1 Mitochondrial genomes resolve the phylogeny 2 of Adephaga (Coleoptera) and confirm tiger 3 beetles (Cicindelidae) as an independent family 4 Alejandro López-López1,2,3 and Alfried P. Vogler1,2 5 1: Department of Life Sciences, Natural History Museum, London SW7 5BD, UK 6 2: Department of Life Sciences, Silwood Park Campus, Imperial College London, Ascot SL5 7PY, UK 7 3: Departamento de Zoología y Antropología Física, Facultad de Veterinaria, Universidad de Murcia, Campus 8 Mare Nostrum, 30100, Murcia, Spain 9 10 Corresponding author: Alejandro López-López ([email protected]) 11 12 Abstract 13 The beetle suborder Adephaga consists of several aquatic (‘Hydradephaga’) and terrestrial 14 (‘Geadephaga’) families whose relationships remain poorly known. In particular, the position 15 of Cicindelidae (tiger beetles) appears problematic, as recent studies have found them either 16 within the Hydradephaga based on mitogenomes, or together with several unlikely relatives 17 in Geadeadephaga based on 18S rRNA genes. We newly sequenced nine mitogenomes of 18 representatives of Cicindelidae and three ground beetles (Carabidae), and conducted 19 phylogenetic analyses together with 29 existing mitogenomes of Adephaga. Our results 20 support a basal split of Geadephaga and Hydradephaga, and reveal Cicindelidae, together 21 with Trachypachidae, as sister to all other Geadephaga, supporting their status as Family. We 22 show that alternative arrangements of basal adephagan relationships coincide with increased 23 rates of evolutionary change and with nucleotide compositional bias, but these confounding 24 factors were overcome by the CAT-Poisson model of PhyloBayes. The mitogenome + 18S 25 rRNA combined matrix supports the same topology only after removal of the hypervariable 26 expansion segments. Dense taxon sampling of mitogenomes, analyzed with site 27 heterogeneous mixture models, produce well-supported trees of the Adephaga. Mitochondrial 28 genomes are an increasingly valuable tool for resolving deep phylogenetic relationships, 29 outperforming the previously most widely used ribosomal RNA markers. 30 31 32 Keywords 33 Coleoptera; Adephaga; Cicindelidae; phylogeny; mitochondrial genomes 34 35 1 36 Introduction 37 Mitochondrial genome sequences are now widely available, and with greater taxon sampling 38 they provide an increasingly complete picture of deep relationships among and within the 39 insect orders (Li et al., 2015; Simon and Hadrys, 2013; Song et al., 2016; Timmermans et al., 40 2016). However, the complexities of rate heterogeneity and compositional heterogeneity that 41 are prevalent in mitochondrial sequence evolution confound phylogenetic inferences from 42 these data (Pons et al., 2010; Song et al., 2010; Talavera and Vila, 2011). Several recent 43 studies have shown that long-branch attraction is partly overcome through the use of 44 Bayesian mixture models implemented in the PhyloBayes that allow multiple independent 45 substitution processes with their own rate estimates and equilibrium frequencies, and whose 46 optimal number is estimated from the data (Lartillot et al., 2007; Lartillot et al., 2009). The 47 use of this software has resolved relationships that could not be estimated with other 48 likelihood and Bayesian models (Song et al., 2016; Talavera and Vila, 2011). For example, in 49 the Coleoptera (beetles), the mitochondrial sequence of Tetraphalerus, a member of the 50 suborder Archostemata, was widely observed in a spurious position within the Polyphaga, but 51 it was correctly placed outside of the other coleopteran orders in trees generated with 52 PhyloBayes (Timmermans et al., 2016; Timmermans et al., 2010). 53 54 Despite these advances in data generation and phylogenetic methodology, on various 55 occasions mitochondrial genomes have failed to establish relationships correctly. In the 56 current study we address the problem of basal relationships in the coleopteran suborder 57 Adephaga, which were recently analyzed as part of a much larger study of coleopteran 58 phylogeny based on some 250 mitogenomes (Timmermans et al., 2016). That study was 59 successful in resolving the relationships of the major superfamilies of Polyphaga (the 60 suborder containing some 90% of all described beetle species), but relationships in Adephaga 61 remained implausible and specifically failed to recover the widely established sister 62 relationship of the terrestrial Geadephaga and aquatic Hydradephaga. This was a surprising 63 finding given the evidence from 18S rRNA data (Maddison et al., 1999; Shull et al., 2001) 64 and a combination of rRNA genes and single-copy nuclear markers (McKenna et al., 2015), 65 but was in accordance with earlier morphological studies that consider the terrestrial groups, 66 including the large family Carabidae (ground beetles), to be derived from within the aquatic 67 lineages. These topologies would imply either a secondary loss of aquatic lifestyle leading to 68 Geadephaga or the independent origins of various aquatic lineages, as proposed previously 69 (Beutel, 1993; Kavanaugh, 1986). The alternative arrangement of a Geadephaga that is 2 70 paraphyletic for Hydradephaga has also been proposed, suggesting a single origin of aquatic 71 life style from a terrestrial ancestor near the base of the Carabidae. This scenario is supported 72 by some characters in Trachypachidae, a lineage presumed to represent a basal split within 73 the Geadephaga that, although entirely terrestrial, exhibits some characteristics that have been 74 interpreted to show similarities to those of the extant aquatic lineages (Bell, 1966; Crowson, 75 1960; Hammond, 1979). 76 77 The question about the relationships among terrestrial and aquatic groups is further 78 complicated by the poor understanding of the basal branches in Geadephaga, in particular 79 with regard to the placement of the small families Cicindelidae, Rhysodidae and Paussidae, 80 which are morphologically divergent from carabids and have been variously considered as a 81 sublineage within Carabidae or as closely related lineages outside of the Carabidae + 82 Trachypachidae clade (Bell, 1994; Beutel, 1992). The uncertainty about their phylogenetic 83 position has also affected the taxonomic status of these lineages, in particular of the 84 Cicindelidae (tiger beetles), which in the literature have been considered as separate family or 85 as subfamily of Carabidae, without reaching a consensus (Bils, 1976; Crowson, 1981; 86 Regenfuss, 1975). They are usually awarded family status due to their highly divergent 87 morphology and adaptations of the larvae, which are sit-and-wait predators hunting from 88 deep burrows in the ground (Cassola, 2001; Pearson and Vogler, 2001). It is still disputed 89 whether these characteristics are derived from carabids, or if they point to an independent 90 origin of the cicindelids leading to their peculiar life style. 91 92 Mitochondrial genomes have produced to date a confusing picture of cicindelid relationships 93 with other taxa, as the only available sequence (Habrodera) was placed as sister taxon to the 94 representative of the aquatic family Dytiscidae (Timmermans et al., 2016), i.e. in a highly 95 improbable position distant from the Carabidae. Equally, the relationships of the aquatic 96 families were highly unexpected, because the burrowing Noteridae were the sister to all other 97 Adephaga, although they are usually considered related to the Dytiscoidea (Dytiscidae, 98 Amphizoidae, Hygrobiidae and Aspidytidae), a monophyletic lineage that includes the great 99 majority of all hydradephagan species. The remaining aquatic lineages including the algal- 100 feeding Haliplidae and surface-hunting Gyrinidae were arranged in a comb-like fashion near 101 the base of the Adephaga (Timmermans et al., 2016). These findings are clearly in 102 contradiction to the topologies from nuclear genes (Maddison et al., 1999; McKenna et al., 103 2015; Shull et al., 2001), and mitochondrial loci may drive trees in combined analyses with 3 104 nuclear genes (Bocak et al., 2014). Yet, the rRNA loci, which remain the most widely 105 available nuclear markers, have themselves resulted in unexpected relationships, in particular 106 in regard to the position of Cicindelidae in a derived lineage of carabids, together with 107 Rhysodinae, Paussinae and Scaritini, to form the so-called CRPS quartet (Maddison et al., 108 1998, 1999). This locus thus groups some of the morphologically and ecologically most 109 derived geadephagan lineages and places them near the species-rich and evolutionarily recent 110 Harpalinae, which was unexpected from morphological analyses, but detailed data 111 exploration could not attribute the CRPS quartet to the result of long-branch attractions 112 (Maddison et al., 1999). 113 114 The current study attempts to resolve the question about basal relationships in Adephaga and 115 to reconcile the differences in topology obtained with mitochondrial and nuclear markers. We 116 address specifically the key discrepancies between the results obtained with both markers 117 regarding the Geadephaga - Hydradephaga split, which is supported by rRNA (Maddison et 118 al., 1999; Shull et al., 2001) and single-copy nuclear genes (McKenna et al., 2015) only, and 119 the position of Cicindelidae, which is placed in a derived position within Carabidae based on 120 rRNA and outside of Carabidae by mitogenomes. Using eight additional mitogenomes for a 121 more complete sample of cicindelids, supplemented with additional carabids including