Course and extent of variation of equine infectious anemia virus during parallel persistent infections.

(1)

0022-538X/87/041266-05$02.00/0

Course and Extent of Variation

of Equine Infectious Anemia Virus

during Parallel Persistent

Infectionst

SUSAN L. PAYNE,' OLIVIASALINOVICH,'t SUZANNE M. NAUMAN,' CHARLES J. ISSEL,2'3 AND RONALD C.

MONTELARO1*

Departments of

Biochemistry' and Veterinary Science,2

and Louisiana

Agricultural Experiment

Station,"

2andthe

Department of

Veterinary

Microbiology and

Parasitology,3

School

of

Veterinary Medicine,

Louisiana State

University,

Baton

Rouge, Louisiana

70803

Received 3 September 1986/Accepted 16 December 1986

Comparisons

of

peptide

and

oligonucleotide

mapsof

glycoproteins

and RNA from nine

isolates

of

equine

infectious

anemiavirus (EIAV) thatweregenerated

during parallel

infections oftwoShetlandponies

revealed

that each

isolate

was

structurally unique. Each

EIAVisolate contained a

unique

subset of variant

peptides,

oligonucleotides,

or

both, indicating

that structural variation in

EIAV

isarandomand noncumulativeprocess

and

thata

large

spectrumof

possible

EIAV variantscanbe

generated

in infected animals.

The

members of the lentivirus subfamily

of

retroviruses,

including equine infectious

anemia virus

(EIAV),

visna

virus, and human immunodeficiency

virus

(HIV),

cause

chronic diseases ina

variety

of animals and

humans

(2, 5, 6,

8, 16).

The

persistent

infections

caused

by

these viruses in some

instances

can

be traced

to

their

ability

to

circumvent

or overcomethe immuneresponses

of

the host

(4, 6, 9,

18).

In thecase

of

EIAV, which

causesa

unique episodic

disease

in

horses, it has been demonstrated that isolates recovered

from

an

infected

animal

during

different febrile

episodes

can

be

distinguished

antigenically

and

structurally by

a

variety

of

biochemical

and

immunological

assays

(11, 15, 18). Results

of these studies have indicated that structural and

antigenic

variation is localized

to

the envelope glycoproteins gp9O and

gp45 (11, 18).

The occurrence of

envelope

glycoprotein

variation has also been

demonstrated for visna virus and

HIV

(1, 4, 7, 19).

A

critical question regarding lentivirus variation is

the number

of variants that

canoccurinnature,a

major

factor

in

assessing potential

vaccines. To analyze the

spectrum

of

structural

variation

among

isolates

of

EIAV,

we used

pep-tide, glycopeppep-tide, and oligonucleotide mapping procedures

to compare nine EIAV

isolates

recovered

during distinct

clinical

episodes

in two

ponies infected with the

same virus.

To

conduct these

experiments, identical virulent virus

inocula

were

used in

parallel infections of

two

Shetland

ponies

as

described

previously (11, 13, 18). The clinical

histories

of these

animals

are

summarized in Fig.

1.

Plasma

samples

taken

during

clinical episodes

were

subjected

to

endpoint dilution in fetal equine kidney cells

torecover

the

predominant virus

population (13). The nine

EIAV

isolates

recovered were

propagated in fetal equine kidney

cells and

purified

by glycerol gradient centrifugation

(10, 11, 14). Each virus

isolate

was

determined

to

be antigenically

distinct by

using

a

variety of

immunoassays employing

serum from

infected

ponies, panels

of monoclonal antibodies, or both

* Corresponding author.

t Louisiana Agricultural Experiment Station paper no. 86-12-0110.

tPresentaddress: Lovelace InhalationToxicologyResearch

In-stitute, Albuquerque, NM 87185.

(18; A. Orrego, Ph.D.

thesis, Louisiana State

University,

Baton

Rouge,

1983).

Theextent

of

structural

variation

in the

viral

glycoproteins

gp9O

and

gp45

for each of the

nine

EIAV isolates was

assayed

by two-dimensional

125I-labeled

tryptic

peptide

(11,

a)

0.

H-cL

-a a

0)

-_

a.0a. a.

r(i re5

a. a.

Doys

Post

Inoculotion

FIG. 1. Clinical histories of experimentally infected ponies showing clinical episodes from which virusisolateswererecovered. Both animalsreceivedidentical.intravenous virus inocula containing

4.8loglo50% tissuecultureinfective doses ofhost-adapted EIAV (13).Allsustainedrectaltemperaturerecordings above 39°C (dashed line) were considered abnormal and were designated febrile

epi-sodes. (A) Pony 91 exhibited classical chronic equine infectious anemia thatwascharacterized by recurring febrile episodes.Pony 91 died after fourfebrileepisodes, onday 180. (B) Pony 127 also experienced chronic equine infectious anemia. The 5 months

be-tweenthe thirdandfourth febrileepisodesrepresentaninapparent

stage of the diseasein which the asymptomatic animal remainsa

viruscamer. 1266

on November 10, 2019 by guest

http://jvi.asm.org/

[image:1.612.327.564.354.605.2]

(2)

B

600

14 015

I0

eo

P

O02

Cj

32 33 03541

330 4 4

U 5

A2

0

*0*4

3

60 9.31

39410001

lo*'

160

2

U

446

onz

170

-.19

230 'ic2

310@033 350

02

32 347@ O

4

10

32

3!

41012

,,

A

U~~~~~~~'O

sbt

b

_

a

Q- O"

[image:2.612.136.494.68.308.2]

OU0

FIG. 2. Compositetwo-dimensionaltryptic peptidepatternsgenerated frommapsof 251I-labeled glycoproteinsgp9O and gp45 from each

EIAVisolate. Thepeptides generated for each virusisolateweremapped in duplicate. The compositemapsweregenerated by comparing

each virus isolate with all other isolates by directly overlaying autoradiographs. Arrowsindicate the directions ofelectrophoresis (a) and

chromatography (b). Arbitrary numberswereassignedtoeachpeptidefor identification.Closed circlesindicatepeptides thatwere common

toallvirusisolates. Open circles indicate variant peptides.Thehatchedarearepresentsradioiodinatedtryptic glycopeptides thatwerenot

resolved bystandardpeptide mapping techniques(11). (A) gp9Ocompositepeptidemap.(B)gp45composite peptidemap.

18)

and

glycopeptide (17) mapping

procedures. All peptide and

glycopeptide

patterns were

reproducible

on

repeated

mappings,

and virus isolates

remained stable during

contin-uedpassagein

tissue

culture. Figure 2 shows the

composite

maps

developed

to

display

conserved and variant peptides

for

gp9O and gp45.

A

total

of 54

peptides

were

identified

on

comparison

of the

gp9O

component

of the nine variants and

the prototype

strain of EIAV.

Twenty-seven

(50%) of the

gp9O

peptides

were

found

to

be

commonto

all

isolates. The

pattern

of variation

of the remaining

peptides

was

used

to

uniquely identify eight of

nine

variant

strains,

as

well

as

the

prototype

strain of the

virus

(Table 1, gp9O peptides). Only

isolates

P3.2-2

and

P3.2-3

could

not

be

distinguished by gp9O

peptide

mapping

analysis. For gp45 only

11

of

45

peptides

(24%)

werecommonto

all

isolates. In

this

case

the

pattern

of

variant

peptides could be used

to

distinguish all isolates

(Table 1, gp45

peptides); P3.2-2 and P3.2-3 could be

distin-guished.

Table 1 lists those

peptides

not common to all virus isolates. Some of these variant

peptides

were present in isolates from onlyone pony. These

included gp9O

peptides

12, 18, 20,

and25

and

gp45

peptides 6, 14, 15, 25, 27, 29,

and 30.

gp9O peptides

18 and 25 were

unique

to

single

virus isolates. Most

other

peptides

were present in and varied

among

isolates

from both

animals,

however. For

example,

gp9O peptide

49 was present in

P3.1-2,

P3.2-1,

and

P3.2-4;

gp9O peptide

52waspresent in

P3.1-1, P3.1-3, P3.2-2, P3.2-3,

and P3.2-4. The data suggest that structural variations

among virus isolates do not accumulate with

time,

as

pep-tides appear in one isolate, only to be lost in subsequent isolates.

Instead,

it appears that a limited set of

peptides

vary

independently, leading

to the presence of a different subset of

peptides

for each isolate.

In contrast to the

envelope

glycoproteins, the peptide

maps

for

the viral core

proteins p9,

p15,

and p26

were

identical for all virus isolates (data

not

shQwn),

as

reported

previously (11, 17).

gp9O

and

gp45

fromeach virus isolatewere

also

analyzed

by

glycopeptide mapping

to compare the

glycosylated

tryptic

peptides

not

resolved by standard

peptide mapping

procedures (11, 17, 18). Four classes of gp9O and

two

classes

of

gp45

glycopeptide

patterns were

observed.

Representa-tive

glycopeptide

maps are

shown

in

Fig.

3, and

in

Table

2 the

glycopeptide

patterns

obtained for each

virus

isolate

are

summarized.

These

data

indicate that

changes

in

the

pattern of

glycosylation

of

gp9O

and

gp45

are

independent of

one

another.

For

example,

all virus

isolates from

pony

91 (P3.2-1

through P3.2-4) share

acommon

gp45

glycosylation

pattern,

while they

exhibit four different gp9O

glycosylation

patterns. Itis

interesting

that virus isolatesP3.2-2 and

P3.2-3,

which

produce

apparently

identical

gp9O

peptide

patterns, exhibit different

glycopeptide

maps.

To

analyze

the

genomic

variation among EIAV isolates, RNAwas

purified

fromeach virus and

analyzed by

oligonu-cleotide

mapping by

previously described techniques (15,

18).

A

composite

map

showing

conserved and variant

oligo-nucleotides

was then

developed (Fig. 4).

A total of 51

oligonucleotides

are

represented

in

Fig. 4,

and 33

(65%)

of these were common toall isolates.

Eight

ofthe nine virus strains and the prototype virus could be

distinguished by

analysis

of the distribution ofvariant

oligonucleotides. Only

isolates

P3.2-1

and

P3.2-4

couldnotbe

distinguished easily.

(Table 3).

As observed with the

peptide

maps described above, there

appeared

to be no accumulation of variant

oligonucleotides

amonglate virus isolates.

Results of the

experiments

presented

here

provide

a

A

b4

120

lSO

24

05

0° o3

ft

0

lm-23

m

on November 10, 2019 by guest

http://jvi.asm.org/

(3)

detailed comparison of nine virus

isolates recovered from

two

ponies

infected with identical

inocula of EIAV. Several

important properties of EIAV variation

can

be concluded

from these observations.

A

distinct virus population predominates during each

febrile episode in

a

persistently infected

pony.

The variant

virus strains examined

in this study

were

recovered from

endpoint

dilutions of plasma and, thus,

are

assumed

to represent

the

predominant virus populations in the infected

animal

at

the time the plasma samples

were

obtained.

[image:3.612.319.556.69.275.2]

There

appearsto

be

a

relatively

large number of structural

variations possible in EIAV. In

addition

to

the nine virus

isolates

described here,

we

recovered five

more

EIAV

isolates from

a

third

pony

that received the

same

initial virus

inoculum (unpublished data). These virus isolates could also

be

distinguished structurally,

bringing the total number of

unique

variants generated from this virus inoculum

to

14. These 14

isolates

were

distinct

from isolates recovered from

TABLE 1. DistributionamongEIAVstrains of variantpeptides from the envelope glycoproteins gp9O and gp45

Distribution of: Virus Peptide

gp9O peptidesa

gp45peptidesb

Prototype 7, 20, 28, 36, 44, 45, 46, 47, 49, 50, 51, 52, 53

P3.1-1 7, 15, 28, 29, 30, 33, 35, 36, 38, 39, 40, 44, 45,46, 47, 51, 52, 53 P3.1-2 7, 12, 15, 21, 29, 35, 36, 38, 39, 43,

46, 47, 48,49, 50, 51

P3.1-3 7, 15, 25, 28, 29, 30, 33, 35, 39, 44, 45,46, 47, 52, 53

P3.1-4 7, 12, 15, 29, 30, 35, 38, 39, 43, 46, 47 P3.2-1 20, 28, 36,38, 39, 40, 44, 45, 46, 49,

51, 53

P3.2-2 7, 15, 20, 21, 28, 29, 30, 36, 38, 39, 40, 43, 44,45, 46, 47, 48, 50, 51, 52, 53

P3.2-3 7, 15, 20, 21, 28, 29, 30, 36, 38, 39, 40, 43, 44,45, 46, 47, 48, 50, 51, 52, 53

P3.2-4 7, 15, 18, 20, 21, 28, 29, 30, 36, 38, 39, 40, 43,44, 45, 46, 47, 49, 50, 51, 53

P3.2-5 7, 20, 21, 28, 29, 30, 33, 35, 40, 43, 44, 45, 47,48, 50, 51, 52, 53

Prototype

P3.1-1

P3.1-2

P3.1-3

P3.1-4

P3.2-1

P3.2-2

P3.2-3

P3.2-4

P3.2-5

3, 6, 12, 17, 21, 24, 26, 35, 36, 39, 41 2, 3, 21, 22, 23, 30, 31, 33, 35, 41, 42,

44,45

2, 3, 10, 18, 20, 22, 23, 29, 33, 38, 41, 42, 44, 45

2, 10, 11, 12, 17, 18, 20, 21, 22, 23, 24, 26, 33,35, 36, 38, 39, 41, 42, 43, 44, 45, 46

2, 3, 10, 17,18, 20, 22, 23, 29, 33, 36, 39, 41, 42, 44,45

3, 6, 11, 12,18, 19, 21, 23, 24, 26, 35, 36, 39, 41, 42, 43

2, 10, 11, 12, 15, 18, 19, 21, 24, 25, 27, 30, 35,39, 42, 45

2, 6, 10, 11,12, 14, 15, 17, 18, 21, 22, 23, 24, 27,30, 36, 39, 41, 43, 44, 45,46

6, 10, 11, 12,18, 21, 23, 24, 26, 30, 35, 36, 38,39, 42, 44, 45, 46 2, 6, 10, 11, 12, 14, 18, 20, 21, 22, 24,

25, 26, 27,30, 33, 35, 36, 38, 39, 42, 44, 45, 46

ICompare with Fig. 2A.

bCompare with Fig. 2B.

A B

1I

L.

[if IV 1I

FIG. 3. Examples ofglycopeptide patterns obtained for lectin-purified

'l25-labeled

gp9O and gp45 glycopeptides. Foranalysis of these patterns, the presence or absence of vertical groups of glycopeptides, as opposed to individual

glycopeptides

were com-pared, such that possible

microheterogeneity

of the carbohydrate moiety did not interfere with the

comparisons

(17). (A) gp9O glycopeptidemaps.Class I,P3.2-1; class II,P3.2-2; class III, P3.1-3; classIV, P3.2-5(B) gp45glycopeptidemaps. Class I, P3.1-1;class II,P3.1-2. For all maps thedirection of

electrophoresis

wasfromleft toright;thedirection ofchromatographywasfrombottomtotop.

independent

serial transmissions of

EIAV between

ponies

(11, 15).

Thus,

a

total of

at

least 17 distinct structural

variants

were

cataloged in our

laboratory. This large range of

EIAV variation is similar to

that

observed

for

HIV,

in

which

no

two virus

isolates examined have been found to be

identical

in restriction enzyme

mapping, DNA sequencing

studies,

or

both

(1).

The

evolution of EIAV variants

during

a

persistent

infec-tion is

evidently random; i.e., no predictable sequence of

virus variation was observed in the two

experimentally

infected

animals, although

each animal

received

identical

virus inocula. This result differs from the data

reported for

parallel

persistent infections with visna

virus,

in which

similar

patterns

of variant

evolution were observed

(3). This

difference

may reflect that a

larger

spectrum

of variation

is

possible

in EIAV

compared

with visna virus.

TABLE 2. Classification ofEIAVisolates byglycopeptide patternsof theenvelope glycoproteinsgp90andgp45

Viruses in: Glycoprotein

ClassI ClassII ClassIII ClassIV

gp9O Prototype P3.1-1 P3.1-3 P3.2-3

P3.2-1 P3.1-2 P3.2-5

P3.2-4 P3.1-4

gp45 Prototype P3.1-2

P3.1-1 P3.1-4

P3.1-3

P3.2-1 toP3.2-5

on November 10, 2019 by guest

http://jvi.asm.org/

[image:3.612.57.300.279.714.2] [image:3.612.315.555.611.718.2]

(4)

10 a

8

D27

_ 51

0 214 a05

60Q4S 25 *5

24 34

0 Q023 0 73 C

40 1iO Q0

(n

2-300 340@ *5

1979 370

ba

39,

o4o

FIG. 4. Composite oligonucleotidefingerprint generated by

com-parisonsofoligonucleotidemapsfromEIAVisolates. Forpurposes

ofcomparison, viralRNAfromeach virusisolatewas purified and

subjected to mapping procedures on at least two separate

occas-sions, and typically three to four maps were used for making

comparisons between any two isolates. Comparisons between all

isolates were made by directly overlaying autoradiographs. The

composite reflectsonly the high-molecular-weight (lower) region of

theoriginalmaps. Arbitrary numbers were assigned toeach

oligo-nucleotidefor identification. Closed circles indicate oligonucleotides

thatwere commontoallvirus isolates. Open circles indicate variant

oligonucleotides. The directions of first-dimension (a) (8%

poly-acrylamide; pH 3.3) and second-dimension (b) (22%

polyacryl-amide;pH 8.2) electrophoresis are indicated byarrows.

The

variations in EIAV genomic

RNA and glycoproteins

are not

necessarily cumulative. Subsequent

virus strains

may not

arise from the virus strain which predominates

in

the

preceding febrile episode. Rather, it

appears

that

random

mutations

generated during virus replication

result in a

variety of

integrated proviruses,

any oneof which may lead to

the

generation of future predominant virus

strains. In contrast to

EIAV,

sequential isolates of visna display

cumu-lative

changes (3). Sequential

HIV

isolates

like EIAV,

however,

fail

to

display

cumulative

changes.

Noncumulative

variation of HIV has been interpreted

to

indicate

parallel

evolution

of variants in the infected individual

(7). An

alternative

explanation for the results obtained for both

EIAV

and HIV would be

a

high

mutation

rate

which masks

any

direct

lineage between

sequential isolates.

The time

required for the evolution of variant populations

of EIAV

is

variable

but

can

be

remarkably rapid. The

shortest

time

observed between clinical episodes and

dis-tinct

EIAV

isolates

was

approximately 15 days (P3.2-1

to

[image:4.612.78.270.72.217.2]

P3.2-2). In

contrast,

the

generation of variants of visna virus

TABLE 3. Distributionof variantoligonucleotides

Virus Oligonucleotidea

Prototype 1, 4, 6, 13, 15, 16, 19, 32,41, 47 P3.1-1 1, 4, 6, 13, 16, 19, 32,36, 40, 47 P3.1-2 1, 4, 6, 16, 19,32, 36

P3.1-3 1, 4, 6, 13, 16, 19, 32,36, 46, 47 P3.1-4 1, 4,6, 13, 16, 19, 32, 33,36 P3.2-1 (1), 3,(4),6, 13, 16, 19,32, 36, 47 P3.2-2 1, 6, (11),18, 19, 32, (36)

P3.2-3 13, 18, 37,47

P3.2-4 1, 3, 4, 6,13, 16, 19, 32, 36,47 P3.2-5 1, 6, 11, 13, 18,(19), 22, 32, (36), 48

aParenthesesindicatethepresenceofaweaksignalin theposition ofthe

oligonucleotideforthat virus isolate.

usually requires at least a year (12), and the generation of

HIV variants is believed to follow a similar time course (7).

In this regard, EIAV provides a unique model to study

rapidly changing virus structure, as well as the dynamic

interaction between host immune responses and evolving

lentivirus antigens.

This work was supported by the Louisiana Agricultural Experi-ment Station, the Louisiana State University School of Veterinary Medicine, and Public Health Service grant CA-38851 from the National Cancer Institute.

LITERATURE CITED

1. Alizon, M., S. Wain-Hobson, L. Montagnier, and P. Sonigo. 1986. Genetic variability of the AIDS virus: nucleotide sequence analysis of two isolates from African patients. Cell 46:63-74. 2. Barre-Sinoussi, F., J. C. Chermann, F. Rey, M. T. Nugeyre, S.

Chamaret, J. Gruest, C. Dauguet, C. Axler-Blin, F. Brun-Vezinet, C. Rouzioux, W. Rozenbaum, and L. Montagnier. 1983. Isolation of a T-lymphotropic retrovirus from a patient at risk for acquired immune deficiency syndrome (AIDS). Science 220:868-871.

3. Clements, J. E., N. D'Antonio, and0. Narayan. 1982. Genomic changes associated with antigenic variation of visna virus. II. Common nucleotide sequence changes detected in variants from independent isolations. J. Mol. Biol. 158:415-434.

4. Clements, J. E., F. S. Pedersen, 0. Narayan, and W. A. Haseltine. 1980. Genomic changes associated with antigenic variation ofvisna virus during persistent infection. Proc. Natl. Acad. Sci. USA 77:4454 4458.

5. Gallo, R. C., P. S. Sarin, E. P. Gelmann, M. Robert-Guroff, E. Richardson, V. S. Kalyanaraman, D. Mann, G. D. Sidhu, R. E. Stahl, S. Zolla-Pazner, J. Leibowitch, and M. Popovic. 1983. Isolation of human T-cell leukemia virus in acquired immune deficiency syndrome (AIDS). Science220:865-867.

6. Haase,A. T.1986.Pathogenesisof lentivirus infections. Nature (London)322:130-136.

7. Hahn, B., G. Shaw, M. Taylor, R. Redfield, P. Markham, S. Salahuddin, F. Wong-Staal, R. Gallo, E. Parks, and W. Parks. 1986.Genetic variation inHTLV-III/LAVover time in patients withAIDS or risk for AIDS. Science 232:1548-1553.

8. Issel, C. J., and L. Coggins. 1979. Equine infectious anemia: currentknowledge. J. Am.Vet. Med. Assoc. 174:727-733. 9. Kono, Y., K.Kobayashi,and Y.Fukunaga. 1973. Antigenic drift

ofequine infectious anemia virus in chronically infected horses. Arch. GesamteVirusforsch. 41:1-10.

10. Montelaro, R. C., N. Lohrey, B. Parekh, E. W. Blakeney, and C. J. Issel. 1982. Isolation and comparative biochemical prop-erties ofthe major internal polypeptides of equine infectious anemia virus. J. Virol. 42:1029-1038.

11. Montelaro, R. C., B. Parekh, A. Orrego, and C. J. Issel. 1984. Antigenic variation during persistent infection by equine infec-tious anemia virus, a retrovirus. J. Biol. Chem. 259:10539-10544.

12. Narayan, O., J. E. Clements, S. Kennedy-Stoskopf, and W. Royal, III. 1986. Antigenic variation in the lentiviruses that causevisna-maedi insheep andarthritis-encephalitisin goats, p. 25-40. In T. H. Birkbeck and C. W. Penn (ed.), Antigenic variation ininfectious diseases. IRL Press, Washington, D.C. 13. Orrego, A., C. J. Issel, R. C.Montelaro, and W. V. Adams, Jr.

1982. Virulence andin vitro growth of acell-adapted strain of equine infectious anemia virus after serial passage in ponies. Am. J. Vet. Res.43:1556-1560.

14. Parekh, B., C. J. Issel, and R. C. Montelaro. 1980. Equine infectious anemia virus, a putative lentivirus, contains polypeptides analogous to prototype-C oncornaviruses. Virol-ogy107:520-525.

15. Payne, S., B. Parekh, R. C. Montelaro, and C. J. Issel. 1984. Genomic alterations associated with persistent infections by equine infectiousanemiavirus,aretrovirus. 1984. J. Gen. Virol. 65:1395-1399.

16. Popovic, M., M. G. Sarngadharan, E. Read, and R. C. Gallo.

on November 10, 2019 by guest

http://jvi.asm.org/

[image:4.612.60.295.593.705.2]

(5)

1984. Detection, isolation, and continuous production of cytopathic retroviruses (HTLV-IIL) from patients with AIDS andpre-AIDS. Science 224:497-500.

17. Salinovich, O., and R. C. Montelaro. 1986. Comparison of glycoproteins by two-dimensional mapping of glycosylated

pep-tides. Anal. Biochem. 157:19-27.

18. Salinovich, O., S. L. Payne, R. C. Montelaro, K. A. Hussain,

C. J.Issel,and K. L. Schnorr. 1986.Rapidemergenceof novel

antigenic and genetic variants ofequineinfectious anemia virus duringpersistentinfection. J. Virol. 57:71-80.

19. Wong-Staal, F., G. M. Shaw, B. H. Hahn, S. Z. Salahuddin, M. Popovic, P. Markham, R. Redfield, and R. C. Gallo. 1985. Genomic diversity of human T-lymphotropic virus type III (HTLV-III).Science 229:759-762.