We assessed the performance of Pseudacteon curvatus Borgmeier with respect to the social form of Solenopsis invicta Buren in Argentina. In the field, we studied the effect the parasitoid on size and proportion of ant foragers. In the laboratory, we evaluated P. curvatus oviposition preferences; host size elected; developmental periods; and sexual size dimorphism, sex ratio, and parasitoid survivorship. P. curvatus affected the average size of foraging workers on both social forms diminishing the proportion of big and increasing the proportion of minor workers. P. curvatus required a shorter orientation time and exhibited a greater number of attacks when ovipositing on monogynes workers. In the laboratory, host sizes elected by P. curvatus were similar between social forms. However, attacks on polygyne colonies were more variable, increasing the number of unviable offspring. Developmental times of females and males of P. curvatus were similar for both social forms, but total developmental periods were shorter for males from monogyne colonies. We did not find differences between sexes in emerging adults’ size by social form and the female: male sex ratio was 1:1 for both social forms. P. curvatus pupae survival and adult emergence per trial from monogyne colonies were greater than from polygyne colonies. The rarity of polygyne S. invicta in its native range may prevent this phorid from adjusting its life history to that social form. Consequences of applying this phorid in biological control are discussed.
Several species of phorid flies in the genus Pseudacteon Coquillett (Diptera: Phoridae) are being used in the United States in biological control efforts against Solenopsis invicta Buren (Gilbert and Patrock 2002, Graham et al. 2003, Mottern et al. 2004, Porter et al. 2004, Thead et al. 2005, Morrison and Porter 2006), an invasive pest fire ant from Argentina and Brazil (Mescher et al. 2003, Ross et al. 2007). Besides the negative effect of Pseudacteon caused by direct mortality (Gilbert and Morrison 1997; Morrison and Gilbert 1999; Folgarait et al. 2002a,2005b,2006), it is assumed that different Pseudacteon species would alter the competitive balance of Solenopsis against other competitive ants, because they negatively affect the foraging pattern of Solenopsis (Feener and Brown 1992, Folgarait and Gilbert 1999, Feener et al. 2008).
At least 22 species of Pseudacteon flies are known to attack workers of the genus Solenopsis (Porter and Pesquero 2001, Brown et al. 2003, Folgarait et al. 2005a, Calcaterra 2007, Kronforst et al. 2007). Details of life history (Porter et al. 1995b; Morrison et al. 1997; Folgarait et al. 2002a,2005b,2006), karyotypes in somatic cells (Chirino et al. 2006,2007), host specificity (Porter et al. 1995a, Gilbert and Morrison 1997, Morrison and Gilbert 1999, Folgarait et al. 2002b, Vazquez et al. 2004, Estrada et al. 2006), and oviposition behavior (Morrison et al. 1997, Wuellner et al. 2002) have been documented. Pseudacteon species vary in size, and in host ant workers sizes preferred for oviposition, by attacking a certain size range of host workers (Feener 1987; Morrison et al. 1997; Morrison and Gilbert 1998; Folgarait et al. 2002a,2005b,2006).
Once introduced into North America, S. invicta populations showed geographical differentiation with respect to social form. The polygyne, rare in the species' native range, became the dominant social form over a vast area from western Louisiana to southern Texas and is now found widely in the introduced range (Greenberg et al. 1985, Porter and Savignano 1990, Macom and Porter 1996). It is known that the social form affects the proportion of workers' subcastes (Wood and Tschinkel 1981, Porter and Tschinkel 1985, Tschinkel et al. 2003) and their aggressivity (Keller and Ross 1998, Bourke 2002, Deheer 2002), with polygyne having less proportion of larger workers (Greenberg et al. 1985). However, the relevancy of the social form of S. invicta on development and survival of parasitoids has been little explored (Morrison and Gilbert 1998). We predicted that a small phorid such as Pseudacteon curvatus Borgmeier would develop successfully on polygyne ants because of the high availability of small workers and because polygyne colonies have on average smaller ants than monogyne. Thus, one aim of this study was to investigate life history traits of P. curvatus in relation to the social form of S. invicta in its native range where both forms occur in sympatry.
The impact of Pseudacteon on foraging workers (Feener and Brown 1992; Orr et al. 1995,1997; Morrison et al. 1997; Morrison 1999; Morrison and Gilbert 1999; Wuellner et al. 2002) and on postinfection host behavior (Henne and Johnson 2007) has been analyzed, but without regard to which subcastes are affected. Folgarait and Gilbert (1999) found that when Pseudacteon parasitoids attacked Solenopsis richteri Forel along foraging trails, the relative frequency of smaller workers increased at the expense of larger worker sizes. Thus, another aim was to extend these earlier studies by comparing responses of foraging S. invicta to a small phorid such as P. curvatus. Moreover, we were interested in gaining insight into how social form might alter outcomes of interactions with phorids in the native range.
Pseudacteon curvatus was released in several states (Porter 2000), and host specificity tests showed that flies from different Argentinean provinces prefer different fire ant species (Graham et al. 2003, Vazquez et al. 2004, Thead et al. 2005, Estrada et al. 2006). P. curvatus from Buenos Aires province has a strong preference for S. richteri and hybrids S. richteri × S. invicta (Porter and Briano 2000, Graham et al. 2003, Thead et al. 2005), whereas P. curvatus from Formosa province prefers to attack the red imported fire ant S. invicta (Vazquez et al. 2004). Given that different biotypes of P. curvatus prefer to attack different host species, it also seemed possible that social form of Solenopsis species could be another key factor for life history evolution of these parasitoids. Currently, three other species (P. tricuspis Borgmeier, P. litoralis Borgmeier, and P. obtusus Borgmeier) have been released and are established in the United States (Vogt and Streett 2003, Porter and Gilbert 2004), and their effects are under study. Releases of P. nocens Borgmeier and others are in progress (Gilbert et al. 2008).
In this study, we compared how P. curvatus differs regarding each social form (monogyne versus polygyne) of S. invicta with respect to, in the field, (1) oviposition and orientation times and (2) their effect on sizes and proportion of foragers; and in the laboratory, (3) worker castes selected for oviposition; (4) developmental periods; (5) characteristics of emerged adults; and (6) pupal survival.
Materials and Methods
Ant and Fly Sources
Experiments were carried out between February and October 2005 at the Parque Nacional Chaco (26°48′ S and 59°34′ W), Chaco Province, Argentina. The study site is a subtropical forest in the Chaqueña phytogeographical province and receives an average of 500 mm of rain a year mainly in summer (Cabrera and Willink 1980). S. invicta occurs in pastures and the trails across the forest. Ants were identified in the laboratory using Trager (1991), Pitts (2002), and Pitts et al. (2005) keys. The social form of S. invicta was determined by the number of queens found in excavated nest. Nests with more than one queen were considered polygyne (P). Queens may not be uncovered in excavations so the social form of nests with zero to one queen was determined using the method of Valles and Porter (2003). If a colony presented the amplification of the specific DNA segment for the Gp-9B gene (BB genotype), we considered it as monogyne (M) (C.M.G., unpublished data). P. curvatus females were field collected, identified with a hand magnifying glass (×20) according to the Porter and Pesquero (2001) key, kept under cool, dark humid conditions in individual tubes, and used for experiments within 48 h after collection. Voucher specimens of P. curvatus and S. invicta are deposited into the Folgarait collection at the Universidad Nacional de Quilmes (Buenos Aires, Argentina) and the Museo de Historia Natural Bernardino Rivadavia. Six mounds with three replicates each were used per social form for all observations in the field and the laboratory. All measurements of ants and phorids were done with an ocular micrometer calibrated in 0.03-mm increments on a Nikon stereoscopic microscope, model type SMZ-1 ESD (×30) (Microlat S.R.L., Buenos Aires, Argentina).
Ovipositions and Orientation Times.
To evaluate P. curvatus oviposition preferences for each social form, we released a P. curvatus female over the bait after the foraging trail was established. All baits and foraging trails were isolated with acrylic transparent tunnels (1 by 0.15 by 0.20 m) to avoid alterations in case other parasitoids or rival ants can contact them. To regulate internal temperature changes, the tunnels were perforated, and the holes were covered by a thin mesh of tulle. Measures of oviposition preference included the number of attacks ("ovipositions") counted during the first 15 min after release and the orientation time (time invested in detecting before beginning oviposition). After the fly was captured, a random ant sample was collected (50-150 individuals) from or next to the bait for measuring their sizes. If the phorid did not oviposit during the first 10 min, it was replaced by another, and if it was captured by the ants, the measurement was discarded.
Orientation times and number of attacks, discriminated per social form, were compared by means of the Student's t-test because data were normally distributed. Relationships between attack rate of parasitoid versus its orientation time and number of attacks versus foraging numbers of workers were analyzed by linear regression.
Effects of P. curvatus on Foraging Rate, Sizes, and Proportion of S. invicta Foragers.
To assess the effect of P. curvatus on S. invicta, we evaluated the foraging rate, the worker sizes, and the foragers' proportion before and after releasing the parasitoid using the following approach. Tuna baits were placed 20 cm away from each nest. Ants were allowed to forage for ≈45 min until a constant flow of ants was achieved without phorids' presence, and the foraging rate was determined by counting the number of workers that passed through one point for 5 min. A random sample of 50-150 ants was collected from the trail and bait. The captured ants were measured and separated by sizes.
Finally, mounds were excavated, and the ants were separated from the dirt by means of a dripping system. In the laboratory, they were discriminated in subcastes by sieving them through different pore sizes (ZONYTEST sieves; Rey & Ronzoni S.R.L., Buenos Aires, Argentina). We used five size categories to better discriminate differences between both social forms of S. invicta (Greenberg et al. 1985) and because several species of Pseudacteon oviposit on workers of different sizes (Feener 1987; Morrison and Gilbert 1998; Folgarait et al. 2002b,2005b,2006). Subcastes were catalogued as major (ants retained in the sieve mesh #20, with a pore size of 1,190 μm); big (ants retained in the sieve mesh #18, with a pore size of 1,000 μm); medium (ants retained in the sieve mesh #20, with a pore size of 840 μm); small (ants retained in the sieve mesh #25, with a pore size of 710 μm), and minor (ants that passed through the sieve mesh #25 with a pore size of 710 μm).
To compare social form with respect to measurements of ant foraging rates, we used the Student's t-test because data were normally distributed. Worker mean sizes "without" (before releasing) and "with" (after releasing; previous section) phorids were compared with the paired t-test. Workers proportions involved in the foraging or defense of the bait were estimated by the proportion test.
Evaluation of the Worker Castes Selected, Parasitoid Developmental Periods, and Characteristics of Emerged Adults.
We evaluated the host sizes selected, developmental times, and adult sizes and sex ratios of parasitoids discriminated by host social form. Flies were allowed to oviposit in attack arenas (each called a trial) where we had placed similar amounts of big, medium, and small ants (667 ants per subcaste). Major and minor were discarded because they are not preferentially chosen by P. curvatus (Morrison et al. 1997, Chirino et al. 2004). We used three to six P. curvatus (4.33 ± 1.21 females), which were allowed to oviposit for 2 h. "Attacked" ants were placed in a climate controlled rearing room (28 ± 1°C; 80 ± 10% RH). When ants died, we collected the heads where the late larval stage was found. They were daily checked until pupae stage occurred; those with pupae were measured (head width across the eyes) and observed for 45 d to allow adult fly emergence. Adults were sexed and measured (maximum width of pronotum).
Sizes of ants and parasitoids were normally distributed and exhibited homoscedasticity; therefore, average sizes of parasitized ants were compared by social form with the Student's t-test. Parasitoid developmental periods discriminated by social form were analyzed with the Mann-Whitney test because data were not normally distributed (Daniel 1990). Sizes of adult parasitoids were analyzed using two-factor analyses of variance (ANOVAs; sex and social form). Sex proportion of adults was determined by the binomial test. The mean size of heads parasitized with viable and unviable pupae were compared by means of the Student's t-t est and their size distribution by the Kolmogorov-Smirnov test. Pupal survival (emerged adults per pupae) was analyzed with the Mann-Whitney test. Statistical analyses were done using the STATISTIX Program (Analytic Software, Tallahassee, FL).
Ovipositions and Orientation Times.
Pseudacteon curvatus had a greater number of ovipositions (36.17 ± 7.36 versus 15.83 ± 6.43 attacks, for M than P ants, respectively; t = 2.20, df = 10, P = 0.026; two-tailed t-test) and shorter orientation times when it attacked M workers (49.83 ± 34.38 versus 120.50 ± 70.62 s for M versus P ants, respectively; t = 5.10, df = 10, P = 0.0005; two-tailed t-test). No significant linear regression was found between the number of ovipositions carried out by P. curvatus and the number of ants present on the trails (F1,4 = 1.70; r2 = 0.20, P = 0.23 and F1,4 = 0.17; r2 = -0.12, P = 0.70 for M and P ants, respectively). No significant linear relationship was observed between the number of attacks by P. curvatus and the orientation time (F1,4 = 0.65; r2 = 0.08, P = 0.45 and F1,4 = 0.18; r2 = -0.11, P = 0.69 for M and P ants, respectively).
Effects of P. curvatus on Foraging Rate, Sizes, and Proportion of Foragers of S. invicta.
Foraging rate was significantly higher by M workers than by P workers in the absence of parasitoids (308.50 ± 42.33 versus 221.33 ± 38.34 ants for M and P nests, respectively; t = 3.74, df = 10, P = 0.0039; two-tailed t-test).
At the moment that P. curvatus started hovering over the ants, workers stopped foraging, although they did not stop protecting the bait, regardless of their social form. Workers at the bait started vibrating their gasters, curling their bodies, or presented stereotypical U-posture; then they started to make tunnels toward the baits to forage in a parasitoid-free environment.
In absence of P. curvatus, the mean size of M workers was statistically smaller than on P workers (0.64 ± 0.02 versus 0.69 ± 0.02 mm for M and P workers, respectively; t = 6.50, df = 10, P < 0.001; two-tailed t-test). In the presence of the parasitoid, ants were smaller in both social forms of S. invicta (t = 14.33, df = 10, P < 0.001 for M colonies; t = 4.20, df = 10, P = 0.003 for P colonies; two-tailed paired t-test). Again M workers were statistically smaller than P workers (0.59 ± 0.02 versus 0.63 ± 0.03 mm for M and P workers, respectively; t = 3.20, df = 10, P = 0.0056; two-tailed t-test).
Relative frequencies of the five subcastes were affected by P. curvatus, although the parasitoid preferred to attack small workers (Fig. 1). In their absence, big workers of M nests represented 5.6% of foragers, whereas 93.1% belonged to three subcastes of smaller size; in comparison, 7.1% of P workers belonged to the big workers and 86.7% corresponded to the medium, small, and minor subcastes (Table 1). In the presence of P. curvatus, the percentage of big workers decreased significantly, whereas the percentage of medium, small, and minor workers increased for M and P nests (Table 1). The percentage of major workers was not significantly modified in M nests, but it diminished in P nests. The percentages of medium and small workers decreased in M nests but were not altered in P nests. Finally, the proportion of minor workers increased in both social forms of S. invicta (Table 1).
Comparisons ”with“ and ”without“ parasitoids were analyzed by means of the paired Student's t test (P < 0.05).
Changes in the proportions for each subcaste, in absence and presence of P. curvatus, were analyzed by the proportion test. Arrows (↑, ↓) indicate significant changes (P < 0.05). Ma, major; B, big; M, medium; S, small; Mi, minor.
Worker Castes Selected.
Pseudacteon curvatus oviposited on workers of similar sizes regardless of the social form of S. invicta (0.63 ± 0.04 mm; t = 0.16, df = 9, P = 0.88; two-tailed t-test), and females and males emerged from workers of similar sizes from M and P colonies (0.64 ± 0.05 versus 0.63 ± 0.05 mm for females, respectively; t = 0.18, df = 7, P = 0.86 and 0.63 ± 0.05 versus 0.64 ± 0.09 mm for males, respectively; t = 0.11, df = 9, P = 0.91; two-tailed t-test; Fig. 2). The mean size of unviable pupae did not differ from those producing viable ones (0.62 ± 0.06 versus 0.63 ± 0.04 mm for M colonies, respectively; t = 0.59, df = 225, P = 0.5574; and 0.62 ± 0.10 versus 0.62 ± 0.06 mm for P colonies, respectively; t = 1.49, df = 239, P = 0.1371; two-tailed t-test). The head size distribution of viable and unviable pupae were similar for M and P colonies (χ2 = 1.223, P > 0.90 for M nests and χ2 = 4.68, P = 0.1926 for P nests, respectively; Kolmogorov-Smirnov test). As a whole, the preferred size class, small workers, had the highest percentage of viable pupae (87 and 80% for M and P nests, respectively), whereas in minors, the percentage of viable pupae was 6.8 and 15%, and in medium workers, it was 6.2 and 5% for M and P nests, respectively.
Reared females and males of P. curvatus exhibited similar developmental times (Table 2). Larval periods of emerged females and males of M workers were significantly shorter (2 d) than those developed from P ants (Table 2). This difference, however, only affected total developmental time of males (Table 2).
Comparisons between sexes within social forms and for each sex between social forms were analyzed with the Mann-Whitney Test (two-tailed).
N indicates the no. of nests from which females and males emerged, and n is the no. of emerged individuals.
Characteristics of Emerged Adults.
Pseudacteon curvatus adults that developed in each social form did not differ in size (0.33 ± 0.02 versus 0.34 ± 0.04 mm for M and P adults, respectively; t = 0.21, df = 9, P = 0.84; two-tailed t-test). ANOVA results did not show sexual size dimorphism (F1, 17 = 2.228, P = 0.1488), significant differences in the sizes of the same sex between social form (F1, 17 = 0.003, P = 0.9606), or an interaction between sex and social form (F1, 17 = 1.054, P = 0.3189). Sex ratio (females per males) did not vary between social forms, and 50% of the emerged adults were females (females = 82 and adults = 171, P = 0.32 for M colonies; females = 89 and adults =181, P = 0.44 for P colonies; binomial test).
All trials performed on M ants produced pupae, whereas 83.3% of trials done with P ants were productive. As a whole, pupal survival was 72.5 and 73.6% for M and P nests, respectively. However, the production of pupae per trial was higher on M ants than on P ants (76.2 versus 14.0% for M and P nests, respectively; Z = 8.16, df = 10, P < 0.0001; Mann-Whitney test). Number of pupae produced per trial was 38.83 ± 36.25 and 40.50 ± 63.91 for M and P nests, respectively; therefore, there was more variability when development took place on P ants. More adults emerged from M colonies than from P colonies (23.50 versus 6.01 adults for M and P ants, respectively; Z = 4.20, df = 9, P < 0.05; Mann-Whitney test).
This study compared oviposition behavior and development of P. curvatus with respect to the social form of S. invicta fire ants in this native range, Argentina. In the field, P. curvatus exhibited lower orientation times and higher oviposition rates when attacking M foragers than when attacking P foragers. In addition, P. curvatus modified sizes and proportions of foragers in favor of minor and in detriment of big workers in both social forms. The percentage of medium and small workers decreased in M nests but not in P nests. However, in M colonies, the same percentage of major workers (also unsuitable hosts) remained at the bait (Table 1). This change in the foraging pattern can be beneficial for S. invicta because the ants continue foraging by releasing workers (minors), which are less preferred by P. curvatus.
Although the average size of M workers was larger than P workers (0.75 versus 0.69 mm for M and P workers, respectively; P < 0.01; C.M.G., unpublished data), we observed that the mean size of M forager ants was smaller than P foragers. We hypothesize that Solenopsis ants showed a trade-off between the size and the number of workers involved in foraging because M workers presented a foraging rate statistically higher than P foragers. It is likely that in mass recruiting, this species might be similarly efficient by having more workers of a smaller size than by having a smaller number of bigger workers.
Interestingly, the overall size of S. invicta in its native range is smaller than in the introduction places (in Texas these were 0.86 and 0.71 mm for M and P colonies, respectively; Morrison and Gilbert 1998). Accordingly, the mean size of worker elected by P. curvatus was smaller in Chaco, Argentina (0.63 ± 0.04 mm for both social forms) than in Texas (0.71 ± 0.06 versus 0.66 ± 0.11 mm for M and P, respectively; Morrison and Gilbert 1998). Concurrently, the P colonies density in the southeastern United States is 3.2 times greater than for M colonies (in Texas is 2.3 times more) (Porter et al. 1991; Porter et al. 1992). Probably, the release from natural enemies in the introduced range (Porter et al. 1997) allowed them, apart from the founder effect that populations of S. invicta suffered in their introduction into novel places (Ross et al. 1996), to allocate more energy into reproduction and workers' vigor. Consequently, this social form became the most important one in the invasive range.
In the laboratory, P. curvatus mainly parasitized small workers, which are smaller than the mean size of the colonies of both social forms (Fig. 1). This preference could also be evaluated by the high percentage of unviable pupae for minors and medium workers. This pattern was also found for P. cultellatus Borgmeier (Folgarait et al. 2002a), another small species that develops on workers of small size of S. invicta and S. richteri. Besides, P. curvatus successfully developed in both social forms, but pupae and adult production per trial were higher on M colonies. It is likely that these differences of P. curvatus on P ants occur because this social form is the least abundant in Argentina (Wojcik 1983, Ross and Trager 1990, Porter et al. 1997, Ross et al. 1997, Mescher et al. 2003). Our field site was composed of three grassland patches separated by 20-40 m between each other. Within these patches, we found that M density was 3.5 greater than P density (C.M.G., unpublished data). Consequently, P. curvatus may be poorly adapted to this social form because of gene flow from surrounding populations that breed on M mounds of S. invicta.
To date, no studies have evaluated the sexual size dimorphism in P. curvatus; we did not find evidence for such phenomenon, and the sex proportion was 50%. No sexual size dimorphism was found for P. cultellatus either (females, 0.31 ± 0.03 mm, N = 28; males, 0.29 ± 0.03 mm, N = 49; unpublished data). The sexual size dimorphism in arthropods is a result of additional larval instars, longer developmental periods, or higher growth rate in the largest sex (Esperck and Tammaru 2006, Stillwell and Fox 2007), although it is dependent on temperature and food quality (Parker and Johnston 2006, Stillwell and Fox 2007). In the genus Pseudacteon, it is proposed that body size is determined by the host size (Morrison et al. 1999). However, in P. curvatus, females and males develop in similar host sizes and have similar developmental periods, and hosts with heads smaller and bigger than the optimal size yield a high percentage of unviable pupae, which may indicate a constraint against the evolution of sexual size dimorphism.
Pseudacteon curvatus is the species with the least host specificity (Gilbert and Morrison 1997, Morrison and Gilbert 1998); it also has the widest distribution and abundance in Argentina and Brazil and is adapted to different biomes and climates (Folgarait et al. 2005a, Calcaterra 2007). Field releases of the biotype from Las Flores (collected on S. richteri) to control hybrid populations of S. richteri × S. invicta showed that the parasitoid oviposited and developed successfully in Alabama but was unsuccessful in the assays conducted in Florida and Tennessee because of lower oviposition rates (Graham et al. 2003). The biotype collected from S. invicta in Formosa showed high specificity and preference toward S. invicta from Florida (Vazquez et al. 2004). In Alabama and Tennessee, M colonies dominate, whereas in Florida, both social forms occur. Therefore, our results suggest that the social form of the host in both source and target populations of Pseudacteon phorids used in biological control efforts should be taken into account when analyzing the success of introductions and impact on target ants. However, the rapid spread of P. curvatus from Corrientes in P populations of fire ants in Texas (Gilbert et al. 2008) shows its plasticity in adjusting to that social form. We do not know if different biotypes are true sibling or cryptic species. The unique case of cryptic species is reported for big and small P. obtusus, which have morphological differences in their body sizes and presence versus absence of an aristae in the antennae; however, molecular data suggest that they should be different species (Kronforst et al. 2007). However, P. curvatus populations from Brazil and Argentina have the same mean sizes (Morrison et al. 1999, Wuellner et al. 2002), but there is regional variability in the shape of their ovipositors. Ovipositors of Argentinean P. curvatus are less sharply curved than on Brazilian P. curvatus, and the former has a medial reinforcement ridge in the ventral tooth, which is absent in the latter (Porter and Pesquero 2001). Thus, it is unclear whether the biotype or the presence, even in small patches, of the M social form explains the success of P. curvatus introductions. It will be interesting to genetically study the different populations of P. curvatus and compare their production in introduced areas/patches where mostly the M or P social forms predominate, as well as to check for postintroduction shifts in life histories.
This study was carried out thanks to the permission of the Administración de Parques Nacionales (APN) and the Parque Nacional Chaco, to which we are very grateful. We also thank R. Patrock and R. Plowes for comments on this paper. This research was supported by grants from the Programa de Investigación en Interacciones Biológicas (PPUNQ-0340/03), Universidad Nacional de Quilmes, Lee and Ramona Bass Foundation, and the Robert J. Kleberg and Helen C. Kleberg Foundation. M.G.C. and P.J.F. thank the Consejo Nacional de Investigaciones Científicas y Técnicas. Finally, we thank anonymous reviewers whose comments helped to improve this manuscript.