Individuals of the new species preserved in ethanol (70 and 100%) were cleared with Nesbitt’s solution and then mounted on glass slides in Hoyer’s medium following the procedures described by Jordana et al. (1997). Specimens in ethanol gel were photographed using a stereomicroscope (M165C) attached to a DFC420 digital camera with a dome, as presented in Kawada and Buffington (2016). The morphological study and raw drawings were made using a Leica DM750 phase-contrast microscope with a drawing tube attached. The mouthparts photograph was taken with a Leica MC170 HD camera attached to this microscope, stacked using LAS v. 4.12 software. Final drawings were vectorized and organized into plates using CorelDraw 2022 software.

The terminology used in the description follows mainly: prelabral and clypeal chaetotaxy after Yoshii and Suhardjono (1992); labral chaetotaxy after Cipola et al. (2014); labial papillae, maxillary palp and basolateral and basomedian labial fields after Fjellberg (1999), using the Gisin’s system (1964) for naming the chaetae rows; postlabial chaetotaxy after Chen and Christiansen (1993); subcoxae outer chaetotaxy after Yosii (1959); trochanteral organ lines of spines after Christiansen (1958a) and South (1961); unguiculus lamellae after Hüther (1986); manubrium ventral chaetotaxy after Yoshii (1982); and male’s genital plate after Christiansen (1958b). The head dorsal chaetotaxy follows Mari-Mutt (1979) and body after Szeptycki (1979), both adapted from Soto-Adames (2010), Cipola et al. (2018), Cipola and Viana (2023) and Zhang et al. (2019); and specialized chaetae (S-chaetae) after Zhang and Deharveng (2015). The macrochaetotaxy simplified formula for Lepidocyrtinae follows Gisin (1964) and Cipola and Viana (2023), modified to depict the inner + outer (lateral) macrochaetae of mesothorax to the fourth abdominal segment. The distribution pattern of pseudopores follows Mateos et al. (2021).

The abbreviations used in the description are: Abd―abdominal segment(s), Ant―antennal segment(s), mac―macrochaeta(e), mes―mesochaeta(e), mic―microchaeta(e), ms―specialized microchaeta(e), psp―pseudopore(s), sens―specialised ordinary chaeta(e), Th―thoracic segment(s); MTO––metatrochanteral organ; BP4―“basal” plate of fourth abdominal segment; for clypeal chaetotaxy: f―frontal, pf―prefrontal, l―lateral; for mouthparts: a.a.―apical appendage of labial papillae, b.c.―basal chaeta of maxillary palp, l.p.―lateral process of papilla E, lpc―labial proximal chaetae, s.b.―appendages of sublobal plate, t.a.―terminal appendage of the maxillary palp; for unguis: b.t.―paired basal teeth, m.t.―unpaired median tooth, a.t.―unpaired apical tooth; for unguiculus: ai―antero-internal lamella, ae―antero-external lamella, pi―postero-internal lamella, pe―postero-external lamella.

Symbols used to depict the chaetotaxy are presented in Fig. 1. Chaetae labels and other important taxonomic abbreviations are marked in bold in the text. Chaetae of uncertain homology are followed by a question mark (?). Chaetotaxy in figures and text is given by one side of the body only (left side in the figures), except for the clypeal and prelabral regions.

The studied material is deposited at the Collembola Collection of the Federal University of Rio Grande do Norte, Rio Grande do Norte, Brazil (CC/UFRN), and the Invertebrate Collection of the National Institute of Amazonian Research, Manaus, Brazil (INPA).

Specimens selected for molecular experiments were stored in absolute ethanol at –20 °C from the collection date until the DNA extraction. DNA was extracted from a single individual using the TIANamp MicroDNA extraction kit (Tiangen Co., Ltd, China). Libraries were constructed using KAPA Hyper Prep Kit (Roche). Shanghai Yao’en Biotechnology Co., Ltd, China, performed all laboratory procedures, including DNA extraction, amplification and library construction, following custom procedures. Whole-genome sequencing was performed by an Illumina NovaSeq platform, producing paired-end reads with 150 bp length. Approximately 57,333,204 reads or 8 Gb of data were generated.

Previous to the mitogenome assembly, the raw sequencing data was analyzed to remove sequencing adapters, reads with low quality and contaminants, and to correct potential errors. To perform the previous steps, we used BBTools (sourceforge.net/projects/bbmap), with the “clumpify.sh” and “bbduk.sh” pipelines. A total of 5Gb of cleaned paired-end data was then inputted into MitoZ v. 3.6 (Meng et al. 2019). The listed integrated tools were called to perform the following steps: MEGAHIT v. 1.2.9 (Li et al. 2015) for assembly; Tiara v. 1.0.1 (Karlicki et al. 2022) and HMMER v. 3.4 (Wheeler and Eddy 2013) for homology searches and sequence alignment. For annotation, we used BLAST+ (Gertz et al. 2006), GeneWise (Birney et al. 2004), Infernal v.1.1.5 (Nawrocki and Eddy 2013), and MiTFi v. 0.1 (Jühling et al. 2012). Visualization of the results of annotation and coverage graphic was done using Circos v. 0.69 (Krzywinski et al. 2009), BWA v. 0.7.17 (Li and Durbin 2009), and SAMtools v. 1.18 (Li et al. 2009). The mitogenome sequence and raw sequencing data will be available in NCBI (at https://www.ncbi.nlm.nih.gov) under the accession numbers PV872867 and SRR34434446, respectively, associated to bioproject number: PRJNA1125622.

To determine the phylogenetic position of the new species of Rhynchocyrtus within Lepidocyrtinae, we incorporated data from eight additional species of the subfamily from Lepidocyrtus and Pseudosinella, spanning representatives of five subgenera of the first genus. Other four species of Seirinae, broadly considered as the sister-group of Lepidocyrtinae, and four species of Entomobryinae (sister group of the cluster Seirinae + Lepidocyrtinae) served as outgroup taxa (Zhang et al. 2014; Zhang and Deharveng 2015; Zhang et al. 2019; Godeiro et al. 2023; Bellini et al. 2023). Complete mitogenomes were downloaded from NCBI (https://www.ncbi.nlm.nih.gov), accession numbers are listed on Table 1. Previous to the phylogenetic matrix generation, all 13 protein-coding genes (PCGs) from each species were organized into separate directories. Translations from nucleotide sequences to amino acids were carried out using TransDecoder v. 5.5.0 (http://transdecoder.github.io). Multiple sequence alignments were performed with MAFFT v. 7.470 (Katoh and Standley 2013) employing the “L-INS-i” algorithm, and alignment trimming was conducted using TrimAl v.1.4.1 (Capella-Gutiérrez et al. 2009) with the “-gappyout” setting. The final aligned sequences were concatenated into a single matrix of 3,358 amino acid sites (including 1,664 parsimony-informative sites) using FASconCAT-G v. 1.04 (Kück and Longo 2014). Phylogenetic inference via maximum likelihood (ML) as optimality criterion was performed with IQ-TREE v.2 (Minh et al. 2020) on a partitioned dataset. ModelFinder (Kalyaanamoorthy et al. 2017) was used to identify the most suitable substitution model for each gene partition, as detailed in Table S1. ML analyses included 1,000 replicates each of SH-aLRT (Guindon et al. 2010) and UFBoot2 (Hoang et al. 2018) to assess branch support. Bayesian inference (BI) was conducted with PhyloBayes-MPI v.1.8 (Lartillot et al. 2013) using the CAT + GTR site-heterogeneous mixture model. Two independent Markov Chain Monte Carlo (MCMC) runs were carried out and stopped once convergence was achieved (maxdiff < 0.1). The first 10% of sampled trees was discarded as burn-in, and the consensus tree was built from the remaining data. Final tree visualizations were generated using FigTree v. 1.4.2 (https://tree.bio.ed.ac.uk/software/figtree).

To calculate the genetic distances between the Lepidocyrtinae species, we first used the complete COI gene (1,539 bp), as it is a more informative marker. Even so, as a complementary approach, we also used the partial COI gene (658 bp), since this barcode marker has been widely applied in previous studies of Collembola, allowing our results to be used for cross-comparison with previous and future works. To construct the matrices, the nucleotide sequences of the complete and parcial COI of the nine species of Lepidocyrtinae listed in Table 1 were aligned in AliView v.1.28 (Larsson 2014) using MUSCLE v.3.8.31 (Edgar 2004), and the gaps were manually trimmed. The final alignment was used as input to calculate the pairwise distances in MEGA v.X (Kumar et al. 2018; Stecher et al. 2020) in both cases. Analyses were conducted using the Kimura 2-parameter model (Kimura 1980) and p-distance. Codon positions included were 1^st+2^nd+3^rd+Noncoding. All ambiguous positions were removed for each sequence pair (pairwise deletion option). A non-parametric bootstrap of 1000 replications (Felsenstein 1985) was used as the variance estimation method.

Family Entomobryidae Tömösvary, 1882

Subfamily LepidocyrtinaeWahlgren 1906sensu Zhang and Deharveng, 2015

The length of the newly obtained mitochondrial DNA of Rhynchocyrtus cleideaesp. nov. was 14,333 bp, but approximately 500 bp corresponding to the beginning of the ND5 gene were not recovered due to an assembling error in the origin of replication (Fig. 12; Table 3). The mitogenome contained the typeset of one control region with 263 bp and 37 genes, which included 13 PCGs, 22 tRNAs, and two rRNAs. Sequence features are given in Fig. 12 and Table 3. The nucleotide composition of the complete sequence was A (38.55%; 5,525) C (20.6%; 2,953) G (10.65%; 1,526) T (30.2%; 4,329). We infer that the gene order arrangement found in the mitogenome of Rhynchocyrtus cleideaesp. nov. was different from the Pancrustacean Ancestral Gene Order (AGO), but similar to the arrangement of two other Lepidocyrtus (Setogaster) species (Fig. 13). The gene tRNA-Thr (trnT) and the region between NAD6 and tRNA-Ser₂ (trnS₂) were translocated to an anterior position in the sequence, located before tRNA-Phe (trnF). Additionally, the gene tRNA-Pro (trnP) was translocated to a posterior position, located before NAD1. The other six mitogenomes of Lepidocyrtinae analyzed here presented the AGO.

The phylogenetic relationships of the sampled taxa, detailed in Table 1, are summarized in Fig. 14. The overall ML and BI node support values (SH-aLRT and bootstrap for ML/posterior probability for BI) was higher than 90/0.99 in most of the tree branches, with a few exceptions like the relationships between some sampled Entomobryinae and part of Lepidocyrtinae.

Regarding the subfamilies, our tree recovered the following topology: Entomobryinae + (Seirinae + Lepidocyrtinae), all with ML and BI absolute node support. Within the Lepidocyrtinae, two main branches were obtained: one clustering the subgenera Acrocyrtus Yosii, 1959, Cinctocyrtus and Lepidocyrtus sensu stricto, together with Pseudosinella, with variable node support values and divergent internal topology between ML and BI inferences; and the second one gathering Setogaster Salmon, 1951 and Lanocyrtus Yoshii and Suhardjono, 1989 taxa, together with Rhynchocyrtus cleideaesp. nov., where the BI node support was absolute to all internal branches, while ML support values were 99 or higher. The only subgenus of Lepidocyrtus with more than one sampled species in our analyses was Setogaster, represented by three Neotropical species, and it was recovered as a paraphyletic taxon (Fig. 14).

Table 4 presents the estimated evolutionary divergence between the complete COX1 sequences of the sampled species. The number of base substitutions per site using Kimura 2-parametrer model ranged from 0.171, between Lepidocyrtus (Setogaster) nigrosetosus and Lepidocyrtus (Lanocyrtus) fimetarius, and 0.316, between R. cleideaesp. nov. and L. (Lepidocyrtus) curvicollis, with an average of 0.266 among all the species. Similarly, based on p-distance values, the number of base differences ranged from 0.151, L. (S.) nigrosetosus x L. (La.) fimetarius to 0.255, Rhynchocyrtus cleideaesp. nov. x L. (L.) curvicollis, with an average of 0.222 across all sampled species. The obtained divergence values, together with the morphological aspects of the sampled taxa are enough to support them as independent species (as recently discussed in Rodrigues et al. 2024), but divergence values between sampled species of Setogaster were inconsistent when compared to R. cleideaesp. nov. and L. (La.) fimetarius. In this context, the Neotropical L. (S.) nigrosetosus exhibited remarkably low estimated evolutionary divergence from the Chinese sample of L. (La.) fimetarius, much lower than that observed with its closely related L. sotoi. Similarly, R. cleideaesp. nov. showed a lower estimated divergence with L. (S.) nigrosetosus than the latter did from an unidentified Brazilian species of the same subgenus, L. (S.) sp. (see Tables 1 and 4, and Fig. 14).

Complementing Table 4 comparison, Table 5 shows the estimated evolutionary divergence between parcial COX1 sequences of the same sampled Lepidocyrtinae. The results were somewhat similar to those obtained using the complete gene. The number of base substitutions per site using Kimura 2-parametrer model ranged from 0.185, between L. (La.) fimetarius and L. (S.) nigrosetosus; and 0.308, between L. (Le.) curvicollis and R. cleidaesp. nov. Alternatively, based on p-distance values, such differences ranged from 0.161, also between L. (La.) fimetarius and L. (S.) nigrosetosus; and 0.229, between L. (Le.) curvicollis and L. (S.) sp. As in the case of the complete COX1, the partial COX1 also supported a very close relationship between L. (S.) nigrosetosus and L. (La.) fimetarius, and between R. cleideaesp. nov. and L. (S.) nigrosetosus (see Table 5).

This study expands the known distribution of Rhynchocyrtus in the northeastern region of Brazil, adding new records for the states of Alagoas and Rio Grande do Norte. The genus is endemic to Brazil and is known to occur in the Atlantic Forest biome, particularly in coastal forest ecosystems (Mendonça and Fernandes 2007; Zeppelini et al. 2025). Its predominant presence in leaf litter microhabitats and on forest floor surfaces suggests an epedaphic or even atmobiotic lifestyle, as observed in representatives of Lepidocyrtus (Soto-Adames 2000; Mateos and Lukić 2019).

It is very likely that the distinctive morphology of the mouthparts in Rhynchocyrtus is associated with the selective consumption of resources found in forested microhabitats, which we could not identify at this time. Collembola can occupy different feeding guilds, exhibiting either generalist or specialized feeding habits (Hopkin 1997; Christiansen et al. 2009; Chahartaghi et al. 2005; Malcicka et al. 2017). The hypothesis of selective consumption of Rhynchocyrtus species is based on the observation of other Collembola, like the Neanuridae, which hold strong modifications on the mouthparts, resulting in distinct feeding habits (Hopkin 1997; Christiansen et al. 2009). A dedicated study on the diet of Rhynchocyrtus species is needed to test such a hypothesis.

Although at first our new species may look somewhat similar to its sole congener and type species, Rhynchocyrtus klausi, R. cleideaesp. nov. differs from the latter in many features, especially: body coloration pattern, with Th II laterally pigmented (Th II to Abd II in R. klausi), Abd III with a lateral spot (absent in R. klausi), and Abd II with a transversal band (lateral spot in R. klausi). On the head, the new species differs in A series with 3 mac (5 in R. klausi), maxilla with 1 sickle-shaped tooth and two lamellae, being the distal lamella apically rounded (2 teeth, 3 lamellae and posterior lamellar oval in R. klausi), and postlabial region and cephalic groove with some smooth chaetae (all ciliate in R. klausi). Rhynchocyrtus cleideaesp. nov. differs from the type species of the genus in dorsal chaetotaxy by: Abd II with m3mac (absent in R. klausi), Abd III with 2 lateral mac (3 R. klausi), and Abd IV with 12–13 lateral mac (15 in R. klausi). The new species also differs by unguis with paired m.t. and a.t. present (m.t. unpaired and a.t. absent in R. klausi), sublobal plate with 3 appendages (devoid in R. klausi), and manubrial plate with 4–5 chaetae (7–8 in R. klausi) (Mendonça and Fernandes 2007). See Table 2 for more comparisons between the species.

We could not compare the new species with R. klausi regarding some features like dorsal microchaetotaxy of head and trunk, entire postlabial and coxae chaetotaxy, complete pattern of body psp, sensilla number on Abd IV and posterior collophore, due to the absence of such information in the original description (Mendonça and Fernandes 2007). A future additional study of R. klausi, whose type material was lost in the devastating fire of the Museu Nacional of Federal University of Rio de Janeiro (MNRJ) in September 2018, may reveal further differences between the two species of the genus.

Our phylogenetic analyses support, with high node values, Rhynchocyrtus as an ingroup of Neotropical Setogaster,a subgenus of Lepidocyrtus, challenging both the status of the first as a generic lineage of Lepidocyrtinae, and the validity of the latter as a monophyletic subgenus. In fact, the Chinese population of Lepidocyrtus (Lanocyrtus) fimetarius was also recovered as part of the Setogaster clade, casting further doubts about the monophyly and diagnostic boundaries of these subgenera. Not only does our phylogenetic tree (Fig. 14) support such close relationships, but the mitochondrial gene order of Rhynchocyrtus also matches the pattern observed in the two sampled Setogaster species (Fig. 13). Furthermore, the evolutionary distances estimate between L. (S.) sotoi and L. (S.) nigrosetosus and R. cleideaesp. nov. show below-average divergence when compared to the full set of sampled species (Tables 4 and 5).

It is also worth noting that our data support that Rhynchocyrtus is not closely related to Lepidocyrtus (Cinctocyrtus), as previously suggested by Mendonça and Fernandes (2007), despite some similarities shared by both lineages like absence of antennal, legs and collophore scales, and presence of the dental basal tubercle (Mendonça and Fernandes 2007; Cipola et al. 2018).

The current systematics of Lepidocyrtinae is mostly based on morphological traits (like in Yoshii 1982; Yoshii and Suhardjono 1989, 1992; Cipola et al. 2018). Even so, several previous studies have highlighted the difficulties in distinguishing monophyletic lineages within the subfamily based on morphology, with some traditional taxa already recognized as para or polyphyletic genera (Christiansen 1960, 1961, 1988; Christiansen and Bellinger 1991; Soto-Adames 2000, 2002; Cipola et al. 2018; Mateos et al. 2018; Winkler et al. 2020; Godeiro et al. 2021; Kováč et al. 2023). A similar situation is likely to occur among the subgenera of Lepidocyrtus, due to overlapping or ambiguous diagnostic morphological characters that do not accurately reflect natural lineages (Christiansen and Bellinger 1991; Soto-Adames 2000; Cipola et al. 2018). For instance, our phylogeny pointed out that lineages lacking the dental tubercle, like Lepidocyrtus sensu stricto, Lanocyrtus and Pseudosinella, are not closely related (Fig. 14). Such an observation supports the idea that the dental tubercle was lost multiple times within Lepidocyrtinae, in line with the notes of Godeiro et al. (2021) on the internal evolution of the subfamily.

Although preliminary in the broader context of global Lepidocyrtinae extant diversity, our data support that certain morphological traits, such as the dental tubercle (observed in Setogaster and Rhynchocyrtus, but mostly absent in Lanocyrtus), the absence of scales on at least Ant. I and femora (shared by Lanocyrtus and Rhynchocyrtus, but variable in Setogaster), and the mucronal chaeta morphology (with spinelet in Setogaster, but without in Lanocyrtus and Rhynchocyrtus) (Yoshii 1982; Yoshii and Suhardjno 1989; Mendonça and Fernandes 2007; Winkler and Traser 2012; Cipola et al. 2018), are variable within the clade that includes these subgenera, and therefore carry limited, if any, phylogenetic signal. A more viable approach to delimiting valid internal divisions within Lepidocyrtus, and likely to other Lepidocyrtinae, is the establishment of species groups based on more detailed dorsal chaetotaxy, and/or on the distribution pattern of psp across the body and appendages. Recent studies like Mateos et al. (2018, 2021, 2023) and Winkler et al. (2020) showed some of these characters, in conjunct, present phylogenetic signals able to circumscribe natural groups within Lepidocyrtus. Indeed, the dorsal chaetotaxy in Entomobryoidea has proven to hold significant phylogenetic value in defining higher suprageneric taxa (Szeptycki 1979; Zhang and Deharveng 2015; Zhang et al. 2019; Godeiro et al. 2021, 2023), while recent studies suggest that the distribution of body and appendages psp may also carry important systematic information for certain genera within the superfamily (Deharveng et al. 2018; Mateos et al. 2021).

Even if our dataset is more representative of Lepidocyrtinae than those used in previous studies based on mitogenomes (e.g., Godeiro et al. 2021; Bellini et al. 2023), and our analyses do not support Rhynchocyrtus as a full genus of Lepidocyrtinae, we prefer to retain its current status at this time. Our sampled species are clearly insufficient to confidently test all subgenera of Lepidocyrtus, one of the largest and most widespread extant springtail genera (Bellinger et al. 1996–2024), or to resolve the paraphyly of Setogaster. Only the inclusion of additional species from diverse zoogeographical regions, as well as from all genera and subgenera within Lepidocyrtinae, will allow for a more accurate understanding of its internal relationships and a clearer delimitation of its generic and subgeneric boundaries.