Abstract
We previously reported the successful design, synthesis and testing of the prototype opioid painkiller NFEPP that does not elicit adverse side effects. The design process of NFEPP was based on mathematical modelling of extracellular interactions between G-protein coupled receptors (GPCRs) and ligands, recognizing that GPCRs function differently under pathological versus healthy conditions. We now present an additional and novel stochastic model of GPCR function that includes intracellular dissociation of G-protein subunits and modulation of plasma membrane calcium channels and their dependence on parameters of inflamed and healthy tissue (pH, radicals). The model is validated against in vitro experimental data for the ligands NFEPP and fentanyl at different pH values and radical concentrations. We observe markedly reduced binding affinity and calcium channel inhibition for NFEPP at normal pH compared to lower pH, in contrast to the effect of fentanyl. For increasing radical concentrations, we find enhanced constitutive G-protein activation but reduced ligand binding affinity. Assessing the different effects, the results suggest that, compared to radicals, low pH is a more important determinant of overall GPCR function in an inflamed environment. Future drug design efforts should take this into account.
Similar content being viewed by others
Introduction
The family of G-protein coupled receptors (GPCRs) represents the largest class of receptors in the human genome and some of the most common drug targets. Located on the cell membrane, they transduce extracellular signals into key physiological effects. Natural GPCR ligands include neurotransmitters, chemokines, hormones, odours or photons. GPCRs are involved in a large number of disorders, such as diabetes, high blood pressure, depression, addiction, pain, arthritis, Parkinson’s and many others1. A prominent member of this family is the \(\mu\)-opioid receptor (MOR). It binds endogenous opioid peptides (e.g. endorphins, enkephalins) as well as exogenous drugs (e.g. morphine, fentanyl), and its activation results in the modulation of intracellular G-proteins and second messengers (e.g. cAMP, ion channels). MOR plays major roles in analgesia (inhibition of pain), addiction, bowel movement, arousal, respiration and other physiological functions2.
Recent works of our group3 led to the development of the novel analgesic compound N-(3-fluoro-1-phenethylpiperidin-4-yl)-N-phenylpropionamide (NFEPP) which activates the MOR preferentially at acidic extracellular pH-levels, as given in injured tissues2. This is of utmost interest because it may preclude the adverse effects of conventional MOR agonists like fentanyl which include constipation, sedation and apnea. These adverse effects are mediated mostly in the brain and the gut, i.e. in healthy tissues (pH 7.4). Since the generation of pain can be effectively inhibited by blocking the electrical excitation of sensory neurons at the site of the injury (i.e. the origin of nociceptive stimulation), this gives rise to the hope that NFEPP might have less or even no adverse effects, which could already be corroborated in animal studies3,4,5,6.
Up to now, the effects of NFEPP and fentanyl were mathematically analysed at the level of their binding rates at relevant amino acid residues accessible from the extracellular side of MOR3,7. To get a more complete picture, we herein present a model of the intracellular second messenger pathways relevant to pain and analgesia. The mechanism underlying the analgesic effect of MOR activation in nociceptive neurons is mainly due to a stabilisation or even lowering of the plasma membrane potential beneath the threshold value required to elicit an action potential2,8. This effect is mediated via intracellular inhibitory G-proteins, which dissociate into \(\alpha\)- and \(\beta \gamma\)-subunits after formation of a receptor–ligand complex9. Among other actions, the \(\beta \gamma\)-subunits bind to calcium channels in the plasma membrane. This leads to closure of the channels, thereby lowering the amount of positive calcium-ion influx and reducing cellular excitability2,8,10,11.
In this paper, we model this pathway to analyse the in vitro effects of fentanyl and NFEPP on the number of closed membrane calcium channels and activated (i.e. dissociated) G-protein complexes at different pH-levels in cultured cells and sensory neurons to investigate basic mechanisms underlying opioid analgesia. We propose a reaction network that connects the receptor–ligand interactions to the G-protein cycle, and further to the signalling cycle of calcium channel opening and closing (see Fig. 1 for an illustration). After careful parametrization and validation of the model using Förster resonance energy transfer (FRET) data from corresponding in vitro experiments, the stochastic reaction process was simulated for different values of the receptor–ligand binding rate, comparing the mean inhibition of calcium currents resulting from these numerical simulations to additional data from in vitro patch clamp experiments. By numerical simulation of the reaction network, we observe that the binding rate has a non-linear effect onto the mean amplitude of deactivated calcium channels, which explains the different effects of NFEPP and fentanyl in inflamed versus healthy environments.
It is important to note that our approach differs from others that have investigated signalling pathways from receptor to the nucleus or to intracellular second messengers (not to the plasma membrane)12,13. In contrast to those studies, we choose a stochastic approach because it delivers more information than deterministic alternatives.
Aside from pH, other inflammatory mediators play important roles. For example, reactive oxygen species (radicals) can modulate the function of GPCRs14,15,16,17. In order to understand the interplay between pH and additional radicals for the signalling, we modelled different scenarios and performed in vitro experiments. In the precense of radicals, G-protein activation can be initiated in the absence of an opioid ligand (so-called constitutive G-protein activation). Motivated by this observation, we included the reaction of constitutive G-protein dissociation in our network.
Altogether, the resulting model allows to study two different inflammatory conditions:
-
(a)
influence of pH, and
-
(b)
influence of radicals.
With regard to (a), lower pH value changes the protonation state of amino acid residues and opioid ligands, and we investigate whether this affects the binding rates and subsequent modulation of calcium channels. Concerning (b), we study whether an increased concentration of radicals may have an effect on the binding affinity of ligands and/or increase the probability for constitutive G-protein dissociation.
A crucial advantage of the model is that it also permits to analyse the combined effects of (1) and (2), and our results suggest that, compared to radicals, low pH is a more important determinant of overall GPCR function in an inflamed environment.
Unlike in our previous work, we here studied these effects at the systems biology instead of molecular level: results from reaction network simulation are integrated with data from in vitro experiments in order to analyse the consequences of environment-dependent ligand binding rates onto the downstream signalling, i.e. calcium channel inhibition.
Models and methods
In this section we introduce a probabilistic model for the signalling pathway from receptor activation over the G-protein cycle to the calcium channel inhibition. We explain the in vitro experiments which were performed to validate the modelling results and estimate the parameter values. Moreover, we motivate our choice of parameter values and specify the numerical approach used for solving the system.
The reaction network
The biochemical reaction network under consideration consists of the following reactions (see Table 1 for an overview and Fig. 1 for an illustration). A ligand L attaches to a receptor R in the membrane, resulting in a receptor–ligand complex RL (reaction \({\mathcal {R}}_1\)). This binding process is dependent on the concentration of protons (i.e. pH) in the microenvironment of the receptor. The protonation of both the ligand and certain residues in the receptor are important determinants for receptor activation, likely due to the pH-dependent formation of hydrogen bonds and/or salt bridges between ligand and receptor18,19.
This receptor–ligand complex RL activates a trimeric G-protein complex which leads to exchange of GDP by GTP and subsequent dissociation into \(\alpha\)- and \(\beta \gamma\)-subunits (reaction \({\mathcal {R}}_2\)). These subunits activate different signalling pathways. Along with the hydrolysis of GTP, another reaction partner M (e.g. arrestin) emerges (reaction \({\mathcal {R}}_3\)), which initiates internalisation of the receptor–ligand complex (reaction \({\mathcal {R}}_4\)). The \(\beta \gamma\)-subunit inhibits a membrane calcium channel by binding to it (reaction \({\mathcal {R}}_5\)) (In other papers20 this is referred to as switching the calcium channel from the “willing” state to the “reluctant” state.). After dissociation of the \(\beta \gamma\)-subunit from the calcium channel, a trimeric G-protein complex is reformed, and the calcium channel is opened (reaction \({\mathcal {R}}_6\)). The internalised receptor \(RL_w\) is either recycled to the cell membrane (reaction \({\mathcal {R}}_7\)) or degraded (reaction \({\mathcal {R}}_8\)). The reaction partner M can itself be degraded (reaction \({\mathcal {R}}_9\)). The ligand L can vanish before it binds to the receptor, e.g. by degradation or unspecific binding to other extracellular components (reaction \({\mathcal {R}}_{10}\)), or it is degraded intracellularly (reactions \({\mathcal {R}}_7\) and \({\mathcal {R}}_8\)). The constitutive G-protein activation is given by reaction \({\mathcal {R}}_{11}\), where we simply use \({\mathcal {R}}_2\) without ligand. That is, we model the constitutive binding reaction as a net reaction consisting of first the switching of GTP and GDP, followed by the dissociation of the \(\beta \gamma\)-subunit from the GPCR complex.
Given these reactions, the stochastic dynamics of the system are mathematically modelled by a reaction jump process characterised by the chemical master equation21,22. The state of the system is given by a vector
counting the number \(x_S\) of molecules of the different species \(S\in {\mathcal {S}}\), where \({\mathcal {S}}\) is the set of species under consideration:
For each reaction \({\mathcal {R}}_j\) there is a stoichiometric vector \(\varvec{\nu }_j\in {\mathbb {Z}}^{11}\) defining the net change in the population state \({\varvec{x}}\) induced by this reaction. That is, each time that reaction \({\mathcal {R}}_j\) occurs, this leads to a jump in the system’s state of the form
E.g., the stoichiometric vector \(\varvec{\nu }_1\) of reaction \({\mathcal {R}}_1\) is given by \(\varvec{\nu }_1=(-1,-1,1,0,\ldots ,0)\). The rates at which the reactions occur are given by propensity functions \(f_j:{\mathbb {N}}_0^{11} \rightarrow [0,\infty )\), which can be found in the right column of Table 1.
The temporal evolution of the system is described by the Markov jump process \(({\varvec{X}}(t))_{t\ge 0}\), \({\varvec{X}}(t)=(X_S(t))_{S\in {\mathcal {S}}}\), where \(X_S(t)\) is the number of molecules of species S at time t. We define the probability \(p({\varvec{x}},t):={\mathbb {P}}({\varvec{X}}(t)={\varvec{x}}|{\varvec{X}}(0)={\varvec{x}}_0)\) to find the system in state \({\varvec{x}}\) at time t given some initial state \({\varvec{x}}_0\). Then, the overall dynamics are characterised by the standard chemical master equation given by
The reaction rate equation characterising the corresponding deterministic reaction system is given by the ordinary differential equation (ODE)
for concentrations \({\varvec{C}}(t)=\frac{1}{V}{\varvec{X}}(t)\), where V is the system’s volume. It is well-known that the volume-rescaled Markov jump process \(({\varvec{X}}(t)/V)_{0\le t\le T}\) governed by the chemical master equation (1) converges to the solution \({\varvec{C}}(t)_{0\le t\le T}\) of the ODE system (2) in the limit of large particle numbers, i.e., for \(V\rightarrow \infty\)22.
Stochastic vs deterministic approach
The stochastic approach has several advantages over the deterministic one. At first, ODEs are an approximation assuming that the higher moments are trivially given by powers of the first moment. Stochastic modelling is exact in the sense that it takes into account all higher moments. Furthermore, the stochastic approach is closer to reality because it assumes a finite set (discrete number) of molecules, while ODEs consider concentrations and only work as approximations for large particle numbers. So the stochastic model is better suited for modelling a small compartment like an axon terminal with a small number of MORs and G-proteins. For our analysis, we will consider comparatively small numbers of molecules for all species (concretely, 20 MORs and 40 G-proteins, see Table 3), such that the stochastic approach is indispensable. Last but not least, a stochastic model delivers more information than ODEs. E.g., it enabled us to analyse the variances of the trajectories or the probability distribution of certain variables like the number of ligand–receptor binding events, which will be done in “Isolated impact of pH value”.
In many situations, however, the ODE model provides a valid approximation of the rescaled first moment of the stochastic process, \({\varvec{C}}(t)\approx {\mathbb {E}}({\varvec{X}}(t))/V\), as it is also the case here. This fact will be exploited in “Parameter estimation” where the less complex ODE model instead of the stochastic one will be used for estimating the reaction rates \(k_1,\ldots ,k_{10}\) based on experimental data.
Laboratory in vitro experiments
In order to validate our model, we performed laboratory experiments measuring G-protein activation and membrane calcium currents in vitro. To determine initial G-protein activation (as reflected by the exchange rate of GDP for GTP), the [\(^{35}\)S]-GTP\(\gamma\)S binding assay was used. Because these experiments require genetic alteration (by transfection) of cells, we performed these measurements in commonly used human embryonic kidney (HEK293) cells. In addition, we extracted data produced by FRET experiments3. These experiments measure ligand-induced G-protein subunit dissociation (which follows G-protein activation). The FRET experiments were used to fit the reaction rates. To mimic the mechanisms underlying in vivo opioid analgesia, we examined calcium currents in sensory neurons harvested from rodents using a patch-clamp protocol (see Supplementary Information for methodological details). The experimental results are shown in Figs. 2 and 5a below, and described in more detail in “Isolated impact of pH value”.
Parameter estimation
Our model includes eleven previously unknown parameters \(k_1,\ldots ,k_{11}\). The determination of appropriate values for these parameters included two main steps: a rough selection of values based on literature, followed by more precise standard parameter estimation.
In Ray et al.7, it has been shown by means of molecular dynamics (MD) simulations that the ligand binding affinity varies for different ligands and pH values. The protonation of both the ligand and certain residues in the receptor are important determinants for receptor activation, likely due to the pH-dependent formation of hydrogen bonds and/or salt bridges between ligand and receptor18,19. We used the relative changes of the rate constant as determined in Ray et al.7 and chose the following different values of \(k_1\) for different ligand/pH combinations: \(k_1=1.25\times 10^{-2}\,\text {s}^{-1}\) for fentanyl and pH 6.5, \(k_1=2.5\times 10^{-2}\,\text {s}^{-1}\) for fentanyl/pH 7.4, \(k_1=2.5\times 10^{-3}\,\text {s}^{-1}\) for NFEPP/pH 6.5, and \(k_1=5\times 10^{-4}\,\text {s}^{-1}\) for NFEPP/pH 7.4. As the in vitro experiments used for the parameter estimation were performed in the absence of radicals, the rate \(k_{11}\), which is responsible for the constitutive G-protein activation \({\mathcal {R}}_{11}\) , was set to zero, because constitutive activation probably does not play an important role in healthy tissue.
The parameters \(k_2,\ldots ,k_{10}\) of the other intracellular reactions were assumed to depend only mildly (if at all) on the ligand/pH combination in the cellular environment; for each of these parameters a single value has been chosen, independent of ligand and pH. This is a reasonable assumption because we chose an intracellular pH value of 7.4 based on well-known mechanisms of cellular homeostasis: although transient (several minutes) changes of intracellular pH may occur with tissue acidosis, intracellular buffer systems and ion pumps in the plasma membrane will rapidly restore physiological pH to ensure cell viability23. Since most previous studies examined situations of longer-lasting inflammation (up to several days)3,4,5,6, we look at this situation, as well. Using results from Zamponi et al.24, we started by setting the rate constant \(k_5\) of the central binding reaction between the \(\beta \gamma\)-subunit and the calcium channel to \(k_5=5\times 10^{-2}\,\text {s}^{-1}\) and proceeded to arrange the other values relative to it according to what is known in the literature. Comparing previous work24,25 it can be deduced that \({\mathcal {R}}_5\) happens at a timescale an order of magnitude shorter than \({\mathcal {R}}_2\) and \({\mathcal {R}}_3\) (with \({\mathcal {R}}_2\) being slightly faster than \({\mathcal {R}}_3\)), so we chose \(k_2\) and \(k_3\) five resp. ten times smaller than \(k_5\). The rate constants of reactions \({\mathcal {R}}_4,{\mathcal {R}}_6,{\mathcal {R}}_9\) were assumed to be of the same magnitude as those of \({\mathcal {R}}_2\) and \({\mathcal {R}}_3\) (with \({\mathcal {R}}_6\) being slightly faster). The recycling and degradation of internalised ligand–receptor complexes are much slower, at a level of minutes (Fig. 1 in Williams et al.26) which leads to comparatively small rate constants \(k_7\) and \(k_8\) for the reactions \({\mathcal {R}}_7\) and \({\mathcal {R}}_8\) of the internalised receptor. The extracellular decay of ligand due to unspecific binding and other incidents (reaction \({\mathcal {R}}_{10}\)) was set to a value at which it showed a first effect on calcium channel inhibition.
After this initial step of selecting rough parameter values based on available information, we fine-tuned the parameters \(k_1,\ldots ,k_{10}\) via standard parameter estimation techniques using the in vitro experimental FRET data (see Fig. 2 below) that consists of four individual time series for G-protein activation for the four cases fentanyl/pH 6.5, fentanyl/pH 7.4, NFEPP/pH 6.5, and NFEPP/pH 7.4. The experimental data was first pre-processed by determining the offset time (time point at which the respective ligand was added), and the linear scaling transformation that maps the number of undissociated G-proteins from the model to the measured FRET signal. Then the residual distance between the solution of the ODE model and the experimental data was minimized by optimally adapting the parameters \(k_2,\ldots ,k_{10}\), starting from the initial values previously chosen (second column of Table 2 termed “Preselected”). Here, the residual is the mean squared distance between ODE solution and data, summed over all four time series. The minimization was done using standard techniques for parameter estimation27,28, within the framework of the software PREDICI29. The resulting parameters values are shown in Table 2, third column termed “Parameter Estimation”. These values of \(k_2,\ldots ,k_{10}\), optimally adapted to all four cases of ligand/pH combination at once, were fixed.
In a final step, individually for each ligand/pH combination, the parameter \(k_1\) was fine-tuned by minimizing the residual function for each single time-series for fixed \(k_2,\ldots ,k_{10}\) by changing the respective \(k_1\). The outcome was:
where colors shall help to identify the respective curves in Figs. 3, 4 and 5a.
Some of the optimal parameter values exhibit mild deviations from the preselected ones, but no stark contrast to the literature was observed; in fact, closer inspection showed that the mean squared deviation between model-based simulation and experiment was significantly reduced by fine-tuning parameters. The resulting best fit is shown in Fig. 3.
It has been tested that, for these parameter values, the residual distance between data and stochastic model is very close to the residual distance for the ODE model.
Numerical simulations of the reaction network
Simulations of the stochastic reaction network were performed using Python 3. For each combination of rate constants, 500 Monte Carlo simulations were carried out and the arithmetic mean was calculated in order to estimate the percentage of closed calcium channels plotted in Figs. 4 and 8. The initial state numbers of receptors, G-proteins and calcium channels were chosen at a ratio of 1:2:4 (see Table 3). These numbers are only a rough estimate since the exact stoichiometry of binding events in relation to the number of activated second messengers is currently not fully understood at the experimental level30,31. However, these numbers should suffice to get some first impressions of the properties of the reaction network.
As a time horizon for each simulation 1200 s were chosen. The results are presented in “Results”. In order to find the steady state of the dynamics under pure constitutive activation (i.e., ignoring the ligand-induced activation of receptors given by reaction \({\mathcal {R}}_1\)), a simulation without ligands (or, equivalently, with \(k_1=0\)) was run. Given a non-zero rate constant \(k_{11}\) for constitutive G-protein activation, we determined the long-term average number a of closed calcium channels under these conditions. This long-term mean a (rounded to natural numbers) of closed calcium channels was then used to determine the initial state for the dynamics including receptor activation (see Table 3 for the results). To check for normal distribution of the mean, the 500 runs were divided into batches of 50 and the respective means then tested. Anderson–Darling test indicated normal distribution with \(P\le 0.05\), so the 95%-confidence interval of the t-distribution is shown in the plots.
Results
We already demonstrated that the reaction network model introduced in “Models and methods” allows to explain the time course of G-protein subunit dissociation correctly for different ligands and pH values. Based on this validation step, the model was used to analyse the impact (a) of different extracellular pH values correlating to those occurring in injured tissues in vivo2 in combination with a conventional or a pH-dependent opioid ligand (fentanyl or NFEPP, respectively) (see “Isolated impact of pH value”), and (b) of radicals (see “Combined impact of pH value and increased radical concentration”) onto the overall signalling pathway. Based on our parameter fitting results, the changing pH value was modelled via varying the rate constant \(k_1\) of ligand–receptor binding. Additional radicals, on the other hand, were assumed to lower the binding rate \(k_1\) of opioids, while increasing the rate \(k_{11}\) of constitutive activation17,32.
Isolated impact of pH value
For the following analysis we set the rate constant \(k_{11}\) to zero. Our goal was to analyse the effect of varying rates \(k_1>0\) for the ligand-induced activation of a receptor (given by the binding reaction \({\mathcal {R}}_1: L+R \rightarrow RL\)) onto the amount of closed calcium channels \(Ca_{off}\). We examined the ligands fentanyl and NFEPP in combination with changing pH levels (see Eq. (3) for the respective rate values \(k_1\)). The rates for the other reactions were left unchanged in all reaction network simulations based on the assumption that intracellular pH remains at 7.4 (see Table 2).
Figure 4a represents the evolution of the mean number of closed calcium channels for the different ligand-binding rates \(k_1\) given in Eq. (3). For all ligand and pH pairs except for NFEPP/pH 7.4 we observed similar amplitudes of closed calcium channels (about \(44 \%\) of all calcium channels), while for NFEPP/pH 7.4 the amplitude is significantly reduced to approximately \(29\%\) (but note that only a maximum of 40 out of the total of 80 channels can be closed since there are only 40 G-proteins, so the maximum calcium channel inhibition is \(50\%\)). In Fig. 4b, the variances of the number of closed calcium channels depending on time is shown. We can observe that this variance is significantly larger for NFEPP at a normal pH value than for all other scenarios. This information will be used later to support the conclusion that the reduction of the calcium channel inhibition is not due to a uniform decline of the trajectories.
Results from in vitro experiments
The maximum inhibition of voltage-induced calcium currents by fentanyl or NFEPP at pH 6.5 and pH 7.4 was measured by patch-clamp experiments in rat sensory neurons. The results are comparable to the scenarios simulated in Fig. 4a in that both fentanyl and NFEPP potently inhibited calcium currents at low pH, whereas NFEPP was significantly less effective than fentanyl at normal pH (Fig. 5a). Figure 2 shows the experiment that was used for data fitting.
Non-linear behaviour and stochastic effects
Seeing that our model resembled the in vitro results quite well, we now sought to get more information about the dependence of the calcium channel inhibition amplitude and the binding rate \(k_1\). Therefore we plotted both against each other; the results are shown in Fig. 5b. We see a non-linear behaviour of the calcium channel inhibition amplitude with a rather sharp drop for \(k_1<2\times 10^{-3}\,\text {s}^{-1}\) where the calcium channel response declines quickly.
In order to investigate the stochastic effects, we calculated the probability distributions of the number of binding events between ligands and receptors that happened during a simulation run over the time interval \([0,1200\,\text {s}]\) (see Fig. 6). For decreasing \(k_1\) (see 3 for the corresponding ligand/pH pairs) the distribution shifts to lower values and gets a wider range. The non-linearity is also represented in the larger shift from (c) to (d) compared to the other shifts.
Combined impact of pH value and increased radical concentration
We next investigated the effect of varying pH values in combination with rising radical levels on the signalling.
Results from in vitro experiments
Reactive oxygen species can be produced in in vitro experiments by adding H\(_{2}\)O\(_{2}\) to the sample (see Supplementary Information). Our experimental data support that increasing radical (H\(_{2}\)O\(_{2}\)) concentrations are correlated with increasing constitutive G-protein activation, see Fig. 7. This implies that we can include the possible influence of radicals by changing the rate \(k_{11}\) for constitutive activation in the stochastic model.
According to the literature, increased radical concentrations promote the forming of disulfide bonds (DSB) in the receptor 14,33. Hypothetically, the constitutive activation observed in our in vitro experiments stems from DSB formation because adding dithiothreitol (DTT)—a reducing agent that disrupts DSB—reversed the effect of the highest concentration (1 mM) of H\(_{2}\)O\(_{2}\) on basal G-protein activation (Fig. 7). More recent reports further support that DSB formation influences the functionality of GPCR15,16, and there is evidence that constitutive MOR activity is associated with enhanced accessibility of cysteine residues34 (which is essential for DSB formation).
The formation of DSB by radicals has also been investigated by other groups, considering the effect onto ligand-based activation of the GPCR (in our model referring to the binding rate \(k_1\)). For example, Zhang et al.32 describe decreased ligand binding after the removal of a DSB in the extracellular part of the MOR. A review article by Wheatley et al.17 mentions decreased agonist affinity at the CXC-chemokine receptor 4 and increased constitutive activity of the angiotensin II type 1 receptor after breaking extracellular DSBs. In all of the mentioned examples in literature, the change of the receptor structure due to a change of the pristine DSB constellation decreases the binding affinity of specifically designed ligands. Concerning our model, these results indicate that when studying inflamed tissue including increased radical concentrations, the values of the ligand activation rate \(k_1\) should be reduced.
Changing the stochastic model accordingly
With progressive inflammation, there are now two effects on \(k_1\), one from the pH and one from radicals. For an analysis of the combined effect, taking into account the effect of additional DSB formation, the values for \(k_1\) given in Eq. (3) were reduced to \(80\%\) for the inflamed scenario (pH 6.5, additional radicals). For the concrete values see Table 4.
As for the constitutive activation, our in vitro experiments imply that we can include the influence of additional radicals by increasing the corresponding rate constant \(k_{11}\). The experimental data available for the scenario including radicals do not allow for a direct parameter estimation of \(k_{11}\). We chose the value \(k_{11}=5 \times 10^{-5}\) in the presence of radicals, inducing a base level of approximately 5 closed calcium channels in the healthy tissue scenario, which appears to be a reasonable value compared to the other model parameter values. In the scenario without radicals, we set \(k_{11}=0\) as before.
According to the resulting model, the base level of closed calcium channels increases with progressive inflammation (i.e. rising radical concentrations). This is seen in Fig. 8 which shows two plots, one for fentanyl and one for NFEPP. The black curve represents the healthy tissue situation (pH 7.4, no additional radicals) while the olive (pH 6.5, additional radicals) curve shows the effects of progressive inflammation. The main difference between NFEPP and fentanyl is the increase of the amplitude/peak value of the closed calcium channels for NFEPP in the healthy scenario, which is not altered by the constitutive receptor activity.
Discussion
We presented a stochastic model of a canonical GPCR signalling pathway linked to plasma membrane function. This pathway is composed of a biochemical reaction network which begins at the receptor, continues with the G-protein and extends to the membrane calcium channels. The respective reaction rates were determined by fitting the model to data from in vitro FRET experiments. In addition, we have studied the role of radicals.
The validated model shows a non-linear behaviour of the calcium channel inhibition response with regard to the change of the receptor–ligand binding rate \(k_1\). A critical value of \(k_1\) at which the response drops markedly, is \(k_1=2\times 10^{-3}\,\text {s}^{-1}\) (see Fig. 5b). In the vicinity of this value, small changes of \(k_1\) lead to large changes of calcium channel inhibition response. For the ligand NFEPP, decreasing pH from its level for healthy tissue to inflamed tissue lets \(k_1\) increase from below to above this critical value, so that increasing pH results in a quickly decreasing effect of NFEPP. For fentanyl, this is not the case. That is, we observe a markedly diminished effect (i.e. a lower number of closed calcium channels) of NFEPP at normal pH compared to all other scenarios (NFEPP at low pH, fentanyl at low or normal pH).
These results support our previous studies demonstrating that the conventional ligand fentanyl activates MOR both in injured (low pH) and non-injured (normal pH) tissues, while NFEPP is not active in non-injured environments (brain, intestinal wall)3,4,5,6. Those findings were attributed to the decreased pKa value of NFEPP (pKa = 6.8). Apparently, the decreased pKa precludes the protonation of NFEPP in healthy tissues (pH > 7.35; e.g. in brain) but facilitates its protonation in injured/inflamed microenvironments (low pH, high concentrations of protons). Consistent with the notion that the protonation of opioid ligands is typically required for the binding to opioid receptors19, NFEPP was active only in injured tissues. In addition, protonation likely plays a significant role in the activation process of most GPCRs35.
We were able to investigate this phenomenon further: for NFEPP and normal pH, we see higher fluctuations around a lower average of closed calcium channels, compared to all other sceanrios (NFEPP at low pH, fentanyl at low or normal pH).
These results are an extension of the findings in our earlier work3. There it was theorised and corroborated in animal studies that a ligand with proper pH-dependent binding rate would exhibit analgesic effects without side effects. Now we can add that the change of binding rates results in reduced calcium channel inhibition. Thus, the present data provide a more detailed explanation by including the intracellular signalling pathway underlying our initial findings. This further supports our concept of targeting disease-specific conformations of MOR to preclude adverse side effects of painkillers.
Moreover, we included an analysis of constitutive MOR activation with increased concentration of radicals. Comparing the influence of two prominent inflammatory mediators (pH and radicals) on ligand-induced opioid receptor function, it seems that pH has a higher impact than radicals under the chosen parameters. When designing novel opioid painkillers devoid of side effects elicited in non-injured environments, pH-sensitivity may be more important than radical-sensitivity which mainly affects the “base line” and not the peak value in Fig. 8. Given the high degree of homology between GPCRs1, our current studies may be applicable to other signalling pathways (e.g. from receptor to nucleus12), to GPCR involved in other diseases (e.g. cancer, high blood pressure, addiction, depression, arthritis) or even to non-human GPCRs in deranged environments (e.g. in animals or plants exposed to ocean acidification36,37,38,39,40,41,42,43).
Data availability
The datasets generated during and/or analysed during the current study are available in the github repository, https://github.com/user3849/MOR1.
References
Congreve, M., de Graaf, C., Swain, N. A. & Tate, C. G. Impact of GPCR structures on drug discovery. Cell 181(1), 81–91 (2020).
Stein, C. New concepts in opioid analgesia. Expert Opin. Investig. Drugs 27(10), 765–775 (2018).
Spahn, V. et al. A nontoxic pain killer designed by modeling of pathological receptor conformations. Science 355, 966–969 (2017).
Rodriguez-Gaztelumendi, A., Spahn, V., Labuz, D., Machelska, H. & Stein, C. Analgesic effects of a novel pH-dependent \(\mu\)-opioid receptor agonist in models of neuropathic and abdominal pain. Pain 159, 2277–84 (2018).
Massaly, N., Temp, J., Machelska, H. & Stein, C. Uncovering the analgesic effects of a pH-dependent \(\mu\)-opioid receptor agonist using a model of non-evoked ongoing pain. Pain 161, 2798–804 (2020).
Jimenez-Vargas, N. N. et al. Agonist that activates the \(\mu\)-opioid receptor in acidified microenvironments inhibits colitis pain without side effects. Gut(Epub ahead of print) (2021).
Ray, S., Sunkara, V., Schütte, C. & Weber, M. How to calculate pH-dependent binding rates for receptor-ligand systems based on thermodynamic simulations with different binding motifs. Mol. Simul. 46(18), 1443–1452 (2020).
Bhave, G. & Gereau, R. W. Posttranslational mechanisms of peripheral sensitization. J. Neurobiol. 61, 88–106 (2004).
Weis, W. I. & Kobilka, B. K. The molecular basis of GPCR activation. Annu. Rev. Biochem. 87, 897–919 (2018).
Proft, J. & Weiss, N. G-protein regulation of neuronal calcium channels: Back to the future. Mol. Pharmacol. 87, 890–906 (2015).
Zamponi, G. W., Striessnig, J., Koschak, A. & Dolphin, A. C. The physiology, pathology, and pharmacology of voltage-gated calcium channels and their future therapeutic potential. Pharmacol. Rev. 67, 821–870 (2015).
Shaw, W. M. et al. Engineering a model cell for rational tuning of GPCR signaling. Cell 177(3), 782–796 (2019).
Bridge, L. J., Mead, J., Frattini, E., Winfield, I. & Ladds, G. Modelling and simulation of biased agonism dynamics at a G-protein-coupled receptor. J. Theor. Biol. 442, 44–65 (2018).
Bowen, W. D. & Pert, C. B. Conformational malleability of opiate receptors: Sulfhydryl modification alters ion-induced \(\mu\)/\(\delta\)-ligand selectivity shifts in rat striatal sections. Cell. Mol. Neurobiol. 2, 115–128 (1982).
Yang, F. et al. Structure, function and pharmacology of human itch receptor complexes. Nature 600(7887), 164–169 (2021).
Cao, C. et al. Structure, function and pharmacology of human itch GPCRs. Nature 600(7887), 170–175 (2021).
Wheatley, M. et al. Lifting the lid on GPCRs: The role of extracellular loops. Br. J. Pharmacol. 165(6), 1688–1703 (2012).
Meyer, J., Del Vecchio, G., Seitz, V., Massaly, N. & Stein, C. Modulation of \(\mu\)-opioid receptor activation by acidic pH is dependent on ligand structure and an ionizable amino acid residue. Br. J. Pharmacol. 176(23), 4510–4520 (2019).
Lešnik, S., Bertalan, E., Bren, U. & Bondar, A.-N. Opioid receptors and protonation-coupled binding of opioid drugs. Int. J. Mol. Sci. 22, 13353 (2021).
Currie, K. P. M. G-protein inhibition of cav2 calcium channels. Channels 4(6), 497–509 (2010).
Gillespie, D. T. A rigorous derivation of the chemical master equation. Phys. A 188(1–3), 404–425 (1992).
Winkelmann, S. & Schütte, C. Stochastic Dynamics in Computational Biology (Springer, 2020).
Boron, F. W. Regulation of intracellular pH. Adv. Physiol. Educ. 28, 160–169 (2004).
Zamponi, G. W. & Snutch, T. P. Decay of prepulse facilitation of N type calcium channels during G-protein inhibition is consistent with binding of a single \(g_{\beta \gamma }\) subunit. Proc. Natl. Acad. Sci. U.S.A. 95, 4035–4039 (1998).
Shea, L. D., Neubig, R. R. & Linderman, J. J. Timing is everything the role of kinetics in G-protein activation. Life Sci. 68, 647–658 (2000).
Williams, J. T. et al. Regulation of \(\mu\)-opioid receptors: Desensitization, phosphorylation, internalization, and tolerance. Pharmacol. Rev. 65, 223–254 (2013).
Bock, H. et al. Model Based Parameter Estimation. Theory and Applications. Based on the Workshop on Parameter Estimation Vol. 4 (Springer, 2013).
Wulkow, N., Telgmann, R., Hungenberg, K.-D. & Schütte, C, & Wulkow, M. Deterministic and Stochastic Parameter Estimation for Polymer Reaction Kinetics I. Theory and Simple Examples. Macromolecular Theory and Simulations (2021).
Wulkow, M. Computer aided modeling of polymer reaction engineering-the status of prediction. I-simulation. Macromol. Reaction Eng. 2(6), 461–494 (2008).
Hilger, D. The role of structural dynamics in GPCR-mediated signaling. FEBS J. 288(8), 2461–2489 (2021).
Latorraca, N. R., Venkatakrishnan, A. J. & Dror, R. O. GPCR dynamics: Structures in motion. Chem. Rev. 117(1), 139–155 (2017).
Zhang, P. et al. Mutation of human \(\mu\)-opioid receptor extracellular “disulfide cysteine’’ residues alters ligand binding but does not prevent receptor targeting to the cell plasma membrane. Mol. Brain Res. 72(2), 195–204 (1999).
Cremers, C. M. & Jakob, U. Oxidant sensing by reversible disulfide bond formation. J. Biol. Chem. 288(37), 26489–26496 (2013).
Xu, W., Sanz, A., Pardo, L. & Liu-Chen, L.-Y. Activation of the \(\mu\)-opioid receptor involves conformational rearrangements of multiple transmembrane domains. Biochemistry 47(40), 10576–10586 (2008).
Zhang, X. C., Cao, C., Zhou, Y. & Zhao, Y. Proton transfer-mediated gpcr activation. Protein. Cell 6(1), 12–17 (2015).
wwPDB consortium,. Protein data bank: The single global archive for 3d macromolecular structure data. Nucleic Acids Res. 47(D1), D520–D528 (2019).
Koehl, A. et al. Structure of the \(\mu\)-opioid receptor-G-protein complex. Nature 558(7711), 547–552 (2018).
Isberg, V. et al. Generic GPCR residue numbers–aligning topology maps while minding the gaps. Trends Pharmacol. Sci. 36(1), 22–31 (2015).
Li, J.-G. et al. Asp147 in the third transmembrane helix of the rat \(\mu\)-opioid receptor forms ion-pairing with morphine and naltrexone. Life Sci. 65(2), 175–185 (1999).
Mahinthichaichan, P., Vo, Q. N., Ellis, C. R. & Shen, J. Kinetics and mechanism of fentanyl dissociation from the \(\mu\)-opioid receptor. JACS Au 1(12), 2208–2215 (2021).
Zhang, H., Kenaan, C., Hamdane, D., Hoa, G. H. B. & Hollenberg, P. F. Effect of conformational dynamics on substrate recognition and specificity as probed by the introduction of a de novo disulfide bond into cytochrome p450 2b1. J. Biol. Chem. 284(38), 25678–25686 (2009).
Dror, R. O. et al. Activation mechanism of the \(\beta\)2-adrenergic receptor. Proc. Natl. Acad. Sci. 108(46), 18684–18689 (2011).
Fowler, C. B., Pogozheva, I. D., Lomize, A. L., LeVine, H. & Mosberg, H. I. Complex of an active \(\mu\)-opioid receptor with a cyclic peptide agonist modeled from experimental constraints. Biochemistry 43(50), 15796–15810 (2004).
Acknowledgements
This research was partially funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy—The Berlin Mathematics Research Center MATH+ (EXC-2046/1 project ID: 390685689), by DFG-STE 477/19-1, DFG-STE 477/21-1, DFG-FOR 5177, and via Project C03 of CRC 1114.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Contributions
Conceptualization: S.W., M.W., C.S., C.S. Formal analysis: S.W., M.W., C.S. Investigation: A.T., S.R., V.S., H.W., M.Ö.C., F.Y. Supervision: C.S., S.W., M.W., C.S. Writing—original draft: all authors. Writing—review and editing: all authors.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Thies, A., Sunkara, V., Ray, S. et al. Modelling altered signalling of G-protein coupled receptors in inflamed environment to advance drug design. Sci Rep 13, 607 (2023). https://doi.org/10.1038/s41598-023-27699-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-023-27699-w
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.