1090Journal of Lipid ResearchVolume 40, 1999
SAA-only HDL formed during the acute phase response
in apoA-I and apoA-I mice
Veneracion G. Cabana,
Catherine A. Reardon, Bo Wei, John R. Lukens, and Godfrey S. Getz
Department of Pathology, the University of Chicago, Chicago, IL 60637
of unknown function that is involved in systemic amyloido-
sis and may also be involved in atherogenesis. The precise
role of SAA in these processes has not been established.
SAA circulates in plasma bound to high density lipoprotein-
). The pathway for the production of SAA-containing
HDL is not known. To test whether apolipoprotein (apo)A-I-
HDL is required in the production of SAA-HDL, we ana-
lyzed the lipopolysaccharide (LPS)-induced changes in
and apoA-I mice. In apoA-I
jection of LPS, remodeling of HDL occurred: total choles-
terol increased and apoA-I decreased slightly and shifted to
lighter density. Dense (density of HDL
) SAA-containing particles were formed. Upon fast
phase liquid chromatography fractionation of plasma,
of SAA eluted with HDL that was enriched in cholesterol
and phospholipid and shifted “leftward” to larger particles.
Non-denaturing immunoprecipitation with anti-mouse apoA-
I precipitated all of the apoA-I but not all of the SAA, con-
firming the presence of SAA-HDL devoid of apoA-I. In the
mice, which normally have very low plasma lipid
levels, LPS injection resulted in significantly increased total
and HDL cholesterol. Greater than 90% of the SAA was
lipid associated and was found on dense but large, spherical
HDL particles essentially devoid of other apolipoproteins.
We conclude that serum amyloid A (SAA) is able to seques-
ter lipid, forming dense but large HDL particles with or
without apoA-I or other apolipoproteins. The capacity to
isolate lipoprotein particles containing SAA as the pre-
dominant or only apolipoprotein provides an important
system to further explore the biological function of SAA.
Cabana, V. G., C. A. Reardon, B. Wei, J. R. Lukens, and G. S.
SAA-only HDL formed during the acute phase re-
sponse in apoA-I
Serum amyloid A (SAA) is an acute phase protein
mice, after in-
) but large (size of
J. Lipid Res.
Supplementary key words
gene knockout mice
mouse plasma lipids
acute phase response
The acute phase response (APR) is a systemic reaction
to infectious and non-infectious stress processes character-
ized by the rapid increase in the concentrations of the
acute phase reactants C-reactive protein and serum amy-
loid A (SAA). Changes in lipid and apolipoprotein levels
also occur during the APR. The most common changes
are decreases of total cholesterol, HDL cholesterol, or
apoA-I as reported in human subjects after surgery (1, 2),
infections (3–6), myocardial infarction (7), septicemia (8,
9), Kawasaki disease (10–12), myeloma and lymphoma
(13, 14), and other minor or critical illnesses (7, 15, 16).
Hypertriglyceridemia has also been reported in some of
these conditions (5, 13). Decreases of total and HDL cho-
lesterol (17, 18) with hypertriglyceridemia (17,19) have
also been described in rabbits and non-human primates
after induction of the APR by various agents. Reduced
HDL is an independent risk factor for heart disease (20).
Whether SAA is directly involved in the reduction of HDL
during the APR has not been established.
SAA is an acute phase reactant which increases
fold within 24 to 48 h after induction of infection and tissue
destructive processes. It is a 12 kD product of a polymor-
phic, highly conserved gene family present in all verte-
brates studied from ducks to humans (21). It is produced
in response to inflammatory cytokines predominantly in
the liver (22) and also in some extra hepatic tissues (23–
26). The biological significance of the massive increase of
SAA during the APR remains unclear.
SAA is the precursor of the AA proteins deposited in re-
active or secondary amyloidosis. The precise mechanism
involved in the conversion of SAA to AA is still not clearly
understood. Recent evidence also suggests that SAA may
be involved in atherogenesis aside from its value as a
marker of the risks of future myocardial infarction, stroke,
or death from cardiovascular causes (27–30). SAA displays
a reciprocal relationship with the antioxidant enzymes
paraoxonase and platelet activating factor acetyl hydrolase
(31). SAA apparently binds cholesterol with high affinity
(32) and could alter HDL-mediated cholesterol efflux
Abbreviations: APR, acute phase response; apoA-I-HDL, HDL parti-
cles containing apoA-I as the major apolipoprotein; FPLC, fast phase
liquid chromatography; HDL, high density lipoprotein; LPS, lipo-
polysaccharide; SAA, serum amyloid A;
of mouse SAA gene and gene product, respectively, as suggested by the
Nomenclature Committee of the International Society of Amyloidosis;
SAA-only HDL or SAA-HDL, HDL particles containing SAA essentially
devoid of other apolipoproteins.
To whom correspondence should be addressed.
Saa1.1 and SAA1.1, designation
by guest, on June 5, 2013
Cabana et al.SAA-only HDL in apoA-I gene knockout and wild-type mice1091
(33). Some SAA isoforms are expressed in atherosclerotic
plaques (34, 35) and SAA induces the migration, adhe-
sion, and tissue infiltration of monocytes (36), cells strongly
implicated in the pathogenesis of atherosclerosis. More-
over, SAA circulates in the plasma associated with apoA-I-
containing HDL (apoA-I-HDL), and in vitro could dis-
place apoA-I from the particles (37). The significance of
its association with HDL is not clearly understood.
From the available information it is not clear whether
the formation of SAA-HDL results from modification of
preassembled HDL, or whether SAA produced during the
APR is able to modify the biogenesis of HDL in the liver
producing SAA-HDL de novo. In this study we have asked
whether SAA-containing HDL can be formed in the ab-
sence of apolipoprotein A-I, i.e., in the apoA-I-deficient
mouse, and whether even in the presence of apoA-I in the
wild-type mouse an SAA-containing HDL particle devoid
of apoA-I can be isolated. This study provides evidence
that SAA-(only)-HDL can be found both in the absence
and in the presence of apoA-I. In this study we will show
i ) the production of SAA-HDL does not require the
presence of apoA-I-HDL,
ii ) SAA-only HDL and SAA-AI-
HDL exist in plasma and are present in dense but large
particles with a very small lipid core, and
sence of apoA-I, SAA-HDL is produced as discrete parti-
cles devoid of other apolipoproteins.
iii ) in the ab-
MATERIALS AND METHODS
Induction of the APR
C57BL/ 6 mice (apoA-I
Jackson Laboratories (Bar Harbor, ME). Mice deficient in the
apoA-I gene (apoA-I
) were obtained from Dr. Noboyu Maeda
(38) and a breeding colony was established at the animal facili-
ties of the University of Chicago. ApoA-I
generations to the C57BL/ 6 strain were obtained from the Jack-
son Laboratories. Plasma lipid and apolipoprotein changes dur-
ing the APR were comparable in the animals obtained from these
The APR was induced by intraperitoneal injection of 50
bacterial lipopolysaccharide (LPS,
At each time point, 4–5 mice were exsanguinated by retro orbital
and/ or cardiac puncture from animals anesthetized by inhala-
tion of metoxyflurane (Methofane, Pitman-Moore) and the
blood was collected into tubes containing 0.1% EDTA, pH 7.4.
The plasma recovered after centrifugation was stored at 4
the presence of the protease inhibitor phenylmethylsulfonyl fluo-
m in methanol), and anti-bacterial agents (per ml
g gentamicin sulfate, 50
phenicol) and used within 1 week. Due to the limited blood vol-
ume of the mouse, plasmas from 4–5 animals were pooled.
, 6–8 wks) were purchased from
mice backcrossed 6
E. coli serotype 0127:B8, Sigma).
HDL was analyzed using three methods of isolation. Most of
the analyses examining the distribution of SAA involved density
gradient centrifugation of whole plasma. For detailed composi-
tional analysis of HDL, HDL was first obtained by sequential flo-
tation at d 1.063–1.25 g/ ml followed by re-isolation by density
gradient centrifugation to separate the different subclasses. As
protein components are sometimes desorbed from lipoproteins
in the high centrifugal field involved in density gradient centrifu-
gation, we also separated lipoproteins by size using FPLC. This al-
lowed us to confirm the increment in size of HDL when SAA was
incorporated into it.
HDL was isolated by one of the three methods. Most of the re-
sults presented are based upon density gradient centrifugation of
whole plasma following procedures as described previously (19).
In this procedure, 1 or 2 ml of plasma was layered at the interface
of a 3–20% NaBr gradient and centrifuged to equilibrium for 66
h at 38,000 rpm in a SW41 Ti rotor (Beckman). After centrifuga-
tion, 30 0.4-ml fractions were collected using a gradient fraction-
ator (ISCO) with UV monitor. The refractive index of each of the
fractions was assessed as an indicator of the density based on the
refractive index of salt solutions of known concentration, density ,
and refractive index. The fractions were dialyzed against Tris-
buffered saline (TBS: 10 m
m Tris, 150 m
7.4) and used for analyses.
For detailed analyses of the composition of the particles, HDL
was isolated from pooled plasma first by sequential flotation at d
1.063–1.25 g/ ml following standard methodology (39) to remove
most of the non-HDL plasma proteins. The isolated HDL was dia-
lyzed against TBS and 2 mg HDL protein/ tube was re-isolated by
density gradient ultracentrifugation (as described above) to sepa-
rate the different HDL subclasses. The HDL fractions were dia-
lyzed against TBS before their lipid and apolipoprotein composi-
tions were determined.
HDL was also isolated from plasma by fast phase liquid chro-
matography (FPLC) using two Superose 6 columns (Pharmacia)
arranged in tandem. Typically, 600
in pre-filtered and degassed PBS, pH 7.5. Seventy fractions (0.4
ml/ fraction) were collected and analyzed.
m NaCl, 10 mm NaN
l of plasma was fractionated
Lipid and protein quantitation
Triglycerides and cholesterol were analyzed using commer-
cially purchased enzymatic kits (Boehringer Mannheim). HDL
cholesterol was determined after precipitation of the apoB-con-
taining lipoproteins by magnesium chloride and phosphotung-
stic acid (Boehringer Mannheim). Phospholipids were quanti-
tated by an enzymatic–colorimetric assay of choline-containing
phospholipids (Wako). All lipoprotein analyses were performed
by methods standardized against CDC furnished standards. Pro-
tein was quantitated according to the procedure of Lowry et al.
(40) with SDS to disrupt the lipid micelles (41) using bovine al-
bumin as standard (Pierce).
ApoA-I was quantitated by radial immunodiffusion (RID)
based on the assay for human apoA-I (42) using anti-mouse
apoA-I produced as described below (see Production of anti-
body). Briefly, a pre-titered amount of antiserum was incorpo-
rated into an agar plate. Two-mm wells were punched into the
agar. Samples and standards pre-incubated for 30 min in PBS
containing 1% Triton-X100 (Kodak) to denature the lipoprotein
were added in duplicate wells at the opposite quadrants of the
plate. After 24 h of incubation at room temperature in a humidi-
fied chamber, the precipitin rings were measured using a mi-
crometer lens. Concentration of the standard ranging between 1
and 20 mg/ dl was plotted against the square of the diameter of
the precipitin rings and apoA-I concentration of the samples was
estimated from the standard curve. The standard used was the
peak fraction of C57BL/ 6 HDL obtained by a two-step isolation
procedure by sequential flotation at d 1.063–1.25 g/ ml followed
by density gradient centrifugation in a 3–20% NaBr gradient. After
dialysis and determination of protein concentration as described
above, the fraction was subjected to SDS PAGE and the relative
proportion of apoA-I (
90%) was estimated by densitometric
SAA was assessed by Western immunoblotting using antibody
against mouse SAA produced as described below .
by guest, on June 5, 2013
1092Journal of Lipid ResearchVolume 40, 1999
Polyacrylamide gel electrophoresis
Molecular weights of proteins were estimated from a 10–20%
SDS PAGE (0.2% SDS) based on the procedure of Weber and Os-
borne (43) using a mini-gel system (Hoeffer). One
g of HDL protein was loaded per lane after reduction
-mercaptoethanol and heating to 100
Tris-buffered sample diluent (10 m
m EDTA, 0.02 mg/ ml bromphenol blue, pH 8.0). Electro-
phoresis was conducted at 25 mA/ gel with a Tris-glycine running
buffer (25 m
m Tris, 192 mm glycine, 1% SDS) until the tracking
dye moved to about 1 cm from the bottom of the gel. The gels
were either used for electrotransfer of the protein band to Immo-
bilon-P membrane (Millipore), or stained with Coomassie R250
(0.25% stain, 40% methanol, 10% acetic acid), destained by dif-
fusion into 40% methanol, 10% acetic acid, and stored in 7%
acetic acid. The approximate proportion of apolipoproteins
present in a given lipoprotein fraction were assessed by scanning
densitometry of the protein bands.
The size of the HDL particles isolated from equilibrium den-
sity centrifugation was estimated from a non-denaturing 4–30%
gradient gel electrophoresis system as described by Blanche et al.
(44) using gels purchased from Dr. David Rainwater (Southwest
Foundation, San Antonio, TX). Seven to 10
used per lane of the 3 mm gel. A mixture of reference proteins
(HMW Standard, Pharmacia) consisting of thyroglobulin (r
8.5 nm), ferritin (r
6.1 nm), catalase (r
4.08 nm), and bovine serum albumin (r
nm) was included in each gel. Electrophoresis was conducted at
150 V for 18 h (or 2500 volt hours) with the temperature main-
tained at 10
C by a circulating water bath (Forma Scientific). The
gels were stained with Coomassie G250 in perchloric acid (0.1%
stain, 50 g [v/ v] perchloric acid) and stored in 7% acetic acid.
Isoelectric focusing (IEF) of gels prepared in 8
4–6.5 ampholines (Pharmacia) was a modification of the proce-
dure published by Kashyap et al. (45). Thirty
was denatured by mixing with an equal volume of tetramethyl
urea, centrifuged in a microfuge (Beckman), and the superna-
tant was applied to an isoelectric focusing tube that had been
pre-focused (1 h at 110 V with 20 m
anodic and cathodic electrolytes, respectively) to establish the
pH gradient. After application of the sample, electrophoresis was
conducted for 1 h at 110 V, then 3.5 h at 400 V using fresh elec-
trolytes of the same concentration. The gels were either stored
frozen in a SDS sample diluent to be used as a second dimension
gel, or stained with Coomassie G250 in perchloric acid as above
and stored in 7% acetic acid. Second dimension gels were 3 mm
thick 12% SDS PAGE slabs prepared and run as described above.
Production of recombinant murine SAA
Recombinant murine SAA1.1 (formerly murine SAA2) (46)
was produced as a fusion pGEX-KG protein with glutathione-S
transferase (GST) in the vector bacteria. The cDNA was pro-
duced by the polymerase chain reaction (PCR) using as template
RNA isolated from apoA-I
mouse liver 24 h after the injection
of LPS. The mRNA was isolated by guanidine thiocyanate/cesium
chloride following procedures outlined by Chirgwin et al. (47).
Oligonucleotides (Integrated DNA Technologies, Iowa) were
prepared to contain an XbaI and a SalI site at the 5
of the murine
Saa1.1 gene (5
, the underlined nucleotides corresponding to
1 in the first oligo and the stop codon in the second oligo).
After sequencing to confirm that we had indeed copied the
Saa1.1 transcript, the cDNA was subcloned into a pGEX-KG vec-
tor, transformed into bacteria, grown, and the protein expression
was induced by isopropylthio-
l of plasma
C for 3 min in a
m Tris, 1% SDS, 10% sucrose,
g of sample was
5.2 nm), lactate de-
m urea with pH
g of HDL protein
m NaOH and 0.01% H
-galactoside (IPTG) (48). Cell
lysates were prepared by sonication and incubated with glu-
tathione-Sepharose beads (Pharmacia) to isolate the fusion pro-
tein (GST-SAA). After thorough washing of the glutathione-
Sepharose beads with PBS (140 m
, 1.8 mm KH
, pH 7.3), the SAA was cleaved from
the fusion protein with thrombin (50 IU/ ml bead volume) while
it remained attached to the beads, yielding a highly concentrated
solution of lipid-free SAA.
m NaCl, 2.7 mm KCl, 10 mm
Production of antibody
Recombinant murine SAA1.1 produced in the pGEX bacterial
system was used to produce antibody in rabbits. Five hundred
of the recombinant SAA was run in a 10–20% SDS PAGE; the
band was visualized with 4
m sodium acetate and cut out. The gel
bands, mashed using an agate mortar and pestle, were emulsified
with complete Freund’s adjuvant and injected into rabbits intra-
dermally at multiple sites and intramuscularly into the hindquar-
ters of the animal. A booster dose of 250
immunogen was given 6 weeks subsequent to the initial injection,
after which a high titer antiserum was obtained. The specificity
of the antibody was tested by immunoblotting. A single band of
12,000 Da molecular mass was observed when tested against
Antibody against murine apoA-I was similarly prepared. HDL
was obtained from pooled C57BL/ 6 plasma by the two-step pro-
cedure of sequential flotation at d 1.063–1.21 g/ ml and density
gradient centrifugation in a 3–20% NaBr gradient. ApoA-I was
isolated from the HDL peak fraction as a 28,000 Da molecular
mass SDS PAGE band. All other procedures are as outlined
above. The specificity of the antibody was tested by immunoblot-
ting. A single band of 28,000 Da molecular mass was observed
when tested against whole plasma.
To test for the presence of apoE, an antibody against rat apoE
was used. This was prepared using as immunogen apoE as a
34,000 dalton protein band in an SDS PAGE of the peak fraction
of rat plasma isolated by density gradient centrifugation. Anti-rat
apoA-IV prepared in goats was a gift from Dr. Patrick Tso of the
University of Cincinnati. Anti-mouse apoA-II was purchased from
commercial sources (Biodesign, Kennebunk, ME).
g of similarly prepared
Western immunoblotting was performed on plasma or HDL
proteins transferred from SDS PAGE based on the procedure of
Towbin and Gordon (49). Electrotransfer to membrane (Immo-
bilon P, Millipore) was conducted at 360 mA/ gel (Transphor
Electrophoresis Unit, Heoffer) for 1 h; the temperature was
maintained at 10
C by a circulating water bath (Forma Scientific).
Unreacted sites in the membrane were blocked with a 5% solution
of powdered milk in washing buffer (80 m
, 100 mm NaCl, pH 7.5). After thorough washing, appro-
priately diluted primary antibody was reacted with membrane-
bound protein for 2 h with shaking. The membrane was washed,
reacted with peroxidase-labeled secondary antibody for 1 h,
washed, then reacted with a chemiluminescence detection sys-
tem (ECL, Amersham). The protein bands were visualized by re-
action with a photographic film (Hyperfilm ECL, Amersham).
, 20 mm
Non-denaturing immunoprecipitation of HDL was performed
to determine whether HDL included particles containing only
SAA. ApoA-I-containing particles were precipitated from C57BL/
6 HDL while apoE-, apoA-IV-, and apoA-II-containing particles
were precipitated from both the C57BL/ 6 and apoA-I
using the respective antisera and SAA-containing fractions ob-
tained by equilibrium density gradient centrifugation or by
FPLC. Precipitation was performed as follows. Fifty
l of sample
by guest, on June 5, 2013
Cabana et al.SAA-only HDL in apoA-I gene knockout and wild-type mice 1093
was first reacted with pre-titered primary antibody overnight at
C with gentle agitation on a rotator. After the incubation, the
precipitating reagent (recombinant protein G attached to aga-
rose beads, Pharmacia) was added in amounts sufficient to pre-
cipitate all the primary antibody based on the manufacturer’s
recommendation. The mixture was further incubated for 1 h at
room temperature with rotation. The supernatant and precipi-
tate were separated by centrifugation and the precipitate was
washed thoroughly with PBS. Both the precipitate and superna-
tant were analyzed for apolipoprotein content by SDS PAGE and
Quantitation of RNA
Total liver RNA was isolated by the guanidine thiocynate pro-
cedure (47) from apoA-I
mice before and at various times
after injection of LPS. Relative SAA mRNA levels were deter-
mined by the slot blot procedure described previously (50) using
32P-labeled murine Saa1.1 cDNA as a probe.
HDL isolated by density gradient centrifugation were exam-
ined in the Philips CM10 electron microscope (University of
Chicago Core Electron Microscopy Facility) after staining with
sodium phosphotungstate as previously described (17). The di-
ameters of 100 particles were measured from enlarged prints of
All results are expressed as mean ? SEM. Significance was
tested using the paired t-test. Data analyses were performed us-
ing the Minitab Statistical Software for Windows 95.
APR of C57BL/6 mice
Plasma lipids and apolipoprotein levels. Our previous re-
sults (17) have described the changes in plasma lipopro-
teins that occur during the APR in C57BL/ 6 mice. In this
study we have performed a more detailed analysis of these
changes as a reference for the changes that occur in apoA-
I?/ ? mice. Plasma lipid levels at baseline, 10 h, and 24 h
after the injection of LPS to induce the APR are shown in
Table 1. In this group of 6 mice, there was a statistically
significant increase in total cholesterol from 79 ? 2.1 at 0
h to 94.7 ? 6.0 at 10 h (P ? 0.004 vs. 0 h) and 112.5 ? 8.2
mg/ dl at 24 h (P ? 0.018 vs. 0 h, 0.048 vs. 10 h). This in-
crease was reflected in a significant increase in the non-
HDL cholesterol from 13.5 ? 1.6 to 22.0 ? 1.3 at 10 h (P ?
0.010 vs. 0 h) and 50.2 ? 7.5 mg/ dl at 24 h (P ? 0.002
vs. 0 h, 0.015 vs. 10 h) without significant changes in the
HDL cholesterol level. The increase in the non-HDL choles-
terol was not further explored. The mean plasma triglycer-
ides showed a slight but not statistically significant increase.
Despite the lack of changes in plasma HDL cholesterol
level, apoA-I decreased from 118.6 ? 7.8 to 86.6 ? 9.1 at 10
h (P ? 0.012 vs. 0 h) and 92.3 ? 4.9 mg/dl at 24 h (P ?
0.012 vs. 0 h, ns vs. 10 h). SAA increased. We have previ-
ously shown (17) that SAA which was undetectable by im-
munoblotting with anti-mouse SAA at baseline was present
at 2 h post injection and reached peak levels 22 h later.
Distribution of apoA-I and SAA. As apoA-I and SAA are
the primary apolipoproteins in the APR HDL, we exam-
ined the distribution of these apolipoproteins in the HDL
subclasses at 0 h and 24 h post LPS injection. The HDL were
isolated by equilibrium density gradient centrifugation
and gel filtration chromatography. Upon separation by
centrifugation (Fig. 1A), apoA-I displayed a symmetrical
distribution with a peak in fraction 19 (d 1.09 g/ ml) at 0
h. At 24 h, apoA-I displayed a bimodal distribution with
the majority of apoA-I located at fractions 17 and 18 (d
1.07–1.08 g/ ml) and the rest in a shoulder in fractions
20–22 (d 1.10–1.12 g/ ml). There was no significant in-
crease of lipid-free apoA-I in the bottom fraction (data
not shown). In contrast, the majority of SAA was found be-
tween fractions 20–25 with the peak of SAA-containing
particles in fraction 23 (d 1.13 g/ ml). These results are
consistent with our previous observations (17) in which
TABLE 1.Plasma lipid and apoA-I levels of C57BL/ 6 and apoA-I?/ ? mice before and after
injection of lipopolysaccharide
79.0 ? 2.1
94.7 ? 2.4a
112.5 ? 8.2b
23.0 ? 1.2
37.0 ? 4.0f
57.7 ? 5.2g
65.5 ? 3.3
72.7 ? 3.4
62.3 ? 7.0
10.1 ? 0.9
10.3 ? 0.3
24.1 ? 2.8h
13.5 ? 1.6
22.0 ? 1.3c
50.2 ? 7.5d
12.9 ? 1.0
26.7 ? 3.7i
34.1 ? 4.5j
46.0 ? 5.5
58.3 ? 9.1
72.8 ? 13.6
33.6 ? 3.0
31.3 ? 1.8
58.7 ? 1.8k
118.6 ? 7.8
86.6 ? 8.1e
92.3 ? 4.9e
Values are expressed as mean ? SEM. Mice were individually analyzed; na, not analyzed.
aP ? 0.004 vs. 0 h.
bP ? 0.018 vs. 0 h; ? 0.048 vs. 10 h.
cP ? 0.010 vs. 0 h.
dP ? 0.002 vs. 0 h; ? 0.015 vs. 10 h.
eP ? 0.012 vs. 0 h.
fP ? 0.080 vs. 0 h.
gP ? 0.0000 vs. 0 h; ? 0.012 vs. 10 h.
hP ? 0.0002 vs. 0 h and 10 h.
iP ? 0.07 vs. 0 h.
jP ? 0.0004 vs. 0 h; ns vs. 10 h.
kP ? 0.021 vs. 0 h; ? 0.012 vs. 10 h.
by guest, on June 5, 2013
1094Journal of Lipid Research Volume 40, 1999
we demonstrated the conversion of the monodisperse
HDL profile in C57BL/ 6 plasma at 0 h to a bimodal HDL
profile with a new peak around d 1.13 g/ ml (fraction 23).
Isoelectric focusing revealed no differences in the propor-
tion of SAA1 and SAA2 isoforms across the gradient (data
The distinct distribution of apoA-I and SAA suggests
that some SAA-containing particles could contain apoA-I
while other particles may contain SAA devoid of apoA-I.
The results also suggest that the apoA-I-containing parti-
cles may undergo remodeling as SAA levels increase in the
APR as the majority of the apoA-I shifted towards lighter
density particles. This ‘leftward’ shift was progressive, with
the greatest change occurring at 24 h (data not shown).
This remodeling influenced primarily apoA-I-containing
particles, as the peak of apoE-containing particles (peak
fraction 13, d 1.05 g/ ml) was essentially unchanged dur-
ing the APR (data not shown).
The distribution of apoA-I and SAA in the HDL fraction
was also examined after separation of the particles on the
basis of size on FPLC columns. The major peak of the HDL
shifted from fraction 50 in the baseline plasma to fraction
47 in the 24-h APR plasma with a shoulder around frac-
tion 50 (Fig. 2A). This indicates that the majority of the
HDL particles are slightly larger in size in the APR plasma.
The peak of apoA-I in the APR HDL shifted towards
slightly larger particles compared to baseline HDL which
is consistent with the results in Fig. 1A that apoA-I in the
APR is on lighter density particles. Similar to the results
obtained by density gradient centrifugation, the distribu-
tion of SAA on the APR-HDL particles is not identical to
that of apoA-I. The majority of the SAA is located on
larger particles than the apoA-I. A discordance between
size and density of SAA-containing particles has been ob-
served previously by our laboratory in other animal mod-
SAA was located in the fractions containing phospho-
lipid (Fig. 1B). To determine whether the majority of the
SAA was lipid associated, FPLC fractions corresponding to
HDL (fractions 41–53) and the non-lipoprotein proteins
(fractions 55–63) were immunoblotted for the presence
of SAA (Fig. 2B). Only about 10% was present in the non-
lipoprotein fractions, indicating that the majority of the
SAA was lipid associated.
Size of the HDL particles. To further explore changes in
HDL particle size during the APR, particles obtained by
density gradient centrifugation of plasma were analyzed
by non-denaturing gradient gel electrophoresis. HDL
from three time points were used: baseline, 10 h, and 24
h. As shown in Fig. 3A, HDL particles obtained from base-
line plasma (0 h) decreased in size as their density in-
creased (i.e., particles in fraction 15 ? fraction 19 ? frac-
tion 23). While there was no enlargement of the particles
in fraction 15, which contained mostly apoE and small
amounts of apoA-I and SAA after the injection of LPS,
there was a progressive enlargement in particle size with
time in the SAA-containing fractions. For example, parti-
cles in fractions 19 and 23 at both 10 and 24 h were larger
than the particles in the corresponding fractions obtained
before injection. The largest increase in particle size from
baseline during the APR was in fraction 23, 24 h after in-
jection of LPS. This is the peak SAA-containing fraction.
That enlargement of the particles is a function of the level
of SAA has been verified by our in vitro incubation of
plasma with recombinant SAA (unpublished results).
SAA-only HDL. The distinct distributions of apoA-I and
SAA in the fractions obtained by density gradient centrifu-
gation (Fig. 1A) as well as those obtained by FPLC (Fig.
1B) show that some of the SAA and some of the apoA-I
are on distinct particles. To determine whether portions
of the two apolipoproteins, apoA-I and SAA, may be on
separate particles, we immunoprecipitated SAA-containing
fractions obtained by density gradient centrifugation (frac-
C57BL/ 6 plasma at 0 and 24 h post LPS injection. (A) Plasma frac-
tions obtained by equilibrium density gradient centrifugation (n ?
10) were analyzed for the relative levels of different apoproteins.
Shown are SAA (?– --- –?), 0 h A-I (? ––– ?), 24 h A-I (? ––– ?),
and apoE (? ––– ?). (B) Plasma fractions obtained by FPLC (n ?
5) were analyzed for the relative levels of apolipoproteins and lip-
ids. Shown are SAA (?– --- –?), 0 h A-I (? ––– ?), 24 h A-I (? ––– ?)
and 24 h phospholipid ( ? . . . . . . ?). ApoA-I was quantitated by ra-
dial immunodiffusion using antibody to mouse apoA-I and ex-
pressed as percentage of total plasma apoA-I. ApoE and SAA were
estimated by densitometric scanning of Coomassie-stained bands
on SDS PAGE.
Distribution of apolipoproteins in the HDL fractions of
by guest, on June 5, 2013
Cabana et al. SAA-only HDL in apoA-I gene knockout and wild-type mice1095
tions 21–27) under non-denaturing conditions. This anal-
ysis is based on the assumption that all of the apolipopro-
teins found in the same particles should be precipitated by
an antibody directed to one of the apolipoproteins. Figure
4A shows a typical result of the immunoprecipitation ex-
periments using rabbit anti-mouse apoA-I. The unbound
proteins in the supernatant (lanes 1 and 3) and the pre-
cipitated proteins (lanes 2 and 4) were separated by SDS
PAGE and subjected to immunoblotting with anti-apoA-I
(lanes 1 and 2) or anti-SAA (lanes 3 and 4). As is shown in
Fig. 4A, apoA-I was found only in the precipitate (lane 2).
There was no detectable apoA-I in the supernatant (lane
1) indicating that the apoA-I was quantitatively immuno-
precipitated. The same samples probed with anti-SAA
(lanes 3 and 4) revealed the presence of SAA both in the
supernatant (lane 3) and the precipitate (lane 4). Thus,
SAA was present in both the apoA-I-containing particles
(i.e., those immunoprecipitated with anti-mouse apoA-I)
and in particles that did not contain apoA-I (i.e., particles
that remained in the supernatant). More SAA was detect-
able in the supernatant than in the pellet of the precipita-
tion with anti-apoA-I, suggesting that in the SAA-contain-
ing fractions obtained by density gradient centrifugation,
more of the SAA was found in HDL lacking apoA-I (SAA-
HDL) than in HDL containing both apoA-I and SAA (AI-
SAA-HDL) particles. Coomassie R250 staining of SDS
PAGE of fractions 21–27 revealed no apolipoproteins
other than apoA-I and SAA (data not shown). If this in-
deed reflects the protein composition of these fractions,
this is evidence that a significant amount of “HDL” parti-
cles containing SAA as essentially the only apolipoprotein
are present in the APR plasma. Particles in this fraction
(fraction 25 analyzed) had a very high protein (79.4%) and
low cholesteryl ester (6.8%) content resulting in a high sur-
face to core ratio (9.2 vs. 3.3 at baseline) (Table 2).
As further evidence for the existence of particles con-
taining only SAA devoid of other apolipoproteins, the
peak of SAA-containing particles from an FPLC fraction-
ation of C57BL/6 plasma was precipitated by a pool of anti-
bodies including anti-apoA-I, anti-apoE, anti-apoA-II, and
anti-apoA-IV using antibody concentrations that quantita-
tively precipitated the respective antigens when used singly
as shown above for apoA-I and apoA-IV (Fig. 4C). Again,
SAA was found both in the supernatant (lane 1) and pre-
cipitate (lane 2), with a greater proportion of the SAA
found in the supernatant than in the precipitate.
APR of APOA-I?/? mice
SAA induction. Prior to determining whether SAA-HDL
can form in the absence of apoA-I, we examined whether
the absence of apoA-I influences the accumulation of SAA
in the plasma. The APR was induced in apoA-I?/ ? mice by
the injection of LPS and the presence of SAA in the
plasma at various times after injection was examined by
immunoblotting of the whole plasma. From non-detect-
able levels at baseline, SAA was detectable at 4 h after the
tion profile of plasma lipoproteins at 0 h (top) and 24 h post LPS
injection (bottom). Six hundred ?l of plasma was applied to two
Superose 6 columns in tandem. The tracings represent absorbance
at OD280 nm. Note the significant increase and “leftward” shift of
the HDL peak suggesting formation of large particles. (B) Distribu-
tion of SAA. One ?l of each of the odd numbered fractions from
the 24 h plasma (A, bottom) was analyzed by Western immunoblot-
ting using antibody against mouse SAA. Molecular mass standards
are shown at the right. Representative of n ? 5 animals.
SAA in HDL particles of C57BL/ 6 plasma. (A) FPLC elu-
by guest, on June 5, 2013
1096 Journal of Lipid ResearchVolume 40, 1999
injection of LPS and progressively increased thereafter to
peak levels at 24 h (Fig. 5A). Although in the C57BL/ 6
plasma SAA was detectable at 2 h post LPS injection (17),
the magnitude of increase of SAA in the apoA-I?/ ? mice
at 24 h was comparable to that which occurs in C57BL/ 6
mice. The increase in plasma SAA concentration was due
to the induction of the SAA gene expression as reflected
in the increases of hepatic SAA mRNA estimated by dot-
blot analyses using a murine Saa1.1 cDNA probe (Fig.
5B). While no SAA mRNA was detectable at 0 h, the mRNA
levels were detectable at 2 h with peak levels at 24 h. By 48
h, the SAA mRNA decreased 50% from the 24 h level.
Plasma lipid and lipoprotein levels. As SAA can accumu-
late in the plasma of apoA-I?/ ? mice, we next examined
what effect this had on plasma lipid and HDL levels. At
baseline, apoA-I?/ ? mice had very low total and HDL cho-
lesterol levels (Table 1). Total plasma cholesterol was 3- to
4-fold lower in the apoA-I?/ ? mice than in the C57BL/ 6
control mice (23.0 ? 1.2 mg/ dl vs. 79.0 ? 2.1 mg/ dl, re-
spectively) and HDL cholesterol was 6-fold lower (10.1 ?
0.9 mg/ dl vs. 65.5 ? 3.3 mg/ dl, respectively). HDL in the
apoA-I?/ ? mice is defined by density or by that fraction of
by density gradient centrifugation and 7 ?g of protein in selected
HDL fractions was loaded to each lane of a 4–30% non-denaturing
electrophoresis gel and stained with Coomassie R250. (A) C57BL/ 6
plasma, fractions 15, 19, and 23 obtained at 0, 10, and 24 h post
LPS injection; (B) apoA-I?/ ? mice plasma, fractions 15, 19, 23, 25,
and 27 obtained at 0 and 24 h post LPS injection. Particle radii
(nm) of known standards are shown at the right.
Enlargement of SAA-containing HDL. Plasma was separated
done under non-denaturing conditions as outlined in Materials
and Methods. (A) Precipitation with anti-mouse apoA-I using
pooled SAA containing fractions (fractions 21–27) obtained by
equilibrium density gradient centrifugation of 24 h plasma from
C57BL/ 6 mice. The supernatant (lanes 1 and 3) and precipitate
(lanes 2 and 4) were probed with anti-mouse apoA-I (lanes 1 and 2)
or anti-mouse SAA (lanes 3 and 4). ApoA-I (left arrow) and SAA
(right arrow) are indicated. X signifies unidentified antigen cross-
reactive with the protein G-agarose beads. (B) Precipitation with
anti-rat apoA-IV. Pooled SAA-containing fractions (fractions 43 and
45) obtained by FPLC fractionation of 24 h plasma from apoA-I?/ ?
mice were precipitated with anti-rat apoA-IV and probed with anti-
rat apoA-IV (lanes 1–5) or anti-mouse SAA (lanes 6–10). Lanes 1
and 6, anti-A-IV supernatant; lanes 2 and 7, anti-A-IV precipitate;
lanes 3 and 8, protein G bead supernatant; lanes 4 and 9, protein G
bead precipitate; lane 5, molecular mass standard 45 kD; lane 10,
molecular mass standard 14.4 kD. (C) Precipitation with a mixture
of antibodies. A mixture of anti-apoA-I, anti-apoE, anti-apoA-II, and
anti-apoA-IV was used to precipitate C57BL/ 6 HDL (FPLC frac-
tions 43–45) (lanes 1 and 2) and a mixture of anti-apoA-IV, anti-
apoE, and anti-apoA-II was used to precipitate apoA-I?/ ? HDL
(FPLC fractions 43–45) (lanes 3 and 4). These antibodies com-
pletely precipitated the respective antigens when used singly as
shown above for apoA-I and apoA-IV. Lanes 1 and 3, supernatant;
lanes 2 and 4, precipitate; lane 5, molecular mass standard 14.4 kD
probed with anti-mouse SAA.
Immunoprecipitation of HDL. All precipitations were
by guest, on June 5, 2013
Cabana et al. SAA-only HDL in apoA-I gene knockout and wild-type mice1097
total plasma cholesterol not precipitated by MgCl2 and
phosphotungstic acid (see Methods) and clearly does not
imply the presence of apoA-I-containing particles in the
apoA-I?/? mice. Non-HDL cholesterol levels were the same
in the two strains of mice. While both total and HDL cho-
lesterol were significantly lower in the apoA-I?/ ? than in
C57BL/ 6 mice, the plasma triglyceride levels of apoA-I?/ ?
and apoA-I?/ ? mice were similar.
After induction of the APR in the apoA-I?/ ? mice, total
plasma cholesterol increased from 23.0 ? 1.2 to 37.0 ?
4.0 at 10 h (P ? 0.08 vs. 0 h) and 57.7 ? 5.2 mg/ dl at 24 h
(P ? 0.0000 vs. 0 h; 0.012 vs. 10 h) (Table 1). There was a
progressive and steady increase in plasma cholesterol lev-
els up to 30 h post injection (Fig. 5C). The early increases
in total cholesterol were not accompanied by correspond-
ing increases in HDL cholesterol. HDL cholesterol began
to increase only at around 18 h (data not shown) and
reached a maximum 2.4-fold increase at 24 h (24.1 ? 2.8
mg/ dl, P ? 0.0002). By 24 h about 45% of the total
plasma cholesterol increment was in HDL. Non-HDL cho-
lesterol increased from 12.9 ? 1.0 to 26.7 ? 3.7 at 10 h
(P ? 0.07 vs. 0 h) and 34.1 ? 4.5 mg/ dl at 24 h (P ?
0.0004 vs. 0 h; ns vs. 10 h). Thus, while only the non-HDL
cholesterol was increased in apoA-I?/ ? (i.e., C57BL/ 6)
mice, both HDL and non-HDL cholesterol increased in
the apoA-I?/ ? mice. This is also reflected in the distribu-
tion of lipids in the fractions obtained by FPLC fraction-
ation of apoA-I?/ ? plasma (Fig. 6). While both phospho-
lipids (A) and cholesterol (B) increased in all the
lipoprotein fractions, the greatest increment was observed
in the fractions corresponding to HDL (fractions 36–46).
There was a slightly statistically significant increase of tri-
glycerides at 24 h (Table 1). Triglyceride levels returned to
baseline at 30 h (data not shown).
The increase in lipid accompanied formation of new li-
poprotein particles as shown by equilibrium density gradi-
ent centrifugation of apoA-I?/ ? plasma (Fig. 7). Denser
HDL particles were present in the 24 h HDL in fractions
22–27 (d 1.12–1.17 g/ ml) which were not present at 0 h.
Similar to C57BL/ 6 (Fig. 1A), these fractions contained
most of the SAA (see below).
Distribution of apolipoproteins and size of the particles. The
apolipoprotein composition of apoA-I?/ ? HDL was ana-
lyzed from fractions obtained by FPLC and by density gra-
dient centrifugation. The FPLC elution profile of apoA-
I?/ ? plasma (Fig. 8A) was very similar to that obtained in
C57BL/ 6 mice (Fig. 3A). There was a progressive “left-
ward” shift towards large HDL particles and a significant
TABLE 2.Composition of HDL
Number ProteinCEFC PLTG
percent by weight
aN ? one pool of 5 mice per time point; HDL analyzed from frac-
tions obtained by equilibrium density gradient centrifugation (3–20%
NaBr gradient) of whole plasma.
bN ?one pool of 15 mice per time point; HDL analyzed from frac-
tions obtained by a two-step procedure: sequential flotation at d 1.063–
1.25 g/ ml followed by equilibrium density gradient centrifugation (3–
20% NaBr gradient).
cEstimated by densitometric scanning of stained gels, this fraction
contains 91% apoA-I, 5% apoA-II.
dThis fraction contains 76% apoA-I, 4% apoA-II/ C, 17% SAA.
eThis fraction contains 21% apoA-I, negligible apoA-II/ C, 63%
fMost of the protein in this fraction (Fig. 9) is apoA-II/ C (72%).
gThe protein in this fraction is a mixture of apoB-48, apoE, apoA-
II/ C with SAA about half (50.2%) of the HDL apolipoproteins (Fig. 9).
hThis fraction contains mostly SAA (90%) with barely detectable
levels of apoE and apoA-IV (Fig. 9).
tected by Western immunoblotting with anti-mouse SAA using 1 ?l
of samples obtained before and at various times after the injection
of LPS. Std, 14,400 Da. (B) Hepatic SAA mRNA estimated by densi-
tometric scanning of slot blots using a random primed murine
Saa1.1 cDNA probe. (C) Total cholesterol quantitated in plasma
obtained before and at various times after the injection of LPS. Six
pools of 4 mice per pool were used for the 0 h and 24 h values, 1
pool each for the 6, 12, and 18 h values, and 2 pools of 15 mice
each for the 30 h values.
APR changes in apoA-I?/ ? mice. (A) SAA in plasma de-
by guest, on June 5, 2013
1098 Journal of Lipid ResearchVolume 40, 1999
increase of the HDL peak. The beginning of this shift was
evident at 10 h post injection of LPS and after 24 h re-
sulted in the appearance of a new peak centered around
fraction 43. SAA was found in this new peak (Fig. 8B).
Similar to the C57BL/ 6, ?90% of the SAA eluted at frac-
tions 41–47 which are within the main phospholipid peak
in the HDL region (fractions 36–46, Fig. 6A). SAA was
barely detectable in the non-lipoprotein fractions (frac-
tions 54–61) suggesting that all the SAA in the plasma was
The apolipoprotein composition of apoA-I?/ ? plasma
fractions that floated in the HDL density range were ana-
lyzed by SDS PAGE (Fig. 9). As expected, before the in-
duction of the APR, apoA-II was the major apolipoprotein
of HDL (Fig. 9, top). ApoA-II was found mainly in frac-
tions 17–23 (d 1.07–1.13 g/ ml) comprising about 72% of
the apolipoprotein present in these fractions (based on
scanning densitometry of Coomassie-stained gels). ApoA-
IV comprising 12.5% of the apolipoproteins was present in
the same density fractions as apoA-II. ApoE was present
mainly in the fractions of slightly lower density, i.e., frac-
tions 15–19 (d 1.05–1.09 g/ ml) with peak levels at fraction
17 (d 1.06 g/ ml). ApoE was also present in the very high
density fractions (fractions 29–30, d ? 1.25 g/ ml) associ-
ated with little or no lipid. Traces of other proteins were
present but have not been identified.
Twenty four hours after the injection of LPS, the con-
centration and relative distribution of the HDL apolipo-
proteins changed (Fig. 9, bottom). There was an apparent
increase in the apoE content of the acute phase HDL.
However, there was no apparent increase in total plasma
apoE as analyzed by Western immunoblotting (data not
shown) which suggests that during the APR, apoE was re-
distributed from other fractions including the non-lipo-
protein fraction to HDL. The density distribution of apoA-
II and apoA-IV remained unchanged. There may also be
an increase in some high molecular mass proteins (ca.
200,000 Da, e.g., fractions 15–21) of unknown identity.
The most striking change is the predominant presence
of SAA (Fig. 9, bottom). The density distribution of SAA
was distinct from that of apoE, apoA-II, or apoA-IV. SAA was
found mainly between fractions 19–27 (d 1.09–1.17 g/ ml
and comprised 90% of the apolipoproteins present in
fractions 25–27 (d 1.15–1.17 g/ ml).
The composition of the SAA-containing particles was
determined in the fractions obtained by density gradient
centrifugation (fraction 26) as fractions obtained by FPLC
contained many non-lipoprotein proteins. As shown in Ta-
ble 2, these particles are protein-rich (61% protein by
weight), low in esterified cholesterol (11.6% by weight),
and have a high surface to core ratio (6.2). Thus the SAA-
containing particles are protein-rich particles with a large
surface and small lipid core.
Non-denaturing gel electrophoresis of fractions ob-
tained by density gradient centrifugation of plasma
showed that the particles in fractions in which ?90% of
the protein is SAA (fractions 23–27) were larger and more
discrete than the corresponding fractions obtained before
mice. A representative profile from 4 mice is shown. Phospholipids
(A) and cholesterol (B) expressed as mg/ dl were analyzed in all the
even numbered fractions obtained by FPLC chromatography of
plasma as shown in Fig. 2, 0 h ( ? ––– ?) and 24 h (?------?) after
the injection of LPS. The position of albumin (alb) is indicated.
Distribution of plasma lipoprotein lipids in apoA-I?/ ?
was first isolated by sequential flotation at d 1.063–1.25 g/ ml from a
plasma pool of 15 mice at each time point before re-isolation by
equilibrium density gradient ultracentrifugal flotation in a 3–20%
NaBr gradient. The number and densities of the fractions are
shown in the bottom. The tracings represent absorbance at OD280
nm. Top, 0 h; bottom, 24 h.
Density profile of HDL in apoA-I?/ ? mice. HDL (2 mg)
by guest, on June 5, 2013
Cabana et al. SAA-only HDL in apoA-I gene knockout and wild-type mice1099
the injection of LPS (Fig. 3B). Fractions 15 and 19, which
do not contain SAA or in which SAA is not the only pro-
tein, did not increase in size during the APR whereas the
SAA-containing fractions (fractions 23–27) were much
larger than the 0-h fractions and more intensely staining,
suggesting an increase in the number of particles. The in-
crease in size of particles in fractions 23–27 correlated
with the increase of SAA in plasma so that the maximum-
sized particles were obtained at 24 h (data not shown).
Thus, in both the apoA-I?/ ? and the apoA-I?/ ? mice, SAA
was found in dense but large particles. Negative staining
electron microscopy revealed that these particles in 24 h
apoA-I?/ ? HDL appear spherical (Fig. 10B) as do mature
HDL lipoproteins. As we have previously shown (17), al-
though the particles appeared large by non-denaturing
gel electrophoresis (Fig. 3), they were small as analyzed by
negative stain electron microscopy (Fig. 10, r ? ?4 nm
for panel B and ?5 to 6 nm in panel A). The high surface
to core ratio of these particles suggests a very small lipid
core (Table 2) consistent with the small spherical particles
seen in electron photomicrographs.
SAA-only HDL. The SDS PAGE analysis of apoA-I?/? APR
HDL showed that fractions 25–27 obtained by density gradi-
ent centrifugation contained mostly SAA (Fig. 9, bottom),
with minor amounts of apoE and apoA-IV detectable. As
apoE and apoA-IV are known to dissociate from lipoprotein
during centrifugation, we confirmed the presence of SAA-
only HDL by non-denaturing immunoprecipitation of the
peak SAA-containing HDL fractions obtained by FPLC (frac-
tions 43 and 45). The results are shown in Fig. 4B. While all
the apoA-IV was precipitated by the antibody (lane 2), SAA
was present in both the supernatant (lane 6) and the precip-
itate (lane 7). These results show that there are particles that
contain both SAA and apoA-IV and particles containing SAA
without apoA-IV in these HDL fractions. Control lanes 3, 4,
8, and 9 show that the precipitating reagent (Protein G aga-
rose) did not precipitate either apoA-IV or SAA. Precipi-
tating the same fractions with a mixture of anti-apoA-IV, anti-
apoA-II, and anti-apoE titered to quantitatively precipitate
the respective antigens when used singly (Fig. 4C, lanes 3
and 4) showed abundant SAA in the supernatant (lane 3)
and a lesser amount in the precipitate (lane 4) similar to the
results with C57BL/6 (lanes 1 and 2). This indicates that in
these animals too, most of the SAA in plasma are in particles
containing only SAA devoid of other apolipoproteins.
The major finding of this study is that the absence of
apoA-I has relatively little influence on the pattern and as-
sembly of HDL particles containing SAA. Very little of the
plasma SAA even in the apoA-I?/ ? mice is found in the
tion profile of plasma lipoproteins at 0 h (top), 10 h (middle), and
24 h post LPS injection (bottom). Six hundred ?l of plasma was ap-
plied to two Superose 6 columns in tandem. The tracings represent
absorbance at OD280 nm. Note the significant increase and “left-
ward” shift of the HDL peak suggesting formation of large particles.
(B) Distribution of SAA. One ?l of each of the odd numbered frac-
tions from the 24 h plasma (A, bottom) was analyzed by Western
immunoblotting using antibody against mouse SAA. Molecular
mass standards are shown at the right.
Elution profile of SAA in apoA-I?/ ? plasma. (A) FPLC elu-
by guest, on June 5, 2013
1100 Journal of Lipid ResearchVolume 40, 1999
lipid-poor or lipid-free fraction when fractionated either
by FPLC or density gradient centrifugation. A similar ob-
servation has recently been made by Hajri et al. (51). The
study of apoA-I-deficient mice has highlighted the fact
that during the APR there are changes in the cholesterol
and phospholipid content of lipoproteins other than HDL.
We also report the appearance of HDL in both the wild-
type apoA-I?/ ? and apoA-I-deficient mice that contain
SAA as the only or predominant apolipoprotein.
The APR is a complex set of physiological reactions that
influence the whole lipoprotein profile, although most of
the attention has been focused on HDL, as this is where
the major apolipoprotein indicator of this response, SAA,
is found. In the experiments reported in this study, we
have observed changes in levels of VLDL/ LDL in both A-
I?/ ? and A-I?/ ? mice. We have not studied the composi-
tion of these non-HDL lipoproteins in detail. We have pre-
viously noted an increase of plasma triglycerides during
the APR of rabbits and baboons, though not in mice (17).
In the apoA-I-deficient mice, there was a statistically signif-
icant triglyceride increase. This is in accord with the in-
crease in cholesterol and phospholipid in the VLDL frac-
tion (see Fig. 6). In the C57BL/ 6 mice, there was a trend
toward elevation of triglycerides though the triglyceride
response in these animals was more variable and hence
not statistically significant (Table 1). The precise mecha-
nism for these changes is not quite clear. However, several
of the cytokines known to be increased in the APR (i.e.,
IL-1, IL-6, and TNF?) have been shown to increase VLDL
production (52). Whether there is any change in the me-
tabolism of these cytokines in apoA-I-deficient mice re-
mains to be explored.
In C57BL/ 6 mice, there is no change in the HDL cho-
lesterol during the APR despite the transient drop in
apoA-I concentration. On the other hand, in the apoA-I-
deficient mice, there is more than a doubling of HDL cho-
lesterol, mostly due to the accumulation of SAA-contain-
ing lipoproteins (Table 1, Figs. 3, 7, 8, 9). This suggests
that the induction of SAA may promote the formation of
an HDL particle containing mostly SAA. Recently, similar
increases in HDL lipids and SAA in apoA-I knockout mice
have been described (51).
Indeed, one of the major goals of this study was to ascer-
tain whether SAA could be incorporated into a lipoprotein
in the absence of apoA-I. Our results seem to indicate quite
unequivocally that this is the case. In the apoA-I-deficient
mice, SAA accumulated in the plasma with approximately
the same kinetics and to approximately the same extent as in
the wild-type mice. Almost all the SAA co-localizes with lipid
in the HDL region, with very little in the lipid-free zone in
agreement with the studies of Hajri et al. (51) who showed
that 90% and 75% of the SAA applied to a filtration column
eluted with cholesterol in apoA-I?/? and apoA-I ?/? mice,
respectively. The distribution of the apolipoproteins A-II, A-
IV, E, and C is distinct from that of SAA in these animals. In-
deed, many of the SAA-containing particles appear to con-
tain very little, if any, of other apolipoproteins. Much of the
SAA in the plasma of an acute phase apoA-I-deficient animal
is in these SAA “only” HDL particles which seem to consti-
tute a new lipoprotein in the acute phase plasma of these an-
imals. This result appears to be in contradiction to the pub-
lished report of Webb et al. (53), who administered the SAA
of CE/J mice to apoA-I-deficient mice mediated by adenovi-
ral vector transfer, i.e., in the absence of the APR context.
They found that much of the SAA was in the lipoprotein-free
fraction. We do not fully understand this discrepancy. It is
possible that the administration of SAA, carried on an ade-
noviral vector, does not fully duplicate the several changes as-
sociated with the APR. Alternatively, the SAA of CE/J mice
does not behave in the plasma like the SAA produced in
C57BL/6 mice. CE/J SAA has a different amino acid se-
quence (54). Accordingly, we analyzed the APR in CE/J mice
and found that the distribution of SAA in the HDL density
regions is fully consistent with the observation reported for
either C57BL/6 mice or the apoA-I?/? mice, i.e., almost all
the SAA is found associated with the HDL particles (unpub-
lished observation). Our preliminary conclusion is that some
other changes associated with lipoprotein biogenesis in the
APR is a requirement for the assembly of an SAA rich
(“only”) high density lipoprotein.
The finding of an SAA (only) HDL in the apoA-I-deficient
mouse stimulated us to seek such a subclass of HDL in
apoA-I-sufficient mice. Two kinds of evidence argue that
such an HDL subclass exists even in wild-type animals.
teins from each of the odd numbered fractions that floats in the
HDL region upon density gradient centrifugation of plasma was
loaded to each lane of a 10–20% SDS PAGE and stained with Coo-
massie R250. Top, before injection; bottom, 24 h after the injection
Apolipoprotein content of apoA-I?/ ? HDL. Five ?g of pro-
by guest, on June 5, 2013
Cabana et al. SAA-only HDL in apoA-I gene knockout and wild-type mice1101
The FPLC profile of acute phase lipoprotein in C57BL/ 6
mice (Fig. 1B) shows that SAA and apoA-I do not fully co-
localize. For example, fractions 42–44 contain SAA but lit-
tle or no apoA-I. These fractions also contain no apoE,
which localizes to an even larger particle (fraction 36).
Even on density gradient fractionation, the peak contain-
ing the highest SAA (fraction 23, Fig. 1) is in the tail of
the apoA-I distribution. Most compelling is the fact that by
immunoprecipitation with anti-apoA-I antibody under
non-denaturing conditions, a substantial proportion of
SAA remains unprecipitated while all of the other apoli-
poproteins are precipitated (Fig. 4). Even a combination
of all four antisera (anti-A-I, anti-A-IV, anti-A-II, and anti-
E) failed to precipitate all the SAA-containing particles
(Fig. 4C). Thus we conclude that in the wild-type mouse
about a third of the SAA is in a particle lacking other apo-
lipoproteins. In the apoA-I-deficient plasma, more than
half of the particles were non-precipitable by either anti-
apoA-IV, anti-apoA-II, or anti-apoE. If these particles are
formed prior to secretion from the liver, this apolipopro-
tein is capable of organizing a lipoprotein particle, even
though it does not have the same canonical structure of
by sequential flotation of lipoproteins at d 1.063–1.25 g/ ml and further fractionated by density gradient cen-
trifugation. (A) The major peak of 0 h control HDL, fraction 19; (B) the SAA-rich fraction of 24 h HDL, frac-
tion 27 (see Fig. 7). Both are shown at ?116,000 magnification.
Electron photomicrograph of negatively stained HDL from apoA-I?/ ? mice. HDL was first isolated
by guest, on June 5, 2013
1102 Journal of Lipid ResearchVolume 40, 1999
the other apolipoproteins. In the case of SAA, the N ter-
minal 11–12 amino acids seem to be most important in its
lipid association (55, 56). The composition of the SAA
“only” HDL (Table 2) from apoA-I-deficient mice suggests
a small particle with a modest core of cholesteryl ester and
triglyceride. This is confirmed by electron microscopy
The distribution of SAA in A-I?/ ? and A-I?/ ? mice is
very similar which is consistent with SAA being an organiz-
ing apolipoprotein. In the apoA-I-deficient mice, a new li-
poprotein appears to be formed demonstrable at higher
densities on density gradient centrifugation. The increase
in phospholipid and cholesterol in the high density range
is consistent with the formation of a new lipoprotein. How-
ever, not all of the SAA is in this new lipoprotein. Some of
it is associated with apoA-I in the C57BL/ 6 mice and
apoA-IV in apoA-I?/ ? mice (Fig. 5). The addition of SAA
to an HDL particle seems to increase the size of the parti-
cle seen by non-denaturing gel electrophoresis (Fig. 3)
and FPLC (Fig. 1B), despite the smaller size seen on elec-
tron microscopy. We have previously commented upon
the discrepancy in the measurement of the size of SAA-
containing lipoproteins, depending upon whether elec-
tron microscopy or other methodologies are used. We
have postulated that SAA is associated with the lipopro-
tein by its N-terminus, forming a stoke-like arrangement
on the surface of the lipoprotein, an arrangement that is
“seen” differently by electron microscopy (the negative
stain penetrates to the core of the lipoprotein) and by hy-
drodynamic measures as in non-denaturing gel electro-
phoresis where the outer fringes are “seen.” This picture
is compatible with the very high protein content of SAA-
rich lipoproteins (Table 2). It is possible that the associa-
tion of SAA with HDL particles also results in some re-
modeling of the lipid components, though this is not fully
clarified by our results.
During the APR, there is an increase of two isoforms of
SAA, designated SAA1 and SAA2, which are different
gene products. We asked whether one of these isoforms was
particularly associated with the SAA “only” HDL. Isoelectric
focusing indicated that no such preference occurred.
This study clearly indicates that the SAA “only” HDL ex-
ists in vivo. For many years the function of this highly con-
served and highly expressed acute phase reactant has
been sought, so far without a clear resolution. In studies
of the function of this protein, it must be borne in mind
that it is an apolipoprotein predominantly if not exclu-
sively carried on an HDL particle. Presumably, it is in this
form that SAA is accessible to the extravascular tissues.
The significance of the lipoprotein association for SAA
function is not clear. In most functional studies, either re-
combinant SAA or SAA in association with total HDL has
been used. Even in the latter case, a mixture of particles
containing different proportions of SAA and apoA-I is pre-
sented to tissue cells. In this study, we highlight the avail-
ability of an HDL that contains SAA as essentially the only
apolipoprotein. Such particles could represent a valuable
reagent for functional studies in which the confounding
issues of lipoprotein mixture and other apolipoproteins
can be avoided. The capacity to isolate an HDL contain-
ing SAA as the major or only apolipoprotein provides an
important system to further explore the biological func-
tion of SAA in its natural context, i.e., as a major compo-
nent of a lipoprotein.
The authors wish to thank Dr. Patrick Tso of the University of
Cincinnati for the gift of anti-rat apoA-IV, Mr. Michael Machura
for technical assistance, and Miss Yimei Chen of the Laboratory
for Image Analysis and Electron Microscopy of the University
of Chicago for assistance in the preparation of electron micro-
graphs. This work is supported by AHA 96015880 (Formation
and Interaction of Acute Phase HDL).
Manuscript received 19 November 1998 and in revised form 26 February 1999.
1. Akgun, S., N. H. Ertel, A. Mosenthal, and W. Osser. 1998. Postsur-
gical reduction of serum lipoproteins: interleukin-6 and the acute
phase response. J. Lab. Clin. Med. 131: 103–108.
2. Kumon, Y., Y. Nakauchi, K. Kidawara, M. Fukushima, S. Kobayashi,
Y. Ikeda, T. Suehiro, K. Hashimoto, and J. D. Sipe. 1998. A longitu-
dinal analysis of alteration in lecithin-cholesterol acyltransferase
and paraoxonase activities following laparoscopic cholecystectomy
relative to other parameters of HDL function and the acute phase
response. Scand. J. Immunol. 48: 419–424.
3. Alvarez, C., and A. Ramos. 1986. Lipids, lipoproteins and apopro-
teins in serum during infection. Clin. Chem. 32: 142–145.
4. Sammalkorpi, K., V. Valtonen, Y. Kerttula, E. Nikkila, and M. R.
Taskinen. 1988. Changes in serum lipoprotein pattern induced by
acute infections. Metabolism. 37: 859–865.
5. de Luis, D. A., M. Lahera, R. Canton, D. Boixeda, A. L. San Roman,
R. Aller, and H. de la Calle. 1998. Association of Helicobacter pylori
infection with cardiovascular and cerebrovascular disease in dia-
betic patients. Diabetes Care. 21: 1129–1132.
6. Memon, R. A., R. Hussain, J. G. Raynes, A. Lateff, and T. J. Chiang.
1996. Alterations in serum lipids in lepromatous leprosy patients
with and without ENL reactions and their relationship to acute
phase proteins. Int. J. Lepr. Other Mycobact. Dis. 64: 115–122.
7. Fahie-Wilson, M., R. Mills, and K. Wilson. 1987. HDL cholesterol
and the acute phase reaction following myocardial infarction and
acute pancreatitis. Clin. Chim. Acta. 167: 197–209.
8. Bienvenu, J., P. Deshaires, H. Bernon, P. Armanet, P. Peristeris, A.
Lepape, and J. P. Perdrix. 1988. Proteine serique amyloide A
(SAA) et HDL. Implication clinique en reanimation chirurgicale.
Ann. Biol. Clin. 46: 343–346.
9. Bentz, M. H., and J. Magnette. 1998. Hypocholesterolemia during
the acute phase of an inflammatory reaction of infectious origin.
Rev. Med. Interne. 19: 168–172.
10. Cabana, V. G., A. A. Gidding, G. S. Getz, J. Chapman, and S. T.
Shulman. 1997. Serum amyloid A and high density lipoprotein
participate in the acute phase response of Kawasaki disease. Ped.
Res. 42: 651–655.
11. Chiang, A. N., B. Hwang, C. C. Shaw, B. C. Lee, J. H. Lu, C. C.
Meng, and P. Chou. 1997. Changes in plasma levels of lipids and li-
poprotein composition in patients with Kawasaki disease. Clin.
Chim. Acta. 260: 15–26.
12. Kim, H., H. Yamaguchi, K. Inamo, T. Okada, and K. Harada. 1995.
Changes in apolipoproteins during the acute phase of Kawasaki
disease. Acta Paediat. Japon. 37: 672–676.
13. Blackman, J. D., V. G. Cabana, and T. Mazzone. 1993. The acute
phase response and associated lipoprotein abnormalities accom-
panying lymphoma. J. Intern. Med. 233: 201–204.
14. Hachem, H., G. Favre, and G. Soula. 1988. Evidence for qualitative
abnormalities in high-density lipoproteins from myeloma patients:
the presence of serum amyloid A protein could explain HDL ab-
normalities. Biochim. Biophys. Acta. 963: 271–277.
15. Jacobs, D. R., B. Hebert, P. J. Schreiner, S. Sidney, C. Iribarren, and
S. Hulley. 1997. Reduced cholesterol is associated with recent mi-
nor illness. The CARDIA study. Am. J. Epidemiol. 146: 558–564.
16. Gordon, B. R., T. S. Parker, D. M. Levine, S. D. Saal, J. C. L. Wang,
by guest, on June 5, 2013
Cabana et al. SAA-only HDL in apoA-I gene knockout and wild-type mice1103 Download full-text
B. H. Sloan, P. S. Parie, and A. L. Rubin. 1996. Low lipid concen-
trations in critical illness: implications for preventing and treating
endotoxemia. Crit. Care Med. 24: 584–589.
17. Cabana, V. G., J. R. Lukens, K. S. Rice, T. J. Hawkins, and G. S.
Getz. 1996. HDL content and composition in APR in three spe-
cies: triglyceride enrichment of HDL a factor in its decrease. J.
Lipid Res. 37: 2662–2674.
18. Parks, J. S., and L. L. Rudel. 1985. Alteration of high density lipo-
protein subfraction distribution with induction of serum amyloid
A protein (SAA) in the non-human primate. J. Lipid Re s. 26: 82–91.
19. Cabana, V. G., J. N. Siegel, and S. M. Sabesin. 1989. Effects of the
acute phase response on the concentration and density distribu-
tion of plasma lipids and apolipoproteins. J. Lipid Res. 30: 39–49.
20. Vega, G. L., and S. M. Grundy. 1996. Hypoalphalipoproteinemia
(low high density lipoprotein) as a risk factor for coronary heart
disease. [Review]. Curr. Opin. Lipidol. 7: 209–216.
21. Betts, J. C., M. R. Edbrooke, R. V. Thakker, and P. Woo. 1991. The
human acute-phase serum amyloid A gene family: structure, evolu-
tion and expression in hepatoma cells. Scand. J. Immunol. 34: 471–
22. Pepys, M. B., and M. L. Baltz. 1983. Acute phase proteins with spe-
cial reference to C-reactive protein and related proteins (pentax-
ins) and serum amyloid A protein. Adv. Immunol. 34: 141–212.
23. Urieli-Shoval, S., R. L. Meek, R. H. Hanson, N. Eriksen, and E. P.
Benditt. 1994. Human serum amyloid A genes are expressed in
monocyte/ macrophage cell lines. Am. J. Pathol. 145: 650–660.
24. Ramadori, G., D. Sipe, and H. R. Colten. 1985. Expression and
regulation of murine serum amyloid A (SAA) gene in extrahepatic
sites. J. Immunol. 135: 3645–3647.
25. Benditt, E. P., and R. L. Meek. 1989. Expression of the third mem-
ber of the serum amyloid A gene family in mouse adipocytes. J.
Exp. Med. 169: 1841–1846.
26. Meek, R. L., N. Eriksen, and E. P. Benditt. 1992. Murine serum
amyloid A3 is a high density apolipoprotein and is secreted by
macrophages. Proc. Natl. Acad. Sci. USA. 89: 7949–7952.
27. Liuzzo, G., L. M. Biasucci, J. R. Gallimore, R. L. Grillo, A. G. Re-
buzzi, M. B. Pepys, and M. D. Maseri. 1994. The prognostic value
of C-reactive protein and serum amyloid A protein in severe unsta-
ble angina. N. Engl. J. Med. 331: 417–424.
28. Casl, M. T., B. Surina, I. Glojnaric-Spasic, E. Pape, N. Jagarinec,
and S. Kranjcevic. 1995. Serum amyloid A protein in patients with
myocardial infarction. Ann. Clin. Biochem. 32: 196–200.
29. Ridker, P. M., N. Rifai, M. A. Pfeffer, F. M. Sacks, L. A. Moye, S.
Goldman, G. C. Flaker, and E. Braunwald. 1998. Inflammation,
pravastatin, and the risk of coronary events after myocardial in-
farction in patients with average cholesterol levels. Cholesterol and
Recurrent Events (CARE) investigators. Circulation. 98: 839–844.
30. Blum, A., G. Kaplan, N. Vardinon, I. Yust, M. Burke, S. Laniado,
and H. Miller. 1998. Serum amyloid type A may be a predictor of
restenosis. Clin. Cardiol. 21: 655–658.
31. Van Lenten, B. J., S. Y. Hama, F. C. de Beer, D. M. Stafforini, T. M.
McIntyre, S. M. Prescott, B. N. La Du, A. M. Fogelman, and M. Na-
vab. 1995. Anti-inflammatory HDL becomes pro-inflammatory
during the acute phase response. J. Clin. Invest. 96: 2758–2767.
32. Liang, J-s., and J. D. Sipe. 1995. Recombinant human serum amy-
loid A (apoSAAp) binds cholesterol and modulates cholesterol
flux. J. Lipid Res. 36: 37–47.
33. Banka, C. L., T. Yuan, M. C. de Beer, M. Kindy, L. K. Curtiss, and
F. C. de Beer. 1995. Serum amyloid A (SAA): influence on HDL-
mediated cellular cholesterol efflux. J. Lipid Res. 36: 1058–1065.
34. Meek, R. L., S. Urieli-Shoval, and E. P. Benditt. 1994. Expression of
apolipoprotein serum amyloid A mRNA in human atherosclerotic
lesions and cultured vascular cells: implications for serum amyloid
A function. Proc. Natl. Acad. Sci. USA. 91: 3186–3190.
35. Yamada, T., T. Kakihara, T. Kamishima, T. Fukuda, and T. Kawai.
1996. Both acute phase and constitutive serum amyloid A are
present in atherosclerotic lesions. Pathol. Int. 46: 797–800.
36. Badolato, R., J. M. Wang, W. J. Murphy, A. R. Lloyd, D. F. Michiel,
L. L. Bausserman, D. J. Kelvin, and J. J. Oppenheim. 1994. Serum
amyloid A is a chemoattractant: induction of migration, adhesion,
and tissue infiltration of monocytes and polymorphonuclear leu-
kocytes. J. Exp. Med. 180: 203–209.
37. Coetzee, G. A., A. F. Strachan, D. R. van der Westhuyzen, H. Hoppe,
M. S. Jeenah, and F. C. de Beer. 1986. Serum amyloid A-containing
human high density lipoprotein 3. J. Biol. Che m. 261: 9644–9651.
38. Williamson, R., D. Lee, J. Hagaman, and N. Maeda. 1992. Marked
reduction of high density lipoprotein cholesterol in mice geneti-
cally modified to lack apolipoprotein A-I. Proc. Natl. Acad. Sci. USA.
39. Schumaker, V. N., and D. L. Puppione. 1986. Sequential flotation
ultracentrifugation. Methods Enzymol. 128: 155–170.
40. Lowry, O. H., N. J. Rosebrough, A. L. Farr, and R. J. Randall. 1951.
Protein measurement with the Folin phenol reagent. J. Biol. Chem.
41. Markwell, M. A., S. M. Hass, L. L. Beiber, and N. E. Tolbert. 1978.
A modification of the Lowry procedure to simplify protein deter-
mination in membrane and lipoprotein samples. Anal. Biochem.
42. Albers, J. J., P. W. Wahl, V. G. Cabana, and W. R. Hazzard. 1975.
Quantitation of human plasma high density lipoprotein: relation-
ship to age, sex and lipid levels. Metabolism. 24: 633–644.
43. Weber, K., and M. Osborne. 1969. The reliability of molecular
weight determination by dodecyl sulfate-polyacrylamide gel elec-
trophoresis. J. Biol. Chem. 244: 4406–4412.
44. Blanche, P., E. Gong, T. Forte, and A. Nichols. 1981. Characteriza-
tion of human high density lipoproteins by gradient gel electro-
phoresis. Biochim. Biophys. Acta. 665: 408–419.
45. Kashyap, M. L., B. A. Hynd, K. Robinson, and P. S. Gartside. 1981.
Abnormal preponderance of sialylated apolipoprotein C-III in
triglyceride-rich lipoproteins in type V hyperlipoproteinemia. Me-
tabolism. 30: 111–118.
46. Nomenclature Committee of the International Society of Amy-
loidosis. J. Sipe, Subcommittee Chairperson. 1999. Int. J. Exp. Clin.
Invest. In press.
47. Chirgwin, J., A. Przybyla, R. MacDonald, and W. Rutter. 1979. Isola-
tion of biologically active ribonucleic acid from sources enriched
in ribonucleases. Biochemistry. 18: 5294–5299.
48. Pharmacia Biotech, Inc. 1977. GST Gene Fusion System, 3rd Edi-
tion, Revision 1.
49. Towbin, H., and J. Gordon. 1984. Immunoblotting and dot blot-
ting—current status and outlook. J. Immunol. Methods. 7: 313–340.
50. Kushwaha, R. S., P. H. Barrett, C. A. Reardon, D. S. Lewis, K. D.
Carey, G. S. Getz, and H. C. McGill, Jr. 1995. Relationships of
plasma and hepatic variables with rates of plasma low-density lipo-
protein apolipoprotein B metabolism in baboons fed low- and
high-fat diets. Metabolism. 44: 1058–1066.
51. Hajri, T., R. Elliott-Bryant, J. D. Sipe, J. S. Liang, K. C. Hayes, and
E. S. Cathcart. 1998. The acute phase response in apolipoprotein
A-I knockout mice: apolipoprotein serum amyloid A and lipid dis-
tribution in plasma high density lipoproteins. Biochim. Biophys.
Acta. 1394: 209–218.
52. Feingold, K. R., I. Hardardottir, and C. Grunfeld. 1998. Beneficial
effects of cytokine induced hyperlipidemia (Review). Z. Ernah-
rung. 37 (Suppl 1): 66–74.
53. Webb, N. R., M. C. de Beer, D. R. van der Westhuyzen, M. S. Kindy,
C. L. Banka, K. Tsukamoto, D. L. Rader, and F. C. de Beer. 1997.
Adenoviral vector-mediated overexpression of serum amyloid A in
apoA-I-deficient mice. J. Lipid Res. 38: 1583–1590.
54. de Beer, M. C., F. C. de Beer, W. D. McCubbin, C. M. Kay, and M. S.
Kindy. 1993. Structural prerequisites for serum amyloid A fibril
formation. J. Biol. Chem. 268: 20606–20612.
55. Turnell, W., R. Sarra, I. D. Glover, J. Baum, D. Capsi, M. L. Baltz,
and M. B. Pepys. 1986. Secondary structure prediction of human
SAA1. Presumptive identification of calcium and lipid binding
sites. Mol. Biol. Med. 3: 387–407.
56. Patel, H., J. Bramall, H. Waters, M. C. de Beer, and P. Woo. 1996.
Expression of recombinant human serum amyloid A in mamma-
lian cells and demonstration of the region necessary for high-density
lipoprotein binding and amyloid fibril formation by site-directed
mutagenesis. Biochem. J. 318: 1041–1049.
by guest, on June 5, 2013