Structure

Article

Structural and Functional Analysisof HtrA1 and Its SubdomainsCharles Eigenbrot,1,* Mark Ultsch,1 Michael T. Lipari,2 Paul Moran,2 S. Jack Lin,2 Rajkumar Ganesan,2 Clifford Quan,2

Jeffrey Tom,2 Wendy Sandoval,3 Menno van Lookeren Campagne,4 and Daniel Kirchhofer2,*1Department of Structural Biology2Department of Early Discovery Biochemistry3Department of Protein Chemistry4Department of Immunology

Genentech, Inc., South San Francisco, CA 94080, USA

*Correspondence: dak@gene.com (D.K.), eigenbrot.c@gene.com (C.E.)DOI 10.1016/j.str.2012.03.021

SUMMARY

The homotrimeric human serine protease HtrA1 ishomologous to bacterial HtrA proteases regardingthe trypsin-like catalytic and PDZ domains butdiffers by the presence of an N-terminal domainwith IGFBP and Kazal homology. The crystal struc-tures and SAXS analysis presented herein reveal therare tandem of IGFBP- and Kazal-like modules, aprotease active site that adopts a competent confor-mation in the absence of substrate or inhibitor andamodel for the intact protein in solution. Highly sensi-tive enzymatic assays and binding studies demon-strate that the N-terminal tandem has no apparenteffect on protease activity, and in accordance withthe structure-based predictions, neither the IGFBP-nor Kazal-like module retains the function of theirprototype proteins. Our structures of the unligandedHtrA1 active site suggest two-state equilibriumand a ‘‘conformational selection’’ model, in whichsubstrate binds to the active conformer.

INTRODUCTION

The serine protease HtrA1 (PRSS11; Clan PA, family S1) belongs

to an evolutionarily conserved family of HtrA proteins (Clausen

et al., 2002, 2011). In humans, HtrA1, 3, and 4 share the same

domain architecture: an N-terminal IGFBP-like module and a

Kazal-like module, a protease domain with trypsin-like fold,

and a C-terminal PDZ domain. The identification of human

loss-of-function mutations, causing familial ischemic cerebral

small-vessel disease (Hara et al., 2009), suggested an important

role for HtrA1 in human physiology. The molecular mechanism

involves deficient TGF-b inhibition by HtrA1, resulting in

increased TGF-b signaling (Hara et al., 2009). Dysregulated

TGF-b signaling by aberrant HtrA1 expression may also

contribute to arthritic disease, perhaps in conjunction with

HtrA1-mediated degradation of various extracellular matrix

components (Chamberland et al., 2009; Grau et al., 2006;

Hadfield et al., 2008; Hu et al., 1998; Oka et al., 2004; Tocharus

et al., 2004; Tsuchiya et al., 2005) or indirectly via upregulation

of matrix metalloproteases (Grau et al., 2006). In addition, human

1040 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All ri

genetic studies identified a strong correlation between the

progression of age-related macular degeneration and a SNP in

the HtrA1 promoter region, which results in increasedHtrA1 tran-

script levels (Dewan et al., 2006; Yang et al., 2006).

The bacterial HtrA proteins DegP, DegS, and DegQ share

homology with HtrA1’s trypsin-like and PDZ domains, but they

lack the IGFBP and Kazal modules at their N-termini (Kim and

Kim, 2005; Meltzer et al., 2008). They form trimers, but DegP

can also form higher-order oligomers, including 24-mers (Krojer

et al., 2008). DegS is an allosteric enzyme with its PDZ domain

acting as a regulatory switch (Hasselblatt et al., 2007; Sohn

et al., 2007). The autoinhibited state was found to be relieved

by binding of small peptides to the PDZ domain or by domain

deletion (Sohn et al., 2007; Wilken et al., 2004). Enzyme regula-

tory function of the PDZ domain has been also reported for

DegP and the human HtrA1 and HtrA2 (Hadfield et al., 2008;

Martins et al., 2003;Meltzer et al., 2008;Murwantokoet al., 2004).

The IGFBP-like/Kazal-like tandem (referred to as the

N-domain) is rare and found only in HtrA1 and five other human

proteins (see The Family of N-Domain-Containing Proteins).

Kazal domains are tight binding inhibitors of serine proteases

(Lu et al., 2001) with three conserved disulfide bonds and are

exemplified by the pancreatic secretory trypsin inhibitor SPINK1

(Hecht et al., 1991), which was originally discovered by Kazal

et al. (1948). A loop between the second and third cysteines

binds to the protease active-site cleft and inserts its P1 residue

into the S1 specificity pocket (Hecht et al., 1991) in a fashion

similar to other standard-mechanism serine protease inhibitors,

such as the Kunitz domains (Arnoux et al., 1995; Kunitz and

Northrop, 1936). The IGFBP-like module of HtrA1 shares

sequence homology with the N-terminal lobe of the high affinity

IGF binding proteins IGFBP-1–IGFBP-6, which regulate IGF

signaling (Hwa et al., 1999). The N-terminal lobe of IGFBP

proteins is connected via a linker to the C-terminal lobe, and

both are usually required for high-affinity IGF binding (Bach

et al., 2005; Hwa et al., 1999; Sitar et al., 2006). A group of

proteins, including HtrA1,Mac-25 (Kato, 2000), the CCNproteins

Holbourn et al., 2008), and ESM-1 (Lassalle et al., 1996), all con-

tain segments with sequence homology to the N-terminal IGFBP

lobe of bona fide IGFBPs and have been categorized as low-

affinity IGF-binding proteins (Hwa et al., 1999; Kim et al., 1997).

Recent structural insight into the trimeric HtrA1 protease

domain suggested that catalytic activity is regulated by a

substrate-induced activation mechanism based on the finding

ghts reserved

A

B C D

Figure 1. HtrA1 Constructs and Their Enzy-

matic Activity

(A) Full-length HtrA1 (HtrA1), its catalytically in-

activated form HtrA1(S328A) and PDZ-deleted

form (HtrA_DPDZ), and the IGFBP/Kazal tandem

(N-domain) were expressed and purified from

insect cells. HtrA1 with N-domain deletion

(HtrA1_DN), the catalytic domain (HtrA1_Cat), and

its inactivated form HtrA1_Cat(S328A) were ex-

pressed and purified from Escherichia coli. All

catalytic domain-containing constructs eluted

as trimers by size-exclusion chromatography,

whereas the N-domain eluted as amonomer. Solid

and open diamonds indicate the positions of

disease-associated missense (A252T and V297M)

and nonsense (R302X and R370X) mutations,

respectively, as reported (Hara et al., 2009). See

also Table S1.

(B) Linear time-dependent cleavage of the fluo-

rescence-quenched peptide substrate H2-Opt

(Mca-IRRVSYSF(Dnp)KK) by increasing concen-

trations of HtrA1 (0.35–4 nM) versus 4 nM of

catalytically inactive HtrA1(S328A).

(C) Effect of N-domain deletion on H2-Opt

cleavage rates as a function of enzyme concen-

tration.

(D) Effect of N-domain deletion on time-dependent

enzyme inactivation by the active site-directed

irreversible inhibitor DFP. Calculated pseudo-first-

order rate constants k for inactivation of HtrA1 and HtrA_DN by DFP were 69.3 3 10�3 min�1 and 64.8 3 10�3 min�1, respectively. Error bars in (C) and (D)

represent the standard deviation of three independent measurements.

See also Table S1.

Structure

HtrA1 Structures

that the competent active-site conformation was only observed

when a peptidic substrate mimic was covalently bound to the

active site (Truebestein et al., 2011). Moreover, the study further

demonstrated that the PDZ domain does not regulate enzymatic

activity, raising the question of whether the N-domain could

serve as an activity modulator in lieu of the PDZ domain. There-

fore, we investigated the N-domain by structural and biochem-

ical approaches. The X-ray crystal structure of the N-domain at

2.0 A resolution showed an unprecedented arrangement of

two common folds. We report on the N-domain’s inability to

bind IGF-I or inhibit serine proteases, including HtrA1. Our

X-ray crystal structures of the apo-form of the HtrA1 protease

domain show a substantial degree of active-site conformational

plasticity. Contrary to a prior report, we see the enzymatically

competent active-site conformation in the absence of substrate,

which demonstrates that HtrA1 does not require substrate for its

activation. In addition, a low-resolution structure of full-length

HtrA1 was determined using small-angle X-ray scattering, which

allows approximate placement of the N-domain and PDZ

domains within the HtrA1 trimer.

RESULTS

Investigation of HtrA1 Activity Regulationby the N-DomainTo investigatewhether theHtrA1N-domainmay act as an activity

regulator in place of or in conjunction with the PDZ domain, we

established a sensitive enzymatic assay using the fluores-

cence-quenched peptide substrate Mca-IRRVSYSF(Dnp)KK

(H2-Opt), originally identified for the related HtrA2 (Martins

Structure 20, 10

et al., 2003), to measure catalytic activities of HtrA1 constructs

having N- and PDZ domain deletions (Figure 1A). H2-Opt was

an excellent substrate for HtrA1, allowing measurements at

pM HtrA1 concentrations (enzyme concentrations refer to the

HtrA1 monomer). Figure 1B shows the linear initial velocities of

H2-Opt cleavage with increasing enzyme concentrations (0.35–

4 nM). Replacing the catalytic serine with alanine (HtrA1(S328A))

completely abolished H2-Opt cleavage. Plots of the initial

velocities as a function of HtrA1 concentration showed a linear

relationship with a slope of 21.6 ± 0.2 nM product min�1 nM�1

enzyme (Figure 1C). The slope from HtrA1 lacking the N-domain

(HtrA1_DN) was identical (21.2 ± 0.4 nM product min�1 nM�1

enzyme), indicating that the N-domain neither enhanced nor

reduced HtrA1 enzymatic activity.

Diisopropyl fluorophosphate (DFP) is an irreversible active-site

inhibitor that forms a covalent bond with the catalytic serine and

occupies the S1 specificity pocket. It is an excellent probe to

investigate conformational states of active sites, because it

only binds to enzymatically competent active sites, as shown

for DegS (Sohn et al., 2007). Figure 1D shows that the pseudo-

first-order rate constants k for inactivation of HtrA1 and

HtrA1_DN by DFP were almost identical (69.3 3 10�3 min�1 vs.

64.8 3 10�3 min�1). In the absence of DFP, the enzyme activity

remained unchanged. The results suggested that the active

site readily adopts a competent conformation and that the

N-domain imparts neither activating nor inhibitory effects.

The recent study by Truebestein and colleagues (Truebestein

et al., 2011) used HtrA1 constructs lacking the N-domain to

assess the role of the PDZ domain for HtrA1 catalysis. Here,

we examined the role of the PDZ domain in the context of

40–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved 1041

Table 1. X-Ray Data Collection and Refinement for HtrA1

Structures

HtrA1_Cat

HtrA1_

Cat(S328A) N-Domain

Data collection ESRF ID-29 ALS 5.0.2 ALS 5.0.1

Space group P6322 P43212 P3121

Unit cell (A) a = b = 108.5,

c = 234.4

a = b = 151.9,

c = 87.54

a = b = 72.08,

c = 41.29

VM (A3/Dalton) 3.0 3.6 2.1

Resolution (A) 50–3.0 50–2.3 50–2.0

Rsyma,b 0.077 (0.732) 0.120 (0.691) 0.106 (0.423)

No. of observations 183,024 596,739 83,839

No. of unique

reflections

17,102 45,897 8,555

Completeness (%)b 99.9 (99.9) 100 (100) 99.2 (93.5)

I/sIb 27 (2.5) 19 (3.9) 19 (2.7)

Wilson B-factor (A2) 87 40 32

Refinement

Resolution (A) 50–3.0 50–2.3 50–2.0

No. of reflections 16292 44800 7676

Final Rc, RFREE 0.248 / 0.300 0.187 / 0.219 0.224 / 0.276

Protein residues 574 593 106

Solvent moleculesd 0 216 (8) 43 (5)

Atomse 4,381 (22) 4,923 (174) 831 (32)

Mean B-factor (A2) 82.4 39.8 34.2

rmsd bonds (A) 0.008 0.009 0.009

rmsd angles (�) 1.1 1.2 1.2

rmsd bonded Bs (A2) 3.6 4.2 4.1

No. of TLS groups 3 3 2

Ramachandran (%) 89 / 9 / 1 / 1 92 / 7 / 0 / 1 86 / 11 / 1 / 1aRsym = SkIj � j < I > k/Sj < I > j, where I is the intensity of a single obser-

vation and < I > the average intensity for symmetry equivalent observa-

tions.bIn parenthesis, for the highest resolution shell, of ten shells.cR = SjFo � Fcj/SjFoj, where Fo and Fc are observed and calculated

structure factor amplitudes, respectively. RFREE is calculated as R for

reflections sequestered from refinement.dIn parentheses, the number of nonwater buffer molecules.eIn parentheses, the number of atoms assigned less than unit occupancy.

shifted apart

clarity

A B

DC

90˚

N

C

N

C

C130

C112

R133

R137

R141

IGFBP-4

SPINK1

IGFBP-like module

-helix from Kazal-like module

R133

Figure 2. HtrA1 N-Domain Structure from X-Ray Crystallography

(A) The HtrA1 N-domain (IGFBP-like plus Kazal-like modules, residues Gly36

to Ala154, not including Gly94-Gln106; beige surface) is relatively thin in one

dimension.

(B) HtrA1 N-domain (beige) superposed with part of IGFBP-4 (blue; pdb 2DSP)

and with part of the SPINK1 Kazal domain (green; pdb 1TGS).

(C) The Kazal-like a helix is remarkably polar and presents its electropositive

face to the IGFBP-like module, where it carries a net negative electrostatic

potential (calculated within Pymol and colored to make +65 kT/e blue

and�65 kT/e red and shown after shifting the two fragments to separate them

by 8 A for clarity). A net negative electrostatic potential also may be present

in other IGFBP-like modules of the N-domain family, as they all carry at least

one glutamate residue homologous to one of the three glutamate side chains

within 4 A of the Kazal a helix from HtrA1 (Figure 5A).

(D) The interface between IGFBP-like and Kazal-like modules is dominated by

interactions between Arg side chains (Kazal-like module) and main-chain

carbonyl oxygens (IGFBP-like module).

Structure

HtrA1 Structures

full-length HtrA1 (HtrA1) containing the N-domain. Substrate

cleavage rates as a function of enzyme concentration for HtrA1

with and without the PDZ domain were 21.6 ± 0.2 and 19.7 ±

0.3 nM product min�1 nM�1 enzyme, respectively. Time-depen-

dent inactivation of HtrA1 with and without the PDZ domain by

the irreversible active site inhibitor DFP gave k values of

69.3 3 10�3 and 78.9 3 10�3 min�1, respectively. Thus, the

PDZ domain deletion (HtrA1_DPDZ) changed neither the enzy-

matic activity nor the rates of enzyme inactivation by DFP.

Structural and Functional Analysis of the N-Domainof HtrA1To gain insight into potential N-domain functions that might be

related to its predicted IGFBP- and Kazal-like modules, we

determined the crystal structure of the N-domain at 2.0 A

1042 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All ri

resolution (Table 1). The overall shape is a slightly curved, dis-

torted rectangle (Figure 2A). There are eight disulfide links—six

in the IGFBP-like module and two in the Kazal-like module.

Because of apparent disorder, the second Cys residue of the

last disulfide in the Kazal-like module is missing from the final

model. The HtrA1 IGFBP-like module faithfully reproduces the

fold of the IGFBP-4 N-terminal lobe as seen in the X-ray structure

of the IGF-I/IGFBP-4 complex (Sitar et al., 2006; Protein Data

Bank [pdb] 2DSP) albeit with a root-mean-square deviation

(rmsd) of 3.2 A (Figure 2B). In contrast, the structure-based

sequence alignment between IGFBP-4 and the HtrA1 N-domain

shows a poor correspondence for IGF-I binding residues and

makes an IGF-binding function for HtrA1 unlikely (Figure S2A

available online). We investigated whether the IGFBP-like

N-terminal module within the HtrA1 N-domain binds to IGF using

biolayer interferometry-based biosensors. Immobilized IGF-I

and IGF-II did not bind to HtrA1 N-domain (Figures 3A and 3B).

The same result was obtained by using a reversed experimental

setup, in which no IGF-I or IGF-II binding to immobilized

N-domain was detected (Figure 3C). Additional experiments

with insulin also showed a complete lack of N-domain binding

(data not shown).

ghts reserved

A

B

C

D

E

Figure 3. N-Domain neither Binds to IGF

nor Inhibits Protease Activities

(A–C) Representative sensorgram with analyte

proteins binding to sensor surfaces immobilized

with biotinylated (A) IGF-I, (B) IGF-II, and (C) HtrA1

N-domain (N-domain). The concentrations of the

soluble analyte proteins used were 500 nM

N-domain, 200 nM IGFBP-1 and IGFBP-3, 500 nM

IGF-I and IGF-II, and rabbit polyclonal anti-N-

domain antibody.

(D) Inhibition of trypsin-like proteases by N-

domain (5 mM), aprotinin (5 mM), or a1-antitrypsin

(A1AT, 5 mM; 20 mM for HtrA1 and HtrA1_Cat).

Enzyme activity was determined by measuring

the rates of cleavage of H2-Opt (for HtrA1 and

HtrA1_Cat) or synthetic paranitroanilide sub-

strates. Activities are expressed as a percentage

of the uninhibited enzyme. Error bars represent

standard deviations of three or more independent

measurements.

(E) Effect of N-domain on b-casein digestion.

b-casein (50 mg/ml) digestion by full-length HtrA1

(FL, 15 nM; left panel) or HtrA1_Cat (15 nM; right

panel) in the presence of N-domain (N, 5 mM) or

DFP (DFP, 7.5 mM) was analyzed after 2 hr at

37�C by SDS-PAGE. Arrows indicate b-casein

cleavage products; arrowhead shows intact b-casein and asterisk shows N-domain. Lanes labeled ‘‘FL – DFP’’ and ‘‘Cat – DFP’’ are formal controls, where all

buffer components used for addition of DFP were added but without DFP.

See also Figure S1.

Structure

HtrA1 Structures

The HtrA1 Kazal-like a helix packs against the IGFBP-like

module, mediated by highly polar interactions. The Kazal-like

module’s helix is remarkably amphipathic, and the polar, electro-

positive side packs against the IGFBP-like module (Figure 2C).

Arginine side chains from residues 133, 137, and 141 are within

H-bonding distance to main-chain carbonyl oxygen atoms in

loops of the IGFBP-like module (Figure 2D). The disulfide link

between Cys112 and Cys130 acts as an additional tether for

the Kazal-like module’s helix.

The HtrA1 Kazal-like module differs from canonical Kazal

modules (e.g., SPINK1) by a longer a helix (one turn more) and

the presence of only two—instead of three—disulfide bonds.

The disulfide pattern of a canonical Kazal module is preserved

for the (Cys#2/Cys#4) and (Cys#3,Cys#6) cystines (Figures 4A

and 5A). The remaining canonical disulfide, (Cys#1/Cys#5), is

lacking in HtrA1 because the first Cys is missing and Cys #50sposition and role is supplanted by Arg133. The activin-binding

protein follistatin also includes noninhibitory Kazal-like modules,

and these include a noncanonical position for Cys#5, which

aligns with HtrA1’s Arg137. Thus, Arg133 and Arg137 serve

a role analogous to Cys residues in SPINK1 and follistatin,

respectively, by linking the Kazal-like a helix to a preceding poly-

peptide segment (Figures 2B, 4A, 4B, and 5A). In SPINK1 and fol-

listatin, the linkage is to the N-terminal part of the Kazal module.

In the N-domain, the Arg-mediated linkages are to the IGFBP-

like module. In this light, we note that other than the covalent

peptide link connecting them (Val111 – Cys112), all interactions

between the N-domain’s IGFBP- and Kazal-like modules are

polar in nature.

Despite sequence and structure differences, the HtrA1 Kazal

superposes with inhibitory Kazal modules (e.g., SPINK1 [pdb

1TGS]) with a rmsd less than 1 A when these features are

Structure 20, 10

ignored. Of 41 residues within our Kazal-like domain, 30 Ca

atoms superpose on SPINK1 with an rmsd of 0.8 A (Figure 2B).

However, the HtrA1 Kazal has two features, which are inconsis-

tent with the protease interactions seen for true inhibitors. The

nominal ‘‘P1’’ residue of HtrA1 (Ser114) is significantly out of

position, as can be seen by comparison to the Rhodniin complex

with thrombin (van de Locht et al., 1995; Figures 4C and S2C). In

addition, the adjacent IGFBP-like module would cause a major

steric clash with the target protease. The inhibitory activity of

the HtrA1 Kazal-like module was examined by treating a panel

of trypsin-like proteases with soluble N-domain at a concentra-

tion of 5 mM. The results showed that it did not interfere

with protease activities, whereas the serine protease inhibitors

a1-antitrypsin and aprotinin displayed inhibitory activities ac-

cording to their known specificities (Figure 3D). Likewise, HtrA1

and HtrA1_Cat were not inhibited by the N-domain but were

inhibited by a1-antitrypsin, as reported (Hara et al., 2009; Hou

et al., 2005; Hu et al., 1998). Additional experiments showed

that, whereas DFP completely inhibited b-casein digestion by

either HtrA1 or HtrA1_Cat, the N-domain had no apparent inhib-

itory activity (Figure 3E).

The Family of N-Domain-Containing ProteinsThe small family of IGFBP/Kazal tandem (= N-domain)-contain-

ing human proteins includes HtrA1, HtrA3, HtrA4, Mac25, Kazal

D1, and IGFBPL1 (Figure 5B). In each case the IGFBP/Kazal

tandem forms the N-terminal portion of the proteins, followed

by a catalytic domain (HtrA proteins) or an Ig-like domain. Other

proteins, such as members of the CCN family and follistatin,

contain one of the modules—IGFBP—or Kazal-like module—

but never both. Based on protein sequence alignment (Fig-

ure 5A), we expect good overall structural homology within the

40–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved 1043

IGFBP-like EGF

A B

C

R137

C112 (Cys#2)

Cys#5

Cys#1 Cys#2

Cys#4

Cys#6

Cys#3

Thrombin catalytic triad

Asp

Ser His

Rhodniin P1 residue (His)

HtrA1 P1 residue (Ser114)

C119 (Cys#3)

C155 (Cys#6)

Ser114

Figure 4. N-Domain Structure Compared to Analogs

(A) The N-domain Kazal-like module uses Cys and Arg residues separated by

six amino acids to associate with the preceding IGFBP-like module. (Cys155 is

not observed because of disorder.)

(B) Follistatin (pdb 2ARP) uses two Cys residues separated by six amino acids

in an analogous fashion.

(C) Thrombin (gray) and Rhodniin’s Kazal1 (light blue) from the X-ray structure

of their complex (pdb 1TBR), showing the P1 residue from Rhodniin inserted in

the S1 pocket of thrombin, which blocks access of substrates to the active

site/catalytic triad. The N-domain (light brown), superposed on Rhodniin,

carries its ‘‘P1’’ residue on a loop, which tracks far from the S1 pocket,

consistent with a lack of inhibitory activity.

See also Figure S2.

Structure

HtrA1 Structures

‘‘N-domains.’’ The reported structure allows precise assignment

of the boundary between the two modules: there is no structural

correspondence between IGFBP-4 and the HtrA1 N-domain

after HtrA1 residue Cys110 and no structural correspondence

between inhibitory Kazal modules and the HtrA1 N-domain

before HtrA1 residue Val111. Therefore, the module boundary

in this family is identifiable by the Cys-X-Cys motif. The other

five N-domains share Kazal a helix sequences rich in Arg resi-

dues, mostly including ones homologous with Arg133 and

Arg137 and putative P1 residues that are either the same as in

HtrA1 (Ser) or other relatively uncommon P1 amino acids (e.g.,

Trp in HtrA3).

Enzymatically Competent and Incompetent Structuresof the Protease Domain Apo-FormWe report X-ray structures of the unliganded HtrA1 catalytic

domain with no added substrate mimic or inhibitor (apo protein;

1044 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All ri

Table 1). Our structures of the wild-type HtrA1 catalytic

domain (HtrA1_Cat, with protomers named chainA, chainB,

and chainD) and mutationally inactivated HtrA1 catalytic domain

(HtrA1_Cat(S328A), with protomers named chainX, chainY, and

chainZ) provide six crystallographically independent apo cata-

lytic domain conformations, and they differ in interesting ways

with respect to their apparent readiness to bind and process

substrate (summarized in Figure 6). A graphical summary of

crystal packing appears in Figure S3. For HtrA1_Cat, one proto-

mer (chainD) is relatively poorly ordered overall, and key residues

forming the oxyanion hole and S1 substrate specificity pocket

are not included in the final model. Protomers chainA and chainB

present two distinct conformations of the active-site region

(Figures 7A, 7B, and S4). For chainA, the side chain of the cata-

lytic triad residueHis220 is out of place relative to othermembers

of the triad (Ser328 and Asp250). Moreover, access to the S1

pocket is completely blocked by loop2, including Leu345, which

has a pronounced conformational difference relative to chainB

(see the next paragraph; Figures 7B and 7C). Residues 343–

348 do not adopt the b strand conformation observed for analo-

gous residues in active serine proteases and therefore cannot

form the usual main-chain H-bonds with substrate. This incom-

petent chainA conformation is essentially the same as seen for

apo HtrA1 structures (Truebestein et al., 2011) in pdb entries

3NUM and 3NWU.

For HtrA1_Cat chainB, the active-site residues are poised to

bind substrate and support amide bond cleavage. The catalytic

triad residues are properly positioned and Leu345 is now part of

an intact b strand and helps define the S1 pocket rather than

occlude it (Figures 7B, 7C, and S4). This chainB arrangement

is very similar to that observed in pdb entry 3NZI (Truebestein

et al., 2011), a structure of the HtrA1 protease with a substrate

surrogate covalently attached to Ser328. The rmsds for superpo-

sition of 190 Ca atom pairs and for 14 Ca pairs nearest the cata-

lytic site are 0.6 and 0.2 A, respectively (Figure 7D).

The three protomers in the HtrA1_Cat(S328A) structure

display the occluded S1 pocket seen for chainA and additional

features inconsistent with catalytic activity (see Supplemental

Results). As a measure of the low similarity of the loop2 confor-

mations of canonical and noncanonical forms of HtrA1 we

observe, the rmsd for superposition of Ca atoms of noncanonical

versions of residues 188–192 (chymotrypsinogen numbering

213–217) onto the canonical version is 1.723 ± 0.006 A. The anal-

ogous calculation for thrombin (pdb codes 1TBR and 1AFQ),

prostasin (pdb codes 3E0P and 3E1X), and HGFA (pdb codes

1YC0 and 1YBW) yields rmsd values of 0.819, 1.300, and

1.745 A, respectively.

Taken together, our apo HtrA1 protease structures reveal

large conformational differences in the active site region and

are easily classified as either incompetent or competent for

substrate binding. Because there are no large differences

between the number of H-bonds or of hydrophobic interactions

in the two conformations, and because of the disorder apparent

in chainD, we conclude that the two conformations are close in

free energy and therefore are both sampled frequently in

solution. We note that, whereas active sites of chainB and

chainD do not have intermolecular packing contacts likely to

have affected their conformations, chainA active sites have

contacts that are similar to those found for all three protomers

ghts reserved

Group 1

Group 2

IGFBP Kaz Catalytic PDZ

N L IGFBP family

HtrA1, HtrA3, HtrA4

Mac25, Kazal D1, IGFBPL1 Ig-like C2 type

N-domain Family

N-domain

IGFBP Kaz

IGFPB vWC

C

TSP-1 CT CCN family

EGF Kaz Follistatin repeat

B

A

Figure 5. N-Domain Family

(A) Sequence alignment of the six human proteins with the IGFBP-Kazal tandem. Included for reference are portions of IGFBP-4 N-lobe, Kazal (SPINK1), and

follistatin sequences, with canonical Kazal Cys residue linkages indicated with black lines. The pattern of disulfide links in HtrA1’s Kazal-like domain is unlike that

of SPINK1 and the follistatin Kazal-like domain for two reasons. First, the HtrA1 Kazal starts after the first canonical cysteine. Second, canonical Cys#50s role is

supplanted by Arg133, and follistatin’s noncanonical Cys#50s role is supplanted by Arg137. HtrA1-IGFBP-like residues within 4 A of the Kazal-like module (d)

include Glu residues, creating net negative charge.

(B) Domain arrangement of proteins with IGFBP- and Kazal-like modules (highlighted in green and blue, respectively). Not included here is ESM-1 (IGFBP-rP6),

which has an IGFPB-like domain followed by a nondefined region. Within the N-domain family, two groups of three proteins share identical domain architecture

(group 1: HtrA1, HtrA3, and HtrA4; group 2:Mac25, Kazal D1, and IGFBPL1). Group 2 differs from the group 1 in that the N-domain is followed by an Ig-like domain

(C2 type) instead of the catalytic/PDZ domains. Alternate names used for N-domain family members: HtrA1, IGFBP-rP5, and L56; Mac25, IGFBP-rP1, TAF, and

angiomodulin; IGFBPL1 and IGFBP-rP4. The six bona fide IGFPBs (IGFBP family) are composed of an N-terminal lobe (homologous to the IGFBP-like module of

the N-domain), a linker region (L), and a C-terminal lobe. The six CCN family proteins (cysteine-rich protein 61, connective tissue growth factor, and nephro-

blastoma overexpressed protein) have an N-terminal IGFBP-like domain, a von Willebrand factor type C repeat (vWC), a thrombospondin type-1 repeat (TSP-1),

and a cysteine-knot-containing module (CT) (except for CCN5).

Structure

HtrA1 Structures

of HtrA1_Cat(S328A) and the isomporphous pdb entry 3NWU.

However, we discount their influence in our analysis because

another recent HtrA1 catalytic domain structure, which lacks

such contacts (pdb 3NUM [Truebestein et al., 2011]),

has essentially the same incompetent conformation seen for

chainA.

Structure 20, 10

Low-Resolution Structure of Full-Length HtrA1by Small-Angle X-Ray ScatteringA low-resolution molecular envelope of intact Htra1 was

calculated using small-angle X-ray scattering (SAXS) analysis.

The central core was assigned to the protease trimer. Three

pronounced features protruding from the central core were

40–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved 1045

distorted disordered/missing residues

distorted distorted + detergent distorted + disordered

competent active site in apo protein

Y325

L345

D250

H220

S328

Y325

L345

D250 H220

A328

Y325

L345

D250 H220

A328

Y325

L345

D250 H220

A328

! -OG

glycerol

Y325

L345

D250 H220

S328

D250

H220

S1

oxyanion hole

HtrA1_Cat chainA HtrA1_Cat chainB HtrA1_Cat chainD

HtrA1_Cat(S328A) chainX HtrA1_Cat(S328A) chainY HtrA1_Cat(S328A) chainZ

Figure 6. Six Crystallographically Unique

Catalytic Domain Protomers

Together, these apo structures demonstrate

significant conformational variability of the HtrA1

protease active-site region, including one version

that is poised for substrate binding (chainB). The

label ‘‘distorted’’ refers to the position of Leu345,

although His220 and Ser328 also adopt confor-

mations inconsistent with catalysis. ‘‘b-OG’’ is

b-octylglucoside.

See also Figure S3.

Structure

HtrA1 Structures

assigned to PDZ domains so that there are no significant

contacts between protease and PDZ domains (Figure 8). This

arrangement is reminiscent of the relationship between protease

and PDZ domains in DegP (pdb 1KY9) and is consistent with the

lack of ordered electron density seen for the PDZ domain of

an HtrA1 protease-PDZ construct (pdb 3NUM). Our best fit to

the experimental scattering curve was obtained when three

N-domains were placed within the central core with their short

dimension roughly parallel to the protease 3-fold axis and allow-

ing for covalent attachment to the protease domain N-termini. In

this position, there are no major contacts between N-domains

and they are adjacent to protease domains on the ‘‘back’’ side,

opposite the active sites. Experimental and calculated radii of

gyration were in good agreement (52.2 and 52.5 A, respectively).

Examination of Substrate SpecificityLike HtrA2 (Martins et al., 2003), HtrA1 cleaved H2-Opt at the

Val-Ser peptide bond, consistent with strong sequence conser-

vation among the specificity-determining residues between the

two proteases. A panel of H2-Opt peptides with variable P1 resi-

dues was tested in enzymatic assays with HtrA1 and HtrA1_Cat

and cleavage sites confirmed bymass spectrometry. The results

indicated a clear P1 preference for small- and medium-sized

hydrophobic residues (Val, Leu, Met, Ala, and Ile) but also for

the two polar amino acids Thr and Gln (Figure S1). None of

the substrates with charged P1 residues were cleaved. There

was no difference in substrate specificity between HtrA1 and

HtrA1_Cat, suggesting that the accessory domains did not influ-

ence substrate interactions. A similar P1 preference was found

for the HtrA1-cleavable macromolecular substrates b-casein,

biglycan, decorin, IGFBP-5, and osteoadherin (Table S1). Only

the P1 residues Glu (decorin) and Tyr (IGFBP-5) were not

observed in the panel of H2-Opt substrates. The P1 specificity

1046 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved

profile is identical (except for Tyr from

IGFBP-5) to that reported recently by

Truebestein et al. (2011) and almost iden-

tical to that of HtrA2, which also has

strong preference for Val, Leu, Met, Ala,

Ile, Gln, and Ser (Martins et al., 2003).

DISCUSSION

The three human HtrA family members

HtrA1, 3, and 4 share an N-domain that

is notably absent in the well-described

bacterial homologs. Previous studies

recognized that the N-domain comprises two distinct modules

with sequences homologous to the N-terminal lobe of IGFBPs

and Kazal domains, respectively, giving rise to the notion

that the N-domain may recapitulate some functions associated

with these modules (Zumbrunn and Trueb, 1996). Furthermore,

a recent study indicated that, unlike bacterial HtrAs, the

HtrA1 catalytic activity may not be regulated by its PDZ

domain (Truebestein et al., 2011), raising the possibility of a regu-

latory role by the N-domain instead. The crystal structure

elucidated herein of the HtrA1 N-domain shows that it indeed

faithfully recapitulates the IGFBP- and Kazal-like folds, but

there are important structural deviations, which have functional

implications.

The Kazal-like module of HtrA1 is atypical in three important

aspects: it has only two instead of the typical three disulfide

bonds, the characteristic a helix is unusually long, and the

nominal P1 residue in the inhibitory loop is in a noncanonical

position. The missing disulfide bond per se does not preclude

protease inhibitory function, as shown by the two disulfide

bond-containing Kazal domain EPI1, which potently inhibits the

serine protease subtilisin A (Tian and Kamoun, 2005). Interest-

ingly, in HtrA1 the missing disulfide bond is replaced by H-bonds

from Arg133 of the Kazal-like module to main-chain atoms of

the IGFBP-like module. This H-bonding interaction and addi-

tional interactions from Arg137, also in the helix, strengthen the

connection between the two modules, thereby effectively re-

stricting access of the Kazal-like module to a target protease.

Indeed, in using relatively high concentrations of the N-domain,

we did not find inhibition of a diverse panel of serine proteases,

including HtrA1. The substitution of a disulfide bond, which is

present in both inhibitory and noninhibitory Kazal domains

(e.g., SPINK1 and follistatin), with H-bonding interactions in the

HtrA1 N-domain is probably also a feature of the other five

Y325

L345

D250

H220

S328

chainA distorted chainB competent

Y325

L345

D250

H220

S328

chainB with no substrate 3NZI with substrate

open S1

blocked S1

loopA

loop2

loop1

loopC

loopD

loopB

loop3

A

B

C

D

chainA

chainB

Figure 7. Competent and Incompetent

Conformations of the HtrA1 Protease

Domain from X-Ray Structures

(A) Apo protease domains have a loop2 confor-

mation that is inconsistent with substrate binding

(chainA [gray]) or that is poised to bind and

process substrate (chainB [orange]) Dotted

triangle and spheres represent Ca atoms of cata-

lytic triad residues His220, Asp250, and Ser328.

(B) Close-up view of active sites for chainA (gray)

and chainB (orange). The positions of His220 and

Leu345 in chainA are incorrect for substrate pro-

cessing. The large difference in the side chain of

Tyr325 is associated with an intermolecular

contact for chainA and is probably irrelevant.

Elements of the oxyanion hole for chainB are

illustrated as competent for catalytic activity, but

the resolution of our X-ray structure is too low to be

certain of this detail.

(C) The incompetence of chainA for substrate

binding arises from blockage of the S1 pocket

(green) by Leu345 (red).

(D) Active-site regions from the substrate-bound

structure (pdb 3NZI, light blue) and from chainB

(orange) show close correspondence of these

competent conformations. The large difference

between the two versions of Tyr325 is associated

with differences in their intermolecular contacts

and is probably irrelevant.

See also Figure S4.

Structure

HtrA1 Structures

human N-domain proteins. This prediction is based on the pres-

ence of a series of Arg residues in the Kazal a helix in all

N-domain proteins as well as the conserved pattern of Cys resi-

dues (Figure 5A). To the extent that these predicted interactions

preserve the relationship between IGFBP- and Kazal-like

modules, the other N-domain proteins would also lack inhibitory

activity.

We did not obtain any experimental evidence to suggest IGF-I,

IGF-II, or insulin binding activity for the N-domain. The IGFBP-

like module in HtrA1 is a reasonable facsimile of the homologous

domain from IGFBP-4 (Sitar et al., 2006) but only with respect to

the overall fold. The structural correspondence is quite imper-

fect, and the side chain character is very poorly analogous.

Despite initial evidence for interaction between another N-

domain family member, Mac25 (IGFBP-7) and IGF-I and IGF-II

(Oh et al., 1996), a sequence alignment of HtrA1 and Mac25

leads us to predict that no biologically relevant IGF-binding func-

tion pertains to Mac25 or other N-domain family members. This

notion is supported by a biosensor binding study that found no

appreciable Mac-25 interaction with IGF-I or IGF-II (Vorwerk

et al., 2002).

Another aspect of HtrA1 N-domain function is its potential

regulation of HtrA1 catalytic activity in the context of full-length

HtrA1. However, enzymatic assays with various HtrA1 con-

structs measuring activities and inhibitor incorporation rates

demonstrated that, like the PDZ domain, the N-domain is

dispensable for catalytic activity. Therefore, HtrA1 significantly

deviates from the bacterial HtrA paradigm in that neither of its

accessory domains, N-domain and PDZ, regulates catalytic

activity. This is further supported by the finding that the speci-

Structure 20, 10

ficity profile of a synthetic substrate panel was not influenced

by the accessory domains.

Our analysis of intact HtrA1 using SAXS yielded a saucer-like,

almost flat low-resolution molecular envelope that is consistent

with an HtrA1 trimer (Figures 8 and S5). In our best model ratio-

nalizing this envelope, the PDZ domains protrude from the

protease trimer and the N-domains lie flat against the side of

the protease trimer that is distal to the active sites, without

large contact areas between N-domains. If strong homotypic

interactions among N-domains exist in HtrA1, they might be

reasonably expected to be present in crystals of the N-domain.

The N-domains in our crystals are closely packed but are in an

arrangement that is inconsistent with integration into the

protease trimer (Figure S2B). Our molecular SAXS model does

not provide an intratrimer direct-contact mechanism by which

N-domains might influence protease activity, in accordance

with our biochemical results showing that the N-domain is

dispensable for activity. Additionally, the simple fact that the

N-terminal residues of the protease domains are found far from

the catalytic sites (Figure S3B) suggests N-domains are also

far from the catalytic sites in intact HtrA1. Our attempts to form

diffracting crystals of HtrA1_DPDZ or of intact HtrA1 were

unsuccessful.

Our finding of an apo catalytic domain with a competent

active-site conformation is contrary to the proposal made by

Truebestein et al. (2011) that HtrA1 is activated by substrate-

induced remodeling. Our experimental observations are

consistent with a low-energy barrier separating competent and

incompetent active-site conformers. First, the inactive con-

former would be unable to accept DFP inactivation because of

40–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved 1047

PDZ

PDZ PDZ

CAT

CAT

CAT

N

N

N

90˚

active site face

loopA

PDZPDZ PDZ

N N

Figure 8. Molecular Envelope of Full-Length HtrA1 from SAXS and

Fit of Protein Domains

Two views rotated by 90�, semitransparent spheres represent the SAXS

envelope, and three HtrA1 molecules are shown schematically in blue, green,

and magenta. ‘‘CAT’’ is catalytic domain; ‘‘N’’ is N-domain.

See also Figure S5.

Structure

HtrA1 Structures

the occluded S1 pocket. However, 10 mM DFP inactivates 90%

of HtrA1 in amere 35min, which is only slightly longer than the 5–

10 min for inactivation of PDZ-peptide activated DegS by 10 mM

DFP (Sohn et al., 2007). Therefore, the active HtrA1 conformer

must be frequently available for DFP interaction. Second, the

mild detergent b-octylglucoside was observed to distort an

a-helical portion at the active site, including the catalytic

His220, suggesting a pronounced conformational lability of the

active site. Furthermore, we and others have observed apo

forms of trypsin-homology proteases with active-site regions in

incompetent active-site conformations, quite different from

those seen in complex with pseudosubstrate inhibitors (Carvalho

et al., 2002; Fujinaga and James, 1987; Gohara and Di Cera,

2011; Huntington, 2008; Johnson et al., 2005; Shia et al., 2005;

Spraggon et al., 2009). Five of the six crystallographically

unique HtrA1 protease domain structures fall into this category.

They share certain features with other apo forms of trypsin-

homology proteases, such as a distorted loop2 and blockade

of the S1 pocket. Therefore, the observed HtrA1 conformations

are best explained by an equilibrium between active (E) and

inactive states (E*), the latter of which may, for HtrA1, formally

include more than one conformation. Such an equilibrium is

a regulatory mechanism commonly observed in trypsin-like

serine proteases (Gohara and Di Cera, 2011). A comprehensive

survey of trypsin homology proteases found that E and E* are

associated with two distinct conformational states of loop2

(220 loop in chymotrypsinogen numbering; Gohara and Di

Cera, 2011), similar to those observed in apo forms of HtrA1

and the related DegS (Sohn et al., 2010). Thus, in following the

allosteric equilibrium scheme of trypsin-like serine proteases

(Gohara and Di Cera, 2011), we propose that the substrate (S)

simply samples the active HtrA1 conformer (HtrA1) out of an

equilibrium with inactive conformers (HtrA1*) to form the enzy-

me:substrate complex (HtrA1:S) for catalysis and product (P)

formation:

HtrA1�4HtrA14+S

HtrA1 : S/HtrA1+P:

In this sense, we view HtrA1 as being subject to ‘‘conforma-

tional selection’’ by substrate (Boehr et al., 2009) rather than

‘‘substrate-induced remodeling,’’ although there may be limits

1048 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All ri

to the substrate and enzyme concentrations over which such

a distinction is useful (Hammes et al., 2009).

Our results suggest that neither the PDZ nor the N-domain of

HtrA1 influence the equilibrium state, raising the possibility that

there exists a yet unknown allosteric regulator of HtrA1 catalytic

activity. The substantial intrinsic catalytic activity of HtrA1 points

toward the possibility that such a putative regulator may serve

as an allosteric switch by stabilizing the inactive rather than the

active state of HtrA1. The HtrA1* 4 HtrA1 equilibrium also

suggests potential therapeutic approaches aimed at HtrA1

neutralization, in that stabilization of HtrA1* may fully inhibit

enzyme activity by a competitive inhibition mechanism (Gohara

and Di Cera, 2011).

Human HtrA1 is linked to a variety of human diseases. In the

studies described herein, we have sought structural and

biochemical insights to its function so that its role in disease

may be better understood. We have determined a low-resolution

envelope for the structure of intact full-length HtrA1, to our

knowledge the first structure of any ‘‘N-domain,’’ and provided

clear evidence that its enzyme active site can exist in a conforma-

tion competent for substrate binding even in the absence of

substrate. Our functional studies have shown that the N-domain

neither regulates HtrA1 enzyme activity nor interacts with IGF or

proteases. As such, our current understanding of HtrA1 function

and regulation is still insufficient to understand its role in patho-

logical, as well as physiological, processes. Among the most

glaring, remaining unknowns about HtrA1 is the means by which

its protease activity is regulated, which would be helpful in

understanding its role in disease.

EXPERIMENTAL PROCEDURES

Experimental procedures, covering crystallographic analysis of the N-domain,

HtrA1_Cat and HtrA1_Cat(S328A), SAXS analysis of HtrA1, sources of

reagents, production of HtrA1 proteins, synthesis of fluorescence-quenched

peptide substrates, enzyme assays, and measurements using biolayer inter-

ferometry appear in the Supplemental Experimental Procedures.

ACCESSION NUMBERS

Coordinates and structure factors for crystal structures of the N-domain,

HtrA1_Cat, and HtrA1_Cat(S328S) have been deposited at the Protein Data

Bank with accession codes 3TJQ, 3TJN, and 3TJO, respectively, for imme-

diate release upon publication.

SUPPLEMENTAL INFORMATION

Supplemental Information includes one table, five figures, Supplemental

Results, and Supplemental Experimental Procedures and can be found with

this article online at doi:10.1016/j.str.2012.03.021.

ACKNOWLEDGMENTS

We thank J. Corn for assistance in building models of the N-domain for use

as molecular replacement probes, Y. Franke for baculovirus constructs,

K. Mortara for insect cell expression, A. Schoeffler for help with SAXS analysis,

P. Liu for mass spectra, and K. Stengel, J. Murray, and staff at MXPress for

collection of diffraction data. Portions of this research were conducted at

the Advanced Light Source operated on behalf of the U.S. Department of

Energy, Office of Basic Energy Sciences. At the time this work was conducted,

all authors were employees of Genentech, Inc.

ghts reserved

Structure

HtrA1 Structures

Received: December 2, 2011

Revised: March 15, 2012

Accepted: March 30, 2012

Published online: May 10, 2012

REFERENCES

Arnoux, B., Merigeau, K., Saludjian, P., Norris, F., Norris, K., Bjørn, S., Olsen,

O., Petersen, L., and Ducruix, A. (1995). The 1.6 A structure of Kunitz-type

domain from the alpha 3 chain of human type VI collagen. J. Mol. Biol. 246,

609–617.

Bach, L.A., Headey, S.J., and Norton, R.S. (2005). IGF-binding proteins—the

pieces are falling into place. Trends Endocrinol. Metab. 16, 228–234.

Boehr, D.D., Nussinov, R., andWright, P.E. (2009). The role of dynamic confor-

mational ensembles in biomolecular recognition. Nat. Chem. Biol. 5, 789–796.

Carvalho, A.L., Sanz, L., Barettino, D., Romero, A., Calvete, J.J., and Romao,

M.J. (2002). Crystal structure of a prostate kallikrein isolated from stallion

seminal plasma: a homologue of human PSA. J. Mol. Biol. 322, 325–337.

Chamberland, A., Wang, E., Jones, A.R., Collins-Racie, L.A., LaVallie, E.R.,

Huang, Y., Liu, L., Morris, E.A., Flannery, C.R., and Yang, Z. (2009).

Identification of a novel HtrA1-susceptible cleavage site in human aggrecan:

evidence for the involvement of HtrA1 in aggrecan proteolysis in vivo.

J. Biol. Chem. 284, 27352–27359.

Clausen, T., Southan, C., and Ehrmann, M. (2002). The HtrA family of prote-

ases: implications for protein composition and cell fate. Mol. Cell 10, 443–455.

Clausen, T., Kaiser, M., Huber, R., and Ehrmann, M. (2011). HTRA proteases:

regulated proteolysis in protein quality control. Nat. Rev. Mol. Cell Biol. 12,

152–162.

Dewan, A., Liu, M., Hartman, S., Zhang, S.S., Liu, D.T., Zhao, C., Tam, P.O.,

Chan, W.M., Lam, D.S., Snyder, M., et al. (2006). HTRA1 promoter polymor-

phism in wet age-related macular degeneration. Science 314, 989–992.

Fujinaga, M., and James, M.N. (1987). Rat submaxillary gland serine protease,

tonin. Structure solution and refinement at 1.8 A resolution. J. Mol. Biol. 195,

373–396.

Gohara, D.W., and Di Cera, E. (2011). Allostery in trypsin-like proteases

suggests new therapeutic strategies. Trends Biotechnol. 29, 577–585.

Grau, S., Richards, P.J., Kerr, B., Hughes, C., Caterson, B., Williams, A.S.,

Junker, U., Jones, S.A., Clausen, T., and Ehrmann, M. (2006). The role of

human HtrA1 in arthritic disease. J. Biol. Chem. 281, 6124–6129.

Hadfield, K.D., Rock, C.F., Inkson, C.A., Dallas, S.L., Sudre, L., Wallis, G.A.,

Boot-Handford, R.P., and Canfield, A.E. (2008). HtrA1 inhibits mineral deposi-

tion by osteoblasts: requirement for the protease and PDZ domains. J. Biol.

Chem. 283, 5928–5938.

Hammes, G.G., Chang, Y.-C., and Oas, T.G. (2009). Conformational selection

or induced fit: a flux description of reaction mechanism. Proc. Natl. Acad. Sci.

USA 106, 13737–13741.

Hara, K., Shiga, A., Fukutake, T., Nozaki, H., Miyashita, A., Yokoseki, A.,

Kawata, H., Koyama, A., Arima, K., Takahashi, T., et al. (2009). Association

of HTRA1 mutations and familial ischemic cerebral small-vessel disease.

N. Engl. J. Med. 360, 1729–1739.

Hasselblatt, H., Kurzbauer, R., Wilken, C., Krojer, T., Sawa, J., Kurt, J., Kirk, R.,

Hasenbein, S., Ehrmann, M., and Clausen, T. (2007). Regulation of the sigmaE

stress response by DegS: how the PDZ domain keeps the protease inactive in

the resting state and allows integration of different OMP-derived stress signals

upon folding stress. Genes Dev. 21, 2659–2670.

Hecht, H.J., Szardenings, M., Collins, J., and Schomburg, D. (1991). Three-

dimensional structure of the complexes between bovine chymotrypsinogen

A and two recombinant variants of human pancreatic secretory trypsin inhib-

itor (Kazal-type). J. Mol. Biol. 220, 711–722.

Holbourn, K.P., Acharya, K.R., and Perbal, B. (2008). The CCN family of

proteins: structure-function relationships. Trends Biochem. Sci. 33, 461–473.

Hou, J., Clemmons, D.R., and Smeekens, S. (2005). Expression and character-

ization of a serine protease that preferentially cleaves insulin-like growth factor

binding protein-5. J. Cell. Biochem. 94, 470–484.

Structure 20, 10

Hu, S.I., Carozza, M., Klein, M., Nantermet, P., Luk, D., and Crowl, R.M. (1998).

Human HtrA, an evolutionarily conserved serine protease identified as a differ-

entially expressed gene product in osteoarthritic cartilage. J. Biol. Chem. 273,

34406–34412.

Huntington, J.A. (2008). How Na+ activates thrombin—a review of the func-

tional and structural data. Biol. Chem. 389, 1025–1035.

Hwa, V., Oh, Y., and Rosenfeld, R.G. (1999). The insulin-like growth factor-

binding protein (IGFBP) superfamily. Endocr. Rev. 20, 761–787.

Johnson, D.J., Adams, T.E., Li, W., and Huntington, J.A. (2005). Crystal

structure of wild-type human thrombin in the Na+-free state. Biochem. J.

392, 21–28.

Kato, M.V. (2000). A secreted tumor-suppressor, mac25, with activin-binding

activity. Mol. Med. 6, 126–135.

Kazal, L.A., Spicer, D.S., and Brahinsky, R.A. (1948). Isolation of a crystalline

trypsin inhibitor-anticoagulant protein from pancreas. J. Am. Chem. Soc. 70,

3034–3040.

Kim, D.Y., and Kim, K.K. (2005). Structure and function of HtrA family proteins,

the key players in protein quality control. J. Biochem. Mol. Biol. 38, 266–274.

Kim, H.S., Nagalla, S.R., Oh, Y., Wilson, E., Roberts, C.T., Jr., and Rosenfeld,

R.G. (1997). Identification of a family of low-affinity insulin-like growth factor

binding proteins (IGFBPs): characterization of connective tissue growth

factor as a member of the IGFBP superfamily. Proc. Natl. Acad. Sci. USA

94, 12981–12986.

Krojer, T., Sawa, J., Schafer, E., Saibil, H.R., Ehrmann, M., and Clausen, T.

(2008). Structural basis for the regulated protease and chaperone function of

DegP. Nature 453, 885–890.

Kunitz, M., and Northrop, J.H. (1936). Isolation from beef pancreas of

crystalline trypsinogen, trypsin, a trypsin inhibitor, and an inhibitor-trypsin

compound. J. Gen. Physiol. 19, 991–1007.

Lassalle, P., Molet, S., Janin, A., Heyden, J.V., Tavernier, J., Fiers, W., Devos,

R., and Tonnel, A.B. (1996). ESM-1 is a novel human endothelial cell-specific

molecule expressed in lung and regulated by cytokines. J. Biol. Chem. 271,

20458–20464.

Lu, S.M., Lu, W., Qasim, M.A., Anderson, S., Apostol, I., Ardelt, W., Bigler, T.,

Chiang, Y.W., Cook, J., James, M.N., et al. (2001). Predicting the reactivity of

proteins from their sequence alone: Kazal family of protein inhibitors of serine

proteinases. Proc. Natl. Acad. Sci. USA 98, 1410–1415.

Martins, L.M., Turk, B.E., Cowling, V., Borg, A., Jarrell, E.T., Cantley, L.C., and

Downward, J. (2003). Binding specificity and regulation of the serine protease

and PDZ domains of HtrA2/Omi. J. Biol. Chem. 278, 49417–49427.

Meltzer, M., Hasenbein, S., Hauske, P., Kucz, N., Merdanovic, M., Grau, S.,

Beil, A., Jones, D., Krojer, T., Clausen, T., et al. (2008). Allosteric activation

of HtrA protease DegP by stress signals during bacterial protein quality

control. Angew. Chem. Int. Ed. Engl. 47, 1332–1334.

Murwantoko, Y., Yano, M., Ueta, Y., Murasaki, A., Kanda, H., Oka, C., and

Kawaichi, M. (2004). Binding of proteins to the PDZ domain regulates proteo-

lytic activity of HtrA1 serine protease. Biochem. J. 381, 895–904.

Oh, Y., Nagalla, S.R., Yamanaka, Y., Kim, H.S., Wilson, E., and Rosenfeld, R.G.

(1996). Synthesis and characterization of insulin-like growth factor-binding

protein (IGFBP)-7. Recombinant human mac25 protein specifically binds

IGF-I and -II. J. Biol. Chem. 271, 30322–30325.

Oka, C., Tsujimoto, R., Kajikawa, M., Koshiba-Takeuchi, K., Ina, J., Yano, M.,

Tsuchiya, A., Ueta, Y., Soma, A., Kanda, H., et al. (2004). HtrA1 serine protease

inhibits signaling mediated by Tgfbeta family proteins. Development 131,

1041–1053.

Shia, S., Stamos, J., Kirchhofer, D., Fan, B., Wu, J., Corpuz, R.T., Santell, L.,

Lazarus, R.A., and Eigenbrot, C. (2005). Conformational lability in serine

protease active sites: structures of hepatocyte growth factor activator

(HGFA) alone and with the inhibitory domain from HGFA inhibitor-1B. J. Mol.

Biol. 346, 1335–1349.

Sitar, T., Popowicz, G.M., Siwanowicz, I., Huber, R., and Holak, T.A. (2006).

Structural basis for the inhibition of insulin-like growth factors by insulin-like

growth factor-binding proteins. Proc. Natl. Acad. Sci. USA 103, 13028–13033.

40–1050, June 6, 2012 ª2012 Elsevier Ltd All rights reserved 1049

Structure

HtrA1 Structures

Sohn, J., Grant, R.A., and Sauer, R.T. (2007). Allosteric activation of DegS,

a stress sensor PDZ protease. Cell 131, 572–583.

Sohn, J., Grant, R.A., and Sauer, R.T. (2010). Allostery is an intrinsic property of

the protease domain of DegS: implications for enzyme function and evolution.

J. Biol. Chem. 285, 34039–34047.

Spraggon, G., Hornsby, M., Shipway, A., Tully, D.C., Bursulaya, B., Danahay,

H., Harris, J.L., and Lesley, S.A. (2009). Active site conformational changes of

prostasin provide a newmechanism of protease regulation by divalent cations.

Protein Sci. 18, 1081–1094.

Tian, M., and Kamoun, S. (2005). A two disulfide bridge Kazal domain from

Phytophthora exhibits stable inhibitory activity against serine proteases of

the subtilisin family. BMC Biochem. 6, 15.

Tocharus, J., Tsuchiya, A., Kajikawa, M., Ueta, Y., Oka, C., and Kawaichi, M.

(2004). Developmentally regulated expression of mouse HtrA3 and its role as

an inhibitor of TGF-beta signaling. Dev. Growth Differ. 46, 257–274.

Truebestein, L., Tennstaedt, A., Monig, T., Krojer, T., Canellas, F., Kaiser, M.,

Clausen, T., and Ehrmann, M. (2011). Substrate-induced remodeling of the

active site regulates humanHTRA1 activity. Nat. Struct. Mol. Biol. 18, 386–388.

Tsuchiya, A., Yano, M., Tocharus, J., Kojima, H., Fukumoto, M., Kawaichi, M.,

and Oka, C. (2005). Expression of mouse HtrA1 serine protease in normal bone

1050 Structure 20, 1040–1050, June 6, 2012 ª2012 Elsevier Ltd All ri

and cartilage and its upregulation in joint cartilage damaged by experimental

arthritis. Bone 37, 323–336.

van de Locht, A., Lamba, D., Bauer, M., Huber, R., Friedrich, T., Kroger, B.,

Hoffken, W., and Bode, W. (1995). Two heads are better than one: crystal

structure of the insect derived double domain Kazal inhibitor rhodniin in

complex with thrombin. EMBO J. 14, 5149–5157.

Vorwerk, P., Hohmann, B., Oh, Y., Rosenfeld, R.G., and Shymko, R.M. (2002).

Binding properties of insulin-like growth factor binding protein-3 (IGFBP-3),

IGFBP-3 N- and C-terminal fragments, and structurally related proteins

mac25 and connective tissue growth factor measured using a biosensor.

Endocrinology 143, 1677–1685.

Wilken, C., Kitzing, K., Kurzbauer, R., Ehrmann, M., and Clausen, T. (2004).

Crystal structure of the DegS stress sensor: How a PDZ domain recognizes

misfolded protein and activates a protease. Cell 117, 483–494.

Yang, Z., Camp, N.J., Sun, H., Tong, Z., Gibbs, D., Cameron, D.J., Chen, H.,

Zhao, Y., Pearson, E., Li, X., et al. (2006). A variant of the HTRA1 gene

increases susceptibility to age-related macular degeneration. Science 314,

992–993.

Zumbrunn, J., and Trueb, B. (1996). Primary structure of a putative serine

protease specific for IGF-binding proteins. FEBS Lett. 398, 187–192.

ghts reserved