Most firefly genera have poorly defined taxonomic boundaries, especially in the Neotropics, where they are more diverse and more difficult to identify. Recent advances that shed light on the diversity of fireflies in South America have focused mainly on Atlantic Rainforest taxa, whereas lampyrids in other biomes remained largely unstudied. We found three new firefly species endemic to the Amazon basin that share unique traits of the male abdomen where sternum VIII and the pygidium are modified and likely work as a copulation clamp. Here we test and confirm the hypothesis that these three species form a monophyletic lineage and propose Haplocauda gen. nov. to accommodate the three new species. Both Maximum Parsimony and probabilistic (Bayesian and Maximum Likelihood) phylogenetic analyses confirmed Haplocauda gen. nov. monophyly, and consistently recovered it as the sister group to Scissicauda, fireflies endemic to the Atlantic Rainforest that also feature a copulation clamp on abdominal segment VIII, although with a different configuration. We provide illustrations, diagnostic descriptions, and keys to species based on males and females. The three new species were sampled from different regions, and are likely allopatric, a common pattern among Amazonian taxa.

Morphology, terminology, and map

Specimens of these distinctive new species were located at two institutions: the Brazilian National Institute of Amazonian Research (INPA) and University of Georgia Collection of Arthropods (UGCA). The abdomen was removed from 2 male and 1–2 female specimens per species and soaked in 10% KOH for 24–36h. This clearing procedure was also applied to two entire specimens representing one of the three new species. The morphology was examined under a Leica M205 C stereomicroscope, and photographs were made with the Leica Application Suite X Auto-montage Software. We follow the classification of Martin et al. (2019), and the anatomical terminology of Silveira et al. (2016a). Distribution maps of the species were made using QGIS 3.10.14 (QGIS 2021). We recorded label data for all type specimens using the following conventions: double quotes (“ ”) for label data quoted verbatim; double forward slashes (// ) separate labels; double comma (,,) for line breaks; brackets [ ] enclose our comments or notes. All labels are typed, unless otherwise noted.

Phylogenetic analyses

We ran phylogenetic analyses with two goals: (i) test the hypothesis that the three new species bearing the distinctive abdominal morphology constituted a monophyletic group; and, if so, (ii) explore the placement of that group within Lampyridae. Of particular interest were the phylogenetic relationships of the new species with Lampyrinae Rafinesque, 1815, a subfamily that includes several taxa sharing similar anatomical features with the new species (see below).

Lampyrid classification remains unstable despite recent advances (e.g., Martin et al., 2019) and relationships among genera, as well as their monophyly, remain largely unexplored outside Luciolinae (Ballantyne et al., 2019; but see Vaz et al., 2020; Vaz et al., 2021; and Silveira et al., 2021). In order to determine the placement of these new species in Lampyridae, we included in our taxon sampling representatives of the four subfamilies that are known to occur in South America, namely: Psilocladinae (Psilocladus miltoderus Blanchard, 1846), Amydetinae (Amydetes fastigiata Illiger, 1807), Cladodinae (Cladodes flabellatus Solier, 1849; Bocakova et al., in press), and Lampyrinae, in addition to taxa incertae sedis (Araucariocladus hiems Silveira & Mermudes, 2017, and Vesta thoracica Olivier, 1790). We included a denser sampling of Lampyrinae because the new species shared many similarities with the lampyrine Costalampys Silveira, Roza, Vaz & Mermudes, 2021 and Scissicauda McDermott, 1964, which was transferred to Lampyrinae in Martin et al. (2019). A summary of the classification adopted here is given in Table 1. Data for the material examined of the new species is given below (see Results), and additional material used in the phylogenetic analyses is provided in Supp. Mat. 1. 

Using MESQUITE (Maddison & Maddison, 2017), we built a matrix with 20 taxa scored for 76 characters following the guidelines of Sereno (2007). The data include 55 characters reanalyzed from Silveira et al. (2021) combined with 21 new characters (see Results; Supp. Mat. 2). Measurement-based characters are taken at the longest or the widest point ​​of the respective structure. Key character states are labeled in figures, abbreviated as C:S, where C and S indicates character and state number, respectively, followed by the numbers of interest.

In order to compare the outcomes of multiple phylogenetic approaches, we performed Maximum Parsimony (MP) analyses in TNT (Goloboff et al., 2016), Bayesian inference (BI) in MrBayes 3.2.7a (Huelsenbeck et al., 2001; Ronquist et al., 2012) (available at The CIPRES Science Gateway V. 3.3 [phylo.org] – Miller et al., 2010), and Maximum Likelihood inference in IQTREE 1.6.12 (Minh et al., 2020).

We ran the MP analyses using New Technology heuristic searches with Tree Bisection and Reconnection, with equal and implied weights (EW and IW, respectively), scaling the k parameter from 0.5–50 to investigate how homoplasy impacts tree topology (Goloboff, 2008; Mirande, 2009). Node support was assessed using the Bremer Index and standard bootstrapping (for MP–EW), and symmetric resampling (for MP–IW) with 1000 replicates. Character evolution was optimized using WINCLADA (Nixon, 2002). 

The BI analysis implemented the MKV model, modified from MK to account for ascertainment bias due to the prior exclusion of invariant characters in our matrix (Lewis, 2001), in addition to among-character rate variation with 8 rate categories for tractability (Wright, 2019). We ran 50,000,000 generations, saving trees each 2000 generations, and discarding the first 25% as burn-in. We checked for convergence using Tracer v1.6 (Rambaut et al., 2014). We compared the fit and the topologies of the BIMKV with gamma and lognormal rate distribution, since each was found to provide dataset-specific improvements to evolutionary models by Harrison & Larsson (2015). For the ML, we obtained estimates of branch support with Ultrafast bootstrapping with 1000 replicates, in IQTREE. To compare the three main topologies obtained (i.e. MLG [shared with MPEQ], BIG, and BILN), we performed pairwise comparisons in IQTREE, and applied the SH (Shimodaira and Hasegawa, 1999) and AU (Shimodaira, 2002) statistical tests.

Trees were read in FigTree version 1.4.4 (obtained at https://github.com/rambaut/figtree/releases), and the different measures of node support obtained (Bremer index, Bootstrap, Ultrafast Bootstrap, and posterior probabilities) were placed on the preferred topology (see below) using Adobe Photoshop 2021. 

Captions:

Haplocauda_BI_Mk_Gamma.con.tre – Consensus Bayesian tree file with node probabilities, MKV model with ascertainment bias, with gamma distribution (eight categories).

Haplocauda_BI_Mk_lnorm_ncat8.con.tre – Consensus Bayesian tree file with node probabilities, MKV model with ascertainment bias, with lognormal distribution (eight categories).

Haplocauda_BI_Mk.con.tre – Consensus Bayesian tree file with node probabilities, MKV model with ascertainment bias, with fixed rates.

Haplocauda_ML_MK_G.treefile – Maximum Likelihood tree file with three measurements of node support (aLRT, Bayesian aLRT, UFBOOT2), MKV model with ascertainment bias, with gamma distribution (eight categories).

Haplocauda_ML_MK.treefile – Maximum Likelihood tree file with three measurements of node support (aLRT, Bayesian aLRT, UFBOOT2), MKV model with ascertainment bias, with fixed rates.

Haplocauda_Parsimony_equal_weight.nex – Parsimony consensus tree after equal weights, with Bremer support and bootstrap values.

Haplocauda_Parsimony_implied_weights_1.nex – Parsimony consensus tree after implied weighing (k=1), with symmetric resampling values.

Haplocauda_Supp_Mat_2_SimpNEX_Dec_2021.nex – Morphological data matrix as simplified nexus