Screening Pyridine Derivatives against Human Hydrogen Sulfide-synthesizing Enzymes by Orthogonal Methods

Biosynthesis of hydrogen sulfide (H2S), a key signalling molecule in human (patho)physiology, is mostly accomplished by the human enzymes cystathionine β-synthase (CBS), cystathionine γ-lyase (CSE) and 3-mercaptopyruvate sulfurtransferase (MST). Several lines of evidence have shown a close correlation between increased H2S production and human diseases, such as several cancer types and amyotrophic lateral sclerosis. Identifying compounds selectively and potently inhibiting the human H2S-synthesizing enzymes may therefore prove beneficial for pharmacological applications. Here, the human enzymes CBS, CSE and MST were expressed and purified from Escherichia coli, and thirty-one pyridine derivatives were synthesized and screened for their ability to bind and inhibit these enzymes. Using differential scanning fluorimetry (DSF), surface plasmon resonance (SPR), circular dichroism spectropolarimetry (CD), and activity assays based on fluorimetric and colorimetric H2S detection, two compounds (C30 and C31) sharing structural similarities were found to weakly inhibit both CBS and CSE: 1 mM C30 inhibited these enzymes by approx. 50% and 40%, respectively, while 0.5 mM C31 accounted for CBS and CSE inhibition by approx. 40% and 60%, respectively. This work, while presenting a robust methodological platform for screening putative inhibitors of the human H2S-synthesizing enzymes, highlights the importance of employing complementary methodologies in compound screenings.

Interaction with human CBS, CSE and MST analyzed by differential scanning fluorimetry (DSF) and surface plasmon resonance (SPR). The newly synthesized derivatives were assayed by two complementary biophysical techniques, namely DSF and SPR, for their ability to bind to the H 2 S-synthesizing human enzymes tCBS, CSE and MST, recombinantly expressed and purified from E. coli. For each target protein, the DSF assays were preliminarily optimized in terms of protein and dye concentration, resulting in the following conditions (final volume: 20 µL in each well): tCBS (2 µg/well; ~2 μM), CSE (1 µg/well; ~1 μM) or MST (2 µg/ well; ~3 μM); final dye concentration: 1x. As shown in Fig. 2 (top panel), the DSF thermal denaturation curve of tCBS (marked with 'a') displays an unusually high fluorescence of the dye at the initial temperature (20 °C), indicating either a partial unfolding of the protein or a possible interference from a protein component in the  n/a n/a n/a n/a n/a 56 C21 COO(CH 2 ) 2 COONa COOCH 3 COO(CH 2 ) 2 COONa n/a n/a 55 C22 COOCH 2 COOCH 3 COOH COOCH 2 COOCH 3 n/a n/a 55 C23 COOCH 2 COOC 2 H 5 thienyl COOCH 2 COOC 2 H 5 n/a n/a 55 C24 COOCH 2 COONa thienyl COOCH 2 COONa n/a n/a 55 C25 n/a n/a n/a COOH C 2 H 5 This work C26 n/a n/a n/a COOC 2 H 5 C 2 H 5 This work C27 n/a n/a n/a thienyl CH 2 COOC 2 H 5 This work C28 n/a n/a n/a thienyl CH 2 COONa This work C29 n/a n/a n/a n/a n/a 52 C30 n/a n/a n/a n/a n/a 50 C31 n/a n/a n/a n/a n/a 50 Table 1. Pyridine derivatives screened against human H 2 S-synthesizing enzymes.  Table S1). The assay displayed Z'-factors of −1.58, + 0.08 and −0.28 for T m1, T m2 and T m_' Ave' , respectively. Compounds C9, C10, C19 and C23 resulted in aberrant thermal denaturation profiles, precluding any type of analysis. The remaining compounds had a limited impact on the tCBS thermal denaturation ( Supplementary Fig. S2). While none of the compounds exhibited |Z-score| ≥ 3.0 for any of the analysed parameters (T m1, T m2 and T m_' Ave' ), four compounds, namely C2, C14, C28 and C29, yielded |ΔT m_' Ave' | ≥ 1.0 °C. The CSE thermal denaturation profile obtained in the DSF experiments (marked with 'a' in Fig. 2, middle panel) presented a basal fluorescence at 20 °C even higher than observed for tCBS. The initial drop preceded two fluorescence increases (at 45-60 °C and 70-80 °C) interspaced by a major drop (at 60-70 °C). By analysing the CSE thermal unfolding monitored by CD spectropolarimetry (marked with 'b' in Fig. 2, middle panel), a single transition was detected, with a T m at 70.4 °C. While precluding reliable estimates of apparent T m values, the DSF thermal denaturation profiles still allow analysing relative changes upon compounds screening, particularly in terms of the high basal fluorescence measured at 20 °C and intensity of each transition. To quantitatively analyse the effect of the pyridine derivatives on CSE thermal denaturation, we established two parameters based on fluorescence ratios along the thermal denaturation profile (marked with Greek letters along the curve in Fig. 2, middle panel, line a): Ratio A = α/β and Ratio B = γ/β. Statistical validation of the assay for CSE was obtained by incubating the enzyme with 200 μM AOAA as the negative control (N = 26; 6 independent experiments), to be compared with the non-incubated enzyme as positive control (N = 26; 6 independent experiments) (Supplementary Fig. S1 and Supplementary Table S2). The assay displayed Z'-factors of + 0.53 and + 0.31 for Ratios A and B, respectively. Compounds C10, C18, C19 and C23 resulted in aberrant thermal denaturation profiles, precluding any type of analysis. By analysing the effect of the other tested pyridine derivatives on the thermal denaturation profile of CSE ( Supplementary Fig. S3), C1, C2, C9, C17, C30 and C31 were identified as the only compounds exhibiting |Z-score| ≥ 3.0 for both Ratio A and Ratio B (Supplementary Table S2).
The MST thermal denaturation profiles obtained by DSF and CD (Fig. 2, bottom panel) exhibited two well separated transitions with nearly matching T m values between the two methods: T m1 = 42.4 °C (60%) and T m2 = 56.2 °C (40%) by DSF yielding T m_' Ave' = 48.2 °C, to be compared with T m1 = 45.0 °C (60%) and T m2 = 58.8 °C (40%) by CD spectropolarimetry yielding T m_ave = 50.5 °C. Possible interactions of the pyridine derivatives with MST were evaluated by DSF ( Supplementary Fig. S4 and Supplementary Table S3). Statistical validation of the assay for MST was obtained by incubating the enzyme with 3-mercaptopyruvate (3MP, at 2 mM) as the negative control (N = 28; 6 independent experiments), to be compared with the non-incubated enzyme as positive control (N = 31; 6 independent experiments) (Supplementary Fig. S1 and Supplementary Table S3). As shown in Supplementary Fig. S1, the major effect of 3MP was to significantly decrease the relative amplitude of the first transition ('Frac' in Supplementary Table S3), while decreasing both T m values. The assay displayed Z'-factors of + 0.50, −0.32, + 0.68, and −0.06, for Frac, T m1, T m2 and T m_' Ave' , respectively. Compounds C19 and C23 resulted in aberrant thermal denaturation profiles, precluding any type of analysis. Compounds C2, C30 and C31 exhibited |ΔT m1 | ≥ 1.0 °C with |Z-score| ≥ 3.0, while all but six compounds exhibited a |Z-score| ≥ 3.0 with respect to T m2 (Supplementary Table S3). Five compounds (C5, C13, C26, C27, and C30) resulted in a shift of |ΔT m_' Ave' | ≥ 1.0 °C and, notably, out of them only compound C30 exhibited a shift in both T m1 and T m2 with |Z-score| ≥ 3.0.
The synthetic compounds were also screened for their ability to bind CBS, CSE and MST by SPR. This is an optical methodology based on the physical principle of refractive index changes at the biospecific sensor surface upon complex formation 31,36 . After immobilizing the enzymes on the chips, the tested pyridine derivatives were assayed at four concentrations: 25, 50, 100 and 200 µM. No Z'-factor could be measured for these assays due to lack of adequate controls under the selected experimental conditions. Under those conditions, several compounds (C9, C10, C18, C19 and C23) displayed poor solubility, thus precluding an analysis of their interaction with the target proteins by SPR. Moreover, the results obtained for several compounds revealed solubility issues affecting differently the sensorgrams for the three enzymes (Supplementary Figs S5-S7): C1 (MST), C11-C14 (tCBS and CSE), C16 (tCBS and CSE), C17 (all), C20-22 (tCBS and CSE), C24-26 (tCBS and CSE), C27-28 (all), and C31 (tCBS). For the remaining compounds, due to the fast kinetics observed for the studied interactions, no association or dissociation rate constants could be determined. We thus identified compounds yielding sensorgrams where the steady-state response units were proportional to the compound concentration (in the tested range) and did not exceed the expected maximal value (Rmax, based on the compound molecular weight and surface density), thus avoiding any misidentification by superstoichiometric binding behaviour. As shown in Fig. 3, five compounds (C1, C3, C5, C6 and C7) proved to interact with tCBS, two (C7 and C31) with CSE, and three (C5, C7, C14) with MST, in a concentration-dependent mode. The fact that the corresponding steady-state response units linearly increase with the compound concentration up to 200 μM points to low-affinity interactions. Similarly, weak interactions were also observed between AOAA and both tCBS and CSE ( Supplementary  Fig. S8).

Inhibition of human CBS, CSE and MST analyzed by fluorimetric and colorimetric assays.
The screening was complemented with activity measurements to test the inhibitory efficacy of the synthetic compounds towards the target enzymes. A fluorescence-based assay using the H 2 S-detecting dye AzMC was initially attempted with tCBS alone. A Z'-factor of + 0.78 was obtained for this assay by measuring the fluorescence signal of the dye in the absence of tCBS (negative control; N = 16; 3 independent experiments) and in its presence (positive control; N = 16; 3 independent experiments) (Supplementary Table S4). All pyridine derivatives were first screened at a single concentration of 200 µM. Under these conditions, sixteen compounds appeared to lower the dye signal (Supplementary Table S4) and were thus assayed in a concentration range between 15.6 µM and 1 mM. As shown in Fig. 4, out of them, twelve seemed rather effective (apparent EC 50 values in the range of 41.6-229.9 µM). However, all the sixteen compounds tested were found to interfere with the AzMc probe, as they markedly decreased the dye signal (see Supplementary Table S5)  To rule out possible interferences in the MB assay and/or direct reactivity of the compounds with H 2 S, each derivative was tested in preliminary control assays using 3 mM GYY4137 as the H 2 S source (Z'-factor = + 0.84, N = 9; 3 independent experiments for both positive and negative controls, respectively carried out in the presence and absence of GYY4137). As shown in Supplementary Table S6, in these assays none of the 31 compounds caused a marked decrease in the colorimetrically detected H 2 S. The MB method was therefore employed to evaluate the effect of each compound on the H 2 S-producing activity of tCBS, CSE and MST (respectively, with Z'-factors = + 0.57, + 0.78 and + 0.51; N ≥ 9 in at least 3 independent experiments for both positive and negative controls, respectively carried out in the presence and absence of the proteins, see Table 2). By screening the effect of the whole compound library on the activity of each target enzyme, it was found that, whereas none of the tested derivatives has inhibitory effect towards MST, C30 and C31 are poor inhibitors of both tCBS and CSE (Table 2). Indeed, while 1 mM C30 inhibits tCBS and CSE by approx. 50% and 40%, respectively, 0.5 mM C31 accounts for tCBS and CSE inhibition by approx. 40% and 60%, respectively.

Discussion
Hydrogen sulfide (H 2 S) is endogenously produced to accomplish the regulation of numerous physiological processes, ranging from neoangiogenesis 37 , vasorelaxation and blood pressure 38 to cardioprotective 39 , antinflammatory 40 and antioxidant 41 effects. Altered H 2 S metabolism is associated with multiple human pathologies, such as cardiovascular 17 and inflammatory 42 disorders, neurodegeneration 19 and cancer 22,43 . Therefore, the development of compounds targeting the three H 2 S-synthesizing enzymes, namely CBS, CSE and MST, may prove beneficial for future therapeutic strategies, as posited for CBS 20,28 . To date, only unspecific or relatively weak inhibitors have been reported for CBS. While AOAA displays a half-maximal inhibitory concentration (IC 50 ) of approximately 3 μM for human recombinant CBS, it also inhibits human CSE with even higher potency, as well as other PLP-dependent enzymes 25 . Whereas benserazide appears more specific towards CBS, it exhibits lower potency 29 . The present work attempted to identify new CBS, CSE and MST inhibitor scaffolds, using combined biophysical and biochemical approaches. The tested compounds were assembled in a composite library of 31 pyridine derivatives. A combination of orthogonal biophysical and biochemical methods was applied to provide a robust and effective platform to identify putative ligands. Despite the fact that both DSF and SPR are able to detect protein-ligand interactions, though through different mechanisms, frequently in compound screenings the lists of hits identified by either method are only partially overlapping. Both methodologies are typically employed due to the ease of setting up such assays, and to the relatively low amounts of protein required (reviewed e.g. in 32,44,45 ). The usage of both methodologies thus affords a more robust approach 46 .
Herein, DSF data required a rather elaborate analysis due to different factors. The DSF denaturation profiles of tCBS presented a relatively high basal fluorescence (unrelated to protein destabilization at the initial temperature as judged by the CD-monitored thermal denaturation profile) and were best fitted with two apparent T m values (Fig. 2). The discrepancy between the CD-and DSF-monitored thermal denaturation profiles and corresponding T m values could be due to interference of the tCBS cofactors, through fluorescence increase and/or quenching by PLP and heme iron, respectively. Regardless of cofactor interference, DSF could still be employed to evaluate the effect produced by each compound on the protein thermal denaturation profile and four compounds (C2, C14, C28 and C29) were found to mildly affect tCBS thermal stability (|ΔT m_' Ave' | ≥ 1.0 °C), though with |Z-score| < 3.0 (Supplementary Table S1).
Similarly to tCBS, CSE showed high basal fluorescence intensity at resting temperature, possibly related to the presence of the PLP cofactor. Indeed, the CD-monitored thermal denaturation profile revealed CSE to be remarkably stable, ruling out protein instability at resting temperature (Fig. 2). Despite the unconventional DSF thermal denaturation profile, it was still employed to survey possible interactions of the pyridine derivatives with CSE. Based on two parameters (Ratio A and B) quantitatively evaluating profile shape changes, both statistically validated by using the CSE inhibitor AOAA as negative control, six compounds (C1, C2, C9, C17, C30 and C31) were identified as putative interactors for CSE (Supplementary Table S2). These compounds indeed proved to be the only ones causing changes in both parameters with |Z-score| ≥ 3.0.
MST revealed a better agreement between the DSF and CD thermal denaturation profiles, with two well defined transitions likely associated with the two homologous domains that compose MST 47 and other transulfurases, like rhodanese 48 . From the four analysed parameters (Frac, T m1 , T m2 , T m_' Ave' ), compounds putatively interacting with MST were identified based on a combination of defined criteria related to T m shifts and corresponding Z-score values. According to these criteria, only two compounds (C2 and C30) were found to cause shifts in both T m1 and T m2 with |Z-score| ≥ 3.0, and notably C30 caused the greatest ΔT m_' Ave' (−1.4 °C) among all tested compounds (Supplementary Table S3).
Despite the limited identification of strong hits within the tested compound library, DSF assays under the experimental conditions used in this study proved to be suitable to screen compounds, at least for CSE and MST, while further optimization of the assay seems to be required in the case of tCBS, particularly for high-throughput screenings. Indeed, for the assay on CSE, according to 49 , Ratio A exhibited a Z'-factor (+0.53) consistent with an excellent assay, while Ratio B exhibited a slightly lower Z'-factor (+0.31), yet consistent with a good assay (Supplementary Table S1). Likewise, for the assay on MST, two parameters (Frac and T m2 ) were found to be characterized by Z'-factors (+0.50 and +0.68, respectively) consistent with an excellent assay (Supplementary  Table S2). In contrast, for the assay on tCBS, only T m2 exhibited a slightly positive Z'-factor (+0.08), unlike T m1 and T m_' Ave' (displaying Z'-factor values of −1.58 and −0.28) (Supplementary Table S3).
In parallel with the DSF assays, putative interactions of the tested compounds with the three H 2 S-synthesizing enzymes were investigated by surface plasmon resonance (SPR). The disadvantages of using a fluorescence-based technique involving an added dye are overcome by SPR, which is a label-free methodology. Herein a CM5 chip was employed to immobilize all target proteins, based on covalent linkage to the carboxymethylated surface of exposed lysine residues distributed through the protein surface, thus offering different possible binding orientations. Pyridine derivatives were screened at four concentrations, from 25 to 200 μM, which already proved to be above the desired solubility for 15 compounds (approximately half of this library). Despite this limitation and the current lack of adequate controls to statistically validate the use of SPR for high-throughput screening, we observed interactions for a limited number of compounds, based on the proportionality between steady-state response units and compound concentration: 15% of the compounds in the library for tCBS, ~6% for CSE and ~10% for MST (Fig. 3). For these compounds, however, it was not possible to extrapolate binding affinities, as the steady-state response units linearly increased with the compound concentration pointing to low-affinity interactions. The results herein obtained for the pyridine derivatives suggest that the use of SPR in future compound screenings targeting CBS, CSE and MST would require further optimization and proper statistical validation.  The 31 pyridine derivatives were then assayed for their ability to inhibit H 2 S production by tCBS using the H 2 S-selective fluorescent probe AzMC 29 . While 16 compounds were selected as 'positive hits' based on a detailed kinetic analysis evaluating these compounds in a wide concentration range (Fig. 4), control experiments were performed replacing tCBS with the H 2 S donor GYY4137, revealing that all the tested compounds strongly interfered with AzMc detection of GYY4137-generated H 2 S (Supplementary Table S5 and Supplementary Fig. S9). This apparent drawback, similar to the interference of reference CBS inhibitor NSC67078 27 with the AzMc probe reported by Druzhyna and co-workers 29 , further highlights the caveat of fluorescence-based methods, particularly when dealing with compounds and/or fragments with aromatic moieties. The discovery that the presumed 16 'hit' pyridine derivatives interfered with the AzMc fluorescence-based method prompted us to employ the MB method in 96-well plates as described in 29 , adapted to prevent H 2 S loss upon addition of zinc acetate. After confirming that none of the 31 compounds interfered with MB detection of H 2 S released by GYY4137 (Supplementary Table S6), the compounds were screened against the three target proteins. Whereas none of the pyridine derivatives affected MST (Table 2), compounds C30 and C31 appeared to partially inhibit both CBS and CSE (Table 2), albeit only at relatively high concentrations (respectively, 1 and 0.5 mM). Interestingly, C30 and C31 share the same molecular scaffold (Fig. 1). It is worth noting that all fluorimetric and colorimetric assays herein described under the experimental conditions used in the present study proved to be suitable even for high-throughput screenings based on the determined Z'-factor values 49 (Table 2 and Supplemental Tables S4-S6).
Herein, orthogonal biophysical screening methods, based both on protein-compound interaction (DSF and SPR) and on enzymatic inhibition (H 2 S-detection assays), were employed for the first time for the three human H 2 S-synthesizing enzymes CBS, CSE and MST altogether. Testing a library of 31 pyridine derivatives, a relatively low overlap was observed between DSF and SPR outputs in terms of compounds possibly interacting with the target enzymes. The possibility of including functional approaches such as enzymatic inhibition assays in screening campaigns increases the robustness of the resulting hit compound identification. Herein, only two out of the 31 tested compounds, namely C30 and C31, displayed a weak inhibitory activity towards both tCBS and CSE. Notably, compound C31 was also identified by DSF as a hit compound for CSE and likely interactor of the same protein by SPR.
In conclusion, the drug discovery process can be viewed as a bumpy ride where methodological limitations may lead to the identification of false positives as hit molecules. In this regard, the present study further highlights the importance in compound screening campaigns of crossing the read-outs from complementary methodological approaches, ranging from the investigation of biophysical dynamic aspects (such as protein-ligand binding) to the implementation of more functional assays (enzymatic activity). From this perspective, we hope that the experimental setup herein presented, that integrates analysis of protein-ligand interactions by both DSF and SPR, and protein thermal denaturation investigation by Far-UV CD, with activity assays based on fluorimetric and colorimetric H 2 S detection methods, offers a robust platform for the discovery of hit compounds to develop selective and potent inhibitors of the three human H 2 S-synthesizing enzymes.

Methods
Chemicals. All chemicals were purchased from Sigma, except GYY4137 (from CAYMAN), the Protein Thermal Shift Dye Kit ™ (from Applied Biosystems) and the SPR Amine Coupling Kit, type 2 (from GE Healthcare).
Differential scanning fluorimetry (DSF). DSF allows to gain information on ligand binding to a target protein from the observed changes in the protein thermal denaturation profile 57 . Using 96-to-1536-well plates in an RT-PCR instrument, protein denaturation is monitored as a function of temperature increase by making use of a fluorescent dye that emits light upon binding to the buried hydrophobic amino acid residues that become SCIENTIfIC RePoRtS | (2019) 9:684 | DOI:10.1038/s41598-018-36994-w exposed as the protein unfolds. Typically, thermal denaturation profiles are sigmoidal and the protein melting temperature (T m ) is estimated from the inflexion point. Ligand binding is evaluated from the ability of a given compound to either stabilize or destabilize the target protein, which is reflected in an increase or decrease in the T m . Here, DSF measurements were carried out in 384-well plates in an Applied Biosystems QuantStudio ™ 7 Flex Real-Time PCR. Prior to testing the compounds, the assay was optimized for each target protein in terms of concentration of the proteins (1 to 6 μM) and the fluorescent dye (1x, 2x and 4x; the Protein Thermal Shift Dye from Applied Biosystems is commercially available in a 1000 x concentration, without disclosing the actual molar concentration

Surface plasmon resonance (SPR). Putative interactions of the tested compounds with tCBS, CSE, and
MST were investigated by surface plasmon resonance (SPR). Assays were carried out in a Biacore 4000 instrument (GE Healthcare) at 25 °C. First, a pH scouting was performed on a CM5 chip for immobilization optimization using the following buffers: 10 mM sodium acetate pH 5.0, 5.5 and 5.8, 10 mM Bis-Tris pH 6.0 and 6.5, 10 mM sodium phosphate pH 7.0, 10 mM HEPES pH 7.5, or 10 mM Tris-HCl pH 7.5. tCBS, CSE and MST were diluted to 5 µg/mL (tCBS) or 10 µg/mL (CSE and MST) in their corresponding immobilization buffer (10 mM sodium acetate pH 5.8 for CBS, and 10 mM Bis-Tris pH 6.0 for CSE and MST) and immobilized onto CM5 sensor chips using the standard amine coupling procedure. HBS-N, which consisted of 10 mM Hepes pH 7.4 and 150 mM NaCl, was used as the background buffer. Prior to immobilization, the carboxymethylated surface of the chip was activated with 20 mM 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide and 5 mM N-hydroxysuccinimide for 1.5 min. Proteins were coupled to the surface with a 5 to 10 min injection time at a flow rate of 10 µL min −1 in order to reach 5000 to 10000 response units (RU). The remaining activated carboxymethylated groups were blocked with a 5 min injection of 1 M ethanolamine pH 8.5.
Compounds (analytes) were directly diluted in running buffer (10 mM HEPES, 150 mM NaCl, 1 mM TCEP, 0.1 mM EDTA, 0.05% (v/v) TWEEN-20, 5 mM MgCl 2 , pH 7.2.) and injected at four different concentrations using 2-fold dilutions series, with the highest concentration tested being 200 µM. Interactions were qualitatively assessed from the obtained plots of steady-state analyte binding levels against the concentration, making use of the provided Biacore 4000 evaluation software (GE Healthcare).
Fluorimetric H 2 S-producing activity assays. Fluorimetric H 2 S production activity assays were adapted from Thorson et al. 33 . Assays were carried out in 96-well black plates, using the H 2 S-selective fluorescent probe AzMC and a FLUOstar Optima BMG Labtech plate reader. The reaction mixture, in 200 mM Tris-HCl pH 8.0, contained 1.12 µg recombinant human tCBS per well (100 nM), 0.5 mM homocysteine, 50 µM PLP, and 50 µM AzMc (diluted from a 49.7 mM stock in DMSO). Compounds dissolved in DMSO were added to each well to yield a final concentration ranging from 15.6 µM to 1 mM (5% DMSO) in a total assay volume of 250 µl. The plate was incubated at 37 °C for 10 min and the reaction was then triggered by adding the 10 mM l-cysteine. The increase in the probe fluorescence (λ excitation = 340 nm; λ emission = 460 nm) was monitored over 1 hour at 37 °C. The reader was set up to automatically shake the plate for 5 seconds prior to each data acquisition. Data were analysed using Excel and activity was calculated from the initial slope of the fluorescence increase after l-cysteine addition. Control experiments to evaluate a possible interference of the tested compounds with the probe were done by replacing tCBS with the H 2 S donor GYY4137 (3 mM in DMSO, final concentration).
Counterscreen using the methylene blue assay. The colorimetric methylene blue assay, commonly used to detect H 2 S, was adapted to 96-well plates, as reported by Druzhyna et al. for CBS 29 , and improved in order to i) avoid H 2 S escape from the reaction mixture and headspace before trapping sulfide with zinc acetate, and ii) be extended to the other two H 2 S-synthesizing enzymes (CSE and MST). Prior to the enzymatic inhibition assays, SCIENTIfIC RePoRtS | (2019) 9:684 | DOI:10.1038/s41598-018-36994-w control absorbance measurements were performed in 96-well plates: at λ = 600 nm to evaluate the compounds solubility under the assays conditions, and at λ = 690 nm to evaluate their interference with the methylene blue read outs (and employ a correction factor whenever needed). For CBS-and CSE-catalyzed H 2 S production assays, 110 µL of reaction mixture contained 0.5 mM or 1 mM of tested compound, tCBS (10 µg/well) or CSE (30 µg/well) in 50 mM Tris-HCl pH 8.0, and 5 µM PLP. In the case of tCBS, 2 mM homocysteine was also added. After keeping the plate on ice for 30 min, the reactions by tCBS and CSE were respectively triggered by adding either l-cysteine (10 mM final concentration) alone or plus homocysteine (2 mM final concentration), The plate was sealed with vinyl adhesive films and incubated for 2 h at 37 °C. Regarding MST, a 110 µL-reaction mixture contained 0.5 mM or 1 mM of each compound, 30 µg/well MST in 50 mM Tris-HCl pH 8.0, and 10 mM dithiothreitol. After keeping the plate on ice for 30 min, the reaction was triggered by adding 1.5 mM sodium 3-mercaptopyruvate. The plate was sealed with vinyl adhesive films and incubated for 1 h at 37 °C. After the above mentioned incubation times, the CBS, CSE and MST reaction plates were kept on ice for 15 min. Then, 110 µl of 4% zinc acetate were dispensed to each well by punching a hole through the strip with a gas-tight Hamilton syringe. After 15-min incubation on ice, the films were removed and 15 µl of 60 mM NNDPD (3.6 mM final concentration) and 15 µl of 90 mM FeCl 3 (5.4 mM final concentration) were dispensed to each well (final volume: 250 µl). The plate was incubated for 10 min at room temperature in the dark and the absorbance measured afterwards at 690 nm. Data were acquired using the plate reader Thermo Scientific Appliskan Multimode. Statistical Analysis. The Z'-factor was evaluated for each assay as described in 49 . Data were reported in the tables with their corresponding Z-score and coefficient of variation (CV). A |Z-score| threshold value of 3 was assumed to evaluate the effect of the screened compounds. All calculations were performed using Excel (Microsoft).

Data Availability
The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.