PMF-GRN: a variational inference approach to single-cell gene regulatory network inference using probabilistic matrix factorization

Inferring gene regulatory networks (GRNs) from single-cell data is challenging due to heuristic limitations. Existing methods also lack estimates of uncertainty. Here we present Probabilistic Matrix Factorization for Gene Regulatory Network Inference (PMF-GRN). Using single-cell expression data, PMF-GRN infers latent factors capturing transcription factor activity and regulatory relationships. Using variational inference allows hyperparameter search for principled model selection and direct comparison to other generative models. We extensively test and benchmark our method using real single-cell datasets and synthetic data. We show that PMF-GRN infers GRNs more accurately than current state-of-the-art single-cell GRN inference methods, offering well-calibrated uncertainty estimates.

Coordinated control of neuronal differentiation and wiring by sustained transcription factors

The large diversity of cell types in nervous systems presents a challenge in identifying the genetic mechanisms that encode it. Here, we report that nearly 200 distinct neurons in the Drosophila visual system can each be defined by unique combinations of on average 10 continuously expressed transcription factors. We show that targeted modifications of this terminal selector code induce predictable conversions of neuronal fates that appear morphologically and transcriptionally complete. Cis-regulatory analysis of open chromatin links one of these genes to an upstream patterning factor that specifies neuronal fates in stem cells. Experimentally validated network models describe the synergistic regulation of downstream effectors by terminal selectors and ecdysone signaling during brain wiring. Our results provide a generalizable framework of how specific fates are implemented in postmitotic neurons.

High performance single-cell gene regulatory network inference at scale: The Inferelator 3.0

Motivation

Gene regulatory networks define regulatory relationships between transcription factors and target genes within a biological system, and reconstructing them is essential for understanding cellular growth and function. Methods for inferring and reconstructing networks from genomics data have evolved rapidly over the last decade in response to advances in sequencing technology and machine learning. The scale of data collection has increased dramatically; the largest genome-wide gene expression datasets have grown from thousands of measurements to millions of single cells, and new technologies are on the horizon to increase to tens of millions of cells and above.

Results

In this work, we present the Inferelator 3.0, which has been significantly updated to integrate data from distinct cell types to learn context-specific regulatory networks and aggregate them into a shared regulatory network, while retaining the functionality of the previous versions. The Inferelator is able to integrate the largest single-cell datasets and learn cell-type-specific gene regulatory networks. Compared to other network inference methods, the Inferelator learns new and informative Saccharomyces cerevisiae networks from single-cell gene expression data, measured by recovery of a known gold standard. We demonstrate its scaling capabilities by learning networks for multiple distinct neuronal and glial cell types in the developing Mus musculus brain at E18 from a large (1.3 million) single-cell gene expression dataset with paired single-cell chromatin accessibility data.

Availability and implementation

The inferelator software is available on GitHub (https://github.com/flatironinstitute/inferelator) under the MIT license and has been released as python packages with associated documentation (https://inferelator.readthedocs.io/).

Supplementary information

Supplementary data are available at Bioinformatics online.

Activity of adefovir dipivoxil against all patterns of lamivudine-resistant hepatitis B viruses in patients

One hundred and thirty-one post-liver transplantation patients with chronic hepatitis B and failing lamivudine therapy with detectable serum hepatitis B virus (HBV) deoxyribonucleic acid by hybridization assays or > or =1 x 10(6) copies/mL by polymerase chain reaction, and elevated alanine transaminase levels despite continuous lamivudine, were enrolled in an open-label study of adefovir dipivoxil. The B and C domains of HBV polymerase were sequenced for baseline samples to determine the presence of lamivudine resistance mutations. The results showed that 98% of the samples had tyrosine-methionine-aspartate-aspartate (YMDD) mutations, indicating a strong correlation between the above clinical definition of lamivudine treatment failure and the presence of YMDD mutations. In addition to the rtM204V/I and the rtL180M mutations, the mutation rtV173L was identified in 19% of patients. Four major patterns of lamivudine-resistant HBV were identified: rtL180M + rtM204V (60%), rtV173L + rtL180M + rtM204V (19%), rtM204I (9%) and rtL180M + rtM204I (9%). Treatment with adefovir dipivoxil showed similar antiviral efficacy in patients with lamivudine-resistant virus from all four patterns.

Comparative effects of adefovir and selected nucleoside inhibitors of hepatitis B virus DNA polymerase on mitochondrial DNA in liver and skeletal muscle cells

Adefovir is a potent nucleotide analog inhibitor of hepatitis B virus (HBV) DNA polymerase. Its oral prodrug adefovir dipivoxil has been approved for the treatment of chronic HBV infection. In this study, adefovir was characterized for its in vitro effects on mitochondrial DNA (mtDNA) synthesis and compared with the nucleoside analogues lamivudine (3TC), fialuridine (FIAU), and zalcitabine (ddC). No substantial changes in mtDNA content were detected in human hepatoblastoma HepG2 cells and normal human skeletal muscle cells following a 9-day treatment with 0.3-30 microm adefovir, concentrations up to 500-fold higher than the peak serum levels in patients treated with adefovir dipivoxil. Similarly, mtDNA was unchanged in both cell types following treatment with 3TC. In contrast, 30-55% and > 90% reductions in mtDNA were observed following incubation with 30 microm FIAU and ddC, respectively. The effects of FIAU on mtDNA became more pronounced following prolonged 18-day treatment of skeletal muscle cells while the effects of other drugs remained unchanged.

Identification of critical residues on thrombin mediating its interaction with fibrin

Thrombin binding to fibrin may be important in localizing thrombin to the site of vascular injury. However, fibrin-bound thrombin retains its catalytic activity toward fibrinogen, and may be prothrombotic under certain conditions. A collection of 52 purified thrombin mutants was used to identify those residues mediating the thrombin-fibrin interaction. Comparison of fibrinogen clotting activity with fibrin binding activity identified twenty residues involved in fibrinogen recognition with four of these residues important in fibrin binding (Lys65, His66, Tyr71, Arg73). No mutant was identified with normal clotting activity and deficient fibrin binding, suggesting that these two properties are not readily dissociable. A DNA thrombin aptamer that binds to these residues was able to inhibit the thrombin-fibrin interaction, and displace thrombin that was already bound. Mapping of these fibrin-binding residues on thrombin revealed that they are localized within exosite I, and comprise a subset of the residues important in fibrinogen recognition.

Safety and efficacy of adefovir dipivoxil in patients co-infected with HIV-1 and lamivudine-resistant hepatitis B virus: an open-label pilot study

BACKGROUND

Lamivudine-resistant hepatitis B virus (HBV) is found in about 15-32% of infected patients with or without co-infection with HIV-1 after 1 year of lamivudine therapy. Adefovir dipivoxil is active in vivo and in vitro against wild-type and lamivudine-resistant HBV. We assessed the safety and efficacy of a once daily dose of adefovir dipivoxil in an open-label trial for the treatment of lamivudine-resistant HBV infection in HIV-1-infected patients.

METHODS

35 HIV-1/HBV co-infected patients receiving lamivudine therapy (150 mg twice daily) as part of their current HIV-1 antiretroviral regimen were enrolled. Patients received a 10 mg once-daily dose of adefovir dipivoxil for48 weeks while maintaining their existing anti-HIV-1 therapy, including lamivudine. Patients were assessed every 4 weeks for safety and efficacy.

FINDINGS

Four patients withdrew from the study (two because of adverse events), leaving 31 patients who received adefovir dipivoxil for a median of 48 weeks (range 44-48). Mean decreases in serum HBV DNA concentrations from baseline (log 8.64 copies/mL [SE log 0.08]) were 2log 3.40 copies/mL [log 0.12] at week 24 (n=31) and 2log 4.01 copies/mL [log 0.17] at week 48 (n=29; p<0.0001). Two patients underwent hepatitis B e antigen seroconversion-one at week 32 and one at week 36. Adefovir dipivoxil was generally well tolerated, but was associated with a transient increase in serum alanine aminotransferase concentrations in 15 patients. We found no significant changes in either HIV-1 RNA or CD4 cell count.

INTERPRETATION

These results indicate that 48 weeks of 10 mg daily adefovir dipivoxil is well tolerated and active against lamivudine-resistant HBV in HIV-1/HBV co-infected patients.

Molecular modeling and biochemical characterization reveal the mechanism of hepatitis B virus polymerase resistance to lamivudine (3TC) and emtricitabine (FTC)

Success in treating hepatitis B virus (HBV) infection with nucleoside analog drugs like lamivudine is limited by the emergence of drug-resistant viral strains upon prolonged therapy. The predominant lamivudine resistance mutations in HBV-infected patients are Met552IIe and Met552Val (Met552Ile/Val), frequently in association with a second mutation, Leu528Met. The effects of Leu528Met, Met552Ile, and Met552Val mutations on the binding of HBV polymerase inhibitors and the natural substrate dCTP were evaluated using an in vitro HBV polymerase assay. Susceptibility to lamivudine triphosphate (3TCTP), emtricitabine triphosphate (FTCTP), adefovir diphosphate, penciclovir triphosphate, and lobucavir triphosphate was assessed by determination of inhibition constants (K(i)). Recognition of the natural substrate, dCTP, was assessed by determination of Km values. The results from the in vitro studies were as follows: (i) dCTP substrate binding was largely unaffected by the mutations, with Km changing moderately, only in a range of 0.6 to 2.6-fold; (ii) K(i)s for 3TCTP and FTCTP against Met552Ile/Val mutant HBV polymerases were increased 8- to 30-fold; and (iii) the Leu528Met mutation had a modest effect on direct binding of these beta-L-oxathiolane ring-containing nucleotide analogs. A three-dimensional homology model of the catalytic core of HBV polymerase was constructed via extrapolation from retroviral reverse transcriptase structures. Molecular modeling studies using the HBV polymerase homology model suggested that steric hindrance between the mutant amino acid side chain and lamivudine or emtricitabine could account for the resistance phenotype. Specifically, steric conflict between the Cgamma2-methyl group of Ile or Val at position 552 in HBV polymerase and the sulfur atom in the oxathiolane ring (common to both beta-L-nucleoside analogs lamivudine and emtricitabine) is proposed to account for the resistance observed upon Met552Ile/Val mutation. The effects of the Leu528Met mutation, which also occurs near the HBV polymerase active site, appeared to be less direct, potentially involving rearrangement of the deoxynucleoside triphosphate-binding pocket residues. These modeling results suggest that nucleotide analogs that are beta-D-enantiomers, that have the sulfur replaced by a smaller atom, or that have modified or acyclic ring systems may retain activity against lamivudine-resistant mutants, consistent with the observed susceptibility of these mutants to adefovir, lobucavir, and penciclovir in vitro and adefovir in vivo.

Combination therapy with lamivudine and famciclovir for chronic hepatitis B-infected Chinese patients: a viral dynamics study

In vitro studies have shown that lamivudine and penciclovir (the active metabolite of famciclovir) act synergistically to inhibit hepatitis B virus (HBV) replication. We compared the effectiveness of HBV viral suppression by lamivudine monotherapy versus lamivudine plus famciclovir combination therapy in Chinese patients with chronic HBV infection. Twenty-one Chinese hepatitis B e antigen (HBeAg)-positive patients, with detectable HBV DNA (Digene Hybrid Capture II), were randomized to receive either lamivudine 150 mg/d orally (group 1, 9 patients) or lamivudine 150 mg/d plus famciclovir 500 mg 3 times a day orally (group 2, 12 patients) for 12 weeks, with a follow-up period of at least 16 weeks. Serial serum HBV-DNA levels were determined and a mathematical model with provision for incomplete inhibition of virus production during therapy was applied to analyze the dynamics of viral clearance. The mean antiviral efficacy was significantly greater in group 2 than in group 1 (0.988 +/- 0.012 vs. 0.94 +/- 0.03, P =.0012). HBV DNA returned to pretreatment level within 16 weeks after the end of initial treatment in 4 patients (66.7%) in group 1 and none in group 2 (P =.08), who remained HBeAg positive and received no further treatment after week 12. Hence, in Chinese chronic HBeAg-positive patients, combination therapy using lamivudine and famciclovir was superior to lamivudine monotherapy in inhibiting HBV replication. Further studies of longer duration are needed to define whether combination therapy will increase the HBeAg seroconversion rate and decrease the rate of emergence of lamivudine-resistant variants.

In vitro evaluation of hepatitis B virus polymerase mutations associated with famciclovir resistance

Several mutations (V521L, P525L, L528M, T532S, and V555I) in the gene for hepatitis B virus (HBV) polymerase have been identified in HBV isolated from patients that displayed break-through viremia during famciclovir treatment. To determine whether these mutations cause phenotypic resistance to famciclovir, we compared the inhibition constants (K(i)) of penciclovir triphosphate (PCVTP, the active metabolite of famciclovir) for recombinant wild-type and mutant HBV polymerases containing these mutations. In in vitro enzymatic assays, the V555I mutation displayed the most resistance (with K(i) increased by 6.2-fold) to PCVTP. The V521L and L528M mutations showed moderately decreased sensitivity to PCVTP (K(i) increased by >3-fold). We also analyzed the cross-resistance profiles of these variants for adefovir and lamivudine, two other antiviral agents that also inhibit DNA replication by HBV polymerase. All 5 famciclovir-associated mutations were sensitive to adefovir diphosphate (ADVDP) in in vitro enzymatic assays (<2.3-fold decreased sensitivity). The V521L, L528M, and T532S mutations were also sensitive to lamivudine triphosphate (LAMTP); however, the P525L and V555I mutations displayed moderately decreased sensitivity to LAMTP in enzymatic assays (3.6-fold decreased sensitivity). The lamivudine-resistant mutations M552I, M552V, and L528M+M552V, which were previously shown to display 8- to 25-fold resistance to LAMTP, were less resistant (< or = 3.1-fold) to PCVTP.

Fulminant hepatic failure resulting from lamivudine-resistant hepatitis B virus in a renal transplant recipient: durable response after orthotopic liver transplantation on adefovir dipivoxil and hepatitis B immune globulin

BACKGROUND

Mutations in the hepatitis B virus (HBV) genome may occur during therapy.

METHODS

We report an asymptomatic HBV carrier who underwent transplantation for end-stage renal disease. She developed an HBV flare 6 months after transplantation and was placed on lamivudine. After initial rapid improvement, she relapsed clinically and virologically. She decompensated with jaundice, peripheral edema, ascites, encephalopathy, coagulopathy, and hepatorenal syndrome. A liver biopsy specimen revealed submassive necrosis.

RESULTS

Emergency liver transplantation was performed: lamivudine was discontinued. Hepatitis B immunoglobulin and adefovir dipivoxil were initiated. Sixteen months after orthotopic liver transplantation, she is HBV DNA seronegative with normal liver enzymes. Sequencing of HBV polymerase gene from preliver transplantation sera did not detect the usual lamivudine resistance mutations in the YMDD motif but instead two other mutations (F514-->L, L528-->M). Lamivudine resistance was demonstrated in vitro.

CONCLUSIONS

Asymptomatic HBV carriers may reactivate following renal transplantation after immunosuppression. Resistance to lamivudine may result in severe hepatic damage in immunocompromised patients.

Biphasic clearance kinetics of hepatitis B virus from patients during adefovir dipivoxil therapy

In a recent phase II clinical study, 13 chronic hepatitis B-infected patients treated daily with 30 mg adefovir dipivoxil for 12 weeks displayed a median 4.1-log10 decrease in plasma hepatitis B virus (HBV)-DNA levels. The decline of viral load during therapy displayed a biphasic kinetic profile that was modeled to determine the efficacy of inhibition of viral production, as well as kinetic constants for the clearance of free virus and the loss of infected cells. Viral production was suppressed with an efficacy of 0.993 +/- 0.008, indicating that only 0.7% of viral production persisted during therapy. The initial, faster phase of viral load decline reflects the clearance of HBV particles from plasma with a half-life of 1.1 +/- 0.3 days, translating to a 48% daily turnover of the free virus. The second, slower phase of viral load decline closely mirrors the rate-limiting process of infected cell loss, with a half-life of 18 +/- 7 days. The duration of therapy required to completely eliminate the virus from plasma or suppress it to levels sufficient to induce seroconversion is a function of the half-life of the free virus, the half-life of infected cells, and the efficacy of inhibition of virus production from infected cells. These quantitative analyses provide a more detailed picture of the dynamics of HBV infection and therapy, and can be used to compare the efficacy of various doses and inhibitors of HBV replication for the treatment of HBV infections.

Mutations in hepatitis B DNA polymerase associated with resistance to lamivudine do not confer resistance to adefovir in vitro

To determine whether adefovir is active against lamivudine-resistant hepatitis B virus (HBV), the inhibition constants of adefovir diphosphate and lamivudine triphosphate for wild-type and mutant human HBV DNA polymerases, which contain amino acid substitutions associated with lamivudine resistance, were compared. Recombinant wild-type and mutant human HBV DNA polymerases were expressed and substantially purified using a baculovirus expression system and immunoaffinity chromatography. HBV DNA polymerase mutants M552I, M552V, and L528M/M552V showed resistance to lamivudine triphosphate with inhibition constants (Ki) increased by 8.0-fold, 19.6-fold, and 25.2-fold compared with that of wild-type HBV DNA polymerase. However, these mutants remained sensitive to adefovir diphosphate with the inhibition constants increasing by 1.3-fold and 2.2-fold or decreasing by 0.79-fold. The L528M single mutation, identified in patients with increasing HBV DNA levels during therapy with famciclovir, also remained sensitive to adefovir diphosphate with the inhibition constant increased by only 2.3-fold.

Allosteric effects of a monoclonal antibody against thrombin exosite II

We previously isolated a monoclonal antithrombin IgG from a patient with multiple myeloma [Colwell et al. (1997) Br. J. Haematol. 97, 219-226]. Using a panel of 55 surface mutants of recombinant thrombin, we now show that the epitope for the IgG most likely includes Arg-101, Arg-233, and Lys-236 in exosite II. The IgG affects the rate at which thrombin cleaves various peptide p-nitroanilide substrates with arginine in the P1 position, increasing the kcat for substrates with a P2 glycine residue but generally decreasing the kcat for substrates with a P2 proline. The allosteric effect of the IgG is altered by deletion of Pro-60b, Pro-60c, and Trp-60d from the 60-loop of thrombin, which lies between exosite II and the catalytic triad. The effect of the IgG, however, does not depend on the presence or absence of sodium ions, a known allosteric regulator of thrombin. The IgG does not affect the conformation of thrombin exosite I as determined by binding of a fluorescent derivative of hirudin54-65. These results provide evidence for a direct allosteric linkage between exosite II and the catalytic site of thrombin.

Strategies for development of novel antithrombotics: modulating thrombin's procoagulant and anticoagulant properties

Thrombin is a serine proteinase that can interact with a large number of diverse macromolecular substrates, which results in either a procoagulant or anticoagulant effect. These divergent properties are physiologically regulated by the endogenous protein thrombomodulin. This review summarizes recent work on a variety of methods used to exploit the allosteric nature of the enzyme. The procoagulant and anticoagulant functions of thrombin can be modulated by sodium binding, site-directed mutagenesis, and a small synthetic molecule. Modulation of thrombin's intrinsic properties represents a novel approach to the development of unique antithrombotic agents.

Modulation of thrombin's procoagulant and anticoagulant properties

The procoagulant and anticoagulant functions of thrombin are controlled physiologically by allosteric changes induced by Na+ and vascular cell-surface TM. Key residues that mediate Na+ interaction with thrombin have been identified. Based on a site-directed mutagenesis approach, E229K thrombin is found to be the most optimal and potent PC activator with a marked shift in substrate specificity for PC over fibrinogen. E229K thrombin demonstrates significant anticoagulant and antithrombotic efficacy in animal models in vivo. Alternatively, a synthetic organic molecule (LY254603) has been discovered which interacts with thrombin and effectively modulates its functions in vitro. This new class of antithrombotic agents exploits the powerful natural PC anticoagulant pathway and may have a superior therapeutic profile than direct thrombin inhibitors.

Functional requirements for inhibition of thrombin by antithrombin III in the presence and absence of heparin

Mutation of 79 highly exposed amino acids that comprise approximately 62% of the solvent accessible surface of thrombin identified residues that modulate the inhibition of thrombin by antithrombin III, the principal physiological inhibitor of thrombin. Mutations that decreased the susceptibility of thrombin to inhibition by antithrombin III in the presence and absence of heparin (W50A, E229A, and R233A) also decreased hydrolysis of a small tripeptidyl substrate. These residues were clustered around the active site cleft of thrombin and were predicted to interact directly with the "substrate loop" of antithrombin III. Despite the large size of antithrombin III (58 kDa), no residues outside of the active cleft were identified that interact directly with antithrombin III. Mutations that decreased the susceptibility of thrombin to inhibition by antithrombin III in the presence but not in the absence of heparin (R89A/R93A/E94A, R98A, R245A, K248A, K252A/D255A/Q256A) in general did not also affect hydrolysis of the tripeptidyl substrate. These residues were clustered among a patch of basic residues on a surface of thrombin perpendicular to the face containing the active site cleft and were predicted to interact directly with heparin. Three mutations (E25A, R178A/R180A/D183A, and E202A) caused a slight enhancement of inhibition by antithrombin III.

Protein engineering thrombin for optimal specificity and potency of anticoagulant activity in vivo

Previous alanine scanning mutagenesis of thrombin revealed that substitution of residues W50, K52, E229, and R233 (W60d, K60f, E217, and R221 in chymotrypsinogen numbering) with alanine altered the substrate specificity of thrombin to favor the anticoagulant substrate protein C. Saturation mutagenesis, in which residues W50, K52, E229, and R233 were each substituted with all 19 naturally occurring amino acids, resulted in the identification of a single mutation, E229K, that shifted the substrate specificity of thrombin by 130-fold to favor the activation of the anticoagulant substrate protein C over the procoagulant substrate fibrinogen. E229K thrombin was also less effective in activating platelets (18-fold), was resistant to inhibition by antithrombin III (33-fold and 22-fold in the presence and absence of heparin), and displayed a prolonged half-life in plasma in vitro (26-fold). Thus E229K thrombin displayed an optimal phenotype to function as a potent and specific activator of endogenous protein C and as an anticoagulant in vivo. Upon infusion in Cynomolgus monkeys E229K thrombin caused an anticoagulant effect through the activation of endogenous protein C without coincidentally stimulating fibrinogen clotting and platelet activation as observed with wild-type thrombin. In addition, E229K thrombin displayed enhanced potency in vivo relative to the prototype protein C activator E229A thrombin. This enhanced potency may be attributable to decreased clearance by antithrombin III, the principal physiological inhibitor of thrombin.

A nickel chelate microtiter plate assay for six histidine-containing proteins

Protein purification has been made significantly easier by the use of affinity tags that can be genetically engineered at either the amino- or carboxyl-terminus of recombinant proteins. One of the most widely used tags is six consecutive histidine residues or 6His tag. These residues bind with high affinity to metal ions immobilized on chelating resins even in the presence of denaturing agents and can be mildly eluted with imidazole. We report the methodology for the immobilization of six histidine-containing proteins onto microtiter plates. A derivative of nitrilotriacetic acid (NTA) was prepared. This derivative, N,N-bis[carboxymethyl]lysine (BCML), was easily coupled to a maleic anhydride-activated polystyrene microtiter plate. The plate was then charged with Ni2+ for the capture of the 6His-tagged proteins. Using two different recombinant proteins with the 6His tag at either the N- or C-terminus, we demonstrated that the binding to the Ni(2+)-NTA plate was specific for six histidine-containing proteins. Proteins lacking the 6His tags did not bind to the plate. The plate was used in a modified enzyme-linked immunoabsorbent assay format to quantitate protein concentrations and determine the affinity of protein-ligand interactions. The technology can also be extended to include high-throughput screening assays for antagonists of protein-protein interactions.

Conversion of thrombin into an anticoagulant by protein engineering

At sites of vascular injury, thrombin interacts with multiple procoagulant substrates, to mediate both fibrin clotting and platelet aggregation. But upon binding to thrombomodulin on the vascular endothelium, thrombin instead activates protein C, thereby functioning as an anticoagulant and attenuating clot formation. Upon infusion in vivo, both the procoagulant and anticoagulant effects of thrombin were observed. Preliminary studies indicating that thrombin's protein C activating and fibrinogen clotting activities could be dissociated by mutagenesis suggested to us that a thrombin variant that lacked procoagulant activity while retaining anticoagulant function might be an attractive antithrombotic agent. Using protein engineering, we introduced a single substitution, E229A, that substantially shifted thrombin's specificity in favour of the anticoagulant substrate, protein C. In monkeys, this modified thrombin functioned as an endogenous protein C activator demonstrating dose-dependent, reversible anticoagulation without any indication of procoagulant activity. Notably, template bleeding times were not prolonged, suggesting a reduced potential for bleeding complications.

A peptide derived from a tissue factor loop region functions as a tissue factor--factor VIIa antagonist

Tissue factor (TF) is a transmembrane protein that functions in the initiation of blood coagulation in vivo. At sites of vascular injury, TF serves as a cell-surface receptor for the serine protease factor VIIa (FVIIa), forming an enzyme--cofactor complex and enhancing the catalytic activity of FVIIa. Tissue factor, along with the receptors for alpha- and gamma-interferons, is a member of the class 2 cytokine receptor superfamily. Crystallographic analysis demonstrated that the extracellular domain of TF consists of two immunoglobulin-like domains joined by a linker region. Each domain is comprised of two antiparallel beta-sheets containing seven conserved beta-strands separated by more variable loop regions. Extensive mutagenesis has been performed in order to map the FVIIa binding site on TF. Results indicated that the discontinuous binding site for FVIIa lies at the domain--domain interface and includes residues from extended loops and beta-strands within both the N- and C-terminal domains. Our previous study provided evidence that three consecutive residues (D44, W45, K46) within the TF loop region between beta-strands C and C' of the N-terminal domain were important for interactions with FVIIa. We have presently extended our alanine-scanning mutagenesis to include the residues within the flanking beta-strands. Thirteen sTF mutants were screened for their ability to enhance FVIIa activity. Three residues within strand C (Y34, Q37, I38) and two residues within C' (K48, Y51) were shown to be important for TF cofactor function.(ABSTRACT TRUNCATED AT 250 WORDS)

Selection of a suppressor mutation that restores affinity of an oligonucleotide inhibitor for thrombin using in vitro genetics

The thrombin aptamer is a single-stranded DNA of 15 nucleotides that was identified by the selection of thrombin-binding molecules from a large combinatorial library of oligonucleotides. This prototype aptamer of thrombin has a unique double G-tetrad structure capable of inhibiting thrombin at nanomolar concentrations through binding to a specific region within thrombin exosite I. Substitution of arginine 70 in thrombin exosite I with glutamic acid effectively eliminated binding of the prototype thrombin aptamer. In contrast, aptamers selected against R70E thrombin were able to bind and inhibit both wild-type and R70E thrombins, and displayed potassium-independent inhibition. Aptamers selected against R70-E thrombin bound to sites identical or overlapping with that of the prototype thrombin aptamer. These aptamers retained the potential to form double G-tetrad structures; however, these structures would be destabilized by a T-->A substitution, disrupting the T4-T13 base pairing found in the prototype. This destabilization appeared to be partially compensated by newly recruited structural elements. Thus, selection against R70E thrombin did not lead to aptamers that bound to alternative sites, but instead to ssDNA structures with a suppressor mutation that accommodated the mutation in thrombin within a double G-tetrad context. These results provide insight into the aptamer-thrombin interaction and suggest that the binding site for the prototype is the dominant aptamorigenic site on thrombin.

Functional mapping of the surface residues of human thrombin

Utilizing site-directed mutagenesis, 77 charged and polar residues that are highly exposed on the surface of human thrombin were systematically substituted with alanine. Functional assays using thrombin mutants identified residues that were required for the recognition and cleavage of the procoagulant substrate fibrinogen (Lys21, Trp50, Lys52, Asn53 + Thr55, Lys65, His66, Arg68, Tyr71, Arg73, Lys77, Lys106 + Lys107, Asp193 + Lys196, Glu202, Glu229, Arg233, Asp234) and the anticoagulant substrate protein C (Lys21, Trp50, Lys65, His66, Arg68, Tyr71, Arg73, Lys77, Lys106 + Lys107, Glu229, Arg233), interactions with the cofactor thrombomodulin (Gln24, Arg70) and inhibition by the thrombin aptamer, an oligonucleotide-based thrombin inhibitor (Lys65, His66, Arg70, Tyr71, Arg73). Although there is considerable overlap between the functional epitopes, distinct and specific residues with unique functions were identified. When the functional residues were mapped on the surface of thrombin, they were located on a single hemisphere of thrombin that included both the active site cleft and the highly basic exosite 1. No functional residues were located on the opposite face of thrombin. Residues with procoagulant or anticoagulant functions were not spatially separated but interdigitated with residues of opposite or shared function. Thus thrombin utilizes the same general surface for substrate recognition regardless of substrate function although the critical contact residues may vary.

Identification of the factor VIIa binding site on tissue factor by homologous loop swap and alanine scanning mutagenesis

Tissue factor (TF) is a membrane-bound glycoprotein that functions as a cofactor for coagulation factor VIIa (VIIa) and initiates blood coagulation at sites of vascular injury. On the basis of sequence alignments, TF was predicted to be a member of the cytokine receptor superfamily. Utilizing the structural information available for the cytokine receptor superfamily, we have used site-directed mutagenesis to identify the binding site on TF for VIIa. The predicted loop regions in TF were systematically replaced with the homologous loops from the gamma-interferon receptor (gamma-IFN-R), the protein most related to TF in the superfamily of cytokine receptors. Six discontinuous regions (residues 16-20, 40-46, 60-69, 101-111, 129-151, 193-207) were identified that are required for interaction with VIIa and enhancement of activity. Individual substitution of 68 residues within these loops with alanine revealed eight residues (K20, D44, W45, K46, Q110, R135, F140, V207) that are required for cofactor activity. These residues fall into two groups, those that are required only for interactions with VIIa (K46, Q110, R135, F140, V207) and those that are also required to induce the conformational change in VIIa required for enhanced activity (K20, D44, W45). The discontinuous regions of TF required for interactions with VIIa form a single binding surface for VIIa that is analogous to the interface defined by the crystal structure of the complex between growth hormone and its receptor.(ABSTRACT TRUNCATED AT 250 WORDS)

A template for the protein kinase family

The crystal structure of the catalytic subunit of cAMP-dependent protein kinase, complexed with ATP and a 20-residue inhibitor peptide, is reviewed and correlated with chemical and genetic data. The striking convergence of the structure with the biochemistry and genetics provides for the first time a molecular basis for understanding how this enzyme functions, as well as an explanation for the highly conserved residues that are scattered throughout the molecule. Because these residues probably serve a common role in all eukaryotic protein kinases, this first protein kinase structure serves as a general template for the entire family of enzymes.

Systematic mutational analysis of cAMP-dependent protein kinase identifies unregulated catalytic subunits and defines regions important for the recognition of the regulatory subunit

A library of mutants of the catalytic subunit of the Saccharomyces cerevisiae cAMP-dependent protein kinase was screened in vitro for mutants defective in the recognition of the regulatory subunit. The mutations identified were mapped onto the three-dimensional structure of the mouse catalytic subunit with a peptide inhibitor. Mutations defective in the recognition of both the regulatory subunit and the peptide substrate Leu-Arg-Arg-Ala-Ser-Leu-Gly (Kemptide) mapped to the peptide-binding site shared by all substrates and inhibitors of the catalytic subunit and functionally define the binding site for the autoinhibitor sequence in the hinge region of the regulatory subunit. Mutants defective only in the recognition of the regulatory subunit identified residues that comprise additional binding sites for the regulatory subunit. The majority of these residues are clustered on the surface of the catalytic subunit in a region flanking the distal portion of the autoinhibitor/peptide-binding site. The simultaneous substitution of Lys233, Asp237, Lys257, and Lys261 in this region caused a 260-fold decrease in affinity for the regulatory subunit, whereas the catalytic efficiency toward Kemptide decreased by only 1.8-fold. The substitution of autophosphorylated Thr241, also in this region, and the 3 residues interacting with the phosphate also caused an unregulated phenotype.

Identification of electrostatic interactions that determine the phosphorylation site specificity of the cAMP-dependent protein kinase

"Charged-to alanine" scanning mutagenesis of the catalytic subunit of the Saccharomyces cerevisiae cAMP-dependent protein kinase (C1) identified three glutamate residues, E171, E214, and E274, that are involved in the recognition of a peptide substrate, kemptide (Leu1Arg2Arg3Ala4Ser5Leu6Gly7). These glutamate residues are conserved or conservatively substituted with asparate in the serine/threonine protein kinases that have a requirement for basic residues on the N-terminal side of their phosphorylation sites. Alanine replacement mutants in C1 were subjected to kinetic analysis using alanine-substituted peptides as substrates. The additivity or nonadditivity of the effects of the alanine substitutions on the catalytic efficiency (kcat/Km) was analyzed. This allowed the identification of electrostatic interactions between the three glutamate residues in the enzyme and the two arginine residues present in the peptide substrate. The data suggest that E171 interacts with Arg2 in the substrate and that E214 and E274 both interact with Arg3. This may be a general method for identifying simple intermolecular interactions involving proteins when there is no three-dimensional structure available of the complex of interacting species. The identification of these interactions provides the potential for rational protein engineering of enzymes with alternative specificities.

Rational scanning mutagenesis of a protein kinase identifies functional regions involved in catalysis and substrate interactions

A systematic mutagenesis strategy was used to identify the functional regions and residues of a protein kinase. Clusters of the charged amino acids in the catalytic subunit of Saccharomyces cerevisiae cAMP-dependent protein kinase, were systematically mutated to alanine, producing a set of mutations that encompassed the entire molecule. Residues indispensable for enzyme activity were identified by testing the ability of the mutants to function in vivo. Active mutants were assayed in vitro, and mutants with reduced specific activity were subsequently analyzed by steady-state kinetics to determine the effects of the mutation on kcat and on Km for MgATP and for a peptide substrate. Specific residues and regions of the enzyme were identified that are likely to be important in catalysis and in binding of MgATP, functions that are common to all protein kinases. Additional regions were identified that are likely to be important in binding a peptide substrate, the recognition of which is likely to be specific to the serine/threonine protein kinases that have a requirement for basic residues around the target hydroxyamino acid. The properties of mutants defective in substrate recognition were consistent with an ordered sequential reaction mechanism. This represents the first comprehensive analysis of a protein kinase by a rational mutagenesis strategy.