Although molecular data have proven indispensable in confidently resolving the phylogeny of many clades across the tree of life, these data may be inaccessible for certain taxa. The resolution of taxonomy in the ant subfamily Leptanillinae is made problematic by the absence of DNA sequence data for leptanilline taxa that are known only from male specimens, including the monotypic genus Phaulomyrma Wheeler & Wheeler. Focusing upon the considerable diversity of undescribed male leptanilline morphospecies, the phylogeny of 34 putative morphospecies sampled from across the Leptanillinae, plus an outgroup, is inferred from 11 nuclear loci and 41 discrete male morphological characters using a Bayesian total-evidence framework, with Phaulomyrma represented by morphological data only. Based upon the results of this analysis Phaulomyrma is synonymized with Leptanilla Emery, and male-based diagnoses for Leptanilla that are grounded in phylogeny are provided, under both broad and narrow circumscriptions of that genus. This demonstrates the potential utility of a total-evidence approach in inferring the phylogeny of rare extant taxa for which molecular data are unavailable and begins a long-overdue systematic revision of the Leptanillinae that is focused on male material.

Part of this dataset was derived from phylogenomic data obtained with the hym-v2 ultra-conserved element probe set of Branstetter et al. (2017), with exons for 11 loci from Leptanilla GR02 (ANT-exon-sequences-40-taxa-reduced.fasta) being used analogously to UCE probes in a modified phyluce workflow (Faircloth 2016). For loci and taxa for which UCE enrichment was unsuccessful, the needed data were derived from a version of the 7,451-bp alignment of Borowiec et al. (2019), obtained via Sanger sequencing, that contains autapomorphic indels and introns (11,090 bp). The concatenated alignment (9,351 bp) was then partitioned according to intronic/exonic nature and codon position, as was a 9,062-bp legacy-locus alignment was created that includes only those data published prior to this study (Ward & Sumnicht 2012; Borowiec et al. 2019), with 20 terminals. These partitions were either used as an ad hoc partitioning scheme or input into PartitionFinder2 as character subsets.