1991); G. M. Kaminska and J. Y. Niederkorn, Investig.
Ophthalmol. Vis. Sci. 34, 222 (1993).
5. A. P. Schachat and R. B. Murphy, Eds., Retina (Mosby,
Baltimore, MD, ed. 2, 1994).
6. H.-J. Xu et al., Cancer Res. 51, 4481 (1991).
7. PEDF was purified from WERI-Rb-27R (6) serum-free
conditioned media by sequential steps consisting of
dialysis (molecular mass cutoff, 30 kD) against dis-
tilled water, 60 to 95% ammonium sulfate precipi-
tation, step elution from lentil lectin Sepharose 4B
(Pharmacia) with 0.5 M ?-methyl-D-mannopyrano-
side, and elution from a HiTrap heparin Sepharose
column (Pharmacia) with increasing NaCl gradient.
Purification was monitored by an endothelial cell
migration assay (26), and the yield was 17.5%. Ed-
man degradation of proteolytically derived internal
peptides of the protein yielded two unambiguous
sequences (TSLEDFYLDEERTVRVPMMXD and IAQL-
PLTGXM) (27). A BLAST protein homology search
revealed that PEDF contains identical sequences.
8. S. P. Beccerra, in Chemistry and Biology of Serpins, F.
C. Church et al., Eds. (Plenum, New York, 1997), pp.
9. J. Tombran-Tink, G. J. Chader, L. V. Johnson, Exp. Eye
Res. 53, 411 (1991); F. R. Steele et al., Proc. Natl.
Acad. Sci. U.S.A. 90, 1526 (1993).
10. T. Tanawaki, S. P. Becerra, G. J. Chader, J. P. Schwartz,
J. Neurochem. 64, 2509 (1995); Y. Sugita et al.,
J. Neurosci. Res. 49, 710 (1997).
11. R. J. Pignolo, V. J. Cristofalo, M. O. Rotenberg, J. Biol.
Chem. 268, 8949 (1993); J. Tombran-Tink et al.,
J. Neurosci. 15, 4992 (1995).
12. S. P. Becerra et al., J. Biol. Chem. 268, 23148 (1993);
S. P. Becerra et al., ibid. 270, 25992 (1995).
13. D. W. Dawson, O. V. Volpert, P. Gillis, unpublished
14. Human PEDF cDNA was engineered by polymerase
chain reaction to encode a COOH-terminal hexa-
histidine tag, cloned into pCEP4 (Invitrogen), and
transfected into human embryonic kidney cells.
Recombinant PEDF was purified from the condi-
tioned media with the Xpress Protein Purification
16. For preparation of stromal extract, corneas were
freed of associated epithelium and as much of the
endothelium as possible, washed extensively in ice-
cold phosphate-buffered saline (PBS, pH 7.4), and
minced into small fragments that were incubated for
24 hours in PBS containing 0.5 mM phenylmethane-
sulfonyl fluoride. The extract was filter sterilized,
stored at –80°C, and tested in migration assays at a
final concentration of 10 ?g of protein per milliliter.
17. Y.-Q. Wu and S. P. Becerra, Investig. Ophthalmol. Vis.
Sci. 37, 1984 (1996).
18. L. P. Aiello et al., N. Engl. J. Med. 331, 1480 (1994);
A. P. Adamis et al., Am. J. Ophthalmol. 118, 445
19. N. Ogata et al., Curr. Eye Res. 16, 9 (1997); K.
Hayasaka et al., Life Sci. 63, 1089 (1998); S. A.
Vinores et al., J. Neuroimmunol. 89, 43 (1998).
20. L. P. Aiello et al., Proc. Natl. Acad. Sci. U.S.A. 92,
10457 (1995); A. P. Adamis et al., Arch. Ophthalmol.
114, 66 (1996); J. M. Provis et al., Exp. Eye Res. 65,
21. L. E. H. Smith et al., Invest. Ophthalmol. Vis. Sci. 35,
22. E. A. Pierce, E. D. Foley, L. E. Smith, Arch. Ophthalmol.
114, 1219 (1996).
23. S. E. Connolly et al., Microvasc. Res. 36, 275 (1988).
24. M. A. Goldberg, S. P. Dunning, H. F. Bunn, Science
242, 1412 (1988).
25. C. J. Gulledge and M. W. Dewhirst, Anticancer Res. 16,
26. Migration assays were performed in quadruplicate for
each sample with bovine adrenal capillary endothelial
cells or human dermal microvascular endothelial cells
(Clonetics, San Diego, CA) as described (28). To com-
bine multiple experiments, we first subtracted back-
ground migration (Bkgd) toward vehicle (0.1% bovine
serum albumin) and then normalized data by setting
maximum migration toward inducer alone to 100%.
All experiments were repeated two to five times.
Statistics were performed on raw data before nor-
malization with the Student’s t test. Standard errors
were converted to percentages.
27. Single-letter abbreviations for the amino acid resi-
dues are as follows: A, Ala; D, Asp; E, Glu; F, Phe; G,
Gly; I, Ile; L, Leu; M, Met; P, Pro; Q, Gln; R, Arg; S, Ser;
T, Thr; V, Val; X, any amino acid; and Y, Tyr.
28. P. J. Polverini, N. P. Bouck, F. Rastinejad, Methods
Enzymol. 198, 440 (1991).
29. PEDF antipeptide antibody (anti-PEDF) was raised in
rabbits against a peptide containing PEDF amino acids
327 to 343, conjugated to Keyhole-limpet hemocyanin,
and affinity-purified on a peptide column. Polyclonal
antisera against bacterial recombinant PEDF/EPC-1 (an-
ti–EPC-1) [B. R. DiPaolo, R. J. Pignolo, V. J. Cristofalo,
Exp. Cell Res. 220, 178 (1995)] and the antiangiogenic
protein angiostatin [M. S. O’Reilly et al., Cell 79, 315
(1994)] were gifts. Purchased reagents included neutral-
izing anti-VEGF (Genzyme, Cambridge, MA), pan anti-
bodies to TGF?, and all angiogenic inducers (R & D
Systems, Minneapolis, MN) except lysophosphatidic
acid (Sigma). All proteins and antibodies were exten-
sively dialyzed against PBS before use in biological
30. Human vitreous fluid was withdrawn from three
cadaveric eyes (refrigerated within 1.4 to 4.5 hours of
death) obtained from individuals without ocular dis-
ease. Fluid was frozen until used. Fresh vitreous fluid
was obtained from bovine and mouse eyes.
31. We thank A. Mountz for VEGF measurements; B.
Kennedy and the Midwest Eye Banks and Transplanta-
tion Center for human eye tissue; M. K. Francis and V.
Cristofalo for anti–EPC-1; M. O’Reilly and J. Folkman for
bovine capillary endothelial cells and angiostatin; and C.
Hawkins, R. O’Grady, and Y. Mu for assistance with
retinoblastomas. Supported by the National Eye Insti-
cer Institute, and the Chicago Baseball Charities.
15 March 1999; accepted 3 June 1999
HMG-1 as a Late Mediator of
Endotoxin Lethality in Mice
Haichao Wang,1,3* Ona Bloom,3Minghuang Zhang,3
Jaideep M. Vishnubhakat,3Michael Ombrellino,2,3Jiantu Che,3
Asia Frazier,2,3Huan Yang,3Svetlana Ivanova,3
Lyudmila Borovikova,3Kirk R. Manogue,3Eugen Faist,4
Edward Abraham,5Jan Andersson,6Ulf Andersson,7
Patricia E. Molina,2Naji N. Abumrad,2Andrew Sama,1
Kevin J. Tracey2,3
release large quantities of tumor necrosis factor (TNF) and interleukin-1 (IL-1),
of TNF and IL-1 have shown limited efficacy in clinical trials, possibly because
these cytokines are early mediators in pathogenesis. Here a potential late
mediator of lethality is identified and characterized in a mouse model. High
mobility group–1 (HMG-1) protein was found to be released by cultured mac-
showed increased serum levels of HMG-1 from 8 to 32 hours after endotoxin
exposure. Delayed administration of antibodies to HMG-1 attenuated endo-
toxin lethality in mice, and administration of HMG-1 itself was lethal. Septic
patients who succumbed to infection had increased serum HMG-1 levels, sug-
gesting that this protein warrants investigation as a therapeutic target.
Mortality rates for systemic bacterial infec-
tion have not declined significantly, despite
advances in antibiotic therapy and intensive
care. Bacteria do not directly cause lethal
shock and tissue injury. Rather, bacterial en-
dotoxin (lipopolysaccharide, LPS) stimulates
the acute, early release of cytokines such as
TNF and IL-1? from macrophages, and it is
these host products that mediate damage (1).
Macrophages from C3H/HeJ mice do not re-
lease TNF and IL-1 when stimulated by LPS;
these animals are resistant to LPS lethality
(2). Normal, LPS-responsive mice can be
protected from lethal endotoxemia by thera-
peutic agents that selectively inhibit cytokine
action or prevent cytokine release (3).
Translating these pathogenic insights into
clinical therapy has proved difficult, in part
because these “early” mediators (TNF and
IL-1) are released within minutes after LPS
exposure (4). Thus, even a minimal delay in
treatment directed against TNF or IL-1 is
ineffective (3, 5). Paradoxically, LPS-respon-
sive mice treated with lethal doses of LPS
succumb at latencies of up to 5 days, long
1Department of Emergency Medicine and
ment of Surgery, North Shore University Hospital–
New York University School of Medicine, Manhasset,
NY 11030, USA.3The Picower Institute for Medical
Research, Manhasset, NY 11030, USA.4Department
of Surgery, Klinicum Grosshadern, Ludwig-Maximil-
ians University, Munich, Germany.5Division of Pul-
monary Sciences and Critical Care Medicine, Univer-
sity of Colorado Health Sciences Center, Denver, CO
80262, USA.6Department of Infectious Disease, Karo-
linska Institute, Huddinge University Hospital, Stock-
holm, Sweden.7Department of Rheumatology, Astrid
Lindgren’s Children’s Hospital, Karolinska Institute,
*To whom correspondence should be addressed. E-
R E P O R T S
9 JULY 1999 VOL 285SCIENCEwww.sciencemag.org
after serum TNF and IL-1 have returned to
basal levels. Moreover, mice deficient in
TNF die within several days of LPS admin-
istration (6), suggesting that mediators other
than TNF might contribute causally to endo-
To identify potential “late” mediators of
endotoxemia, we stimulated murine macro-
phage-like RAW 264.7 cells with LPS and
analyzed the conditioned culture medium
by SDS–polyacrylamide gel electrophore-
sis (PAGE). LPS stimulation for 18 hours
induced the appearance of a 30-kD protein
that was not apparent at earlier time points.
The NH2-terminal sequence of this late-ap-
pearing factor (Gly-Lys-Gly-Asp-Pro-Lys-
identical to murine HMG-1, a 30-kD mem-
ber of the high mobility group (HMG)
nonhistone chromosomal protein family (7,
8). Based on the HMG-1 sequence in Gen-
Bank (accession no. M64986), we designed
primers and isolated HMG-1 cDNA after
polymerase chain reaction (PCR) amplifi-
cation. Recombinant HMG-1 (rHMG-1)
protein was expressed in Escherichia coli,
purified to homogeneity, and used to gen-
erate polyclonal antibodies (9).
Immunoblot analysis revealed that large
amounts of HMG-1 were released from RAW
264.7 cells in a time-dependent manner (Fig.
1A), beginning 6 to 8 hours after stimulation
with LPS. Cell viability, as judged by trypan
blue exclusion and lactate dehydrogenase re-
lease, was unaffected by LPS concentrations
that induced the release of HMG-1, indicat-
ing that HMG-1 release was not due to cell
death. HMG-1 mRNA levels were unaffected
by LPS treatment (Fig. 1B), indicating that
HMG-1 release is unlikely to be linked to
increased transcription of the gene. Stimula-
tion of RAW 264.7 cells for 18 hours with
TNF (5 to 100 ng/ml) or IL-1? (5 to 100
ng/ml) also induced HMG-1 release in a cy-
tokine dose-dependent manner. In contrast,
stimulation with interferon-? (IFN-?) alone
did not induce HMG-1 release, even at con-
centrations up to 100 U/ml; however, IFN-?
increased by three- to fivefold the amount of
HMG-1 released by stimulation with either
TNF or IL-1 (10, 11). Pulse labeling experi-
most of the HMG-1 released during the first
12 hours after TNF and IFN-? stimulation
was derived from a preformed protein pool.
Radioactivity was incorporated into newly
synthesized HMG-1 from 12 to 36 hours after
macrophage stimulation (10, 11).
We next examined the inducible release of
HMG-1 from other cell types. LPS triggered
HMG-1 release from human primary peripheral
blood mononuclear cells and primary macro-
phages from LPS-sensitive mice (C3H/HeN),
but not from macrophages from LPS-resistant
C3H/HeJ mice (11, 12). Human primary T
35S-methionine revealed that
cells, rat adrenal (PC-12) cells, and rat primary
kidney cells did not release HMG-1 after stim-
ulation with LPS, TNF, or IL-1?. Like other
macrophage products (for example, TNF, IL-
1?, and macrophage migration inhibitory fac-
tor), HMG-1 lacks a classical secretion signal
sequence, so the mechanism of release remains
to be determined.
To determine if HMG-1 was released sys-
temically during endotoxemia in mice, we
measured serum HMG-1 levels after LPS
administration. Serum HMG-1 was readily
detectable 8 hours after administration of a
median lethal dose (LD50) of LPS and was
maintained at peak, plateau levels from 16 to
32 hours after LPS treatment (Fig. 1C).
About 20 to 50 ?g of HMG-1 was released
into the murine circulation within 24 hours
after endotoxin administration [assuming a
distribution half-life (t1/2) of 3 min and an
elimination t1/2of 20 min]; this is compara-
ble to the quantity of TNF and IL-1 released
by LPS treatment. The kinetics of HMG-1
appearance in the blood of LPS-treated mice
differs from that of previously described le-
thal LPS-induced mediators.
Passive immunization of unanesthetized
(anti–HMG-1) 30 min before a lethal dose
(LD100) of LPS did not prevent LPS-induced
death (Fig. 2A). Based on the kinetics of
HMG-1 accumulation in serum (Fig. 1C), and
the relatively short biological half-life of anti-
bodies to cytokines (3, 13), we reasoned that
complete neutralization of a late-appearing me-
diator might require repeated dosing. Adminis-
tration of anti–HMG-1 in two doses (one 30
min before LPS and one 12 hours after LPS)
increased the survival rate of the mice to 30%.
With three doses of antiserum (?30 min, ?12
hours, ?36 hours), 70% of the treated mice
survived, as compared with 0% survival in
controls treated with three matched doses of
preimmune serum (P ? 0.05). No late death
occurred over 2 weeks, indicating that anti–
HMG-1 did not merely delay the onset of LPS
lethality, but provided lasting protection.
To investigate whether antibody treatment
could be delayed until after administration of
LPS, we injected anti–HMG-1 beginning 2
hours after LPS (followed by additional doses
at 12 and 36 hours after LPS). This delayed
against an LD100of LPS (Fig. 2B). Preimmune
Fig. 1. (A) Release of HMG-1 from cultured
macrophages after stimulation with LPS. Mu-
rine macrophage-like RAW 264.7 cells (Ameri-
can Type Culture Collection, Rockville, Mary-
land) were cultured in RPMI 1640 medium,
10% FBS, and 1% glutamine. When 70 to 80%
confluence was reached, cells were resuspend-
ed in serum-free OPTI-MEM I medium and
seeded onto tissue culture plates (5 ? 106cells
per well). After 2 hours, RAW 264.7 cells were
treated with LPS (E. coli 0111:B4, 100 ng/ml)
and proteins in the cell-conditioned medium
were fractionated by SDS-PAGE, excised from
Coomassie- stained SDS-PAGE gels, and subject-
ed to NH2-terminal sequencing analysis (Com-
monwealth Biotechnologies, Richmond, Virginia).
Polyclonal antisera against purified recombinant
HMG-1 were generated in rabbits (Biosynthesis,
Lewisville, Texas); immunoblotting showed that
antiserum reacted with native HMG-1 released
by RAW cells (inset). HMG-1 levels were mea-
sured by optical intensity of bands on immuno-
blots with NIH 1.59 image software, with refer-
ence to standard curves generated with purified
rHMG-1. Data are shown as the mean ? SE (n ?
3). (B) Expression of HMG-1 mRNA in macro-
phages. Murine macrophage-like RAW 264.7 cells
were cultured in RPMI 1640, 10% FBS, and 1%
glutamine, and stimulated with LPS (1 ?g/ml) for
isolated with the SV Total RNA Isolation System
(Promega) and levels of HMG-1 mRNA were de-
termined by reverse transcriptase (RT)–PCR with
the Access RT-PCR System (Promega; ?-actin
T-3? and 5?-CCTAGAAGCATTTGCGGTGCAC-
GATG-3?; and HMG-1 primers, 5?-ATGGGCA-
AAGGAGATCCTA-3? and 5?-ATTCATCATCAT-
CATCTTCT-3?). (C) Accumulation of HMG-1 in
serum of LPS-treated mice. Male Balb/C mice (20 to 23 g) were treated with LPS [10 mg/kg,
intraperitoneally (ip)]. Serum was assayed for HMG-1 by immunoblotting; the detection limit is?50 pg.
Data are shown as the mean ? SE (n ? 3).
R E P O R T S
www.sciencemag.orgSCIENCE VOL 2859 JULY 1999
serum–treated controls all developed lethargy,
groomed and active, had no diarrhea, and were
viable. To clarify that anti–HMG-1 protected
mice from LPS lethality, we purified the immu-
noglobulin G (IgG) fraction from anti–HMG-1
and administered it to mice exposed to an
LD100of LPS (11, 14). The highest dose of
anti–HMG-1 IgG tested, 5 mg per mouse, con-
ferred complete protection against an LD100of
LPS, whereas all control mice given compara-
ble doses of rabbit IgG died (Table 1). Treat-
ment with anti–HMG-1 IgG (2 mg per mouse)
significantly reduced serum HMG-1 levels,
whereas no reduction was observed after treat-
ment with a lower dose of antibodies (0.5 mg
per mouse) or with control IgG (5 mg per
mouse). Antiserum against a chemically syn-
thesized peptide corresponding to the first 12
amino acids of HMG-1 also significantly atten-
uated the lethality of endotoxemia in mice (15).
To determine if HMG-1 was toxic, we
administered highly purified rHMG-1 to un-
anesthetized Balb/C mice (10 to 50 ?g per
mouse). Within 2 hours, the mice developed
signs of endotoxemia, including lethargy, pi-
loerection, and diarrhea. At higher doses (500
?g per mouse), three of five mice died at 18,
30, and 36 hours after rHMG-1 administra-
tion. Toxicity and lethality were not observed
in control mice treated with a protein fraction
purified from E. coli transformed with a plas-
mid devoid of HMG-1 cDNA (9), indicating
that the toxicity we observed was specific to
HMG-1. To exclude the possibility that en-
dotoxin contamination of HMG-1 prepara-
tions mediated lethality, we injected rHMG-1
into LPS-resistant mice. rHMG-1 (500 ?g
per mouse) was lethal within 16 hours both to
C3H/HeJ (n ? 4) and C3H/HeN (n ? 3)
mice, indicating that HMG-1 itself is toxic
even in the absence of LPS signal transduc-
tion. When sublethal doses of rHMG-1 were
injected into Balb/C mice together with sub-
lethal doses of LPS, the combined challenge
was lethal to 90% of the mice, as compared
with 0% lethality in mice exposed to LPS or
HMG-1 alone (Fig. 2C). Thus, HMG-1 itself
mediates lethality in both LPS-sensitive and
Animal models of human sepsis, including
the murine endotoxemia model used here, have
inherent limitations (16). As an initial step in
determining whether HMG-1 participates in the
pathogenesis of human sepsis, we studied 8
Table 1. Protection against LPS lethality by anti–
HMG-1 IgG. Balb/C mice (male, 20 to 23 g, three
to six mice per group) were injected intraperito-
neally with IgG purified from anti–HMG-1 or con-
trol rabbit IgG 30 min before injection of an LD100
of LPS. All mice were then treated with additional
doses of anti–HMG-1 IgG or control IgG at 12 and
24 hours after LPS. Serum HMG-1 levels were
determined by immunoblots (under denaturing
conditions) 14 hours after LPS challenge (n ? 3
per group). ND, not determined.
Dose of HMG-1–
1003 ? 180
1070 ? 20
415 ? 240
*P ? 0.05 versus treatment with control rabbit IgG.
Fig. 2. (A) Anti–HMG-1 protect against
LPS lethality in mice. Polyclonal anti-
bodies against rHMG-1 were generated
in rabbits, and antiserum was assayed
for specificity and titer by enzyme-
linked immunosorbent assay and im-
munoblotting. Antibodies reacted spe-
cifically with HMG-1 and did not cross-
react with LPS, other bacterial proteins,
TNF, or IL-1?. Immunoblots of lysates
of macrophages or E. coli transformed
with plasmid containing HMG-1 cDNA
revealed only one band of immunore-
activity. Male Balb/C mice (20 to 23 g)
were randomly grouped (10 mice per
group) and treated with an LD100of LPS
(25 mg/kg). Anti–HMG-1 (Ab) or preim-
mune serum (0.2 ml per mouse, ip) was
administered 30 min before LPS. Additional doses of preimmune (0.4 ml, ip) or anti–HMG-1 (0.4
ml, ip) were administered at 12 and 36 hours after LPS as indicated. (B) Delayed administration of
anti–HMG-1 protects against LPS lethality in mice. Male Balb/C mice (20 to 23 g) were randomly
grouped (seven mice per group) and treated with an LD100of LPS. Anti–HMG-1 or preimmune
serum (0.4 ml per mouse) was administered at 2, 24, and 36 hours after LPS. (C) Administration of
rHMG-1 is lethal to mice. Recombinant HMG-1 was purified and LPS content determined by the
Limulus Amoebocyte Lysate Test (Bio-Whittaker, Walkersville, Maryland). Purified rHMG-1 protein
contained ?2.5 ng of LPS per microgram of rHMG-1. Male Balb/C mice (20 to 23, 10 animals per
group) were injected with a nonlethal dose of LPS (3.1 mg/kg, ip). Purified rHMG-1 protein was
administered intraperitoneally in the doses indicated at 2, 16, 28, and 40 hours after LPS.
Fig. 3. Increased serum HMG-1
levels in human sepsis. Serum was
obtained from 8 healthy subjects
and 25 septic patients infected
with Gram-positive [Bacillus fragi-
lis (one patient), Enterococcus fae-
calis (one patient), Streptococcus
pneumoniae (four patients), Liste-
ria monocytogenes (one patient),
or Staphylococcus aureus (two pa-
tients)], Gram-negative [E. coli
(seven patients), Klebsiella pneu-
moniae (one patient), Acineto-
bacter calcoaceticus (one patient),
Pseudomonas aeroginosa (one pa-
tient), Fusobacterium nucleatum
(one patient), Citrobacter freundii
(one patient)], or unidentified
pathogens (five patients). Serum
was fractionated by SDS-PAGE, and HMG-1 levels were determined by immunoblotting analysis with
reference to standard curves of purified rHMG-1 diluted in normal human serum; the detection limit is
?50 pg. *P ? 0.05 versus normal. **P ? 0.05 versus survivors.
R E P O R T S
9 JULY 1999VOL 285 SCIENCEwww.sciencemag.org
normal subjects and 25 critically ill septic pa-
dysfunction. HMG-1 was not detectable in the
serum of normal subjects, but significant levels
were observed in critically ill patients with sep-
sis (Fig. 3), and these levels were higher in
patients who succumbed as compared to pa-
tients with nonlethal infection.
HMG-1 is a highly conserved protein with
?95% amino acid identity between rodent and
human (17–20). It has previously been charac-
terized as a nuclear protein that binds to cruci-
form DNA (21), and as a membrane-associated
protein termed “amphoterin” that mediates neu-
rite outgrowth (19, 20). Extracellular HMG-1
interacts directly with plasminogen and tissue
type plasminogen activator (tPA), which en-
hances plasmin generation at the cell surface;
this system plays a role in extracellular prote-
olysis during cell invasion and tissue injury
(19). In addition, HMG-1 has been suggested to
bind to the receptor for advanced glycation end
products (RAGE) (22).
As with other inflammatory mediators
such as TNF and IL-1, there may be protec-
tive advantages of extracellular HMG-1 when
released in nontoxic amounts. Macrophages
release HMG-1 when exposed to the early,
acute cytokines, indicating that HMG-1 is
also positioned as a mediator of other inflam-
matory conditions associated with increased
levels of TNF and IL-1 (for example, rheu-
matoid arthritis and inflammatory bowel dis-
ease). Indeed, in most inflammatory scenari-
os, LPS is probably not the primary stimulus
for HMG-1 release; it seems more likely that
TNF and IL-1 function as upstream regula-
tors of HMG-1 release. The delayed kinetics
of HMG-1 release suggest that serum HMG-1
levels may be a convenient marker of disease
severity. Moreover, the observations that
HMG-1 itself is toxic, and that anti–HMG-1
prevents LPS lethality, point to HMG-1 as a
potential target for therapeutic intervention.
References and Notes
1. K. J. Tracey et al., Science 234, 470 (1986); C. A.
Dinarello, FASEB J. 8, 1314 (1994); D. C. Morrison and
J. L. Ryan, Annu. Rev. Med. 38, 417 (1987); L. L.
Moldawer, Crit. Care Med. 22, S3 (1994).
2. B. Beutler, N. Krochin, I. W. Milsark, C. Luedke, A.
Cerami, Science 232, 977 (1986); J. W. Killion and
D. C. Morrison, Infect. Immun. 54, 1 (1986); A. Pol-
torak et al., Science 282, 2085 (1998).
3. K. J. Tracey et al., Nature 330, 662 (1987); M. Bianchi
et al., J. Exp. Med. 83, 927 (1996); H. R. Alexander,
G. M. Doherty, C. M. Buresh, D. J. Venzon, J. A.
Norton, J. Exp. Med 173, 1029 (1991); G. Wakaba-
yashi, J. A. Gelfand, J. F. Burke, R. C. Thompson, C. A.
Dinarello, FASEB J. 5, 338 (1991).
4. D. G. Hesse et al., Surg. Gynecol. Obstet. 166, 147
(1988); H. R. Michie et al., N. Engl. J. Med. 318, 1481
5. M. P. Fink, Crit. Care Med. 23, 989 (1995); E. Abra-
ham, Chest 113 (suppl.), 224S (1998).
6. M. W. Marino et al., Proc. Natl. Acad. Sci. U.S.A. 94,
8093 (1997); F. Amiot, C. Fitting, K. J. Tracey, J. M.
Cavaillon, F. Dautry, Mol. Med. 3, 864 (1997).
7. M. E. Bianchi, M. Beltrame, G. Paonessa, Science 243,
1056 (1989); J. M. Walker, K. Gooderham, J. R. Hastings,
E. Mayes, E. W. Johns, FEBS Lett. 122, 264 (1980); L.
Einck and M. Bustin, Exp. Cell Res. 156, 295 (1985).
8. HMG-1 has also been termed “amphoterin” [J. Park-
kinen et al., J. Biol. Chem. 268, 19726 (1993)].
HMG-1 and amphoterin are the same protein; we use
the name “HMG-1” to reflect the original description
of this protein as the first member of the HMG
9. HMG-1 was cloned by DNA amplification of the
648–base pair (bp) open reading frame from Rat
Brain Quick-Clone cDNA (5 ng; Clontech, Palo Alto,
CA) with the following primers: 5?-CCCGCGGATC-
3? (PCR at 94°C for 1 min, 56°C for 2 min, 72°C for
45 s; 30 cycles). The 680-bp PCR product was
digested with Bam HI and Hind III and subcloned into
the Bam HI–Hind III cloning sites of the pCAL-n
vector (Stratagene, La Jolla, CA). The recombinant
plasmid was transformed into E. coli BL21(DE3)pLysS
(Novagen, Madison, WI), and positive clones were
confirmed by DNA sequencing of both strands. Trans-
formed cells were induced with isopropyl-D-
thiogalactopyranoside, and rHMG-1 protein was pu-
rified with a calmodulin-binding resin column (Strat-
agene). As controls for experiments involving admin-
istration of rHMG-1 to mice, we purified proteins
from E. coli BL21(DE3)pLysS that had been trans-
formed with a plasmid that lacks the HMG-1 cDNA
insert (pCAL-n). The amount of control material
administered to mice was normalized to the number
of E. coli that produce 0.5 mg of rHMG-1.
10. H. Wang and K. J. Tracey, unpublished observations.
11. Supplementary data can be found on Science Online
12. Macrophages were obtained from the peritoneal cavity
of LPS-sensitive (C3H/HeN and Balb/C) or LPS-resistant
(C3H/HeJ) mice 4 days after intraperitoneal injection
with 2.0 ml of thioglycollate broth (4%; Difco, Detroit,
MI). Macrophages were pooled from four mice, resus-
pended into RPMI 1640, 10% fetal bovine solution
(FBS), and 1% glutamine, and plated at a density of 4 ?
106cells per well in six-well Falcon Primaria tissue
culture plates. After 24 hours, the culture medium was
replaced with serum-free OPTI-MEM-I medium, and LPS
(1 ?g/ml) was added. The level of HMG-1 in the
culture medium was determined 18 hours later by
immunoblotting. HMG-1 was not detectable in
culture medium of LPS-stimulated C3H/HeJ murine
macrophages; HMG-1 levels reached 1 ?g/106cells
in the culture medium of LPS-stimulated C3H/HeN
13. L. B. Hinshaw et al., Circ. Shock 30, 279 (1990); C.
Nathan and M. Sporn, J. Cell Biol. 113, 981 (1991);
D. G. Remick et al., Am. J. Pathol. 136, 49 (1990); R. E.
Walker et al., J. Infect. Dis. 174, 63 (1996).
14. IgG was purified from anti–HMG-1 by Protein G–
Sepharose HiTrap affinity chromatography (Phar-
macia Biotech). Purified IgG fractions were de-
salted by ultrafiltration through Centricon-10 (Mil-
lipore), followed by two washes with 1? phos-
phate-buffered saline. The specificity of IgG was
confirmed by immunoblot analysis of macrophage
lysates, which revealed one band of 30 kD. Anti–
HMG-1 IgG did not cross-react with LPS, TNF, IL-1,
or bacterial proteins on immunoblots.
15. H. Wang and K. J. Tracey, unpublished observations.
16. M. P. Fink and S. O. Heard, J. Surg. Res. 49, 186
(1990); K. J. Tracey, and E. Abraham, SHOCK 11, 224
(1999); E. A. Deitch, ibid. 9, 1 (1998).
17. K. L. Lee et al., Nucleic. Acids Res. 15, 5051 (1987); S.
Ferrari, L. Ronfani, S. Calogero, M. E. Bianchi, J. Biol.
Chem. 269, 28803 (1994); S. Ferrari, P. Finelli, M.
Rocchi, M. E. Bianchi, Genomics 35, 367 (1996).
18. J. F. Maher and D. Nathans, Proc. Natl. Acad. Sci.
U.S.A. 93, 6716 (1996).
19. J. Parkkinen et al., J. Biol. Chem. 268, 19726 (1993).
20. M. Salmivirta, H. Rauvala, K. Elenius, M. Jalkanen, Exp.
Cell Res. 200, 444 (1992).
21. M. E. Bianchi and M. Beltrame, Am. J. Hum. Genet. 63,
1573 (1998); D. Landsman and M. Bustin, Bioessays
15, 539 (1993); M. Bustin and N. K. Neihart, Cell 16,
22. O. Hori et al., J. Biol. Chem. 270, 25752 (1995); A. M.
Schmidt et al., Diabetes 45, S77 (1996).
23. We thank C. Dang for technical assistance; J. Eaton, J.
Roth, B. Sherry, M. Bukrinsky, and M. Symons for
critical reading of the manuscript; and D. Prieto for
11 December 1998; accepted 3 May 1999
Ploidy Regulation of Gene
Timothy Galitski,1Alok J. Saldanha,1,2Cora A. Styles,1
Eric S. Lander,1,2Gerald R. Fink1,2*
Microarray-based gene expression analysis identified genes showing ploidy-
dependent expression in isogenic Saccharomyces cerevisiae strains that varied
in ploidy from haploid to tetraploid. These genes were induced or repressed in
proportion to the number of chromosome sets, regardless of the mating type.
Ploidy-dependent repression of some G1cyclins can explain the greater cell size
associated with higher ploidies, and suggests ploidy-dependent modifications
of cell cycle progression. Moreover, ploidy regulation of the FLO11 gene had
direct consequences for yeast development.
Changes in the number of chromosome sets
occur during the sexual cycle, during meta-
zoan development, and during tumor progres-
sion. Organisms with a sexual cycle double
their ploidy upon fertilization and reduce
their ploidy by half at meiosis. In the devel-
opment of almost all plants and animals, spe-
cialized polyploid and polytene cell types
arise though endocycles, cell cycles lacking
cell division (1). Aberrant cell cycle control
during tumor progression is thought to result
in polyploidy and altered cell behavior (2).
Cells of different ploidy typically show very
different developmental, morphological, and
physiological characteristics. However, a lack
1Whitehead Institute for Biomedical Research, 9 Cam-
bridge Center, Cambridge, MA 02142, USA.2Depart-
ment of Biology, Massachusetts Institute of Technol-
ogy, Cambridge, MA 02139, USA.
*To whom correspondence should be addressed. E-
R E P O R T S
www.sciencemag.orgSCIENCEVOL 2859 JULY 1999