Amer. J. Bot. 75(4): 589-593. 1988. SHORT COMMUNICATION APHIDS LIMIT FECUNDITY OF A WEEDY ANNUAL (RAPHANUS SATIVUS)1 ALLISON A. SNOW2 AND MAUREEN L. STANTON Botany Department, University of California, Davis, California 95616 ABSTRACT Few previous studies document effects of herbivores on the reproduction of wild plants in situ. We examined the impact of aphids' on seed production in wild radish (Raphanus sativus L.), an annual herb. Aphid infestation increased during the three-month flowering period. Flower and fruit production declined during the season, in part due to aphids. Inflorescences of late? blooming plants more than doubled their fruit production when aphids were removed. Thus, aphids curtailed the blooming period of wild radish, perhaps conferring a selective advantage on early-flowering plants. A few individuals were not susceptible to aphid colonization. HERBIVORES cause dramatic reductions in the yield of agricultural species, yet few previous studies concern their effects on the fecundity of wild plants. Plant reproductive success is obviously affected when severe damage results in mortality (e.g., Fedde, 1973; Schmitt and Antonovics, 1986). More often, herbivores merely reduce plant growth, usually leading to a decrease in fecundity (Rausher and Feeny, 1980; Louda, 1984; Marquis, 1984; Whitham and Mopper, 1985; but see Hendrix, 1979; Paige and Whitham, 1987). The detrimental effects of sucking insects, such as aphids, are generally less visible than the damage caused by leaf-chewing ?herbivores. However, these sap-feeders can severely reduce plant growth and reproduction (in agricultural systems: Kennedy and Stroyan, 1959; Harper, 1963; Banks and Macaulay, 1967; Dixon, 1971a, b; Vereijken, 1979). Further ecological studies are needed to un? derstand how herbivory affects lifetime repro? duction in wild plants, thereby promoting evo? lutionary change. Relatively little is known about the impact ofsap-feeding insects on wild species. Here we describe the effects of aphid feeding and variation in flowering time on some fitness components of Raphanus sativus. 1 Received for publication 30 March 1987; revision ac? cepted 16 October 1987. We thank Dr. W. H. Lange ofUC/Davis for identifying aphids, H. Hasbrouck and C. McGee for technical assis? tance, and R. Nakamura, M. Rausher, and T. Whitham for reviewing a draft ofthis paper. This work was supported by NSF grants BSR 84-11077 to AAS and DEB 82-14508 to MLS. 2 Current address: Smithsonian Environmental Re? search Center, PO Box 28, Edgewater, MD 21037. MATERIALS AND METHODS- Wild radish is a cosmopolitan annual that became naturalized in California in the late 1800's (Panetsos and Baker, 1967). This economically important weed is largely self-incompatible, relying on insect vectors for pollen transfer and seed set. Flowers open sequentially on indeterminately growing branches. Local pollinators include honeybees, lepidoptera, syrphid flies, and other insects (Stanton, 1987a). In this and several previously studied California populations, seed production was not limited by pollinator ser? vice (Stanton, 1987b). We studied patterns of flowering and fruit set during the 1983 growing season (March? May). Field work was conducted at a popu? lation of several thousand wild radish plants growing in an abandoned field at the University ofCalifornia at Davis, CA. (See Stanton, 1987a, for description of "Arboretum" site.) The 1? 1.5 m tall plants each had several hundred inflorescences (also referred to as flowering branches) that began producing fruits in mid? March. Aphids did not appear until later in the season, so their effects on plant reproduc? tion were determined using plants that reached peak flowering in late April. These plants will be referred to as late-flowering plants. Prior to the arrival of aphids, we measured fruit set on 18 haphazardly chosen plants that flowered early in the season. Starting on 19 March and 20 April we labelled all open flowers on each of 3 randomly selected flowering branches per plant. This procedure was re? peated 3 times over a period of 10 days. flow? ers were marked by attaching a small adhesive label to the stem section below each pedicel. Fruit set from these flowers was recorded ap- 589 590 AMERICAN JOURNAL OF BOTANY [Vol. 75 proximately 1 month later. The frequency of naturally occurring aphid colonies on these in? florescences was noted on 12 April and 20 April. By late April, when most early-flowering plants had stopped producing flowers, aphids were common on many later flowering indi? viduals. Ten such plants were haphazardly chosen for an aphid removal experiment. It was not possible to exclude aphids from entire plants without also affecting pollinators, so aphids were continually removed from a subset ofbranches on each plant. The late plants had just begun setting fruit and each flowering branch had many remaining buds. All natu? rally occurring aphids were manually removed from 8 branches per plant on 26 April. On a random sample of 4 of these branches, any aphids that appeared were removed daily. Aphids were reintroduced to each of the re? maining 4 branches by cutting off an aphid? infested inflorescence (usually from the same plant) and attaching it to the experimental in? florescence with a wire twist-tie. Ifaphids failed to move onto the new branch, this procedure was repeated. By 4 May, colonies of at least 20 aphids had been established on all 4 aphid? treatment branches of 7 plants. Three plants had few aphids on any branches, even after repeated introductions, and data from these plants were not included in analyses of treat? ment effects. Beginning 1 May, the youngest flower on each experimental branch was la? belled weekly until flower production ceased. Fruit set from flowers that opened during these intervals was recorded on 25 May, when most fruits had reached their final size. Branches were then collected for counts of seed number per fruit. A portion of the fruits fell off before their seeds were counted, so sample sizes for seed counts are less than those for fruit set measurements. (A total of 158 fruits were col? lected.) Data were analyzed using nonparametric statistical tests because variances were not ho? mogeneous and distributions were not normal. To test for effects of aphids on plant repro? duction, we averaged data from 4 inflores? cences per treatment on each plant. For flower, fruit, and seed counts, the paired means from 7 plants were compared using Wilcoxon's signed ranks test, which is analogous to a paired t test (Sokal and Rohlf, 1981). Percent fruit set of individual plants was analyzed using log? linear frequency analysis (CATMOD proce? dure in SAS; Freund and Little, 1981). RESULTs-Seasonal changes injloweringand fruit set-Flowering of individual plants was staggered over the blooming period, and early plants were not heavily colonized by aphids (see below). The 18 early study plants ceased flowering by late April, when individuals in the later study group had reached their peak. Phe? nological differences among co-occurring wild radish plants have also been noted in other years (personal observation). Fruit set from early- and late-blooming plants reached over 70%, but on the late-blooming plants fruit set declined rapidly over time. On the early plants, 75% of361 flowers that opened on 19-29 March set fruit. One month later, fruit set on these plants was still 64% (N = 437, April 20-29). On uninfested branches of late? blooming plants, fruit set declined from 71 % (N = 222) on May 1-4 to 20% (N = 236) on May 13-19. This decrease in fruit set corre? sponded to increasingly dry soil conditions (personal observation). Stanton (1987b) ob? served a similar decrease in fruit set in 1984, and showed that this was not due to insufficient visitation by pollinators. Effects ofaphids-As in other years, aphids were not observed at the beginning ofthe flow? ering season. By late March, they were seen feeding on the buds of a few inflorescences. Common aphid species were turnip, cabbage, and potato aphids (Hyadaphis erysimi, Brevi? coryne brassicae, and Macrosiphum euphor? biae). The frequency of aphids on branches of early plants increased from 50% on 12 April to 78% on 20 April (N = 54 branches on 18 plants; most parasitized branches had 5-10 aphids). By May, much larger colonies of > 30 aphids were common, often feeding on both buds and young fruits of late-flowering plants. Daily inspections were necessary to maintain branches free of aphids. Even at the height of aphid abundance, however, 3 of the 10 late study plants were not colonized. Removal ofaphids led to a dramatic increase in fecundity. Aphids inhibited flower produc? tion as well as the proportion of flowers setting fruit (Fig. 1, Table 1). During the first 3 days of aphid establishment, flower production on branches with aphids was not significantly dif? ferent from that on aphid-free branches (Fig. 1). Initial fruit production (prior to abortion) was also similar (6.5 vs. 6.8 fruits per inflo- rescence). After 4 May, aphid removal caused a 46% increase in the number of flowers pro? duced (Fig. 1), and an 80% increase in percent fruit set. Combining these effects, we see that removing aphids resulted in 2.4 x more fruits per branch after 4 May (Fig. 2). Individual plants showed significant variation in overall fruit set, and the interaction term from log? linear frequency analysis was almost significant April 1988] SNOW AND STANTON- FECUNDITY OF RAPHANUS 591 261912 ///////+-------i----------t 4MAY I a::: w? CD ~ :::> z 15 I <..) z ? a::: CD a:: w 10 a.. en I- ::> a::: l.J.. ~ 5 261912 .. / / ..- .. / / ..- .. / / . .. ,. ~ /,-1---------? ./ ./ ./ ./ ./ ./ ~APHtOS FULLY ESTABLISHED 4MAY t 30 :I: U Z z FLOWERING DATE FLOWERING DATE Fig. 1. Effects ofaphids on the total number of flowers per inflorescence. Branches without aphids (solid line) had significantly more flowers and fruits than those with aphids (dashed line; P < 0.005; N = 7). Means ?SE are shown for 7 plants, using the mean of 4 inflorescences for each plant. Fig. 2. Effects of aphids on the total number of fruits. Branches without aphids (solid line) had significantly more flowers and fruits than those with aphids (dashed line; P < 0.005; N = 7). Means ?SE are shown for 7 plants, using the mean of 4 inflorescences for each plant. (Table 1). Individuals were statistically ho? mogeneous in their response to aphid herbiv? ory (P < 0.08). Aphid feeding had no effect on the number ofseeds per fruit (Wilcoxon's signed rank test; both treatments averaged about 3 seeds per fruit). Seed set data were also analyzed using the General Linear Model procedure of SAS (Freund and Little, 1981), but even when vari? ation due to maternal plant and flowering date was accounted for, aphid effects were not sig? nificant. DISCUSSION- Naturally occurring aphid her? bivores can significantly reduce the fecundity of wild radish. Aphid feeding limited female reproductive success by reducing flower and fruit production. Male-based reproduction probably suffered also because few flowers were available to act as pollen-donating organs, and pollen from these flowers was less likely to sire seeds. Models of sexual reproduction in plants often assume that paternal success will increase with pollen and/or flower production (Lloyd, 1984)~Thereare few data to support this seem- ingly reasonable assumption, but Schoen and Stewart (1986) showed that paternal success in white spruce increased with production ofmale cones. Because wild radish is an outcrossing annual, we expect that decreases in flower? pro? duction strongly influence both male and fe? male fitness of plants subjected to different levels of infestation. The plant-to-plant variation in aphid attack seen in this study can be attributed to several factors. First, individuals flowering before aphid populations reached their peak escaped the brunt offeeding that occurred later. Second, susceptibility to aphid colonization varied among synchronously blooming neighbors, as we could not establish colonies on three out of ten late-blooming plants. Resistance to aphids could be due to environmental and/or genetic factors (Maddox and Cappucino, 1986). To the extent that variation in flowering time and re? sistance have a genetic basis (e.g., Harper, 1964; Murfet, 1977; McIntyre and Best, 1978), and ifthese traits are additively controlled (Eenink, Dieleman, and Groenwold; 1982), herbivores could act as a strong selective agent in natural populations. For example, in wild parsnip ear? ly-flowering genotypes experienced less her? bivory than did late-flowering genotypes (Ber? enbaum, Zangeri, and Nitao, 1986). Schemske (1984) showed that seed predation within for- 592 AMERICAN JOURNAL OF BOTANY [Vol. 75 TABLE 1. Effect ofaphids onfruit set ofindividual plantsa Aphids removed Aphids presentPlant num- Proportion (Number of Proportion (Number of ber set flowers) set flowers) 1 0.33 (70) 0.02 (43) 2 0.23 (88) 0.07 (59) 4 0.60 (105) 0.41 (68) 5 0.36 (77) 0.19 (64) 7 0.28 (74) 0.13 (66) 8 0.26 (86) 0.24 (41) 9 0.20 (89) 0.19 (70) Total 0.33 (589) 0.19 (411) Source df X square P Individual plant 6 65.24 0.0001 Aphid treatment 1 23.19 0.0001 Plant x treatment 6 11.26 0.081 a Frequencies offruit set after 4 May were analyzed using the log-linear method in the CATMOD procedure ofSAS in a 7 x 2 x 2 design. Plants #3, 6, and 10 were not included in this analysis because aphids could not be es? tablished on them. est habitats selected for earlier flowering eco? types of Impatiens pallida. Flowering time has been shown to be under additive genetic con? trol in another radish species, Raphanus raph? anistrum (Mazer, 1987), which is closely re? lated to R. sativus (Panetsos and Baker, 1967). Genetically distinct families ofR. sativus, also vary in phenology (M. Stanton and H. Young, unpublished). We suggest that selective pres? sures due to both climatic constraints and her? bivores may influence the optimum flowering time for wild radish. In this study, effects of aphids were dem? onstrated experimentally by removing aphids from a random subset of branches on infested plants. One problem with having two different herbivory treatments on the same plant is that resource-sharing among branches could con? found our results. (As described earlier, it was impossible to apply treatments to separate plants.) Ifbranches with aphids used more en? ergy and nutrients than uninfested branches, we may have underestimated the impact of aphids. We suspect that even greater effects might be seen if aphids were removed from entire plants, which would presumably be healthier as a result. On the other hand, ifaphid feeding caused unused resources to be allocated to uninfested shoots, the effect of aphids was overestimated. This could occur if aphids did not constitute a strong metabolic sink, or if feedipg impaired normal meristem develop? ment through hormonal or other effects. Un? fortunately, the relative amount of resources used by aphids and uninfested shoots is not known. In many species, however, aphid feed- ing represents a major sink, depriving other parts of the plant of resources (Kennedy and Stroyan, 1959). We saw no evidence of me? chanical damage (e.g., necrosis) at aphid feed? ing sites, but possible effects on morphogenesis cannot be ruled out. In summary, we think it is unlikely that resources were redirected from branches attacked by aphids to those without them. Morrow and LaMarche (1978) used a similar experimental design to show that chronic her? bivory depressed growth in subalpine Euca? lyptus trees. They compared growth ofbranch? es that were sprayed with insecticide with the growth ofunsprayed branches on the same trees. Sprayed branches grew faster than unsprayed ones, and both showed much better growth than branches of untreated trees. Two conclu? sions were drawn. First, sprayed branches ex? ported photosynthate to other parts ofthe tree, improving subsequent growth ofbranches that were exposed to herbivores. Second, since the average growth of sprayed and unsprayed branches combined was significantly greater than growth of untreated trees, the damaging effect of herbivores was not simply an artifact of using branches rather than individuals as experimental units. Within-plant comparisons underestimated the impact of phytophagous insects. Whenever possible, future investi? gators should assign different treatments to dif? ferent individuals to avoid uncertainty about the role of resource-sharing among branches. LITERATURE CITED BANKS, C. J., AND E. D. M. MACAULAY. 1967. Effects of Aphis fabae Scop. and of its attendant ants and insect predators on yields of field bean (Vicia faba L.) Ann. Appl. BioI. 60: 445-453. BERENBAUM, M. R., A. R. ZANGERI, AND J. K. NITAO. 1986. Constraints on chemical coevolution: wild par? snips and the parsnip webworm. Evolution 40: 1215? 1228. DIXON, A. F. G. 1971a. The role of aphids in wood formation. I. The effect of the sycamore aphid, Drep? anosiphumplatanoides (Schr.) (Aphidae), on thegrowth of sycamore, Acer pseudoplatanus L. J. Appl. Ecol. 8: 165-179. --. 1971 b. The role of aphids in wood formation. II. The effect of the lime aphid, Eucallipterus tiliae L. (Aphidae), on the growth of lime, Tilia x Vulgaris Hayne. J. Appi. Ecoi. 8: 393-399. EENINK, A. H., F. D. DIELEMAN, AND R. GROEWOLD. 1982. Resistance of lettuce (Lactuca) to the leaf aphid Na? sonovia ribis nigri. 2. Inheritance of the resistance. Euphytica 31: 301-304. FEDDE, G. F. 1973. Impact of the balsam wooly aphid on 'cones and seed production in infected Fraser fir. Canad. Entomol. 105: 673-680. FREUND, R. J., AND R. C. LITTELL. 1981. SAS for linear models. SAS Institute, Cary, NC. HARPER, A. M. 1963. Sugar beet root aphid, Pemphigus April 1988] SNOW AND STANTON - FECUNDITY OF RAPHANUS 593 betae Doane (Homoptera: Aphidae), in southern Al? berta. Canad. Entomol. 95: 863-873. ---. 1964. Varietal resistance of sugar beets to the sugar beet root aphid Pemphigus betae Doane (Ho? moptera: Aphidae). Canad. Entomol. 96: 520-522. HENDRIX, S. D. 1979. Compensatory reproduction in a biennial herb following insect defloration. Oecologia (Berl.) 42: 107-118. KENNEDY, J. S., AND H. L. G. STROYAN. 1959. Biology of aphids. Ann. Rev. Entomol. 4: 139-160. LLOYD, D. G. 1984. Gender allocations in outcrossing cosexual plants. In R. Dirzo and J. Sarukhan [eds.], Perspectives on plant population ecology. Sinauer, Sunderland, MA. LOUDA, S. M. 1984. Herbivore effects on stature, fruiting, and leaf dynamics of a native crucifer. Ecology 65: 1379-1386. MADDOX, G. D., AND N. CAPPUCCINO. 1986. Genetic determination of plant susceptibility to an herbivo? rous insect depends on environmental context. Evo? lution 40: 863~866. MARQUIS, R. J. 1984. Leafherbivores decrease fitness of a tropical plant. Science 226: 537-539. MAZER, S. J. 1987. The quantitative genetics of life? characters in Raphanus raphanistrum: ecological and evolutionary consequences of seed weight variation. Amer. Naturalist 130: 891-914. McINTYRE, G. I., AND K. F. BEST. 1978. Studies on the flowering of Thlaspi arvense L. IV. Genetic and eco? logical differences between early- and late-flowering strains. Bot. Gaz. 139: 190-195. MORROW, P. A., AND V. C. LAMARCHE, JR. 1978. Tree ring evidence for chronic insect suppression of pro? ductivity in subalpine Eucalyptus. Science 210: 1244? 1246. MURFET, I. C. 1977. Environmental interaction and the genetics of flowering. Ann. Rev. PI. Physiol. 28: 253? 278. PAIGE, K. N., AND T. G. WHITHAM. 1987. Overcom- pensation in response to herbivory: the advantage of being eaten. Amer. Naturalist 129: 406-417. PANETSOS, C. P., AND H. G. BAKER. 1967. The origin of variation in "wild" Raphanus sativus (Cruciferae) in California. "Genetica 38: 243-274. RAUSHER, M. D., AND P. FEENY. 1980. Herbivory, plant density, and plant reproductive success: the effects of Battus phi/enor on Aristi/ochia reticulata. Ecology 61 : 905-917. SCHEMSKE, D. W. 1984. Population structure and local selection in Impatiens pallida (Balsaminaceae), a self? ing annual. Evolution 38: 817-832. SCHMITT, J., AND J. ANTONOVICS. 1986. Experimental studies of the evolutionary significance of sexual re? production. IV. Effect ofneighbor relatedness and aphid infestation to seedling performance. Evolution 40: 830? 836. SCHOEN, D. J., AND S. C. STEWART. 1986. Variation in male reproductive investment and male reproductive success in white spruce. Evolution 40: 1109-1121. SOKAL, R. R., AND F. J. ROHLF. 1981. Biometry. W. H. Freeman, New York. STANTON, M. L. 1987a. The reproductive biology ofpetal color variants in wild populations ofRaphanus sativus L. I. Pollinator response to color morphs. Amer. J. Bot. 74: 178-187. ---. 1987b. The reproductive biology of petal color variants in wild populations of Raphanus sativus L. II. Factors limiting seed production. Amer. J. Bot. 74: 188-196. VEREIJKEN, P. H. 1979. Feeding and multiplication of three cereal aphid species and their effect on the yield of winter wheat. Agricultural Research Reports 888: 1-58, Centre for Agricultural Publishing and Docu? mentation, Wageningen, Netherlands. WHITHAM, T. G., AND S. MOPPER. 1985. Chronic her? bivory: impacts on architecture and sex expression of pinyon pine. Science 228: 1089-1091.