REPORTS

the life of neutrophils (11), can induce NETs

Neutrophil Extracellular Traps efficiently. (iii) Incubation with DNA interca-
lating dyes before neutrophil activation pre-

Kill Bacteria vents NET formation but has no effect on the

induction of apoptosis by staurosporine or tu-
mor necrosis factor ␣ (8). (iv) NETs are formed
Volker Brinkmann,1 Ulrike Reichard,1,2 Christian Goosmann,1,2 as early as 10 min after activation, a time course
Beatrix Fauler,1 Yvonne Uhlemann,2 David S. Weiss,2 faster than apoptosis (fig. S1). (v) Time-lapse
Yvette Weinrauch,3 Arturo Zychlinsky2* video microscopy (movie S2) shows that motile
cells make NETs. Taken together, these data
Neutrophils engulf and kill bacteria when their antimicrobial granules fuse with strongly indicate that NETs are not the result of
the phagosome. Here, we describe that, upon activation, neutrophils release leakage during cellular disintegration. We cannot
granule proteins and chromatin that together form extracellular fibers that bind exclude, however, the possibility that NET forma-
Gram-positive and -negative bacteria. These neutrophil extracellular traps tion is an early event in the neutrophil program for
(NETs) degrade virulence factors and kill bacteria. NETs are abundant in vivo cell death. Neutrophils are terminally differentiat-
in experimental dysentery and spontaneous human appendicitis, two examples ed cells that are programmed to die a few hours
of acute inflammation. NETs appear to be a form of innate response that binds after they enter into circulation. Furthermore, iso-
microorganisms, prevents them from spreading, and ensures a high local con- lated neutrophils are a heterogeneous population
centration of antimicrobial agents to degrade virulence factors and kill bacteria. with respect to age, and a small portion of this
“aged” subpopulation is expected to die. Neutro-
In response to inflammatory stimuli, neutro- from azurophilic (primary) granules (5, 6) such phils can undergo caspase-dependent (12) and
phils migrate from the circulating blood to as neutrophil elastase (Fig. 2A), cathepsin G, -independent apoptosis in vitro (13), but the pro-
infected tissues, where they efficiently bind, and myeloperoxidase (table S1). Proteins from cess that leads to neutrophil death in vivo is not
engulf, and inactivate bacteria. Phagocytosed specific (secondary) granules and tertiary gran- known. It is conceivable that NET formation is an
bacteria are killed rapidly by proteolytic en- ules, such as lactoferrin and gelatinase, respec- early event in cell death.
zymes, antimicrobial proteins, and reactive tively, were also present (table S1). In contrast, NETs associate with both Gram-positive
oxygen species (1, 2). Neutrophils also de- CD63, a granule membrane protein, the cyto- (Staphylococcus aureus, shown in Fig. 3A) and
granulate, releasing antimicrobial factors into plasmic markers annexin I (7), actin, tubulin, Gram-negative pathogens (Salmonella typhi-
the extracellular medium (3). Here, we show that and various other cytoplasmic proteins were murium and Shigella flexneri, shown in Fig. 3,
neutrophils generate extracellular fibers, or neu- excluded from NETs (table S1). B and C, respectively). We have previously
trophil extracellular traps (NETs), which are struc- DNA is a major structural component of shown that neutrophil elastase degrades viru-
tures composed of granule and nuclear constitu- NETs, because several DNA intercalating dyes lence factors of Gram-negative bacteria (14).
ents that disarm and kill bacteria extracellularly. stained NETs strongly (Fig. 2B) and a brief Our finding that bacteria are trapped in NETs
NETs were made by activated neutrophils. treatment with deoxyribonuclease (DNase) re- decorated with neutrophil elastase prompted us
Although naı̈ve cells were round with some sulted in the disintegration of NETs (movie S1). to test whether bacterial virulence factors were
membrane folds (Fig. 1, A and C), neutrophils Conversely, protease treatment left the DNA of targeted extracellularly. Immunofluorescence
stimulated with interleukin-8 (IL-8), phorbol the NETs intact (8). The NETs reacted with staining of IpaB, a virulence factor of S. flex-
myristate acetate (PMA), or lipopolysaccharide antibodies against histones H1, H2A, H2B, H3, neri, was weaker in bacteria trapped in NETs
(LPS) became flat and formed membrane pro- and H4 (table S1) and against the H2A-H2B- compared to free Shigella (Fig. 3D, top left),
trusions (Fig. 1B) as previously described (4). DNA complex (9, 10) (Fig. 2C). although the bacteria and the NETs were clear-
Surprisingly, we found that activated neutro- Double immunostaining of ultrathin cryo- ly visible when DNA was stained. In contrast,
phils but not naı̈ve cells made prominent extra- sections for TEM (Fig. 2D) confirmed the pres- when neutrophil protease activity was blocked
cellular structures (arrows, Fig. 1, B and D). ence of neutrophil elastase (small gold particles, by the secretory leukocyte proteinase inhibitor
These fibers, or NETs, were very fragile, and arrowheads) and H2A-H2B-DNA complexes (SLPI), bacteria trapped in NETs contained
specimens had to be washed and fixed carefully (large gold particles, arrows) in NETs. Histone high amounts of IpaB (Fig. 3D, bottom left).
to preserve them. High-resolution scanning and neutrophil elastase staining was found on Interestingly, virulence factors from Gram-
electron microscopy (SEM) showed that the globular NET domains. Furthermore, immuno- positive bacteria were also susceptible to neu-
NETs contained smooth stretches with a diam- staining of SEM samples (Fig. 2E) corroborated trophil proteases. Lower amounts of the S.
eter of 15 to 17 nm (Fig. 1E, arrowheads) and the localization of neutrophil elastase to the aureus virulence factor ␣ toxin were found in
globular domains of around 25 nm (Fig. 1E, globular domains of NETs. These data demon- NET-associated bacteria compared to that of free
arrows) that aggregated into larger threads with strate that the structures visualized by different bacteria or when neutrophil proteases were
diameters of up to 50 nm. Analysis of cross microscopy approaches (immunofluorescence, blocked with SLPI (fig. S2). These results suggest
sections of the NETs by transmission electron TEM, and SEM) are identical. NET formation that NETs can disarm a wide range of pathogens.
microscopy (TEM) revealed they were not sur- was quantified in a fluorometer with the use of We corroborated that extracellular proteases
rounded by membranes (Fig. 1F). a DNA dye that is excluded from cells. Neutro- degrade bacterial virulence factors by inhibiting
The composition of NETs was analyzed by phils release NETs as early as 10 min after neutrophil phagocytosis. This was accom-
immunofluorescence. NETs contained proteins activation, and the release depends on the dose plished by incubating activated neutrophils with
of the activator (fig. S1). cytochalasin D. In the presence of cytochalasin
Several lines of evidence indicate that neu- D, an inhibitor of actin polymerization, NETs
1
Microscopy Core Facility and 2Department of Cellular trophils make NETs actively: (i) Stimuli that persisted and phagocytosis was blocked. We
Microbiology, Max Planck Institute for Infection Biology,
Schumannstrasse 21/22, 10117 Berlin, Germany. 3Depart-
induce NETs do not promote the release of the infected these neutrophils that have NETs but
ment of Microbiology, New York University School of cytoplasmic marker lactate dehydrogenase cannot phagocytose with S. flexneri. Extracel-
Medicine, 540 First Avenue, New York, NY 10016, USA. (LDH), and activated cells exclude vital dyes lular neutrophil elastase, like purified elastase
*To whom correspondence should be addressed. E- for at least two hours after stimulation (8). (ii) (14), degraded the virulence factors IcsA and
mail: zychlinsky@mpiib-berlin.mpg.de Stimuli such as IL-8 and LPS, which prolong IpaB but not the control OmpA, an outer mem-

1532 5 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org

 REPORTS
brane protein (Fig. 3E). This confirms that neu- antimicrobial defense. Also, these data are in ious proteins like proteases from diffusing
trophil elastase presented in NETs actively tar- accord with recent findings proposing that away and inducing damage in tissue adja-
gets bacterial virulence factors. oxygen-independent mechanisms play an im- cent to the site of inflammation (20). NETs
Activated neutrophils incubated with cy- portant role in the control of infections (19). might also have a deleterious effect on the
tochalasin D after formation of the NETs can The data presented here indicate that granule host, because the exposure of extracellular
kill about 30% of a S. flexneri or S. aureus proteins and chromatin together form an ex- histone complexes could play a role during
inoculum (Fig. 3F, without DNase). We pro- tracellular structure that amplifies the effec- the development of autoimmune diseases
pose the hypothesis that the NET structure is tiveness of its antimicrobial substances by like lupus erythematosus.
necessary for this extracellular bactericidal ensuring a high local concentration. NETs
activity. Indeed, when NETs were dismantled degrade virulence factors and/or kill bacteria References and Notes
1. P. Elsbach, J. Weiss, in Inflammation: Basic Principles and
with DNase (movie S1), the killing of bacte- even before the microorganisms are engulfed Clinical Correlates, J. I. Gallin, I. M. Goldstein, R. Snyderman,
ria was negligible (Fig. 3F). In these experi- by neutrophils. In addition to their antimicro- Eds. (Raven Press, New York, ed. 2, 1992), pp. 603–636.
ments, the cultures were not washed after bial properties, NETs may serve as a physical 2. S. J. Klebanoff, in Inflammation: Basic Principles and
treatment with protease-free DNase, leaving barrier that prevents further spread of bacte- Clinical Correlates, J. I. Gallin, R. Snyderman, Eds.
(Lippincott Williams & Wilkens, Philadelphia, PA, ed.
the total protein concentration unchanged. ria. Moreover, sequestering the granule pro- 3, 1999), pp. 721–768.
Hence, these data strongly suggest that the teins into NETs may keep potentially nox- 3. P. E. Henson et al., in Inflammation: Basic Principles
fibrous structure of NETs is necessary for the
sequestration and killing of bacteria by deliv-
ering a high local concentration of antimicro-
bial molecules to the bound microbes.
In an alternative approach to demonstrate
the antibacterial activity of NETs, we showed
that a monoclonal antibody against the H2A-
H2B-DNA complex abrogated S. flexneri and
S. aureus killing in infections of neutrophils
pretreated with cytochalasin D after NET for-
mation (Fig. 3G). An isotype control antibody
had no effect on killing. The factors responsible
for bacterial killing are likely to include granule
proteins like bactericidal permeability increas-
ing protein (BPI) (table S1) and histones. The
antimicrobial activity of histones (15), evolu-
tionarily conserved proteins that bind DNA to
form the nucleosome complex, and peptides
derived from histones, is well established (16,
17) Indeed, purified H2A killed S. flexneri, S.
typhimurium, and S. aureus cultures with con-
centrations as low as 2 ␮g/ml (140 nM) in 30
min (fig. S3). The concentration of H2A re-
quired to kill bacteria is low compared with
other antimicrobial proteins (18).
To determine whether NETs are present in
vivo, we analyzed samples from experimental
shigellosis in rabbits and spontaneous appendi-
citis in humans. Staining of histological sec-
tions clearly showed extracellular fibrous ma-
terial that contains NET components: histones
(Fig. 4, A and F), DNA (Fig. 4, C and G), and
neutrophil elastase (Fig. 4E). In vivo, NETs trap
bacteria as shown by the localization of Shigel-
la (Fig. 4B) to the NETs. These results indicate
that NETs are abundant at inflammatory sites.
Neutrophils make NETs through an active
mechanism that remains to be understood.
NETs disarm pathogens with proteases such
as neutrophil elastase. NETs also kill bacteria Fig. 1. Electron microscopical analysis of resting and activated neutrophils. (A) Resting neutrophils
efficiently, and at least one of the NET com- are round and devoid of fibers. (B) Upon stimulation with 25 nM PMA for 30 min, the cells flatten,
ponents, histones, exerts antimicrobial activ- make many membrane protrusions, and form fibers (NETs), arrows in (B) and (D). (C) TEM analysis
ity at surprisingly low concentrations. These of naı̈ve neutrophils in suspension. (D) Ultrathin section of neutrophils stimulated in suspension
data correlate with previous findings showing with 10 ng of IL-8 for 45 min. Bars in (A) to (D) indicate 10 ␮m. The multilobular nuclei and
that neutrophil degranulation releases antimi- different granules are clearly visible in both figures. The activated cells in (D) have many
pseudopods and show NETs (arrow). (E) High-resolution SEM analysis of NETs that consist of
crobial factors extracellularly (3) and the ob- smooth fibers (diameters of 15 to 17 nm, arrowheads) and globular domains (diameter around 25
servation that inflammatory exudates rich in nm, arrow). Globular complexes can be aggregated to thick bundles or fibers. (F) Ultrathin sections
neutrophils, like pus, contain DNA, which of NETs show that they are not membrane-bound. Neutrophils were stimulated as in (D). Bars in
was not known to play an active role in (E) and (F), 500 ␮m.

www.sciencemag.org SCIENCE VOL 303 5 MARCH 2004 1533

 REPORTS
Fig. 2. Immunostaining of
NETs. Neutrophils were acti-
vated with 10 ng of IL-8 for 30
min and stained for neutrophil
elastase (A), DNA (B), and the
complex formed by H2A-H2B-
DNA (C). Extracellular fibrous
material is stained brightly. As
expected, we found granular
staining for neutrophil elastase
(A) and nuclear staining for hi-
stones and DNA [(B) and (C)].
Samples were analyzed with
the use of a Leica TCS-SP
(Beusheim, Germany) confocal
microscope. The images are
projections of a z stack (origi-
nal dimensions: x and y, 85.5
␮m; z ⫽ 6.3 ␮m). Bar, 10 ␮m.
(D) Immunodetection of his-
tones (large gold particles, ar-
rows) and neutrophil elastase
(small gold particles, arrow-
heads) in ultrathin cryosec-
tions of neutrophils stimulated
with IL-8 (10 ng, 1 hour). Bar,
200 nm. (E) Immuno-SEM,
pseudocolored, of neutrophils
treated as in (A) to (C). Over-
lay of images from secondary
electron detector (red, topog-
raphy) and backscattered elec-
tron detector (green, element
sensitive, most back-scattered
electrons from the site of gold binding). Bright yellow dots (arrows) show localization of 12-nm gold particles detecting neutrophil elastase. Bar, 200 nm.

Fig. 3. Gram-positive and Gram-negative

bacteria associate with neutrophil fibers. SEM
of S. aureus (A), S. typhimurium (B), and S.
flexneri (C) trapped by NETs. Neutrophils
were treated with 100 ng of IL-8 for 40 min
before infection. Bar, 500 nm. (D) Immuno-
fluorescence of neutrophils infected with S.
flexneri stained for the virulence factor IpaB
and DNA. IpaB is degraded by neutrophil
elastase and is only detectable on the bacte-
ria (arrows) when neutrophil elastase is
blocked with SLPI. DNA staining shows NETs
and bacteria (arrows). (E) Western blot
showing that the virulence factors IcsA and
IpaB but not OmpA were degraded by cy-
tochalasin D–treated neutrophils incubated
with S. flexneri. Lane 1, bacteria alone. Lane
2, bacteria incubated with cytochalasin D–treated neutrophils. (F) reduced by addition of antibodies against histones. Neutrophils were
Extracellular bactericidal activity was greatly reduced in both S. treated with cytochalasin D to prevent phagocytosis and infected
flexneri and S. aureus infections after incubation with DNase, which with S. flexneri or S. aureus. In the presence of antibody against H2A,
dissociates NETs. (G) Extracellular bacterial killing by neutrophils was bacterial killing was abrogated.

1534 5 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org


Fig. 4. Analysis of tissue sections from experimental shigellosis in
rabbits (A to D) and spontaneous human appendicitis (E to H). (A)
Immunofluorescence staining of histones reveals nuclear and extra-
cellular localization that largely overlaps with staining for DNA (C).
(B) Staining with an antibody against Shigella-specific LPS. (D) The
overlay indicates that numerous Shigellae are closely associated to
4.
and Clinical Correlates, J. I. Gallin, R Snyderman, Eds.
(Raven Press, New York, ed. 2, 1992), pp. 511–539.
L. C. Junqueira, J. Carneiro, R. O. Kelley, Basic Histol-
5.
6.
ogy (Appelton & Lange, Norwalk, CT, ed. 8, 1995).
D. F. Bainton, J. Immunol. Methods 232, 153 (1999).
N. Borregaard, J. B. Cowland, Blood 89, 3503 (1997).
7. C. Movitz, C. Dahlgren, Cell Biol. Int. 25, 963 (2001).
8. V. Brinkmann et al., data not shown.
9. M. J. Losman, T. M. Fasy, K. E. Novick, M. Monestier,
J. Immunol. 148, 1561 (1992).
10. P. Salgame et al., Nucleic Acids Res. 25, 680 (1997).
11. P. L. Sabroe et al., J. Immunol. 170, 5268 (2003).
12. B. Fadeel et al., Blood 92, 4808 (1998).
13. N. A. Maianski, D. Roos, T. W. Kuijpers, Blood 101,
1987 (2002).
14. Y. Weinrauch et al., Nature 417, 91 (2002).
15. J. G. Hirsch, J. Exp. Med. 108, 925 (1958).
16. S. M. Zhdan-Pushkina, N. V. Dronova, Mikrobiologiia
45, 60 (1976).
17. H. S. Kim, C. B. Park, M. S. Kim, S. C. Kim, Biochem.
Biophys. Res. Commun. 229, 381 (1996).
18. P. Elsbach, J. Weiss, O. Levy, Trends Microbiol. 2, 324
(1994).
19. E. P. Reeves et al., Nature 416, 291 (2002).
20. V. Balloy et al., Am. J. Respir. Cell Mol. Biol. 28, 746 (2003).
21. The gift of M. Monestier, Temple University, the monoclo-
nal antibody against H2A-H2B-DNA, is gratefully acknowl-
edged. The authors thank the help of M. Ingersoll, B.
Raupach, C. Scharff, C. Heinz, and members of the Depart-
ment of Cellular Microbiology, Max Planck Institute for
Infection Biology. Supported in part by NIH grant
AI037720.
Supporting Online Material
www.sciencemag.org/cgi/content/full/303/5663/1532/DC1
Materials and Methods
Figs. S1 to S3
Table S1
Movies S1 and S2
9 October 2003; accepted 24 December 2003
www.sciencemag.org SCIENCE VOL 303 5 MARCH 2004
REPORTS
fibrous material staining for histones and DNA. (E) Staining for
neutrophil elastase in an area of neutrophil exudate in human spon-
taneous appendicitis reveals fibrous extracellular material that also
stains for histone (F) and DNA (G). (H) Overlay of the images. The
images are projections of confocal z stacks generated from sections of
5 to 6 ␮m thickness. Bar, 50 ␮m.
Emerging Vectors in the Culex
pipiens Complex
Dina M. Fonseca,1,2* Nusha Keyghobadi,1 Colin A. Malcolm,3
Ceylan Mehmet,3 Francis Schaffner,4 Motoyoshi Mogi,5
Robert C. Fleischer,1 Richard C. Wilkerson2
In the Old World, some mosquitoes in the Culex pipiens complex are excellent
enzootic vectors of West Nile virus, circulating the virus among birds, whereas
others bite mainly humans and other mammals. Here we show that, in northern
Europe, such forms differing in behavior and physiology have unique micro-
satellite fingerprints with no evidence of gene flow between them, as would be
expected from distinct species. In the United States, however, hybrids between
these forms are ubiquitous. Such hybrids between human-biters and bird-biters
may be the bridge vectors contributing to the unprecedented severity and range
of the West Nile virus epidemic in North America.
Species in the Culex pipiens complex are overwintering mosquitoes can serve as a
considered to be the primary vectors of source of WNV to initiate an infection
West Nile virus (WNV) in North America cycle in the spring (7 ). Blood-meal analysis
because they are often the most common has revealed that Cx. pipiens in the United
mosquitoes in urban areas (1), because dis- States bite both humans (anthropophagy)
ease outbreaks occur during their peak and birds, suggesting they may serve as
abundance period (2), because they are bridge vectors of the disease from birds to
competent laboratory vectors of WNV (3), humans (2). Human WNV epidemics re-
and because field populations in the United quire bridge vectors, because humans and
States have repeatedly been found infected other mammals do not usually generate
with the virus (4, 5). In addition, they can high enough viremia to infect biting mos-
transmit the virus transovarially (6 ), so quitoes (8). Although Cx. pipiens has been
1535