E Ar~:~hiv~ of
Arch. Environ. Contain. Toxicol. 21, 126-134 (1991)
c
nvironmental ontamination
9 1991 Springer-Verlag New York Inc.
Toxicity of Mercury, Copper, Nickel, Lead, and Cobalt to Embryos and Larvae of Zebrafish, Brachydanio rerio G 6 r a n D a v e .1 a n d R u i q i n Xiu** * Department of Zoophysiology, University of G6teborg, Box 25059, 400 31 G6teborg, Sweden and ** Department of Microbiology and Aquabiology, Institute of Environmental Health and Engineering, Chinese Academy of Preventive Medicine, 29 Nan Wei Road, 100 50 Beijing, China
Abstract. The toxicity of mercury (HgCI2), copper (CuCI2:5 H20), nickel (NiSO4:6 H20), lead (Pb(CH3COO)2:3 H20) and cobalt (COC12:6 H20) was studied under standardized conditions in embryos and larvae of the zebrafish, Brachydanio rerio. Exposures were started at the blastula stage (2-4 h after spawning) and the effects on hatching and survival were monitored daily for 16 days. Copper and nickel were more specific inhibitors of hatching than cobalt, lead, and mercury. Nominal " n o effect" concentrations determined from the dose-response relationships (ZEPs, Zero Equivalent Points) for effect on hatching time were 0.05 p,g Cu/L, 10 Ixg Hg/L, 20 txg Pb/L, 40 Ixg Ni/L and 3,840 Ixg Co/L, and those for effect on survival time were 0,25 ~g Cu/L, 1.2 rzg Hg/L, 30 ~g Pb/L, 80 txg Ni/L, and 60 fxg Co/L. The " n o effect" concentrations for Ni, Hg and Pb are consistent with previously reported MATC values for sensitive species of fish. The " n o effect" concentrations for copper are 1-2 orders of magnitude lower than previously reported values. The major reason for the latter discrepancy was considered to be the absence of organics that can complex copper ions in the reconstituted water that we used, which had a hardness of 100 mg/L (as CaCO3) and a pH of 7.5-7.7. Unexposed controls were started with embryos from different parental zebrafishes and the parental-caused variability in early embryo mortality, median hatching time and median survival time were estimated.
Embryo-larval (EL) toxicity tests are generally more sensitive than toxicity tests with juvenile and adult fish. F o r many pollutants, E L toxicity tests are as sensitive or almost as sensitive as chronic (life cycle) toxicity tests (Macek and Sleight 1977; McKim 1977). Hatchability, growth, and survival are the most frequently used response variables in E L toxicity tests. Survival and hatching are discrete (all or none) responses that can be evaluated by ordinary dose-response procedures, eg log-probit analysis. Growth, being a suble-
Address correspondence to: GOran Dave, Department of Zoophysiology, Box 250 59, 400 31 G6teborg, Sweden.
thal response, is often believed to be more sensitive than survival. However, among 173 tests, reviewed by Woltering (1984), larval survival was affected at LOEC (Lowest Observed Effect Concentration) in 57% of these tests, growth in only 36% and hatching in 11%. In a later review of 136 tests, Kristensen (1990) concluded that growth was the most sensitive variable in 70% of the tests when the juvenile stage was included. Furthermore, the embryo stage was found to contribute very little to the sensitivity of the E L toxicity test. However, if hatching time and not only percentage hatch is considered, then the embryo stage is sometimes more sensitive than the larval stage (Dave 1986; Dave et al. 1987). In the p r e s e n t E L t o x i c i t y t e s t s with the z e b r a f i s h (Brachydanio rerio), only hatching and survival were used as response variables. Since the larvae were not fed, it was assumed that toxicant-induced stress responses would ultimately show up as a reduced survival time. Hatching time can be affected secondarily as a result of primary effects on developmental rate of the embryo or early larval stages (Rosenthal and Alderdice 1976). Or toxicants may interfere with the hatching process directly. Hatching often results from a combined effort of the hatching enzyme (chorionase), increased perivitelline pressure, muscular contractions or active uptake of water by the embryo (Denuc6 1985). By these means environmental factors, including toxicants, may affect hatching time in two directions (stimulation and inhibition). The aim of the present study was to investigate the effects of five heavy metals (mercury, copper, nickel, lead and cobalt) on embryos and larvae of zebrafish under standardized conditions (SIS, 1988). Furthermore, we wanted to investigate possible parental influence on hatching and survival time in unexposed fish (controls).
Materials and methods Animals The procedures for production of zebrafish eggs and exposure of eggs and larvae have been described previously (SIS 1988; Dave et al. 1987). An English translation of the standard protocol (SIS 1988)
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
is available upon request from one of the authors (G. Dave). Briefly, the test is started by addition of 30 eggs in the blastula stage (2-4 h after spawning) to a geometric series of concentrations (dilution factor 0.5). Test solutions (50 ml in a Petri dish) are renewed daily. After 24 h, the number of eggs is reduced to 20 in order to compensate for control mortality and to provide equal numbers (N) for effect variable estimates. Since the embryos (eggs) can't swim, this culling is random. No food is provided and the test is terminated when at least 90% of the larvae have died in the control. The larvae are not fed because of lack of a suitable synthetic diet. Daily records of the number of live eggs (embryos) and larvae are used to determine the median time for hatch and survival (MHT and MST).
Exposure to Mercury Eggs from two pairs of spawners (one male and one female in each pair) were distributed randomly (five by five) in dishes containing nominal concentrations of 1-512 txg Hg/L (dilution factor 0.5). Each concentration was duplicated and each concentration series contained a duplicated control (4 controls altogether). The source of mercury was mercuric chloride (HgClz), pro analysi 99.5%, Merck, lot no 8556740. All concentrations are expressed as ~xg total Hg/L and are based on nominal values.
Exposure to Copper, Nickel, Lead, and Cobalt Eggs of a defined age (2-4 h after spawning) were obtained from 5 spawning pairs (1 male plus 1 female). Eggs from two pairs were used for the copper exposure unit, and eggs from the remaining three pairs were used for the nickel, lead and cobalt exposure units. Each exposure unit consisted of 11 concentrations (dilution factor 0.5) and two controls. All exposures were started on the same day with 2-4 h old eggs (blastula stage). Thirty eggs were added, five by five, to each concentration (50 ml test solution in a Petri dish with i.d. 10 cm). After 24 h, the number of living eggs (transparent) was reduced to 20 in the copper exposure unit. In the other exposure units (Ni, Pb, Co) the number was reduced to 10 because of higher initial mortality. The toxicant sources were CuSO4:5 H20 (Fisher reagent grade), N i S O 4 : 6 H20 (Merck analytical grade), Pb(CH3COO)2:3 H20 (Merck analytical grade) and CoC12:6 H20 (Baker analytical grade). Test concentrations given are expressed as ~g/L of the metal only.
Measurements and Ambient Conditions Test solutions were renewed every 24 h. Living embryos (transparent) and larve (motile) were recorded on these occasions, and more frequently around hatching time. The dishes were incubated at 26~ at a photoperiod fo 12 h light and 12 h darkness controlled by an incandescent tube and a timer. The dilution water was prepared from Milli Q Reagent Water (Millipore) and analytical wade CaC12: 2 HzO (117.6 rag/L), MgSO4:7 H20 (49.3 mg/L), NaHCO 3 (25.9 rag/L) and KCI (2.3 rag/L). This water has a hardness of 100 mg/L expressed a s C a C O 3. After aeration, the pH was adjusted, if necessary, to 7.5. Daily recordings of temperature, dissolved oxygen and pH in new and old test solutions ranged from 25.4 to 26.4~ 7.5-7.7, and 88-100% of air saturation, respectively.
Statistical Treatment Median times for hatching (MHT) and survival (MST) were calculated by probit analysis (Peltier and Weber 1985). The Lowest Observed Effect Concentrations (LOECs) for effects on hatch and survival were estimated by comparing MHTs and MSTs. If 95% con-
127
fidence limits for controls and exposed animals were sepmated an effect was assumed. The next lower concentration was defined as the No Observed Effect Concentration (NOEC). The geometric mean of LOEC and NOEC was used as one estimate of the "no effect" concentration~ Another estimate of the "no effect" concentration was obtained by inspection of the log (MHT) = f (tog C) and the log (MST) = f(log C) relationships. From these dose-response relationships the Zero Equivalent Point (ZEP) was determined graphically as shown in Figures 1-5.
Results Early embryo mortality (after 24 h) and the median times tbr hatch and survival with each metal (Hg, Cu, Ni, Pb, Co) are shown in Figures 1-5. Control values are summarized in Table 1.
Sources of Variability in Controls The influence of individual embryo variability was counteracted by randomization to treatment (= Petri dish). In order to save time and equalize the time for start of exposure, eggs were distributed 5 by 5 to each dish until they contained 30 eggs. This randomization was appropriate because initial mortality among replicates was not biased. As expected, the variability between replicated controls (from the Hg experiment) was less than between simultaneously treated experimental units (from the Cu, Ni, Pb, and Co experiments), which were started with eggs from different parental zebrafishes. A comparison with variability within and between laboratories (from Dave et al. 1987) is atso shown in Table 1. As expected, variability (CV) associated with replication is considerably less than the variability associated with repetition, and part of the latter is explained by parental variability. The higher variability for M H T compared to MST is consistent with previously estimated variability within and between laboratories (Dave et al. 1987).
Mercury Early embryo mortality (within 24 h), MH T (Median Hatching Time) and MST (Median Survival time) in various concentrations of mercury are give in Figure 1. The results from the two replicates (A and B) are presented separately because determinations of " n o - e f f e c t " concentrations were made separately in order to evaluate replicability. The L O E C (Lowest Observed Effect Concentration) for effects on hatching time was 16 Ixg Hg/L. In 32 wg/L hatching was completely inhibited, but in lower concentrations of Hg there was a slight stimulation of hatching. This stimulation was most pronounced in 4 Ixg Hg/L. Median survival time in this concentration was only 11 days, which is 2.5 days less than the control value (13.5 d). The dose response relationship for effects on survival (Figure l) can be separated into four phases. Early embryos were killed in concentrations ~>5t2 ~g/L, late embryos between 512 and 32 ~g/L, early larval stages between 32 and 8 Ixg/L, and late larval stages between 8 and 1 txg/L. These four phases were seen during continuous exposure, and are the result from both prolonged exposure and shift in developmental stage. Furthermore, in the transition from early to late larval stages (after 10 to 11 days) there is atso an effect of starvation, since the larvae were not fed (discussed below).
128
G. Dave and R. Xiu
16 14 12
REPLICATE A ZEP 1.18JJg/L _ _ _ ..,_~_ _ .................
16 14
cont r_o/MST _
_
%" ~
,; ..... +_z___/_
............
I0
12 to
-~
o
8
/'ZEP O.05.ug/L
\ -L. control MHT
. ." .~.'./ ........ .. .. ......... .. .. ......... .. .. .... . . .
+ m ~.~ .~
4 IOO
o
oli
i
16
~b so mE
2
0
I00
~b ~E
I
I0
IO0
n n n/1, I] I1 n,n n n i
o
~o
ioo
m
,fl fl fl n, n n n,~
0
0.1 I I0 Nominal conc.of copper, ug / L
I000
so
0
0
iooo
ido
HI 16o
Fig. 2. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to copper under standardized conditions
Nominal conc.of mercury, ~g/L
A
16 14
REPLICATE B ZEP I /5 ug/L .... ~................... co_ntr_d_MS_T___
to
ZEP 80 u g / L , / control MST
/
14 12
12
10
I0
Z
~_~
g
---,--,--,--,--,--,--, ---
-
"8~
8
8.-c
6
-
9
g~
.~=
4
l
i
_
I00
i,- .................. ~ ;
,
,
0
control MHT
---s
....
2
I
I0
IO0
o
,
I00
~
rln o
B
.
I000
~b 50 a~:
[1, ~ ~ I
4
I0 I00 Nominal conc. of mercury, ug/L
.
.
+_!_!
0
i
o
I
6
16o
iobo
10
I00
1000
so
IOC)O
Fig. 1. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to mercury (Hg z+) in replicated treatments (A and B) under standardized conditions Note: ZEPs, Zero Equivalent Points, for effects on hatching and survival were estimated graphically at intersections between the fitted (by the eye) dose-response relationship and the median times for hatch and survival in the controls, respectively. The interpretation of ZEP is that it is the highest concentration where the measured response would be the same as in the control
.
Nominol conc.of nickel, ug/L Fig. 3. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to nickel under standardized conditions
Copper The highest c o n c e n t r a t i o n (128 I~g Cu/L) killed all e m b r y o s within 24 h (Figure 2), and less than 50% hatched in concentrations d o w n to 1 ixg Cu/L. F u r t h e r m o r e , hatching was sig-
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
ZEP 6 0
16
i,/"
i4
~z
.....
8 6
o
12
o;ro,
*+
,ug/L
. . . . . 4__,__ _/_. . . . . . . . . . . . .
_c2'L~__MS_T__
/
I0 84
~
/_
+ .... .--.--.--.--.--.--.-...
12
Z
129
8~,
8
s~:
6
m~
~
g~= ~_~
4
4
,; .... :__:__hi__
_oo:,:o,
I
ZEP 5 8 4 0 ,ug/l
2iO0
$-t
o
o
1
0
,b
I
,do
]() IO0 Nominal conc of leod, Jag/L
,obo
o
IO0
rE s~ >,u
50
o
O
IO00
Fig. 4. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to lead under standardized conditions
i
I
~do
,o'oo
I0 lO0 Nominel conc of cobcllt,j..ig/L
,o&o
I000
Fig. 5. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio re~'/o) embryos and larvae exposed to cobalt under standardized conditions
Cobalt nificantly delayed in all of the tested concentrations, while survival was not affected in the lowest concentration (0.125 p.g Cu/L). Thus, hatching was more sensitive to copper than was larval survival. This, unexpected, high sensitivity of hatching to copper was confirmed in another (unpublished) experiment were the LOEC for effects of hatching was 0.63 txg Cu/L and the NOEC was 0.063 Ixg Cu/L (dilution factor 0.1). The extrapolated " n o effect" concentration (ZEP) for hatching in this experiment was as low as 0.05 txg Cu/L (Figure 1).
The effects of cobalt were qualitatively different from those seen with mercury, copper, nickel, and lead. Hatching was at most minimally affected even at the highest concentrations. The ZEP for MHT shown in Figure 5 (3,840 txg Co/L) is only vaguely supported by our results at the two highest concentrations. Survival was not affected until after 9 days, but after that time survival was affected from 15,360 ~g/L down to 60 txg/L (ZEP for MST in Figure 5).
Discussion Nickel Early embryos were not killed even at the highest concentration (l,024 t~g Ni/L), but hatching was delayed down to 40 Ixg/L (ZEP for MHT in Figure 3). The "no effect" concentration for survival was estimated to he 80 Ixg/L (ZEP for MST in Figure 3).
Lead High concentrations of lead (480 and 960 Ixg Pb/L) inhibited hatching nonspecifically by killing the embryos. Lower concentrations delayed hatching slightly, but no clear doseresponse relationship was found for this delay (Figure 4). The " n o effect" concentrations for effects on hatching and survival were estimated to be 20 Ixg/L and 30 t~g/L, respectively (ZEPs for MHT and MST in Figure 4).
RepIicability, Repeatability, and Reproducibility The variability between replicates was low (high repiicability). The coefficient for variation (CV; SD x 100/2) for the logarithms MSTs for the 4 controls shown in Table 1 is 0.4%. Corresponding CVs for repeated tests at five laboratories were 1.2% or 1.5% (repeatability). The reproducibility for these five laboratories' controls (CVs for control log MST in tests with zinc sulphate and potassium bichromate) was 2.0% and 3.3%. Values for repeatability and reproducibility were given by Dave et al. (1987). These comparative figures for replicability (0.4%), repeatability (1.2-1.5%) and reproducibility (2.0-3.3%), suggest that replicated treatments are not needed in this test. This is because the variability associated with replication is small compared to that associated with repeating the experiment or reproducing it by someone else. However, a replicated control is advisable as an internal control for variability and repeatability at the laboratory,
130
G. Dave and R. Xiu
Table 1. Variability in Early Embryo Mortality (EEM), Median Hatching Time (MHT) and Median Survival Time (MST) in controls due to replication, parental variation, repetition at the same laboratory (repeatability) and at another laboratory (reproducibility) Variable (unit)
Source of variation
Mean value
SD
CV (%)
N
Data from
EEM (dead/total)
Replication Parental Repetition c Laboratoryc Replication Parental Repetition
7/30 13.6/30 16.5% 16.5% 0.465 0.506 0.453 0.411 0.453 0.411 1.128 1.127 1.126 1.137 1.126 1.137
2.6/30 5.9/30 12.1% 12.1% 0.036 0.090 0.045 0.069 0.051 0.069 0.005 0.019 0.014 0.018 0.023 0.037
37 43 73 73 8 18 10 17 11 17 0.4 1.7 1.2 1.5 2.0 3.3
4 4 37 37 4 4 19 22 19 22 4 4 19 22 19 22
Fig. Ia Figs. 2-5b Ref.c Ref# Fig. 1a Figs. 2-5b Table Id Table 2~ Table 1d Table 2~ Fig. 1 Figs. 2-5 Table 1a Table 2~ Table 1d Table 2e
Log MHT (days)
Laboratory Log MST (days)
Replication Parental Repetition Laboratory
Replicates shown in Figure 1 of this study b Control values shown in Figures 2-5 of this study c Results taken from Dave et al. (1987) for total variation found in all tests at all laboratories d Results by Dave et al. (1987) from ring test with K2Cr207 presented in their Table 1 Results given by Dave et al. (1987) for ring test with Z n S O 4 • 7H20 shown in their Table 2 a
Furthermore, the estimated variability due to parental fish emphasizes the importance of randomization, especially when eggs from several parental fishes are used in a test.
E s t i m a t i n g the " N o - E f f e c t " for Mercury
Concentration
The objectives for most EL toxicity tests are to determine " n o effect concentrations" for chemicals, products and effluents. Two approaches, discussed previously by Dave (1984) and by Stephan and Rodgers (1985), may be used to derive " n o effect concentrations" from chronic and EL toxicity tests. One approach is to determine LOEC by an hypothesis testing procedure, such as analysis of variance with a multiple comparison procedure, and to use the geometric mean of LOEC and NOEC as an estimate of the " n o effect concentration". The other approach is regression analysis based upon the dose-response relationship found in the actual experiment. The latter procedure is depicted in Figure 1. The Zero Equivalent Point (ZEP) as defined by Luckey (1975) is the concentration corresponding to the intersection between the dose-response relationship and the control MST line. The " n o effect concentrations" estimated by these two approaches are summarized in Table 2. Because of the wide 95% confidence limits of the MST for concentration 2 Ixg/L of replicate B, the LOEC is determined to be 4 Ixg/L instead of 2 ixg/L as in replicate A. The statistical meaning of confidence limits has been discussed previously by Hodson et al. (1977), Stephan (1977), Gersich et al. (1986), and Kaiser (1989). In general practice, most people are using the confidence limits as an aid in the interpretation of differences in toxicity between chemicals or treatments. In our set of data, we have been using them to compare
Table 2. Estimated "no effect concentrations" for divalent inorganic mercury 0zg Hg/L) in a replicated embryo-larval toxicity test with zebrafish (Brachydanio rerio) Estimated value
Replicate A
Replicate B
Geometric mean
LOEC a NOEC b Geometric mean ZEPc
2 1 1.4 1.18
4 2 2.8 1.13
2.8 1.4 2.0 1.15
a Lowest Observed Effect Concentration derived from data shown in Figure 1 b No Observed Effect Concentration derived from data shown in Figure 1 c Zero Equivalent Point derived graphically as shown in Figure 1
MHTs and MSTs from different treatments. This may not be statistically defendable, but we feel that this is the way most people are using the confidence limits today. If the same people had been using regression analysis by simply fitting the dose-response relationships to derive the ZEPs for these sets of data (replicate A and B) as in Figure 1, their "no effect concentrations" would not only have been more consistent, their " n o effect concentrations" would also have been somewhat lower (Table 2). Thirdly, their arguments in favor of their decision are easier to explain to most people, because the concentration they have determined to be the " n o effect concentration" is actually the highest concentration, that according to their combined data should not give any detectable effect on survival. Thus, in the practical use of EL toxicity tests for regulatory purposes, long and inconclusive debates on differences betwen biological and statistical significance can be avoided. This difference is certainly
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
an important issue when putting the result of a toxicity test into its environmental perspective, but the question can not be answered by results from a toxicity test with a surrogate species. If the test had been terminated after 10 days, which has been a d v o c a t e d by some p e o p l e , then the " n o effect concentration" for Hg would have been between 8 and 16 Fxg Hg/L. The geometric mean (11.3 Ixg Hg/L) is one order of magnitude higher than the ZEP derived after 14 days (1.t5 p~g Hg/L, Table 2).
Comparison between F e d and N o n - F e d Larvae Since we don't have comparative data for fed and nonfed larvae, a comparison is presented with a previous study (Snarski and Olson 1982) of another cyprinid fish, the fathead minnow (Pimephales promelas). Our results showed that inorganic mercury (HgC12) affected early developmental stages of the zebrafish at concentrations down to 1.1 lag Hg/L. In the fathead minnow, chronic effects of HgC12 were found at even lower concentrations. No spawning occurred in 1.02 pxg Hg/L and the number of spawning pairs were reduced in 0.26 and 0.50 lag Hg/L. Larval survival after 30 and 60 days was not reduced when they were exposed to concentrations below 4.5 ~xg Hg/L, while body length and weight were reduced in 0.58 Ixg Hg/L after 60 days in troutstarter-fed larvae. In Artemia-fed larvae, body length and weight were not affected in 1.02 ixg Hg/L but they were modified in 2.01 p.g Hg/L. Thus, the " n o effect" concentrations derived for the fathead minnow E L exposures are similar to those found in the zebrafish EL exposures. In the absence of directly comparable data for fed and non-fed larvae, the similarity in " n o effect concentrations" for growth in fed fathead minnow larvae and survival in non-fed zebrafish larvae suggests, that the late mortality seen in our experiment (Figure I) reflects a decreased growth in fed larvae. The reason for not feeding the larvae is that we don't have any suitable diet to provide. If an inferior diet is given, then the variability in growth increases, which may obscure effects on growth. If an hypothesis testing procedure is used for determination of LOEC, this would make the test tess sensitive. The variability in growth among trout starter-fed minnow larvae was greater than that in Artemia-fed larvae (Snarski and Olson 1982). Furthermore, the CV for body weight was higher in sheeshead minnow (Cyprinodon variegatus) larvae fed restricted rations than in those fed unrestricted rations of Artemia (Cripe et af. 1986). Therefore, for nutritional reasons and practical matters related to inconsistencies in live food production and optimal numbers of feeding per day, it seems more appropriate not to feed the zebrafish larvae than to feed them an inferior diet and/or at a suboptimal ration.
Copper The most unexpected finding in this study was that hatching was delayed at such low concentrations of copper and nickel. In particular, the extreme toxicity of copper and its
t.3[ Table 3, Summary of "no-effect" concentrations 0xg/L) of mercury, copper, nickel lead and cobalt to embryos and larvae of zebrafish (Brachydanio rerio) Response variable Effect criterium
Hg Cu (p.g/L) (txg/L)
Hatching time GM NOEC-LOEC a 13 ZEP b 10 Survival time GM NOEC-LOEC 2.0 ZEP 1.15 Guidelinec 0. I d
Ni (vg/L)
Pb Co (Ixg/L) (~xg/L)
Arch. Environ. Contain. Toxicol. 21, 126-134 (1991)
c
nvironmental ontamination
9 1991 Springer-Verlag New York Inc.
Toxicity of Mercury, Copper, Nickel, Lead, and Cobalt to Embryos and Larvae of Zebrafish, Brachydanio rerio G 6 r a n D a v e .1 a n d R u i q i n Xiu** * Department of Zoophysiology, University of G6teborg, Box 25059, 400 31 G6teborg, Sweden and ** Department of Microbiology and Aquabiology, Institute of Environmental Health and Engineering, Chinese Academy of Preventive Medicine, 29 Nan Wei Road, 100 50 Beijing, China
Abstract. The toxicity of mercury (HgCI2), copper (CuCI2:5 H20), nickel (NiSO4:6 H20), lead (Pb(CH3COO)2:3 H20) and cobalt (COC12:6 H20) was studied under standardized conditions in embryos and larvae of the zebrafish, Brachydanio rerio. Exposures were started at the blastula stage (2-4 h after spawning) and the effects on hatching and survival were monitored daily for 16 days. Copper and nickel were more specific inhibitors of hatching than cobalt, lead, and mercury. Nominal " n o effect" concentrations determined from the dose-response relationships (ZEPs, Zero Equivalent Points) for effect on hatching time were 0.05 p,g Cu/L, 10 Ixg Hg/L, 20 txg Pb/L, 40 Ixg Ni/L and 3,840 Ixg Co/L, and those for effect on survival time were 0,25 ~g Cu/L, 1.2 rzg Hg/L, 30 ~g Pb/L, 80 txg Ni/L, and 60 fxg Co/L. The " n o effect" concentrations for Ni, Hg and Pb are consistent with previously reported MATC values for sensitive species of fish. The " n o effect" concentrations for copper are 1-2 orders of magnitude lower than previously reported values. The major reason for the latter discrepancy was considered to be the absence of organics that can complex copper ions in the reconstituted water that we used, which had a hardness of 100 mg/L (as CaCO3) and a pH of 7.5-7.7. Unexposed controls were started with embryos from different parental zebrafishes and the parental-caused variability in early embryo mortality, median hatching time and median survival time were estimated.
Embryo-larval (EL) toxicity tests are generally more sensitive than toxicity tests with juvenile and adult fish. F o r many pollutants, E L toxicity tests are as sensitive or almost as sensitive as chronic (life cycle) toxicity tests (Macek and Sleight 1977; McKim 1977). Hatchability, growth, and survival are the most frequently used response variables in E L toxicity tests. Survival and hatching are discrete (all or none) responses that can be evaluated by ordinary dose-response procedures, eg log-probit analysis. Growth, being a suble-
Address correspondence to: GOran Dave, Department of Zoophysiology, Box 250 59, 400 31 G6teborg, Sweden.
thal response, is often believed to be more sensitive than survival. However, among 173 tests, reviewed by Woltering (1984), larval survival was affected at LOEC (Lowest Observed Effect Concentration) in 57% of these tests, growth in only 36% and hatching in 11%. In a later review of 136 tests, Kristensen (1990) concluded that growth was the most sensitive variable in 70% of the tests when the juvenile stage was included. Furthermore, the embryo stage was found to contribute very little to the sensitivity of the E L toxicity test. However, if hatching time and not only percentage hatch is considered, then the embryo stage is sometimes more sensitive than the larval stage (Dave 1986; Dave et al. 1987). In the p r e s e n t E L t o x i c i t y t e s t s with the z e b r a f i s h (Brachydanio rerio), only hatching and survival were used as response variables. Since the larvae were not fed, it was assumed that toxicant-induced stress responses would ultimately show up as a reduced survival time. Hatching time can be affected secondarily as a result of primary effects on developmental rate of the embryo or early larval stages (Rosenthal and Alderdice 1976). Or toxicants may interfere with the hatching process directly. Hatching often results from a combined effort of the hatching enzyme (chorionase), increased perivitelline pressure, muscular contractions or active uptake of water by the embryo (Denuc6 1985). By these means environmental factors, including toxicants, may affect hatching time in two directions (stimulation and inhibition). The aim of the present study was to investigate the effects of five heavy metals (mercury, copper, nickel, lead and cobalt) on embryos and larvae of zebrafish under standardized conditions (SIS, 1988). Furthermore, we wanted to investigate possible parental influence on hatching and survival time in unexposed fish (controls).
Materials and methods Animals The procedures for production of zebrafish eggs and exposure of eggs and larvae have been described previously (SIS 1988; Dave et al. 1987). An English translation of the standard protocol (SIS 1988)
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
is available upon request from one of the authors (G. Dave). Briefly, the test is started by addition of 30 eggs in the blastula stage (2-4 h after spawning) to a geometric series of concentrations (dilution factor 0.5). Test solutions (50 ml in a Petri dish) are renewed daily. After 24 h, the number of eggs is reduced to 20 in order to compensate for control mortality and to provide equal numbers (N) for effect variable estimates. Since the embryos (eggs) can't swim, this culling is random. No food is provided and the test is terminated when at least 90% of the larvae have died in the control. The larvae are not fed because of lack of a suitable synthetic diet. Daily records of the number of live eggs (embryos) and larvae are used to determine the median time for hatch and survival (MHT and MST).
Exposure to Mercury Eggs from two pairs of spawners (one male and one female in each pair) were distributed randomly (five by five) in dishes containing nominal concentrations of 1-512 txg Hg/L (dilution factor 0.5). Each concentration was duplicated and each concentration series contained a duplicated control (4 controls altogether). The source of mercury was mercuric chloride (HgClz), pro analysi 99.5%, Merck, lot no 8556740. All concentrations are expressed as ~xg total Hg/L and are based on nominal values.
Exposure to Copper, Nickel, Lead, and Cobalt Eggs of a defined age (2-4 h after spawning) were obtained from 5 spawning pairs (1 male plus 1 female). Eggs from two pairs were used for the copper exposure unit, and eggs from the remaining three pairs were used for the nickel, lead and cobalt exposure units. Each exposure unit consisted of 11 concentrations (dilution factor 0.5) and two controls. All exposures were started on the same day with 2-4 h old eggs (blastula stage). Thirty eggs were added, five by five, to each concentration (50 ml test solution in a Petri dish with i.d. 10 cm). After 24 h, the number of living eggs (transparent) was reduced to 20 in the copper exposure unit. In the other exposure units (Ni, Pb, Co) the number was reduced to 10 because of higher initial mortality. The toxicant sources were CuSO4:5 H20 (Fisher reagent grade), N i S O 4 : 6 H20 (Merck analytical grade), Pb(CH3COO)2:3 H20 (Merck analytical grade) and CoC12:6 H20 (Baker analytical grade). Test concentrations given are expressed as ~g/L of the metal only.
Measurements and Ambient Conditions Test solutions were renewed every 24 h. Living embryos (transparent) and larve (motile) were recorded on these occasions, and more frequently around hatching time. The dishes were incubated at 26~ at a photoperiod fo 12 h light and 12 h darkness controlled by an incandescent tube and a timer. The dilution water was prepared from Milli Q Reagent Water (Millipore) and analytical wade CaC12: 2 HzO (117.6 rag/L), MgSO4:7 H20 (49.3 mg/L), NaHCO 3 (25.9 rag/L) and KCI (2.3 rag/L). This water has a hardness of 100 mg/L expressed a s C a C O 3. After aeration, the pH was adjusted, if necessary, to 7.5. Daily recordings of temperature, dissolved oxygen and pH in new and old test solutions ranged from 25.4 to 26.4~ 7.5-7.7, and 88-100% of air saturation, respectively.
Statistical Treatment Median times for hatching (MHT) and survival (MST) were calculated by probit analysis (Peltier and Weber 1985). The Lowest Observed Effect Concentrations (LOECs) for effects on hatch and survival were estimated by comparing MHTs and MSTs. If 95% con-
127
fidence limits for controls and exposed animals were sepmated an effect was assumed. The next lower concentration was defined as the No Observed Effect Concentration (NOEC). The geometric mean of LOEC and NOEC was used as one estimate of the "no effect" concentration~ Another estimate of the "no effect" concentration was obtained by inspection of the log (MHT) = f (tog C) and the log (MST) = f(log C) relationships. From these dose-response relationships the Zero Equivalent Point (ZEP) was determined graphically as shown in Figures 1-5.
Results Early embryo mortality (after 24 h) and the median times tbr hatch and survival with each metal (Hg, Cu, Ni, Pb, Co) are shown in Figures 1-5. Control values are summarized in Table 1.
Sources of Variability in Controls The influence of individual embryo variability was counteracted by randomization to treatment (= Petri dish). In order to save time and equalize the time for start of exposure, eggs were distributed 5 by 5 to each dish until they contained 30 eggs. This randomization was appropriate because initial mortality among replicates was not biased. As expected, the variability between replicated controls (from the Hg experiment) was less than between simultaneously treated experimental units (from the Cu, Ni, Pb, and Co experiments), which were started with eggs from different parental zebrafishes. A comparison with variability within and between laboratories (from Dave et al. 1987) is atso shown in Table 1. As expected, variability (CV) associated with replication is considerably less than the variability associated with repetition, and part of the latter is explained by parental variability. The higher variability for M H T compared to MST is consistent with previously estimated variability within and between laboratories (Dave et al. 1987).
Mercury Early embryo mortality (within 24 h), MH T (Median Hatching Time) and MST (Median Survival time) in various concentrations of mercury are give in Figure 1. The results from the two replicates (A and B) are presented separately because determinations of " n o - e f f e c t " concentrations were made separately in order to evaluate replicability. The L O E C (Lowest Observed Effect Concentration) for effects on hatching time was 16 Ixg Hg/L. In 32 wg/L hatching was completely inhibited, but in lower concentrations of Hg there was a slight stimulation of hatching. This stimulation was most pronounced in 4 Ixg Hg/L. Median survival time in this concentration was only 11 days, which is 2.5 days less than the control value (13.5 d). The dose response relationship for effects on survival (Figure l) can be separated into four phases. Early embryos were killed in concentrations ~>5t2 ~g/L, late embryos between 512 and 32 ~g/L, early larval stages between 32 and 8 Ixg/L, and late larval stages between 8 and 1 txg/L. These four phases were seen during continuous exposure, and are the result from both prolonged exposure and shift in developmental stage. Furthermore, in the transition from early to late larval stages (after 10 to 11 days) there is atso an effect of starvation, since the larvae were not fed (discussed below).
128
G. Dave and R. Xiu
16 14 12
REPLICATE A ZEP 1.18JJg/L _ _ _ ..,_~_ _ .................
16 14
cont r_o/MST _
_
%" ~
,; ..... +_z___/_
............
I0
12 to
-~
o
8
/'ZEP O.05.ug/L
\ -L. control MHT
. ." .~.'./ ........ .. .. ......... .. .. ......... .. .. .... . . .
+ m ~.~ .~
4 IOO
o
oli
i
16
~b so mE
2
0
I00
~b ~E
I
I0
IO0
n n n/1, I] I1 n,n n n i
o
~o
ioo
m
,fl fl fl n, n n n,~
0
0.1 I I0 Nominal conc.of copper, ug / L
I000
so
0
0
iooo
ido
HI 16o
Fig. 2. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to copper under standardized conditions
Nominal conc.of mercury, ~g/L
A
16 14
REPLICATE B ZEP I /5 ug/L .... ~................... co_ntr_d_MS_T___
to
ZEP 80 u g / L , / control MST
/
14 12
12
10
I0
Z
~_~
g
---,--,--,--,--,--,--, ---
-
"8~
8
8.-c
6
-
9
g~
.~=
4
l
i
_
I00
i,- .................. ~ ;
,
,
0
control MHT
---s
....
2
I
I0
IO0
o
,
I00
~
rln o
B
.
I000
~b 50 a~:
[1, ~ ~ I
4
I0 I00 Nominal conc. of mercury, ug/L
.
.
+_!_!
0
i
o
I
6
16o
iobo
10
I00
1000
so
IOC)O
Fig. 1. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to mercury (Hg z+) in replicated treatments (A and B) under standardized conditions Note: ZEPs, Zero Equivalent Points, for effects on hatching and survival were estimated graphically at intersections between the fitted (by the eye) dose-response relationship and the median times for hatch and survival in the controls, respectively. The interpretation of ZEP is that it is the highest concentration where the measured response would be the same as in the control
.
Nominol conc.of nickel, ug/L Fig. 3. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to nickel under standardized conditions
Copper The highest c o n c e n t r a t i o n (128 I~g Cu/L) killed all e m b r y o s within 24 h (Figure 2), and less than 50% hatched in concentrations d o w n to 1 ixg Cu/L. F u r t h e r m o r e , hatching was sig-
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
ZEP 6 0
16
i,/"
i4
~z
.....
8 6
o
12
o;ro,
*+
,ug/L
. . . . . 4__,__ _/_. . . . . . . . . . . . .
_c2'L~__MS_T__
/
I0 84
~
/_
+ .... .--.--.--.--.--.--.-...
12
Z
129
8~,
8
s~:
6
m~
~
g~= ~_~
4
4
,; .... :__:__hi__
_oo:,:o,
I
ZEP 5 8 4 0 ,ug/l
2iO0
$-t
o
o
1
0
,b
I
,do
]() IO0 Nominal conc of leod, Jag/L
,obo
o
IO0
rE s~ >,u
50
o
O
IO00
Fig. 4. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio rerio) embryos and larvae exposed to lead under standardized conditions
i
I
~do
,o'oo
I0 lO0 Nominel conc of cobcllt,j..ig/L
,o&o
I000
Fig. 5. Early embryo mortality, Median Hatching Times (MHTs) and Median Survival Times (MSTs) for zebrafish (Brachydanio re~'/o) embryos and larvae exposed to cobalt under standardized conditions
Cobalt nificantly delayed in all of the tested concentrations, while survival was not affected in the lowest concentration (0.125 p.g Cu/L). Thus, hatching was more sensitive to copper than was larval survival. This, unexpected, high sensitivity of hatching to copper was confirmed in another (unpublished) experiment were the LOEC for effects of hatching was 0.63 txg Cu/L and the NOEC was 0.063 Ixg Cu/L (dilution factor 0.1). The extrapolated " n o effect" concentration (ZEP) for hatching in this experiment was as low as 0.05 txg Cu/L (Figure 1).
The effects of cobalt were qualitatively different from those seen with mercury, copper, nickel, and lead. Hatching was at most minimally affected even at the highest concentrations. The ZEP for MHT shown in Figure 5 (3,840 txg Co/L) is only vaguely supported by our results at the two highest concentrations. Survival was not affected until after 9 days, but after that time survival was affected from 15,360 ~g/L down to 60 txg/L (ZEP for MST in Figure 5).
Discussion Nickel Early embryos were not killed even at the highest concentration (l,024 t~g Ni/L), but hatching was delayed down to 40 Ixg/L (ZEP for MHT in Figure 3). The "no effect" concentration for survival was estimated to he 80 Ixg/L (ZEP for MST in Figure 3).
Lead High concentrations of lead (480 and 960 Ixg Pb/L) inhibited hatching nonspecifically by killing the embryos. Lower concentrations delayed hatching slightly, but no clear doseresponse relationship was found for this delay (Figure 4). The " n o effect" concentrations for effects on hatching and survival were estimated to be 20 Ixg/L and 30 t~g/L, respectively (ZEPs for MHT and MST in Figure 4).
RepIicability, Repeatability, and Reproducibility The variability between replicates was low (high repiicability). The coefficient for variation (CV; SD x 100/2) for the logarithms MSTs for the 4 controls shown in Table 1 is 0.4%. Corresponding CVs for repeated tests at five laboratories were 1.2% or 1.5% (repeatability). The reproducibility for these five laboratories' controls (CVs for control log MST in tests with zinc sulphate and potassium bichromate) was 2.0% and 3.3%. Values for repeatability and reproducibility were given by Dave et al. (1987). These comparative figures for replicability (0.4%), repeatability (1.2-1.5%) and reproducibility (2.0-3.3%), suggest that replicated treatments are not needed in this test. This is because the variability associated with replication is small compared to that associated with repeating the experiment or reproducing it by someone else. However, a replicated control is advisable as an internal control for variability and repeatability at the laboratory,
130
G. Dave and R. Xiu
Table 1. Variability in Early Embryo Mortality (EEM), Median Hatching Time (MHT) and Median Survival Time (MST) in controls due to replication, parental variation, repetition at the same laboratory (repeatability) and at another laboratory (reproducibility) Variable (unit)
Source of variation
Mean value
SD
CV (%)
N
Data from
EEM (dead/total)
Replication Parental Repetition c Laboratoryc Replication Parental Repetition
7/30 13.6/30 16.5% 16.5% 0.465 0.506 0.453 0.411 0.453 0.411 1.128 1.127 1.126 1.137 1.126 1.137
2.6/30 5.9/30 12.1% 12.1% 0.036 0.090 0.045 0.069 0.051 0.069 0.005 0.019 0.014 0.018 0.023 0.037
37 43 73 73 8 18 10 17 11 17 0.4 1.7 1.2 1.5 2.0 3.3
4 4 37 37 4 4 19 22 19 22 4 4 19 22 19 22
Fig. Ia Figs. 2-5b Ref.c Ref# Fig. 1a Figs. 2-5b Table Id Table 2~ Table 1d Table 2~ Fig. 1 Figs. 2-5 Table 1a Table 2~ Table 1d Table 2e
Log MHT (days)
Laboratory Log MST (days)
Replication Parental Repetition Laboratory
Replicates shown in Figure 1 of this study b Control values shown in Figures 2-5 of this study c Results taken from Dave et al. (1987) for total variation found in all tests at all laboratories d Results by Dave et al. (1987) from ring test with K2Cr207 presented in their Table 1 Results given by Dave et al. (1987) for ring test with Z n S O 4 • 7H20 shown in their Table 2 a
Furthermore, the estimated variability due to parental fish emphasizes the importance of randomization, especially when eggs from several parental fishes are used in a test.
E s t i m a t i n g the " N o - E f f e c t " for Mercury
Concentration
The objectives for most EL toxicity tests are to determine " n o effect concentrations" for chemicals, products and effluents. Two approaches, discussed previously by Dave (1984) and by Stephan and Rodgers (1985), may be used to derive " n o effect concentrations" from chronic and EL toxicity tests. One approach is to determine LOEC by an hypothesis testing procedure, such as analysis of variance with a multiple comparison procedure, and to use the geometric mean of LOEC and NOEC as an estimate of the " n o effect concentration". The other approach is regression analysis based upon the dose-response relationship found in the actual experiment. The latter procedure is depicted in Figure 1. The Zero Equivalent Point (ZEP) as defined by Luckey (1975) is the concentration corresponding to the intersection between the dose-response relationship and the control MST line. The " n o effect concentrations" estimated by these two approaches are summarized in Table 2. Because of the wide 95% confidence limits of the MST for concentration 2 Ixg/L of replicate B, the LOEC is determined to be 4 Ixg/L instead of 2 ixg/L as in replicate A. The statistical meaning of confidence limits has been discussed previously by Hodson et al. (1977), Stephan (1977), Gersich et al. (1986), and Kaiser (1989). In general practice, most people are using the confidence limits as an aid in the interpretation of differences in toxicity between chemicals or treatments. In our set of data, we have been using them to compare
Table 2. Estimated "no effect concentrations" for divalent inorganic mercury 0zg Hg/L) in a replicated embryo-larval toxicity test with zebrafish (Brachydanio rerio) Estimated value
Replicate A
Replicate B
Geometric mean
LOEC a NOEC b Geometric mean ZEPc
2 1 1.4 1.18
4 2 2.8 1.13
2.8 1.4 2.0 1.15
a Lowest Observed Effect Concentration derived from data shown in Figure 1 b No Observed Effect Concentration derived from data shown in Figure 1 c Zero Equivalent Point derived graphically as shown in Figure 1
MHTs and MSTs from different treatments. This may not be statistically defendable, but we feel that this is the way most people are using the confidence limits today. If the same people had been using regression analysis by simply fitting the dose-response relationships to derive the ZEPs for these sets of data (replicate A and B) as in Figure 1, their "no effect concentrations" would not only have been more consistent, their " n o effect concentrations" would also have been somewhat lower (Table 2). Thirdly, their arguments in favor of their decision are easier to explain to most people, because the concentration they have determined to be the " n o effect concentration" is actually the highest concentration, that according to their combined data should not give any detectable effect on survival. Thus, in the practical use of EL toxicity tests for regulatory purposes, long and inconclusive debates on differences betwen biological and statistical significance can be avoided. This difference is certainly
Toxicity of Hg, Cu, Ni, Pb, and Co to Zebrafish
an important issue when putting the result of a toxicity test into its environmental perspective, but the question can not be answered by results from a toxicity test with a surrogate species. If the test had been terminated after 10 days, which has been a d v o c a t e d by some p e o p l e , then the " n o effect concentration" for Hg would have been between 8 and 16 Fxg Hg/L. The geometric mean (11.3 Ixg Hg/L) is one order of magnitude higher than the ZEP derived after 14 days (1.t5 p~g Hg/L, Table 2).
Comparison between F e d and N o n - F e d Larvae Since we don't have comparative data for fed and nonfed larvae, a comparison is presented with a previous study (Snarski and Olson 1982) of another cyprinid fish, the fathead minnow (Pimephales promelas). Our results showed that inorganic mercury (HgC12) affected early developmental stages of the zebrafish at concentrations down to 1.1 lag Hg/L. In the fathead minnow, chronic effects of HgC12 were found at even lower concentrations. No spawning occurred in 1.02 pxg Hg/L and the number of spawning pairs were reduced in 0.26 and 0.50 lag Hg/L. Larval survival after 30 and 60 days was not reduced when they were exposed to concentrations below 4.5 ~xg Hg/L, while body length and weight were reduced in 0.58 Ixg Hg/L after 60 days in troutstarter-fed larvae. In Artemia-fed larvae, body length and weight were not affected in 1.02 ixg Hg/L but they were modified in 2.01 p.g Hg/L. Thus, the " n o effect" concentrations derived for the fathead minnow E L exposures are similar to those found in the zebrafish EL exposures. In the absence of directly comparable data for fed and non-fed larvae, the similarity in " n o effect concentrations" for growth in fed fathead minnow larvae and survival in non-fed zebrafish larvae suggests, that the late mortality seen in our experiment (Figure I) reflects a decreased growth in fed larvae. The reason for not feeding the larvae is that we don't have any suitable diet to provide. If an inferior diet is given, then the variability in growth increases, which may obscure effects on growth. If an hypothesis testing procedure is used for determination of LOEC, this would make the test tess sensitive. The variability in growth among trout starter-fed minnow larvae was greater than that in Artemia-fed larvae (Snarski and Olson 1982). Furthermore, the CV for body weight was higher in sheeshead minnow (Cyprinodon variegatus) larvae fed restricted rations than in those fed unrestricted rations of Artemia (Cripe et af. 1986). Therefore, for nutritional reasons and practical matters related to inconsistencies in live food production and optimal numbers of feeding per day, it seems more appropriate not to feed the zebrafish larvae than to feed them an inferior diet and/or at a suboptimal ration.
Copper The most unexpected finding in this study was that hatching was delayed at such low concentrations of copper and nickel. In particular, the extreme toxicity of copper and its
t.3[ Table 3, Summary of "no-effect" concentrations 0xg/L) of mercury, copper, nickel lead and cobalt to embryos and larvae of zebrafish (Brachydanio rerio) Response variable Effect criterium
Hg Cu (p.g/L) (txg/L)
Hatching time GM NOEC-LOEC a 13 ZEP b 10 Survival time GM NOEC-LOEC 2.0 ZEP 1.15 Guidelinec 0. I d
Ni (vg/L)
Pb Co (Ixg/L) (~xg/L)
