Tiny individuals attached to a new Silurian arthropod suggest a unique mode of brood care

Significance

The paper reports a remarkable arthropod from the Silurian Herefordshire Lagerstätte of England. The fossil reveals a unique association in an early Paleozoic arthropod involving tethering of 10 tiny individuals each by a single thread to the tergites so that their appearance is reminiscent of kites. The evidence suggests that these are juveniles and that the specimen records a unique brooding strategy. This is part of a diversity of complex brooding behaviors in early arthropods heralding the variety that occurs today. The possibility that the small individuals represent a different arthropod, possibly parasitic, which colonized the larger individual, seems less likely.

Abstract

The ∼430-My-old Herefordshire, United Kingdom, Lagerstätte has yielded a diversity of remarkably preserved invertebrates, many of which provide fundamental insights into the evolutionary history and ecology of particular taxa. Here we report a new arthropod with 10 tiny arthropods tethered to its tergites by long individual threads. The head of the host, which is covered by a shield that projects anteriorly, bears a long stout uniramous antenna and a chelate limb followed by two biramous appendages. The trunk comprises 11 segments, all bearing limbs and covered by tergites with long slender lateral spines. A short telson bears long parallel cerci. Our phylogenetic analysis resolves the new arthropod as a stem-group mandibulate. The evidence suggests that the tethered individuals are juveniles and the association represents a complex brooding behavior. Alternative possibilities—that the tethered individuals represent a different epizoic or parasitic arthropod—appear less likely.

Evidence of brooding in fossil arthropods is unusual and normally confined to eggs and early juveniles: later stage juveniles are rarely encountered. Among the highlights described from the Silurian Herefordshire Lagerstätte are ostracods preserving soft parts, including evidence of a brooding strategy that persists today: eggs and possible early juveniles are held within the space at the rear of the carapace (1). Here we report a new larger arthropod from the same fauna, with smaller arthropods attached to the tergites by means of long threads. These smaller individuals lie within or are associated with a cuticular capsule, the largest about 2 mm in length, with a gape through which the appendages emerged. They preserve evidence of ∼6 pairs of appendages in contrast to 15 (four of them in the head) in the adult. The evidence suggests that the attached individuals are juveniles that must have added segments during the transition to an adult morphology, a strategy established in trilobites, eucrustaceans, pycnogonids, and other “Orsten” forms and in short great appendage arthropods by the early Cambrian (2–4). If so, the parent may be a female, although male brood care is known in arthropods (in pycnogonids eggs are carried by the male, which is equipped with ovigers).

Results

Aquilonifer spinosus is a new genus and species of arthropod from the Herefordshire Lagerstätte, a late Wenlock (mid-Silurian) volcaniclastic deposit in Herefordshire, United Kingdom (5, 6). It is preserved, as are the other fossils from this Lagerstätte, in three dimensions as a calcitic void fill in a carbonate concretion (7). The name of the new taxon refers to the fancied resemblance between the tethered individuals and kites, and echoes the title of the 2003 novel The Kite Runner by Khaled Hosseini (aquila, eagle or kite; -fer, suffix meaning carry; thus aquilonifer, kite bearer; spinosus, spiny, referring to the long lateral spines on the tergites). The material is a single specimen, the holotype OUMNH C.29695, registered at the Oxford University Museum of Natural History (Fig. 1 and Movie S1).

Fig. 1.

Holotype of A. spinosus gen. et sp. nov., “virtual” reconstructions. (A) Dorsal view. (B) Ventral view with juveniles omitted. (C) Ventral oblique view of right head appendages and hypostome (stereo pair). (D) Ventral–oblique (stereo pair). (E) Juvenile 5, oblique view with associated capsule. (F) Juvenile 5, lateral view. (G) Lateral view with juveniles removed. (H) Anterior–oblique view (stereo pair) of posteriormost head appendage and anterior trunk appendages showing exopods. (I) Proximal part of antenna showing spines (stereo pair). (J) Dorsal view without appendages (stereo pair) with juveniles numbered as referred to in text. (K) Anterodorsal–oblique view. (L) Anterior view (stereo pair). (M) Dorsal view of posterior of trunk (stereo pair). (N) Anterior view of trunk limb 9 (stereo pair). ap, juvenile appendages; b, basis; c, claw; ca, capsule; e, endopod; g, gut; h1–4, head appendages; hy, hypostome; t, telson; t1–11, trunk appendages; x, exopod. Numbers refer to trunk tergites, attached juveniles, or appendage podomeres as appropriate. (Scale bars, 1 mm.)

Diagnosis.

Features include a head shield with rostrum-like anterior projection, large uniramous antenna, chelate limb, and two other biramous appendages in the head, the last similar to those of the trunk; an elongated trunk with long, slender lateral spines on the 11 tergites, with all trunk somites bearing limbs of which all but the last are biramous; and a short telson and long cerci.

Description.

The head shield is subtriangular in dorsal view (Fig. 1 A and J); the margins are incompletely preserved. The posterior area is raised medially into a broad axial ridge, which is also present along the length of the trunk (Fig. 1 J and K). An anterior rostrum-like projection extends forward and somewhat ventrally a distance similar to the length of the rest of the head shield (Fig. 1 J and K). An apparent series of four or five short slender lateral spines near the base of this projection are artifacts of preservation (Fig. 1J). The sides of the head shield bear a paired series of at least four long slender spines, projecting antero-laterally and curved convex dorsally (Fig. 1J). The spines increase slightly in length from anterior to posterior and are similar in morphology to those on the trunk tergites. A swelling in the axial area on the ventral side of the head, which is aligned with the attachment of appendage 3, is interpreted as a hypostome (Fig. 1 B and C).

There is no evidence of eyes. The first three head appendages are morphologically differentiated, whereas the fourth appears very similar to those of the trunk (Fig. 1 B–D and H).

Due to incomplete preservation proximally and lack of information on the interior morphology of the head, it is not possible to determine the sequence in which the first two head appendages insert. The relative position of antenna and chelate appendages in other Paleozoic arthropods, however, suggests that the uniramous nonchelate appendage (the antenna) is anteriormost (8).

Head appendage 1 (green; Fig. 1 A, B, D, G, I, and K), as designated here, is uniramous, antenniform, and large. The right appendage is the better preserved (the reconstruction of the left is incomplete distally). The angle of the slices (see Methods), subparallel to the length of the appendage, makes the proximal part difficult to interpret, but it may consist of three or four segments similar in length to the more distal ones or, perhaps less likely, a long basal segment. The appendage tapers gradually to a point. The individual podomeres are narrower proximally and expand distally (Fig. 1I) to a point about their midlength where they bear two short narrow spines, which project dorsolaterally relative to the orientation of the trunk; more spines may have been present. The podomeres taper distally beyond the spines to their articulation with the next podomere. Only the segments in the proximal half of the appendage are easy to enumerate—neither spines nor podomere boundaries are evident more distally. Spine bases are evident on the left appendage but not the spines themselves. Extrapolation suggests that the total number of podomeres is about 25. This first appendage is about the same length as the body, including the “rostrum” but excluding the cerci.

Head appendage 2 (pink; Fig. 1 A–D, K, and L) extends forward but not beyond the anterior projection of the head shield. Subtle changes in direction along the length of the right limb suggest that there may be as many as five proximal podomeres, but this is not certain (the slices run along the length of the limb, rather than transverse to it, obscuring details). The appendage terminates in a laterally directed swollen chela-like structure, which terminates distally in two slender curved finger-like projections. A poorly preserved laterally directed projection from near the base of appendage 2 (better preserved on the left example but difficult to reconstruct; Fig. 1 A and B) may represent a slender exopod, but its nature is uncertain.

Appendage 3 (blue; Fig. 1 A–D, G, K, and L) is biramous. A broad basis expands adaxially and may project into a gnathobase. It gives rise to an endopod, which extends abaxially and then curves axially so that the distal and proximal podomeres are subparallel. Geniculations suggest the presence of five or six podomeres and a terminal spine. The exopod is much longer, is more slender, and projects laterally. That on the right appendage bends sharply ventrally and curves outward distally; it may end in a series of short podomeres (Fig. 1L).

Appendage 4 (yellow–green; Fig. 1 A–D, G, H, K, and L) bears an endopod similar to that of appendage 3, likewise with evidence of five or six podomeres. The exopod is evident on the left side, where it is very short and projects just a short distance anteriorly (this ramus is incompletely preserved and has been lost on the right limb, but the data available are consistent with the morphology of the trunk appendage exopods).

The trunk consists of 11 divisions (tergites) of similar length; the first two and the last one are slightly shorter than the rest (Fig. 1A), which may reflect a gradient in growth rate along the trunk axis (9). The trunk is near parallel sided, tapering markedly only in the last three tergites (Fig. 1M). Each tergite is comprised of a broad, gently convex axial ridge occupying about half its width (excluding the long slender lateral spines), flanked by lateral areas that are slightly concave dorsally (Fig. 1 J–L). Two short triangular lateral projections of trunk tergites 1–10 bear long slender spines, curved concave dorsally. These lateral spines are approximately evenly spaced along the length of the trunk. Only the posterior spine is preserved on the left side of tergite 10, and the anterior spine, together with a hint of the posterior one, on the right side. Tergite 11 appears to bear just one spine on each side, which projects posteriorly (Fig. 1M). The boundaries between the tergites are marked by transverse grooves in the axial area (Fig. 1 J and K). The position of the maximum height of a tergite lies progressively further posteriorly in tergites 7–9 (Fig. 1K).

The first trunk appendage (appendage 5, blue–green) is similar to the posteriormost appendage of the head (Fig. 1 B–D and H). A broad basis expands adaxially; it may project into a gnathobase, but there is a significant gap between the opposing members of the pair here and in successive limbs. The basal podomere gives rise to an endopod, which extends abaxially and then curves axially so that the distal and proximal podomeres are subparallel. Geniculations in the right appendage suggest the presence of six podomeres and a terminal spine. The right appendage preserves a short incompletely preserved exopod projecting forward.

Trunk appendages 2–10 (appendages 6–14) are similar in morphology to the first trunk appendage. They increase slightly in size to trunk appendage 6 and decrease slightly in the more posterior appendages (Fig. 1 B and D). The basis projects adaxially, and the right limb of trunk appendage 7 preserves delicate spines. Left trunk appendages 8 and 9 preserve possible evidence of segmentation in the endopod (Fig. 1N). Four stout proximal podomeres are evident followed by a distal section of apparently two podomeres (left trunk endopod appendage 9) terminating in a slender claw (i.e., six podomeres + claw). The exopod is a long, flat, forward projection. The orientation of the slices combined with indifferent preservation makes it appear filamentous, but its structure is unknown. This exopod is not evident in appendage 11 (this is unlikely to be a preservational artifact, as the exopod is clearly present in appendage 10).

The trunk terminates in a small conical projection that extends beyond the last trunk tergite (Fig. 1M). This projection (referred to here as the telson) bears a pair of slender parallel cerci that are about three-quarters the length of the rest of the body (Fig. 1 A and G). It is unclear whether these structures are annulated.

The gut is preserved as an impersistent sediment fill; it becomes visible dorso-medially in the head because it lies too close to the head shield for the intervening material to be visualized (Fig. 1 A, J, and K). The position of the anus is unknown.

The length of the body from the tip of the rostrum-like projection of the head shield to the posterior margin of the telson is 9.5 mm. The large first antenniform appendage is about the same length (9.5 mm), and the cerci are about 7.3 mm long.

Apart from its unusual morphology, the other remarkable feature of the arthropod is the attachment of multiple individuals to the trunk tergites (Fig. 1 A, D, and J–L). These 10 individuals, which are best seen when the trunk limbs are removed (Fig. 1J), are enumerated clockwise in what follows starting from the anteriormost on the right side (Fig. 1J). They are shaped like flattened lemons. They consist of an outer “shell” (here referred to as a capsule) that does not appear to be calcified. The shell is generally ∼15–20 µm thick where it is thinnest (Fig. S1B) but may be thicker in places perhaps as a result of soft tissue adhering to the inner surface or the orientation of the capsule to the grinding plane. The capsule opens distally exposing filamentous structures within. Some capsules, such as that of individual 3, show a narrow ridge along one margin, which may represent a kind of hinge (Fig. 1J and Fig. S1A). The largest capsules (individuals 3, 6, and 9) are about 2 mm in length (Fig. 1J). In some cases the filamentous internal structures are separated from the capsule, particularly in individuals 1 and 5 (Fig. 1J). The smallest capsules (individuals 4 and 10) are less than 0.6 mm long (Fig. 1J). Thus, the capsules are characterized by a significant size range (the largest is ∼4× the length of the smallest). Most individuals preserve a mass of tissue associated with the capsule, and individuals 2, 3, and 5 in particular preserve evidence of multiple paired slender projections that represent limbs (Fig. 1 E, F, and J and Fig. S1), although the details are difficult to interpret due to their small size relative to the spacing of slices. Individual 5, which is preserved outside its capsule, shows at least six pairs, some of them evident as curved lines on a surface exposed during grinding (Fig. 1 E and F and Fig. S1B). The body extends and tapers beyond the obvious appendages through a length similar to the appendage-bearing part. Individual 3 sits within its capsule and shows at least three pairs of limbs projecting out of the gape (Fig. S1C). The smallest capsules (individuals 4 and 10) preserve hints of soft tissue within the capsule but no evidence of specific structures.

Fig. S1.

Juveniles attached to holotype of A. spinosus gen. et sp. nov. (A) surface with juveniles 2 and 3 captured during reconstruction, (B) surface with juveniles 5 and 6, and (C) “virtual” reconstruction (stereo pair) of juvenile 3. (Scale bar, 1 mm.) ap, appendages; ca, capsule.

Each capsule is borne by a slender flexible thread that originates where the capsule tapers to a point. This proximal area of the capsule is thickened (Fig. S1 A and B). The thread expands abruptly just beyond the capsule and tapers gradually to a long slender portion that affixes to the host (Fig. 1J). Some of the threads appear to be discontinuous (e.g., that of individual 1), but this is interpreted as a reconstruction artifact. The threads vary in length from about 1.5 mm (that of individual 5) to 3.3 mm (that of individual 9) (Fig. 1J). The threads are attached to the slender lateral spines on the tergites, except for those of capsules 4 and 10, which are attached to the main part of a tergite (Fig. 1J).

Discussion

Phylogenetic Position of the New Genus and Species.

The combination of characters in A. spinosus differs from that in any other known arthropod, living or fossil, and we therefore assign it to a new genus and species. Aquilonifer shows some similarity to Artiopoda, but when added to the analysis of Legg et al. (10) (with minor modifications; see Methods), it falls out as a stem-group mandibulate lying above the Marrellomorpha and below those “Orsten” forms that cannot be placed in crown-group Crustacea (Fig. 2). Transposing the order of head appendages 1 and 2 (see Description) yields longer trees (142.60540 steps versus 142.16612 steps) with a largely unresolved topology. Modes of development are coded in the phylogenetic analysis (see Methods and citations therein), but more derived brooding strategies are very diverse, particularly among Eucrustacea (11), and provide little constraint on phylogenetic position.

Fig. 2.

Cladogram showing the phylogenetic position of A. spinosus gen. et sp. nov. Shown is a strict consensus of the 12 most parsimonious trees of 142.16612 steps (consistency index = 0.513; retention index = 0.870), produced using New Technology search options in TNT (tree analysis using new technology) and using implied character weighting with a concavity constant of three. Numbers above nodes are GC support values. 1, Euarthropoda (crown-group); 2, total-group Chelicerata; 3, Artiopoda; 4, total-group Mandibulata; 5, Mandibulata (crown-group).

The Nature of the Attached Individuals.

The very small size and consequent lack of detail revealed by the grinding technique make the individuals attached to Aquilonifer difficult to interpret. However, their size and morphology are inconsistent with protozoan ciliates such as peritrichs or with epiphytic algae. The outer covering of the capsules resembles a carapace that encloses the body and opens at one extremity. The absence of a mineralized shell, and presence of soft tissue beyond the capsule, together with the apparent symmetry, eliminates brachiopods. The serially arranged paired structures within the capsules, about six in number (Fig. 1 E and F and Fig. S1) and sometimes projecting out or separated from the capsule, represent segmented appendages. Thus, the evidence indicates that the attached individuals are arthropods.

Arthropods attached by a thread are likely to represent one of three possible strategies: they are either parasites, epizoans, or brooded juveniles. Comparative behaviors are most readily sought among living crustaceans because they are by far the most diverse group of aquatic arthropods today. Parasitic forms may retain appendages for a motile phase in the life cycle. Living tantulocarids develop in a sac-like structure derived from the tantalus larva to which they are connected by a kind of umbilical cord, and the larva in turn is attached to the host crustacean (12, 13). Parasitic thoracican barnacles may retain cirri even though they feed by absorption through the peduncle (14). A variety of parasitic copepods use a system of rootlets, some threadlike, to absorb nutrients from a variety of different hosts (13). The individuals attached to Aquilonifer, however, are unlikely to be parasitic because there would be no advantage in such long threads for absorbtion, and their most common attachment position, on the slender lateral spines of the host, is not a favorable site for accessing nutrients.

The gape at the distal end of the capsules attached to Aquilonifer would have facilitated feeding with the appendages. Among living epizoans, thoracican cirripedes such as Octolasmis, which infest larger crustaceans today (15), are similar to these attached individuals. Some thoracicans, such as Pagurolepas, which live in association with hermit crabs, have reduced the calcified plates that armor the capitulum (16). The threads that tether the capsules to Aquilonifer, however, are much more slender and longer than the robust muscular peduncle of thoracican cirripedes. The attached individuals are also different to the larval stages of the cirripede Rhamphoverritor reduncus from the Herefordshire Lagerstätte, which are about twice the size, even though they represent developmental stages before attachment to a substrate (17). Given the potential for diversification among arthropods, as exemplified by living crustaceans, the individuals attached to Aquilonifer could represent an unknown type of epizoan; other lines of evidence, however, argue against this possibility. Epizoans have been reported from the Herefordshire biota—on brachiopods (18, 19)—but similar capsules to those described here have not been observed tethered to any other animal from the fauna. Furthermore, the “host” arthropod was clearly living when the capsules became attached: it is unlikely to have tolerated the presence of so many drag-inducing epizoans, and head appendage 1 is long enough to have cleaned the trunk tergites (“general body grooming” as in some living crustaceans) (20). Thus, the attached individuals are more likely to be juveniles.

Tethering of capsule-like structures containing tiny individuals is consistent with a brooding strategy, albeit one with no exact parallel among living arthropods; it would have protected the juveniles from predation by keeping them close to the parent. Attachment by a stalk occurs in the embryos of freshwater crayfish (Astacida), for example, which are tethered to the adult (21, 22). Some of the individuals attached to Aquilonifer show evidence of limbs: about six pairs are evident in individual 5, for example (Fig. 1 E and F). The length of the body in individual 5 would accommodate sufficient pairs to make up the number in the host: they may not be preserved or have not yet developed fully. Release of the juveniles would have to have occurred within a molt cycle of the adult, but this may have been extended to avoid them being discarded.

The size of the capsules varies from ∼0.5–2.00 mm. A diversity of larval sizes is also known in recent ostracods (23), and eggs and juveniles have been reported together in individuals of the ostracod Nymphatelina from the Herefordshire Lagerstätte (1) and in ostracods from the Ordovician Beecher’s Bed (24). Embryos brooded by the living crayfish Procambarus pass through the earliest stages “rather synchronously,” whereas rates of development vary thereafter so that stage 3, 4, and 5 juveniles from the same batch may occur together under a mother’s abdomen (21, p. 573). Similar patterns could explain the variation in size of the capsules attached to Aquilonifer. Alternatively the range in size may indicate that the breeding adult accommodated more than one generation by molting at long intervals.

The correlation between the size of each capsule and the length of its thread is not statistically significant (Figs. S2 and S3 and Table S1). The correlation becomes stronger (although it is only significant at P < 0.10) when the thread length is augmented by the distance between its point of attachment and the lateral margin of the trunk of the host (i.e., the base of the slender tergal spines). Thus, molting may have included lengthening of the thread in a manner similar to epizoic thoracican barnacles (25) perhaps to improve access to particulate food (Table S1).

Fig. S2.

Plot of length of thread versus length of capsule (estimated dimensions in mm).

Fig. S3.

Plot of length of thread plus “advantage” versus length of capsule (estimated dimensions in mm). For explanation, see Table S1.

Table S1.

Dimensions of capsules, threads, and their attachment to the adult

	A	B	C
Juvenile number	Capsule length	Thread length	“Advantage”	B + C
1	1.25	2.13	1.26	3.39
2	1.48	2.9	0.27	3.17
3	1.77	1.89	1.55	3.44
4	0.51	2.27	−0.44	1.83
5	1.25	1.5	0.44	1.94
6	2	1.89	0.41	2.3
7	1.43	1.69	0.41	2.1
8	0.93	1.69	0.41	2.1
9	1.94	3.29	0.41	3.7
10	0.57	2.13	0	2.13

All measurements are in mm. The juveniles are designated by numbers (see Fig. 1J). The dimensions of the capsules and the attaching threads were estimated by rotating these structures into the plane of the screen (in the case of the thread, increments of the length were rotated and measured separately as necessary). The capsule length (A) was measured along the long axis, and the threads (B) were measured from the point where they emerge from the capsule to their attachment to the adult. The advantage (C) is the distance from the point at which the thread is attached to the spine measured along its length to the base of the spine (i.e., where the spine connects with the axial part of the tergite). Including the advantage (B + C) provides a measure of the maximum distance of the juvenile from the lateral edge of the “body” of the adult. In the case of juvenile 4, the thread is attached to the axis of the tergite, not a lateral spine, and the advantage was computed as negative. The correlation coefficient (Figs. S1 and S2) is higher when the advantage is included, suggesting that the size of the capsule and its distance from the lateral edge of the body may be related, perhaps implying some functional advantage in feeding. Note that in neither case does the trend line pass near the origin, suggesting that the threads may have achieved some length before the capsules began to enlarge.

Mode of Life.

The morphology of the adult Aquilonifer provides limited evidence of mode of life. The first head appendage shows a superficial similarity to that of the Cambrian arthropod Kiisortoqia soperi from Sirius Passet, Greenland, which is antenna-like but armed with paired spines along its adaxial margin interpreted as “possibly suitable for capturing prey” (26, p. 495). The spines on the equivalent appendage in Aquilonifer, however, are relatively short, are more widely spread, and do not face adaxially. Furthermore the appendage in Aquilonifer tapers to a slender extremity and does not appear suitable for a grasping function. This first appendage may, in contrast, have been sensory or functioned in sweeping sediment in search of food. The second appendage is chelate and presumably functioned in manipulating food.

Both right and left (in a less pronounced fashion) biramous limbs curve adaxially at their distal extremity (Fig. 1 B and D). This position may be a response to burial. However, the segmented nature and flexibility of the endopods suggest that they could have functioned as walking limbs. Neither their morphology, nor that of the exopods, appear to be primarily adapted for swimming, indicating that Aquilonifer was benthic. The basipods were weakly spinose, but there is no evidence that they met in the midline. Food was presumably transferred directly to the mouth rather than transported anteriorly by the trunk limbs. The long cerci that project from the telson were presumably sensory.

The juveniles would have operated at low Reynolds numbers and likely used movement of the appendages to elevate them during feeding (27) rather than relying on forward locomotion of the adult to generate lift. The vast majority of crustacean larvae, for example, filter phyto- or zooplankton from the surrounding water (28). It is less likely that the juveniles attached to Aquilonifer were feeding on the sediment surface, as there would be no obvious advantage in a longer thread once the substrate was reached. Although the threads are preserved curving ventrally, none of them reaches below the appendages; their arrangement may be partly a result of the parent being overwhelmed by sediment (capsules 7 and 8 are in a position where they might impede movement of the trunk appendages). There is no evidence of the position of the oviduct in Aquilonifer or how the arthropod transferred or attached the offspring to its dorsal side. The long antenna may have been involved, or one parent may have attached the eggs to the other.

Brooding in Early Arthropods.

Evidence of parental care is rarely preserved in fossil taxa and is largely restricted, as here, to brood care of eggs and early juveniles (1, 24, 29). All examples reported to date in early Paleozoic arthropods involve protection within a bivalved carapace, a strategy that evolved independently in bradoriids (30), Waptia (29), and myodocope ostracods (1, 24). Extended parental care (31) has yet to be clearly demonstrated in invertebrate fossils.

Analogs for brood care in aquatic arthropods today are found in crustaceans and pycnogonids. Several strategies exist: enclosure by the thoracopods, by attaching eggs to the pleopods or ovigers (in the case of pycnogonids), within a dorsal brood pouch, within a marsupium formed by oostegites, and protection using an elongated first pleopod (32). The distribution of these methods among crustaceans suggests that most or all of them may have evolved independently (32). Brooding in pycnogonids is different in that the male rather than the female carries the eggs. Aquilonifer adopted yet another strategy that includes a dorsal position and attachment by a thread to a tergite. Among living crustaceans, a dorsal position for the embryos is confined to Thermosbaenacea, blind shrimp-like forms that live in caves and other underground systems (33). Their dorsal brood pouch is formed from an extension of the carapace in the female, and the embryos are transported there by currents generated by the thoracopods or transferred within a membrane that subsequently dissolves. The embryos of Thermosbaenacea are free within the dorsal brood pouch (33). The embryos of freshwater crayfish (Astacida) are tethered to the adult by a stalk (21, 22). The egg cases are attached to the pleopods by a stalk secreted by cement glands on the sternum and pleopods. When the hatchling emerges, it remains tethered to the egg case by a telson thread composed of the inner lining of the egg capsule. This maintains the attachment to the parent until the hatchling can use the hooks on the first pereiopod to grip the adult. In some crayfish, an anal thread performs the same function as the telson thread. Thus, among the diversity of brooding strategies in living aquatic arthropods are devices analogous, but very different, to that in A. spinosus.

Our interpretation of this remarkable specimen as representing 10 juveniles tethered to the parent A. spinosus, combined with its phylogenetic position among early arthropods, indicates that a complexity of brooding strategies evolved early in the history of the group.

Methods

The holotype of A. spinosus (OUMNH C. 29695) was ground at 30-µm intervals, in two separate pieces. Surfaces were imaged digitally and image stacks used to generate a 3D “virtual fossil” using the custom SPIERS software suite (www.spiers-software.org) (34, 35). The virtual fossil (VAXML) was studied on-screen using the manipulation, virtual dissection, and stereoscopic-viewing capabilities of SPIERS. Images in Fig. 1 were rendered as ray-traced virtual photographs using the open-source Blender package (https://www.blender.org). The data are housed at OUMNH.

The holotype of A. spinosus (OUMNH C.29695) was studied as an interactive virtual model, in VAXML format. VAXML models (36) consist of a series of STL- or PLY-format files describing morphology, together with an XML-based file providing metadata. They can be imported into any 3D graphics package that supports STL/PLY files or more conveniently can be viewed directly using the SPIERSview component of the freely available SPIERS software suite.

To understand the affinities of Aquilonifer, it was coded into the extensive phylogenetic dataset of Legg et al. (10), including subsequent modifications by Siveter et al. (37), and a single additional character from Legg (38): the possession of an extensive posterior transverse ridge on the trunk tergites, which was coded as present in some cheloniellids (see the supplemental section in ref. 38 for discussion). This new dataset of 315 taxa and 754 characters (Dataset S1) was analyzed under general parsimony in TNT (tree analysis using new technology) v.1.1 (39). All characters were unordered and weighted using implied weighting with a concavity constant of three. Tree searches used 100 Random Addition Sequences with Parsimony Ratchet (40), Sectorial Searches, Tree Drifting, and Tree Fusing (41). Nodal support was measured using Symmetric Resampling (each search used New Technology Searches with a change probability of 33%) and is reported as GC values.