Leveraging museum specimens, genomics and legacy datasets to unravel the phylogeny and biogeography of cryptin wasps (Hymenoptera, Ichneumonidae, Cryptini)

Modern genomic techniques have enabled the generation of phylogenetic datasets of unprecedented scale. However, there are also troves of molecular data accumulated from past studies using Sanger sequencing, often at fine taxonomic scales. Combining both sources of data is an obviously appealing possibility, but it can also lead to inconsistency due to high levels of missing data, disparities in the scale of Sanger versus genomic datasets, and little overlap in sequences across terminals. To provide an empirical investigation of the potential of such ‘hybrid’ datasets, we combined data from ultraconserved elements (UCEs) for 183 species of Cryptini (Ichneumonidae, Hymenoptera) with a previously existing dataset of 7 loci and morphological data including 308 species plus outgroup taxa. Bioinformatics pipelines allowed recovery of ‘legacy’ markers from the bycatch of UCE sequencing, reducing the problem of limited character overlap. The resulting tree combining Sanger and UCE data is highly supported and includes dense taxon sampling of the group, allowing for a better understanding of the global radiation of Cryptini. The Neotropical region had the highest phylogenetic diversity but the lowest level of phylogenetic dispersion when corrected for standardized effect size, while the Oriental fauna showed the highest level of phylogenetic dispersion. Our results highlight the potential of hybrid datasets to produce a more complete picture of the Tree of Life combining affordability, robust support and deep taxonomic sampling.

problematic in the past (e.g.Allio, Scornavacca, et al., 2020;Heckenhauer et al., 2018;Matschiner et al., 2020;Ran et al., 2018).Some genomic approaches, such as ultraconserved elements (UCEs: Faircloth et al., 2012), have the added advantage of being suitable for the enrichment and sequencing of DNA from old or suboptimally preserved museum specimens (e.g.Blaimer et al., 2016;McCormack et al., 2015), a crucial necessity if we are to obtain a reasonably complete picture of the Tree of Life.
In spite of these advances, most genomic approaches are still more costly on a per-specimen base than traditional Sanger sequencing, such that including all possible species in genomic datasets is still financially out of range for most research groups (Zhang et al., 2019).At the same time, a wealth of previously generated Sanger-based datasets is already available resulting from decades of research, often providing extensive taxonomic coverage for the groups of interest.An obviously appealing possibility is combining both sources of data, producing integrated sets of Sanger-NGS data (henceforth 'hybrid datasets').Ideally, this would have the potential of allying the robust relationships uncovered by large-scale genomic data with the fine-scale taxonomic resolution from affordable and/ or previously existing Sanger data.
While combining high support and deep taxon sampling is undoubtedly desirable, such integrated datasets can suffer from inconsistencies derived from their 'hybrid' nature.First, hybrid datasets are bound to have a significant proportion of missing data, since the whole purpose of combining the two types of data is to include in the analyses taxa for which genomic data are not available.Second, there is usually a tremendous disparity in the scale of Sanger and genomic datasets, which makes the missing data potentially more problematic for the taxa covered only by Sanger data.Finally, in some cases, there may be little or no character overlap between taxa for which only Sanger data are available versus taxa that were sequenced only with genomic approaches.All these issues can conceivably lead to artefacts and inconsistencies in phylogenetic analyses, and a more thorough investigation of the empirical value of such hybrid datasets is warranted, especially in the specific context of combining Sanger with UCE data.
The issue of character overlap is of particular concern, but fortunately, there are options available for making genomic datasets compatible with previously existing Sanger matrices.In fact, it was aiming at just such an integration that the ant-specific UCE probe set of Branstetter et al. (2017) included baits for 'legacy' markers, comprising commonly sequenced nuclear genes in ants.Even in the absence of specific baits, mitochondrial DNA is commonly recovered as bycatch in raw reads resulting from targeted enrichment (do Amaral et al., 2015;Picardi & Pesole, 2012).There are multiple bioinformatic pipelines aiming at recovering whole or partial mitogenomes from raw sequencing reads (e.g.Allio, Schomaker-Bastos, et al., 2020;Jin et al., 2020), as well as a built-in program in PHYLUCE (Faircloth, 2016) to retrieve DNA barcoding sequences from UCE reads (phyluce_assembly_match_contigs_to_ barcodes).Obtaining at least partial coverage of Sanger loci from genomic datasets can help alleviate the problem of character overlap and allow for sound analysis of hybrid datasets.
This study aims to investigate the potential of hybrid datasets using a hyperdiverse group of parasitic wasps, the Cryptini (Hymenoptera, Ichneumonidae, Cryptinae) as a study group.Cryptins 1 are a cosmopolitan lineage currently with 250 genera and over 2400 species (Yu et al., 2016, and recent additions).Species of Cryptini are usually parasitoids of holometabolous insects, and the tribe collectively attacks species of Lepidoptera, Coleoptera, Hymenoptera and Neuroptera, with some species acting as predators on spider egg sacs (Gauld, 2006).
The most comprehensive appraisal of the phylogeny of the group was the recent work of Santos (2017) based on seven molecular loci (two mitochondrial and five nuclear) and 109 morphological characters.The taxon sampling included 308 species belonging to 182 of the 250 cryptin genera (72.8%), with many genera represented by multiple species.Although it represented an important advance in our understanding of cryptin phylogeny, this study had two limitations.First, overall support for the recovered relationships was low, with most of the nodes in the backbone of the Cryptini tree poorly supported; several relatively stable major clades were identified, but the relationships among them were equivocal.Second, many taxa could not be included in the analyses due to the lack of fresh specimens for PCR and Sanger sequencing.
Herein we leverage the power of UCEs in generating large-scale datasets and in sequencing old museum specimens, with the goal of improving backbone support of the Cryptini tree and placing previously unavailable taxa in the phylogeny.We then integrate this new phylogenetic framework with the previously available dataset of Santos (2017) to provide fine-scale taxonomic resolution at shallow nodes.
Two corrections should be made regarding the taxon sampling of this previous work.Upon closer examination, it was observed that the specimen identified as Paragambrus sapporoni is in fact a male of Trychosis sp.This was among the few specimens that had not been determined by the author in the study of Santos (2017), and the determination of male specimens is often challenging in Cryptini.Likewise, re-examination of synoptic material and the relevant literature showed that the specimen previously assigned to Neocryptopteryx blanchardi (=Trachysphyrus blanchardi in Porter, 2008) is actually conspecific with Trachysphyrus riojanus Porter.
Apart from this previously existing dataset, Sanger sequencing data for all seven loci were added for the recently described Cryptoxenodon metamorphus (Supeleto, Santos, Basílio, & Aguiar, 2020).In addition, morphological characters were scored for 21 of the 42 newly sequenced taxa (see Appendix S1).

| UCE dataset
To increase compatibility between the Sanger/morphology and genomic datasets, UCE sequencing was performed whenever possible using the same extracts used in the Santos (2017) study: a total of 139 Cryptini and 52 outgroup taxa from that work were sequenced.Furthermore, 44 new Cryptini species from 40 genera were added; 31 of these 40 genera were not represented in the original 2017 work, mostly representing taxa for which specimens were recently obtained or were too old to be easily sequenced in a PCR-Sanger framework.Part of the UCE dataset has been used in other recent studies focused on testing the phylogenetic placement of specific cryptin taxa (Santos et al., 2019(Santos et al., , 2021;;Supeleto, Santos, Brady, & Aguiar, 2020).Lab protocols for library preparation, enrichment and sequencing are described in detail in those works, representing well-established protocols for UCEs with Hymenoptera, and are summarized as follows.Genomic DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) and an aliquot of <5-150 ng was used as input for library preparation.When necessary, DNA was sheared by sonication (Q800; Qsonica Inc., Newtown, CT, USA) to obtain fragments with a size range of approximately 200-600 bp.Sheared samples were dried down to 13 μL and prepared using the Kapa Hyper Prep Kit (Kapa Biosystems, Wilmington, MA, USA), with custom, dual-indexing adapter-primers (Glenn et al., 2019) and with bead cleaning steps using an AMPure substitute (Rohland & Reich, 2012) in between enzymatic steps.Pools of 8-10 libraries were used for UCE enrichment using a custom probe library targeting 2590 loci (Branstetter et al., 2017) using the MYBaits kit procedure (Blumenstiel et al., 2010), followed by a final PCR step using Kapa reagents.After enrichment, 10-16 pools were combined at equimolar ratios for a total of 96-120 samples per sequencing run on Illumina HiSeq 2500 or HiSeq 4000 platforms (2 × 150; Illumina Inc., San Diego, CA, USA).Raw sequence reads for all samples are available from the NCBI Sequence Read Archive under BioProject accession PRJNA632862.
Informatic processing and analysis were conducted using the Smithsonian's High-Performance Computing cluster (Smithsonian Institution, 2020).Illumina sequencing reads were filtered and trimmed using Illumiprocessor (Bolger et al., 2014;Faircloth, 2013) and assembled using either Trinity v. r2013-02-25 (Grabherr et al., 2011) or SPAdes (Bankevich et al., 2012).Assembled contigs were processed with Phyluce as follows.First, contigs were queried against a FASTA file of all enrichment baits, creating a relational database with the location of the UCE loci.Individual loci were then extracted to separate FASTA files, and each locus was aligned using MAFFT v. 7.130b (Katoh et al., 2002) and trimmed with GBLOCKS v. 0.91b (Castresana, 2000;Talavera & Castresana, 2007).Alignments were then filtered to include only loci available for at least 50% of the taxa, aiming at a compromise between low missing data and high number of loci; the resulting matrix included 1448 concatenated alignments for a total of 337,851 base pairs (Appendix S2).This particular cutoff of 50% was shown in previous works to be suitable for datasets of Ichneumonoidea, either recovering similar results as more complete datasets (Santos et al., 2021) or outperforming less stringent datasets with higher levels of missing data (Jasso-Martínez et al., 2022).

| Data mining for sanger data
To a certain extent, the compatibility of the two datasets is ensured by the use of the same samples from the Sanger-morphology dataset to generate UCE data, providing a level of character overlap to the hybrid dataset.In order to increase this compatibility, we attempted to bioinformatically extract sequences for the Sanger loci from the genomic data generated for UCE sequencing.We did this using the script assembly_match_contigs_to_ barcodes from the Phyluce v1.5 pipeline (Faircloth, 2016).
The script takes a template FASTA file with representative sequences for a given locus and extracts matching regions from all contigs in which it is present.As it might be expected, capture success was higher for mitochondrial and ribosomal loci and relatively low for protein-coding nuclear genes, but the use of this approach allowed us to obtain data from at least some Sanger loci for almost all species for which only UCE data was available (Table 1;  Appendix S3).In addition, for some samples it allowed us to recover sequences for Sanger loci that had not been successfully sequenced in the previous study (Santos, 2017).
Captured sequences for each locus were added to the respective alignments from the Santos (2017) work and realigned using MAFFT under the same parameters of the original work.A rapid Neighbour-Joining tree produced by Geneious (Biomatters) was used to identify abnormally long branches, which in COI and 16S samples usually corresponded to contaminations by the endoparasite Wolbachia (Rickettsiales).Sequences representing contaminations or containing stop codons (for proteincoding genes) were removed and the samples were realigned once more prior to phylogenetic analyses.A final combined matrix was then produced including a total of 410 terminal taxa (Figure 1; Appendix S4); with the addition of 31 cryptin genera sequenced with UCEs to the original dataset of Santos (2017), generic representation of the combined dataset was increased from 181 to 209 out of 250 genera of Cryptini (83.6%).

| Phylogenetic analyses
Phylogenetic analyses were conducted first on the UCE data alone, comprising a matrix of 235 taxa and 1448 loci with 30.8% missing data, analysed as follows.The concatenated dataset was initially partitioned using the SWSC-EN algorithm (Tagliacollo & Lanfear, 2018), which uses individual site characteristics to define partitions that account for rate heterogeneity and patterns of molecular evolution within each UCE locus.The resulting data blocks were then partitioned by partition schemes defined by PartitionFinder2 (Lanfear et al., 2016).Maximum-likelihood analyses were run with IQTREE v1.6.12 (Nguyen et al., 2015), with 10,000 rounds of ultra-fast bootstrapping (Hoang et al., 2018) and an implementation of the SH-like approximate likelihood ratio test (SH-aLRT; Guindon et al., 2010) to assess clade support and using ModelFinder (Kalyaanamoorthy et al., 2017) to choose the best model for each partition via the option -MFP.Analyses were run with the safe numerical mode (option -safe) to avoid numerical underflow that can result from large datasets.The tree was rooted with the xoridine Odontocolon albotibiale (Bradley) as in the original analyses of Santos (2017).
The total evidence dataset (UCE, Sanger and morphology) was also analysed with IQTREE under similar parameters, except that the topology was constrained to conform to the results of the UCE-only analysis.The reasoning for performing a constrained tree search was that such an approach (a) helps to avoid potential inconsistencies caused by the larger amount of missing data in the combined dataset, by fixing the backbone of the tree to the (putatively superior) analysis of UCE data alone; (b) greatly reduces computational time, making tree search up to 10 times quicker in preliminary analyses.The constrained tree search was implemented in IQTREE via the -g option, which inputs a guiding cladogram as part of the analysis.
To provide a preliminary assessment of how the global radiation of Cryptini is distributed across biogeographic regions, we calculated two metrics of phylogenetic diversity for each region: (1) Faith's (1992) phylogenetic diversity (PD), which is the sum of the lengths of the branches of the tree corresponding to a community or region; (2) a standardized effect size of the phylogenetic diversity against a background of null (randomized) communities.The latter controls for the fact that PD is largely contingent on the overall species richness, and therefore presents a metric of how phylogenetically dispersed or concentrated a community is relative to their overall species richness (Webb, 2000;Webb et al., 2002).Both metrics were calculated using the 'ses.pd'function in the R package picante (Kembel et al., 2010), with the 'taxa.labels' null model, which shuffles taxa labels across tips of phylogeny as part of the statistical testing.
T A B L E 1 Success metrics for the capture of Sanger loci as bycatch from the UCE sequencing raw reads.Note: # target sequences: total number of taxa targeted for capture, i.e. the 41 taxa sequenced only with UCEs plus the taxa for which that locus had not been successfully sequenced in the study of Santos (2017).# recovered sequences: number of sequences recovered and validated through BLAST searches and Neighbour-Joining branch lengths.

| RESULTS
The analysis of the UCE data alone resulted in a tree with high support (average bootstrap support [BTP] = 97.06,average SH-aLRT = 95.33);all relationships described below have full support unless otherwise specified (Appendix S5).In terms of the placement of outgroup taxa (Figure 2), both Ophioniformes (with representatives from four subfamilies) and Ichneumoniformes were recovered as monophyletic.The Agriotypinae were recovered as sister to all other Ichneumoniformes.The Ichneumoninae were sister to a clade containing most species of Phygadeuontinae plus Microleptes Gravenhorst (Microleptinae).Ateleutinae were sister to the phygadeuontine genus Hemiteles Gravenhorst, with this clade sister to Cryptinae + Adelognathinae.The Cryptinae (sensu Santos, 2017) were monophyletic, as were its two tribes, Aptesini and Cryptini, with the exception that an unidentified species of Paragambrus Uchida (Cryptini) was recovered within the Aptesini, as sister to Mansa Tosquinet.Within Cryptini, Helcostizus Förster was recovered as sister to all remaining taxa, which were grouped in large clades closely matching the informal genus groups defined by Santos (2017); see Appendix S6).The analysis of the total-evidence datasets also resulted in a highly supported tree (average BTP = 94.02,average SH-aLRT = 92.04),suggesting that such a 'hybrid' approach can successfully combine the robustness of phylogenomic analyses with fine-scale taxonomic sampling.Higher-level relationships essentially follow those of the UCE-only tree since almost all of the non-Cryptini taxa were represented in the UCE dataset (Appendix S7).Within Cryptini, almost all of the newly added taxa were placed with maximum support in the tree, even in cases where relatively few UCE loci were recovered (Figures 3-6).For instance, Spathacantha sp.(106 loci recovered) was recovered within a group of other Afrotropical genera in the Gabunia group; Dagathia philippina (103 loci) was sister to Amrapalia multimaculata; the Malagasy Afretha exavata (117 loci) was sister to the Malagasy Mesophragis inermis; and the Chilean Caenopelte pallinorsa (227 loci) was sister to the morphologically similar, Chilean species Anacis rubripes.In contrast, some taxa Absolute metrics of phylogenetic diversity (Faith's PD) were almost fully explained by species richness in each biogeographic region (R 2 = .96;see Figures 3-6), with the Neotropical region showing the highest number of sampled taxa ( 114) and the highest PD (2.589), followed by the Oriental region (81 spp., PD = 2.294), Afrotropical (53, 1.312), Nearctic (38, 1.116), Australasian (31, 0.851) and Palearctic (27, 0.709).When standardized for effect sizes, the Neotropical region was shown to be the area with the highest concentration of taxa in closely related lineages (PD.obs.z= −5.274).In fact, all regions had communities that were less phylogenetically dispersed than expected by a null model, except the Oriental region, which can thus be considered the area with the highest phylogenetic diversity when controlled for sample size (PD.obs.z= −1.114).

| UCE dataset
The topology of the tree resulting from the UCE-only dataset is largely congruent with the total-evidence maximumlikelihood analysis of Santos (2017) based on seven molecular loci and 109 morphological characters.In terms of the placement of outgroup taxa, both trees show a monophyletic Cryptinae with two tribes (Aptesini and Cryptini); the Ateleutinae as separate from Cryptini and closely related to Hemiteles; a polyphyletic Phygadeuontinae; and the Agriotypinae as the sister to all other Ichneumoniformes.The results of both studies differ with regards to the placement of Ichneumoninae and Phygadeuontinae.In the topology of Santos (2017), the Ichneumoninae were more closely related to Cryptinae than the bulk of the Phygadeuontinae (called 'Phygadeuontinae sensu stricto' in Santos, 2017); in the current study, the Ichneumoninae were sister to a clade containing the bulk of the representatives of Phygadeuontinae, a result also recovered by Bennett et al. (2019) though with much smaller taxon sampling.Meanwhile, another recent study (Santos et al., 2021) using UCEs to investigate specifically the placement of Ichneumoninae recovered the subfamily as sister to all other Ichneumoniformes minus Agriotypinae, while the Phygadeuontinae were closer to Cryptinae.Hence, there are three different 'three-taxon statements' for the grouping of Cryptinae, Ichneumoninae and the 'Phygadeuontinae sensu stricto', all supported by different molecular datasets (Figure 7).
The conflict in the higher-level topologies between this study and that of Santos et al. ( 2021) is surprising considering both studies used similar UCE-based matrices.This suggests that differences in taxon sampling can play a major role in the topologies recovered by the analysis of UCE data.The degree to which taxon sampling can impact phylogenetic reconstruction has been a topic of considerable debate (see Nabhan & Sarkar, 2012 for a review), especially in the context of a trade-off between adding more taxa versus more genes to molecular analyses (e.g.Crawley & Hilu, 2012;Cummings & Meyer, 2005;Peloso et al., 2016;Rokas & Carroll, 2005).Clearly, further analyses with a better sampling of Phygadeuontinae remain necessary to clarify the higher-level relationships among the Ichneumoniformes subfamilies.
The placement of Paragambrus in Aptesini was previously unsuspected; the genus was originally assigned to Cryptini (Uchida, 1936) and this placement was maintained by Townes and Townes (1962), Townes et al. (1965) and Townes (1970), who classified Paragambrus in the 'Baryceratina', a subtribe characterized by the dorsal valve of ovipositor having distinct teeth or ridges, mandible distinctly tapered towards apex and for their use of pupae of slug moths (Lepidoptera, Limacodidae) as hosts.It is worth pointing out, however, that Paragambrus shows some notable similarities with the aptesin Mansa, recovered as its sister group in our analyses: both genera have the dorsal valve of ovipositor with distinct ridges, a very large areolet in the fore wing, elongate propodeal spiracle and T1 much longer than T2.These morphological similarities provide more credibility to the sister group relationship of both genera and since Mansa is well-established as a member of Aptesini, to the placement of Paragambrus in this tribe.
Within Cryptini, the included taxa were grouped in large clades with taxonomic composition almost entirely matching the informal genus groups defined in Santos (2017).The relationships among clades were also similar, except for a swap in relationships in a few clades (Figure 8).Within each clade, relationships often changed considerably between the two trees, with the consensus inside each genus group mostly collapsed.The overall SPR distance between the two trees, when both trees were pruned to include only the overlapping taxa, was of 40 SPR moves.The trend of UCE data confirming the major genus groups observed by Santos (2017) had already been shown in recent studies (Santos et al., 2019;Supeleto, Santos, Brady, & Aguiar, 2020).
The use of UCE data resulted in considerable improvement of clade support for the backbone of the Cryptini tree when compared to the analyses of Santos (2017).The latter maximum-likelihood tree had an average BTP of 83.7, but many deep nodes were poorly supported.The tree resulting from the analysis of UCE data has an average BTP of 97.2, including full or 99.0 support for the relationships among major genus groups.

| Combining datasets
The use of hybrid datasets achieved the desired goal of combining strongly supported results with fine-scale taxonomic coverage.The addition of UCE data to the Sanger/ morphology dataset of Santos (2017) resulted in the elevation of average BTP from 67.3 (mistakenly reported then as 83.7) to 94.0, as well as a robust average SH-aLRT of 92.0.Conversely, integrating the finer-scale Sanger/morphology data with the genomic dataset led to an increase of 43% in taxon sampling, allowing for a more complete assessment of the evolutionary history of Cryptini.
The addition of large-scale genomic data has been shown to improve support and resolution to the backbone of the tree even when available only for a subset of the taxa in the overall dataset (Baker et al., 2020;Leaché et al., 2014;Li et al., 2022;Peloso et al., 2016).Meanwhile, improving taxon sampling is also known to enhance the quality of phylogenetic inference (Heath et al., 2008;Pollock et al., 2002;Zwickl & Hillis, 2002).Finally, Azevedo et al. (2022) showed that adding morphological data still improved support levels even when incorporated into a dataset including genomic data, lending further support to the notion that combining the maximum number of data sources is the ideal approach.
This combined approach inevitably results in substantial amounts of missing data (e.g.66% missing data in Baker et al., 2020; 60.4% missing data in this study), but both empirical and simulations-based studies have shown that increased missing data often do not strongly affect topology (de La Torre- Barcena et al., 2009;Jiang et al., 2014;Li et al., 2022;Wiens & Morrill, 2011), particularly when the topology can be constrained to conform to the genomically-inferred backbone.Missing data may be a more serious concern in cases of extreme heterogeneity in taxon coverage: Letsch et al. (2021) added transcriptome data for 30 species to a Sanger dataset of 498 taxa and found that a supertree approach was favourable to concatenation.In our case, species that had both genomic and Sanger data comprised 46.5% of the OTUs, spanning all of the subclades of the cryptine tree.Approximate evenness in higher-level taxonomic coverage is a potentially important factor in combined studies, as it allows the more restricted Sanger data to drive the resolution of mostly shallow nodes, where relatively few characters may be enough for highly supported inferences.Overall, we concur with Peloso et al. (2016) in that background knowledge of the phylogenetic tree can be used to identify those taxa that should be more thoroughly sampled (e.g. by using genomic techniques) versus those for which limited character sampling suffices for unambiguous placement.
Given the advantages of combined analyses, the abundance of published and unpublished 'legacy' Sanger data and the ease with which both data sources can be made compatible (e.g.direct enrichment of legacy loci, bioinformatic mining of bycatch), it is somewhat surprising that this approach has not been used more often in UCE studies - particularly with Hymenoptera, which have enjoyed an explosive growth in the use of UCE data (Zhang et al., 2019).While Branstetter et al. (2017) provided an early example of the successful combination of both sources of data with a topology constrained to that of the UCE tree backbone, subsequent studies that incorporate both UCEs and Sanger data tend to analyse them as separate datasets (Branstetter & Longino, 2019;Longino & Branstetter, 2020;Ward & Branstetter, 2022).The use of hybrid datasets has the potential to simultaneously integrate thousands of species into the Hymenoptera tree of life while maintaining robust support for higher-level evolutionary relationships.

Cryptini
When comparing the cryptin fauna of different biogeographic zones, the Neotropical region had the highest level of phylogenetic diversity, but this was due to the larger number of species from that region represented in the tree; raw phylogenetic diversity is strongly dependent on the overall species richness because a larger number of species results in more branch lengths to be added up to the total index.When standardized by effect size, however, the Neotropical region showed relatively little phylogenetic dispersion, likely related to most of its representatives being concentrated in the Lymeon groupwhich is comprised almost exclusively of Neotropical taxa, with a few Nearctic offshoots (Figure 6).While there is no direct estimation of the timing of Cryptini diversification, existing divergence dating analyses including cryptins as outgroup taxa suggest that the group may be approximately 45-57 million years old (Santos et al., 2021;Spasojevic et al., 2021), thus originating at a time when the continents were close to their current configuration.The distribution of Neotropical taxa in the cryptin tree is consistent with the idea that the group would have originated elsewhere and dispersed into the Neotropics through a few colonization events, each resulting in relatively large radiations.This would suggest that the current large number of species in the region results from a higher rate of net diversification (Mittelbach et al., 2007), contra explanations that invoke a 'time-for-speciation' effect (Jetz & Fine, 2012;Stephens & Wiens, 2003).
Indeed, some degree of 'phylogenetic conservatism' can be observed in the fauna of almost all regions, with communities significantly more phylogenetically clustered than it would be expected by a null, random model.The Oriental region showed the largest phylogenetic dispersion relative to its sample size, the only region where this metric did not differ significantly from the null model (p = .133).Members of almost all major clades are found in the Oriental region, which could potentially be an indicator that cryptins have been present in the area since the origin of the group, though investigating such a hypothesis would depend on formal historical biogeography analyses and a more thorough taxon sampling.
Our analyses of phylogenetic diversity should be viewed as preliminary: we have not achieved complete sampling at the generic level and the specieslevel taxonomic sampling represents only about 14% of the described diversity of Cryptini, with no specific sampling design to include the same proportion of the species found in each biogeographic region.It is possible, however, that our taxon sampling is not far from approximately reflecting the real species richness in each region: while that is not necessarily the case for all Ichneumonidae (Quicke, 2012), the Cryptini are certainly more species-rich in the tropics.Among tropical regions, Townes (1970) had estimated that the Neotropical region contains more species of Ichneumonidae than the Oriental or Afrotropical regions.There is also evidence that the Afrotropics are somewhat depauperate when compared to tropical regions in the Americas or Asia (Raven et al., 2020).The representation in our dataset, however, certainly underestimates the fauna from the Palearctic region at the species level.
In our study, species of Cryptini were grouped in almost exactly the same larger groups as recovered by Santos (2017), though the relationships among some of the groups changed (Figure 8).Likewise, the relationships within groups were also often different than those in Santos (2017) - the SPR distance between the two trees when both trees were pruned to include only the overlapping taxa (i.e. the 370 original terminals of Santos, 2017) was of 64 SPR moves (subtree prune and regraft; see Bordewich & Semple, 2005).Herein we discuss the composition of each clade, with particular reference to the relationships of newly added taxa.

| Helcostizus
The genus Helcostizus was recovered as sister to all other Cryptini, a result consistent with the maximum likelihood analyses of Santos (2017) (hereafter abbreviated as 'S17'); parsimony analyses in S17 had recovered Helcostizus as part of the Gabunia group (see below).Until 2017, Helcostizus was part of the Phygadeuontini (now Phygadeuontinae), placed by Townes (1970) in the subtribe Mastrina.Previous analyses using a single molecular locus (Laurenne et al., 2006;Quicke, 2012) had already recovered Helcostizus within Cryptini, but that placement was then considered 'anomalous' (Laurenne et al., 2006).Future work with more refined taxon sampling is needed to test whether Helcostizus is indeed an isolated offshoot of early cryptin diversification or if it is part of a clade with other, as of yet unknown, representatives.

| Clade A: Gabunia group
This clade roughly corresponds to Townes' subtribe Gabuniina: composed exclusively of parasitoids of deeply concealed hosts.Species of this group show what Santos and Perrard (2018) termed the 'Dutilleul syndrome', a series of morphological adaptations to find and reach hosts under wood or other deep, hard substrates.Such traits include a hammer-like antennal tip, subspherical head, swollen fore tibia, strong oviposition muscles on T7-T8 and a stout, straight ovipositor with the tip of the ventral valve partially overlapping the dorsal valve.These traits, however, have convergently evolved in a number of other lineages of Cryptini, some of which had been previously assigned to Townes' Gabuniina.Herein we added members of four additional genera previously assigned to 'Gabuniina'.(1) Spathacantha Townes was so far known only from the holotype of S. apicalis Townes; we obtained a specimen that matches well the description and taxonomic delimitation of the genus but clearly belongs to a different species.It was recovered inside the Gabunia clade, closely related to other exclusively or largely Afrotropical genera such as Gabunia, Gerdius and Schreineria.(2) Trypha Townes was recovered as sister to Agonocryptus Cushman.This Neotropical genus is unusual among members of the Gabunia group for its slender body and thin ovipositor, and was so far known only from the type series; additional specimens were rediscovered in the same locality where the type series had been collected.(3) Dagathia Cameron was recovered as sister to Amrapalia Gupta; the latter genus was erected to accommodate part of the species previously assigned to Dagathia, and the two genera are morphologically similar (Gupta & Jonathan, 1970).( 4) Tanepomidos Gupta & Jonathan, originally described in Townes' Gabuniina, was not recovered in this clade but instead as part of the Mesostenus group, within a small clade of Oriental species with overall similar morphology, particularly Perjiva.
The internal topology observed for the Gabunia group differs substantially from that of S17: notably, Kemalia Koçak and Echthrus Gravenhorst form a grade relative to the remaining members of the clade rather than being sister groups; and Anepomias Seyrig was sister to all other taxa except the two aforementioned genera.Among the remaining taxa of the clade, the Neotropical genera are found in two clades that form a grade leading to a single clade containing exclusively Old World taxa. 4.3.3| Clade B: Ischnus group   As in S17, this group is divided into two subclades: one includes the large and widespread genus Ischnus Gravenhorst and four other somewhat similar genera, and the other includes a heterogeneous assemblage of New World taxa.Topologies recovered with our four different analyses are largely similar, except for differences in the internal phylogeny of Ischnus and the placement of Whymperia Cameron.Relative to the dataset of S17, only two additional species of Camera were added to provide deeper molecular coverage for the genus, represented by only three Sanger loci in the previous analyses. 4.3.4| Clade C: Xylophrurus group   This small clade had the same taxonomic sampling as in S17, but its topology changed: in all four analyses, Dihelus was sister to Enclisis + Xylophrurus, whereas S17 had recovered (Dihelus + Enclisis) + Xylophrurus. 4.3.5 | Clade D: Cryptus group   Relative to S17, two genera were added to this clade: Synechocryptus Schmiedeknecht was always recovered as sister to the Afrotropical Gessia Townes.The monotypic Nearctic genus Reptatrix Townes was sister to the Eastern Palearctic Calaminus Townes.The internal topology of this clade largely agrees with the results from S17.All results indicate the need for further work to better understand the taxonomic limits among the morphologically similar Cryptus Fabricius, Buathra Cameron, Meringopus Förster and Nippocryptus Uchida. 4.3.6 | Clade E: Mesostenus group This was the largest individual genus group recovered in S17, mostly comprised of Old World taxa but also including some widespread genera (e.g.Mesostenus) and a few speciose Neotropical genera (e.g.Polycyrtus, Cryptanura).Herein, representatives of seven genera were added to the dataset.Some relationships for the newly included genera were particularly congruent with biogeography.The Malagasy genus Afretha was recovered as sister to the sympatric Mesophragis.Buysmania oxymora, which is widespread in the Oriental region, was nested within Euchalinus, an Oriental genus for which the studied specimens came from Indonesia and the Philippines.The two small Neotropical genera Mecistum and Hercana were recovered as sister groups, both belonging to Townes' former subtribe Mesostenina.Tanepomidos assamensis, from India, was recovered nested in a clade that includes other Oriental genera such as Goryphus, Skeatia, Perjiva, Euchalinus and Buysmania.Given the paraphyletic nature of Goryphus, the generic limits within this clade need to be revised to identify monophyletic and diagnosable supra-specific units.
Conversely, the Oriental genus Lipoprion was recovered as sister to the Neotropical Cryptanura.In this case, both genera share distinctive morphological traits such as the transversely elongated areolet and S1 ending at spiracle level, as well as general similarity in the shape of the head and mesosoma.
Contrary to the results of S17, we found Mesostenus as a polyphyletic group.This is not entirely surprising, as in the early taxonomic history of Ichneumonidae this genus used to be a general repository for all species of Cryptini with a small areolet (the species with large areolet being housed in Cryptus).While Townes' extensive taxonomic revisions (e.g.Townes, 1970;Townes & Townes, 1962, 1966) have greatly contributed to improve the delimitation of these traditional 'wastebasket genera', these older and larger groups are often the ones that have been recovered as para-or polyphyletic in largescale phylogenies. 4.3.7 | Clade F: Ceratomansa group Three genera were added to this small clade of exclusively Australasian taxa.Junctivena gallowayi Gauld was sister to Ceratomansa + Wuda, and Aprix nutatoria and Neaprix sp. were sister to each other.The internal topology of the clade differs substantially from that of S17; notably, the taxa showing the 'Dutilleul syndrome' traits (Wuda, Lorio and Lophoglutus) appear in two separate clades rather than forming a monophyletic group; since Aprix also shares some of these morphological traits, taxa with putative adaptations to attack deeply concealed hosts seem to have evolved at least three times within the clade.Their monophyletic status in S17 probably derives from the higher weight of the morphological characters in the final matrix as compared to their small proportion of total character composition when UCE data are added.It is also noteworthy that Lophoglutus appeared as monophyletic in our current tree, as opposed to being paraphyletic with respect to Lorio in S17.Our tree also confirms that the Australian representatives of the widely distributed genera Anacis and Myrmeleonostenus actually represent separate Australasian lineages that need to be described as their own genera pending more detailed taxonomic work. 4.3.8| Clade G: Glodianus group   This small clade included five exclusively Neotropical genera in the analyses of S17: the three genera formerly assigned to Townes' Glodianina plus Polyphrix and Eknomia.In our study, a clade of the 'Glodianina' + Eknomia was once again recovered (albeit with relatively low support; BTP = 60, SH-aLRT = 94.8); in addition, Polyphrix was sister to a small subclade including three other Neotropical genera: Diapetimorpha, Debilos and Cryptoxenodon.The former two were included in S17 but not assigned to any particular genus group as their placement was unstable across the different analyses (either as their own clade or as part of the Lymeon group).Supeleto, Santos, Brady, and Aguiar (2020) described Cryptoxenodon and used Sanger sequencing and morphological data to show it was unambiguously the sister group to Debilos, and this three-genus clade appears as firmly placed in the Glodianus group in our analyses. 4.3.9| Clade H: Lymeon group   This large clade of mostly Neotropical species was recovered with the same generic composition as in S17.Within the group, one large subclade is composed of species that show the 'Dutilleul syndrome' characters to varying extent, with a tendency towards subspherical heads, modified antennal tips, inflated fore tibiae and stout ovipositors.Some of these taxa, such as Digonocryptus, Distictus and Fortipalpa, had been previously assigned to Townes' Gabuniina, which roughly corresponds to the 'Gabunia group' of S17.Supeleto, Santos, Brady, and Aguiar (2020) had found that the newly described Acrosnemus was part of this subgroup and a sister to Melanocryptus.Notably, our study shows that the large Neotropical genus Digonocryptus seems to be non-monophyletic, with D. pontagus and D. zatheos appearing outside of the main Digonocryptus clade.Indeed, D. pontagus had been only 'tentatively assigned' to Digonocryptus, and D. zatheos lacks some of the most distinct traits of the genus such as the teeth or tubercles on the clypeal margin.In addition, the limits between Digonocryptus and the largely Andean Cyclaulus need to be verified carefully in future studies and it may be that the two genera need to be synonymized.
As previously shown by the tree in S17, the synonymy of Biconus with Anacis performed by Porter (2004) is unjustified in phylogenetic grounds; since the former genus seems to be monophyletic, phylogenetically distant from Anacis and clearly diagnosable (see key in Townes, 1970), we hereby revalidate Biconus stat.rev.
Harpura is a monotypic genus that bears a strong similarity to the Neotropical genera of Townes' Mesostenina (e.g.Polycyrtus, Acorystus, Cryptanura): small and elongated ovipositor; polished and terete T1; and body overall shiny and polished.In addition, its distinctly downcurved ovipositor is similar to that of some Neotropical species of Mesostenus.Surprisingly, the genus was not recovered as closely related to any of these taxa but as a member of the Lymeon group, in the same clade of other morphologically unusual genera such as Latosculum, Lagarosoma and Prosthoporus.
The small and apparently rare Leptarthron, another new generic addition to the Cryptini tree, was recovered as sister to Golbachiella; both genera are generally small-bodied but have little similarity in more specific morphological characters.Similarly, the singular genus Hypsanacis was recovered as a somewhat isolated lineage within the Lymeon group, which is in accordance with the lack of particularly similar taxa among Cryptini.
Another addition to the Lymeon group was the small genus Strabotes; it was recovered as sister to the morphologically similar Mallochia; the two genera are keyed out together in the key of Townes (1970) and both have a generally subcylindric body, T1 short and stout and a small tooth on the clypeal margin.As in S17, our tree suggests the paraphyly of Baryceros with respect to Lamprocryptidea, but a more refined sampling of the two genera is needed to settle the taxonomic limits and identity of both genera.Likewise, Lymeon again appears as a polyphyletic taxon, but a proper assessment of this large Neotropical lineage will depend on much deeper taxon sampling. 4.3.10| Clade I: Osprynchotus group   This small clade includes five genera that are exclusively parasitoids of mud-nesting aculeate wasps, four of which are in Townes' former 'Osprynchotina', in addition to Neocryptopteryx and the newly added Picrocryptoides, which appear as sister groups.Hosts of Picrocryptoides are still unknown.Neocryptopteryx hypodineri is a parasitoid of Hypodynerus (Vespidae, Eumeninae; Porter, 1967a), though others of its species attack Lepidoptera (Blanchard, 1947).The taxonomic limits of Neocryptopteryx are still unclear, as Porter (2008) moved many species previously assigned to the genus back to Trachysphyrus without providing an explanation.Following the placement of T. riojanus in a distinct clade separate from the more typically Chilean Trachysphyrus and its similarity to taxa previously in Neocryptopteryx, we assign it to the latter genus following the taxonomic delimitation in Porter (1987).Otherwise, the topology of the Osprynchotus clade mirrors exactly that of S17. 4.3.11| Clade J: Trachysphyrus group Six new genera were added to this almost exclusively Chilean or Andean clade.Most of the species in this group have at some point been placed in the genus Trachysphyrus, later broken up by Porter (1967aPorter ( , 1967bPorter ( , 1987) ) into a profusion of mostly small genera.Caenopelte Porter and Dochmidium Porter seem to be nested within Anacis Porter, with this four-species group appearing as sister to all other members of the clade.The monotypic Nelophia was nested inside the larger genus Phycitiplex.Morphologically, the two genera are similar and are both unusual among members of the Trachysphyrus group in being found in southern South America east of the Andes.The position of this group, nested within a broader background of taxa that essentially occurs west of the Andes, suggests that this may represent a secondary colonization event of the main South American lowlands.
The placement of Periplasma within the Trachysphyrus group may appear surprising given its superficial similarity to Dotocryptus Brèthes (from the Osprynchotus group) in general aspect and sculpturing and particularly in sharing an extremely long ovipositor.However, Porter (1987) had already observed that Periplasma should be only distantly related to Dotocryptus, noticing substantial differences in ovipositor tip structure, propodeal shape and areolation, mandibular proportions and other traits.Hence, the similarity between both genera seems to be another example of striking morphological convergence within Cryptinae.
4.3.12| Clade K: Agrothereutes group No additional genera were added to this clade, which showed a similar topology to that found in S17, with Gyropyga nigra and Etha sp.forming a grade with respect to all other species in the clade, and the group of spiderparasitizing genera (Idiolispa, Hidryta and Trychosis) recovered as monophyletic.

F
I G U R E 2 Relationships among the outgroup taxa and Cryptini as recovered with the UCE matrix with 50% completeness partitioned by PartitionFinder.Numbers at nodes correspond to support values, respectively SH-aLRT (before the slash) and bootstrap (after the slash); values missing from nodes 100% clade support for that metric.F I G U R E 3 (Legend on next page) had relatively high SH-aLRT support but low BTP support: Tanepomidos assamensis (82 loci; SH-aLRT = 97.4,BTP = 35) was placed within a clade including other Oriental taxa; and Buysmania oxymora (393 loci; SH-aLRT = 92.7,BTP = 41) was nested within Euchalinus.

F
Relationships among species of Cryptini (clades A-D) recovered from maximum-likelihood analysis of the hybrid dataset of morphology, Sanger data and UCE loci, with the topology constrained to conform to the UCE-only tree.Inset box includes diversity metrics for each biogeographic region: # spp., total number of species included in the phylogeny; PD.obs, observed PD in community; PD.obs.z,standardized effect size of PD vs. null communities; p, p-value indicating whether standardized effect size values are significantly different from a null mode.F I G U R E 4 Relationships among species of Cryptini (clades F and K) recovered from maximum-likelihood analysis of the hybrid dataset of morphology, Sanger data and UCE loci, with the topology constrained to conform to the UCE-only tree.F I G U R E 5 Relationships among species of Cryptini (clade E) recovered from maximum-likelihood analysis of the hybrid dataset of morphology, Sanger data and UCE loci, with the topology constrained to conform to the UCE-only tree.F I G U R E 6 Relationships among species of Cryptini (clades G-J) recovered from maximum-likelihood analysis of the hybrid dataset of morphology, Sanger data and UCE loci, with the topology constrained to conform to the UCE-only tree.

F
Comparison of different 'three-taxon statements' with the relationships between Cryptinae, Ichneumoninae and Phygadeuontinae sensu stricto in three recent phylogenetic studies.Small subfamilies (Adelognathinae, Ateleutinae) and individual clades of the Phygadeuontinae sensu lato ignored for simplicity.F I G U R E 8 Comparison of relationships among Cryptini genus groups in this study versus the previous work of Santos (2017).