Interactions Within Pairs of Biological Control Agents on Water Hyacinth, Eichhornia Crassipes ⇑ Danica Marlin A, , Martin P

Biological Control 67 (2013) 483–490

Contents lists available at ScienceDirect

Biological Control

journal homepage: www.elsevier.com/locate/ybcon

Interactions within pairs of biological control agents on water hyacinth, Eichhornia crassipes ⇑ Danica Marlin a, , Martin P. Hill a, Marcus J. Byrne b a Department of Zoology and Entomology, Rhodes University, P.O. Box 94, Grahamstown 6140, South Africa b School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Private Bag X3, Wits 2050, Johannesburg, South Africa highlights graphical abstract

Water hyacinth was exposed to pair- wise combinations of three biocontrol agents. + = SYNERGISTIC The mite O. terebrantis alone had the EFFECT least impact on plant growth. The mirid E. catarinensis and weevil N. eichhorniae each had negative + = EQUIVALENT EFFECT impacts on plant growth. The mite performed poorly in the presence of mirids and weevils. EQUIVALENT + = EFFECT The mite possibly enhanced the performance of the other two agents. article info abstract

Article history: Water hyacinth, Eichhornia crassipes, is an invasive aquatic plant in South Africa where seven biological Received 3 May 2013 control agents have been released on the weed. Combined herbivory by these multiple agents may cause Accepted 11 October 2013 greater damage than any of the agents acting alone. This study examined the effects of herbivory by the Available online 21 October 2013 water hyacinth mite Orthogalumna terebrantis, the mirid Eccritotarsus catarinensis, and the weevil Neoche- tina eichhorniae, singly or in paired combinations, on the weed’s growth. Plants were subjected to herbiv- Keywords: ory and plant growth parameters e.g. production of leaves, and the percentage of the leaf surface area Herbivory damaged by herbivory, were measured every 14 days. Plants subjected to herbivory by mirids only, or Insect interactions weevils only, produced signiﬁcantly fewer ramets than the control (herbivory-free) plants. Plants sub- Invasive weed Multiple agents jected to mirids only, or a combination of mites with weevils, produced the least number of leaves. Plant Plant–insect interactions petioles were shortest on plants subjected to a combination of mites with weevils, and increases in plant biomass were lowest in plants fed on by weevils or the combination of weevils with mirids. The combination of mites with mirids caused the greatest damage to the leaf surface area. The results suggest that different plant growth parameters are not impacted equally by herbivory, but are dependent on the agent or combination of agents causing damage. Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction forts having been initiated against the weed more than 20 years ago (Cilliers, 1991). To date, seven biocontrol agents have been re- Water hyacinth, Eichhornia crassipes (Martius) Solms-Laubach leased on water hyacinth in South Africa (Coetzee et al., 2011). Pre- (Pontederiaceae), is considered to be the worst aquatic weed in vious studies have investigated the interactions of various South Africa (Hill and Cilliers, 1999), despite biological control ef- biocontrol agents of water hyacinth (Delfosse, 1978a; Caunter and Mohamed, 1990; Moran, 2005; Ajuonu et al., 2009). In general,

⇑ Corresponding author. Present address: Department of Zoology and Entomol- these studies found the use of multiple agents increases stress on ogy, University of Pretoria, Private Bag X20, 0028, South Africa. Fax: +27 12 362 the plant, which reduces its growth more so than when a single 5242. agent is used E-mail addresses: [email protected], [email protected] (D. Marlin).

Interspecific interactions influence the performance and fitness Adults of all three species feed on water hyacinth leaves and the of phytophagous insects (Kaplan and Denno, 2007) and also play females of the three species oviposit in the leaf blades (Warner, an important role in structuring insect communities (Denno 1970; Perkins, 1973; Hill et al., 1999), where the mite and mirid et al., 2000). A key manner in which phytophagous insects interact eggs can potentially be removed by the feeding of adult weevils, is through indirect effects involving shared host plants (Kaplan and increasing the possibility of agent interaction. The overall aim Denno, 2007). Herbivore damage may induce plant resistance was thus to examine the interactive effects of O. terebrantis, mechanisms such as the release of toxic secondary metabolites, N. eichhorniae and E. catarinensis on water hyacinth growth. which can indirectly reduce herbivore populations as well as a species preference for that plant (Karban and Baldwin, 1997; 2. Materials and methods Rodriguez-Saona and Thaler, 2005). Insect and mite induced responses of plants to herbivory have been studied in various 2.1. Experimental set-up systems, including water hyacinth, and have yielded a range of conclusions (Karban and Carey, 1984; Agrawal, 1998; Bounfour Healthy, agent-free and undamaged water hyacinth plants of and Tanigoshi, 2001; Coetzee et al., 2007a). similar size were selected from stock cultures of the Agricultural The use of multiple agents occupying different niches to control Research Council – Plant Protection Research Institute (PPRI), Pre- an invasive plant species has been supported both in theory and in toria, South Africa, and used in the experiment. The maintenance of practice (Delfosse, 1978a; Hoffmann and Moran, 1998; Jiménez the plant stock cultures is described elsewhere (Marlin et al., and Balandra, 2007), but also has been criticised as opportunistic 2013). The trial was conducted during summer inside a glasshouse and unscientific (Myers, 1985; Denoth et al., 2002; Crowe and Bou- at the PPRI. The average day and night temperatures in the glass- chier, 2006). A meta-analysis of 51 papers revealed that in only house were 30.72 ± 6.74 °C and 20.65 ± 2.66 °C, respectively. one-quarter of cases was plant performance reduced more by mul- All plants had between four to six leaves at the beginning of the tiple enemies than was predicted from each enemy alone (Ste- experiments. All daughter plants (ramets) and dead matter were phens et al., 2013). However, Denoth et al. (2002) found that in removed from the plants before they were weighed and placed the majority of projects for the biological control of weeds, the like- in pairs into plastic tubs (70 Â 40 Â 35 cm depth). The plants were lihood of control increased as more species of agents were re- held upright by wire rings attached by hooks to the sides of the tub. leased, even though in some instances there might be The tubs were filled with 16 L of water. Potassium nitrate (KNO ) competition between herbivore species on the same plant. In addi- 3 and potassium dihydrogen orthophosphate (KH PO ) were added tion, competition between biocontrol agents used against weeds is 2 4 to each tub as nitrogen (2.5 mg LÀ1) and phosphorous bases not common because the biocontrol agents do not represent the (0.4 mg LÀ1), respectively. These concentrations of KNO and KH full suite of natural enemies present in the country of origin, and 3 2- PO were used because they are representative of the high nutrient because the weed often provides a variety of host niches, which 4 concentrations that are common in many South African fresh may potentially reduce competition for space and food (Denoth water systems (Holmes, 1996). Commercial iron chelate (13% Fe) et al., 2002). also was added to each tub at 1.4 g/16 L water. The water and In comparing the damage caused by multiple agents to that of nutrients in each tub were replaced weekly. individual agents, Hatcher (1995) classified four levels of interac- After four weeks of acclimation to the experimental conditions, tions as follows: (a) synergistic, where the interaction causes a sig- both plants in each tub were randomly selected for infestation nificantly greater reduction in a plant variable than would the with one of the following combinations of agents (namely the mite damage of a single agent, (b) additive, where the interaction causes O. terebrantis, the weevil N. eichhorniae, and the mirid E. catarinen- the same reduction in a plant variable as would the combined sis), as either: (1) only mites, (2) only weevils, (3) only mirids, (4) damage of the agents, (c) equivalent, where the interactions causes mites and weevils, (5) mites and mirids, and (6) weevils and mir- an equivalent reduction in a plant variable as would the damage of ids. No agents were added to the control plants. Each treatment either agent alone, and (d) inhibitory, where the interaction causes was replicated seven times. Prior to introducing the agents, the a significantly lower reduction in a plant variable as would damage weevils and mirids were sexed to ensure a 1:1 sex ratio. Male of the weaker of two agents. The additive classification was ad- and female O. terebrantis are morphologically indistinguishable justed by Turner et al. (2010) so that an interaction is additive if (Perkins, 1973) and were therefore not sexed. In natural field pop- the impact of multiple agents is greater than that of the most dam- ulations, mites occur in approximately equal proportions of adult aging agent acting alone but less than or equal to the added im- females and males (Walter, 2009), so a 1:1 sex ratio was assumed. pacts of each agent acting alone. Mites were added at 150 mites/plant, weevils were added at 2 For water hyacinth, the interactions between the sap-sucking adult pairs/plant and mirids were added at 15 adult pairs/plant. mirid Eccritotarsus catarinensis (Carvalho) (Hemiptera: Miridae) These insect stocking densities were sufficient to cause visible and the petiole-mining weevils Neochetina eichhorniae Warner damage to the plant and reduce some of the plant growth param- and Neochetina bruchi Hustache (Coleoptera: Curculionidae) have eters (Marlin, 2011). In the single-agent treatments, the number of been examined by Ajuonu et al. (2007, 2009) and Weyl and Hill individuals added was doubled, e.g. in the ‘‘only mites’’ treatment, (2012), while the interactions between the leaf-mining mite Ortho- 300 mites were placed onto a plant in order to balance the total in- galumna terebrantis Wallwork (Acari: Galumnidae) and N. eichhor- sect load. A net (mesh size 0.5 Â 0.5 mm) was secured over the top niae, have been examined by Delfosse (1977, 1978a,b). The weevils of each tub to prevent the agents from escaping, and to create the and the mirid were found to be compatible (Ajuonu et al., 2007; same experimental light conditions. Weyl and Hill, 2012), although N. eichhorniae appeared to feed less when combined with the mirid (Weyl and Hill, 2012). Delfosse (1977) showed that the weevils laid more eggs and fed more in 2.1.1. Plant growth parameters the presence of the mites, but he did not examine plant growth. The experiment was conducted over an 11 week period and No previous studies have looked at the interactions between O. plant growth parameters (number of leaves, number of ramets, terebrantis and E. catarinensis. the length of the longest petiole and wet biomass) were measured In this study, the impact of O. terebrantis, N. eichhorniae and E. every two weeks. The change in plant wet biomass was obtained catarinensis on water hyacinth growth parameters was investi- by weighing the plants before and after the experiment, where gated for the agents occurring alone or in paired combinations. the end wet biomass included ramets. This weight was used as a D. Marlin et al. / Biological Control 67 (2013) 483–490 485 measure of the relative growth for each treatment. The growth for the leaf surface area damaged, and then a one-way ANOVA was parameters were averaged for the two plants in each tub to obtain used to test for the differences between the leaves at the end of the a mean response per tub. sample period. Thereafter, a one-way ANOVA was used to test for differences in the surface area damaged by herbivory between 2.1.2. Damage to the leaf surface area the treatments, on each of leaves 2, 4 and 5 separately. Damage The total damage caused by the agents also was measured every on leaf 5 was further analysed; in the paired combination treat- two weeks by recording the percentage of the abaxial leaf surface ments the damage caused by each individual species was noted area damaged on leaves 2, 4 and 5. The total damage on leaf 5 was to reveal which agent was responsible for the majority of the dam- further separated into damage caused by each agent species sepa- age. A one-way ANOVA was used to test for differences between rately to reveal which agent was responsible for the majority of the the damage caused to leaf 5 by the individual agent species. damage. Leaf 5 was chosen because damage is more distinct on older leaves as these have been exposed to herbivory for a longer 2.2.3. Agent performance time. The leaf damage was averaged for the two plants in each The percentage of the surface area damaged by the agents was tub to obtain a mean response per tub. used as an indication of their abundance and a measure of their performance. Three one-way ANOVAs, one for each of the three 2.1.3. Agent performance agent species, were used to determine whether the agents per- Since damage to leaves by agent herbivory reflects the presence formed better when on their own or in combination with another of the agent, and increases as the number of individual agents in- agent; the damage on leaf 5 in the paired treatments was separated creases (Center and Jubinski, 1996), the percentage of leaf surface into damage caused by each species individually and compared to area damaged on leaf 5 by each agent species separately was used the damage caused by that same species in a single-agent treat- to indicate the presence and abundance of that agent species, in ment. To make this comparison, the damage in the single-agent the final week of the experiment. The agent abundance was then treatments was divided by two since double the inoculum of used as a measure of their performance to determine whether each agents had been used in these treatments. For data of the damage agent species performed better when on its own or in combination caused by the mirids, the transformed data were used to run the with either of the other two agents. Additionally, because mirids analyses but the actual percentage damage caused to the leaf sur- are easily seen without handling or destroying the plants, the num- face is presented in the figures. ber of mirids (adults and nymphs) was counted at the end of the experiment and compared between treatments. Mites and weevils 3. Results were not counted directly as the physical disturbance of handling the plants dislodges these agents and thus hinders counting. 3.1. Effect of herbivory on plant growth parameters Wright and Center (1984) found a close relationship between the number of adult weevils on a water hyacinth plant and the number Differences were observed between most treatments (single- of feeding scars on leaves (r2 = 0.72) and therefore the damage to agent treatments i.e. mites, weevils or mirids alone, and multi- the leaf surface area caused by weevil feeding was used as a predic- ple-agent treatments i.e. combinations of mites + weevils, tor for their abundance. Similarly, the percentage of the leaf surface mites + mirids or weevils + mirids, and control treatments) for all area damaged by mite herbivory was taken to be an indication of of the plant parameters measured except the number of leaves mite abundance. produced (Table 1). When compared to the control, significantly fewer ramets were produced by plants fed on by mirids only, fol- 2.2. Statistical analyses lowed by plants fed on by weevils only (F41,7 = 3.35, P = 0.001; Fig. 1A). Similarly, the lengths of the longest petioles were signifi- All data met the requirements of homogeneity (Levene’s test) cantly shorter on plants that had been fed on by weevils only, or a and normality (Kolmogorov–Smirnov test), except the number of combination of weevils and mites (F = 307.27, P = 0.001), mirids counted per plant and the percentage damage on leaf 5 41,7 whereas petioles of plants that had been fed on by only mites or caused by mirid herbivory, and these data were therefore log(x) a combination of mites and mirids were as long as the control transformed. Fisher’s protected least significant difference (LSD) plants’ petioles (Fig. 1B). A summary of the plant growth parame- test was used to separate the treatment means if the differences ters, recorded at the end of the experiment, is presented in Table 2. were significant at the 5% level (Snedecor and Cochran, 1980). Data were analysed using the statistical programme STATISTICA Version 7.0 (ÓStatSoft, Inc., USA). 3.2. Damage to the leaf surface area

2.2.1. Plant growth parameters The percentage of the surface area damaged on the different A one-way ANOVA was used to test for differences in plant leaves of a water hyacinth plants was statistically different growth parameters (number of leaves, number of ramets, length between some of the leaves (2, 4 and 5), and treatments of the longest petiole and wet biomass) between the treatments, at the end of the sample period (at week 11). At the start of the Table 1 experiment all the growth parameters were similar between treat- Differences in plant growth parameters of water hyacinth, between treatments, after ments (no signiﬁcant difference in growth parameters between the 11 weeks when plants were either exposed, or not exposed (control treatment), to treatments, P > 0.05). The overall effect on plant growth, rather pair-wise combinations (herbivory treatments) of the mite Orthogalumna terebrantis, the weevil Neochetina eichhorniae and the mirid Eccritotarsus catarinensis. df = 41 and than the rate of change, was considered most important and thus n = 7 for all treatments. the end point values, measured in the ﬁnal week of the experiment, were analysed. Plant growth parameter SS MS F value P value Number of leaves 24.31 4.05 2.03 0.084 2.2.2. Damage to the leaf surface area Number of ramets 12.14 2.02 3.35 0.001 To test for differences in the surface area damaged by herbivory Length of longest petiole 1843.62 307.27 4.57 0.001 Change in wet biomass 0.63 0.11 2.412 0.043 between leaves 2, 4 and 5, the total damage caused by both agents in the paired combination treatments was pooled to obtain a total Statistical differences (P < 0.05) are highlighted in bold. 486 D. Marlin et al. / Biological Control 67 (2013) 483–490

mites + mirids (F5,33 = 3.122, P = 0.020; Fig. 2). Similarly on leaf 5, a combination of mites + weevils caused the least amount of damage to the leaf, and this was signiﬁcantly less than the damage caused by the mites alone, the mirids alone and a combination of

mites + mirids (F5,33 = 5.617, P < 0.001; Fig. 2). The combination of mites + mirids caused the most damage to the leaf on both leaf 4 and leaf 5 (Fig. 2). When the damage to the leaf surface area on leaf 5 was separated into damage caused by the individual agents, the mirids were responsible for the majority of the damage in both the weevils + mirids and the mites + mirids treatments (mirids were responsible for 25% and 55% of total damage in the weevils + mirids and mites + mirids treatments, respectively; Fig. 3).

3.3. Agent performance in single and paired treatments

The number of mirids did not differ between treatments

(F2,13 = 3.575, P = 0.058), although mirid numbers were slightly higher when combined with the mites (13 mirids) compared to when they occurred singly (7 mirids) or in combination with the weevils (6 mirids). The mites alone caused signiﬁcantly more damage to the surface of leaf 5 compared to when they were com-

bined with the weevils or the mirids (F2,18 = 8.553, P = 002; Fig. 4A). In contrast, the weevils caused similar damage to the leaf surface

in all of the treatments (F2,15 = 0.155, P = 0.858; Fig 4B). In the treatments which included mirids, the mirids alone caused significantly less damage to the leaf surface than when they were combined with the mites, but similar damage when they were

combined with the weevils (F2,15 = 4.443, P = 0.031; Fig. 4C).

4. Discussion

Feeding by mirids or weevils, or a combination of mites and weevils, had a negative impact on water hyacinth ramet production, and this has been observed in previous studies (Heard and Winterton, 2000; Coetzee et al., 2007a). When this and similar studies are compared using Hatcher’s (1995) classiﬁcation, the best control of water hyacinth i.e. synergistic effect, is provided when Fig. 1. The effect of herbivory by combinations of three agents (mite Orthogalumna terebrantis, weevil Neochetina eichhorniae and mirid Eccritotarsus catarinensis)on the two Neochetina weevils act together, or when O. terebrantis acts the total number of ramets produced (A) and on the lengths of the longest petioles with either N. eichhorniae or E. catarinensis (Table 3). In one case (B) of water hyacinth plants, after 11 weeks. Error bars represent the standard error was there an inhibitory interaction: a combination of the weevil of the means. Means followed by the same letter are not statistically different with the mirid showed fewer feeding scars than the weevils alone (Fisher’s LSD test, P < 0.05). n = 7 for all treatments. (Weyl and Hill, 2012; Table 3). This is in contrast with the present study and that of Ajuonu et al. (2009) where the interaction was

(F2,113 = 17.415, P < 0.001; Fig. 2). For leaf 2, there were no differ- equivalent. However, it is difficult to compare the studies because ences observed in the leaf surface area damaged by agent herbiv- they did not all measure the same plant growth parameters. This ory between any of the treatments (F5,32 = 1.393, P = 0.253). highlights the importance of standardising measures of agent ef- However, differences were observed in the leaf surface area fects, especially for water hyacinth where plant growth is strongly damaged by agent herbivory between some of the treatments on influenced by experimental conditions such as the nutrient condi- leaves 4 and leaf 5. On leaf 4, a combination of mites + weevils tions under which the plants are grown (e.g. Reddy et al., 1990; caused the least damage to the leaf, and this was significantly less Coetzee et al., 2007a; Marlin et al., 2013). If the true measure of than the damage caused by the mirids alone or a combination of an agent’s effectiveness is that it should reduce the density and/

Table 2 Means (±S.D.) of water hyacinth plant growth parameters at the end of an 11-week experiment, when plants were either exposed, or not exposed (control treatment), to pair-wise combinations (herbivory treatments) of three agents (mite Orthogalumna terebrantis, weevil Neochetina eichhorniae and mirid Eccritotarsus catarinensis). df = 41 and n = 7 for all treatments.

Treatment Total number of leaves produced in Total number of ramets produced in Maximum petiole length Change in wet biomass 11 weeks 11 weeks (cm) (kg) Mites only 9.29 ± 1.55 a 2.93 ± 0.89 a 37.71 ± 7.28 a 0.82 ± 0.19 a Weevils only 8.07 ± 1.67 a 1.5 ± 0.58 c 22.71 ± 8.72 b 0.55 ± 0.12 bc Mirids only 7.29 ± 0.91 a 1.43 ± 0.67 c 33.29 ± 10.94 a 0.60 ± 0.16 abc Mites + weevils 7.86 ± 1.25 a 1.79 ± 0.95 bc 21.89 ± 4.61 b 0.60 ± 0.27 abc Mites + mirids 8.36 ± 1.57 a 2.36 ± 0.90 ab 37.57 ± 3.59 a 0.76 ± 0.21 ab Weevils + mirids 8.75 ± 1.41 a 1.92 ± 0.49 bc 29.0 ± 10.05 ab 0.52 ± 0.17 c Control 7.14 ± 1.41 a 2.36 ± 0.81 ab 34.86 ± 9.58 a 0.81 ± 0.28 a

Means in columns followed by the same letter are not statistically different at the 5% level (Fisher’s protected least signiﬁcant difference test). D. Marlin et al. / Biological Control 67 (2013) 483–490 487

Fig. 2. Water hyacinth leaf surface area damaged on leaves 2, 4 and 5, after an 11- week exposure to herbivory by various combinations of three agents (mite Orthogalumna terebrantis, weevil Neochetina eichhorniae and mirid Eccritotarsus catarinensis). Error bars represent the standard error of the means. Means followed by the same lower case letters for leaf 4, and the same upper case letters for leaf 5, are not statistically different (Fisher’s LSD test, P > 0.05). The asterisk (⁄) indicates statistical differences between leaves within a particular treatment. n = 7 for all treatments.

Fig. 3. Leaf surface area damaged on leaf 5 of water hyacinth plants exposed to herbivory by different combinations of three agents (mite Orthogalumna terebrantis, weevil Neochetina eichhorniae and mirid Eccritotarsus catarinensis), after 11 weeks . The total damage was separated into damage caused by each agent species separately. Error bars represent the standard error of the means. Means followed by the same letter are not statistically different (Fisher’s LSD test, P < 0.05). n = 7 for all treatments.

or rate of spread of the target weed (Hoffmann and Moran, 1998), Fig. 4. Leaf surface area damage by herbivory of the mite Orthogalumna terebrantis then changes in plant biomass could serve as the most important (A), the weevil Neochetina eichhorniae (B) and the mirid Eccritotarsus catarinensis (C), measure of an agent’s impact on the plant, and the number of ra- on leaf 5 of water hyacinth plants, at the end of an 11-week experiment. Total mets could serve as an indicator for spread. None of the agents damage was separated into damage caused by each agent species separately. Error bars represent the standard error of the means. Means followed by the same letter or combinations tested here caused a significant reduction in bio- are not statistically different (Fisher’s LSD test, P < 0.05). n = 2 for all treatments. mass during the course of the experiment and only the mirids and weevils alone caused a significant reduction in ramet numbers. In general, no single agent or combination of agents can be se- additional agents cause an ‘‘equivalent’’ effect meaning that extra lected as having the greatest negative impact on all of the plant agents do not add to the negative impact on plant growth, but growth parameters. Rather, it seems that a single agent or a com- importantly they do not detract from it. The results shown here bination of agents is particularly damaging to only one specific thus refute Myers’ ‘‘lottery’’ model (Myers, 1985) which assumes growth parameter. Table 3 shows that in the majority of cases that one agent will have the main impact on a weed, but data from 488 D. Marlin et al. / Biological Control 67 (2013) 483–490

Table 3 A review of the interactive effects (interactions) of herbivory by the weevils Neochetina eichhorniae (NE) and N. bruchi (NB), the mirid Eccritotarsus catarinensis (EC) and the mite Orthogalumna terebrantis (OT), on water hyacinth growth (either at a whole plant level or on a speciﬁc growth parameter), compared to the effects of herbivory by an agent acting alone, with the interactions being deﬁned according to Hatcher (1995).

Impact on weed or speciﬁc growth parameter Classiﬁcation of interactions according to Reference Hatcher’s system (1995) NE with OT together reduced plant size and density n/a, did not compare interactive effect of Delfosse agents with the effect of individual agent (1978a) Combined action of NE with OT reduced size and density of the weed more than the effect of each agent Possibly synergistic Delfosse individually (this is mentioned in the text though not shown in the results) (1978b) After eight weeks the petiole length was reduced: Ajuonu et al. NE with EC = NE alone, Equivalent (2009) NE with EC > EC alone Synergistic Shoot (ramet) production was reduced: NE with EC = NE or EC alone Equivalent

No differences in plant growth were obtained between treatments. However, assuming increased Weyl and numbers of weevil feeding scars increase damage to the plant, then after eight weeks the number of Hill (2012) scars on the second leaf: NE with NB = NE alone, NB with EC = NB alone, Equivalent NE with NB > NB alone, Synergistic NE with EC < NE alone Inhibitory After 11 weeks ramets were reduced by: This study OT with NE > OT alone, Synergistic OT with EC = OT alone, OT with NE = NE alone, Equivalent NE with EC = NE alone and EC alone, Equivalent EC with OT > EC alone. Synergistic Petiole lengths were reduced by: OT with NE > OT alone, Synergistic OT with EC = OT alone and EC alone, Equivalent OT with NE = NE alone, NE with EC = NE alone and EC alone Equivalent

Meaning of symbols: the interaction of the agents has the same effect (=) on plant growth as would a single agent; or a greater (>) effect than either agent alone; or a lesser (<) effect than the weaker of the two agents. the present study show that each agent and each combination of Herbivory damage to the leaf surface area by each agent indi- agents affect the plant differently and no single agent is suitable cated that E. catarinensis and N. eichhorniae had no effect on each for every situation. other, probably because weevil larvae develop inside leaf petioles The leaf surface was most damaged when the plants were ex- (Warner, 1970; DeLoach and Cordo, 1976) whereas the mirid posed to herbivory by a combination of mites and mirids. There- nymphs develop on the surface of the leaf blades (Hill et al., fore, the mites and mirids cause superficial damage to the leaves, 1999), thus minimizing competitive interactions. In addition, the but this is not translated into damage that affects plant growth diel activity patterns of weevils and mirids are different, as are i.e. ramet production or biomass reduction. Although adult weevil their life-cycles; the weevils requiring around 13–17 weeks and feeding does damage the leaf surface, most of the weevil damage is the mirids three weeks (Hill et al., 1999). Ajuonu et al. (2007) found caused by the larvae burrowing inside the petioles (Hill and Cil- that old weevil feeding scars had a negative effect on adult mirid liers, 1999). Measuring feeding damage to leaves does not in itself survival. However, at the end of this study mirid numbers were indicate that agents are having an impact on plant growth, but the same on plants with mirids and weevils combined or with mir- rather indicates the presence or abundance of the agents (Wright ids alone, suggesting that there is little competition between these and Center, 1984). In addition, insect herbivory does not simply re- two agents. move leaf area, but also causes physiological changes in the plant. Of the three agents examined, O. terebrantis had the least im- For example, Ripley et al. (2008) showed that feeding by N. eichhor- pact on plant growth, whereas N. eichhorniae and E. catarinensis niae reduced water hyacinth photosynthetic rates, but this reduc- both had negative impacts on the plant, but on different growth tion was not related to leaf area removal, and could be in part parameters. In combination with weevils and mirids, the mites due to microbes that have been transported by the insect control generally had no additional effect on plant growth. Although the agents (Venter et al., 2013). Similarly, feeding by O. terebrantis mites performed poorly in the presence of the weevils and the mir- has been shown to have a negative impact on water hyacinth pho- ids, they improved the performance of the mirids, and have been tosynthetic rates but not on the plant’s growth (Marlin et al., 2013). thought to increase weevil numbers and feeding in other studies The general findings of this study show that O. terebrantis, N. (Delfosse, 1977). Therefore, this study is in agreement with authors eichhorniae and E. catarinensis can co-exist with little negative who believe that interspecific competition is relatively rare and interaction. The mirids and the weevils performed better, as shown thus is not an important factor structuring ecological communities by the amount of damage they caused to the leaf surface area, (Price, 1984; Strong, 1984, 2008). when in combination with the mites. The number of mirids was Orthogalumna terebrantis, N. eichhorniae and E. catarinensis co- significantly higher on plants that had mites on them compared exist in the field in many countries including their countries of ori- to plants that only had mirids on them. In contrast, O. terebrantis gin, with variable impact on water hyacinth populations (Julien performed better in the absence of N. eichhorniae or E. catarinensis. and Griffiths, 1998) and some researchers (Ajuonu et al., 2009) This could be because mite eggs are small (0.1. Â 0.14 mm, Cordo have shown that they are compatible. Although these agents do ap- and DeLoach, 1976) and adult weevils and mirids could easily pear to be compatible, their combined effects on water hyacinth damage them during their feeding. Further, the feeding of adult are neither ‘‘additive’’ nor ‘‘synergistic’’, and this could be due to weevils could remove entire mite galleries. the plant’s incredibly high growth rate (Edwards and Musil, D. Marlin et al. / Biological Control 67 (2013) 483–490 489

1975). Thus, the plant simply outgrows the agents whose popula- Delfosse, E.S., 1977. Effect of Orthogalumna terebrantis (Acari: Galumnidae) on tion growth rates are much slower and more sensitive to environ- Neochetina eichhorniae (Col.: Curculionidae) eggs and oviposition. Entomophaga 22, 359–363. mental variables such as nutrient levels and temperatures. Delfosse, E.S., 1978a. Effect on waterhyacinth of Neochetina eichhorniae (Col.: In conclusion, the present study shows that co-existing species Curculionidae) combined with Orthogalumna terebrantis (Acari: Galumnidae). may help each other rather than hinder each other. In addition, dif- Entomophaga 23, 379–387. Delfosse, E.S., 1978b. Interaction between the mottled waterhyacinth weevil, ferent agents are better suited to different areas; for example N. Neochetina eichhorniae Warner and the waterhyacinth mite, Orthogalumna eichhorniae can survive colder temperatures than E. catarinensis terebrantis Wallwork. In: Freeman, T.E. (Ed.), Proceedings of the 4th (Coetzee et al., 2007b), and thus the release of a new agent should International Symposium on the Biological Control of Weeds. University of Florida, Gainesville, Florida, pp. 93–97. be supported when an additional agent provides an ‘‘equivalent’’ DeLoach, C.J., Cordo, H.A., 1976. Life cycle and biology of Neochetina bruchi, a weevil effect because this ensures that biological control is not dependent attacking water hyacinth in Argentina, with notes on N. eichhorniae. Ann. on one agent only. In South Africa, for example, the weevils have Entomol. Soc. Am. 69, 643–652. Denno, R.F., Peterson, M.A., Gratton, C., Cheng, J., Langellotto, G.S., Hunerty, A.F., established at all of the water hyacinth infestations where they Finke, D.L., 2000. Feeding-induced changes in plant quality mediate interspecific have been released, but the mirids and mites have not been re- competition between sap-feeding herbivores. Ecology 81, 1814–1827. leased at as many sites (Byrne et al., 2010). Based on the results Denoth, M., Frid, L., Myers, J.H., 2002. Multiple agents in biological control: of this study, augmentative releases or new releases at sites where improving the odds? Biol. Control 24, 20–30. Edwards, D., Musil, C.J., 1975. Eichhornia crassipes in South Africa – a general review. mirids and mites do not yet occur are necessary to enhance the J. Limnol. Soc. S. Afr. 1, 23–27. chances of achieving effective biological control of water hyacinth Hatcher, P.E., 1995. Three-way interactions between plant pathogenic fungi, in South Africa. herbivorous insects and their host plant. Biol. Rev. 70, 639–694. Heard, T.A., Winterton, S.T., 2000. Interactions between nutrient status and weevil herbivory in the biological control of water hyacinth. J. Appl. Ecol. 37, 117–127. Hill, M.P., Cilliers, C.J., 1999. A review of the arthropod natural enemies, and factors Acknowledgments that influence their efficacy, in the biological control of the water hyacinth, Eichhornia crassipes (Mart.) Solms-Laubach (Pontederiaceae), in South Africa. In: We thank the Working for Water Programme of the Depart- Olckers, T., Hill, M.P. (Eds.), Biological Control of Weeds in South Africa (1990– 1998). African Entomology Memoir 1. The Entomological Society of South ment of Environmental Affairs, South Africa, for providing funding Africa, Pretoria, pp. 103–112. for this research, and the ARC-PPRI for the use of facilities. Photos Hill, M.P., Cilliers, C.J., Neser, S., 1999. Life history and laboratory host range of used in the graphical abstract were kindly provided by A. King and Eccritotarsus catarinensis (Carvalho) (Heteroptera: Miridae), a new natural enemy released on water hyacinth (Eichhornia crassipes (Mart.) Solms-Laub.) D. Kruger. We thank M. Smith for statistical advice. (Pontederiaceae) in South Africa. Biol. Control 14, 127–133. Hoffmann, J.H., Moran, V.C., 1998. The population dynamics of an introduced tree, Sesbania punicea, in South Africa, in response to long-term damage caused by References different combinations of three species of biological control agents. Oecologia 114, 343–348. Holmes, S., 1996. South African Water Quality Guidelines, vol. 7. Aquatic Agrawal, A.A., 1998. Induced responses to herbivory and increased plant Ecosystems, Department of Water Affairs and Forestry, South Africa. performance. Science 279, 1201–1202. Jiménez, M.M., Balandra, Ma., 2007. Integrated control of Eichhornia crassipes by Ajuonu, O., Byrne, M., Hill, M., Neuenschwander, P., Korie, S., 2007. Survival of the using insects and plant pathogens in Mexico. Crop Prot. 26, 1234–1238. mirid Eccritotarsus catarinensis as influenced by Neochetina eichhorniae and Julien, M.H., Griffiths, M.W. (Eds.), 1998. Biological Control of Weeds: A World Neochetina bruchi feeding scars on leaves of water hyacinth Eichhornia crassipes. Catalogue of Agents and Their Target Weeds, fourth ed. Commonwealth Biocontrol 52, 193–205. Agricultural Bureau International, Wallingford, United Kingdom. Ajuonu, O., Byrne, M., Hill, M., Neuenschwander, P., Korie, S., 2009. The effect of two Kaplan, I., Denno, R.F., 2007. Interspecific interactions in phytophagous insects biological control agents, the weevil Neochetina eichhorniae and the mirid revisited: a quantitative assessment of competition theory. Ecol. Lett. 10, 977– Eccritotarsus catarinensis on water hyacinth, Eichhornia crassipes, grown in 994. culture with water lettuce, Pistia stratiotes. Biocontrol 54, 155–162. Karban, R., Baldwin, I.T., 1997. Induced Responses to Herbivory. University of Bounfour, M., Tanigoshi, L.K., 2001. Host plant-mediated interactions between Chicago Press, Chicago, Illinois. Tetranychus urticae and Eotetranychus carpini borealis (Acari: Tetranychidae). Karban, R., Carey, J.R., 1984. Induced resistance of cotton seedlings to mites. Science Exp. Appl. Acarol. 25, 13–24. 225, 53–54. Byrne, M., Hill, M., Robertson, M., King, A., Jadhav, A., Katembo, N., Wilson, J., Marlin, D., 2011. The Role of the mite Orthogalumna terebrantis in the biological Brudvig, R., Fisher, J., 2010. Integrated management of water hyacinth in South control programme for water hyacinth, Eichhornia crassipes, in South Africa. Africa: Development of an integrated management plan for water hyacinth PhD Thesis. Rhodes University, Grahamstown, South Africa. control, combining biological control, herbicidal control and nutrient control, Marlin, D., Hill, M.P., Ripley, B.S., Strauss, A.J., Byrne, M.J., 2013. The effect of tailored to the climatic regions of South Africa. Water Res. Counc. Rep. No. TT herbivory by the mite Orthogalumna terebrantis on the growth and 454/10. photosynthetic performance of water hyacinth (Eichhornia crassipes). Aquat. Caunter, I.G., Mohamed, S., 1990. The effect of Neochetina eichhorniae and Bot. 104, 60–69. Myrothecium roridum alone and in combination for the initial biological Moran, P.J., 2005. Leaf scarring by the weevils Neochetina eichhorniae and N. bruchi control of water hyacinth. J. Plant Prot. Tropics 7, 133–140. enhances infection by the fungus Cercospora piaropi on water hyacinth, Center, T.D., Jubinski, G.P., 1996. The effects of augmented water hyacinth weevil Eichhornia crassipes. BioControl 50, 511–524. (Neochetina eichhorniae) populations on water hyacinth (Eichhornia crassipes) Myers, J.H., 1985. How many insect species are necessary for successful biocontrol mat expansion rates (Abstract). In: Moran, V.C., Hoffman, J.H. (Eds.), of weeds? In: Delfosse, E.S. (Ed.), Proceedings of the sixth International Proceedings of the 9th International Symposium on the Biological Control of Symposium on the Biological Control of Weeds, 19–25 August 1984. Weeds. Stellenbosch, University of Cape Town, South Africa, p. 507. University of British Columbia, Vancouver, Agriculture Canada, Ottawa, pp. Cilliers, C.J., 1991. Biological control of water hyacinth, Eichhornia crassipes 77–82. (Pontederiaceae), in South Africa. Agr. Ecosyst. Environ. 37, 207–217. Perkins, B.D., 1973. Preliminary studies on a strain of the waterhyacinth mite from Coetzee, J.A., Byrne, M.J., Hill, M.P., 2007a. Impact of nutrients and herbivory by Argentina. In: Dunn, P.H. (Ed.), Proceedings of the second International Eccritotarsus catarinensis on the biological control of water hyacinth, Eichhornia Symposium on the Biological Control of Weeds. Commonwealth Agricultural crassipes. Aquat. Bot. 86, 179–186. Bureau, Rome, Italy, pp. 179–184. Coetzee, J.A., Byrne, M.J., Hill, M.P., 2007b. Predicting the distribution of Eccritotarsus Price, P.W., 1984. Communities of specialists: vacant niches in ecological and catarinensis, a natural enemy released on water hyacinth in South Africa. evolutionary time. In: Strong, D.R., Jr., Simberloff, D., Abele, L.G., Thistle, A.B. Entomol. Exp. Appl. 125, 237–247. (Eds.), Ecological Communities: Conceptual Issues and the Evidence. Princeton Coetzee, J.A., Hill, M.P., Byrne, M.J., Bownes, A., 2011. A review of the biological University Press, Surrey, UK, pp. 510–523. control programme on Eichhornia crassipes (C.Mart) Solms (Pontederiaceae), Reddy, K.R., Agami, M., Tucker, J.C., 1990. Influence of phosphorus on growth and Salvinia molesta D.S.Mitch. (Salviniaceae), Pistia stratiotes L. (Araceae), nutrient storage by water hyacinth (Eichhornia crassipes (Mart.) Solms) plants. Myriophyllum aquaticum (Vell.) Verdc. (Haloragaceae) and Azolla filiculoides Aquat. Bot. 37, 355–365. Lam. (Azollacaea) in South Africa. Afr. Entomol. 19, 451–468. Ripley, B.S., DeWet, L., Hill, M.P., 2008. Herbivory-induced reduction in Cordo, H.A., DeLoach, C.J., 1976. Biology of the water hyacinth mite Orthogalumna photosynthetic productivity of water hyacinth, Eichhornia crassipes (Martius) terebrantis in Argentina. Weed Sci. 24, 245–249. Solms-Laubach (Pontederiaceae), is not directly related to reduction in Crowe, M.L., Bouchier, R.S., 2006. Interspecific interactions between the gall-fly photosynthetic leaf area. Afr. Entomol. 16, 140–142. Urophora affinis Frfld. (Diptera: Tephritidae) and the weevil Larinus minutus Rodriguez-Saona, C., Thaler, J.S., 2005. Herbivore-induced responses and patch Gyll. (Coleoptera: Curculionidae), two biological control agents released against heterogeneity affect abundance of arthropods on plants. Ecol. Entomol. 30, 156– spotted knapweed, Centaurea stobe L. ssp. micranthos. Biocontrol Sci. Technol. 163. 16, 417–430. 490 D. Marlin et al. / Biological Control 67 (2013) 483–490

Snedecor, G.W., Cochran, W.G., 1980. Statistical Methods, seventh ed. Iowa State Venter, N., Hill, M.P., Hutchingson, S.-L., Ripley, B.S., 2013. Weevil borne microbes University Press, Ames. contribute as much to the reduction of photosynthesis in water hyacinth as Stephens, A.E.A., Srivastava, D.S., Myers, J.H., 2013. Strength in numbers? Effects of does herbivory. Biol. Control 64, 138–142. multiple natural enemy species on plant performance. Proc. R. Soc. B 280, Walter, D.E., 2009. Reproduction and embryogenesis. In: Krantz, G.W., Walter, D.E. dx.doi.org/10.1098/rspb.2012.2756. (Eds.), A Manual of Acarology, third ed. Texas Tech University Press, Texas, USA, Strong Jr., D.R., 1984. Exorcising the ghost of competition past: phytophagous pp. 54–61. insects. In: Strong, D.R., Jr., Simberloff, D., Abele, L.G., Thistle, A.B. (Eds.), Warner, R.E., 1970. Neochetina eichhorniae, a new species of weevil from water hyacinth, Ecological Communities: Conceptual Issues and the Evidence. Princeton and biological notes on it and N. bruchi. Proc. Entomol. Soc. Wash. 72, 487–496. University Press, Surrey, UK, pp. 28–41. Weyl, P.S.R., Hill, M.P., 2012. The effect of insect–insect interactions on the Strong Jr., D.R., 2008. Food chains and food webs. Encycl. Ecol. 2, 1627–1636. performance of three biological control agents released against water hyacinth. Turner, P.J., Morin, L., Williams, D.G., Kriticos, D.J., 2010. Interactions between a Biocontrol Sci. Technol. 22, 883–897. leafhopper and rust fungus on the invasive plant Asparagus asparagoides in Wright, A.D., Center, T.D., 1984. Predicting population intensity of adult Neochetina Australia: a case of two agents being better than one for biological control. Biol. eichhorniae (Coleoptera: Curculionidae) from incidence of feeding on leaves of Control 54, 322–330. water hyacinth, Eichhornia crassipes. Environ. Entomol. 13, 1478–1481.