Evolution Icon Evolution
Paleontology Icon Paleontology

Fossil Friday: How the Caterpillar Got Its Legs, or Not

Photo: Caterpillar in Baltic amber, by Manukyan Andranik via Wikimedia, CC BY-SA 4.0 International).

This Fossil Friday features a caterpillar trapped in 45-million-year-old tree resin of Eocene Baltic amber. A caterpillar of course looks very much different from the butterfly into which it eventually develops. The wonderful metamorphosis of caterpillars into butterflies was first discovered by the British physician William Harvey (1651) and Dutch biologist Jan Swammerdam (1669), and famously featured in paintings by the pioneer entomologist Maria Sybilla Merian in her book Metamorphosis insectorum Surinamensium (Merian 1705).

The more primitive groups of insects like roaches, locusts, cicadas, and bugs have a so-called hemimetabolous development, where the nymph is similar in body plan to the adult insect, and with each molting grows in size and especially in length of the wing sheaths. However, most insect species, and indeed most animals on our planet, belong to Holometabola, the clade of insects with complete metamorphosis, which includes lace wings, beetles, bees and wasps, mosquitoes and flies, scorpionflies and fleas, as well as caddisflies and butterflies. In these insects the larva has a very different body plan from the adult insect. After the final larval stage there is a resting stage called pupa or chrysalis, in which the larval body is mostly dissolved into a kind of cell soup and rearranged into the adult body plan. This miraculous development was featured in the Illustra Media documentary Metamorphosis (Illustra Media 2011Klinghoffer 2011) and poses a considerable conundrum for evolutionary biologists.

The Nature of the Conundrum

Only three hypotheses for the evolution of metamorphosis in insects have been presented: one, which suggests that the holometabolan larva is equivalent to the hemimetabolan nymph, fell out of favor decades ago. Another hypothesis suggested that the holometabolan caterpillar originated from a weird hybridization event of an insect with a velvet worm (Williamson 2009), which is generally considered as preposterous nonsense (Giribet 2009; also see Evolution News 2011). The currently preferred hypothesis is based on very old ideas of Harvey (1651) and Berlese (1913), which were further developed and elaborated by Truman & Riddiford (1999200220192022). Their so-called pronymph hypothesis suggests that the juvenile stages of hemimetabolous and holometabolous insects are not homologous, but that only the hemimetabolan pronymph is equivalent to the caterpillar larva, and the multiple nymphal instars are all equivalent to the pupal stage (also see Grimaldi & Engel 2005). However this hypothesis faces two formidable challenges:

  1. The proymph is a non-feeding final embryonic stage, lacking functional mouth parts, which hatches from the egg and immediately molts into the first nymphal instar. The caterpillar larva is a pure feeding stage, basically a gut with legs. How could one evolve into the other with functional and advantageous intermediate forms?
  2. Likewise, how could a single pupal stage, in which the complete body plan is dissolved and rearranged (including even the brain, see Truman et al. 2023 and Saplakoglu 2023), evolve via viable transitional forms from a normal series of nymphal instars that gradually transform into the adult with each molting? This appears to be not just unlikely but inconceivable and virtually impossible. Therefore, this hypothesis is controversial even among mainstream biologists, who have raised many objections to the interpretation of the pupa as only nymphal stage (e.g., DuPorte 1958). All that evolutionists have to offer are vague speculations such as this: “Perhaps 280 million years ago, through a chance mutation, some pro-nymphs failed to absorb all the yolk in their eggs, leaving a precious resource unused. In response to this unfavorable situation, some pro-nymphs gained a new talent: the ability to actively feed” (Jabr 2012). Easy peasy.

Anyway, we should expect that such a marvellous mode of development evolved from normal nymphal stages, if at all, after hundreds of millions of years of gradual change. However, that is not at all what the fossil record shows.

Actually, the first holometabolan insects are recorded from the same Pennsylvanian period as the first flying insects. Molecular clock data even suggest that Holometabola are at least as ancient (about 328-318 mya) as the earliest fossil record of flying insects (Labandeira 2011), or place “the origin of Holometabola in the Carboniferous (355 Ma), a date significantly older than previous paleontological and morphological phylogenetic reconstructions” (Wiegmann et al. 2009a2009bMisof et al. 2014). My dear colleague and frequent co-author André Nel (2019) recently commented that “the late Carboniferous was also the time of the oldest known holometabolous insects, with complete metamorphosis (wasps, beetles, scorpionflies).” Indeed, fossils from larval and adult holometabolous insects of different orders have been found in late Carboniferous layers (see Kukalová-Peck 1997, Nel et al. 20072013Béthoux 2009Kirejtshuk & Nel 2013Kirejtshuk et al. 2014).

Early and Abrupt Appearance

This very early and abrupt appearance of the highly complex holometabolan metamorphosis represents one of the many examples of the waiting time problem, because it certainly required many coordinated mutations, which again required orders of magnitude more time to originate and spread than was available.

But any theory for the origin of the caterpillar larva needs to explain a lot more than that. While hemimetabolan nymphs and all adult winged insects have only three pairs of thoracic legs, the caterpillar larvae of butterflies and plant wasps additionally possess several pairs of chubby abdominal leglets called prolegs. “These prolegs pose an evolutionary mystery, and scientists have long grappled over how and why they got them” (Pallardy 2023). Where did those prolegs come from? There are three alternative hypotheses on the table:

  1. Prolegs are serially homologous with thoracic legs, and thus derived from reactivated abdominal legs of crustaceans. This alternative was challenged and arguably refuted by previous evo-devo studies like Yue & Hua (2010) and Oka et al. (2010).
  2. Prolegs are novel adaptations without immediate precursor structures.
  3. Prolegs are derived from endites, internally facing structures of the crustacean limbs (e.g., Oka et al. 2010).

Now, a new study by Matsuoka et al. (2023) tested these three hypotheses with evo-devo data. The authors suggest that prolegs are novel traits, but based on the re-activation of pre-existing endite genes. The press release makes it very clear: prolegs “seem to be modified endites. As crustaceans evolved into insects, endites were largely lost. But in butterflies and moths, the gene for them got reactivated, providing caterpillars with their prolegs.” (Pallardy 2023).

To evaluate the feasibility of this hypothesis we first have to look at the distribution of prolegs within holometabolan insects, because this character is not hierarchically distributed as would be predicted by Darwinism, but instead is very incongruent (homoplastic): prolegs occur in the larvae of plant wasps, scorpionflies and fleas, caddisflies and butterflies, and some families of flies, but are absent in all other holometabolan groups. This incongruent pattern implies that prolegs were either reduced multiple times, or instead originated independently as a convergence, which is also suggested by developmental data (Suzuki & Palopoli 2001). Actually, Hinton (1955) proposed an independent origin of prolegs 27 times within Diptera. This alone is a grandiose empirical failure of Darwinian theory, because the unique anatomical similarity does not seem to be plausibly based on inheritance from a common ancestor. For the sake of the argument we will let this pass and just look at the new study.

Assessing the New Study

As we have seen above, Darwinists now explain the origin of caterpillar leglets with the reactivation of a crustacean gene, that was dormant for maybe 100 million years. Seriously? After such a long period without function and without adaptive pressure to eliminate deleterious mutations, this gene should still have remained functional instead of been degraded by random genetic noise? This would be akin to a genuine miracle and arguably would violate an assumed law of evolution known as Dollo’s Law, which is based on the simple fact that history does not repeat itself (Gould 1970). Could this law be broken on some realistic time scale?

As shown by Rana (2017), there were several studies that evaluated the time frame in which the function of a gene is degraded and lost, so that it cannot be reactivated:

  1. The study by Marshall et al. (1994) suggested that reactivation is reasonable over time scales of 0.5-6 million years. The authors concluded that “the reactivation of long (>10 million years)-unexpressed genes and dormant developmental pathways is not possible unless function is maintained by other selective constraints.”
  2. Lynch & Conery (2000) showed that duplicated genes lose function by stochastic silencing within a few million years. Rana (2017) mentions that such duplicated genes can serve as proxy for dormant genes, because they are no longer under the influence of selection. Lynch and Conery found a half-life of 4 million years, which implies that function is lost after 16-24 million years.
  3. Protas et al. (2007) showed that such a loss of function happens much more quickly, in about 1 million years, if it is advantageous and thus influenced by selection.

Horne (2010) studied the reactivation of eye sight in blind ostracods and commented that “there appear to be several well-documented examples of the reactivation of dormant genes, allowing the reappearance of ‘lost’ characters, in some cases after several [my emphasis] million years.” 

So, we have a realistic time frame of roughly 1-24 million years for the reactivation of a dormant gene. Indeed, short term reversals can be observed in lab experiments, e.g., concerning drug resistance among germs (Gouda et al. 2019). Anything longer than the mentioned time constraint is prohibited by Dollo’s Law of irreversibility (Gould 1970Bull & Charnov 1985). Any apparent reactivation on longer time frames (see examples mentioned by Fryer 1999Dingle 2003Cruickshank & Paterson 2006Horne 2010, and Rana 2017) cannot be reasonably explained with Darwinian processes, but requires intelligent design as more plausible and causally adequate explanation. Rana (2017) correctly emphasized that “it is not unusual for engineers to reuse the same design or to revisit a previously used design feature in a new prototype.”

But There Might Be a Loophole

Lynch (2022) recently found that “the long half‐life of enhancers, transcription factor binding sites, and protein−protein interaction motifs suggest that evolutionary reversals are possible after much longer periods of loss than previously suspected.” He concluded that “these data indicate that reactivation of these smaller functional units is possible after many millions of years and suggest that re-evolution of complex traits may occur through their loss and regain. Thus these data suggest that organisms need not surmount “the sheer statistical improbability … of evolution ever arriving at the same complex genic end‐result twice” (Müller in Gould 1970), rather “organisms might only need to retrace a single step such as the reacquisition of a transcription factor binding site in a cis‐regulatory element of a protein−protein interaction motif.”

However, there is a caveat, because the longer half-life does not apply to silenced protein coding genes, which would degrade much faster. Lynch (2022) explicitly admitted that “it seems unlikely that the genetic information for the development and function of the character can be maintained for long periods of time in the absence of the character (Bull & Charnov, 1985).” This could only be avoided in cases of serially homologous characters, when at least one instance of expression of this character would remain, so that selection could work against the deterioration of function by random copy errors. Therefore, Lynch (2022) suggested as a loophole for the violation of Dollo’s law “that the developmental programs required for the establishment of serially homologous characters may never really be lost so long as a single instance of the character remains.”

How Would Darwinists Argue?

So, let’s have a look at the possibility that this loophole could allow for the reactivation of the endite gene in caterpillars as suggested in the new study by Matsuoka et al. (2023). Probably, Darwinists would argue as follows: putative homologs of abdominal leg endites are present as pairs of eversible vesicles on the abdomen in some primitive wingless insects (apterygotes) like diplurans, bristletails, and silverfish that are known from (controversial) Devonian and Carboniferous fossils. Such vesicles are absent in all known winged insects and thus were reduced in the stem species of crown group pterygotes, which lived at least 323 million years ago in the earliest Pennsylvanian (Namurian) period according to the oldest fossil record (Brauckmann et al. 1994Brauckmann & Schneider 1996Prokop et al. 2005Prokop & Hörnschemeyer 2016Wolfe et al. 2016), and 410 million years ago in the Late Devonian period according to molecular clock estimates (Wiegmann et al. 2009bMisof et al. 2014). This is 10 million years prior to the oldest fossil record of holometabolans (313.7 mya, Wolfe et al. 2016) and 60 million years prior to molecular clock estimates of their origin (350 mya, Wiegmann et al. 2009a2009bMisof et al. 2014 / 328-318 mya according to Labandeira 2011). Therefore, the transformation would have occurred after 60-10 million years of gene suppression if the prolegs would belong to the ground plan of holometabolan insects. This would still reach or exceed the above mentioned limits imposed by Dollo’s law.

But it gets much worse for the evolutionist hypothesis. As we have seen above, larval prolegs do not belong to the ground plan of holometabolan insects (Peters et al. 2014), but developed independently multiple times in several crown groups among them. Therefore, we have to look at the age of those crown groups and not the age of Holometabola as a whole to evaluate the available window of time. Let’s be maximally generous and assume that larval prolegs are at least homologous in caddisflies (Trichoptera) and butterflies (Lepidoptera), so that they could belong to the ground plan of their common amphiesmenopteran ancestor. According to The Timetree of Life (Wiegmann et al. 2009a2009b) the relevant crown groups originated in Permian and Triassic periods: Lepidoptera, for example, about 230 million years ago and Amphiesmenoptera (the clade of Trichoptera+Lepidoptera) 282 million years ago. This molecular dating roughly agrees with the early fossil record of these groups. This implies that the reactivation of the dormant gene would have occurred after 128-41 million years (410/323-282 mya) of absence of any instantiated serially homologous character, which is simply impossible according to the limits proposed by mainstream evolutionary biology itself.

Is This Hard Science? Really?

Of course, such inconvenient facts do not bother evolutionary biologists at all, because the law obviously must have been broken, because we know it happened. Apparently laws do not mean much in evolutionary biology and can be suspended whenever a just-so story requires it. Sounds like hard science — not!

Why is it that you cannot find such simple calculations as we just made above anywhere in the mainstream scientific literature, to check if a scenario is plausible and compatible with other claims of evolutionary theory? Are the scientists really interested in testing their hypotheses and eventually finding out that they don’t hold water? It certainly doesn’t look like that to me. In spite of all the scientific efforts by Darwinists, the origin of complete metamorphosis in holometabolan insects remains an unsolved mystery, which is much better and causally more adequately explained by intelligent design.