Transitional changes in the CRP structure lead to the exposure of proinflammatory binding sites

Abstract

C-reactive protein (CRP) concentrations rise in response to tissue injury or infection. Circulating pentameric CRP (pCRP) localizes to damaged tissue where it leads to complement activation and further tissue damage. In-depth knowledge of the pCRP activation mechanism is essential to develop therapeutic strategies to minimize tissue injury. Here we demonstrate that pCRP by binding to cell-derived microvesicles undergoes a structural change without disrupting the pentameric symmetry (pCRP*). pCRP* constitutes the major CRP species in human-inflamed tissue and allows binding of complement factor 1q (C1q) and activation of the classical complement pathway. pCRP*-microvesicle complexes lead to enhanced recruitment of leukocytes to inflamed tissue. A small-molecule inhibitor of pCRP (1,6-bis(phosphocholine)-hexane), which blocks the pCRP-microvesicle interactions, abrogates these proinflammatory effects. Reducing inflammation-mediated tissue injury by therapeutic inhibition might improve the outcome of myocardial infarction, stroke and other inflammatory conditions.

ARTICLE

Received 2 Aug 2016 | Accepted 7 Dec 2016 | Published 23 Jan 2017

DOI: 10.1038/ncomms14188 OPEN

Transitional changes in the CRP structure lead to the exposure of proinammatory binding sites

David Braig1,2, Tracy L. Nero3, Hans-Georg Koch4, Benedict Kaiser1, Xiaowei Wang2,5, Jan R. Thiele1,Craig J. Morton3, Johannes Zeller1, Jurij Kiefer1, Lawrence A. Potempa6, Natalie A. Mellett2, Luke A. Miles3,7, Xiao-Jun Du2,5, Peter J. Meikle2,7, Markus Huber-Lang8, G. Bjrn Stark1, Michael W. Parker3,7, Karlheinz Peter2,5,* & Steffen U. Eisenhardt1,*

C-reactive protein (CRP) concentrations rise in response to tissue injury or infection. Circulating pentameric CRP (pCRP) localizes to damaged tissue where it leads to complement activation and further tissue damage. In-depth knowledge of the pCRP activation mechanism is essential to develop therapeutic strategies to minimize tissue injury. Here we demonstrate that pCRP by binding to cell-derived microvesicles undergoes a structural change without disrupting the pentameric symmetry (pCRP*). pCRP* constitutes the major CRP species in human-inamed tissue and allows binding of complement factor 1q (C1q) and activation of the classical complement pathway. pCRP*microvesicle complexes lead to enhanced recruitment of leukocytes to inamed tissue. A small-molecule inhibitor of pCRP (1,6-bis(phosphocholine)-hexane), which blocks the pCRPmicrovesicle interactions, abrogates these proinammatory effects. Reducing inammation-mediated tissue injury by therapeutic inhibition might improve the outcome of myocardial infarction, stroke and other inammatory conditions.

1 Department of Plastic and Hand Surgery, University of Freiburg Medical Centre, Medical Faculty of the University of Freiburg, 79106 Freiburg, Germany.

2 Baker IDI Heart and Diabetes Institute, Melbourne, Victoria 3004, Australia. 3 ACRF Rational Drug Discovery Centre, St Vincents Institute of Medical Research, Melbourne, Victoria 3065, Australia. 4 Institute for Biochemistry and Molecular Biology and Spemann-Graduate School for Biology and Medicine University of Freiburg, Medical Faculty of the University of Freiburg, 79104 Freiburg, Germany. 5 Departments of Medicine and Immunology, Monash University, Melbourne, Victoria 3800, Australia. 6 College of Pharmacy, Roosevelt University, Schaumburg, Illinois 60173, USA. 7 Department of Biochemistry and Molecular Biology, Bio21 Molecular Science and Biotechnology Institute, University of Melbourne, Melbourne, Victoria 3052, Australia. 8 Department of Traumatology, Hand, Plastic, and Reconstructive Surgery, Center of Surgery, University of Ulm, 89081 Ulm, Germany. * These authors jointly supervised the work. Correspondence and requests for materials should be addressed to S.U.E. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

C-reactive protein (CRP) is a homopentameric protein (pCRP), which is synthesized by the liver in response to tissue injury and inammation1. Although circulating

pCRP is not proinammatory in healthy subjects, it exacerbates existing tissue injury in a complement-dependent manner. This has been shown in animal models of rat myocardial infarction, lipopolysaccharide (LPS)-mediated tissue inammation and ischemic cerebral injury26.

The disc-shaped pCRP consists of ve identical subunits: each subunit weighs B23 kDa. The subunits are non-covalently bound by numerous electrostatic and hydrophobic interactions7. The two exposed faces of the pentamer are called the A face (or the effector face) and the B face (or the binding face), respectively. The B face binds damaged or apoptotic cell membranes and bacterial cell walls. Residues on the A face of pCRP are known to interact with complement factor 1q (C1q) and the Fcg receptors; however, the location of these residues in the circulating pentamer suggests that this is not the interacting form of CRP712.pCRP localizes to injured tissue where it undergoes conformational changes (that is, formation of monomeric subunits, monomeric human CRP (mCRP)) leading to neoepitope exposure (that is, residues 199206 become accessible to conformationspecic antibodies 9C9 or 3H12) and complement activation. An initial structural change in the pentameric protein produces pCRP*, a CRP isoform expressing the neoepitope while maintaining an overall pentameric conguration. pCRP* can then dissociate into neoepitope-expressing mCRP4,11,13. Deposits of neoepitope-expressing CRP have been found in various conditions of inammation. Conformation-specic antibodies are commonly used to detect neoepitope-expressing CRP; however, they cannot distinguish between pCRP* and mCRP1419.

Although the link of CRP tissue deposits, complement activation and increased leukocyte inltration is well established, little is known about the sequence of molecular interactions that take place in injured tissue. We therefore investigated the interplay of circulating pCRP with human monocytes and closely followed structural changes and protein interactions to identify potential targets for the reduction of CRP-mediated tissue injury. We demonstrate that pCRP binds to LPS-activated monocytes, and is subsequently released on microvesicles where it undergoes structural changes while maintaining pentameric symmetry (pCRP*). Our data further show that pCRP* constitutes the major CRP species in human-inamed tissue and activates the complement system, which exacerbates the inammatory response. 1,6-bis(phosphocholine)-hexane (1,6-bis-PC), a small-molecule inhibitor that binds to the pCRP phosphocholine binding site20, can inhibit the pCRPmicrovesicle interactions and thereby abrogate CRP-mediated tissue injury.

ResultspCRP binds to activated monocytes. THP-1 monocytic cells were incubated with LPS and/or pCRP. Binding was quantied by ow cytometry with conformation-specic antibodies (anti-pCRP-8D8). Unstimulated cells bound little pCRP over time. In contrast, LPS-activated cells showed a sharp increase in pCRP attachment, which peaked at 15 min and then gradually decreased over time, until it reached the level of resting cells at about 120 min (Fig. 1a,b). Human monocytes, isolated from peripheral blood of healthy subjects, showed similar binding characteristics (Fig. 1c). The pCRPmonocyte interaction was Ca2 -dependent and occurred within physiological pCRP serum concentrations (Supplementary Fig. 1). Overall, these results suggest a specic interaction of pCRP via its Ca2 -dependent binding sites with exposed ligands on LPS-activated monocytes.

pCRP is released on cell-derived microvesicles. Binding was further characterized by confocal uorescence microscopy, which revealed clusters of pCRP on the cell plasma membrane of activated monocytes and THP-1 cells. Shedding of pCRP-bearing membrane areas was visible and some clusters could be observed in close vicinity to the cells (Fig. 1d,e, arrows). These latter clusters ranged in diameter between 100 and 500 nm, which were the typical size of plasma membrane-derived microvesicles. We will adhere to the convention outlined by Buzas et al.21 whereby microvesicles are 1001,000 nm in diameter, exosomes are smaller microvesicles with diameters of 50100 nm and the larger apoptotic vesicles have diameters of 1005,000 nm.

We further investigated the time correlation between pCRP binding and activation of the nuclear factor-kB (NF-kB) pathway, which is switched on in LPS-stimulated immune cells. In LPS-stimulated cells, degradation of IkBa and phosphorylation of p65 increased simultaneously with pCRP binding (Fig. 1f). Likewise, when cells entered a quiescence state after 120 min, pCRP was completely released from the cell plasma membrane. Western blots to detect CRP in the cell-free supernatant before (t 0 min) and after LPS stimulation (t 120 min) conrmed

that pCRP was not consumed by the cells (Fig. 1g). To conrm that binding of pCRP requires membrane changes, which are only present in active cells, THP-1 cells were rst stimulated with LPS for 120 min and only then incubated with pCRP. Binding in this group was similar to cells that were simultaneously incubated with LPS and pCRP for 120 min (Fig. 1h).

These results indicate that circulating pCRP binds to the perturbed plasma membrane of activated monocytes and is released into the surrounding tissue bound to cell-derived microvesicles.

Binding of pCRP to activated leukocytes in the microcirculation. To conrm the in vivo relevance of our ndings, we performed intravital microscopic tracking of pCRP in LPS-induced cremasteric muscle inammation in rats. Shortly after intravenous injection, the Alexa Fluor 594-labelled pCRP could be detected in the microcirculation. During the course of the inammation, pCRP bound to transmigrating leukocytes in postcapillary venules and was transported in the inamed perivascular tissue. If pCRP was preincubated with the small-molecule inhibitor 1,6-bis-PC that blocks the phosphocholine binding sites of pCRP, no transmigrating leukocytes with bound Alexa Fluor 594-labelled pCRP were detected (Fig. 1i).

Stimulated cells release LPC-enriched microvesicles. To further assess the LPS-induced membrane changes, we determined the lipid composition of LPS-stimulated monocytic cells and cell-derived microvesicles by lipidomics. THP-1 cells were incubated with LPS for different time periods, total lipids extracted and analysed by tandem mass spectrometry. Most lipids in the cells decreased over time; the decrease was most pronounced for lysophosphatidylcholine (LPC) and cholesterol (CHO), which dropped signicantly between 20 and 120 min after stimulation. This corresponds to the decrement of pCRP attachment and microvesicle release (Fig. 2a and Supplementary Fig. 2A). As loss of lipids might be due to shedding of micro-vesicles, we examined the lipid content of the 1,500 g cell-free supernatant of resting and LPS-activated cells. The supernatant uids of unstimulated cells contained only 873 pmol ml 1 of phospholipids, whereas 1,20744 pmol ml 1 could be detected in the LPS-stimulated group (Fig. 2b). Subsequent differential centrifugation of the 1,500 g supernatant revealed that the majority of microvesicles could be pelleted at 16,000 g. A minor species of even smaller vesicles could be pelleted from the 16,000 g

2 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

supernatant at 100,000 g, as well as at 400,000 g (Fig. 2b). We observed similar results with ADP-stimulated platelets. The supernatant of unstimulated platelets contained 626 5 pmol ml 1 of phospholipids, whereas 2,25218 pmol ml 1 were detected after stimulation. Again, differential centrifugation

pelleted most lipids at 16,000 g and smaller fractions at 100,000 g and 400,000 g (Fig. 2c).

Determination of the lipid composition of microvesicle populations revealed considerable differences when compared with THP-1 cells, as well as among the individual microvesicle

a b

%

80

60

40

20

100

LPS+pCRP

LPS

pCRP

100

80 60

15 min

120 min

*

8D8-FITC

%

40

20

Normalized to mode

Normalized to mode

40 20

0

*

40 20

0 0 20 40 60 80 100 120

80 60

105

min

102 103 104

FITC-A

105 102 103 104

FITC-A

c d

DAPI 8D8-FITC Merge 3D

DAPI 8D8-FITC Merge 3D

%

60

P = 0.003

LPS+pCRP

LPS

0

**

60 ** ** **

8D8-FITC

**

e

0 0 20 40 60 80 100 120

min

h

P = 0.012

P = 0.012

f g

0 5 10 20 30 60 120

t (min)

IB

50 40 30

37 kDa 75 kDa

75 kDa

37 kDa

- p65

8D8-FITC

20 10

0

P = 0.552

p65

t (min) 0 120

Anti-CRP-8

25 kDa

GAPDH

LPS 120 min LPS + pCRP 120 min

LPS + pCRP 30 min

pCRP 30 min

i

30 min 45 min 60 min

LPS + Alexa Fluor

594-labelled pCRP

LPS + Alexa Fluor

594-labelled pCRP + bisPC

Alexa Fluor 594-labelled pCRP after 20 min

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

As expected, pCRP bound to LPC containing liposomes but not to PC-only liposomes (Fig. 4a). We also observed pCRP binding to THP-1 and Jurkat cell-derived microvesicles (Fig. 4a). EDTA, as well as 1,6-bis-PC, inhibited the interaction (Fig. 4b), which suggests a Ca2 -dependent interaction via the phosphocholine binding sites with membrane phospholipids. BNPAGE allows separation of pCRP and mCRP. Bound CRP migrated at the same velocity as control pCRP, suggesting that it maintained pentameric symmetry while bound to microvesicles (Fig. 4a,b).

As THP-1 cells harbour Fcg receptors, which might provide an additional interaction site for CRP on THP-1 cell-derived microvesicles, binding was further quantied on Jurkat cell-derived microvesicles. This T-cell-derived cell line does not harbour Fcg receptors (Supplementary Fig. 3A,B). There was no signicant difference in the amount of pCRP that bound to liposomes, THP-1 or Jurkat cell-derived microvesicles. This suggests the perturbed phospholipid membrane as the primary binding site for pCRP on microvesicles (Fig. 4c).

The dynamics of the pCRPmicrovesicle interaction was characterized in real time based on repeated measurements of Alexa Fluor 488-labelled pCRP to immobilized microvesicles.Binding of pCRP to microvesicles was essentially complete after 30 min, and the addition of EDTA led to a rapid dissociation (Fig. 4d).

The release of pCRPmicrovesicle complexes by THP-1 cells was examined by ow cytometry. Cells were incubated with LPS and pCRP and the uorescence of released microvesicles directly quantied with conformation-specic anti-pCRP-8D8 antibodies by gating on the microvesicle population (Fig. 4e and Supplementary Fig. 4A,B). pCRPmicrovesicle complexes were detectable within minutes and the uorescent signal on the vesicles slightly increased over time.

To determine if the lipid composition of microvesicles alone was sufcient to allow binding of pCRP, multilamellar liposomes were created to mimic either large (P 16,000 g) or small (SN 100,000 g) THP-1 cell-derived microvesicles. These mimics also contained physiological amounts of LPC and lysophosphatidylethanolamine (LPE), as determined by the lipid mass spectrometry experiments. pCRP was found to bind to both microvesicle mimics, and this interaction was EDTA-sensitive (Fig. 4f).

We next tested whether pCRP binds to microvesicles of different cellular original. Microvesicles were puried from stimulated monocytes, polymorphonuclear leukocytes and platelets. All microvesicle species bound pCRP in a Ca2-dependent manner, and monocytic microvesicles showed the highest afnity.

Figure 1 | pCRP binds to activated monocytes. (a) THP-1 cells were incubated with LPS. Where indicated pCRP (100 mg ml 1) was added simultaneously with LPS. Cells were stained with anti-pCRP-8D8/FITC. Fluorescence was recorded by ow cytometry. *P valueo0.05 compared to resting cells (paired t-test). Shaded areas display the s.e.m. (nZ3 for each group and time point). (b) Representative uorograms of LPS-stimulated THP-1 cells in the presence of pCRP at t 0 (blue curves) and at indicated time points (red curves). (c) pCRP binding to human monocytes was analysed as in (a). **P value

o0.01 (paired t-test). Displayed are meanss.e.m. (n 3 for each group and time point). Shaded areas dene the s.e.m. (d) Confocal uorescence

microscopy of LPS-activated human monocytes in the presence of pCRP. Depicted is a plane from the centre of the cell and a 3D reconstruction of the whole cell. The nucleus was stained with DAPI (blue) and pCRP with anti-pCRP-8D8/FITC (green). pCRP localizes in clusters on the plasma membrane. pCRPmicrovesicle complexes separate from the membrane (arrows). Scale bars, 10 mm. (e) Confocal uorescence microscopy of LPS-activated THP-1 cells. Cells were treated as in (d). Scale bars, 10 mm. (f) Western blot of NF-kB pathway screen of LPS-activated THP-1 cells. Cells were incubated for indicated time periods with LPS, lysed and separated on SDSPAGE. Uncropped images of western blots are shown in Supplementary Fig. 12. (g) Western blot of cell-free supernatants after incubation of THP-1 cells with LPS and pCRP for indicated time periods. CRP was detected after separation on SDSPAGE, which induced dissociation of the pentamer. The detected size thus corresponds to a CRP monomer (23 kDa). (h) pCRP binding to THP-1 cells. Cells were treated as described under (a); however, one sample was simultaneously incubated with pCRP and LPS for 120 min, whereas the other sample was rst incubated with LPS for 120 min and only then with pCRP for 30 min. Bars indicate means.e.m. P values were calculated with a paired t-test (n 3). (i) In vivo visualization of the pCRPleukocyte interaction. Rat cremasteric muscle tissue was stimulated via superfusion with 1 mg ml 1 LPS. Alexa

Fluor 594-labelled pCRP (25 mg ml 1 serum concentration) or pCRP preincubated with 1,6-bis-PC (bisPC) was added 20 min later. At t 45 min binding of

pCRP to transmigrated leukocytes can be observed (arrows). In the presence of 1,6-bis-PC, labelled pCRP can be seen in the microcirculation, but not on transmigrating leukocytes. Scale bars, 100 mm.

species. Microvesicles in the 16,000 g pellet were highly enriched in CHO and the plasma membrane phospholipids sphingomyelin (SM), ceramides (Cer) and phosphatidylserine (PS), when compared to THP-1 cells (Fig. 3a). Their size and composition are characteristic for microvesicles, which are released from the cell plasma membrane22. The 100,000 g pellet and 100,000 g supernatant contained high amounts of phosphatidylethanolamine (PE) and lower amounts of PS, SM and Cer, when compared to microvesicles in the 16,000 g pellet (Fig. 3b). In addition, these smaller microvesicles contained increased amounts of LPC (Fig. 3c). This composition reects their smaller size and indicates their origin from the endoplasmic reticulum. Platelet-derived microvesicles contained a similar CHO content, were enriched in SM, but contained less phosphatidylinositol (PI) than platelets (Fig. 3d). Again, the smaller vesicles contained increased amounts of LPC (Supplementary Fig. 2B).

We further determined the inuence of pCRP on microvesicle release by ow cytometry. Microvesicles were identied by their characteristic size and PS content in the outer membrane leaet23. THP-1 cells rapidly released microvesicles after LPS stimulation and no further increase could be detected after 120 min. There was no signicant difference between LPS and LPS pCRP-

treated cells. pCRP alone failed to induce microvesicle release (Supplementary Fig. 2C). Microvesicle production was due to cell activation and not the result of cell apoptosis, as no increase in Annexin V binding to LPS-treated cells was observed (Supplementary Fig. 2D). We additionally differentiated human monocytes into M1 and M2 macrophages and analysed the ability of these macrophages to release microvesicles on stimulation with phorbol 12-myristate 13-acetate (PMA), monophosphoryl lipid A (MPLA) and LPS (Supplementary Fig. 2E,F). In contrast to monocytes and platelets, which show a marked increase of microvesicle release on stimulation, both macrophage subsets showed a constant release of microvesicles over time, which did not increase on stimulation.

pCRP binding to microvesicles induces structural changes. We analysed the interaction of pCRP with microvesicles in more detail. pCRP was incubated with either THP-1 or Jurkat cell-derived microvesicles. Bound pCRP was pelleted and analysed by blue nativepolyacrylamide gel electrophoresis (BNPAGE).pCRP can attach to exposed PC and PE head groups of perturbed membrane phospholipids. Introducing LPC into the lipid bilayer of liposomes can mimic this phenomenon24.

4 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

a

P < 0.001

P < 0.001 P < 0.001

NS P < 0.001

P < 0.001

P < 0.01

P < 0.01

P < 0.05

Lipids (pmol per mg protein) PE (pmol per mg protein)

P < 0.05

P < 0.01

250,000

200,000

150,000

100,000

50,000

25,000

20,000

15,000

10,000

5,000

PC (pmol per mg protein)

100,000

80,000

60,000

40,000

20,000

0 0 min 20 min

THP-1 cells - PC

120 min 0 min 20 min

THP-1 cells - LPC

120 min

LPC (pmol per mg protein)

1,500

1,000

500

0

LPE (pmol per mg protein)

3,000

2,000

1,000

0

20,000

15,000

10,000

5,000

0

PS (pmol per mg protein)

6,000

4,000

2,000

0

P < 0.05

120 min

0 0 min 20 min

THP-1 cells - PE

120 min 0 min 20 min

THP-1 cells - LPE

120 min 0 min 20 min

THP-1 cells - PS

120 min

P < 0.001

P < 0.001

30,000

20,000

10,000

0

0 0 min 20 min

THP-1 cells - total lipids

NS NS

NS

NS

P < 0.05

SM (pmol per mg protein)

CHO (pmol per mg protein)

80,000

60,000

40,000

20,000

0

NS

PI (pmol per mg protein) Phospholipids (pmol ml

1 )

0 min 20 min

THP-1 cells - PI

120 min 0 min 20 min

THP-1 cells - SM THP-1 cells - CHO

120 min 0 min 20 min 120 min

b c

THP-1 cells Platelets

1,500

1,000

500

0 Control

Phospholipids (pmol ml1 )

1,500

1,000

500

0

Phospholipids (pmol ml1 )

3,000

2,000

1,000

Phospholipids (pmol ml1 )

2,000

1,500

1,000

500

0

LPS

P 16,000 g

P 100,000 g

P 400,000 g

0 Control

ADP

P 16,000 g

P 100,000 g

P 400,000 g

THP-1 supernatant

Centrifugation of LPS - SN Platelet supernatant Centrifugation of ADP - SN

Figure 2 | LPS-activated THP-1 cells release microvesicles. THP-1 cells were stimulated with LPS for indicated time periods. Total lipids were extracted and subjected to mass spectrometric analysis. (a) THP-1 cells lost signicant amounts of lipids after LPS stimulation. Most lipids decreased signicantly over time, with LPC and CHO showing the most distinctive decline, which occurred between 20 and 120 min after stimulation. Lipid values are expressed as pmol per mg protein. Other lipids identied were: LPE, phosphatidylcholine (PC), PE, PI, PS and SMs (n 12). P values were calculated with one-way

analysis of variance (ANOVA). (b) Lipid concentrations in the 1,500 g supernatant of LPS stimulated and resting THP-1 cells were determined by mass spectrometric analysis (n 6). The 1,500 g supernatant after LPS stimulation was further separated by differential centrifugation. Microvesicles were

collected at 16,000 g, and smaller microvesicles pelleted at 100,000 g and 400,000 g. Displayed lipid values are expressed as pmol ml 1 supernatant (SN) and the s.d. (n 6). (c) Lipid concentrations in the 1,500 g supernatant of ADP-stimulated and resting platelets (n 2). The 1,500 g supernatant

after ADP stimulation was further separated by differential centrifugation as described in (b). Displayed lipid values are expressed in pmol ml 1 supernatants.d. (n 2).

Binding was independent of the stimulant used to trigger the release of microvesicles (Fig. 4g and Supplementary Fig. 4C,D).

Several reports described conformational changes in pCRP upon membrane binding, which may lead to neoepitope expression or even dissociation of the pentameric pCRP into its

subunits13,16,25. Conformational changes on interaction with phospholipid bilayers were monitored by tryptophan uorescence analysis. Of the six tryptophan residues in each CRP monomer, Trp110 and Trp205 (part of the neoepitope dened by residues 199206) are buried at the monomermonomer interface

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

a

PS 3%

Cer

4%

P < 0.0001

60

LPC0.8%

SM

9%

PC 53%

LPE

2%

LPE

2%

PI 9%

PS8% Cer

14%

% CHO of total lipids

40

20

0

PE 14%

PI 14%

PE 14%

SM 18%

LPC

1%

Cells

P 16,000 g

PC 34%

P 16,000 g (THP-1)

THP-1 cells

b c

THP-1

PI 3%

PS 3%

LPE

4%

8

6

4

2

LPE

1%

PS 3%

PI 6%

Cer

5% SM

11%

Cer

8%

LPC

6%

PE 27%

SM 17%

% LPC of phospholipids

LPC

3%

PE 31%

PC 38%

PC 34%

0 Cells

P 16,000 g

P 100,000 g

SN 100,000 g

P 100,000 g (THP-1)

Platelets

SN 100,000 g (THP-1)

d

Cer

2% Cer

3%

P = 0.33

60

40

20

0

PS 18%

PI 4%

SM14% SM

22%

PC41% PC

44%

LPC

1%

PS 13%

% CHO of total lipids

PE 12%

LPE

1%

PI 8%

LPC

2%

PE 15%

P 16,000 g

Platelets

P 16,000 g (platelets)

Figure 3 | Lipid composition of THP-1 cells and microvesicles. (a) THP-1 cells and cell-derived microvesicles (16,000 g pellet) were lipid extracted and subjected to mass spectrometric analysis. Microvesicles contain large amounts of CHO, phosphatidylcholine (PC), SMs, ceramides (Cer) and PS, which are predominantly found in the cell plasma membrane. The other lipids identied were: LPC, LPE, PE and PI. Displayed are means and s.d. P values were calculated with an unpaired t-test (n 12). (b) Smaller microvesicles contain increasing amounts of PE, reecting their endoplasmic reticulum origins.

(c) LPC content increases in small microvesicles and induces a positive membrane curvature due to the exibility of the single-chain fatty acid(nZ4 for each group). In (a), CHO is expressed in % of total lipids. Other membrane lipids are expressed as per cent of total lipids (excluding CHO). Lipids, which comprises o0.5% of total lipids, were excluded from the analysis. (d) The lipid composition of platelets and platelet-derived microvesicles (16,000 g) was determined as described above (ac) (n 6). P values were calculated with an unpaired t-test.

in pCRP. Freshly prepared samples of pCRP, pCRP liposomes

(40% CHO, 40% PC and 20% LPC) and mCRP showed iodide quenching of the tryptophan uorescence that is in agreement with a simple SternVolmer relationship (Supplementary Fig. 5A) which suggests that the tryptophan residues within each protein are all accessible to quenching to the same degree. The lower SternVolmer rate constant (KSV) for mCRP

(0.5 M 1) indicates that the uorescence quencher has greater accessibility to the tryptophan residues than in either the pCRP or pCRP liposome samples (B0.9 M 1), that is,

the tryptophan residues are more solvent exposed in mCRP. After incubation overnight to allow full interaction between

pCRP and the liposomes13, the pCRP and mCRP samples remained in agreement with the simple SternVolmer relationship by continuing to show a linear dependence of the F0/F ratio to [I ] (Supplementary Fig. 5B). The pCRP liposome sample, however, has lost the linear

relationship and now shows a curved F0/F versus [I ] plot. Plotting the data according to the modied SternVolmer equation (Supplementary Fig. 5C) indicates that 25% of the tryptophan uorescence is now accessible to quenching, with the remaining 75% protected. This is consistent with a change in the effective solvent accessibility for at least one tryptophan residue in each monomer within pCRP.

6 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

We also analysed neoepitope expression on THP-1 cells and microvesicles with conformation-specic antibodies. As described above (Fig. 1), the anti-pCRP-8D8 antibody detected pCRP on activated monocytic cells. In contrast, anti-mCRP-9C9 antibodies did not detect cell-bound CRP at 30 min and only a minimal signal was observed after 120 min (Fig. 4h). When bound to microvesicles, CRP exhibited reactivity with anti-pCRP-8D8 as well as anti-mCRP-9C9 antibodies (Fig. 4h). In addition, we examined the deposition of native and neoepitope-expressing CRP on the surface of

microvesicles from the circulation of patients with ST-elevation myocardial infarction. Microvesicle subsets were identied by CD11b (leukocytes), CD41 (platelets) and CD62P (activated platelets). Similar to the results with puried microvesicles, we identied CRP on microvesicles of different cellular origin, which could be detected by both anti-pCRP-8D8 and anti-mCRP-9C9 antibodies (Fig. 4i). Together with the BNPAGE data for microvesicle-bound CRP, these data suggest that both pCRP and pCRP* are present on cell-derived microvesicles.

a

b

THP1 Jurkat

MV SN SN

mCRP pCRP Liposomes

EDTA

THP-1 MV Jurkat MV

THP-1 MV Jurkat MV

pCRP

mCRP PC

PC LPC

MV mCRP pCRP

Liposomes bisPC

EDTA

EDTA

bisPC

bisPC

pCRP

mCRP

Anti-CRP-8

Anti-CRP-8 Anti-CRP-8

c d

e

NS

% pCRP

Control

% NS

** ** **

80

pCRP recovered form input

40

20

2,000

5 g ml1 pCRP

2.5 g ml1 pCRP EDTA

1 g ml1 pCRP

1,500

1,000

500

0 0 10 20

Min Min

Signal (l) fluo corrected

40

30

20

*

60

*

*

*

EDTA

EDTA

0

bisPC

bisPC

EDTA

bisPC

30 40 50

10 0 0 20 40 60 80 100

120

PC PC/LPC THP-1 MV Jurkat MV

f g

P 16,000 g

Liposomes

SN 100,000 g

Liposomes

NS

mCRP pCRP

PC/LPC

Liposomes EDTA

% 100

EDTA

EDTA

*

pCRP recovered from input

*

NS

80

60

40

20

pCRP recovered from input

8D8-FITC-positive MV

100

**

pCRP

80

60

40

20

* ** *

0 PC/LPC

P 16,000 g

SN 100,000 g

0 + + + + + + EDTA

Liposomes Mono (LPS)

Mono (MPLA)

mCRP

Anti-CRP-8

PMNL (PMA)

PMNL (fMLP)

platelets

Liposomes

h i

THP1-cells

% 60

50

40

30

THP1-MV

% 50

FITC-positive THP-1 cells

P<0.01

% P<0.05

P=0.01

NS

P<0.01

P<0.01

NS

% %

90 80 70 60 50 40 30

8D8

NS

40

30

20

10

P<0.01

FITC-positive MV

FITC-positive MV

100 100 100

90 80 70 60 50 40 30

90

80 70 60 50 40 30

NS

NS

FITC-positive MV

NS

20

10

0

FITC-positive MV

8D8

9C9

8D8

9C9

0 8D8

9C9

9C9

8D8

9C9

8D8

9C9

8D8

9C9

8D8

9C9

8D8

9C9

LPS + pCRP 30 min

LPS + pCRP 120 min

LPS + pCRP 120 min

Control STEMI CD11b

Control

STEMI Control STEMI CD41 CD62P

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

the tissue lysates or precipitation of mCRP (Supplementary Fig. 6B,C). Since immunohistochemistry of burn wounds demonstrated that the neoepitope is exposed, the pentameric form in burn wounds and CEA specimens can be identied as pCRP*.

As our assay only indirectly identies pCRP, we further analysed tissue lysates of burn wounds on SDSPAGE with 1/20 of the commonly used SDS content. The low SDS concentrations preserve pCRP and allow separation of pCRP and mCRP28. Lysates showed a distinct band, migrating to the same height as control pCRP. This band was not visible in healthy skin tissue lysates and we did not observe a band that migrated at the height of mCRP (Fig. 5e). Gels were boiled in SDSPAGE buffer before western transfer to enable uniform detection of CRP isoforms by anti-CRP-8 antibodies.These results conrm that CRP tissue deposits contain large amounts of pentameric CRP, mostly pCRP*.

As we observed a structural change from pCRP to pCRP* on the surface of liposomes and microvesicles, but no dissociation into mCRP by BNPAGE (Fig. 4a,b), we also employed the red/ox-assay to these specimens. In addition to a higher sensitivity in mCRP detection (compare Fig. 5d to Fig. 5e), the assay analyses soluble CRP in the supernatant as well as the membrane-bound fractions. A conversion of pCRP to mCRP in the presence of liposomes or microvesicles was not found, conrming our BNPAGE results (Fig. 5f). Liposomes did not prevent reduction of mCRP by DTT, as mCRP, which being added to liposomes could be effectively reduced (Fig. 5g).

Microvesicle-bound pCRP* activates the complement system.

Whereas circulating pCRP is a rather inert molecule, CRP bound to phospholipid membranes has been shown to activate the complement cascade3,10,29. We investigated whether pCRP* bound to microvesicles initiates the complement cascade. Binding of 5-chloromethyluorescein diacetate (CMFDA)-labelled microvesicles to immobilized C1q was studied in the presence of mCRP and pCRP. Only few microvesicles bound to C1q in the absence of CRP. In contrast, both CRP isoforms led to increased binding, and pCRP was considerably more potent than mCRP (Fig. 6a).The interaction of microvesicles with C1q revealed a visually slower on-rate of microvesicles than the interaction with pCRP (Fig. 6a versus Fig. 4d). We also immobilized

Figure 4 | pCRP binding to microvesicles induces structural changes. pCRP was incubated with liposomes (PC or PC/LPC) and microvesicles (MV) from THP-1 and Jurkat cells. The supernatant (SN) of the last wash step after MV purication was used as control. Vesicles were pelleted and subjected to BNPAGE. (a) pCRP bound to PC/LPC liposomes, THP-1 and Jurkat cell-derived MV. No binding occurred to PC-only liposomes and no CRP could be found in the pellet fraction of the supernatant. mCRP was not found on the surface of MV and liposomes. (b) Ca2 -depletion (EDTA) or 1,6-bis-PC (bisPC)

reduces binding of pCRP to MV. (c) Quantication of pCRP binding. There was no signicant difference in the amount of pCRP that bound to either MV species or PC/LPC liposomes (n 5). EDTA and 1,6-bis-PC signicantly reduced binding (n 2). P values were calculated with a Students t-test.

**P value of o0.01, NS not signicant. Displayed are means and s.e.m. (d) Kinetics of the pCRPMV interaction were determined by measuring binding

of Alexa Fluor 488-labelled pCRP to immobilized THP-1 MV. (e) Binding of pCRP to THP-1 cell-derived MV was analysed by ow cytometry with conformation-specic antibodies for pCRP (anti-pCRP-8D8/FITC). P values were calculated with a paired t-test. *P value of o0.05 to control MV and shaded areas display the s.e.m. (n 4). (f) Binding of pCRP to liposome mimics of THP-1 MV (P 16,000 g and SN 100,000 g) was analysed as described

above. Binding was Ca2 -dependent and we did not observe dissociation of pCRP into its monomeric subunits. Displayed are means and s.e.m. (n 3).

(g) Binding of pCRP to different cell-derived MV was analysed on BNPAGE as described in (c). Monocytes (Mono), polymorphonuclear leukocytes (PMNL), phorbol 12-myristate 13-acetate (PMA), monophosphoryl lipid A (MPLA), N-formylmethionyl-leucyl-phenylalanine (fMLP). Displayed are means and s.e.m. (n 3). P values were calculated with a Students t-test. *P value o0.05 and **P value o0.01. (h) Conformational changes in pCRP on binding to

THP-1 cells and MV were assessed with conformation-specic antibodies by ow cytometry. CRP bound to THP-1 cells could only be detected by anti-pCRP-8D8 antibodies, whereas CRP on MV was recognized by anti-pCRP-8D8 and anti-mCRP-9C9 antibodies. Displayed are means and s.d. (n 3).

(i) Deposition of CRP on MV of different cellular origin in the circulation of patients with ST-elevation myocardial infarction (STEMI) was analysed by FACS with anti-pCRP-8D8/FITC and anti-mCRP-9C9/FITC antibodies. Displayed are the percentages of FITC-positive MV of each subset. Bars indicate means and s.d. (n 4). P values were calculated with an unpaired t-test.

Structural isoforms of CRP tissue deposits. Tissue-deposited CRP is preferentially recognized by anti-mCRP antibodies (that is, the neoepitope is exposed) and it is therefore thought to be structurally distinct from circulating pCRP (where the neoepitope is not exposed). Most previous studies have not investigated the quaternary structure of deposited CRP and thus the relative proportions of pCRP* and mCRP are unknown4,1417,26.

As a model of tissue inammation, we evaluated the previously described CRP deposition in burn wounds15 for its quaternary structure and immune reactivity. CRP in burn tissue lysates migrated on denaturing SDSpolyacrylamide gel electrophoresis (SDSPAGE)-like control CRP at the predicted weight-averaged molecular mass of a CRP monomer (23 kDa) (Fig. 5a). Immunohistochemistry of burn wounds with conformationspecic anti-pCRP-8D8 and anti-mCRP-9C9 antibodies showed distinct staining with anti-mCRP-9C9, but only minimal staining with anti-pCRP-8D8 (Fig. 5b).

Each CRP subunit contains an intrasubunit disulde bond. Reduced CRP migrates slower on SDSPAGE than oxidized CRP, and both isoforms can thus be identied by separation of N-ethylmaleimide (NEM)-stabilized CRP under oxidizing conditions27. To determine the relative proportions of reduced and oxidized CRP in inamed tissue, fresh lysates of burn wounds were treated with NEM and separated on SDSPAGE. Burn wound lysates contained no detectable amounts of reduced CRP (Fig. 5c).

The intrasubunit disulde bond can also be used to distinguish pentameric forms of CRP from mCRP, as low concentrations of DTT (dithiothreitol) are able to reduce mCRP but not pCRP (Supplementary Fig. 6A)27. We employed this assay to tissue lysates of human burn wounds, skeletal muscle tissue with ischemia/reperfusion injury (I/R injury) and carotid endarterectomy (CEA) specimens to determine the relative amounts of pentameric and monomeric CRP. Whereas control mCRP was efciently reduced at 37 C, only small amounts of tissue-deposited CRP could be reduced under the same conditions, indicating that the majority of CRP is in a pentameric form. If the samples were boiled, which induces pCRP dissociation, we observed nearly complete reduction of the intrasubunit disulde bond (Fig. 5d). Quantication of western blots revealed that while burn wounds and CEA specimens contain signicant amounts of mCRP, the majority of deposited CRP remained in a pentameric form (Fig. 5d). Detection of mCRP was not impaired by unknown components within

8 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

microvesicles and monitored the interaction with Alexa Fluor 488-labelled C1q. Only a small amount of C1q bound in the absence of pCRP, but we observed a similar increase and analogous binding characteristics in the presence of pCRP as that seen with immobilized C1q (Supplementary Fig. 7A).

We further studied complement activation on microvesicles by FACS with different complement component 3 (C3) antibodies directed against C3b (recognizes C3, C3b and iC3b), C3c (recognizes C3, C3b, iC3b and C3c), iC3b (iC3b neoepitope specic) and C3d (recognizes C3, C3b, iC3b and C3d). Microvesicles were incubated with normal human sera (NHS) and deposition of complement C3 determined by the C3b antibody. pCRP* signicantly increased deposition of

complement C3b on microvesicles, when compared to NHS only. In contrast, heat-inactivated sera and pCRP alone did not lead to C3b deposition (Fig. 6b). The subtype-specic antibodies conrmed that C3 is deposited on microvesicles even in the absence of CRP and further processed to iC3b. mCRP led only to a minor increase in C3 binding. However, in the presence of pCRP large amounts of C3, deposited mainly in the form of iC3b, were detected on the surface of microvesicles (Fig. 6b and Supplementary Fig. 7B).

pCRP*microvesicle complexes activate endothelial cells. Monocytic microvesicles activate endothelial cells in an

a b c

100 m 100 m

Healthy skin

CRP control

Burn wound

Control CRP

Burn wound-tissue lysate Red Ox

25 kDa

25 kDa

Red

Ox

Anti-CRP-8

Anti-mCRP-9C9 Anti-pCRP-8D8

Anti-CRP-8

d

Muscle tissue

Control pCRP Control mCRP

37C 95C

37C 95C

37C 95C

Control

pCRP

pre - I/R

post - I/R

25 kDa

I/R muscle

DTT DTT DTT DTT

Red

25 kDa

Ox

Anti-CRP-8

** **

Anti-CRP-8

Control pCRP Control mCRP Burn wound

DTT DTT DTT DTT

NS

80

***

Red

*

Reduced CRP

%

60

40

20

0

25 kDa

Ox

Anti-CRP-8

control pCRP control mCRP

DTT DTT DTT DTT

CEA plaque

*

Red

25 kDa

Control pCRP

Control mCRP

Ox

I/R muscle

Burn wound

CEA plaque

Anti-CRP-8

Burn wound

e

f

Healthy skin

Control pCRP

Control mCRP

MV

pCRP mCRP

PC/LPC

DTT DTT DTT DTT

pCRP

25 kDa

25 kDa

pCRP

#

Anti-CRP-8

g

pCRP mCRPPC PC/LPC PC PC/LPC

DTT DTT DTT DTT

mCRP

Anti-CRP-8

Anti-CRP-8

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

and mCRP (recombinant C36A,C97A mutant) demonstrate that these two CRP isoforms have different secondary structures (Supplementary Fig. 9A). The difference in the CD spectrum indicates that the mCRP protein is signicantly more disordered than pCRP. This partially unfolded state might account for the rapid clearance of mCRP from circulation32,33. We examined whether mCRP and pCRP* inuence uptake of CMFDA-labelled microvesicles in human macrophages. Only small amounts were engulfed in the absence of NHS, pCRP* and mCRP. NHS alone led to increased phagocytosis, which increased further in the presence of pCRP*.The effect of mCRP was even more pronounced, but contrary to pCRP*, NHS could not amplify its effects. If complement was inactivated by heat pretreatment, the sera failed to increase phagocytosis of microvesicles (Supplementary Fig 10).

Model of pCRP* interaction with complement C1q. The crystal structure of native pCRP in complex with phosphocholine is shown in Fig. 7a. The phosphocholine (or phosphoethanolamine) ligand binding site is located in the groove of a b-sheet on the B face of pCRP. In Fig. 7b the phosphocholine head group of an LPC molecule has been docked into the ligand binding site. The phosphocholine moieties of the inhibitor 1,6-bis-PC have been shown to interact in a likewise manner20. When Ca2 is not present (that is, in conditions of Ca2 deciency or addition of

EDTA), residues 140150 form a loop that projects away from the body of the CRP subunit exposing a normally hidden proteolysis site and the pCRP subunits dissociate as the individual subunit conformation unfolds7,34,35.

Activation of monocytic cells by LPS stimulation perturbs the cell membrane resulting in the exposure of PC, LPC, SM, PE and LPE lipid phosphocholine or phosphoethanolamine head groups, enabling pCRP to bind to the lipid surface (Fig. 1d,e). Only pCRP was detected on the surface of activated THP-1 cells or human monocytes (Fig. 4h). In contrast, both pCRP and pCRP* were detected on the surface of microvesicles (Fig. 4h,i). To demonstrate the interaction of pCRP with the surface of an LPS-activated THP-1 cell, or microvesicle, pCRP was docked to a model lipid bilayer containing CHO and 1-palmitoyl-2-deoylphosphatidylcholine (POPC) (Fig. 7c,d).The phosphocholine head groups of POPC interact in a similar manner to that depicted in Fig. 7b for LPC. There are distinct hydrophobic patches on the A and B faces of pCRP. These patches on the pCRP B face can interact with lipid rafts (that is, regions rich in CHO, Cer, PI, PS and SM) on the activated cell or microvesicle surface. A CHO binding

Figure 5 | Quaternary structure of CRP in inamed tissue. (a) Tissue lysates of burn wounds and healthy skin were separated on denaturing SDSPAGE, which induces dissociation of pCRP. Anti-CRP-8 antibodies detected a band in burn wound lysates, which had the size of a CRP monomer (23 kDa). (b) Immunohistochemistry of burn wounds revealed distinct staining (green) by anti-mCRP-9C9 antibodies, but only minimal staining with anti-pCRP-8D8 antibodies. Scale bars, 100 mm. (c) The oxidation state of CRP in burn wounds was assessed by non-reducing SDSPAGE. Reduced CRP migrates slower and can thus be distinguished from oxidized CRP. We did not observe reduced CRP in tissue lysates of burn wounds. (d) The quaternary structure of CRP in human skeletal muscle after I/R injury (two patients, total of six samples), burn wounds (three patients, total of ve samples) and CEA specimens (one patient, total of two samples) was determined by assessing the accessibility of the intrasubunit disulde bond. Tissue lysates were incubated with 10 mM DTT at 37 C to assess the accessibility of the intrasubunit disulde bond. DTT can efciently reduce the disulde bond in mCRP, but not in pCRP/pCRP*. Small amounts of tissue-deposited CRP could be reduced by DTT (indicated by an asterisk). Reduction was effective if lysates were heated, which induces pCRP dissociation. Quantication of reduced CRP revealed a signicant difference between control pCRP and CRP in tissue lysates of burn wounds and CEA plaques. Displayed are means and s.e.m. P values were calculated with an unpaired t-test. *P value o0.05, **P value o0.01 and ***P value o0.001. (e) Tissue lysates were separated on 1/20 SDSPAGE. Anti-CRP-8 antibodies recognized a distinct band in burn wound lysates that migrated at the same height as control pCRP. It did not recognize a band with similar size to mCRP. A band, which could be observed in burn wound lysates and healthy skin, was also recognized (#) and most likely reects binding to a common skin epitope. (f) Accessibility of the intrasubunit disulde bond in CRP in the presence of PC/LPC liposomes and microvesicles (MV). Control mCRP was effectively reduced. No dissociation of pCRP to mCRP was observed. (g) Lipid binding does not affect reduction of mCRP by DTT, as reduction of control mCRP incubated with liposomes was unimpeded. Uncropped images of western blots are shown in Supplementary Fig. 13.

IL-1b-dependent manner30. Incubation of human umbilical vein endothelial cells (HUVECs) with monocytic microvesicles leads to upregulation of adhesion molecules (for example, intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1)), which mediate recruitment of leukocytes to the site of tissue injury. Our study conrmed previous results that monocytic microvesicles lead to enhanced expression of ICAM-1 and VCAM-1 and that native pCRP does not increase their expression (Supplementary Fig. 7C).

We further assessed whether NHS and pCRP* inuence the proinammatory effects of microvesicles. Microvesicles from LPS-activated THP-1 cells were incubated with pCRP, NHS or pCRP NHS, and their effects on VCAM-1 and ICAM-1

expression assessed by quantitative real-time PCR with reverse transcription and western blot. pCRP*-covered microvesicles led to an increase in ICAM-1 and VCAM-1 expression. This proinammatory effect of pCRP* was even more pronounced in the presence of NHS, which led to a twofold increase of both adhesion molecules when compared to THP-1 cell-derived microvesicles only (Fig. 6c). The proinammatory effects of pCRP* were not restricted to monocytic microvesicles and could also be observed with Jurkat cell-derived microvesicles. However, their potential to induce expression of cell adhesion molecules was about 2030-fold lower (Fig. 6c).

To demonstrate the in vivo relevance of the proinammatory pCRP* effects, we quantied leukocyte transmigration in the microcirculation of the rat cremaster muscle by intravital microscopy. In an established model for localized inammation, the muscle tissue was superfused with low levels of LPS to induce cell activation and microvesicle release4. LPS triggered leukocyte evasion into the perivascular tissue and pCRP* signicantly increased the number of transmigrated leukocytes. This proinammatory effect of pCRP* could be inhibited by 1,6-bis-PC (Fig. 6d). Depletion of the complement system by pretreatment of rats with cobra venom factor abrogated most of the pCRP*-mediated proinammatory effects, further highlighting the relevance of CRP*-mediated complement activation (Supplementary Fig. 8).

Clearance of mCRP in circulation. Generating monomeric CRP by treating pCRP with 8 M urea in the presence of 10 mM EDTA for 2 h at 37 C or by heating pCRP for 5 min at 95 C in 0.1% SDS results in an unfolded protein similar in size, solubility, antigenicity and in vitro activity as the recombinant human mCRP C36A,C97A-double mutant protein31. The circular dichroism (CD) spectra for pCRP

10 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

motif (L/V-C-X(0-3)-YX(2-6)-R/K36) has been identied in CRP (residues 3547). In pCRP, part of the CHO-binding motif is exposed on the surface (residues 4247, coloured white in Fig. 7a) and is able to interact with CHO molecules in the lipid bilayer. PC, LPC, SM, PE and LPE lipid head groups also poke up into the pCRP subunit interface, where the phosphocholine (or phosphoethanolamine) amino group can break existing intersubunit interactions to form new interactions

between the lipid head group and residues at the subunit interface.

The increased CHO and LPC content of microvesicles, compared to their cells of origin (Fig. 3ad), increases the uid nature (CHO and LPC) and the curvature (LPC) of the lipid bilayer further exposing the phosphocholine and phosphoethanolamine head groups of PC, LPC, SM, PE and LPE lipids3739. Adjacent pCRP subunits (that is, CRP monomers)

a b

%

60

C1q immobilized

#

**

##

**

MV + pCRP

MV + mCRP

MV + pCRP + bisPC

MV

%

60 50 40 30 20 10

100

++

**

Signal (l) fluo corrected

50 40 30 20 10

80

60

C3-PE-positive MV

APC positive MV

40

20

0

0 20 40

Min

Control

0 NHS

HIS

pCRP pCRP + NHS

pCRP + HIS

0 C3c

iC3b

C3d

C3c

iC3b

C3d

C3c

iC3b

C3d

60 80

NHS pCRP + NHS mCRP + NHS

c 200

150

100

50

0

Co NHS Co pCRP NHS

**

***

THP1-MV

Anti-GAPDH

Relative ICAM-1

mRNA (fold)

3,000

2,000

1,000

0

Relative ICAM-1

mRNA (fold)

#

kDa

100

10

8 6 4 2 0

100

80 60 40 20

0

LPS co co pCRP NHS

NHS pCRP

70

35

40

Anti-ICAM1

co co NHS NHS

pCRP Jurkat MV

NHS pCRP

THP-1 MV

**

***

Relative VCAM-1

mRNA (fold)

THP1-MV

Relative VCAM-1

mRNA (fold)

kDa

LPS co co pCRP NHS

NHS pCRP

100

Anti-VCAM1

40

35

Anti-GAPDH

Co NHS Co pCRP NHS

THP-1 MV

NHS pCRP

co co NHS NHS

pCRP Jurkat MV

d

*

*

LPS+pCRP; 20 min

LPS+pCRP+bisPC; 20 min LPS+pCRP+bisPC; 60 min

100 m

LPS+pCRP; 60 min

Transmigrated cells-% of basal level

1,000

800

600

400

200

*

0 LPS LPS+pCRP LPS+pCRP+bisPC

bisPC

LPS LPS+pCRP LPS+pCRP+bisPC

bisPC

LPS LPS+pCRP LPS+pCRP+bisPC

bisPC

20 min 60 min 120 min

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

showing that microvesicle-bound pCRP* recruits C1q and is responsible for activating the complement system (Fig. 6a,b).

The neoepitope (residues 199206) lies at the pCRP intersubunit interface and is not available for anti-mCRP antibody 9C9 or 3H12 binding. Once the subunits dissociate to form pCRP* (Fig. 7e and Supplementary Fig. 9B,C), tryptophan uorescence emission studies indicate that Trp205 ips out (red position in Supplementary Fig. 9C) and becomes solvent exposed. The increase in Trp205 solvent accessibility is consistent with our tryptophan uorescence quenching data for pCRP bound to liposomes. The neoepitope is now accessible by the conformation specic anti-mCRP antibodies 9C9 and 3H12. In pCRP the C36C97 intrasubunit disulde bond (location is shown in Supplementary Fig. 9C) is protected from reduction, but once the neoepitope is exposed the disulde bond is vulnerable to reducing conditions (for example, addition of DTT). Breaking of the intrasubunit disulde bond results in the unfolding of the CRP subunit to produce the proinammatory mCRP, that is, the conformation of CRP becomes disordered (Supplementary Fig. 9A), and all of the CHO-binding motif (residues 3547) is now exposed. mCRP is eventually released and cleared from the circulation by the bodys protein quality control mechanisms (Supplementary Figs 9D and 10).

DiscussionWe have identied a novel, proinammatory mechanism of CRP (Fig. 8). This is supported by the following ndings:(1) circulating pCRP binds to the cell membrane of activated, but not resting, monocytes. Binding is observed in vitro and in vivo. (2) Stimulated cells release pCRP on LPC-enriched, proinammatory microvesicles. Binding is specic, and can be inhibited by Ca2 depletion and the CRP inhibitor 1,6-bis-

PC. (3) Microvesicle-bound pCRP undergoes structural changes, leading to the expression of neoepitopes (pCRP*). pCRP* is the major isoform of CRP in human-inamed tissue. (4) In contrast to pCRP, microvesicle-bound pCRP* exerts proinammatory properties. It activates the complement system, amplies the microvesicle-induced expression of cell adhesion molecules on endothelial cells and leads to increased recruitment of leukocytes to the site of tissue damage. (5) Molecular modelling conrms that activation of the complement system requires a structural change in the native protein, as complement factor C1q can be docked onto pCRP*, but not onto pCRP.

Binding of CRP to membrane phospholipids of apoptotic cells and Fcg receptors of resting leukocytes has been studied in

Figure 6 | Proinammatory effects of CRP and microvesicles. (a) Binding of CMFDA-labelled microvesicles (MV) to immobilized C1q was determined in the presence of pCRP, 1,6-bis-PC-pCRP and mCRP by repeated uorescence measurements. 1,6-bis-PC is labelled as bisPC. Shaded areas display s.e.m. (n 3). (b) Binding of complement C3 to MV was assessed by ow cytometry with anti-C3b-PE antibodies (n 5), which recognizes C3, C3b and iC3b or

three different C3 antibodies that recognize either the C3c part, the iC3b neoepitope or the C3d part, and a second APC-labelled antibody (n 3). pCRP*

led to signicantly increased C3 deposition that is present on the MV surface mainly in the form of iC3b. No deposition of complement C3 was observed in the absence of NHS or in the presence of heat-inactivated sera (HIS). Displayed are the means and s.e.m. P values were calculated with one-way analysis of variance (ANOVA). **Po0.01 to NHS, Po0.01 to HIS, #Po0.05 to mCRP NHS and ##Po0.01 to mCRP NHS. (c) Expression of VCAM-1 and

ICAM-1 by HUVECs was determined by qRT-PCR (quantitative real-time PCR) and western blot. pCRP NHS signicantly increased the expression of

ICAM-1 and VCAM1 on HUVECs in the presence of cell-derived MV. Compared to THP-1 MV (n 4), Jurkat MV led only to a small increase in ICAM-1 and

VCAM-1 expression (n 4). Displayed are the means and s.e.m. P values were calculated with one-way ANOVA. **P value o0.01 and ***P value o0.001

compared to NHS. #P value o0.05 compared to THP-1-MV. Uncropped images of western blots are shown in Supplementary Fig. 14. (d) pCRP signicantly increases the number of transmigrated leukocytes in LPS-induced inammation in rat cremasteric postcapillary venules. This effect can be masked by preventing pCRP dissociation with 1,6-bis-PC. Transmigration of leukocytes was analysed by intravital microscopy under superfusion with LPS(25 ng ml 1)intravenous injection of pCRP (25 mg ml 1) that had been preincubated with 1,6-bis-PC in some groups. After labelling with rhodamine 6G, the number of transmigrated leukocytes was quantied in a visual eld of 200 mm in length and of 100 mm in width in the immediate vicinity ofa postcapillary venule after 20, 60 and 120 min. Values are means.e.m. of six rats per group; P values were calculated with one-way ANOVA, *Po0.05. Images shown are typical examples for postcapillary venules under LPS and pCRP in the presence and absence of 1,6-bis-PC after20 and 60 min. Scale bar, 100 mm and applies to the whole panel.

are held together by nine salt bridges, three hydrogen bonds and ten nonpolar interactions. The pKa values of buried/partially buried aspartic and glutamic acid side-chains are signicantly raised from their intrinsic values of 3.9 and 4.2, respectively. The experimentally determined average pKa values for aspartic and glutamic acids in folded proteins are 3.51.2 (range 0.59.2) and 4.20.9 (range 2.18.8), respectively40. We have predicted the pKa of the acidic residues on the intersubunit interface to be 8.4, 3.9, 4.6, 3.7, 6.1 and 4.0 for Asp155, Asp169, Glu42, Glu101, Glu108 and Glu197, respectively. The acidic pH observed at sites of inammation and tissue injury (in vitro and in vivo, pH 3.65.7)4146 will weaken the electrostatic intersubunit interactions (that is, salt bridges and hydrogen bonds) by protonating the acidic aspartic and glutamic acid side chains and the increased curvature of the surface will also assist to break the intersubunit cohesive forces (electrostatic and nonpolar interactions). The uid motion of the microvesicle lipid bilayer provides a mechanical shear force and, coupled with the weakening of the pCRP intersubunit interactions, within 30 min on the surface pCRP starts to dissociate to pCRP* (ref. 13) as illustrated in Fig. 7e (refs 9,13,41). As the CRP subunits move apart, they initially maintain an overall pentameric shape and the neoepitope (residues 199206, coloured yellow in Fig. 7e and Supplementary Fig. 9B,C) become exposed13,39. The CRP subunits are only loosely associated with each other in pentameric pCRP*, with the electrostatic interaction distances being 44 . At this point, pCRP* has three main options: (1) to reform pCRP; (2) remain as pCRP* for interaction with complement C1q (or other interacting proteins) or (3) dissociate into mCRP (Supplementary Fig. 9BD).

Complement C1q is a hexameric protein consisting of three separate chains47. The interaction between pCRP* and complement C1q is known to occur via the C1q globular heads (Fig. 7fh). One of the collagen-like bres and globular head is shown in Fig. 7h. Site-directed mutagenesis has identied residues on the C1q globular head involved in CRP binding47,48. The CRP residues determined by site-directed mutagenesis to interact with C1q49 all lie on the surface of the interior of the pentameric ring. The cross-sectional diameter of the C1q globular head is B50 (Fig. 7f), too large to t into the interior void of pCRP (B30 diameter; Fig. 7c); it is however, able to force the loosely associated pCRP* subunits further apart and thereby interact with the interior void of pCRP* (Fig. 7g,h). The C1q globular head was manually docked to pCRP* and the previously identied interacting residues lie on the C1qpCRP* interface, consistent with our data

12 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

a

b

B face

pCRP B face

LPC

Q150 E138

L64 F66

A face

Ca2+

D60

E147

~100

~30

~108

~120

Figure 7 | Model of the pCRP conformational rearrangement to pCRP* and interaction of pCRP* with complement C1q. (a) The individual subunits of human pCRP (PDB ID: 1B09; ref. 7) have been displayed as a molecular surface and coloured to highlight the subunit boundaries. View shown is the membrane binding B face of pCRP, with phosphocholine (cream spheres) and Ca2 ions (black spheres) located in the ligand binding site on each subunit.

The exposed region of the CHO-binding motif (residues 3547) is coloured white in each subunit. The cross-sectional size of the pentamer is B100 (B10 nm). (b) Close-up view of a phosphocholine head group from a microvesicle LPC molecule (light grey, orange, red, blue and white sticks) bound in the ligand binding site of one pCRP subunit (cream cartoon). Some of the residues lining the ligand binding site are labelled (using one letter amino-acid codes) and shown as sticks, the two Ca2 ions are green spheres. Salt bridge interactions between the phosphocholine amine group and Glu81 (E81) are indicated by black dashed lines. (c) Interaction of pCRP with a model CHO:POPC bilayer (CHO shown as cream coloured sticks; POPC as light grey, orange, red, blue and white sticks). The ligand binding site on each pCRP subunit can bind to a phosphocholine head group of POPC as shown in (b) for an LPC molecule. View is from above, looking down onto the pCRP A face and showing the pentameric conformation. The diameter of the pentamer inner annular void is B30 (B3 nm). (d) Side view showing the interaction between the B face of pCRP and the POPC phosphocholine head groups in the bilayer. The conversion of pCRP to pCRP* is a reversible process. (e) pCRP* in a pentameric conformation, same view as in (c). As the individual

CRP subunits move apart, the neoepitope (residues 199206, coloured yellow) is exposed and available for anti-mCRP-specic antibody (9C9 or 3H12) binding. The cross-sectional size of the pCRP* pentamer is B108 (B10.8 nm). (f) The cross-section size of the complement C1q globular head group (PDB ID: 1PK6; ref. 47) is B50 (B5 nm), this is the C1q domain that interacts with pCRP*. (g) and (h) The globular head of C1q inserts itself into the inner annular void of pCRP* forcing the subunits further apart. The interacting residues lie on the inner annular surface of the pCRP* pentamer and on the

C-terminal surface of the C1q globular head group. The globular head group of C1q is physically unable to bind to pCRP. In (h) only one collagen-like bre is shown (labelled as C1q collagen-like stem). We used part of a free art image for the coiled spring in (h) from http://www.vecteezy.com

Web End =www.vecteezy.com .

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 13

E81

180

c d

pCRP

90

N61

pCRP A face

pCRP* A face

e

f

~50

C1q globular head

90

C1q stem N terminus

C1q Collagen-like stem

A face

g h

90

pCRP* pCRP*

A face

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

Conformational change from pCRP to pCRP*

Interaction of pCRP* with complement C1q

Release of pCRP on LPC-enriched microvesicles

Deposition of mCRP and complement cleavage products

Figure 8 | Model of CRPmicrovesicle interaction. pCRP (yellow) binds to the plasma membrane of activated monocytes. It is subsequently released on membrane-derived microvesicles (blue). Microvesicle-associated pCRP is converted to pCRP* (red). pCRP* can activate the complement system by binding C1q (light green) or dissociate into individual mCRP subunits.

detail8,11,50. Our analysis additionally reveals activation-induced pCRP binding to monocytes, which explains why pCRP is not proinammatory in the absence of existing tissue injury3,6,29.

Alterations of the LPC plasma membrane content, and membrane perturbations have been implicated in promoting pCRP binding to apoptotic cells11,13,16. We demonstrate binding to LPC-enriched microvesicles. pCRP attachment to microvesicles occurs via exposed phosphocholine or phosphoethanolamine head groups of phospholipids24,37. Blocking the binding sites in pCRP by 1,6-bis-PC, or removing the Ca2 ions from their binding sites, precluded a stable interaction between pCRP and the microvesicles. This suggests an interaction with the B face of pCRP, leaving the A face accessible for C1q binding.

CRP bound to monocytes maintains the native pentameric structure of circulating pCRP. In contrast, CRP bound to micro-vesicles maintains pentameric symmetry but also reacts with neoepitope-specic antibodies (pCRP*). This is in line with previous reports, which show expression of neoepitopes on circulating microvesicles in patients with acute coronary syndrome25.

The concept that an enhanced membrane curvature is essential for neoepitope expression of CRP had been postulated previously39. This would explain why CRP does not elicit a proinammatory response during the clearance of apoptotic bodies50. These vesicles are B15 mm in size and thus larger than microvesicles51. It is likely that pCRP* formation does not occur on apoptotic bodies because of their larger diameter, similar to intact cells. Thus, although pCRP binds to apoptotic fragments, the slight curvature of their membrane prevents the formation of pCRP*, and thus classical pathway activation. Still, some smaller apoptotic fragments might allow pCRP* formation and decoration of their surface with complement cleavage products. As pCRP* also recruits complement inhibitory factors, which lead

to the formation of iC3b, only small amounts of anaphylatoxins are created. In addition, the inherent properties of the vesicles seem to be the major factors that determine the pCRP* effects. Apoptotic microvesicles elicit only a minimal inammatory response, which is marginally augmented by pCRP* and complement. This is in contrast to the highly proinammatory actions of monocytic microvesicles (Fig. 6).

Consistent with the published literature, it is likely that pCRP binding to multivalent ligands induces structural alterations in the pentamer to produce pCRP* (ref. 52). In addition, acidic pH, as occurring at sites of inammation, can lead to exposure of buried epitopes9,41. pCRP* allows binding and activation of the complement system6,13,52. In reducing and/or Ca2 -depleted conditions often found at sites of tissue injury or inammation, pCRP* further dissociates into individual unfolded subunits (mCRP), which might contribute to the clearance of cell debris and activation of phagocytic cells4,16. As pCRP* is abundantly found in injured tissues, it might represent the inammatory active isoform that is later cleared as mCRP. Previous data showing deposition of mCRP in inamed tissue might therefore reect a snapshot of the end point of the proinammatory activation cascade that has taken place.

This is consistent with our data, which show that pCRP* is more potent in binding C1q and activating the classical complement pathway than mCRP. Still, mCRP maintains partial complement-activating activity and is a strong opsonin, facilitating efcient uptake of mCRP-decorated vesicles. These properties might explain the short half-life of mCRP in the circulation. Rather than reecting a conict with previous data, we hypothesize that the involvement of both isoforms represents different time points in the proinammatory cascade (Fig. 8)4,32. The data presented here conrm previous studies that the proinammatory properties of CRP are regulated by conformational changes4,13.

14 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

Besides the above-mentioned proinammatory effects, anti-inammatory properties have been attributed to pCRP. pCRP has been shown to protect mice against endotoxin shock53,54. Although the mechanisms of this controversy are not entirely clear, it is possible that the pro- or anti-inammatory properties of pCRP depend on the extent of pre-existing inammation. The intravital microscopy model that we adopted (Fig. 6d) uses only very low concentrations of locally applied LPS solution (25 ng ml 1). In contrast, mice in the abovementioned endotoxin shock model were challenged with much higher doses of LPS (15,00027,000 ng per mg body weight). Thus, pCRP might exert a dual role; it might aggravate low-level inammatory responses as seen in sterile tissue injury (or low-level LPS administration), and also reduce excessive inammation, as occurring in endotoxin shock.

Still, the inammatory response in an endotoxin shock is a complex interplay of pro- and anti-inammatory mechanisms. The compensatory anti-inammatory response syndrome, which leads to a systemic deactivation of the immune system, can lead to failure of proinammatory defense mechanisms against infectious organisms55. In this scenario, the ability of CRP to promote a proinammatory response would be benecial.

Taken together, our results unravel a mechanism of how CRP exerts its proinammatory effects in damaged tissue on a molecular level. The small-molecule inhibitor 1,6-bis-PC blocked binding of pCRP to microvesicles, inhibited pCRP-dependent C1q attachment and abrogated CRP-mediated leukocyte recruitment. Small-molecule CRP inhibitors may therefore be a promising approach for the treatment of inammatory conditions such as acute myocardial infarction and stroke2,4,20. However, the limited potency and pharmacokinetic properties of 1,6-bis-PC currently restricts its use as a therapeutic agent. Our data warrant further research towards the development of more potent and orally bioavailable CRP inhibitors. The potential benets for a broad scope of inammatory diseases, as well as various I/R injuries, make this a highly attractive novel drug class.

Methods

CRP preparations. Human pCRP was purchased from Calbiochem (Nottingham, UK) and thoroughly dialysed against Dulbeccos PBS (D-PBS) supplemented with 2 mM Ca2 . Structural and functional integrity of pCRP was veried by native gel electrophoresis and liposome binding studies. Commercial pCRP migrated at the same height as serum CRP, showed no contaminations with mCRP and bound in a Ca2-dependent manner to LPC-containing liposomes (Supplementary Fig. 11A,B). mCRP was generated by treating pCRP with 8 M urea in the presence of 10 mM EDTA for 2 h at 37 C or by heating for 5 min at 95 C in 0.1% SDS. mCRP was thoroughly dialysed in low salt phosphate buffer (10 mM Na2HPO4, 10 mM NaH2PO4, and 15 mM NaCl, pH 7.4)14,56.

Recombinant human C36A,C97A mCRP was puried from inclusion bodies using reversible anhydride modication57. To detect any bacterial contamination of CRP preparations, all reagents were tested for LPS contamination with a Limulus assay (Sigma-Aldrich, Taufkirchen, Germany) and it was found to be below the detection limit (0.125 U ml 1 or 0.01 ng ml 1 LPS).

Antibodies. Anti-mCRP-9C9 and anti-pCRP-8D8 antibodies were puried from hybridoma culture supernatants58. Anti-CRP antibody clone CRP-8, mouseIgG FITC-conjugated F(ab0)2 and human IgG-puried immunoglobulins were purchased from Sigma-Aldrich. CD16-FITC, CD32-FITC, CD64-FITC, CD14-PE, CD68-PE, CD80-FITC, CD163-FITC and respective control antibodies were from Miltenyi Biotec (Bergisch Gladbach, Germany). Anti-ICAM-1 (#4915) and NF-kB pathway antibodies (NF-kB Pathway Sampler Kit; #9936) were from Cell Signaling Technology (Danvers, MA, USA). Anti-VCAM-1-horseradish peroxidase (anti-VCAM-1-HRP) (E-10) and anti-GAPDH-HRP (0411) were purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Complement C3-PE antibody (clone 6C9), which reacts with human C3 as well as the breakdown products C3b and iC3b, was purchased from LifeSpan BioSciences (Seattle, WA, USA) and Alexa Fluor 488 Annexin V from Life Technologies (Carlsbad, CA, USA).

APC goat anti-mouse IgG was from BD Biosciences (San Diego, CA, USA). Anti-iC3b (neoepitope), anti-C3c (recognizes C3, C3b, iC3b, C3c) and anti-C3d

(recognizes C3, C3b, iC3b, C3d) antibodies were purchased from Quidel (San Diego, CA, USA).

Cells. Monocytes were isolated from peripheral venous blood of healthy human volunteers by density gradient centrifugation. All volunteers gave informed consent before enrolment and approval from the relevant ethics committee was obtained (ethics review boards, University of Freiburg, Freiburg, Germany; 343/13). Monocytes, THP-1 acute monocytic leukemia cells (source: DSMZ, Braunschweig, Germany) and Jurkat T-cell leukemia cells (source: Center for Chronic Immunodeciency, Freiburg, Germany) were cultured in 90% RPMI-1640 media, 10% FCS, 2 mM L-glutamine with 50 U ml 1 penicillin and 50 mg ml 1 streptomycin. Identity of cell lines was conrmed by Multiplex human Cell line

Authentication Test (Multiplexion, Heidelberg, Germany). HUVECs were bought from PromoCell (Heidelberg, Germany) and cultured in Endothelial Cell Basal Medium with SupplementMix (PromoCell), 10% FCS, 50 U ml 1 penicillin and 50 mg ml 1 streptomycin.

Intravital microscopy of rat cremaster muscle. Animal studies were in compliance with ethical regulations and approved by the institutional animal care committee (35-9185.81/G-10/114) at the University of Freiburg. Male Wistar rats (Charles River, Sulzfeld, Germany) weighing 120180 g (age: 56 weeks) were anaesthetized with 1.5 vol% isourane (Abbott, Wiesbaden, Germany). After tracheotomy, respiration was volume controlled (frequency, 3545 breaths per min; tidal volume, 4.55 ml; FiO2, 0.350.50; Servo Ventilator 900C, Maquet,

Rastatt, Germany). The right carotid artery and the left jugular vein were cannulated for continuous monitoring of mean arterial pressure, heart rate, blood gases and acid/base status. The right cremaster muscle was externalized overa heated aluminium platform and prepared for visualization of the microcirculation as previously described59. After coverage with a square coverslip, the cremaster muscle was continuously superfused at 1 ml min 1 with PBS-Ca2 -Mg2 warmed to 37 C. The tissue was then allowed to stabilize for 20 min. Digital intravital epiuorescence microscopy of the cremaster muscle was performed using a modied Zeiss microscope (Axio Scope A-1 MAT) equipped with a water immersion objective lens ( 20/1.0) and a Colibry 2 system for uorescence

records (Zeiss). Observations were recorded by a high-resolution digital camera (AxioCam MRm Rev. 3 FireWire, Zeiss). The cremaster muscle was then constantly superfused with indicated concentrations of LPS for 60 min. After20 min, an L-pCRP solution bolus (25 mg ml 1 serum concentration) was injected intravenously. pCRP was uorescently labelled with Alexa Fluor 594-labelled reactive dye that has a succinimidyl ester moiety reacting with primary amines of pCRP to form stable dye-protein conjugates (L-pCRP). Purication of the labelled protein and determination of the degree of labelling were performed according to the manufacturers protocol (Protein Labelling Kit; Life Technologies). For decameric stabilization, 1 mg ml 1 pCRP solution was incubated with 1,6-bis-PC in 1:100 molar ratio for 30 min at 37 C. Weight-dependent rat serum volume was calculated and CRP serum levels were veried by a particle-enhanced immunoturbidimetric assay (Modular Analytics; Roche Diagnostics, Basel, Switzerland). The uorescence signal was measured at 15 min intervals.

Leukocyte endothelial interactions and leukocyte transmigration in the cremaster muscle were observed by intravital microscopy as previously described4. Transmigrated leukocytes were measured after indicated times for 20 s in an area of 20,000 mm2 (200 mm length along the postcapillary venule and 100 mm depth into the cremasteric muscle tissue). To investigate the role of the complement system in CRP-mediated inammation in vivo, rats were treated with cobra venom factor (250 U kg 1 body weight via intraperitoneal injection) 24 h before intravital imaging. We used the AxioVision software (Zeiss, Jena, Germany) to measure transmigration of rhodamine 6G-labelled leukocytes.

Sample collection and preparation of tissue lysates. All patient recruitment was performed following approval from the relevant institutional human ethics committees (ethics review board, Freiburg, Germany, projects 24/13 and 67/08) in accordance with the Declaration of Helsinki. All patients gave informed consent before enrolment. Burned skin was collected from patients with deep second-degree to full-thickness burn wounds, who needed surgical excision on days48 following injury. Further patient characteristics can be found in Supplementary Table 1. Healthy control skin was taken from excess material of abdominoplasty procedures. Atherosclerotic plaques were taken from patients needing carotid endarterectomy and human skeletal muscle biopsies were taken from patients needing soft tissue reconstruction by means of free muscle transfer as previously described4. Tissue specimens were directly frozen in liquid nitrogen and storedat 80 C.

Frozen specimens were cut into small cubes and incubated in lysis buffer (25 mmol l 1 Tris-HCl (pH 7.4), 150 mmol l 1 NaCl, 2 mmol l 1 CaCl2,0.5% Triton X-100, 1 mM orthovanadate, 10 mg ml 1 leupeptine, 10 mg ml 1 aprotinin and 1 mM phenylmethylsulfonyl uoride) for 30 min on ice. Tissue was homogenized with an Ultra-Turrax disperser from IKA (Staufen, Germany) and centrifuged at 17,000 g for 10 min at 4 C to remove insoluble components. To assess the oxidation state of the intrasubunit disulde bond, 100 mmol l 1

NEM (Sigma-Aldrich) was added to the lysis buffer to stabilize reduced

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 15

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

CRP monomers. Protein concentrations were determined with a BCA Protein Assay Kit from Pierce (Rockford, IL, USA).

Histology of human skin. Staining was performed on native tissue sections16. After incubation with the primary antibody (clone CRP-8 1:200; anti-mCRP9C9 1:20; anti-pCRP 8D8 1:20) for 1 h at room temperature, slides were incubated with the undiluted HRP-labelled anti-mouse antibody for 30 min. Reaction products were stained with HistoGreen Substrate Kit for peroxidase from Linaris (Dossenheim, Germany) resulting in a green reaction product.

SDSPAGE, BNPAGE and western blot. For SDSPAGE, tissue lysates were precipitated with an equal volume of 10% trichloroacetic acid on ice. Protein pellets were denatured at 95 C for 5 min in SDS loading dye (with or without DTT) and separated on 15% SDSpolyacrylamide gels. Native electrophoresis was performed by BNPAGE as previously described60. Proteins were separated on 416% native PAGE Bis-Tris gels from Life Technologies (Darmstadt, Germany). Gels were subsequently incubated for 5 min at 95 C in 1x gel buffer containing 10 mM DTT and 1% SDS. After western blot, PVDF membranes were blocked in 5% milk powder in TBS-T and incubated with clone CRP-8 antibodies (1:500), followed by a HRP-conjugated anti-mouse antibody (1:5,000) in 1% bovine serum albumin (BSA) TBS-T. For western blot, we uniformly used anti-CRP-8 antibodies. Native gels (or 1/20 SDSPAGE gels) were always boiled before western transfer, which induces pCRP dissociation and thus enables a uniform transfer and detection of all CRP isoforms by anti-CRP-8 antibodies. These experiments are thus independent of conformation specicity of antibodies. Respective pCRP or mCRP controls conrm the feasibility of this approach. ECL Western blotting Detection Reagents (GE Healthcare, Buckinghamshire, UK) were used to visualize protein bands in a Fusion Fx7 chamber (Peqlab, Erlangen, Germany). Band intensity was determined with ImageQuant 5.2 (Molecular Dynamics, Sunnyvale, CA USA) and CRP binding was calculated relative to an input control.

NF-jB pathway screen. THP-1 cells (1 106 per ml) were incubated with

LPS (10 mg ml 1) for indicated time periods and the reaction stopped by putting the cells on ice. Cells were washed once in ice-cold PBS (520 g for 5 min) and lysed in RIPA buffer with protease/phosphatase inhibitors (Cell Signalling Technology). The lysate was briey sonicated and centrifuged at 14,000 g for 10 min. Supernatant uids were collected and the protein concentration determined by BCA assay. Equal amounts of protein were mixed with 2 SDS loading buffer and separated

on 12.5% SDSPAGE. Detection occurred after western blotting as described above; however, blocking of the PVDF membrane was performed with1% BSA TBS-T instead of milk powder due to the subsequent use of phosphor-ylation-specic antibodies.

Determination of CRP isoforms in tissue. pCRP and monomeric CRP were distinguished by two different methods. Electrophoresis on polyacrylamide gels containing reduced amounts of SDS can separate pCRP (pCRP*) from mCRPas described by Taylor and van den Berg28. The low SDS content preserves the pentameric structure of pCRP/pCRP* and allows its separation from mCRP due to its slower migration velocity on polyacrylamide gels. Tissue lysates were prepared as described above, loaded in native buffer and separated on813% polyacrylamide gels with 1/20 SDS content.

A method to differentiate between pCRP/pCRP* and mCRP has recently been published by Wang et al.27 Each CRP subunit contains a disulde bond, which is accessible to low concentrations of DTT in mCRP but not in pCRP/pCRP*. As reduced CRP subunits migrate slower on SDSPAGE than oxidized subunits,the accessibility of the intrasubunit disulde bond can be used to identify mCRP. Tissue lysates, control pCRP and mCRP were incubated in 25 mmol l 1 Tris-HCl (pH 7.4), 150 mmol l 1 NaCl and 2 mmol l 1 CaCl2 in the presence and absence of 10 mM DTT at 37 C for 2 h. The reaction was stopped by adding 100 mmol l 1

NEM for 15 min at 25 C, which stabilizes reduced disulde bonds during gel electrophoresis. Samples were precipitated with an equal volume of 10% trichloroacetic acid at 4 C. The pellet was washed once in ice-cold acetone, resuspended in non-reducing SDS-loading dye and separated on a 15% non-reducing SDSPAGE.

Microvesicle isolation. Microvesicles originate from cell membranes and exhibit a size of B1001,000 nm. THP-1 microvesicles were puried by differential centrifugation according to a recently published protocol with slight modications30. Briey, THP-1 cells (1 107 per ml) were treated with 10 mg ml 1

LPS (Escherichia coli serotype O127:B8, Sigma-Aldrich) and Jurkat cells with 1 mM staurosporine (Sigma Aldrich) for 5 h in FCS-free media. Cells were centrifuged at 500 g for 5 min at room temperature. The supernatant was centrifuged at 1,500 g for 15 min. The supernatant containing microvesicles was centrifuged at 16,000 g for 60 min at 4 C to pellet the microvesicles (P 16,000 g). Platelet microvesicles were puried as recently described61. Human polymorphonuclear leukocytes (PMNLs) were puried as described62. PMNLs (1 107 per ml) were

treated with 10 nM PMA or 1 mM fMLP (N-formylmethionyl-leucylphenylalanine) for 20 min at 37 C. Cells were double centrifuged at 4,000 g for

20 min at room temperature. The supernatant containing microvesicles was processed as described above for THP-1 microvesicles. Human monocytes were isolated from peripheral blood mononuclear cells (PBMCs) by negative selection (EasySep Human Monocyte Enrichment Kit; Stemcell Technologies, Kln, Germany). Puried monocytes (1 107 per ml) were incubated with 10 mg ml 1

LPS for 2 h. Microvesicles were puried as described for THP-1 cells.

All microvesicle preparations were washed twice with PBS supplemented with 2 mM Ca2 and stored at 80 C until use. The amount of microvesicles was

quantied by the total protein concentration determined by a BCA Protein Assay Kit. Microvesicles were highly pure, as analysed by ow cytometry (Supplementary Fig. 11C). Smaller vesicles, including exosomes (50100 nm), were isolated from the 16,000 g supernatant by further ultracentrifugation (TLA 100.2 rotor; Beckman Coulter) at 100,000 g for 60 min at 4 C (P 100,000 g), and of the 100,000 g supernatant (SN 100,000 g) at 400,000 g for 60 min at 4 C (P 400,000 g).

Flow cytometry. Human monocytes or THP-1 cells (2 106 per ml) were incu

bated with 100 mg ml 1 pCRP for indicated time periods in FACS buffer (PBS-Ca-Mg, 1% BSA) at 37 C. Cells were pelleted (520 g, 5 min) and the reaction stopped by resuspending the cells in ice-cold FACS buffer with 0.1% sodium azide. Cells were stained with anti-pCRP-8D8 (1:50) or anti-mCRP-9C9 (1:50) antibodies for 30 min. After washing, cells were incubated with 500 nM 4,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich) and a FITC-labelled anti-mouse F(ab)2 (1:200).

Cells were again washed and subsequently analysed by ow cytometry(BD LSR Fortessa Cell Analyzer). Only DAPI-negative cells were analysed to exclude non-viable cells. Microvesicles were stained as described for cells; however, each centrifugation step occurred at 16,000 g for 30 min. The microvesicle gate was dened by size, using 0.88, 1.1 and 1.34 mm beads (Spherotech, Lake Forest, USA), and the forward scatter threshold set to 400 to exclude gating exosomes.

Quantication of CRP on microvesicles from ST elevation myocardial infarction patients. All patient recruitment was performed following approval from the relevant institutional human ethics committees (ethics review boards, Alfred Medical Research and Education Precinct Project 516/13) in accordance with the Declaration of Helsinki. All patients gave informed consent before enrolment. We recruited patients who presented with acute ST elevation myocardial infarction and underwent primary percutaneous intervention of a coronary artery. The clinical inclusion criteria were age between 18 and 80 years, coronary artery disease status established by angiography and willing and able to provide informed consent. To enable sufcient time for serum pCRP to rise and enhance the probability of pCRP*/mCRP detection, all patients were enrolled 24 to 48 h after angiography. Healthy volunteers were recruited as controls.

Blood samples were collected using vacationer tubes with heparin as the anticoagulant. The blood was centrifuged at 3,000 g for 15 min within 30 min of phlebotomy. The plasma was carefully collected and centrifuged at 12,000 g for2 min. About 90% of the supernatant (platelet-free plasma) was carefully collected into fresh tubes and snap frozen in liquid nitrogen before storing them at 80 C.

Frozen samples were thawed in a 37 C water bath for 5 min, vortexed and then centrifuged at 3,000 g for 15 min. Once again B90% of the supernatant (platelet-free plasma) was carefully collected into fresh tubes and centrifuged at 14,000 g for 60 min. The supernatant was discarded and the remaining pellets (microvesicles) were reconstituted with PBS. The buffer was ltered twice through a 0.2 mm membrane lter before use. Twenty microliters of microvesicles were incubated with either 5 ml of anti-mCRP-9C9 or 5 ml of anti-pCRP-8D8. A measure of 0.2 ml of anti-mouse IgG-FITC antibody (Sigma, USA) was used as secondary antibody. Dual staining was performed by adding PE-conjugated anti-human CD41, CD11b or CD62P into the sample. Control samples included isotype controls and single stain samples. All ow cytometry analyses were performed using the LSRFortessa (BD Bioscience, USA) and the system is set to collection of all events for 90 s.

Complement deposition on microvesicles. THP-1 microvesicles (25 mg ml 1) were incubated with or without pCRP (40 mg ml 1) for 30 min at 37 C in25 mmol l 1 Tris-HCl (pH 7.4), 150 mmol l 1 NaCl, 2 mmol l 1 CaCl2. Ten per cent NHS or HIS (56 C, 45 min) was added for a further 15 min before micro-vesicles were pelleted for 30 min at 16,000 g at 4 C. Microvesicles were resus-pended in Annexin V-binding buffer and stained with C3-PE (1:200) and Alexa Fluor 488 Annexin V (1:200) for 30 min at 20 C. Subsequently, microvesicles were pelleted (16,000 g, 30 min, 4 C), resuspended in Annexin V-binding buffer and analysed by two-colour ow cytometry.

For C3c, iC3b and C3d analysis, THP-1 cells were labelled with CMFDA (Life Technologies) according to the manufacturers instructions before microvesicle release and purication. CMFDA-labelled microvesicles were treated as described above. Staining was performed with the respective C3 antibodies (1:200) and a second APC anti-mouse IgG (1:200). Microvesicle gates were dened as described under ow cytometry.

Quantication of microvesicles by ow cytometry. THP-1 cells were washed twice at 530 g for 5 min in PBS to remove contaminating microvesicles from the media and nally resuspended in Annexin V-binding buffer (BD Biosciences)

16 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

at a concentration of 1 106 per ml. Cells were transferred in FACS tubes and

Alexa Fluor 488 Annexin V (1:100) added. LPS (10 mg ml 1) and pCRP(40 mg ml 1) were added and cells were left at room temperature. Released microvesicles were analysed as described in ow cytometry. Microvesicles were dened as Annexin V-positive particles o1.1 mm.

Microvesicle release by M1/M2 macrophages. Human PBMCs were differentiated into M1 or M2 macrophages with PromoCell Macrophage Generation Media (PromoCell GmbH, Heidelberg, Germany) according to the manufacturers instructions. Differentiation into M1 or M2 macrophages was veried by ow cytometry with CD68-PE/CD80-FITC (M1 macrophages) and CD68-PE/CD163-FITC (M2 macrophages) according to the manufacturers recommendations. Target antibodies and control antibodies were obtained from Miltenyi Biotec. Cells were washed twice with PBS and subsequently incubated with PMA (50 ng ml 1), MPLA (1 mg ml 1), LPS (1 mg ml 1) or PBS in

Annexin-binding buffer supplemented with 10% FCS. The supernatant was collected at indicated time points and centrifuged at 1,500 g for 15 min to pellet contaminating cells and cell debris. Microvesicles were stained with AnnexinVFITC (1:200; BD Pharmingen) and counted by ow cytometry23. True count beads (BD Biosciences, San Jose, CA, USA) were used for absolute quantication.

Fluorescence microscopy. Cells were stained as described in ow cytometry and immobilized on adhesion microscope slides (Paul Marienfeld, Laura-Knigshofen, Germany) according to the manufacturers protocol. Cells were xed with2% formaldehyde, permeabilized with 0.05% PBS-Tween and mounted with Vectrashield Mounting Media with DAPI (Vector Laboratories, Burlingame,CA, USA). Image acquisition was performed by confocal microscopy(Leica TCS SP2 AOBS; Leica Microsystems, Wetzlar, Germany) and processed with the Leica Confocal Software. Magnication of the objective was 63, scanner

speed set to 800 Hz and the pinhole diameter was 1 airy unit. FITC was excited at 488 nm and DAPI at 405 nm.

Liposome- and microvesicle-binding studies. Lipids were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Large multilamellar liposomes were prepared as previously described63. Liposomes had the following compositions: PC liposomes (40% CHO, 60% PC); PC/LPC liposomes (40% CHO, 40% PC, 20% LPC); P 16,000 g liposomes (50% CHO, 7% Cer, 10% SM, 4% PS, 1% LPE, 1% LPC, 27% PC); SN 100,000 g liposomes (6% Cer, 11% SM, 6% LPC, 5% LPE,3.5% PS, 68.5% PC).pCRP of 40 mg ml 1 was incubated with 200 mg ml 1 liposomes or

200 mg ml 1 microvesicles (P 16,000 g) at 37 C for 1 h in 25 mM Tris-HCl(pH 7.4), 150 mM NaCl, 2 mM CaCl2 and 1 mg ml albumin. Ten millimoles of EDTA was added to inhibit Ca2 -dependent CRP binding. A specic small-molecule inhibitor of CRP, 1,6-bis-PC, was synthesized by Syngene International (Bangalore, India) and used at a concentration of 10 mM20. Liposomes and microvesicles were subsequently pelleted for 30 min at 110,000 g at 4 C. The pellet was resuspended in lysis buffer and separated on 416% Bis-Tris gels as described under BNPAGE. pCRP and mCRP controls were directly dissolved in lysis buffer.

Lipididomics. To determine the lipid prole of whole THP-1 cells, 1 106 per ml

cells were stimulated with LPS (10 mg ml) in RPMI-1640 media supplemented with 10% FCS for indicated time periods. Cells were then washed in PBS, resuspended in 20 mM Tris-HCl (pH 7.4), 500 mM NaCl and lysed with a tip sonicator. The protein concentration of the lysate was determined and adjusted to 1 mg per ml. THP-1 microvesicles were puried from FCS-free media, as FCS contains micro-vesicles and high concentrations of LPC. Purication was performed as described in microvesicle isolation; however, cells were only stimulated for 2 h and the cell concentration was adjusted to 2 106 per ml. Platelet microvesicles were

puried as described above. Pelleted microvesicles were washed in PBS, and nally resuspended in 20 mM Tris-HCl (pH 7.4) and 500 mM NaCl. The supernatant (200 ml) of the individual centrifugation steps was lyophilized and resuspendedin 10 ml of 20 mM Tris-HCl (pH 7.4) and 500 mM NaCl.

Samples underwent total lipid extraction, using a single-phase chloroform/ methanol (2:1) technique as described previously64. A 10 ml aliquot of celllysate, microvesicles or lyophilized cell supernatant was combined with 200 ml CHCl3/MeOH (2:1) and 15 ml of internal standard mix and then briey vortexed.

Samples were mixed (rotary mixer, 10 min), sonicated (water bath, 30 min) and then allowed to stand (20 min) at room temperature. Samples were centrifuged (16,000 g, 10 min) and the supernatant was dried under a stream of nitrogen at 40 C. The extracted lipids were then resuspended in 50 ml H2O saturated

BuOH with sonication (10 min), followed by 50 ml of 10 mM NH4COOH in MeOH. Extracts were centrifuged (3,350 g, 5 min) and the supernatant transferred into 0.2 ml glass vials with Teon insert caps. Mass spectrometric analysis was performed using 5 ml injections of the lipid extracts. Lipid analysis was performed by liquid chromatography, electrospray ionizationtandem mass spectrometry using an Agilent 1,200 liquid chromatography system combined with an Applied Biosystems API 4000 Q/TRAP mass spectrometer with a turbo ion-spray source (350 C) and the Analyst 1.5 data system. Quantication of lipids was based

on signal intensity relative to the corresponding internal standard as described previously65. Results were then given in pmol ml 1 or pmol mg 1. For lyophilized samples, background values of buffer or media were subtracted. Values displayed for each lipid class were calculated as the sum of each individual species within the glass.

ICAM-1 and VCAM-1 expression of HUVECs. HUVECs were cultured in 6-well plates. Conuent HUVECs were serum-starved for 4 h and then incubated with microvesicles (25 mg ml 1 of total protein), NHS (10%), pCRP (40 mg ml 1), mCRP (40 mg ml 1) or LPS (10 mg ml 1) for 2 h at 37 C in serum-free endothelial cell growth medium. Cells were then lysed in RLT buffer (Qiagen, Venlo, The

Netherlands), and RNA was isolated using the RNeasy Mini Kit (Qiagen). Total RNA was quantied by NanoDrop (Thermo Scientic, Waltham, MA, USA), and digested with DNase I (Life Technologies). Total RNA of 0.5 mg was reverse transcribed into cDNA using the AfnityScript cDNA Synthesis Kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturers instruction. mRNA levels were analysed by real-time PCR using ABsolute QPCR ROX Mix (Thermo Scientic). Primers and probes for human ICAM-1 (Hs00164932_m1), VCAM-1 (Hs01003372_m1) and RNA18S5 (Hs03928985_g1) were purchased from Applied Biosystems (Life Technologies). Amplication of RNA18S5 was used for normalization. All experiments are expressed as relative expression normalized to untreated HUVECs according to the 2 DCT method66. For western blots cells were treated as described above. However, incubation with the respective microvesicle preparation was conducted for 4 h. Cells were directly lysed in ice-cold RIPA buffer, processes as described under NF-kB pathway screen and stained with the respective antibodies against ICAM-1 (1:1,000), VCAM-1 (1:200) and GAPDH (1:200).

Real-time kinetic measurements. Kinetic measurements were obtained by direct detection of uorescently labelled compounds. C1q and pCRP were labelled with Alexa Fluor 488 (Life Technologies) and microvesicles with CMFDA, as described above. To monitor the interaction of uorescently labelled pCRP/C1q with unlabelled microvesicles, puried THP-1 microvesicles (0.5 mg ml 1) were spotted on a nitrocellulose membrane, left to dry for 2 min and subsequently blocked with 5% BSA in TBS-T. After washing, labelled pCRP/C1q was added at the indicated concentrations and repeated uorescence measurements of microvesicle-associated pCRP/C1q and respective reference area values were obtained by a LigandTracer Green with a FITC-compatible detector and analysed using Trace Drawer1.6 (Ridgeview Instruments AB, Vange, Sweden). The differential signal (microvesicle area minus reference area) becomes a background-corrected measure for the amount of pCRP/C1q attached to the microvesicles. When indicated,50 mM EDTA was added to the Alexa Fluor 488-labelled pCRP solution. To determine the binding of CMFDA-labelled microvesicles to unlabelled immobilized C1q, 60 mg ml 1 human C1q (Merck Chemicals GmbH, Darmstadt, Germany)

was spotted on a nitrocellulose membrane. If indicated, microvesicles(0.5 mg ml 1) were preincubated for 2 h at 37 C with pCRP (100 mg ml 1), mCRP (100 mg ml 1) or 1,6-bis-PCpCRP complexes (100:1 molar ratio 1,6-bis-PCpCRP) before they were added at a nal concentration of 20 mg ml 1.

Repeated uorescence measurements were obtained as described above.

Modelling. All modelling and geometry optimization was performed using SYBYL-X 2.1 (Certara LP, http://www.tripos.com

Web End =http://www.tripos.com ). LPC molecules were constructed using standard bond lengths and bond angles. The crystal structure of pCRP in complex with phosphocholine (PDB ID: 1B09; ref. 7) was manually docked to either a model 1:2 CHO:POPC lipid bilayer67 or LPC molecules by aligning the phosphocholine head groups of POPC or LPC to the crystal bound phosphocholine molecules. The crystal-bound phosphocholine ligands were then removed to create either the LPCpCRP complex or pCRPbilayer complex.

The complexes were geometry optimized for at least 2,000 iterations (or until the gradient of successive iterations was o0.05 kcal mol 1 ) using the molecular

mechanics MMFF94s force eld and partial atomic charges, and the conjugate gradient minimization method (all other parameters were at default values) to relieve any steric conicts that may have arisen during the docking or modelling process. The pentameric pCRP* model was constructed by manually movingthe individual subunits of pCRP apart until the interacting side chains were44 apart (Fig. 7e) or the globular head of complement C1q (PDB ID: 1PK6;

ref. 47) was able to dock inside the inner annular void, bringing the known interacting residues into contact (Fig. 7g)4749. The C1q-pCRP* complex was then manually docked to the model CHO:POPC lipid bilayer and geometry optimized using the protocol described above. A rotamer library of experimentally determined tryptophan side-chain conformations was used to place Trp205 in the pCRP* model into a solvent-exposed orientation (Supplementary Fig. 9C), consistent with previous tryptophan uorescence emission studies13. One subunit from the pCRP crystal structure was used to generate the unfolded mCRP molecule (Supplementary Fig. 9D). The number of salt bridges, hydrogen bonds and nonpolar interactions at the pCRP intersubunit interface were calculated using PISA v.1.51 (http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html

Web End =http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html). pKa predictions for the acidic residues located at the intersubunit interface were performed using

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 17

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188

the Depth web server (http://mspc.bii.a-star.edu.sg/depth/

Web End =http://mspc.bii.a-star.edu.sg/depth/) 68. Figures were created

using PyMOL version 1.7.4.1 (Schrodinger LLC; http://www.pymol.org

Web End =http://www.pymol.org ).

CRP uorescence analysis. Fluorescence quenching experiments were performed in 20 mM Tris (pH 7.4) containing 100 mM NaCl and 2 mM CaCl2 (TBSCa)

at a protein concentration of 5 mg ml 1. In the samples containing liposomes (40% CHO, 40% PC and 20% LPC), premade liposomes (see below for information on liposome preparation) were included at a nal concentration of 100 mg ml 1.

Samples were prepared by the addition of aliquots of a 2 M KI stock containing 10 mM Na2S2O3 with an equivalent stock of KCl used to correct for any salt-induced effects. Fluorescence intensity was measured on a Perkin-Elmer EnSpire plate reader with an excitation wavelength of 295 nm and an emission wavelength of 340 nm. Fluorescence quenching data were analysed as previously described69. Liposomes were prepared from Egg-phosphatidylcholine (eggPC) and LPC (Avanti Polar Lipid). The lipids were dissolved in chloroform/methanol (3/1, v v 1) and mixed in a ratio of 4:1 (eggPC:LPC; mol), were dried under nitrogen gas and then evaporated by vacuum overnight. The lipid was then hydrated with MilliQ water to a concentration of 200 mg ml 1 at 60 C,shaking them occasionally for 2 h. The lipid mixture was sonicated for 30 s intervals 10 times and passed through two pieces of polycarbonate membrane (0.2 mm pore size; Avanti Polar Lipids) with a Mini-Extruder (Avanti Polar Lipids). Thirty passes were performed with each membrane. Liposomes were stored at 4 C until use.

Microvesicle phagocytosis. Engulfment of microvesicles by human macrophages was analysed by a modied ow cytometric assay that has been described previously50. THP-1 cells were labelled with CMFDA according to the manufacturers instructions before microvesicle purication. Labelled microvesicles (25 mg ml 1) were incubated with 0.3 106 adherent macrophages for 2 h at 37 C

in the presence of 10% autologous human serum and 25 mg ml 1 of either pCRP or mCRP. After washing, cells were released, stained with anti-CD14-PE antibodies (1:100) and analysed by two-colour ow cytometry. Macrophages were gatedby forward, side scatter and CD14 positivity. Unlabelled microvesicles servedas a negative control. The percentage of macrophages that bound or ingested labelled microvesicles was calculated.

CD spectroscopy. Generating mCRP by treating pCRP with 8 M urea in the presence of 10 mM EDTA for 2 h at 37 C or by heating pCRP for 5 min at 95 C in0.1% SDS results in an unfolded protein similar in size, solubility, antigenicity and in vitro activity as the recombinant human mCRP C36A,C97A-double mutant protein31. The CD spectra of pCRP and recombinant C36A,C97A mCRP(both at 190 mg ml 1) were recorded at 25 C using a Jasco J-815 CD spectrometer equipped with a Peltier-type temperature control system (JASCO model

PTC-423S/15) and interfaced to a personal computer. The CD spectra were measured from 190 nm to 260 nm every 0.1 nm with 5 s averaging per pointand a 2 nm bandwidth. A 0.1 cm path length cell was used for obtaining the spectra. The CD spectra were signal averaged by adding four scans and baseline corrected. The difference spectrum was obtained by subtracting the averaged pCRP curve from the averaged mCRP curve with the Jasco suite of software.

Statistical analysis. Statistical analysis was performed with GraphPad Prism v.5.0 (GraphPad Software, La Jolla, CA, USA). For comparison of two groups,a two-tailed t-test was employed. A P value of o0.05 was considered statistically signicant. All experiments were performed at least three times. The data are expressed as meanstandard error of the mean (s.e.m.) or standard deviation (s.d.) when indicated. A one-way analysis of variance to compare effects of different treatments was used, if more than two groups were compared. In case of signicance, Tukeys test was used for pairwise comparison. Only signicant results for both analysis of variance and Turkeys test are presented.

Data availability. All relevant data are available from the authors on request.

References

1. Volanakis, J. E. Human C-reactive protein: expression, structure, and function. Mol. Immunol. 38, 189197 (2001).

2. Gill, R., Kemp, J. A., Sabin, C. & Pepys, M. B. Human C-reactive protein increases cerebral infarct size after middle cerebral artery occlusion in adult rats. J. Cerebr. Blood Flow Metab. 24, 12141218 (2004).

3. Lane, T. et al. Infusion of pharmaceutical grade natural human C reactive protein is not pro inammatory in healthy adult human volunteers. Circ. Res. 114, 672676 (2013).

4. Thiele, J. R. et al. Dissociation of pentameric to monomeric C-reactive protein localizes and aggravates inammation: in vivo proof of a powerful proinammatory mechanism and a new anti-inammatory strategy. Circulation 130, 3550 (2014).

5. Griselli, M. et al. C-reactive protein and complement are important mediators of tissue damage in acute myocardial infarction. J. Exp Med. 190, 17331740 (1999).

6. Agrawal, A., Gang, T. B. & Rusinol, A. E. Recognition functions of pentameric C-reactive protein in cardiovascular disease. Mediat. Inamm. 2014, 319215 (2014).

7. Thompson, D., Pepys, M. B. & Wood, S. P. The physiological structure of human C-reactive protein and its complex with phosphocholine. Structure 7, 169177 (1999).

8. Lu, J. et al. Structural recognition and functional activation of FcgammaR by innate pentraxins. Nature 456, 989992 (2008).

9. Singh, S. K. et al. Exposing a hidden functional site of C-reactive protein by site-directed mutagenesis. J. Biol. Chem. 287, 35503558 (2012).

10. Volanakis, J. E. & Narkates, A. J. Interaction of C-reactive protein with articial phosphatidylcholine bilayers and complement. J. Immunol. 126, 18201825 (1981).

11. Chang, M. K., Binder, C. J., Torzewski, M. & Witztum, J. L. C-reactive protein binds to both oxidized LDL and apoptotic cells through recognition ofa common ligand: phosphorylcholine of oxidized phospholipids. Proc. Natl Acad. Sci. USA 99, 1304313048 (2002).

12. Agrawal, A., Shrive, A. K., Greenhough, T. J. & Volanakis, J. E. Topology and structure of the C1q-binding site on C-reactive protein. J. Immunol. 166, 39984004 (2001).

13. Ji, S. R. et al. Cell membranes and liposomes dissociate C-reactive protein (CRP) to form a new, biologically active structural intermediate: mCRP(m). FASEB J. 21, 284294 (2007).

14. Mihlan, M. et al. Monomeric C-reactive protein modulates classic complement activation on necrotic cells. FASEB J. 25, 41984210 (2011).

15. Braig, D. et al. A conformational change of C-reactive protein in burn wounds unmasks its proinammatory properties. Int. Immunol. 26, 467478 (2014).

16. Eisenhardt, S. U. et al. Dissociation of pentameric to monomeric C-reactive protein on activated platelets localizes inammation to atherosclerotic plaques. Circ. Res. 105, 128137 (2009).

17. Strang, F. et al. Amyloid plaques dissociate pentameric to monomeric C-reactive protein: a novel pathomechanism driving cortical inammation in Alzheimers disease? Brain Pathol. 22, 337346 (2012).

18. Yang, X. W., Tan, Y., Yu, F. & Zhao, M. H. Interference of antimodied C-reactive protein autoantibodies from lupus nephritis in the biofunctions of modied C-reactive protein. Hum. Immunol. 73, 156163 (2012).

19. Slevin, M. et al. Modied C-reactive protein is expressed by stroke neovessels and is a potent activator of angiogenesis in vitro. Brain Pathol. 20, 151165 (2010).

20. Pepys, M. B. et al. Targeting C-reactive protein for the treatment of cardiovascular disease. Nature 440, 12171221 (2006).

21. Buzas, E. I., Gyorgy, B., Nagy, G., Falus, A. & Gay, S. Emerging role of extracellular vesicles in inammatory diseases. Nat. Rev. Rheumatol. 10, 356364 (2014).

22. Hugel, B., Martinez, M. C., Kunzelmann, C. & Freyssinet, J. M. Membrane microparticles: two sides of the coin. Physiology 20, 2227 (2005).

23. Jayachandran, M., Miller, V. M., Heit, J. A. & Owen, W. G. Methodology for isolation, identication and characterization of microvesicles in peripheral blood. J. Immunol. Methods 375, 207214 (2012).

24. Volanakis, J. E. & Wirtz, K. W. Interaction of C-reactive protein with articial phosphatidylcholine bilayers. Nature 281, 155157 (1979).

25. Habersberger, J. et al. Circulating microparticles generate and transport monomeric C-reactive protein in patients with myocardial infarction. Cardiovasc. Res. 96, 6472 (2012).

26. Shi, P. et al. Immunohistochemical staining reveals C-reactive protein existing predominantly as altered conformation forms in inammatory lesions. Acta Biol. Hungar. 65, 265273 (2014).

27. Wang, M. Y. et al. A redox switch in C-reactive protein modulates activation of endothelial cells. FASEB J. 25, 31863196 (2011).

28. Taylor, K. E. & van den Berg, C. W. Structural and functional comparison of native pentameric, denatured monomeric and biotinylated C-reactive protein. Immunology 120, 404411 (2007).

29. Pepys, M. B. et al. Isolation and characterization of pharmaceutical grade human pentraxins, serum amyloid P component and C-reactive protein, for clinical use. J. Immunol. Methods 384, 92102 (2012).

30. Wang, J. G. et al. Monocytic microparticles activate endothelial cells in an IL-1beta-dependent manner. Blood 118, 23662374 (2011).

31. Zouki, C., Haas, B., Chan, J. S., Potempa, L. A. & Filep, J. G. Loss of pentameric symmetry of C-reactive protein is associated with promotion of neutrophilendothelial cell adhesion. J. Immunol. 167, 53555361 (2001).

32. Li, H. Y. et al. Topological localization of monomeric C-reactive protein determines proinammatory endothelial cell responses. J. Biol. Chem. 289, 1428314290 (2014).

33. Motie, M., Schaul, K. W. & Potempa, L. A. Biodistribution and clearance of

125I-labeled C-reactive protein and 125I-labeled modied C-reactive protein in CD-1 mice. Drug Metab. Dispos. 26, 977981 (1998).

18 NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14188 ARTICLE

34. Christopeit, T., Gossas, T. & Danielson, U. H. Characterization of Ca2 and phosphocholine interactions with C-reactive protein using a surface plasmon resonance biosensor. Anal. Biochem. 391, 3944 (2009).

35. Mikolajek, H. et al. Structural basis of ligand specicity in the human pentraxins, C-reactive protein and serum amyloid P component. J. Mol. Recogn. 24, 371377 (2011).

36. Ji, S. R. et al. Monomeric C-reactive protein activates endothelial cells via interaction with lipid raft microdomains. FASEB J. 23, 18061816 (2009).

37. Bigay, J. & Antonny, B. Curvature, lipid packing, and electrostatics of membrane organelles: dening cellular territories in determining specicity. Dev. Cell 23, 886895 (2012).

38. Mouritsen, O. G. Lipids, curvature, and nano-medicine. Eur. J. Lipid Sci. Technol. 113, 11741187 (2011).

39. Wang, M. S., Messersmith, R. E. & Reed, S. M. Membrane curvature recognition by C-reactive protein using lipoprotein mimics. Soft Matter 8, 79097918 (2012).

40. Pace, C. N., Grimsley, G. R. & Scholtz, J. M. Protein ionizable groups: pK values and their contribution to protein stability and solubility. J. Biol. Chem. 284, 1328513289 (2009).

41. Hammond, Jr D. J. et al. Identication of acidic pH-dependent ligands of pentameric C-reactive protein. J. Biol. Chem. 285, 3623536244 (2010).

42. Silver, I. A., Murrills, R. J. & Etherington, D. J. Microelectrode studies on the acid microenvironment beneath adherent macrophages and osteoclasts. Exp. Cell Res. 175, 266276 (1988).

43. Sharpe, J. R. et al. Progression of wound pH during the course of healing in burns. J. Burn Care Res. 34, e201e208 (2013).

44. Schneider, L. A., Korber, A., Grabbe, S. & Dissemond, J. Inuence of pH on wound-healing: a new perspective for wound-therapy? Arch. Dermatol. Res. 298, 413420 (2007).

45. Ohkuma, S. & Poole, B. Fluorescence probe measurement of the intralysosomal pH in living cells and the perturbation of pH by various agents. Proc. Natl Acad Sci USA 75, 33273331 (1978).

46. Leake, D. S. Does an acidic pH explain why low density lipoprotein is oxidised in atherosclerotic lesions? Atherosclerosis 129, 149157 (1997).

47. Gaboriaud, C. et al. The crystal structure of the globular head of complement protein C1q provides a basis for its versatile recognition properties. J. Biol. Chem. 278, 4697446982 (2003).

48. Roumenina, L. T. et al. Interaction of C1q with IgG1, C-reactive protein and pentraxin 3: mutational studies using recombinant globular head modules of human C1q A, B, and C chains. Biochemistry 45, 40934104 (2006).

49. Bang, R. et al. Analysis of binding sites in human C-reactive protein for Fc{gamma}RI, Fc{gamma}RIIA, and C1q by site-directed mutagenesis. J. Biol. Chem. 280, 2509525102 (2005).

50. Gershov, D., Kim, S., Brot, N. & Elkon, K. B. C-reactive protein binds to apoptotic cells, protects the cells from assembly of the terminal complement components, and sustains an antiinammatory innate immune response: implications for systemic autoimmunity. J. Exp. Med. 192, 13531364 (2000).

51. Poon, I. K., Lucas, C. D., Rossi, A. G. & Ravichandran, K. S. Apoptotic cell clearance: basic biology and therapeutic potential. Nat. Rev. Immunol. 14, 166180 (2014).

52. Agrawal, A. Not only immunoglobulins, C-reactive protein too. Mol. Immunol. 56, 561562 (2013).

53. Xia, D. & Samols, D. Transgenic mice expressing rabbit C-reactive protein are resistant to endotoxemia. Proc. Natl Acad. Sci. USA 94, 25752580 (1997).54. Mold, C., Rodriguez, W., Rodic-Polic, B. & Du Clos, T. W. C-reactive protein mediates protection from lipopolysaccharide through interactions with Fc gamma R. J. Immunol. 169, 70197025 (2002).

55. Ward, N. S., Casserly, B. & Ayala, A. The compensatory anti-inammatory response syndrome (CARS) in critically ill patients. Clin. Chest Med. 29, 617625 viii (2008).

56. Potempa, L. A., Siegel, J. N., Fiedel, B. A., Potempa, R. T. & Gewurz, H. Expression, detection and assay of a neoantigen (Neo-CRP) associated witha free, human C-reactive protein subunit. Mol. Immunol. 24, 531541 (1987).

57. Potempa, L. A., Yao, Z. Y., Ji, S. R., Filep, J. G. & Wu, Y. Solubilization and purication of recombinant modied C-reactive protein from inclusion bodies using reversible anhydride modication. Biophys. Rep. 1, 1833 (2015).

58. Ying, S. C., Gewurz, H., Kinoshita, C. M., Potempa, L. A. & Siegel, J. N. Identication and partial characterization of multiple native and neoantigenic epitopes of human C-reactive protein by using monoclonal antibodies.J. Immunol. 143, 221228 (1989).59. Thiele, J. R., Goerendt, K., Stark, G. B. & Eisenhardt, S. U. Real-time digital imaging of leukocyte-endothelial interaction in ischemia-reperfusion injury (IRI) of the rat cremaster muscle. J. Visual. Exp. 66, e3973 (2012).

60. Swamy, M., Siegers, G. M., Minguet, S., Wollscheid, B. & Schamel, W. W. Blue native polyacrylamide gel electrophoresis (BN-PAGE) for the identication and analysis of multiprotein complexes. Sci. STKE 2006, pl4 (2006).

61. Diehl, P. et al. Microparticles: major transport vehicles for distinct microRNAs in circulation. Cardiovasc. Res. 93, 633644 (2012).

62. Nauseef, W. M. Isolation of human neutrophils from venous blood. Methods Mol. Biol. 1124, 1318 (2014).

63. Braig, D., Bar, C., Thumfart, J. O. & Koch, H. G. Two cooperating helices constitute the lipid-binding domain of the bacterial SRP receptor. J. Mol. Biol. 390, 401413 (2009).

64. Meikle, P. J. et al. Plasma lipidomic analysis of stable and unstable coronary artery disease. Arterioscler. Thromb. Vasc. Biol. 31, 27232732 (2011).

65. Weir, J. M. et al. Plasma lipid proling in a large population-based cohort.J. Lipid Res. 54, 28982908 (2013).66. Schmittgen, T. D. & Livak, K. J. Analyzing real-time PCR data by the comparative C(T) method. Nat. Protocols 3, 11011108 (2008).

67. Hong, C., Tieleman, D. P. & Wang, Y. Microsecond molecular dynamics simulations of lipid mixing. Langmuir 30, 1199312001 (2014).

68. Tan, K. P., Nguyen, T. B., Patel, S., Varadarajan, R. & Madhusudhan, M. S. Depth: a web server to compute depth, cavity sizes, detect potential small-molecule ligand-binding cavities and predict the pKa of ionizable residues in proteins. Nucleic Acids Res. 41, W314W321 (2013).

69. Lehrer, S. S. Solute perturbation of protein uorescence. The quenching of the tryptophyl uorescence of model compounds and of lysozyme by iodide ion. Biochemistry 10, 32543263 (1971).

Acknowledgements

This work was supported by a grant from the National Health and Medical Research Council of Australia (NHMRC) to K.P. and grants from the German Research Foundation (DFG) to S.U.E. (EI 866/1-1, EI 866/2-1, and EI 866/5-1), M.H.-L. (EI 866/5-1) and H.-G. K. (EI 866/2-1). Funding from the Victorian Government Operational Infrastructure Support Scheme to Baker IDI and St Vincents Institute is acknowledged. S.U.E. holds a Heisenberg fellowship of the German Research Foundation (DFG)(EI 866/4-1) and has been granted a Heisenberg professorship. X.W. is an Australian Heart Foundation Fellow, X.J.D., P.A.M., M.W.P. and K.P. are NHMRC Research Fellows.

Author contributions

D.B.study design, acquisition, analysis and interpretation of data, preparation of manuscript. T.L.N. acquisition, analysis and interpretation of data, preparation of manuscript. H.G.K. analysis and interpretation of data, preparation of manuscript. B.K.acquisition of data. X.W.acquisition of data, interpretation of data. J.R.T. acquisition of data. C.J.M.acquisition and analysis/interpretation of data. J.K.acquisition of data. L.A.P.interpretation of data, preparation of manuscript. N.A.M.acquisition and analysis of data. L.A.M.acquisition and analysis of data. X.J.D. interpretation of data, preparation of manuscript. J.Z.acquisition of data, preparation of manuscript. M.H.L.interpretation of data, preparation of manuscript. P.J.M.interpretation of data, preparation of manuscript. G.B.S.interpretation of data, preparation of manuscript. M.W.P.analysis and interpretation of data, preparation of manuscript. K.P.study design, analysis and interpretation of data, preparation of manuscript. S.U.E.study design, analysis and interpretation of data, preparation of 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: The 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: Braig, D. et al. Transitional changes in the CRP structure lead to the exposure of proinammatory binding sites. Nat. Commun. 8, 14188 doi: 10.1038/ncomms14188 (2017).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

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) 2017

NATURE COMMUNICATIONS | 8:14188 | DOI: 10.1038/ncomms14188 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 19