Aquaculture, I13 (1993) 269-289Elsevier Science Publishers B.V., Amsterdam

AQUA 3006s

269

( Thunberg )

Inviable hybrids of Crassostrea virginica(Gmelin) with C. rivularis (Gould) and C. gigas

Standish K. Allen Jr.u, Patrick M. Gaffneyb, John Scarpau and David Busheku

'Rutgers University, Institute of Marine and Coastal Sciences, Haskin Shellfish Research Laboratory,Port Norris, NJ, USA

bCollege of Marine Studies, University of Delaware, Lewes, DE, USA

(Accepted 23 October 1992)

ABSTRACT

Interspecific hybridization may provide important tools for selective breeding programs in oysterculture, especially for enhancement ofdisease resistance, and may have a bearing on debates concern-ing the introduction ofnon-native species. Factorial crosses of C. virginica with C. rivularis (C. ar-akiensis) and C. gigas were made, producing control and hybrid larvae. Larval survival and growth

were documented. After several replicated experiments il became apparent that diploid hybrids were

inviable, and so triploid hybrids were also tested and found to be equally inviable. Feeding studieswirh hybrids were initiated to determine if lack of growth and viability were related to capture offood. Overall, hybrids of C. virginicawith C. rivulans and C. gigas can be readily produced, but are

inviable after 8- I 0 days and grow little. With regard to the species examined here, previous reportsofsuccessful hybridization should be questioned. Introduction of C. gigas to the native range of Cvirginicawill not have direct genetic effects on C. virginica.

INTRODUCTION

One promising alternative to conventional selective breeding is the produc-tion by interspecific hybridization of novel genetic types that miglrt yieldquantum changes in the commercial traits of a species. These traits, e.9., dis-ease resistance, growth rate and hardiness, can then be refined by selectionand backcrossing. An important feature of this strategy for aquaculture is therapid generation of novel genotypes, in contrast to the gradual progress ofconventional breeding. Hybridization can be modified further with recentlydeveloped techniques of induced polyploidy to create interspecific polyploidhybrids. Our research program is currently focusing on the evaluation of in-

Correspondence to:5.K. Allen Jr., Rutgers University, Institute of Marine and Coastal Sciences,Haskin Shellfish Research Laboratory, Port Norris, NJ 08349, USA.

0044-8486 /93/$06.00 @ 1993 Elsevier Science Publishers B.V. All rights reserved.

INVIABLE HYBRIDS IN CRISSOSTRE/

Experimental design

Because of the uncertainty of the fate of hybrids in Crassostreawe designedthe crosses so that we could trace the parentage ofprogeny produced, whethertrue hybrids or of uniparental origin. Six types of cross were conducted (Ta-ble l). Abbreviations for oyster species are as follows: G--C. gigas, R=C.rivularis and V: C. virginica, with females listed first.

In each cross, one to three experiments were done. Each experiment con-tained from two to I I replicates. Each replicate of a mating consisted of eithera full factorial mating of two species (Fig. l, top) or, in most instances, onehalf of the factorial consisting of oocytes from one species fertilized by spermfrom two species (Fig. 1, bottom ). Within an experiment, a single spenn sus-pension taken from two males was used for fertilizing oocytes, but differentfemales were used for each replicate. Replicates within an experiment, there-fore, were paternal half-sib groups. Table I lists the total number of experi-ments and replicates within them for crosses among C. virginica, C. gigas andC. rivularis. After spawning, parents were frozen at -80'C for subsequentelectrophoresis.

Production of diploids and larval development

All crosses were made and reared at 22-23 ppt salinity and 23-25" C, ex-cept that salinity for crosses with C. rivularis oocytes was 20 ppt (Breese andMalouf, 1977). A maximum of I million fertilized eggs per cross were incu-bated in 15 I culture vessels ( <67 eggs/ml). This density is higherthan usu-ally recommended. However, we overstocked our cultures because we wantedto ensure sufficient numbers of progeny, especially if they originated from an

TABLE 1

Summary of experimental groupings, the basis for paired comparisons in hybrid and control crosses

of C. virginica with C. gigas and C. rivularis. Control and test groups were produced as illustrated inFig. l. For diploid hybrids, pure species crosses served as controls; for triploid hybrids, controls werediploid hybrids

Control/test cross No. experiments No. replicatesper experiment

271

GG/GVw/vG

RR/RVvv/vR

VG/WGGV/GGV

3/l3/3/3

4/3/33/3/2

4164

INVIABLEHYBRIDSIN CMSSOSTR',4 273

(Table I ). In each cross 2-4 million eggs of one species were fertilized withpooled sperm from two males of the other species and immediately split intotwo groups. Early development was monitored microscopically until approx-imately 500/o of first polar bodies appeared. At this time, one group of eggs

was treated with cytochalasin B (CB).We also produced two mass spawns of diploid and triploid hybrids between

C. virginica and C. gigas. Eggs were stripped from 20 and 2l females of C.

gigas and C- virginica, respectively. We combined 2 million eggs from eachfemale and then split the pool into 20 equal aliquots. Each aliquot was fertil-ized by a separate male of the opposite species. After l0 min, fertilized eggs

were recombined and split equally into two groups (diploid and triploid).When 500/o of the polar bodies were visible the triploid group was treated withCB.

CB treatment consisted of adding 1.0 mg CB, dissolved in I ml dimethyl-sulfoxide (DMSO), per 2l of seawater. After 20 min, the eggs were drainedonto a 15 pmscreen, rinsed with I pm filtered seawater, then soaked in 0.050/o

DMSO (v /v) dissolved in I pm filtered seawater for20 min. Afterwards eggs

were added to culture tanks.

Genetic conJirmation

Flow cytometryApproximately 10000 48-h-old (straight hinge) larvae were sampled from

most cultures for flow cytometric analysis. In preparation for flow cytometry,larvae were concentrated into a I ml suspension, pelleted by centrifugation atl500xgfor l0 s in a microcentrifuge, and the supernatant discarded. Onehalf ml of DAPI/detergent/DMSO solution (Allen and Bushek, 1992 ) wasadded to the tube and the pellet resuspended. Larval suspensions were storedat -80"C. For analysis, larvae were thawed and disaggregated by repeatedaspiration through a 1 ml syringe fitted with a26 G needle. Cell suspensionswere passed through a25 pm screen immediately before the assay.

Flow cytometry was conducted on a Partec CA II cytometer. Analysis oflarval samples yielded data from a population of larvae (i.e., a sample of 10000larvae prepared by disaggregation ) (Allen and Bushek, 1992). Apparent DNAcontent was analyzed by the curve fi tting program MODFIT ( Verity SoftwareHouse, Topsham, Maine, USA). In the hybrid cultures, we were looking forevidence of ( I ) intermediate DNA content of the hybrid compared to thepure species and (2) increased variance in DNA content in hybrid crosses

that might indicate either mosaicism/aneuploidy resulting from chromo-some losses, or uniparental inheritance ( e.g., through gynogenesis ) . (The lat-ter would result in increased variance in larval samples if the distributions ofDNA content in uniparental and hybrid larvae were not distinguishable as

discrete peaks. ) In triploid hybrid cultures, we used larval DNA analysis to

INVIABLE HYBRIDS IN CR,{S,SOSTNEI 275

were provided with algal species as described. For starved groups, incomingwater was filtered through a0.4 pm bacterial filter (because filtration to Ilm, as above, did not remove all food) and no food was added. Growth andsurvival were monitored as previously described.

Statistical analyses

All data were analyzed with the computer program SYSTAT (Wilkinson,1990 ). Fertilization success and yield data were arcsine transformed prior tostatistical analysis (Sokal and Rohlf, 1981).

Each experiment generally consisted of 3-4 replications of hybrids and ap-propriate controls (Table I ). For overall comparison of performance amongthe various crosses, all replicates were combined to obtain a mean perform-ance for the cross in question (Table 2). For growth data, the mean larvallength in a replicate was calculated from individual measurements (N:20)taken on days 2, 4,6, etc. Mean size of larvae in a particular cross was calcu-lated as the mean of all replicates.

We compared larval performance in pure crosses (e.g., GG vs W) andreciprocal hybrids (e.g., VG vs GV) using /-tests or Mann Whitney U-testswhen the assumptions of r-tests were not met. We then compared perform-ance of hybrids to their controls using a paired comparisons design. Each ex-periment was analyzed as a mixed model two-way ANOVA without replica-

TABLE 2

Fertilization rate and survival to 48 h of pure and hybrid crosses of C. virginica, C. gigas, and C.rivularis. Fertilization rate was observed directly by scoring 100 eggs 60-90 min after fertilization.Yield was estimated directly by counting numbers of straight-hinge larvae at 48 h. Estimated embryosurvival represents the proportion offertilized eggs developing to straight-hinge stage

Cross Fertilizationrate (o/o)

Yield (o/o) Estimated embryosurvival (o/o)

wVGGV

WGGGV

GG

VRRV

RR

38

3545

420

42

9

l6

42

33

2225

2

t6

37

5

l5

36

86

6256

4478

87

5394

86

INVIABLE HYBRIDS IN CN,{.',SOSTREI 277

for GV vs GG (consensus P=0.0037, Table 3) but not for VG and W(P:0.057, Table 3). Estimated embryo survival was about the same in thepure and in hybrid cultures.

C. virginicax C. rivularisFor crosses of C. virginica and C. rivularis, fertilization rate was lowest in

the VR cross (Table 2). We found no difference in fertilization rate betweenRR (860/o) and W (r:0.31; d.f.:19, P=0.76). However, fertilization ratein VR was significantly lower than RV (t:2.24, d.f.: 10, P:0.049). Pairedcomparisons of fertilization rate in RR vs RV and W vs VR (Table 3) re-vealed no significant differences (consensus P=0.452 and 0.083,respectively ).

Yield at 48 h in W and RR crosses was about 330/o (Table 2), not signifi-cantly different (t:0.26, d.f.=28, P:0.80). Hybrid crosses had lower sur-vival and were significantly different from one another (t:2.27, d.f.:16,P:0.037). Hybrid crosses yielded signiflrcantly fewer healthy larvae than theirpure controls according to consensus tests of paired comparisons (Table 3 ).Estimated embryo survival was also greater in the pure crosses (Table 2 ).

Triploid hybridsWe have no data on fertilization rate in triploid crosses and their controls.

Absolute survival to 48 h was very low for WG (20/o) and GGV ( 160/o ) crosses(Table 2), both of which were significantly different from their respectivediploid controls (consensus P:0.0097 and P:0.0301, respectively). Esti-mated embryo survival of triploid hybrids was lower than that of diploid hy-brids (Table 2).

Larval survival and growth

DiploidsAfter 48 h, survival rates in all hybrid crosses exhibited a common pattern

(with the exception of one replicate, VG,o). For all hybrids except VG,u,mortality was severe and steady for 10-12 days, so that virtually no hybridlarvae survived after this period. VG and VR crosses tended to expire beforetheir reciprocals (Fig. 2). There was no particular time period when mortal-ities were worse than others. The sole exception to this pattern was VG16, inwhich about 1000 larvae survived to eyed stage (0.10/o of the total number ofeggs incubated) and several hundred spat were obtained.

Pure crosses also had steep mortality curves, probably attributable to thehigh densities maintained in culture. By the time hybrids had died off, larvaldensities in the pure crosses had diminished to about l-4/ml, an acceptabledensity for mid-sized larvae.

The pattern of growth in all hybrids, except VG,u, was the same (Fig. 3).

INVIABLE HYBRIDS IN CRI.SSOSTRT,{

spawns, despite the high initial numbers of eggs and the large number ( > 400 )of family combinations produced, virtually complete mortality occurred byday 10. VG and WG crosses that began with 19.2 x 106 eggs each had I 500and 625larvae, respectively, by day 10; GV and GGV crosses that began with18.9X 106larvae each declined to 150 each in the same time period. All re-maining hybrid larvae were clearly abnormal and cultures were terminated.

The pattern of growth in triploid hybrids was essentially the same as dip-loids, increasing in size only about 100/o from the straight-hinge stage. We testedthe difference in larval shell length between diploids and triploids in the rep-licated hybrid cultures at each 2 day interval with a paired comparisons test.Although mean larval size was greater in WG at each time period (Table 4),triploids were not significantly larger than diploids (consensus P: Q.229;. 1,the mass spawn of WG hybrids, mean shell lengths of diploids and triploidsat2 day intervals were compared with a Mann Whitney U-test. On day 6 only,the WG culture had significantly larger larvae than VG (P<0.001), al-though the absolute difference was only 4 pm.

GGV and GV cultures were compared in the same way. From the pairedcomparisons test for replicated cultures, there was no time at which the two

TABLE 4

Difference in mean larval shell length, by day of culture, between diploid and triploid hybrid culturesof C. virginica and C. gigas

Day Paired replicatesr Mass spawn2

WG-VG GGV_GV VVG_VG GGV_GV

2 Diff. of means (pm)Probability

4 Diff. of means (pm)Probability

6 Diff. of means (pm)Probability

8 Diff. of means (pm)Probability

l0 Diff. of means (pm)Probability

Consensus P value

1.320.117

t.t20.s96

2.370.44'7

4. l00.047

2.420.862

0.229

2.800.141

5.000.063

5.30.160

4.30.220

0.015

0.540.586

l.s00.r54

3.98<0.001

0.500.649

<0.032

r.420.01 7

2.740.004

9.48< 0.0001

4.70<0.0001

<0.000001

tPaired replicates were lested by paired comparisons test.2Mass spawns were tested by Mann Whitney U-tests, because of skewed distributions of data on some

days.

INVIABLE HYBRIDS IN CR,{SSOSTRE,4 28r

1.75

FztrlFzo(J

zatr1

kF]trld

1.50 -

1.45 -

1.30 -

1.15;

(18) (1 1)

QL)-T

I

oI

Ti-L

I

l

l

!

I

_l

i

(e)

TOlr

i

i

6?iI

I

L (15)UIT1l

?I

1.00 L-- r I I L

GG GV VG VVL

VRii

RV R.R.

cRoss

Fig. 4. Mean DNA content of hybrid and pure crosses of C. virginica, C. gigas and C. rivularis lawaedetermined by flow cytometry. Bars are comparison limits of means (Sokal and Rohll 198 I ).

unambiguous discrimination of hybrids from parental species based on DNAcontent would be difficult, at least in larvae.

The other feature of DNA content examined was the variance of DNA con-tent in hybrid and pure crosses. Paired comparisons of coefficients of varia-tion (CV: [standard deviationx 100]/mean) showed significant differ-ences for GG vs GV only, with the hybrids showing greater variation in DNAcontent (P:0.0201).

El ec tr o phor e ti c analy s i sAll48-h pooled larval samples showed maternal gene expression. We were

particularly interested in paternal gene expression as conflnnation ofhybridization.

C. gigasxC. virginica. Ten replicates of GV crosses and their GG controlswere examined at 48 h after fertilization. Each replicate larval pool exhibiteda staining pattern at the Gpi locus identical to the mother (i.e., BB in 9 repli-cates, AB in one ). Because the two males providing the sperm used to fertilizeall of the replicates had genotypes BB and AB, the patterns in larvae wereconsistent with either a lack of paternal gene expression at 48 h or a failure ofthe AB male to contribute to the larval pool. Coincidentally, the same situa-tion obtained for the GG controls. Therefore it was not possible to confirmhybridization using this locus.

INYIABLE HYBRIDS IN CR,4,S.SOSTR'I

Parents

283

Progeny (pooled larvae)

Day 2 Day 4 Day 6

I

I

I

I

r

-

rI

-III

II

II

II

I

I

II

I

I

Fig. 6. Representative zymograms of pooled larvai samples from single pair matings of C. virginica(female) XC. gigas at 2, 4, and 6 days after fertilization. Staining intensity was typically weak byday 7.

samples were re-run. Lap-2 showed paternal alleles on days 2 and 4 for allthree crosses, and on day 6 in family YG+. Lap-l stained weakly, but showeda hybrid pattern on all days. Gpi showed paternal alleles at days 2 and 4. Est'2 was not well resolved in the larvae of these crosses.

In only one cross (VG,u) did any progeny survive to metamorphose andbecome juveniles. We scored both parents and l0 spat. At the Gpi locus, be-cause the males and the female were homozygous for different alleles, we ex-pected all progeny to be identical heterozygotes. Instead, four were homozy-gous for the maternal (C. virginica) allele, and six were heterozygous for twoC. virginica alleles, only one of which could have come from the parents usedin the cross. All ten progeny showed the C. virginica allele only at Mdh-l,confirming the non-hybrid nature of the progeny. Lap- I and Lap-2 also showedtypical C. virginica patterns, with none of the bands expected from the C.gigas father. These successful "hybrids" were clearly contaminant C. virgin-icathat originated from other spawns made in the hatchery.

C. virginicaXC. rivularui. Larvae from two crosses were examined at 48 h.Staining was poor for most enzymes. Est-l and,.Es/-2 showed bands charac-teristic of both species, whereas Gpi exhibited the maternal pattern only.

INVIABLE HYBRIDS IN CRI.SSO,STREI

7

DAY

285

150

130

t75

150

? 110

EINi;90

? 125

tr]NA loo

7

DAY

l3ll

Fig. 7. Mean size ( 3 replicates) of larvae from feeding experiment with fed and starved larvae of Cvirginica and C. virginicaxC. gigas hybrids (left ), and C, rivularis and C. virginicaxC. rivularis hy-brids (right) from day 2 to day 12. Left - VV fed (l) and VV starved (I), VG fed (O) and VGstarved ( O ). Fed laryae were given normal daily algal ration. Starved larvae were cultured in I pmfilteredwaterwith no algaeadded. Right-RR fed (V) and RR starved (V), RVfed (I) and RVstarved ( I ). Fed larvae were given normal daily algal ration. Starved larvae were cultured in 0.4 pmfiltered waler with no algae added.

DISCUSSION

C. gigas and C. rivularis cross readily with C. virginica. Mean fertilizationrate, while lower than pure crosses, was > 500/o in hybrid matings. Fertiliza-tion seems to be symmetrical for VG and GV. Fertilization between C. virgin-ica and C. rivularis was clearly asymmetrical, as was yield. The highest fertil-ization rate of all crosses produced was obtained in C. rivularis (female) x C.

virginica matings (940/o). Fertilization success in the reciprocal cross, VR,was highly variable, ranging from < lo/oto 97o/o (data not shown).

A rough comparison of embryo survival can be obtained from Table 2.Hy-brids between C. virginica and C. gigaswere as viable in the first 48 h as purecrosses. Triploid hybrid cultures were less viable than diploid hybrids. GGVsurvived half as well as GV. Additional mortalities were undoubtedly causedby CB treatments, which are known to be toxic; on average, triploid culturesof C. gigas are expected to be half as viable as diploid ones (Allen et al., 1989 ).Embryo survival in WG cultures was lower than expected from CB treatedeggs. Allen and Bushek (1992) reported mean survival in l4 triploid culturesof C. virginica of 22o/o. Timing of early meiotic events is not disrupted in hy-brids (Scarpa and Allen,1992), so it is not clear why mortalities were higherin WG crosses. Diploid hybrids of C. virginica and C. rivularis were clearlyless viable than their respective control crosses.

INVIABLE HYBRIDS IN CR,{SSOSTREI 287

Electrophoretic analysis confirmed the presence of hybrid larvae in allcrosses of C. virginicawith C. rivularis and C gigas. The limitations of proteinelectrophoresis also relate to the need to analyze larval populations ratherthan individuals. While electrophoresis confirms the expression of alleles fromboth species, it cannot confirm the presence oflarvae ofuniparental origin.Neither flow cytometry nor electrophoresis could exclude uniparental inher-itance. Two other lines of evidence suggest that none occurs, however. First,Scarpa and Allen (1992) demonstrated that there was no cytogenetic evi-dence to support uniparental development, using samples from replicatesproduced for this study. Second, apart from VG,u, no hybrid larvae were mea-sured that exceeded 100 pm. Uniparental progeny, if present, would likelyhave been observed as larger larvae.

Patterns of gene expression were not consistent among crosses, replicates,or enzyme loci. This heterogeneity may have arisen from several causes: ( I )the number and physiological state of the larvae in the larval pools subjectedto electrophoresis, (2) variation in enzyme stability on different cytoplasmicbackgrounds, (3) locus and allele-specific gene expression, and (4) geneticvariation at modifier loci affecting gene expression or stability. Clearly, thesehybrid oyster larvae are not the ideal material for the study of gene expres-sion, by comparison with organisms in which individual offspring can be re-liably scored at numerous loci.

Despite these limitations, the need for genetic confirmation is clear. VG,6produced survivors from a culture that was characteized as hybrid. Yet pro-tein electrophoresis demonstrated conclusively that the survivors were con-taminants. We were able to trace these contaminants to a culture that hadbeen produced on the same day (although in another part of the hatchery)and were maintained concurrently with the hybrid cultures; We think it isunlikely that contamination occurred while we were making the matings;rather, we suspect the contaminants were transferred by screens in the courseofhatchery rearing.

The case is strong for the inviability of hybrids of C. virginicawith C. gigasand C. rivularis. We question the validity of reports that indicate otherwise,especially since no genetic confirmation was provided. Indeed, except for thefollowing paper (Allen and Gaffney, 1993), no interspecific hybridizationswithin the genus Crassostrea have been confirmed.

Lack of hybridization between C. virginica and C. gigashas ecological im-plications in the debate over the introduction of the latter to the mid-Atlanticcoast of the United States. Failure to hybridize means that an introduction ofC. gigas would have no direct genetic consequences on the native C. virginica(cf. Gaffney and Allen, 1993 ). But this same failure also means that strategiesto use hybrids of C. virginica with C. rivularis or C. gigas for breeding in aqua-culture will be fruitless. Viable interspecific hybrids (e.9., C. rivularisxC. gi-

I}.IVIABLE HYBRIDS IN CR,{.SSOSTREI 289

Gaffney, P.M. and Allen, Jr., S.K., 1992. Genetic aspects of introduction and transfer of mol-Iuscs. J. Shellfrsh Res., I l: 535-538.

Gaffney, P.M. and Allen, Jr., S.K., 1993. Hybridization among Crassostrea species: a review.Aquaculture, in press.

Gaffney, P.M. and Scott, T.M., 1984. Genetic heterozygosity and production traits in naturaland hatchery populations of bivalves. Aquaculture, 42: 289-302.

Rice, W.R., 1990. A consensus combinedP-value test and the family-wide significance of com-ponents tests. Biometrics, 46: 303-308.

Scarpa, J. and Allen, Jr., S.IL, I 992. Comparative kinetics of meiosis in hybrid crosses of Crass-

ostreavirginicaard C. rivularis with C. virginica. J. Exp. Zool., 263:316-322.Selander, R.K., Smith, M.H., Yang S.Y., Johnson, W.E. and Gentry, J.B., 1971. Biochemical

polymorphism and systematics in the genus Peromyscas. I. Variation in the old-field mouse(Peromyscas polionotus\. Studies in Genetics, 6:49-91. (Univ. Texas Publ. No. 7103).

Small, S.M. and Benfey, T.J., I 987. Cell size in triploid salmon. I . Exp. Zrl,ol., 241: 339-342.Sokal, R.R. and Rohlf, F.J., 1981. Biometry, 2nd edition. W.H. Freeman and Co., San Fran-

cisco, CA, 859 pp.Wilkinson, L., I 990. SYSTAT: The System for Statistics. SYSTAT, Inc., Evanston ,lL, 67 6 pp.

270 S.K. ALLEN Jr. ET AL.

terspecific hybridization and polyploidy in species of oysters in the genus

Crassostrea.The record on interspecifrc hybridization among Crassostrea species is by

no measure clear. Gaffney and Allen ( 1993 ) concluded that most reports ofinterspecific hybridization in oysters must be interpreted with caution. Typ-ically, the occurrence of interspecific fertilization, development, and survivalof the larvae is taken as evidence of successful hybridization, without direct(genetic) confirmation. The few attempts at confirmation have shown that"hybrids" were in fact pure species, suggesting that contamination may be acommon occurrence in hybrid experiments, or alternatively, that interspe-cific hybridization stimulates uniparental development. These questions canbe resolved through a combination of careful experimental design and genetictechniques.

By using parents of known genotypes, we may use techniques (e.g., cyto-genetics, flow cytometry, protein electrophoresis) to discriminate among thebiological alternatives of gynogenesis, androgenesis, polyploidy, and contam-ination. In this paper we describe our experiments with crosses of C. virginicawith C. gigas and C. rivularis, document larval growth and survival, and dem-onstrate genetic evidence of hybridization using flow cytometry and electro-phoresis. We began the work by simply making interspecific matings in anattempt to produce viable offspring. As it became clear that diploid hybridswere inviable, we tried to "rescue" them by inducing triploidy. Finally, weconducted a set of feeding experiments to determine whether hybrid larvaecould capture and assimilate food. Cytogenetic evaluation of early develop-ment in hybrid crosses is documented elsewhere (Scarpa and Allen, 1992).

MATERIALS AND METHODS

Oysters and gametes

Sexually mature Crassostrea gigas were obtained from Washington, C. ri-vularis from Oregon, and C. virginica from Maine and New Jersey. Gameteswere obtained by dissection as follows. All surfaces and instruments contact-ing the oysters were cleaned with dilute bleach and rinsed with freshwaterbetween handling and opening of different individuals. Each oyster wasopened and a gonad biopsy taken for determining the sex of the individual bylight microscopy. Gametes from each oyster were dissected directly into in-dividual beakers. Oocytes were then separated from gonadal tissue and otherdebris by passing the suspension through a70 pm nylon screen. The oocyteswere caught on a 25 pm nylon screen and resuspended in seawater. Spermwere separated from debris by passing the suspension through a 15 pm nylonscreen.

272

$e

S.K. ALLEN Jr. ET AL.

CIAB

ffi 3'

oA-rB

dABJ-I- I

r^ffiIr-1-,--lL2dAB

Fig. 1. Experimental design for crosses made between species A and B. Crosses were either done as

full factorial (top) or, most often, with females from one species (bottom). An experiment consistedofa number ofreplicates ofthese matings. Individual females were used for each replication. Spermfrom two males of each species was pooled to ensure fertilization. For statistical analysis, we com-pared pure species (i.e., AA vs BB) and hybrid crosses (i.e., BA vs AB) using t-test. We comparedpure vs hybrid crosses using a paired comparison test (also see text).

uncommon event, e.g., gynogenesis. Pure species crosses (controls), i.e., GG,RR, and W, were stocked at the same initial densities as hybrids.

Fertilization success was assessed by examining at least 100 oocytes by lightmicroscopy at 60-90 min post-insemination. Fertilization was consideredsuccessful if the oocyte was at or beyond polar body I formation.

Survival to straight-hinge stage (48 h) was deternined by sieving culturesonto a 25 pm mesh screen and counting larvae of normal appearance. Yieldat 48 h was calculated as

(no. of straight-hingeX 100)/no. of eggs incubated

Embryo survival was estimated as the proportion of fertilized eggs that pro-duced larvae at 48 h.

Seawater in the larval cultures was renewed every 2 days as long as larvaewere alive. During these changes, numbers of remaining larvae were esti-mated and shell length was measured for 20 individuals, except in the massspawns of triploid hybrids (described below) where 50 individuals weremeasured.

Production of triploid hybrids

We conducted two experiments to produce triploid C. virginicaXC. gigas( = \rVG ) and one experiment with triploi d C. gigasx C. virginica ( = GGV )

274 S.K.ALLENJT. ETAL.

confirm the production of triploids (Allen and Bushek,1992). All DNA val-ues were calculated relative to a standard, in this case, spenn from a singledwarf surf clam, Mulinia lateralis.

ElectrophoresisSamples of adductor muscle and digestive gland from adult oysters (par-

ents) were homogenized for electrophoresis in grinding buffer (0.1 M Tris-HCI pH 7.0 buffer containing 0.t0/o ftmercaptoethanol). Samples of progeny

from individual families were obtained by centrifugation ( 1500 Xg for 10 s )of dense larval suspensions ( approximately 1 0000-20000 48-h larvae in I ml,fewer larvae in older samples). Each pellet was resuspended, homogenized inan equal volume of grinding buffer, then centrifuged at 5000 g for 6 min toprovide a supernatant for starch gel electrophoresis. Five polymorphic allo-zyme loci stained reliably in adults and larvae: Est-2 (esterase, E.C. 3. 1.1.I ),Gpi (glucosephosphate isomerase, E.C. 5.3.1.9), Lap-1 and Lap-2 (leucineaminopeptidase, E.C. 3.4.11.-'), and Mdh-L (malate dehydrogenase, E.C.1.1. 1.37 ). Both the LiOH buffer system of Selander et al. ( l97l) and Tris-citrate pH 7.0 gels (Gaffney and Scott, 1984) were adequate for all systems.Other enzymes proved difficult to resolve regularly in larvae, or were not suf-ficiently diagnostic to be useful in detecting hybrids.

Feeding experiments

Three separate feeding experiments were conducted.

( 1 ) Fluorescent bead ingestion. On days 2, 4 and 6,about 900 larvae from eachstarved W and VG replicate (see below) were split into 3 groups of about300 each. Larvae were held in small vials containing l0 ml of seawater witheither0.3pm (106/ml), 1.66 pm (10slml), or9.33pm (5x 104/ml) fluores-cent polystyrene beads at room temperature. After 30 min, larvae were fixedby the addition of I ml 100/o formalin and observed with epi-fluorescent mi-croscopy. The proportion of larvae with fluorescent beads ingested was scored.

(2) Fed and starved VG.Each of three replicates ofboth W and VG was sub-divided into a fed and a starved group, forming 12 groups: Wr"o (l-3),W.t"*"d ( I -3 ), VGr"a ( l-3 ), VG.r.*"a ( I -3 ). Fed groups were provided withIsochrysis aff. galbana (clone t-ISO or clone c-ISO) and Chaetoceros calci-trans at feeding densities of 30000 cells/ml daily. For starved groups, incom-ing water was filtered through a I pm cartridge filter and no food was added.Growth and survival were monitored as described above.

(3) Fed and starved RV.Each of three replicates of both RR and RV was sub-divided into a fed and a starved group, forming 12 groups as above. Fed groups

276 S.K. ALLEN Jr. ET AL.

TABLE 3

Paired comparison tests (see Materials and Methods) of fertilization success and yield of larvae at 48h in matings of C. virginicawith C. gigas and C. rivularis. Consensus tests were performed on individ-ual probabilities from paired comparisons. ND: no data

Experiment Fertilization Yield

Probability Consensus Probability Consensus

GG vs GV

WvsVG

RR vs RV

VV vs VR

0.00020.055

0.0170.s240.48 r

0.606ND0.330

0.425ND0.038

0.0001

0.2474

0.4s20

0.0830

0.00150.205

0.0730.1870.344

0.0037

0.0569

0.0054

o.0423

I2

I23

I2

J

I2

3

0.lll0.0080.217

0.6040.0700.038

tion, with individual females ( replicates ) representing the random factor andmale species, the fixed factor. For analysis of larval growth, we used the meanlarval size in each replicate at each time period to avoid pseudoreplication.Probability values from individual experiments were combined to yield con-sensusprobabilities (Rice, 1990) foreachcase (Table 3).

RESULTS

Fertilization rate and 48 h survival

C. virginicaX,C. gigasMean fertilization rate in the pure crosses was high: 860/o in W and 870/oin

GG (Table 2).Fertilization rate in diploid matings of C. virginica and C.gigas showed no significant difference among either pure crosses (GG vs VV,t:0.13, d,.f .-29,P:0.90 ) or between reciprocal hybrids (GY vs VG, /= 0.85,d.f.=29, P:0.40). Analysis of mean fertilization rates by paired compari-sons indicated fertilization was lower in GV matings than GG matings (con-sensus P:0.0001, Table 3). There was no detectable difference in fertiliza-tion success between W and VG (Table 3).

Yields were similar in the pure crosses (t=0.62, d.f.=38, P:0.54) andreciprocal hybrids (r=0.58, d.f.=41, P:0.56); however, hybrid yields werealways lower than yields from pure crosses. This was statistically significant

278 S-K. ALLEN Jr. ET AL.

DAY

5 7 9 ll t3

DAY

Fig. 2. Mean survival (of all replicates, except VG16) of hybrid larvae and their respective controlsfrom day 2 to day 12.Left- C. virginicaxC. gigasiyV (I), GG (0), GV (A ), VG (V ). Right -C. virginicaXC. rivularis:YY (1t ), RR (t), RV (A ), VR (V ).

1trIN.A

FINv)

7

DAY

7

DAY

Fig.3. Meansize (overall replicates, exceptVGl6) ofhybridlarvae andtheirrespectivecontrolsfromday 2 to day 12. Left - C. virginicaxC. gigas:YY (f ), GG (l), GV (A ), VG (V ). Right - C.

virginicaxC. rivularis: vV (f ), RR (O ), Rv ( A ), vR ( V ).

Hybrid larvae increased in size only about 100/o from the straight-hinge stage.Hybrid larvae whose mother had the larger egg were always slightly larger. Ingeneral, pure crosses grew normally to 150-200 pm by the time hybrids hadexpired, whereupon cultures were tenninated.

TriploidsThe mortality patterns observed in WG and GGV larvae were essentially

identical to those seen in diploid VG and GV larvae (above). In the mass

I3579il13

280 S.K. ALLEN Jr. ET AL.

groups were significantly different, but again at each period the triploids werelarger (Table 4). However, a consensus test was significant (P:0.014) inthis case. In the mass spawn of GGV hybrids, Mann Whitney U-tests revealedtriploids to be significantly larger at every time period (Table 4).

Genetic confirmation

Flow cytometryThe proportion of polyploids was estimated by analyzing cell populations

obtained from the dissociation of a subsample of cultured larvae. Inductionof triploid hybrids was successful in all attempts, including the mass spawns(Table 5 ). Percent triploidy in WG averaged 770/0, ranging from 38 to 980/0,

and averagedT60/o in GGV, ranging from 65 to 850/0. Frequently,low propor-tions of tetraploid cells were observed in the triploid hybrids; some triploidcells were observed in diploid hybrids.

We also estimated DNA content, standardized against Mulinia lateralisspenn, of 48-h-old larval cultures in hybrid and pure crosses. Mean relativeDNA content of reciprocal hybrids was intermediate to the parental species(Fig. a). Comparison limits of all hybrids overlapped with those of the paren-tal species (Fig. a). In view ofthe relatively small differences in DNA contentamong species and the variability observed among replicates, it appears that

TABLE 5

Percent polyploidy in cells dissociated from pooled samples of48-h-old larvae in triploid crosses be-tween C. virginica (V) and C. gigas (G), and their diploid hybrid controls. Female of the cross islisted first. Rep #-replicate number; M=total mortality or too few larvae to sample; NS=not sampled

Cross Rep # o/o 3N o/o 4N Cross Rep # o/o 3N o/o 4N

vGl2

3

45

67

8

9

10

Mass

0NS06

24NS22J

2

0NS0000NS0000

95M90767593M9881

3849

0M

0t4

5

2

M0009

vvc I2

3

45

6

7

8

9

l0Mass

GGV MMt4

9

7

IM2M379485

Mass 65

INS2NS3046

Mass 3

GV NSNS000

282 S.K. ALLEN Jr. ET AL.

Progeny (pooled larvae)

Day 1 Day 3 Day 5 Day 7

-I

-II

II

Fig. 5. Representative zymograms of pooled larval samples from a single pair mating of C. gigas(female)xC. virginica at l, 3, 5 and 7 days after fertilization. Staining intensity was typically weakby day 7.".

Paternal gene expression in GV larvae at 48 h was not evident atthe Mdh-1 locus; Mdh-2 stained weakly but was not diagnostic. Paternal expressionwas seen at Est-1in all replicates except GV7, which showed only the maternalband. Lap-l and Lap-2 were poorly resolved in these larvae, but appeared toexhibit solely maternal patterns.

Pooled GV larvae from an additional cross were scored at l,'3, 5, andTdays after fertilization (Fig. 5 ). Paternal gene expression was evident for Lay1 at days 3 and 5, and for Gpi at day 3, but was not observe d for Mdh- I at anytime. Other loci did not stain clearly.

C. virginicaXC. gigas. Larvae from four crosses were examined. In cross VG1,pooled 24-hlaruae showed only the maternal genotype at Lap-1, Lap-2, Est-2,Gpi,andMdh-L.

In crosses YG2, VG3, and VGo, larvae were sampled at days 2, 4, and 6(Fig. 6). Cross VGo was poorly resolved and stained faintly after day 2.Theabsence of staining in later larval life probably reflected the small numbersand poor condition ofhybrid larvae rather than the lack ofgene expression.The Mdh-l locus showed the paternal allele and the hybrid heterodimer bandon day 4 in all three crosses. These proteins may have had reduced stabilityin the hybrid larvae, because they were absent from the gels when the same

284 S.K. ALLEN Jr. ET AL.

C. rivularisXC. virginica.Larvae from six crosses were examined at 48 h. Allshowed hybrid patterns at Est- 1 and Est-2 and Gpi. Other loci were not clearlyresolved.

Feeding trials

Fluorescent bead ingestionTo get some idea of why hybrids were not growing, we conducted several

feeding tests. On the second day of culture there was no difference in the ap-parent ability of W and VG larvae to capture particles (Table 6). By day 4,some larvae in pure crosses captured fluorescent beads in the largest size class(9.3 pm), but no hybrids did. Finally, by day 6,23o/o of larvae from the purecross captured the largest beads, while nearly all captured smaller ones. Incontrast, only half of the hybrids were feeding, and none on the largest beads,on day 6.

Starved VG and RVStarved W crosses grew more slowly than fed W ones, but apparently

were able to extract a reasonable ration from water nominally filtered to I pm(Fig. 7 left). For VG cultures, however, there was no difference in the size offed or unfed larvae throughout the experiment. The same was the case for RVcultures: both RV1"6 and RY.,u*g6 gr€w at identical rates (Fig. 7 right). Forthis experiment, water was filtered to 0.4 pm (bacterial filter); that all foodwas removed was indicated by the lack of growth in the starved pure cross,RR.

TABLE 6

Mean percent (3 replicates) oflarvae ingesting fluorescent beads ofvarious sizes at days 2, 4, and 6

of starvation in hybrid and control crosses of C. virginica (V) and C. Sigas (G)

Day:

Cross:

Beaddensiry(no./ml)

VGVGvvVG

Beadsize(pm)

9't 97 4797 95 520230

94 9894 973 l1

0.31.7

9.3

106

105

5x 104

9393

2

286 S.K. ALLEN Jr. ET AL.

Larval growth and survival was remarkably similar among all hybrid cul-tures. Survival was characterized by the gradual attrition of otherwise healthylooking larvae, beginning at about day 6 and ending in complete mortalitybetween days 10 and 12. Growth was characterized. by an initial increase inlength quickly reaching an asymptote. After day 6, larvae grew no more. Thislack of growth was apparently not due to an inability to capture food particles.Our experiments with fluorescent beads demonstrated that, initially, hybridscaptured beads as successfully as pure crosses. However, by day 6, only about500/o of hybrid larvae could capture even the smallest beads, corresponding toan overall decrease in activity. Starved and fed hybrid larvae grew at identicalrates even though fed hybrids were feeding (data not shown).

We hypothesize that larvae grow until they run out of energy reserves avail-able from the egg. At that point, despite their apparent ability to ingest parti-cles, they can not assimilate them. If this is true, then production of non-viable hybrids of this sort might be useful in the study of conditioning in bi-valves. For example, effects of conditioning diet on quantity and quality ofnutrients in the egg might be quantified by documenting the early growth rateor maximum size of inviable hybrid larvae.

Hybrid triploid larvae sampled were on average larger than their respectivediploid controls, although this difference was statistically significant only inGGV cultures. We hypothesize that early development is not subject to cellnumber regulation as is apparently the case in adult triploids (Small and Ben-fey, 1987; Aliah et al., 1990 ). Therefore, given that triploids have a larger cellsize (to maintain a constant nuclear to cytoplasmic volume ratio), we mightexpect triploid larvae to be larger. Although the effects of polyploidy on nu-clear and cell volumes have been studied extensively in other organisms(Bungenberg de Jong, 1957), we are not aware of studies documenting cellsize and number in early developmental stages of triploid bivalves.

We presented genetic evidence that larvae formed from these matings werehybrid, using flow cytometry and starch gel electrophoresis. Both techniqueshave their limitations primarily because they were employed on larval popu-lations, that became increasingly moribund, rather than individuals. For flowcytometry, mean DNA contents of all samples from hybrid cultures were in-termediate to means from pure cultures, but there was variation among cul-tures. The second use for flow cytometric data was to detect excessive varia-tion within a sample of larvae from a particular culture. Such variation mightindicate a cryptic population of cells of uniparental origin, or aneuploidy. Noclear evidence was found for increased variation although GY cultures hadsignifrcantly higher CVs than GG controls. But here also there was limitationto the resolution of internal variation. Besides experimental artifacts, such as

machine error and differential staining, variation in apparent DNA contentcould arise within a sample because of the presence of dead or dying larvae orthe heterogeneity of cell types in crude larval preparations.

288 S.K. ALLEN Jr. ET AL.

gas, Allen and Gaffney, 1993) may still play an important role in oysteraquaculture.

SUMMARY

( I ) For interspecific matings of C. virginica with C. gigas and C. rivularis,barriers to artificial fertilization are slight, but without exception progeny areinviable.(2 ) Hybrid larvae appear normal and can ingest food, but nevertheless ceasegrowing and begin to die approximately I week after fertilization.(3) Contamination of cultures will always be a problem in the hatchery set-ting. Genetic conflrrmation of hybrids is essential.(4) With regard to the species examined here, previous reports of successfulhybridization should be questioned.(5) Lack of hybridization between C. virginica and C. gigas has bearing onthe question of non-native introductions. Introduction of C. gigas to the na-tive range of C. virginica will not have direct genetic effects on C. virginica.

ACKNOWLEDGEMENTS

We thank Greg DeBrosse and Sandra Scarpa for technical help. This re-search was supported by the Northeast Regional Aquaculture Center throughgrants 88-38500-4070 and 89-38500-4356 from the Cooperative State Re-search Service, U.S. Department of Agriculture and grant NA90AA-D-FM46Ifrom the National Marine Fisheries Service Oyster Disease Research Pro-gram to PMG and SKA. Any opinions, findings, conclusions, or recommen-dations expressed in this publication are those ofthe authors and do not nec-essarily reflect the view of the U.S. Department of Agriculture. NJAESpublication No. D-32100-2-92 and contribution #92-15 of the Institute ofMarine and Coastal Sciences, Rutgers University.

REFERENCES

Aliah, R.S., Yamaoka, K., Inada, Y. and Taniguchi, N., 1990. Effects of triploidy on tissuestructure of some organs in ayu. Nippon Suisan Gakkaishi, 56: 569-57 5.

Allen, Jr., S.K. and Bushek, D., 1992. Large scale spawns of triploid Crassostrea virginica( Gmelin ) using "stripped" gametes. Aquaculture, I 03: I - 1 I .

Allen, Jr., S.K. and Gaffney, P.M., 1993. Genetic confirmation of hybridization between Crass-ostrea gigas (Thunberg) and C. rivulans (Gould ). Aquaculture, I 13: 291-300.

Allen, Jr., S.K., Downing, S.L. and Chew, K.K., 1989. Hatchery Manual for ProducingTriploidOysters. University of Washington Press, Seattle, WA,27 pp.

Breese, W.P. and Malouf, R.E., 1977. Hatchery rearing techniques for the oyster Crassostrearivularis Gould. Aquaculture, 12:. 123-126.

Bungenberg de Jong, C.M., 1957. Polyploidy in animals. Bibliographical Genetica, 27: lll-228.