Genetic caste determination does not impose growth rate costs in Pogonomyrmex harvester ants


Correspondence author. E-mail:


1.  Caste determination in eusocial insects is the process by which individuals differentiate into reproducer or helper phenotypes.

2.  Environmental caste determination (ECD) is predicted to be more efficient than genetic caste determination (GCD), yet GCD occurs in several populations of Pogonomyrmex harvester ants.

3.  We tested whether GCD reduces efficiency by comparing colony growth rates of two GCD lineage pairs (H and J) with two closely related ECD congeners, P. rugosus and P. barbatus, under laboratory conditions over a range of temperatures. In addition, we directly compared metabolic rates of GCD and ECD larvae using flow-through respirometry.

4.  Unexpectedly, colonies from GCD lineages grew faster than colonies of P. rugosus across all temperatures, and grew at the same rate as colonies of P. barbatus. Slower colony growth rates of P. rugosus were caused by lower queen fecundity and slower larval development. Variation in developmental rate was not due to differences in larval metabolic rates, which did not differ among taxa.

5.  These results suggest that GCD in Pogonomyrmex does not impose significant productivity costs during colony growth. Instead, efficiency costs are compensated by other physiological mechanisms which may or may not be directly related to the mode of caste determination.

6.  Persistence of GCD populations in contact with ECD competitors likely stems from a life-history trade-off favouring different taxa across the geographic range of the complex: the slow growing but starvation-resistant P. rugosus dominates in resource-poor regions, while faster growing and competitive GCD populations predominate in more mesic habitats.


In many ways, social insect colonies function as ‘superorganisms’. Colony members collectively accomplish the same tasks as those performed by solitary individuals (e.g. foraging, reproduction, brood care), and colonies undergo similar ontogenetic shifts from growth to reproduction (Seeley 1989). Like cells, colony members differentiate into morphologically and behaviourally distinct forms, or castes, that have different functions. As a colony’s needs change over its lifecycle, so too do the number and ratios of castes. How this process of specialization is regulated is likely to be subject to natural selection, affecting both the equilibrium frequencies of different castes under a given environmental regime (e.g. Yang, Martin & Nijhout 2004) and the degree of flexibility of individual caste potential and colony-level caste ratios (Anderson, Linksvayer & Smith 2008).

An important determinant of fitness for perennial societies is the ability to colonize and defend territories against conspecific and heterospecific competitors (Ryti & Case 1992; Petraitis 1995). Life-history theory predicts that conditions of high competition favour rapid growth (Grime 1977). In ants, colony survival, competitiveness and reproductive success are primarily functions of numerical dominance, particularly during colony initiation and growth (Adams 1990; Tschinkel, Adams & Macom 1995; Cole & Wiernasz 1999; Holway 1999; Palmer 2004). Most ant species show a characteristic shift in reproductive caste allocation that reflects the importance of rapid growth for fitness, in which colonies delay allocation into sexual offspring until a critical size threshold is reached (Oster & Wilson 1978; Cassill 2002).

Colonies modify caste ratios by altering the developmental trajectories of offspring. Adult reproductive caste is presumed to depend on environmental conditions experienced as a larva (Wheeler 1986), with additional maternal factors that may predispose an individual towards becoming a queen or worker (Schwander et al. 2008). Such environmental determination of caste (ECD) should allow the colony to effectively shift caste ratios as required to maximize fitness by altering the cues provided to the developing brood. However, there is also growing evidence for direct genetic effects on caste propensity in ants (e.g. Winter & Buschinger 1986; Helms Cahan & Vinson 2003; Fournier et al. 2005; Hughes & Boomsma 2008), ranging from weak genetic effects on caste predisposition of individuals to cases where caste is almost exclusively determined by genotype (Anderson, Linksvayer & Smith 2008). While weak genetic effects on caste should have relatively little negative impact on resource allocation efficiency, genetically hardwired caste determination is predicted to impose significant costs because offspring biased towards a particular caste regardless of colony needs should waste resources and reduce colony growth.

Genetically hardwired caste determination (genetic caste determination, or GCD) is overall very rare, but very widespread among Pogonomyrmex harvester ant populations (Helms Cahan et al. 2002; Julian et al. 2002; Volny & Gordon 2002). In GCD populations, queens and workers are derived from distinct mating crosses (Helms Cahan et al. 2002; Julian et al. 2002; Volny & Gordon 2002). Populations contain two genetically distinct lineages, and queens may multiply with males from either lineage. Inter-lineage crosses become workers, and intra-lineage crosses become queens.

In contrast to ECD colonies in which all eggs are totipotent, eggs in GCD colonies either fail to develop or develop into an inappropriate caste if laid at a time when that particular caste (queen or worker) is not required (Helms Cahan et al. 2004; Clark et al. 2006; Schwander, Helms Cahan & Keller 2006; Volny, Greene & Gordon 2006). This imposes significant energy costs for queens when founding colonies (Anderson et al. 2006b; Schwander, Helms Cahan & Keller 2006). Significantly, the cost increases as a lineage becomes more common in the population, where queens of the common lineage acquire fewer alternate-lineage sperm appropriate for worker production and thus carry a larger inappropriate egg load. Negative frequency dependence driven by GCD has been proposed as an important force promoting persistence of the system as a whole (Helms Cahan et al. 2004).

The persistence of GCD is influenced not only by internal lineage dynamics, but also potentially by costs relative to alternative modes of determination. Under GCD, the efficiency of conversion of eggs to adults is low provided that the queen does not selectively use sperm. Estimates from colony founding suggest that even at an ‘equilibrium’ lineage frequency of 50 : 50, GCD queens would face a nearly 50% productivity penalty compared with a queen whose eggs could develop into either caste (Schwander, Helms Cahan & Keller 2006), suggesting that an ECD strategy should be able to invade a GCD population. GCD colonies compete directly with colonies of two ECD species from which the lineages themselves are derived, P. rugosus and P. barbatus (Helms Cahan & Keller 2003; Anderson et al. 2006a; Schwander, Helms Cahan & Keller 2007). Thus, any inherent costs of GCD are likely to have important ecological effects with regard to inter-specific competition.

In line with the hypothesis that GCD decreases colony productivity, direct comparison of one GCD lineage pair with P. rugosus revealed a significant overall reduction in colony initiation success and productivity in the GCD population relative to its ECD competitor (Anderson et al. 2006b). However, all comparative studies to date have been conducted on the initial brood produced by colony-founding queens. Colony founding is an unusual period in the life cycle because it is an energetically closed system (Rissing & Pollock 1986); during founding, the queen does not forage and instead produces initial workers from stored reserves. Therefore, efficiency costs associated with GCD would directly reduce initial productivity and could not be recouped through increased resource inputs that might be available once the colony begins foraging.

In this study, we tested whether GCD imposes a productivity cost due to inappropriate egg load by quantifying growth rates of GCD and ECD colonies under controlled laboratory conditions. Growth rates were measured at three different temperatures, as we predicted productivity may be more strongly depressed under high-temperature, high growth conditions. Variation in rearing temperature in the field is expected across seasons (early spring to mid-summer) as well as across the distributional ranges of the study species, from the hotter, lower altitude Sonoran desert occupied by P. rugosus, through the transitional grassland region occupied by the GCD lineages, to the cooler, more mesic Chihuahuan desert and sky island mountain ranges occupied by P. barbatus (Anderson et al. 2006a; Schwander, Helms Cahan & Keller 2007a). We focused on the period after colony founding, when colony growth rate is the most critical for success but colonies are actively foraging and not intrinsically energetically constrained. We conducted two types of comparisons. First, we compared growth rates of colonies headed by queens of each lineage of an interbreeding lineage pair to test whether predicted differences in inappropriate egg load translated into differences in colony growth. Second, we tested for an overall cost of GCD by comparing colony growth rates of the two pairs of GCD lineages, the H and J lineages, to their ECD relatives, P. rugosus and P. barbatus. To test whether putative growth rate differences among groups stem from GCD costs, we investigated four mechanisms likely to cause variation in colony growth rates. First, we tested for the role of conversion efficiency (the proportion of eggs that develop successfully into adults), because the production of genetically inappropriate eggs under GCD should result in a very low conversion efficiency in GCD colonies compared with ECD colonies. Three other mechanisms, unrelated to GCD per se, included queen fecundity, speed at which eggs hatch into larvae and rate of larval development. To investigate whether increased rates of larval development were driven by increased metabolic rates, we also used flow-through respirometry to test whether GCD and ECD larvae differ in mass-specific metabolic rate.

Materials and methods

Colony growth rates

Naturally mated foundress queens of P. rugosus, P. barbatus, and the H and J lineage pairs were collected from mating swarms in July or August 2003, 2005 and 2006 from sites in Arizona (site CG, Casa Grande, Maricopa Co.; site F, Schwander, Helms Cahan & Keller 2007a; site J, Helms Cahan & Keller 2003), New Mexico (site H, Helms Cahan & Keller 2003) and Texas (site WWR, Welder Wildlife Foundation, Sinton, San Patricio Co.) and reared in the laboratory at 30 °C. For the two GCD populations, the relative egg load predicted for queens of each interbreeding lineage was inferred from the relative frequency of queens of that lineage collected from the swarm (Table 1).

Table 1.   Relative frequencies of interbreeding lineages during the mating swarm for the two GCD lineage pairs in each collection year. Frequencies are based on collections of queens leaving the mating swarm; sample sizes are indicated in parentheses. Bold values indicate the more common lineage
Lineage pair20042007
  1. GCD, genetic caste determination.

H1/H20·44/0·56 (738)0·29/0·71 (79)
J1/J20·31/0·69 (151)0·14/0·86 (52)

Growth rate comparisons were conducted in 2004 and in 2007. Colonies of uniform size were established by isolating one queen, five workers and three advanced-stage worker larvae from laboratory colonies of each lineage; including larvae was necessary because only larvae are capable of digesting protein and redistributing such resources to adults (Sorensen, Kamas & Vinson 1983). Colonies were fed weekly with ad libitum 50 : 50 wheat germ : cornmeal mixture and one mealworm or cricket. Uneaten or discarded items were removed from colonies weekly. In 2004, all experimental queens were 1 year old, and each colony was housed in an individual 17 cm (L) × 12 cm (W) × 6 cm (H) plastic box with a single 16 × 100 mm plastic test tube with a water reservoir at one end stopped with cotton to provide a humid nest chamber and maintained at 60% humidity in an incubator (Sanyo, Osaka, Japan) at either 26 or 34 °C constant temperature. Queens of P. barbatus were not available in 2004 and were not included in the 2004 comparisons. In 2007, we repeated the 34 °C treatment and conducted an additional comparison at 30 °C. Because almost one-fourth of the queens died in the 2004 experiment at both 26 °C (18 of 70 queens) and 34 °C (19 of 75 queens), we furnished colonies with two 16 × 150 mm glass test tubes as a nest substrate to reduce desiccation stress in 2007. In 2007, none of 103 queens died in the 30 °C treatment and 2 out of the 103 queens died in the 34 °C treatment. The 2007 replicates were conducted with 2-year-old queens except for P. barbatus, which were 1 year old. To rule out any potential confounding effect of queen age on fecundity, and thus colony growth, we compared the 2004 and 2007 replicates of the 34 °C treatment for the number of larvae produced, as this measure of growth is most sensitive to queen fecundity (see Results). Comparison across years revealed no main effect of year/queen age on productivity (F1,145 = 1·03, P = 0·31). Sample sizes for all treatments and replicates are presented in Table 2.

Table 2.   Sample sizes for growth rate comparisons for each temperature regime. The year in which the comparison was conducted is indicated in parentheses
Type26 °C (2004)30 °C (2007)34 °C (2004)34 °C (2007)
Pogonomyrmex rugosus7141014
Pogonomyrmex barbatus77

All colonies were acclimatized to the experimental colony size and temperature for 1 week, after which eggs were removed (week 0). For the 30 °C treatment only, total brood production (eggs, larvae, pupae, and callows) was counted in week 2 to quantify initial fecundity. Each week thereafter, the number of larvae, pupae, callows (newly eclosed unpigmented adults) and pigmented workers was counted for all treatments. Colonies were monitored for 7 (34 °C), 8 (30 °C) or 9 (26°) weeks, enough time for two cohorts of workers to be raised. Colonies in which the queen died during the experiment were discarded. There were no significant differences among lineages and species in queen mortality rates in either temperature treatment, even when only the high-mortality year (2004) was considered (G-tests of homogeneity, 26 °C: G4 = 0·54, P = 0·97; 34 °C: G4 = 7·34, P = 0·12).


Colony growth rate can increase by two alternative mechanisms: more individuals can be reared per unit time, or individuals can move through development and join the workforce at a faster rate. To capture both of these mechanisms, we measured colony growth rate in two ways. First, to characterize total colony growth, the number of live workers at the end of the experiment was compared across lineages. This measure includes individuals added to the workforce and original workers and brood minus any workers that died during the experiment. Second, we compared the number of larvae in the final week as a direct measure of the number of individuals reared per unit time. For GCD lineages, colonies were identified by the lineage of the queen (H1 or H2, J1 or J2) rather than by the ancestry of the workers (H1/H2, J1/J2), so that we could identify effects of queen-specific characters, including fecundity and relative frequency in the mating swarm. To test for a frequency-dependent effect on productivity within each lineage pair, we first compared colony growth rates between interbreeding lineages (H1 vs. H2, J1 vs. J2). There were few significant differences (see Results), and we pooled interbreeding lineages (i.e. into an H lineage pair and a J lineage pair) for all statistical comparisons with the ECD species. Lineages and species were compared with anova, followed by post-hoc pairwise comparisons.

Mechanisms underlying colony growth rate

We evaluated the mechanisms underlying growth rate variation by calculating proxy measures from the detailed census data collected for the 30 °C treatment.

To test for an efficiency cost imposed by GCD, we calculated an egg-to-adult conversion efficiency ratio for each colony. To control for any differences in developmental rate across colonies, we used the number of pupae present 1 week after the first new pupa was observed in the colony as the number of eggs ‘converted’ into adults in the first cohort. This was divided by the corrected total fecundity (see correction below) observed in week 2 to yield a conversion efficiency value. Colonies that failed to produce a pupa by the penultimate census were excluded.

Total fecundity, or queen productivity, was calculated from the week 2 census by subtracting the number of initially provided brood from the total number censused. A potential problem with this measure is that some eggs have already hatched by week 2, and larvae are known to feed on unhatched eggs, lowering the apparent fecundity (Baroni Urbani 1988). Indeed, total number of offspring and proportion of brood that were larvae at week 2 were negatively correlated (Pearson correlation coefficient = −0·201, P < 0·05). To correct for larval cannibalism, we regressed total fecundity on proportion larvae for each lineage or species type separately and used the slope of the regression to correct for the number of eggs missing from cannibalism (corrected fecundity = original fecundity − (slope)(proportion larvae)).

Egg hatching speed was estimated as the proportion of fecundity at week 2 that were larvae rather than eggs, using the corrected measure of fecundity. anova on normalized data was used to test for significant differences across lineage and species types in each of these measures.

The week in which the first pupa was observed, excluding those developing from initially provided brood, was used as a proxy for larval developmental rate. The week of pupal appearance could be due to both earlier hatching and faster development, and we took the residuals of a regression of egg hatching rate on pupal appearance week to generate an uncorrelated variable representing rate of development from hatching to the pupal stage. Because some colonies failed to produce a pupa during the experiment, some values had to be censored at 9 weeks. To accommodate such right-censored data, a Kaplan–Meier survival test was used to compare the distribution of developmental rates across groups.

To assess the contribution of each mechanism to colony growth rate, the proportions of variance explained by fecundity, hatching speed, conversion efficiency and developmental rate were estimated with a stepwise forward regression of these variables on total worker number and number of larvae at week 8.

Larval metabolic rates

To test whether increased developmental rate is supported by higher metabolic rates, we measured mass-specific metabolic rates of individual larvae using flow-through respirometry. Larvae of similar age (c. 3 weeks old; N = 13 for GCD larvae of H1 and J1 maternal ancestry, N = 14 for GCD larvae of H2 and J2 ancestry, P. barbatus and P. rugosus) were isolated without food for c. 24 h prior to each experiment. Immediately before trials, larvae were weighed (±10 μg; Sartorius ME5, Goettingen, Germany) and photographed next to a reference grid using a DS-5M digital camera attached to a Nikon SMZ1500 stereomicroscope (Nikon Instruments, Tokyo, Japan). Larvae were then placed individually into respirometry chambers (a loop of ¼ inch copper tubing) and submerged into a constant-temperature water bath set to 30 °C. Two additional chambers without larvae were used as blank controls. The eight total chambers were sampled sequentially using a gas multiplexer (TR-RM8; Sable Systems, Las Vegas, NV, USA). CO2-free air at 55% relative humidity was flushed over all larvae at 30 mL min−1 for a 30 min acclimation period prior to trials, after which air was sampled at 30 mL min−1 for 10 min from each chamber sequentially. Chambers not actively sampled were flushed with humidified air at c. 30 mL min−1. Carbon dioxide output was measured with a calibrated infrared gas analyser (LI-7000; Li-Cor Biosciences, Lincoln, NE, USA) and logged (using ExpeData software; Sable Systems) onto a computer. Measured CO2 values were baselined using control chamber values, and converted to μmol CO2 h−1. Finally, these values were divided by larval mass to yield μmol CO2 h−1 mg−1 ant. Mass-specific metabolic rate was compared across species using a single-factor ancova with mass as the covariate.


Overall effects of temperature

Colony growth rates increased in both the rate of worker accumulation and the number of offspring reared with each increase in rearing temperature (Figs 1 and 2). Over 9 weeks at 26 °C, colony workforce increased on average 1·28-fold, with 6·38 ± 0·66 (SE) larvae in the final brood. At 30 °C, colonies increased on average 2·60-fold in 8 weeks, with 18·48 ± 1·23 larvae. At 34 °C, colonies increased on average 3·49-fold in 7 weeks, with 21·90 ± 0·72 larvae.

Figure 1.

 Average number of workers (±SE) per week at each temperature regime in (a) the H1 (solid lines) and H2 (dashed lines) lineages and (b) the J1 (solid lines) and J2 (dashed lines) lineages. For the 34 °C treatment, results for 2004 are indicated with short dashed lines and results for 2007 are indicated with broad dashed lines. Asterisks indicate significant pairwise differences at < 0·05. See Table 2 for sample sizes.

Figure 2.

 Average number of workers (±SE) in the two pairs of genetic caste determination lineages and the two environmental caste determination species at the three experimental temperatures. Asterisks indicate significant pairwise differences at < 0·05. For the 34 °C treatment, results for the J lineage pair in 2004 are indicated with short dashed lines and results for 2007 are indicated with broad dashed lines.

Interbreeding lineage comparisons: the H lineages

In 2004, the H1 and H2 lineages were nearly equal in frequency, while in 2007, the H2 lineage was approximately twice as common as the H1 lineage (Table 1). Interbreeding GCD lineages are expected to suffer reduced colony growth as a lineage becomes more common; thus, the a priori expectation was that the H2 lineage would show reduced colony growth rates relative to H1 only in those treatments conducted in 2007 (the 30 °C treatment and the 2007 trial of the 34 °C treatment). Contrary to our expectation, the only treatment in which the H2 lineage produced significantly fewer workers than the H1 lineage was in 2004, in the 26 °C treatment (F1,19 = 4·52, < 0·05; Fig. 1a). They did not differ in number of larvae at this temperature (F1,19 = 0·25, = 0·62). No differences were found in numbers of workers or larvae at 30 °C. At 34 °C, there were no significant differences in worker number, but in both years, the H2 lineage had a significantly larger larval cohort than the H1 lineage (main effect of lineage: F1,64 = 10·39, = 0·002).

Interbreeding lineage comparisons: the J lineages

The relative frequencies of the two J lineages favoured the J2 lineage in both years, though J2 was found at a higher frequency in 2007 than in 2004 (Table 1). Thus, we expected reduced growth in the J2 relative to the J1 lineage in all cases. However, as with the H lineages, lineage frequency did not have the expected impact on colony growth rates. The J lineages did not differ significantly from one another in worker number at any of the three temperatures (Fig. 1b). The lineages did not differ significantly in the size of the final larval cohort at either 26 or 34 °C; however, there was a trend towards a larger number of larvae in the J1 lineage at 30 °C (F1,35 = 2·99, = 0·09).

Unlike all other taxa, productivity of the J lineages at 34 °C differed significantly between years, with much lower worker and larval production for both the J1 and J2 lineages in 2004 than in 2007 (worker number: F1,57 = 29·71, < 0·001; larvae number: F1,57 = 12·88, = 0·001; Fig. 1b). This appeared to be due to loss of the initial introduced larvae during the first few weeks in 2004, possibly as a result of increased susceptibility of J1/J2 larvae to dry nest conditions. Because the two replicates differed significantly, we separated the results for the J lineages by year and included both in all subsequent GCD vs. ECD comparisons.

GCD vs. ECD comparisons

Across temperatures and for both measures, P. rugosus colonies grew significantly slower than the GCD lineages (Fig. 2). At 26 °C, the difference in growth trajectories was particularly apparent: although all GCD lineages produced a stable or increasing number of workers and larvae over time, P. rugosus colonies declined in both of these measures, significantly lower than the H lineage pair (Tukey’s pairwise comparisons, worker number: P < 0·04; larvae: < 0·001), and marginally non-significantly fewer larvae than the J lineage pair (= 0·07). At 30 °C, P. rugosus had significantly fewer workers than both GCD lineage pairs (Fig. 2b), as well as fewer larvae (P. rugosus vs. H: < 0·002; P. rugosus vs. J: < 0·02). At 34 °C, P. rugosus had significantly fewer workers than the H lineage pair, the 2007 replicate of the J lineage pair and P. barbatus (Fig. 2c), and fewer larvae than the H lineage pair and the 2007 replicate of the J lineage pair (P. rugosus vs. H: P < 0·001; P. rugosus vs. J: P < 0·002). Pogonomyrmex rugosus did not differ significantly from the 2004 replicate of the J lineage pair (Fig. 2c).

Pogonomyrmex barbatus did not differ significantly from either GCD lineage pair in either measure of colony growth at either of the two temperatures (30 and 34 °C) in which it was included (Fig. 2). The only exception was the 2004 replicate of the J lineage pair, which had significantly lower numbers of workers than P. barbatus and the H lineages (2004 J vs. P. barbatus: P < 0·001; 2004 J vs. H: P < 0·001; Fig. 2c).

Mechanisms underlying colony growth rate

Conversion efficiency

To look for evidence that GCD directly reduces productivity, we tested for a difference among taxa in conversion efficiency from eggs to pupae. Conversion efficiency varied significantly (F5,94 = 7·99, < 0·04). This overall effect was driven by a particularly low conversion efficiency in the J2 lineage, which was the most common GCD lineage (86%). The remaining GCD lineages were similar in conversion efficiency to the two ECD parental species (Fig. 3).

Figure 3.

 Conversion efficiencies across lineages and species from egg to pupa in the first cohort in the 30 °C treatment. Letters indicate sets that did not differ significantly in post-hoc Tukey’s pairwise comparisons.


The number of new offspring present during the second week of the 30 °C trial, corrected for larval consumption of eggs, also varied significantly by lineage type (F5,98 = 7·99, < 0·001; Fig. 4), as did the uncorrected fecundity values (F5,98 = 5·43, < 0·001). Pogonomyrmex rugosus was the least fecund, significantly lower than the two GCD lineages with the highest fecundities, H2 and J2 (P. rugosus vs. H2, < 0·02; P. rugosus vs. J2, < 0·001).

Figure 4.

 Mean queen fecundity (corrected for larval egg consumption, see text) for queens of each lineage and species at week 2 in the 30 °C treatment. Letters indicate sets that did not differ significantly in post-hoc Tukey’s pairwise comparisons.

Egg hatching speed

In contrast to fecundity, the proportion of eggs hatched into larvae did not vary significantly among lineages (Fig. 5), although there was a tendency for slower hatching in P. barbatus that may not have been detectable due to the limited sample size (n = 7).

Figure 5.

 The proportions of hatched eggs at week 2 across lineages and species in the 30 °C treatment. None of the groups differed significantly.

Developmental rate

The first pupae appeared in colonies beginning in the third week, with most colonies producing a pupa by week 6. Only four colonies failed to produce a pupa by the end of the 8-week experimental trial. Pogonomyrmex rugosus development was delayed significantly relative to all other groups, which did not differ from one another (Kaplan–Meier test with right-censored data: P. rugosus vs. H1, χ2 = 4·43, < 0·04; P. rugosus vs. H2, χ2 = 4·74, = 0·03; P. rugosus vs. J1, χ2 = 6·63, = 0·01; P. rugosus vs. J2, χ2 = 11·55, = 0·001; P. rugosus vs. P. barbatus, χ2 = 4·89, < 0·03; Fig. 6).

Figure 6.

 Cumulative distribution functions of production of the first pupa over the course of the 8-week experiment for the four genetic caste determination lineages and two environmental caste determination species.

Combined analysis

All four of the putative mechanisms underlying colony growth rate – fecundity, hatching speed, conversion efficiency and developmental rate – contributed significantly to final worker number, together explaining 67·5% of the variance (Table 3). Developmental rate was the most important variable, explaining 26·9% of the variance in total worker number, followed by fecundity, hatching speed and conversion efficiency.

Table 3.   Proportion of variance explained by each putative growth mechanism
MechanismTotal worker number (%)Number of larvae (%)
  1. *< 0·05; **< 0·01; ***< 0·001.

Pupal development26·94***4·44**
hatching speed14·26***2·62*
Conversion efficiency10·07***22·95***

In contrast to overall worker number, the number of larvae present on week 8 was affected much more strongly by conversion efficiency and queen fecundity, and much less, though still significantly, by developmental rate and hatching speed (Table 3).

Larval metabolic rate

To test whether the slow rate of larval development in P. rugosus reflected a low metabolic rate, we compared metabolic rates across taxa at 30 °C using flow-through respirometry. The technique was sensitive to individual differences in larval metabolic rate; regression analysis showed larval mass to be a very significant predictor of total metabolic rate (R2 = 56·6%, F1,80 = 104·31, P < 0·001). However, larvae did not differ among groups in their mass-specific metabolic rate (F3,77 = 0·42, P = 0·742; Fig. 7).

Figure 7.

 Box plot of mass-specific metabolic rates of larvae of the four genetic caste determination lineages and two environmental caste determination species. The central line indicates the median value, with the box bracketing 50% of the observations. Outliers are indicated with asterisks. There were no significant differences among groups.


This study indicates that Pogonomyrmex lineages do not grow slower when their caste determination system is genetic, at least during the initial growth phase of the colony life cycle. Rather, GCD lineages (H and J) outgrew one of their ECD congeners, P. rugosus, with which they compete at their western range margin, and grew at the same rate as colonies of P. barbatus, which border their range to the east.

Genetic caste determination is predicted to be costly because it does not allow colonies to adjust caste proportions to colony need (Clark et al. 2006; Schwander, Helms Cahan & Keller 2006). Moreover, the cost should rise as a lineage becomes more frequent – queens of the common lineage obtain a lower proportion of alternate-lineage sperm, reducing the proportion of eggs capable of developing into workers (Schwander, Helms Cahan & Keller 2006). Our results for colony growth after initial founding suggest that conversion efficiency is indeed reduced as a GCD lineage becomes more common (Fig. 3): the common J2 lineage was significantly less efficient at converting eggs into viable pupae, as expected if the majority of eggs laid by J2 queens were not of the appropriate genotype. However, this effect translated into little or no difference in colony growth rates between interbreeding lineages, even when frequencies were highly skewed. Conversion efficiency costs appear to be counteracted by other characteristics that may be unrelated to the GCD mechanism. This study and others (Schwander, Helms Cahan & Keller 2006; Schwander, Keller & Helms Cahan 2007) have demonstrated higher fecundity in the H2 and J2 lineages relative to their interbreeding partners (Fig. 4). Higher fecundity increases productivity (Table 3), and would therefore mitigate or exaggerate conversion efficiency costs depending on which lineage is more frequent in the population. The J1 and J2 lineages, for example, have been shown to be equivalent in initial brood production not when they are at equal frequencies, but when the J2 lineage is approximately two-thirds of the total population, when it suffers efficiency costs that match its relative fecundity advantage (Schwander, Helms Cahan & Keller 2006). A similar balance between fecundity advantage favouring the J2 lineage and conversion efficiency advantage favouring the J1 lineage occurred in this study (Figs 3 and 4), leading to equivalent colony-level growth rates. Interestingly, this frequency ratio appears to be most commonly seen in natural J populations (Schwander, Keller & Helms Cahan 2007b), suggesting that equilibrium frequencies may result from a local balance of frequency-independent and frequency-dependent forces.

The inefficiency associated with GCD is also predicted to depress the performance of GCD populations in competition with ECD populations; however, we did not observe reduced conversion efficiencies in the GCD lineages relative to either ECD species (Fig. 3), and GCD lineages did not grow more slowly than either ECD species regardless of rearing temperature (Fig. 2). Two factors appear to be responsible for this result. First, overall conversion efficiencies even in the ECD species were low: only c. 20% of eggs developed into workers even though colonies were given resources ad libitum. There are a number of possible reasons for such uniformly low conversion efficiencies. Even non-GCD eggs are not necessarily fully totipotent; a recent controlled breeding experiment in an ECD population of P. rugosus demonstrated significant genetic compatibility effects on caste potential, suggesting that ECD colonies may themselves bear some low degree of inappropriate egg load (Schwander & Keller 2008). Young colonies may also feed the majority of eggs to developing larvae, either by producing specialized trophic eggs or cannibalizing otherwise viable embryos. GCD colonies would not be at a disadvantage if genetically inappropriate eggs were recycled. Finally, colony productivity may be limited more by workforce than by egg availability. In small colonies, the queen can probably lay far more eggs than the number of larvae that can be successfully reared, buffering the colony from the effects of GCD-associated reduction in egg viability. If small colony size does limit the costs of GCD, we would predict those costs to increase as colonies grow and total brood size approaches the maximal egg-laying rate of the queen. Interestingly, direct competition experiments between adult P. rugosus and H-lineage colonies demonstrated a clear advantage of P. rugosus in interference competition over large seed piles (Julian & Helms Cahan 2006). The outcome of direct competition in ants is often determined by numerical advantage (e.g. Palmer 2004; Buczkowski & Bennett 2008), suggesting that adult colonies of P. rugosus may maintain a larger standing workforce once they reach the adult colony stage.

The second factor contributing to high GCD growth rates was the importance of mechanisms other than conversion efficiency to overall growth, for which GCD lineages actually exceeded P. rugosus. Higher fecundity of H and J queens relative to P. rugosus led to a larger standing number of brood reared per unit time, and the rate of larval development was significantly enhanced in the GCD lineages and P. barbatus relative to P. rugosus (Fig. 7). To some extent, both advantages may be explained by the ancestry of the GCD lineages. At least three of the four lineages are derived from historical interspecific hybridization between P. rugosus and P. barbatus, and the fourth (J2) is either also a hybrid or derived solely from P. barbatus (Anderson et al. 2006a; Schwander, Helms Cahan & Keller 2007a). Assuming that increased fecundity was advantageous when hybridization occurred, fixation of P. barbatus– typical alleles at loci controlling fecundity would have been highly likely even if background levels of introgression were low (Barton 2001). Rapid development in GCD worker offspring may also be derived from P. barbatus. Alternatively, GCD larvae may develop rapidly due to a genetic side effect of the GCD system: workers are invariably inter-lineage crosses, significantly increasing heterozygosity. Enhanced growth rate of hybrids due to heterosis is common, occurring in insects (Houle 1989), fish (Moav, Hulata & Wohlfarth 1975) and plants (Rood et al. 1990).

If selection favours high intrinsic growth rates, why do P. rugosus have low growth rates? Moreover, why has the range of P. rugosus not been invaded by the GCD lineages, given that it is slower growing? We hypothesize that rapid growth may be disfavoured in some environments (Arendt 1997). Low or unpredictable resources can select for stress-tolerant, slow-growth strategies that resist starvation (Gotthard, Nylin & Wiklund 1994; Hoffmann & Harshman 1999), and increase allocation to storage, immunity and defence (Schiesari, Peacor & Werner 2006). Slow growth as an adaptation to stressful environments, in combination with higher temperature requirements for successful development, may explain the parapatric distributions of P. rugosus and the GCD lineages in south-eastern Arizona and southern New Mexico. In desert areas prone to drought, the slow growth rate and low resource demands of P. rugosus would allow it to survive periodic starvation; low resource demands would also allow it, during productive times, to store more seeds than competing GCD lineages. In contrast, GCD lineages must use a greater fraction of acquired resources to support high growth rates, leaving less leftover for storage. Although individually, GCD and P. rugosus larvae have similar metabolic rates (Fig. 7), the larger total number of larvae in GCD colonies would increase resource needs per unit time two to fourfold depending on the temperature at which the brood is reared. We therefore predict that P. rugosus dominates resource-poor areas of its possible range, giving way to GCD lineages in resource-rich areas. A previous study of a contact site between P. rugosus and the H lineages revealed significant spatial clumping of the two colony types across the landscape, suggesting that local habitat features such as moisture or resource availability may affect which type is most successful at colony establishment and persistence (Helms Cahan et al. 2006).

In summary, the loss of individual developmental plasticity associated with GCD does not necessarily translate into a significant constraint on the success of the colony ‘superorganism’. Although there are inherent efficiency costs, these costs can be compensated for by other mechanisms, such as increased fecundity or more rapid development. The source of such mechanisms may come from the specific set of interacting species studied here and their associated life-history adaptations, or may be an inherent genetic consequence of the hybridization-based GCD system. A better understanding of the causes of growth rate compensation should provide important insights into when alternative modes of caste determination might be expected to evolve and persist.