Proc. Natl. Acad. Sci. USAVol. 92, pp. 7202-7206, August
1995Medical Sciences
Activation of a 15-kDa endonuclease in
hypoxia/reoxygenationinjury without morphologic features of
apoptosis
(kidney/proximal tubules/DNA damage/aurintricarboxylic acid)
NORISHI UEDA*, PATRICK D. WALKERt, SU-MING Hsut, AND SUDHIR V.
SHAH*t*Department of Medicine, Division of Nephrology, and
tDepartment of Pathology, University of Arkansas for Medical
Sciences, 4301 West Markham and John L.McClellan Memorial Veterans
Hospital, 4300 West Seventh Street, Little Rock, AR 72205
Communicated by Carl W Gottschalk University of North Carolina,
Chapel Hill, NC, April 4, 1995 (received for review January 6,
1995)
ABSTRACT Hypoxia/reoxygenation is an importantcause of tissue
injury in a variety of organs and is classicallyconsidered to be a
necrotic form of cell death. We examined therole of endonuclease
activation, considered a characteristicfeature of apoptosis, in
hypoxia/reoxygenation injury. Wedemonstrate that subjecting rat
renal proximal tubules tohypoxia/reoxygenation results in DNA
strand breaks andDNA fragmentation (both by an in situ technique
and byagarose gel electrophoresis), which precedes cell death.
Hy-poxia/reoxygenation resulted in an increase in
DNA-degradingactivity with an apparent molecular mass of 15 kDa on
asubstrate gel. This DNA-degrading activity was entirely
calciumdependent and was blocked by the endonuclease inhibitor
aur-intricarboxylic acid. The protein extract from tubules
subjectedto hypoxia/reoxygenation cleaved intact nuclear DNA
obtainedfrom normal proximal tubules into small fragments,
whichfurther supports the presence of endonuclease activity.
Despiteunequivocal evidence of endonuclease activation, the
morpho-logic features of apoptosis, including chromatin
condensation,were not observed by light and electron microscopy.
Endonucle-ase inhibitors, aurintricarboxylic acid and Evans blue,
providedcomplete protection against DNA damage induced by
hy-poxia/reoxygenation but only partial protection against
celldeath. Taken together, our data provide strong evidence for
arole ofendonuclease activation as an early event, which is
entirelyresponsible for theDNAdamage and partially responsible for
thecell death that occurs during hypoxia/reoxygenation injury.
Ourdata also indicate that in hypoxia/reoxygenation injury
endo-nuclease activation and DNA fragmentation occur without
themorphological features of apoptosis.
Complete or partial cessation followed by restoration of
bloodflow is a -serious event that affects many organs and
isparticularly common in the heart, the brain, and the kidney(1-5).
Ischemia/reperfusion in vivo and hypoxia/reoxygen-ation in vitro
are classically considered to result in a necroticform of cell
death in which there is rapid collapse of internalhomeostasis of
the cell (6-8) associated with membrane lysisand inflammation (9).
In contrast, apoptosis, or programmedcell death, is considered
operationally, morphologically, andbiochemically distinct from
necrotic cell death (6, 10) and is aprocess in which the cell
actively participates in its own demise.Apoptosis is characterized
by several early morphologicalalterations, including plasma
membrane blebbing and chro-matin condensation (11, 12). Endogenous
endonuclease acti-vation, resulting in the cleavage of host
chromatin into oligo-nucleosome-length DNA fragments, has been
considered acharacteristic biochemical marker for apoptosis (13,
14). In thepresent study we examined the role of endonuclease
activationas an early event responsible for DNA damage and/or
celldeath in hypoxia/reoxygenation injury to isolated rat renal
proximal tubules. We also examined the relationship betweenDNA
fragmentation and the histologic features of apoptosis
inhypoxia/reoxygenation injury.
MATERIALS AND METHODSProximal tubules from kidneys of male
Sprague-Dawley rats(250-300 g) were isolated by Percoll density
gradients asdescribed (15) with minor modifications described in
our previousstudy (16) and had respiratory parameters similar to
previousstudies (17). Each experiment was initiated with a 30-min
pre-equilibration of the tubule suspension at 37°C bubbled with
95%02/5% CO2. Hypoxia was achieved by gassing with 95% N2/5%CO2,
and reoxygenation was attained by reintroducing 95%02/5% CO2 at 37C
in a shaker bath. Control tubules weremaintained under 95% 02/5%
C02. Cell death was measured bylactate dehydrogenase release as
described (15).The residual double-stranded DNA was measured by
the
alkaline unwinding assay, and determination of ethidiumbromide
fluorescence was as described (18-21). For detectionof DNA
fragmentation, the cells were lysed with a buffercontaining 0.5%
Triton X-100, 10 mM Tris HCl, and 20 mMEDTA, to prevent DNA
fragmentation during isolation (22).We further verified our results
by utilizing another method(23), which includes 1 M NaCl and in
which we included 100mM EDTA in the buffer (data not shown).
Fragmented DNAwas obtained by centrifugation at 13,000 x g for 20
min asdescribed (20, 24) and subjected to electrophoresis, and
thenthe gel was incubated with RNase A (20 gg/ml) at 37°C for 4hr
before staining with ethidium bromide.
In situ assay for DNA fragmentation was performed usingterminal
deoxynucleotidyltransferase (0.3 unit/ml; Oncor)and biotinylated
dUTP as described (25). Endonuclease ac-tivity was detected by
SDS/PAGE containing heat-denatured(100°C for 10 min) calf thymus
DNA at 10 ,ug/ml as described(26). Intact nuclei were isolated from
control tubules bysucrose gradients (27) and resuspended in buffer
A (15 mMTris HCl, pH 7.5/60 mM KCl/15 mM NaCl/14 mM
2-mer-captoethanol/0.15 mM spermine/0.5 mM spermidine) con-taining
10% (wt/vol) sucrose. Light and electron microscopywere carried out
using routine methods.
RESULTSProximal tubules isolated from rat kidney were exposed
tovariable periods of hypoxia (5-30 min) followed by a
30-minreoxygenation. As shown in Fig. LA, DNA strand breaks
wereobserved with as little as 5 min of hypoxia. Five minutes
ofhypoxia followed by reoxygenation for 30 min resulted in amarked
increase in the amount of DNA damage (residualdouble-stranded DNA,
43% ± 1%). DNA strand breaksoccurred in a time-dependent manner
and, consistent withother measures of cell injury (1-3, 5, 15),
were much higherafter reoxygenation at every time point examined.
Significantcell death was first observed after 15 min of hypoxia
followed
ITo whom reprint requests should be addressed.
7202
The publication costs of this article were defrayed in part by
page chargepayment. This article must therefore be hereby marked
"advertisement" inaccordance with 18 U.S.C. §1734 solely to
indicate this fact.
Dow
nloa
ded
by g
uest
on
June
15,
202
1
Medical Sciences: Ueda et al.
A Time control100 0 0
80
z
1 60-
S0 Alt40 -W
Proc. Natl. Acad. Sci. USA 92 (1995) 7203
B** t
C H /R H /HR HHIR H H/R C H H/fR H H/fR H H/ft5 10 15 30 10 15
30
Period of Hypoxia (min) Period of Hypoxia (min)
FIG. 1. (A) Time course of the effect of hypoxia (5-30
min)/reoxygenation (30 min) on DNA strand breaks in proximal
tubules. Results aremeans ± SE (n = 4). *, P < 0.05 and **, P
< 0.01 compared with the time control; t, P < 0.005 compared
with hypoxia alone. (B) Time courseof the effect of hypoxia (10-30
min)/reoxygenation (30 min) on cell death in proximal tubules.
Results are means ± SE (n = 7). *, P < 0.02 and**, P < 0.0001
compared with the time control; t, P < 0.02 compared with
hypoxia alone. C, control; H, hypoxia; H/R,
hypoxia/reoxygenation;LDH, lactate dehydrogenase.
by 30 min of reoxygenation compared to its own time control(Fig.
1B), indicating that the DNA damage clearly precedescell
death.Hypoxia alone or hypoxia/reoxygenation caused DNA
fragmentation (multiples of 180 bp) in a time-dependentmanner
(Fig. 2 A and B). The DNA fragmentation wasobserved with as little
as 5 min of hypoxia and clearlypreceded cell death. Hypoxia alone
or a period of hypoxia of
A Bbp 1 2 3 4
2036-1636-
1018-
512-39a-220-
154-
30 min or less followed by reoxygenation (30 min) resultedin a
ladder pattern of DNA fragmentation. By contrast, aperiod of
hypoxia of 30 min followed by 60 min of reoxy-genation resulted in
not only DNA fragments but also asmear pattern of DNA degradation
that is often seen in thenecrotic form of cell death (11). These
data suggest that theendonuclease activation is an early event
during hypoxia/reoxygenation injury.
C3 4
Ebp
- 14.5 kDa 2036-1636-
1018 -
D MN C H/f c c MR
14.5 kDa-
+Ca -Ca +Ca, .ATA
512-396 -344 -298 -220 -
154 -
FIG. 2. (A) Effect of hypoxia/reoxygenation on DNA fragmentation
in proximal tubules. The nucleic acid (20 JLg) was subjected
toelectrophoresis on a 1.6% agarose gel. Lanes: 1, control (45
min); 2, hypoxia for 5 min; 3, hypoxia for 10 min; 4, hypoxia for
15 min; 5, hypoxiafor 5 min and reoxygenation for 30 min; 6,
hypoxia for 10 min and reoxygenation for 30 min; 7, hypoxia for 15
min and reoxygenation for 30 min.(B) Lanes: 1, molecular size
standards; 2, control (90 min); 3, hypoxia for 30 min; 4, hypoxia
for 30 min followed by 60 min of reoxygenation.Oligonucleosomal
fragments appear as ladders whose molecular sizes are approximately
multiples of 180 bp. Note that there is a mixture of a ladderand a
smear pattern of DNA from tubules exposed to 30 min of hypoxia and
60 min of reoxygenation (lane 4) in contrast to only a ladder
patternobserved from tubules exposed to a shorter period of hypoxia
and/or reoxygenation. (C) Effect of hypoxia/reoxygenation on
endonuclease activityin proximal tubules. Each protein extract (8
,ug per lane) was subjected to electrophoresis on a DNA substrate
gel. Lanes: 1, control (60 min); 2,hypoxia for 30 min; 3, hypoxia
for 30 min and reoxygenation for 30 min; 4, micrococcal nuclease
(0.1 ng per lane). (D) Effect of either calciumor the endonuclease
inhibitor aurintricarboxylic acid on endonuclease activity in
proximal tubules exposed to hypoxia (30 min)/reoxygenation (30min).
Each protein extract (10 ,ug per lane) was subjected to
electrophoresis on a DNA substrate gel. MN, micrococcal nuclease
(0.1 ng per lane);C, control; H/R, hypoxia for 30 min followed by
reoxygenation 30 min; ATA, aurintricarboxylic acid. (E) Effect of
protein extracts from tubulesexposed to hypoxia/reoxygenation on
intact nuclear DNA. The protein extract (10 j,g) from tubules
exposed to hypoxia for 30 min andreoxygenation for 30 min was
incubated with intact nuclei from isolated control tubules in
buffer A containing 10% sucrose and 1 mM CaCl2 for0, 30, and 60
min. The nucleic acid (20 ,ug) extracted as described in Materials
and Methods was subjected to electrophoresis on a 1.6% agarosegel.
Lanes: 1, molecular size standards; 2, 0-min incubation; 3, 30-min
incubation; 4, 60-min incubation.
l 2
Dow
nloa
ded
by g
uest
on
June
15,
202
1
Proc. Natl. Acad. Sci. USA 92 (1995) 7205
z
0
80.
60.
40-
20.
0. -
I9
C H/R
*
K,,....0.,Z.%o,.oIt8>
'a
-~LLIZZATA FA
I91
FIG. 4. (A) Effect of an endonuclease inhibitor,
aurintricarboxylic acid, on hypoxia/reoxygenation-induced DNA
strand breaks in proximaltubules. Tubules were incubated with
aurintricarboxylic acid (ATA, 100 ,M) or an inactive analog of
aurintricarboxylic acid, fuchsin acid (FA, 100,uM) for 30 min.
Tubules were washed and resuspended in an incubation buffer and
then exposed to 30 min of hypoxia followed by 30 min
ofreoxygenation. Results are means ± SE (n = 4). *, P < 0.005
compared with hypoxia (30 min) and reoxygenation (30 min) alone.
(B) Effect ofan endonuclease inhibitor, aurintricarboxylic acid, on
hypoxia/reoxygenation-induced cell death in proximal tubules.
Results are means ± SE (n= 4). *, P < 0.05 compared with hypoxia
(30 min) and reoxygenation (30 min) alone. C, control; H/R,
hypoxia/reoxygenation; ATA,aurintricarboxylic acid; FA, fuchsin
acid; LDH, lactate dehydrogenase. (C) Effect of endonuclease
inhibitor, Evans blue (50 ,ug/ml), onhypoxia/reoxygenation-induced
DNA strand breaks in proximal tubules. Results are means ± SE (n =
5). *, P < 0.0005 compared with hypoxia(30 min) and
reoxygenation (30 min) alone. (D) Effect of an endonuclease
inhibitor, Evans blue, on hypoxia/reoxygenation-induced cell death
inproximal tubules. Results are means ± SE (n = 4). *, P < 0.01
compared with hypoxia (30 min) and reoxygenation (30 min) alone. C,
control;H/R, hypoxia/reoxygenation; EB, Evans blue.
15 min of hypoxia (Fig. 3B). Similar changes were seen intubules
exposed to 30 min of hypoxia and tubules exposed to15 or 30 min of
hypoxia and 30 min of reoxygenation.To provide further evidence
that DNA fragmentation was
indeed present at a time when no morphological features
ofapoptosis were present, an in situ assay forDNA fragmentationwas
performed (25). Control tubules contained 3 ± 1 positivenuclei
(staining dense black) per hundred cells (Fig. 3C),whereas tubules
subjected to 30 min of hypoxia and 60 min ofreoxygenation showed a
marked increase to 21 ± 3 positivenuclei (n = 4,P < 0.001) (Fig.
3D). Taken together, these datademonstrate unequivocal evidence of
endonuclease activationwithout morphological evidence of
apoptosis.
Aurintricarboxylic acid and Evans blue have been shown toinhibit
endogenous endonuclease activity and have been uti-lized to
delineate a role for endonucleases in DNA damage (20,28-31). The
endonuclease inhibitor aurintricarboxylic acidprovided almost
complete protection against 30 min of hypoxiaand 30 min of
reoxygenation-induced DNA strand breaks inproximal tubules (Fig.
4A). In contrast, fuchsin acid, a struc-tural analog of
aurintricarboxylic acid with little nucleaseinhibitory activity
(29), did not have any protective effect.Aurintricarboxylic acid
provided significant but only partialprotection against
hypoxia/reoxygenation-induced cell death(Fig. 4B), whereas fuchsin
acid was without effect. Evans blue,
another endonuclease inhibitor, also provided complete
pro-tection against hypoxia/reoxygenation-induced DNA damage(Fig.
4C) and partial protection against cell death in proximaltubules
(Fig. 4D). When proximal tubules treated with thesame endonuclease
inhibitors were exposed to a shorter periodof hypoxia (15 min)
followed by reoxygenation (30 min),similar marked protection
against DNA damage and a greaterprotection against cell death were
achieved (data not shown).
DISCUSSIONHypoxia/reoxygenation is an important cause of tissue
injuryin a variety of organs and is classically considered to be
anecrotic form of cell death. In a recent in vivo study
ofischemia/reperfusion to kidney, DNA fragmentation in thekidney
cortex was detected 12 hr after reperfusion (32).Although this
demonstrates that this form of DNA damageoccurs in vivo, it is not
known whether the DNA damage is theresult of cell death or is an
early event prior to loss of cellviability. In addition, the role
of endonuclease activation in theDNA damage or cell death in
hypoxia/reoxygenation injuryhas not been previously examined. The
four major pointsderived from the data in this study are as
follows. First, DNAdamage is a very early event in
hypoxia/reoxygenation injuryand clearly precedes any evidence of
cell death. Second, our
-r
di
.0
.00
.0
.00
10.0
.0
.0
10.0.0.0
Medical Sciences: Ueda et al.
Dow
nloa
ded
by g
uest
on
June
15,
202
1
7206 Medical Sciences: Ueda et al.
data provide compelling evidence for endonuclease activationin
hypoxia/reoxygenation injury. This evidence includes dem-onstration
ofDNA fragmentation by agarose gel electrophore-sis as well as by
an in situ method; an increase in DNA-degrading activity on a
substrate gel that is calcium dependent,is inhibited by an
endonuclease inhibitor, and has an apparentmolecular mass of 15
kDa; and cleavage of nuclear DNAobtained from normal tubules into
small oligonucleosome-length fragments by protein extract from
tubules subjected tohypoxia/reoxygenation. Third, we show that
endonucleaseactivation is entirely responsible for all the DNA
damage andpartially responsible for the cell death that occurs
duringhypoxia/reoxygenation injury. Fourth, despite
unequivocalevidence for endonuclease activation, considered a
character-istic early feature of the apoptotic mode of cell death,
therewere no morphological features typical of apoptosis by
eitherlight or electron microscopy.Endonuclease activation is
considered a hallmark of apo-
ptosis and is frequently used as the sole criteria for
itsdetection (33). Cells undergoing apoptosis show a sequenceof
cardinal morphological features including membraneblebbing,
cellular shrinkage, and condensation and margin-ation of nuclear
chromatin. Endonuclease activation is oneof the earliest events
observed during apoptosis, and it isgenerally assumed that both the
biochemical and morpho-logical features of apoptosis depend on the
activation of anendonuclease (12). Despite overwhelming evidence
for en-donuclease activation, we did not detect any of the
charac-teristic morphological features of apoptosis by light
andelectron microscopy. It is unlikely that this finding inour
study may be due to the transient nature of apoptoticmorphological
change (6, 9, 25, 34) because we examined forthe morphological
change of apoptosis during variable pe-riods of hypoxia (15-30 min)
followed by periods of reoxy-genation (30 min to 4 hr). In
addition, using an in situ assay,we demonstrated DNA fragmentation
at the same timepoints at which morphological features of apoptosis
werelacking. As noted above, DNA fragmentation by agarose
gelelectrophoresis and an increase in endonuclease activitywere
also demonstrated at the same time points. Our datathus indicate
that despite unequivocal evidence of endonu-clease activation in
hypoxia/reoxygenation injury, the mor-phologic features of
apoptosis were not observed at the sametime points. Recent findings
indicate that DNA fragmenta-tion and chromatin condensation may be
triggered throughseparate metabolic pathways (33, 35, 36).The
importance of the endonuclease activation in DNA
damage and cell death was examined by utilizing
endonucleaseinhibitors. Endonuclease inhibitors provided almost
completeprotection against DNA damage and partial protection
againstcell death, especially after a longer period of
hypoxia/reoxy-genation. The partial protection against cell death
provided bythe endonuclease inhibitors suggests that in addition to
endonu-clease activation, other mechanisms including necrosis are
likelyto participate in cell death (33, 37) during longer periods
ofhypoxia/reoxygenation. The traditional view that
ischemia/reperfusion or hypoxia/reoxygenation leads to a necrotic
form ofcell death may be reconciled with our findings of
endonucleaseactivation, considered characteristic of the apoptotic
mode of celldeath. It is likely that when the degree of ischemia is
very severe,the tissue exhibits a necrotic form of cell death; when
it is mild,the cells undergo complete repair. However, with
moderate-to-severe injury, when the damage is such that complete
repaircannot be accomplished, the cell initiates a cascade of
events thatleads to endonuclease-mediated cell injury.
We thank Dainette Powell and Sara G. Shareef for
technicalassistance and Ellen Satter for secretarial assistance.
This work wassupported in part by National Institutes of Health
Grant R01-DK-41480.
1. McCord, J. M. (1985) N. Engl. J. Med. 312, 159-163.2. Paller,
M. S., Hoidal, J. R. & Ferris, T. F. (1984) J. Clin.
Invest.
74, 1156-1164.3. Halliwell, B. & Gutteridge, J. M. C. (1990)
Methods Enzymol.
186, 1-85.4. Nath, K. A. & Paller, M. S. (1990) Kidney Int.
38, 1109-1117.5. Gonzalez-Flecha, B., Cutrin, J. C. & Boveris,
A. (1993) J. Clin.
Invest. 91, 456-464.6. Kerr, J. F. R., Wyllie, A. H. &
Currie, A. R. (1972) Br. J. Cancer
26, 239-257.7. Weinberg, J. M. (1991) Kidney Int. 39, 476-500.8.
Burke, T. J. & Schrier, R. W. (1993) in Diseases ofthe Kidney,
eds.
Schrier, R. W. & Gottschalk, C. (Little Brown, Boston), Vol.
2,1257-1286.
9. Wyllie, A. H., Kerr, J. F. K. & Currie, A. R. (1980) Int.
Rev. Cytol.68, 251-306.
10. Searle, R., Kerr, J. F. R. & Bishop, C. J. (1982)
Pathol. Ann. 17,229-259.
11. Duvall, E. & Wyllie, A. H. (1986) Immunol. Today 7,
115-119.12. Arends, M. J., Morris, R. G. & Wyllie, A. H.
(1990)Am. J. Pathol.
136, 593-608.13. Wyllie, A. H. (1980) Nature (London) 284,
555-556.14. Cohen, J. J. & Duke, R. C. (1984) J. Immunol. 132,
38-42.15. Doctor, R. B. & Mandel, L. J. (1991) J. Am. Soc.
Nephrol. 1,
959-969.16. Walker, P. D., Kaushal, G. P. & Shah, S. V.
(1994) J. Am. Soc.
Nephrol. 5, 55-61.17. Borkan, S. C. & Schwartz, J. H. (1989)
Am. J. Physiol. 257,
F114-F125.18. Bimboim, H. C. & Jevcak, J. J. (1981)
CancerRes. 41, 1889-1892.19. Schraufstatter, I. U., Hinshaw, D. B.,
Hyslop, P. A., Spragg, R. G.
& Cochrane, C. G. (1986) J. Clin. Invest. 77, 1312-1320.20.
Ueda, N. & Shah, S. V. (1992) J. Clin. Invest. 90,
2593-2597.21. Golconda, M. S., Ueda, N. & Shah, S. V. (1993)
Kidney Int. 44,
1228-1234.22. Enright, H., Hebbel, R. P. & Nath, K. A.
(1994)J. Lab. Clin. Med.
124, 63-68.23. McMaster, G. K., Beard, P., Engers, H. D. &
Hirt, B. (1981) J.
Virol. 38, 317-326.24. Peitsch, M. C., Polzar, B., Stephan, H.,
Crompton, T., Mac-
Donald, H. R., Mannherz, H. G. & Tschopp, J. (1993) EMBO
J.12, 371-377.
25. Gavrieli, Y., Sherman, Y. & Ben-Sasson, S. A. (1992) J.
Cell Biol.119, 493-501.
26. Rosenthal, A. L. & Lacks, S. A. (1977)Anal. Biochem.
80,76-90.27. Peitsch, M. C., Hesterkamp, T., Polzar, B., Mannherz,
H. G. &
Tschopp, J. (1992) Biochem. Biophys. Res. Commun. 186,
739-745.
28. Hallick, R. B., Chelm, B. K., Gray, P. W. & Orozco, E.
M., Jr.(1977) Nucleic Acids Res. 4, 3055-3064.
29. Baba, M., Schols, D., Pauwels, R., Balzarini, J. & De
Clercq, E.(1988) Biochem. Biophys. Res. Commun. 155, 1404-1411.
30. Nakane, H., Balzarini, J., De Clercq, E. & Ono, K (1988)
Eur.J. Biochem. 177, 91-96.
31. Batistatou, A. & Greene, L. A. (1991) J. Cell Biol. 115,
461-471.32. Schumer, M., Colombel, M. C., Sawczuk, I. S., Gobe, G.,
Connor,
J., O'Toole, K M., Olsson, C. A., Wise, G. J. & Buttyan,
R.(1992) Am. J. Pathol. 140, 831-838.
33. Schulze-Osthoff, K., Walczak, H., Droge, W. & Krammer,
P. H.(1994) J. Cell Biol. 127, 15-20.
34. Russell, S. W., Rosenau, W. & Lee, J. C. (1972)Am. J.
Pathol. 69,103-118:
35. Sun, D. Y., Jiang, S., Zheng, L.-M., Ojcius, D. M. &
Young,J. D.-E. (1994) J. Exp. Med. 179, 559-568.
36. Cohen, G. M., Sun, X. M., Snowden, R. T., Dinsdale, D.
&Skilleter, D. N. (1992) Biochem. J. 286, 331-334.
37. Schwartz, L. M., Smith, S. W., Jones, M. E. E. &
Osborne, B. A.(1993) Proc. Natl. Acad. Sci. USA 90, 980-984.
Proc. Natl. Acad. Sci. USA 92 (1995)
Dow
nloa
ded
by g
uest
on
June
15,
202
1
