L KCl demonstrates that the charge on the PEG, and not a structural change from cation−PEG binding, is the dominant cause of the residence time enhancement. The correlation between the KCl concentration and the degree of residence time enhancement, the greatly reduced enhancement in LiCl, and the PEG−diamine enhancement at [KCl] = 1 mol/L all suggest that the residence time enhancement for the PEG−cluster−pore system is dominated by the Coulombic attraction between the PEG and the cluster. The LiCl results suggest that the cluster may serve as a barrier to translocation through the pore and this could play a role in

increase seen for PEG in 3.5 mol/L KCl (Figure 7B inset). The slight residence time enhancement, in LiCl and low ionic strength KCl, most likely results from the cluster restricting the flow of neutral PEG out the cis side of the pore. The critical result here is that the neutral PEG shows only a small degree of residence time enhancement. This illustrates the importance of the Coulombic interaction in the residence time enhancement mechanism. It is clear from Figures 6 and 7 that bound cations play an important role in the enhancement effect. However, it may be possible that the charged PEG is not responsible for the residence time enhancement, but rather the bound cations, G

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry



the enhancement, but the slight increase in the residence time for PEG in LiCl suggests this effect is limited. To a first approximation, it appears the Coulombic attraction between the analyte and cluster plays an important role in the enhancement. This suggests that the cluster-based approach for increasing residence times should be applicable to a wide variety of charged analytes.

CONCLUSION The goal of this work is to establish a new method for increasing the mean residence time of a charged analyte in αHL pores. The technique utilizes charged clusters that enter the cis side of the pore and interact with oppositely charged molecules that enter from the trans side of the pore. The principle of operation was demonstrated with PEG in high-ionic strength KCl solution where PEG has been shown to behave like a polycation.5,11 Specifically, we observe an over 1 order of magnitude increase in the PEG residence time for a wide range of PEG sizes and KCl concentrations. We modeled the dependence of the PEG residence time on the peak widths of the current blockade distribution, which showed that increasing the PEG residence time leads to narrower peak widths in the current blockade distributions. This in turn improves the ability of the nanopore sensing to accurately correlate each blockade with the correct analyte molecule. Finally, we showed that the residence time enhancement depends on the charge of the analyte molecule, which implies that the cluster-based enhancement should work for a wide variety of charged analytes. Further optimization of the residence time enhancement should be possible with adjustments to the analyte charge and the applied transmembrane voltage. Future studies will explore these dependencies. ASSOCIATED CONTENT

S Supporting Information *

Description of a control experiment to verify shallow blockades result from clusters entering the pore, EIS-MS measurements of the clusters to verify the clusters are negatively charged and monodisperse, and residence time distributions of the PEG in the pore to justify the exponential residence time distribution used to derive eq 3. This material is available free of charge via the Internet at http://pubs.acs.org.



REFERENCES

(1) Kasianowicz, J. J.; Robertson, J. W. F.; Chan, E. R.; Reiner, J. E.; Stanford, V. M. Annu. Rev. Anal. Chem. 2008, 1, 737−766. (2) Howarka, S.; Siwy, Z. S. Chem. Soc. Rev. 2009, 38, 2360−2384. (3) Bayley, H.; Martin, C. R. Chem. Rev. 2000, 100, 2575−2594. (4) Krasilnikov, O. V.; Rodrigues, C. G.; Bezrukov, S. M. Phys. Rev. Lett. 2006, 97, 018301. (5) Reiner, J. E.; Kasianowicz, J. J.; Nablo, B. J.; Robertson, J. W. F. Proc. Natl. Acad. Sci. U. S. A. 2010, 107, 12080−12085. (6) Balijepalli, A.; Robertson, J. W. F.; Reiner, J. E.; Kasianowicz, J. J.; Pastor, R. W. J. Am. Chem. Soc. 2013, 135, 7064−7072. (7) Breton, M. F.; Discala, F.; Bacri, L.; Foster, D.; Pelta, J.; Oukhaled, A. J. Phys. Chem. Lett. 2013, 4, 2202−2208. (8) Keyser, U. F.; Koeleman, B. N.; Van Dorp, S.; Krapf, D.; Smeets, R. M. M.; Lemay, S. G.; Dekker, N. H.; Dekker, C. Nat. Phys. 2006, 2, 473−477. (9) Krasilnikov, O. V.; Sabirov, R. Z.; Ternovsky, V. I.; Merzliak, P. G.; Muratkhodjaev, J. N. FEMS Microbiol. Immunol. 1992, 105, 93− 100. (10) Merzlyak, P. G.; Yuldasheva, L. N.; Rodrigues, C. G.; Carneiro, C. M. M.; Krasilnikov, O. V.; Bezrukov, S. M. Biophys. J. 1999, 77, 3023−3033. (11) Robertson, J. W. F.; Rodrigues, C. G.; Stanford, V. M.; Rubinson, K. A.; Krasilnikov, O. V.; Kasianowicz, J. J. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 8207−8211. (12) Rodrigues, C. G.; Machado, D. C.; Chevtchenko, S. F.; Krasilnikov, O. V. Biophys. J. 2008, 95, 5186−5192. (13) Baaken, G.; Ankri, N.; Schuler, A.-K.; Rühe, J.; Behrends, J. C. ACS Nano 2011, 5, 8080−8088. (14) Talaga, D. S.; Li, J. J. Am. Chem. Soc. 2009, 131, 9287−9297. (15) Zhang, Y.; Cremer, P. S. Curr. Opin. Chem. Biol. 2006, 10, 658− 663. (16) Borgia, A.; Williams, P. M.; Clarke, J. Annu. Rev. Biochem. 2008, 77, 101−125. (17) Fologea, D.; Uplinger, J.; Thomas, B.; McNabb, D. S.; Li, J. Nano Lett. 2005, 5, 1734−1737. (18) Berg, H. C.; Purcell, E. M. Biophys. J. 1977, 20, 193−219. (19) van Dorp, S.; Keyser, U. F.; Dekker, N. H.; Dekker, C.; Lemay, S. G. Nat. Phys. 2009, 5, 347−351. (20) Peng, H. B.; Ling, X. S. S. Nanotechnology 2009, 20, 185101. (21) King, G. M.; Golovchenko, J. A. Phys. Rev. Lett. 2005, 95, 216103. (22) Henrickson, S. E.; Misakian, M.; Robertson, B.; Kasianowicz, J. J. Phys. Rev. Lett. 2000, 85, 3057−3060. (23) Rincon-Restrepo, M.; Mikhailova, M.; Bayley, H.; Maglia, G. Nano Lett. 2011, 11, 746−750. (24) Henrickson, S. E.; DiMarzio, E. A.; Wang, Q.; Stanford, V. M.; Kasianowicz, J. J. J. Chem. Phys. 2010, 132, 135101. (25) Clarke, J.; Wu, H.-C.; Jayasinghe, L.; Patel, A.; Reid, S.; Bayley, H. Nat. Nanotechnol. 2009, 4, 265−270. (26) Dahl, J. M.; Mai, A. H.; Cherf, G. M.; Jetha, N. N.; Garalde, D. R.; Marziali, A.; Akeson, M.; Wang, H.; Lieberman, K. R. J. Biol. Chem. 2012, 287, 13407−13421. (27) Ayub, M.; Hardwick, S. W.; Luisi, B. F.; Bayley, H. Nano Lett. 2013, 13, 6144−6150. (28) Keyser, U. F. J. R. Soc., Interface 2011, 8, 1369−1378. (29) Squires, A. H.; Hersey, J. S.; Grinstaff, M. W.; Meller, A. J. Am. Chem. Soc. 2013, 135, 16304−16307. (30) Astier, Y.; Uzun, O.; Stellacci, F. Small 2009, 5, 1273−1278. (31) Heaven, M. W.; Dass, A.; White, P. S.; Holt, K. M.; Murray, R. W. J. Am. Chem. Soc. 2008, 130, 3754−3755. (32) Zhu, M.; Aikens, C. M.; Hollander, F. J.; Schatz, G. C.; Jin, R. J. Am. Chem. Soc. 2008, 130, 5883−5885. (33) Shichibu, Y.; Negishi, Y.; Tsunoyama, H.; Kanehara, M.; Teranishi, T.; Tsukuda, T. Small 2007, 3, 835−839. (34) Xie, S.; Tsunoyama, H.; Kurashige, W.; Negishi, Y.; Tsukuda, T. ACS Catal. 2012, 2, 1519−1523. (35) Liu, Y.; Tsunoyama, H.; Akita, T.; Tsukuda, T. Chem. Commun. 2010, 46, 550−552.





Article

AUTHOR INFORMATION

Corresponding Author

*Joseph E. Reiner. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS J.E.R. acknowledges funding from the VCU Presidential Research Quest Fund and VCU start-up funding, a generous equipment donation from NIST in Gaithersburg, MD, along with helpful discussions with Arvind Balijepalli, Joseph Robertson, and John Kasianowicz during the preparation of this paper. We acknowledge Grace Cummings who helped with the LiCl and PEG−diamine experiments and we also acknowledge the anonymous reviewer who suggested the LiCl experiment to help explain the role bound cations play in the residence time enhancement mechanism. H

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Article

(36) Zhu, Y.; Qian, H.; Drake, B. A.; Jin, R. Angew. Chem., Int. Ed. 2010, 49, 1295−8. (37) Zhu, M. Z.; Aikens, C. M.; Hendrich, M. P.; Gupta, R.; Qian, H. F.; Schatz, G. C.; Jin, R. J. Am. Chem. Soc. 2009, 131, 2490−2492. (38) Reiner, J. E.; Robertson, J. W. F.; Burden, D. L.; Burden, L. K.; Balijepalli, A.; Kasianowicz, J. J. J. Am. Chem. Soc. 2013, 135, 3087− 3094. (39) Mueller, P.; Rudin, D. O.; Tien, H. T.; Wescott, W. C. J. Phys. Chem. 1963, 67, 534−535. (40) Kothalawala, N.; West, J. L., IV; Dass, A. Nanoscale. 2014, 6, 683−687. (41) Felinger, A. Anal. Chem. 1997, 69, 2976−2979. (42) Song, L.; Hobaugh, M. R.; Shustak, C.; Cheley, S.; Bayley, H.; Gouaux, J. E. Science 1996, 274, 1859−1865. (43) Campos, E.; Asandei, A.; McVey, C. E.; Dias, J. C.; Oliveira, A. S. F.; Soares, C. M.; Luchian, T.; Astier, Y. Langmuir 2012, 28, 15643− 15650. (44) Campos, E.; McVey, C. E.; Carney, R. P.; Stellacci, F.; Astier, Y.; Yates, J. Anal. Chem. 2013, 85, 10149−10158. (45) Huang, W.; Pal, R.; Wang, L.-M.; Zeng, X. C.; Wang, L.-S. J. Chem. Phys. 2010, 132, 054305. (46) Aksimentiev, A.; Schulten, K. Biophys. J. 2005, 88, 3745−3761. (47) Robertson, J. W. F.; Kasianowicz, J. J.; Reiner, J. E. J. Phys.: Condens. Matter 2010, 22, 454108. (48) Robertson, J. W. F.; Kasianowicz, J. J.; Banerjee, S. Chem. Rev. 2012, 112, 6227−6249. (49) Flyvbjerg, H.; Petersen, H. G. J. Chem. Phys. 1989, 91, 461−466.

I

dx.doi.org/10.1021/ac503425g | Anal. Chem. XXXX, XXX, XXX−XXX