Synchronized moulting behaviour in trilobites from the Cambrian Series 2 of South China

The study of moulting behaviour in the fossil record is relatively well known in arthropods and this is especially true for trilobites. Nevertheless, while studies focusing on the style of moulting in social and semi-social groups of modern animals (e.g. arthropods) are common, very few works investigate moulting adaptations in deep time. Here we report a trilobite assemblage from the Cambrian Series 2 “Tsinghsutung” Formation of South China. Around 850 specimens were used for this study from three different levels across one section near Balang (SE Guizhou Province, South China). These levels preserve numerous trilobite clusters in some cases containing around 400 individual specimens. Up to four species have been found in these clusters, but two species are more common. Trilobite clusters bear a high percentage of disarticulated specimens that we interpret as moults. Additionally, measurements of bioclast orientation and the dorsoventral attitude suggests very quiet water conditions followed by rapid burial events, prior to scavenger disturbance. Together, this indicates that the fossil assemblages were a result of a biological phenomenon rather than mechanical processes, allowing us to interpret the position of the fossil parts as different moulting configurations. Since the trilobite assemblage seems to be in situ, the large number of exuviae suggests a local place of migration. This was triggered by the need for group protection while moulting, which is suggestive of gregarious behaviour, possibly synchronized. These trilobites from the Cambrian Epoch 2, Age 4 constitute one of the earliest known gregarious community of trilobites and has important implications for understanding the ecology of this group during their emergence in the Cambrian.

www.nature.com/scientificreports/ such as predators, age, reproductive status or feeding (see 8,17,22 ). In this study we report and discuss the early evidence for gregarious moulting of two oryctocephalid species from the 'Tsinghsutung' Formation, Cambrian Series 2 (Stage 4) of Balang, South China.

Results
trilobite preservation. Specimens in this study come from three levels in the 'Tsinghsutung' Formation, Q51, Q52 and Q53, and are preserved in the majority of cases as complete individuals (i.e. articulated), although some sclerites are absent in a few cases (see moulting configurations and discussion below) (Fig. 1, see Fig. 1S supplementary material). Trilobites from these levels are typically preserved through replacement by the mineral illite. The illite degrades into iron oxides, and after dissolution has the ability to produce good quality internal and external moulds 23,24 . Bioclast abundance, composition and articulation. Bioclasts in the slabs are densely packed aggregates of semi-articulated specimens, sometimes concentrated in a specific region of the slab and are occasionally overlapping (Fig. 1). Clusters consist of 4 to 407 individuals and, although isolated specimens were also found, they were unusual in these three beds. A total of 857 individuals were included in this analysis (114 in Q51, 488 in Q52 and 255 in Q53). The clusters are mostly formed by complete or nearly complete individuals with a very high percentage of articulation, the majority being complete specimens with absent ventral cephalic structures (Fig. 2). The absence of ventral cephalic structures indicates that the majority of these specimens are moults. The abundance of bioclasts also varies within slabs with the density of individuals fluctuating across the slabs (see Fig. 1), but those were never compacted enough to suggest a hash-surface sensu Webster et al. 23 . Here we www.nature.com/scientificreports/ divided the bioclast composition into five different categories (Fig. 2) predominantly based on the definitions outlined by Webster et al. 23 : complete cephalon-trunk (CCT), disarticulated cephalon-trunk (DCT), isolated part of cephalon (IPC), isolate trunk fragment (ITF) and isolated sclerites (IS). Even though isolated sclerites of trilobites are common in the fossil record, especially in Lagerstätten 25 , this study shows the opposite trend, as the most common kind of bioclasts are highly articulated specimens (Fig. 2 Bioclast size distribution and ontogeny. In order to obtain the average size of each individual within the sample we measured three linear measurements: (i) maximum length; (ii) maximum width and (iii) maximum cephalic width. Although, Protoryctocephalus arcticus and Duyunaspis duyunensis were nearly equally abundant (359 and 336 individuals respectively), the size distribution of these species shows clear differences. D. duyunensis had a normal distribution while P. arcticus shows a bimodal distribution, most noticeably in specimens between 0.5 and 1.5 mm in width and specimens between 2.5 and 3.5 mm in width (Fig. 4). This bimodal distribution in P. arcticus may be explained by differences between the sample sites within the 'Tsinghsutung' Formation, however this seems to be well supported throughout Q52 and Q53, regardless of the type of measurement or the quantity of D. duyunensis (Fig. 5).
The ontogenetic series of these trilobites have not been studied yet but some of their developmental characters are noted. Both species show heminanamorphic development 29 in which an anamorphic phase of development (i.e. sequential appearance of additional segments) is followed by an epimorphic phase (i.e. sequential moults retain a constant number of segments). This is used to separate the meraspid and holaspid stages. Axial length in the meraspid stage ranges from 0.86 to 5.68 mm in P. arcticus and from 1.72 to 5.18 mm in D. duyunensis; and maximum width from 0.59 to 3.30 mm in P. arcticus and from 0.98 to 3.09 mm in D. duyunensis. Dorso-ventral attitude and orientation. Dorso-ventral attitude and orientation are important features that can be employed to distinguish between trilobites that were accumulated biologically versus mechanically. The dorsal up attitude was dominant in two of the three beds (Q51 and Q53) and half of the samples in the case www.nature.com/scientificreports/ of Q52 are also dorsal up (Fig. 2). Given the size difference of the articulated specimens, samples displaying a width between 0.59 and 5.04 mm were divided into six different categories. No preferred orientation in any of the categories has been found in any of the three beds ( Fig. 6). Although the sample from the Q52 bed shows a slightly denser orientation of ENE-WSW, the small K value (von Mises distribution) indicates a uniform distribution of the sample directions (0.01563, 0.03699, 0.08031).

Discussion and conclusions
Justification of biological clustering behaviour. Trilobite clusters from the 'Tsinghsutung' Formation contain up to 857 specimens and occur in densely packed aggregates. Clusters are multitaxa but species are clearly segregated. The size distribution shows a trend towards very small sizes in all taxa (size segregated); consequently, the individuals in the clusters mostly represent early and late meraspid and early holaspid stages www.nature.com/scientificreports/ in the case of the oryctocephalids, and early holaspid stages in B. balangensis, although the latter is a small trilobite 28 . However, the abundance of each taxon shows a clear bias. The percentage of the two oryctocephalids is never equal, showing antagonist behaviour, and B. balangensis constitutes only a negligible percentage in all samples. Hydrodynamic sorting or current-aligned specimens would indicate at least local transportation 30 .
The narrow size range of specimens preserved in all our samples from the three beds suggests that size-sorting, inferred from current transportation, is a possibility 31 . However, the absence of a preferred orientation within any sample suggests either the current velocity was not strong enough to re-orientate bioclasts, or there was a lack of consistency in current direction. Overall, the considerable size-sorting, the absence of current-induced sedimentary structures within each bed, and the presence of non-disrupted exuviae is suggestive of a low current velocity or even an absence of currents. On the other hand, the preservation of most of the trilobite bioclasts in a dorsal-up attitude may indicate that they were either (i) subject to unidirectional currents and adopted a hydrodynamically stable disposition; or (ii) were rapidly buried, since bioturbation can flip the exuviae. The dorsal-up attitude was the life attitude, but also resulted from normal exuviation in these trilobites. A number of studies on modern arthropods have demonstrated that disarticulation occurs within hours to weeks after death www.nature.com/scientificreports/ or ecdysis 25,30,[32][33][34][35][36] . Therefore, given all this evidence, mechanical aggregation can be dismissed since there was no strong preferential currents and the bioturbation index is very low. Thus, it seems reasonable to state that our sample indicates behavioural congregation and that the specimens were buried rapidly after the ecdysis without any disturbance of the exuviae. Trilobite clusters are commonly documented from the fossil record (e.g. 16,24,[37][38][39][40] ). Speyer and Brett 37 suggested three possible scenarios to explain such behavioural congregation: (i) stress-related aggregation behaviour (e.g. storm disturbances); (ii) non-reproductive gregarious behaviour (e.g. moulting); and, (iii) reproductive clustering behaviour. The size sorting of the assemblage (i.e. meraspid stages c. 89%, n = 401) does not fit the profile  www.nature.com/scientificreports/ of stress-related aggregation behaviour. If trilobites of all sizes were looking for shelter, as a result of stress, we should find a wide size range in the assemblage. Additionally, the taxonomic composition is very low, which is not indicative of a stress-related aggregation behaviour. Clusters or assemblages with high taxonomic diversification are related to faster and instantaneous burial such as obruption events 41,42 . On the other hand, most of the specimens represent early ontogenetic stages (mainly meraspid). Therefore, it seems very unlikely that these clusters represent reproductive clustering behaviour. The small size of specimens and the high number of moults is therefore suggestive that these clusters represent protection during moulting, reducing the risk of predation by moulting in large numbers in a quiet environment away from strong currents. Ecdysis (i.e. moulting) is a critical stage for arthropods that determines its growth in the early stages and, in adults, allows reproduction as the soft body of the females allows for copulation and/or produces offspring 10,17,37 . Many extant groups of arthropods (e.g. crustacean, insects) undertake coordinated migrations 17,43 and also possess a synchronized moulting cycle (a cycle also observed in trilobites (sensu 12,16,17,44 ). The migration is triggered by the need for protection during this period of weakness (e.g. moulting), or alternatively increases the ability to find mates, as migrations bring many individuals together in one restricted geographic area. Trenchard et al. 45 suggested that other trilobite genera such as oryctocephalids may have exhibited such behaviour because the deep-water areas are close to the exaerobic zones where the number of predators is reduced 39 . Clusters of juveniles or larvae are also common among arthropods and have been interpreted as a nursery 12,16,17 . Maintaining this gregarious behaviour could also carry other advantages for early larvae and juveniles, such as maternal food consumption (trophic egg consumption and/or matriphagy) and cooperative prey capture as observed in modern spiders 11,46,47 . Although our sample bears a high percentage of larvae, it is difficult to support such a hypothesis at this time, so further evidence is necessary to address this. Given the morphological features known in trilobites, they were unlikely to be aggressive with conspecifics. Although spines are present in their anatomy, these were likely for protection from larger predators and not from their own species 7 . Thus, they were apparently able to establish a gregarious behaviour and not suffer injuries among them. On the contrary, if they were aggressive, the moult synchrony would have helped to maintain the social coexistence given the inactivity before and during ecdysis reported in actual arthropods 5,11,12 . Moulting behaviour. The order Corynexochida was previously known to have four moulting behaviours: (i) Facial sutures opened; (ii) Rostral plate removed; (iii) Cephalon removed, and (iv) Cranidium removed (see 2 for more details). The majority of moulting behaviours found here seem to be associated with a break between the cephalon and thorax along with a rupture of the facial suture produced by a downward dorsal flexure, which seems to be the trend within this group 2,48 . Nonetheless, we have found several moulting configurations which confirm that the rest of the moulting behaviours described by Daley and Drage 2 and recently by Drage 6 (but not figured) are also present in this trilobite order, suggesting that these behaviours were common across many trilobite orders. The available data shows that the most common moulting behaviour is the opening of the facial sutures, and the rest of the recognized behaviours are present in lower proportions. Our findings agree with the www.nature.com/scientificreports/ results reported by Daley and Drage 2 and Drage 6 . However, given the high number of other behaviours reported in trilobites it seems unlikely that a relationship between configuration and behaviour exists. The sedimentological setting of the 'Tsinghsutung' Formation suggests a depositional environment with very little current action, where the trilobites were rapidly buried but not in a turbulent manner. Consequently, the biological meaning of the moulting configurations is uncertain. Further fieldwork to collect specimens within a sedimentological framework is necessary to assess the implications of such moulting behaviours for trilobite evolution. This work shows one of the earliest synchronized moulting behaviours in the fossil record, and demonstrates that trilobites had complex social behaviours, providing a new insight into the ecological evolution of these early arthropods.

Material and methods
Geological setting and fossil assemblage. More than 10,000 specimens of Protoryctocephalus arcticus, Burlingia balangensis and Duyanaspis duyonenesis have been collected throughout the 'Tsinghsutung' Formation (123 to 250 m thick). The 'Tsinghsutung' Formation is within the transitional Jiangnan Slope Belt near the locality of Balang in the east of Guizhou Province, South China 49-51 . Our samples (857 specimens) were collected from three levels: 51, 52 and 53 m below the Kaili Formation (see Fig. 1S supplementary material and 52 ). The lithology at this locality is unlike that of the Tsinghsutung or Wuxun formations studied previously in the transitional Jiangnan Slope Belt (see 51,52 ). This is why we have used inverted commas throughout the manuscript to indicate that the facies is similar to the Tsinghsutung Formation but with different lithological features. This formation in Balang is about 208 m thick and is mainly composed of grey to dark grey thick-bedded to massive limestone, grey thick-bedded dolomitic limestone, calcareous dolomite and grey thin-to thick-bedded dolomite and argillaceous dolomite. The samples in this study come from the Protoryctocephalus arcticus Zone. This zone occupies the middle-upper part of the 'Tsinghsutung' Formation and has a highly diverse trilobite assemblage: Protoryctocephalus arcticus, Protoryctocephalus balangensis, Duyunaspis duyunensis, Redlichia noblis (= Redlichia guizhouensis), Burlingia balangensis, Panxinella angustilimbata, Dinesus panxinensis, Nangops danzhaicnsis, Balangcunaspis cransversu, Mufushania cf. nakingensis, Eosoptychoparia gaodongensis, Eosoptychopariasp. cf. yunnanensis, Olenoides constrictus, and Olenoides. cf. hubeiensis 53 .
All the samples were coated with ammonium chlorite prior to being photographed with a Nikon D300 camera. Length, width and angular measurements of the specimens were made using ImageJ 54 . Drawings were made using Adobe Illustrator CC2019.
taphonomy. Clusters can represent behavioural aggregation of individuals, but they can also be accumulated mechanically by means of currents. In this study we have followed several criteria (see below) for the identification a particular cluster type and, in turn, type of aggregation (behavioural or mechanical): (1) Type of preservation. The type of preservation tells us about the early diagenesis suffered by the trilobites, as dissolution or deep cementation can change morphological features 55 . (2) Percentage of articulation. Exoskeletons can be either fully articulated, which includes unbroken dorsal and ventral sutures, or partially articulated (as discussed above). Partially articulated individuals that do not represent obvious moult assemblages were also examined in terms of the proportion of various elements, i.e. cephala, axial shields, cephalothoraces and trunks. Those clusters with a high percentage of disarticulated sclerites likely represent a background time-averaging due to sediment starvation prior to burial. (3) Percentage of taxonomic composition. Clusters can be composed of either a single species (monotaxic) or two or more species (polytaxic). Those clusters with highly polytaxics and a high abundance of each taxon are suggestive of mechanical accumulation. (4) Size distribution. Individuals may be size-segregated (unimodal; but not necessarily implying a single cohort) or show a range of sizes (polymodal). A size segregation could be indicative of a stage-related biological behaviour or a size sorting due to environment conditions. (5) Orientation. Horizontal orientation is the facing direction of an individual within 360° on a horizontal bedding plane. Bioclast alignment in the azimuth orientation can provide an indication of current or wave transportation. This is especially accurate for the long axes of articulated exoskeletons 33 . Those unidirectional currents can re-orientate highly articulated specimens such that its sagittal axis is aligned with the current direction (anterior end pointing up-current), or such that its sagittal axis is perpendicular to the current direction, with the lateral margin facing into the current 56 . (6) Bioclast dorso-ventral attitude. Several experiments have demonstrated that complete, isolated cephala and cranidia adopt a hydrodynamically stable dorsal-up (convex-up) attitude when subjected to surface currents; in the absence of such currents, isolated cephala settle in an inverted (convex-down) attitude 25,56-59 . The dorso-ventral attitude of these sclerites can therefore offer biostratinomic insight [60][61][62] , with the caveats that any current-induced signal can be subsequently modified by surface scavenging (which results in predominantly convex-down sclerite attitudes; see 25,60 ) and extensive bioturbation (which may result in random sclerite attitudes, including perpendicular or oblique to the sediment surface; see 25 ). www.nature.com/scientificreports/ (9) Fossils assemblage, paleoecology. The presence of other taxa may also be useful biostratinomic tools, especially biostratinomically sensitive, sessile multielement (e.g., echinoderms, Wiwaxia) and bivalved taxa (e.g., brachiopds) 68 . (10) Moults Vs Carcasses. In order to distinguish moults and carcasses we follow several criteria. The most important is to observe whether the specimens show suture lines with evidence of having been opened in one or several parts of the exoskeleton (e.g. both librigena) 2,48,60,63 . Other criteria, such as the lack of predation or scavenging evidences and lack of internal carcass features (i.e. guts) are also important criteria to distinguish moults and carcasses 2,13,25,37,39,40 . Statistical analysis. The data (measures, angles and morphology) were obtained and analysed using the ImageJ software (see supplementary material, dataset 1) and the resulting plots were made using Python 2D plotting library Matplotlib 65 . All quantitative data (percentages) were obtained using Python software. The K value of the Von Mises distribution was obtained using the open access Past software that uses the Rayleigh's test for directional data according to Davis 66 . www.nature.com/scientificreports/