Targeting collagen biosynthesis by small molecules inhibiting the function of the peptide-substrate-binding domain of collagen prolyl 4-hydroxylases Authors: Bukunmi Adediran, Carlos Vela-Rodríguez, Sudarshan Murthy, M Mubinur Rahman, Rik K Wierenga, Lari Lehtiö, M Kristian Koski Published: 2026 ABSTRACT AbstractCollagen prolyl 4-hydroxylase (C-P4H) is an essential enzyme in collagen synthesis and known to be a potential target for drugs that prevent excess collagen formation. Currently known C-P4H inhibitors target the catalytic site of C-P4H, being analogues of 2-oxoglutarate (2OG). However, in mammalian cells there are many other 2OG-dependent dioxygenases with a highly similar catalytic domain, which limits the selective specificity of drugs targeting C-P4H activity. The peptide-substrate-binding (PSB) domain is unique for the C-P4H family and known to be important for the catalytic efficiency of C-P4H. Therefore, interfering with peptide binding to the PSB domain might allow more specific inhibition of the hydroxylation activity of C-P4Hs. We developed a robust FRET-based high-throughput screening assay (Z’ > 0.73) based on PSB-peptide interactions. This assay was used to screen a peptidomimetic library of 15614 compounds with the PSB domains of C-P4H isoforms I and II. A hit compound (OUL-PSBi-001) was identified with IC50s of 40 µM and 62 µM for PSB-I and PSB-II, respectively. We also showed that this compound indeed inhibits the catalytic activity of the full-length CP4H-I and -II. This FRET assay provides a new strategy for finding selective inhibitors for the treatment of fibrotic diseases and cancer.Keywords: Peptidomimetics, Fibrotic diseases, FRET, High-throughput screening, Drug discoverySubject terms: Biochemistry, Chemical biology, Chemistry, Drug discovery Introduction Collagens are extracellular matrix (ECM) proteins, which provide mechanical support, participate in cellular development and communication, and are important for tissue repair and wound healing1. Common for all collagen polypeptide chains is a region that has repeating X-Y-Gly sequence triplets, which allows the oligomerization of collagens into the typical triple helical structure1,2. The proline 4-hydroxylation at position Y is catalyzed by enzymes called collagen prolyl 4-hydroxylases (C-P4Hs)3–5. Three C-P4H isoforms in humans have been identified, which are referred to as C-P4H-I, -II, and -III3. C-P4Hs require 2-oxoglutarate (2OG), iron (Fe2+), molecular oxygen (O2) for its prolyl hydroxylase activity, whereas ascorbic acid (vitamin C) is required for being able to catalyze multiple hydroxylation reactions of procollagen6,7. C-P4H-I is the most abundant and most important isoform, whereas C-P4H-II is more tissue and cell type specific8,9. During embryonic development collagen synthesis has a critical role, and C-P4H-I knock-out mice have been shown to be embryonically lethal10. In healthy adults, collagen synthesis is less critical, although prolonged C-P4H deficiency, due to the lack of vitamin C, leads to scurvy disease5. Excessive C-P4H activity is associated with fibrosis caused by the accumulation of collagens11. Fibrosis is a major health issue, and an effective way to prevent collagen accumulation is to inhibit collagen synthesis by inactivating the C-P4Hs5,6,12. Cancer metastasis is correlated with significant changes of the ECM properties, associated with increased collagen synthesis13,14, and it has been reported that inhibition of C-P4H-I and C-P4H-II activity dramatically decreases spontaneous metastasis in many cancers14–16. C-P4Hs in vertebrates are α2β2 heterotetramers of about 240 kDa, consisting of two catalytic α-subunits and two protein disulfide isomerases (PDI), which act as β-subunits (Fig. 1a)3,5. The C-terminal half of the α-subunit contains the key residues for Fe2+and 2OG binding and is referred to as the catalytic (CAT) domain, as it is responsible for catalyzing the hydroxylation process17,18. The CAT domain interacts tightly with three thioredoxin domains of the β-subunit, namely with domains a, b’ and a’18 (Fig. 1a). The N-terminal half of the α-subunit contains two helical domains referred to as the N-terminal dimerization domain and the middle peptide-substrate-binding domain (PSB). In the current structural model of the α2β2 tetramer, the PSB domain does not interact with the β-subunit but instead protrudes out of the elongated tetrameric molecule, which is capped by the β-subunits (Fig. 1a)18. Organization of the C-P4H complex. (a) The C-P4H tetramer is an elongated molecule, where the α2-dimer (monomers in magenta and red) forms the core of the complex via its dimerization domains (or N domain), whereas the CAT domains of the α-subunit interact with the PDI/β-subunits (light blue), which consist of four domains (a,b,b’ and a’)18. The N, PSB, and CAT domains of the “upper” chain of the α2-dimer are colored in magenta, purple and magenta, respectively, whereas the domains of the “lower” α chain are red. The PSB domains are protruding out of the core of the tetramer complex and interact with the dimerization (N) domain and the CAT domain. The proposed binding mode of the procollagen peptide (yellow) to the PSB and CAT domains is also indicated. (b) Crystal structure of PSB-II with bound PAG peptide (PDB ID 6EVN)25. The PSB domain is composed of five α-helices H7-H11. The PAG peptide binds to this domain in such a way that the proline of the central Pro-Ala-Gly triplet occupies the P5 pocket (lined by Tyr196, Phe199 and Tyr230), whereas the first proline of the following Pro-Pro-Gly triplet is buried in the P8 pocket (lined by Tyr158, Tyr193, Tyr196, and the salt bridge Asp192-Arg223). The protein-peptide interactions concern mainly the stacking interactions between tyrosine side chains of the domain and prolines of the peptide. The side chains of Asn227, Arg223 and Tyr158 form the key protein-peptide hydrogen bonds (as visualized by dotted lines). The three C-P4H isoforms all have different α-subunits, each encoded by its own gene19,20. The α-subunit of each isoform contains the PSB domain which consists of five α-helices (Fig. 1b) and it is thought to be important in anchoring the long procollagen chain during its hydroxylation18,21–24. In the current C-P4H model, the peptide binding sites of the PSB and CAT domains are properly aligned for simultaneous binding to an extended peptide (Fig. 1a)18. It has been shown that tyrosine-to-alanine point-mutations of the tyrosines that line the peptide-substrate binding groove of the PSB domain (especially tyrosines 193, 196 and 230, see Fig. 1b) lower the activity of the full-length C-P4H-I for the proline-rich model peptide substrate (PPG)10 indicating that the PSB domain is important for efficient proline hydroxylation by the CAT domain23. The PSB domains of C-P4H-I and C-P4H-II (henceforth referred to as PSB-I and PSB-II, respectively), have a sequence identity of 65%, and they have somewhat different peptide binding specificity and properties21,24–26. Crystal structures of both PSB-I and PSB-II in apo form and in complex with short proline-rich peptides have been reported (Fig. 1b)23–26. Both PSB domains have been co-crystallized with a procollagen-like peptide PPG-PAG-PPG (also referred to as the PAG peptide), which has the binding affinity (KD) to PSB-I and PSB-II of 91 µM and 20 µM, respectively25,26. Since its discovery, C-P4H has been thought to be a possible pharmaceutical target due to its role in promoting fibrosis. The identified inhibitors so far described in the literature are metal ions which replace the active site Fe2+, or cofactor analogues which replace 2OG in the active site of the CAT domain3,5,6. However, because C-P4Hs share a common catalytic domain with other P4H homologs and other 2OG-dependent dioxygenases27,28, designing specific and selective inhibitors for the CAT domain of C-P4Hs is challenging. Currently there are no C-P4H inhibiting drugs in the market. We rationalized that a feasible alternative approach to inhibit C-P4H activity could be to inhibit the binding of procollagen to the PSB domain as this domain is specific for C-P4Hs and important for its catalytic function. Herein, we present a high-throughput (HTS) screening assay for finding inhibitors of the PSB domain function based on Förster resonance energy transfer (FRET), which has been used successfully to probe protein-protein interactions of other complexes29. After screening a library of 15,614 peptidomimetic compounds with PSB-I and PSB-II, we identified one potential compound which binds to the PSB domain of both C-P4H orthologues. Our findings support the concept that inhibitors of PSB peptide binding are also inhibitors of the C-P4H catalytic activity. The discovered compound is the first reported C-P4H inhibitor targeting the C-P4H PSB domain. Results CFP-tagged PSB-I and PSB-II bind to the YFP-PAG peptide resulting in FRET signals As a starting point for developing the FRET screening assay targeted to find binders for the PSB domain, it was known from the previous studies that PSB-I and PSB-II domains bind polyproline peptides and peptides containing only PPG triplets with different affinities21–25. However, it was also shown that both constructs bind the PAG peptide by using the same mode of binding and with affinity better than 100 µM25,26, which would allow making a screening assay suitable for both isoforms. The development of the screening assay was started with the PSB-II domain. The N-termini of the PSB-II domain and the PAG peptide were fused to the C-terminus of the monomeric variants of CFP and YFP, respectively (Fig. 2). This fusion allows for the use of FRET to monitor and analyze peptide binding. When the CFP-fused protein and the YFP-fused peptide form a complex, the proximity of the fluorophores facilitates energy transfer from CFP to YFP. The presence of an inhibitor of peptide binding disrupts this complex formation, leading to a reduction in energy transfer (Fig. 2). This inhibition of peptide binding is manifested therefore as a higher emission at the CFP emission maximum wavelength (477 nm) and a lower emission at the YFP emission maximum wavelength (527 nm). To quantify the amount of energy transferred from CFP-PSB-II to YFP-PAG when present in the complex, the ratiometric FRET value (rFRET) is used. This rFRET value is obtained by dividing the fluorescence measured at 527 nm by the fluorescence measured at 477 nm. A lower rFRET value indicates that a lower amount of the complex is present in the reaction mixture, because of the binding of an inhibiting small molecule (Fig. 2). Schematic representation of the FRET principle: the binding of a small molecule (indicated by a star) disrupts the interaction between the PSB domain (tagged with CFP at its N-terminus) and the PAG peptide (tagged with YFP at its N-terminus). The rFRET value, the ratio of the emission at 527 nm divided by the emission at 477 nm, is low in the presence of an inhibitor. Assay parameters such as concentration, molar ratio, buffer, and reaction volume were tested to achieve a reproducible and robust signal for the CFP-PSB-II – YFP-PAG pair. The optimized conditions of the assay mixture included 10 mM Bis–Tris-Propane (pH 7.0), 3% (w/v) PEG 20K, 0.01% (v/v) Triton-X100 and 0.5 mM TCEP working buffer, 100 nM CFP-PSB-II concentration, 1:20 CFP: YFP ratio, and reaction volumes ranging 10 to 30 µL. These assay conditions were found to be suitable also for the CFP-tagged PSB-I construct. The statistical fluctuations observed in the measurements for both assays were found to be very similar (Figs. 3a, b). Based on the Z’ number calculated from the plate validation (0.79 for PSB-I and 0.73 for PSB-II), the assay can be considered excellent for inhibitor screening30. Given that compounds are typically dissolved in DMSO, its effect on the assay was investigated with the PSB-II construct. We observed that the assay can tolerate DMSO up to 5%, which exceeds the percentage of DMSO used in these studies (Fig. 3c). Initial experiments utilized free YFP to confirm that the recorded rFRET signal was indeed a consequence of the specific interaction between PSB and the peptide of the PSB-peptide complex, rather than non-specific energy transfer. We have earlier reported for another system that in the presence of 1 M guanidine hydrochloride (GndHCl), the protein-protein interactions are effectively disrupted with little to no effect on fluorescence properties of the stable YFP and CFP tags31. Therefore, this condition was used as a positive control in the inhibitor screening. The FRET-based assay is suitable for screening of small molecules and determine affinity towards a model binding peptide. (a) Plate validation for CFP-PSB-I. (b) Plate validation of CFP-PSB-II. (c) DMSO tolerance test of the assay for the CFP-PSB-II / YFP-PAG FRET pair. The reaction buffer supplemented with 1 M GdnHCl was used as a positive inhibition control. The data presented in (a) and (b) corresponds to the mean and SD of 176 replicates per plate. The data in (c) corresponds to the mean and SD of four replicates. The concentration of CFP-PSB used was 100 nM and for YFP-PAG was 2 µM. (d) FRET-based saturation curves for binding of the PAG peptide to PSB-I and PSB-II. The measurements show the mean and standard deviation from three independent experiments with 4 technical replicates wherein the concentration of CFP-PSB-I (triangle) and CFP-PSB-II (dot) was kept constant at 200 nM. Dissociation constants for the PSB: PAG interactions measured with FRET To validate the assay and confirm that the fluorescent tags did not interfere with peptide binding, the FRET method was used to measure binding affinity of the CFP-tagged PSBs and YFP-tagged PAG peptide. While FRET has been used to determine dissociation constants (KD) for synthetic fluorophores32, it has also been adapted for fluorescent proteins31. The rFRET value was calculated from the fluorescence data and the saturation curve was plotted with a KD values of 12.1 µM and 17.5 µM for PSB-I: PAG and PSB-II: PAG complexes, respectively (Fig. 3d). These confirm that the CFP and YFP tags do not reduce binding affinity and that both PSBs would have similar affinity to the peptide. It can be noted that the previously reported values of PSB-I: PAG and PSB-II: PAG are 92 µM and 19 µM, respectively25,26. The latter values were obtained by calorimetry using the untagged PSB construct and peptide, suggesting that at least for PSB-I the fusion has an influence on the peptide binding affinity. The FRET-based inhibitor screening identifies a compound which binds to both PSB domains with high affinity A small-molecule library consisting of 15,614 peptidomimetic compounds in assay plates provided by the Institute for Molecular Medicine Finland (FIMM) and used for the initial screening step of both PSB-I and PSB-II (Figs. 4a, b). The used compound concentration in the screen was 20 µM and the DMSO concentration was less than 0.2%. The screening data were analyzed with multiple filters to eliminate non-specific FRET signals. Assay mixtures for which the fluorescence with compounds was more than 20% greater than the fluorescence of the negative control at 527 nm were considered outliers and eliminated. Furthermore, fluorescence of samples with compounds more than 20% greater than the positive control at 477 nm were considered as outliers as well. Also, compounds showing signs of quenched fluorescence were eliminated. After applying these filters, we set the hit limit of the rFRET value to 40% (Figs. 4a, b). The initial screening yielded 57 positive hits, and these compounds were rescreened subsequently with PSB-I and PSB-II (Fig. 4c). The 57 hit compounds mostly inhibited both PSBs but there is some clear variation between the inhibition efficacies (Fig. 4c). Identification of chemical compounds that inhibit the interaction between PSB and a proline-rich peptide using a FRET-based assay. Screening of the peptidomimetic library for: (a) PSB-I and (b) PSB-II. The red dotted line represents the inhibition threshold used to select compounds from the primary screening. (c) Correlation of the apparent inhibition of compounds selected from (a) and (a), plotted with PSB-I on the x-axis and PSB-II on the y-axis. Open circles highlight the compounds which were ordered and studied further. The red open circle corresponds to the OUL-PSBi-001 (d) Covalent structure of the hit compound OUL-PSBi-001. The quinoxaline core is on the left side, and the N-phenyl moiety is on the right side. (e) Determination of the IC50 values by plotting rFRET versus the logarithm of inhibitor concentration (M) for the hit compound against PSB-I; (f) corresponding IC50 determination for PSB-II. To identify hits with significant potency, we purchased the six most potent hit compounds, which were available from commercial vendors, for potency measurements (Fig. 4c). The best hit was a compound named 2-[6,7-dimethyl-2-oxo-3-(trifluoromethyl)-1,2-dihydroquinoxalin-1-yl]-N-phenylacetamide (Fig. 4d, referred to as OUL-PSBi-001 from now on), which showed full inhibition at 100 µM. The measured IC50 for PSB-I was 40 µM and for PSB-II 62 µM (Figs. 4e, f). The binding of OUL-PSBi-001 to the PSB domains was also validated with the nano differential scanning fluorimetry (NanoDSF) by using the PSB domains without the CFP tag. Both PSB isoforms showed slightly higher melting temperatures in the presence of 100 µM OUL-PSBi-001 in the solution. The Tm of the PSB-I domain increased from 63.8 °C to 64.2 °C, and the Tm of the PSB-II domain increased from 62.4 °C to 62.9 °C. Although the increases in thermal stabilities are small they are consistent with binding of this compound to the PSB domains and show that OUL-PSBi-001 does not cause inhibition of peptide binding by unfolding of the PSB-I or PSB-II domains. OUL-PSBi-001 inhibits the catalytic hydroxylase activity of C-P4H-I and C-P4H-II To measure the effect of OUL-PSBi-001 on the hydroxylase activity of C-P4H-I and C-P4H-II, we established a microplate-based bioluminescence assay using full-length human C-P4H-I and C-P4H-II, purified from an insect-cell expression system33. The assay is the same as used previously for measuring the catalytic activity of lysyl hydroxylase enzymes34. In this assay, succinate formation is monitored with the Succinate-Glo JmjC demethylase/hydroxylase assay kit (Promega Corboration), which converts succinate generated by the C-P4H hydroxylase reaction into a luminescent signal (Fig. 5a). The proline-rich peptide (PPG)₁₀ was used as the substrate, as it is a well-characterized C-P4H substrate9. The assay mixture including C-P4H-I (5 µg/mL) or C-P4H-II (5 µg/mL) and peptide substrate (100 µM) produced a high luminescent signal after 30 min reaction time (Fig. 5b). Including 100 µM OUL-PSBi-001 in the reaction mixture reduces the background corrected signal for both C-P4H-I as well as C-P4H-II by about 50% (Fig. 5b) showing that OUL-PSBi-001 inhibits the activity of both full-length C-P4H isoforms. Inhibition of full-length C-P4Hs with OUL-PSBi-001 measured using a bioluminescence-based C-P4H activity assay. (a) Schematic overview of the assay principle. (b) Activity assay results of C-P4H-I (upper panel) and C-P4H-II (lower panel) showing the uninhibited enzyme reaction (“E”), the control reaction not including any enzyme (“No E”), and the inhibited reaction including enzyme and 100 µM OUL-PSBi-001 (“E + I”). Docking studies show that the OUL-PSBi-001 compound fits in the peptide-binding groove of the PSB domain Docking studies were performed using the crystal structures of PSB-I (PDB ID 9HT8) and PSB-II (PDB ID 6EVN) complexed with the PAG peptide25,26. The docking results propose that the compound binds in the peptide-binding groove of both PSB isoforms in such a way that the quinoxaline core binds to the P5 pocket and the N-phenyl moiety occupies the P8 pocket (Figs. 6a, c). The two methyl substituents of the quinoxaline moiety point inwards and have van der Waals interactions with the protein and the trifluoro methyl group points to the solvent region. Superimpositions of the OUL-PSBi-001 docking models with the crystal structures of the PAG-complexed PSB-I (Fig. 6b) and PSB-II (Fig. 6d) show that the quinoxaline core aligns well with the peptidyl proline bound to the P5 pocket, whereas the N-phenyl group aligns well with the peptidyl proline bound in the P8 pocket. In the docking models of both PSB isoforms, two direct hydrogen bonds are found between, respectively, the side chains of Asn227 and Arg223 and the compound carbonyl oxygen (Figs. 6b, d). These hydrogen bonds correspond to the direct hydrogen bonds seen in the crystal structures of the PSB-peptide complexes. In these complexes, the side chains of Asn227 and Arg223 are hydrogen bonded to the back-bone carbonyl oxygen of the alanine of the PAG peptide. The other critically important hydrogen bond found in the PSB-peptide complexes, between the Asn227 side chain and the backbone peptide NH of the alanine of the PAG peptide (Fig. 6), is not found in the docking models of PSB and OUL-PSBi-001 due to the absence of a hydrogen bond donor (corresponding to the free peptide NH of the alanine residue of the bound peptide) in the docked compound. Docking models of compound OUL-PSBi-001 with the PSB domains. (a) The OUL-PSBi-001 compound (yellow sticks) modelled in the peptide binding groove of PSB-I (light brown surface) in such a way that the quinoxaline moiety (left part) and the N-phenyl group (right part) of this compound occupy the P5 and P8 pockets, respectively. (b) The proposed mode of binding of the OUL-PSBi-001 compound mimics the binding of the PAG-peptide [white sticks, taken from the PSB-I PAG complex structure (PDB ID 9HT8)]. The model also proposes two direct hydrogen bonds between the carbonyl oxygen of the inhibitor and the PSB domain as highlighted in the figure with the black dashed lines. (c) The docking model of OUL-PSBi-001 with PSB-II is highly similar to the PSB-I docking model. (d) The proposed hydrogen bond interactions between PSB-II and OUL-PSBi-001 (black dotted lines) are the same as in the corresponding complex of PSB-I and OUL-PSBi-001 shown in panel (b). This view, slightly rotated compared to panel (b), shows that the second amide moiety of the compound (black arrow) is not free to form a hydrogen bond with Asn227 like seen in the peptide complexes. Also highlighted in panels (b) and (d) (by orange dotted line) are the hydrogen bond interactions between the alanine of the peptide and the side chain of Asn227 of PSB. Discussion C-P4Hs are long-known potential targets for treatment of fibrotic diseases12 and more recently also for treatment of cancer progression14. Many 2OG analogues (e.g. pyridine 2,4-dicarboxylate, 3-hydroxypyridine-2-carbonyl-glycine, oxalylglycine, 3-carboxy-4-oxo-3,4-dihydro-1,10-phenanthroline) and metal ions (e.g. Zn2+, Co2+) are known C-P4H inhibitors. These compounds bind, respectively, to the 2OG and Fe2+ binding sites of the CAT domain, and prevent the hydroxylation reaction, but not the peptide binding6. These compounds also inhibit other mammalian P4Hs like hypoxia inducible factor (HIF) prolyl 4-hydroxylase (HIF-P4H, also referred to as PHD2)6, which have a highly homologous CAT domain35. Inhibitors against HIF-P4H have been studied intensively during the past twenty years. The first inhibitor, a 2OG analogue Roxadustat/Evrenzo, has been accepted for treatment of renal anemia in several countries36,37. Roxadustat has also been shown to inhibit C-P4Hs with similar efficacy38. In addition, there are several other 2OG-dependent dioxygenases in human cells with a similar CAT domain fold27, and inhibitors targeted to the C-P4H CAT domain can also reduce the activity of these other enzymes. There is therefore an urgent need to develop new strategies for finding inhibitors of the C-P4H enzymes and targeting their PSB domain would be an excellent approach. The PSB domain is not found in HIF-P4H, and other 2OG-dependent dioxygenases and its peptide binding properties have been shown to be important for the catalytic activity of C-P4H. In particular it was found that point mutation variants of C-P4H-I in which tyrosine residues of the P5 and P8 pocket were mutated into an alanine have KM values of the model substrate (PPG)10 that are increased from 30 µM for wild type C-P4H-I to about 250 µM23. The precise role of the PSB domain in the reaction mechanism of C-P4H is not understood, but it is thought to act by providing an anchoring binding site for the procollagen chain during hydroxylation by the catalytic domain18. This information has provided the basis for our new approach aimed at finding, ultimately, drugs or peptide analogues that inhibit the C-P4H enzymes by binding to the PSB domain. Noteworthy, a recent study on the interaction between EVH1 (Ena/VASP homology domain 1) and its cognate proline-rich peptide has shown that high affinity, and membrane-permeable inhibitors of binding of such peptides to their acceptor indeed can be developed, despite the notion that the binding site of a proline rich peptide is a shallow groove39,40. For this approach, we developed a new FRET-based HTS platform to identify compounds that bind to the PSB domain. The isolated PSB domains are suitable for HTS as they can be produced efficiently recombinantly in the E. coli host, and they are small and stable proteins21–23. Furthermore, our extensive structural and biophysical peptide-binding studies have shown that the PAG peptide binds to both PSB domains in a similar way and with similar affinity25,26. The YFP-tagged PAG peptide can be produced in a straightforward manner in E. coli and therefore this tagged peptide is an optimal component in this screening method with the CFP-tagged PSB-I and PSB-II domains. The assay was optimized and validated leading to the generation of a highly robust and reproducible signal with a Z’-factor > 0.73. In contrast to other developed inhibitor screening assays of C-P4H-I based on the quantification of succinate produced by the hydroxylation reaction41, this FRET method directly quantifies the interaction of the PSB domain and the procollagen peptide. Therefore, the identified hit is expected to inhibit the PSB-peptide interaction and not the catalytic activity of the CAT domain of C-P4H. Because the PSB-(PPG)10 interaction is important for the full C-P4H hydroxylation activity, it is expected that PSB binders will be inhibitors of the hydroxylation activity by C-P4H. Indeed, our results confirm that the hit compound OUL-PSBi-001 inhibits the C-P4H reaction (Fig. 5b) likely due to inhibition of the peptide binding to the PSB domain although at high concentration needed for the inhibition it could also affect the activity of the full complex in some other manner. It is not known how the PSB-(PPG)10 interaction facilitates the hydroxylation activity by the C-P4H active site, which could be mediated by large conformational changes of the C-P4H complex. In any case the IC50 value of inhibition of the C-P4H activity by OUL-PSBi-001 (about 100 µM) is in the same range as the IC50 value of inhibition of peptide binding by OUL-PSBi-001 (40 µM and 62 µM for PSB-I and PSB-II, respectively), which is consistent with the notion that both effects are caused by the same binding phenomenon. The hit compound, OUL-PSBi-001 (PubChem CID: 50760839) is a peptidomimetic with two amide bonds and a quinoxaline core (Fig. 4d). Quinoxaline-based analogues are known drug molecules. They are used to inhibit tyrosine kinases in cancer therapy42 and known to bind to the ATP-binding pocket of the kinases as shown by the crystal structure of JAK243. Our studies show that the developed FRET assay can identify small molecule inhibitors of the formation of PSB-peptide complexes from a large library of compounds and that it can be used with the two most abundant PSB isoforms (PSB-I and PSB-II) for the discovery of compounds that inhibit the PSB function. The identified hit compound provides the first step in the development of potent PSB binders. Structural studies of the PSB domains complexed with this inhibitor will provide more insight into its binding mode and allow for a structure-based design of tighter binding analogues. As the compound also shows inhibition of the C-P4H enzymatic activity with the peptide (PPG)10 as substrate, it also allows for mechanistic studies of how the two C-P4H activities, peptide binding by the PSB domain and hydroxylation activity by the CAT domain, are acting in concert to hydroxylate procollagen. Methods Cloning Plasmids containing the α-subunits of human C-P4H-I and -II18, were used as template for the generation of amplicons coding for the PSB domains. Amplicons were then cloned by sequence and ligation independent cloning (SLIC) to a linearized pNIC-CFP vector31. The generation of the YFP-PAG peptide was done by amplifying the mCitrine sequence with a 3’ extension coding for the desired amino acid sequence; the amplicon was subsequently cloned by SLIC to a linearized pNIC28-BSA4 vector. Both vectors contain N-terminal 6x His tags for purification purposes. The SLIC reaction was done with a 1:3 ratio transforming the product to XL-1 blue competent cells, which were grown overnight at 37 °C on LB agar plates containing 5% sucrose and 50 µg/mL kanamycin. Plasmids were extracted from a single colony, and the cloning was confirmed by sequencing in the Biocenter Oulu sequencing core facility. Protein expression and purification The fusion proteins were expressed in E. coli BL21(DE3) in 1 L of terrific broth-base auto-induction media supplemented with glycerol (8% w/v). The cultures were incubated at 37 °C and upon reaching an OD600 value of 0.8 the temperature was decreased to 18 °C for overnight incubation. Cells were harvested by centrifugation at 5020 ×g and re-suspending in 15 mL of lysis buffer [50 mM HEPES (pH 7.5), 500 mM NaCl, 10% v/v glycerol, 0.5 mM TCEP, 10 mM imidazole]. The re-suspended cells were flash frozen in liquid nitrogen and stored at -20 °C until purified. The fusion proteins were purified using a 5 mL Hi-Trap IMAC HP column (cytiva) charged with NiSO4 and equilibrated with a 20 mL lysis buffer. Proteins were eluted with elution buffer (lysis buffer with 300 mM imidazole). YFP-PAG and CFP-PSB-I eluted as pure protein, as judged by SDS-PAGE, whereas CFP-PSB-II was further purified with size-exclusion chromatography (SEC). The fusion proteins were flash frozen and stored at -70 °C. FRET-assay development and optimization The FRET assay herein presented was developed using a previously validated approach31, in which the interaction of two binding partners bring the fluorescent proteins to proximity thus promoting the transfer of energy from CFP to YFP (Fig. 2). For the assay, the fluorescence intensities at 477 nm and 527 nm were measured with an Infinite Pro M1000 microplate reader (Tecan) using an excitation 430 nm as the excitation wavelength. The rFRET value was calculated by dividing the intensity at 527 nm by the intensity at 477 nm. Initially, we determined the effect of the ratio between CFP-PSB-II and YFP-PAG based on previous reports of this interaction25. For this, we maintained CFP-PSB-II at 0.5 µM while varying the concentration of YFP-PAG in ratios of 1:2, 1:5, 1:10, 1:15 and 1:20 (PSB: PAG) in a total volume of 30 µL in Nunc 384-well flat bottom polypropylen plates (Fisherbrand). YFP lacking the PAG-peptide was used to assess whether rFRET value was specific to the PSB: PAG interaction or otherwise caused by unspecific interaction of the fluorophores at high concentrations. We subsequently optimised the protein concentration by determining the rFRET value as previously while varying the CFP concentration (25, 50, 100, 200, 400 and 500 nM) at a constant 1:20 ratio. As a final parameter before proceeding with the compound screening, we measured the rFRET value with reactions containing 100 nM CFP-PSB-II (1:20 CF: YFP ratio) in 10, 20 and 30 µL. For all stages of the optimisation we used 10 mM Bis–Tris-Propane (pH 7.0), 3% (w/v) PEG20K, 0.01%(v/v) Triton-X100 and 0.5 mM TCEP as the reaction buffer. Dissociation constant determination To determine the dissociation constant of the CFP-PSB-II: YFP-PAG pair, the concentration of the CFP-PSB-II was kept constant (200 nM) while the concentration of YFP-PAG was gradually increased from 0 to 15 μM. 10 µL reaction volume and 4 replicates were used per concentration condition. Each sample was excited at 430 nm and the emission fluorescence was measured at 477 nm and 527 nm. In addition, samples were further excited at 477 nm and emission fluorescence was measured at 527 nm. The data was analyzed using the GraphPad prism 9.0 as described by Song et al.32. Assay validation The majority of HTS systems only allow for single testing of compounds, and therefore adequate sensitivity and high degree of accuracy are major requirements of the assays30. Assays for HTS also require reproducibility to accurately identify hit compounds among a very large number of compounds. After optimizing the different constituting parameters of the assay, the validation experiment was done by adapting the 3-day plate uniformity test. This uniformity test requires that for newly developed assays, the optimized assay conditions are tested for 3 consecutive days with no change in the conditions44. A total of five 384-well plates were used for the experiment by testing one plate each on day one and day 2 while three plates were tested on day 3 to understand the day-to-day and plate-to-plate variability of the assay. The interleaved-signal plate layout format was used for dispensing both positive and negative control into the wells. This means that the positive and negative controls were dispensed in alternate columns. Z’-factor was the statistical parameter used in the determination of the variability of the assay. Z’-factor is a dimensionless and simple parameter which considers the difference in signal from samples and control as well as difference in the standard deviation of the controls. Z’-factor is used to assess the quality of the assay itself which makes use of the controls alone, therefore it was used in the validation of this assay30. 10 µL sample volume, mixture of CFP-PSB II (100 nM) and YFP-PAG (2 µM) was the negative control while the addition of 1 M GdnHCl to the mixture serves as the positive control. A DMSO tolerance test was carried out on the assay to ascertain the amount of DMSO that the assay can withstand with no compromise to the FRET signal. Eight different DMSO concentrations were tested (0, 1, 1.5, 2, 3, 5, 7.5, and 10%). Compound screening Peptidomimetics library of 15614 peptide-mimicking small molecule compounds (ChemDiv) was purchased from FIMM High Throughput Biomedicine Unit, University of Helsinki. Compounds were screened at 20 µM concentration with less than 1% DMSO. Compounds were dispensed with the ECHO acoustic liquid handler (Beckman Coulter, Inc.). After which the FRET pair was dispensed with a multichannel pipette to test for possible inhibition of the protein-peptide interaction. Furthermore, both positive and negative controls were measured alongside the compounds to ascertain the Z’-factor value during screening. The fluorescence intensities at 477 nm and 527 nm were recorded with an Infinite Pro M1000 microplate reader (Tecan) using 430 nm as excitation wavelength. The apparent inhibition was calculated as a percentage of PSB binding to the model peptide, where the negative control of the reaction was taken as 100% binding and the positive control as 0%. IC50 measurement For IC50 measurement, an increasing concentration of the compounds was added to the FRET pair. A log dilution series from 100 µM to 3.25 nM was considered for the compound concentration. The result was analyzed using GraphPad Prism 9.0. The FRET signal was plotted against the log of the inhibitor concentrations, and a nonlinear regression fit was used for the IC50 calculation. Differential scanning fluorimetry (DSF) DSF measurements were carried out using Prometheus Nt.48 NanoDSF instrument (NanoTemper Technologies GmbH). The melting temperatures (Tm) were determined for both PSB domains with and without the selected hit compound OUL-PSBi-001. For the DSF analysis, the purified PSBs (without CFP-tag)25,26 were diluted with either 20 mM Bis-Tris pH 6.8, 100 mM glycine (PSB-I) or 20 mM Tris pH 8, 50 mM NaCl, 50 mM glycine (PSB-II) to 10 µM. The analysis in the presence of the hit compound included 100 µM OUL-PSBi-001. The final DMSO concentration in all reactions was 1.3%. After preparation, samples were incubated at room temperature for 30 min for equilibration. Then the samples were loaded into standard capillary tubes, measurement was carried out from 20 °C to 90 °C, with increment of 1 °C per min. Five datasets were collected for each sample. An online tool MoltenProt45,46 was used to calculate the Tm from NanoDSF data. Microsoft Excel (Microsoft Corporation) was used to calculate the average values, and to draw the final figures. Bioluminescence based C-P4H activity assay The C-P4H catalytic activity was measured with the insect-cell expressed full-length C-P4H-I and -II8,33 and using a bioluminescent Succinate-Glo JmjC demethylase/hydroxylase assay kit (Promega Corboration), which quantifies the amount of succinate formed in the assay mixture from 2OG by C-P4H. Reaction mixtures (5 µL total volume) contain 5 µg/mL C-P4H, 100 µM peptide [(PPG)10], and a freshly prepared stock solution of 250 µM FeCl₂, 2.5 mM ascorbate, and 500 µM 2OG (final concentrations in the assay mixture: 50 µM FeCl₂, 500 µM ascorbate, and 100 µM 2OG) in 50 mM Tris buffer (pH 7.8). Enzyme and peptide were pre-incubated for 15 min at room temperature, and reactions were initiated by addition of the cofactor premix, followed by incubation for 30 min The reactions were then stopped by adding Reagent I (containing Succinate Buffer, Succinate-Glo Solution, and acetoacetyl-CoA), followed by a 1-hour incubation at 25 °C, allowing for the succinate dependent formation of ATP. Subsequently, reagent II, containing luciferase, was then added, and the amount of luminescence was measured after a 10-minute incubation time using a Spark microplate reader (Tecan). All measurements were performed in three independent experiments carried out on different days; error bars indicate the standard deviation between experiments. Docking of OUL-PSBi-001 Protein–ligand docking was performed using GOLD (The Cambridge Crystallographic Data Centre, CCDC)47 to investigate the mode of binding of OUL-PSBi-001 to the PSB domain. Docking was carried out using chain A of the PSB-I structure (PDB ID 9HT8) and chain A of the PSB-II structure (PDB ID 6EVN). To define the binding site, the peptide present in the structure (PPGPAGPPG) was extracted and used as a reference; the binding site was defined using a 10 Å radius around the peptide-defined pocket. Prior to docking, hydrogen atoms were added to the protein, and all non-protein atoms, including the peptide, other crystallographic ligands, and water molecules, were removed. The ligand was prepared using Marvin (Marvin 19.1, Chemaxon). Docking was run using GOLD default settings, and poses were scored using the ChemPLP scoring function. Docking poses were inspected and visualized in Hermes 2021.3.0 (The Cambridge Crystallographic Data Centre, CCDC). Acknowledgements The use of the facilities and expertise of the Biocenter Oulu Structural Biology core facility (a member of Biocenter Finland, Instruct-ERIC Centre Finland and FINStruct), Proteomics and Protein Analysis core facility (a member of Biocenter Finland) and Biocenter Oulu sequencing center are gratefully acknowledged. The assistance of Dr Albert Galera-Prat during the recombinant protein expression is gratefully acknowledged, and Professor Johanna Myllyharju and Dr Antti Salo for stimulating discussions. Author contributions L.L. M.K.K. conceptualization; B.A., C.V.-R., S.M., M.M.R. investigation; B.A., C.V.-R. writing-original draft; All authors, writing, reviewing, and editing; L.L. and R.W. funding acquisition; C.V.R., R.W., L.L., M.K.K. supervision. Funding We acknowledge funding from Biocenter Oulu spearhead project (LL) and Jane and Aatos Erkko Foundation (RW). Data availability All data generated or analyzed during this study are included in this published article. The raw data generated and analyzed during the current study are available from the corresponding authors on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Bukunmi Adediran, Carlos Vela-Rodríguez and Sudarshan Murthy. Contributor Information Lari Lehtiö, Email: lari.lehtio@oulu.fi. M. Kristian Koski, Email: kristian.koski@oulu.fi. References Associated Data Data Availability Statement All data generated or analyzed during this study are included in this published article. The raw data generated and analyzed during the current study are available from the corresponding authors on reasonable request.