The Burmese amber specimens studied herein (Figs 1–5, S1) originated from amber mines near Noije Bum (26°20′ N, 96°36′ E), Hukawng Valley, Kachin State, northern Myanmar. The holotype is deposited in the Nanjing Institute of Geology and Palaeontology, Chinese Academy of Sciences, Nanjing, China. The two paratypes are deposited in the Leibniz-Institut zur Analyse des Biodiversitätswandels (formerly the Geologisch-Paläontologisches Institut und Museum der Universität Hamburg), Germany. The amber pieces were trimmed with a small table saw, ground with emery papers of different grit sizes, and finally polished with polishing powder.

Photographs under incident light were taken with a Zeiss Discovery V20 stereo microscope or a Leica M205A stereomicroscope. Confocal images were obtained with a Zeiss LSM710 confocal laser scanning microscope, using the 488 nm Argon laser excitation line. Images under incident light were stacked in Zerene Stacker 1.04. Confocal images were stacked with Helicon Focus 7.0.2 and Adobe Photoshop CC. Microtomographic data were obtained with a Zeiss Xradia 520 Versa 3D X-ray microscope at the micro-CT laboratory of NIGP and analyzed in VGStudio MAX 3.0. Scanning parameters were as follows: isotropic voxel size, 1.6106 μm; power, 3 W; accele­ration voltage, 40 kV; exposure time, 1.5 s; projections, 3001. Images were further processed in Adobe Photoshop CC to adjust brightness and contrast.

To test the robustness of the molecular phylogeny of Corylophidae, we reanalyzed the data compiled by Robertson et al. (2013) with a site-heterogeneous model. Eight genes were included, namely the nuclear genes 18S, 28S, H3 and CAD, and the mitochondrial 12S, 16S, COI and COII. All sequences were obtained from GenBank using the Batch Entrez tool. The accession numbers were the same as provided by Robertson et al. (2013). Sequence alignment generally followed the procedure of Robertson et al. (2013), though with slight modifications. In brief, the translated alignments of protein-coding genes were done using MUSCLE (Edgar 2004) module in Geneious 4.8.4 with default parameters. The rRNA genes were aligned using MAFFT 7.49 (Katoh and Standley 2013) Q-INS-i option. The ambiguously aligned regions of 28S (bp 2081–3196 in the alignment result) and CAD (bp 241–360 in the alignment result) were removed.

The site-heterogeneous mixture model CAT-GTR+G4 was run in PhyloBayes mpi 1.7 (Lartillot et al. 2009). Two independent Markov chain Monte Carlo (MCMC) chains were run until convergence (maxdiff <0.3). Convergence was assessed by using the bpcomp program to generate output of the largest (maxdiff) and mean (meandiff) discrepancies observed across all bipartitions.

The tree was drawn with the online tool iTOL 5.7 (Letunic and Bork 2019) and graphically edited with Adobe Illustrator CC 2017.

To evaluate the systematic placement of the new species, a morphology-based phylogenetic analysis was performed. The data matrix was mainly derived from a previously published dataset (Ślipiński et al. 2009; Robertson et al. 2013).

The unconstrained parsimony analysis was performed under implied weights using the program TNT 1.5 (Goloboff et al. 2008, 2016). Parsimony analyses achieve highest accuracy under a moderate weighting scheme (i.e., when concavity constants, K, are between 5 and 20) (Goloboff et al. 2018; Smith 2019). Therefore, the concavity constant was set to 12 here, as suggested by Goloboff et al. 2018. Most parameters were set as default in the “new technology search”, while the value for “find min. length” was changed from 1 to 100.

Since the morphology-based phylogeny of Corylophidae was somewhat discordant with the molecular phylogeny, we additionally conducted a constrained analysis (e.g., Slater 2013; Fikáček et al. 2020). For taxa with both morphological and molecular data, their interrelationships were fixed according to the molecular tree. The fossil and other extant taxa without molecular data were allowed to move freely across the reference tree. The constrained parsimony analysis was performed under implied weights (K = 12) with R 4.1.0 (R Core Team 2021) and the R package TreeSearch 1.0.1 (Smith 2021).

Character states were mapped onto the trees using unambiguous optimization with WinClada 1.0 (Nixon 2002).

The following abbreviations of institution are used: CCGG – Collection Carsten Gröhn, Glinde. GPIH – Geo­logisch-Paläontologisches Institut und Museum der Universität Hamburg. NIGP – Nanjing Institute of Geo­logy and Palaeontology, Chinese Academy of Sciences. The following abbreviations of morphological characters are used: BL – apparent body length in dorsal view; BW – body width; EL – elytral length; HL – head length; HW – head width; PL – pronotal length; PW – pronotal width. The following abbreviation is used in the phylogenetic analysis: PP – posterior probability.

The result under site-heterogeneous model (Fig. S2) was well consistent with the result under site-homogeneous model by Robertson et al. (2013). The monophyly of Corylophidae was strongly supported (PP = 1.00). Corylophinae excluding Holopsis Broun (Peltinodini) was strongly supported (PP = 1.00), while Corylophinae as a whole was only moderately supported (PP = 0.83). All currently recognized tribes (sensuRobertson et al. 2013) were recovered as monophyletic groups. Foadiini (represented by Foadia Pakaluk and Priamima Pakaluk & Lawrence) was moderately supported (PP = 0.74), while all other tribes were strongly supported (PP = 1.00).

The result of unconstrained analysis (Fig. S3) is very similar to that of Ślipiński et al. (2009), only with the position of Sericoderini changed. Xenostanus was resolved as sister to the group consisting of Othoperini, Peltinodini, Sericoderini, Corylophini, Teplinini and Rypobiini. The tribes Aenigmaticini and Stanini were clustered together.

In the constrained analysis (Fig. 6), Aenigmaticini and Stanini were only distantly related in the reference tree. Partly affected by this topology, Xenostanus was recovered as the sister group of Stanini. Teplinini grouped together with Rypobiini, and Cleidostethini grouped together with Orthoperini.

In the combined morphological and molecular data analyses by Robertson et al. (2013), Coccinellidae was sister to Corylophidae. However, in both phylogenies based on molecular data alone (present study; Robertson et al. 2013), Anamorphidae (represented by Bystus Guérin-Méneville and Symbiotes Redtenbacher) was resolved as the sister group of Corylophidae. The sister-group relationship between Anamorphidae and Corylophidae have been further supported by the large-scale phylogenomic study (McKenna et al. 2019). Thus, the frequently advocated combined analysis may not be an ideal solution to utilize the information from both morphological and molecular data as it seems to be.

Compared with the site-homogeneous models, site-he­terogeneous models account for the unequal rate of evolution in sequences and have been proved to be more insensitive to phylogenetic artifacts such as long branch attraction (Lartillot et al. 2007; Pisani et al. 2015; Cai et al. 2020, 2022). Site-heterogeneous models may generate improved results even for analyses with limited gene fragments sampled (Li et al. 2021). For the present study, the internal phylogeny of Corylophidae generated with the site-heterogeneous model (Fig. S2) was essentially identical to that generated with a site-homogeneous model by Robertson et al. (2013). Thus, the current classification scheme of Corylophidae based on this phylogeny is generally satisfactory, and could serve as a framework for determining the position of the Xenostanus fossil.

As discussed by Robertson et al. (2013), the unconstrained morphology-based phylogeny of Corylophidae is largely inconsistent with the molecular one and therefore unreliable. Nevertheless, morphology may still provide some valuable information in cases where molecular sequences are hard or impossible to obtain. The positions of Teplinini and Cleidostethini were not evaluated by Robertson et al. (2013) due to the lack of molecular data. In our constrained morphological analysis, Teplinini was resolved as sister to Rypobiini, which is consistent with the unconstrained analysis by Ślipiński et al. (2009). However, the aberrant tribe Cleidostethini turned out to be the sister group of Orthoperini, and they together were sister to Aenigmaticini, while in the unconstrained analysis Cleidostethini was sister to the whole Corylophinae except Foadiini, suggesting the requirement for further analyses.

Corylophidae is currently divided into two subfamilies, Periptyctinae and Corylophinae (Ślipiński et al. 2009; Robertson et al. 2013). Periptyctinae is composed of only three genera, and was once placed in family Endomychidae (Ślipiński et al. 2001, 2009). Xenostanus clearly does not belong to Periptyctinae, based on its pedicel shorter than scape (pedicel longer in Periptyctinae), absence of prosternal carinae (prosternal carinae present in Periptyctinae), and anterior pronotal margin unemarginated (anterior pronotal margin deeply emarginate in Periptyctinae). Based on the morphological and molecular phylogenetic analyses, 11 tribes have been recognized within Corylophinae (Ślipiński et al. 2009; Robertson et al. 2013). The unique character combination of Xenostanus does not fit well into any of the existing tribes. Xenostanus has an elongate body shape. While most corylophids have an oval to circular body, the tribes Foadiini, Aenigmaticini and Stanini also have an elongate body, and are sometimes referred to as latridiid-like taxa. Xenostanus can be distinguished from Foadiini in the straight anterior margin of pronotum (medially emarginate in Foadiini), and from Aenigmaticini in the antennae with 10 antennomeres (with nine antennomeres in Aenigmaticini; Pakaluk 1985) and the basally widest prothorax (posteriorly narrowed in Aenigmaticini). The phylogenetic analysis suggests a close relationship between Xenostanus and Stanini (Fig. 6). Both taxa share a similar pronotal shape (not constricted posteriorly). Nevertheless, Xenostanus can also be distinguished from Stanini based on the antennae (with 11 antennomeres in Stanini) and the absence of transverse mesoventral carina (present in Stanini).

The postcoxal lines on metaventrite and abdominal ventrite 1 are important diagnostic characters for Xenostanus. These postcoxal lines are usually present in Coccinellidae and some other related taxa (Ślipiński and Tomaszewska 2010; Robertson et al. 2015). However, most corylophids do not possess such lines (e.g., Furukawa 2010: fig. 4D). In extant Corylophidae, only Holopsis (Peltinodini) is known to have metaventral and abdominal postcoxal lines at the same time (e.g., Furukawa 2012). Considering the basal position of Holopsis within Corylophidae, the presence of postcoxal lines have been suggested to be pleisiomorphic for it (Robertson et al. 2013). By contrast, Xenostanus occupied a more derived position in Corylophidae, and its postcoxal lines are therefore likely gained secondarily.

Based on the results of phylogenetic analysis and the above discussion on the morphological characters, we suggest that Xenostanus should be placed in a new tribe, Xenostaninitrib. nov. The discovery of Xenostanus greatly extends the earliest record of Corylophidae, which implies this family had already been diversified by mid-Cretaceous.