A minimalist biosensor: Quantitation of cyclic di-GMP using the conformational change of a riboswitch aptamer

Abstract

Cyclic di-GMP (c-di-GMP) is a second messenger that is important in regulating bacterial physiology and behavior, including motility and virulence. Many questions remain about the role and regulation of this signaling molecule, but current methods of detection are limited by either modest sensitivity or requirements for extensive sample purification. We have taken advantage of a natural, high affinity receptor of c-di-GMP, the Vc2 riboswitch aptamer, to develop a sensitive and rapid electrophoretic mobility shift assay (EMSA) for c-di-GMP quantitation that required minimal engineering of the RNA.

Introduction

Cyclic di-GMP (c-di-GMP) is a ubiquitous signaling molecule in bacteria that plays a role in virulence, motility and biofilm formation.¹ The molecule was first discovered by Benziman and coworkers in 1987 and has been the subject of extensive research, yet many questions remain about the components and mechanisms of signaling.^2-4 Recently, c-di-GMP and the other natural cyclic dinucleotides were shown to trigger an innate immune response in mammalian cells and thus, c-di-GMP is also a promising small molecule adjuvant.^5,6 In order to make further advances in studying c-di-GMP biology it is necessary that facile and accurate methods exist for measuring the concentration of this second messenger.

Key areas of c-di-GMP signaling that remain poorly understood include input signals that lead to c-di-GMP-dependent regulation, temporal resolution of c-di-GMP signaling, and an explanation for the high multiplicity of c-di-GMP metabolism enzymes in various bacteria.⁷ Bioinformatic studies have shown that diguanylate cyclases and phosphodiesterases, which synthesize and degrade c-di-GMP, respectively, are highly abundant in bacteria; for example 29 synthase and/or degradase enzymes have been predicted in E. coli alone.⁸ However, a full understanding of the regulation of these enzymes as well as how the global intracellular pool of c-di-GMP is affected by environmental signals is lacking. A general in vitro method for detecting c-di-GMP is advantageous because it can be applied for the study of virtually any organism, including genetically intractable organisms or mixed microbial populations, and ideally could detect changes in c-di-GMP concentration with high sensitivity.

Previously developed methods of detecting c-di-GMP in solution require extensive sample purification or have modest sensitivity. The commonly used HPLC-MS method has a lower limit of detection (LLOD) of 2 ng/mL or ∼2 nM c-di-GMP but typically requires significant sample purification and is limited in high-throughput capability.⁹ Other methods, including circular dichroism or small molecule-based detection methods, are more accessible but are based on the dimerization or G-quadruplex formation of c-di-GMP, and thus detect c-di-GMP at low micromolar concentrations.^10-12 Potential c-di-GMP binding proteins have been identified from whole cell lysates using the Differential Radial Capillary Action of Ligand Assay (DRaCALA), however this method uses radiolabeled ligand and is aimed at identification of protein-ligand interactions, not for quantification of endogenous ligand.¹³

Highly selective c-di-GMP fluorescent biosensors have been developed that take advantage of protein or RNA receptors that naturally bind c-di-GMP.^14-16 Protein-based biosensors use fusions of fluorescent proteins to PilZ domain-containing proteins, such as YcgR from Salmonella enterica and MrkH from Klebsiella pneumoniae, which give changes in Förster resonance energy transfer (FRET) upon binding c-di-GMP with reported dissociation constants (K_d) of 195 and 120 nM, respectively.^14,15 While these protein-based biosensors are useful for in vivo imaging of c-di-GMP, their use in vitro requires rigorous purification of the expressed protein from cellular extracts in order to remove any pre-bound c-di-GMP.

In contrast, one distinct advantage of riboswitch-based biosensors is the ability to synthesize RNAs by in vitro transcription, which eliminates the possibility of contaminating c-di-GMP from cellular extracts. A GEMM-I class riboswitch aptamer upstream of a Tfox-like gene in Vibrio cholerae called Vc2 exhibits remarkably high affinity for c-di-GMP (K_d ∼11 pM)^17,18 and has been fused to various output domains to create c-di-GMP sensors.^19,20 We¹⁶ and others²¹ have created fluorescent sensors by coupling the Vc2 aptamer to the Spinach aptamer,^22,23 an RNA mimic of GFP. Fusion of the Vc2 aptamer to Spinach either through its natural P1 stem (K_d ∼8 ± 1 nM at 25°C, 10 mM MgCl₂)¹⁶ or an unnatural stem (K_d not reported, estimated high nanomolar)²¹ results in a loss of affinity for c-di-GMP compared to the WT Vc2 aptamer alone. Alternatively, Breaker and coworkers have fused the Vc2 aptamer to a hammerhead ribozyme sequence through a random transducer stem and selected for allosteric c-di-GMP ribozyme sequences (affinities range from 70 to 500 nM).²⁰

Together, these prior studies indicate the potential use of riboswitch aptamers for in vitro detection of c-di-GMP. However, a minimalist approach that uses the conformational change of the Vc2 aptamer alone had not yet been explored. We were interested to see whether it was possible to develop a ligand quantitation assay with such a “minimalist biosensor,” because it could accelerate the time between riboswitch discovery and application as a biosensor.

Here, we show that it is possible to perform in vitro quantitation of c-di-GMP using the Vc2 aptamer alone. A tetraloop mutant of the Vc2 aptamer (A33U Vc2) was found to display an increased mobility shift upon ligand binding that enabled detection down to ∼100 pM of c-di-GMP, which is at least 20-fold more sensitive than previous methods. However, due to slow ligand equilibration rates and poor resolution of traditional slab gel EMSAs toward the separation of conformational states, this initial assay required nearly 4 days to complete. Through optimizing the binding reaction parameters to considerably reduce equilibration time and converting to a microfluidic mobility shift assay (μMSA) to further reduce electrophoretic separation time,^24,25 the total assay time was improved to 30 min while low nanomolar sensitivity to c-di-GMP was maintained. Thus, in this technical paper, we illustrate a strategy for adapting riboswitch aptamers to the μMSA format as minimalist biosensors for detection of their target ligands.

Materials and Methods

Reagents and oligonucleotides

DNA oligonucleotides were purchased from Elim Biopharmaceuticals (Hayward, CA) and IDT (Coralville, IA). Cyclic di-GMP was purchased from Axxora, LLC (Farmingdale, NY). Commercially available reagents were used without further purification. Alexa Fluor 488-labeled trypsin inhibitor (TI) was purchased from Invitrogen (Waltham, MA).

In vitro transcription

DNA templates were made through PCR amplification using primers that added the T7 polymerase promoter sequence. The templates were transcribed using T7 RNA polymerase (NEB, Ipswich, MA) in 40 mM Tris-HCl, pH 8.0, 6 mM MgCl₂, 2 mM spermidine, and 10 mM DTT. RNA was purified by denaturing (7.5 M urea) 6% polyacrylamide gel electrophoresis (PAGE) and was extracted from gel pieces using Crush Soak buffer (10 mM Tris-HCl, pH 7.5, 200 mM NaCl and 1 mM EDTA, pH 8.0). RNAs were precipitated with ethanol and resuspended in TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA). Accurate RNA concentrations were determined via a thermal hydrolysis assay.²⁶

Radiolabeling of RNA

In vitro transcribed RNA was dephosphorylated with alkaline phosphatase (Roche, Basel, Switzerland). The RNA was phenol/chloroform extracted and then ethanol precipitated overnight. RNAs were 5′-end labeled with Y-³²P ATP (Perkin Elmer, Waltham, MA) using T4 polynucleotide kinase (NEB) and were purified via PAGE as described previously.²⁷

Native slab gel EMSAs

Reactions (10 µL) were prepared by mixing renatured radiolabeled RNA (<100 pM or 1 uM, heated to 70°C for 3 min and cooled at room temperature for 10 min) with c-di-GMP in buffer containing 89 mM Tris, pH 8.0, 89 mM boric acid, 1–10 mM MgCl₂, 100 mM KCl, and 1 mg/mL yeast extract. Reactions were incubated at room temperature for various times, then were mixed with 3 µL 50% glycerol before loading on a 10% polyacrylamide native gel containing the same buffer. Gels were run for 14 h at 4°C, then were dried, exposed on a phosphor cassette, and imaged using a Typhoon laser-scanning system (GE Healthcare).

Fluorescent RNA labeling

Alexa Fluor 488-labeled RNA was prepared as described previously.²⁸ Briefly, the 3′ end of the RNA was oxidized with sodium periodate and then AlexaFluor488 hydrazide (Life Technologies, Carlsbad, CA) was coupled to the RNA in sodium acetate buffer. The RNA was precipitated in ethanol and then purified by PAGE as previously described.

Microfluidic device fabrication

As previously described,²⁴ glass microfluidic chips were designed in-house and fabricated with standard wet etching process by Perkin Elmer (Hopkinton, MA). Prior to gel polymerization in-chip, each microchannel was washed with 1 M NaOH and silanized with a degased solution of 3-(trimethoxysilyl)-propyl methacrylate (Sigma Aldrich, St. Louis, MO), glacial acetic acid (Sigma Aldrich), and water (Mediatech, Menassas, VA) (2:3:5, v/v/v). The solution was introduced to the chip by capillary force. Photopatterning of polyacrylamide in the glass channels was conducted in-house according to our previously described process.²⁴ A 3%T to 18%T discontinuous polyacrylamide gel (%T indicates the total concentration of acrylamide:bis, w/v; 3.3% was the weight percentage of bis-acrylamide in total monomer, w/w) was employed to enhance the separation resolution.

µMSA operation

The µMSA device is a glass microfluidic device comprised of 2 intersecting microchannels (80 µm deep × 20 µm wide), each terminating in fluid wells (Fig. 4A). Fluid wells are notated as: reagent reservoir (RR), reagent waste (RW), buffer reservoir (BR), and buffer waste (BW). Pipette tips (1 cm long) were cut and placed into the wells to expand the reservoir volume. Run buffer (1X TBMK) contained 89 mM Tris, 89 mM boric acid, 3 mM MgCl₂, and 10 mM KCl. The binding reactions (10 µL) were prepared in 0.5 mL Eppendorf Lo-Bind tubes with the following final concentration of components: 1X TBMK, 1 μM riboswitch RNA, 100 μg/mL yeast tRNA, 500 nM AlexaFluor 488 (AF488)-conjugated Trypsin Inhibitor (TI*) (Life Technologies, Carlsbad, CA), and indicated concentrations of c-di-GMP. Yeast tRNA is used to protect the riboswitch RNA from degradation by RNases. TI* was used as an internal standard. The mixture was heated at 70°C for 3 min and cooled for 10 min at room temperature to renature the RNAs before the addition of c-di-GMP and TI*. The riboswitch binding reaction was incubated off-chip in a tube, and a 4 µL sample was pipetted into the RR. All other wells were filled with 4 µL of 1X TBMK buffer. Platinum electrodes were inserted into each well and were connected to a custom built, 8-channel high voltage power supply with current/voltage feedback control. The µMSA was performed in constant current mode. First, samples were loaded into the microchannel by applying −12 µA to the RR, and electrophoresed into the RR-to-RW microchannel by applying −2 µA to BR and BW as pinching voltage and grounding RW for 1 min. After the loading process was stabilized, the voltage plan was switched to sample injection/separation, by applying −32 µA to BR, 10 µA to RR and RW as pulling voltage and grounding BW. The result is injection of a plug of sample into the separation channel at the t-junction and the sample is electrophoresed into the BR-to-BW microchannel, which houses a 3–18 %T discontinuous sample stacking and separation gel.

Figure 4.

Conversion to a μMSA allows for rapid separation of conformational states and low nanomolar c-di-GMP detection. (A) Upper is photograph of the µMSA glass chip. Microchannel network is comprised of 2 intersecting microchannels, each terminating in a fluid reservoir. Reservoirs are: buffer reservoir, BR; riboswitch reservoir, RR; riboswitch waste, RW and buffer waste, BW. Bottom is simplified schematic of the microchannel network. (B) Example full-field image of binding experiment. A binding reaction containing 100 nM c-di-GMP and 1 μM Alexa Fluor 488-labeled RNA was incubated for 60 min and was then loaded on a polyacrylamide gel run for 30 s at RT. (C) Analysis of equilibration time required for an excess of RNA to equilibrate with c-di-GMP. Binding reactions containing 100 nM c-di-GMP and 1 μM Alexa Fluor 488-labeled RNA were incubated for various times and were then loaded on a polyacrylamide gel run for 30 s at RT. (D) Standard curve of A33U Vc2 binding c-di-GMP. A binding reaction containing 1 μM Alexa Fluor 488-labeled RNA with various c-di-GMP concentrations was incubated for 60 min and was then loaded on a polyacrylamide gel run for 30 s at RT. In all studies, TI was used as an internal standard. A.F.U. stands for arbitrary fluorescence units.

µMSA imaging and data analysis

Full-field images of peak migration and fluorescence were collected via an Olympus IX-70 inverted epi-fluorescence microscope equipped with a Peltier cooled charge-coupled device (CCD) camera (CoolSNAP HQ2, Roper Scientific, Trenton, NJ) and a 10X objective. An X-Cite1 exacte mercury lamp (Lumen Dynamics, Mississauga, Canada) provided the illumination source, which was then filtered by an XF100-3 filter (Omega Optical, Brattleboro, VT). In equilibrium time studies, the camera exposure time was 100 ms. For concentration calibration, electrophoresis was monitored in real time and halted when the unshifted RNA band migrated to the 1 mm marker, followed by collection of a 500 ms exposure image.

Images were analyzed using ImageJ software (NIH, Bethesda, MD). Fluorescence intensity was collected along the axis of the separation channel and averaged in the transverse direction. Post-processing was performed with OriginPro 8.0 (OriginLab, Northampton, MA). A built-in non-linear Gaussian peak-fitting algorithm was employed to calculate peak position and area-under-the-curve (AUC) from the axial fluorescence signal. To quantify the amount of bound riboswitch, we introduced TI as an internal standard molecule. Instead taking the absolute value of AUC of the bound riboswitch, we normalized it to that of TI therefore minimizing the run-to-run variation. Signal-to-noise ratio (SNR) was calculated by dividing the maximum intensity value of the Gaussian fit by the root mean square (r.m.s) noise of neighboring regions of signal. Peaks were considered detectable when SNR >3, as per convention.

Cell extraction of E. coli

E. coli strains from the Keio collection²⁹ were obtained from the E. coli Genetic Stock Center (Yale University). The wild-type E. coli K-12 strain was BW25113 [F-, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ–, rph-1, Δ(rhaD-rhaB)568, hsdR514]. Mutant strains were isogenic derivatives of BW25113 and included JW1278-1 (Δgmr-722::kan), JW2052-1 (ΔyegE-765::kan), JW3493-2 (ΔyhjH-780::kan), JW5291-1 (ΔyeaJ-783::kan). Fresh cultures of 15 mL of LB (BW25113) or LB/kanamycin (mutant strains, 50 µg/mL kanamycin) were inoculated 1:40 with overnight cultures grown at 37°C. Cultures were grown in a 37°C incubator with shaking for approx. 7 h, then the OD₆₀₀ was measured and cell pellets were collected by centrifugation at 6,500 rpm for 10 min at 4°C. An organic-aqueous extraction of cell pellets was performed as described previously.⁹ Briefly, the pellets were resuspended using 600 µL acetonitrile/methanol/water (40/40/20, v/v/v) on ice for 20 min with occasional vortexing. The suspension was further extracted at ambient temperature for 20 min, then cell debris was pelleted with centrifugation at 13,200 rpm for 20 min at 4°C. The supernatant was carefully removed and stored on ice and the pellet was extracted twice more with 300 µL extraction buffer. The combined supernatants were evaporated to dryness by rotary evaporation and the dried material was resuspended in 200 µL deionized H₂O. For analysis by µMSA, 5 µL of concentrated extract was added to bring the final volume of the riboswitch binding reaction to 10 µL (final concentrations of other components are 1 µM AF488-labeled RNA, 100 µg/ml tRNA, 1X TBMK buffer, and 500 nM TI*). The reported c-di-GMP concentration values are for the concentrated cell extract, take into account the 2X dilution factor, and are normalized to OD₆₀₀.

Results

A tetraloop mutant of the Vc2 aptamer demonstrates an increased electrophoretic mobility shift

Conventional EMSAs are based upon separation of different molecular weight species. For example, formation of the riboswitch aptamer-ligand complex using Vc2 RNA and radiolabeled c-di-GMP has been observed by native EMSA, which relies on the large electrophoretic mobility difference between free and complexed c-di-GMP.^18,19,30 Smith et al. employed a competition binding experiment to analyze binding of unlabeled ligands.¹⁸ However, these types of assays typically have poorer limits of detection than direct binding assays, since low levels of c-di-GMP inefficiently compete with the labeled ligand.

Although less common, native EMSAs also can be used to separate different radiolabeled RNA conformations induced by ligand binding (Fig. 1A). This method requires labeled RNA but enables unlabeled ligands to be interrogated by direct binding experiments, which is important for quantifying c-di-GMP from natural sources, evaluating potential ligand candidates,³¹ or identifying small molecule inhibitors of riboswitch activity.³² Previous studies using small angle x-ray scattering (SAXS) and single molecule FRET (smFRET) experiments have shown that the Vc2 aptamer undergoes a significant structural compaction upon binding c-di-GMP.^19,30 Specifically, in an unbound state, the majority of RNAs are in an elongated conformation with distal P2 and P3 stems, but when bound to c-di-GMP, these 2 stems form tertiary interactions that result in a more globular architecture. This structural compaction has been used by Kulshina et al. to determine the specificity of the RNA for c-di-GMP over other nucleotide analogs through native EMSA.³⁰

Figure 1.

An A33U mutation leads to a larger electrophoretic mobility shift of the Vc2 aptamer with dependence on Mg²⁺ concentration. (A) Model of the conformational change resolved by an electrophoretic mobility shift assay. The global RNA conformations are outlines of structures previously solved by SAXS. (B) Sequence and secondary structure of the Vc2 aptamer. The A33U mutation is boxed and c-di-GMP is in large letters. The tetraloop-tetraloop receptor interaction is shown in dashed lines and nucleotides involved in this tertiary interaction are in gray. (C) EMSA of WT and A33U Vc2 RNAs with or without c-di-GMP run in 10 mM, 3 mM, or 1 mM MgCl₂. Approximately 1 nM ³²P A33U Vc2 was incubated with 10 μM c-di-GMP for 4 h before loading on 10% polyacrylamide gels run for the same amount of time (14 h) at 4°C. Images encompass the area 8.5 – 13.5 cm from the top of the gels.

In contrast to a binary plus/minus binding assay, we sought to develop an assay that would allow for highly sensitive quantitation of c-di-GMP. Determination of c-di-GMP concentrations through EMSA requires resolution of the bound and unbound bands for accurate quantitation. Thus, optimization efforts sought to increase the observed mobility shift upon ligand binding. Specifically, we analyzed 2 strategies for affecting RNA structure: the effects of altering Mg²⁺ concentrations and of weakening a tertiary interaction. Magnesium ions mediate binding of c-di-GMP to the Vc2 aptamer and have been shown to affect mobility shifts of another riboswitch aptamer.^19,33 In addition, one tertiary interaction that stabilizes the globular structure of the bound Vc2 riboswitch is a tetraloop/tetraloop receptor interaction between the P2 and P3 stems.^18,30 An A33U mutation to the tetraloop in the P2 stem loop of Vc2 impairs a base-stacking interaction with the tetraloop receptor in the P3 stem, therefore weakening the ability of this tertiary interaction to form (Fig. 1B).^19,34

Radiolabeled WT and A33U Vc2 aptamer RNAs were incubated in buffers with varied MgCl₂ concentrations either with or without an excess of c-di-GMP, then were separated on native slab polyacrylamide gels containing the corresponding concentration of MgCl₂ (Fig. 1C; Fig. S1). Under all 3 Mg²⁺ conditions tested, the WT Vc2 aptamer shows a smaller relative mobility shift upon binding c-di-GMP than the A33U Vc2 aptamer. Interestingly, the main difference in relative mobility shift comes from decreased mobility of the A33U aptamer in the unbound state, an effect that is enhanced with lower magnesium concentrations (Table 1). In contrast, the mobility of the c-di-GMP bound WT and A33U Vc2 RNAs is nearly equal within each gel, suggesting that the A33U mutation reduces compactness of the unbound aptamer, but that addition of c-di-GMP restores the structure to match the mobility of WT Vc2.

Table 1.

Distance traveled by WT and A33U Vc2 RNAs under various MgCl₂ concentrations. Data were obtained by measuring the distance of the bands shown in Figure 1C from the well of the gel using ImageJ software

	10 mM MgCl2	3 mM MgCl2	1 mM MgCl2
WT Vc2 – c-di-GMP	9.82 cm	11.14 cm	12.75 cm
WT Vc2 + c-di-GMP	9.93 cm	11.35 cm	13.11 cm
A33U Vc2 – c-di-GMP	9.37 cm	10.54 cm	11.77 cm
A33U Vc2 + c-di-GMP	9.98 cm	11.28 cm	12.95 cm

In addition, Mg²⁺ concentration impacts both the absolute and relative electrophoretic mobility of the aptamers (Fig. 1C; Fig. S1). First, decreasing the MgCl₂ concentration led to increased electrophoretic velocities for all RNAs. Intuitively, this effect is due to reduced shielding by magnesium ions of the negatively charged RNA backbone. Theoretically, decreasing ionic strength increases the Debye length of the electric double layer around the molecule, as reflected in an altered zeta potential (${\displaystyle {\varphi }_0}$). The magnitude of the electrophoretic mobility is proportional with ${\displaystyle {\varphi }_0}$, given by ${\displaystyle \mu =\epsilon E{\varphi }_0/\eta }$, in which ${\displaystyle \epsilon }$ is the permittivity of the electrolyte solution, E is the electric field strength and ${\displaystyle \eta }$ is the dynamic viscosity. Increased ionic strength can also lead to Joule heating, which may alter the dynamic viscosity. Second, decreasing the Mg²⁺ concentration also yielded larger separations between the bound and unbound RNA conformations for both WT and A33U Vc2 RNAs (Table 1), although the effect is more pronounced for the A33U mutant. Thus, lowering the MgCl₂ concentration affects both the absolute electrophoretic velocities and the relative mobility shift of each species, thus enhancing the separation resolution between bound and unbound species.

The electrophoretic mobility shift assay allows for high picomolar detection of c-di-GMP

A slow off-rate (1.1 × 10⁻⁵ min⁻¹) has been reported for c-di-GMP dissociation from the WT riboswitch aptamer under buffer conditions with 10 mM MgCl₂.¹⁸ Thus, not only does the WT aptamer have a reduced mobility shift compared to A33U Vc2, but the time for the binding reaction to reach equilibrium (ca. months) renders WT Vc2 impractical for c-di-GMP quantitation. In contrast, the binding reaction between A33U Vc2 (<100 pM) and non-saturating concentrations of c-di-GMP reaches equilibrium after ∼75 h of incubation (Fig. 2A). Similar equilibration times were observed for reactions whether incubated in buffer with 10 mM MgCl₂ and 10 mM KCl or 3 mM MgCl₂ and 3 mM KCl, indicating that in this range of ionic concentrations there was little effect on kinetics (Fig. 2A; Fig. S2). Thus, we used a 75 h incubation time for experiments to determine the sensitivity of c-di-GMP detection.

Figure 2.

The A33U Vc2 aptamer detects high picomolar levels of c-di-GMP. (A) Analysis of A33U Vc2 equilibration time with c-di-GMP. Less than 100 pM ³²P A33U Vc2 RNA was incubated with 1 nM c-di-GMP for various times before loading on a 10% polyacrylamide gel run for 14 h at 4°C in buffer containing 10 mM MgCl₂ and 10 mM KCl. (B) EMSAs of A33U Vc2 with c-di-GMP under different MgCl₂ concentrations. Less than 100 pM ³²P A33U Vc2 was incubated with various concentrations of c-di-GMP for 75 h before separating on a 10% polyacrylamide gel run for 14 h at 4°C. (C) Analysis of A33U Vc2 affinity for c-di-GMP under various MgCl₂ concentrations. Data from 2 or 3 independent replicates and the best-fit curves are shown.

While lowering Mg²⁺ levels leads to a larger electrophoretic mobility shift between bound and unbound species, reduced Mg²⁺ levels also result in poorer binding affinity for c-di-GMP (Fig. 2B, C). At 1 mM MgCl₂ the affinity of A33U Vc2 for c-di-GMP is in the nanomolar regime (K_d value of 1.6 ± 0.1 nM), whereas sub-nanomolar affinity is achieved at 3 and 10 mM MgCl₂ (K_d values of 520 ± 30 pM and 210 ± 10 pM, respectively). The LLOD for c-di-GMP is ∼100 pM with 10 mM MgCl₂ (SNR > 3) (Fig. 2B), which is an order of magnitude lower than the most sensitive existing methods.^9,35 However, as a compromise between resolution and binding affinity, further optimization of the assay was performed with 3 mM MgCl₂. Under these conditions, the dynamic range of detection between 10–90% of the fraction bound is 55–4400 pM c-di-GMP, providing a method for quantifying low levels of c-di-GMP with high accuracy. Furthermore, preliminary data shows that mutants of the Vc2 aptamer enable adaptation of this assay for detection of the recently discovered second messenger 3′–5′, 3′–5′ c-AMP-GMP (Fig. S3), which has been implicated in the regulation of Vibrio cholerae virulence³⁶ and Geobacter extracellular electron transport.^37,38

Optimization of the slab gel EMSA and conversion to µMSA allows for more rapid and sensitive detection of c-di-GMP

We considered that the long equilibrium incubation time (∼75 h) is not only a function of the rate constants but is also dependent on the concentrations of RNA and ligand in the binding reaction. In order to reduce the time to reach equilibrium, the addition of excess unlabeled riboswitch aptamer in the radiolabeled binding reactions was analyzed. For example, in a binding reaction containing a total RNA concentration of 1.1 μM A33U Vc2 with 100 nM c-di-GMP, equilibrium was reached within 30 min (Fig. 3A), which is a 150-fold decrease in incubation time compared to previous experiments using <100 pM of A33U Vc2 with 1 nM c-di-GMP, albeit under slightly different MgCl₂ concentrations (Fig. 2A).

Figure 3.

Addition of excess RNA decreases equilibration time while maintaining low nanomolar sensitivity. (A) Analysis of equilibration time required for an excess of A33U Vc2 to bind c-di-GMP. Approximately 100 pM ³²P A33U Vc2 was mixed with 1 μM unlabeled RNA and 100 nM c-di-GMP then was incubated for various times before loading on a 10% native polyacrylamide gel run for 14 h at 4°C. The gel image insert shows the data points analyzed. (B) Standard curve of c-di-GMP binding A33U Vc2. Various concentrations of c-di-GMP were incubated with approximately 100 pM ³²P A33U Vc2 and 1 μM unlabeled RNA and 100 nM c-di-GMP for 30 min before loading on a 10% native polyacrylamide gel run for 14 h at 4°C. The image of the gel analyzes is shown in the insert; the 6^th and 7^th lanes were excluded from analysis as they were a result of a gel loading error.

With the addition of excess RNA, the amount of the Vc2 aptamer greatly exceeds the dissociation constant, so the signal from bound RNA becomes directly proportional to the amount of c-di-GMP present in the sample. Importantly, the measurement of c-di-GMP concentrations no longer depends on the fraction of RNA bound, but solely on the amount of RNA that is shifted to the bound state (as excess RNA remains largely unbound at low levels of c-di-GMP). This method was used to generate a standard curve for c-di-GMP detection (Fig. 3B). The LLOD for c-di-GMP under these conditions is ∼5 nM, which matches that of current HPLC-MS methods,^9,35 and has a linear response to increased c-di-GMP levels up to 400 nM as shown here.

While the reaction equilibration time was decreased, performing the separation on the slab gel still required 12–14 h of gel electrophoresis at 4°C. To reduce electrophoresis time, we converted to the microfluidic mobility shift assay (μMSA) format (Fig. 4A). This technique provides rapid, resource-sparing, and reproducible results using fluorescently-labeled RNA without the precautions required for radiolabeled materials.²⁴ As shown, separation of bound and unbound A33U Vc2 aptamer labeled at the 3′ end with Alexa Fluor 488 was achieved in 30 seconds at room temperature, a ∼1,440-fold decrease in the separation time (Fig. 4B). Meanwhile, the same equilibration time of 30 min was found when monitoring the binding reaction by μMSA (Fig. 4C).

A standard curve for c-di-GMP detection was generated via μMSA using 1 μM Alexa Fluor 488-labeled RNA with c-di-GMP concentrations ranging from 1–1000 nM (Fig. 4D). The amount of bound RNA was observed to increase linearly with the concentration of c-di-GMP, and the LLOD is 10 nM c-di-GMP (S/N ratio ∼6.7) (Fig. S4). Thus, on the microfluidics platform, we are able to achieve comparable sensitivity and similar linearity of response relative to the slab gel format. The aforementioned advantages of the μMSA format along with the 174-fold reduction in experiment time makes this assay advantageous for highly sensitive and rapid detection of unlabeled c-di-GMP.

Previously, deletion mutants of 2 phosphodiesterases, Gmr (also called YciR) and YhjH, and 2 diguanylate cyclases, YegE and YeaJ, in different E. coli strains have been shown to affect biofilm formation and/or motility phenotype.^39-41 Specifically, we expected that knocking out the phosphodiesterases would increase c-di-GMP concentrations, leading to increased biofilm staining and decreased motility in soft agar assays. In contrast, we expected that knocking out the diguanylate cyclases would decrease c-di-GMP concentrations, leading to the opposite phenotypes. The phenotypic observations for these mutants and, in the case of Δgmr, allosteric ribozyme reporter assays,⁴² have been consistent with the predicted changes in c-di-GMP concentrations. However, to our knowledge, the c-di-GMP concentrations for these mutant strains had not been directly measured.

To analyze endogenous c-di-GMP levels, we performed an organic-aqueous extraction of cell pellets from Keio collection strains grown to stationary phase (OD₆₀₀ ∼3) in 15 mL of LB liquid culture without or with kanamycin (for mutants), following previously described procedures.^9,37 The strains analyzed were BW25113, the wild-type E. coli K-12 strain, and isogenic derivatives of BW25113 with the following single gene deletions: Δgmr, ΔyhjH, ΔyegE, and ΔyeaJ. We then attempted to quantify c-di-GMP levels directly from these crude cell extracts via µMSA with the fluorescently labeled A33U Vc2 aptamer. Triplicate runs on the µMSA platform gave consistent values for each sample, but we unfortunately observed high variability between 2 independent biological replicates from experiments performed on different days (Fig. S5). Also in retrospect, the Keio strains were not the same strains used in the phenotype assays, and the growth conditions were different as well. These differences may change the expected c-di-GMP concentrations, as E. coli harbor multiple diguanylate cyclases and phosphodiesterases (up to 29) whose expression levels are influenced by growth conditions.¹

Discussion

Here we present a minimalist approach toward biosensor development that focused on in vitro detection of c-di-GMP using the Vc2 riboswitch aptamer alone. Numerous riboswitch aptamers exhibit low nanomolar affinities toward their cognate ligands^17,43 and undergo a conformational change upon ligand binding, so this study details the assay development steps necessary and serves as a proof-of-principle for c-di-GMP detection. Recently, we also have used a SAM-I riboswitch aptamer to verify S-adenosylmethionine levels in yeast cell extracts through standard EMSA.⁴⁴ Together, these studies show that the conformational shift of riboswitch aptamers, which has previously been used to study RNA structure and folding,⁴⁵ can be used to quantify their small molecule ligands.

One key characteristic of the Vc2 aptamer is its unusually slow ligand off-rate,¹⁸ which is over 4 orders of magnitude slower than the reported off-rates for the adenine and FMN riboswitches.^46,47 Slow ligand dissociation contributes to the high sensitivity of the aptamer for c-di-GMP and the observation of distinct bands for ligand-bound and unbound states, as ligand dissociation is slower than the rate of separation by electrophoresis. We found that the A33U Vc2 tetraloop was useful to enhance electrophoretic separation of the 2 species. Previous affinity studies of the A33U Vc2 aptamer reported a 200-fold decrease in c-di-GMP affinity,³⁴ attributed to an increase in the off-rate. This A33U Vc2 mutation destabilizes an important tertiary interaction whereas other disruptive mutations mainly facilitate unbinding of the ligand.³⁴ Single molecule studies have shown that this tetraloop mutation results in an increase in the amount of statically undocked RNAs, in which the P2 and P3 stems are distal, agreeing with this RNA being in a less compact formation when not bound to c-di-GMP.¹⁹

We found that an intermediate (3 mM) magnesium concentration maintains strong c-di-GMP binding affinity and still provides sufficient destabilization of the unbound aptamer structure to produce an improved mobility shift. Previous studies of the Vc2 aptamer have linked low magnesium concentration to reduced ligand binding affinity.^16,19,30 The crystal structure of the c-di-GMP bound riboswitch aptamer revealed that magnesium coordinates with c-di-GMP in the ligand binding pocket and SAXS studies showed low magnesium concentration correlates with increased global size.^30,34

Finally, in order to reduce the assay time, we switched from a “fraction bound” type of binding assay that uses RNAs at concentrations below the K_d value to a “standard curve” type of binding assay that uses RNAs at stoichiometric concentrations. This modification in assay protocol, in combination with the adoption of the μMSA format, provides an assay that can be accomplished in practical time-scales and with low amounts of fluorescently labeled RNAs. The fluorophore labeling procedure can be performed on in vitro transcribed RNAs of any length, and the sequence at the 3′ end can be extended as necessary to avoid perturbation of ligand binding to the aptamer (Fig. 1B). We demonstrate that μMSA matches the limit of detection of the most sensitive existing techniques for c-di-GMP quantification.

In the future, we expect that further method optimization and normalization controls will be required to quantify c-di-GMP accurately from crude bacterial extracts, as opposed to extensively purified standards. However, our preliminary results do suggest that the assay will have the sensitivity to detect c-di-GMP from small batches of cell extracts. The high variability in independent replicates point out that a main challenge in fact may be accounting for differences in biological samples and in extraction efficiency. A potential solution would be to measure a larger number of replicates so that any statistically significant differences can be reliably assessed. Promisingly, the format offers the possibility for high-throughput screening, as a free-standing polyacrylamide gel device recently has been described that enables 96-well multiplexing of μMSA for sample analysis.²⁵

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This work was supported by Department of Defense NDSEG Fellowship (C.A.K.), NIH New Innovator Award (Grant No. 1DP2OD008677 to M.C.H.) and NSF CAREER Award (Grant CBET-1056035 to A.E.H.). M.C.H. holds a Career Award at the Scientific Interface from the Burroughs Wellcome Fund. A.E.H. is an Alfred P. Sloan Foundation Research Fellow in chemistry.