Alanine-scanning Mutagenesis of the Epidermal Growth Factor-like Domains of Human Thrombomodulin Identifies Critical Residues for Its Cofactor Activity*

Thrombomodulin (TM) is an endothelial cell surface-bound cofactor in thrombin-dependent formation of activated protein C, a potent anticoagulant. Cofactor activity has been localized to the carboxyl-terminal half of the six epidermal growth factor-like (EGF) domains of TM (TMz). To identify residues in TME that are critical for activity, 77 alanine point mutants were made between Cys-333 and Cys-462 by site-directed mutagenesis (all residues except Ala, Cys, Gly, and Pro). Mutants were expressed in Escherichia coli and cofactor activity measured directly in periplasmic ex- tracts obtained by osmotic shock. Critical residues were defined as those which when mutated had less than 25% cofactor activity of a reference TME. Western blots of non-reduced samples confirmed that alanine substitutions did not significantly decrease expression levels or result in the formation of multimers. In EGF4, which is essential for protein C activation by the thrombin-TM complex, critical residues were: Glu-357, Tyr-358, and Phe-376. In EGFB-EGF6, critical residues within a proposed acidic thrombin-binding region were: Glu-408, Tyr-413, Leu- 415, Asp-416, Asp-417, Asp-423, Asp-425, and

Alanine-scanning Mutagenesis of the Epidermal Growth Factor-like Domains of Human Thrombomodulin Identifies Critical Residues for Its Cofactor Activity* (Received for publication, October 1, 1992) Mariko Nagashima, Erik Lundh, Jill C. Leonard, John Morser, and John F. Parkinson$ From Berlew Biosciences Znc.,Richmond, Thrombomodulin (TM) is an endothelial cell surfacebound cofactor in thrombin-dependent formation of activated protein C, a potent anticoagulant. Cofactor activity has been localized to the carboxyl-terminal half of the six epidermal growth factor-like (EGF) domains of TM (TMz). To identify residues in TME that are critical for activity, 77 alanine point mutants were made between Cys-333 and Cys-462 by site-directed mutagenesis (all residues except Ala, Cys, Gly, and Pro). Mutants were expressed in Escherichia coli and cofactor activity measured directly in periplasmic extracts obtained by osmotic shock. Critical residues were defined as those which when mutated had less than 25% cofactor activity of a reference TME. Western blots of non-reduced samples confirmed that alanine substitutions did not significantly decrease expression levels or result in the formation of multimers. In EGF4, which is essential for protein C activation by the thrombin-TM complex, critical residues were: Asp-349,  EGFB-EGF6, critical residues within a proposed acidic thrombinbinding region were: . A potential Ca2+-binding site, which is comprised of residues Asp-423, , was also identified and overlaps the thrombin-binding region. Asp-461, in the C-loop of EGF6 previously shown to be critical for thrombin binding, was also critical. Asp-390, Asp-400,  in EGF5 were also critical. Thus, rapid alanine-scanning mutagenesis of TME has identified 22 critical residues in the region comprising EGF4-6, which is essential for thrombin binding and protein C activation by the thrombin-TM complex.
Thrombomodulin (TM)' is an endothelial cell surface glycoprotein that binds thrombin with high affinity. Binding of thrombin to TM alters the conformation of thrombin, leading t o accelerated formation of activated protein C (1)(2)(3). In the presence of its cofactor, protein S, activated protein C inactivates clotting factors Va and VIIIa, thereby down-regulating the coagulation cascade (4)(5)(6). Complex formation between * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' The abbreviations used are: TM, thrombomodulin; EGF, epidermal grbwth factor; TME, wild-type T M comprising EGF domains 1-6, T M E -M~~~L , TME containing Leu instead of Met at position 388 Tricine, N-[2-hydroxy-l,l-bis(hydroxymethyl)ethyl]glycine. thrombin and T M also results in direct inhibition of the procoagulant activities of thrombin, such as fibrin formation and platelet activation (7)(8)(9). Thus, TM plays an important role in maintaining the anticoagulant properties of the endothelial cell surface.
Human TM is a single polypeptide chain of 557 residues consisting of five domains: an NH2-terminal "lectin-like" domain, six epidermal growth factor-like (EGF) repeats, an 0-glycosylation domain, the transmembrane domain, and a cytoplasmic domain (10)(11)(12)(13). Structure-function studies using proteolytic fragments of rabbit TM or deletion mutants of recombinant human T M (14-18) have localized cofactor activity to EGF repeats 4, 5, and 6 (EGF4-6), although the exact thrombin-TM interaction has not been elucidated.
In order to identify residues that are critical for interaction with thrombin and in protein C activation by the thrombin-T M complex, residues within EGF4-6 were systematically replaced with alanine by site-directed mutagenesis, and the effect of each substitution on cofactor activity was determined. Alanine was used as a substitution, since it would not alter the main chain conformation nor would it impose extreme electrostatic or steric effects (19). T o facilitate the investigation, the domain consisting of the six EGF-like repeats of human TM (TME) was expressed in Escherichia coli, and cofactor activity was measured directly in periplasmic extracts prepared by osmotic shock. A total of 77 mutants was constructed in the same expression plasmid, and their relative activities were compared with the activity of the reference TME. Here, we report the results of a preliminary screen that identified 22 residues within EGF4, -5, and -6 of human TM as critical for thrombin-dependent activation of protein C.

MATERIALS AND METHODS
T4 DNA ligase, pSELECT-1 vector, and pGEM-3Zf were from Promega Corp., Madison, WI. E. coli strain DH5a was from GIBCO/ BRL, Grand Island, NY. T4 DNA polymerase and E. coli strain CJ236 (dut-ung-) were from Bio-Rad. Restriction enzymes were from either New England Biolabs, Beverly, MA or Boehringer Mannheim. Plasmid pKT279 was from Stratagene, La Jolla, CA. Human a-thrombin (-4000 NIH units/mg) and hirudin (300-1000 units/mg) were from Sigma. Recombinant human protein C was from Dr. John McPherson, Genzyme Corp., Framingham, MA and was purified as described (18). Chromogenic substrate D-valyl-L-leucyl-L-arginine-pnitroanilide (S-2266) was from Kabi Vitrum, Franklin, OH. Rabbit polyclonal antiserum raised against purified reduced and alkylated TME expressed in insect cells was kindly provided by Julia Gray, Berlex Biosciences. Biotinylated goat anti-rabbit IgG and ECL detection kit were from Amersham Corp. Vectastain ABC kit was from Vector Laboratories, Burlingame, CA.

Construction of E. coli Expression Plasmids-A
DNA fragment coding for TME (amino acids 227-462)' was obtained by polymerase chain reaction of human genomic DNA using primers 5'CCGGG-Residue numbering is according to Suzuki et al. (12). ATCCTCAACAGTCGGTGCCAATGTGGCG3' and 5'CCGGG-ATCCTGCAGCGTGGAGAACGGCGGCTGC3'. This fragment, through a series of intermediate constructs, was placed under the control of the @-lactamase promoter and signal sequence in pKT279. An EcoR5-Bgl2 fragment of the resultant plasmid and a ScaI-Sac1 fragment of pGEM-3Zf DNA containing the fl origin of replication were then inserted into pSELECT-1 vector at EcoR5-BamHI and ScaI-Sac1 sites, respectively, to construct an E. coli expression vector, pTHR211, coding for TME. Plasmid pTHR211 was used to generate plasmid pMJM57 containing a Leu substitution for Met-388 (TME-M388L) using site-directed mutagenesis as described below.
Site-directed Mutagenesis-Plasmids coding for T M E -M~~~L alanine mutants were constructed using site-directed mutagenesis as described by Kunkel et al. (20). A single-stranded uracil DNA template, prepared from E. coli strain CJ236 with R408 helper phage, was used for synthesis of the mutagenic strand in the presence of specific oligonucleotides with T4 DNA polymerase and T4 DNA ligase. We incorporated a restriction enzyme recognition sequence in each oligonucleotide without changing the amino acid sequence, and the resultant transformants were characterized by restriction enzyme digests of isolated plasmids. Three independent positive clones were isolated for each mutant except Y368A, E408A, E411A, and I414A, for which only two positive clones were obtained. In cases where there were large discrepancies between cofactor activities of replicates, plasmids were further characterized by dideoxy sequencing (21).
Preparation of Periplasmic Extracts"DH5a cells expressing TME-M388L alanine mutants were grown overnight in 1.5 ml of L-broth containing ampicillin (50 pg/ml) at 37 "C. Cells were harvested by centrifuging at 14,000 rpm in a microcentrifuge for 25 s, washed once with 0.5 ml of 100 mM Tris-HCI, pH 8.0,50 mM NaC1, and suspended for 10 min in 0.5 ml of 300 mM Tris, pH 8.0, 20% sucrose, 1 mM EDTA, 0.5 mM MgC12. Cells were centrifuged as before, and periplasmic extracts were obtained by a 10-min incubation at 4 "C in 150 pl of 0.5 mM MgCI,, followed by a 5-min centrifugation at full speed at 4 "C in a microcentrifuge.

7" Cofactor Activity Assay-25
pl each of periplasmic extract, protein C (0.5 p~ final), and thrombin (1 nM final) were mixed and assayed for TM cofactor activity as described (18). The final Ca2+ concentration was 2.5 mM. All assays contained triplicate extracts each of DH5a cells transfected with either pSELECT-1 vector (no TM& wild-type TME (pTHR211), and l"~"388L (pMJM57) as internal controls. A standard curve of purified T M E -M~~~L expressed in insect Sf9 cells (provided by Michael McCaman, Berlex Biosciences) was used so that milli-absorbance units/min at 405 nm could be converted to units/ml. Purified Th&-M388L has a specific activity of 802,000 units/mg, where 1 unit is defined as 0.06 pmol of activated protein C formed per min at 37 'C under the conditions of the assay. Cofactor activities of T M E -M~~~L alanine mutants were expressed as percentages of T M E -M~~~L .
Apparent Kd values for thrombin binding to T&M388L alanine mutants were determined from the dependence of cofactor activity on thrombin concentration in the 1-60 nM range. Linear regression analysis of double-reciprocal plots of l/cofactor activity verszu 1/ [thrombin] gave an intercept on the x axis of -I/&. Apparent Kd values were only determined for mutants where significant activity could be measured over the thrombin concentration range used.
Statistical Analysis-Each clone of each mutant was assayed for activity at least twice (three times for those mutants for which only two positive clones were isolated), and all the data were included in the determination of the significance of difference using a Student's t test.
Western Blot Analysis-Non-reduced samples of periplasmic extracts were electrophoresed on 10% Tris-Tricine SDS-PAGE gels (Novex Inc., San Diego, CA). Proteins were transferred to a nitrocellulose filter in transfer buffer (192 mM glycine, 25 m M Tris, pH 8.3, 20% methanol) at 4 "C. The filter was blocked with 1% bovine serum albumin in 10 mM Tris, pH 7.5, 0.9% NaCI, 0.05% NaN3 and then incubated with polyclonal anti-TME serum in blocking buffer (1:2000 dilution). After washing in 10 mM Tris, pH 7.5, 0.9% NaCI, 0.05% NaN3, 0.05% Tween 20, the filter was incubated with biotinylated second antibody in blocking buffer containing 0.05% Tween 20. Following washing as before, proteins were detected using the Vectastain ABC solution and ECL detection system according to the manufacturer's specifications.

RESULTS AND DISCUSSION
TM Secreted in E. coli Is Functionally Active-TME was expressed in E. coli under the control of the p-lactamase promoter and directed to the periplasmic space with the plactamase signal sequence. Secreted TME cofactor activity in thrombin-dependent activation of protein C could be measured directly in the periplasmic extracts without prior protein purification. Using this system, wild-type TME had a cofactor activity of 8.6 k 2.8 units/ml ( n = 18). Assays were performed at 2.5 mM Ca2+ to reduce the background rate of protein C activation by thrombin alone. Background rates of protein C activation in extracts of E. coli transfected with control plasmid pSelect-1 were <0.5% of those obtained with wildtype TME. Glaser et al. (22) have demonstrated that substitution of Met-388 with Leu (TME-M388L) in purified TME results in approximately 2-fold higher specific activity when compared with wild-type TME. In periplasmic extracts, 'l"~"388L had a cofactor activity of 20.1 k 5.2 units/ml ( n = 18). The ratio of TME-M388L/wi1d-type TME cofactor activities was 2.35 k 0.39, with an interassay coefficient of variation of 16.7% ( n = 18). Since total expression levels of wild-type TME and TME-M388L in extracts were similar by Western blot analysis using polyclonal anti-TME antibody (see Fig. 2), the ratio of T M E -M~~~L to wild-type TME activities in extracts was consistent with those of purified proteins. This result was confirmed when specific activities for TME and T&-M388L were determined by measuring antigen levels in E. coli shockates using purified recombinant human thrombomodulin expressed in mammalian cells and a sandwich enzyme-linked immunosorbent assay comprising two monoclonal anti-TME antibodies (43B and 531). The specific activities of TME and TME-M388L were 437,000 f 24,000 and 1,019,000 & 87,000 units/mg of protein, respectively, giving a Th&-M388L/TME specific activity ratio of 2.33 ( n = 2).3 Thus, the current system provided a fast and easy method of comparing cofactor activities of different mutants.
Identification of TME Residues That Are Critical for Cofactor Activity-We previously identified the C-loop of EGF3 to EGF6 as the smallest TME fragment with full cofactor activity (18). In order to investigate the contribution of individual residues in this region to cofactor activity, each residue between Cys-333 and Cys-462 (except Ala, Cys, Gly, and Pro) was replaced with alanine by site-directed mutagenesis. A total of 77 mutants were constructed using TME-M388L as a template. Three independent clones were isolated for each alanine mutant, and each was assayed at least twice. Results were expressed as mean percentages of TME-M388L (Fig. 1).
Of the 77 mutants, 22 had <25% of T M E -M~~~L cofactor activity. Other mutations produced proteins with activity ranging between 25 and 100% of Thl~"388L. Two mutants, Q365A and L369A, were found to have 10-15% higher activity than T M E -M~~~L .
These increases were found to be additive in a double mutant (results not shown). None of the 22 critical mutations caused gross changes in the growth rate of DH5a cells. To eliminate the possibility that loss of activity in the 22 critical mutants could be accounted for by gross changes in disulfide bonding or by reduced expression levels, a Western blot of periplasmic extracts was performed under nonreducing conditions (Fig. 2). No evidence for the formation of multimers was found, indicating that none of the low activity mutants arose from disulfide scrambling. In contrast to TME-M388L and TME, which migrated as single bands, some mutants migrated as doublets. The basis for these doublets is not clear, and the possibility that they arise from alternative disulfide pairing cannot be excluded. Most of the mutants were also expressed at similar levels to the TME and and T M E -M~~~L standards, again indicating that loss of activity could not be accounted for by decreased expression. For those E. Campbell and M. McCaman, personal communication. M388L. In all cases, the normalized cofactor activity was still <25% of T M E -M~~~L (data not shown).

Structure-Function Studies
Of the 22 critical residues identified, there were 11 negatively charged residues but no positively charged residues. Furthermore, all but 4 of the 22 residues (Asn-402, Glu-408, Asn-429, and Leu-440) were either identical to or conservative substitutions of the amino acids in the corresponding positions in human, mouse, and bovine TM.
Residues in the Thrombin-binding Region-Six of the negatively charged residues (Glu-408, Asp-416, Asp-417, Asp-423, Asp-425, and Glu-426) were found within a region spanning the C-loop of EGF5 to the interdomain between EGF5 and EGF6. This region is known to play a key role in thrombin binding (16,17,23). While overall negative charges were presumably important for the binding to the anion-binding exosite of thrombin, substitutions of larger hydrophobic residues in this region with a smaller hydrophobic residue also resulted in the loss of activity (I414A, I415A, 1424A). Preliminary experiments indicated that I424A had a much lower thrombin affinity than T M E -M~~~L (see Table I). The bulky side chains of leucine and isoleucine may be required for the correct spatial arrangement of TM to be recognized by thrombin. Decreased cofactor activity of the D461A mutant might also be due, in part, to the decreased affinity for thrombin. In earlier studies using purified deletion mutants, it was shown that the deletion of residues 447-462 from wild-type TME increased Kd for thrombin about 6-fold without affecting kcat/ Km (18).
Identification of a Putative Ca2+-binding Site-As shown in Fig. 3, the region of TME spanning residues 423-444 contains a consensus sequence for a Ca2+-binding site, Asp-Ile-Asp-

Glu-Cys-X-X-X-X-Cys-X-X-X-X-Cys-X-Asn-X-X-
Phe-X-Cys-X-Cys, which is highly homologous to known Ca2+-binding EGF domains in protein S and factor IX (24,25). In the present study we found that alanine substitution of , and Phe-444 led to large decreases in TME cofactor activity. Identification of every single non-cysteine residue within the consensus sequence for the Ca2+-binding site as being critical for TME activity is a significant finding. Although indirect, these are the first experimental results supporting the presence of a Ca2+-binding site in this region of TME and suggest that this may be an important structural feature of this thrombin-binding region of TME (see below). Residues That May Alter the Substrate Specificity of Thrombin-Binding of TM to thrombin causes multiple conformational changes around the catalytic site of thrombin, leading to alteration in its substrate specificity. Using fluorescent labeled thrombin, Ye et al. (26) have shown that a fragment consisting of EGF5 and EGFG can induce only partial confor-mational changes, which do not lead to the activation of protein C. In the current studies, we found 4 residues outside EGF5 and -6 that were critical for cofactor activity . While it remains possible that these alanine mutations introduced major structural perturbations, these residues could be interacting with thrombin to bring about additional conformational changes within thrombin required for protein C activation. Preliminary kinetic data (see Table I) suggest that Phe-376 may be involved in thrombin binding.
Le Bonniec et al. (27,28) have reported that Glu-192 and Glu-39 near the active site of thrombin were responsible for its inability to accept Asp residues in the P3 and P3' positions of the protein C cleavage sequence. They also postulated that TM could function, in part, by alleviating these inhibitory effects. It would be interesting to investigate if any of these 4 residues in EGF4 were responsible for these effects. In our deletion studies, the mutant consisting of residues 351-462 had the same Kd for thrombin, but 90% reduction in kCat/K,,,, compared with wild-type TME (18). Deletion or mutation of Asp-349 might be responsible for this loss of catalytic activity (29).
Decreased Thrombin Binding Affinity of Alanine Mutants-In the present study, critical residues were defined as those which when mutated to alanine had 25% or less of M388L cofactor activity. This loss of activity could arise from direct effects on & for thrombin, k,,/K,,, of the thrombin-TM complex for protein C, or from effects on both K d and kcat/ K,,,. To address this issue we have performed preliminary kinetic experiments in which the apparent K d of the low activity mutants for thrombin was determined directly in extracts from the dependence of cofactor activity on thrombin concentration. The data are summarized in Table I. The data show that many of the alanine mutations in the thrombinbinding region and the putative Ca2+-binding site cause increases in the apparent Kd for thrombin. These data suggest that the effects of these mutations are on the function of the correctly folded protein and not on protein folding per se. This conclusion is supported by the Western blot analysis. Definitive kinetic analyses to differentiate between functional loss of activity and loss of structural integrity in all 22 critical mutants will require larger amounts of purified proteins expressed in mammalian cells. Our results do suggest, however, that both EGF5 and EGFG are important in thrombin binding. The presence of bound Ca2+ in EGFG may serve to coordinate the carboxylate groups of Asp-423, Asp-425, and Glu-426 at the protein surface. Such a structure, together with the anionic residues in the C-loop of EGF5, could form an extended negatively charged surface for interaction with the anionbinding exosite of thrombin.
In conclusion, we have utilized alanine-scanning mutagenesis in an E. coli expression system as a rapid screen to evaluate the contribution of individual residues in TME to its cofactor activity. While we were unable to perform detailed kinetic studies using this system, it allowed us to identify 22 critical residues in EGF4, EGF5, and EGF6, including those reported earlier by others (16,17,23,29). Some of these residues are likely to be surface contact residues that interact directly with thrombin and regulate Kd for thrombin and kc,,/ K,,, for protein C of the thrombin-TM complex.

Structure-Function Studies
of Human Thrombomodulin