Identification and validation of novel and more effective choline kinase inhibitors against Streptococcus pneumoniae

Streptococcus pneumoniae choline kinase (sChoK) has previously been proposed as a drug target, yet the effectiveness of the first and only known inhibitor of sChoK, HC-3, is in the millimolar range. The aim of this study was thus to further validate sChoK as a potential therapeutic target by discovering more powerful sChoK inhibitors. LDH/PK and colorimetric enzymatic assays revealed two promising sChoK inhibitor leads RSM-932A and MN58b that were discovered with IC50 of 0.5 and 150 μM, respectively, and were shown to be 2–4 magnitudes more potent than the previously discovered inhibitor HC-3. Culture assays showed that the minimum inhibitory concentration (MIC) of RSM-932A and MN58b for S. pneumoniae was 0.4 μM and 10 μM, respectively, and the minimum lethal concentration (MLC) was 1.6 μM and 20 μM, respectively. Western blot monitoring of teichoic acid production revealed differential patterns in response to each inhibitor. In addition, both inhibitors possessed a bacteriostatic mechanism of action, and neither interfered with the autolytic effects of vancomycin. Cells treated with MN58b but not RSM-932A were more sensitive to a phosphate induced autolysis with respect to the untreated cells. SEM studies revealed that MN58b distorted the cell wall, a result consistent with the apparent teichoic acid changes. Two novel and more highly potent putative inhibitors of sChoK, MN58b and RSM-932A, were characterized in this study. However, the effects of sChoK inhibitors can vary at the cellular level. sChoK inhibition is a promising avenue to follow in the development of therapeutics for treatment of S. pneumoniae.


Scientific RepoRtS
| (2020) 10:15418 | https://doi.org/10.1038/s41598-020-72165-6 www.nature.com/scientificreports/ residues are then polymerized by an unidentified protein to form teichoic acid which is 11 transported across the cell membrane acid flippase TacF. The TacL ligase attaches teichoic polymers to a glycolipid anchor to form LTA 7 . Meanwhile, teichoic acid polymers are cross-linked to peptidoglycan to form WTA by the action of LCP phosphotransferases 11 . The enzyme of S. pneumoniae has previously been proposed as a drug target 3 . However, the sChoK inhibitor originally tested, Hemicholinium-3 (HC-3), was weak: HC-3 had IC 50 12 and minimum inhibitory concentration (MIC) values in the millimolar range 3 . The aim of this study was to discover stronger inhibitors of sChoK in order to further establish sChoK as a target. Here, we present two compounds which strongly inhibit sChoK activity and S. pneumoniae growth and viability: MN58b and RSM-932A (Fig. 1C) 13,14 , two compounds known to inhibit human choline kinase (hChoK) 14 .
Using S. pneumoniae as a model, we present evidence to demonstrate that employing sChoK inhibitors is a promising strategy for blocking the growth of S. pneumoniae cells. These inhibitors, in turn, were found to have downstream effects on teichoic acid production and assembly, on cell wall shape, and on the sensitivity of the cell to autolysis. Determining MN58b and RSM-932A competitivity by LDH/pK. The mechanism of action of the inhibitors with respect to each substrate was determined using a previously described method 13,14 . Briefly, using the IC 50 of each inhibitor determined at the K m of choline, the concentration of one substrate was kept constant, and initial velocities were measured while varying the second substrate in the presence and absence of the inhibitor. The percentage of inhibition (%I) was determined by dividing each initial velocity in the presence of the inhibitor with its corresponding initial velocity in the absence of compound. %I was plotted against concentration and fitted using a two-parameter-fit nonlinear regression algorithm of GnuPlot v 4.3 into Eq. (2).
where [S] is the concentration of the varied substrate, [I] is the concentration of inhibitor, Ks is the dissociation constant of the substrate (assumed to be equal to the K m ), Ki is the dissociation constant of the inhibitor, and α is a numerical constant that measures the effect the substrate has on the binding of the inhibitor and vice versa.
Docking of MN58b and RSM-932A onto sChoK. With regard to the kinetic data, MN58b and RSM-932A were docked into the active site and choline site alone, respectively using apo-sChoK structure as a model (RCSB accession #4R77). The entire active site was defined as all of the residues of sChoK known from the crystal structure of sChoK to make contact with both ATP (43, 90, 183, 89, 91, 31, 818, and 194) and choline (residue numbers 197, 251, 254, 268, 178, 213, 29, 176), while the choline site was defined as only those residues making contact with choline.
The molecular structure of MN58b was defined using the JSME molecular editor (https ://cactu s.nci.nih.gov/ trans late/edito r.html), and a .pdb file for this molecule was generated using the PRODRG server (https ://davap c1.bioch .dunde e.ac.uk/cgi-bin/prodr g/submi t.html). The resulting structure was then inspected using Pymol. The RSM-932A .SDF structure file was downloaded from the RCSB Protein Data bank (https ://www.rcsb.org, accession number 5W60) and converted to a .pdb using the online SMILES translator (https ://cactu s.nci.nih. gov/trans late/). Docking was performed using the Patchdock server. (https ://bioin fo3d.cs.tau.ac.il/Patch Dock/). The top ten structures were then refined using Firedock (https ://bioin fo3d.cs.tau.ac.il/FireD ock/), and the top solution from this output was selected as the most likely final structure of the complexes. Protein alignments and figures were performed using Pymol, and Fig. 1A and B were produced with this program as well.
Determining minimum inhibitory concentration (MIC) and minimum lethal concentration (MLC). Seven μL of a glycerol stock of the S. pneumoniae R6 strain was used to inoculate 7 mL of Brain Heart Infusion broth (BHI, Accumedia, Sydney, Australia) supplemented with 5 units/mL catalase (BHI-CAT). This starter culture was incubated at 37 °C in a water bath until an optical density at 610 nm (OD 610 ) was reached. Forty μL of this starter culture was used to inoculate 40 mL of BHI-CAT media with or without varying concentrations of inhibitor. This culture was then incubated at 37 °C in water bath until an OD 610 of about 1.0 was reached in the control. Western blot detection of lipoteichoic acid. R6 cells were cultured to saturation in either BHI-cat alone or BHI-cat 0.5 MIC of MN58b or RSM-932A. One milliliter (1 mL) of cell samples was centrifuged at 3,000g, sonicated 3 × 2 min on ices. Western blots were performed as described 3 while total protein on the membrane was measured by fluorescence according to Quickstain instructions.
Scanning electron microscopy monitoring of S. pneumoniae cells treated with MN58b or RSM-932A. Scanning electron microscopy was performed as previously reported 17 .
Results sChok kinetic constants. Kinetic constants were measured for sChoK: K m , and k cat values were 130 ± 36 µM, 3.14 ± 0.24 s −1 , respectively. These results were comparable to previously reported values 15 . This initial information was used to design additional steady state experiments, such as IC 50 measurements and determinations of drug competitivity.
ic 50 results by LDH/PK and colorimetric methods. The IC 50 s of each drug against sChoK were measured using two methods: the indirect ldh/pk method that relied on recombinant sChoK and which indirectly measured consumption of ATP 14 and a previously reported colorimetric method that relied on BL21 (DE3) extracts expressing recombinant sChoK, which directly measured the consumption of choline and the production of phosphocholine 16 . Results are shown in Table 1. Inhibitors were similarly effective in a complex cell extract and in a simplified system relying on a purified enzyme. Both MN58b and RSM-932A had IC 50s that fell within the µM range, which was an improvement over the IC 50 previously reported for the sChoK inhibitor HC-3 (IC 50 of 2.7 mM) 12 . RSM-932A was stronger that MN58b by 1-2 orders of magnitude.
MIC and MLC results. Both drugs had MICs and MLCs in the low micromolar ranges (see Table 1), making them much more powerful inhibitors of S. pneumoniae growth than the previously reported sChoK inhibitor HC-3, which had an MIC of 5,400 µM 3 . To our knowledge, other than HC-3, no other sChoK inhibitors have ever been reported, making MN58b and RSM-932A the most effective sChoK inhibitors thus far reported both on the level of enzyme inhibition and inhibition of cell growth.
MN58b is competitive with both choline and ATP while RSM-932A is competitive with choline alone. A mechanistic analysis was performed for each inhibitor by performing steady-state enzymatic reactions with sChoK in which one substrate was varied while the other substrate was fixed both in assays with and without inhibitor fixed at a concentration equal to its IC 50 . sChoK inhibition was calculated as a percentage (%I) calculated by dividing the initial velocity at each substrate concentration with inhibitor by counterpart initial velocities in the absence of inhibitor. %I values were plotted against substrate concentration. The resulting data was fitting to an Eq. (2) and K i and α values calculated. In the case of MN58b, %I went down as the concentration of substrates choline ( Fig. 2A) or ATP (Fig. 2B) went up, indicating that inhibitor binding was antagonistic to the binding of both choline and ATP (and vice versa) In contrast, with RSM-932A, an increase in choline marked a parallel increase in %I, while an increase in ATP showed no observable trend (Fig. 2C,D, respectively), meaning that inhibitor binding was antagonistic to choline and non-competitive with ATP.
As a measure of the effect of each substrate on inhibitor binding, α and K i (dissociation constant) values were derived for both inhibitors by fitting %I values according to Eq. (1) (see Table 2, "Methods and materials"). The α values calculated for MN58b were low with respect to both choline and ATP; that is, they did not approach infinity. These results supported a mixed-inhibition mechanism for these compounds (not completely competitive). Changes in the steady state constants K m and V max were also determined in the presence and absence of drug by fitting initial velocity and substrate values according to Eq. (1). Observed increases in K m values coupled with decreases in V max values supported a mixed inhibition model in which the inhibitor binding displaced both choline and ATP. Meanwhile, α values derived for RSM-932A were low in the case of choline (indicating a mixed inhibition model) but approached a value of 1 in the case of ATP, which was consistent with a noncompetitive mechanism of action. In the case of ATP, only a slight drop in K m values was observed between the presence and absence of inhibitor coupled with a drop in V max , which supports a non-competitive model. In the case of choline, a decrease in both V max and K m was observed, supporting a mixed-mode of inhibition in which only choline is displaced by the inhibitor.  (Fig. 1A,B) The drugs were docked onto the sChok protein structure using the program Patchdock 18 . While both MN58b and RSM-932A could be seen binding to a site overlapping with the binding site of choline, only MN58b entered and entirely overlapped with an AMP molecule used to model the structure of sChoK with ATP.  www.nature.com/scientificreports/ MN58b and RSM-932A do not provoke or prevent autolysis. By testing the effects of incubation on optical densities of cell cultures grown to an O.D. of 0.500 at different concentrations of inhibitor, we found that that MN58b and RSM-932A were bacteriostatic up to a concentration of 50 × MIC after which they become bacteriolytic (Fig. 3A). Autolysis is a process of cell lysis and death that is mediated by LTAs via choline binding autolysins 19 such as LytA. The choline analog and sChoK inhibitor HC-3 had previously been found to attenuate the autolytic process in Lactobacillus reuteri cells 17 . In light of the non-specificity observed with HC-3 (being able to interfere with both ChoKs and the autolytic process), it seemed logical to check the specificity of MN58b and RSM-932A for sChoK inhibition. The interaction of MN58b, RSM-932A, and HC-3 with bacteriolytic antibiotic vancomycin was determined (Fig. 3B). As expected, HC-3 prevented cell lysis provoked by vancomycin, but MN58b and RSM-932A had minimal to no effect, indicating that the latter drugs were more specific sChoK inhibitors than HC-3. The non-specificity of HC-3 may help to explain its weakness against sChoK.

MN58b sensitizes S. pneumoniae cells to autolysis while RSM-932A does not. Given the obser-
vation that MN58b and RSM-932A were sChoK inhibitors, our expectation was that cells treated with these inhibitors would show a decrease in overall LTA content because a decrease in phosphocholine availability would reduce the overall production of LTA residues and, by extension, LTA molecules. A reduction in LTA content, in turn, should help desensitize cells to autolytic induction. Although autolysin proteins would be released, they would not be able to dock on the cell surface because the autolysin choline binding domains would have fewer choline containing LTA molecules with which to interact. Both LTA reduction and cell sensitization had previously been observed with HC-3 17 . Surprisingly, cells grown in the presence of 0.5 MIC RSM-932A were equally sensitive to autolysis as untreated controls, while MN58b actually primed the cells to be autolyzed (Fig. 3C,D). These unexpected results demonstrated that the effect of MN58b and RSM-932A on LTA production and sensitivity to autolysis were not as predictable as the results observed with HC-3.

Western blot detection of LTA in cells treated with MN58 and RSM-932A.
In order to quantify LTA production, western blots were performed on extracts of cells grown to saturation in the presence of 0.5 MIC MN58b or RSM-932A. LTA band patterns clearly differed between treatments. With MN58b, a marked increase in total LTA production was observed (Fig. 4A,B  www.nature.com/scientificreports/ a marked increase in the smaller sizes of LTA (Fig. 4A-C). It is important to note that each size band in an LTA western blot differs from the lower one by a single LTA residue (Fig. 4A) 19 . This result indicated that MN58b was dysregulating both LTA production and subunit polymerization. Meanwhile, with RSM-932A, overall production remained unchanged (Fig. 4D) while larger size polymers were favored (Fig. 4C). However, this result was unexpected given our original hypothesis that sChoK inhibitors should limit production of LTAs and is even more surprising because each inhibitor has a distinct effect on LTA patterns observed.

MN58b but not RSM-932A causes deformation of the S. pneumoniae cell wall.
In order to find out if these distinct effects on LTA production extended to effects on the shape of the cell wall, scanning electron microscopy was performed on cells treated with either MN58 or RSM-932A (Fig. 5). MN58 was found to modulate the shape of the cell wall giving the wall a bumpy surface (Fig. 5B) while RSM-932A did not greatly affect the normally smooth surface of the cell (Fig. 5C). Taken together with the data from the autolysis experiments,  www.nature.com/scientificreports/ it seems that smaller assembly sizes of LTA and/or increased LTA production leads to distortion of the cell wall and sensitivity to autolysis. Meanwhile, increasing the quantity of larger assembly sizes of LTA does not lead to cell wall distortion and does not affect the level of sensitivity of the cell to autolytic induction.

In enzymatic assays with S. pneumoniae extracts MN58b is still an inhibitor but RSM-932A becomes an agonist. In light of the inconsistent results between sChoK inhibition by MN58b and RSM-
932A which are both known choline kinase inhibitors, we repeated the colorimetric experiments using S. pneumoniae extracts as the source of sChoK, The goal of the experiment was to determine whether or not differences in sChoK inhibition could be detected in a more native environment than that provided by extracts of E. coli overexpressing sChoK or purified sChoK. MN58b continued to function as an inhibitor of choline kinase activity under these conditions (Fig. 6, bottom) Surprisingly we found that, while RSM-932A was somewhat inhibitory at lower concentrations, it functioned as an agonist of choline kinase activity at higher concentrations in a dose dependent manner (Fig. 6, top). This agonism was confirmed by mass spectrometry (data not shown) and was also found to be ATP dependent, thereby supporting the inference that RSM-932A was still affecting sChoK activity, although in the opposite direction. This suggested that there was a factor found in the S. pneumoniae cells that was not found in E. coli cells nor in the purified enzyme that modified how sChoK interacted with RSM-932A.

Discussion
We demonstrated in this study that purified sChoK could potentially be used as a screening system to detect inhibitors of S. pneumoniae growth, and we presented evidence of the strongest sChoK inhibitors of enzymatic action and cell growth described to date. In the case of MN58b, the MIC is at least a magnitude of order smaller than its IC 50 , which suggests that either the effect of this drug is occurring on other proteins or that the effect of the drug increases as a consequence of being accumulated at a high concentration inside a cell. In the case of RSM-932A, both the IC 50 and MIC are in the sub-micromolar range, which would normally suggest that the inhibitory effect of the drug was due to its binding to its enzyme target alone. However, in a more native environment, RSM-932A actually functions as an agonist. Importantly, we did not observe the suppression of downstream lipoteichoic acid production that is the expected consequence of choline kinase inhibition 17 due to the putative attenuation of phosphocholine production which would remove one of the building blocks of LTA Instead, we observed increased phosphocholine production (in the case of MN58b) and changes in lipoteichic acid sizes to either smaller (MN58b) or larger sizes (RSM-932A). This change appears to have an effect on cell sensitivity to autolysis and on the shape of the cell wall. Dysregulation of LTA should lead to differences in sensitivity because prior to cleaving peptidoglycan, autolysins must first bind the phosphocholine head of LTA. However, the question remains as to whether changes in LTA assembly is a consequence of inhibitor action on choline kinase or rather a separate process, such as direct binding of RSM-932A and MN58b to a yet unidentified protein responsible for polymerization of LTA chains or even to a separate choline binding protein entirely.
Interestingly, there is a key difference in autolysis, which may occur because of overexpression of LTA and/ or overexpression of smaller LTA sizes, leading to differences in cell wall shape as well. Larger sizes of LTA do not appear to lead to changes in sensitivity to autolysis nor to changes in the cell wall. However, in both cases, changes in the size of LTA are associated with the effectiveness of the drugs in inhibiting cell growth.
We would thus like to propose a possible model that could both explain the disparate results with RSM-932A in its activity as either an inhibitor or an agonist and the differences in LTA polymerization and production observed between MN58b and RSM-932A: choline kinase forms part of a larger complex that includes the www.nature.com/scientificreports/ protein responsible for LTA chain assembly. In fact, hChoKα has also been proposed to be a scaffolding protein in eukaryotic cells 20 . A similar function could be anticipated for sChoK. The differences in RSM-932A action cannot be explained by positing that this action differs depending on whether or not sChoK is in the presence of a cell extract because, in the presence of an E. coli cell extract, this drug remains an inhibitor. In the presence of an S. pneumoniae extract, however, this drug becomes an agonist. One possibility is that sChoK, which localizes to the cytoplasm 21 , forms a complex with another cellular element that is not found in E. coli, and this interaction is modulating the effects of RSM-932A. We thus propose that this other cellular element is an LTA polymerase (which may or may not localize to the cytoplasm) and that this interaction is a two-way process. Inhibitor binding to sChoK may also modulate this LTA polymerase indirectly via this putative protein-protein interaction. Meanwhile, MN58b and RSM-932A, while both inhibitors of sChoK, do not have the same mechanism of action. As seen from the steady-state experiments, these two inhibitors bind to sChoK differently. MN58b is antagonistic to both ATP and choline, while RSM-932A is antagonistic only to choline. A different mode of binding would have a different effect on the structure of sChoK and therefore lead to disparate effects of these drugs on the putative interaction with a putative LTA polymerase, which would, in turn, lead to outputs in LTA molecule sizes that differ. In fact, MN58b and RSM-932A are found to have a differential mode of interaction with hChoKα as well 14,22 .
It remains to be seen whether this model can be supported through proteomic studies that could identify the putative sChoK complex or whether or not these experiments could lead to identification of the enzyme responsible for LTA polymerization and the mechanism whereby dysregulating production and polymerization of LTA leads to inhibition of cell division.
However, the mechanism of action of any choline analog could include simultaneously (or exclusively) inhibiting the action of bacterial choline binding proteins (CBP) other than choline kinase. These include lysozyme (LytC), Amidase (CbpD), choline binding protein (CbpF), pneumococcal surface protein A and C (PspA, PspC), choline transporters, and Phosphorylcholine esterase (Pce). LytC, CbpD, and CbpF are all involved in the autolytic process. Our study has shown that MN58b and RSM932A neither enhance, nor block the process of autolysis, making the possibility that LytC, CbpD and CbpD are direct targets unlikely. PspA, PspC, and Pce are known to manage cell virulence, but it is unclear if these are essential proteins and none have been validated as drug targets for blocking cell growth. Nevertheless it is certainly possible that there is either MN58b and RSM932A directly targets some choline binding protein or even modulates expression of these proteins to the point where growth is disabled 23 . Though we continue to favor the model that choline kinase is a target of choline kinase inhibitors, to truly begin distinguishing between the different possible targets, a good first step would be to identify mutations in strains resistant to MN58b and RSM932A and/or monitor changes in expression that are a result of treatment. Nevertheless, it is clear that by using recombinant purified or unpurified sChoK as a system of screening, one can identify promising new inhibitors of S. pneumoniae growth to be developed as possible antibiotics.
A valid concern of applying sChoK inhibitors to treat an infection is that hChoK activity could be affected, leading to secondary effects. This concern is particularly valid for those inhibitors that have been developed to block hChoK. As a starting point, we deliberately chose the best characterized two hChoK inhibitors available for study, MN58b and RSM-932A. This includes pre-clinical toxicity studies, that give us an idea of dosage limits 24 . The effectiveness of these drugs against infection in animal and human models within the limits of toxicity remains to be assessed. We do know that these drugs are non-toxic to human primary cell lines and selective to tumor cells. RSM-932A even reached phase I drug trials for treating tumors (https ://clini caltr ials.gov/ct2/ show/NCT01 21586 4). In addition, these drugs are being examined for use inflammatory disorders 25,26 . What we have determined here is that that the usefulness of sChoK inhibitors, including those designed for hChoK, could extend to use as antibiotics for treating S. pneumoniae infections. In addition, possible toxicity could be prevented in future studies by designing sChoK inhibitors that are selective for bacterial isoforms.