Search

Successful medicinal chemistry campaigns to discover and optimize sphingosine kinase inhibitors require a robust assay for screening chemical libraries and for determining rank order potencies. Existing assays for these enzymes are laborious, expensive and/or low throughput. The toxicity of excessive levels of phosphorylated sphingoid bases for the budding yeast, Saccharomyces cerevisiae, affords an assay wherein inhibitors added to the culture media rescue growth in a dose-dependent fashion. Herein, we describe our adaptation of a simple, inexpensive, and high throughput assay for assessing inhibitors of sphingosine kinase types 1 and 2 as well as ceramide kinase and for testing enzymatic activity of sphingosine kinase type 2 mutants. The assay was validated using recombinant enzymes and generally agrees with the rank order of potencies of existing inhibitors. NIH/NIGMS [R01 GM104366, R01 GM121075]; NIH [R01 GM075240]; NIGMS; University of Virginia Summer Research Internship Program fellowships This work was supported by NIH/NIGMS R01 GM104366 & NIH/NIGMS R01 GM121075 to KRL & WLS, NIH R01 GM075240 to JSS, NIGMS: https://www.nigms.nih.gov/Pages/default.aspx; University of Virginia Summer Research Internship Program fellowships to JG and PBB https://med.virginia.edu/diversity/programs/srip/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Simultaneous replacement of Asp-94 with serine and Gly-98 with glutamate in rat alpha-parvalbumin creates a CD-site ligand array in the context of the EF-site binding loop. Previous work has shown that, relative to the wild-type CD site, this engineered site has markedly reduced Ca(2+) affinity. Seeking an explanation for this phenomenon, we have obtained the crystal structure of the alpha D94S/G98E variant. The Ca(2+) coordination within the engineered EF site of the 94/98E variant is nearly identical to that within the CD site, suggesting that the attenuated affinity of the EF site in 94/98E is not a consequence of suboptimal coordination geometry. We have also examined the divalent ion binding properties of the alpha 94/98E variant in both Na(+)- and K(+)-containing buffers. Although the Ca(2+) and Mg(2+) affinities are higher in K(+) solution, the increases are comparable to those observed for wild-type alpha. Consistent with that finding, the apparent Na(+) stoichiometry, estimated from stability studies conducted as a function of Na(+) concentration, is 1.0 +/- 0.1, identical to that of wild-type alpha. Thus, the reduced affinity for divalent ions is evidently not the result of heightened monovalent ion competition. The thermodynamic analysis indicates that the less favorable Gibbs free energy of binding reflects a substantial enthalpic penalty. Significantly, the crystal structure reveals a steric clash between Phe-57 and the C(gamma) atom of Glu-98. The consequent displacement of Phe-57 also produces a close contact with Ser-55. Thus, steric interference may be the source of the enthalpic penalty.

Introduction of a fifth carboxylate into the ligand array of the CD site (via the combined S55D and E59D mutations) or the EF site (G98D) of rat alpha-parvalbumin substantially increases divalent ion affinity. This behavior, in conflict with that seen in model peptide systems, agrees with existing data for rat beta-parvalbumin [Henzl et al. (1996) Biochemistry 35, 5856-5869]. The complete analysis of the S55D/E59D double variant necessitated characterization of alpha E59D. Whereas the D59E mutation has minimal influence on beta CD site affinity, E59D has a major impact on the alpha CD site, lowering the apparent association constant by a factor of 14. The thermodynamic consequences of exchanging the rat alpha CD and EF site ligand arrays, which differ at the +z and -x coordination positions, were also examined. When the alpha CD array is imported into the EF site, it acquires a low-affinity phenotype, in agreement with previous findings for beta [Henzl et al. (1998) Biochemistry 37, 9101-9111]. However, when the EF ligand array is introduced into the alpha CD binding loop, it retains a high-affinity signature. This result, contrary to that observed in beta, suggests that the influence of the parvalbumin CD site environment supersedes the intrinsic behavior of the ligand array, a conclusion further supported by the disparate impact of the beta D59E and alpha E59D mutations.

Rat alpha- and beta-parvalbumins have distinct monovalent cation-binding properties [Henzl et al. (2000) Biochemistry 39, 5859-5867]. Beta binds two Na(+) or one K(+), and alpha binds one Na(+) and no K(+). Ca(2+) abolishes these binding events, suggesting that the monovalent ions occupy the EF-hand motifs. This study compares alpha and beta divalent ion affinities in Na(+) and K(+) solutions. Solvent cation identity seriously affects alpha. In Hepes-buffered NaCl, at 5 degrees C, the macroscopic Ca(2+)-binding constants are 2.6 x 10(8) and 6.4 x 10(7) M(-1) and the Mg(2+) constants, 1.8 x 10(4) and 4.3 x 10(3) M(-1). In Hepes-buffered KCl, the Ca(2+) values increase to 2.9 x 10(9) and 6.6 x 10(8) M(-1) and the Mg(2+) values to 2.2 x 10(5) and 3.7 x 10(4) M(-1). Monte Carlo simulation of alpha binding data-employing site-specific constants and explicitly considering Na(+) binding-yields a K(Na) of 630 M(-1) and indicates that divalent ion-binding is positively cooperative. NMR data suggest that the lone Na(+) ion occupies the CD loop. Solvent cation identity has a smaller impact on beta. In Na(+), the Ca(2+) constants for the EF and CD sites are 2.3 x 10(7) and 1.5 x 10(6) M(-1), respectively; the Mg(2+) constants are 9.2 x 10(3) and 1.7 x 10(2) M(-1). In K(+), these values shift to 3.1 x 10(7) and 3.8 x 10(6) M(-1) and the latter to 1.4 x 10(4) and 2.9 x 10(2) M(-1). These data suggest that parvalbumin divalent ion affinity, particularly that of rat alpha, can be significantly attenuated by increased intracellular Na(+) levels.

Despite its higher net charge and reduced opportunities for favorable tertiary interactions, Ca(2+)-free rat beta-parvalbumin is more stable than rat alpha-parvalbumin. Under conditions wherein alpha denatures at 45.8 degrees C, beta denatures at 53.6 degrees. The homologous chicken beta isoform known as CPV3 also exhibits heightened stability-prompting an inquiry into the stabilizing influence of Pro-21 and Pro-26. Individual P21A and P26A mutations lower the T(m) of rat beta by 3.2 degrees, decreasing conformational stability by 0.74 kcal/mol. Simultaneous replacement of Pro-21 and Pro-26 essentially abolishes the excess stability (DeltaT(m) = -7.6 degrees; DeltaDeltaG(conf) = -1.77 kcal/mol). Significantly, the P21A/P26A variant displays Ca(2+) affinity virtually indistinguishable from wild-type beta, implying that structural alterations in the AB domain do not necessarily influence the divalent ion affinity of the CD-EF domain. The consequences of introducing proline at positions 21 and 26 in rat alpha were also examined. Whereas the H26P mutation raises the T(m) by 5.6 degrees (DeltaDeltaG(conf) = 1.25 kcal/mol), A21P lowers the T(m) by 8.5 degrees (DeltaDeltaG(conf) = -1.9 kcal/mol). Replacement of Ala-21 by proline in an alpha AB/beta CD-EF chimera increases the T(m) by 5.8 degrees (DeltaDeltaG(conf) = 0.95 kcal/mol), implying that the destabilization of alpha by Pro-21 results from steric conflict with a residue in the CD-EF domain. Consistent with that hypothesis, the K80S mutation markedly stabilizes alpha A21P, yielding a protein with a T(m) 2.0 degrees higher than wild-type alpha. The observed differences in stability resulting from proline addition/removal are largely consistent with alterations in main-chain and side-chain conformational entropy.

Birds express three parvalbumins, one a isoform and two p isoforms. The latter are known as avian thymic hormone (ATH) and avian parvalbumin 3. Although both were discovered in thymus tissue, and presumably function in T-cell maturation, they have been detected in other tissue settings. We have conducted detailed Ca 2 + - and Mg2 + -binding studies on recombinant ATH and the C72S variant of CPV3, employing global analysis of isothermal titration calorimetry data. In Hepesbuffered saline, ATH binds Ca 2 + with apparent microscopic binding constants of 2.4 ′ 0.2 x 10 8 and 1.0 ′ 0.1 x 10 8 M - 1 . The corresponding values for CPV3-C72S are substantially lower, 4.5 ′ 0.5 x 10 7 and 2.4 ′ 0.2 x 10 7 M - 1 , a 1.9-kcal/mol difference in binding free energy. Thus, the β-parvalbumin lineage displays a spectrum of Ca 2 + -binding affinity, with ATH and the mammalian β isoform at the high-and low-affinity extremes and CPV3 in the middle. Interestingly, despite its decreased Ca 2 + affinity, CPV3-C72S exhibits increased affinity for Mg 2 + , relative to ATH. Whereas the latter displays Mg 2 + -binding constants of 2.2 ′ 0.2 x 10 4 and 1.2 ′ 0.1 x 10 4 M - 1 , CPV3-C72S yields values of 5.0 ′ 0.8 x 10 4 and 2.1 ′ 0.3 x 10 4 M - 1 .

We have examined the metal ion-binding domains from rat alpha and beta parvalbumin. We find that the CD-EF fragments differ markedly in their tendency to self-associate. Whereas Ca(2+)-free alpha CD-EF is monomeric, the Ca(2+)-free beta peptide dimerizes weakly (K(2) = 2400 +/- 200 M(-1)). In buffer containing 1.0 mM Ca(2+), the apparent dimerization constant for beta CD-EF (191,000 +/- 29,000 M(-1)) is more than 50 times that of alpha (3400 +/- 200 M(-1)). Alpha CD-EF binds two Ca(2+) with positive cooperativity. Titration calorimetry data afford binding constants of 3.7(0.1) x 10(3) M(-1) and 8.6(0.2) x 10(4) M(-1). Beta CD-EF also binds two Ca(2+) cooperatively but with lower affinity. Equilibrium dialysis yields Adair constants of 4.2(0.1) x 10(3) and 6.1(0.2) x 10(3) M(-1). Significantly, the difference in Ca(2+) affinity is substantially smaller than that observed for the full-length proteins-suggesting that the AB domain can modulate divalent ion affinity. Analysis of beta calorimetry data requires explicit consideration of the self-association behavior. Data collected at low CD-EF concentration are consistent with preferential occupation of the EF site, dimerization of singly bound monomers, and cooperative filling of the CD sites. At higher concentrations, apo-protein dimerization can apparently precede cooperative occupation of the EF sites. In the presence of Ca(2+), alpha CD-EF exhibits higher thermal stability, consistent with its higher Ca(2+) affinity. However, the beta melting temperature shows greater concentration dependence, consistent with its greater tendency to dimerize. Neither fragment exhibits a sigmoidal melting curve in the Ca(2+)-free state, suggesting that the apo-peptides are disordered.

Ca2+ is an obligatory factor for both the extrinsic and intrinsic pathways of coagulation. In majority of in vitro studies, investigators use saturating concentrations of Ca2+ (5 to 10 mM) for FVIIa/tissue factor (TF) activation of factor IX (FIX), and factor X (FX) (extrinsic coagulation), as well as for the activation of FIX by FXIa, FX by FIXa/FVIIIa, and prothrombin by FXa/FVa (intrinsic coagulation). However, the concentration of Ca2+ in plasma is only 1.1 mM, which is considerably below the saturating concentration needed for optimal coagulation. Importantly, plasma also contains 0.6 mM Mg2+ that could compensate for subsaturating concentrations of Ca2+ in promoting coagulation. Previous studies have attempted to clarify this concept in FIX, FX and prothrombin activation. However, these studies are sparse and in virtually all cases not detailed. We have systematically examined the role of plasma concentration of Mg2+ (in addition to the plasma concentration of Ca2+) in promoting all Ca2+ dependent steps of extrinsic and intrinsic coagulation and compared it with the saturating concentration of Ca2+. The Km (~ 100 nM) for activation of FIX by FXIa was similar in the presence of plasma concentrations of Ca2+/Mg2+ or 5 mM Ca2+. Furthermore, the Km and Vmax for the activation of FX and FIX by FVIIa/TF were essentially similar for both conditions. The Km and Vmax for the activation of FX by FVIIIa/FIXa ± phospholipid, and prothrombin by FXa/FVa ± phospholipid were also indistinguishable in these two different metal ion conditions. Notably, when only plasma concentrations of Ca2+ (1.1mM or 1.7mM) were used in all reactions mentioned above, coagulation proceeded at suboptimal rates. In further studies, we used Biacore to investigate the binding of FXIa and FVIII to FIXa, soluble TF to FVIIa, and FVa to FXa. Soluble TF, dansyl-Glu-Gly-Arg (dEGR)-IXa and dEGR-Xa were coupled to CM5 chips in the presence of 10 mM Ca2+. At 5 mM Ca2+, the binding of FXIa to dEGR-IXa was characterized by a Kd of ~40 nM, binding of FVIII to dEGR-IXa by a Kd of ~100 nM, and FVa to dEGR-Xa by a Kd of ~120 nM. In the presence of plasma concentrations of Ca2+ and Mg2+, binding constants were similar to those obtained in the presence of 5 mM saturating Ca2+ concentration. Additional 45Ca2+ binding studies using equilibrium dialysis and prothrombin fragment 1, dEGR-VIIa and decarboxylated dEGR-VIIa, FIX and decarboxylated FIX, and FX and decarboxylated FX, indicated that in the g-carboxyglutamic acid (Gla) domain, 2-3 Ca2+ binding sites (Shikimoto, et al., J. Biol. Chem. 278, 24090-24094, 2003; Wang, et al., Biochemistry42, 7959-7966, 2003; Bajaj, et al., J. Biol. Chem.281, 24873-24888, 2006) out of seven core divalent ion binding sites (Soriano-Garcia et al., Biochemistry31, 2554-2566, 1992) could be replaced by Mg2+. Conversely, Mg2+ could not displace the Ca2+ binding sites in the epidermal growth factor-like domain 1 (EGF1) and protease domains of FIX or FX. Overall these studies indicate that (1) saturating concentrations of Ca2+ used in in vitro investigations are valid representations of coagulation studies, except for that Mg2+ compensates for suboptimal concentrations of Ca2+ under physiological conditions; (2) two of the Ca2+-binding sites in the Gla domain (numbers 1 and 7, per Tulinsky numbering (Soriano-Garcia et al., Biochemistry31, 2554-2566, 1992)), and possibly a third site (number 4) are specific for Mg2+ under physiologic conditions; and (3) the Ca2+-binding sites in the EGF1 and protease domains are specific for Ca2+ and can not be occupied by Mg2+ under physiologic conditions. In conclusion, Ca2+ and Mg2+ act in concert to promote optimal coagulation under physiologic conditions. Mg2+ alone does not promote coagulation since it cannot bind to the Ca2+ specific sites in the Gla domain necessary for folding of the Gla domain omega loop.

Factor VII (FVII) consists of an N-terminal γ-carboxyglutamic acid (Gla) domain followed by two epidermal growth factor-like (EGF1 and EGF2) domains and the C-terminal protease domain. Activation of FVII results in a two-chain FVIIa molecule consisting of a light chain (Gla-EGF1-EGF2 domains) and a heavy chain (protease domain) held together by a single disulfide bond. The complex of tissue factor (TF) and FVIIa activates FIX and FX during coagulation. FVIIa on its own is structurally more “zymogen-like” and when bound to TF it is more “active enzyme-like.” We obtained crystal structures of EGR-VIIa/soluble (s) TF (2.9 Å resolution), dansyl-EGR-VIIa/sTF (1.9 Å resolution) and benzamidine-VIIa/sTF (1.6 Å resolution). We also investigated the effect of TF binding on the S1, S2, and S3/S4 subsites (Schechter and Berger, BBRC, 27:157-162, 1967) in FVIIa. The affinity of variously inhibited FVIIa to sTF was also measured using Biacore technology. For obtaining second order inhibition rate constants, FVIIa ± soluble (s) TF was incubated with each inhibitor for various times, diluted several fold and assayed for the residual FVIIa activity. The second order rate constants were obtained by plotting the first order rate constants versus the inhibitor concentrations. These data are summarized in the table below. From these data it appears that all subsites are affected upon FVIIa binding to sTF. Since in the crystal structure of EGR-VIIa/sTF the P1 Arg residue is the only residue that makes contact with FVIIa, it follows that the S1 site is affected ~10-fold upon binding to sTF. Adding a dansyl group that partially occupies the S3/S4 position (1.9 Å structure) increases the second order rate constant 7-fold (2.41 versus 0.35) over that of EGR-ck. Moreover the addition of Pro (DFPR-ck) or Phe (DFFR-ck) residue occupying the S2 position increases the second order rate constant 357-fold and 1500- fold, respectively (125 and 525 versus 0.35). Thus, comparison of dEGR, DFPR, DFFR inhibition suggests that FVIIa prefers Phe at S2 and at S3/S4 positions, and that TF opens up the S1/S2/S3/S4 sites for substrate or inhibitor occupancy. These data are consistent with LTR (P3/P2/P1) residues in FX at its activation cleavage site as well as with LTR (P3/P2/P1) residues and FTR (P3/P2/P1) residues at the 145-146 and 180-181 FIX activation cleavage sites, respectively. Thus, these studies with chloromethylketone inhibitors have biologic relevance. For Biacore studies, sTF was amine coupled to a CM5 chip. The binding of unoccupied active site FVIIa in 5 mM calcium to sTF was characterized by a KD of 7 nM. Benzamidine (10 mM)-VIIa, p-aminobenzamidine (pAB, 1 mM)-VIIa, EGR-VIIa, dEGR-VIIa, DFPR-VIIa and DFFR-VIIa each bound to sTF with KD values ranging from 1- 2 nM. These affinity measurements indicate that the S1 site occupied FVIIa molecule (benzamidine-FVIIa, pAB-VIIa) has essentially the same conformation as the S1/S2/S3/S4 occupied FVIIa. This conclusion is consistent with similar crystal structures of variously inhibited FVIIa molecules complexed with sTF. The differential rates of incorporation of various chloromethylketone inhibitors could be due to the interaction of various residues (P1, P2, P3, P4) with the corresponding active subsites (S1/S2/S3/S4) of FVIIa. Additionally, the rate of incorporation of chloromethylketone inhibitors into FVIIa also involves the irreversible alkylation step, which could be faster for DFFR-ck and DFPR-ck. Once these inhibitors are incorporated, it appears that they induce the same conformation in FVIIa as achieved by S1 site occupancy alone. Thus S1 site occupancy in FVIIa induces the required conformation to modestly increase the affinity for TF. Second Order Rate Constants for Inhibition of FVIIa ± sTF with Various Chloromethylketone (ck) Inhibitors Inhibitor Minus sTF k (min−1 mM−1) Plus sTF k (min−1 mM−1) Fold Difference EGR-ck 0.04 0.35 8.8 dansyl EGR-ck 0.07 2.41 34.4 (D)FPR-ck 2.3 125 54.3 D)FFR-ck 5.6 525 93.8

In blood coagulation serine proteases, amino acid 192 (chymotrypsin numbering) influences substrate and inhibitor specificity, and also forms a peptide bond with Gly 193 , a critical part of the protease oxyanion hole. Residue 192 is either Glu or Gln in several serine proteases including thrombin, factor (F) Xa, FIXa, and activated protein C. However, it is Lys in factor FXIa and FVIIa. Both FXIa and FVIIa/tissue factor (TF) share FIX as their substrate. In addition, FVIIa/TF activates FX and is inhibited by TF pathway inhibitor (TFPI), whereas FXIa is inhibited by diisopropylfluorophosphate (DFP), p -aminobenzamidine ( p- AB), antithrombin and amyloid-b precursor protein Kunitz domain (AbPP). We investigated the importance of Lys 192 in FXIa and FVIIa for substrate/inhibitor specificity and oxyanion hole formation. Recombinant FXIa was prepared with Pro (FXIa K192P ), Glu (FXIa K192E ) or Gln (FXIa K192Q ) substitutions for Lys 192 . Rate constants for inhibition by DFP were similar for wild type FXIa (FXIa WT ), FXIa K192E and FXIa K192Q , but 80-fold slower for FXIa K192P . Only FXIa K192P differed in binding to p AB compared to FXIa WT . These data indicate that Pro 192 causes distortion of the oxyanion hole and S1 binding site. All mutants bound the FXIa substrate FIX normally. However, catalysis of FIX was impaired 400, 35, and 110-fold for FXIa K192P , FXIa K192E , and FXIa K192Q , respectively, while k 2 for inhibition by antithrombin was reduced 160, 15 and 16-fold. FXIa K192P bound to AbPP with 800-fold decreased affinity; while FXIa K192E and FXIa K192Q bound with 6-10-fold higher affinity. Modeling studies indicate that there is loss of the normal interaction between Lys 192 and the substrate/inhibitor P3’ residue in interactions between FXIa K192E and FXIa K192Q and factor IX and antithrombin. The K i for Glu and Gln mutant inhibition by AbPP is slightly better probably due to the ability of these two amino acids to make two H-bonds as compared to only one for Lys 192 . Lys 192 makes H-bond with the carbonyl O of Gly 12 in AbPP, whereas Glu and Gln at this position make H-bonds with the carbonyl O of Cys 14 and Ala 16 in AbPP. In FXIa K192P , the oxyanion hole is impaired due to a flip in the 192-193 peptide bond causing the amide N of Gly 193 to point away from the oxyanion hole, resulting in a severe impairment of all catalytic functions. Thus a non-proline residue 192 is essential for proper oxyanion hole formation in serine proteases, and Lys 192 in FXIa contributes significantly towards macromolecular substrate catalysis and inhibitor binding. In contrast to FXIa WT , the 192-193 peptide bond in FVIIa WT ± TF is already in a nonstandard conformation in which the amide N of Gly 193 points away, and the carbonyl O of the Lys 192 points into the oxyanion hole (Bajaj, et al., J. Biol. Chem.281, 24873-24888, 2006). The H-bond between the amide N of Gly 193 and the carbonyl side chain of Gln 143 , a residue unique to FVIIa, maintains this nonstandard conformation in FVIIa. Notably, upon substrate binding the 192-193 bond flips 180° such that the oxyanion hole is fully formed. Now the H-bond is formed between the amide N of Gly 193 and the carbonyl oxygen of the transition state intermediate. For this to occur, the H-bond between the amide N of Gly 193 and the carbonyl side chain of Gln 143 must first be broken. Previous studies have shown that changing Lys 192 to Glu or Gln in FVIIa severely impairs its biologic activity. However, the structural basis for this impairment is not understood. We performed molecular modeling studies which reveal that the side chain of the Glu 192 carboxyl group or of the Gln 192 carbonyl group makes strong H-bond with the side chain NH 2 group of Gln 143 ; this is not possible for Lys 192 . Thus in Glu 192 or Gln 192 mutant of FVIIa, it is energetically difficult to break the H-bond between the amide N of Gly 193 and the carbonyl side chain of Gln 143 because the H-bond involving the side chain of Glu 192 or Gln 192 and the Gln 143 must be simultaneously broken. This provides a structural rational for the impaired function of FVIIa K192E or FVIIa K192Q . To test this concept further, we prepared FVIIa in which Lys 192 was replaced with alanine. All properties of this alanine mutant were normal substantiating the proposed structural role of Lys 192 in FVIIa. This structural role is unique to FVIIa among all known serine proteases, as is the presence of residue Gln 143 that contributes to this phenomenon.