GPIT Geologisch-Paläontologisches Institut, Universität Tübingen (Institute of Geology and Paleontology, University of Tübingen), Germany

ISER Institutul de Speologie Emil Racoviță (Emil Racoviță Institute of Speleology), Bucharest, Romania

LPB (FGGUB) Laboratorul de Paleontologie (Facul­tatea de Geologie și Geofizică, Universitatea din București) (Laboratory of Paleontology, Faculty of Geology and Geophysics, University of Bucharest), Bucharest, Romania

UBB Laboratory of Paleotheriology and Quaternary Geology, Universitatea Babeș-Bolyai (Babeș-Bolyai University), Cluj-Napoca, Romania

The Haţeg Basin is located in western Romania, in the southwestern Carpathians (Fig. 1). It is an intramontane basin formed during the terminal Cretaceous phase of the Carpathian orogeny (e.g., Bojar et al. 1998; Willingshofer et al. 2001). The sediment infill from this time interval comprises continental deposits originating from the erosion of the surrounding mountain ranges (Poiana Ruscă and Retezat mountains) and accumulated beginning in the middle part of the late Campanian and continuing well into the Maastrichtian (Bojar et al. 2010; Albert et al. 2025). At this time, the Haţeg Basin (as part of the wider Transylvanian area) was a subtropical island at a paleolatitude of 20–30°N with a warm and seasonally variable climate (Van Itterbeeck et al. 2004; Panaiotu and Panaiotu 2010).

The rocks of this uppermost Cretaceous sediment infill crop out discontinuously in the southcentral, central (Sibişel Valley, respectively Râul Mare Valley) and northwestern parts of the basin (Csiki-Sava et al. 2016). Previously, the uppermost Cretaceous continental rocks from these four areas were assigned to the Densuş–Ciula Formation (northwest) and the Sînpetru Formation (central–southern areas), respectively (Grigorescu 1992). The relationships of these successions are, however, poorly understood, and they are currently tentatively assigned to four lithostratigraphic units: the Sînpetru Formation (south–central), the Densuş–Ciula Formation (northwest), the ‘Râul Mare River section’ (central), and the ‘Pui Beds’ (southeast) (Csiki-Sava et al. 2016) (Figs 1, 2). Specimens described herein are from the ‘Pui Beds’, the ‘Râul Mare River section’, and the Middle Member of the Densuş–Ciula Formation cropping out near Tuştea.

The ‘Pui Beds’ are exposed along the Bărbat River Valley near the village of Pui in the southeastern part of the Haţeg Basin (Csiki-Sava et al. 2016). These consist mainly of red pedogenic silty mudstones interbedded with gray-greenish conglomeratic sandstone channel fillings, although sometimes dark-gray silty mudstone horizons also occur (Van Itterbeeck et al. 2004; Therrien 2005; Csiki-Sava et al. 2016). They were deposited in a meandering river floodplain with a seasonal and semi-arid climate, and are usually pedogenically overprinted (Van Itterbeeck et al. 2004; Bojar et al. 2005; Therrien 2005; Csiki-Sava et al. 2016). The rocks are from the 'middle' part of the Maastrichtian, dated roughly to the early to late Maastrichtian boundary based on palynomorphs (Van Itterbeeck et al. 2005).

Vertebrate fossils from the ‘Pui Beds’ are most commonly encountered in isolation, dispersed especially throughout the red-colored floodplain fines (e.g., Van Itterbeeck et al. 2004; Csiki et al. 2010a; Mocho et al. 2024), although occasional examples of mono- or oligotaxic accumulations yielding more complete partial to incomplete individual skeletons (e.g., Van Itterbeeck et al. 2004; Venczel and Codrea 2016, 2019; Augustin et al. 2022; Díez Díaz et al. 2025, as well as the Kallokibotion turtle specimen LPB (FGGUB) R.2593 discussed herein) or of small-sized attritional multitaxic bonebeds (pers. observ.) are also reported from this lithotype. Isolated occurrences are most often represented by one single complete to highly fragmentary specimen or, sometimes, by a very small number of skeletal elements preserved in close proximity that can be either associated but disarticulated or else may even belong to different taxa. In these types of occurrences, the vertebrate remains are often bleached white to yellowish or reddish cream-colored, with a coarse surface texture suggesting at least partial removal of the periosteum. They are also incomplete to different degrees, broken and cracked, with grey-haloed red silty mudstone matrix lining the surface and finer cracks, and filling in both the exposed interior of the bones and the wider cracks themselves, suggesting that this type of damage to them is produced mainly pre-burial. Fossils recovered from small mono/oligotaxic or politaxic bonebeds hosted by the red floodplain fines differ mainly in that they are often more complete, better preserved with the periosteum still largely in place, and with examples of skeletal association, even in situ articulation, being known (e.g., Augustin et al. 2022). More rarely, most likely largely attritional macrovertebrate assemblages have also been reported from coarser, grey-greenish sandstone channel fills (e.g., Solomon and Codrea 2015) and from dark grey muddy siltstones, where these are associated with microvertebrate remains (Codrea and Solomon 2012); here, the elements mostly display darker, dark brown to blackish colors, better-preserved periosteum, and less widely open pre-burial cracks, although advanced degrees of pre-burial fragmentation and skeletal disarticulation are still noted.

The sample of specimens reported here to show signs of bioerosion were recovered exclusively from the red fine-grained floodplain facies, and most of them were found either isolated through surface collecting (e.g., LPB (FGGUB) R.2428 and R.2429), or else through small-scale digs around surface-collected specimens exposed previously by erosion and which yielded a small number of disparate specimens (e.g., LPB (FGGUB) R.2046 and R.2370); the sole exception to this is the quasi-complete Kallokibotion skeleton LPB (FGGUB) R.2593 preserving the largest part of the shell together with the skull and part of the limbs that represents a case of incomplete individual skeleton preservation. Although some of the specimens, discovered and excavated either by former students or by collaborators and then deposited in the LPB (FGGUB) collections, lack exact details of origin and field occurrence conditions (e.g., LPB (FGGUB) R.2428 and R.2593; see Table 1), they display the same set of overall taphonomic features characteristic of the red floodplain fine lithofacies (see above and Materials and methods) as well as a still-adhering matrix type as do other specimens for which recovery conditions were assessed by us first-hand during excavation (e.g., LPB (FGGUB) R.2046), suggesting preservation under largely similar paleoenvironmental settings.

The ‘Râul Mare River section’ crops out along the Râul Mare River, including near Nălaţ-Vad, in the central part of Haţeg Basin (Codrea et al. 2002; Smith et al. 2002; Csiki-Sava et al. 2016). Here, the uppermost Cretaceous continental beds are composed of mainly grayish siliciclastics, including dark gray calcic clay-rich deposits interbedded with coarser-grained and lighter grey sandstones (Van Itterbeeck et al. 2004; Săsăran et al. 2011; Csiki-Sava et al. 2016). The depositional setting was likely a meandering fluvial system, dotted by locally occurring freshwater ponds (Van Itterbeeck et al. 2004; Ciobanete et al. 2011; Săsăran et al. 2011; Csiki-Sava et al. 2016). Based on palynology and magnetostratigraphy, the section has been dated to the ‘middle’ Maastrichtian (Van Itterbeeck et al. 2005) or else to the early part of the late Maastrichtian (Ciobanete et al. 2011; Panaiotu et al. 2011).

As first noted previously by Smith et al. (2002) and Van Itterbeeck et al. (2004), and also supported by subsequent reports on additional vertebrate fossil discoveries from Nălaț-Vad locality by Godefroit et al. (2009), Csiki-Sava et al. (2016), Brusatte et al. (2017) and Mocho et al. (2023, 2024), the local fossiliferous succession is characterized by the presence of several different taphonomic modes of vertebrate remain occurrences – common instances of isolated remains (either just one single fairly complete to fragmentary specimen, or at most a small number of potentially associated remains belonging to the same individual) and dinosaur egg clutches, alongside with much rarer cases of largely monotaxic accumulations containing incomplete individual skeletons (e.g. Godefroit et al. 2009; Csiki et al. 2010b) or microvertebrate bonebeds associated with eggshell coquinae. In most cases, the fossils are recovered from dark grey fine-grained, micaceous sandy-silty floodplain deposits, although occasionally they can also be found in light grey poorly consolidated sandstone channel fills. In both cases, however, the color of the macrovertebrate remains is dark purple to reddish brown to dark brown, with generally shiny bone surfaces showing well-preserved periosteum. Extensive pre-burial cracking as that reported at Pui (see above), as well as advanced degrees of subaerial weathering or abrasion are usually absent, even if specimens are often incompletely preserved and disarticulated, although cases of in situ articulation are also reported occasionally, both in isolated occurrences (Csiki-Sava et al. 2016) or within lenticular bonebeds preserving incomplete skeletons (Csiki et al. 2010b). In several instances, the bones are partly covered by a dark red iron oxide-bearing concretionary coating, very rarely reported or completely absent from other uppermost Cretaceous vertebrate occurrence of Haţeg Basin.

The sole skeletal element showing bioerosional traces reported by us from this locality was also found isolated, and it displays the hallmark features of this taphonomic mode at Nălaţ-Vad, being deep purple in color, incomplete but otherwise well preserved, largely unabraded and unweathered; some of these taphonomic features are also shared by LPB (FGGUB) R.2167, an isolated partial femur displaying a potentially dermestid-related boring (Brusatte et al. 2017), that nonetheless appears to show a more advanced abrasion stage at its proximal end. Meanwhile, the second femur showing largely similar preservational characteristics and reported by Brusatte et al. (2017) as featuring dermestid borings (LPB (FGGUB) R.2592) belongs to an (otherwise skeletally poorly represented) incomplete rhabdodontid individual whose remains are covered by iron oxide-rich encrusting. Remarkably, no other known skeletal element of this individual appears to bear bioerosional traces, probably suggesting that dermestid infestation was localized and accidental instead of being widespread within this particular skeleton (although support for such a scenario is admittedly weakened by the small number of preserved bones for individual R.2592).

The Densuş–Ciula Formation, exposed in the northwestern part of Haţeg Basin, spans the middle Campanian to upper Maastrichtian interval and is composed predominantly of reddish and greenish siliciclastic sedimentary rocks with a variable volcaniclastic component in its lower part (Grigorescu 1992; Bojar et al. 2011; Csiki-Sava et al. 2016). The unit is subdivided into three members (Lower, Middle, and Upper) that differ, among other things, in the content of volcanoclastic material as well as the abundance of vertebrate fossils (Grigorescu 1992; Botfalvai et al. 2017). Sedimentation occurred primarily within alluvial fan and braided river systems (Therrien 2005; Albert et al. 2025). The Tuştea local section (belonging to the upper part of the Middle Member), where a large number of vertebrate fossils (including the bioerosionally marked elements from this study) and dinosaur nests were found, is made up mainly of alternating layers of pale red, pedogenetically modified mudstones with calcareous nodules and massive dark red silty mudstones, representing distal, well-drained floodplain environments, and is topped by a thick bed of fluvial channel-laid cross-stratified pebbly-cobbly coarse sandstones (Botfalvai et al. 2017). The local succession can be loosely constrained to the early part of the late Maastrichtian (Csiki-Sava et al. 2016).

The sedimentology and taphonomy of the Tuştea –Oltoane locality that yielded all the herein discussed vertebrate remains preserving bioerosional traces from the Densuş-Ciula Formation, in addition to a caudal vertebra bearing tooth puncture marks that was reported previously by Csiki et al. (2010a), had been thoroughly documented by Botfalvai et al. (2017). The locality consists of a multitaxic, attritional macrovertebrate bonebed that is over 3 m thick and covers a surface area of over 150 m². It was interpreted as having a polyphasic genetic history, and most remarkably contains two superposed levels, each with several dinosaur egg nests and associated hatchling remains, interpreted as successive nesting horizons (see also Grigorescu 2010, 2017; Grigorescu et al. 2010). Vertebrate remains are dispersed throughout the thick fossiliferous red mudstones, being present both in the calcrete level-bearing paler reddish horizons (interpreted as pedogenically modified floodplain sediments) and in the darker purple-red, blocky and micaceous massive silty mudstones (i.e. flood-related, aggradationally added floodplain sediment sheets); these appear to be absent, however, from the slightly coarser, sandy sheet-splay deposit interbedded within the red silty mudstones, while only a very few highly abraded indeterminate bone fragments had been identified in the overlying cross-bedded and poorly sorted coarse pebbly sandstones. A large part of the vertebrate remains recovered from the locality are disarticulated, isolated, often highly fragmented, weathered and abraded (sometimes to the extreme condition of unidentifiable bone ‘pebbles’), although these are often closely associated within the same level with well-preserved, virtually unabraded and unweathered, even occasionally associated/articulated skeletal elements. Altogether they point to a lengthy, attritional accumulation of vertebrate remains undergoing distinct taphonomic histories, from highly scattered and disarticulated remains exposed for long periods of time across the floodplain surface to those covered rather rapidly after the demise of the animals during peak flood stages (Botfalvai et al. 2017). The color of the bones varies from light cream to yellowish-reddish cream, and are frequently coated by a grey-green halo (in the dark purplish red massive mudstones) or covered by light greenish, white-colored concretionary masses (in the calcrete-associated paler red silty mudstones).

Remarkably, all Tuştea specimens from our bioerosional traces-marked sample (Table 1) were discovered in the aggradational flood-related sediment packages (dark, massive mudstones), whereas the previously reported tooth puncture marks-inflicted caudal vertebra came from a calcrete-dominated horizon. Furthermore, most specimens are isolated (but see below) and fragmentary, albeit neither heavily weathered nor strongly abraded. Nonetheless, it is worth emphasizing that despite their preservational similarities, these specimens originate from different levels and spatial positions within the fossiliferous bed, indicating that they are not related to each other in their specific taphonomic history or time of burial. One notable exception to this is represented by the two lithostrothian titanosaur chevrons (LPB (FGGUB) R.1881 and R.2025) which, although discovered at some years distance, came from the same horizon and were rather closely spaced, and thus were interpreted recently as associated elements of a partial titanosaur skeleton by Díez Díaz et al. (2025), an interpretation with which we concur and accept.

The Haţeg Basin yielded one of the most diverse vertebrate assemblages from the entire Upper Cretaceous of Europe including fishes, amphibians, several species of kogaionid multituberculate mammals, squamates, kallokibotionine and dortokid turtles, five different crocodyliforms, azhdarchid pterosaurs, struthiosaurid ankylosaurs, rhabdodontid and hadrosauroid ornithischians titanosaurian sauropods, non-avian coelurosaurian theropods, and birds (Nopcsa 1900, 1902, 1923, 1928, 1929; Huene 1932; Rădulescu and Samson 1986; Weishampel et al. 1993; Rădulescu and Samson 1996; Buffetaut et al. 2002; Weishampel et al. 2003; Martin et al. 2006; Csiki et al. 2010b, 2010c; Martin et al. 2010; Vasile et al. 2013a; Csiki-Sava et al. 2015, 2016; Venczel and Codrea 2016; Venczel et al. 2016; Csiki-Sava et al. 2018; Vremir et al. 2018; Venczel and Codrea 2019; Augustin et al. 2021b, 2022). The vertebrate remains occur mainly as isolated elements, partly associated specimens, microvertebrate accumulations, or in small-sized multitaxic bonebeds (‘fossil pockets’) (Csiki et al. 2010a).

Twelve new examples of bioerosion were identified in the collections of the University of Bucharest (Fig. 3) on specimens belonging to dinosaurs and turtles (Table 1) from Upper Cretaceous (Maastrichtian) continental deposits of the Haţeg Basin.

In order to identify skeletal elements harboring bioerosional traces, we have surveyed and surface investigated three sets of vertebrate fossils represented by specimens discovered at Nălaţ-Vad, Pui and Tuştea, respectively. In our survey, we have focused specifically on larger-sized specimens that could be checked preliminarily for bioerosion by macroscopic means, and have excluded a priori the following categories of fossils: eggs and eggshells; isolated teeth; and microvertebrate remains (loosely meaning specimens that are mm-sized and were recovered mostly by bulk screen-washing of sediments, regardless whether they originate from small-sized animals or represent small-sized elements from larger-bodied taxa). By using this sampling protocol, as the first step of our investigation, we have preliminarily hand-checked 34 individual specimens from the LPB (FGGUB) collections coming from Nălaț-Vad, with a total yield of 3 specimens showing signs of bioerosion (one reported here and two noted previously by Brusatte et al. 2017 – 8.82%), 215 specimens from Pui, with 6 newly documented occurrences (2.79%), and 294 specimens from Tuștea, with a grand total of 6 bioerosion instances (5 newly reported herein and one documented previously by Csiki et al. 2010a; Botfalvai et al. 2017 – 2.04%).

We note, nevertheless, that our specimen samples are biased in their size and composition by differences in field occurrence details of the recovered specimens (isolated or associated), invested collecting efforts, methods of recovery used, and dimensions of the available outcrops (see also Geological setting for more details) among other factors, biases that may influence the preservational characteristics of the surveyed elements, as well as both the overall percentage representation of the bioeroded specimens within the local samples and that of the different bioerosional trace types we identified. Thus, whereas Tuştea locality corresponds to one single large-sized attritional bonebed that had been systematically quarried and collected for several years (e.g., Botfalvai et al. 2017), the Bucharest collection from Nălaţ-Vad was amassed mainly during occasional field surveys involving mainly surface collecting of newly exposed specimens from different spots across a rather large outcrop area (e.g., Smith et al. 2002; Csiki-Sava et al. 2016; Brusatte et al. 2017), and at most minor excavations around these specimens that rarely returned further remains from within the same site (Brusatte et al. 2017). Meanwhile, Pui represents an intermediate case where more intense and frequent field surveys and efforts of surface collecting in the rapidly eroding outcrops had been supplemented by occasional moderate excavations of small-sized lenticular bonebeds (‘bone pockets’; e.g., Csiki et al. 2010a) that yielded either partial individual skeletons (e.g., Augustin et al. 2022) or assemblages of isolated/associated remains of several individuals (e.g., Díez Díaz et al. 2025; Treiber et al. 2025). Finally, we acknowledge that sizeable collections of specimens from some of these localities (especially Pui and Nălaț-Vad; e.g., Smith et al. 2002; Van Itterbeeck et al. 2004; Godefroit et al. 2009; Csiki et al. 2010b; Mocho et al. 2023, 2024) are also available in other institutions (e.g., UBB), and that surveys similar to ours reported here but concerning those collections may nuance, or even alter, some of the conclusions reached by our investigation.

In the studied sample, two specimens (LPB [FGGUB] R.1883 and R.2593) are attributed to the stem-testudine Kallokibotion (Nopcsa 1923; Pérez-Garcia and Codrea 2018). The first of these, (LPB [FGGUB] R.2593), is a partial skeleton from Pui (currently under study), comprising a nearly complete skull, a well-preserved plastron and partial carapace, as well as associated limb elements (Augustin et al. 2025). Bioerosional traces are limited to the left, posterior, peripheral region of the carapace. LPB (FGGUB) R.1883 from Tuştea represents an isolated, partial peripheral with good surface preservation, and slight edge abrasion; the bioerosional traces occur on the ventral (i.e. inner) aspect.

The remaining 10 specimens represent a range of disarticulated dinosaur postcranial skeletal elements (Table 1). Preservation varies from well preserved and nearly complete elements (e.g., a hadrosauroid radius LPB (FGGUB) R.2370, and a titanosaurian metacarpal LPB (FGGUB) R.2597) to fragmentary and poorly preserved specimens (e.g., a large, indeterminate long bone fragments LPB [FGGUB] R.2046 and R.2428).

Most elements have a good preservation of the bone surface (e.g., LPB [FGGUB] R.2046, R.2330, R.2403), while in others it is only moderately well-preserved (e.g., LPB [FGGUB] R.1619, R.2428, R.2429). Several bones (e.g., LPB [FGGUB] R.1619, R.2046, R.2330) are represented only by the diaphysis. Bioerosion is often concentrated or limited to one side of the bone, although some specimens show bioerosion across all surfaces (e.g., LPB [FGGUB] R.1619, R.2046, R.2370, R.2597). Most specimens exhibit minor to moderate cracking, and the cracks are often filled with sediment; some cracks were subsequently restored using epoxy resin or plaster. Coloration ranges from reddish to greenish and differs considerably between localities (see also above, Geological setting). The scapula from Nălaţ-Vad (LPB [FGGUB] R.2403) shows a dark reddish color, also partially present on specimens from Tuştea that are often multi-colored (LPB [FGGUB] R.1619 and R.1881); meanwhile, the remaining bones from Tuştea (LPB [FGGUB] R.1883, R.2025 and R.2330) display a slightly greenish color. In the sample from Pui, the color of the bones ranges from slightly greenish (LPB [FGGUB] R.2046, R.2429 and R.2957), to light red/beige color (LPB [FGGUB] R.2428 and R.2593). Two of the surveyed elements show clearly visible preparation marks (LPB [FGGUB] R.2428 and R.2593, both from Pui), and thus bioerosional traces on them must be interpreted with caution, as certain preparation techniques can produce marks that resemble those caused by bioerosion (e.g., White and Toth 1989; James and Thompson 2015; Marin-Monfort et al. 2018; Wiest et al. 2018). However, in these specimens (as in other specimens in this study), the bioerosional traces are still partially covered by original sedimentary matrix, demonstrating that the traces must have been produced before burial.

Initial identification of bones showing bioerosional traces was made macroscopically in the collections of the University of Bucharest, and for this purpose, several hundred specimens were screened. The subsequent detailed analysis involved thorough macroscopic examination of the potentially bio-eroded specimens and the use of reflected light microscopy (Zeiss Stemi 2000) to characterize the morphology of the bioerosional traces found and to differentiate between trace morphotypes, in order to identify potential tracemakers.

Bioerosional traces were photographed using a Canon EOS R6. Traces were measured using ImageJ (version 1.54p) calibrated with scale bars in the photographs. The incisor tip width for the different Hațeg Basin multituberculate specimens available to us (see below) were measured directly on the fossils using a 0.1 mm-precision digital caliber; to estimate the same parameter for other previously described and illustrated specimens, their published images were similarly scaled in ImageJ using the provided original scale bars. However, only a limited number of such specimens are figured in occlusal, anterior, or posterior view as well, which restricted our ability to obtain accurate width measurements for these and allows only approximate estimates.

Computed tomography (CT) was employed to scan the carapace of Kallokibotion sp. (LPB [FGGUB] R.2593) using a Nikon XT H 320 μCT scanner, aimed to image possible internal traces. The specimen was μCT-scanned at the 3D Imaging Lab of the University of Tübingen, Germany with 4,476 projections (with four frames per projection) at a voltage of 215 kV and a current of 195 μA, using a 1.5 mm copper filter, generating data at a voxel size of 0.1082859 mm. The dataset was manually segmented with Amira v2021.1 (ThermoFisher Scientific).

For the description of the bioerosional trace fossils described below we employed the terminology proposed by Pirrone et al. (2014), supplemented in certain cases (when suitable categories were not formulated by Pirrone et al. 2014) with relevant termini published previously, in which cases the respective references are specifically mentioned.

Based on their shape and spatial arrangement, the traces observed in the studied sample can be classified into six macroscopic morphotypes: (I) star-shaped pits (e.g., Kaiser 2000; Pomi and Tonni 2011; Backwell et al. 2012; Augustin et al. 2019), (II) tunnels (e.g., Csiki 2006; Huchet et al. 2013; Brusatte et al. 2017), (III) irregular excavations, (IV) holes (sensuPirrone et al. 2014), (V) opposed paired grooves (e.g., Longrich and Ryan 2010; Valais et al. 2012; Augustin et al. 2019; grooves sensuPirrone et al. 2014), and (VI) irregularly gnawed surfaces (e.g., Longrich and Ryan 2010; Augustin et al. 2019, 2020b; Skutschas et al. 2025) (Fig. 4). All traces occur on cortical (outer) bone except for morphotypes (II), (III), and (IV) on specimens LPB (FGGUB) R.2429, R.2593 and R.2597, all of which extend into the cancellous bone.

Star-shaped pits

This type of trace occurs isolated on the bone surface of several specimens – LPB (FGGUB) R.1619, R.1881, R.2025, R.2046, and R.2403. The traces are restricted to the cortical bone layer and do not penetrate the underlying cancellous bone (Fig. 5). Each star-shaped pit comprises a shallow, bowl-shaped central depression with an uneven, sometimes slightly rugged surface, from which narrow groove- or furrow-like (sensuPirrone et al. 2014) extensions radiate outward, giving the pit a star-like outline. The overall shape of the trace is circular to slightly elliptical. Most star-shaped pits have a diameter of approximately 5 mm, although a single larger example measuring 10 mm is present on specimen LPB (FGGUB) R.2046. In the latter, the radiating grooves/furrows reach a width of approximately 0.8 mm, while in the smaller pits, the furrows are smaller, too, averaging around 0.2 mm in width. Currently, there is no formally erected ichnotaxon that encompasses this type of bioerosional trace.

Tunnels

This type of trace fossil was observed exclusively on the dorsal surface of the carapace of specimen LPB (FGGUB) R.2593 and on the titanosaur tibia LPB (FGGUB) R.2429 (Figs 6A–C, 7B). The boreholes not only affect the compact bone but extend into the underlying spongy tissue and even fully perforate the element. Their shapes range from nearly circular (LPB [FGGUB] R.2429, R.2593) to slightly elliptical (LPB [FGGUB] R.2429). The tunnel in LPB (FGGUB) R.2593 measures 12.7 x 12.1 mm, compared to the slightly larger dimensions found in LPB (FGGUB) R.2429 in which the medial side opening of the tunnel measures 15.76 x 13.3 mm, while the lateral side opening shows measurements of 15.17 x 11.75 mm (Fig. 6A, B). All tunnels are filled with matrix, which limits the definition of their margins and prevents detailed observation of their internal morphology (e.g., bioglyphs). This morphotype is assignable to the ichnotaxon Karethraichnus fiale, defined as “a cylindrical to bi-convex or flask-shaped profile, axis approximately perpendicular to the substrate surface” (Zonneveld et al. 2015, p. 809).

Irregular excavations

This morphotype category comprises only one trace that is present on the titanosaur metacarpal LPB (FGGUB) R.2597 from Pui (Fig. 6D). More specifically, it is located on the transition between the diaphysis and epiphysis, near the distal end. The trace extends into the spongy bone tissue, not being confined to the compact bone layer. It has an irregularly shaped, elongated, and slightly elliptical outline, as well as scalloped margins. Its length is 25 mm, with a width (perpendicular to the long-axis) of 3–5 mm and a depth (extending into the bone) of around 3 mm. The irregular excavation is still partially filled with sedimentary matrix and thus it is impossible to assess whether internal structures on the walls of the trace (i.e. bioglyphs) are present. Currently, there is no formally defined ichnotaxon that encompasses this bioerosional trace.

Holes

This morphotype is represented by bioerosion examples found on the titanosaur tibia LPB (FGGUB) R.2429, on the titanosaur metacarpal LPB (FGGUB) R.2597, and on the turtle carapace LPB (FGGUB) R.2593, all from Pui (Figs 7A, 7C, 8). The hole (sensuPirrone et al. 2014) on LPB (FGGUB) R.2597 is positioned at the transition between the diaphysis and the distal epiphysis. It is not restricted to the outer compact bone layer but extends into the spongy bone as well. It has a roughly oval outline (3.8 × 6.1 mm), rounded margins without an overhang, and a depth of around 3 mm. In addition, the hole seems to taper towards its bottom. On the turtle carapace LPB (FGGUB) R.2593, one hole with a diameter of 12 mm (11.8 × 12.1 mm) displays a similar shape and outline to that found on LPB (FGGUB) R.2597. Another hole in the carapace, measuring 8.2 × 6.5 mm in diameter, reveals a chamber-like structure penetrating into the bone, as shown by the µCT-scan (Fig. 7A). The hole on the titanosaur tibia LPB (FGGUB) R.2429 measures 13.15 × 11.27 mm. The interior of all these holes is still partially filled by sedimentary matrix, obscuring any potential bioglyphs. The morphotype can be referred to Karethraichnus lakkos. Originally, the ichnogenus Karethraichnus included three ichnospecies: K. lakkos and K. kulindros for non-penetrative holes, as well as K. fiale for penetrative holes (tunnels in our usage; see above). The former two (K. lakkos and K. kulindros) of these were, however, synonymized by Wisshak et al. (2019), and their diagnosis was emended to shallow (non-penetrating) Karethraichnus having a simple hemispherical to cylindrical profile with a rounded to flattened, hemispherical terminus (Zonneveld et al. 2015; Wisshak et al. 2019).

Opposed paired grooves

This morphotype comprises only two examples in the studied sample, identified on the fibula fragment LPB (FGGUB) R.2330 from Tuştea (Fig. 9). This type of trace consists of grooves (sensuPirrone et al. 2014) that are confined to the cortical bone, exhibit a uniform shape, and are always arranged as opposed pairs, with each pair consisting of two parallel grooves. These parallel-arranged grooves show identical depth profiles. The paired sets of two parallel grooves occur isolated and are positioned exclusively on edges or fractured margins (never occurring on smooth and/or flat bone surfaces). The individual grooves are teardrop-shaped, with the deepest point located at their broader, rounded ends that are facing towards one another between the pairs. The length of individual grooves is up to 3 mm, with widths ranging between 0.16 and 0.52 mm; the grooves within one pair display similar widths. The spacing between the paired parallel grooves varies from approximately 1.01 mm to 1.30 mm, whereas the spacing between the opposed pairs is roughly 0.54 mm. Currently, there is no formally defined ichnotaxon that would correspond to this type of bioerosional trace.

Gnawed surfaces

This morphotype was identified on several different specimens (LPB [FGGUB] R.1619, R.1883, R.2046, R.2330, R.2370, R.2428, and R.2597) (Fig. 10). The gnawed surfaces appear as multiple scratches, grooves, or notches (sensuPirrone et al. 2014), often forming clusters or accumulations. In these clusters, the individual bioerosional features sometimes overlap each other, making it impossible to count them; however, they seem to range from a minimum of 7 to more than 50 within one cluster. Affected areas can extend over larger surfaces, such as approximately 10 × 40 mm on specimen LPB (FGGUB) R.2046, where broad grooves measuring up to 1 mm in width and 3 mm in length are present. A similar pattern is visible on LPB (FGGUB) R.1619, with the groove width up to 0.7 mm and a length up to 3 mm, as well as on LPB (FGGUB) R.2330, with a groove width of 0.5 mm and length up to 2 mm. Smaller gnawed areas, measuring around 10 × 20 mm, contain numerous narrower furrows. These furrows are approximately 0.05 mm wide and up to 3 mm long in LPB (FGGUB) R.1619, and similar cases were observed on LPB (FGGUB) R.1883 (furrow width of 0.2 mm and length of 3 mm), as well as on LPB (FGGUB) R.2957 (approximate width of up to 0.1 mm and length of 2 mm). Additionally, certain specimens (e.g., LPB [FGGUB] R.2370, R.2428) display shallow clusters of scratches with a width up to 0.2 mm and a length up to 2 mm, covering surfaces of approximately 5 × 5 mm, and forming a shape of rounded to slightly oval outline. Accordingly, this morphotype is characterized by a high degree of variability, concerning both the number of clustered scratches/grooves/notches, as well as the size of the affected area, and that of the individual bioerosional features. This morphotype is assignable to Machichnus bohemicus, which is defined as “shallow, thin, discrete, parallel to subparallel, smooth-bottomed scratches. The scratches occur in small groups or series of corresponding width and length” (Mikuláš et al. 2006, p. 122).

The overall arrangement and morphology of the morphotypes described above strongly suggest a biogenic origin consistent with terrestrial bioerosion. Such traces have previously been attributed to a range of trace-maker organisms, including invertebrates, vertebrates, and plants (e.g., Andrews and Cook 1985; Abler 1992; Erickson and Olson 1996; Kaiser 2000; Bader 2003; Bader et al. 2009; Longrich and Ryan 2010; Backwell et al. 2012; Augustin et al. 2019). To identify potential tracemakers, the structures described herein were compared with previously reported examples of bone-modifying bioerosion in continental depositional settings. These tracemakers were furthermore considered in correlation with the documented latest Cretaceous faunal assemblage of the Haţeg Basin (e.g., Csiki-Sava et al. 2016), with particular attention to bone-modifying taxa capable of producing trace patterns similar to those described in the current sample. Although no one has reported any insect body fossils so far from the uppermost Cretaceous of the Haţeg Basin, a few examples of potential insect feeding traces have been described (e.g., Csiki and Grigorescu 2002; Csiki 2006; Vremir 2009; Brusatte et al. 2017; Augustin et al. 2019), as were purported insect eggs (Bodor et al. 2014; Hermanová et al. 2017) and coprolites (Vasile et al. 2013b). For a general overview of bone-modifying agents among invertebrates, vertebrates, and plants see Augustin et al. (2019), Backwell et al. (2022), and Fernández-Jalvo and Andrews (2016).

Termites as tracemakers of star-shaped pits

The star-shaped pits observed at Nălaţ-Vad, Pui, and Tuştea are similar to traces attributed to termites by Kaiser (2000) from the Plio-Pleistocene of Northern Tanzania; by Pomi and Tonni (2011) from the Late Pleistocene from Argentina; by Backwell et al. (2012) from the Middle and Upper Stone age from South Africa and Ethiopia; and by Augustin et al. (2019) from the Upper Cretaceous from Romania. Notably, and not unexpectedly, the star-shaped traces described here closely resemble examples reported recently by Augustin et al. (2019) and found in nearby and roughly coeval deposits (the Sînpetru Formation of the Haţeg Basin).

The interpretation of these structures as termite-related bioerosion is reinforced by comparisons with extant termite feeding traces. In particular, recent experimental bioerosion studies have documented that termites regularly produce, among other types of traces, star-shaped pits (e.g., Watson and Abbey 1986; Backwell et al. 2012), lending additional weight to the hypothesis that these fossil traces found all across the Haţeg Basin were created by termites. Indeed, this morphological continuity between modern and fossil traces has been widely recognized in the ichnological literature (Kaiser 2000; Backwell et al. 2012; Matu et al. 2017; Augustin et al. 2019). The reported presence of purported termite frass in the continental uppermost Cretaceous of the Haţeg Basin (Vasile et al. 2013b) may offer further support to the interpretation of the star-shaped pits as feeding traces of termites. Star-shaped pits are also known to be produced by dermestid beetles (Parkinson 2013, 2023; Backwell et al. 2022), but they are significantly smaller (0.55–1.38 mm) than those produced by termites (Backwell et al. 2022).

Dermestidae as tracemakers for holes, irregular excavations, and tunnels

Tunnels and holes in fossil bones have often been described in both archaeological and paleontological contexts, and they are attributed either to indeterminate insects (Gates 2005; Britt et al. 2009; Kirkland and Bader 2010; Parkinson 2016; Augustin et al. 2019, 2021a; Perea et al. 2020) or, more commonly, to dermestid beetles (Kitching 1980; Martin and West 1995; Hasiotis et al. 1999; Paik 2000; Csiki 2006; Roberts et al. 2007; Britt et al. 2008; Cabral et al. 2011; Huchet et al. 2013; Serrano-Brañas et al. 2018; McHugh et al. 2020; Venegas-Gómez et al. 2024). Boreholes in fossilized bones are usually associated with pupation behavior. Although Parkinson (2023) recently contested that dermestids actually produce such traces, actualistic experiments have shown that they build their pupation chambers in bone, thereby creating comparable traces (Zanetti et al. 2018; Charabidzé et al. 2022). In these experiments, the holes produced by dermestids to build pupation chambers have a circular to slightly elliptical opening (Zanetti et al. 2018: figs 1, 2), while the chambers themselves are spherical to ovoidal, measuring 4.5–19.0 mm in length and 3.9–13.8 mm in width (Zanetti et al. 2018). The size and shape of the traces found on the turtle carapace LPB (FGGUB) R.2593 and the titanosaur tibia LPB (FGGUB) R.2429 from Pui are consistent with previously described dermestid beetle borings in fossil bones, which are rounded to oval in cross-section and typically measure around 2–16 mm in diameter (e.g., Kitching 1980; Hasiotis et al. 1999; Csiki 2006; Britt et al. 2008; Huchet et al. 2013; Paik 2000; McHugh et al. 2020). In the Hațeg Basin, Csiki (2006) described cavities and a tunnel preserved in a titanosaur osteoderm from the Sînpetru Formation, with a diameter of 10 mm, which closely resemble the tunnels from Pui noted earlier, whereas Brusatte et al. (2017) briefly noted the presence of round tunnels with a diameter of around 12–16 mm in rhabdodontid femora from Nălaţ-Vad.

Holes and tunnels in turtle shells are regularly ascribed to the action of ectoparasites and in fact Karethraichnus (to which we ascribed our holes and tunnels as well) was erected for purported traces of ectoparasites on turtle shells from the Eocene (Zonneveld et al. 2015). However, the examples described herein are much larger than those described by Zonneveld et al. (2015), which have a maximum diameter of 9 mm, and occur not only on turtle shells, but morphologically identical traces also affect titanosaur bones. In addition, the holes and tunnels described herein do not form tight clusters and never show any signs of healing (or reaction of the bone) and thus are completely incompatible with an ectoparasitic origin. Nonetheless, they fulfill the diagnosis for Karethraichnus lakkos and K. fiale, respectively (for diagnoses, see above). Notably Alves-Silva et al. (2024) recently assigned holes on Quaternary mammal bones from Brazil to Karethraichnus and interpreted them as dermestid traces. Furthermore, the irregular excavation identified on the titanosaur metacarpal LPB (FGGUB) R.2597, also from Pui, appears to have formed through the coalescence of multiple (likely three) smaller holes, each showing the same morphology as the singular hole found on the same specimen, and therefore likely share the same origin with the hole itself.

In previous studies, pupation chambers of dermestid beetle larvae are cited as the most common interpretation of similar tunnels and holes (and thereby probably also of the irregular excavations described herein from Pui) (e.g., Kitching 1980; Hasiotis et al. 1999; Csiki 2006; Britt et al. 2008; Paik 2000; Huchet et al. 2013). This interpretation is supported by field observations (Kreyenberg 1928; Gabel 1955; Tobien 1965; Martin and West 1995) and controlled feeding/pupation experiments (Zanetti et al. 2018; Charabidzé et al. 2022), which documented that dermestids do indeed produce such holes and tunnels (often in hard durable substances such as bone) when building their pupation chambers. The close similarity in size, shape, and distribution between bioerosion morphotypes II, III, and IV described herein and the different previously reported examples strongly indicates that the respective traces were likewise produced by dermestid beetle larvae.

The range of morphological and size variation observed between the tunnels and holes described in our sample could be linked to either behavioral differences or to different properties of the substrate. It has, for example, been noted that dermestid beetle larvae of the same species can produce within-bone pupation chambers with greatly varying sizes (Zanetti et al. 2018). Moreover, it is possible that overlying desiccated soft tissues were originally part of the borehole but subsequently decayed and are not preserved in the fossil specimen, thus altering its original size and morphology (see Martin and West 199Huchet et al. 2013). Also, trace morphology depends on the initial angle of penetration, which in turn is presumably influenced by behavioral differences and substrate properties. Consequently, it may result in a rounded outline of the tunnel on one side of the bone, compared to an elliptical outline at the exit, as observed in the titanosaur tibia LPB (FGGUB) R.2429.

Multituberculate mammals as tracemakers of opposed paired grooves

The opposed paired grooves observed on the Tuştea specimen also closely resemble certain traces previously described in the fossil record. Their teardrop-shaped morphology and occurrence in opposed pairs are similar to examples attributed to multituberculate mammals, and reported by Augustin et al. (2020b: fig. 2) from the Upper Jurassic of China, by Longrich and Ryan (2010: fig. 1) from the Upper Cretaceous of Canada, by Valais et al. (2012: fig. 3) from the ‘mid’-Cretaceous of Argentina; and, most importantly, those by Augustin et al. (2019: fig. 7) from the Upper Cretaceous of Romania, who described a remarkably similar trace fossil from nearby and roughly coeval strata of the Sînpetru Formation from the central Haţeg Basin. However, the opposed paired grooves reported here differ from those described by Augustin et al. (2019) in both the spacing between adjacent parallel grooves, and the width of the individual grooves (see below), suggesting potential differences in dental morphology, body size, and/or relative incisor position of the tracemakers.

Associated bite marks from upper and/or lower jaws could plausibly be produced by one of the following vertebrate clades (all of which are known from the continental Upper Cretaceous from Transylvania or Europe): squamates, crocodyliforms, mammals, or small-to-medium sized theropod dinosaurs (for a taxonomic overview, see Csiki-Sava et al. 2015, 2016). Among these clades, bite marks have been described for squamates (e.g., D’Amore and Blumensehine 2009), crocodyliforms (e.g., Drumheller and Brochu 2014), theropod dinosaurs (Augustin et al. 2020a; Brown et al. 2021), as well as various different mammals, including multituberculates, which are known from the Upper Cretaceous of Transylvania (Longrich and Ryan 2010; Fernández-Jalvo and Andrews 2016; Augustin et al. 2019, 2020b). However, squamates, crocodyliforms, and theropod dinosaurs all have a homodont dentition with closely packed teeth and thus are incapable of producing opposed paired grooves, for which an occluding pair of upper and lower incisor-like teeth would be necessary. This leaves only mammals as potential producers of these traces, and so far, multituberculates are the only ones known from the Upper Cretaceous of Transylvania.

As we noted above, the single such trace described previously by Augustin et al. (2019) on specimen GPIT/RE/12857 from Sînpetru has a groove width of 0.26 ± 0.02 mm, a groove length of 1.20 ± 0.36 mm, a spacing between the paired grooves (measured between their midlines) of 0.81 ± 0.01 mm, and a distance between the pairs themselves, of 0.51 ± 0.05 mm. These measurements depart from those of the two otherwise morphologically similar bitemarks from Tuştea reported here on specimen LPB (FGGUB) R.2330 (Fig. 9), which themselves differ in size from each other (see Table 2). The larger of these latter bitemarks shows a groove width of 0.48 mm for the pair of longer grooves (for a length of 2.96 mm) and, respectively, 0.51 mm for the shorter ones (for 1.46 mm in length), a spacing between the grooves of 1.08 ± 0.06 mm, and a distance between the pairs of 0.67 ± 0.05 mm. Meanwhile, the groove widths in the smaller bitemark from Tuştea (displaying four equally long grooves of 1.32 ± 0.19 mm) is 0.30 mm and 0.21 mm, respectively, with the spacing between the paired grooves being 1.29 ± 0.02 mm, and the distance between the opposed pairs, 0.44 ± 0.07 mm. Therefore, the previously described bitemark from Sînpetru and the smaller one reported here from Tuştea share almost the same groove width, but differ in the spacing between the grooves, with those of the Sînpetru bitemark being more closely spaced than those seen in the Tuştea bitemark. Conversely, the larger bitemark from Tuştea shows grooves that are at least 41% wider than in the other two examples (Table 2).

Different dimensions and morphologies of the purported multituberculate bitemarks are not unexpected, because several different multituberculate taxa are known from the Upper Cretaceous of Transylvania. In total, five different nominal multituberculate taxa (Barbatodon oardaensis Codrea et al., 2014, B. transylvanicus Rădulescu & Samson, 1986, Kogaionon radulescui Smith et al., 2022, K. ungureanui Rădulescu & Samson, 1996, and Litovoi tholocephalos Csiki-Sava et al., 2018) have been reported from the Upper Cretaceous deposits of the Haţeg Basin (see also Solomon et al. 2016; Codrea et al. 2017), besides several other occurrences of indeterminate kogaionids (e.g., Codrea et al. 2002; Csiki and Grigorescu 2002; Smith et al. 2022), and some of these differ considerably in body size (Csiki-Sava et al. 2022). By comparisons with the incisors known at least partially from all five nominal taxa, the tracemaker of these paired parallel grooves can potentially be identified.

All of these taxa belong to the endemic multituberculate family Kogaionidae, and, as documented in several examples of well-preserved cranial specimens bearing in situ dentition (e.g., Rădulescu and Samson 1996; Csiki et al. 2005; Smith and Codrea 2015; Csiki-Sava et al. 2018; Smith et al. 2022), most likely all of them had one pair of lower incisors (i1) and two pairs of upper incisors; as for the latter, the first pair (I2) is much larger than the other pair (I3) and opposes the lower incisors, as commonly seen in derived multituberculates (e.g., Kielan-Jaworowska et al. 2004; Csiki-Sava et al. 2022). Thus, the bitemarks identified here were produced most likely by the I2 (of the upper jaw) occluding against the i1 (of the lower jaw). Four of these five taxa even preserve (at least some of) the incisors in situ, but relevant measurements (esp. of the incisor width and their spacing) are most often not provided or are unavailable due to the incomplete nature of their fossils, and only a very few of them were previously figured in occlusal/anterior/posterior view as well, allowing the approximate width measurements to be taken from published photographs. In the following section, some of the reported incisor widths were measured directly on specimens available to us, while other values were derived from measuring the tip of the incisors in the figures provided in published descriptions (using the corresponding scale). Accordingly, these latter measurement values should be treated with some caution, but they nevertheless allow a rough assessment of the appropriate dental dimensions of interest here.

Overall, Csiki-Sava et al. (2022) classifies the Transylvanian kogaionid species in three different estimated body size classes: large sized (Barbatodon transylvanicus, Kogaionon ungureanui, Litovoi tholocephalos), medium sized (Kogaionon radulescui), and small sized (Barbatodon oardaensis). Body size estimates based on dentition provide an overview of their relative dimensions: Barbatodon oardaensis is about 35%–50% smaller than Barbatodon transylvanicus (Codrea et al. 2014; Smith and Codrea 2015; Solomon et al. 2022) and 25% smaller than Kogaionon radulescui (e.g., Smith et al. 2022), whereas Kogaionon ungureanui is approximately 15% larger than Barbatodon transylvanicus (e.g., Smith and Codrea 2015) and 35% larger than Kogaionon radulescui (e.g., Smith et al. 2022); the latter taxon is also roughly 20% smaller than Barbatodon transylvanicus and Litovoi tholocephalos (e.g., Smith et al. 2022).

The measurements from the large-sized taxon Barbatodon transylvanicus from Pui show an I2-tip width of 0.97 mm (UBB P-Mt1, figured in Smith et al. (2022): fig. 9) and, respectively, an i1-tip width of 0.71 ± 0.06 mm (for specimen UBB P-Mt2-1, figured in Solomon et al. 2016) and 1.16 mm (for specimen LPB [FGGUB] M.1635; Csiki et al. 2005). The even larger Kogaionon ungureanui (ISER SPT/001 from Sânpetru) shows measurements of 1.26 mm for the I2-tip width and an estimated inter-incisor gap of 2.65 mm. The middle-sized taxon Kogaionon radulescui (UBB NV-Mt1 from Nălaţ-Vad, figured in Smith et al. 2022) shows an i1-tip width of 0.543 ± 0.13 mm. Other mid-sized I2s from indeterminate kogaionids (e.g., ISER PUI.003 and LPB (FGGUB) M.1706, both from Pui and figured in Csiki-Sava et al. (2022), as well as an isolated I2 mentioned in Grigorescu (1984) from Sânpetru), show I2-tip widths of 0.63 mm, 0.73 ± 0.08 mm, and 0.44 ± 0.07 mm, respectively.

The measurements of the mid-sized Kogaionon radulescui fit size-wise the larger bitemark from Tuştea, with groove widths of 0.51 and 0.48 mm (Fig. 11), respectively. Codrea et al. (2017) had referred multituberculate specimen reported previously from this locality by Csiki and Grigorescu (2002) to the small-sized Barbatodon oardaensis based on their assessed dimensions. However, one of the more diagnostic Tuștea specimens, an as yet unpublished blade-like lower premolar (p4) is marginally larger even than that of K. radulescui, let alone that of B. oardaensis (Csiki-Sava et al. 2022), indicating that a most probably medium-sized, as yet undetermined kogaionid was instead represented at this locality. Accordingly, even though K. radulescui is as-yet unreported at Tuştea and thus probably can be excluded as the specific tracemaker, an animal of roughly comparable body size represented by this currently taxonomically indeterminate kogaionid may have produced the larger bitemarks.

Smaller-sized incisors were available for measurements mainly for different indeterminate Kogaionidae, with the isolated I2 LPB (FGGUB) M.1705 from Pui, figured in Vasile et al. (2019) having a tip width of 0.22 ± 0.04 mm, and the isolated i1 ISER PUI.004, also from Pui, and figured in Csiki-Sava et al. (2022), showing a tip width of 0.46 mm. These values fit quite well to the groove widths recorded in the Sânpetru (0.26 mm and 0.18 mm) and the smaller Tuştea (0.30 mm and 0.21 mm) bitemarks. The smaller Tuştea trace was then likely produced by an individual representing a small-sized kogaionid, potentially maybe either a small-sized Kogaionon sp. (known from Pui; Csiki-Sava et al. 2022) or Barbatodon oardaensis, although none of these taxa had been reliably reported as yet from this particular locality (see above); a similar origin can be assessed for the bitemark from Sînpetru as well, where the presence of small-sized indeterminate kogaionids was indeed noted by Csiki-Sava et al. (2022).

Furthermore, assignment of the proposed multituberculate trace-maker(s) to small or medium-sized taxa may also be supported by the noted differences in the inter-groove width. The Sînpetru bitemark suggests an inter-incisor gap that is 70% smaller than that estimated for the large-sized Kogaionon ungureanui, while the Tuştea bitemarks are also smaller by 60% (large bitemark) and 51% (smaller bitemark), respectively.

Notably, however, differences recorded in trace size and morphology between and within sites (i.e., the larger and smaller bitemarks from Tuştea, respectively the smaller bitemark in Sânpetru) may also reflect different ontogenetic stages of the mammalian tracemakers, the intensity of the impact itself and/or the stage of wear displayed by the incisors, instead of taxonomic differences. Unfortunately, multituberculate mammals are extinct, making it impossible to draw comparisons with bite marks of extant members of the group derived from field observations or feeding experiments.

Gnawed surface caused by termites and multituberculate mammals

The gnawed surface-type bioerosions reported here also resemble previously described examples from the fossil record. Indeed, the gnawed areas with broad grooves (see above) show strong similarities with those described by Augustin et al. (2020b) from the Upper Jurassic of China, by Longrich and Ryan (2010) from the Upper Cretaceous of Canada, and by Skutschas et al. (2025) from the Upper Cretaceous of Russia. Meanwhile, the narrower grooves/furrows are reminiscent of those described by Augustin et al. (2019) from a roughly coeval and nearby Upper Cretaceous site of the Haţeg Basin in Romania, at Sânpetru, but also of those reported by Backwell et al. (2012) from the Middle and Later Stone Age from South Africa, and by Pomi and Tonni (2011) from the Late Pleistocene from Argentina. These gnawed surfaces have been variably attributed to termites (Pomi and Tonni 2011; Backwell et al. 2012; Augustin et al. 2019) or to multituberculate mammals (Longrich and Ryan 2010; Augustin et al. 2020b; Skutschas et al. 2025), respectively, indicating that differentiating between traces left by termites and mammals is not always straightforward. However, a distinction between the potential trace-makers is likely possible based on the size (esp. the width) and the arrangement (esp. pairing) of the grooves. The gnawed surfaces displaying a broader groove width, for instance, are remarkably similar to those produced by extant rodents (Klippel and Synstelien 2007; Pokines et al. 2017; Marginedas et al. 2023), which are often regarded the modern ecological analogues for the extinct multituberculates due to their overall size and similarity in dentition (e.g., van Valen and Sloan 1966; Krause 1982; Kemp 2004), thus supporting our suggestion that these particular bioerosional traces were indeed produced by multituberculate mammals.

It is worth noting that gnawing behavior in extant rodents emerges during the juvenile phase, typically when individuals have reached approximately 30–50% of their adult body size (Gall and Kyle 1968; Thiels et al. 1990; Katahira and Ohwada 1994; Ginot et al. 2020) indicating that small-bodied juveniles are already capable of producing recognizable gnaw marks relatively early in ontogeny. When this is applied to the smallest measured Transylvanian multituberculate incisor width of 0.22 mm, then any grooves or furrows narrower than 0.06 mm can most likely exclude mammals as tracemakers. Nonetheless, it should be noted that such an extrapolation from rodents to multituberculates remains far from certain, as the two clades are only distantly related. In addition, the occurrence of paired, parallel grooves of roughly equal width may serve as another potential indicator of a multituberculate origin, due to the paired arrangement of their incisors (Kielan-Jaworowska and Hurum 2001; Csiki-Sava et al. 2022), especially if noted in combination with relatively wide grooves.

Accordingly, the gnawed area observed on the long bone fragment LPB (FGGUB) R.2428 from Pui, which exhibits paired, parallel grooves with an average and rather constant width of 0.19 ± 0.05 mm, may plausibly have been produced by a small multituberculate mammal, just as the gnawed surface found on LPB (FGGUB) 2046 from Pui that bears grooves with a width of 0.58 ± 0.09 mm may also originate from the activity of a (possibly larger) multituberculate. In this regard, it is worth noting that Csiki-Sava et al. (2022) documented the presence of both small- and large-sized multituberculates at Pui, throughout the local uppermost Cretaceous continental succession. In comparison, LPB (FGGUB) 1619 from Tuştea shows grooves/furrows with a width of only 0.05 ± 0.008 mm and thus termites could have more likely produced this particular gnawed surfaceinstead.

In general, the precise identification of gnawed surfaces remains difficult, and more data from feeding experiments and/or observations of feeding behavior of modern representatives (or ecological correspondents) of these ancient organisms is certainly needed to identify these traces with more confidence.

The observed bioerosional traces demonstrate a remarkable diversity, suggesting the activity of at least three distinct groups of tracemakers. Firstly, the evidence supports the activity of termites, the presence of which was proposed before in the Upper Cretaceous of the Haţeg Basin based on feeding traces and coprolites (Vremir 2009; Vasile et al. 2013b; Augustin et al. 2019). Then, the traces described here also documents the presence of dermestid beetles in the latest Cretaceous of Haţeg Island, as already indicated previously by the occurrence of specific tunnels and holes in fossil bones (Csiki 2006; Brusatte et al. 2017). Additionally, and most importantly, we found more examples of multituberculate bite marks, which were first documented in the general study area by Augustin et al. (2019).

Each of these types of bioerosional modification carries its own taphonomic and paleoecological significance, reflecting discrete processes that acted on skeletal remains after the death of a given animal and prior to their final burial. Differences in the distribution and characteristics of bioerosional traces among localities, that themselves differ with respect to the lithologies represented and thus in their reconstructed depositional setting, point to somewhat divergent taphonomic pathways operating within the various assemblages and paleoenvironment. When considered together, bioerosional trace fossils and their distribution (Fig. 12) can help to reconstruct a complex sequence of processes, including subaerial exposure, scavenging, and burial of the bones.

When comparing the different sites and the distribution of the types of bioerosion described, it becomes evident that bioerosion produced by termites has been reported from all localities that yielded bioerosional trace fossils thus far, which notably represents all four uppermost Cretaceous lithostratigraphic units of the Haţeg Basin (see Fig. 12, Table 1). In the present study, termite and dermestid bioerosional traces were for the first time identified on bones from Pui. Possible traces produced by multituberculate mammals may also be present, as suggested by gnawed surfaces that show a large groove width; however, this identifications remain inconclusive for the time being. At Nălaţ-Vad, only termite bioerosion was identified in this study, likewise for the first time. Previously, Brusatte et al. (2017) described tunnels in rhabdodontid bones from Nălaţ-Vad that they attributed to dermestids. So far, no multituberculate-related bioerosion has been reported from this site. At Tuştea, the present study revealed for the first time bioerosion produced by termites and multituberculate mammals; previously, only one purported bite mark of the eusuchian Allodaposuchus was reported from this locality (Csiki et al. 2010a; Botfalvai et al. 2017). Meanwhile, no tunnels or holes (present in many other sites) were documented from Tuştea. Although not studied herein, the very productive sites near Sânpetru yielded a diverse set of bioerosional trace fossils, including a tunnel in a titanosaur osteoderm ascribed to dermestid beetles (Csiki 2006; Csiki et al. 2010a), as well as probable feeding traces of termites (‘star-shaped pits’), purported bite marks of a multituberculate mammal (‘opposed paired grooves’), and scores that are likely attributable to a ziphodont carnivorous reptile (either a crocodyliform or a theropod dinosaur) (Augustin et al. 2019).

Most of the bioeroded bones display traces produced by a single agent. Only two specimens from Pui exhibit two types of bioerosional traces produced by different culprits: LPB (FGGUB) R.2330 shows bioerosional traces attributed to a multituberculate mammal and, likely, a termite (gnawed surfaces with narrow grooves), while LPB (FGGUB) R.2957 displays traces produced by both dermestid beetles (holes and irregular excavations) and termites (star-shaped pits). At Sânpetru, several different types of bioerosional traces were documented occasionally on the same bone (Augustin et al. 2019), and in some of these cases it was even possible to establish a specific succession of the trace genesis, since they were superimposed.

Bite marks of multituberculate mammals have been reported before from the Late Cretaceous but only rarely. Longrich and Ryan (2010) have previously proposed the occasional consumption of bones by multituberculates, although these authors suggested that this was no regular behavior and that the consumption of carrion by multituberculates was only a supplementary dietary source for them. As posited by Longrich and Ryan (2010), the dense cortex of the inflicted bones is indicative of the consumption of the mineral content of the bone itself, rather than of the removal of soft tissues from within the bone. The pattern of trace distribution concentrated on natural margins and ridges would also support such an assumption, since it follows the pattern of feeding traces produced by squirrels, which feed on bones after the dry stage of carcass decay is reached in order to consume minerals instead of greased or mummified tissue (Klippel and Synstelien 2007). Alternatively, as the multituberculate-related bite marks from the uppermost Cretaceous of Transylvania are only superficial, it can be hypothesized that they were inadvertently produced during the ingestion of the soft tissue (previously, also suggested by Augustin et al. 2019 for the Sânpetru bite mark). In general, multituberculate bite marks are a recurrent feature in the Upper Cretaceous continental deposits of Transylvania (esp. if the gnawed surfaces with wide grooves were indeed inflicted by multituberculates). Contrary to the previous assertion by Longrich and Ryan (2010), scavenging of large vertebrate carcasses seems to have been a regular behavior of the kogaionid multituberculates of Haţeg Island.

Purported termite feeding traces in the form of star-shaped pits have been reported before from the uppermost Cretaceous of the Haţeg Basin, from the site ‘La Scoabă’ of the Sînpetru Formation near Sânpetru (Augustin et al. 2019). Termites today are known to feed on vertebrate carcasses of large to small sizes, as well as on bones ranging from fresh to sub-fossil (Coe 1978; Thorne and Kimsey 1983; Watson and Abbey 1986; Freymann et al. 2007; Backwell et al. 2012; Prestes et al. 2014). Such carnivorous and osteophagous behavior for termites is generally interpreted as a strategy to supplement nitrogen (Watson and Abbey 1986; Freymann et al. 2007; Prestes et al. 2014). In addition to nitrogen, animal tissues may provide other essential nutrients such as proteins, lipids, and certain trace elements (Backwell et al. 2012). The morphology and size of termite feeding traces on bones can indicate how long they were exposed to the feeding activity. More specifically, star-shaped pits appear to develop especially after prolonged exposure – experimental observations with a mound-building termite have shown that star-shaped pits are encountered more than twice as frequently after 6–12 months than after up to 6 months of exposure (Backwell et al. 2012). Thus, the occurrences of termite feeding traces in the uppermost Cretaceous of Haţeg indicate that the bones experienced rather prolonged exposure (subaerial or within the near-surface soil horizon) before final burial. In addition, mound-building termites (the only ones documented to produce star-shaped pits so far; Backwell et al. 2022) do not build their nests in frequently flooded or waterlogged environments and typically forage within a distance of several meters (max. 40 m) around the nest (Augustin et al. 2019, and references therein), meaning that the bioerosion likely happened in relatively well-drained micro-habitats within the paleoenvironments recorded by the ‘Pui Beds’, the ‘Râul Mare River section’ and the Middle Member of the Densuş–Ciula Formation.

Similar to the termite feeding traces and the multituberculate bite marks, tunnels and holes, presumably produced by dermestid beetle larvae while seeking shelter for pupation, have been reported before from the uppermost Cretaceous of the Haţeg Basin, namely the locality ‘La Cărare’ (Sînpetru Formation, near Sânpetru) and the Râul Mare River section near Nălaţ-Vad (Csiki 2006; Brusatte et al. 2017). In modern ecosystems, dermestid beetles are important scavengers, primarily consuming soft tissues of carcasses (Hefti et al. 1980; Kirkland and Bader 2010; Holden et al. 2013; Zanetti et al. 2014). When soft tissues such as skin, muscles, tendons, or cartilage are no longer available, or else under stressful environmental conditions, these beetles can also feed on bone (Hefti et al. 1980; Roberts et al. 2003; Kirkland and Bader 2010; Holden et al. 2013; Zanetti et al. 2014). Dermestid larvae generally colonize carcasses during the dry stage of decomposition, approximately two weeks after the death of the scavenged animal (Bornemissza 1957; Reed 1958; Coe 1978; Timm 1982), and they feed only on tissues exposed to air (Martin and West 1995). The exact amount of time after which a carcass enters the dry stage of decomposition is dependent on several factors, including climate, size of the carcass, and the taxon concerned. Regarding climate, it has been observed that it takes longer for a given carcass to enter the dry stage under lower temperatures (all other parameters being equal). In actualistic surveys of decay rates, it has been demonstrated that mammal carcasses enter the dry stage between approximately two weeks in warm/temperate conditions and up to several months in cold conditions (Komar 1998; Bunch 2009; Michaud and Moreau 2011). The paleoclimate of Haţeg Island, located at 20–30°N during the Late Cretaceous (Van Itterbeck et al. 2004; Panaiotu and Panaiotu 2010), has been reconstructed as warm and subtropical and thus the period upon which carcasses may have entered the dry stage was likely towards the lower end of that spectrum (i.e. few weeks rather than several months). However, it must be noted that most actualistic surveys of decay and decomposition rates use mammal carcasses (and usually pigs), while the herein considered Late Cretaceous taxa (titanosaurian dinosaurs and kallokibotionine turtles) were notably different. Nonetheless, actualistic surveys of decay/decomposition rates exist for different turtles (including semiaquatic and terrestrial ones) exposed to warm conditions on land, under which the carcasses entered the dry stage after 12–14 days (Abell et al. 1982; Brand et al. 2003). Since the holes and tunnels (just as those observed in extant feeding experiments, see above) are likely produced by the dermestid larvae aiming to find or create suitable environments for pupation and they infest carcasses after the latter entered the dry stage of decomposition, the minimum amount of subaerial exposure time can be estimated to range between ten weeks and one year (based on the time it takes dermestids to hatch from their eggs and reach the pupation stage; Smith 1986). Thus, a similar minimal duration of subaerial exposure can be assumed for the uppermost Cretaceous fossil bones from Haţeg Basin as well.

Intriguingly, actualistic decomposition surveys have also noted so-called disarticulation stages that incorporate skeletal articulation and completeness of turtles (Dodd 1995). Using this information, the Kallokibotion sp. carcass described herein (and showing holes and tunnels), can be assigned to disarticulation stage 2 (sensuDodd 1995), with limbs and the head still articulated to the shell. This disarticulation stage occurs between 1 month (in case of the semiaquatic turtles) and 3 months (in case of the terrestrial tortoise), suggesting a maximum duration of exposure of only several months for the Kallokibotion carcass before burial; note that Kallokibotion had a supposedly semi-terrestrial lifestyle (Rabi et al. 2013; Pérez-Garcia and Codrea 2018; Augustin et al. 2021), and thus the upper range of 3 months of this stage is considered more likely.

Taken together, these observations suggest the exposure of the remains showing holes/tunnels span at least ten weeks (but was possibly even much longer), while in the case of the Kallokibotion carcass, a maximum exposure of a few months seems likely. Nevertheless, it must be emphasized at this point that the exact duration of exposure is difficult to determine due to uncertainties regarding the local latest Cretaceous climate (i.e. exact values of temperature and amounts of precipitation, degree of seasonality) combined with a lack of experimental data available for the clades under consideration here. The uppermost Cretaceous continental sedimentary succession at Pui (i.e., ‘Pui Beds’) is interpreted to represent a meandering river floodplain environment under a seasonal, semi-arid climate, typically exhibiting evidences of advanced pedogenic modifications (Van Itterbeeck et al. 2004; Bojar et al. 2005; Therrien 2005; Csiki-Sava et al. 2016). Bioerosion patterns observed in this locality vary depending on bone size and exposure. Some specimens, such as LPB (FGGUB) R.2046, R.2370, R.2428, and R.2597, show bioerosion distributed over the entire bone surface. In contrast, larger bones, such as the titanosaur tibia LPB (FGGUB) R.2429 and the turtle carapace LPB (FGGUB) R.2593, exhibit bioerosion predominantly on one side. Overall, Pui demonstrates more extensive bioerosion compared to other localities, likely reflecting an extended duration of subaerial exposure of the vertebrate remains, and thus the longer time available for tracemakers to act, which is also in general agreement with the higher degree of pedogenesis reported at this locality (Csiki-Sava et al. 2016). The one-sided bioerosion on larger bones may result from their limited movement (after bioerosion has commenced) and partial sinking into the soil layer before final burial, thus restricting exposure to the bioerosional agents.

The latest Cretaceous depositional setting at Nălaţ-Vad (i.e., ‘Râul Mare River section’) is interpreted as a meandering fluvial system, punctuated by locally occurring freshwater ponds (Van Itterbeeck et al. 2004; Ciobanete et al. 2011; Săsăran et al. 2011; Csiki-Sava et al. 2016). The presence of a relatively high groundwater table is associated with coarser-grained channel and crevasse sandstones (Csiki-Sava et al. 2016). Paleoclimatic evidence suggests a subhumid environment with alternating wet and dry periods, likely of monsoonal character (Therrien 2005). Bioerosion at Nălaţ-Vad is generally limited and predominantly one-sided, as observed in specimen LPB (FGGUB) R.2403. The relative rarity of bioerosional traces may be related to the more humid conditions, which might have promoted faster burial of bones, thereby limiting exposure to tracemakers. In addition, the reconstructed wetter settings might have been a comparably unfavorable habitat for some tracemakers, such as termites (which prefer drier areas for building their mounds, see Augustin et al. (2019) or dermestids (which preferably infest dried carcasses, see Csiki 2006).

The fossiliferous succession from Tuştea locality (Middle Member, Densuş–Ciula Formation) represents distal, well-drained floodplain environments, and is capped by a thick bed of fluvial channel deposits (Botfalvai et al. 2017). The broader paleo-landscape included alluvial fans and adjoining alluvial plains, draining a northwesterly source area (Bojar et al. 2010), under a seasonally dry climate (Bojar et al. 2005; Therrien 2005; Csiki-Sava et al. 2016). Bioerosion at Tuştea is generally one-sided, affecting specific regions of the bones, as seen in LPB (FGGUB) R.1883, R.2025, and R.2330. However, some specimens, such as LPB (FGGUB) R.1619 and R.1881, show bioerosion over the entire bone surface, coinciding with a darker reddish coloration. This pattern suggests that many bones became partially embedded relatively quickly, limiting exposure and leading to one-sided bioerosion, whereas others remained exposed more completely and for longer periods, allowing bioeroding organisms to act across the entire bone surface and to produce a higher quantity of traces.

Taphonomic evidence across all studied localities indicates that bones and carcasses were often exposed on the surface for prolonged periods before being transported, sorted, and concentrated by recurring flooding events during the rainy season (Therrien et al. 2009; Csiki et al. 2010a). In an extensive taphonomic survey of the vertebrate assemblages from Haţeg Basin, Csiki et al. (2010a) established the presence of four main taphonomic modes with several sub-categories (each of these characterized by specific sets of features of vertebrate remain preservation and occurrence), including: (i) isolated bones (complete vs. incomplete ones), (ii) associated specimens (partial vs. incomplete skeletons), (iii) macrofossil bonebeds, and (iv) microfossil bonebeds. Among these, three taphonomic modes A1, A2, and A3 (sensuCsiki et al. 2010a) have been identified across the specimens studied herein: complete and incomplete isolated bones, as well as incomplete skeletons. At Pui, our sample includes mostly isolated bones, ranging from complete isolated bones (mode A1, showing low degrees of weathering, abrasion, and breakage), to incomplete ones (mode A2, showing advanced stages of weathering, abrasion and breakage); moreover, an incomplete skeleton (mode B2) of the turtle Kallokibotion is also represented within the sample with bioerosional trace fossils. At Tuştea, the sample of bones showing bioerosion includes only isolated bones (mostly incomplete and heavily weathered ones, mode A2), and at Nălaţ-Vad, the single specimen with bioerosional trace fossils is similarly an isolated incomplete bone (mode A2, but showing a low degree of weathering). Overall, isolated bones are the most frequent type of fossil occurrence in all uppermost Cretaceous continental sedimentary facies of Haţeg Basin, due to fluvial transport and reworking, whereas partial skeletons are found only rarely and are largely confined to certain sites and depositional environments (Csiki et al. 2010a). Moreover, taphonomic histories leading to isolated bone preservation are assumed to be lengthier and process-intensive, which can explain the observed widespread nature as well as higher rates of bioerosion trace occurrence on these elements, compared to the more completely preserved skeletal remains.

Notably, bioerosion has been documented thus far only on turtle (Kallokibotion) and dinosaur (titanosaur, rhabdodontid, hadrosauroid) bones, besides unidentified bone fragments. Although this could be explained by habitat segregation/preferences (meaning that the taxa that show bioerosion on their remains preferred better drained micro-habitats), it could plausibly also just represent a sampling bias, since the taxa in question are the most common and abundant vertebrates from Haţeg Island (at least with respect to identifiable specimens).