Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Am Chem Soc. Author manuscript; available in PMC 2010 July 8.
Published in final edited form as:
PMCID: PMC2717167

Single-molecule Protein Unfolding in Solid State Nanopores


We use single silicon nitride nanopores to study folded, partially folded and unfolded single proteins by measuring their excluded volumes. The DNA-calibrated translocation signals of β-lactoglobulin and histidine-containing phosphocarrier protein match quantitatively with that predicted by a simple sum of the partial volumes of the amino acids in the polypeptide segment inside the pore when translocation stalls due to the primary charge sequence. Our analysis suggests that the majority of the protein molecules were linear or looped during translocation and that the electrical forces present under physiologically relevant potentials can unfold proteins. Our results show that the nanopore translocation signals are sensitive enough to distinguish the folding state of a protein and distinguish between proteins based on the excluded volume of a local segment of the polypeptide chain that transiently stalls in the nanopore due to the primary sequence of charges.


Protein translocation across nanometer-scale pores is of fundamental importance both in basic science and in biotechnology. Nanopore translocation is a common event in trafficking of proteins between eukaryotic organelles. Emerging technologies have enabled progress in understanding protein translocation and our understanding is still at an early stage1. Particles suspended in solution can be characterized24 by the drop in ionic current they cause during translocation through a small pore. This current drop is caused by the decrease in conductivity resulting from the displacement of a specific volume of electrolyte from the pore. The current drop persists for as long as the particle remains in the pore; this sojourn time depends on particle transport properties and charge as well as nanopore shape and electrical bias. For a protein particle, the current drop signal contains information on the protein’s shape or folding state. If the protein molecule is unfolded, the signal depends on the amino acid sequence through the volume and charge of the polypeptide segment present inside the nanopore. Ion channels or protein pores reconstituted in lipid membranes have been utilized to characterize single polymers, DNA/RNA molecules57, and proteins811. More recently, single nanometer-scale solid-state pores have been fabricated in insulating silicon nitride (SiNx) and silicon dioxide membranes1216 and used to study chain-like DNA molecules1722 and native-state protein molecules.23,24

Recent studies of protein translocation by Han et al.24,25 through larger-diameter synthetic nanopores, 55 nm24 and 24–28 nm25 were performed in salt solutions without denaturing agents and without presenting event distributions. Fologea et al.23 used 16–18 nm diameter pores to study larger proteins BSA and Fibrinogen observing that both gave two clusters of events, though only the larger ΔIb cluster was analyzed.

Protein nanopores have been used to study proteins and peptides 810,26. Most of these results showed multiple peaks in current drop amplitudes. Oukhaled et al.10 observed short and long duration current blockages. They concluded that short blockages are due to the passage of completely unfolded proteins as their frequency increases as the concentration of the denaturing agent increases. The duration of the long passages were attributed to the wait time for protein unfolding.

In this article, we seek to measure single protein molecules at different concentrations of denaturant (urea) as they translocate through voltage-biased silicon nitride nanopores. We aimed to distinguish the native, partially folded, and unfolded states of bovine β-lactoglobulin variant a (βLGa) and to evaluate the ability of a nanopore to determine and/or influence the conformational state of a protein.

βLGa is a lipocalin present in whey, the stability and folding of which has been extensively studied27. Its principle biological role is apparently to provide a source of protein in milk. It may also enhance solubility of fat and fat-soluble nutrients. Interest in βLGa unfolded structures and aggregation has been driven by basic science and by the importance of βLGa to the dairy and food processing industries. Transolcation of lipocalin-ligand adducts through nanochannels provides a potential mechanism for remediation of hydrophobic and amphiphilic targets.

The unfolding of βLGa by urea has been studied by fluorescence28,29 dynamic light scattering (DLS)29 and urea gradient gel electrophoresis30. Urea can be used to change the conformational state of βLGa from folded at 0M to partially folded intermediate at 5M and unfolded at 8M.28,29 Based on DLS diffusion measurements we can infer a mean hydrodynamic volume of 24.4±4.1, 51.0±16.6 and 165±44 nm3 at 0, 5, and 8 M urea respectively, with the errors reflecting dispersion in the measurement29. Such changes are easily within the resolution of nanopore measurements. Since urea can lower the energy barrier to unfolding we expect this could influence the conformational integrity of the protein during translocation.

The main component of a nanopore sensing system was a single nanopore fabricated in a silicon nitride membrane that separated two salt-solution-filled chambers whose only electrical connection was via the electrolyte solution inside the nanopore (Fig. 1A). A pair of Ag/AgCl electrodes was embedded in each chamber solution. When charged protein molecules were added to the cis chamber and the correct polarity voltage was applied to the electrodes, the protein molecules was captured by the electric field near the nanopore, and driven through the nanopore to the trans chamber. The protein molecule interacting with, or translocating through, the nanopore caused a transient current drop (blockage). The current blockage events were recorded with an Axopatch 200B (Molecular Devices) integrated system (10 kHz low pass 4-pole Bessel filter, event-driven mode). At this setting, the nanopore measuring system was tested and calibrated with synthetic current blockages: ideal square pulses of pulse height 100 pA and width from 25 to 300 μsec, generated from a function generator (Agilent 33250A). (See Supplement for calibration details.) The recorded data was analyzed with the same MatLab routines as used for DNA and protein translocation. When the pulse width was between 25 and 100 μs, the pulse height was attenuated, but the time durations remained correct under our data analysis procedure. The preservation of the square pulse shape is the design property of multi-pole low-pass Bessel filters. The current blockage amplitudes (ΔIb as shown in Fig. 1B) presented in this work were corrected with this calibration. Nanopores were fabricated by ion beam sculpting12,15. The cis and trans chambers were cast in PDMS (polydimethylsiloxane)17.

Figure 1
A) Schematic diagram of a nanopore translocation experiment. The three-dimensional structure of βLGa dimer derived from PDB file 2AKQ is shown approximately to scale with typical SiNx nanopores. Positive and negative residues at pH 7 are colored ...

Previous studies with different shaped particles translocation in pores of varying sizes24,31, have shown that the instantaneous amplitude of current blockage ΔIb(t) is approximately proportional to the instantaneous excluded atomic volume Λ(t) of a translocation particle inside the pore and can be written as


Here σ is the solution conductivity, and ψ is the applied voltage to the electrodes. Over the range of 0.5–3.0 M KCl the conductivity of a single nanopore is nearly linearly proportional to the KCl concentration. 18,32, Consequently, we used the measured bulk conductivity of the buffered salt solutions for the nanopore conductivities in this work. These measured values were σ=112, 169, 144, and 110 mS/cm for 1 M KCl, 2 M KCl, 2 M KCl+5 M urea, and 2 M KCl+8 M urea respectively. Urea reduces the magnitude of current blockade values in proportion to its effect on solution conductivity in Eq. 1. As illustrated in Fig. 1A, dm is the diameter and lm is the length of a protein molecule, Dp is the average diameter of a cylindrical nanopore. f(dm/Dp, lm/Heff) is a correction factor that depends primarily on the relative geometry of the particle and the pore, but also includes other possible parameters we have ignored. The physical thickness of the nanopores fabricated by ion beam sculpting was estimated to be 10 to 15 nm 15,33. The calibrated effective thickness, Heff, accounts for any extension of the electric field lines beyond the physical limits of the nanopore. Using Equation (1), the instantaneous excluded volume of a molecule can be estimated from ΔIb(t) with Λ(t)≈(ΔIb(t)* Heff2)/(σ ψ). For a long (lm[dbl greater-than sign]Heff) approximately cylindrical dsDNA molecule, equation (1) can be simplified as ΔIbσ ψADNA/Heff, here ADNA is the mean atomic volume per unit length of a dsDNA molecule and its value can be calculated from the four different nucleic acids.34,35 A value of ADNA =1.8 nm2 per nm is used in this work (see SI for details). We calibrated Heff using ΔIb measured from a linear 2,706 base pair dsDNA (pNEB206, NEB) to allow the excluded atomic volume of a protein in a nanopore can be estimated.

The current blockage duration, td, is taken to be the sojourn time of a protein in a nanopore. The distribution of td reflects the fundamental physics of translocation. If the translocation is dominated by viscous diffusion under a constant bias, the distribution of charged particle sojourn times can be solved analytically. We modelled the sojourn time as the first passage time (fpt) for one-dimensional diffusion of a charged particle in a constant electric field (ψ/Heff) from the entrance to the exit of the nanopore. The electrophoretic drift velocity (v) and the diffusion constant (D) allow derivation (see supplement) of the sojourn time distribution:


Equation (2) is appropriate for translocating objects that are approximately uniformly charged. Equations (1) and (2) have ignored many complex issues like electro-osmotic flow in and near the pore, interactions between a charged molecule and the pore, and variable charge along the length of the translocating molecule.


2.1 βLGa Current Blockage Events and Event Polarity

Translocation event distribution under different conditions

The structure and primary sequence of βLGa appear in Fig. 1A and C. The reported pI of βLGa ranges from 4.8–5.5 and it has charge ~8e at pH 7.27 βLGa was measured in 2 M KCl 10 mM buffer above (pH 7.0) and below (pH 4.6) the pI. After adding 35±15 nM βLGa to the cis chamber at pH 7, we initially observed current blockage events only when the trans chamber electrode was positively biased, consistent with negative βLGa. Transient blockage events were typically detected at an average rate of ~10 s−1. The distribution of event parameters did not change over the range of protein concentrations used in the cis chamber. Fig. 1B shows typical current blockage events recorded in event driven mode (i.e. gaps between events are not recorded); the depth and duration of the events varied significantly. If event recording had previously been conducted for several hours, translocation events with the trans chamber negatively biased could be observed at a greatly reduced rate, indicating βLGa had translocated to the trans chamber.

Each recorded blockage event was characterized by its time duration td and its mean current drop amplitude ΔIb as defined by the levels in Fig. 1B. The joint distributions of recorded events for βLGa in 2 M KCl are shown in Fig. 2A for ψ=120 mV and Fig. 2D for ψ=60 mV. Each dot in the figures represents the td and ΔIb of one blockage event. The joint distribution of ΔIb and td for both biases showed two major clusters of events labelled as cluster 1 for smaller ΔIb and cluster 2 for larger ΔIb. The values of ΔIb were smaller for ψ=60 mV compared to ψ=120 mV. The population of cluster 2 was ~50% of the total events for ψ=60 mV and it was ~20% for ψ=120 mV in this measurement. The sojourn time histograms at the bottom axis show that the dwell time for βLGa at ψ=60 mV was somewhat shorter than those at 120 mV. The distribution of 60 mV events may be limited by the open pore noise.

Figure 2
Event distributions. βLGa in: (A) 2 M KCl with no urea at pH 7 120 mV, (B) 2 M KCl 5 M urea at pH 7 120 mV, (C) 2 M KCl 8 M urea at pH 7 120 mV, (D) 2 M KCl with no urea at pH 7 60 mV inset: Sample events, (E) 2 M KCl with no urea at pH 4.6 –120 ...

Polarity changes when pH<pI

βLGa at pH 4.6 in a clean apparatus produced current blockage events only when the trans chamber electrode was negatively biased, consistent with positively charged βLGa. The results of βLGa at pH 4.6 (ψ =−120 mV) is shown in Fig. 2E. The lowest values of the excluded volume were consistent with the volumes observed for cluster 1a and 1b of the pH 7.0 experiments. The volumes of clusters 2 and 3 are also consistent with the volume observed at pH 7.0 however the translocation times are substantially longer suggesting that the dynamics and mechanism of translocation may be very different at pH 4.6.

Nanopore Diameter

Similar measurements performed for βLGa proteins with different diameter (Dp) nanopores showed that the peak values of ΔIb varied with Dp, suggesting changes in Heff or in the correction factor f in equation (1) (data not shown). In this work, only results from the same nanopore are directly compared.

2.2 Event analysis and βLGa protein in no urea, 5M and 8M urea

Following the measurement shown in Fig. 2A and 2D, the cis chamber solution was flushed out with 2 M KCl+5 M urea at pH 7; βLGa was added and the event distribution from this measurement is shown in Fig. 2B. The same procedure was repeated for 2 M KCl + 8 M urea and the event distribution is shown in Fig. 2C. In summary, these urea concentration measurements (ψ =120 mV) showed that: Without urea (2a), the distribution of ΔIb comprised two major clusters of events. Cluster 1 had a most probable value of ΔIb ~40 pA with td across multiple scales from ~70 to 200 μs. Cluster 2 had a most probable ΔIb ~100 pA with a very broad distribution in ΔIb. The population of cluster 2 was about 20% of the total events in this measurement.

In 5 M urea (2B), the most probable ΔIb of cluster 1 was at ~36 pA and the population of cluster 2 events had decreased to about 10%. A third low-probability cluster was observed for only the 5 M urea data with ΔIb 2–3 times that of cluster 2.

In 8 M urea (2C), the most probable value of ΔIb of cluster 1 was at ~28 pA. Very few cluster 2 events were observed.

The decrease in ΔIb of cluster 1 was consistent with the decrease in solution conductivity σ with increasing urea concentration. Using the same data analysis parameters, no events were detected in a recorded trace of open pore current before βLGa was added, consistent with good discrimination of open pore noise from bona-fide current blockages. However, we observed that adding βLGa to the cis chamber caused an increase in the open pore current fluctuations and generated short lived spike events as shown in Fig. 1B.

2.3 Calibration of nanopore Heff using dsDNA

After the βLGa measurements, we calibrated the same nanopore using 2,706 bp dsDNA in 1M KCl at pH 7. Fig. 2F shows the most probable ΔIb for the dsDNA was ΔIb≈120 pA with td~70 μs. This was consistent with our previous results on dsDNAs measured in nanopores fabricated the same way. Using σ=0.112/(Ωcm), ψ=120 mV, and ADNA =1.8 nm2 35, a value of Heff=20±2 nm was calculated (ΛdsDNA=36 nm3) for this nanopore. Assuming Λ≈(ΔIb* Heff2)/(σ ψ) holds for both DNA and βLGa protein translocations, we converted the ΔIb to an estimated excluded volume Λ for βLGa as labeled on the right axes of Fig. 2. The differences in current drop scales arise from the different solution conductivities (σ) used in Eq. 1.

2.4 The distribution of the ΔIb or Λ of βLGa molecules

The marginal distribution of ΔIb or Λ showed multiple peaks. We fit the histograms of ΔIb with multiple Gaussian peaks, labeled in the Figures as 1a, ,1b,1b, ,2a,2a, ,2b,2b, and and3.3. The fitted peak positions of Λ values are listed in Table 1. The multiple peak fitting worked well to characterize the DNA translocation in two geometries. One peak located at ΔIb=120 pA and td=70 μs represents linear dsDNA translocation. The other peak located at ΔIb=133 pA and td=60 μs represents partially folded dsDNA translocation 19.

Figure 3
HPr (right panels) and βLGa (left panels) measured in the same nanopore. The nanopore had an average diameter of Dp=4±1 nm and the open pore current was I0=3.2 nA in 8 M urea and 2 M KCl.
Table 1
The excluded volumes and sojourn time distribution fit parameters. Excluded volume distributions were fit with enough (2–4) Gaussian components to give flat residuals. Protein sojourn time distributions were fit to both the biased diffusion model ...

For βLGa in 2 M KCl, the most probable excluded volume for the cluster 2 events was ΛβLGa ~20 nm3. The most probable excluded volume for cluster 1 events in 0, 5, and 8 M urea was ΛβLGa~8 nm3. The excluded volume of the cluster 1 events was about 40% of the cluster 2 events.

2.5 Comparison of unfolded βLGa and HPr

Fig. 3 shows results from two different proteins, βLGa and HPr (85 amino acids, −2e at pH 7) 36 measured in the same nanopore at 2 M KCl and 8 M urea. HPr has no cysteine and is completely denatured in 8 M urea. The open pore current was the same (I0=3.2 nA) during these measurements. Only cluster 1 events were observed for both of these proteins in 8 M urea. The most probable current drop values were ΔIb(HPr)=22±3 pA and ΔIb(βLGa)=26±6 pA. The position and width of the peaks in the βLGa current drop distribution found for the 8 M urea matched well between data sets (Fig. 2C and and3B)3B) taken on different days with different nanopores, suggesting that these two different pores have similar Heff. The same value of Heff=20 nm was used to convert ΔIb to Λ (right axes).

The marginal distribution of HPr ΔIb events measured at 8 M urea showed an asymmetric peak that could be decomposed into two Gaussian contributions. The marginal distribution of βLGa ΔIb events measured at 8 M urea was also asymmetric and required 3 Gaussian contributions for a satisfactory fit. Peak 1 of HPr and βLGa gave similar excluded volumes even though βLGa is nearly twice the size of HPr. Fit parameters for Fig. 2 and and33 are summarized in Table 1.

2.6 Multiple time scales for dwell time distributions

Fits of the marginal distribution of sojourn times to Eq. 2 appear as the solid lines through the experimental distributions at the bottom of the panels in figures 2 and and3.3. Fit parameters are summarized in Table 1. For peak 1 of DNA, the parameters for linear translocation through the nanopore were D1≈162 nm2/μs with v1≈14 nm/μs (td≈70μs). For the peak 2 of folded dsDNA 19, D2≈150 nm2/μs with v2≈17 nm/μs.

The distributions of nanopore sojourn times for proteins were substantially more heterogeneous than we observed for the DNA control. Fits of the observed distributions of sojourn times for the proteins required extremely low values for the diffusion constant and electrophoretic velocities. For βLGa, the fitting parameters are on the order of D ~10−1 nm2/μs and v ~10−1 nm/μs (t ≈200μs). Both parameters are 3 orders of magnitude smaller than the bulk values 29 for diffusion constants (~102 nm2/μs) and for drift velocity (~102 nm/μs), indicating that they are either significantly altered in the nanopore or the translocation process is not consistent with the biased diffusion mechanism. Using βLGa bulk values 37 for D and v in Eq. 2 predicts average sojourn times on the order of <1 μs.

The anomalously long sojourn times suggested that barriers to exiting the nanopore exist and that thermal activation may be the limiting step for translocation. Activated sequential-barrier-crossing predicts sequential first-order kinetic steps resulting in a multi-exponential first passage time distribution. Therefore we also fit the sojourn time distributions to such functions. Average exponential decay time parameters (t1 and t2) associated with the long time side of the distribution appear in Table 1. The short time scale rise (td<30μs) in the exponential first passage time distribution was not accurately resolved due to the 10 kHz Bessel filter which reduced the amplitude of the resistive pulses to below the threshold for detection.

Both cluster 1 and cluster 2 in the βLGa event histograms required 2 Gaussian contributions to fit the excluded volumes (or ΔIb) and either two biased diffusion contributions or two exponentials to fit sojourn times (td). The parameters describing the sojourn times for cluster 1 and cluster 2 were similar except for the relative weights in a given urea concentration. The presence of multiple sojourn time scales and the observation of multiple current blockage levels during translocation lead us to conclude that multiple states of βLGa occur during nanopore translocation.

3. Analysis and Discussion

In this section we discuss the nature of folded and unfolded βLGa and how this gives rise to the heterogeneous translocation event distributions shown in Fig. 2 and and3.3. We relate these observations to the sojourn time of the translocating molecule in the nanopore and the instantaneous volume displaced by the molecule during the translocation.

3.1 Excluded volume is not consistent with globular protein translocation

Adding the volume of all 161 βLGa amino acids gives an excluded volume of Λ =22.8 nm3 per βLGa monomer.38,39 This volume is expected to generate at least ΔIb~110 pA in 2 M KCl solution (Fig. 2A). Folded βLGa has a large void calyx that should increase the true excluded volume from this value. Partial or complete unfolding increases a protein’s hydrodynamic radius. One might naïvely expect that nanopore translocation volume would similarly increase, but the opposite trend was observed.

The cluster 1 events for different urea concentrations have similar excluded volumes of ΛβLGa =6.7–8.8 nm3, approximately 33% that of the entire protein. Early studies2 in large pores found that the current blockage was larger for spheres than for cylinders of the same volume. By neglecting this correction (f in Eq. 1) our DNA calibration procedure should overestimate the volume of globular βLGa. Under all conditions βLGa translocation was dominated by events with excluded volumes smaller than that of the monomeric folded protein.

The partial volumes of the 85 amino acids in HPr add to 11.2 nm3, or about half that of βLGa. The difference in blockage depth between HPr and βLGa at 8 M urea was only 10%. The similarity of blockage depth of βLGa and HPr is more like the linear translocation of DNA/RNA. For linear translocation (Fig. 4D), the depth of current blockage is expected to be largely independent of contour length.

Figure 4
A) The net charge at pH 7.0 of βGLa and HPr as predicted by treating the ionization of the individual amino acids as independent as a function of the number of residues translocated.

Figure 4B compares the excluded volume calculated from a 20 nm contour length segment of HPr and βLGa as a function of number of amino acids translocated (0.38 nm/residue). The values predicted by this analysis (ΛHPr =6.75 nm3, ΛβLGa=7.48 nm3) match quantitatively with the values of ΛHPr =6.8±0.1 nm3 and ΛβLGa=7.5±0.2 nm3 observed in Fig. 3 for the smallest-volume contribution to the HPr and βLGa event distributions. The observed differences in excluded volume arise because βLGa has bulkier amino acids and therefore a larger excluded volume for a segment inside the nanopore. The variability of the βLGa translocation volume is also bigger, consistent with observations. This suggests that the cluster 1 translocation events were dominated by protein that was essentially linear inside the pore (Fig. 4D), and the folded or globular protein translocation could account for no more than 20% of the events in 2 M KCl even when no urea is present (Fig. 2A).

We have found that at 5 M urea βLGa forms disulfide-linked dimers29. Cluster 3 in figure 2B is consistent with partially unfolded disulfide-linked dimers. The broad distribution of sojourn times suggests that the translocating dimer is not in the folded state. Crosslinking at 5 M urea is consistent with the observation that disulfide crosslinking does not occur when βLGa is folded.29

3.2 Protein Charge Sequence and Stalling

In the βLGa nanopore translocation measurement, we have observed low displaced volume values and very slow translocation times. One explanation for the observations is that the protein is being unfolded and pulled through as a linear polymer. For linear translocation, the primary sequence of charged amino acids will be more important than that the net charge of the protein.

Figure 4A shows the calculated charge and force of βLGa inside the pore as a function of the number of residues translocated for ψ=120 mV and Heff=20 nm. This analysis predicted several locations where the net force is zero, suggesting stalling points or locations of metastability. Two positive peaks in the curve suggest that at two points during the linear translocation the electrophoretic force opposes translocation. This can be more readily visualized in the translocation potential plot of Fig. 4C.

The presence of multiple stall points provides an explanation for the anomalous td results when fitting to Eq. 2; Eq. 2 is only valid in the absence of energetic barriers. If the barriers are large enough, Kramers reaction rate theory40 predicts that the distribution of sojourn times should be multi-exponential according to the number of barriers present. One or two exponential contributions were adequate in all cases, consistent with the presence of no more than two activated barrier crossings during translocation.

Since the barrier at ψ=60 mV is smaller comparing to ψ=120 mV, a shorter td is predicted by the thermally activated barrier model. The peak values of td for βLGa appear to decrease as the voltage was decreased from ψ=120 mV (Fig. 2A) to ψ=60 mV (Fig. 2D) instead of increase, as predicted by voltage driven translocation. This also supports the thermally activated barrier model of the sojourn time distribution.

The presence of positive charges at the N terminus and negative charges at the C terminus suggests that the insertion of βLGa into the nanopore will favor the C-terminus. The dipole of βLGa has been experimentally measured to be ~700 Debye 41. Using PDB entry 2akq, a computational estimate for the monomeric dipole is 795 Debye 42 and is approximately oriented from the C-terminus to the N-terminus of the folded protein. This also suggests that dipolar orientation of the intact monomer would favor C-terminal-first insertion. An expansion of this analysis of the electric field orientation of βLGa appears in the supplemental materials.

Fig. 4C shows that the barriers during translocation are maximized for thin pores and become smoother for thicker pores. The pattern of translocation barriers is highly dependent on the sequence of charged residues. These observations are in sharp contrast to DNA translocation. When calculating the driving force for DNA translocation, the thickness of the pore cancels out because of the uniform charge density. There is essentially no DNA sequence effect on the translocation driving force. By contrast, our analysis suggests that the translocation of proteins is highly sequence dependent. The distribution of translocation times depends strongly on protein sequence, applied voltage, as well as Heff.

We have neglected two potentially important contributions to the translocation potential (Fig 4C): entropy and folding. βLGa is unfolded in 8M urea. Random coil polymers are expected to resist confinement. The protein must give up conformational entropy to enter the confined region of the nanopore. If the unfolded state is behaving as a random coil, then entropic considerations would suggest a barrier centered at ~ residue 81 where the chain entropy would be most-reduced by the confinement of a region in the nanopore.

For a folded protein, both partial unfolding at the cis chamber side and refolding at the trans chamber side could create additional energetic contributions to translocation.

3.3 The Electrical force exerted on a βLGa may partially unfold it

Our observations are consistent with bulk measurements that showed that βLGa protein is completely denatured in 8M urea and partially denatured in 5M urea. 2830 Looped translocation geometries for the unfolded protein (8M urea) suggest some persistence of structure. In 5M urea, oxidative dimerization of βLGa (peak 3) starts as soon as the sample solution was made. The presence of unfolded translocation at 0M urea suggests electric-field induced unfolding of βLGa.

Research on force unfolding of proteins has shown some proteins rupture at elongation forces larger than ~5 pN.43,44 The electrical field strength in the nanopores is E ≈ ψ/Heff = 6×104V/cm; this electric field is at least two orders of magnitude larger than it is in a regular electrophoresis experiment. The force per charge estimated is eE= 1.0 pN/e. βLGa has 27 negative and 18 positive residues at pH 7.0. When a βLGa molecule is entering the nanopore, opposite charges in the protein will be driven in opposite directions by the electric field (see Fig. 4A.) This electric force is much larger than the strength of the hydrogen bonds that hold a protein in a folded shape, thus some of the hydrogen bonds in a βLGa could be broken at the entrance of a nanopore, at least partially denaturing it. The increased population of cluster 2 events (folded or looped βLGa) at lower voltage, ~50% at ψ =60 mV (~20% at ψ =120 mV), supports our force unfolding hypothesis above since the weaker electrical field strength provides less driving force to unfold a βLGa protein.

Figure 4C predicts that a thicker nanopore would manifest smaller electrostatic energy barriers to linear translocation and smaller td as a result. A thicker nanopore would have more amino acids present at the stall point. For a nanopore of a given thickness, this can also predict the effect of the protein being less than fully extended near the stall points. That is, the effective number of amino acids in the pore would larger. Fig. 4C shows that, at a stall point, the energy is higher if more amino acids are in the pore. Therefore increasing the extension would be energetically favorable near the stall points. This suggests that proteins at a stall point will tend to elongate to their most extended form.

This argument neglects entropy. Elongation in the pore would decrease the entropy of the segment inside the pore, but increase the entropy of segments outside the pore. The net entropic effect will depend, to first approximation, on the translocation position through its effect on the number of residues in the different regions of the nanopore apparatus. Thus different stall points may have different entropic contributions to the free energy.

3.4 pH effects on translocation

Several changes in the 2D event distribution were observed at pH 4.6 as compared to pH 7.0. The principle effect of reducing the pH is the partial protonation of carboxylic acid amino acid side chains. At pH 4.6 the acidic residues are approximately 50% protonated. This substantially changes the distribution of stall points and the dipole moment of the protein. The polarity of translocation events is observed to be opposite of those at pH 7.0 in accordance with expectations. Because there are fewer total charges on the protein the driving force for electrostatic unfolding is expected to be reduced. Overall the distribution of cluster 1 translocation events is peaked earlier in td, consistent with smaller electrostatic barriers to translocation. The distribution of events appears to favor larger excluded volumes, suggesting a greater contribution from loped translocation structure. This is consistent with a reduced driving force for electrostatic extension of the protein. The increase in translocation time for cluster 2 and 3 events may be due to the specific arrangement of charges in the loops of translocating protein or due to other effects. For example, since the protonation of the carboxylic acid is near the midpoint of the equilibrium, any electric-field effect on the pKa of the side groups could substantially alter the electrostatics of the translocating protein. Resolution of these effects awaits future investigation.

3.5 Evaluation of bumping/collision hypothesis

The shallow blockage shown by the proteins is consistent with ~30% of the protein volume contributing to the blocked current. One hypothesis is that a globular βLGa molecule only partially entered the pore and then went back to the cis chamber. The cluster at the lowest current blockage values, therefore, might be expected to be due to collisions. In DNA experiments, we observed short spikes with small ΔIb and td, they usually appear at the left bottom corner in an event distribution plot. We attributed this type of event to long DNA molecules (~μm) that were captured in the middle of the long chain rather than an end. The DNA must then be bent at the entrance by the electric field. If the bending failed, the DNA would be bounced back to the cis chamber.

In this work, βLGa and HPr proteins were smaller than the size of nanopores used, and short events did not appear in the histogram. For the cluster 1 events to be due to globular protein that partially entered a pore, persisted for the long sojourn times, and then return to the cis chamber, there must be some trap for the protein in the vicinity of the nanopore. Otherwise, diffusion would normally carry the protein away far more rapidly than the time scales we observed in the sojourn time distributions. Given the net charge of a βLGa protein (−9e), and the large electrical field strength in the nanopore, a trap deep enough to localize globular βLGa protein for the times observed is unlikely.

3.6 Protein Adsorption

An alternative interpretation of the presence of the multiple timescales for translocation is that proteins adsorb to the nanopore surface causing a change in current. The unusually low translocation volume would then arise because, for the vast majority of the translocation time, the protein is only measured when in the surface layer of the nanopore. This hypothesis cannot easily account for the similarity of the βGLa translocation volumes and the HPr translocation volumes. Though the calculation of the net effect on the current of a protein adsorbed to the nanopore is not well understood, one would expect the effect to be approximately in proportion to either the volume displaced or the surface area contacted on the nanopore. In either of these cases we expect the difference between HPr and βGLa to be larger than we observed. The translocation time distribution would be dominated by the desorption reaction. To explain the observed translocation time distributions multiple adsorption modes would be required with different desorption rates. The protein is not expected to adsorb to the SiNx surface in the presence of 8 M urea. Protein adsorption to the surface of the nanopore/membrane may be involved in the long-time changes (~hours) of the nanopore electrical properties, however it does not appear to be a likely explanation for the observed translocation signals.

3.7 Translocation event heterogeneity and protein sequence

Linear translocation of dsDNA was much less heterogeneous than that of proteins. In this section we discuss how the increase in translocation event heterogeneity can be related to the details of the protein sequence.

Primary sequence effects

The individual amino acids vary substantially in their partial molecular volumes. This difference was quantitatively observed in comparing unfolded βLGa and HPr. This suggests that nanopores could, in principle, distinguish proteins based on the primary sequence variability of the excluded volume and the location of stall points.


Highly charged loops present on the surface of βLGa were identified based on the folded structure and the topological constraints imposed by disulfide bonds. These putative structures are a possible explanation for the range of excluded volumes observed. Two mechanisms could make βLGa be looped during translocation. 1) Unreduced native disulfide bridges will enforce the presence of a loop. 2) Nanopores could preferentially capture negatively charged turns/clusters on the surface of βLGa as shown by the red circles in Fig. 1C. As with translocation of linearized protein, the sequence details of the loop inserted into the nanopore will dictate the dynamic nature of the signal measured. The translocation potentials and excluded volume profiles for several likely loops appear in Fig. 5. Fig. 5A shows the capture of the loop at residue 133. Stall points are predicted with ΛβLGa=14 and 8 nm3. The measured current (and therefore the excluded volume) of such an event would depend on the time-weighted mean of the volumes Λ1 and Λ2 at the two stall points:

Figure 5
(A, B) Possible looped translocation geometries for the unfolded protein with no disulfide bonds intact. These loops are pre-formed in the native structure as shown in Fig. 1. (A) The E51 loop insertion has two stall locations with different excluded ...

P(ti) is the distribution of stall times at each metastable position in the translocation. A close examination of individual events shows that many of them have multiple levels (Fig. 1B, 2D, 2E, ,3).3). Since the preent analysis did not attempt to distinguish substeps from noise, only the average excluded volume was reported. Fig. 5B shows the capture of a loop at residue 51. Both stall points give similar translocation volumes. Events like these would be expected to give similar volumes to the completely linear translocation. Fig. 5C shows the capture of a cluster at residue 113. The native disulfide bridges are compatible with this translocation and two stall points are predicted with ΛβLGa=15 and 8 nm3 similar to Fig. 5A. These events are consistent with the larger excluded volume contributions to cluster 1 in the βLGa event distributions. The upper edge of the cluster 1 ΛβLGa≈15 nm3 and the lower edge is ΛβLGa≈7.5. The time at each stall point will produce a range of observed average volumes between these two limits according to Eq. 3, consistent with our observed distribution. Fig. 5D shows the capture of a charged cluster of amino acids at residue 51 and the resulting loops that are present due to intact native disulfide bonds. Fig. 5E shows the capture of a loop at residue 63 that is favored by native disulfide bonds. These geometries are examples of how partially folded translocation events can give rise to the displaced volumes of 17–20 nm3 observed in cluster 2 of the 0M and 5 M urea experiments. Fig. 5F shows disulfide linked dimers will have more complicated translocations patterns with peak volumes contributions from 4–6 strands, consistent with the events in cluster 3 in Fig. 2D. Dimers formed at 5 M urea are likely to be one of the early structures formed when βLGa is incubated under amyloidogenic conditions though formation of higher order oxidative aggregates is relatively unimportant.29

3.8 Correlation between ΔIb and td

The various translocation events illustrated in Fig. 5 not only provide an explanation for the heterogeneity, they also provide a mechanism for correlation between ΔIb and the charge sequence of a protein which determines td. The presence of multiple translocation shapes provides multiple clusters of events providing an overall correlation between ΔIb and td. However, the presence of multiple stall points with different excluded volumes suggests a mechanism of correlation between ΔIb and td within a single cluster through Eq. 3. As the time spent in the deeper well increases, the mean will become increasingly weighted by the volume at that position.

When a non-flexible particle translocates, the current drop amplitude and sojourn time are independent random variables. There should be no correlation between ΔIb and td. Correlation between ΔIb and td requires that there be multiple species with different electrophoretic mobilities and excluded volumes. For example, a shorter td represents partially folded DNA with larger ΔIb (Fig. 2F). In the present case correlation can only arise due to conformation changes occurring in the protein either prior to or during the translocation process. These changes must affect both the electrophoretic mobility and the excluded volume of the protein.

We discuss two possible explanations for the presence of multiple current blockage levels in a single transient. One possibility is the simultaneous translocation of multiple molecules that reside in the pore for different lengths of time. In this case there could be up to three levels observed corresponding to the excluded volume of each molecule alone and the molecules together. The other possibility is that a single molecule could show multiple current blockage values. We find the second explanation more compelling since the presence of steps in the current blockage profile depended on the character of the molecule being translocated and not on the concentration of protein. Furthermore, the translocation of a protein as an unfolded chain leads to multiple translocation current levels in a single event.

3.9 Comparison with polynucleotide translocation

The average current blockage value, ΔIb, is a time-averaged quantity. This average depends on the amount of time the translocating molecule spends at each different value of excluded volume during the translocation. A significant difference between polynucleotide translocation and polypeptide translocation is the degree of variability inherent in the polymer. The charge variability of proteins can result in a variable translocation driving force as a function of displacement -- an effect not important to polynucleotide translocation. The excluded volume of amino acids varies by a factor of 3.8 whereas that of dsDNA (or ssDNA) varies by less than 2% (or 10%). This variability in volume maps directly to variability in ΔIb. Furthermore, proteins are structurally more complex than DNA. The persistence length of DNA is much longer than that of proteins suggesting that proteins are more likely to loop. Induced and persistent tertiary and secondary structure in proteins will increase the heterogeneity of protein translocation signals in much the same way as is observed for polynucleotides. However, for proteins, the importance and variety of such contributions is expected, and is observed to be greatly increased.

3.10 Single-molecule protein unfolding

The electrical detection of single protein molecules in various states of folding invites comparison to other single molecule approaches to folding such as fluorescence45,46 and force measurements47, 48. The different physical principles that underlie the different methods make them not directly comparable. The use of fluorescence energy transfer, for example, allows probing one or more folding coordinates.4952 AFM studies of beta-sheet proteins have suggested that the direction of force applied can influence the rupture threshold of the protein48. The tendency for proteins with (large) net dipoles to orient in an electric field (see Supplement) would provide an optimal geometry for electric-field induced unfolding. The strong electric field in the nanopore and distribution of positive and negative charges along the sequence of the protein provides the driving force for unfolding.

One of the challenges of studying protein folding on the single molecule level remains being able to observe both the folded and unfolded states under a given set of experimental conditions. A significant benefit for protein folding studies using nanopores is the electrical signal that is synchronized with the translocation and thus unfolding process. Translocation of unfolded proteins by nanopores provides a Maxwell’s Daemon-like synchronization signal that could be very useful for incorporation into fluorescence-electrical hybrid experiments.

The present study suggests that nanopore approaches to protein folding may have some advantages that are complementary to existing methods. The distribution of the current drop appears to contain information about the stable loops that are present in partially folded states. Sampling of theses loops is at the electrostatic stall points. The nanopore experiment places structurally-disrupting forces onto the protein according to its distributions of charges rather than at the location of a tether. This is either a benefit or a limitation depending on one’s perspective

During translocation, the nanopore appears to perturb the distribution of protein conformation states in favor of extended states. The similarity of this to biological nanopore protein translocation may be important. However, the natural electrochemical potential across biological membranes often can produce electric fields comparable to those expected in synthetic nanopores. The commonplace translocation of unfolded proteins through natural nanopores suggests that folding following nanopore translocation may be a particularly physiologically relevant approach to single molecule protein unfolding/refolding studies. All single molecule approaches to single molecule protein folding require artificial introduction of an additional driving force to manipulate the distribution of folded and unfolded states. This can be any combination of solution conditions including chemical denaturants and/or the application of an external mechanical force. It is not well understood if these driving forces leave intact the free energy landscape upon which the folding reaction occurs. The nanopore accomplishes this with an electrostatic potential that is comparable to those present across biological membranes.


We used SiNx nanopores to examine the differences between folded, partially unfolded, and unfolded single protein molecules. βLGa translocations were heterogeneous showing multiple time scales and multiple translocation current blockages. The calibrated excluded volume was smaller in most cases than that expected for globular βLGa translocations. The sojourn times for βLGa translocations in all cases were 2–3 orders of magnitude longer than expected in folded globular states. This suggests that the general understanding of nanopore translocation that has been formulated based on polynucleotides needs to be modified for proteins.

Our analysis suggests that the events we measure are consistent with linear translocation and looped translocation of proteins even under folded conditions. Evaluating the potential as a function of linear translocation predicts that the protein will stall at different sequence positions during a translocation. This arises because of the heterogeneity of charge along the amino acid sequence, an effect not present in polynucleotides. Stalling at different locations can explain the long sojourn times, the heterogeneity in the current blockage histogram, and correlation between sojourn times and excluded volumes.

The results in this work have demonstrated that a solid-state nanopore sensor can be used to evaluate the folding state or shape of a protein by measuring its excluded volume. The excluded volume of the amino acids present at the stall points can vary enough to distinguish βLGa and HPr. Our experiments and analysis open the way to study the shape and sequence variability of single protein molecules in solid-state nanopores. Since translocation appears to entirely disrupt the calyx, further investigation is required to allow intact nanopore transport of βLGa-ligand complexes for nanobioremediation applications.

Supplementary Material



We thank Bradley Ledden and Ryan Rollings for nanopore preparation, Professor Jene A. Golovchenko for assisting with FIB hole fabrication, Professor Jeremy S. Lee for the HPr protein. This work is supported by NIH R21HG003290 and NIH R01GM071684.


Supporting Information Available.

The contents of Supporting Information include the following: Details of the protein and nanopore Materials used; Methods used to extract the event parameters, calibrate the size of the nanopore, correct for the effects of Bessel-filtering, and calculate translocation profiles; detailed analysis and discussion of dipolar orientation effects in nanopores; and a derivation of the first passage time distribution for biased diffusion to a sink. This information is available free of charge via the Internet at


1. Wickner W, Schekman R. Science. 2005;310:1452 – 1456. [PubMed]
2. Gregg EC, Steidley KD. Biophys J. 1965;5:393–405. [PubMed]
3. DeBlois RW, Bean CP. Rev Sci Instrum. 1970;41:909–916.
4. Bezrukov SM. J Membr Biol. 2000;174:1–13. [PubMed]
5. Bezrukov SM, Vodyanoy I, Parasegian VA. Nature. 1994;370:279–281. [PubMed]
6. Kasianowicz JJ, Brandin E, Branton D, Deamer DW. Proc Natl Acad Sci USA. 1996;93:13770–13773. [PubMed]
7. Meller A, Nivon L, Brandin E, Golovchenko J, Branton D. Proc Nat Acad Sci USA. 2000;97:1079–1084. [PubMed]
8. Stefureac R, Waldner L, Howard P, Lee JS. small. 2004
9. Movileanu L, Schmittschmitt JP, Scholtz JM, Bayley H. Biophys J. 2005;89:1030–1045. [PubMed]
10. Oukhaled G, Mathe J, Biance A-L, Bacri L, Betton J-M, Lairez D, Pelta J, Auvray L. Phys Rev Lett. 2007:98. [PubMed]
11. Robertson JWF, Rodrigues CG, Stanford VM, Rubinson LA, Krasilinikov OV, Kasianowicz JJ. Proc Nat Acad Sci USA. 2007;104:8207–8211. [PubMed]
12. Li J, Stein D, McMullan C, Branton D, Aziz MJ, Golovchenko JA. Nature. 2001;412:166–169. [PubMed]
13. Storm AJ, Chen JH, Ling XS, Zandbergen HW, Dekker C. Nat Mater. 2003;2:537–540. [PubMed]
14. Heng JB, Ho C, Kim T, Timp R, Aksimentiev A, Grinkova YV, Sligar S, Schulten K, Timp G. Biophys J. 2004;87:2905–11. [PubMed]
15. Cai Q, Ledden B, Krueger E, Golovchenko JA, Li J. J Appl Phys. 2006:100. [PMC free article] [PubMed]
16. Dekker C. Nat Nanotechnol. 2007;2:209–215. [PubMed]
17. Fologea D, Gershow M, Ledden B, McNabb DS, Golovchenko JA, Li J. Nano Lett. 2005;5:1905–1909. [PMC free article] [PubMed]
18. Fologea D, Uplinger J, Thomas B, McNabb DS, Li J. Nano Lett. 2005;5:1734 – 1737. [PMC free article] [PubMed]
19. Li J, Gershow M, Stein D, Brandin E, Golovchenko JA. Nat Mater. 2003;2:611–615. [PubMed]
20. Storm AJ, Chen JH, Zandbergen HW, Dekker C. Phys Rev E. 2005:71. [PubMed]
21. Storm AJ, Storm C, Chen J, Zandbergen H, Joanny JF, Dekker C. Nano Lett. 2005;5:1193–1197. [PubMed]
22. Gershow M, Golovchenko JA. Nat Nanotechnol. 2007;2:775–779. [PMC free article] [PubMed]
23. Fologea D, Ledden B, McNabb DS, Li J. Appl Phys Lett. 2007:91. [PMC free article] [PubMed]
24. Han A, Schurmann G, Mondin G, Bitterli RA, Hegelbach NG, de Rooij NF, Staufer U. Appl Phys Lett. 2006;88:093901–093903.
25. Han A, Creus M, Schurmann G, Linder V, Ward TR, Rooij NFd, Staufer U. Anal Chem. 2008;80:4651–4658. [PubMed]
26. Sutherland TC, Long YT, Stefureac RI, Bediako-Amoa I, Kraatz HB, Lee JS. Nano Lett. 2004;4:1273–1277.
27. Sawyer L, Kontopidis G. Biochim Biophys Acta. 2000;1482:136–148. [PubMed]
28. Yagi M, Sakurai K, Kalidas C, Batt CA, Goto Y. J Biol Chem. 2003;278:47009–47015. [PubMed]
29. Giurleo JT, He X, Talaga DS. J Mol Bio. 2008;381:1332–1348. [PubMed]
30. Beringhelli T, Eberini I, Galliano M, Pedoto A, Perduca M, Sportiello A, Fontana E, Monaco HL, Gianazza E. Biochemistry. 2002;41:15415–15422. [PubMed]
31. Henriquez RR, Ito T, Sun L, Crooks RM. Analyst. 2004;129:478–482. [PubMed]
32. Smeets RM, Keyser UF, Krapf D, Wu M-Y, Nynke HD, Dekker C. Nano Lett. 2006;6:89–95. [PubMed]
33. King GM, Golovchenko JA. Phys Rev Lett. 2005:95. [PubMed]
34. Zwolak M, Ventra MD. Rev Mod Phys. 2008:80.
35. Nadassy K, Tomás-Oliveira I, Alberts I, Janin J, Wodak SJ. Nucleic Acids Res. 2001;29:3362–3376. [PMC free article] [PubMed]
36. Jia Z, Quail JW, Waygood EB, Delbaeresn LTJ. J Biol Chem. 1993;268:22490–22501. [PubMed]
37. Guzey D, McClements DJ. Food Hydrocolloids. 2006;20:124–131.
38. Perkins SJ. Eur J Biochem. 1986;157:169–180. [PubMed]
39. Zamyatnin AA. Prog Biophys Mol Biol. 1972;24:109–123. [PubMed]
40. Kramers HA. Physica (Utrecht) 1940;7:284–304.
41. Ferry JD, Oncley JL. J Am Chem Soc. 1941;63:272– 278.
42. Felder CE, Prilusky J, Silman I, Sussman JL. Nucleic Acids Res. 2007;35:W512–W521. [PMC free article] [PubMed]
43. Lellermayer MSZ, Smith SB, Granzer HL, Bustamante C. Science. 1997;276:1112. [PubMed]
44. Bechtluft P, Leeuwen RGHv, Tyeman M, Tomkiewicz D, Nouwen N, Tepper HL, Driessen AJM, Tans SJ. Science. 2007;318:1458–1461. [PubMed]
45. Michalet X, Weiss S, Jger M. Chem Rev. 2006;106:1785–813. [PMC free article] [PubMed]
46. Schuler B. ChemPhysChem. 2005;6:1206–1220. [PubMed]
47. Borgia A, Williams P, Clarke J. Annu Rev Biochem. 2008;77:101–25. [PubMed]
48. Brockwell DJ, Paci E, Zinober RC, Beddard GS, Olmsted PD, Smith DA, Perham RN, Radford SE. Nat Struct Biol. 2003;10:731–737. [PubMed]
49. Deniz A, Laurence T, Dahan M, Chemla D, Schultz P, Weiss S. Annu Rev Phys Chem. 2001;52:233–53. [PubMed]
50. Jia Y, Talaga DS, Lau W, Lu H, DeGrado W, Hochstrasser RM. Chemical Physics. 1999;247:69–83.
51. Schuler B, Eaton WA. Curr Opin Struct Biol. 2008;18:16–26. [PMC free article] [PubMed]
52. Talaga DS, Lau W, Roder H, Tang J, Jia Y, DeGrado WF, Hochstrasser RM. Proc Nat Acad Sci USA. 2000;97:13021–13026. [PubMed]