The first juvenile dromaeosaurid (Dinosauria: Theropoda) from Arctic Alaska

Compared to the osteological record of herbivorous dinosaurs from the Late Cretaceous Prince Creek Formation of northern Alaska, there are relatively fewer remains of theropods. The theropod record from this unit is mostly comprised of isolated teeth, and the only non-dental remains known can be attributed to the troodontid cf. Troodon and the tyrannosaurid Nanuqsaurus. Thus far, the presence of members of Dromaeosauridae has been limited to isolated teeth. Here we describe a symphyseal portion of a small dentary with two ziphodont teeth. Based on tooth shape, denticle morphology, and the position of the Meckelian groove, we attribute this partial dentary to a saurornitholestine dromaeosaurid. The fibrous bone surface, small size, and higher number of mesial denticles compared to distal ones point to a juvenile growth stage for this individual. Multivariate comparison of theropod teeth morphospace by means of principal component analysis reveals an overlap between this dentary and Saurornitholestinae dromaeosaurid morphospace, a result supported by phylogenetic analyses. This is the first confirmed non-dental fossil specimen from a member of Dromaeosauridae in the Arctic, expanding on the role of Beringia as a dispersal route for this clade between Asia and North America. Furthermore, the juvenile nature of this individual adds to a growing body of data that suggests Cretaceous Arctic dinosaurs of Alaska did not undergo long-distance migration, but rather they were year-round residents of these paleopolar latitudes.


Introduction
Dromaeosauridae [1][2][3] (S1 Table) is a group of predatory theropod dinosaurs evolutionarily close to the origin of birds [4,5]. This clade likely originated in the Middle Jurassic [6,7], with the first definitive dromaeosaurids recovered from Cretaceous deposits [8]. By the Late Cretaceous they reached a virtually cosmopolitan distribution [9], so far remaining unknown only in Antarctica. Given the small to medium size of most dromaeosaurids, and their fragile, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 characters, and another on tooth-crown-only characters. Details on character choice and rationale can be found in Hendrickx et al. [35], where original datasets are also reported as supplementary material. The dentition-based matrix consists of 107 taxa and 146 characters. The tooth-crown-based matrix includes 102 taxa and 91 characters. These dental datasets were analyzed by additionally forcing a topological constraint following Hendrickx et al. [35], reflecting a previously recovered tree topology for Theropoda (e.g. Rauhut and Carrano [36] and Brusatte et al. [37]), with DMNH 21183 as a floating OTU. Character scorings for DMNH 21183 in all three phylogenetic datasets are reported in Supplementary Information Document (S1 Dataset). Characters were all equally-weighted and treated as unordered or ordered following the source literature. The phylogenetic analyses were run using the software TNT [38]. For each dataset, we performed a "New Technology" search that included a combination of random and constraint sectorial searches, ratchet, tree-drifting, and tree-fusing, with ten search replications as the starting point for each hit and searches carried out until 100 hits of the same minimum tree length were achieved (TNT command used is "xmult = hits 100 replic 10 css rss PLOS ONE ratchet 5 drift 5 fuse 5"). The most parsimonious trees (MPTs) obtained were subjected to two rounds of TBR branch swapping (command "bbreak = TBR"). Strict consensus trees were generated from the resulting set of MPTs Nodal support was calculated for the consensus by running a standard bootstrap analysis with 1000 replicates.
To test the qualitative observations reported in the anatomical description section, we performed multivariate analyses of theropod tooth measurements including those retrieved by the erupted tooth in DMNH 21183 (S2 Dataset). Although this tooth is not fully erupted, the breaks on the medial side of the dentary around the area of the relative alveolus and observations under microscopy allowed measurement of the crown height of the tooth. Measurements on tooth morphometrics are from Gerke and Wings [39] and a modified dataset from Larson and Currie [40] (S2 Dataset). The dataset from Gerke and Wings [39] includes measurement data for 335 theropod teeth for which entries were modified to reflect a higher systematic rank or clade compared to the original genus-level classification (e.g. Abelisauridae, Dromaeosauridae, Tyrannosauridae). Measurements for this dataset include crown basal length (CBL), crown height (CH), crown basal width (CBW), and the ratio between mesial (anterior) and distal (posterior) denticles (DSDI).
The dataset from Larson and Currie [40] includes measurement data for over 1200 small theropod teeth, mainly from the latest Cretaceous of the Western Interior Basin of North America, and has also been used for similar studies in the past [e.g. 41, 42]. Principal measurements included fore-aft basal length (FABL), crown height (CH), basal width (BW), and anterior (ADM) and posterior denticles per millimeter (PDM). This dataset was modified by removing all entries with at least one variable missing or unable to be evaluated, such as the lack of a measurement caused by the absence of the related structure (e.g. ADM in teeth without mesial denticles or carina). In this dataset, the taxonomy of the identified teeth was referred to the "family-ranked" clade with the exception of Richardoestesia. For example, Saurornitholestes langstoni and Atrociraptor marshalli are both referred to Saurornitholestinae, following [14]. The Milk River cf. Zapsalis is referred to Saurornitholestes following Currie and Evans [43]. The Aquilan cf. Richardoestesia gilmorei, the Oldman cf. Richardoestesia gilmorei, the Horseshoe Canyon cf. Richardoestesia, the Lancian cf. Richardoestesia, the Aquilan cf. Richardoestesia isosceles, and Richardoestesia isosceles are all referred to Richardoestesia. The subclades of theropods included in this dataset, apart from the specimen studied firsthand (DMNH 21183) are: Dromaeosaurinae, Richardoestesia, Saurornitholestinae and Troodontidae.
We performed two principal components analyses (PCA), a multivariate technique that takes a number of measurements and converts them into a smaller set of values that represent the variability of the sample plotted in a multivariate space. Following the example of similar studies (e.g. [9,35,39]), the same measurements were also used for a linear discriminant function analyses (DFA) on both the datasets. This method provides a value to assess the degree of confidence on the classification of the clusters in the morphospace, where 0.5 is no better than random in model accuracy while 1 represents perfect accuracy (100% accuracy [44]). Both these methods allow the dataset of teeth to be plotted in a morphospace, and to quantitatively compare the degree of overlap and the relative position of DMNH 21183 with 1) the main clades of Theropoda from the Gerke and Wings [39] dataset and 2) the main deinonychosaurian subclades (e.g., Dromaeosaurinae, Troodontidae) from the Larson

Locality
The specimen was collected from exposures of the Prince Creek Formation, at the Pediomys Point locality (Figs 1 and 2). The locality is along the Colville River, 8 km upstream from the Liscomb bonebed, North Slope Borough, Alaska, USA (Fig 1). Bulk sediment was collected at the site over multiple field seasons between 2005-2007, 2012, and 2014, with screenwashing and sorting of the material conducted at the Perot Museum of Nature and Science (DMNH) in Dallas, Texas, USA.

Geological setting and depositional environments
The Prince Creek Formation (PCF) was deposited in the Colville Basin of northern Alaska and provides us with the largest collection of polar dinosaur bones in the world [25]. The PCF was originally sub-divided into two subunits or tongues: an older Tuluvak Tongue and a younger Kogosukruk Tongue [46]. However, Mull et al. [47] revised this nomenclature based on regional stratigraphic correlations and the Prince Creek Formation was redefined to include only the former Kogosukruk Tongue along with some younger, Paleocene strata. The total thickness of the PCF along the Colville River is approximately 450 m [48,49]. Biostratigraphic [50][51][52][53][54][55][56][57][58][59], and isotopic analyses [60] indicate that the age of the Prince Creek Formation ranges from Campanian to Paleocene. All deposits containing evidence of dinosaurs are Early Maastrichtian in age and approximately 68.5-70 million years old [57,59,[61][62][63].
The PCF is an alluvial succession composed of conglomerate, sandstone, siltstone, mudstone, carbonaceous shale, coal, and bentonite [59,61,64]. The mostly fine-grained sediments record deposition in low energy, suspended load channels and on organic-rich floodplains on a low-gradient coastal plain. Thicker, multi-story meandering trunk channels contain the largest grain sizes and record the highest flow velocities in the area. Smaller meandering and fixed or anastomosed distributary channels formed crevasse splay-complexes adjacent to trunk channels. Abundant organic-rich facies were deposited in low-lying areas between the large channels and splay-complexes. Floodplains contained levees, splays, lakes, ponds, swamps, and soil-forming environments. Volcanic ashfall was common and smectite-rich bentonites were commonly preserved in wet floodplain environments [65]. Trampling of sediments by dinosaurs was common [63]. Weakly-developed paleosols similar to modern Entisols, Inceptisols, andic soils, and potential acid sulfate soils formed on levees, point bars, crevasse splays, and along the margins of floodplain lakes, ponds, and swamps that also supported lowland trees, shrubs, herbs, ferns, moss, and algae [61,65]. Macroscopic and micromorphological features, and illite-smectite mixed-layer clay minerals in paleosols indicate predominantly waterlogged, reducing conditions interrupted by oxidizing conditions and periodic drying out of some soils [61,62,64,65]. Soils experienced repeated sediment influx from overbank flooding of nearby distributary channels, and periodic deposition of hyperconcentrated flows [62].
Sediments deposited in the most distal areas of the coastal plain contain evidence of marine influence that includes inclined heterolithic stratification in channel point bars and pyrite, jarosite mottles, jarosite halos surrounding root-traces, and brackish-water fauna in floodplain facies [64,66].
The 750 m-long outcrop belt at Pediomys Point is located near the upriver end of a slough off the main course of the Colville River (Fig 1). Cretaceous environments preserved at Pediomys Point include a meandering distributary channel that transitions laterally into a silt and mud-filled abandoned channel along with floodplain environments that include crevasse splays, levees, small lakes and swamps, floodplain paleosols, and ashfall deposits ( [67], Fig 2). Trampled sediments found above a bentonite (Fig 2) are similar to those described from strata 4.25 kilometers downstream along the Colville River that are interpreted as adult and juvenile hadrosaur tracks along a swamp margin [63]. Rare brackish-water clams (most likely Nucula aff. N. percrassa Conrad; see [66]) and gastropods found near the top of the stratigraphic succession at Pediomys Point suggest an estuarine or lagoonal environment for those deposits [66]. Interfingering of continental-terrestrial and shallow marine deposits, including those of flood basins, interdistributary bays and estuaries were identified in older deposits of the PCF along the Colville River [68] and in younger deposits above the Liscomb Bonebed near Ocean Point [69]. This suggests that the environments at Pediomys Point were transitional between the subaerial coastal plain or delta plain, and shallow marine habitats (Fig 2).

Referred specimen
DMNH 21183. The anterior portion of a right dentary, preserving two teeth and four alveoli (Fig 3).

Description
DMNH 21183 (Fig 3) (Table 1) is an anterior portion of a right dentary with an unerupted mesial tooth (rdt2; Fig 3A-3C) and a distally placed partially erupted tooth (rdt3; Fig 3). The anterior surface of the dentary is damaged, obscuring details of the symphyseal region. Given the position of the erupted tooth (rdt3; Fig 3) above the Meckelian foramina (on the medial surface of the dentary, Fig 3B), and above an the anteroventral process of the dentary (ave; Fig  3A and 3B), we identify this as the 3 rd dentary tooth in the dentary [e.g. 43, 72], with the more mesially positioned, unerupted tooth (rdt2) being identified as the 2 nd . Alveoli 2-4 are preserved ( Fig 3A-3C), although the margins of the latter (fourth) are obscured by erosion. Both the medial and lateral surfaces of the dentary present a parallel, anteroposteriorly-oriented, fibrous bone texture. The anterior margin of the 3 rd alveolus (a3; Fig 3A-3C) has a raised anterior rim that extends dorsally up to approximately mid-height of the crown (a3). There is a well-preserved subtriangular interdental plate lingual to the septum separating the 2 nd and 3 rd alveoli (a2-a3), and a bigger, semicircular one between the 3 rd and 4 th alveoli (a3-a4). The paradental space is relatively dorsoventrally short. Laterally, the dorsal rim of the dentary rises into a triangular ridge, which creates a convex surface anterolaterally that sinks at its base into a circular fossa. The alveoli are elliptically shaped, being slightly anteroposteriorly longer than mediolaterally wide (Fig 3C). The medial side of the dentary features a shallow Meckelian groove (Fig 3B), set slightly more ventrally than mid-height of the bone. The medial side of the dentary features an anteriorly located Meckelian foramen (Fig 3B), which may be paired with a second, more ventrally positioned foramen, but damage to the bone surface in this area makes the identification of this feature uncertain. There is an anteroventral expansion in the alveolar margin of the dentary, visible both medially and laterally (Fig 3A and 3B) that is excavated by a sub-oval fossa on the lateral side ( Fig 3A).

PLOS ONE
most half of portion of the 3 rd tooth (rdt3) is erupted, with 12 denticles visible on the distal carina ( Fig 3A, 3B and 3D).
The teeth have larger distal denticles than the mesial ones. The unerupted portion of the third tooth crown can be seen through a fracture on the medial side of the dentary ( Fig 3B). Taking into account the base of this crown, an estimate of~30 denticles per serrated carina can be inferred for the tooth.
Unfortunately, wear of the mesial carina destroyed some details of the denticles. However, apart from some denticles, most of the interdenticular grooves between adjacent denticles are clear, and they are shallow incisions rather than deep sulci ( Fig 3D, S1 Fig). In the 2 nd tooth this comparison between mesial and distal carina is not possible because the mesial margin of the tooth is hidden by sediment and the labial wall of the alveolus (Fig 4A and 4B). The shape of the apices of the distal denticles is slightly hooked, with an orientation toward the apex of the crown and with an externally rounded margin rather than with a sharp tip (Fig 4A-4C). DMNH 21183 is recovered as the sister OTU of Saurornitholestes, in a partially resolved Eudromaeosauria (sensu Longrich and Currie [14]). This dromaeosaurid clade is represented by a polytomy between Atrociraptor, a monophyletic clade with Deinonychus, Tsaagan and Velociraptor, and another monophyletic clade with Dromaeosaurus and Bambiraptor as successively closer taxa to the clade Saurornitholestes + DMNH 21183.

Analysis
Lastly, the analysis that includes DMNH 21183 in the tooth-crown-based data matrix of Hendrickx et al.

Multivariate analysis results
The dentition of DMNH 21183 (the 3 rd and more exposed-better preserved tooth; rdt3 in Fig  3D) was assessed in a morphometric dataset of theropod teeth [39], with a PCA analysis returning four axes with the following eigenvalues and percentages of total variance explained by each axis: Axis 1 (0.377, 96.583%), Axis 2 (0.006, 1.594%), Axis 3 (0.004, 1.13%), Axis 4 (0.003, 0.69%). Coefficients for the five measurements on each axis are given in Table 2. The majority of the variance is captured in the first two axes of the principal components. The highest variable contribution is represented by CBW (37.79%), followed by CH (~33.70%), CBL (~28.12%), and lastly DSDI (0.38%). The position of DMNH 21183 in the first two axes of the theropod teeth morphospace is shown in Fig 7. DMNH 21183 overlaps the dromaeosaurid morphospace in the lower left quadrant of the plot. This convex hull partially overlaps with those of troodontids, noasaurids, and basal theropods among others. This cluster is set on the opposite side from the centroids of allosauroids, ceratosaurids, and spinosaurids, which occupy most of the center and right area of the teeth morphospace. DMNH 21183 is in the opposite area of the morphospace than tyrannosaurids (Fig 7). A similar spatial arrangement  Table 3.
Coefficients of the PCA analysis run on the theropod teeth dataset published by Gerke and Wings [39] with the addition of DMNH 21183. ADM, anterior denticles per millimeter; BW, basal width; CBL, crown base length (in mm); CBW: crown base width; CDA: crown distal angle; CH: crown height (in mm); CH, crown height; DSDI: denticle size difference index, denticles in mesial carina divided by those in distal carina.
Coefficients of the PCA analysis run on the deinonychosaurian teeth dataset published by Larson and Currie [40] with the addition of DMNH 21183. ADM, anterior denticles per
The majority of the variance is explained by the first two axes of the principal components. The highest variable contribution (similarly to previous analyses on this dataset [40-42]) is represented by CH (~60%) followed by PDM (~30%), FABL (~15%) and lastly ADM (<5%). The position of DMNH 21183 in the first two axes of the deinonychosaurian teeth morphospace is shown in Fig 8. DMNH 21183 overlaps the Saurornitholestinae morphospace, being set further away from the centroid clustering the other analyzed taxon with hooked denticles, Troodontidae. It is outside the convex hull comprising Dromaeosaurinae, which has mostly subequal, rectangular denticles, and is set within the Saurornitholestinae morphospace in a position slightly toward the Richardoestesia morphotype. DFA results in a spatial arrangement of data points similar to the one of the PCA (S3 Fig), and the analyses provides a model accuracy of 0.8 for classification of DMNH 21183 as a saurornitholestine deinonychosaur. This outcome is comparable to those generated by the phylogenetic analyses, and we confidently refer DMNH 21183 to the Saurornitholestinae. We do not at this time assign it to any currently recognized species within the clade.
One major caveat of these analyses regards the potential bias due to the different ontogenetic stages of the teeth included in the sample [41], and most importantly in relation to the likely juvenile growth stage of DMNH 21183. Because PCA analysis would simply group observations based on measurements, it is likely that teeth belonging to juvenile and adult individuals will cluster in separate areas of the morphospace (see Discussion for a more in depth comparison of denticle size). This issue has been shown in the literature (e.g. [73,74]) to affect previous PCAs of tyrannosauroid teeth, so we use this morphometric line of evidence as a complementary tool to assess the systematic identification of DMNH 21183.

Morphological comparisons and the phylogenetic position of DMNH 21183
While DMNH 21183 is fragmentary, there are enough anatomical characters preserved in it to indicate its likely systematic position. Although very few non-dental theropod remains have been found in the Prince Creek Formation of Alaska, a general comparison with other contemporaneous theropod taxa, with a particular focus on those clades previously recognized in the formation (Dromaeosauridae, Tyrannosauridae, Troodontidae) is attempted here. The

PLOS ONE
combination of ziphodont dentition, presence of interdental plates (e.g. [75]), a Meckelian groove, and Meckelian foramina are characteristic of theropod dinosaurs (e.g. [76]). The orientation of the anterior-most tooth socket and its unerupted tooth (both anteromedially oriented), the presence of paired Meckelian foramina, and a ventral expansion laterally pierced by a fossa, point toward the identification of the specimen as an anterior portion (almost symphyseal, as also shown by the ventral enlargement of the dentary in lateral view) of a theropod dentary. The shallow Meckelian groove present in the specimen is a derived maniraptoran feature [77], in contrast to the deep groove seen in basal tyrannoraptorans (e.g.

PLOS ONE
Tyrannosauroidea) [78]. The Meckelian foramen (Fig 3B) is slit-like and not enlarged and rounded like in Tyrannosauridae (e.g. oral mandibular foramen in Nanuqsaurus [23] and foramen intramandibularis oralis in Tyrannosaurus [79]), a feature seen in Dromaeosauridae (e.g. Dromaeosaurus [80]; Acheroraptor [72]). The pair of anterior foramina nearby the symphysis is similar to the condition found in velociraptorines like Acheroraptor [72]. The shallow paradental space, a Meckelian groove that is set in the lower half of the medial side of the dentary but not directly on the ventral margin, and the hooked distal denticles of the teeth are coelurosaurian features [81,82]. The lenticular shape of the alveoli from dorsal view (Fig 3C; S1 Fig), rather than box shaped (or squared off), is more similar to the condition in derived coelurosaurians (i.e. Maniraptora) rather than in Tyrannosauroidea [83]. The presence of distinct alveoli rather than a connected dorsal furrow excludes derived troodontids [83]. The triangular ridge on the lateral margin of the dentary (Figs 3A and 4D) is shared with Buitreraptor [84], Velociraptor (AMNH 6515) and Tsaagan [85], although in all these cases the presence of this structure may be an artifact of the damaged lateral rim of the dentary. The raised rim in front of the well-preserved alveolus 4 ( Fig 3A) resembles the condition of the anterior alveoli in Saurornitholestes [43].

PLOS ONE
Interdental plates are notoriously rare in paravian theropods, and fusion or reduction of these structures has been largely considered synapomorphic for Dromaeosauridae or Deinonychosauria as a whole, except for some taxa like Microraptorinae [86], Austroraptor [87], Acheroraptor [72], and Archaeopteryx [88,89]. The larger of the preserved interdental plates of DMNH 21183 is triangular (like Acheroraptor), with a somewhat arched apex and a broader base. The absence of medial crenulation or dorsoventral furrows in the interdental plate excludes it from assignment to Tyrannosauridae. The presence of interdental plates fused to the margin of the dentary is shared with Atrociraptor [90] and Saurornitholestes [43].
The small size of mesial denticles in DMNH 21183 is more similar to the condition in Saurornitholestinae rather than Dromaeosaurinae [43, 80,[115][116][117][118]. The shape of the denticles is slightly pointed toward the apex of the crown as in Dromaeosauridae and Troodontidae and not C-or U-shaped like in Tyrannosauridae [83]. The reduction or lack of mesial denticulation is shared with some Asian Velociraptorinae like Tsaagan [85]. A similar condition to DMNH 21183 where mesial teeth bear unserrated mesial carinae and denticulated distal carinae is present in many other theropod clades. This feature is also seen in the dromaeosaurid Tsaagan [85], the troodontids Linhevenator [83,119] and possibly Saurornithoides (AMNH 6516 [120]).
The number of denticles (12 to 13) in the partially exposed crowns of DMNH 21183 indicates that the total number of denticles per carina is greater than the~15 present in many troodontids, particularly in taxa more derived than Sinovenator [26]. As Hendrickx [83] pointed out, while some tooth crowns of Saurornitholestes appear to have less than 15 denticles on the carina [40, 92,123] quantitative analyses by Larson and Currie [40] indicates that the large majority of Saurornitholestes teeth have many more than 15 denticles on the crown (for other remarks on morphological variation in the dental series of this taxon see [43]). Within Deinonychosauria, taxa with a large number of denticles (�6 per 1 mm) include Richardoestesia, Saurornitholestinae (including Saurornitholestes), Sinovenator, and Velociraptor [40, 83,[124][125][126][127].
While the position of the mesial margin of the teeth preserved in DMNH 21183 and the degree of surface abrasion preclude detailed morphological observation, the finely serrated mesial carina, with denticles smaller than their distal serial homologues, can be seen particularly well in the 3 rd tooth (rdt3; Fig 3D). The presence of distal denticles larger than mesial ones was long thought to characterize the dentition of Dromaeosauridae, and this criterion was used to identify velociraptorine teeth (e.g., [83,[115][116][117][118]). Teeth with fine mesial serrations are usually characterized by a denticle size index (DSDI: the ratio between number of mesial and distal denticles) higher than 1.2 while teeth that bear carinae with subequal denticle size have usually a DSDI close to 1. These arbitrary values were proposed by Rauhut and Werner [115] and corresponds in the case of DSDI�1.2 to approximately more than six mesial denticles for five distal serrations [83]. DMNH 21183 has a quite high DSDI (~2.3; Fig 9), which is well beyond the range of many deinonychosaurian taxa, and in the range of the most finely serrated saurornitholestine teeth (between 1 and 2.5; Fig 9). A DSDI >1.2 has been reported in the majority of eudromaeosaurians ( [83] and references therein). Some lateral teeth of the troodontid Zanabazar [128] and some isolated crowns assigned to Troodon have also a very high DSDI (outliers with DSDI around 2-3; Fig 9 [40, 83,91]. This is the opposite condition than that typically found in Dromaeosaurinae (DSDI~1), like in the eponymous taxon Dromaeosaurus [83]. The surprisingly high DSDI of DMNH 21183 may be a juvenile trait, since many juvenile theropods have been shown to exhibit particularly fine mesial serrations relative to the distal ones, and evidence from tyrannosaurids [78,129,130] shows progressively decreasing DSDIs through ontogeny [135].
External textural features in dinosaur surface bone have been shown to change in relation to ontogeny [32], in a process mirroring internal microscopic remodeling [131]. For example, the skulls of ceratopsian dinosaurs show surface textural changes during ontogeny, from lightly striated to deeply rugose textures [30 ; 132]. Cortical bone texture with fine-grained, longitudinally striated pattern [31] is considered a size-independent criterion as an indication of relative immaturity in non-avian archosaurs [30,31,133]. Lightly striated cortical bone grain express nascent ontogenetic characters in theropods [31], as it is particularly clear from Tyrannosaurus [31], Scipionyx [134], and Juravenator [135,136]. The same striated, fibrous bone grain texture described in relatively immature individuals of these other theropod taxa is also present in DMNH 21183 (Fig 3A and 3B). This textural feature, in combination with the diminutive size of the specimen (Table 1), is evidence that DMNH 21183 is a juvenile.
After morphological, morphometric, and phylogenetic analyses, DMNH 21183 is here interpreted as specimen of Saurornitholestinae. The recurrent clustering of DMNH 21183 with Eurasian dromaeosaurids (a clade including Eudromaeosauria + Microraptorinae; Figs 5 and 6) and excluding Unenlaginae (in [33]), the sister-taxon relationship with Saurornitholestes (Fig 6A) in the analysis based on the dentition-only matrix from Hendrickx et al. [83], as well as in a eudromaeosaur clade more exclusive than that containing Dromaeosaurus in the analysis based on the tooth-crown-only matrix from Hendrickx et al. [33], all strengthen the saurornitholestine interpretation of the specimen based on morphological comparisons and teeth morphometrics.
Saurornitholestinae was first named by Longrich and Currie [14] as a subclade of Dromaeosauridae including Saurornitholestes, Bambiraptor and Atrociraptor. Not all phylogenetic analyses focused on deinonychosaurian interrelationships recover this clade (e.g. [101,137]), but others do [34,37] or partially do so [138], and we recognize this clade. Saurornitholestinae is that lineage of eudromaeosaurs closer to Saurornitholestes than to Dromaeosaurus and Velociraptor, and this more inclusive, stem-based concept of the name is useful when discussing isolated dental material which has often been assigned to Saurornitholestes or cf. Saurornitholestes  96,139]). There are several referrals to saurornitholestine dromaeosaurs in the latest Cretaceous of North America, including: the Milk River saurornitholestine [96] from the latest Santonian-earliest Campanian; the Early Campanian Menefee Formation saurornitholestine [41]; the Early Campanian Foremost Formation saurornitholestine [140]; and the Middle-Late Campanian Oldman Formation saurornitholestine [140] (S3 Dataset).
More relevant for comparative purposes to taxa in the Prince Creek Formation, Atrociraptor marshalli was recovered in phylogenetic analyses of paravian interrelationships as a derived member of Dromaeosauridae and close to the node comprising Saurornitholestes (e.g. [37,60]). In the dental-characters-only phylogenetic analysis, DMNH 21183 clusters separately from Atrociraptor (Fig 6A), while the crown-only-characters and full-osteological results recover both Atrociraptor and DMNH 21183 in a polytomy within Eudromaeosauria (Figs 5  and 6B). There are similarities in the apically-hooked denticles on the teeth of DMNH 21183, Saurornitholestes, and Atrociraptor, but the generally larger denticles of the latter, particularly in the mesial carina [90,141], sets it apart from Saurornitholestes and DMNH 21183. However, the implications of ontogenetic stage on relative denticle size should also be considered in this case. On the other hand, the interdental plates in DMNH 21183 more closely resemble those of Atrociraptor (TMP 95.166.1 [90]) than those of Saurornitholestes (TMP 1988.121.0039 [60]) in having a narrower base. Atrociraptor and Saurornitholestes are recovered in a sister-group relationship by Currie and Evans [60], and the shared similarities between DMNH 21183 and these two taxa may prove predictive should more complete dromaeosaurid material be found in the PCF.

Paleoecological and paleobiogeographical implications of a juvenile Arctic Saurornitholestinae
The discovery of dinosaur remains at high latitudes (i.e. higher than 66˚), and in particular the abundant dinosaur bone record from the PCF, challenged traditional reptilian models for dinosaurian physiologies and inspired debate centered on the potential for long-distance migrations by dinosaurs [20,[142][143][144]. Given the breadth of migration patterns in extant animals (e.g. [145]), to focus our discussion it is relevant to point out that these migrations for Arctic dinosaurs were inferred to cover latitudinal distances rather than trans-Arctic migrations, which are not even observed with animals today [146]. Increased subsequent interest showed, through a variety of methods such as biomechanic, isotopic analyses, and osteohistology, that these dinosaurs likely had the necessary adaptations for overwintering in the ancient Arctic [20,147,148] and need not have moved to more southerly latitudes. Thus far, these discussions have focused almost exclusively on herbivorous taxa, with one exception that was based on the argument that if the dinosaurian prey did not migrate, then the predators were non-migratory as well [28].
In their review of the adaptive benefits of migration for modern mammals, Avgar and others [149] showed that in the terrestrial realm there is a decided preference for mammalian long-distance migrations to occur among large-bodied herbivores rather than carnivores. One of the suggested reasons for the rarity of migration by mammalian carnivores is that for purposes of energy consumption and mating, these animals need to establish, maintain, and defend territories. Such behavior would preclude the ability to migrate.
Dromaeosaurids were evolutionarily close to avians, and there is some discussion of the flight capabilities of some taxa within this group [150]. At best, some taxa may have been able to exhibit some rudimentary flight skills (e.g. Microraptor and Zhenyuanlong [151][152][153]). Further, the anatomy of most non-micraraptorine dromaeosaurids, especially larger species, argues for a flightless lifestyle [89]. That eliminates flight as an energy efficient means for a predatory animal to cover the vast distances required to migrate across large geographic areas.
The fibrous bone texture of DMNH 21183 strongly suggests that the individual was very young at the time of its death. Given that there is also a demonstrable positive correlation between tooth size and body size, at least in some theropods (such as dromaeosaurids, but does not apply to basal ornithomimosaurs and therizinosaurs with extremely tiny teeth compared to body size [Hendrickx pers. comm. 2020]) and other diapsids [154,155], it is reasonable to conclude that the DMNH 21183 belonged to a very small individual. Further, similar (if not stronger) biomechanical constraints for long-range migrations would apply for these small theropods, as has been advocated for ornithischian dinosaurs [156,157]. Therefore, this specimen of a young, small individual suggests that dromaeosaurs likely nested in the ancient Arctic or in the close proximity, behavior different than long-distance migratory animals. Artistic restoration by scientific illustrator Andrey Atuchin depicts a riparian setting in the Prince Creek Formation, matching the geological evidence described in this paper. DMNH 21183 comes from the juvenile dromaeosaurid on the branch close to the adult, while a subadult (foreground) stalks an individual of Unnuakomys hutchisoni [67], a methatherian known from this locality. Individuals of the sympatric ceratopsid Pachyrhinosaurus perotorum [22] rest in the background. https://doi.org/10.1371/journal.pone.0235078.g010

PLOS ONE
The identification of the first osteological remains attributed to Saurornitholestinae from the North Slope supports previous observations on the paleoecology and ecosystem structure of Late Cretaceous communities in Arctic Alaska. Previous workers suggested that gregarious herbivorous dinosaur groups inhabiting the Cretaceous Arctic were non-migratory [157]. The specimen described here (DMNH 21183) is particularly important because it represents a very small and young individual with little or limited ability to engage in long-distance travel or migrations. In modern migratory birds, it is often the case that young birds lack experience and they must rely on learning migratory habits from the adults [158]. Also, as another analog, there is a decided preference for modern mammalian long-distance migrations to occur among large-bodied herbivores rather than carnivores. Taken together, we infer that DMNH 21183 implies a perennial residency of this dromaeosaur clade (Saurornitholestinae) in the Arctic [20,25,26,28]. This Alaskan Saurornitholestinae would have lived in a biotope featuring a coniferous open woodland (dominated by taxodiaceous conifers) with an angiospermfern understory [61,159,160]. Herbaceous vegetation included ferns, angiosperms, abundant horsetails and other sphenophytes [61,159,160]. This ancient Arctic ecosystem would have included animals such as basal ornithopods [21], the hadrosaurid Edmontosaurus [161], the centrosaurine Pachyrhinosaurus [22], the diminuitive tyrannosaurid Nanuqsaurus [23], a large troodontid [26] and at least another dromaeosaurid taxon closer to Dromaeosaurus [28] than to Saurornitholestes. Small body-sized animals representing potential prey for the Arctic saurornitholestine (Fig 10) might have been mammals such as the methatherian Unnuakomys [67], a Gypsonictopidae and the multituberculate Cimolodon [25,162].
DMNH 21183 adds further weight to the paleobiogeographical connection between closely related Asian and North American eudromaeosaur taxa (with sister clades present in both Asia and the Western Interior Basin of North America). DMNH 21183 is too fragmentary to provide more specific taxonomic distinction within the current record of known dromaeosaurids, but is most similar to saurornitholestines. Given the geological age of Saurornitholestes langstoni and Atrociraptor marshalli, and the wide-ranging tooth-form taxon Richardoestesia (Late Campanian-Late Maastrichtian), we predict that additional specimens and data may eventually provide evidence supporting the establishment of a new dromaeosaurid taxon in the Early to early Late Maastrichtian of Arctic Alaska.