Modifying the pH sensitivity of OmpG nanopore for improved detection at acidic pH

Abstract

The outer membrane protein G (OmpG) nanopore is a monomeric β-barrel channel consisting of seven flexible extracellular loops. Its most flexible loop, loop 6, can be used to host high-affinity binding ligands for the capture of protein analytes, which induces characteristic current patterns for protein identification. At acidic pH, the ability of OmpG to detect protein analytes is hampered by its tendency toward the closed state, which renders the nanopore unable to reveal current signal changes induced by bound analytes. In this work, critical residues that control the pH-dependent gating of loop 6 were identified, and an OmpG nanopore that can stay predominantly open at a broad range of pHs was created by mutating these pH-sensitive residues. A short single-stranded DNA was chemically tethered to the pH-insensitive OmpG to demonstrate the utility of the OmpG nanopore for sensing complementary DNA and a DNA binding protein at an acidic pH.

Significance

Nanopore-based sensing is emerging as a powerful technique for biomolecular analysis. Although quite robust for nucleic acid sequencing in genomics research, there are still limitations in fully exploiting nanopore tools for proteomics research. Here, we expand the utility of a nanopore based on outer membrane protein G, one of the rare nanopores that can distinguish native proteins homologs or isoforms from complex biological mixtures such as cell lysate or serum. We identified and then eliminated specific residues contributing to its limited resolution at low pH, thereby resulting in an improved nanopore platform upon which to further develop specific and sensitive protein sensors in a broader range of operating conditions.

Introduction

Technological advancement in genomics and its sub-fields have aided in our study of the biomolecular complexities within the cell. However, most cellular processes are controlled by the dynamic variations of proteins in the cell. Proteomics not only reveals the identity and abundance of cellular proteins, whose expression levels may not be directly correlated to mRNA transcripts, but also allows us to map the specific status of the protein under distinct conditions (1). For example, proteomics can reveal proteins’ post-translational modifications and the presence of isoforms due to alternative splicing or alternative initiation and can show whether proteins form complexes with other proteins or ligands. Such molecular information constitutes the foundation of our growing understanding of cellular function and disease.

Currently, proteomics research relies predominantly on methods based on mass spectrometry (MS) that continue to make significant technological advances (2, 3, 4). Major challenges of protein identification by MS are cellular heterogeneity of proteins, leading to signal averaging, and low abundance of samples (5,6). To address these challenges, orthogonal, empirical and computational methods are currently being developed to improve single-cell proteomics and single-molecule protein analysis (7, 8, 9). Specifically in the field of nanopore technology (10,11), which exploits the current readout of a single nano-sized pore, recent works have made remarkable progress in protein sequencing (12,13), identification of unfolded or linearized polypeptides (12,14, 15, 16, 17), probing the post-translational modification state (17, 18, 19), and analyzing specific “fingerprints” of folded proteins in biological or solid-state nanopores (20, 21, 22, 23). However, unlike linearized polypeptides, analysis of folded proteins within a nanopore is limited by the size of the nanopore lumen. Thus, synthetic or biological pores in different diameters have been explored to adapt to protein analytes of various sizes (24, 25, 26, 27, 28).

One protein nanopore of interest, outer membrane protein G (OmpG), has recently been used to analyze natively folded proteins (29, 30, 31, 32). Unlike conventional nanopores that constantly stay open and exhibit stable current in the analyte free state, OmpG is an inherent gating pore, rapidly transitioning between open and closed states. The dynamic movement of its flexible loops flipping in and out of the OmpG lumen, particularly the longest loop, loop 6 (L6), is the cause for ∼90% of this gating activity (33, 34, 35). Previous engineering on OmpG focused on either restricting the movement of L6 or shortening L6 to reduce the gating for sensing purposes (34,36), while we untraditionally exploit this dynamic movement of L6 for highly selective detection of protein homologues in a complex mixture (29,32). Many factors, such as ligand linker length, ionic strength, pH, and voltage, can alter the sensitivity and specificity of the OmpG nanopore (19,37). However, analytes that are distinguishable at pH ≥ 6 are no longer resolvable in acidic conditions (19). This loss in detection sensitivity reduces the utility of OmpG as a robust tool for proteomics research.

In this study, we found that substituting L6 charged residues with neutral ones effectively created a stably open OmpG nanopore at a broader pH range. On the basis of this finding, we constructed an OmpG pore with a tethered single-stranded DNA (ssDNA) and tested its ability to detect complementary ssDNA or nuclease analytes at pH 5 and observed a notable improvement in sensitivity and reliability from the wild-type nanopore.

Materials and methods

Materials

All reagents were purchased from Fisher Scientific unless specified. Thiolated ssDNA (5′-TTGGTCATGATACTGCTGATTGC-3′) was purchased from Integrated DNA Technologies. Tris (2-carboxyethyl) phosphine (TCEP) and octyl glucoside were purchased from Gold Biotechnology. 1,2-Diphytanoyl-sn-glycerol-3-phosphocholine lipid was purchased from Avanti Polar Lipids. Nuclease S1 was purchased from Worthington Biochemicals.

Cloning, expression, and purification of OmpG proteins

Mutations were introduced into pT7-OmpG construct by mutagenesis polymerase chain reaction using a previously described protocol (32,38). The primers used are listed in Table S1. The desired polymerase chain reaction product was DpnI, digested and transformed into chemically competent Escherichia coli DH5α cells with ampicillin selection. The plasmid was purified from miniprep and the mutations verified by DNA sequencing.

All OmpG proteins were expressed and purified according to a previously described protocol (19,33). Briefly, a 250-mL LB BL21 (DE3) culture transformed with the desired plasmid was induced at 0.6–0.8 OD₆₀₀ with 0.5 mM β-D-1-thiogalactopyranoside at 37°C for 3 h, shaken at 200 rpm. The induced cells were pelleted at 20,000g, resuspended in 50 mM Tris-HCl (pH 8.0), and frozen at −20°C. Later, the cell pellets were thawed briefly at 37°C and sonicated on ice. The lysate was centrifuged at 20,000g to pellet the insoluble inclusion body. The inclusion body was solubilized in 50 mM Tris-HCl (pH 8.0) and 8 M urea and centrifuged, and the cleared solution was applied to a Q Sepharose anion exchange column. The proteins were washed with 50 mM Tris-HCl (pH 8.0), 75 mM NaCl, and 7 M urea and later eluted with elution buffer: 50 mM Tris-HCl (pH 8.0), 200 mM NaCl, 7 M urea. For cysteine-containing constructs, a final concentration of 3.0 mM TCEP was added to all buffers.

ssDNA labeling of OmpG-C224 proteins

The labeling method used was as previously described with modification (39). The 3′ end C3 thiolated ssDNA ligand was dissolved in TE buffer (10 mM Tris-HCl [pH 8], 0.1 mM EDTA) to a final concentration of 5 mM. An equal volume of 200 mM TCEP (pH to 8 with NaOH) was added to the thiolated ssDNA and incubated for 2 h at room temperature to deprotect the thiol group. Thereafter, the ssDNA in TCEP was stored at −20°C until ready to activate with 2-dipyridyl disulfide (2-PDS). The excess TCEP was removed by desalting with fresh TE buffer using a 3 kDa cutoff centricon. The ssDNA was activated by incubating with 50 mM 2-PDS at a 1:1 ratio (v/v) for 2 h at room temperature. The excess 2-PDS was removed by mixing equal volume of diethyl ether into the DNA/2-PDS mixture. The mixture was centrifuged at 10,000g for 30 s, and the top diethyl ether layer was discarded. The diethyl ether extraction procedure was repeated nine times. The residual diethyl ether was left to evaporate for 12–16 h in a chemical fume hood. The 2-PDS DNA molar concentration was determined by Nanodrop and then stored at −20°C until ready to use for labeling OmpG-C224 proteins.

After the OmpG-C224 proteins were purified, they were desalted into 50 mM Tris-HCl (pH 8.0), 200 mM NaCl, and 7 M urea (no TCEP). The OmpG-C224 protein was mixed in a 1:1 mole ratio with activated 2-PDS DNA at room temperature for 2 h. The reaction mixture was run on a non-reducing 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel to separate labeled and unlabeled OmpG. The labeled protein band, which runs more slowly than the unconjugated protein, was excised from the gel and crushed with a pestle. The protein was eluted twice by incubating the crushed gel in 3 times the gel volume of non-reducing elution buffer (50 mM Tris-HCl [pH 8.0], 200 mM NaCl, 7 M urea) with gentle agitation at 25°C for 2–6 h. The crushed gel mixture was pelleted at 20,000g for 15 min at room temperature. The eluted protein solution was pooled and subjected to the refolding process.

Refolding of OmpG proteins

Approximately 35 μM denatured OmpG proteins were diluted in a 3:2 buffer-to-protein (v/v) ratio with refolding buffer (50 mM Tris-HCl [pH 9.0], 3.25% w/v octylglucoside). Samples were then incubated in a 37°C bath for 2–3 days. The refolding and ssDNA labeling efficiencies were determined by mobility shift on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (Figs. S1 and S8). The refolded proteins were flash-frozen in 20% glycerol (final concentration) for long-term storage at −80°C.

Single-channel recording of OmpG proteins

Single-channel recording of OmpG was performed as previously reported (38). Current signals were filtered using a Bessel filter at 2 kHz and acquired at a sampling rate of 100 μs after digitization with a Digidata 1320A/D board (Axon Instruments). Buffers used for the various pH conditions were 1 M KCl with either 20 mM sodium acetate (pH 5) or 20 mM sodium phosphate (pH 6) or 20 mM Tris-HCl (pH 7). Specifically, for detection of nuclease S1, buffer containing 1 M KCl, 5 mM CaCl₂, and 20 mM sodium acetate (pH 5) was used. Once a stable bilayer was formed, OmpG proteins (1–5 nM) were added into the cis (grounded) chamber and a high voltage (≥±200 mV) was used to induce pore insertion. After a single pore inserted into the membrane, the voltage was lowered to ±50 mV for recording. To test various analyte binding, either complementary ssDNA 5′-GCAATCAGCAGTATCATGACCAA-3′ (1 μM) or nuclease S1 (0.5–1.0 U/μL) was added into the chamber in which the OmpG loops were located. The orientation of the OmpG nanopore was determined by its gating behavior at both positive and negative applied voltage (Fig. S2) (40). At least five independent pores were examined for each construct in the absence or presence of target analytes.

Single-channel gating analysis

To characterize the OmpG ionic recordings in the absence or presence of a target analyte, we generated all-amplitude histograms from at least 30 s of the raw trace in Clampfit 10.7 (Molecular Devices). The all-amplitude histograms revealed the distributions of the current states of OmpG for the various OmpG proteins. Using single-channel search in Clampfit 10.7, open probability (P_o), dwell time of the open state (τ_open), and dwell time of the closed state (τ_closed) were calculated as previously described (31,38).

Results and discussion

OmpG nanopore with a positively charged L6 is pH insensitive but responsive to voltage polarity

The equilibrium between the open and closed states of OmpG strongly depends on the pH (33). At pH 5, the pore switches to a primarily closed state with a P_o of ∼20%, which accounts for the loss of detection sensitivity. To maintain its high selectivity at a wide pH range, OmpG needs to stay in a predominantly open state, similar to its gating behavior at pH ≥ 6. Our previous work has shown that OmpG pH gating is controlled mostly by the electrostatic interaction (EI) network around L6 (Fig. 1 A) (33). Specifically, we found that repulsive force between the negatively charged L6 and the negative patch located at the opposite side of the barrel (Fig. 1 A) prevents the invasion of L6 into the lumen and thus destabilizes closed states. In contrast, the attraction between L6 and the positive patches (Fig. 1 A) promotes closed conformations. Acidic pH partially protonates residues of the negative patch (and L6) and consequently attenuates the repulsion force, which triggers OmpG to switch to a more closed state. Here, we substituted the negatively charged amino acids of L6 with positively charged lysines (K) or arginines (R) to create a mutant termed OmpG_L6KR (Figs. 1 B and S3). The lumen patch residues are expected to play an opposite role in the control of OmpG_L6KR gating (i.e., the negative patches attract the L6KR while the positive patches repel the L6KR from the lumen).

Figure 1.

The pH-dependent gating behavior of OmpG_L6KR construct versus OmpG_wt. (A) Structural models (PDB: 2IWV) of the lumen charges that form an electrostatic interaction network with L6. (B) Location of the charges on loop 6 (individually labeled blue and red spheres) and the charge residues found in the OmpG lumen (unlabeled blue and red spheres). (C) Single-channel traces of OmpG constructs at increasing pH conditions and (D) open probability, τ_open, and τ_closed characteristics of each construct at −50 mV. (E) Schematic of the two main forces, electrophoretic force (EP) and electrostatic interaction (EI), acting on the charges of loop 6 at −50 mV. (F) Single-channel traces of OmpG constructs at increasing pH conditions and (G) open probability, τ_open, and τ_closed characteristics of each construct at +50 mV. (H) Schematic of EP and EI, acting on L6 at +50 mV. The arrows in (E) and (H) depict the direction in which loop 6 will move when acted upon by either EP or EI. Negative charges are depicted as red spheres and positive charges as blue spheres. Data shown here were from at least five independent pores. To see this figure in color, go online.

We then tested how the gating of OmpG_L6KR responded to pH change with single-channel current recording experiments at ±50 mV (Fig. 1, C, D, F, and G). The potential polarity is defined by the potential of the chamber in which the OmpG loops are located (Fig. 1, E and H). At −50 mV, OmpG_L6KR stayed preferentially in the open state at all pH conditions tested, in contrast to OmpG_wt, which switched from an open to a mostly closed state when the pH decreased from 6 to 5 (Fig. 1 C). At pH 5, the P_o of OmpG_L6KR was ∼4.5-fold higher than that of OmpG_wt. A major contributor to the high P_o of OmpG_L6KR is a 12-fold reduction in τ_closed in comparison with that of OmpG_wt (Fig. 1 D; Table S2). Under the experimental conditions, the loop conformations of OmpG are governed by the EI between L6 and the barrel residues as well as the electrophoretic force (EP) that either attracts or repels L6 in and out of the lumen depending on the polarity of the applied potential across the bilayer. At −50 mV, the EP pushed the positively charged L6 of OmpG_L6KR away from the lumen (Fig. 1 E). The acidic pH of 5 weakens the attraction between the negative patch and L6KR; therefore, the repulsion by the positive patches may be the dominant EI force. Hence, both EP and EI act synergistically to drive L6KR out of lumen, which explains the drastically reduced τ_closed in comparison with OmpG_wt, whose negatively charged L6 is attracted into the lumen by both EP and EI at −50 mV and pH 5 (Fig. 1 E).

At −50 mV and pH > 5, the P_o values of both OmpG_wt and OmpG_L6KR were similar (Fig. 1 D; Table S2). For OmpG_wt, L6 is still attracted to the lumen under EP; however, the strong repulsion between the less protonated negative patch and L6 dominates the EI, resulting in an increase in P_o and a decrease in τ_closed. For OmpG_L6KR, we observed a slight increase in τ_closed, which could be due to an electrostatic attraction of L6KR toward the less protonated negative patch (Fig. 1, D and E).

At +50 mV, OmpG_L6KR gating no longer responded significantly to change in pH. Notably, the mutant was preferentially in the closed state at +50 mV, which contrasts with its behavior at −50 mV (Fig. 1, F and G; Table S2). The P_o of OmpG_L6KR was significantly reduced by 1.6-fold at pH 5, by 3.8-fold at pH 6, and by 3.1-fold at pH 7 in comparison with OmpG_wt. This is likely due to the EP that attracts the positively charged L6KR toward the lumen.

In addition, at +50 mV and pH > 5, the strong attraction of the negative patch to L6KR dominates the EI that aligns with the direction of EP, causing a drastically increased τ_closed and reduced τ_open in comparison with OmpG_wt (Fig. 1 H). At the acidic pH 5, because the L6KR loop/negative patch attraction weakens, the L6KR/positive patch repulsion may dominate, leading to the EI being in the opposite direction of EP (Fig. 1 H). This would explain the slightly decreased τ_closed at pH 5 versus the higher τ_closed at pH 6 and 7 in comparison with that of OmpG_wt (Fig. 1 G).

Notably, the pH-sensitive OmpG variants that we have created exhibit the tendency for τ_open to increase with higher pH, which indicates that other auxiliary β-barrel properties that may stabilize L6 in an open conformation, such as side-chain interactions, may grow stronger at higher pH (41). Overall, we observed that the gating of OmpG pores with a charged L6 is controlled by the collective effect of electrophoresis and the EIs of L6.

OmpG pores with a neutral L6 show lower gating activity

Because both the EP and intramolecular EIs modulate the OmpG through its charged L6 (whether negative or positive), we decided to create OmpG mutants with a neutral L6. In this way, we hypothesized that the neutral L6 would have reduced interactions with the charged residues in the lumen or neighboring loops and be devoid of the electrophoretic and electrostatic effect and thus would be an open OmpG construct with reduced pH-dependent gating. We generated two OmpG constructs: OmpG_L6NQ, in which glutamate, aspartate, and arginine were replaced with glutamine or asparagine, and OmpG_L6SGT, in which the negative residues were replaced with threonine or serine and arginine with glycine (Figs. 2 A and S3).

Figure 2.

The gating behavior of OmpG constructs with neutral loop 6 substitutions. (A) Structural models (PDB: 2IWV) of neutral loop 6 mutants in comparison with wild-type. (B) Single-channel traces of OmpG constructs at pH 5, 6, and 7 at −50 mV. (C) Box-and-whisker plot of the open probability, τ_open, and τ_closed characteristics of each construct at −50 mV. To see this figure in color, go online.

At −50 mV, both OmpG constructs with neutral L6 exhibited an open pore at all pH conditions tested (Fig. 2, B and C). However, OmpG_L6SGT always exhibited the higher P_o, higher τ_open, and lower τ_closed over OmpG_L6NQ. At pH 5, OmpG_L6NQ and OmpG_L6SGT had ∼5.1- and ∼5.7-fold increases in P_o, respectively, over OmpG_wt (Fig. 2, B and C; Table S3). Most notably, τ_closed decreased by ∼10- and ∼15-fold for OmpG_L6NQ and OmpG_L6SGT, whereas τ_open increased by ∼2.3- and ∼7.4-fold, respectively (Fig. 2 C; Table S3). The simultaneous increase in τ_open and decrease in τ_closed significantly contribute to the drastic increase in the overall P_o of the OmpG_L6NQ and OmpG_L6SGT constructs. At pH 6, τ_closed for OmpG_L6NQ and OmpG_L6SGT decreased by ∼2.6- and ∼3.8-fold, whereas τ_open increased by ∼2.1- and ∼4.5-fold, respectively, over that of OmpG_wt. At pH 7, τ_closed was similar for the mutants and wild-type, whereas τ_open increased by ∼2.5- and ∼3.8-fold for OmpG_L6NQ and OmpG_L6SGT, respectively, in comparison with OmpG_wt.

At +50 mV, OmpG_L6NQ and OmpG_L6SGT also showed a largely open pore at all pH levels. For example, at pH 5 the P_o of OmpG_L6NQ and OmpG_L6SGT increased by ∼1.5- and ∼1.8-fold, respectively, over OmpG_wt, which is less drastic than the change at −50 mV (Fig. S4; Table S3).

Overall, our data demonstrate that removing the charges at L6 greatly shifts the equilibrium toward the open state, and the neutral L6 constructs are only slightly sensitive to pH changes. Interestingly, the OmpG_L6NQ construct generally showed lower P_o than OmpG_L6SGT, which can be explained by the greater ability of L6NQ to interact with the lumen residues through hydrogen bonding, instead of EI. This hydrogen bonding can be facilitated by the longer side chain and the amine-rich nature of asparagine (N) and glutamine (Q) residues.

Gating effect of knockin mutations on neutral OmpG_L6SGT pores

To further pinpoint the contribution of specific residues on L6 in the wild-type pH-dependent gating, we used the OmpG_L6SGT as the reference construct and generated six knockin mutants, in which a single wild-type charged amino acid was re-introduced (Figs. 3, S3, S5, and S6; Tables 1 and S4). Upon re-introduction of the original charged residue, all knockin constructs except L6SGT_T229E remain predominantly open at all pH levels, retaining an P_o of at least 0.88 ± 0.05 at an applied potential of −50 mV (Fig. 3; Table 1). The τ_closed values for the five constructs, L6SGT_S221D, L6SGT_S224D, L6SGT_S225D, L6SGT_T227E, and L6SGT_G228R, were similar to that of OmpG_L6SGT, whereas their τ_open values were decreased moderately by 1.4- to 2.4-fold in comparison with that of OmpG_L6SGT at all pH conditions (Table 1). In contrast, the gating behavior of L6SGT_T229E deviated greatly from that of OmpG_L6SGT (Fig. 3, orange box). For example, at pH > 5, τ_open of L6SGT_T229E decreased by ∼5-fold compared with OmpG_L6SGT (Table 1). At pH 5, L6SGT_T229E exhibited even greater tendency toward the closed state: τ_open reduced by ∼7.6-fold and τ_closed increased by 7.2-fold in comparison with OmpG_L6SGT. Consequently, the P_o of L6SGT_T229E decreased from 95% at pH 7 to 30% at pH 5, thus showing a partially restored pH-sensitive gating, like OmpG_wt.

Figure 3.

Gating behavior of representative knockin mutants on loop 6 in the OmpG_L6SGT background. Knockin point mutants were substituted into loop 6 of OmpG_L6SGT to determine the effect of individual charges on gating at pH 5, 6, and 7. Single-channel recording of L6SGT_S221D is shown to represent the knockin mutants that did not significantly restore wild-type-like pH-sensitive gating at pH 5. Cartoon representations (right) depict the OmpG pores tested. Single-channel recording of all OmpG constructs was performed at −50 mV. To see this figure in color, go online.

Table 1.

Gating characteristics of knockin constructs in OmpG_L6SGT background at −50 mV

Construct	pH	Open Probability	τopen (ms)	τclosed (ms)
OmpG_L6SGT	5	0.96 ± 0.01	25.2 ± 4.2	1.0 ± 0.1
L6SGT_S221D		0.88 ± 0.05	10.5 ± 2.4	1.3 ± 0.3
L6SGT_S224D		0.94 ± 0.01	13.9 ± 0.8	0.8 ± 0.1
L6SGT_S225D		0.94 ± 0.03	18.0 ± 5.5	0.9 ± 0.1
L6SGT_T227E		0.89 ± 0.04	9.7 ± 1.3	1.3 ± 0.2
L6SGT_G228R		0.93 ± 0.03	24.9 ± 3.4	1.3 ± 0.1
L6SGT_T229E		0.30 ± 0.05	3.3 ± 0.6	7.2 ± 0.5
OmpG_wt		0.17 ± 0.04	3.4 ± 0.7	15.8 ± 2.6
OmpG_L6SGT	6	0.97 ± 0.03	70.9 ± 7.5	1.2 ± 0.2
L6SGT_S221D		0.97 ± 0.01	38.4 ± 1.7	1.4 ± 0.2
L6SGT_S224D		0.97 ± 0.02	49.7 ± 8.9	0.9 ± 0.3
L6SGT_S225D		0.99 ± 0.01	75.0 ± 8.4	1.1 ± 0.1
L6SGT_T227E		0.98 ± 0.01	74.4 ± 12.7	1.2 ± 0.1
L6SGT_G228R		0.97 ± 0.01	72.8 ± 12.2	1.7 ± 0.2
L6SGT_T229E		0.92 ± 0.03	17.1 ± 0.9	1.4 ± 0.1
OmpG_wt		0.73 ± 0.10	15.6 ± 3.3	4.5 ± 0.8
OmpG_L6SGT	7	0.99 ± 0.01	112.4 ± 28.3	1.3 ± 0.1
L6SGT_S221D		0.98 ± 0.01	71.8 ± 15.7	1.4 ± 0.1
L6SGT_S224D		0.99 ± 0.01	139.6 ± 34.6	0.9 ± 0.2
L6SGT_S225D		0.99 ± 0.01	81.1 ± 9.0	1.0 ± 0.1
L6SGT_T227E		0.99 ± 0.01	106.8 ± 21.5	1.2 ± 0.4
L6SGT_G228R		0.98 ± 0.02	130.3 ± 18.8	1.9 ± 0.2
L6SGT_T229E		0.95 ± 0.01	21.0 ± 0.5	0.8 ± 0.2
OmpG_wt		0.94 ± 0.02	29.3 ± 4.7	1.9 ± 0.5

Furthermore, we observed a similar trend for these knockin constructs at +50 mV (Fig. S6; Table S4). At pH 6 and 7, however, τ_closed for L6SGT_G228R increased by ∼2-fold in comparison with OmpG_L6SGT and OmpG_wt. This increase could be due to the EP, as the negative potential in the opposite-facing chamber attracts G228R toward the lumen, facilitating the interaction of R228 with the deprotonated negative patch at high pH (Fig. 1, A and H).

In summary, the significant effect of the E229 knockin mutation suggests that E229 may play a dominant role in OmpG’s pH-dependent gating.

Gating effect of E229 and R228 mutations in OmpG_wt L6 background

As the residue E229 largely rescued the pH-dependent gating behavior when introduced to OmpG_L6SGT, we asked to what extent removing this single charge from the wild-type would affect its gating. We substituted E229 with either threonine or arginine to generate L6wt_E229T and L6wt_E229R, with the other four negative residues remaining on the loop (Figs. 4 A, S3, and S7). We observed that both the neutral and positively charged residue substitution for E229 resulted in an increase in P_o of about 3.4-fold over OmpG_wt at pH 5 and −50 mV (Fig. 4). The increased P_o is mostly explained by a 5- to 5.8-fold decrease in τ_closed with little change to τ_open in comparison with OmpG_wt (Table 2). This result indicates that E229 stabilizes the closed state at pH 5, presumably by interacting with the positive patch in the lumen. The fact that the other four negatively charged L6 residues cannot remedy the loss of a single E229 suggests that the residue 229 may form stronger interactions with the positive patch than the other negative L6 residues, presumably because of closer proximity.

Figure 4.

Analysis of knockout point mutations of E229 and R228 on OmpG_wt background at −50 mV. (A) Single-channel recording and cartoon representation of the OmpG constructs at pH 5, 6, and 7 conditions. (B) Box-and-whisker plot of open probability of the OmpG constructs at pH 5, 6, and 7. To see this figure in color, go online.

Table 2.

Gating characteristics of knockout constructs in OmpG_wt background at −50 mV

Construct	pH	Open Probability	τopen (ms)	τclosed (ms)
OmpG_wt	5	0.17 ± 0.04	3.4 ± 0.7	15.8 ± 2.6
L6wt_E229T		0.58 ± 0.05	3.7 ± 0.7	2.7 ± 0.7
L6wt_E229R		0.49 ± 0.09	2.7 ± 0.5	3.0 ± 0.8
L6wt_R228G		0.58 ± 0.06	3.3 ± 0.3	2.3 ± 0.4
L6wt_E229T/R228G		0.79 ± 0.04	5.6 ± 0.9	1.6 ± 0.2
OmpG_L6SGT		0.96 ± 0.01	25.2 ± 4.2	1.0 ± 0.1
OmpG_wt	6	0.73 ± 0.10	15.6 ± 3.3	4.5 ± 0.8
L6wt_E229T		0.94 ± 0.02	17.0 ± 3.5	1.0 ± 0.2
L6wt_E229R		0.88 ± 0.02	11.7 ± 2.6	1.6 ± 0.1
L6wt_R228G		0.95 ± 0.02	26.1 ± 4.4	1.1 ± 0.1
L6wt_E229T/R228G		0.93 ± 0.02	21.0 ± 4.7	1.2 ± 0.1
OmpG_L6SGT		0.97 ± 0.03	70.9 ± 7.5	1.2 ± 0.2
OmpG_wt	7	0.94 ± 0.02	29.3 ± 4.7	1.9 ± 0.5
L6wt_E229T		0.96 ± 0.01	28.1 ± 4.8	1.2 ± 0.2
L6wt_E229R		0.90 ± 0.02	17.6 ± 3.0	2.0 ± 0.4
L6wt_R228G		0.97 ± 0.01	32.7 ± 3.1	1.2 ± 0.2
L6wt_E229T/R228G		0.96 ± 0.01	36.5 ± 10.3	1.3 ± 0.1
OmpG_L6SGT		0.99 ± 0.01	112.4 ± 28.3	1.3 ± 0.1

Because of its proximity to E229, as well as a voltage effect observed at +50 mV (Table S4), residue R228 was also selected for substitution with neutral residue glycine (Figs. 4 and S7; Tables 2 and S5). The resulting L6wt_R228G mutant showed an increase in P_o of 3.4-fold in comparison with OmpG_wt at pH 5, which is the result of a ∼7-fold decrease in τ_closed. Disruption of a potential interaction between R228 and the negative lumen patch could explain the increase in P_o. When we generated a double neutral substitution L6wt_E229T/R228G, the mutant exhibited a 10-fold decrease in τ_closed in comparison with the wild-type, which is a greater decrease than we saw with any of the single mutants (Fig. 4 A, orange box; Table 2). Consequently, at pH 5, the P_o of L6wt_E229T/R228G shifted from 0.17, a mostly closed pore, to 0.79, a mostly open pore, with the substitution of E229 and R228. We observed a similar trend in increased P_o and decreased τ_closed of the single and double mutants at pH 6, whereas at pH 7, the mutants’ behavior was similar to that of OmpG_wt (Fig. 4 B).

Of note, at +50 mV pH 5, only L6wt_R228G and L6wt_E229T/R228G showed significant increase in P_o (Fig. S7, Table S5). This suggests that the other four negatively charged L6 residues affect pH sensitivity more significantly at +50 mV, as loss of E229 alone only slightly increased P_o. Additionally at +50 mV, L6wt_E229R decreased in P_o at pH > 5, most likely because of the attraction to the negative patch in the lumen.

Taken together, E229 and R228, located on the C-terminal side of L6, contribute significantly to the pH sensitivity of OmpG_wt at pH < 7 and −50 mV. However, it is clear from OmpG_L6SGT that the pH sensitivity on L6 is due to the aggregate of all six of its charged residues (five negative, one positive), although each individual charge has varying degrees of impact on the pH-dependent gating.

OmpG^ssDNA pore for detection of complementary DNA and nuclease S1

Because the OmpG_L6SGT nanopore had improved P_o at pH 5 and both voltage polarities (±50 mV), we tested its utility in detecting and discriminating among biomolecule analytes at acidic pH. A cysteine residue was introduced at position 224 of OmpG_wt and OmpG_L6SGT and conjugated with a ssDNA 23-mer, resulting in two OmpG nanopore sensor constructs: OmpG_wt^ssDNA and OmpG_L6SGT^ssDNA (Figs. S3 and S8). We then tested the ability of the two OmpG^ssDNA sensors to detect the binding of either a complementary ssDNA or nuclease S1, an enzyme that digests ssDNA (or RNA) optimally at an acidic pH (Fig. 5). At +50 mV, little change in either OmpG construct was observable when either complementary ssDNA or nuclease S1 was added to the recording chamber (Fig. S9). In contrast, at −50 mV, both OmpG_wt^ssDNA and OmpG_L6SGT^ssDNA were permanently blocked, as the ssDNA was driven into the lumen by electrophoresis (Fig. 5, A and B). This blocked state could be transiently relieved by quickly switching the polarity to +50 mV (or 0 mV) and back again to −50 mV (Fig. S10). Once a single complementary ssDNA was bound to the OmpG^ssDNA, both pore constructs became unblocked and adopted a gating pattern similar to that of the unlabeled state, albeit with lowered P_o (Fig. 5, A and C). For example, when complementary ssDNA was bound, the P_o of OmpG_wt^ssDNA increased from blocked (0) to 0.06 ± 0.02 and that of OmpG_L6SGT^ssDNA increased to 0.84 ± 0.09 (Fig. 5, B and C; Table S6). The increased P_o is likely a result of the inability of the double-stranded DNA (dsDNA) to enter the nanopore lumen. Therefore, the gating is mediated primarily by L6, which explains the similar gating pattern between the unlabeled state and the dsDNA-labeled state of both OmpG^ssDNA constructs. Nevertheless, our experiments show that both OmpG_wt^ssDNA and OmpG_L6SGT^ssDNA are similarly effective in detecting complementary ssDNA at −50 mV.

Figure 5.

Comparison of the detection ability of OmpG_wt^ssDNA and OmpG_L6SGT^ssDNA at pH 5 and −50 mV. Top panel shows cartoon depictions of the OmpG nanopore conditions tested. The second and third panels show single-channel recordings and all-amplitude histograms for OmpG_wt^ssDNA. The bottom two panels show single-channel recordings and all-amplitude histograms for OmpG_L6SGT^ssDNA. OmpG gating analysis is shown for (A) unlabeled OmpG constructs, (B) OmpG^ssDNA constructs only, (C) OmpG^ssDNA after complementary DNA binding, (D) OmpG^ssDNA after nuclease S1 binding, and (E) OmpG^ssDNA after final cleavage of the ssDNA ligand conjugated to L6. The open probability values for each OmpG signal trace is shown in the all-amplitude histograms. The errors represent the SDs from three independent pores. The nanopores were recorded at −50 mV loop voltage in buffer containing 1 M KCl and 20 mM sodium acetate (pH 5). An additional 5 mM CaCl₂ was used during nuclease S1 binding and cleavage reactions (D and E). To see this figure in color, go online.

Next, we tested how the two OmpG^ssDNA nanopores responded to potential interactions with nuclease S1. Upon nuclease S1 binding to the ssDNA ligand, both OmpG^ssDNA constructs changed from a permanently closed unbound state at −50 mV to a partially open state (Fig. 5, B and D; Table S6), with the P_o of OmpG_wt^ssDNA increasing to 0.06 ± 0.02 and that of OmpG_L6SGT^ssDNA increasing to 0.27 ± 0.05. We did not observe any measureable tendency to switch from the partially open state to the permanently blocked state, which suggests that dissociation of nuclease S1 from OmpG^ssDNA was extremely slow. Nuclease S1 binding could be clearly discriminated from complementary ssDNA binding only in OmpG_L6SGT^ssDNA. In contrast, OmpG_wt^ssDNA exhibited almost identical current gating patterns in both the nuclease S1 and complementary ssDNA-bound states (Fig. 5, C and D, Table S6). We confirmed that binding of complementary ssDNA or nuclease S1 to OmpG^ssDNA is specific because neither a non-complementary ssDNA nor unlabeled OmpG-C224 induced a signal change at −50 mV (Fig. S11). Binding of nuclease S1 to the nanopore led to the digestion of the ssDNA ligand. After the complete cleavage of ssDNA from the nanopore, the gating pattern of the OmpG_L6SGT^ssDNA showed a drastic increase in P_o from 0.27 ± 0.05 (nuclease-S1-bound state) to 0.64 ± 0.08, which allowed the pre- and post-cleavage states to be unambiguously distinguished (Figs. 5, C–E, and S12; Table S6). In comparison, the P_o of OmpG_wt^ssDNA exhibited only a moderate increase from 0.06 ± 0.03 (nuclease-S1-bound state) to 0.15 ± 0.04 (Fig. 5, C–E; Table S6). Thus, at acidic pH, the OmpG_L6SGT^ssDNA nanopore exhibits improved ability in sensing and discriminating complementary ssDNA and a DNA processing enzyme over the OmpG_wt^ssDNA nanopore.

Conclusions

In this study, we identified the specific charged residues on L6 that mediate pH sensitivity and created an OmpG pore with a neutral L6 that remains open at all pH conditions tested. We also demonstrated the use of the OmpG nanopore for sensing ssDNA and a DNA nuclease at acidic pH and thus extended the utility of the OmpG platform to a wider range of analytes and operating pH conditions, which could be exploited for proteomics study.

Author contributions

M.A.V.F. conceived, designed, and performed the experiments; analyzed data; and wrote the manuscript. F.L. performed experiments and improved the method for DNA labeling. C.P. and C.Y. performed experiments. M.C. conceived, designed, and coordinated the study and edited the manuscript. All authors gave final approval of the manuscript for publication.

Editor: Sudha Chakrapani.