The physical soldier caste of an invasive, human-infecting flatworm is morphologically extreme and obligately sterile

Significance

We provide substantial evidence that a globally invasive, human-infectious trematode (parasitic flatworm) in its first intermediate host possesses a physically specialized and obligately sterile soldier caste. Unlike similar-sized immature reproductives, soldier worms both lack reproductive organs and attack and kill competitor trematodes. The soldiers also possess the most extreme morphological specialization for defense yet observed in trematodes, having a giant muscular pharynx analogous to the enlarged mandibles of soldier ants. Partly due to the effectiveness of its specialized soldiers in killing competitors, Haplorchis pumilio substantially structures the ecological community of trematodes infecting its invasive host snail. This work indicates that obligate sterility, akin to the complete division of labor in the social insects, has independently evolved in the flatworms.

Abstract

We show that the globally invasive, human-infectious flatworm, Haplorchis pumilio, possesses the most physically specialized soldier caste yet documented in trematodes. Soldiers occur in colonies infecting the first intermediate host, the freshwater snail Melanoides tuberculata, and are readily distinguishable from immature and mature reproductive worms. Soldiers possess a pharynx five times absolutely larger than those of immature and mature reproductives, lack a germinal mass, and have a different developmental trajectory than reproductives, indicating that H. pumilio soldiers constitute a reproductively sterile physical caste. Neither immature nor mature reproductives showed aggression in in vitro trials, but soldiers readily attacked heterospecific trematodes that coinfect their host. Ecologically, we calculate that H. pumilio caused ~94% of the competitive deaths in the guild of trematodes infecting its host snail in its invasive range in southern California. Despite being a dominant competitor, H. pumilio soldiers did not attack conspecifics from other colonies. All prior reports documenting division of labor and a trematode soldier caste have involved soldiers that may be able to metamorphose to the reproductive stage and have been from nonhuman-infectious marine species; this study provides clear evidence for an obligately sterile trematode soldier, while extending the phenomenon of a trematode soldier caste to freshwater and to an invasive species of global public health concern.

The fortress-defense hypothesis, originally proposed to explain the evolution of a soldier caste in eusocial insect colonies (1, 2), also applies to some parasitic organisms. For many parasites, the host body, or part of the body, is both nest and food source. Encroachment on that resource could select for defense involving the formation of a soldier caste in resident parasites. Indeed, if we adopt a taxon-agnostic, trophic definition of parasitism (3), we see that many taxa recognized as having a soldier caste are, in fact, parasitic. Certain species of gall thrips and aphids, which are parasites of plants, defend their colonies against predators and competitors with a dedicated soldier caste (4–6). Some clonal wasps produce sterile soldiers inside their lepidopteran host body that attack and kill competing parasitoids to protect their reproductively immature siblings (7–9). Some snapping shrimp that are possibly parasitic on sponges have workers that defend their colony from attack (10–12). Most recently, soldiers have been reported worldwide in the first intermediate host infections of trematode parasitic flatworms (13–18). Here, we document the most extreme physical specialization yet seen in trematode soldiers.

Digenean trematodes (Platyhelminthes: Trematoda: Digenea) are parasitic flatworms that typically cycle between infecting a vertebrate final host and two intermediate hosts (19). In the first intermediate host, typically a mollusk, a single, sexually produced founder establishes a colony comprising dozens to thousands of cloned worms that collectively block host reproduction and commandeer the host body to serve the colony (15, 20). Trematode colonies often face a high risk of invasion and competitive displacement by other trematodes (21, 22). Such antagonism has led to a range of defensive adaptations, ranging from “indirect” chemical inhibition of competitors (22), to the production of a single, specialized defensive reproductive worm (23), to the division of labor between a reproductive caste and a caste of soldier worms that protect the colony from attack (13–18).

Trematode soldiers are much smaller than mature reproductives, more mobile, and use their relatively large mouthparts (pharynxes) to attack and kill competing trematodes that may attempt to infect their host snail and invade their colony (13–18). Trematode soldiers are also at least temporarily nonreproductive—they lack the brood chamber and free-floating embryos that characterize active reproductives. However, in the two cases where carefully examined, trematode soldiers have been observed to have a germinal mass embedded in body tissues (24). Because this organ produces trematode embryos, it is quite possible that soldiers maintain the physiological ability to reproductively mature. Further, a separate class of small, immature reproductives has not been documented for any trematode species with soldiers. Yet, trematode colonies must create new reproductives to replace those that die and to add new members to the colony as the host grows. It is therefore possible that some or most soldiers eventually transition to the reproductive stage (13–16, 25). In other words, previously known trematode soldiers may not represent an obligate, permanent nonreproductive caste, such as those that characterize the most advanced eusocial insects (26).

Here, we provide evidence for division of labor in a trematode involving a permanent, obligately nonreproductive soldier caste. We first document the occurrence of soldiers in the globally invasive and medically important freshwater trematode, Haplorchis pumilio Looss, 1896. This flatworm uses the snail Melanoides tuberculata (Müller, 1774) as first intermediate host, fishes as second intermediate host, and birds and mammals as final hosts (27). H. pumilio is a zoonotic parasite of humans in Southeast Asia (27). However, H. pumilio has successfully invaded the Americas along with its first intermediate host snail, where it is emerging as a potential public health threat (28–30). In addition to showing that H. pumilio possesses soldiers in the snail host, we document that its soldiers exhibit the most extreme morphological specialization for defense yet observed in trematodes. We then demonstrate the existence of a separate class of immature reproductives, and that the soldiers and reproductives follow distinct developmental paths. Hence, the soldiers of H. pumilio appear to be an obligately sterile physical caste, akin to that of the most advanced social insects. As in other organisms with a soldier caste, H. pumilio occupies hotly contested habitat: at least seven other trematode species infect the M. tuberculata host snail in the United States (29, 31–34). H. pumilio soldiers appear to effectively defend their colony from these competing trematodes. This partly explains our last finding that this species dominates the ecological structure of its trematode community where it is introduced throughout California.

Results

Soldiers, Immatures, and Mature Reproductives Are Morphologically and Developmentally Distinct.

H. pumilio forms a colony in the first intermediate host that consists of several thousand worms (rediae) (4,115 to 9,498 worms, n = 5 colonies). As in previously known trematodes with soldiers (13–17, 25), we found that the size–frequency distribution of the H. pumilio colony is bimodal, with relatively discrete small and large size classes corresponding to soldiers and reproductives (Figs. 1 and 2 and SI Appendix, Figs. S1–S3). However, unlike other trematodes, the bimodality in H. pumilio colonies involves soldiers and reproductives with obviously distinct developmental trajectories.

Fig. 1.

H. pumilio soldiers are morphologically distinct from other colony-mates. (A) A soldier redia with its greatly enlarged pharynx (bracket), which leads to a long gut possibly containing host tissue (arrow). (B) An immature reproductive redia at the same scale and magnification as (A). The small pharynx (bracket) leads to a characteristically short and empty gut (arrow). Developing embryos in the germinal mass (arrowhead) are not yet free in the brood chamber, which is still unformed. (C) An immature reproductive redia, not yet born and still developing alongside other embryos inside the brood chamber of a mature reproductive, already contains a short, empty gut (arrow) and a germinal mass (arrowhead) from which embryos will arise. All photos taken with differential interference contrast microscopy. (Scale bar, 50 μm.)

Fig. 2.

H. pumilio soldiers are morphologically distinct from immature and mature reproductives. (A) A size–frequency distribution of worms characterizing a typical colony, constructed from the average of five randomly sampled colonies. (B) Absolute pharynx volume of soldiers (red) is much larger than that of immature (green) and mature reproductive (blue) rediae, despite reproductives achieving much larger body sizes. Log–log regression of pharynx- to body-volume (regression lines and 95% confidence bands) shows that soldier pharynx size is proportional to soldier body size, while pharynx size in reproductives has no significant relationship with body size (SI Appendix, Fig. S4B and Table S2).

The primary trait distinguishing H. pumilio reproductives, both mature and immature, is that they contain offspring and/or a germinal mass (Fig. 1 B and C and SI Appendix, Fig. S3). Most reproductives are large and mature, having a well-developed brood chamber containing free-floating offspring (mostly dispersive cercariae) at various stages of development (Figs. 1C and 2 and SI Appendix, Fig. S3A). Immature reproductives lack these free-floating offspring, but do possess a clear germinal mass that often contains developing early-stage embryos (Fig. 1B). Although spanning a range of sizes, immature reproductives averaged 17% the body volume of a typical mature reproductive (95% CI: 13 to 21%), with the least developed immatures being as little as 4.7% the volume of the average mature reproductive (Fig. 2A). The range of pharynx volumes of immature reproductives was completely contained within the range characterizing mature reproductives (immatures:10^3.8 to 10^4.3 µm³; matures: 10^3.1 to 10^4.5 µm³) (Fig. 2B and SI Appendix, Fig. S4). However, on average, the pharynx volume of mature reproductives was 67% that of immatures (95% CI: 59 to 76%). The immature pharynx- to body-volume ratio was 0.017 (95% CI: 0.013 to 0.023, n = 27), which is approximately 10× larger than that of mature reproductives (0.0019, 95% CI: 0.0018 to 0.0020, n = 352). There were very few immature reproductives in a typical H. pumilio colony, composing less than 1% of all worms in an average colony (0.6%, 95% CI: 0.30 to 0.87%, n = 5 colonies) (Fig. 2 and SI Appendix, Figs. S1 and S2).

Soldiers, in contrast, were relatively common, composing 12% of the worms in an average colony (95% CI: 3.8 to 19.5, n = 5 colonies) (Fig. 2 and SI Appendix, Figs. S1 and S2). Soldiers were 4.5% the volume of mature reproductives (95% CI: 4.2 to 4.8%). Despite being the same size as the smallest immature reproductives (Figs. 1 and 2), H. pumilio soldiers are clearly morphologically distinct. First, in the 89 soldiers closely examined at high magnification for morphometrics, we never observed either developing embryos or a germinal mass. Likewise, we saw no evidence of developing embryos or a brood chamber in the 4,239 soldiers examined at lower magnification during colony censusing. Second, the gut of soldiers appeared to be much longer than that of immature reproductives, often extending to the posterior-most region of the body, and it is filled with pigmented granules instead of being empty (Fig. 1 A and B and SI Appendix, Fig. S3). Third, the soldiers possessed an enormous pharynx, 5.1 times larger (95% CI: 4.7 to 5.5) in absolute terms than that of both immature and mature reproductive worms for all sampled worms pooled (Figs. 1B and 2B and SI Appendix, Figs. S3 and S4). This relative soldier-to-reproductive pharynx size difference was consistent among four of the five colonies studied for morphometrics, which had overlapping confidence limits and a pooled soldier-to-reproductive pharynx volume ratio of 5.6 (95% CI: 5.1 to 6.2), while one colony (colony F) had a soldier-to-reproductive pharynx volume ratio smaller than that of the others (ratio = 3.7, 95% CI: 3.2 to 4.3). Despite a small degree of size overlap between the smallest soldier pharynxes and the largest mature reproductives pharynxes, not a single encountered immature reproductive had a pharynx as large as even the smallest soldier (respective ranges: 6.7 × 10³ to 20 × 10³ μm³, n = 27, versus 21 × 10³ to 94 × 10³ μm³, n = 89). Concerning pharynx- to body-volume ratios, the pharynx of soldiers was nearly a fourth of their body size (0.23, 95% CI: 0.22 to 0.25, n = 89), which was 13× greater than that of immature reproductives and 124× greater than that of mature reproductives (reported above). Finally, while soldier pharynx volume increased with body volume (log–log mixed-effects regression, slope P < 0.0001, ΔAIC versus null = −23.6), there was no evidence of a relationship between pharynx and body volume in pooled reproductives ($\mathrm{\Delta }$AIC versus null = +0.1) (Fig. 2B).

H. pumilio Soldiers Attack Competitors While Immature and Mature Reproductives Do Not.

Previously documented trematode soldiers use a muscular pharynx to attack competitors, with evidence indicating that a larger pharynx is a more effective weapon for rediae (22, 35). We tested whether H. pumilio soldiers use their enormous pharynx to attack competitors in a series of in vitro experimental trials. Although soldiers never attacked conspecifics originating from other H. pumilio colonies (95% CI: 0 to 3.6%, n = 11 trials, total worms = 104), they regularly attacked heterospecifics. They attacked heterospecific trematodes at a per-capita rate of 25% (95% CI: 20.1 to 30.4%, n = 18 trials, total worms = 265), with attacks occurring in 16 of 18 trials (Fig. 3D). Soldiers attached their mouth to heterospecific worms and used the pharynx to rupture the tegument and swallow their innards (Fig. 3 A and B). In contrast, reproductives rarely attacked heterospecifics, being observed to attack in only 1 of 18 trials and at a very low a per-capita rate of 0.74% (95% CI: 0.0 to 1.8%, n = 18, total worms = 270), or 34 times less frequently than soldiers (paired t-test, t = 9.2, df = 10, P < 0.0001).

Fig. 3.

H. pumilio soldiers aggressively attack other trematodes. (A) Three soldiers (arrowheads) attack a Philophthalmus gralli reproductive. (B) A soldier uses its muscular pharynx (arrow) to rupture the tegument of a P. gralli reproductive and suck the innards (arrowhead) into its gut (hollow arrowhead). (C) Only soldiers attacked heterospecific trematodes in a series of 3-h attack trials using matched numbers of soldiers, immatures, and mature reproductives. Within each experimental block, the numbers of H. pumilio “attackers” were constant across all stages, with each group of attackers challenged against half as many P. gralli reproductives. Among blocks, attacker numbers varied from 2 to 40. Soldier attack rate increased with increasing attacker density (logistic regression P < 0.0001). (D) Soldiers attacked heterospecifics (Cefo = Centrocestus formosanus, Phgr = Philophthalmus gralli) much more frequently than did reproductive rediae in a series of 2-h attack trials, but nearly never attacked colony-mates (“self”) or conspecifics. We interpret the single attack of a soldier against a colony-mate as aberrant behavior given stress of dissection in vitro conditions [attack rate = 0.56% (95% CI 0.03 to 3.1%, n = 179).]

To test whether immature reproductives attack at higher rates than mature reproductives, we compared the attack rates of immature reproductives, soldiers, and mature reproductives in a second series of attack trials. Higher soldier attack rates occurred in trials containing higher densities of soldiers and attackers (binomial regression, n = 8 trials, P < 0.0001), In contrast, we observed no attacks on heterospecific trematodes by either immature or mature reproductives, regardless of worm density (each n = 134).

H. pumilio Dominates Its Ecological Guild.

Because we found that H. pumilio has soldiers and is the most common trematode species infecting M. tuberculata throughout southern California (29), we predicted that it would have a large impact on the abundance of colonies of other trematodes that use the same host species. Indeed, when we compared indicators of competitive dominance across the trematodes, H. pumilio emerged as the dominant competitor. Pharynx volume is a functional correlate of competitive dominance (22); that of H. pumilio soldiers exceeded all others (Tukey HSD P < 0.05) (SI Appendix, Table S5 and Fig. 4B). Likewise, occupation of larger and older snails can indicate competitive dominance, as this reflects an ability to defend against invasion by other trematodes and/or invade already-established colonies (35); we found that H. pumilio is substantially more likely than other trematode species in its guild to occupy large snails (multinomial logistic regression, P << 0.0001) (Fig. 4 A and C). In fact, H. pumilio shifted from composing zero percent of the trematode colonies found in the smallest quarter of hosts (0 of 52 infected snails) to composing 66% of the colonies occupying the largest quarter of hosts (380 of 579 infected snails). Last, we compared the frequency of double occupancies (colonies of two trematode species infecting the same host body) across the three most common trematode species (H. pumilio, Philophthalmus gralli, and “Renicolidae sp. 1”) in southern California (29). We detected nonrandom assortment (multinomial exact goodness-of-fit P < 0.0001), as expected when competition or facilitation influence trematode colony establishment (35–37). In all cases, double occupancies involving H. pumilio were either indistinguishable from chance (post hoc binomial test with Bonferroni adjustment P > 0.05) or occurred at a lower frequency than expected (P < 0.05) (SI Appendix, Table S4). In contrast, P. gralli and “Renicolidae sp. 1” were frequently found in a greater number of double occupancies than expected by chance.

Fig. 4.

H. pumilio dominates its guild in its invasive range in southern California, USA. (A) The postulated dominance hierarchy illustrating the outcomes of competitive interactions in the M. tuberculata trematode guild of the Americas. Arrows point from dominant to subordinate species. Dominants are dominant to all subordinates of their direct subordinates. The dashed line indicates no interaction; each species behaves as if the other were not present. Species codes: Hapu = H. pumilio, Cefo = Centrocestus formosanus, Phgr = Philophthalmus gralli, Ren1 = “Renicolidae sp. 1,” Ren2 = “Renicolidae sp. 2,” Ren3 = “Renicolidae sp. 3,” “Leci” = “Lecithodendriidae gen. sp.” (B) Pharynx volumes (µm³) of the three trematode species in the USA M. tuberculata guild that possess a mouth and pharynx in the first intermediate host stage. H. pumilio soldiers had a larger mean pharynx size than other trematodes in the sample, in contrast to H. pumilio reproductives. Letters indicate post hoc Tukey significant differences (P < 0.05). H. pumilio soldiers and immature reproductives were pooled in these samples, as evidenced by the three low-pharynx-volume points in the Immature/Soldier group for that species. (C) Stacked size–frequency distributions and multinomial logistic regression (P << 0.0001, McFadden’s pseudo-R² = 0.267, n = 2,734) show that, as snails increase in size, the probability that they will be infected by H. pumilio steadily increases. Shading around regression curves indicates 95% CI. (D) Estimates of lethal competitive displacement by H. pumilio indicate that H. pumilio dramatically altered the total abundance of colonies of the other seven trematode species in the guild infecting the host snail, M. tuberculata. The observed abundances of all non-H. pumilio trematode colonies at each locality reflect the result of all competitive displacements combined. n = 487 snails across six localities (see SI Appendix, Table S3 for full locality names). Species codes: Hapu = H. pumilio, other = trematode other than H. pumilio, none = uninfected.

Combining the above information (SI Appendix, Table S5), we inferred a dominance hierarchy for the guild of trematodes infecting M. tuberculata (Fig. 4A). This hierarchy allowed us to use established methods (36) to quantify how H. pumilio structures its community. We estimated that lethal competition among trematodes reduced the abundance of competing trematode colonies to 32 to 77% of the abundances expected if lethal competition did not occur (Fig. 4D and SI Appendix, Table S3). As the dominant competitor, H. pumilio accounted for an average of 94% (range: 78 to 100%) of these lethal interactions (Fig. 4B and SI Appendix, Table S3). Removing the effect of competition of all non-H. pumilio trematodes with each other only marginally increased their expected colony abundances. Removing only the effect of competition with H. pumilio greatly increased expected colony abundances of all other trematodes, nearly to the level expected in the absence of any competitive interactions (Fig. 4B).

Discussion

Before this work, a permanent sterile soldier caste had not been documented for trematodes. Although some demographic and in vitro evidence has been interpreted as favoring caste permanency for some trematode soldiers (15, 16, 38), the reality is that all prior reports may involve species whose soldiers can, and perhaps regularly do, metamorphose to become a member of the reproductive caste (13, 14, 16–18, 25). A missing line of evidence that would help rule out such “temporal” or facultative caste formation would be the documentation of a separate class of immature reproductives clearly distinguishable from soldiers. Here, we provide such evidence for H. pumilio. This species has the most physically extreme polymorphic soldiers yet found for trematodes. Further, the soldiers differ substantially from the separately identified class of immature reproductives along morphological, developmental, and behavioral grounds. These differences are apparent even between soldiers and the most immature reproductives, suggesting that H. pumilio has a permanent, nonreproductive caste of soldiers.

Reproductive, Morphological, and Developmental Discrimination of Soldiers and Reproductives.

H. pumilio soldiers differ both reproductively and morphologically from both mature and even the most immature reproductives. Mature reproductives have a well-developed brood chamber filled with offspring, which soldiers lack. In addition, soldiers had no sign of a germinal mass, the organ in which embryos initially develop (19). In contrast, immature reproductives, even those equal in size to soldiers and lacking a brood chamber, did have a readily identifiable germinal mass containing early-stage embryos that was embedded in the posterior body tissues. This organ was even detectable in utero; the germinal mass thus forms very early in the development of a reproductive, even before birth. Yet not a single examined soldier possessed a germinal mass, indicating separate developmental trajectories and reproductive outcomes for H. pumilio soldiers and reproductives.

The documentation of a lack of a germinal mass in H. pumilio soldiers contrasts with findings on other trematode species. Although earlier research has shown that soldiers lack a developed brood chamber with developing embryos, that research did not investigate whether soldiers had a young germinal mass embedded in body tissues (13–18). In fact, such an embedded germinal mass has been detected upon close inspection for two of those species documented to have soldiers (24), suggesting that those soldiers maintain the capacity to mature. There is no such hint for H. pumilio, whose soldiers fully lack a germinal mass, consistent with them being permanently nonreproductive.

H. pumilio soldiers and reproductives—even immature reproductives—were also readily distinguishable based on the absolute size of their pharynxes. Pharynx size is functionally important, as soldiers use the muscular pharynx to attack competitors (22, 35). H. pumilio soldiers had a massive pharynx, over five times larger in absolute terms than that of reproductives and composing nearly a quarter of the soldier body volume. Soldier pharynx size scarcely overlapped with that of reproductives and never overlapped with immature reproductive pharynx size. In contrast, the pharynxes of previously documented trematode soldiers, despite being proportionally larger (relative to body size) than reproductives, are roughly equal or smaller in absolute size in comparison to the pharynxes of reproductives (13–15, 17).

Comparing these relative pharynx sizes to those of H. pumilio is particularly interesting in the context of characterizing differences between soldiers versus reproductives for other trematode species, whose soldiers may be developmentally arrested, immature reproductives. The relative pharynx size of H. pumilio soldiers was 124× greater than that of reproductives. This difference is slightly greater than the largest observed between soldiers and reproductives for other species, which range from 5× to 114× (13–15, 17). In contrast, the relative pharynx size of H. pumilio immature reproductives was only 13× greater than that of mature reproductives. That the nondefensive, immature reproductive H. pumilio possessed a relative pharynx size within the range reported for soldiers of other species clarifies the utility of cross-species studies asking whether the relative pharynx size of soldiers that may metamorphose to reproductives tend to have a greater relative pharynx sizes than immatures of species lacking soldiers (39). In any case, the substantial difference in pharynx sizes in absolute and relative terms between H. pumilio soldiers and even the immature reproductives speaks to the substantial physical specialization and division of labor that starts early in development for the soldier and reproductive caste for this species.

Different developmental trajectories are also evidenced by differing pharynx- to body-size allometries between soldiers and reproductives. For reproductives, pharynx size had no relationship with body size: the pharynx of reproductives appears to grow early in development and does not grow as the rest of the body grows. There is even a hint in our data that the pharynxes H. pumilio reproductives may shrink as they mature. In contrast, soldiers appear to start off with large pharynxes 24% of their body size and which grow proportionately with the rest of the body. Such allometric differences are a hallmark of physical castes (26).

These substantial reproductive, morphological, and developmental differences indicate that the division of labor in H. pumilio involves a caste of soldiers and reproductives with separate developmental trajectories. The finding of an immature reproductive in utero shows that reproductives can be directly birthed, ruling out that metamorphosing soldiers are the sole source of reproductives (as may be the case for other trematodes known to have soldiers). But do H. pumilio soldiers ever metamorphose to become reproductive? Although conclusive evidence is still lacking that the soldiers never shrink their pharynx size by half, shorten their long gut by approximately half, and develop a germinal mass from undifferentiated cells to become an immature reproductive, the above evidence combines with the lack of any worms in our surveys that appear intermediate between soldiers and immatures to suggest that they do not. Hence, the reproductive, morphological, developmental, and survey data indicate that H. pumilio possesses a group of obligately nonreproductive, defensive worms: a true sterile soldier caste.

Behavioral Discrimination of Soldiers and Reproductives.

Across all attack trials, mature reproductive worms displayed only a single instance of aggression. In contrast, soldiers consistently attacked heterospecific competitors, and did so at a rate comparable to that documented for other trematodes with soldiers (13–18). Interestingly, we found that soldier attack rate was density-dependent, with a greater frequency of attacks when more soldiers and more competitors were present. This finding is consistent with competitors releasing chemical cues to which soldiers respond and/or attacking soldiers releasing alarm cues that instigate other soldiers to attack. In contrast, density had no effect on reproductive attack rate, which was nearly nonexistent.

Like mature reproductives, immature reproductives were also not aggressive. This finding is particularly noteworthy in light of Galaktionov et al. (24), who suggested that immature trematode rediae typically have a different feeding mode than older, larger rediae, and directly consume host tissues rather than absorb nutrition across the tegument. In fact, they posited this as an alternative explanation to defense for the increased aggression documented for soldiers in prior studies (24). Yet, in striking contrast to the consistent attacks by H. pumilio soldiers, we observed no attacks by immature reproductives (each n = 134). We also found no evidence of a switch in feeding mode for reproductives; that is, we never observed host tissues or any solid material in the guts of immature or mature reproductives; it appears that H. pumilio immature reproductives may absorb food directly across their body wall, which is the general expectation for mature reproductives (19). Hence, our data indicate that neither eating host tissue nor aggression against competing trematodes is an inherent property of all immature trematode rediae, further clarifying the defensive nature of trematode soldier castes.

H. pumilio Soldiers Do Not Attack Conspecifics from Other Colonies.

In contrast to consistently attacking heterospecifics, H. pumilio soldiers never attacked conspecifics from other colonies (nonkin). This result is consistent with what has been observed in some other trematode species with soldiers (17), and in some other social animals with soldiers, e.g., some gall thrips (40) and aphids (41, 42). However, this finding is contrary to the strong aggression exhibited toward conspecifics in other species, such as that observed in another trematode (15), in a sea anemone (43), and in clonal wasp larvae (44).

A lack of aggression toward conspecifics is puzzling, as invasion by a conspecific trematode results in strong within-host resource competition (45–47). Further, given the high proportion of snails infected by H. pumilio in southern California (29), we would expect contact between conspecific colonies to be the most frequent antagonistic interaction experienced by this species in this geographic area. Why does H. pumilio appear to lack the ability to detect and attack nonkin conspecifics?

Interestingly, a similar lack of the ability to recognize nonkin has been reported for another introduced colonial organism, the Argentine ant (Linepithema humile). This ant appears to have lost the ability to discriminate kin in its introduced ranges as a consequence of population bottlenecks reducing genetic variation required to recognize kin (48–50). Perhaps introduced H. pumilio populations have similarly lost the high diversity of marker alleles that are theoretically required for kin recognition to operate (51, 52). Studies on H. pumilio population genetics and antagonistic interactions in the native range could shed light on this interesting issue.

H. pumilio Soldiers Are Most Common at Within-Host Invasion Fronts.

H. pumilio soldiers were not evenly distributed throughout the host body. We found relatively few soldiers in the mantle or the digestive gland/gonad region of the body, while they occurred at the highest numbers in the basal visceral mass (SI Appendix, Figs. S1 and S2). Many trematodes, such as the echinostomoids and heterophyids that frequently infect M. tuberculata in the study area, proliferate in the basal visceral mass during the early phase of infection (19); H. pumilio soldiers thus follow the general pattern documented for trematode soldiers of concentrating in areas where they will most likely encounter invaders (13–15, 17, 18), which is similar to that documented for other types of animals with soldiers [e.g., an aphid (53) and an ant (54)].

Ecological Consequences of a Strong Soldier Caste.

The finding that H. pumilio typically inhabits the largest (and presumably oldest) snails is further consistent with it having a strong ability to invade prior established colonies and to defend against usurpation by other trematodes. Previous work has shown that H. pumilio also causes its snail host to grow a thicker shell and attain a larger body mass than uninfected snails of the same shell length (55). This is consistent with studies showing that trematode species with lower risk of being killed by dominant competitors invest more resources into maintaining their “stolen” host body (56). These observations give further evidence that H. pumilio is engaged in the fortification and defense of the host snail as its “factory-fortress.”

We also found that, as the dominant competitor in its guild of parasites, H. pumilio plays a major role in shaping the patterns of infection in the southern California localities where it is present and common. Such competitive dominance must partly arise from the use of soldiers by H. pumilio to defend its habitat—the host snail—from invasion by competing trematodes. As a consequence of its ability to shape the structure of the guild of trematodes exploiting M. tuberculata as a first intermediate host, H. pumilio may have an outsized impact on “downstream” disease ecology (e.g., patterns of infection in final hosts).

It remains unclear the extent to which our findings concerning the ecological dominance of H. pumilio apply to its native range (or even elsewhere in introduced areas). In its native range, overall infection prevalence of trematodes can be as high as it can be in the Americas (e.g., refs. 57–61), indicating that within-host competition can also be intense in the native range. However, there is no work examining the competitive interactions that occur among the much greater diversity of trematode species that infect M. tuberculata in the native range (62). At least 10 of these trematodes are also human-infectious (62), further highlighting the utility of understanding the ecological role of H. pumilio and its soldier caste in its native range.

The Morphologically Extreme H. pumilio Has Direct Human Importance.

Adult H. pumilio are human-infectious parasites of great importance in their native range (27) and are an emerging public health threat in the Americas (28–30). The nonhuman hosts are of great epidemiological importance to this species. The morphologically extreme soldier caste reported here serves to protect the snail-inhabiting colony from invasion and destruction by competitors. This permits reproductive rediae in the colony to produce more infectious propagules that may eventually infect humans than would otherwise be possible. The aggressive soldier caste also closes a potential avenue of disease prevention. While some other human-infectious trematodes (such as Schistosoma blood flukes) are susceptible to biological control through competition with dominant trematodes (22, 63), H. pumilio may be too well-defended to attack with such natural enemies. As H. pumilio is an ongoing threat to human health, trematodes must now join the ants, bees, wasps, and termites in the list of socially organized animals that directly and substantially intersect with human life.

Materials and Methods

Parasites, Hosts, and Husbandry.

M. tuberculata is a parthenogenic freshwater caenogastropod native to Asia and Africa that has been introduced worldwide (64, 65). As part of an effort to assess the public health threat of the trematodes coinvading with the snail, we collected 3,164 snails in the spring, summer, and fall of 2021 from 47 fishing localities in southern California, a newly recognized area of their introduced range (29). Collections were authorized under California Department of Fish and Wildlife permit SC-7273. We used data from this initial sampling for the ecological aspects of this study and to guide additional collections. After observing the likely presence of a soldier caste in H. pumilio, we collected additional snails from two San Diego localities where trematodes were common to obtain colonies for use in attack trials and morphological analyses (130 snails from Chollas Park Lake in June 2021 and 100 snails from Lake Murray in September 2022). We housed snails for a maximum of 6 wk in plastic 1-gallon aquaria containing aerated, dechlorinated water, feeding boiled lettuce ad libitum, with weekly water changes.

Host Dissections, Colony Censuses, and Worm Morphology.

To identify and segregate trematode-infected snails for study, we placed snails individually in translucent, partitioned plastic boxes with approximately 25 mL of filtered, dechlorinated tap water and exposed them to a bank of 4T12 fluorescent bulbs 10 cm above (14,060 lx, as measured by an Apogee MQ-200 PAR meter) for 4 h. As typical, infected snails so exposed released dispersive cercariae into the water, which we identified to species as described elsewhere (29). We maintained a stock of infected snails in the laboratory from which we selected snails for study as needed. To obtain rediae (the first intermediate host stage forming the colony in the snail) for experiments, we first measured M. tuberculata shell length with Vernier calipers, then deshelled each snail by gently cracking with a hammer. We isolated worms by teasing apart snail tissues in modified Helix pomatia saline (NaCl 97.5 mM, KCl 2.0 mM, CaCl₂ 1.0 mM) (66).

For colony censuses, we used a scalpel and fine forceps to separate each snail into three regions, as previously described (13–15): the head/foot and mantle, the basal visceral mass (defined here as the region between the anterior tip of the kidney and the posterior margin of the stomach), and the digestive gland/gonad complex (also termed the “apical visceral mass” or “distal visceral mass”). Placing each region on a separate 75 × 125 mm glass plate, we used fine forceps to tease apart tissues and free worms into snail saline, then covered the teased tissues with a 50 × 75 mm glass plate inscribed with horizontal guidelines spaced 4.2 mm apart. We counted all worms at 40× magnification, scoring worms as soldiers, small “immature” reproductives (body length < 250 μm), medium-sized reproductives (250 to 500 μm), and large reproductives (body length > 500 μm) (n = 31,732 worms across five colonies). We did not attempt to count worms within the tough tissues of the head or foot, as careful exploration of the pedal blood sinus, esophagus, and associated ganglia, and the brood chamber of three snails revealed only three soldier worms around the esophagus of one individual.

We selected worms for morphometric measurements from each body region by teasing apart tissues as above and, using a clean, drawn glass pipet, exhaustively sampling all worms from a haphazardly selected (67) area of the plate until we collected approximately 70 large reproductives. We then pipetted all worms onto a 75 × 50 mm glass slide, added a 25 × 25 mm coverslip, and ensured that the coverslip was floating above the worms and not distorting them with pressure. We then heat-killed the worms by passing the slide over a flame. We photographed all collected worms with a trinocular-mounted Canon t8i DSLR. Using ImageJ 1.53t, we measured body length and maximum width and pharynx length and width. To estimate volume, we mapped simple geometric shapes onto representative photographs of soldiers, immature reproductives, and mature reproductives and compared the area encompassed by these shapes to the “true” areas obtained by tracing the outline of the body in ImageJ. On identifying a simple shape or combination of shapes that consistently yielded an area within ±10% of the true area of the body, we then parameterized that shape with morphometric measurements (e.g., body length and width) and calculated the volume associated with a symmetrical rotation of that shape. We found that a prolate spheroid well approximated the volume of reproductive worms, regardless of size. The major axis of the spheroid was the body total length, while the minor axis was the maximum width. Soldier body shape was well approximated by using two shapes in combination: a cylinder for the anterior pharyngeal region, with diameter equal to the maximum body diameter in that area and length equal to pharynx length, and a cone for the posterior region, with the base diameter equal to the maximum body diameter and the height equal to the body length minus the pharynx length. A prolate spheroid best approximated pharynx volume in both soldiers and reproductives.

We compared body and pharynx volume across colonies by one-way ANOVA of logged values, with a post hoc Tukey test of honest significant differences, as appropriate. We constructed body-size (volume) frequency distributions for each colony by applying the relative size–frequency distribution characterizing the randomly sampled individuals from the colony to the total counts in the colony. For example, if 6% of sampled soldiers (3 of 50) occurred in the 10^5.2 to 10^5.3 µm³ size range, and 500 total soldiers were recorded in the colony census, we estimated that 30 soldiers in the whole colony occurred in that size range. As we censused individual regions of the host body, we also applied this method to obtain size–frequency distributions within each body region.

To bootstrap the 95% CI around the body-volume ratios of different worm stages, we used Efron’s percentile method (68) with 10,000 draws of 89 soldiers and 379 reproductives pooled from all colonies. We used the same method for soldier-to-reproductive pharynx size ratios, both across colonies and within colonies (soldier, reproductive n for colony A: 11, 53; colony B: 33, 89; C: 13, 70; D: 8, 79; F: 24, 88), and for immature (n = 27) to mature reproductive (n = 352) pharynx size ratios. As colony A had no immatures in the random sampling for morphometrics, and colony B had only 1, we did not compare immature to mature pharynx volumes within colonies.

We used a log–log linear mixed-effects model to explore the relationship between pharynx volume and body volume. Because soldiers and reproductives clearly occupied two discrete areas of morphospace, we analyzed each stage separately for statistical simplicity and biological clarity. We set individual colony as a random effect in each analysis (SI Appendix, Table S2). For the reproductives-only model, including parthenita stage (immature versus mature) as a predictive factor yielded a model with good AIC support (SI Appendix, Fig. S4B and Table S2). However, given poor sampling of the immature reproductive stage across colonies and a biological justification for considering them to be morphologically contiguous with mature reproductives, we pooled both together into a single “reproductive” stage for analysis. Comparing models using AIC scores showed that neither random intercepts nor random slopes were favored in either model (SI Appendix, Table S2). We used parametric bootstrapping from the bootMer function in the lme4 package (69) to obtain 95% CI considering only fixed-effects-associated error. For each of the 468 observations in the dataset, we generated a sampling distribution with 500 bootstrap iterations of the model-predicted mean pharynx volume associated with that observation. We then found the SD of this sampling distribution at each point and calculated the 95% CI around the model-predicted mean at that point (Fig. 2B).

Attack Trials.

Following established methods (13–15), we challenged H. pumilio rediae against a variety of competitors in two separate attack trial experiments. In each experiment, attack trials took place in experimental blocks comprising simultaneous, different attack combinations that occurred in separate wells of a 96-well round-bottomed plate. Each block pitted rediae from a single H. pumilio colony against colony-mates (negative control) and against rediae from 1 to 3 colonies from other snails. Each experiment contained several blocks. All trials took place in darkness at a temperature of 21 to 22 °C.

The first attack trial experiment let us compare attack rates of soldiers and mature reproductives. We separately isolated soldiers and large, mature reproductives from H. pumilio colonies infecting M. tuberculata snails. Snails were dissected as above and worms were freed from tissues by teasing with fine forceps in snail saline. Following isolation, we rinsed worms three times in at least 10 mL of fresh saline. We likewise isolated and rinsed mature reproductives from colonies infecting other M. tuberculata for use as “enemies.” Three trematode species also found in southern California (29) served as enemies in these trials: noncolony-mate H. pumilio, the confamilial heterophyid Centrocestus formosanus, and the philophthalmid P. gralli. Each single-well trial within a block consisted of 15 attackers, either H. pumilio soldiers or reproductives, pooled with an equal number of enemies from a single source colony in 100 µL of snail saline; such control of count density permits a standardized method for quantifying trematode caste differences (13, 14, 17, 56). A block consisted of at least four trials: H. pumilio soldiers versus reproductive colony-mates, H. pumilio reproductives versus reproductive colony-mates, H. pumilio soldiers versus enemy worms, and H. pumilio reproductives versus enemies. Each additional enemy added another 2 trials to the block. The composition of each block is detailed in SI Appendix, Table S1: the entire experiment comprised 75 individual trials. At the start of each block, we used an eyelash pick to gather all worms with each well into a cluster to ensure maximum contact between individuals. To control for any effect of that manipulation on the results, we ensured that each worm within each well was touched by the pick. Once all worms were clustered, we adjusted the saline level in each well to 150 μL, covered the 96-well plate with a silicone sealing mat, and placed the plate in darkness for 2 h. We then scored attacks at a dissecting scope, counting a worm as “attacking” if it was firmly attached by its mouth to an enemy’s tegument.

Complete separation in the above data [all instances of some factor combinations yielded zero attacks (Fig. 3D)] precluded the use of a binomial regression model for analysis (70). Therefore, to characterize attack rates of soldiers and reproductives, we calculated means and CI using trial as the replicated data. To statistically compare attack rates of soldiers and reproductives, we used paired two-tailed t-tests (pairing by block).

In a second attack trial experiment, we compared attack rates of soldiers to those of both immature and mature reproductives. We subdivided H. pumilio attackers into soldiers, reproductives less than 500 μm body length (approximately 5.5 × 10⁶ μm³ body volume), and reproductives greater than 500 μm length. We pitted these three groups against colony-mates (reproductives greater than 500 μm length) and mature P. gralli rediae as enemies. This experiment comprised eight blocks, each with six single-well trials, for a total of 48 trials. These trials were conducted as above, save that the number of attackers and enemy worms per trial varied among blocks given variability in supply of the rare immature reproductives (but was constant within a block); the number of attackers ranged from 2 to 40 worms per well, always with half as many enemies. We waited 3 h before scoring attacks. We calculated 95% CI around observed attack rates by the Wilson score method (71). The attack rate of soldiers increased with increasing soldier density (Fig. 3C); we quantified the significance of this relationship using logistic regression.

Dominance Hierarchy Estimation.

We inferred a dominance hierarchy for H. pumilio and its competitors using the criteria of a) presence or absence of a pharynx, b) absolute pharynx size, c) patterns of infection in snails of different sizes, and d) frequency of co-occurrence with competitors.

Logic and empirical evidence indicate that the size of the muscular pharynx is important for the ability to kill competing trematodes (22). We obtained pharynx volume measurements for the three species with redia-type parthenitae (possessing a mouth and pharynx). We sampled six colonies of H. pumilio, 5 of P. gralli, and 3 of C. formosanus, employing dissection and morphometrics techniques as above (see Host Dissections, Colony Censuses, and Worm Morphology). Per colony, we targeted 10 “immatures” and 10 to 20 “reproductives.” At this time, we pooled H. pumilio soldiers and same-sized immature reproductives (if they were encountered) as we were unaware of the distinction between them. We initially compared log-transformed pharynx volumes using a linear mixed model in which trematode species and parthenita type were crossed, and individual colony was a random effect. As a fixed-effects model was more parsimonious ($\mathrm{\Delta }$AIC $-$ 2.5), we pooled all colonies within a species. We used a Tukey post hoc analysis to quantify the differences between pharynx sizes across species and among immature/soldier worms and reproductives.

The colonies of competitively dominant trematode species can accumulate in larger snail size classes (35). We measured the shell height of all 3,164 snails from the initial survey of 47 localities throughout southern California (29). Each snail was subsequently dissected and trematode infections identified. We then constructed size–frequency distributions of uninfected snails, snails infected with H. pumilio, and snails infected with any of the other seven encountered trematode species. We estimated the probabilities that a snail of a given size would be found with no trematode, H. pumilio, or any other trematode using multinomial logistic regression implemented in the nnet package in R (72). We assessed model significance with a likelihood ratio test and verified goodness-of-fit with McFadden’s Pseudo-R² (values between 0.20 and 0.40 indicate excellent fit) (73).

To evaluate co-occurrence frequencies with competitors, we compared observed colony frequencies to the null expectation of no competition for the three most common trematodes in the guild (H. pumilio, P. gralli, and “Renicolidae sp. 1”). This analysis was restricted to three localities in southern California in which all three species occurred at a prevalence of ~5% or greater (29) to maximize our ability to detect double occupancies (coinfections of the same snail) among those species. To calculate expected frequencies of double occupancies, we first found the expected prevalence of double occupancies by multiplying the locality-level prevalence of each species in a pair. Under an assumption of random assortment, the prevalence of double occupancies should be equal to the product of each species’ prevalence. We then multiplied the estimated prevalence by the total number of snails in each locality’s sample to obtain the expected number of double occupancies for each species pair. We used the R package XNomial (74) to calculate a multinomial goodness-of-fit test of nonrandom assortment within each locality. To assess whether frequencies of double occupancies differed significantly from expectations at the level of each species pair, we employed post hoc binomial tests with Bonferroni correction.

Combining the above data (SI Appendix, Table S5) permitted us to assign each trematode species a dominance rank.

Impact of Within-Host Competition on Trematode Colony Abundances.

Having a dominance hierarchy allowed us to use established methods (36) to estimate the effect of lethal within-host competition on the abundances of colonies of the different trematode species in the M. tuberculata guild. To employ this method, we combined the observed prevalences of trematodes in their first intermediate host snails (percent hosts harboring a trematode colony) with our estimated dominance hierarchy to estimate the abundance of trematode colonies expected if competition were not present (see “$e_i$, the expected prevalence of each species i in the absence of any competitive interaction” below). By comparing this expected value to observed colony prevalences, we then quantified the effect of within-snail lethal competition on trematode colony abundances.

In a previous study, we quantified the prevalence of trematode colonies infecting M. tuberculata in southern California lakes and ponds (29). Of the 20 localities with infected snails, 6 had sufficient overall prevalence (>20%), sufficient H. pumilio prevalence (>5%), and high enough sample size (>50) to merit competition analysis. We restricted our analysis to snails between 25 and 35 mm in length to help control for host age (21). We estimated that this size range included snails with approximately 1 y of postmaturity growth, with a growth year estimated as 8 mo out of the year. No growth curves for the M. tuberculata clone(s) present in southern California have been published. Such curves do exist for clones invading Florida (75) and the West Indies (76), but those snails attained smaller asymptotic sizes (27 and 22 mm, respectively) than what we have observed in California (35 to 52 mm). We therefore fit a Richards curve to estimate the growth of southern California M. tuberculata, assuming that initial size is 2.4 mm (76), size at first reproduction is 15 mm (75, 76), time to first reproduction is 100 d (76, 77), and asymptotic shell length between 35 and 50 mm. From this curve, we estimated that snails between 25 and 35 mm in shell length roughly represented a same-year cohort, spanning approximately 120 d at a minimum and 460 d at maximum. To account for spatial heterogeneity in parasite recruitment, we performed our competition analysis at the scale of localities (e.g., lakes), as that was the scale at which we had sufficient sample size (>100) to permit competition analysis. Although spatial heterogeneity at finer spatial scales (e.g., sample sites within localities) can influence estimates of competitive loss, the prevalence between sample sites within a locality was statistically different at only two of the six study localities and the effect of pooling sites within a locality typically results in underestimating the degree of competitive loss in trematode communities (21). Our results therefore provide conservative lower bounds of the true loss of trematode colony abundance due to competition at the study localities.

We began by finding $e_i$, the expected prevalence of each species i in the absence of any competitive interaction, with the equation

$$e_i=\frac{{(O}_i-D_i)/N}{1-{\sum }_{j=1}^n{[m_{i,j}(O}_j-D_j)/N]},$$

where $O_i$ was the observed number of colonies of trematode i, $N$ was the number of hosts examined for infection, and $D_i$ was the observed number of double occupancies (two trematode species infecting the same host) where the cohabiting trematode is dominant to species i. The element $m_{i,j}$ of the n x n competition matrix, $\mathbf{M}$, described the outcome of species j invading a snail occupied by species i, as indicated by the postulated dominance hierarchy (Fig. 4A and SI Appendix, Table S5): in an encounter between a dominant and a subordinate species, the dominant had a competition outcome of 1 and the subordinate had an outcome of 0; a trematode encountering an equal-ranked competitor had an outcome of 0.5. We made the assumptions that 1) priority effects did not influence the outcome of interspecific competitive interactions, and 2) invaders always failed to displace conspecifics ($m_{i,i}=0$). In the above, ${\sum }_{j=1}^n{[m_{i,j}(O}_j-D_j)/N]$ represents the summed prevalences of all trematode species dominant to species i plus half the prevalences of equal-ranked competitors, excepting double occupancies with dominant trematodes (36). We then calculated $E_i$, the number of colonies of species i we would expect to see in the absence of all competition:

$$E_i=e_{i}N.$$

Being a theoretical expectation of abundance calculated from prevalence, $E_i$ could take on fractional values. The relative proportion of trematodes of species i that were killed due to H. pumilio, ${\rho }_i$, was

$${\rho }_i=\frac{E_{i,H}-O_{i,}}{E_i-O_i},$$

where $E_{i,H}$ was the abundance of trematodes of species i that we would expect in the absence of competition with H. pumilio. We directly calculated $E_{i,H}-O_i$, the estimated number of colonies of trematode species i lost to competition with H. pumilio, by applying the expected prevalence of species i in competition-free space to the space occupied by H. pumilio:

$$E_{i,H}-O_i=e_i{O}_H.$$

To find the total relative proportion of all trematode colonies killed by H. pumilio at a locality, ${\rho }_{\mathit{total}}$, we took the ratio of the expected colony losses due to H. pumilio and the expected colony losses due to all competition:

$${\rho }_{\mathit{total}}=\frac{{\sum }_{i=1}^n{e}_i{O}_H}{{\sum }_{i=1}^n(E_i-O_i)}.$$

Image Processing.

We lightly edited photographs in GIMP 2.10.32 by adjusting brightness and/or contrast and to remove background clutter, as needed.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

R code, morphometrics data, colony census data have been deposited in Dryad (https://doi.org/10.5061/dryad.2280gb5zf). Previously published data were used for this work (29).