Proc. Nati. Acad. Sci. USAVol. 83, pp. 7956-7960, October 1986Microbiology

The double-stranded RNA in Trichomonas vaginalis may originatefrom virus-like particlesALICE L. WANG AND CHING C. WANGDepartment of Pharmaceutical Chemistry, University of California, School of Pharmacy, San Francisco, CA 94143

Communicated by H. A. Barker, June 23, 1986

ABSTRACT A linear 5.5-kilobase double-stranded RNA,identified in many strains and isolates of the parasitic proto-zoan Trichomonas vaginalis in a previous study, is found largelyintact in ribonuclease-treated homogenates of the parasite. Itcan be pelleted with membranes from the homogenate at 12,500x g and further purified in CsCl buoyant density-gradientcentrifugations. The purified sample contains the double-stranded RNA as well as one major protein with an estimatedmolecular mass of 85 kDa in NaDodSO4/PAGE. Electronmicroscopic examinations indicated the presence of icosahedralvirus-like particles of 33-nm diameter in the purified prepa-ration. The exact location of the virus in T. vaginalis is notclear, except that it is not found in the nuclear fraction and isprobably membrane-bound. No free virus can be recoveredfrom the culture medium of T. vaginalis, and no successfulinfection of virus-free T. vaginalis strains by purified virus hasyet been accomplished. There is no viral genomic sequenceidentifiable in host DNA. So far as we know, it is the first timea double-stranded RNA virus has been identified in a proto-zoan.

Trichomonas vaginalis is a sexually transmitted protozoanparasite found primarily in the human vaginal tract. Recentstudies indicated the presence of a double-stranded RNA (dsRNA) in the nucleic acid extract of this organism (1). The dsRNA has a linear structure with an estimated contour lengthof 1.5 gm (1). It consists of 23.4% guanine, 23.4% cytosine,23.0% adenine, and 30.3% uracil, and has a transitiontemperature of 81.7°C with a hyperchromicity of7-15% in 75mM NaCl/7.5 mM sodium citrate, pH 7.0 (0.5x SSC; lxSSC = 0.15 M NaCl/0.015 M sodium citrate). It migrates in0.8% agarose gel electrophoresis, with a mobility equivalentto a DNA size of 5.5 kilobases (kb), and it can be readilystained by ethidium. It is digestable by alkali and sensitive toRNases A and T1 in low-salt solutions. The sensitivitytoward RNase T1 is, however, much reduced in high-saltsolutions, suggesting a structure of ds RNA.

Since this report (1) appears to be the first to identify thepresence of a ds RNA in protozoa, the biological significancein this finding is not immediately clear. This ds RNA has sincebeen identified in some 40 different strains or isolates of T.vaginalis at densities ranging from 280 to 1380 copies per cell.There have been, however, four strains of T. vaginalis foundnot to contain the ds RNA. Three of the four strains turnedout to be resistant to the anti-trichomonial agent metronida-zole (2). There seems, thus, a connection between thepresence of this ds RNA and the sensitivity of T. vaginalistoward metronidazole, since all the 40 ds RNA-containing T.vaginalis samples are also sensitive to metronidazole. How-ever, the one T. vaginalis strain 375, which has no detectableds RNA but remains susceptible to metronidazole, arguesagainst a possible role of the ds RNA in metronidazolesensitivity. Furthermore, a cattle protozoan parasite Tritri-

chomonasfoetus, which is closely related to T. vaginalis andalso highly susceptible to metronidazole, contains no ds RNA(1). It is thus conceivable that the ds RNA could be eliminatedfrom T. vaginalis under the pressure of metronidazole. In theyeast Saccharomyces cerevisiae (3), two ds RNA species, L(4.3-4.8 kb) and M (1.7-1.9 kb), were found associated withthe virus-like particles. L encodes primarily the capsidprotein of the virus-like particles, whereas M encodes apolypeptide toxin of 109 amino acids capable of causing anirreversible change in the plasma membrane of S. cerevisiaeand, consequently, the loss of intracellular ions and ATP (4).The ds RNA in T. vaginalis has apparently no killing effect,and no virus-like structures have yet been identified insectioned T. vaginalis by electron microscopy (1). Thepurpose of the present investigations is to further test apossible association between the ds RNA and a virus in T.vaginalis similar to that observed in yeast.

EXPERIMENTAL PROCEDURESCultures. T. vaginalis ATCC 30001 and two metronidazole-

resistant strains IR78 and CDC85 (2) were maintained andcultivated in TYM medium (pH 6.2) supplemented with 10%heat-inactivated horse serum as described (5, 6). All strainsof T. vaginalis have been cloned and verified to be free ofbacteria, mycoplasma, and plasmids (7).

Electron Microscopy. Samples of cell suspension werefixed with 2.5% glutaraldehyde and pelleted by a briefcentrifugation (7). The pellets were then fixed with 1%osmium tetroxide, dehydrated, embedded in epon, sec-tioned, and stained with uranyl acetate in searching forvirus-like particles in situ. The purified virus-like particleswere fixed and negatively stained by a similar proceduredescribed by Haschmeyer and Meyers (8).

Radiolabeling of ds RNA and Hybridization with DNADigests. The ds RNA was extracted from purified virus-likeparticles with phenol by a standard procedure (9) and labeledat its 3'-end with [32P]pCp by T4 RNA ligase (10). Theremaining free [32P]pCp was removed by Sephadex G-25chromatography.

Restriction fragments of T. vaginalis DNA were fraction-ated by 0.8% agarose gel electrophoresis (7) and transferredto a nitrocellulose filter (10). The filter was prehybridized in50% deionized formamide/0.9 M NaCl/50 mM sodium phos-phate, pH 7.4/5 mM EDTA/0.02% Ficoll/0.02% polyvinyl-pyrrolidone/0.02% bovine serum albumin/denatured salmonsperm DNA (100 ,g/ml) at 42°C for 4 hr. One microgram ofthe purified 32P-labeled ds RNA (50-60 x 106 cpm) was boiledfor 5 min, chilled, and added to the hybridization solution.After 16 hr at 42°C, the filter was washed three times in 2xSSC and 0.1% NaDodSO4 at room temperature followed byincubation with lx SSC and 0.1% NaDodSO4 at 50°C for 1.5hr. It was then rinsed twice with 1x SSC at room tempera-ture, dried, and exposed to Kodak XAR-5 film betweenintensifying screens at -70°C.

Abbreviations: ds RNA, double-stranded RNA; kb, kilobase(s).

7956

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

Proc. Natl. Acad. Sci. USA 83 (1986) 7957

Protein Analysis. Samples were boiled for 5 min in a buffercontaining 62.5 mM Tris HCl, 2% NaDodSO4, 10% (vol/vol)glycerol, and 5% 2-mercaptoethanol (pH 6.8). NaDodSO4/PAGE was carried out in 11% gels according to the methodof Laemmli (12). The protein bands were identified by silverstaining (13).

Materials. Pancreatic RNase A and proteinase K werepurchased from Sigma. Restriction enzymes and X DNA werefrom Bethesda Research Laboratories. T4 RNA ligase wasobtained from New England Biolabs, and [32P]pCp was fromAmersham. All other chemicals were of the highest puritycommercially available.

RESULTSLocalization of the ds RNA. T. vaginalis grown to the late

logarithmic phase were harvested, washed in phosphate-buffered saline (PBS), and resuspended in 10 vol of PBS. Thecell suspension was homogenized at 0C-40C in aPotter-Elvehejm tissue grinder driven by a lab stirring motor(H. K. Heller, Bellerose, NY) for 20 min. The crudehomogenate was then subjected to differential centrifuga-tions at 3000 x g for 15 min and at 12,500 x g for 90 min, withthe pellets collected after each spin designated P1 and P2 (Fig.1). The P2 fraction was resuspended in 10 vol of PBS,homogenized again for 10 min to free the virus-like particlesfrom membrane fragments, and centrifuged at 12,500 x g for90 min (S3 and P3). Small samples of the original crudehomogenate and the separated fractions were extracted withphenol, precipitated with ethanol, and analyzed in 0.8%agarose gel electrophoresis. The results (Fig. 2) indicate thatthe ds RNA remained largely intact in the crude homogenate.It was not in the nuclear fraction P1, but it was found in S1,P2, S2, P3, and S3 (Fig. 2). Ribosomal RNA was found in crudehomogenate and S2 only. The ds RNA was originally broughtdown at 12,500 x g (P2) but released back to the supernatantfraction (S3) upon further grindings, which may suggest anoriginal association of the ds RNA with particles or mem-brane vesicles sedimenting at 12,500 x g (P2). Electronmicrographs ofthe P2 fraction indeed indicate the presence ofaggregated virus-like particles inside membrane vesicles(Fig. 3) not readily penetrated by negative stain.

Identification of the ds RNA with Virus-Like Particles.Pancreatic RNase A was added to the crude homogenate ofT. vaginalis to a final concentration of 10 ug/ml and incu-

CELL SUSPENSION

I Homogenization, 20 min

CRUDE HOMOGENATE

| Centrifugation, 3,000 xg, 15 min

PELLET (Pi) SUPERNATANT (S.)

PELLET (P2)

PELLET (P3)

Centrifugation, 12,500 xg, 90 min

SUPERNATANT (S2)

Homogenization, 10 min

Centrifugation, 12,500 xg, 90 min

SUPERNATANT (S3)

FIG. 1. The protocol of purifying virus-like particles from T.vaginalis.

H Pi P2 S2

-23.5- 9.7- 6.6- 4.3

- 2.2- 2.1

FIG. 2. Agarose gel electrophoretic analysis of samples from thefractionation of T. vaginalis crude homogenates. Gels of 0.8%agarose were cast in 89 mM Tris borate, pH 8.3/2.5 mM EDTA.Bands of nucleic acids were identified by staining with ethidiumbromide (0.5 ,ug/ml). The size markers (in kb) are from the X DNAHindIII fragments. H, crude homogenate; P1, 3000 x g pellet, thislane was purposely overloaded to show the absence of ds RNA; P2,12,500 x g pellet; S2, 12,500 x g supernatant.

bated at 37°C for 30 min. The treated homogenate showed thepresence of intact ds RNA after phenol extraction, suggestingthat the ds RNA remained largely unaffected by the RNasepretreatment. However, when 0.1% NaDodSO4 or protein-ase K (100 ,g/ml) was added to the crude homogenate withor without the RNase treatment, the ds RNA became nolonger detectable in gel electrophoresis (data not shown).Apparently, the ds RNA in the crude homogenate wasprotected from both exogenous and endogenous RNase, butthe protection could be removed by NaDodSO4 or proteinaseK.The S2 and S3 fractions were pooled and CsCl was added

to a final density of 1.39 g/ml. The mixture was centrifugedin an SW41 rotor at 34,000 x g and 20°C for 16 hr. Fractionsof the CsCl buoyant-density gradient were collected in a0.5-ml vol, and the absorbance at 260 nm of each fraction wasmonitored with a Beckman DU7 spectrophotometer (Fig. 4).Samples of individual fractions were extracted with phenoland analyzed in gel electrophoresis. Fig. 4 indicates that thepeak fractions ofabsorbance at 260 nm identified at the centerof the gradient (p = 1.468 g/ml) are also the fractionscontaining the ds RNA. Ribosomal RNA is found exclusivelyat the bottom of the gradient together with a substantialamount of the ds RNA (Fig. 4). Electron microscopicexaminations of this bottom fraction again indicated thevirus-like particles entrapped in membrane vesicles as shownin Fig. 3. Fractions 9-11 were pooled, diluted with 15 vol ofPBS, and centrifuged in an SW41 rotor at 109,000 x g and 4°Cfor 2 hr. The pellets were negatively stained and examined byelectron microscopy (8). Fig. 5 shows that the pellets consistofuniformly virus-like particles. They are icosahedral with anestimated diameter of 33 nm. No such virus-like particles

Microbiology: Wang and Wang

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

7958 Microbiology: Wang and Wang

0 I' okMA

FIG. 3.Eeto irgaho egtvl tie 2fato (seFg ) Br 10n

were detectable in those fractions without detectable A260 ords RNA in gel electrophoresis (Fig. 4). Similar experimentswere performed on the metronidazole-resistant T. vaginalisstrains IR78 and CDC85, which are known to contain nodetectable ds RNA (1). No virus-like particles were found inthese two strains.

Characterization of the Virus-Like Particles. The purifiedvirus-like particles were named T. vaginalis virus. An aver-age yield equivalent to 3-4 ng of T. vaginalis virus protein per106 T. vaginalis has been established. There is, however, nofree T. vaginalis virus detectable in the culture medium up tothe stationary phase. Many sections of a T. vaginalis cellhave been examined under transmission electron micro-scope, but no evidence of the presence of mature T. vaginalisvirus has yet been found.The purified T. vaginalis virus preparation (-1 gg of ds

RNA) was extracted with phenol and the extracted nucleicacid was precipitated with ethanol. Electrophoretic analysis(not shown) indicated the presence of a single ethidium-stainable band with a migration in the gel corresponding to asize of 5.5 kb of DNA. It is thus concluded once again thatthe ds RNA is the nucleic acid component of T. vaginalisvirus.The same purified T. vaginalis virus (=4 ;kg of protein) was

then analyzed in NaDodSO4/PAGE for its protein composi-tion. The data presented in Fig. 6 show one major band withan estimated molecular mass of 85 kDa. This band isdetectable by both silver staining and Coomassie blue stain-ing and is thus most likely a protein.The ds RNA extracted from purified T. vaginalis virus was

labeled with [32P]pCp at the 3' end by T4 RNA ligase and wasused as a probe to hybridize with DNA fragments of T.vaginalis on Southern blots (10). EcoRI and HindIII digestsof T. vaginalis DNA derived from both the T. vaginalisvirus-containing strain ATCC 30001 and the virus-free strainIR78 were used in these experiments. The results indicatedno sign of hybridization between the DNA fragments and theds RNA, thus suggesting the absence of homology between

ds RNA and host DNA. Control experiments with denaturedT. vaginalis virus ds RNA on dot blots found highly efficienthybridization with the radiolabeled probe.

Finally, an attempt was made to use the purified T.vaginalis virus to infect the two strains of T. vaginalis IR78and CDC85, known to contain no such virus-like particles (1).Purified T. vaginalis virus was added to the in vitro culturesof the two T. vaginalis strains to a final concentration of 10Ag of ds RNA per ml. The two strains were found growing atnormal rates under such conditions. The late logarithmicphase cells were harvested, homogenized, and fractionatedby differential and CsCl gradient centrifugations and checkedfor virus-like particles and ds RNA. No such particles or dsRNA were detected in the cell homogenates, suggesting thefailure of viral infection of T. vaginalis.

DISCUSSIONThe present investigation has provided unequivocal evidencethat virus-like particles (T. vaginalis virus) containing the1.5-gm linear ds RNA (1) and a major protein of 85 kDa arepresent in T. vaginalis. Since the ds RNA has been detectedin some 40 independent isolates and strains of T. vaginalis (1),the virus must have a prevalent presence in the parasite. Theabsence of T. vaginalis virus in the culture medium of T.vaginalis may reflect instability of the virus in an externalenvironment, which may partly explain the failure of purifiedT. vaginalis virus to infect virus-free T. vaginalis. Thepurified T. vaginalis virus, however, has apparently retainedits nucleic acid component in spite of the relatively intensenegative stain of the virus-like particles (Fig. 5), because theds RNA was successfully extracted from the purified T.vaginalis virus. The finding of one major 85-kDa proteinsuggests that the latter could be the capsid protein of T.vaginalis virus, but the presence of other viral proteins atlower levels such as the RNA-dependent RNA replicasecannot be ruled out. T. vaginalis virus is undoubtedlymultiplied in the growing T. vaginalis. But there is no

Proc. Natl. Acad. Sci. USA 83 (1986)

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

Proc. Natl. Acad. Sci. USA 83 (1986) 7959

A

3.9

Fractions2 4 6 8 10 12 14 16

\' ~~~~i\0~~~~~

\I%\

, \

.

1 3 5 7

0

0

I11

x 0

A1

9 11 13 15 17Fractions

18

FIG. 4. CsCl buoyant-density gradient centrifugation analysis ofthe S2 and S3 fractions of T. vaginalis crude homogenate. (A) Agarosegel electrophoresis profile of individual fractions. Size markers are inkb. (B) Absorbance at 260 nm profile of individual fractions. Fraction1 is the bottom of the gradient.

evidence of symbiosis between the virus and the protozoan,since the latter can grow normally without the virus. It is notknown whether certain aspects of the pathogenicity in T.vaginalis infections could be attributed to the presence of T.vaginalis virus. It is also unclear whether T. vaginalis virushas a specific location inside the infected T. vaginalis, but itmay have a close association with the membrane fraction(Fig. 3). The difficulty in identifying T. vaginalis virus insectioned T. vaginalis under the electron microscope couldbe due to the relatively small number of T. vaginalis virus(280-1380; see ref. 1) in a T. vaginalis cell, which has a ratherlarge diameter of 15 ,um. More intensive in situ searches forthe virus are currently under way.

This is the only ds RNA virus we know of that has beenidentified in a protozoan. Mattern et al. (14) observed thepresence of virus-like particles in Entamoeba histolytica. Theparticles have two morphological types-a filamentous formand a polyhedral (icosahedral) form, mostly 75-85 nm indiameter (14, 15). The latter apparently consists ofDNA andis lytic to certain strains of E. histolytica (16). Later studiesrevealed the presence ofa third type ofbeaded particles in thenuclei of a few amebal strains (17). None of the threevirus-like particles bears much resemblance to T. vaginalisvirus. There is, however, considerable resemblance betweenT. vaginalis virus and the killer ds RNA viruses of S.cerevisiae, especially species L (3). Both S. cerevisiae dsRNA virus species L and T. vaginalis virus containnonsegmented ds RNA genomes, and their shapes, sizes,buoyant densities, and nucleic acid and protein componentsare all quite similar (Table 1). They are transmitted withinindividually defined genetic systems for the host-i.e., theyare infectious by heredity. Their propagation relies on celldivision, a feature normally associated with the naturallyoccurring plasmids. One interesting possibility would be thatboth T. vaginalis and S. cerevisiae may infect the vaginaltract to allow exchange of the same ds RNA virus betweenthe two species.

Studies of protozoa have been hampered in the past

1. Xe;i t~~~~4t Y

-Y;-m+

0 0 ;'. , + ;<0 '0;4

,,<w~~~~~~jr,;~$ S t :,; , ,, 0 0 ';' vit''s> : si."1> '',i-""t;S ++V~~iaMfdilit,'.?g~~~tf~~t ^ 4. '''

FIG. 5. Electron micrograph of the purified virus-like particles (T. vaginalis virus). (Bar = 100 nm.)

S

2.01

1.5

Microbiology: Wang and Wang

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020

7960 Microbiology: Wang and Wang

kDaA B

92.5- _ 85

45.0

31.0 _

21.5 A_

FIG. 6. NaDodSO4/PAGE analysis of the protein profile ofpurified virus-like particles (T. vaginalis virus). Lanes: A, proteinstandards; B, T. vaginalis virus protein. The hazy band correspond-ing to 66 kDa is a staining artefact from the silver stain in alkalinesolution due to the inclusion of 2-mercaptoethanol (12).

because of the lack of vectors capable of performing genetictransformation for the organisms. This deficiency has beenparticularly limiting on the investigations of the type ofprotozoa like T. vaginalis, which lacks sexual differentiationduring its development and, thus, has no apparent means ofgenetic recombination. Our discovery of T. vaginalis virusmay make genetic manipulation of T. vaginalis possiblethrough viral transfection. The absence of a viral genomicsequence in the host DNA suggests a lack of reversetranscription and integration of the viral genome. In ourprevious investigation with the ds RNA purified directly fromT. vaginalis by NACS-37 column chromatography and twocycles of agarose gel electrophoresis, some hybridizationbetween the labeled RNA sample and T. vaginalis DNAfragments was observed on Southern blots (1). This wasattributed to the presence of some labeled ribosomal

Table 1. Comparisons of S. cerevisiae virus L* andT. vaginalis virus

S. cerevisiae virus T. vaginali$ virus

Shape Icosahedral IcosahedralDiameter, nm 33-41 33Density in CsCl 1.40-1.42 1.468ds RNA

Lengtht, am 1.31 1.50Estimated size, kb 4.3-4.8 5.5t

Major capsid protein, kDa 88 85

All information regarding S. cerevisiae virus is cited from ref. 3.*5. cereviziae ds RNA virus species L.tDetermined by electron microscopy.*This is calculated from electron microscopy data assuming 3 A foreach base pair of ds RNA (18).

RNA,which was removed from the purified T. vaginalis virusin the present studies. The T. vaginalis DNA fragmentsidentified in the previous Southern hybridization experi-ments were cloned in the pUC18 plasmid and identified as theT. vaginalis ribosomal gene by use of a cloned Schistosomalmansoni ribosomal gene (a gift from Philip Loverde, NewYork State University, Buffalo) as the probe for Southernhybridization (unpublished observation).

Finally, in our recent investigations on another relatedanaerobic parasitic protozoan, Giardia lamblia, a similar butdistinctive ds RNA virus specific to the parasite was identi-fied, purified, and characterized (19). It is found in the twonuclei of G. lamblia trophozoite and appears as a sphere 33nm in diameter. It contains a linear ds RNA equivalent to thesize ofa 7.0-kbDNA with little homology to T. vaginalis virusds RNA, and it has one major 66-kDa protein. In contrast toT. vaginalis virus, this virus can be harvested from theculture medium of G. lamblia and used to infect the virus-freestrains of G. lamblia. Thus, the prospect of finding a vectorfor genetic transformation in protozoa may soon come veryclose to reality.

The authors would like to express their gratitude to Dr. Harold E.Varmus (University of California, San Francisco), Drs. AustinNewton and Noriko Ohta (Princeton University), and Dr. J. A.Bruenn (New York State University, Buffalo) for their interest in ourstudies, their stimulating discussions, and their helpful suggestions.Thanks also go to Dr. Philip LoVerde for his generous gift of clonedS. mansoni ribosomal gene, and Ms. Mei Lie Wong (University ofCalifornia, San Francisco) for her invaluable help in preparing theelectron micrographs of the virus-like particles. This work wassupported by Grant AI-19391 from the National Institutes of Health.C.C.W. is a Burroughs-Wellcome Scholar in Molecular Parasitology.

1. Wang, A. L. & Wang, C. C. (1985) J. Biol. Chem. 260,3697-3702.

2. Mflller, M. & Gorell, T. E. (1983) Antimicrob. Agents Chemo-ther. 24, 667-673.

3. Bruenn, J. A. (1980) Annu. Rev. Microbiol. 34, 49-68.4. Skipper, N. & Bussey, H. (1977) J. Bacteriol. 129, 668-677.5. Diamond, L. S. (1957) J. Parasitol. 43, 488-490.6. Wang, C. C. & Cheng, H.-W. (1983) Mol. Biochem. Parasitol.

10, 171-184.7. Wang, A. L. & Wang, C. C. (1985) Mol. Biochem. Parasitol.

14, 323-336.8. Haschmeyer, R. H. & Meyers, R. J. (1972) in Principles and

Techniques of Electron Microscopy, ed. Hayat, M. A. (VanNostrand-Reinhold, New York), Vol. 2, pp. 101-147.

9. Blin, N. & Stafford, D. W. (1976) Nucleic Acids Res. 3,2303-2308.

10. Southern, E. M. (1975) J. Mol. Biol. 98, 503-517.11. Bruenn, J. & Brennan, V. (1980) Cell 19, 923-933.12. Laemmli, U. K. (1970) Nature (London) 227, 680-685.13. Wray, W., Boulikas, T., Wray, V. P. & Hancock, R. (1981)

Anal. Biochem. 118, 197-203.14. Mattern, C. F. T., Diamond, L. S. & Daniel, W. A. (1972) J.

Virol. 9, 342-358.15. Diamond, L. S., Mattern, C. F. T. & Bartgis, I. L. (1972) J.

Virol. 9, 326-341.16. Hruska, J. F., Mattern, C. F. T. & Diamond, L. S. (1974) J.

Virol. 13, 205-210.17. Diamond, L. S. & Mattern, C. F. T. (1976) Adv. Virus Res. 20,

87-112.18. Langridge, R. & Gomatos, R. J. (1%3) Science 141, 694-698.19. Wang, A. L. & Wang, C. C. (1986) Mol. Biochem. Parasitol.,

in press.

Proc. Natl. Acad. Sci. USA 83 (1986)

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

12, 2

020