Microbiol. Immunol. Vol. 36 (10), 1029-1039, 1992
SeroepidemiologicalSurvey for Antibody to Borreliaburgdorferiin Cows Hiroshi ISOGAI,*,1Emiko ISOGAI,2Toshiyuki MASUZAWA,3 Yasutake YANAGIHARA ,3 Mitsunori MATSUBARA,4 Masamichi SHIMANUKI,4 Toshiyuki SETA,4Kenji FUKAI, 4 Nobumichi KUROSAWA,4 Masafumi ENOKIDANI,4 Toshio NAKAMURA, 4 Motoshi TAJIMA,4Kiyoshi TAKAHASHI,5 Koichi TAKAHASHI, 6 and Nobuhiro FUJII7 1Division of Animal
Experimentation
, Sapporo
Medical College,
Sapporo,
Hokkaido
060, Japan,
2Departmentof Public Health, Higashi Nippon Gakuen University,Hokkaido 061-02, Japan, 3Departmentof Microbiology, Universityof Shizuoka, Shizuoka, Shizuoka 442, Japan, 4KushiroAgricultural Station, Hokkaido 085-12, Japan, 5Departmentof Veterinary Internal Medicine I, Rakuno Gakuen University,Ebetsu, Hokkaido 069, Japan, 6YakumoHealth Center, Hokkaido 049-31, Japan, and 7Departmentof Microbiology, Sapporo Medical College,Sapporo, Hokkaido 060, Japan (Accepted for publication, July 15, 1992)
Abstract Antibody to Borrelia burgdorferiwas examined in 380 healthy and 38 clinical cases of cows from Hokkaido and Shizuoka in Japan. In healthy animals, IgG and IgM antibody to B. burgdorferiHO14 strain were found in 44 cows (14.6%) and 24 cows (8.0%) from Hokkaido. In contrast, antibody-positive case was not observed except for only 1 case which was IgM positive (1/79: 1.3%) in cows from Shizuoka. Mean antibody levels of healthy animals in Hokkaido and Shizuoka were 0.651 and 0.263 (IgG antibody to HO14 strain), 0.642 and 0.169 (IgG to HP3 strain), 0.613 and 0.367 (IgM to HO14 strain) and 0.582 and 0.286 (IgM to HP3 strain). The differences of the antibody levels between cows from Hokkaido and Shizuoka were significant. Seasonal difference was found in seropositive cows from Hokkaido. The rate of seropositive cows was high in summer (23.4% in June and 11.8% in July) but low in winter (0% in January and February). The pattern was discussed to be associated with activation of ticks. One of 4 cows with arthritis showed significantly higher IgG antibody level than that of healthy cows and cows with some disease, although the serum was collected from Shizuoka where antibodypositive animals for B. burgdorferiwere rare among healthy cows. This high IgG antibody may suggest that the arthritis of such cows was caused by infection with B. burgdorferi. Two of 7 cows with unclassified abortion showed positive antibody reaction in Hokkaido. These cases, however, may not be related to the B. burgdorferi infection because the positive rate was similar to those of healthy cows in the same
season.
Lyme disease is an acute and chronic inflammatory syndrome in human beings (17, 18). Borreliaburgdorferi,the etiological agent of Lyme disease, infects a number of vertebrates, including humans and animals (1). The spirochetes, transmitted 102 9
103 0
H.
ISOGAI
ET AL
primarily by ticks of the genus Ixodes,were first isolated from the midgut of deer tick, Ixodes dammini (3). B. burgdorferiwas isolated from 2 species of ixodid ticks, Ixodes ovatus and Ixodes persulcatus, in Hokkaido, Japan (15). Lyme disease has been found in human beings in Hokkaido (12). Additionally, the area was considered to be an endemic area of Lyme disease from the serologic survey in human and animals (7-9). Arthritis and systemic diseases caused by B. burgdorferi infection in domestic animals have been reported (4, 5). The Lyme disease in these animals was similar that in human beings (17, 18). Domestic animals could be suitable hosts for several species of hard-body ticks, and in communities where Lyme disease in human beings is prevalent, they may be exposed to B. burgdorferi. Serological studies have shown that many species of mammals in Ixodes tickinfected area are infected with B. burgdorferi(4, 14). In Japan, it has been reported in only humans (6, 8, 15), dogs (2, 7) and sika deer (9). Our previous examination (7) for the antibody to B. burgdorferi in dogs showed that the animals often have antibody to the organism without clinical signs of Lyme disease. Therefore, even an apparently healthy animal may be exposed to this agent and produce antibody. In Japan, Lyme disease has not been noticed by almost all veterinarians until quite recently. None of the bacteriological or serological examinations for B. burgdorferi have been done when veterinarians meet animal patients for medical treatment. Furthermore, even public organizations for domestic animal health have not surveyed this disease. It is possible that antibody-positive sera for Lyme disease is present in the serum samples collected for examination of various diseases other than this disease. In this study, we report sero-epidemiology of healthy and clinical cases in cows. MATERIALS
AND
METHODS
Two Japanese strains of B. burgdorferi,H014 and HP3, which were isolated from the midgut of I. ovatusand I. persulcatusin Hokkaido, Japan, in 1989 (14), were used for study. Borrelial strains were cultivated in BSK II medium at 32 C for 5 to 7 days (3). Serum specimens were obtained from 418 cows grazing in Hokkaido and Shizuoka Prefectures. Ages and breeds of them were unclear. Sera from healthy cows were collected from 301 cows in Hokkaido and from 79 cows in Shizuoka, Japan. Sera from 38 clinical cases including 4 cases of arthritis which was diagnosed histopathologically, 7 cases of abortion and other diseases were also examined. Four serum samples from cows with arthritis were obtained from Shizuoka Prefecture. Seven sera were obtained from cows after abortion in Abashiri, Hokkaido. These sera were examined for antibodies to several microbial agents causing abortion but the cause of it was not determined. The other twenty-seven samples were obtained from cows showing various disorders including myocarditis, fracture, pneumonia and others. All of the samples were screened for antibody to Leptospiraby microscopic agglu-
ANTIBODY
tination
test.
and
L.
cow
was
None
autumnalis
Immune
at
serum
and
109).
after
IgG
HP3
were
HP3
strain
of
also
cow
at
5
days
by
Russell
washing
added
was
washed.
Ltd., at
follows. sample
were
final
for
0.447
levels
immunization
were
obtained
strain.
sample
from of
Fetal
sera
were
to
respectively.
standardization
bacterial
at
and 0.393 and against strain
0.731,
Antisera
used
IgM and
antibody
and
after
to
each ELISA
calf
sera
stored
by
and
Ace
(15
at -80
plates
C
each
(pH by
50
PBS
sera
were highest
After
phosphatebuffer
time. After kit (Bio-Rad
for
1-N
492
with
10%
and assessed values
37
for
to
and were
picked
well in
to
well.
each
0.05%
coloriCo., 5 times
Tween-20. determined
from
M
well.
The
repeated
was 0.1
Electric
values up
IgM
Hokkaido,
each
added
were
absorbance
were
or
dissolved
was and
wells
C
was
Birmingham,
Corona
reaction
37
PBS
1: 500,
Co.,
each
Washing Ace
C,
1 hr,
was
(MTP-22,
Block
37
in
4 C.
immunoglobulin
mg/ml)
H2SO4
negative
at
Limited,
C
at
100
sheep
H2O2
nm,.
1:
1 hr
Site
at
reader
Japan)
(Yukijirushi
(4
of
at
Hokkaido,
diluted
Binding Ace
overnight
incubation
diluted
England)
Block
pl
Co.,
(100 ƒÊl,
0.02%
at
Next,
IgG
incubation
absorbance
positive
in
were
carbonate
extracts
sample
The
ELISA
by
by
medium
for 30 sec each Protein assay
incubated
10%
an
between
in
reaction.
serum
containing
adding by
the
3 times
(Yukijirushi
o-phenylenediamine
5.0)
process
collected.
of
After
of
in
washed
bacterial
Birmingham,
well. pl
sample
the
were
with
method
10 ,ƒÊg/ml.
fraction, PBS
the
essentially
suspended
three times the Bio-Rad
anti-cow
in
and
by
performed
grown
was
non-specific
Limited,
with
min
solution
100 ƒÊl
the
stopped
all
pellet
at
determined
Japan)
20
final ice by
performed
was strains
g for
coated Block
was
bacterial
on
blocking
hundred
was
(ELISA) preparation
adjusted
for
1: 250
values First,
inoculation,
0.223
IgG
0.308
days
All
was
Site
experimental
Cut-off
were
strain.
each
were
After
was
Ibaraki,
each
The
25%
buffer
value
10
The
immunization.
and
and
were
The
10,000 •~
washed
added
reaction
at
antigen
(14).
immunoglobulin
One
metric
cow
assay
7.2).
it
Binding
citrate-phosphate The
0.282,
controls.
The
at
well
diluted
Japan)
strain
cow.
bacterial
final
IgM
and
for
before
a
phenol-inactivated
the 1st immunization, respectively. Antisera
bacterial
serum
al
pH
well.
sheep
England)
each
peroxidase-conjugated
The
(100 ƒÊl,
et
were
each
fraction,
icterohaemorrhagiae
from
with
after
HO14
after
inoculation
8).
well,
each
and
L.
prepared
cow
days
manner.
each
and then sonicated protein concentration
each
wells
in
washed
(7,
U.S.A.),
to
0.360
negative
microplates
1 hr,
similar
immune
9.6) of
added
for
the
0.255,
first
for
(PBS,
ELISA
to
intervals
the
10
of
at 5 days immunization,
from
centrifugation
Laboratories,
antigen
for
the
at
was
5-day
from
immunosorbent
saline
(0.05 M, pH determination
1031
performed.
of
buffered
was
by and
described
pelleted
COWS
reaction
HO14 at
and
0.623 final
control
used
was
method
strain
collected
the
obtained
Enzyme-linked
After
and the
after
were
previously the
to
IgM-positive
assay
was
them
antisera IgG-positive
as
IN
positive
5 times
inoculation
0.161
as
samples)
showed
bur gdorferi
prepared
were
used
until
first
0.428 days after
Therefore,
assay
B.
Serum
the
levels
preinoculation, 0.821 at 10
to
samples
intracutaneously
(1 •~
5 days
the
B. BURGDORFERI
.
injected
bacteria
of
TO
values
as
from
each of
assay
103 2
H.
groups
for
70%
of
Positive
Ig
class
highest
absorbances
against
HO The
were
each
these
each in
0.687
for
obtained
analyzed
bacterial each
were •„0.687
14, •„ data
and values
by
from
against each
ET AL
strain. group
for
IgG
unpaired
ISOGAI
IgG HP3
experiment
An
was
absorbance
determined
against
value as
HO14
strain,
and •„0.700
for
were
collected
IgM
positive 0.698 against
and
the
more
than
reaction. for
IgM
HP3. differences
t-test.
RESULTS
Standardizationof ELISA Assay Preliminary assays were conducted to use immune sera having a reactivity in the IgG and IgM classes of antibody with B. burgdorferi. The sera were aliquoted for subsequent use in serial dilutions with each further assay as a positive sera. A titration curve for determination of the antibodies reactive with the 2 Japanese strains of B. burgdorferiwas thus obtained for each class of immunoglobulin by plotting
Fig.
1. using
Standard the
curves
immune
for
sera.
determination
of class-specific
antibody
to B. burgdorferi
ANTIBODY
TO
B. BURGDORFERI
IN
COWS
103 3
the mean absorbances obtained from each positive immune serum (Fig. 1). Results were obtained as absorbance values at 492 nm and corrected by relating the individual sera to standard curve obtained with the reference positive serum. No antibodies were detected in fetal calf sera (negative controls) and the level was less than 0.050. Assay was repeated 3 times for each Ig class and bacterial strain. The standard deviations of them were within 0.02 for each assay. SerologicalFindings in Hokkaidoand Shizuoka Serum samples obtained from 380 healthy cows, including 301 cows from Hokkaido and 79 cows from Shizuoka, were examined. Prevalence of specimens seropositive for the Japanese strains of B. burgdorferiis shown in Table 1. IgG and IgM anitbodies to B. burgdorferi H014 was detected in 14.6% and 8.0%, respectively, in Hokkaido. In contrast, IgG-positive cases were not observed in Shizuoka, and only one case was IgM positive (1.3%). Antibodies against strain HP3 were detected in the sera from Hokkaido, but not in those from Shizuoka. The mean levels of antibody to the bacteria detected in the sera from cows are listed in Table 2. Mean level of IgG or IgM antibody were distinctly distributed between the 2 areas. SeasonalDifferencein Hokkaido The seasonal incidence of B. burgdorferi antibody-positive cows is shown in
Table 1.
Table 2.
Prevalence of serum antibody to B. burgdorferi in healthy cows
Comparison of antibody level to B. burgdorferibetween Hokkaido and Shizuoka in Japan
103 4
H.
ISOGAI
ET AL
ANTIBODY
Table 4.
TO
B. BURGDORFERI
IN COWS
103 5
IgG antibody level to B. burgdorferi HO14 in cows with arthritis and other diseases in Shizuoka
Table 5.
IgG antibody level to B. burgdorferiHO14 in the cows with abortion in Hokkaido
Table 3. The seropositive percentage was high in summer (June and July) but low in winter (January and February). The mean antibody level increased gradually from spring to summer (Table 3). Furthermore, the deviation of the level in summer was larger than that in spring. Relationto Clinical Signs One case in 4 cows with arthritis (1/4, 25.0%) and a case with myocarditis in group of other diseases showed high IgG antibody value (0.626 and 0.624) although these sera were obtained from the animals in Shizuoka where almost all of healthy cows examined were seronegative to B. burgdorferi(Table 4). Two cows with unclassified abortion (2/7: 28.6%) showed positive antibody reaction in Hokkaido (Table 5). However, the positive rate was similar to those of healthy cows in the same season.
103 6
H.
ISOGAI
ET AL
DISCUSSION
The cows that appeared to be healthy and had no history of limb/joints disorder had antibodies to B. burgdorferi. This finding was consistent with earlier reports in dogs (7) and human beings (15) in Japan. Because potentially infective Ixodes ticks feed from spring through summer, the cows must have been exposed repeatedly to many infected ticks. In Japan, a patient with Lyme disease was firstly reported by Kawabata et al in 1987 (10), and B. burgdorferiwas isolated from 2 species of ixodid ticks, I. ovatus and I. persulcatusin 1990 (15). The ticks are now found abundantly in Hokkaido and various regions in Japan (13). Infection rate of these ixodid ticks with Borreliae in Hokkaido is presumed to be 10-20% (unpublished data). Recently, cases of human infection with B. burgdorferihave been reported in Hokkaido. Because of the present sero-epidemiological results and other previous data, it can be stated that Hokkaido is a highly endemic area for Lyme disease. Furthermore, in comparison of sero-epidemiology between Hokkaido and Shizuoka, it is indicated that the antibody-positive rate varies among areas in Japan. There is seasonal distribution of antibody-positive cases in cows. Similar antibody pattern has been found in Sika deer (9). In the deer, the pattern was observed not only in a mass but also in individual animals. This pattern suggests that adult ticks are most likely transmitted to cows. The ticks emerge in Hokkaido during the season from spring to summer. Percentages of antibody-positive cows were 14.6 and 9.0% in IgG antibody and 8.0 and 6.3% in IgM antibody to H014 and HP3 strains, respectively (Table 1). The degrees of IgG antibody were lower than those obtained from dogs (7) in 1990. In contrast, IgM antibodies were higher than those of dogs. Precise reasons for them could not be explained at this stage. There are many factors which affect these degrees in cows and dogs, such as area, age, breed and others. Positive percentage, for instance, was 33.3% in dogs more than 5 years old. The degree was higher than that of cows. This problem requires closer analysis. The most frequent clinical signs in antibody-positive cows have been reported to be lameness and swollen joints, but many also had stiffness, laminitis, abortion and fevers (5). In our cases, a cow with arthritis (1/4 : 25.0%) showed high IgG antibody level although almost all of healthy cows in same area showed low antibody levels. Not all antibody-positive animals showed clinical signs. These findings suggest that B. burgdorferiinfection in cows can cause clinical illness in some but not all animals. The number of patients diagnosed as having Lyme disease was not many in Japan but also severe clinical signs have not been reported. Furthermore, none of the animals have been clearly diagnosed as having Lyme disease. Some possibilities may be indicated to account for these facts. First is that bacterial pathogens are different among bacteria prevailing in Japan and Europe and America. Indeed, immunoblot analysis has indicated antigenic differences between bacterial strains isolated from ixodid ticks in Japan and those in the United States (11). Such anti-
ANTIBODY
TO
B. BURGDORFERI
IN
COWS
103 7
genic differences can affect the bacterial pathogenicity and clinical signs for Lyme disease. Next is the differences of host reactivity to B. burgdorferiamong human races and animal species. It may be associated with differences of host defense mechanisms including the differences among individual host. The third possibility is that clinical signs have failed to be noticed in Japan. In other words, clinical signs caused by Lyme disease were possibly diagnosed as those of another disease because the disease was not well known in Japan. Clinical doctors and veterinarians should become familiar with the manifestation of Lyme disease in Japan. Isolation of B. burgdorferi has been unsuccessful from these domestic animals in Japan. The isolation will provide important information for the etiology and pathogenesis of Lyme disease in various animals and, furthermore, for determination of the reservoir agent of B. burgdorferi. We are now trying to isolate B. burgdorferi from various domestic and wild animals. It has been reported that transplacental transmission of B. burgdorferiwas demonstrated (5, 16). B. burgdorferiwas cultured from the blood of a newborn calf, and an aborted calf had antibodies to B. burgdorferi,indicating in utero infection (5). Our cases with abortion showed positive antibody reaction but the positive rate (28.6%) was similar to those of healthy cows in the same season. The antibody levels of the cases were also similar to those of healthy ones. Therefore, the clinical cases could not be determined as having been caused by infection with B. burgdorferi. ELISA for detection of antibodies specific to B. burgdorferihas been considered one of the effective diagnostic tools for Lyme disease (7, 8, 14). In ELISA assay, careful attention must be paid to a cross-reactivity to other spirochetes. The serum samples used in the present study were negative to 2 serovars of Leptospira. Furthermore, leptospirosis and syphilis in domestic cows have been not reported in Hokkaido. The only rare case has been that Leptospirahajor was isolated from an imported cow. Therefore, cross-reactivity, at least, to Leptospira species and Treponemapallidum was not a diagnostic problem in our cases. It has been known that Treponema species are isolated from oral cavity of human and some animal species. These bacteria have a possibility to show cross-reactivity to B. burgdorferi. However, oral Treponemapresent in the cavity of cows has been unclear. Further studies are needed in this field. IgM antibody to B. burgdorferiwas low in both rate and levels in cows. Similar results have been observed in dogs (7). It was possible that levels of IgG antibodies to B. burgdorferi could continue for a long time with some rise and fall, but IgM antibodies could increase temporarily in animals. The other possibility is that IgM antibody may be difficult to increase against this organism. Antigen which was associated with increase in IgM antibody may be hardly recognized by defense mechanisms of certain hosts. Low levels of IgM antisera to both strains could suggest this mechanism when inactivated bacteria were used. However, it was not sure because there was only scant information about this mechanism. We suggested that the Japanese strains coated on ELISA plates are more suitable for serodiagnosis of Lyme disease as an antigen source in Japan than the American. or European strains, because the antigenic properties of Japanese strains were dif-
103 8
H.
ISOGAI
ET AL
ferent from that of the strains in the United States and Europe (9). There are 2 major species of Ixodesticks which are considered as vectors of B. burgdorferiin Japan, such as Ixodes ovatusand I. persulcatus. Two strains of B. burgdorefriused in this study were isolated from these ticks. Furthermore, some characteristics of them have been already reported in several papers (11, 15). Because of the reasons described above, HO14 and HP3 strains of B. burgdorfericould be useful for one of the reference strains in Japan. The
authors
thank
Dr.
T. Hirai,
Abashiri
Livestock
Hygiene
Service
Center,
for kind
support.
REFERENCES
1) Anderson, J.F., Johnson, R.C., Magnareli, L.A., and Hyde, F.W. 1984. Avian and mammalian hosts for spirochete-infected ticks and insects in a Lyme disease focus in Connecticut. Yale J. Biol. Med. 57 : 627-641. 2) Arashima, Y. 1991. Anti-Borrelia burgdorferi antibody in dogs: lyme disease as zoonosis. Jpn. J. Clin. Pathol. 39 : 869-874 (in Japanese with English summary). 3) Burgdorfer, W., Barbour, A.G., Hayes, S.F., Peter, O., and Aeschlemann, A. 1982. Lyme disease -atick-borne
4) 5)
6)
7) 8)
9)
10) 11)
12)
13)
14)
15)
spirochetosis ?
Science
216:1317-1319
.
Burgess, E.C., Gendron-Fitzpatrick, A., and Wright, W.O. 1987. Arthritis and systemic disease caused by Borrelia burgdorferi infection in a cow. Am. J. Vet. Med. Assoc. 191: 1468-1470. Burgess, E.C. 1988. Borrelia burgdorferi infection in Wisconsin horses and cows, p. 235-243. In Annals of the New York Academy of Science, Lyme disease and related disorders, Vol. 539, The New York Academy of Sciences, New York. Iguchi, K., Yagura, S., Maruta, H., Kawabata, M., and Kawano, K. 1989. Studies on the relationship between tick-bite and Lyme disease in members of JSDF with special reference to the serological diagnosis of Lyme disease. Natl. Def. Med. J. 36: 225-230 (in Japanese with English summary). Isogai, E., Isogai, H., Sato, N., Yazawa, M., and Kawakami, M. 1990. Antibodies to Borrelia burgdorferi in dogs in Hokkaido. Microbiol. Immunol. 34: 1005-1012. Isogai, E., Isogai, H., Kotake, S., Yoshikawa, K., Ichiishi, A., Kosaka, S., Sato, N., Hayashi, S., Oguma, K., and Ohno, S. 1991. Detection of antibodies against Borrelia burgdorferi in patients with uveitis. Am. J. Ophthalmol. 112: 23-30. Isogai, E., Isogai, H., Masuzawa, T., Yanagihara, Y., Sato, N., Hayashi, S., Maki, T., and Mori, M. 1991. Serological survey for Lyme disease in Sika deer (Germs nipponyesoensis) by enzymelinked immunosorbent assay (ELISA). Microbiol. Immunol. 35: 695-703. Kawabata, M., Baba, S., Iguchi, K., Yamaguchi, N., and Russell, H. 1987. Lyme disease in Japan and its possible incriminated tick vector, Ixodespersulcatus.J. Infect. Dis. 156: 854. Masuzawa, T., Okada, Y., Yanagihara, Y., and Sato, N. 1991. Antigenic properties of Borrelia burgdorferiisolated from Ixodes ovatusand Ixodespersulcatusin Hokkaido, Japan. J. Clin. Microbiol. 29: 1568-1573. Miyamoto, K., Takahashi, K., Sato, N., Uraguchi, K., Matsuo, S., Iizuka, H., Mori, M., Tsuboi, Y., and Ohtsuka, K. 1990. Cases of erythema and Lyme disease associated with tick bite in Hokkaido, Japan. Jpn. J. Sanit. Zool. 41: 63-65. Namba, N. 1963. Research on the distribution of the ticks and their habitat in pasture of Hokkaido. Res. Bul. Hokkaido Natl. Agr. Exp. Stat. 80: 103-114 (in Japanese with English summary). Russell, H., Sampson, J.S., Schmid, G.P., Wilkinson, H.W., and Pikaytis, B. 1984. Enzymelinked immunosorbent assay and indirect immunofluorescence assay for Lyme disease. J. Infect. Dis. 149: 465-470. Sato, N., Ito, T., Kameyama, K., Takahashi, K., and Honma, H. 1990. Investigations of ticks
ANTIBODY
16)
17)
18)
TO B. BURGDORFERI
IN COWS
103 9
for the presence of Borrelia and the prevalence of positive antibodies to Borrelia burgdorferi among residents in Hokkaido, Japan. Proc. Int. Congr. Infect. Dis. 489: 166. Schlesinger, P.A., Duray, P.H., Burke, B.A., Steere, A.C., and Stillman, M.T. 1985. Maternal fetal transmission of the Lyme disease spirochete, Borrelia burgdorferi. Ann. Intern. Med. 103: 67-68. Steere, A.C., Malawista, S.E., Snydman, D.R., Shope, R.E., Andiman, W.A., Ross, M.R., and Steel, F.M. 1977. Lyme arthritis; an endemic of oligoarticular arthritis in children and adults in three Connecticut communities. Arthritis Rheum. 20: 7-17. Steere, A.C., Taylor, E., Wilson, M.L., Levine, J.F., and Spielman, A. 1986. Longitudinal assessment of the clinical and epidemiological features of Lyme disease in a defined population. J. Infect. Dis. 154: 295-300. (Received for publication, February 24, 1992; in revised form, July 7, 1992)