Caste determination in a polyembryonic wasp involves inheritance of germ cells

Abstract

Social insects are characterized by the development of castes in which some colony members reproduce whereas others function as altruistic helpers. The conditional switch controlling caste formation usually involves environmental stimuli that act on processes that regulate development of individuals. Unlike other social species, embryos of polyembryonic wasps develop clonally to produce large numbers of genetically identical offspring and two morphologically distinct castes. All embryos in a clone exist in an identical environment, the host, yet develop into either reproductive larvae that mature into adult wasps or soldier larvae whose function is defense. Here, we report that caste determination in Copidosoma floridanum involves inheritance of germ cells. Expression of a C. floridanum homolog (Cf-vas) of the germ cell marker Vasa indicated that the B₄ blastomere in four cell-stage embryos is specified as a primordial germ cell. Vas expression later in development further indicated that embryos developing into reproductive larvae possess primordial germ cells whereas embryos developing into soldier larvae do not. Ablation of the B₄ blastomere resulted in most broods containing only soldiers whereas ablation of other blastomeres produced broods containing both castes. These results indicate that soldier larvae are obligately sterile and reveal a previously unknown role for germ cells in caste formation.

Genetic relatedness has long been considered a key factor in the transition from individuality to sociality (1–4). Most social species are insects like ants, bees, and termites that form colonies of close relatives comprised of reproductive and altruistic (nonreproducing) helper castes. Castes have also evolved in selected groups of aphids, thrips, and polyembryonic wasps (5–10). The conditional switch controlling caste formation is usually an environmental stimulus (i.e., diet, temperature, density, pheromones) that acts on internal processes regulating development (11–14). Genetic factors have also been implicated in how individuals respond to environmental stimuli (15–18). The collective effect of these inputs is that reproductive caste members have well developed gonads whereas altruistic helpers (workers and soldiers) are functionally sterile or are capable of only limited reproduction (7, 14). Despite the pronounced differences in behavior and morphology of different castes, the developmental processes regulating caste formation are largely unknown (11–20).

Unlike other social insects, polyembryonic wasps in the family Encyrtidae are parasites whose eggs develop clonally (21, 22). Copidosoma floridanum oviposits into the eggs of the moth Trichoplusia ni. The C. floridanum egg initially forms a single morula stage embryo comprised of ≈200 embryonic cells surrounded by a polar body-derived extraembryonic membrane. After hatching of the host egg, the primary morula gives rise to an increasing number of embryos, called secondary morulae, which together form an assemblage called a polymorula (23, 24). Secondary morulae arise from repeated invagination of the extraembryonic membrane, which partitions clusters of mitotically active embryonic cells into >1,000 secondary morulae by the end of the host's fourth instar. Up to 24% of these embryos develop during the larval host's first-fourth instar into soldier (i.e., precocious) larvae with fighting mandibles and an elongate body (Fig. 1A). The primary function of these soldiers is defense against inter- and intraspecific competitors (25–28). The remaining embryos develop during the host's fifth (final) instar into reproductive larvae that have reduced mandibles and a distinctly more rounded body than soldier larvae (Fig. 1B). Reproductive larvae consume the host, pupate, and mature into adult wasps. Soldier larvae in contrast always die after consumption of the host by their reproductive siblings (24).

Fig. 1.

Light micrographs of a C. floridanum soldier (A) and reproductive (B) larva. (Scale bar = 40 μm.)

How polyembryonic wasps produce distinct castes is unclear, given that progeny develop clonally from the same egg and are exposed to the same environmental conditions in the host. Here, we report that caste determination in C. floridanum involves inheritance of germ cells. Drosophila melanogaster and other holometabolous insects in the orders Diptera, Lepidoptera, and Hymenoptera specify primordial germ cells (PGCs) early in embryogenesis by means of inheritance of germ plasm prepackaged into the oocyte during oogenesis (29). Germ plasm consists of several factors, including the DEAD-box RNA helicase Vasa, which is the most widely recognized molecular marker of PGCs (29, 30). Expression of a vasa homolog by C. floridanum (Cf-vas) indicated that germ plasm asymmetrically segregates to the B₄ blastomere of four-cell-stage embryos and that reproductive larvae possess germ cells whereas soldier larvae do not. Ablation of the B₄ blastomere usually resulted in production of only soldiers, which suggests that germ cell inheritance is a key factor regulating caste formation by polyembryonic wasps.

Methods

Insects. C. floridanum was reared in T. ni at 27°C and a photoperiod of 16 h light/8 h dark as described (24, 31).

Cloning and Sequencing. Genomic DNA and total RNA were isolated from C. floridanum embryos by using TRIzol reagent (Invitrogen). RNA served as template to synthesize first-strand cDNA by using an oligo(dT)_12–18 primer and SuperScript II (Invitrogen) reverse transcriptase according to the manufacturer's recommendations. PCR cloning was performed initially from genomic DNA with degenerate primers designed to match conserved motifs in known Vasa proteins. The forward primers were 5′-TCIGGICGIGAIYTIATGGC-3′ (F1), and and 5′-CAGACGGGITCIGGIAARAC-3′ (F2), and the reverse primers were 5′-TGGAGACGRTCICCRTGDAT-3′ (R1) and 5′GARAAICCCATRTCIARCAT-3′ (R2). For the first primer set (F1-R1), reactions were performed in 25 μl total volume and were run for 30 cycles of denaturation at 95°C for 45 s, annealing at 48°C for 45 s, and extension at 72°C for 1 min, followed by a final extension at 72°C for 10 min. One microliter of the reaction was then used as a template for reamplification under the same conditions by using a second primer set (F2-R2). The products of these reactions were blunt cloned (TOPO TA, Invitrogen) and sequenced (University of Wisconsin Biotechnology Center, Madison, WI). These sequence data were then used to design specific primers, which were used with total RNA from embryos in 5′ and 3′ RACE, by using the Invitrogen 5′ and 3′ RACE kits. All sequences were analyzed by using lasergene 5 sequence analysis software (DNASTAR, Madison, WI). The sequence for Cfvas was submitted to GenBank under accession number AY604008. The RNA helicase domain of Cf-vas was aligned by using the clustalw method with gap creation penalties of 10.00 and gap extension penalties of 0.20. Phylogenetic trees were generated by using the neighbor joining clustalw algorithm with no outgroups designated. Bootstrap values were for 1,000 iterations, and graphics were constructed by using treeview.

Immunocytochemistry and in Situ Hybridization. Recombinant protein was expressed in Esherichia coli by subcloning Cf-vas into the His-tag vector pET-30 Ek/LIC vector system (Novagen). The fusion protein was used to inject rabbits by using standard immunization methods (32). The serum from boosted rabbits was affinity purified by using the His-tagged fusion protein. The antibodies were tested by Western blot analysis by using the fusion protein and embryonic extracts as described (33). In addition, a rabbit polyclonal antibody to Vasa from D. melanogaster (Df-Vas) was a generous gift from P. Lasko (McGill University, Montreal). Preliminary studies using anti-Vas antibodies generated against Df-Vas and Cf-Vas both recognized the Cf-Vas fusion protein on Western blots and yielded similar results when used in immunocytochemical studies (data not presented). Wasp eggs, polymorulae, and larvae were dissected from hosts at selected intervals for use in immunostaining or in situ hybridization studies. Progeny from a minimum of 20 hosts per time point were examined. Early-stage embryos were counterstained with phalloidin (Molecular Probes) whereas polymorulae and larvae were double labeled with an anti-histone H1 antibody (Santa Cruz Biotechnology), which localizes to all embryonic nuclei. Bound primary antibodies were visualized by using Texas red or Alexa 488-conjugated goat anti-rabbit secondary antibodies (The Jackson Laboratory, Molecular Probes). To improve visualization of the embryonic cells in individual embryos, the extraembryonic membrane surrounding embryos was removed by using 2 mg/ml Dispase (Invitrogen) for 15 min in PBS. Embryos were then fixed in 4% paraformaldehyde in PBS for 20 min. Immunostaining and in situ hybridization were then done as described (23, 34, 35). For in situ hybridizations, a single-stranded antisense probe was synthesized that corresponded to nucleotides 1089–2253 of Cf-vas (see below).

Laser Ablation. Four-cell-stage embryos were collected from parasitized host eggs 3–4 h postoviposition and placed in 60 μl of TC100 medium (Sigma) on glass slides with Teflon wells (Polysciences). Embryos were visualized by using a ×40 differential interference contrast microscopy (DIC) water immersion lens on a Leica TCS microscope. At the four-cell stage, the C. floridanum embryo consists of three blastomeres (B_1–3) of similar size that cannot be distinguished from one another and a smaller blastomere designated as the B₄ cell (see below). We used a Micropoint (Arlington Heights, IL) laser interfaced with the microscope to ablate the B₄ blastomere or one of the B_1–3 cells in individual embryos. Under this magnification and our adjustment settings, the damaging sphere of the laser beam was ≈500 nm in diameter. We scored a targeted cell as ablated if we observed its lysis. Individual embryos from each treatment were implanted into third instar hosts as described (26). Hosts were allowed to develop for 13 days and were dissected, and the number and type of progeny present were determined. We compared the broods produced from embryos in which the B₄ cell was ablated to broods from normal four-cell-stage embryos (nonablated) or embryos in which we randomly ablated one of the three large blastomeres (B_1–3). More than ≈200 embryos per treatment were implanted into hosts, and we recovered C. floridanum progeny from ≈15% of dissected hosts. The effect of treatment on brood composition (soldier larvae only or larvae of both castes) and number of offspring per host were determined by logistic regression by using jmp 3.0 software.

Results

A vasa-Like Gene from C. floridanum. Insect PGCs can often be distinguished from somatic cells during early development by using morphology and molecular markers like Vasa (29, 30). We designed degenerate primers to amplify a fragment of the vasa gene flanked by motifs that are conserved among vasa family members. These primers amplified an ≈400-bp fragment from C. floridanum genomic DNA. Twelve cloned inserts were sequenced. A sequence obtained from two identical clones corresponding to a vasa homolog was then used for specific primer design. Subsequent 5′ and 3′ RACE reactions resulted in cloning of a putatively full-length cDNA (2,550 bp) with an ORF of 2,124 bp coding for a predicted 708-aa protein designated as Cf-Vas (Fig. 2).

Fig. 2.

Sequence analysis of Cf-vas. (A) Multiple sequence alignment of the deduced RNA helicase domains of Cf-vas and Vasa-like proteins from other animals: Am-vas, Apis mellifera; Dm-vas, D. melanogaster; Bm-vas, Bombyx mori; Sg-vas, Schistocerca gregaria; Dj-vas, Dugesia japonica VasaA; Hm-vas, Hydra magnipapillata; Cs-vas, Ciona savignyi; Dr-vas, Danio rerio; Mm-vas, Mus musculus. Residues shared by a majority of genes in the alignment are shaded in green. Motifs conserved among DEAD box helicases are boxed in black. The EXRK domain conserved among Vasa and PL10 proteins but not other DEAD box helicases is boxed in red. The GIVGXA motif shared among known insect Vasa proteins is boxed in blue. (B) Schematic structure of Cf-vas, Am-vas, Dm-vas, Bm-vas, and Sg-vas. The numbers refer to the percent identity in the RNA-helicase domain (red) shown in detail in A. Outside the helicase domain, all five Vas-like proteins contain RGG repeats in their N termini. The E-X-E/D-E/D-E-X-W motif present in the C termini of insect Vasa proteins is also shown (see Results). (C) Phylogenetic tree for the multiple alignment shown in A. Sequences were aligned by using clustalw. Numbers above or below branches represent the percentage of 1,000 bootstrap iterations supporting the branch. GenBank accession numbers are as follows: Cf-vas (AY604008), Dm-vas (1054723), Bm-vas (1944405), Sg-vas (27463689), Dj-vasA (3986285), Hm-vas (10039327), Cs-vas (9955400), Dr-vas (18859541), and Mm-vas (286075). The putative Am-vas sequence was obtained by a blast search of the recently assembled A. mellifera genome accessible from the Human Genome Sequencing Center, Baylor College of Medicine (www.hgsc.bcm.tmc.edu). Only one Vasa-like protein with high identity to the Cf-vas sequence was obtained from this analysis.

Cf-Vas contained an ATP-dependent RNA helicase domain that included eight conserved motifs shared among DEAD-box proteins along with an EXRK motif present in Vasa and PL10 proteins but not other DEAD-box helicases (Fig. 2 A). The protein contained RGG repeats in the N-terminal region as present in other Vasa proteins (Fig. 2B) as well as the motif GIVGXA (residues 498–503), which seems to be unique to insect Vasa family members (36, 37) (Fig. 2 A). A second motif specific to insect Vasa proteins, E-X-E/D-E/D-X-W (residues 655–660 in Dm-Vas), was also present in Cf-Vas (Fig. 2B). Based on sequence identity, Cf-Vas was most similar to other insect Vasa proteins (Fig. 2C). We conclude that Cf-vas is a vasa homolog on the basis of mRNA and protein distribution in C. floridanum (see below).

Cf-Vas Is Asymmetrically Segregated During Early Cleavage. Unlike most insects, C. floridanum lays small, yolkless eggs that undergo total cleavage (21–24). The zygote nucleus initially migrates to the posterior pole of the egg along with a polar granule-like region that early workers studying hymenopteran embryogenesis called an oosome or nucleolo (38–40). The oosome does not stain with either nuclear dyes or phalloidin and appears as a dark area when viewed by confocal microscopy (21, 22, 24). First cleavage of a C. floridanum egg yields two blastomeres of equal size whereas second cleavage is unequal, with one blastomere forming a large and small daughter blastomere and the other forming two blastomeres of equal size (24). The oosome always segregates to the small blastomere, which we designate as the B₄ cell. Subsequent cleavage events are asynchronous and produce a morula stage embryo comprised of ≈200 cells surrounded by an extraembryonic membrane of polar body origin (24). After first cleavage, anti-Vas antibody stained the oosome that segregated to one blastomere (Fig. 3A). After second cleavage, anti-Vas stained cytoplasmic granules in the B₄ cell and in the daughters of the B₄ cell at third cleavage (Fig. 3 B and C). At the primary morula stage, anti-Vas staining labeled cytoplasmic granules in 6–12 cells (Fig. 3D).

Fig. 3.

Cf-Vas expression during early cleavage and the proliferation phase of embryogenesis. (A) After first cleavage (two-cell stage), expression of Vas protein (red) is detected only in the oosome (white arrow) inherited by only one blastomere. (Scale bar = 20 μm.) The nucleus of the blastomere is indicated with a white arrowhead. The blastomere's nucleus (arrowhead) is unstained. (B) After second cleavage (four-cell stage), the oosome segregates to the small blastomere (B₄) (white arrow), resulting in a granular staining pattern of the cytoplasm. After third cleavage (eight-cell stage), Vas is detected in two cells that are the progeny of the B₄ blastomere. (Scale bar in C = 30 μm, with scales in A and B the same.) (D) Primary morula. Vas expression is restricted to four to eight embryonic cells. (Scale bar = 36 μm.) Phalloidin labels cellular outlines (green) in A–D. (E) Polymorula explanted from a second instar host. Most secondary morulae (SM) are ≈20μm in diameter and contain ≈20–30 loosely adhered, rounded cells. Two to eight of these cells express Vas in their cytoplasm whereas the remaining cells express only H1 histone in their nuclei. A small number of larger embryos are undergoing morphogenesis to form soldier larvae (S). No cells expressing Vas are detected in these embryos. A small number of secondary morulae without Vas-expressing cells are also present. These are putative soldier embryos (PS) that have not yet initiated morphogenesis. (Scale bar = 50 μm.) (F) Higher-magnification view of secondary morulae from a third instar host. Vas-expressing cells are loosely aggregated in each embryo. The embryo in the upper right is in the process of partitioning into two new embryos. Correspondingly, four to eight Vas-expressing cells are parceled to each embryo (arrows). (Scale bar = 20 μm.) (G) Higher magnification of an embryo from a fourth instar host at the onset of morphogenesis. Embryo is oriented with dorsal up and posterior to the right. Vas-expressing cells are more tightly compacted (compare with F) and localized to the posterior of the embryo. (Scale bar = 25 μm.) Nuclei are counterstained for H1 histone in E–G.

Cells Expressing Cf-Vas Are Asymmetrically Distributed Among Embryos During Proliferation. A more complex pattern of Cf-Vas expression was observed during the proliferation phase of embryogenesis that occurs during the first-fourth instars of the host. As previously noted, the number of secondary morulae in the polymorula progressively increases during this period due to mitosis of the embryonic cells within each secondary morula and ingrowth of the extraembryonic membrane that partitions individual secondary morulae into new morulae. Some morulae during this phase of development also undergo morphogenesis to form soldiers (23–25). During this period of embryogenesis, most secondary morulae are 20–30 μm in diameter and contain 30–40 rounded embryonic cells loosely adhered to one another. Embryonic cells in these secondary morulae all have large, rounded nuclei and cytoplasm relatively clear of organelles. Embryos undergoing morphogenesis to form soldiers in contrast are larger (40–150 μm) and consist of thousands of cells in different stages of differentiation (see below). Here, we found that most secondary morulae contained 2–8 loosely aggregated cells whose cytoplasm was stained by anti-Vas (Fig. 3 E and F). In contrast, a small proportion of secondary morulae contained no cells labeled by anti-Vas (Fig. 3E). Anti-Vas also did not label any cells in larger embryos undergoing morphogenesis to form soldier larvae (Fig. 3E) (see below). A similar pattern of 2–8 labeled cells in small embryos and no labeled cells in large (soldier) embryos undergoing morphogenesis was observed in in situ hybridization experiments using a Cf-vas antisense probe (data not presented). By the middle of the host's fourth instar, the majority of secondary morulae initiate morphogenesis to form reproductive larvae. The onset of morphogenesis is first recognized by compaction of the embryonic cells in each morula to form an embryonic primordium (20). At the onset of compaction, cells stained by anti-Vas became more tightly aggregated and localized to the posterior of embryos (Fig. 3G).

Reproductive Larvae Possess PGCs Whereas Soldier Larvae Do Not. After compaction of the embryonic primordium, C. floridanum embryos of both castes undergo gastrulation and germ band extension. Progeny that develop into reproductive larvae exhibit a process resembling germ band retraction which shortens the embryo and results in a flatter, more rounded larval body. In contrast, the germ band of embryos developing into soldier larvae never shortens, which results in an elongate larval body (23–25). By the end of the germ band extension stage, embryos developing into reproductive larvae contained two clusters of 20–30 cells in the posterior of the embryo that stained with anti-Vas and hybridized with a Cf-vas antisense probe (Fig. 4 A–C). The location of these cells was fully consistent with where the future gonads would develop. Correspondingly, Anti-Vas stained the gonads present in reproductive larvae (Fig. 4D). In contrast, antibody (Fig. 4 E and F) and in situ hybridization (data not presented) did not detect any cells expressing vasa in soldier embryos or larvae (Fig. 4 E and F).

Fig. 4.

Cf-Vas protein and mRNA expression during morphogenesis of embryos developing into reproductive and soldier larvae. (A) Lateral view of a reproductive caste embryo at the onset of visible segmentation. The embryo remains coiled with the anterior (a) and posterior (p) extremities of the germ band oriented to the right. Vas-stained cells (red) are clustered in the posterior of the embryo in a location corresponding to the future gonad (arrow). Embryonic nuclei are counterstained with H1 histone (green). (B) Lateral view of an embryo hybridized with an antisense RNA probe for Cf-vas (arrow). Hybridization is restricted to the same aggregation of cells in the posterior of the embryo. (C) Ventral view of a reproductive caste embryo after uncoiling of the germ band. Embryo is oriented with anterior (a) up. Vas-stained cells (red) are restricted to two clusters of cells in the posterior corresponding to the location of the future gonads. (D) Dorsal view of a newly enclosed reproductive larva. Anterior is oriented up. Vas-stained cells corresponding to the paired gonads (arrows) localize to the abdomen. (Scale bar = 25 μm.) (E) Lateral view of a soldier caste embryo at the onset of visible segmentation. The anterior (a) and posterior (p) extremities of the germ band are in contact, but no Vas-stained cells are present. (F) Lateral view of a soldier larva. The anterior (a) and posterior (p) of the larva are oriented to the right. No Vas-stained cells are present. (Scale bars = 50 μm.)

Ablation of the B₄ Blastomere Produces Broods Comprised of Soldiers. The previous results suggested that PGCs are specified early in embryogenesis of C. floridanum by means of inheritance of the oosome by the B₄ blastomere and that development of embryos into the two castes correlates with the presence (reproductive) or absence (soldier) of germ cells. To explore these results further, we compared broods produced from the following treatments: (i) four-cell-stage embryos in which we targeted the B₄ blastomere for ablation, (ii) four-cell-stage embryos in which we targeted one of the three large (B_1–3) blastomeres for ablation, or (iii) four-cell-stage embryos in which no blastomeres were ablated (Table 1). The B₄ blastomere treatment usually resulted in broods comprised only of soldiers whereas the B_1–3 and nonablated treatments usually produced broods comprised of both castes. The B₄ and B_1–3 treatments also resulted in broods that were significantly smaller than those from nonablated embryos (Table 1). Anti-Vas staining of progeny from these treatments indicated that reproductive larvae had germ cells whereas soldier larvae did not (data not shown).

Table 1. The effect of cell ablation at the four-cell stage of embryogenesis on production of soldier and reproductive larvae.

Treatment	N	Proportion of broods with soldiers only	Mean number (±SE) of progeny per host
Ablation of the B4 cell	30	0.93	17.5 ± 20.1a
Ablation of one of the B1-3 cells	29	0.14	61.2 ± 14.2b
No cell ablated	33	0.09	243.6 ± 52.8c

The broods produced from each treatment consisted of either soldiers only or progeny of both castes (soldiers and reproductives). The proportion of broods containing only soldiers varied significantly with treatment (full model, G² = 51.2; df = 2; P < 0.0001). Calculation of likelihood ratios for each effect indicated this was due to ablation of the B₄ cell because the proportion of broods containing only soldiers did not significantly differ between ablation of one of the B_1-3 cells and control embryos in which no cell was ablated (G² = 4.22; df = 1; P > 0.20). The mean number of progeny produced per host differed significantly among treatments [F_2,92 = 24.6; P < 0.0001; log transformed data; means with the same letter were not significantly different (Tukey–Kramer method)].

Discussion

The first descriptions of a larval polyphenism in polyembryonic wasps were in the early 1900s (38, 41), but it was much later before studies demonstrated that reproductive and soldier larvae are functionally distinct castes (25–28, 42). Early workers proposed several ideas for how dimorphic larvae could develop in these clonal organisms (summarized in ref. 43), including the astute suggestion of Silvestri (38) that soldier larvae may arise from embryos lacking a “germ cell determinant.” Silvestri's explanation was criticized by several workers (44–47), but our results support his idea by showing that (i) embryos with PGCs develop into reproductive larvae whereas embryos lacking PGCs develop into soldiers, and (ii) ablation of the B₄ blastomere usually results in production of only soldiers. C. floridanum eggs and blastomeres are very small, and it is possible in our ablation experiments that we sometimes failed to kill the target cell, damaged a neighboring cell, or damaged the egg itself during injection into a host. These factors likely account for the small number of replicates where we produced broods containing both castes in the B₄ treatment and reciprocally in the B_1–3 treatments where we produced only soldiers. Ablation of any blastomere also resulted in smaller brood sizes compared with nonablated controls. Whether this reduction reflects a specific role for each blastomere in the number of progeny produced per host is unclear. Regardless, our results strongly suggest the B₄ cell becomes the first PGC by means of inheritance of germ cell determinants and ablation of this cell almost always results in broods comprised of only soldiers.

Two distinct modes of germ line specification occur in animals: localization of maternally inherited determinants early in embryogenesis (preformation) and specification of the germ line later in development by means of inductive signals (epigenesis) (29). Comparative studies suggest few clear patterns for germ cell specification in arthropods generally (29, 48). However, detailed studies in D. melanogaster and more descriptive analyses of other insects suggest germ cell specification is likely mediated by preformation in the orders Diptera, Lepidoptera, and Hymenoptera (29, 48). If so, germ cells were likely specified by preformation in the Hymenoptera before the evolution of polyembryony itself. Phylogenetic evidence indicates that soldier-producing polyembryonic wasps form a monophyletic group that arose from a monoembryonic, endoparasitic ancestor whose eggs underwent total rather than syncytial cleavage (21, 49). Other embryological adaptations that have likely favored the evolution of polyembryony include the loss of yolk and a rigid chorion that allows embryos to increase greatly in size in the nutrient-rich, aquatic environment of their insect hosts (21).

Early sequestration of the germ line potentially reduces the risk of heritable deleterious variants arising in multicellular organisms, but it comes at the cost of making asexual reproduction almost impossible (3, 4). Polyembryonic wasps are among the few animals that have overcome this constraint and reestablished secondary asexuality in the life cycle. Buss (3) speculated that secondary asexuality in polyembryonic species evolved through a heterochronic shift in the timing of germ cell determination. However, our results indicate this explanation is not the case in C. floridanum because ablation of the B₄ cell usually resulted in only soldiers, which lack germ cells but possess other larval tissues, whereas ablation of the B_1–3 cells produced broods of both castes. This finding suggests that the germ line is specified at the four-cell stage in C. floridanum, which is as early as any animal described in the literature (29). Our results also suggest that two-cell lineages, PGCs and somatic stem cells, exist during the proliferation stage of embryogenesis and that evolution of a soldier caste is due to a change in the parceling of PGCs to some but not all embryos.

The evolutionary mechanism driving these changes is likely natural selection because functional studies indicate that C. floridanum broods suffer significant fitness costs when soldiers are absent or too few reproductive larvae are produced (25–28, 50). In contrast, the developmental mechanisms regulating proliferation and asymmetric inheritance of germ cells by embryos in a brood are unclear. Monoembryonic insects lack any comparable stage of embryogenesis, but we would speculate that genes involved in the maintenance and proliferation of stem cells potentially play an important role in the proliferative stage of embryogenesis in C. floridanum. Studies on mammalian stem cells have identified several signaling molecules, transcription factors, and cell-surface proteins of potential importance in self-renewal and differentiation (51–53). Studies of germ-line stem cells (GSCs) in the gonads of D. melanogaster likewise implicate secreted signaling molecules like Decapentaplegic (Dpp) or Wingless (Wnt) in GSC self-renewal whereas membrane proteins like β-catenin and DE-cadherens are involved in controlling stem cell fate and migration (54, 55). Determining whether any of these factors have been coopted for use in polyembryony will require much more analysis.

Germ cell specification in D. melanogaster is coupled with posterior patterning during localization of determinants like nanos (56). Germ cell specification is obviously uncoupled from posterior patterning in C. floridanum because axial polarity must be reestablished in each embryo after the proliferative phase of embryogenesis (22–24). Furthermore, results of the current study suggest that PGCs likely play no role in axial patterning given their absence in embryos that develop into soldiers. Our results do reveal, however, a previously unknown function for germ cells in a complex polyphenism. Excluding gonads, soldier larvae seem to possess all major tissues and organ systems, but they develop much earlier, do not molt, and have a distinctly different morphology from reproductive larvae (22–26). Expression patterns of several major anterior-posterior patterning genes are conserved between the two castes (23), but differences in expression patterns of genes involved in other developmental pathways likely exist. Recent studies with ants, honey bees, and termites also identify differences between castes in expression patterns of genes associated with wing development and behavior (16, 19, 20, 57).

As previously noted, early sequestration of the germ line has been explained as an adaptation for controlling genetic conflicts in multicellular animals (3, 58, 59). Limiting the mitotic activity of cells that form future gametes reduces the risk of deleterious cell variants arising that would adversely affect survival of the multicellular individual. Some workers, however, have suggested that germ cell lineages may have originated during the evolution of multicellular organisms as a consequence of other cell types altruistically removing themselves from the germ line to perform somatic functions (4). C. floridanum is interesting in this regard because its broods have a clonal genetic structure that resembles cooperating cells of a multicellular individual more than the diversity of genotypes present in colonies of social insects composed of close kin. Although not physically connected to their reproductive siblings, soldiers are no different genetically from an organ performing a somatic function in the body of a multicellular individual. Viewed from this perspective, it is notable that the germ line in C. floridanum is determined early in development, yet the controls that normally limit mitosis of germ cells in monoembryonic species have clearly been relaxed given that PGCs must proliferate substantially to accommodate the thousands of progeny ultimately produced. This finding would suggest that any benefit of sequestering the germ line has been compromised by polyembryony itself. In contrast, soldier larvae do seem to be an example of caste evolution in which some individuals are removed from the germ line to perform altruistic functions of benefit to clonemates capable of reproducing.