Proteases and Signaling

1
Proteases and Signaling

• Sherwin Wilk, Ph.D.
• Mount Sinai School of Medicine
• Department of Pharmacology Biological Chemistry
• Cell Signaling Systems Course

2
Phosphorylation and proteolysis are required in
the activation of NF-kB
Traenckner et al, EMBO J. 1994 Nov
1513(22)5433-5441.
3
Proteolysis is a hydrolytic reaction
4
Classes of proteolytic enzymes

1. Serine
2. Cysteine
3. Metallo
4. Aspartyl
5. Threonine

5
Protease-Activated Receptors
6
Fig. 1.   Mechanism of G protein-coupled receptor
(GPCR) activation by a reversible binding of a
soluble ligand, such as a neuropeptide, and
irreversible cleavage by a protease.
Déry et al, Am J Physiol. 1998 Jun274(6 Pt
1)C1429-1452.
7
Fig. 2.   A protein structure of
protein-activated receptor (PAR)-1, PAR-2, and
PAR-3. Amino acid sequences in NH2-terminus and
second extracellular loop that are important for
receptor activation are shown. Boxed residues
indicate tethered ligand domains (PAR-1, PAR-2,
and PAR-3) and anion binding sites (PAR-1 and
PAR-3). Arrows indicate cleavage sites. Bold
residues in second extracellular loop are
conserved.  Intron/exon border. Glycosylation
site. B genomic organization and chromosomal
localization of PAR-1 and PAR-2. Both genes
consist of 2 exons and 1 large intron. Exon
1 encodes NH2-terminal domains proximal to
cleavage sites, and exon 2 encodes the rest of
the receptors. Both receptors are localized
within 100 kb on chromosome 5q13.
Déry et al, Am J Physiol. 1998 Jun274(6 Pt
1)C1429-1452.
8
TABLE 1Structure/activity relationships for TRAPs TABLE 1Structure/activity relationships for TRAPs TABLE 1Structure/activity relationships for TRAPs TABLE 1Structure/activity relationships for TRAPs

Peptide Cell/Tissue Type Effect (EC50/IC50) References

SFLLRNPNDKYEPF (TRAP-14) Xenopus oocytes, platelets, rat aortic rings (endothelium denuded or intact), guinea pig gastric longitudinal smooth muscle, CCL39 hamster fibroblasts, rat glomerular mesangial cells, rat astrocytes 4-30 µM Chao et al., 1992 Coller et al., 1992 Kawabata et al., 1999c Vouret-Craviari et al., 1992 Vu et al., 1991a Yang et al., 1992
SFLLR-NH2 Platelets, transfected mammalian cells, endothelium denuded and intact RA, CCL39 fibroblasts 0.5-6 µM Ceruso et al., 1999 Hollenberg et al., 1996 Kawabata et al., 1999c Laniyonu and Hollenberg, 1995 Natarajan et al., 1995 Scarborough et al., 1992
NPNDKYEPF short peptides (lt5) PlateletsCCL39 fibroblasts, various species platelets, rat glomerular mesangial cells, endothelium denuded RA and gastric LM gt200 µMLoss of function Vassallo et al., 1992Albrightson et al., 1994 Bernatowicz et al., 1996 Chao et al., 1992 Connolly et al., 1994 Vouret-Craviari et al., 1992
Macfarlane et al, Pharmacol Rev. 2001
Jun53(2)245-282.
9
Substituted peptides
Acetyl-SFLLR-NH2 Platelets, transfected mammalian cells, rat gomerular mesengial, SH-EP cells gt1000 µM Albrightson et al., 1994 Coller et al., 1992 Sakaguchi et al., 1994 Scarborough et al., 1992
H-SFLLR-NH2 Platelets, transfected mammalian cells, SH-EP cells Low activity Scarborough et al., 1992 Shimohigashi et al., 1994 Van Obberghen-Schilling et al., 1993
XFFLR-NH2 Various cell types Charged amino acids not tolerated, size/shape important Thr substitution yields a PAR-1-specific peptide Bischoff et al., 1994 Ceruso et al., 1999 Chao et al., 1992 Hollenberg et al., 1997 Natarajan et al., 1995 Sakaguchi et al., 1994 Scarborough et al., 1992 Van Obberghen-Schilling et al., 1993 Vassallo et al., 1992 Yang et al., 1992
SXLLR-NH2 Various cell types Only aromatic residues tolerated Albrightson et al., 1994 Ceruso et al., 1999 Chao et al., 1992 Natarajan et al., 1995 Nose et al., 1993 Scarborough et al., 1992 Van Obberghen-Schilling and Pouyssegur, 1993 Vassallo et al., 1992
SFXLR-NH2 Various cell types No loss of activity, more active with 3-(2-naphthyl)-L-alanine Bischoff et al., 1994 Blackhart et al., 1996 Ceruso et al., 1999 Chao et al., 1992 Laniyonu and Hollenberg, 1995 Natarajan et al., 1995 Scarborough et al., 1992 Shimohigashi et al., 1994 Van Obberghen-Schilling and Pouyssegur, 1993 Vassallo et al., 1992
SFLXR-NH2 Various cell types Acidic and basic amino acids not tolerated Blackhart et al., 1996 Ceruso et al., 1999 Chao et al., 1992 Natarajan et al., 1995 Scarborough et al., 1992 Vassallo et al., 1992
SFLLX-NH2 Various cell types Wide range of residues tolerated, but reduced activity Blackhart et al., 1996 Ceruso et al., 1999 Chao et al., 1992 Hollenberg et al., 1997 Natarajan et al., 1995 Nose et al., 1998b Scarborough et al., 1992 Vassallo et al., 1992
Monocyclic SFLLRN analogues Platelets Significant loss of activity McComsey et al., 1999
(Table 1 continued)
Macfarlane et al, Pharmacol Rev. 2001
Jun53(2)245-282.
10
Improved agonists
S(p-F)FFLRNP Platelets, SH-EP cells 1.7 µM Nose et al., 1998a Nose et al., 1993 Shimohigashi et al., 1994
S(p-F)FpGuFLR-NH2 Platelets, smooth muscle, mouse fibroblasts 0.4 µM Bernatowicz et al., 1996
A(p-F)FRCha-HarY-NH2 Platelets 0.01-0.12 µM Ahn et al., 1997 Debeir et al., 1997 Feng et al., 1995 Kawabata et al., 1999c
A(p-F)FRChaCitY-NH2 Platelets 0.2 µM Kawabata et al., 1999c
S(p-F)FHarLRK-NH2 Xenopus oocytes 0.05-0.1 µM Blackhart et al., 1996
S(p-F)F2NaALR-NH2 Platelets, CHRF-288 membranes 80 nM Seiler et al., 1996
Macfarlane et al, Pharmacol Rev. 2001
Jun53(2)245-282.
(Table 1 continued)
11
Antagonists IC50
1-phenylacetyl-4-(6-guanidohexanoyl)-piperazine Platelets 50 inhibition of SFLLRNP Alexopoulos et al., 1998
1-(6-guanidohexanoyl)- 4-(phenylacetylamido-methyl)-piperazidine Platelets 40 inhibition of SFLLRN Alexopoulos et al., 1998
BMS-197525 N-trans(p-F)FpGuFLR-NH2 Platelets, smooth muscle, mouse fibroblasts 0.2 µM of SFLLRNP Bernatowicz et al., 1996
BMS-200261 N-trans(p-F)FpGuFLRR-NH2 Platelets 20 nM (SFLLRN) 1.6 µM (thrombin) Bernatowicz et al., 1996 Kawabata et al., 1999c
3-Mercapto-propionyl-FChaChaRKNDK-NH2 Platelets 0.7-6.4 µM (thrombin) Kawabata et al., 1999c Seiler et al., 1995
LVR(D-)CGKHSR Rat astrocytes 180 µM (3Hthymidine incorporation, by TRAP14 or thrombin) Debeir et al., 1997
Oxazole-30 Platelets, CHRF membranes 25 µM (thrombin), 6.6 µM (SFLLRN) Hoekstra et al., 1998
S(Npys)- Mp-(p-F)F-NHCH(C6H5)2 Platelets 52 µM (SFLLRNP) Fujita et al., 1999
S(Npys)- Mp-(p-F)F-NHCH2CH(C6H5)2 Platelets 54 µM (SFLLRNP) Fujita et al., 1999
RWJ-56110 series Platelets 0.34 µM (thrombin) 0.16 µM (SFLLRN) Andrade-Gordon et al., 1999
SCH 79797 Platelets, smooth muscle 70 nM (3HhaTRAP binding) Ahn et al., 2000
SCH 203099 Platelets, smooth muscle 45 nM (3HhaTRAP binding) Ahn et al., 2000
FR171113 Platelets 0.29 µM (thrombin) Kato et al., 1999
Macfarlane et al, Pharmacol Rev. 2001
Jun53(2)245-282.
(Table 1 continued)
12
Labels IC50/Kd
SFp-azido-FLRNPKGGK-biotin HEL cells No effect Bischoff et al., 1994
3HhaTRAP 3HA(p-F)FRChaHarY-NH2 Platelet membranes, platelets 0.15 µM, Kd  15 nM Ahn et al., 1997
A(p-F)FRChaHar(125I)Y-NH2 Platelets 0.03 µM Feng et al., 1995
BMS-200661 N-trans(p-F)FpGuFLROrn Platelets Kd  10-30 nM Bernatowicz et al., 1996
SFLLRNPNDKYEPF-biotin BHK cells Kd  3 µmol/l Takada et al., 1995
BMS-197525, 3H and biotinylated derivatives CHRF-288 cells Kd  80 nM or less Elliott et al., 1999
Amino acid residues X indicates amino acid scan
Har, homoarginine (pF)F, parafluorophenylalanine
Cha, cyclohexylalanine N-trans,
trans-cinnomoyl pGuF, p-guanidino-phenylalanine
Cit, citrulline Orn, ornithine 2NaA,
2-naphthylalanine Npys, S-3-nitro-2-pyridinesulph
enyl ß-Mp, ß -mercaptopropionyl 3HhaTRAP,
3HA(p-F)FRChaHarY-NH2. All EC50 values refer to
platelet aggregation unless indicated otherwise.
Likewise, all IC50 values refer to inhibition of
TRAP-induced platelet aggregation unless
indicated otherwise.
Macfarlane et al, Pharmacol Rev. 2001
Jun53(2)245-282.
(Table 1 continued)
13
FIG. 2. Structural and functional domains of
PARs. The figure shows alignment of domains of
human PAR1, PAR2, PAR3, and PAR4. A and B
mechanism of cleavage and interaction of the
tethered ligand with extracellular binding
domains. C functionally important domains in the
amino terminus, second extracellular loop, and
carboxy terminus. Conserved residues in loop II
are in bold. Adapted from Derian et al. (89) and
Macfarlane et al. (182).
Ossovskaya and Bunnett, Physiol Rev. 2004
Apr84(2)579-621.
14
FIG. 5. Summary of PAR1 signal transduction. PAR1
couples to Gi?, G12/13?, and Gq11?. Gi? inhibits
adenylyl cyclase (AC) to reduce cAMP. G12/13?
couples to guanine nucleotide exchange factors
(GEF), resulting in activation of Rho, Rho-kinase
(ROK), and serum response elements (SRE). Gq11?
activates phospholipase C?(PLC?) to generate
inositol trisphosphate, which mobilizes Ca2, and
diacylglycerol (DAG), which activates protein
kinase C (PKC). PAR1 can activate the
mitogen-activated protein kinase cascade by
transactivation of the EGF receptor, through
activation of PKC, phosphatidylinositol 3-kinase
(PI3K), Pyk2, and other mechanisms. G?? subunits
couple PAR1 to other pathways, such as activation
of G proteins receptor kinases (GRKs), potassium
channels (Ki), and nonreceptor tyrosine kinases
(TK). Modified from Coughlin (68).
Ossovskaya and Bunnett, Physiol Rev. 2004
Apr84(2)579-621.
15
Cells have capitalized on two highly specific
processes --- phosphorylation and ubiquitination
--- to control complex signal-transduction
pathways, says Maniatis. Theyve basically
exploited every means of regulating signaling at
their disposal.
16
The Ubiquitin-Proteasome System

• Ubiquitin

17
Ciechanover et al, J Biol Chem. 1980 Aug
25255(16)7525-7528.
18
Ciechanover et al, J Biol Chem. 1980 Aug
25255(16)7525-7528.
19
E1SH Ub ATP E1S-Ub PPi AMP
20
E1S-Ub E2SH ES2-Ub E1SH
21
E2-Ub E3-protein Ub-protein conjugate
E2 E3
22
Pickart, Cell. 2004 Jan 23116(2)181-190.
23
The Ubiquitin-Proteasome System

• The Proteasome

24
Wilk and Orlowski, J Neurochem. 1983
Mar40(3)842-849.
25
Groll et al, Nature. 1997 Apr 3386(6624)463-471.

26
Wilk and Orlowski, Arch Biochem Biophys. 2000 Nov
1383(1)1-16.
27
Ferrell et al, Trends Biochem Sci. 2000
Feb25(2)83-88.
28
Tanaka et al, Biochem (Tokyo). 1998
Feb123(2)195-204.
29
Impairment of signaling by proteolysis

• Anthrax lethal factor

30
Composition of anthrax toxin

• protective antigen
• edema factor
• lethal factor

31
Collier and Young, Annu Rev Cell Dev Biol.
20031945-70.
32
Figure 1 Stereo ribbon representation of LF,
coloured by domain. The MAPKK-2 substrate is
shown as a red ball-and-stick model, and the Zn2
ion is labelled. The RMSD (C?) between the cubic
and monoclinic crystal forms is 1.18 ?. The
principal differences lie in the position of the
helical bundle of domain IV, which undergoes a
rigid body shift of 2 ? relative to the ß-sheet,
and a smaller shift of the ß-sheet of domain I.
The third helical element of domain III is
invisible in the monoclinic crystals, but can be
seen in cubic crystals, albeit with high
B-factors. It is possible that this mobile
amphipathic helix has a role in the
membrane-inserting properties of LF observed in
vitro28. Figure prepared with MOLSCRIPT, RENDER
and RASTER3D2931.
Pannifer et al, Nature. 2001 Nov
8414(6860)229-233.
33
Tonello et al, Nature. 2002 Jul 25418(6896)386.
34
Montecucco et al, Trends Biochem Sci. 2004
Jun29(6)282-285.
35
Inhibition of signaling by the Yersinia effector
YopJ
36
Fig. 1. Profile of YopJ inhibition. YopJ is
delivered into the target host cytosol via a type
III secretion system. YopJ blocks activation of
the superfamily of MAPK kinases, including MKKs
(which activate the MAPK pathways), and IKKß
(which activates the NF?B pathway). The
inhibition results in the inability of the cell
to produce cytokines and anti-apoptotic
machinery. The possibility exists that YopJ may
directly activate the cell death machinery, as
denoted by the red arrow with the question mark.
Orth, Curr Opin Microbiol. 2002 Feb5(1)38-43.
37
Fig. 2. Amino acid sequence alignment of AVP, the
family of YopJ homologues and Ulp1. The figure
shows alignment of the catalytic core of cysteine
proteases with the catalytic triad (denoted in
red) and identities or similarities (outlined and
shaded light blue). The alignment includes
(protein accession numbers in parentheses) AVP
(2781331) YopJ (P31498) AvrA (AAB83970) AvrBst
(AAD39255) AvrRxv (AAA27595) AvrXv4 (AAG39033)
AvrPpiG1 (CAC16700) ORF5 (AAF71492) ORFB
(AAF62400) Y4LO (P55555) and Ulp1 (NP_015305).
Orth, Curr Opin Microbiol. 2002 Feb5(1)38-43.
38
Fig. 3. Modification of proteins by ubiquitin or
SUMO. Ubiquitin and SUMO are linked to target
proteins by an isopeptide bond. The carboxyl
terminus of these modifying proteins (blue) is
conjugated to the -amine of a lysine residue
(LYS) in the target protein. Ubiquitin, unlike
SUMO, can conjugate more monomers to itself to
form a polyubiquitin chain. De-ubiquitinating
enzymes (DUB) and ubiquitin-like protein
proteases (for example, Ulp1) can cleave the
isopeptide bond in the ubiquitin-conjugates
(green arrows) or SUMO-conjugates (pink arrow),
respectively.
Orth, Curr Opin Microbiol. 2002 Feb5(1)38-43.
39
Fig. 4. Similarities between two reversible
post-translational modifications. Phosphorylation
is shown on the left and ubiquitination is shown
on the right. Addition of phophorylation (P) by
kinases and ubiquitin (Ub) by the E1E2E3
conjugation machinery to target proteins is shown
in blue. The requirement for energy is noted by
the purple ATP. The removal of phosphorylation by
a phosphatase or of ubiquitin by a
de-ubiquitinating enzyme is shown in red. The
target protein (green) can cycle from an
unmodified to a modified state. The model can be
used for ubiquitin-like proteins by replacing the
modification with, for example, SUMO or NEDD8,
and replacing the protease with a ubiquitin-like
protein protease.
Orth, Curr Opin Microbiol. 2002 Feb5(1)38-43.
40
Regulated Intramembranous Proteolysis

• Notch Pathway

41
Mumm and Kopan, Dev Biol. 2000 Dec
15228(2)151-165.
42
Fig. 2. Presenilin and presenilin-like MpMPs
regulate cleavage of diverse substrates. IP by
?-secretase (presenilin) requires prior JP by
either ß-secretase (shown) or ?-secretase (not
shown). The combination of ß- and ?-secretase
cleavages produce the Alzheimer's
disease-associated peptide Aß, whereas the
combination of ?- and ? -secretase cleavages
produce a peptide known as P3 (not shown), whose
role in the disease process in currently unknown.
Both sequential cleavage events produce CTF?
also known as APP intracellular domain (AID or
AICD), which can translocate to the nucleus to
form a transcriptionally active complex with Fe65
and Tip60. The nuclear targets of the CTF? are
not known. Although a single ?-secretase
cleavage is shown, ?-secretase cleaves the
transmembrane domain of APP at multiple sites
both near the middle of the membrane and near the
cytoplasmic face of the membrane.
Golde and Eckman, Sci STKE. 2003 Mar
042003(172)RE4.
43
Fig. 3. The presenilin ?-secretase appears to
be an aspartyl MpMP. The eight-transmembrane
domain model of PS is shown. Although the
topology of the first six transmembrane domains
is accepted, the exact topology of the
COOH-terminal transmembrane remains
controversial. The red and orange stars indicate
the location of the conserved active site
residues within the transmembrane domains 6 and
7. Mutation of either of these aspartates (D)
results in a dominant-negative PS that inhibits
?-secretase activity. A red triangle also
indicates the location of the evolutionarily
conserved Pro-Ala-Leu (PAL) motif essential for
PS stability. The box to the right shows the
consensus sequences surrounding these conserved
residues. The putative location of these active
site aspartates places them in location to cleave
near the COOH-terminus of Aß. Y, Trp G, Gly.
Golde and Eckman, Sci STKE. 2003 Mar
042003(172)RE4.
44
Fig. 4. ?-Secretase also cleaves the type 1
surface receptor Notch and its family members.
Following binding of a ligand present on the
surface of the adjacent cell (delta, serrate, or
Lag-2) to the extracellular domain of Notch, a
conformational change occurs permitting JP by the
metalloprotease disintegrin (ADAM17). The Notch
COOH-terminal fragment (NEXT) is then cleaved by
? -secretase to release the Notch intracellular
domain (NICD) into the cytoplasm. Upon release,
the NICD translocates to the nucleus, where it
acts as a transcriptional regulator through its
interactions with the transcription factor CSL.
Golde and Eckman, Sci STKE. 2003 Mar
042003(172)RE4.
45
Regulated Intramembranous Proteolysis

• Sterol regulatory element binding protein (SREBP)

46
Fig. 1. (A) RIP of SREBP involves an additional
protein, SCAP, that regulates SREBP transport.
SREBP exists in a hairpin-like conformation with
a small luminal loop and two large cytoplasmic
domains. The NH2-terminal domain is a basic
helix-loop-helix (bHLH) transcription factor and
the COOH-terminal domain is a regulatory factor
(REG) that binds SCAP. When cells are loaded with
sterols, the complex of SCAP and SREBP is
sequestered in the ER. Upon sterol deprivation,
the SCAP-SREBP complex is transported to a
post-ER compartment (thought to be the cis or
medial Golgi) where S1P-mediated
juxtamemembranous proteolysis (JP) of the small
luminal loop of SREBP occurs. The bHLH domain of
SREBP is then translocated to the nucleus, where
it binds to sterol regulatory elements and
controls transcription of a number of genes
involved in sterol metabolism. (B) S2P appears to
be a metalloprotease-type intramembranous
cleaving protease (MpMP). A schematic of the
overall topology of S2P is indicated. The
position of residues involved in catalysis is
indicted by the stars. The conserved residues
indicated by these stars are shown in the box.
The conserved active site histidines (red star)
appear to be located at a site in the membrane
that permits cleavage of SREBP within its
transmembrane domain close to the cytoplasmic
face of the membrane. These residues, along with
the remote aspartate, are hypothesized to form an
active site similar to that formed by classic
metalloprotease. The true three-dimensional
structure of S2P is not known, and location of
residues is strictly based on hydropathy plots
and topology studies.
Golde and Eckman, Sci STKE. 2003 Mar
042003(172)RE4.
47
Regulated Intramembranous Proteolysis

• Rhomboid and EGF signaling

48
Figure 2. Identification of Putative Catalytic
Residues of Rhomboid-1(A) Conserved residues were
individually mutated to alanine and the ability
of the mutant proteins to mediate Spitz cleavage
was determined. The upper panel shows Western
blots of cleaved GFP-Spitz in the medium of cells
transfected with GFP-Spitz, Star, and mutant or
wild-type Rhomboid-1. Mutation of W151, R152,
N169, G215, S217, and H281 abolished detectable
Rhomboid-1 activitycompare with no Rhomboid-1
control (-). The lower panel shows Rhomboid-1
levels in cells, assessed using N-terminally
HA-tagged Rhomboid-1 mutants.(B) All mutants that
were unable to mediate Spitz cleavage were, like
the wild-type protein, localized to the Golgi
apparatus. HA-tagged S217A in a COS cell is shown
in green anti-p115 (Transduction Labs), a Golgi
marker, in red.(C) Comparison of the conserved
GASGG motif surrounding the putative Rhomboid-1
active serine (S217) with that of serine
proteases (family S1A according to MEROPS
classification). The subscripts represent the
percentage conservation at each residue (total
nonredundant sequences for each family are
reported as "n" values).(D) Rhomboid-1 with
conservative mutations S217C and S217T did not
catalyze Spitz cleavage above background levels
panels as in (A).(E) Diagram of Rhomboid-1
residues essential for Spitz cleavage conserved
but nonessential residues are shown in green,
essential residues in red. The TMD predictions
were agreed by a variety of algorithms, including
TMHMM and TMPred (available at http//ca.expasy.or
g/tools/transmem)
Urban et al, Cell. 2001 Oct 19107(2)173-182.
49
Figure 7. Model of Rhomboid-1 ActionWe propose
that the conserved asparagine (N), histidine (H),
and serine (S) form a serine protease catalytic
triad that hydrolyses the Spitz polypeptide
within its TMD (although the role of the
asparagine is uncertainsee text). Our data imply
that this occurs approximately one-third of the
distance into the membrane bilayer from the
lumenal surface. The lumenal tryptophan (W) and
arginine (R) are also essential (although the
tryptophan appears less critical in human
RHBDL2), but their function in the catalytic
mechanism remains to be determined
Urban et al, Cell. 2001 Oct 19107(2)173-182.
50
Fig. 7. RIP by a novel serine MpMP, Rhomboid.
(A) Drosophila Rhomboid-1 promotes cleavage of at
least three epidermal growth factor (EGF)-like
type 1membrane proteins, Spitz, Keren, and
Gurken. Rhomboid cleavage does not appear to
require prior JP. It also is unique in that it
releases the ectodomain from the membrane rather
than releasing a cytoplasmic domain. In their
uncleaved forms, these EGF-like ligands are
sequestered in the ER. In activated cells, the
transmembrane protein Star facilitates
trafficking to the Golgi where the active
Rhomboid protease resides. In the Golgi the
ligands are cleaved near the luminal border of
their transmembrane domains, releasing them from
the membrane. Once released, the ligand is
secreted where it binds the EGFR. (B) The
predicted topology of Drosophila Rhomboid-1.
Rhomboids are serine MpMPs. The stars show the
location of the putative catalytic triad and the
box describes these consensus sequences (N, Asn
G, Gly A, Ala S, Ser H, His). As is the case
for other MpMPs, the active site residues appear
to be positioned in an appropriate location to
carry out their cleavage, which in this case
occurs near the luminal face of the membrane. In
addition, the red triangle shows the location of
another sequence (W, Trp R, Arg) essential for
function.
Golde and Eckman, Sci STKE. 2003 Mar
042003(172)RE4.