A substrate-driven allosteric switch that enhances PDI catalytic activity

Abstract

Protein disulfide isomerase (PDI) is an oxidoreductase essential for folding proteins in the endoplasmic reticulum. The domain structure of PDI is a-b-b'-x-a', wherein the thioredoxin-like a and a' domains mediate disulfide bond shuffling and b and b' domains are substrate binding. The b' and a' domains are connected via the x-linker, a 19-amino-acid flexible peptide. Here we identify a class of compounds, termed bepristats, that target the substrate-binding pocket of b'. Bepristats reversibly block substrate binding and inhibit platelet aggregation and thrombus formation in vivo. Ligation of the substrate-binding pocket by bepristats paradoxically enhances catalytic activity of a and a' by displacing the x-linker, which acts as an allosteric switch to augment reductase activity in the catalytic domains. This substrate-driven allosteric switch is also activated by peptides and proteins and is present in other thiol isomerases. Our results demonstrate a mechanism whereby binding of a substrate to thiol isomerases enhances catalytic activity of remote domains.

ARTICLE

Received 31 Mar 2016 | Accepted 14 Jul 2016 | Published 30 Aug 2016

Roelof H. Bekendam1, Pavan K. Bendapudi1, Lin Lin1, Partha P. Nag2,3, Jun Pu2, Daniel R. Kennedy4, Alexandra Feldenzer4, Joyce Chiu5,6, Kristina M. Cook5, Bruce Furie1, Mingdong Huang1, Philip J. Hogg5,6 & Robert Flaumenhaft1

Protein disulde isomerase (PDI) is an oxidoreductase essential for folding proteins in the endoplasmic reticulum. The domain structure of PDI is abb0xa0, wherein the thioredoxin-like a and a0 domains mediate disulde bond shufing and b and b0 domains are substrate binding. The b0 and a0 domains are connected via the x-linker, a 19-amino-acid exible peptide. Here we identify a class of compounds, termed bepristats, that target the substrate-binding pocket of b0. Bepristats reversibly block substrate binding and inhibit platelet aggregation and thrombus formation in vivo. Ligation of the substrate-binding pocket by bepristats paradoxically enhances catalytic activity of a and a0 by displacing the x-linker, which acts as an allosteric switch to augment reductase activity in the catalytic domains. This substrate-driven allosteric switch is also activated by peptides and proteins and is present in other thiol isomerases. Our results demonstrate a mechanism whereby binding of a substrate to thiol isomerases enhances catalytic activity of remote domains.

DOI: 10.1038/ncomms12579 OPEN

A substrate-driven allosteric switch that enhances PDI catalytic activity

1 Division of Hemostasis and Thrombosis, Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts 02115, USA. 2 Department of Pharmaceutical and Administrative Sciences, The Broad Institute Probe Development Center, Cambridge, Massachusetts 02142, USA. 3 Center for the Science of Therapeutics, Broad Institute, Cambridge, Massachusetts 02142, USA. 4 College of Pharmacy, Western New England University, Springeld, Massachusetts 01119, USA. 5 The Centenary Institute, Newtown, Sydney, New South Wales 2050, Australia. 6 National Health and Medical Research Council Clinical Trials Centre, University of Sydney, Sydney, New South Wales 2050, Australia. Correspondence and requests for materials should be addressed to R.F. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579

Protein disulde isomerase (PDI) is the founding member of a large family of thiol isomerases responsible for catalysing the folding of nascent proteins in the endoplasmic

reticulum (ER). Although highly concentrated in the ER, a small percentage of PDI escapes ER retention and trafcs to secretory granules and plasma membrane, where it participates in modication of surface proteins. The domain structure of PDI is abb0xa0 (refs 1,2). Disulde shufing required for protein folding is accomplished by active site cysteines within a CGHC motif in the thioredoxin-like a and a0 domains. Substrate binding is accomplished by domains b and b0, the latter of which contains a deep hydrophobic binding pocket1,2. The x-linker is a exible 19amino-acid peptide that can adopt at least two conformations. One is a capped conformation in which the x-linker covers the hydrophobic pocket3. PDI can also assume an uncapped conformation in which the x-linker is displaced from the hydrophobic binding site. Although distinct functions have been described for the different domains of PDI, there is cooperativity among them1,4,5. The presence of the b and b0 domains in PDI augments reductase activity of the a and a0 domains5. Conversely, the a and a0 domains are important for binding larger substrates.

In addition to its physiological role in protein folding, PDI has been implicated in a wide variety of pathophysiological processes. PDI expression is upregulated in several cancers6, and PDI expression levels correlate with clinical outcomes7,8. Silencing or inhibition of PDI in animal models of tumour progression suppresses tumour growth and extends survival9,10. PDI has also been shown to participate in neurodegenerative processes11, and blocking PDI is protective in a cell-based model of Huntingtons disease (refs 12,13). Several pathogens subvert extracellular PDI activity to achieve cellular invasion1417. For example, PDI mediates cleavage of disulde bonds in glycoprotein 120 that are required for HIV-1 entry18,19, and its inhibition interferes with the ability of HIV-1 to infect T cells20. Extracellular PDI also serves a critical role in thrombus formation, the underlying pathology in myocardial infarction, stroke, peripheral artery disease and deep vein thrombosis. Inhibition of extracellular PDI blocks injury-induced formation of thrombi in multiple animal models of thrombus formation2125. Platelet-specic knockdown of PDI inhibits thrombus formation, demonstrating a role for platelet-derived PDI in thrombus formation26. Two other members of the thiol isomerase family, ERp5 and ERp57 (which, together with PDI, are termed vascular thiol isomerases) are important for thrombosis2731.

Several novel PDI inhibitors have been identied over the past decade. The majority of these antagonists act at the catalytic cysteine within the CxxC motif, blocking all catalytic activity of PDI, and most of these antagonists act irreversibly6,32. As a class, compounds that interact with the catalytic cysteines of PDI are not selective among thiol isomerases. The substantial homology in the thioredoxin folds of the catalytic domains of thiol isomerases33 complicates efforts to develop compounds that are selective among this large enzyme family. As substrate-binding domains are less homologous and the function of PDI relies on cooperative activity of distinct domains, alternative strategies for blocking PDI that do not involve inhibition of the catalytic cysteines could be effective.

With support from the Molecular Libraries Program, we screened over 348,505 compounds from the Molecule Libraries Small Molecule Repository to identify inhibitors of PDI reductase activity34,35. We characterized two compounds, termed bepristats, based on their unusual activity at the b0 domain. We nd that binding of bepristats to the hydrophobic pocket in the b0 domain of PDI interferes with insulin binding and displaces the x-linker that covers this pocket. Displacement of the x-linker induces a

conformation change, modifying the redox state of the catalytic motif and enhancing PDI catalytic activity as measured by di-eosin-GSSG cleavage. Peptides known to bind PDI, such as mastoparan and somatostatin and protein substrates (for example, cathespin G) also triggered this substrate-driven allosteric switch. These results support a general mechanism wherein ligation of the substrate-binding domain of PDI enhances its catalytic activity.

ResultsCharacterization of novel PDI antagonists. High-throughput screening of 348,505 compounds identied two scaffolds that inhibited PDI reductase activity in the insulin turbidimetric assay. The rst scaffold, represented by CID:23723882/ML359 (refs 34,35) and a synthetically prepared analogue, termed bepristat 1a (CID: 70701262), blocked PDI activity with a half-maximal inhibitory concentration (IC50) of B0.7 mM (Fig. 1a).

A second scaffold characterized by CID:1043221 (ref. 29) and a synthetically prepared analogue bepristat 2a (CID: 71627360) was found to have an IC50 of B1.2 mM (Fig. 1b). Yet despite potent

PDI inhibitory activity in the insulin turbidimetric assay, neither bepristat 1a nor bepristat 2a inhibited in a di-eosin-GSSG-based assay that measures reductase activity at the catalytic cysteines36 (Fig. 1a,b). In fact, rather than inhibiting reductase activity at the catalytic cysteines, bepristat 1a and bepristat 2a both enhanced cleavage of the di-eosin-GSSG probe by PDI (Fig. 1a,b). Neither bepristat 1a nor bepristat 2a affected the uorescence of di-eosin-GSSG in the absence of PDI, nor did these compounds affect di-eosin-GSSG uorescence in the presence of a catalytically inactive PDI (Fig. 1c and Supplementary Fig. 1). MichaelisMenten analysis performed in the absence of bepristats showed an apparent KM of 4,168 nM and an apparent kcat of 1,135 min 1 for cleavage of di-eosin-GSSG by PDI. In the presence of bepristat 1a, the apparent KM was decreased to 1,814 nM and apparent kcat was unaffected, while incubation with bepristat 2a decreased the apparent KM to 1,949 nM and increased the apparent kcat to 1,995 min 1 (Fig. 1d). Bepristats demonstrated more potent inhibitory activity in the insulin turbidimetric assay compared with quercetin-3-rutinoside (rutin), a glycosylated avonoid quercetin shown to block PDI activity and inhibit thrombus formation in vivo22,37; PACMA-31, an irreversible inhibitor of PDI that binds the catalytic cysteines and impairs tumour growth in a murine model of ovarian cancer9; and bacitracin, a dodecapeptide that has been used for decades as the standard PDI inhibitor38 (Fig. 1e). Yet these other inhibitors all blocked PDI-mediated cleavage of di-eosin-GSSG when used at concentrations required to inhibit activity in the insulin turbidimetric assay (Fig. 1e). These results indicate that bepristats block PDI reductase activity via a mechanism that differs from that of previously described PDI inhibitors.

Selectivity and reversibility of bepristats. Developing selective, reversible inhibitors of PDI has been problematic owing to homology among catalytic thioredoxin domains and the ability of thiol-reactive antagonists to interact with multiple targets. We compared bepristats and thiol-reactive PDI inhibitors in the insulin turbidimetric assay to assess their ability to inhibit the insulin reductase activities of ERp5, ERp57 and thioredoxin. We found that bepristats were selective among vascular thiol isomerases, even at concentrations 10-fold higher than their IC50s

(Fig. 2a and Supplementary Fig. 1). In contrast, PACMA-31 demonstrated inhibitory activity against both ERp5 and thioredoxin when used at 10-fold their IC50 (Fig. 2b). Bacitracin showed activity against all thiol isomerases tested. Both PACMA-31 and bacitracin interfered with the reaction of

2 NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579 ARTICLE

a

c d

Bepristat 1a

Bepristat 2a

Et N

Me

% PDI activity

O

20 NS

AGHA

% Maximal

response

(RFU min1 )

[EGSH] nM min1

O

100

100

3,000

15

10

2,000

50

50

N

5

1,000

CI

0 8 7 6 5 4

0

0

0 0 2,500 5,000

HO

Log [Bep1a] (M)

Vehicle

Bep1a

Bep2a

Di-eo-GSSG (nM)

b

e

100

% PDI activity

% Maximal

response

% PDI activity

100

100

EtO2C

HO

Br

100

100

N

Et

Et

HN

50

50

% PDI activity

50

50

% PDI activity

50

N

0 8 7 6 5 4

0 7 6 5 4 3

0 6 5 4 3 2

0

0

7 6 5 4 3

Log [Bep2a] (M)

Log [Q-3-R] (M)

Log [PACMA-31] (M)

Log [bacitracin] (M)

Figure 1 | Novel PDI antagonists paradoxically augment catalytic activity. The effect of (a) bepristat 1a (red) and (b) bepristat 2a (blue) on PDI activity as measured in the insulin turbidimetric assay (dark shades) or the di-eosin-GSSG assay (light shades). Curves represent absorbance values taken from the 25 min time point after insulin aggregation started in the vehicle only sample. (c) Bepristat 1a (Bep1a) and bepristat 2a (Bep2a) fail to augment activity of a PDI mutant in which the CGHC motif is mutated to AGHA. Data are represented as cleavage of di-eosin-GSSG probe in RFU min 1. One-way analysis of variance with Dunnetts post test: NS, non-signicant. (d) The effect of bepristat 1a (red) and bepristat 2a (blue) on KM and kcat was evaluated in the di-eosin-GSSG assay using increasing concentrations of di-eosin-GSSG. MichaelisMenten kinetics were calculated using Graphpad Prism 5.0 and compared with vehicle only (black). (e) Effect of quercetin-3-rutinoside (Q-3-R), PACMA-31 or bacitracin on PDI activity as measured by the insulin turbidimetric assay (black) or the di-eosin-GSSG assay (grey). Values for the insulin turbidimetric assay represent per cent of PDI activity compared with a control exposed to vehicle alone, means.e.m. (n 36). Values for the di-eosin-GSSG assay represent the rate of di-eosin-GSSG uorescence generation

measured for 20 mins.e.m. (n 3). RFU, relative uorescence units.

maleimide-polyethylene glycol-2-biotin (MPB) with the catalytic cysteines of ERp5, ERp57 and thioredoxin (Fig. 2c and Supplementary Figs 2 and 3), demonstrating lack of selectivity by virtue of non-selective interactions with catalytic cysteines on thiol isomerases. When the assay is performed in the presence of N-ethylmaleimide, an irreversible free thiol blocker, no MPB binding to PDI is observed, indicating absence of non-specic MPB labelling of PDI. None of the PDI inhibitors interfered with the MPB reaction with bovine serum albumin (BSA), demonstrating that the inhibitors are not non-selectively reacting with MPB or with any free cysteine.

Inhibition of PDI by either blocking antibodies or small-molecule inhibitors interferes with agonist-induced platelet activation21,22. Platelet-specic knockdown of PDI also impairs platelet activation26. We determined whether blocking PDI using bepristats inhibits platelet activation. Platelets were incubated with either bepristat 1b (a bepristat 1 analogue with enhanced esterase resistance necessary to overcome intrinsic platelet esterase activity), bepristat 2a or PACMA-31, and the response to the PAR1 peptide agonist SFLLRN was evaluated by light transmission aggregometry. Bepristat 1b, bepristat 2a and PACMA-31 all inhibited platelet aggregation (Fig. 2d). We next evaluated whether or not the bepristats functioned as global inhibitors of platelet activation. Neither bepristat signicantly altered expression of CD62P on stimulated platelets, indicating that platelet activation per se is not blocked by the bepristats in this assay (Supplementary Fig. 4; Supplementary Methods). Rather, bepristats appear to block aggregation by interfering with functions downstream of platelet activation. To evaluate reversibility of inhibition using the platelet aggregation assay, platelets were incubated with PDI antagonists for 30 min, washed and then stimulated with SFLLRN. Inhibition of platelet aggregation by bepristat 1b and bepristat 2a was restored following washing. In contrast, platelet aggregation by PACMA-31 was irreversibly inhibited under these conditions (Fig. 2). To conrm that bepristats are reversible inhibitors of PDI, we

evaluated reversibility in the insulin turbidimetric assay. These studies demonstrated that the inhibitory effect of bepristats was readily reversed by dilution to a subinhibitory concentration, while that of PACMA-31 was largely preserved (Supplementary Fig. 5).

Bepristats inhibit thrombus formation. Inhibition of PDI using anti-PDI antibodies or by small molecules such as bacitracin or quercetin-3-rutinoside inhibits thrombus formation in vivo2125. Similarly, platelet-selective deletion of PDI interferes with thrombus formation in mouse arterioles following vascular injury26. We therefore evaluated the effect of bepristats on thrombus formation in cremaster arterioles following laser-induced vascular injury (Fig. 3a and Supplementary Movie 1). Infusion of 15 mg kg 1 of bepristat 1a resulted in a79.7% reduction in platelet accumulation at sites of vascular injury compared with mice infused with vehicle alone (P 0.02;

Fig. 3b and Supplementary Movie 2). Bepristat 2a infusion inhibited platelet accumulation by 85.1% at sites of laser-induced injury compared with vehicle controls (P 0.02; Fig. 3c and

Supplementary Movie 3). These results demonstrate that bepristats are tolerated in vivo and potently inhibit thrombus formation.

Bepristats associate with the b0 domain. To determine the mechanism by which bepristat 1a and bepristat 2a modulate PDI activity, we tested the compounds against PDI fragments containing the a or a0 domains using the insulin turbidimetric assay. These fragments included the a domain, a0 domain, ab domains, abb0 domains and b0xa0 domains (Fig. 4a and Supplementary Fig. 6). Although the isolated domains had diminished insulin reductase activity compared with full-length PDI, their activity could be quantied and the effects of antagonists on their activity tested. Neither bepristat 1a nor bepristat 2a had activity against the isolated a, a0 or ab domains

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579

a b

PDI ERp5 ERp57 TRX

100

100

100

100

% Activity

50

% Activity

% Activity

PDI ERp5 ERp57 TRX

50

50

% Activity

50

0 7 6 5 4

0 7 6 5 4

0 6 5 4 3

0 5 4 3 2

Log [Bep1a] (M)

Log [Bep2a] (M)

Log [PACMA-31] (M)

Log [bacitracin] (M)

c d

kDa

68 50 48 70

68 12 10

70 68

Vehicle NEM Q-3-R Bep1a Bep2a PACMA Bacitracin

PDI

ERp5

ERp57

TRX

BSA

Bepristat 1b

Bepristat 2a

PACMA-31

10 0

10

70

% Aggregation

20

30

40

50

60

70

% Maximal aggregation

80

60

40

20

***

***

***

***

0

Vehicle

Bep1b

Bep1b rev

Vehicle

Bep2a

Bep2a rev

Vehicle

PACMA-31

PACMA-31 rev

Figure 2 | Selectivity and reversibility of bepristats. The ability of (a) bepristat 1a (Bep1a) and bepristat 2a (Bep2a), or (b) PACMA-31 and bacitracin to inhibit the reductase activity of PDI (black), ERp5 (red), ERp57 (blue) and thioredoxin (TRX) (green) as measured in the insulin turbidimetric assay was evaluated. Values represent the per cent activity compared with samples exposed to vehicle alones.e.m. (n 23). (c) PDI, ERp5, ERp57, ERp72, TRX or

BSA were incubated with vehicle, 300 mM N-ethylmaleimide (NEM), 150 mM quercetin-3-rutinoside (Q-3-R), 150 mM Bep1a, 150 mM Bep2a, 150 mM PACMA-31 or 150 mM bacitracin before exposure to MPB. Samples were then analysed by SDSpolyacrylamide gel electrophoresis, and MPB was detected by Cy5-labelled streptavidin. (d) Washed human platelets (2 108 platelets per ml) were incubated with vehicle and stimulated with 3 mM SFLLRN (black

tracings); incubated with either 30 mM Bep1b, 30 mM Bep2a or 30 mM PACMA-31, and then stimulated with 3 mM SFLLRN (dark tracings); or incubated with 30 mM Bep1b rev, 30 mM Bep2a rev or 30 mM PACMA-31, washed and subsequently stimulated with 3 mM SFLLRN (light tracings). Aggregation was monitored by light transmission aggregometry. Bar graphs represent the average maximal per cent aggregations.e.m. (n 37). One-way analysis of

variance with Dunnetts post test: ***Po0.001.

(Fig. 4a). In contrast, they both blocked activity of the abb0 and b0xa0 domains. PACMA-31 inhibited reductase activity of all PDI fragments in the insulin turbidimetric assay (Fig. 4a). These results demonstrate that bepristat 1a and bepristat 2a inhibit PDI reductase activity in the insulin turbidimetric assay by binding outside the catalytic motif, at b0.

The C-terminal end of the b0 domain is connected to an x-linker that covers a deep hydrophobic pocket in b0 and is thought to mediate the movement of the a0 domain relative to the rest of the protein39. In a b0x fragment in which isoleucine 272 is mutated to alanine, the x-linker is constitutively associated with the hydrophobic patch on the b0 domain40. 1-anilinonaphthalene-8-sulfonic acid (ANS) uorescence was used to evaluate binding to hydrophobic regions on PDI in wild-type and mutant constructs. Binding of ANS to hydrophobic regions results in a marked increase in uorescence when evaluated at lex 370 nm (Fig. 4b).

ANS uorescence was prominent on incubation with the isolated b0x domain and weak on incubation with isolated a, a0 or b domains (Fig. 4b). ANS uorescence was not observed on

incubation with I272A mutant (Supplementary Fig. 7), indicating that obstruction of binding pocket by the x-linker prevented ANS binding. Both bepristat 1a and bepristat 2a also interfered with the increase in ANS uorescence observed on incubation with PDI (Fig. 4c). In contrast, incubation with PACMA-31 failed to block ANS uorescence (Fig. 4c). Similar results were obtained when binding of ANS to the isolated b0x domain was evaluated (Fig. 4c).

The x-linker is critical for augmenting PDI catalytic activity. Bepristats bind the b0 domain and enhance catalytic activity at the a and a0 domains, raising the question of how binding of a small molecule targeted to one domain modies enzymatic activity at remote domains. To determine which domains contribute to the ability of bepristats to augment activity at the catalytic cysteines of PDI, the effect of bepristats on di-eosin-GSSG cleavage was tested using full-length PDI, abb0 and b0xa0. Bepristats signicantly augmented the catalytic activity of the b0xa0 fragment (Fig. 5a). In contrast, bepristats failed to augment activity of the abb0 fragment, which is missing the x-linker.

4 NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579 ARTICLE

a b

Vehicle Bepristat 2a

Bepristat 1a

2.5107

2.0107

1.5107

1.0107

5.0106

2.5107

2.0107

1.5107

1.0107

5.0106

0 0 50 150

100 Time (s)

Fluorescence (RFU) Fluorescence (RFU)

Vehicle

Bepristat 2a

0 s

30 s

60 s

90 s

120 s

180 s

Bepristat 1a

200

c

Vehicle

0 0 50 150

100 Time (s)

200

Figure 3 | Bepristats inhibit thrombus formation following vascular injury. (a) Platelet-specic anti-GPIbb antibodies conjugated to Dylight 649 (0.1 mg per g body weight) were infused into mice. Mice were subsequently infused with either bepristat 1a (15 mg per kg body weight) or bepristat 2a (15 mg per kg body weight) as indicated. Thrombi were induced by laser injury of cremaster arterioles before (n 35) and after (n 35) infusion of bepristat 1a or

before (n 42) and after (n 28) infusion of bepristat 2a. Thrombus formation was visualized by video microscopy for 180 s after injury. Fluorescence

intensity scale bars, vehicle8887,884; bepristat 1a466905; bepristat 2a497986. Representative binarized images of platelets at the injury site before (vehicle), after bepristat 1a infusion (bepristat 1a) or after bepristat 2a infusion (bepristat 2a) are shown. Median integrated platelet-uorescence intensity at the injury site in mice before (dark red) and after (light red) infusion of (b) bepristat 1a or (c) bepristat 2a infusion is plotted over time. RFU, relative uorescence units. Scale bars, 20 mm.

The observation that bepristats target the same hydrophobic pocket on the b0 domain that the x-linker associates with suggested that binding of bepristats results in displacement of the x-linker. To evaluate this possibility, we tested the protease sensitivity of abb0x in the presence and absence of bepristats.

Proteinase K cleaves PDI from the C-terminal end41. Cleavage of abb0x by proteinase K occurred more rapidly in the presence of bepristat 1a and bepristat 2a than in their absence (Fig. 5b and Supplementary Fig. 8). The remaining fragment of proteinase K-treated abb0x had a molecular weight similar to abb0, consistent with the premise that proteinase K cleaves the x-linker. Bepristats did not interfere with proteinase K activity as evidenced by the fact that proteinase K cleaved ERp5 equally well in the presence or absence of bepristats. The effect of bepristats on movement of the x-linker was also evaluated using intrinsic uorescence measurements. The x-linker includes a tryptophan residue (Trp-347) that associates with the hydrophobic binding pocket on the b0 domain, resulting in an increase in intensity and a blue shift41. Incubation with bepristats resulted in a loss of intensity of intrinsic tryptophan uorescence and a red shift (Fig. 5c), indicating that bepristats elicit movement of the x-linker on incubation with b0x. These results support a model whereby ligation of bepristats at the hydrophobic binding pocket results in displacement of the x-linker.

Small-angle X-ray scattering (SAXS) was used to determine whether the local effects of bepristats on interactions between the

b0 domain and the x-linker had global consequences on PDI conformation. PDI is a exible protein that exhibits dynamic behaviour in aqueous solution. We used SAXS to measure the gyration radius (Rg) of PDI in aqueous solution. An overall molecular envelope of PDI was derived from these measurements. We found the gyration radius for full-length oxidized PDI to be40.0 , whereas the gyration radius for reduced PDI was 34.0 . PDI complexed with bepristat 1b had a gyration radius of 35.3 and PDI complexed with bepristat 2b showed a gyration radius of35.9 (Fig. 5d). These data suggest that bepristats constrain the dynamic behaviour of PDI. In the presence of bepristats, PDI adopts a more compact conformation that closely approximates the overall envelope of reduced PDI.

By binding the b0 domain and eliciting a change in the global conformation in PDI, bepristats could modify disulde bond formation at the CGHC motifs. Differential cysteine alkylation followed by mass spectroscopy under varying reduced glutathione (glutathione sulfhydryl, GSH):oxidized glutathione (glutathione disulde, GSSG) ratios was used to quantify unpaired thiols and disulde bonds within the CGHC motifs of the a and a0 domains. This technique (Supplementary Fig. 9) uses 12C-IPA to label unpaired thiols. Disulde bonds are then reduced using dithiothreitol (DTT) and the resulting free thiols are labelled using 13C-IPA. Differential cysteine alkylation showed that in a relatively oxidizing environment (high GSSG to GSH ratio), the fraction of reduced PDI was increased following incubation with

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579

a

PDI

a a ab abb bxa

100

Insulin

aggregation in %

* *

Vehicle Bepristat 1a Bepristat 2a PACMA-31

50

***

***

***

***

***

*** **

**

**

***

0

b c

PDI

bx

500

800 VehicleBep1a Bep2a PACMA-31

Wavelength (nm)

500 VehicleBep1a Bep2a PACMA-31

Wavelength (nm)

400

x

600

400

Vehicle

300

300

RFU

RFU

400

RFU

200

200

100

200

100

0 400 500 600

0 400 500 600

0 400 500 600

Wavelength (nm)

Figure 4 | Bepristats associate with the substrate-binding pocket of the b0 domain. (a) Evaluation of domain targets used either full-length PDI or PDI domains including a (residues 1117), a0 (residues 342462), ab (residues 1-218), abb0 (residues 1331) or b0xa0 (residues 219462) as indicated. Full-length

PDI or PDI domains were incubated with vehicle (grey), 30 mM bepristat 1a (red), 30 mM bepristat 2a (blue) or 100 mM PACMA-31 (green) for 30 min. Proteins were then assayed for activity in the insulin turbidimetric assay. Values represent per cent vehicle controls.e.m. (n 4). One-way analysis of

variance with Dunnetts post test: *Po0.05; **Po0.01; ***Po0.001. (b) Fluorescence monitored at lex 370 nm following incubation of 50 mM ANS with isolated a (gold), a0 (green), b (orange), b0x (purple) or vehicle alone (black). (c) PDI or b0x ,as indicated, was incubated in the presence of either vehicle (black), 100 mM bepristat 1a (red), 100 mM bepristat 2a (blue) or 100 mM PACMA-31 (green). Samples were then exposed to 50 mM ANS and uorescence monitored following excitation at lex 370 nm. RFU, relative uorescence units.

bepristats, as indicated by the baseline offset (Fig. 5e,f). This difference between samples exposed to bepristats and control samples is lost under reducing conditions (high GSH to GSSG ratio). Calculation of the redox potential of the two active sites demonstrated no difference between bepristat-exposed and control samples (Supplementary Table 1), since the effect of bepristats is overcome under more reducing conditions. However, these data show that the conformational change induced by bepristats impair active site disulde bond formation under equilibrium conditions.

Substrates stimulate thiol isomerase catalytic activity. The bepristats affect PDI activity by a previously unrecognized mechanism involving engagement of a hydrophobic pocket on the b0 domain with displacement of the x-linker resulting in a conformation change and enhanced reductase activity in the di-eosin-GSSG assay. Peptides such as mastoparan42 have previously been shown to associate with the hydrophobic pocket on b0. Incubation of PDI with mastoparan enhanced the ability of PDI to cleave di-eosin-GSSG in a dose-dependent manner (Fig. 6a), even though mastoparan inhibits PDI activity in an RNase refolding assay42. Somatostatin, a 14-amino-acid peptide hormone, also binds PDI and stimulates cleavage of di-eosin-GSSG (Fig. 6b). In addition, a protein substrate of PDI, cathepsin G, elicits enhanced reductase activity in the di-eosin-GSSG assay (Fig. 6c). Like the bepristats, mastoparan, somatostatin and cathepsin G all decreased the apparent KM in

the di-eosin-GSSG assay of PDI reductase activity (Fig. 6d). None of them altered the apparent kcat.

Peptides and protein substrates associate with substrate-binding domains of other thiol isomerases43. To evaluate whether substrate-driven augmentation of catalytic activity is observed in other thiol isomerases, we tested the ability of mastoparan, somatostatin and cathepsin G to enhance cleavage of di-eosin-GSSG by ERp72, ERp57 and ERp5. All three substrates augmented the catalytic activity of ERp72 (Fig. 6ac). In contrast, neither mastoparan, somatostatin nor cathepsin G was able to augment the catalytic activity of ERp57 or ERp5.

DiscussionPDI family enzymes are multifunctional thiol isomerases that share a common ancestral thioredoxin-like domain. Gene duplication resulted in the formation of multidomain enzymes in eukaryotes in which some thioredoxin-like domains retained catalytic function while others evolved into substrate-binding domains33. While the catalytic domains within PDI are capable of autonomous activity, they demonstrate cooperative and enhanced behaviour within the context of full-length PDI1,4,5. The mechanism of this cooperativity, however, is poorly understood. We have identied an allosteric switch that modulates the catalytic activity of PDI. Our results indicate that binding of small molecules, peptides or proteins to the primary substrate-binding pocket on the b0 domain displaces the x-linker from the hydrophobic pocket causing a conformational change that enhances catalytic activity of remote domains.

This mechanism was uncovered using newly identied small molecules, termed bepristats, that target the b0 domain.

The catalytic thioredoxin-like domains of PDI family proteins

6 NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579 ARTICLE

a b

Proteinase K

% Di-eosin GSSG cleavage

250

** ** **

**

200

Vehicle

Vehicle

Bepristat 1a

NS Bepristat 1a

Bepristat 2a

1 m 2 m 3 m 5 m 10 m 20 m kDa 40 38

40 38 40 38

Bepristat 2a

150

100

50

PDI

bxa abb

c

d

Bepristat 1b

0 0 25 50 75 100 125

Bepristat 2b

0 0 25 50 75 100 125

1,500

1,000

500

310

0 320

330

340

350

360

370

380

390

400

10

10

Fluorescence

P (r)

5

P (r)

5

Wavelength (nm)

Rg Rg

e f

Vehicle

Bepristat 1a

Bepristat 2b

1.0

1.0

1.0

0.4

0.8

0.8

Reduced fraction

Reduced fraction

Reduced fraction

Reduced fraction

Vehicle Bepristat 1a Bepristat 2b

**

* *

0.8

0.3

0.6

0.6

0.6

0.2

0.4

0.4

0.4

0.2

0.2

0.2

0.1

0.0 105 104 103 102 101 100 105 104 103 102 101 100 105 104 103 102 101 100

0.0

0.0

0.0

a a

[GSH]2/[GSSG], M [GSH]2/[GSSG], M

[GSH]2/[GSSG], M

Figure 5 | The x-linker is critical for bepristat-driven augmentation of PDI catalytic activity. (a) Full-length PDI, b0xa0, and abb0 were incubated with either vehicle (white) 30 mM bepristat 1a (red) or 30 mM bepristat 2a (blue) as indicated. PDI and PDI fragments were subsequently evaluated for activity in the di-eosin-GSSG assay. Values represent per cent vehicle controls.e.m. (n 34). One-way analysis of variance with Dunnetts post test: *Po0.05;

**Po0.01; NS, non-signicant. (b) Detection of abb0x by silver staining following incubation with proteinase K for the indicated times in absence and presence of bepristat 1a and bepristat 2a. (c) Intrinsic uorescence emission spectra of PDI in the absence (black) or presence of bepristat 1a (red) or bepristat 2a (blue). (d) SAXS proles of PDI incubated in the presence of vehicle (black), bepristat 1b (red) or bepristat 2b (blue). (e) Plots of the ratio of reduced to oxidized PDI as a function of the ratio of GSH to GSSG in the absence (black) or presence of either 50 mM bepristat 1a (red) or 50 mM bepristat 2b (blue). The lines represent the best nonlinear least squares t of the data. The calculated equilibrium constants were used to determine the standard redox potentials (Supplementary Fig. 9). Data points are the mean values from analysis of 24 peptides encompassing the active site cysteine residues. (f) Fraction of the active site dithiols/disuldes in the reduced state under oxidizing conditions in the absence (black) or presence of either 50 mM bepristat 1a (red) or 50 mM bepristat 2b (blue). The baseline offsets in the plots shown in e were a tted parameter in the nonlinear least squares analysis. The error bars represent 1 s.e.

demonstrate a higher degree of homology than do the substrate-binding domains, which have evolved to associate with different substrate classes. As a result, compounds targeting catalytic domains generally lack selectivity among thiol isomerases. RL90, a monoclonal antibody that targets PDI, has been shown to cross-react with other closely related thiol isomerases, such as ERp57 (ref. 31). Similarly, we nd that PACMA-31 is not entirely selective for PDI among thiol isomerases (Fig. 2). Although bepristats target the substrate-binding domain of PDI and elicit augmentation of di-eosin-GSSG cleavage, they demonstrate no such activity when tested against other thiol isomerases (Supplementary Fig. 1) and do not inhibit insulin aggregation mediated by ERp5, ERp57 or thioredoxin (Fig. 2). Bepristats were found to be active in o1% of several hundred assays tested, as

described in the PubChem database, attesting to their selectivity against a broader range of proteins. A second advantage of targeting the b0 domain is reversibility. Compounds that target the catalytic domains tend to bind irreversibly via catalytic cysteines. By targeting the substrate-binding site, bepristats act as reversible inhibitors of PDI (Fig. 2 and Supplementary Fig. 5). Bepristats block platelet activation in vitro and impair platelet accumulation at sites of vascular injury in an in vivo model of thrombus formation (Fig. 3). These in vivo studies provide proof of principle for targeting the hydrophobic binding site of the b0 domain of PDI in a clinical setting.

Bepristats are also useful in evaluating the role of the x-linker in modulating PDI activity. Protease digestion experiments and studies using the intrinsic uorescence of Trp-347 to monitor

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579

a b c

300

Log [mastoparan] (M)

300

Log [somatostatin] (M)

300

Log [cathepsin G] (M)

% Thiol isomerase activity

% Thiol isomerase activity

% Thiol isomerase activity

200

200

200

100

100

100

0 8 7 6 5 4

0 8 7 6 5 4

0 10 9 8 7 6

d e

Unligated

Ligated

15,000

Vehicle Mastoparan Somatostatin Cathepsin G

a

b b

[EGSH] nM min1

a a

b

a

10,000

b

5,000

Capped Uncapped

x-linker

0 0 2,500 5,000

Disulfide bond

Substrate

Di-eo-GSSG (nM)

Unpaired thiols

Hydrophobic binding pocket

Figure 6 | Protein substrates activate the allosteric switch mechanism in PDI and ERp72. PDI (black), ERp5 (red), ERp57 (blue) or ERp72 (green) were incubated in the absence or presence of (a) mastoparan, (b) somatostatin or (c) cathepsin G at the indicated concentrations, and its effect on reductase activity monitored in the di-eosin-GSSG assay. (d) The effect of mastoparan, somatostatin and cathespin G on KM and kcat was evaluated in the di-eosin-

GSSG assay using increasing concentrations of di-eosin-GSSG. (e) Model of a substrate-driven allosteric switch that activates thiol isomerase catalytic activity. In the unligated state, the hydrophobic binding pocket within the b0 domain of PDI is capped and disulde bond formation within the CGHC motif is favoured. Binding of substrate to the hydrophobic binding pocket results in displacement of the x-linker and the a0 domain, resulting in a more constrained conformation and favours unpaired cysteines within the CGHC motif. PDI reductase activity is enhanced in the ligated state.

movement of the x-linker conrmed displacement with bepristat exposure (Fig. 5). Displacement of the x-linker by bepristats is associated with a more constrained conformation, as demonstrated by SAXS. These studies indicate that binding of bepristats results in displacement of the x-linker and induces a conformational change in PDI. The net consequence appears to be a smaller binding pocket that cannot accommodate large substrates, and an a0-domain conformation that increases thiol-reductase activity for those substrates that can enter the smaller substrate-binding pocket.

While bepristats served as a convenient tool to evaluate this allosteric switch mechanism, peptides known to displace the x-linker demonstrated similar activity. Mastoparan and somatostatin both induced substantial augmentation of PDI-mediated dieosin-GSSG cleavage (Fig. 6). Nuclear magnetic resonance spectroscopy showed that these peptides associate with the hydrophobic binding site on b0 that consists primarily of residues from a-helices 1 and 3, as well as from the core b-sheet42,44.

Chemical shifts that occur on binding of either mastoparan or somatostatin have been mapped to hydrophobic residues adjacent to or within the substrate-binding pocket42. In the capped conformation of PDI, the x-linker binds this site. Peptide ligands such as mastoparan and somatostatin compete with and displace the x-linker, promoting an uncapped conformation44. The full range of substrates capable of augmenting PDI catalytic activity by associating with this binding pocket remains to be determined.

The observation that interactions at the hydrophobic binding pocket can inuence the reductase activity at the CGHC motif

(Fig. 6d) demonstrates that PDI conformation is controlled in two distinct directions. In one direction, redox environment controls PDI conformation in a previously described mechanism that is initiated at the catalytic domains41,45. Reduction of the catalytic cystines in the CGHC motif is thought to trigger rotation of Trp-396, enabling it to interact with Arg-300 on the b0 domain, initiating a series of interactions at the a0b0 interface that positions the a0 domain over the hydrophobic binding site on the b0 domain41,45. This constrained conformation is thought to be a means to limit substrate binding under reducing conditions41. Our results demonstrate a second mechanism controlling PDI conformation. This second mechanism is driven by substrate binding rather than by the redox environment. Binding of a substrate to the hydrophobic binding pocket results in displacement of the x-linker, an increase in active site cysteine thiolates and enhanced reduction of the substrate (Fig. 6e). This mechanism of substrate-driven augmentation of catalytic activity appears to be shared by some thiol isomerases, but not by others. Human ERp72 demonstrates augmentation of catalytic activity on binding mastoparan, somatostatin or cathepsin G while ERp57 and ERp5 do not (Fig. 6). Further studies will be required to identify the extent that a substrate-driven allosteric switch mechanism is used among thiol isomerases.

Methods

Protein purication. Recombinant double-tagged (streptavidin-binding protein-tagged and FLAG-tagged) full-length PDI, ERp57, recombinant His-tagged full-length ERp5, ERp72 and PDI domain fragments were cloned into a pET-15b

8 NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579 ARTICLE

vector at the NdeI and BamHI sites and transformed into Escherichia coli Origami B (DE3) cells (EMD Chemicals). The recombinant proteins were expressed and isolated by afnity chromatography with complete His-Tag purication resin (Roche Applied Science) or Pierce High Capacity Streptavidin Agarose beads and puried on a Superdex 200 (GE Healthcare). Human thioredoxin was purchased from Abcam.

Small-angle x-ray scattering. Further purication of full-length PDI was achieved by gel ltration. PDI (1.5, 3.0 and 4.5 mg ml 1) was dialysed against20 mM Tris, 150 mM NaCl, 5% glycerol (pH: 8.0) containing 0.5 mM bepristat 1b or 2b or dimethylsulfoxide control at 4 C overnight. Evaluation by SAXS was performed on the SIBYLS beamline in the Advanced Light Source using a high-throughput data collection method.

Redox potential determination. The redox potentials of the a (Cys53-Cys56) and a0 (Cys397-Cys400) active-site dithiols/disuldes of human wild-type PDI in the absence or presence of bepristat 1a or bepristat 2b were determined by differential cysteine alkylation and mass spectrometry. Recombinant PDI (5 mM) was incubated in the absence or presence of bepristat 1a or bepristat 2b (50 mM) in argon-ushed phosphate-buffered saline containing 0.1 mM EDTA, 0.2 mM oxidized glutathione (GSSG, Sigma) and indicated concentrations of reduced glutathione (GSH, Sigma) for 18 h at room temperature. Microcentrifuge tubes were ushed with argon before sealing to prevent oxidation by ambient air during the incubation period. Unpaired cysteine thiols in PDI and mutants were alkylated with 5 mM 2-iodo-N-phenylacetamide (12C-IPA, Cambridge Isotopes) for 1 h at room temperature. The proteins were resolved on SDSPAGE and stained with SYPRO Ruby (Invitrogen). The PDI bands were excised, destained, dried, incubated with 100 mM DTT and washed. The fully reduced proteins were alkylated with 5 mM 2-iodo-N-phenylacetamide where all six carbon atoms of the phenyl ring have a mass of 13 (13C-IPA) (Cambridge Isotopes). The gel slices were washed and dried before digestion of proteins with 12 ng ml 1 of chymotrypsin (Roche) in 25 mM NH4CO2 overnight at 25 C. Peptides were eluted from the slices with 5% formic acid and 50% acetonitrile. Liquid chromatography, mass spectrometry and data analysis were performed as described46.

The fraction of reduced active-site disulde bond was measured from the relative ion abundance of peptides containing 12C-IPA and 13C-IPA. To calculate ion abundance of peptides, extracted ion chromatograms were generated using the XCalibur Qual Browser software (v2.1.0; Thermo Scientic). The area was calculated using the automated peak detection function built into the software. The ratio of 12C-IPA and 13C-IPA alkylation represents the fraction of the cysteine in the population that is in the reduced state. The results were expressed as the ratio of reduced to oxidized protein and tted to equation 1:

R

B 1 B

n GSH GSSG o

Keq

GSH

GSSG

1

where R is the fraction of reduced protein at equilibrium, B is the baseline fraction of the cysteine in the population that is in the reduced state and Keq is the

equilibrium constant. The standard redox potential (E00) of the PDI active-site disuldes were calculated using the Nernst equation (equation (2)):

E0 E0GSSG

RT

2F ln Keq 2

using a value of 240 mV for the standard redox potential of the GSSG disulde

bond.

In vivo experiment. All protocols involving the use of animals were in compliance with the National Institutes of Healths Guide for the Care and the Beth Israel Deaconess Medical Center Institutional Animal Care and Use. Intravital video microscopy of the cremaster muscle microcirculation was performed, and digital images were captured with an Orca Flash 4.0v2 sCMOS camera (Hamamatsu Photonics K.K., Shizuoka Pref., Japan)22,47. Representative images are presented, but the median curves include the full data. The kinetics of platelet thrombus formation were analysed by determining median uorescence values over time in B3040 thrombi in three mice (all 8-week-old males, C57/BL6J background).

Two hours before surgery, 100 mg kg 1 of the suicide P450 inhibitor 1-aminobenzotriazole (ABT) was administered intraperitoneally to each mouse.

The cremaster muscle was then surgically exposed. Before arteriolar wall injury, DyLight-labelled antibodies were infused intravenously together with bepristats or vehicle control. Injury to a cremaster arteriolar vessel (3050 mm diameter) was induced with a MicroPoint laser system, Andor Technology, Ltd., Belfast Ireland) focused through the microscope objective, parfocal with the focal plane and tuned to 440 nm through a dye cell containing 5 mM coumarin in methanol. Data were captured digitally from two uorescence channels, 488/520 nm and 647/670 nm, as well as a brighteld channel. Data acquisition was initiated both before and following an ablation laser pulse for each injury. The microscope system was controlled and images were collected and analysed using SlideBook 6.0 (Intelligent Imaging Innovations, Denver, CO).

Before induction of the thrombus but after injection of the uorescently labelled antibody, B30 time points were recorded. Subsequently, a thrombus is initiated and recorded for B170 s. Post capture, an upstream region is dened near the site of each thrombus. The maximum pixel intensity in this region is extracted for each time point. The mean of maximal intensity values in the upstream region for each frame is calculated and used as the threshold to dene those pixels containing signal. Extracting the values of these pixels and summing them, we obtained the uncorrected integrated intensity for each time point. The area of this region(in pixels) is also reported. Using this information, the actual integrated intensity for each frame is calculated according to the following formula:actual integrated intensity (sum of the uncorrected integrated

intensity) (mean of the maxima from the upstream region area of the

uncorrected integrated intensity).

Tryptophan uorescence. Intrinsic uorescence spectra were performed in a reaction volume of 50 ml with 5 mM of PDI in 50 mM Tris-HCl buffer containing 150 mM NaCl (pH 7.6). Emission spectra were recorded at 310400 nm with excitation at 290 nm. Bepristats were used at a concentration of 50 mM.

MPB binding. MPB-binding experiments were performed in a reaction volume of37.5 ml with 5 mM of thiol isomerase or bovine serum albumin (BSA) in Tris-buffered saline (TBS), in the presence and absence of 1 mM of N-ethylmaleimide or bacitracin, and 150 mM of the mentioned other inhibitors. The reaction mixture was incubated at 37 C for 1 h. Subsequently, the reaction mixtures were incubated with 25 mM of MPB. The labelling was performed for 20 min at 25 C. A total of 12.5 ml of 4 Laemmli Sample Buffer with 5% b-mercaptoethanol was added to each of the

samples, followed by heating at 95 C for 10 min. From each sample, 10 ml was loaded on a 12% SDSPAGE gel, followed by transfer on a nitrocellulose membrane. After blocking the membrane with TBS-T containing 5% BSA for 1 h, the membrane was incubated with a 1:2,000 dilution of Cy5-tagged streptavidin for an hour in the dark. Detection of the immunoblots was performed using ImageQuant LAS 4000, and analysis was performed using ImageQuant TL software.

1-Anilinonaphthalene-8-sulfonic acid uorescence. The binding of 1-anilinonaphthalene-8-sulfonic acid (ANS) to full-length PDI and the different domains of PDI was assessed by incubating 5 mM of protein in the presence or absence of 100 mM of the indicated inhibitors in 175 ml of TBS at 37 C for 1 h. Subsequently, 50 mM of ANS was added and the mixture was incubated in the dark at 25 C for 20 min. Fluorescence spectrum (Ex: 370 nm, Em: 400700 nm) was measured in a 384-well plate. The experiment was performed in triplicate.

Platelet aggregation. Platelet-rich plasma was obtained from healthy volunteers. Informed consent was obtained from all volunteers. Platelets were isolated by centrifugation at 2,000g and resuspended in Hepes-Tyrode buffer (134 mM NaCl,0.34 mM sodium phosphate, 2.9 mM KCl, 12 mM sodium bicarbonate, 20 mM HEPES, 1 mM magnesium chloride, 5 mM glucose (pH 7.3)). Washed human platelets (2.5 108 platelets per ml) were incubated with the indicated concentra

tions of bepristats and PACMA-31 at 37 C for 20 min and then exposed to 3 mM PAR-1-activating peptide SFLLRN. Aggregation was measured using a Chrono-Log 680 Aggregation System.

Insulin turbidimetric assay. The thiol isomerase-catalysed reduction of insulin was assayed by measuring the increase in turbidity as detected at an optical density of 650 nm using a Spectramax M3 (Molecular Devices, Sunnyvale, CA). The validation assay consisted of 175 nM of PDI in a solution containing 100 mM potassium phosphate (pH 7.4) containing 0.2 mM bovine insulin, 2 mM EDTA and0.3 mM DTT (all purchased from Sigma Aldrich, St Louis, MO), inhibitors were used at the concentrations indicated. The reaction was performed at 25 C for 1 h and 30 min. For assays of thiol isomerase selectivity,11 nM PDI, ERp57 or thioredoxin, or 33 nM ERp5 were assayed in similar buffer conditions as described above and inhibitors were used at the indicated concentration. The reaction was performed at 37 C for 45 min. For assays of isolated domains studies, 400 nM protein was used, except for a, a0 and ab, in which 800 nM protein was used. The assay was performed in similar buffer conditions as described above. Bepristat 1a, 2a and PACMA-31 were used at a nal concentration of 15 mM in these studies.

For calculation of the IC50, we used absorbance values recorded at 25 min after insulin aggregation is rst detectable in the vehicle control compared with the corresponding value at that time point for specimens containing the study drug.

Reversibility studies were performed by incubating 20 mM PDI with 3 mM of bepristat 1a, 6 mM of bepristat 2a or 300 mM of PACMA-31. After 30 min equilibration, the PDI-inhibitor mixture was diluted 100-fold into the abovementioned assay buffer. The ability to reduce insulin of these mixtures was compared with PDI in the presence or absence of 3 mM bepristat 1a or 0.03 mM bepristat 1a, 6 mM bepristat 2a or 0.06 mM bepristat 2a or 300 mM PACMA-31 or 3 mM PACMA-31.

Di-eosin-GSSG disulde reductase assay. The probe di-eosin glutathione disulde, di-eosin- GSSG, was prepared incubating 100 mM of GSSG with 1 mM of

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579

eosin isothiocyanate in 100 mM potassium phosphate containing 2 mM EDTA (pH 7.4), at 25 C overnight in the dark36. The mixture was passed down a PD-10 column and different fractions were collected. A fold change of uorescence(Ex: 520 nm, Em: 550 nm) was calculated using samples subjected to either vehicle or 20 mM of DTT. Any fraction with a fold change 45 was kept. Using a stock of 10 mM eosin isothiocyanate a standard curve was generated and concentration of samples were determined accordingly. Di-eosin-GSSG cleavage of puried thiol isomerases and PDI domains was monitored in a 96-well uorescence plate format. PDI, AGHA-PDI, ERp5, ERp57, ERp72 and PDI domains were assayed at 50 nM in the presence or absence of the indicated small molecules, peptides or cathepsin G. The assay included 100 mM potassium phosphate (pH 7.4) containing 2 mM EDTA, 5 mM DTT and 150 nM of the di-eosin-GSSG probe. The increase in uorescence was determined for 20 min by excitation at 520 nm and emission at 550 nm in a Synergy Biotek 4. The reduction of 150 nM di-eosin-GSSG by 5 mM

DTT in the presence or absence of mentioned small molecules, peptides or proteins served as a negative control.

MichaelisMenten analysis was assayed using PDI at 20 nM and small molecules, peptides or cathepsin G at concentrations causing maximal augmentation under similar buffer conditions as described above. Eight-point response curves were generated using a range of different di-eosin-GSSG concentrations (505,000 nM). Enzyme kinetics analysis was performed using Graphpad Prism 5.0.

Proteolysis assay. Proteinase K (2 mg ml 1) was incubated with 1.25 mg of the abb0x fragment in 50 mM Tris-HCl containing 5 mM CaCl2 and 10 mM DTT. Reactions were aborted with the addition of 0.5 mM phenylmethanesulfonyl uoride. Subsequently, samples were subjected to Laemmli Sample buffer with 5% b-mercaptoethanol and heated at 95 C for 10 min. Each sample was loaded on a 12% SDSPAGE gel, followed by silver staining using the Pierce Silver Stain Kit (Thermo Scientic).

Data availability. Data supporting the ndings of this study are available within the article (and its Supplementary Information les) and from the corresponding author on reasonable request.

References

1. Hatahet, F. & Ruddock, L. W. Protein disulde isomerase: a critical evaluation of its function in disulde bond formation. Antioxid. Redox Signal. 11, 28072850 (2009).

2. Ellgaard, L. & Ruddock, L. W. The human protein disulphide isomerase family: substrate interactions and functional properties. EMBO Rep. 6, 2832 (2005).

3. Pirneskoski, A. et al. Molecular characterization of the principal substrate binding site of the ubiquitous folding catalyst protein disulde isomerase.J. Biol. Chem. 279, 1037410381 (2004).4. Freedman, R. B., Klappa, P. & Ruddock, L. W. Protein disulde isomerases exploit synergy between catalytic and specic binding domains. EMBO Rep. 3, 136140 (2002).

5. Darby, N. J., Penka, E. & Vincentelli, R. The multi-domain structure of protein disulde isomerase is essential for high catalytic efciency. J. Mol. Biol. 276, 239247 (1998).

6. Xu, S., Sankar, S. & Neamati, N. Protein disulde isomerase: a promising target for cancer therapy. Drug Discov. Today 19, 222240 (2014).

7. Cancer Genome Atlas Research Network. Comprehensive genomic characterization denes human glioblastoma genes and core pathways. Nature 455, 10611068 (2008).

8. Shai, R. et al. Gene expression proling identies molecular subtypes of gliomas. Oncogene 22, 49184923 (2003).

9. Xu, S. et al. Discovery of an orally active small-molecule irreversible inhibitor of protein disulde isomerase for ovarian cancer treatment. Proc. Natl Acad. Sci. USA 109, 1634816353 (2012).

10. Yu, S. J. et al. Enhancement of hexokinase II inhibitor-induced apoptosis in hepatocellular carcinoma cells via augmenting ER stress and anti-angiogenesis by protein disulde isomerase inhibition. J. Bioenerg. Biomembr. 44, 101115 (2012).

11. Uehara, T. et al. S-nitrosylated protein-disulphide isomerase links protein misfolding to neurodegeneration. Nature 441, 513517 (2006).

12. Hoffstrom, B. G. et al. Inhibitors of protein disulde isomerase suppress apoptosis induced by misfolded proteins. Nat. Chem. Biol. 6, 900906 (2010).

13. Kaplan, A. et al. Small molecule-induced oxidation of protein disulde isomerase is neuroprotective. Proc. Natl Acad. Sci. USA 112, E2245E2252 (2015).

14. Khan, M. M. G., Simizu, S., Kawatani, M. & Osada, H. The potential of protein disulde isomerase as a therapeutic drug target. Oncol. Res. 19, 445453 (2011).

15. Diwaker, D., Mishra, K. P., Ganju, L. & Singh, S. B. Protein disulde isomerase mediates dengue virus entry in association with lipid rafts. Viral Immunol. 28, 153160 (2015).

16. Walczak, C. P. & Tsai, B. A PDI family network acts distinctly and coordinately with ERp29 to facilitate polyomavirus infection. J. Virol. 85, 23862396 (2011).

17. Stolf, B. S. et al. Protein disulde isomerase and host-pathogen interaction. ScienticWorldJournal 11, 17491761 (2011).

18. Fenouillet, E., Barbouche, R., Courageot, J. & Miquelis, R. The catalytic activity of protein disulde isomerase is involved in human immunodeciency virus envelope-mediated membrane fusion after CD4 cell binding. J. Infect. Dis. 183, 744752 (2001).

19. Gallina, A. et al. Inhibitors of protein-disulde isomerase prevent cleavage of disulde bonds in receptor-bound glycoprotein 120 and prevent HIV-1 entry. J. Biol. Chem. 277, 5057950588 (2002).

20. Bi, S., Hong, P. W., Lee, B. & Baum, L. G. Galectin-9 binding to cell surface protein disulde isomerase regulates the redox environment to enhance T-cell migration and HIV entry. Proc. Natl Acad. Sci. USA 108, 1065010655 (2011).

21. Cho, J., Furie, B. C., Coughlin, S. R. & Furie, B. A critical role for extracellular protein disulde isomerase during thrombus formation in mice. J. Clin. Invest. 118, 11231131 (2008).

22. Jasuja, R. et al. Protein disulde isomerase inhibitors constitute a new class of antithrombotic agents. J. Clin. Invest. 122, 21042113 (2012).

23. Sharda, A. et al. Defective PDI release from platelets and endothelial cells impairs thrombus formation in Hermansky-Pudlak syndrome. Blood 125, 16331642 (2015).

24. Furie, B. & Flaumenhaft, R. Thiol isomerases in thrombus formation. Circ. Res. 114, 11621173 (2014).

25. Reinhardt, C. et al. Protein disulde isomerase acts as an injury response signal that enhances brin generation via tissue factor activation. J. Clin. Invest. 118, 11101122 (2008).

26. Kim, K. et al. Platelet protein disulde isomerase is required for thrombus formation but not for hemostasis in mice. Blood 122, 10521061 (2013).

27. Jordan, P. A. et al. A role for the thiol isomerase protein ERP5 in platelet function. Blood 105, 15001507 (2005).

28. Passam, F. H. et al. Both platelet- and endothelial cell-derived ERp5 support thrombus formation in a laser-induced mouse model of thrombosis. Blood 125, 22762285 (2015).

29. Holbrook, L. M. et al. The platelet-surface thiol isomerase enzyme ERp57 modulates platelet function. J. Thromb. Haemost. 10, 278288 (2012).

30. Wang, L. et al. Platelet-derived ERp57 mediates platelet incorporation into a growing thrombus by regulation of the aIIbb3 integrin. Blood 122, 36423650 (2013).

31. Wu, Y. et al. The disulde isomerase ERp57 mediates platelet aggregation, hemostasis, and thrombosis. Blood 119, 17371746 (2012).

32. Flaumenhaft, R., Furie, B. & Zwicker, J. I. Therapeutic implications of protein disulde isomerase inhibition in thrombotic disease. Arterioscler. Thromb. Vasc. Biol. 35, 1623 (2015).

33. McArthur, A. G. et al. The evolutionary origins of eukaryotic protein disulde isomerase domains: new evidence from the Amitochondriate protist Giardia lamblia. Mol. Biol. Evol. 18, 14551463 (2001).

34. Khodier, C. et al.Identication of ML359 as a Small Molecule Inhibitor of Protein Disulde Isomerase. http://www-ncbi-nlm-nih-gov.ezp-prod1.hul.harvard.edu/pubmed/24624466

Web End =http://www-ncbi-nlm-nih-gov.ezp-prod1.hul.harvard.edu/ http://www-ncbi-nlm-nih-gov.ezp-prod1.hul.harvard.edu/pubmed/24624466

Web End =pubmed/24624466 .

35. Dockendorff, C. et al. Discovery of 1,3-diaminobenzenes as selective inhibitors of platelet activation at the PAR1 receptor. ACS Med. Chem. Lett. 3, 232237 (2012).

36. Raturi, A. & Mutus, B. Characterization of redox state and reductase activity of protein disulde isomerase under different redox environments using a sensitive uorescent assay. Free Radic. Biol. Med. 43, 6270 (2007).

37. Lin, L. et al. Quercetin-3-rutinoside inhibits protein disulde isomerase by binding to its bx domain. J. Biol. Chem. 290, 2354323552 (2015).38. Mandel, R., Ryser, H. J., Ghani, F., Wu, M. & Peak, D. Inhibition of a reductive function of the plasma membrane by bacitracin and antibodies against protein disulde-isomerase. Proc. Natl Acad. Sci. USA 90, 41124116 (1993).

39. Wang, C. et al. Plasticity of human protein disulde isomerase: evidence for mobility around the X-linker region and its functional signicance. J. Biol. Chem. 285, 2678826797 (2010).

40. Nguyen, V. D. et al. Alternative conformations of the x region of human protein disulphide-isomerase modulate exposure of the substrate binding b domain. J. Mol. Biol. 383, 11441155 (2008).

41. Wang, C. et al. Human protein-disulde isomerase is a redox-regulated chaperone activated by oxidation of domain a. J. Biol. Chem. 287, 11391149 (2012).

42. Denisov, A. Y. et al. Solution structure of the bb domains of human protein disulde isomerase. FEBS J 276, 14401449 (2009).

43. Brophy, T. M., Coller, B. S. & Ahamed, J. Identication of the thiol isomerase-binding peptide, mastoparan, as a novel inhibitor of shear-induced transforming growth factor b1 (TGF-b1) activation. J. Biol. Chem. 288, 1062810639 (2013).

44. Byrne, L. J. et al. Mapping of the ligand-binding site on the b domain of human PDI: interaction with peptide ligands and the x-linker region. Biochem. J. 423, 209217 (2009).

10 NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12579 ARTICLE

45. Wang, C. et al. Structural insights into the redox-regulated dynamic conformations of human protein disulde isomerase. Antioxid. Redox Signal. 19, 3645 (2013).

46. Cook, K. M., McNeil, H. P. & Hogg, P. J. Allosteric control of bII-tryptase by a redox active disulde bond. J. Biol. Chem. 288, 3492034929 (2013).

47. Falati, S., Gross, P., Merrill-Skoloff, G., Furie, B. C. & Furie, B. Real-time in vivo imaging of platelets, tissue factor and brin during arterial thrombus formation in the mouse. Nat. Med. 8, 11751181 (2002).

Acknowledgements

This work was funded by grants from NHLBI (U54 HL112302, R01 HL125275,R01 HL112809, T32 HL007917, T32 HL16324-02, and HL116324), NIDA (DA032476), NIH-MLPCN programme (U54 HG005032), and the Hemostasis and Thrombosis Research Society. The authors are grateful for the SAXS studies conducted at the Advanced Light Source (ALS), a national user facility operated by Lawrence Berkeley National Laboratory on behalf of the Department of Energy, Ofce of Basic Energy Sciences, through the Integrated Diffraction Analysis Technologies (IDAT) programme, supported by DOE Ofce of Biological and Environmental Research. Additional support comes from the National Institutes of Health project MINOS (R01GM105404).

Author contributions

R.H.B., P.B., L.L., A.F. and D.R.K. performed functional assays of thiol isomerase activity; P.P.N. and J.P. performed chemical synthesis; R.H.B., P.B. and L.L. performed thiol isomerase production and characterization; P.B. performed intravital microscopy; L.L. performed intrinsic uorescence experiments; R.H.B performed platelet function studies and proteinase digestions; J.C. and K.M.C. performed and P.J.H. supervised differential alkylation followed by mass spectroscopy; M.H. performed computational modelling and

interpreted SAXS data; R.F. conceived the project; R.H.B. and R.F. prepared the manuscript; R.H.B., P.B., B.F., P.J.H., P.P.N. and R.F. edited the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: R.F. and P.P.N. are co-inventors on pending patents describing bepristats 1 and 2. The remaining authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Bekendam, R. H. et al. A substrate-driven allosteric switch that enhances PDI catalytic activity. Nat. Commun. 7:12579 doi: 10.1038/ncomms12579 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2016

NATURE COMMUNICATIONS | 7:12579 | DOI: 10.1038/ncomms12579 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11