Enzymatic Syntheses and Applications of Fluorescent Cyclic Dinucleotides

Abstract

Bacterial cyclic dinucleotides (CDNs) play important roles in regulating biofilm formation, motility and virulence. In eukaryotic cells, theses bacterial CDNs are recognized as pathogen-associated molecular patterns (PAMPs) and trigger an innate immune response. We report the photophysical analyses of a novel group of enzymatically synthesized emissive CDN analogues comprised of two families of isomorphic ribonucleotides. The highly favorable photophysical features of the CDN analogues, when compared to their non-emissive natural counterparts, are used to monitor in real time the dinucleotide cyclase-mediated synthesis and phosphodiesterase (PDE)-mediated hydrolysis of homodimeric and mixed CDNs, providing effective means to probe the activities of two classes of bacterial enzymes and insight into their biomolecular recognition and catalytic features.

Graphical Abstract

CDNs are second messengers that regulate central processes in all bacteria. The enzymatic syntheses and applications of novel isomorphic fluorescent CDN analogues are reported, illustrating their utility in probing the activities of two classes of bacterial enzymes and providing insight into their biomolecular recognition and catalytic features.

Introduction

Cyclic dinucleotides (CDNs) are second messengers that regulate central processes in all bacteria, including biofilm formation, motility and virulence.^[1] The most studied derivatives include the purine-containing c-di-GMP, c-di-AMP and 3’,3’-c-GAMP.^[2] The recent discovery of c-di-UAMP in E. coli suggests, however, that pyrimidine nucleotides can also be engaged in forming CDNs in vivo.^[3] Bacterial CDNs are recognized as pathogen-associated molecular patterns (PAMPs) and trigger the innate immune response in eukaryotic cells.^{[2–3, 4]} As their regulatory roles are gradually being recognized,^{[1a, 5]} analogues that can shed light on the pathways involved in their production, degradation and recognition are of significance.

CDNs are produced from two nucleotide triphosphates (NTPs) by dinucleotide cyclases. GGDEF and DisA domains are responsible for the synthesis of c-di-GMP and c-di-AMP. DncV, cGAS and the recently discovered cdnE are able to produce the heterodimeric CDNs, 3’,3’-c-GAMP, 2’,3’-c-GAMP and 3’,3’-c-UAMP, respectively.^[6] Whereas cGAS or cdnE have not been shown to generate any homodimeric CDNs, DncV was reported to be promiscuous and produce c-di-GMP and, c-di-AMP, in addition to 3’,3’-c-GAMP, when provided with purine nucleotides in vitro.^[7]

The global and local concentrations of bacterial CDNs are tightly regulated in response to intracellular and extracellular signals.^{[1a, 8]} This control is achieved by the orchestrated activity of dinucleotide cyclases and phosphodiesterases (PDEs), which cleave CDNs into either linear pNpNs or monomeric nucleotides.^[5b] For example, PDEs containing EAL domains, (i.e. rocR from P. aeruginosa) hydrolyze c-di-GMP into pGpG while those with HD-GYP domain cleaves it into GMP. ^{[1a, 2]}

Interrogating the in vitro syntheses and hydrolyses of CDNs often relies on chromatographic techniques. While reliable and informative, such tools are not amenable to real-time monitoring and high throughput analyses, a feature frequently afforded by fluorescence-based methods. Applying the latter is challenging, however, as the natives CDNs are non-emissive and modifications aimed at conferring favorable photophysical features might perturb the processes of interest. Indeed, while using synthetic CDNs containing common emissive nucleosides to fluorescently monitor PDE-mediated c-di-GMP degradation and hydrolysis has been demonstrated (using 2-aminopurine),^[9] applying such strategies to the study of dinucleotide cyclases demands higher isomorphicity and substrate recognition of the NTP precursors, criteria not regularly found among common emissive nucleoside analogues.^[10]

Recent efforts in our laboratory have yielded two families of isomorphic emissive ribonucleosides, the thiopheno- and isothiazolo-based RNA alphabets (^thN and ^tzN, respectively; see examples in Figure 1).^[11] The two families display distinctive photophysical properties and have been used to investigate catalytic RNAs, nucleoside metabolizing enzymes and diverse nucleotide-based cofactors.^[12] We surmised that the isomorphic fluorescent purines analogues’ visible emission and sensitivity to environmental changes could be exploited to study the enzymes involved in CDNs biosynthesis and degradation, as schematically illustrated in Figure 1.

Figure 1.

Top: Guanosine and its emissive surrogates ^thG and ^tzG; Bottom: a hypothetical model predicting fluorescence changes upon CDN formation and hydrolysis.

While polymorphic, prone to aggregation and highly dependent on conditions, c-di-GMP displays stacking of the two nucleobases in solution.^[13] This suggests distinct photophysics for any CDN compared to that displayed by its precursors (NTPs) or degraded products (pN or pNpN). We further hypothesized that when ^thGTP or ^tzGTP would cyclize to the corresponding CDNs, the fluorescence would likely diminish and, conversely, phosphodiester hydrolysis of emissive CDNs would lead to fluorescence enhancement (Figure 1). Such signal change can thus be used to monitor the process in real-time, determine reaction kinetics and, in principle, be used to facilitate inhibitor discovery. Herein we report the fluorescence-based monitoring of CDN synthesis and hydrolysis using ^thG and ^tzG, two isomorphic guanosine surrogates. The resulting structure-activity relationship provides insight into the substrate recognition and catalytic mechanisms of the CDN-specific PDEs studied.

Results and Discussion

Enzymatic synthesis of c-di-GMP analogues with DncV.

CDN analogues have greatly facilitated mechanistic, biochemical and structural studies, particularly in the context of CDN-binding riboswitches and protein receptors.^{[7b, 9, 14]} The biggest hindrance to such studies has frequently been the preparation of analogues, as they have been predominately stepwise synthesized using phosphoramidite chemistry.^{[7b, 15]} We employed DncV, a cyclic dinucleotide synthetase from Vibrio cholerae, to enzymatically produce a series of c-di-GMP analogues. While this enzyme primarily synthesizes 3’,5-’c-GAMP in vivo, when provided with only GTP or ATP in vitro it is also capable of making c-di-GMP and c-di-AMP.^[16] Since DncV can accept both guanosine and adenosine, we postulated it would tolerate the thiopheno and isothiazolo G surrogates, members of our previously synthesized emissive RNA alphabets (Figure 1).^[11]

To benchmark the enzymatic synthesis of c-di-GMP and the corresponding emissive analogues, DncV was incubated with GTP, ^thGTP and ^tzGTP (Figure 2). The reactions were analyzed by HPLC and mass spectrometry (Figure 3, S1–S4). After 40 minutes of incubation with 500 μM GTP, ^tzGTP or ^thGTP at 37 °C, c-di-GMP, c-di-^tzGMP and c-di-^thGMP were obtained in 94%, 81% and 11% yields, respectively. DncV also produced the mixed c-G^tzGMP when incubated with 500 μM each of GTP and ^tzGTP (Figure S5a). When incubated with a mixture of 500 μM of GTP and ^thGTP, DncV also produces all three plausible products c-diGMP, c-G^thGMP and c-di-^thGMP (Figure S5b). Based on HPLC analyses shown in Figure 3, the overall yields of c-di-GMP (40 mins), c-di-^tzGMP (80 mins), and c-di-^thGMP (300 mins) were 94%, 85% and 62%, respectively. These results have indeed confirmed the tolerance level of DncV to ^thGTP and ^tzGTP, our emissive NTPs. Both symmetric and mixed CDN analogues were successfully synthesized in this fashion (Figure 2).

Figure 2.

Structures of the enzymatically synthesized c-di-GMP analogues. DncV is able to convert two NTPs into the corresponding homo- and heterocyclic dinucleotides. B₁, B₂ stand for nucleobases.

Figure 3.

DncV-mediated enzymatic synthesis of c-di-GMP and its analogues c-di-^tzGMP and c-di-^thGMP. (a) UV-monitored HPLC chromatograms (260 nm) of the DncV-mediated synthesis of c-di-GMP; (b) HR-MS of the intermediates from CIAP-treated reaction; (c) Kinetic analysis of the HPLC-integrated relative concentration and fitted curve of the starting materials (measured as nucleosides), products and intermediates for the DncV-mediated reactions of GTP; (d)–(f) HPLC chromatograms (333 nm), HR-MS of the intermediate, and kinetic analysis of DncV-mediated synthesis of c-di-^tzGMP; (g)–(i) HPLC chromatograms (321 nm), HR-MS of the intermediate, and kinetic analysis of DncV-mediated synthesis of c-di-^thGMP. Aliquots were treated with calf intestinal alkaline phosphatase (CIAP) at designated times, therefore the starting materials were presented as G, ^tzG and ^thG, and intermediates were presented as GpG, ^tzGp^tzG and ^thGp^thG. Assays were done in duplicates. Error bars indicate standard deviation (SD).

DncV generates CDNs through a series of sequential reactions, which are thought to involve a release-and-rebound process of the intermediate.^{[7b, 17]} After the first 3’,5’-phosphordiester bond formation, the linear dinucleotide pppNpN is released and oppositely rebound to the enzyme, after which the second 3’,5’-phosphodiester linkage is made.^{[7b, 16b, 18]} To ultimately apply and interpret real-time fluorescence measurements, a better understanding of the enzymatic conversion of GTP and its analogues into the corresponding CDNs is therefore required. The reaction time course was consequently analyzed, paying particular attention to the accumulation and consumption of intermediates. DncV was incubated with GTP, ^tzGTP and ^thGTP, and the reactions containing each of the corresponding substrates were quenched with calf intestinal alkaline phosphatase (CIAP) at designated time points.^[19] The relative concentrations of the starting material (S), uncyclized intermediate (I) and product (P) were then monitored by HPLC. The integrated area under the peak for each species was corrected using the corresponding extinction coefficient (Methods and Supporting experiments sections in SI) and normalized to its relative concentration, which were then plotted against time. Taking into account the unique release-andrebound process of the uncyclized dinucleotide intermediate,^{[16b, 20]} the kinetics of the DncV-mediated CDN syntheses was therefore analyzed according to the model in Scheme 1a and the differential equations (Eq 1–4), where S represents the starting NTP, I₁ and I₂ respectively represent intermediates pppNpN and degraded intermediate pNpN (I= I₁ + I₂), and P represents the product CDN:

$$\text{d}[\text{S}]/\text{dt}=-2k_{\mathit{1}}{[\text{S}]}^2$$ (1)

$$\text{d}\left[{\text{I}}_1\right]/\text{dt}=k_{\mathit{1}}{[\text{S}]}^2-k_{\mathit{2}}[1/1]-k_{\mathit{3}}\left[{\text{l}}_1\right]$$ (2)

$$\text{d}[\text{P}]/\text{dt}=k_{\mathit{2}}\left[{\text{I}}_1\right]$$ (3)

$$\text{d}\left[{\text{I}}_2\right]/\text{dt}=k_{\mathit{3}}\left[{\text{I}}_1\right]$$ (4)

A small amount of uncyclized intermediate was observed during the synthesis of c-di-GMP (Figure 3a–c), and the calculated k_1, k₂ and k₃ values were found to be (7.1±2.3)×10⁻⁶ μM⁻¹s⁻¹, (7.4±2.1)×10⁻² and (6.4±0.3)×10⁻⁴ s⁻¹, respectively (Table 1). When incubated with ^tzGTP, a larger amount of the intermediate was accumulated in the first 10 minutes resulting in an S-shaped product formation curve (Figure 3d–f).^[21] The calculated k₁, k₂ and k₃ values were (1.26±0.07)×10⁻⁶ μM⁻¹s⁻¹, (3.4±0.2)×10⁻³ and (1.3±0.1)×10⁻⁴ s⁻¹, respectively. DncV was also found to convert ^thGTP into the corresponding c-di- ^thGMP with k₁, k₂ and k₃ values of (3.08±0.03)×10⁻⁷ μM⁻¹s⁻¹, (2.32±0.05)×10⁻⁴ and (2.91±0.01)×10⁻⁵ s⁻¹, respectively (Figure 3g–i). After 5 h of incubation, 62% of ^thGTP was converted to c-di-^thGMP. Obvious accumulation of the reaction intermediate is seen for the first 80 minutes (Figure 3g, 3i). The relatively fast degradation of the intermediate compared to product formation (k₂/k₃ was calculated to be 8.0 for c-di-^thGMP synthesis, compared to 26 for c-di-^tzGMP and 116 for c-di-GMP) resulted in relatively high concentration (13% after 300 min) of persistent uncyclized intermediates (Figure 3i).done in duplicates. Error bars indicate standard deviation (SD).

Scheme 1.

Kinetics schemes for DncV-mediated CDN synthesis. S represents the starting NTP, I₁ and I₂ respectively represent intermediates pppNpN and degraded intermediate pNpN (I= I₁ + I₂), and P represents the product CDN. P_app (apparent product) stands for the sum of all the dinucleotide species, including linear ones and cyclized products.

Table 1.

Reaction rate constants of DncV-mediated CDN syntheses^[a]

	k1 (μM−1S−1)	k2 (s−1)	k3 (s−1)
c-di-GMP	(7.1 ±2.3) ×10−6	(7.4±2.1) ×10−2	(6.4±0.3) ×10−4
c-di-tzGMP	(1.26±0.07) ×10−6	(3.4±0.2) ×10−3	(1.3±0.1) ×10−4
c-di-thGMP	(3.08±0.03) ×10−7	(2.32±0.05) ×10−4	(2.91±0.01) ×10−5

^[a]Data were presented as mean ± SD. See Methods in SI for details.

The guanosine surrogates used here illuminate the key functional elements in the purine scaffold that affect the formation and consumption of the reaction intermediate. The kinetic constants listed in Table 1 illustrate that the formation of the first phosphodiester linkage is the rate-limiting step for the syntheses of c-di-GMP, c-di-^tzGMP and c-di-^thGMP.^[22] Furthermore, a certain fraction of the uncyclized intermediate is not entirely consumed in all three reactions (Figure 3c, 3f, 3i). We speculate that the hydrolysis of the open intermediate pppNpN (I₁ in Scheme 1a) to the unreactive pNpN (I₂ in Scheme 1a) could take place.^[23] Additionally, the consumption of ^thGTP in c-di-^thGMP synthesis was observed to be slower than calculated. We speculate that this might be caused by a non-productive DncV-mediated hydrolysis of ^thGTP to ^thG monophosphate (p^thG), circumventing the formation of the inter-nucleotide phosphodiester bond. Our findings support the hypothesis that the absence of the purine’s N-7 can alter the reaction kinetics, although previous structural studies with DncV have not revealed a direct contact between N-7 of GTP and any protein residues.^{[16b, 18, 20]}.

Photophysical properties of emissive c-di-GMP analogues.

Steady state absorption and emission measurements show hypsochromic shifts in the absorption maxima of ^thG and ^tzG upon incorporation into the corresponding cyclic dinucleotides (Figure 4, Table 2). While relatively small (< 10 nm), this trend is suggestive of multichromophoric arrangements reminiscent of Haggregates.^[24] The emission energy of the ^thG- and ^tzG-containing CDNs remains relatively close to that of the parent nucleosides and, intriguingly, the ^thG-containing derivatives still display the high energy shoulder associated with its tautomeric forms.^[25] Comparing the emission quantum yields (Φ) of the fluorescent CDNs to those of the monomeric nucleosides show significant quenching of the former, as predicted (Table 2). Interestingly, ^thG exhibits a significantly enhanced self-quenching effect relative to guanosine; the emission quantum yield values for c-di-^thGMP and c-G^thGMP were about 17% and 39% of Φ^th_G, respectively. The self-quenching effect of ^tzG resulted in the low quantum yield for c-di-^tzGMP (16% of Φ^tz_G), which is comparable to that seen for c-G^tzGMP (19% of Φ^tz_G). These results match the model illustrated in Figure 1, where decreased emission intensity was expected upon the conversion of the emissive NTPs into CDNs. Further analysis below provides additional insight into this model and its potential applications.

Figure 4.

Absorption spectra (a) and emission spectra (b) of ^thG (black), ^tzG (grey), c-di-^thGMP (red), c-di-^tzGMP (indigo), c-G^thGMP (orange) c-G^tzGMP (light blue) dissolved in water. The emission spectra were normalized to optical density of 0.1 at the excitation wavelengths (320 and 330 nm for species containing ^thG and ^tzG, respectively).

Table 2.

Photophysical properties of emissive nucleosides and CDN analogues^[a]

	λabsmax (nm)	ε (M−1 cm−1)	λemmax (nm)	ϕ	ϕε
thGM[a]	321	4.15×103	453	4.6×10−1	1909
tzG[a]	333	4.87×103	459	2.5×10−1	1203
c-di-thGMP	317	7.47×103	457	7.7×10−2	575
c-di-tzGMP	331	8.77×103	456	3.9×10−2	342
c-GthGMP	317	3.54×103	458	1.8×10−1	636
c-GtzGMP	331	4.38×103	456	4.8×10−2	210

^[a]Values for ^thG and ^tzG were obtained from previous publications.^[11]

Monitoring DncV-mediated synthesis of c-di-GMP analogues with fluorescence.

The significant difference in emission seen for nucleotides compared to the corresponding CDNs, can be exploited for monitoring the enzymatic transformations. The bacterial enzyme DncV was thus incubated with either GTP, ^thGTP, ^tzGTP or mixtures thereof under the same conditions as the HPLC-monitored reactions. Aliquots were treated with calf intestinal alkaline phosphatase (CIAP) at designated times, and emission spectra were taken after appropriate dilution (see Methods in SI for details). Rewardingly, significantly diminished fluorescence intensities were observed for the syntheses of all four homodimeric and mixed fluorescent c-di-GMP analogs, as hypothesized (Figure 5).

Figure 5.

DncV-mediated cyclization monitored by emission spectra for (a) c-di-^tzGMP, (b) c-G^tzGMP, (c) c-di-^thGMP and (d) c-G^thGMP. Excitation wavelength was 380 nm for all emission spectra.

Unlike integrate chromatographic analyses, which can separately account for all individual species present, the observed fluorescence signal reflects the sum of all emissive species. To facilitate quantitative analyses of fluorescence data, a conversion factor (a) was introduced to the simulation of the fluorescence-monitored reaction kinetics. This factor bridges the integrated fluorescent spectrum and the concentration of a given chromophore at a given set of conditions and can be calculated from the integrated emission spectrum (FL_int) of a given chromophore with known concentration ([C]) using Eq 5. Since photophysical properties can be affected by reaction conditions, such as pH, ionic strength and temperature, a values are measured under identical conditions as the enzymatic reactions (see Table S1 for a values for all emissive CDN analogues). To compare the fluorescence- and HPLC-monitored reaction kinetics, Eq 6 was used to correlate the recorded fluorescence spectrum with the concentrations of different species in the reaction mixture. As a first attempt, the k values extracted from the HPLC analysis were used to fit the fluorescence data (referred to as FL Model 1).

$${\text{FL}}_{\text{int}}=a[\text{C}]$$ (5)

$${\text{FL}}_{\text{int}}=a_{\mathit{1}}[\text{S}]+a_{\mathit{2}}\left[{\text{I}}_1\right]+a_{\mathit{3}}[\text{P}]+a_{\mathit{4}}\left[{\text{I}}_2\right]$$ (6)

Whereas a₁ and a₃ could be successfully measured for the starting materials and final products, a₂ and a₄ are difficult to obtain, since the intermediates are produced in small quantities. We therefore first assume that a₂ ≈ a₃ ≈ a₄ in Eq 6. The calculated curve fitted well to the experimental data for c-di-^tzGMP synthesis with R² = 0.976, indicating excellent agreement between the reactions monitored by HPLC and fluorescence (Figure 6a). For c-di-^thGMP synthesis, the observed fluorescence signal decreased faster than modeled for the first 40 minutes (Figure 6b), which resulted in a poorer fit (R² =0.833). In addition to the previously discussed reasons, potentially leading to poorly simulated curves for this HPLC analyzed reaction (Figure 3i), the significant accumulation of intermediates (p^thGp^thG or ppp^thGp^thG) during the first 80 reaction minutes, hampers the simulation of its fluorescence response. Possible differences between the photophysical properties of the intermediates and the product are thus more influential on the analysis of c-di-^thGMP synthesis as our assumption of a₂ ≈ a₃ ≈ a₄ in Eq 6 is challenged.

Figure 6.

Kinetics analysis of DncV-mediated synthesis of (a) c-di-^tzGMP (R² = 0.9759, 0.995 for FL model 1 and FL model 2, respectively) and (b) c-di^thGMP (R² = 0.8325, 0.9205 for FL model 1 and FL model 2, respectively).

In addition to using chromatographically determined k values (Scheme 1a), we also simplified the analysis of the homodimeric CDNs synthesis to a pseudo-second order reaction (Scheme 1b). This approach relies on the observation that the intermediates are present in relatively low concentrations (and hence contribute less to the overall fluorescence signal). The apparent kinetics rate constant (k_app) was extracted with equations 7–9 (referred to as FL Model 2).

$$\text{d}[\text{S}]/\text{dt}=-2k_{app}{[\text{S}]}^2$$ (7)

$$\text{d}\left[{\text{P}}_{\text{app}}\right]/\text{dt}=k_{app}{[\text{S}]}^2$$ (8)

$${\text{FL}}_{\text{int}}=a_{\mathit{1}}[\text{S}]+a_{\mathit{3}}\left[{\text{P}}_{\text{app}}\right]$$ (9)

Based on our hypothesis, k_app in Scheme 1b should be comparable to k₁ in Scheme 1a, as the major contributor to the change in fluorescence signal is the formation of the first intermediate (I₁). Rewardingly, using FL Model 2, the simulated curve fits well the experimental data for c-di-^tzGMP with R² =0.995, and derived k_app=1.57±0.15 M⁻¹ s⁻¹, which is close to the k₁ derived from HPLC analysis (1.26±0.07 M⁻¹ s^{− 1}) (Figure 6a, Table S2). Similar to FL model 1, the simulation curve did not fit as well for the synthesis of c-di-^thGMP (R² = 0.921, Figure 6b). The k_app value derived from the pseudo-second order simulation was slightly larger than the k₁ values derived from HPLC analysis (0.530±0.05 and 0.308±0.03 M⁻¹ s⁻¹, respectively), but overall the two k values were still comparable (Table S3). The accumulation of intermediates during c-di-^thGMP synthesis makes the simplification to a pseudo-second order reaction challenging. Adjusting the apparent rate constant yields only minor improvement for the simulation. Nevertheless, these observations reflect our previous observations, highlighting the higher isofunctionality of the isothiazolo family of purine surrogates compared to the thiopheno family.^{[12d, 26]}

RocR-mediated CDN hydrolysis monitored by HPLC.

Beside the cyclases that synthesize c-di-GMP, its hydrolysis to linear pGpG or GMP by PDEs is key to controlling the global and local concentration of such messengers in bacteria and hence to regulating downstream processes.^{[1a, 27]} rocR is a PDE that contains an EAL domain and specifically recognizes and cleaves c-di-GMP into a dinucleotide monophosphate pGpG.^[28] It is one of P. aeruginosa’s most active and well-studied PDEs.^{[28a, 29]} To shed light on the substrate–enzyme interactions and its suitability for fluorescence monitoring, enzymatically synthesized c-di-GMP, c-di-^tzGMP, c-di-^thGMP, c-G^thGMP and c-G^tzGMP were incubated with rocR, and the relative concentrations of starting material (CDN) and product (pNpN) at designated time points were monitored by HPLC (Figure S6). Notably, the assigned identity of the products was confirmed by LC-ESI-TOFMS (Figure S7–S12).

RocR is reported to follow Michaelis-Menten kinetics.^{[27c, 28a]} Scheme 2a was therefore used to analyze the reaction kinetics, where k₁ and k₋₁ describe the enzyme/substrate association and dissociation, respectively, while k₂ reflects the cleavage reaction. Equations 10–13 were thus used to extract the rate constants. When “asymmetrical” mixed c-di-GMP analogues are treated with rocR, the enzyme may, in principle, recognize and cleave either phosphodiester bonds, producing pN₁pN₂ and pN₂pN₁. Assuming the two products result from different binding orientation of the heterodimeric CDNs, Scheme 2b is therefore introduced to model the cleavage reactions, where k₁, k₋₁ and k₂ reflect the association/dissociation and cleavage of the heterodimeric CDN in one orientation (ES1), respectively, and k₃, k₋₃ and k₄ reflect the other (ES2). Equations 14–19 were used to extract the rate constants.

Scheme 2.

Kinetics scheme of rocR-mediated hydrolysis of CDNs. E stands for the enzyme S stands for the starting CDN, and P stands for the product. (a) Kinetics scheme of rocR-mediated hydrolysis of c-di-GMP, c-di-^tzGMP and c-di-^thGMP. (b) Kinetics scheme of rocR-mediated hydrolysis of mixed analogues, including c-G^tzGMP and c-G^thGMP.

c-di-GMP was completely cleaved by rocR to pGpG in nearly 3 minutes with k₁ = 3.96 ± 0.07 μM⁻¹s⁻¹ and k₂ = 0.76 ± 0.03 s⁻¹ (Figure 7a, Table 3). Enzymatic cleavage of c-di-^tzGMP was slower compared to c-di-GMP, but was completed within 20 min (Figure 7b) with k₁ = 0.043 ± 0.02 μM⁻¹s⁻¹ and k₂ = 0.83 ± 0.55 s⁻¹ (Table 3). No cleaved c-di-^thGMP was observed after 40 min of incubation with rocR (Figure 7c). The rocR-mediated cleavage of the mixed c-G^thGMP yielded a single product (Figure 7d), either pGp^thG or p^thGpG, while two products in nearly the same amount were observed for cG^tzGMP (Figure S6), illustrating that both pGp^tzG and p^tzGpG were produced (Figure 7e). Overall, a gradually reduced hydrolysis rate was observed as the structures progressively deviated from the parent CDN in the order: c-di-GMP, c-di^tzGMP, c-G^tzGMP, c-G^thGMP and lastly c-di-^thGMP (Figure 7f).

Figure 7.

Kinetics analyses of rocR-mediated cleavage of c-di-GMP analogues. The enzyme (100 nM) was incubated with 10 μM of (a) c-di-GMP, (b) c-di-^tzGMP, (c) c-di-^thGMP, (d) c-G^thGMP and (e) c-G^tzGMP and the reactions were quenched using 100 mM CaCl₂ at designated time point and analyzed by HPLC. Assays were done in duplicates. Error bars indicate SD.

Table 3.

Reaction rate constants for rocR-mediated CDNs cleavage^[a]

	k1 (μM−1s−1)	k−1 (s−1)	k2 (s−1)	k3 (μM−1s−1)	k−3 (s−1)	k4 (s−1)
c-di-GMP	3.96 ± 0.07	1.15 ± 0.21	0.76 ± 0.03	NA	NA	NA
c-di-tzGMP	0.043 ± 0.02	0.09n± 0.07	0.83 ± 0.55	NA	NA	NA
c-GthGMP	0.021 ± 0.001	0.033 ± 0.001	5 ± 0.57	NA	NA	NA
c-GtzGMP	1.85 ± 0.07	0.096 ± 0.091	3.07 ± 0.25	1.8 ± 0.1	0.135 ± 0.007	0.34 ± 0.01

^[a]Data were presented as mean ± SD.

$$\text{d}[\text{S}]/\text{dt}=-k_{\mathit{1}}[\text{E}][\text{S}]+k_{-\mathit{1}}[\text{ES}]$$ (10)

$$\text{d}[\text{E}]/\text{dt}=-k_{\mathit{1}}[\text{E}][\text{S}]+k_{-\mathit{1}}[\text{ES}]+k_{\mathit{2}}[\text{ES}]$$ (11)

$$\text{d}[\text{P}]/\text{dt}=k_{\mathit{2}}[\text{ES}]$$ (12)

$$\text{d}[\text{ES}]/\text{dt}=k_{\mathit{1}}[\text{E}][\text{S}]-k_{-\mathit{1}}[\text{ES}]-k_{\mathit{2}}[\text{ES}]$$ (13)

$$\text{d}[\text{S}]/\text{dt}=-k_{\mathit{1}}[\text{E}][\text{S}]+k_{-\mathit{1}}\left[{\text{ES}}_1\right]-k_{\mathit{3}}[\text{E}][\text{S}]+k_{-\mathit{3}}\left[{\text{ES}}_2\right]$$ (14)

$$\text{d}\left[{\text{P}}_1\right]/\text{dt}=k_{\mathit{2}}\left[{\text{ES}}_1\right]$$ (15)

$$\text{d}\left[{\text{P}}_2\right]/\text{dt}=k_{\mathit{4}}\left[{\text{ES}}_2\right]$$ (16)

$$\text{d}[\text{E}]/\text{dt}=-\left(k_{\mathit{1}}+k_{\mathit{3}}\right)[\text{E}][\text{S}]+\left(k_{-\mathit{1}}+k_{\mathit{2}}\right)\left[{\text{ES}}_1\right]+(k_{-\mathit{3}}+k_{\mathit{4}})\left[{\text{ES}}_2\right]$$ (17)

$$\text{d}\left[{\text{ES}}_1\right]/\text{dt}=k_{\mathit{1}}[\text{E}][\text{S}]-k_{-\mathit{1}}\left[{\text{ES}}_1\right]-k_{\mathit{2}}\left[{\text{ES}}_1\right]$$ (18)

$$\text{d}\left[{\text{ES}}_2\right]/\text{dt}=k_{\mathit{3}}[\text{E}][\text{S}]-k_{-\mathit{3}}\left[{\text{ES}}_2\right]-k_{\mathit{4}}\left[{\text{ES}}_2\right]$$ (19)

The observed relative rates of rocR-mediated hydrolyses of the CDN analogues indicate that the presence of a nitrogen at the purine’s N-7 position on at least one nucleobase is necessary for efficient substrate recognition and cleavage. The enzyme did not produce observable amounts of cleaved c-di-^thGMP products after 40 min, while the majority of c-di-^tzGMP was found to be cleaved within 20 min, indicating that altering both N-7 positions is likely detrimental to rocR-mediated hydrolytic cleavage. With a single N-7-containing nucleobase such as in the mixed c-G^thGMP, only one phosphodiester bond is cleaved by rocR. In contrast, rocR was able to recognize the mixed c-G^tzGMP, where donor nitrogens are present on both nucleobases, from both orientations, leading to cleavage of either phosphodiester bond and the release of pGp^tzG and p^tzGpG (Figure 7e).

Crystal structure of rocR with a bound ligand has not yet been reported, thus systematically modified CDN analogues as studied here can illustrate the importance of the nucleobases in substrate recognition and cleavage. The structure of Ykul-bound c-di-GMP shows the amide group of the highly conserved Q16 (similar to Q161 of rocR) to be hydrogen bonded to the N-7 of the guanosine found 5’ to the cleavage site^{[27c, 29–30]}. Indeed, mutation of rocR’s Q161 caused a 5-fold decrease in k_cat, and 2-fold increase in K_M, indicating that Q161 is involved in substrate recognition.^[27c] We thus conjecture that the cleavage product of c-G^thGMP is pGp^thG and submit that there is little bias in rocR binding/cleaving of c-G^tzGMP in either orientation, as k₁ ≈ k_3, and k₋₁ ≈ k₋₃ (Table 3). The difference in the final percentage of the two products P₁ (58 %) and P₂ (42 %) (Scheme 2b) might be caused by different efficiencies for the phosphodiester bond cleavage, as k₂ (3.07 ± 0.25 s⁻¹) is much bigger than k₄ (0.34 ± 0.01s⁻¹).

Monitoring CDN hydrolysis with fluorescence.

To fluorescently monitor the rocR-mediated CDN hydrolysis, the enzymatic reactions with c-G^thGMP, c-G^tzGMP and c-di-^tzGMP were executed in cuvettes under the same conditions as the reactions monitored with HPLC, and emission spectra were taken at designated time points. As shown in Figures 8 and S13, significant decrease of emission was observed during c-G^thGMP and c-G^tzGMP hydrolysis. The fitted curves were generated with the same rate constants derived from HPLC analysis. In addition to Equations10–19, Equations 20 and 21 were used to correlate fluorescence signal with concentration for reactions described in scheme 2a and 2b, respectively.^[31] The fluorescence conversion factors (a) are listed in Table S4. The resulting R² values for c-G^thGMP and c-G^tzGMP hydrolysis monitored by fluorescence were 0.986 and 0.916, respectively. On the other hand, rocR-mediated c-di-^tzGMP hydrolysis did not lead to significant change in emission intensity (Figure S13c). It is possible that p^tzGp^tzG and c-di-^tzGMP have similar photophysical properties, suggesting that our assumption of a₂ ≈ a₃ ≈ a₄ in Eq 6 when monitoring CDN synthesis with fluorescence stands well.

Figure 8.

rocR-mediated CDN hydrolysis monitored with fluorescence. a) rocR-mediated c-G^thGMP hydrolysis monitored with steady-state emission spectra. b) rocR-mediated c-G^tzGMP hydrolysis monitored with steady-state emission spectra. The y-scale was integrated emission intensity (area under the curve). Assays were done in triplicates or duplicates (error bars, indicating SD, were smaller than the points shown).

The relatively small differences in emission intensities of the CDNs and the corresponding rocR-cleaved linear products, reflects the complexity in foreseeing how changes in molecular structure impact the photophysical features. The quenching effect of fluorescent nucleosides in CDNs and their degradation products is indeed dependent on several factors associated with their building blocks, including their stacking distance and orientation, as well as their collisional dynamics and possibly aggregation. To further test the proposed model shown in Figure 1, we spectroscopically assessed the complete cycle of CDN synthesis and hydrolysis.

$${\text{FL}}_{\text{int}}=a_1\left[\text{S}\right]+a_2\left[\text{P}\right]$$ (20)

$${\text{FL}}_{\text{int}}=a_1[\text{S}]+a_2\left(\left[{\text{P}}_1\right]+\left[{\text{P}}_2\right]\right)$$ (21)

The fluorescent NTPs were thus incubated with DncV and quenched with CIAP at designated time points (under the same conditions as DncV-mediated synthesis above). The samples were then digested with S1 nuclease to yield the free nucleosides.^[32] This hydrolysis process was monitored by fluorescence spectroscopy. Similar to the results described above, the fluorescence intensity decreased upon incubation with DncV for both ^tzGTP (dashed or solid black lines in Figure 9a–c) and ^thGTP (dashed or solid blank lines Figure 9d–f). Increasing fluorescence intensity was then observed upon administration of S1 nuclease to the DncV-mediated reaction mixtures, and the emission intensity at S1 endpoint is comparable to that of DncV at time 0 (colored lines in Figure 9), as one would predict based on the model presented in Figure 1. The same trends were observed for samples incubated with DncV for different amount of time (Figure 9c, f). The recovery of emission after S1 nuclease mediated hydrolysis reactions further supports our model, suggesting that the fluorescence quenching observed during DncV-mediated CDN syntheses is caused by the spatial proximity of chromophores. While unlikely under our reaction conditions, intermolecular aromatic interactions could also contribute under certain circumstances, as CDNs have been reported to equilibrate between monomeric and intercalated dimeric structures.^[27a]

Figure 9.

DncV-mediated synthesis and S1 nuclease-mediated hydrolysis of CDNs monitored with emission spectra.

Conclusions

Since the discovery of c-di-GMP in 1987, the landscape representing the signaling mechanisms and biological significance of CDNs has continued to expand.^{[1a, 5b, 27a, 33]} As second messengers, the intracellular concentration of CDNs is tightly modulated by multiple dinucleotide cyclases and phosphodiesterases (PDEs), which further impact bacterial homeostasis and virulence through CDN-regulated signaling machineries, including riboswitches and protein receptors.^{[5b, 27a]} In this study we have demonstrated the utility of novel isomorphic fluorescent analogues of this key bacterial messenger for monitoring the activity of a cyclase and both specific and non-specific phosphodiesterases in real-time. A comparison to traditional methods (e.g., HPLC) shows the fluorescence-based approach is reliable and much faster, making it amenable to further optimization and potential applications requiring higher throughput. The subtle structural differences between G and its emissive surrogates ^thG and ^tzG, provided insight into the molecular signatures governing the enzyme–substrate recognition.

Previous structural studies have illustrated that DncV belongs to a large family of nucleotidyltransferases (CD-NTases) that is responsible of synthesizing various types of CDNs as well as cyclic trinucleotides.^[6] Recent reports have indicated that theses CD-NTases and their products might play distinctive roles in host-pathogen interactions, though detailed analyses are needed to further illustrate their function and signaling mechanism.^[6] The isomorphic fluorescent NTPs and their CDNs described in this contribution could serve as powerful tools for the study such processes in vitro, shed light on their recognition features and accelerate the fabrication of high throughput discovery assays for agonists and antagonists.