We found a large, conserved protrusion on the surface of the PCSK9 catalytic domain and have performed site-directed mutagenesis experiments for 13 residues on this protrusion.. Whereas
Trang 1reveal a functionally important protrusion
Jamie Cameron1, Øystein L Holla1, Knut Erik Berge1,*, Mari Ann Kulseth1, Trine Ranheim1,
Trond P Leren1and Jon K Laerdahl2
1 Medical Genetics Laboratory, Department of Medical Genetics, Rikshospitalet University Hospital, Oslo, Norway
2 Centre for Molecular Biology and Neuroscience (CMBN), Institute of Medical Microbiology, Rikshospitalet University Hospital, Oslo, Norway
An elevated level of plasma low-density lipoprotein
(LDL) cholesterol is a major risk factor for coronary
heart disease The key factor regulating the level of
LDL cholesterol is the cell surface LDL receptor
(LDLR) [1] The number of LDLRs is regulated at the
transcriptional level [1] but is also
post-transcription-ally regulated by proprotein convertase subtilisin⁄ kexin
type 9 (PCSK9) [2], also known as NARC-1 [3]
Over-expression of PCSK9 in mice leads to reduced levels of
LDLR and increased levels of LDL cholesterol [2,4,5], whereas mice with no functional PCSK9 have increased levels of LDLR and reduced levels of LDL cholesterol [6]
Some aspects of the mechanism by which PCSK9 regulates the number of LDLRs have recently been identified Secreted PCSK9 binds to the epidermal growth factor-like repeat A (EGF-A) of the extra-cellular domain of the LDLR [7] PCSK9 bound to the
Keywords
evolutionary conservation; LDL cholesterol;
LDL receptor; PCSK9; structural
bioinformatics
Correspondence
J K Laerdahl, Centre for Molecular Biology
and Neuroscience (CMBN), Institute of
Medical Microbiology, Rikshospitalet
University Hospital, NO-0027 Oslo, Norway
Fax: +47 22 84 47 82
Tel: +47 22 84 47 84
E-mail: j.k.lardahl@medisin.uio.no
(Received 3 April 2008, revised 9 May 2008,
accepted 16 June 2008)
doi:10.1111/j.1742-4658.2008.06553.x
Proprotein convertase subtilisin⁄ kexin type 9 (PCSK9) interferes with the recycling of low-density lipoprotein (LDL) receptor (LDLR) This leads to LDLR degradation and reduced cellular uptake of plasma LDL Naturally occurring human PCSK9 loss-of-function mutations are associated with low levels of plasma LDL cholesterol and a reduced risk of coronary heart disease PCSK9 gain-of-function mutations result in lower LDL clearance and increased risk of atherosclerosis The exact mechanism by which PCSK9 disrupts the normal recycling of LDLR remains to be determined
In this study, we have assembled homologs of human PCSK9 from 20 ver-tebrates, a cephalochordate and mollusks in order to search for conserved regions of PCSK9 that may be important for the PCSK9-mediated degra-dation of LDLR We found a large, conserved protrusion on the surface of the PCSK9 catalytic domain and have performed site-directed mutagenesis experiments for 13 residues on this protrusion A cluster of residues that is important for the degradation of LDLR by PCSK9 was identified Another cluster of residues, at the opposite end of the conserved protrusion, appears
to be involved in the physical interaction with a putative inhibitor of PCSK9 This study identifies the residues, sequence segments and surface patches of PCSK9 that are under strong purifying selection and provides important information for future studies of PCSK9 mutants and for inves-tigations on the function of this regulator of cholesterol homeostasis
Abbreviations
CRD, cysteine-rich domain of PCSK9, i.e the C-terminal domain; EGF-A, epidermal growth factor-like repeat A of LDLR; EST, expressed sequence tag; LDL, low-density lipoprotein; LDLR, low-density lipoprotein receptor; PC, proprotein convertase; PCSK9, proprotein
convertase subtilisin ⁄ kexin type 9; WT, wild-type.
*[Correction added on 16 July 2008, after first online publication: the author name has been amended]
Trang 2LDLR is internalized by endocytosis [7,8], and bound
PCSK9 somehow disrupts the recycling of the LDLR
As a consequence, the LDLR is transferred to the
lysosomes for degradation [7]
PCSK9 belongs to a superfamily of subtilisin-like
serine proteases and is the ninth mammalian member
identified in the proprotein convertase (PC) family [3]
The PC zymogens have an N-terminal signal sequence,
a prodomain, a catalytic subtilisin-like domain, and a
C-terminal domain [9] They undergo autocatalytic
cleavage in the endoplasmic reticulum, but the
prodo-main reprodo-mains noncovalently bound to the catalytic
domain In PCSK9, the backbone is cut between
Gln152 and Ser153 [4], and autocatalysis, as well as
correct folding of the protein, is necessary for secretion
of PCSK9 [3] Unlike other convertases, PCSK9 does
not appear to undergo a second autocatalytic event
resulting in the release of an active protease [4,10]
Instead, the prodomain remains tightly bound to the
mature, cleaved PCSK9 after secretion It has been
shown that the enzymatic activity of PCSK9 is not
necessary for its regulation of the LDLR [11,12] The
finding that individuals without any detectable plasma
PCSK9 are healthy and develop normally [13,14]
sug-gests that drugs targeting PCSK9 might represent a
promising new class of LDL cholesterol-lowering
drugs
The crystal structure of free PCSK9 has recently
been determined, and showed the catalytic domain to
have high structural similarity to other subtilisin-like
serine proteases [10,15,16] The prodomain
(resi-dues 31–152) is tightly bound to the catalytic domain
(residues 153–449), hindering access to the active site
catalytic triad, Asp186, His226, and Ser386 The
N-ter-minal part of the prodomain (residues 31–60) is
struc-turally disordered The C-terminal cysteine-rich
domain (CRD) is built from three modules arranged
with quasi-three-fold rotational symmetry, where each
module forms a two-sheet b-sandwich comprising six
b-strands Each b-sandwich of this pseudo-propeller
fold is structurally homologous to the C-terminal
region of resistin [15] and is held together by three
structurally conserved disulfide bonds
In humans, various mutations in the PCSK9 gene
have been found to cause autosomal dominant
hypo-cholesterolemia or hyperhypo-cholesterolemia [4,13,17–24]
For mutations that do not affect PCSK9 folding or
secretion, these effects appear to be largely mediated
by different affinities of the mutant PCSK9s for the
LDLR [11,25] However, another level of complexity
has been added with the recent finding that PCSK9
itself is cleaved by the PC furin, and, to a lesser extent,
by PC5⁄ 6A [26] PCSK9 is cleaved between
resi-dues 218 and 219 in what has been shown to be a structurally disordered loop on the surface of the PCSK9 catalytic domain [10,15,16] Furin-cleaved PCSK9 is inactive in degrading LDLR, and naturally occurring gain-of-function mutations such as R215H [24], F216L and R218S [17,27] are likely to be gain-of-function mutations due to reduced furin cleavage The exact mechanism by which PCSK9 binds to the LDLR and disrupts the normal recycling of the LDLR remains to be determined One strategy to elucidate the underlying mechanism is to study how mutations
in the PCSK9 gene affect the PCSK9-mediated degra-dation of the LDLR Candidate residues for being of functional importance for macromolecular interactions involving PCSK9 are those that are highly conserved between different species, especially conserved residues that are solvent-exposed in unbound PCSK9 and that
do not appear to be important for protein folding Specific and functionally important protein–protein interactions between PCSK9 and other macromole-cules are likely to be mediated through a contact area with complementary shape, hydrophobicity and charges for the two protein surfaces Mutations that change the properties of the interacting PCSK9 surface will result in altered, usually weakened and less specific interaction with LDLR or another binding partner Consequently, residues involved in protein–protein interactions will be more conserved during evolution than other surface-exposed residues We therefore extracted the sequences of homologs of human PCSK9 from public sequence databases in order to search for functional regions by mapping phylogenetic informa-tion onto the known protein structure
PCSK9 is present in the proteome of most verte-brates as well as in the invertebrate Branchiostoma flor-idae Whereas most residues exposed on the PCSK9 surface appear to be under limited selective pressure in vertebrates, a large protrusion on the catalytic domain contains a number of absolutely conserved residues This protrusion could play an important role in specific macromolecular interactions, e.g for the inter-action with and degradation of the LDLRs We have therefore performed site-directed mutagenesis of 13 residues within this protrusion in order to study how the mutant PCSK9s affect uptake of LDL We found that the conserved residues cluster in two groups: one group causes PCSK9 gain-of-function mutations, whereas the remaining residues are located in a small patch giving rise to mutations of the loss-of-function type Our data suggest that the conserved protrusion is involved in two separate specific macromolecular inter-actions of importance for the PCSK9-mediated degra-dation of the LDLRs
Trang 3PCSK9 has homologs in chordates and mollusks
Homologs of human PCSK9 were extracted from a
number of public databases, including the NCBI
non-redundant protein and expressed sequence tag (EST)
databases [28], uniprot [29], the ensembl resources
[30], and some sequencing project databases Protein
sequences homologous to full-length human PCSK9
from many vertebrates were found, including primates,
rat, mouse, squirrel (Spermophilus tridecemlineatus),
and a number of other placental mammals, opossum
(Monodelphis domestica), chicken, the Carolina anole
lizard (Anolis carolinensis), frogs (Xenopus tropicalis⁄
Xenopus laevis), and the fish species Oryzias latipes,
Danio rerio, Tetraodon nigroviridis, and Takifugu
rubri-pes (see supplementary Doc S1) No vertebrate with
more than a single PCSK9 homolog was found
Data from the Florida lancelet (B floridae)
sequenc-ing project, containing both genomic and EST
sequences, indicate at least two potential homologs of
full-length PCSK9 in this organism B floridae is a
representative of the cephalochordates, one of the
three chordate subphyla, the other two being
verte-brates and urochordates No homologs of full-length
PCSK9 were detected in any urochordate, e.g in the
fairly well-studied Ciona intestinalis, or in any other
invertebrates The C-terminal domain of PCSK9, the
CRD, was not found in any vertebrate protein apart
from PCSK9 itself However, homologs of the CRD
appear to be present in proteins in the marine
Califor-nia sea slug (Aplysia californica) and in the freshwater
snail Biomphalaria glabrata These CRD homologs
were extracted from ESTs from the Aplysia EST
project and the Biomphalaria sequencing project (see
supplementary Doc S1)
The sequence data for the PCSK9 homologs are
given in the supplementary Table S1 Multiple
sequence alignments for all PCSK9 homologs were
generated (Fig 1 and supplementary Figs S1 and S2)
Bovines might be lacking a functional PCSK9
Sequence searching with human PCSK9 in the NCBI
EST databases gave highly significant hits in human,
mouse, rat, dog, chicken, frog, fish and lancelet ESTs
However, there was not a single detected PCSK9
homolog in 1.3 million bovine EST sequences In a
recent Bos taurus genome assembly (Btau_3.1) from
the Baylor College of Medicine Human Genome
Sequencing Center, we detected a genomic sequence on
Bos taurus chromosome 3 with high similarity to
human PCSK9 exons 8–12 These putative PCSK9 exons have insertions⁄ deletions and nonsynonymous mutations in regions that are absolutely conserved in all other vertebrates, including fish, and there appears
to be a premature stop codon in putative bovine exon 10 Traces sequenced in both directions on the genome are available in the NCBI Trace Archive that supports the stop codon in exon 10 The Btau_3.1 ver-sion of the bovine genome is a preliminary assembly based on approximately 26 million reads and 7· sequence coverage Close to 95% of bovine ESTs were contained in the assembled contigs, indicating that less than one in 20 Bos taurus protein-coding genes are missing in this assembly
The above findings suggest that the region on bovine chromosome 3 with homology to PCSK9 is a remnant
of a PCSK9 pseudogene, and that extant Bos taurus might be lacking functional PCSK9
Site-directed mutagenesis of residues in a conserved protrusion on PCSK9
On the basis of a multiple sequence alignment of 18 vertebrate PCSK9 homologs, residue conservation was mapped onto a PCSK9 structural model with the consurf tool [31,32] (Fig 2) Residue conservation
on the solvent-exposed PCSK9 surface is limited The exception is a large protrusion on the catalytic domain with a surface area of roughly 1500 A˚2 (Fig 2B) Approximately half of this protrusion, the part closest to the prodomain, is built from the struc-turally disordered loop Gly213–Arg218 (Fig 2A) and residues partially covered by this loop (Fig 2C) Evo-lutionarily conserved regions on protein surfaces are likely to be of functional importance, such as being involved in specific interactions with other macro-molecules We therefore performed site-directed muta-genesis of the 13 most conserved residues in this region (i.e residues on yellow, blue, pink and green background in Fig 2C) and investigated how these mutations affected PCSK9 secretion and the internali-zation of LDL
To study whether the mutant PCSK9s were autocat-alytically cleaved and secreted in a normal fashion, HepG2 cells were transiently transfected with mutant PCSK9 plasmids harboring each of the 13 different mutations The amounts of pro-PCSK9 and mature PCSK9 in cell lysates were determined by western blot analysis using an antibody to PCSK9 In cells express-ing wild-type (WT) PCSK9, two bands of 73 kDa and
64 kDa were observed, which correspond to pro-PCSK9 and the mature form of pro-PCSK9, respectively (Fig 3) Unlike the catalytically inactive mutant
Trang 4S386A-PCSK9, the 13 new PCSK9 mutants appeared
to be autocatalytically cleaved in a fashion similar to
that of WT-PCSK9 Moreover, all 13 mutant PCSK9s,
except for C375A-PCSK9 and C378A-PCSK9, were
secreted in a normal fashion (Fig 3) The amount of
C378A-PCSK9 in the culture media was markedly
reduced, whereas no C375A-PCSK9 was observed in
the media It is likely that C375A-PCSK9 and
C378A-PCSK9 are completely or partially, respectively,
retained in the endoplasmic reticulum due to abnormal
protein folding caused by disruption of the disulfide
bond bridging the residues Cys375 and Cys378
Effect of PCSK9 mutants on the internalization
of LDL and on PCSK9 cleavage by furin
To study the effects of the 13 PCSK9 mutants on the PCSK9-mediated degradation of the LDLR, we used transiently transfected HepG2 cells and studied the amount of LDL internalization by flow cytometry HepG2 cells transfected with WT-PCSK9, empty plas-mid, the catalytically inactive S386A-PCSK9 plasmid [3,23] or one of the two gain-of-function plasmids, S127R-PCSK9 and D374Y-PCSK9 [23], were used as controls (Fig 4) Internalization of LDL by cells
A
B
C
Fig 1 Multiple sequence alignments of human PCSK9 homologs from vertebrates, a cephalochordate (B floridae) and the mollusks Aplysia and Biomphalaria, showing the signal sequence and N-terminus of the prodomain (A), two segments of the catalytic domain (B), and the full CRD, the C-terminal domain (C) Conserved residues are indicated by numbering referring to human PCSK9 The catalytic triad residues are marked with an asterisk The full alignment and sequence data are given in supplementary Figs S1, S2 and supplementary Table S1.
Trang 5expressing these control plasmids was comparable to
previous findings [23,24]
Cells expressing the two mutants C375A-PCSK9
and C378A-PCSK9 internalized 19% and 14% more
LDL, respectively, than cells expressing WT-PCSK9
Thus, as expected for PCSK9 mutants that are secreted
at markedly reduced levels, the two mutants present as
loss-of-function type Cells expressing R194A-PCSK9,
D238A-PCSK9, T377A-PCSK9 or F379A-PCSK9
also internalized more LDL than cells expressing
WT-PCSK9 The amounts of LDL internalized by these cells were higher or similar to those of cells expressing C375A-PCSK9 or C378A-PCSK9 (Fig 4) Thus, we also consider these four to be loss-of-func-tion mutants
Cells expressing R237A-PCSK9 did not show any significant difference in LDL internalization as com-pared with cells expressing WT-PCSK9 (Fig 4) R237A-PCSK9 is therefore a neutral variant Cells expressing one of the remaining six PCSK9 mutants
B
Fig 2 Structural model of human PCSK9 with the conserved protrusion (A) The structurally disordered loop Gly213–Arg218 (orange) with the furin recognition motif is located on the catalytic domain (green) Also shown is the prodomain (gray) blocking the active site and the CRD, the C-terminal domain (pink) (B) Amino acid residue conservation in 18 vertebrate PCSK9 homologs mapped, employing CONSURF [31], onto the space-filling representation of the PCSK9 model The spatial orientation is identical to (A) (upper), and rotated 180 around a vertical axis (lower) The color scale extends from cyan (highly variable residues), through white (intermediate) to magenta (highly conserved) Yellow residues are of intermediate variability, but with low statistical confidence [31] The conserved protrusion is visible in the panel as an extended patch in magenta (upper part, right-hand side) (C) Magnification of the conserved protrusion showing the 13 residues of the cur-rent study giving mutants with a low level of protein secretion (yellow background), loss-of-function mutants (blue), gain-of-function mutants (pink), no change (green), as well as three residues of the disordered loop Gly213–Arg218 previously shown to give rise to gain-of-function mutants (orange) The model is rotated 50 around a vertical axis with respect to (B), upper panel.
Trang 6S153A-PCSK9, Q190A-PCSK9, D204A-PCSK9,
K222A-PCSK9, D374A-PCSK9 and S376A-PCSK9
internalized less LDL than cells expressing WT-PCSK9
(Fig 4) The amounts of LDL internalized by cells
expressing these mutants were in the same range or
lower than in cells expressing the gain-of-function
mutant S127R-PCSK9 Thus, we consider these
mutants to be gain-of-function mutants
The four loss-of-function mutants involving Arg194,
Asp238, Thr377 or Phe379 are located close together
on the conserved protrusion (Fig 2C) Four
gain-of-function mutants, involving Gln190, Lys222, Asp374
and Ser376, are located together in a separate region
between the loss-of-function patch and the disordered
loop Gly213–Arg218 The two remaining
gain-of-func-tion mutants, involving Ser153 and Asp204, are
located on opposing edges of the conserved protrusion
(Fig 2C)
Five of the six gain-of-function mutant residues are clustered in the vicinity of the disordered loop consist-ing of residues Gly213–Arg218 (Fig 2C) This loop contains the furin cleavage site RFHR218 [26], and cleavage by furin at this site results in PCSK9 that is inactive in degrading the LDLR [26] To determine whether the gain-of-function mutants had reduced furin cleavage, the amounts of furin-cleaved PCSK9 in the media of HEK293 cells transiently transfected with the different PCSK9 plasmids were determined by wes-tern blot analysis HEK293 cells were chosen for these analyses because truncated PCSK9 due to cleavage by furin is more prominent than in the medium of HepG2 cells R215H-PCSK9 was included as a negative con-trol Furin-cut PCSK9 was present in small amounts
in the media of cells transfected with gain-of-function plasmids as well as in the media of cells transfected with WT-PCSK9 plasmid or with loss-of-function
Fig 4 Internalization of LDL by HepG2 cells transiently transfected with mutant PCSK9 plasmids The effects of mutants S153A-PCSK9, Q190A-PCSK9, R194A-PCSK9, D204A-PCSK9, K222A-PCSK9, R237A-PCSK9, D238A-PCSK9, D374A-PCSK9, C375A-PCSK9, S376A-PCSK9, T377A-PCSK9, C378A-PCSK9 and F379A-PCSK9 on the internalization of fluorescently labeled LDL (10 lgÆmL)1) were studied in transiently transfected HepG2 cells by flow cytometry WT-PCSK9 plasmid, empty plasmid, and the catalytically inactive S386A-PCSK9 plasmid, as well
as D374Y-PCSK9 and S127R-PCSK9, were used as controls The values relative to WT-PCSK9 are given as the mean from three experi-ments (± SEM) The amount of LDL internalized by cells transfected with WT-PCSK9 was assigned a value of 100.
Fig 3 Autocatalytic cleavage of the mutants S153A-PCSK9, Q190A-PCSK9, R194A-PCSK9, D204A-PCSK9, K222A-PCSK9, R237A-PCSK9, D238A-PCSK9, D374A-PCSK9, C375A-PCSK9, S376A-PCSK9, T377A-PCSK9, C378A-PCSK9 and F379A-PCSK9 was determined by western blot analysis of cell lysates from HepG2 cells transiently transfected with the mutant PCSK9 plasmids WT-PCSK9 plasmid, empty plasmid and the catalytically inactive S386A-PCSK9 plasmid were used as controls Uncleaved, pro-PCSK9 and mature, cleaved PCSK9 are indicated (upper panel) The lower panel shows the amount of mature, cleaved PCSK9 in the media Three separate experiments were performed; one representative experiment is shown.
Trang 7plasmids (Fig 5) Clearly, the relative levels of
full-length mature and furin-cut PCSK9 in the media do
not correlate with mutants being gain-of-function
or not
R194A-PCSK9 and D204A-PCSK9 are
post-translationally modified
As can be seen from Figs 3 and 5, abnormal migration
of mature, cleaved PCSK9 was observed in lysates and
media of HepG2 cells and HEK293 cells transfected
with the R194A-PCSK9 plasmid or the D204A-PCSK9
plasmid However, the corresponding uncleaved
pro-PCSK9 (Fig 3) and the furin-cleaved pro-PCSK9 (Fig 5)
appeared to migrate normally To study whether the
abnormal migration of the mature forms of
R194A-PCSK9 and D204A-R194A-PCSK9 was due to altered
auto-catalytic cleavage, western blot analyses of media from
transfected HepG2 cells were performed using an
anti-body against the prodomain of PCSK9 The
prodo-mains of R194A-PCSK9 and D204A-PCSK9 migrated
normally (Fig 6) Thus, the two mutants were
autocat-alytically cleaved in a normal fashion
To study whether the abnormal migration of
mature, cleaved R194A-PCSK9 and D204A-PCSK9
was due to abnormal glycosylation, the sensitivities of
R194A-PCSK9 and D204A-PCSK9 to an enzyme mix
designed to remove all sugars were determined in cell
lysates of transiently transfected HepG2 cells The
results showed that the differences in the migration of
mature PCSK9 remained after the enzyme treatment
(Fig 7) Thus, an abnormal post-translational
modifi-cation other than glycosylation appears to be
respon-sible for the abnormal migration of R194A-PCSK9 and D204A-PCSK9
Discussion
Vertebrate genome sequencing projects are currently supplying the research community with sequence data from a large number of species that have varying evo-lutionary relationships with humans The data from these projects make it possible to study in detail the level of evolutionary residue conservation in proteins, including estimates of statistical significance In the present study, we have extracted protein sequence data from 21 chordate proteomes and mapped the degree of residue conservation onto the surface of a PCSK9 structure model We found that most of the residues
on the surface of this LDLR-degrading protein appear
to be tolerant to substitutions However, a single large protrusion on the catalytic domain contains a number
of residues that are highly conserved (Fig 2B) We have performed site-directed mutagenesis of 13 resi-dues contributing to this protrusion (Fig 2C), and show that most mutants have either increased or decreased ability to degrade the LDLR and internalize
Fig 5 Amounts of furin-cleaved PCSK9 in the media of HEK293 cells transiently transfected with the mutant PCSK9 plasmids S153A-PCSK9, Q190A-PCSK9, R194A-PCSK9, D204A-PCSK9, K222A-PCSK9, R237A-PCSK9, D238A-PCSK9, D374A-PCSK9, C375A-PCSK9, S376A-PCSK9, T377A-PCSK9, C378A-PCSK9 and F379A-PCSK9 were determined by western blot analysis WT-PCSK9 plasmid, and the gain-of-function plasmids D374Y-PCSK9 and R215H-PCSK9, were used as controls R215H-PCSK9 is not cleaved by furin Three separate experiments were performed; one representative experiment is shown.
WT R194A D204A
PCSK9 prodomain
Fig 6 Western blot analysis using an antibody to PCSK9
recogniz-ing the prodomain was used to identify the prodomains of
WT-PCSK9, R194A-PCSK9 and D204A-PCSK9 in the media of
HepG2 cells.
Fig 7 Western blot analysis was performed for deglycosylated cell lysates The figure shows a representative western blot of cell lysates of HepG2 cells transiently transfected with WT plasmid or plasmids containing R194A-PCSK9 or D204A-PCSK9 with or with-out prior treatment with the Glycoprotein Deglycosylation Kit A horizontal dotted line is included to show that all the mature PCSK9s after deglycosylation have increased mobility due to degly-cosylation.
Trang 8LDL as compared to WT-PCSK9 (Fig 4) Only one
of these residues, Asp374, has previously been
associ-ated with hypercholesterolemia in human populations
[18,19,33]
Previous studies have described a number of
natu-rally occurring loss-of-function mutations in PCSK9
that result in proteins that are not autocatalytically
cleaved and⁄ or not folded properly [4,10,13,23,24]
Apart from the previously studied active site mutant
S386A-PCSK9 [23], the only mutants of the present
study that clearly have impaired secretion are
C375A-PCSK9 and C378A-C375A-PCSK9 (Fig 3) Both residues are
absolutely conserved in all chordates (Fig 1B and
sup-plementary Fig S1), demonstrating their importance
for the formation of a disulfide bridge stabilizing the
conformation and overall shape of the conserved
protrusion (Fig 2)
Four of the mutants, R194A-PCSK9,
D238A-PCSK9, T377A-D238A-PCSK9, and F379A-D238A-PCSK9, were
secreted in a similar fashion as WT-PCSK9 (Fig 3),
but nevertheless present as loss-of-function mutants
(Fig 4) The residues involved are located close
together on the conserved protrusion, at the far end
from both the prodomain and the disordered loop
Gly213–Arg218 (Fig 2C) All four residues are highly
conserved in chordates (Fig 1B and supplementary
Fig S1), and are clearly under strong purifying
selec-tive pressure; that is to say, there is substantial
nega-tive natural selection against amino acid replacements
for these residues Arg194 is conserved in all
verte-brates, but is replaced by Glu in B floridae, whereas
Asp238 is conserved in all vertebrates except for the
fish Fugu (T rubripes), where it is replaced by Glu,
a residue with similar properties to Asp Thr377 is
conserved in every single chordate, whereas Phe379 is
conserved in all vertebrates except for the rat, where it
is substituted by another aromatic residue, Tyr As
discussed above, conservation of these residues is not
necessary for protein expression, folding, autocatalysis
or secretion, strongly indicating that this patch on the
PCSK9 surface is, instead, of importance for the direct
physical protein–protein interactions leading to
PCSK9-mediated degradation of the LDLR
During the preparation of this manuscript, Kwon
et al [34] published the crystal structure of the protein
complex formed between PCSK9 and the EGF-A
domain of the LDLR They found that the interaction
between PCSK9 and EGF-A was primarily
hydropho-bic, with some additional specific polar interactions
Phe379 was in the center of the hydrophobic surface,
and Arg194, Asp238 and Thr377 were involved in the
polar interactions with EGF-A [34] Thus, the surface
region of PCSK9 involved in this interaction coincides
with the part of the conserved protrusion associated with loss-of-function mutants in the present study Our findings that mutations R194A, D238A, T377A and F379A were loss-of-function mutations are in agree-ment with the notion that they diminish the binding of PCSK9 to EGF-A
The crystal structure obtained by Kwon et al [34] shows that the N-terminal amine of mature PCSK9 Ser153 forms a salt bridge with a residue in EGF-A, but that the Ser153 side-chain does not directly contact the binding partner Correspondingly, our results showed that S153A-PCSK9 is not a loss-of-function mutant Instead, S153A-PCSK9 appears to lead to decreased internalization of LDL as compared to WT-PCSK9 One might speculate that this could be due to
a slight change in the ability of residue 153 to form a salt bridge to EGF-A, e.g through an inductive effect All the other residues that were associated with gain-of-function mutations in the present study were located either between the EGF-A binding patch and the disordered loop Gly213–Arg218 (Gln190, Lys222, Asp374, and Ser376), or between the disordered loop and the prodomain (Asp204) (Fig 2C) These residues are under selective pressure, with Asp204 and Asp374 being conserved in all vertebrates Lys222 is conserved
in nonfish vertebrates, whereas the conservation of Ser376 appears to be slightly lower (Fig 1B and sup-plementary Fig S1) Mutations of the disordered loop residues, Arg215 [24], Phe216, and Arg218 [26], located
in this part of the conserved protrusion, have previ-ously been shown to be associated with resistance to furin cleavage We therefore investigated whether the gain-of-function mutants of the present study showed reduced cleavage by furin However, we did not find any difference in the amounts of the furin-cleaved bands when investigating the media of cells transfected with loss-of-function mutants as compared with gain-of-function mutants (Fig 5.)
Some caution should be exercised when interpreting these data, as in our study overexpressed PCSK9 has
to be cut by endogenous furin, and these conditions may not be physiologically relevant for the in vivo situation Interestingly, the basic residues in the furin recognition sequence RFHR218 of the disordered loop Gly213–Arg218 are conserved in all vertebrates, but not in the pufferfish T rubripes and T nigroviridis or
in the opossum (Fig 1B) The cephalochordate homo-logs have a deletion of four residues in this loop as compared with the human PCSK9, and appear to completely lack the disordered loop It appears that
PC regulation of PCSK9 is a vertebrate invention and that this level of regulation has subsequently been lost
in some vertebrate subgroups
Trang 9If the gain-of-function character of the mutants of
the present study is not due to reduced furin affinity,
another possibility is that a different, unknown,
mac-romolecule is interacting in a specific manner with the
relevant part of the conserved protrusion of PCSK9
This macromolecule may be competing with EGF-A
binding or may inhibit the PCSK9-mediated
degrada-tion of the LDLR by another mechanism Fan et al
[35] have recently suggested that multimerization of
PCSK9 is important for its LDLR-regulating activity
They found that mutation of Asp374, a residue in the
conserved protrusion, affected PCSK9 self-association
It is, however, not obvious that PCSK9 self-association
is important in vivo when PCSK9 is secreted at low
concentrations Previous studies have found no
indica-tions of multimerization for mature PCSK9 [3]
Earlier studies have shown that the naturally
occur-ring mutant D374Y-PCSK9 binds LDLR more
effi-ciently than WT-PCSK9 [8,10,25], and Kwon et al [34]
suggested that this was due to an additional hydrogen
bond between PCSK9 Tyr374 and His306 of the
EGF-A We now show that D374A-PCSK9, which results in
a residue Ala374 that clearly cannot form any
hydro-gen bonds with its side-chain, is also a gain-of-function
mutant, although it is only half as potent as
D374Y-PCSK9 This may indicate that the naturally occurring
D374Y-PCSK9 is a gain-of-function mutant due to two
different mechanisms: one is to strengthen the
interac-tion between PCSK9 and EGF-A, and the other is to
disrupt the binding to PCSK9 of a putative inhibitory
macromolecule
The sequence data that were gathered for the present
study reveal interesting phylogenetic relationships in
addition to the identification of the conserved
protru-sion discussed above Homologs of full-length PCSK9
were found in a single copy in a number of
verte-brates, and at least in duplicate in the cephalochordate
B floridae PCSK9 thus appears to be restricted to
chordates, and possibly limited to the Cephalochordata
and Vertebrata On the basis of the presumed
diver-gence of the chordate subphyla, one might speculate
on a Cambrian or late Proterozoic origin of PCSK9
Interestingly, we were unable to find any bovine
PCSK9 homolog that appears to be functional, but the
Bos taurus genome does appear to contain a
PCSK9-like pseudogene One might speculate that the cow,
with its diet of mainly grasses and plant material,
might be thriving without PCSK9, as do some human
individuals without functional PCSK9 [13,14]
The CRD, whose function still appears to be a
mystery, is the C-terminal PCSK9 domain It does not
appear to occur in any vertebrate protein apart from
PCSK9 itself, but was detected in sequence data from
two mollusk proteins of unknown function Although the available data are limited, these mollusk CRDs do not appear to be present in proteins that contain pro-tease domains (see supplementary Doc S1) This might indicate that mollusks employ the CRD for a different purpose than vertebrates do in PCSK9 It is possible that investigations on mollusk CRD-containing pro-teins might give indications on the function of the PCSK9 CRD
The multiple sequence alignments of the PCSK9 homologs (Fig 1 and supplementary Figs S1 and S2) clearly show the catalytic domain to be more con-served than the CRD This is also the case for PCSK9 conservation within the group of primates [36] Resi-due identities between human and opossum are 76% and 53% for the catalytic domain and CRD, respec-tively The prodomain is also fairly well conserved, apart from the structurally disordered region compris-ing the N-terminal 30 residues (Fig 1A) This segment
is very rich in acidic residues, with seven of 10 N-ter-minal residues of human PCSK9 being Asp or Glu This is immediately followed by five small aliphatic residues and a segment with five more acidic residues The N-terminal region of the prodomain will clearly interact strongly and nonspecifically with a positively charged moiety The signal sequence is not conserved, except for a Leu-rich segment
The three catalytic residues are absolutely conserved
in all PCSK9 homologs (Fig 1B), as is the last residue
of the prodomain, Gln152, supporting the notion that these residues are essential for autocatalysis and effi-cient secretion of PCSK9 The 18 Cys residues of the PCSK9 CRD are conserved in all chordates, as well as
in the mollusk CRDs (Fig 1C and supplementary Fig S2) This clearly demonstrates that the nine disul-fide bridges covalently stabilizing the three modules of this domain are essential for its processing and func-tion The CRD also contains a number of conserved Ser, Thr and small aliphatic residues These are mainly located deep in the structure, and are most likely essential for correct folding of the CRD There is a single patch of conserved residues on the surface of the CRD, comprising Arg458, Thr459, Trp461, and Glu481 These residues all contribute to the part of the CRD surface that is interacting with the catalytic domain The evolutionary conservation of these resi-dues indicates that although this interaction might be weak [16], it appears to be of functional importance Piper et al [16] have noted a large number of His residues in the CRD With a pKa value for His between 6 and 7, it is likely that the net charge of the CRD will become substantially more positive at endosomal pH 5–5.5 than at pH 7.4 at the plasma
Trang 10membrane It is tempting to speculate that this could
result in an altered interaction with the strongly acidic
N-terminal region of the prodomain This might be the
reason why PCSK9 binds more strongly to the LDLR
in the endosomal⁄ lysosomal compartments than in the
plasma [7,10] However, the positions and number of
His residues are not particularly conserved in the CRD
(Fig 1C) Whereas human PCSK9 has 15 His residues
out of a total of 245 CRD residues (6.1%), the
propor-tions are 5.3% for rat, 5.0% for opossum, 3.0% for
medaka fish, and 1.3% for the mollusk Aplysia
Asn533, which is glycosylated in human PCSK9, has
been shown not to be essential for PCSK9 secretion
[13,26] It is conserved in placental mammals only The
corresponding residue is not likely to be glycosylated
in other vertebrates
In conclusion, there is a single, large, evolutionarily
conserved protrusion on the surface of the catalytic
domain of PCSK9 The lack of other residue
conserva-tion on the PCSK9 surface makes it less likely that
there are other parts of PCSK9 that interact with high
specificity with other macromolecules as part of the
PCSK9-mediated degradation of the LDLR A cluster
of residues on the conserved protrusion is involved in
the binding of PCSK9 to the EGF-A domain of the
LDLR, and mutations of these residues lead to loss of
function, as found in our study and in the study of
Kwon et al [34] The part of the protrusion located
around the disordered loop Gly213–Arg218 contains a
number of conserved residues for which site-directed
mutagenesis produced gain-of-function mutants These
residues appear to be involved in some form of
inhibi-tion of the PCSK9-mediated degradainhibi-tion of the
LDLR However, our data do not clearly support a
model that solely involves reduced cleavage by furin
Thus, further studies are needed to clarify whether
these residues are involved in the binding of a different
macromolecule that inhibits the degradation of the
LDLR by PCSK9
Experimental procedures
Data collection and bioinformatics analysis
Database resources provided by the NCBI [28], uniprot
[29], the ensembl project [30], the DOE Joint Genome
Institute (http://genome.jgi-psf.org), the Baylor College of
Medicine Human Genome Sequencing Center (http://
www.hgsc.bcm.tmc.edu/projects/bovine), the Aplysia EST
project (http://aplysia.cu-genome.org) and the B glabrata
Genome Initiative
(http://biology.unm.edu/biomphalaria-genome) were searched for homologs of human PCSK9 A
major proportion of the extracted 24 PCSK9 homologs from
23 species is due to automatic gene searching in genomic data from early-stage sequencing projects This necessitated some manual trimming and manipulation of the sequences (see supplementary Doc S1 and supplementary Table S1) Multiple sequence alignments were generated with mus-cle[37], and the multiple sequence alignments were viewed and manipulated with jalview [38] A PCSK9 structural model was generated from a published experimental struc-ture [10] as described previously [24] Amino acid conserva-tion in all vertebrate PCSK9 homologs, but excluding the two pufferfish species, was mapped onto the PCSK9 model employing consurf [31,32] The protein structure illustra-tions were generated with pymol [39]
Cell cultures
HepG2 cells and HEK293 cells, obtained from the Euro-pean Collection of Cell Cultures (Porton Down, UK), were cultured in MEM (Gibco, Carlsbad, CA, USA) containing penicillin (50 UÆmL)1), streptomycin (50 lgÆmL)1), l-gluta-mine (2 mm) and 10% fetal bovine serum (Invitrogen, Carlsbad, CA, USA), in a humidified atmosphere (37C, 5% CO2)
Mutagenesis, cloning and expression of PCSK9
R237A, D238A, D374A, C375A, S376A, T377A, C378A or F379A were introduced into a pCMV–PCSK9–FLAG plas-mid kindly provided by J D Horton (University of Texas Southwestern Medical Center, Dallas, TX, USA), using QuickChange XL Mutagenesis Kit (Stratagene, La Jolla,
CA, USA) according to the manufacturer’s instructions The primer sequences used for the mutagenesis are given in supplementary Table S2 The resulting mutant plasmids are referred to as S153A-PCSK9, Q190A-PCSK9, R194A-PCSK9, D204A-R194A-PCSK9, K222A-R194A-PCSK9, R237A-R194A-PCSK9, D238A-PCSK9, D374A-PCSK9, C375A-PCSK9,
F379A-PCSK9 The integrity of each plasmid was confirmed by DNA sequencing An empty plasmid, pcDNA3.1⁄ myc his-c (Invitrogen), as well as four previously published mutant PCSK9 plasmids containing mutations S386A, S127R, R215H or D374Y [23,24], were used as controls in the transfection experiments together with WT-PCSK9 plasmid Transient transfections of HepG2 cells and HEK293 cells with WT-PCSK9 plasmid or mutant PCSK9 plasmids were performed as described by Cameron et al [24]
Western blot analysis of transfected HepG2 and HEK293 cells
Western blot analyses of cell lysates and culture media of transiently transfected HepG2 cells or HEK293 cells were