Quantifying the unquantifiable: why Hymenoptera, not Coleoptera, is the most speciose animal order

The beetles (order Coleoptera), have historically [2, 3] and more recently [4–7] been described as the most speciose order of animals on Earth. The great diversity of beetles was sufficiently established by the middle of last century such that Haldane (possibly apocryphally)¹ quipped that an intelligent creator of life must have had “…an inordinate fondness for beetles” [8]. However, what evidence underlies the claim that the Coleoptera are more species-rich than the other insect orders? Certainly, more species of beetles (> 350,000) have been described than any other order of animal, insect or otherwise [11, 12], but does this reflect their actual diversity relative to other insects? Though this may seem purely an academic question, its resolution informs our understanding of patterns and mechanisms of insect evolution (e.g., [13–15]) and, to the extent that species richness is a proxy for ecological import [16, 17], how conservation efforts might best be directed.

Why are beetles thought to be so diverse in the first place? In part, historical biases in beetle collecting and an associated accumulation of taxonomic resources for the Coleoptera may have had an outsized influence on our perception of diversity. In the mid-to-late 1800s, beetles were prized among insects for their collectability. Many gentlemen of leisure—including, notably, Charles Darwin—collected beetles for sport and would make a great show of comparing the sizes of their respective collections [18, 19]. This preconception was then reinforced by studies that extrapolated from specific, targeted collections of insect diversity that focused on beetles. Of these, perhaps the highest in profile was a study conducted by Terry Erwin. Erwin [20] used an insecticide to fog the canopies of 19 individual Luehea seemannii trees in a Panamanian rainforest and then collected and identified the insect species that fell out of those trees. After having identified the proportion of the beetle species that were apparently host-specific to L. seemannii (163 of 955), he estimated that there might be as many as 12.2 million beetle species in the tropics. Similar studies seeking to estimate global insect diversity have also tended to emphasize beetles (e.g., [21, 22]).

Nevertheless, some previous work has challenged the canon, with various authors suggesting—though never quite insisting—that the Hymenoptera may be more speciose than the Coleoptera [23–26]. The premise behind this suggestion is that most of the larvae of the Parasitica (one of the two infraorders of apocritan Hymenoptera; the other is the Aculeata, which includes ants, bees, and stinging wasps), are obligate parasites of insect and other arthropod hosts that feed on the host’s tissue until the host dies (≈ “parasitoids”). Why is this parasitic life history relevant to the Hymenoptera’s proportional contribution to insect diversity? Simply put, species of parasitoid Hymenoptera (including the Parasitica, as well as some other groups such as the Orussidae and some Chrysidoidea) attack all orders of insects as well as some non-insect arthropods [27–29], and, reciprocally, most holometabolous insect species are attacked by at least one—and often many more than one—species of hymenopteran parasitoid [30, 31]. For instance, Hawkins and Lawton [32] examined parasitoid communities associated with 158 genera of British insects across five different orders, and found that parasitoid species richness ranged from 2.64 to 9.40 per host species across different host insect orders.

If parasitoid wasps are ubiquitous and most hosts are attacked by many different species, why is there any debate at all about the Hymenoptera being more diverse than other orders? One reason may be that estimates of the regional and global species-richness of parasitoid wasps remain elusive. Their small size and a relative paucity of taxonomic resources have left the parasitoid Hymenoptera relatively under-described compared to other insect orders [25, 33]. As a consequence, when collection-based estimates of regional insect diversity have been attempted, they have often excluded all but the largest and easiest-to identify families of parasitic Hymenoptera (e.g., [34–36]; though see [37, 38]).

A second reason for uncertainty regarding the species richness of the parasitoid Hymenoptera is that their host ranges are often unknown. While it may be true that most insects harbor many parasitoid species, the question remains whether these parasitoid communities are exclusively composed of oligophagous or polyphagous wasps that attack many hosts, or if instead the average insect host tends to have some number of specialist wasps among its many predators (Fig. 1). Only in the latter case would one be able to confidently assert that the Hymenoptera is the largest of the insect orders.

How then to approach this question without asking the micro-hymenopterists (and the coleopterists, dipterists, lepidopterists, etc.) to hurry up and describe all of the world’s insect species? We suggest two complementary approaches: (1) mathematically describing the values of parasitoid-to-host (“P:H”) ratios that would support—or contradict—the notion that the Hymenoptera is the most speciose insect order and (2) tabulating—wherever possible—actual P:H ratios for various genera of host insects.

For the Hymenoptera to be the largest order of insects, the global ratio of wasp parasitoids to hosts (P:H) need not—in fact—equal or exceed 1.0. Indeed, a global P:H of 1.0 (i.e., an average of one unique hymenopteran parasitoid species for each other insect species) would mean that parasitoids account for a full half of all insects. Instead, P:H ratios need only reach values such that the Hymenoptera are more species-rich than the next largest order (which, for the sake of argument, we will assume is the Coleoptera). Here, we work towards finding parameters that describe that space. First, it will be true that:

$$I\, = \,1\, - \,\left( {P\, + \,C} \right)$$

(1)

where P is the proportion of all insect species that are parasitoid Hymenoptera, C is the proportion of insects that are Coleoptera, and I is the remaining proportion of insect species, including the non-parasitoid Hymenoptera (Fig. 2a). Note that for the sake of simplicity we entirely exclude the many Hymenoptera that are parasitic on other parasitoids (“hyperparasitoids”).

Equation 3 allows us to find the values of \(p_{C}\) and \(p_{I}\) that result in a P > C or vice versa. As shown in Fig. 2, the space where P > C includes a substantial area where \(p_{C}\) or \(p_{I}\) (or both) can be < 1. For instance, if the Coleoptera make up 25% of all insects, as suggested by many contemporary authors [22, 39], a \(p_{C}\) of only 0.25 (or one species-specialist parasitoid for every four beetle species), coupled with a \(p_{I}\) of 0.50, results in P = C (and the many tens of thousands of non-parasitoid Hymenoptera will then tip the scale in their favor). Even if the Coleoptera amount to 40% of the insects, which reflects the percentage of currently-described insect species that are beetles, there will be more parasitoid Hymenoptera than beetles if \(p_{C}\) and \(p_{I}\) are equal to or in excess of 0.67 (two specialist parasitoid species for every three host species).

We can then plot \(p_{C}\) vs \(p_{I}\) for values of C between 0 and 0.5 (Fig. 3). Here, each line represents circumstances when P = C, such that the area above and to the right of each line represents values of \(p_{C}\) and \(p_{I}\) that result in a P > C. Here again, \(p_{C}\) and \(p_{I}\) need not be particularly large for the parasitoid Hymenoptera to exceed the species richness of the Coleoptera. For instance, if one quarter of all insects are beetles, \(p_{C}\) and \(p_{I}\) need only exceed 0.4 (the equivalent of two parasitoid species for every five host species).

The next question becomes: can we estimate parasitoid: host ratios (e.g., \(p_{C} , \;p_{I}\)) for different host insects? Quantifying global P:H ratios for entire insect orders is as unapproachable as the task of counting all of the living insect species: not only are most Hymenoptera undescribed, host records for described species are often incomplete, such that multiplying each host species by its supposed number of specialist parasitoids may often inadvertently include parasitoids that share hosts (Fig. 4). While this is problematic, recognition of the problem helps present paths forward. For indeed, some host–parasitoid systems are exceedingly well studied and well-understood, such that we can be reasonably confident about the completeness of the host records of at least some parasitoids. With this information, we can calculate a metric that we call the genus-specialist parasitoid:host ratio. This metric interrogates all members of a host insect genus in the same geographic region and identifies all of the parasitoids known to attack only members of that genus (the “genus-specialist” parasitoids). Because this P:H ratio ignores all parasitoids known to attack any extra-generic host—as well as those whose host range is unknown or has been incompletely studied—it is therefore an extremely conservative estimate of the overall P:H ratio for an insect genus.

Below, we present four case studies, representing host–parasitoid systems with records sufficiently complete to allow for calculation of genus-specialist parasitoid:host ratios. For each system, we focus on a single host genus in North America. We restricted geography so that parasitoid numbers would not be inflated by large biogeographic differences between hosts in their parasitoid assemblages. North America was chosen because sampling is relatively strong, and several robust resources exist for Nearctic parasitoids (e.g., [29, 40, 41]).

For each system, we searched for all literature that mentioned the name of the host genus (or historical synonyms) and either “parasite” or “parasitoid” and compiled a database of records, performing reticulated searches on each parasitoid species name as it was added to the database in order to determine known parasitoids host ranges. From among all parasitoid records, we classified parasitoids as “genus-specialists” if they had only ever been reared from hosts in this same genus. We then split these “genus-specialists” into two groups: those for which an argument can be made that they do not have unknown extra-generic hosts, and those that were “possible genus-specialists” but for which records were less complete. Non-hymenopteran parasitoids (e.g., Tachinidae) were excluded, but in any case were only present for two of the four hosts we examined (Malacosoma and Neodiprion), and generally do not have the taxonomically cosmopolitan host ranges of the hymenopteran parasitoids. For cases where host genera were found on multiple continents, only host species in North America were included in the study, and to be conservative, a parasitoid was still considered “generalist” if it occurred on an extra-generic host species outside of North America. Introduced host species were noted but not counted in host lists, as they do not represent long-term host–parasite relationships. Introduced parasitoid species were included in generalist lists, regardless of whether they were specialists on that genus in North America or elsewhere. We describe each system below and refer the reader to additional materials for species lists, specialist/generalist classifications, and citations. A summary of data across the four genera and references for these data can be found in Additional files 1 and 2.

Note that P:H ratios might be measured more accurately and/or calculated in different ways, most of which we would expect to increase the estimates of P:H reported here. For instance, rather than ignoring all of the so-called “generalist” parasitoids, one could identify those for which host ranges are known (e.g., Fig. 4), divide each by the total number of host genera attacked, and add that fraction to the numerator of the P:H ratio for the focal host genus. As one example, the “generalist” parasitoids Phygadeuon epochrae and Coptera evansi both attack only Rhagoletis flies and the currant fly Epochra canadensis. These would each add an additional 0.5 to the other 24 “genus-specialist” parasitoids of Rhagoletis, giving a revised P:H of 1.56. For Malacosoma, Dendroctonus, and Neodiprion, which all have many “generalist” parasitoids with host ranges that include only a few other extra-generic hosts in the same respective family, such additions should increase P:H ratio estimates by a considerable margin.

Another way to calculate P:H would be to focus not on a host genus but on hosts sharing the same habitat. For instance, Dendroctonus bark beetles share their habitat niche with several other species of beetle, and many of their parasitoids are “specialists” in the sense that they attack more than one bark beetle, but all within the same tree habitat [60]. One could, therefore, calculate a P:H where H is the number of potential beetle host species in the habitat, and P is the number of “habitat-specialist” parasitoid species (those that attack one or more of the hosts in that habitat and no other hosts in other habitats).

Our analyses largely ignore the increasingly common finding that many apparently polyphagous insects—both herbivores and parasitoids—show evidence of additional host-associated genetic structure that might, if considered here as distinct lineages, change P:H ratios (e.g., [76–81]). Indeed, all four of our focal host genera have named subspecies or show evidence for host-associated, reproductively-isolated lineages [57, 82–84]. Though we chose to “lump” subspecies and other reproductively isolated lineages together for this analysis, it is interesting to consider how a detailed study of genetic diversity and reproductive isolation among a host genus and all of its associated parasitoids might change P:H ratios. Studies of the flies in the Rhagoletis pomonella species complex and three of their associated parasitoids suggest that where additional host-associated lineages are found in a phytophagous insect, this cryptic diversity may be multiplied many times over in its specialist parasitoid community [51, 85]. If broadly true, this implies that genus-specific P:H ratios may often be much higher than we report here.

One sensible criticism will surely be: to what extent are the P:H ratios for these four genera reflective of global P:H ratios for their respective orders (Coleoptera, Lepidoptera, Diptera, and the non-parasitoid Hymenoptera)? Surely some insect genera escape parasitism, and perhaps the examples chosen here simply have exceptionally large, or unusually specialized, parasitoid communities. As to the former, it may be that such escape artists exist, but they also may be relatively rare. After all, there are parasitoids that attack aquatic insects [86, 87], that parasitize insects in Arctic communities (e.g., [88]), and even those that dig down into soils to unearth and oviposit into pupae [50]. The list of potential hosts for parasitoids also extends to many non-insect arthropods [89, 90]. As to the four example genera being representative of overly large parasitoid communities, all of their “overall” P:H numbers (Table 1) are actually below the means found for their respective orders in an extensive study of parasitoid communities in Britain [32], suggesting that these communities are of average, or slightly below-average, size.

A second, equally sensible, criticism is that a sample size of four offers only a limited glimpse of genus-specialist P:H ratios, and that the same data should be collected from additional host insect genera. Having started out with a list of nearly 50 potential host genera and finding host–parasitoid records to be inadequate for all but four, we wholeheartedly agree. There is a great need to assemble datasets of the type presented here for additional hosts, including from insects using other feeding niches, from other insect orders, and from insects outside of North America, especially in tropical forests. In the future, perhaps a more comprehensive analysis of P:H ratios will be possible.

While it may indeed be premature to claim that the Hymenoptera is the largest order of insects based solely on our data, many other studies offer support for the same conclusion. In fact, the preponderance of evidence suggests that the common wisdom about the Coleoptera being the most speciose is the more dubious claim. Studies of insect diversity that reduce taxonomic biases have found the Hymenoptera to be the most species-rich in both temperate [37] and tropical [38] forests, as well as in other habits (e.g., [91, 92]) and across the entirety of the British Isles [17]. In addition, a mass-barcoding study of Canadian insects found both Hymenoptera and Diptera were more diverse than Coleoptera [93]. After Hymenoptera, the Coleoptera may not even be the second most-speciose order; several recent inventories of species diversity suggest that the Diptera may hold that title [17, 94, 95]. Moreover, other historically-accepted ideas about diversity of parasitoid hymenopterans have recently been questioned, including the apparent myth that parasitoids are one of only a few groups whose diversity decreases towards the tropics [96–98]. In any case, we hope this commentary results in a redoubled effort to understand and describe the ecology and natural histories of parasitoid wasps, including host ranges and cryptic host-associated diversity, such that estimates of P:H can be made for additional host genera. We also hope to see similar efforts in other animal groups that may harbor great diversity but for which far too little is known about host ranges, such as in some particularly speciose orders of mites and nematodes (e.g., [99, 100]. In other words, and to again quote Erwin [20], we hope that “…someone will challenge these figures with more data.”