Assessed specimens were mostly obtained from Colección de Mamíferos, Universidad Austral de Chile, Valdivia, Chile (UACH) and secondarily from the following collections: Fundación M. Lillo- Facultad de Ciencias Naturales e Instituto Miguel Lillo, Argentina (CML), Instituto Argentino de Investigaciones de las Zonas Áridas-CONICET, Argentina (IADIZA), Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”, Argentina (MACN), Museo Nacional de Historia Natural de Chile, Santiago, Chile (MNHN), Museo de Historia Natural de la Universidad Nacional de San Agustín, Arequipa, Peru (MUSA), Museo de Historia Natural de la Universidad Nacional Mayor de San Marcos, Lima, Peru (MUSM), and Museum of Southwestern Biology (MSB). No specimen was collected during this study. Additionally, several mitochondrial DNA sequences and some UCE data were retrieved from public repositories (see below). The list of specimens used in the molecular and/or morphologic based analyses, along with data on their geographic provenance, is presented in Tables S1, S2.

Of relevance for taxonomic purposes, our sampling includes specimens from the type locality, or adjacent areas (up to ca. 50km), of several nominal forms historically associated with Abrothrix andina, including Mus andinus (see below to where this locality is here restricted), Akodon andinus polius (Salinas, Arequipa, Peru), Akodon gossei (Puente del Inca, Mendoza, Argentina), and Hesperomys dolichonyx (San Pedro de Atacama, Antofagasta, Chile), as well as of Abrothrix olivacea, including Mus olivaceus (Valparaíso, Chile) and Abrothrix olivacea tarapacensis (Quebrada Tarapacá, Tarapacá, Chile). In addition, our sampling includes a specimen, currently housed at MNHN, likely used in the original description of Hesperomys dolichonyx Philippi, 1896, herein designated as lectotype (see below).

This study is based on molecular and morphological evidence. Two molecular datasets were analyzed. The first dataset consists of 1839 ultraconserved element loci obtained from 26 individuals. Of these, 16 correspond to specimens of subgenus Angelomys, seven to other nominal forms of the genus Abrothrix, and three to additional representatives of Abrotrichini, which were used to root the tree. Data for eight specimens were obtained from Parada et al. (2021), whereas sequences for the remaining 18 specimens were generated in this study as described below. Specimens of Angelomys were selected following the mitochondrial cytochrome b (cyt b) genealogy of Quiroga-Carmona et al. (2022) to ensure adequate representation of the main mitochondrial lineages of Ab. olivacea. This sampling was complemented with specimens of Ab. andina from the highlands near Santiago and the Andean highlands of Chile and Argentina. Detailed specimen information is provided in Tables S1, S2, as well as information per UCE locus in Table S3.

The second molecular dataset comprises a large alignment corresponding to the first 801 base pairs of the cyt b gene, including specimens spanning most of the distribution of the subgenus Angelomys, with emphasis on populations from its central and northern range. This dataset includes sequences from 346 specimens of Angelomys collected at 129 localities. Of these, 203 new sequences from Argentine, Chilean, and Peruvian specimens were generated in this study. Sampling includes representatives of the main lineages identified within Ab. olivacea in the geographically dense study of Quiroga-Carmona et al. (2022), particularly from Patagonian populations, as well as sequences of Ab. andina made available by Tammone et al. (2024, 2025). Additionally, the dataset includes 35 sequences representing all other species of the tribe Abrotrichini, along with sequences of Phyllotis darwini, P. xanthopygus, and Calomys venustus, which were used as outgroups. Detailed specimen information for Angelomys is provided in Table SS1, whereas information for the remaining Abrotrichini taxa and the outgroup taxa is provided in Table S4.

Finally, morphological analyses were restricted to specimens included in the molecular datasets, comprising 170 individuals collected at 129 localities and spanning most of the geographic range of the subgenus. Comparative assessments were guided by the phylogenetic relationships inferred from the UCE dataset and by the genealogical structure recovered from cyt b.

Genomic DNA was extracted with the Qiagen DNeasy Tissue and Blood Kit and quantified using the Qubit BR dsDNA kit (Thermo Fisher). Libraries were enriched for UCE loci using the UCE-5Kv1 probe set (Mycroarray), and sequencing was performed with Illumina Technology considering 2x250 paired-end sequencing at Rapidgenomics service.

UCEs were processed with the python package PHYLUCE 1.5.0 (Faircloth 2016). Clean reads were assembled with SPAdes genome assembler v3.14.1 (Prjibelski et al. 2020) prior to extracting the contigs matching UCEs. Each locus was aligned with MAFFT (Katoh et al. 2019). All alignments recovered in our procedure contained sequences from all the individuals considered. Raw sequence reads are available at EMBL (study PRJEB43612; Parada et al. 2021), for newly generated data, all new sequences are deposited in the European Nucleotide Archive (ENA) in study project PRJEB108495 (accession codes are listed in Table SS2).

UCEs were analyzed under two different approaches: As a concatenated matrix including all loci and under a coalescent approach with each locus first analyzed separately. The concatenated matrix was analyzed under Maximum Likelihood (ML) considering the best partition scheme selected by ModelFinder (Kalyaanamoorthy et al. 2017) with the option “-rcluster 1 -MF+MERGE”. This strategy starts with the full partition model and tries merging two loci until the model fit cannot be increased. For each partition, the best substitution model was selected using the Bayesian Information Criterion (BIC) according to ModelFinder (Table S3). Initial partitions and selected best partition scheme and substitution models are provided in Tables S5, S6. The ML inference was done with IQ-TREE version 2.2.6 (Minh et al. 2020). Ultrafast Bootstrap (BS; Minh et al. 2013; Hoang et al. 2018) was used as a measure of branch support using 5000 replicates. The coalescent-based analysis was performed with ASTRAL-III v5.7.8 (Zhang et al. 2018), which constructs a statistically consistent species tree by maximizing the quartet support found within the input set of gene trees. Nodal support in the species tree was summarized as quartet scores (QS). Input gene trees, for each locus, were reconstructed via ML with IQ-TREE using the substitution models selected above and under default parameters.

Sequences of the cyt b gene gathered here were generated from DNA isolated from tissue (muscle or liver) samples preserved in ethanol 96°C using PROMEGA DNA Kit (Wizard SV Genomic DNA Purification System) and following the manufacture Instructions. Sequences were generated in one fragment of 801 bp using primers MVZ05 and MVZ16 (Smith and Patton 1993). The thermal profile for DNA amplification was: 1) initial denaturation at 94 °C for 5min, 2) 35 cycles of denaturation at 94 °C for 45sec, primer annealing at 49 °C for 30sec, and extension at 72° for 1.5min, and 3) a final extension at 72 °C for 10min. Amplicons were sent to Macrogen Inc for sequencing. Sequences were edited using CodonCode Aligner v7.1.2 and submitted to GenBank (Table SS1).

Sequence alignment was performed in the online version of Mafft v7 (https://mafft.cbrc.jp/alignment/server/index.html, Katoh et al. 2019) considering default parameter values; the resulting alignment was visually revised in MEGA v11 (Tamura et al. 2021) in search of internal codon stops and changes in the reading frame. A non-redundant matrix (i.e., one sequence per haplotypic class) was created using the online version of the program ALTER (https://www.sing-group.org/ALTER/, Glez-Peña et al. 2010). This non-redundant matrix, consisting of 233 sequences, was used to infer the cyt b genealogy using ML and Bayesian inference (BI). The nucleotide substitution model, TPM2+F+I+G4, was selected with the Bayesian information criterion using ModelFinder as implemented in IQ-TREE v2.0.7.

The ML analysis was carried out in IQ-TREE settings for perturbation strength (--perturb 200) and stopping rule (--nstop 0.5). Branch support was assessed using 1000 replicates of Ultrafast Bootstrap approximation and 1000 replicates of the SH-like approximate likelihood ratio test (SH-aLRT; Guindon et al. 2010). For the BI analysis we used MrBayes v3.2.7a (Ronquist et al. 2012). All model parameters, including the substitution matrix, base frequencies, proportion of invariant sites, and gamma shape parameter, were fixed to the maximum-likelihood estimates obtained from the best-fitting model selected in IQ-TREE. Two independent runs of four Markov chains each were executed for 20 million generations, sampling every 2000 generations. Convergence and stationarity were assessed by confirming that the standard deviation of split frequencies ( < 0.01) and the potential scale reduction factor (PSRF = 1.00) met the thresholds recommended in the MrBayes manual (Ronquist et al. 2020), and by comparing the independent runs in Tracer v.1.7.2 (Rambaut et al. 2018). The first 2500 samples were discarded as burn-in for both the parameter (sump) and topology (sumt) summaries.

To further explore relationships among closely related mitochondrial haplotypes and visualize geographic patterns of variation, haplotype networks were constructed in PopART using the median-joining network method (Bandelt et al. 1999). Networks were generated separately for each major lineage, with samples assigned to their corresponding cyt b phylogroup.

Uncorrected genetic distances, p distance, within and between cyt b phylogroups were estimated in MEGA v11.

Morphological comparisons were guided by the UCE-based phylogeny and the genealogical structure recovered with the cyt b gene. Additionally, to assess the distinction of Abrothrix andina from the highlands of Santiago with respect to other populations of Angelomys from coastal and mid-elevation regions, geographic groups were also considered.

Specimens were sorted by age-classes. Age 1: M3 not at level of alveolus. Age 2: M3 above alveolus but not at occlusal level of the other molars. Age 3: M3 fully erupted but unworn and showing a well-developed protocone. Age 4: slight wear on maxillary teeth but accessory styles on M1 and M2 still evident; M3, the protocone starts getting worn. Age 5: Moderate wear on maxillary, M1 and M2 with flexus still evident, M3 flat without flexus. Age 6: much wear on maxillary teeth, boundaries between major cones obliterated, and becoming concave. Only adult specimens, age-classes 3–6, were included in the quantitative and qualitative analyses (n = 170 analyzed specimens; age 3 = 27, age 4 = 70, age 5 = 41, age 6 = 33).

Qualitative comparisons included mostly external features (dorso-ventral coloration pattern, hind- and forefeet morphology, and tail features) and some craniodental traits. Features were selected considering the original descriptions provided by Philippi and Landbeck (1858) for Ab. andina, Thomas (1920) for Ak. gossei, Thomas (1913) for Ak. jucundus, Philippi (1896) for Hesperomys dolichonyx and H. d. cinnamomea, as well as others like Patterson et al. (2015), Teta et al. (2017), Teta and Jayat (2022), and Pacheco et al. (2024). Qualitative comparisons focused primarily on evaluating the distinction of the specimens from the highlands of Santiago, the general area of the type locality of Ab. andina, in relation to other central-northern populations of Angelomys.

Standard external measurements (total length, ToL; length of tail, LT; hind foot length, HFL; ear length, EL) and weight (W) were taken from skin labels and or field catalogues. Head and body length (HBL) was also estimated as ToL – LT. The following 20 craniodental measurements were taken with a digital caliper and rounded to the nearest 0.01 mm, according to the definitions provided in Patterson (1992) and Teta and Pardiñas (2014): Skull length (SL); Condyle-incisive length (CIL); Palatilar length (PL); Diastema length (DL); Incisive length (IL); Toothrow length (TrL); Nasal length (NL); Frontal length (FL); Parietal length (PrL); Incisive width (IW); Palatal width at M1 (M1-M1); Palatal width at M3 (M3-M3); Wide of mesopterygoid fossa (WFM); Zygomatic plate width (ZPW); Nasal width (NW); Rostrum width (RW); Frontal sinus width (FSW); Interorbital breadth (IOB); Zygomatic breadth (ZB); and Braincase breadth (BB).

The morphometric analyses were based only on the craniodental measurements, which were log₁₀-transformed prior to conducting transformation to linearize allometric relationships (Huxley 1932) and to improve approximation to multivariate normality (Jolicoeur 1963). Missing values represented a small fraction of the dataset (1.34%) and were imputed using the iterative principal component method implemented in the imputePCA function of the missMDA R package (Josse and Husson 2016). Inspection of the missing value heatmap indicated that missing data were sparsely distributed across individuals and variables, with no systematic pattern of loss (Fig. S1). The multivariate normality was tested using MVN R package (Korkmaz et al. 2014) considering Mardia statistic, which shown that our dataset does not approximate a multivariate normality (skewness = 2182.88, p = 3.94; kurtosis = 3.36, p < 0.001). Therefore, a nonparametric test, the permutational multivariate analysis of variance (PERMANOVA), was used considering Euclidean distances, 9999 iterations, and Bonferroni adjustment to test the significance of the differences among groups. Also, the beta dispersion for our dataset was tested to avoid misinterpretation of the PERMANOVA results and it was found statistically not significant (p > 0.05). These analyses were done using vegan R package (Oksanen et al. 2025). A principal component analysis (PCA) was performed to visualize distribution of the specimens of Angelomys in the morphometric space using FactoMineR R package (Lê et al. 2008). A discriminant analysis of principal components (DAPC) was used to test the classification of the specimens. For DAPC, we used the adegenet R package (Jombart 2008; Jombart and Ahmed 2011). All analyses were performed in R v. 4.2.3 (R Core Team 2024).

Phylogenetic analyses of UCE loci inferred using ML and ASTRAL recovered three well-supported lineages within the subgenus Angelomys (BS = 100; QS = 1.00; Fig. 2A). However, these three lineages do not correspond to the traditionally delimited species of the subgenus, Abrothrix andina and Ab. olivacea due to both are not recovered monophyletic (Fig. 2A, sky blue branches denote specimens traditionally assigned to Ab. andina and olive branches to Ab. olivacea). ML and ASTRAL approaches yielded largely congruent topologies, differing only in the relationships among subclades within one lineage (Fig. 2A).

Santiago-Coastal lineage (SCL). This lineage comprises specimens collected in the highlands of Santiago (the area originally consigned as the type locality of Mus andinus) in central Chile, as well as, from northern coastal areas, and mid-elevation regions of northern and central Chile, and southern Peru. All specimens composing this lineage form a strongly supported monophyletic group (BS = 100; QS = 1.00; Fig. 2A), which is further divided into two well-supported subclades. The first includes the two analyzed specimens from the highlands of Santiago (localities 69 and 71), and the two analyzed specimens from the coasts of southern Antofagasta and Coquimbo Regions (localities 32 and 47). The second subclade includes specimens from the northern Chilean coast, Antofagasta (localities 17 and 19), Arica y Parinacota (locality 11), and Tarapacá Regions (locality 14). Within this second subclade, the ML tree recovers specimens from Antofagasta as sister to specimens from Arica y Parinacota and Tarapacá (BS = 100), whereas the ASTRAL species tree recovers the specimen from Arica y Parinacota as sister to a clade conformed by specimens from Tarapacá and Antofagasta (QS = 1.00); the alternative topology inferred by ASTRAL is shown in the inset box of Figure 2A.

High Andean lineage (HAL). This highly supported monophyletic (BS = 100; QS = 1.00; Fig. 2A) UCE lineage includes specimens collected in the highland regions of northern Chile and central and northwestern Argentina. HAL is further divided into two well-supported subclades. The first one (BS = 100; QS = 1.00) comprises two specimens collected in the high Andes region of Chile in Arica y Parinacota and Antofagasta regions (localities 9 and 25). The second subclade (BS = 100; QS = 1.00) includes three specimens, one collected in the highlands of Atacama region of Chile (locality 37) and two in the highlands of La Rioja and Mendoza provinces of Argentina (localities 42 and 64).

Southern lineage (SL). This lineage is also strongly supported (BS = 100; QS = 1.00; Fig. 2A) and is represented by three specimens collected in central-southern Chile (localities 73, 86, and 115).

The mitochondrial cyt b genealogy inferred using ML and Bayesian analyses also recovers a strongly supported (BS = 100; PP = 1.00; Fig. 2B) monophyletic subgenus Angelomys. However, as in the UCEs tree, the haplotypes of specimens traditionally assigned to Ab. andina and Ab. olivacea do not form reciprocally monophyletic groups. As mentioned above, because cyt b haplotypes from specimens included in the UCE analysis do not reflect the nuclear relationships recovered in the UCE tree, cyt b phylogroups were designated according to the nuclear lineage assignment of specimens with UCE data (Fig. 2A, B). Specimens represented only by cyt b sequences were subsequently associated with these cyt b phylogroups based on their position within the cyt b genealogy. Patterns of mitochondrial variation are therefore described within the framework of the three major nuclear lineages defined by the UCE-based tree.

Santiago-Coastal lineage (SCL). Cyt b haplotypes of specimens assigned to this lineage are mainly recovered in six well-supported phylogroups (SCL1–6; Figs 2, 3). Phylogroup SCL1 (BS = 100, PP = 0.95) contains haplotypes from two disjunct areas (Fig. 3): central Chile (localities 65-68, 71; 33°S–34°S) and the extreme north of Chile (localities 12, 13, 15; 19°S–20°S). Most haplotypes (n = 13) come from central Chile and are restricted to elevations above 1600 m a.s.l., reaching 2800 m a.s.l. in the Highlands of Santiago at Valle del Yeso (locality 66; Fig. 3). Specimens collected in this area correspond to Ab. andina sensu stricto. The remaining haplotypes (n = 5) come from the extreme north of Chile at elevations between 60 and 1200 m a.s.l. and included specimens of the type series (holotype and paratypes) of Ab. o. tarapacensis (localities 13 and 15; the latter is the type locality), as well as another specimen from Arica y Parinacota, in northernmost Chile (locality 12; Fig. 3).

Figure 3. 

Mitochondrial DNA tree and geographic distribution of specimens assigned to the Santiago-Coastal lineage (SCL). Left: cyt b genealogy of Angelomys with the SCL1–SCL6 phylogroups expanded. Tip labels include haplotype number and collection localities of specimens bearing the given haplotypes (Table SS1). Tip labels with number in parenthesis correspond to haplotypes from specimens collected in the highlands of Santiago (1) and to haplotypes from specimens of the type series of Ab. o. tarapacensis (2). Support values above and below branches correspond, respectively, to Ultrafast Bootstraps and posterior probabilities. Bottom left: haplotype network where circle size is proportional to haplotype frequency. Right: geographic distribution of SCL haplotypes across southern Peru and northern-central Chile. Panels A–D magnify areas where multiple haplotypes from different phylogroups occur in sympatry or in close geographic proximity. Localities with sympatric phylogroups are indicated by asterisks (*). Branches in the genealogy, haplotypes in network, and localities on the map are colored according to phylogroup identity. SCL_SL1 and SCL_HAL2 were recovered from specimens carrying SL1 and HAL2 mitochondrial haplotypes, respectively, but recovered within the SCL lineage based on morphological and nuclear evidence.

Phylogroup SCL2 (BS = 100, PP = 0.98) is one of the most widely distributed groups (24°S–32°S; Fig. 3). Across its broad latitudinal range, haplotypes occur from coastal and mid-elevation sites (ca. 300–1800 m a.s.l.; localities 29, 30, 32, 44–52, and 54) to a single high-elevation haplotype from Vallecitos, Atacama Region, collected at 3100 m a.s.l. (locality 35; Fig. 3). It also includes one haplotype, hap134, from the highlands of Santiago (locality 61; Fig. 2). This phylogroup is recovered as a sister to SCL1 with strong support (BS = 99, PP = 1.00; Figs 2B, 3). Together, SCL1 and SCL2 form a clade that is sister to phylogroup SL1 of the Southern Lineage, also with strong support (BS = 99, PP = 0.97; Fig. 2B).

The other phylogroups of the SCL lineage, SCL3 (BS = 98, PP = 0.93), SCL4 (BS = 98, PP = 1.00), SCL5 (BS = 99, PP = 1.00), and SCL6, include haplotypes recovered from specimens collected at coastal and mid-elevation areas (localities 11, 12, 14, 16–19; 17°S–22°S) between Tacna (southern Peru) and northern coastal Antofagasta (northern Chile; Fig. 3A–C). SCL3 includes haplotypes from all these localities, except 17. In the genealogy, this phylogroup is recovered as sister to all other haplotypes of Angelomys except those of phylogroup HAL2 of the High Andean lineage (Fig. 3). In contrast, phylogroups SCL4–6 comprise haplotypes from only one or two localities: SCL4 (locality 14), SCL5 (localities 17 and 19), and SCL6 (locality 8). Phylogroups SCL4–6 are closely related but do not form a monophyletic clade. Rather, they represent a paraphyletic group, because phylogroup HAL1 of the High Andean lineage is nested within this assemblage and is recovered as sister to SCL4 (Figs 2B, 3). Despite its large divergence, the phylogroup SCL3 geographically overlaps with the other phylogroups, with SCL1 at locality 12 (Fig. 3A), with SCL4 at locality 14 (Fig. 3B), and with SCL5 at locality 19 (Fig. 3C).

The haplotype network also reveals substantial internal structuring within SCL (Figs 3, S2). Haplotypes cluster intro three main groups, one corresponding to haplotypes of SCL1 from central Chile, another composed primarily of haplotypes of SCL2 distributed across north-central Chile, and a third dominated by phylogroups SCL3–6 from southern Peru and northern Chile. While the haplotypes of specimens showing mito-nuclear discordances (SCL_HAL2 and SCL_SL1; localities 45 and 69–71, respectively) are recovered as highly divergent mitochondrial haplotypes and outside the main groups in the haplotype network (Figs 3, S2).

Regarding genetic distance, the six phylogroups of SCL differ on average from each other by 4.14% (range: 1.67–5.51%; see details in Table S7).

High Andean lineage (HAL). Cyt b haplotypes of this lineage are recovered in two well-supported phylogroups, HAL1 (BS = 100; PP = 1.00) and HAL2 (BS = 100; PP = 1.00), which are not recovered as sister to each other and show a notable average of genetic distance of 10.47% (Fig. 2B; Table S7). Moreover, both phylogroups are disjunct (Fig. 4). Phylogroup HAL1 comprises haplotypes from specimens distributed across the highland regions of southern Peru, northern Argentina (15°S–25°S) and northern Chile at elevations above 3500 m a.s.l., except those from San Pedro de Atacama, Antofagasta (Chile, locality 21; 2370 m a.s.l.). The highest record for this clade is at an elevation of 4620 m a.s.l. on the flanks of Volcán Llullaillaco also in Antofagasta (locality 28; Fig. 4). This phylogroup includes haplotypes recovered from specimens collected near the type localities of Akodon andinus polius (locality 1) and Hesperomys dolichonyx (locality 21; Fig. 4). In the mitochondrial genealogy (Fig. 2B) this phylogroup is recovered in a highly supported relationship as sister of phylogroup SCL4 (BS = 99, PP = 1.00).

Figure 4. 

Mitochondrial DNA tree and geographic distribution of specimens assigned to the High Andean lineage (HAL). Left: cyt b genealogy of Angelomys with the HAL1 and HAL2 phylogroups expanded. Tip labels include haplotype number and collection localities of specimens bearing the given haplotype (Table SS1). Tip label with a number in parenthesis indicate haplotypes of specimens collected near to the type locality of (1) Akodon gossei, (2) Akodon andinus polius, and (3) Hesperomys dolichonyx. Support values above and below branches correspond to Ultrafast Bootstraps and posterior probabilities, respectively. Upper left: haplotype network with circle sizes proportional to haplotype frequencies. Right: geographic distribution of specimens from which HAL haplotypes were recovered. Branches in the genealogy, haplotypes in network, and localities on the map are colored according to phylogroup identity. Haplotypes indicated as SCL_HAL2 were recovered from specimens carrying HAL2 mitochondrial haplotypes but recovered within SCL based on morphological evidence (see Fig. 5 and the morphological Results section).

Phylogroup HAL2 distributes south of HAL1 and includes all haplotypes from specimens from the Chilean and Argentinean Puna and southern Andean steppe environments between 26°S and 35°S (Fig. 4) at elevations generally between 3000 and 4000 m a.s.l., with the lowest record at 1700 m a.s.l. in Mendoza, Argentina (locality 56), and the highest record at 5837 m a.s.l. on the Chilean flanks of the Ojos del Salado Volcano in the Parque Nacional Nevado Tres Cruces, Atacama (locality 36; Fig. 4). This phylogroup contains haplotypes corresponding to specimens collected nearby the type locality of Ak. gossei (Fig. 4). Two haplotypes of this phylogroup are found in sympatry with haplotypes of the phylogroup SCL2 at locality 45 (Fig. 4). In the mitochondrial genealogy HAL2 is recovered as the sister of all other haplotypes of Angelomys, exhibiting a large average genetic divergence (9.69%) from the rest of phylogroups Angelomys (Table S7).

The haplotype network for the High Andean lineage also shows two highly divergent haplogroups corresponding to HAL1 and HAL2 (Figs 4, S3). The haplotypes from specimens showing mito-nuclear discordance (SCL_HAL2) were recovered within the HAL2 despite its assignment to the SCL based on morphological evidence (see below the section on morphological Results).

Southern lineage (SL). Cyt b haplotypes of this lineage are recovered in three well-supported mitochondrial groups, SL1 (BS = 100, PP = 1.00), SL2 (BS = 100, PP = 0.94), and SL3 (BS = 100, PP = 1.00). However, as in the other cases, these phylogroups do not form a monophyletic group (Figs 2B, 5).

Figure 5. 

Mitochondrial DNA tree and geographic distribution of specimens assigned to the Southern Lineage (SL). Left: cyt b genealogy of Angelomys with the SL1–SL3 phylogroups expanded. Tip labels include haplotype number and collection localities of specimens bearing the given haplotype (Table SS1). Tip labels with number in parenthesis correspond to specimens collected in the vicinity of the type locality of (1) Mus olivaceus and (2) of a topotype of Ab. o. markhami. Support values above and below branches correspond to Ultrafast Bootstrap values and posterior probabilities, respectively. Bottom left: median-joining haplotype network with circle sizes proportional to haplotype frequencies. Right: geographic distribution of SL haplotypes. Inset panel magnifies an area where multiple haplotypes from different phylogroups occurs in sympatry or in close geographic proximity. Localities with sympatric phylogroups are indicated by asterisks (*). Branches in the genealogy, haplotypes in network, and localities on the map are colored according to phylogroup identity. Haplotypes indicated as SCL_SL1 were recovered from specimens carrying SL1 mitochondrial haplotypes but recovered within SCL based on morphological and nuclear evidence(see UCEs tree and morphological Results section).

Phylogroup SL1 includes haplotypes from specimens collected in the lowland region of central Chile (32°S–35°S), including a specimen collected in the vicinity of the type locality (Valparaíso) of Ab. olivacea (locality 60; Fig. 5). Also, four haplotypes (haps 155, 156, 159, and 160) were recovered from specimens collected in the vicinity of the type locality of Ab. andina, the highlands of Santiago, (UACH9241–9244; localities 69–71). However, UCEs data is available for only one of these individuals (UACH9241, hap155) which placed it within the Santiago-Coastal lineage in the UCE-based phylogeny. UACH9241 was also recovered as sister to UACH8303, another specimen from the highlands of Santiago, but whose cyt b haplotype (hap157) was recovered within the phylogroup SCL1 of the Santiago-Coastal lineage (Figs 2A, 3). Although the remaining specimens bearing hap156, hap159, and hap160 were not included in the UCE dataset, their morphology is more consistent with SCL1 than with SL1 (see below the mitochondrial-morphological discordance section of the Results).

Phylogroup SL2 shows the broadest geographic range (35°S–54°S), extending from central Chile through the temperate forests and Patagonian steppes in both Chile and Argentina (Figs 2C, 5). This phylogroup contains a haplotype from a topotype of Abrothrix olivacea markhami (SSUC_MA00329) from Puerto Eden, Isla Wellington (locality 111; Fig. 5). Moreover, haplotypes 170, 171, and 174 were recovered in sympatry with haplotypes 172 and 173 of phylogroup SL1 at Humedales de Putú, on the Pacific coast of the Maule Region, Chile (locality 77; Fig. 5). Phylogroup SL2 is recovered with strong support (BS = 100, PP = 1.00) as sister to the clade formed by SL1+(SCL1+SCL2), also with strong support (BS = 100, PP = 1.00; Fig. 1B).

Phylogroup SL3 encompasses haplotypes recovered from specimens collected in two geographically distant and disjunct areas. The first group (BS = 69, PP = 0.57) comprises haplotypes from the southernmost area of Chile and Argentina, including Isla Riesco, Punta Arenas, Tierra del Fuego, and Cape Horn (localities 113–129; Fig. 5). The second group (BS = 100, PP = 1.00) comprises haplotypes from Mendoza Province, Argentina (localities 58, 59, and 76; Fig. 5). Remarkably, haplotypes 208 and 212 are in sympatry with haplotypes 209–211, and haplotype 213 of phylogroup SL2 in two localities on the north margin of the Magellan Strait, Punta Arenas and Puerto de Hambre (localities 114 and 115; Fig. 5). Phylogroup SL3 is recovered as sister to the clade SCL6+(SCL5+(SCL4+HAL1)); however, this relationship receives only weak support (BS = 80, PP = 0.64), whereas the nested clade itself is strongly supported (BS = 100, PP = 1.00; Fig. 1B).

The haplotype network for the Southern lineage also reveals substantial internal structure within this lineage (Figs 5, S4). Haplotypes cluster into three main haplogroups, one corresponding to haplotypes of SL1 from central Chile, another composed of haplotypes of SL2 from central-south Chile, and a third by haplotypes of SL3 from southernmost Chile and Argentina, including Tierra del Fuego, and from Mendoza (Argentina). The haplotypes of specimens showing mito-nuclear discordances (SCL_SL1) are recovered within haplogroup of SL1.

The average genetic distance among the three main phylogroups is 4.09% (range from 2.98% to 4.92%; Table S7).

Cranial variation among specimens of the subgenus Angelomys reveals a structured pattern that generally matches the nuclear lineages identified in the UCE-based phylogeny. Below, for each lineage we describe the overall pattern of variation and their distinction from one another. A summary of morphometric measurements is provided in Table 1.

In the PCA, the first two principal components explain 76.2% of the total morphometric variance (PC1 = 65.4%, PC2 = 10.8%). PC1 exhibits uniformly positive loadings across all variables and largely reflects overall size variation, with the greatest contributions from zygomatic width (ZPW), followed by incisive length (IL), palatal length (PL), and toothrow length (TrL). In contrast, PC2 is structured by a combination of positive (PrL, NW, IL) and negative loadings (WFM, FL, ZPW; Table S8). Specimens of the HAL mostly occupy negative values of PC1 and positive values of PC2, whereas those of the Southern lineage are mainly distributed in the positive quadrant of both axes. Specimens of the SCL show a greater dispersion but mostly occupy positive values for PC1 and negative values along PC2 (Fig. 6A).

Figure 6. 

Morphometric variation among specimens of Angelomys based on cranial measurements. Specimens are coded by mitochondrial phylogroups (symbols and colors). Individuals collected in the highlands of Santiago, Chile (the general area of the type locality of Abrothrix andina) are highlighted with an orange outline. Specimens showing mito-morphological discordance are indicated by diamond symbols. A Principal Component Analysis (PCA); shaded polygons represent minimum convex hulls enclosing specimens assigned to each UCE-defined lineage B Discriminant Analysis of Principal Components (DAPC) based on the same morphometric dataset; ellipses represent 95% confidence intervals around group centroids.

Although some overlap is observed among the three nuclear-defined lineages, particularly between SCL and SL, PERMANOVA indicates that SCL, HAL, and SL are significantly differentiated in multivariate space when analyzed jointly (Bonferroni-adjusted p < 0.05 for all pairwise comparisons). However, when morphometric differentiation was analyzed at the level of mitochondrial phylogroups, patterns become more heterogeneous (Table S9). Within HAL, no significant morphometric differentiation was detected between HAL1 and HAL2. In contrast, both HAL phylogroups are significatively differentiated from SCL and SL phylogroups (Table S9). Similarly, within SL, no significant differentiation is detected between SL1 and SL2, although both phylogroups also lack differentiation from some SCL phylogroups: SL1 does not differ from phylogroups SCL1 and SL2 does not differ from SCL2 or SCL6.

In contrast, SCL exhibits greater internal structuring. Phylogroups SCL1, SCL2, and SCL4 are significantly differentiated from other SCL phylogroups, whereas no significant differences were detected among SCL3, SCL5, and SCL6. Moreover, when specimens of the SCL were grouped geographically rather than by cyt b phylogroups, size variation among population becomes much more evident, with northern specimens smaller than those from the south (Fig. S5). In general, specimens from southern Peru and northernmost Chile tend to be smaller than those from central Chile.

DAPC reveals a clear separation between HAL and SCL and SL but greater overlap between SCL and SL (Fig. 6B). The analysis retained 14 principal components, accounting for 98.9% of the total morphometric variance, and yielded eight discriminant functions (Table S10). The first discriminant function primarily separates specimens of the HAL from those of SCL and SL, whereas the second discriminant function captures variation largely shared between the latter two lineages, resulting in their partial overlap. Variables contributing most strongly to discrimination along the first discriminant function include IL, FSW, IOB, whereas discrimination along the second function is driven mainly by DL, NL, NW, IOB, and ZB (Table S10). Overall classification accuracy was 85.7%. Group-specific recall and precision values indicate heterogeneous classification performance among lineages (Table S11). Misclassified specimens are predominantly assigned among phylogroups of the same nuclear lineage (Fig. S6) but see mito-morphological discordance cases below.

Qualitative examination, together with morphometric analyses, revealed differences among the three lineages defined by the UCE tree (Figs 7, 8, 9, S7, S8; Tables 1, 2). Specimens of the Santiago-Coastal lineage (SCL) generally exhibit larger and more elongate skulls, enlarged auditory bullae with a well-developed eustachian tube, and marked variation in dorsal coloration and tail morphology across their geographic range. In contrast, specimens of the High Andean lineage (HAL) are the smallest within the subgenus and are characterized by globose auditory bullae with a reduced eustachian tube, densely haired feet with lateral fringes, and short tails. Specimens of the Southern lineage (SL) exhibit a more homogeneous morphology, generally resembling some populations of SCL in coloration pattern, but differing by their larger skull size, less hairy feet lacking lateral fringes, and more globose bullae with narrower eustachian tubes. Additional qualitative descriptions and comparisons are provided below in the species accounts. A detailed presentation of qualitative descriptions and comparisons is provided below in the species accounts.

Figure 7. 

Specimens of the subgenus Angelomys illustrating variation in auditory bullae morphology among the Santiago-Coastal lineage (SCL: A–C), High Andean lineage (HAL: D, E), and Southern lineage (SL: F). An elongated bulla with a well-developed eustachian tube is observed in the three morphotypes of SCL: A Highlands near Santiago (UACH 9242). B Coastal southern Peru-northern Chile (UACH 8710) and C north-central Chile (UACH 8773). More rounded bulla with a reduced eustachian tube characterizes HAL specimens from its D northern (UACH 8667) and E southern distribution (UACH 8813). Slightly elongate bulla with a well-developed eustachian tube is observed in SL specimens from F lowland Central Chile (UACH 9038). Dashed lines outline the general shape of the bulla, and arrows indicate the eustachian tube.

Figure 8. 

Specimens of Angelomys illustrating variation in the hair pattern coloration of the dorsal fur and the ventral fur projection onto the muzzle between Santiago-Coastal lineage (SCL: A–D, F–I) and High Andean lineage (HAL: E and J). From reddish to pale yellowish dorsal coloration with bicolored hairs and ventral fur that does not strongly reach the half of the muzzle is observed in SCL from the Andean highlands near Santiago morphotype (A, UACH9243 and F, UACH9245), Coastal southern Peru-northern Chile morphotype (B, UACH8733; C, UACH8711; G, UACH8731; H, UACH8714), and north-central Chile morphotype (D and I, UACH8772). Vibrant orangish-brown coloration with tricolored hairs and ventral fur that strongly reaches the half of the muzzle is observed in specimens of HAL (E, UACH8672 and J, UACH 8820). Although not illustrated, examined specimens of Southern lineage (SL) exhibit similar characteristics like those in A, D, F, and I.

Figure 9. 

Specimens of Angelomys illustrating variation in forefoot morphology Santiago-Coastal lineage (SCL: A–C, F–H), High Andean lineage (HAL: D and I), and Southern lineage (SL: E and J). A more densely haired forefoot, with well-developed claws and a conspicuous fringe of hair, is observed in specimens of SCL from the Andean highlands near Santiago morphotype (A and F, UACH 9244). In contrast, the other two morphotypes of SCL from southern Peru-northern Chile (B and G, UACH 8717) and north-central Chile (C and H; UACH 8770), exhibit a less densely haired forefoot, with poorly developed claws and a reduced fringe of hair. All examined specimens of HAL display a densely haired forefoot with well-developed claws and a prominent hair fringe, as illustrated by specimens from its northern clade (D, UACH 8802; I, UACH 8672). In contrast, specimens of SL show a less densely haired forefoot, with poorly developed claws and a reduced fringe of hair, as exemplified by specimens from lowlands of central Chile (E and J; UACH 9039). White arrows indicate the morphological features described.

Morphological variation within the Santiago-Coastal lineage. Specimens in the SCL exhibit a highly morphological variation along its broad geographic distribution and can be grouped in three major morphotypes that generally correspond to a north-south geographic groups: southern Peru-northernmost Chile, north-central Chile, and the Andean highlands of central Chile (Tables 1, 2). Despite substantial variation, described below, all of them share the following combination of characters: bicolored hairs on the dorsal fur, presence of an anteroloph on M1; incisive foramina extending to the level of the anterior border of the protocone on M1; and auditory bullae mostly enlarged and with a well-defined, shorter, or longer but wider; eustachian tube (Figs 7A–C, 8A–D; Table 2).

Fur coloration pattern exhibit by specimens of this lineage is probably the most variable feature and ranges from dull grayish-yellow to bright orangish-brown dorsal fur with bicolored hairs (Figs 8A–D, S7A–F), whereas the ventral fur varies from whitish to pale buffy and may extend to the gula region only, to the lower lips, or gradually to the half of the muzzle (Fig. 8F–I). Specimens from southern Peru and northernmost Chile (mainly phylogroups SCL3–6) tend to exhibit brighter and more uniform reddish to orangish dorsal coloration (Fig. 8B, C), comparatively longer tails (Fig. S7B–E; Table 2), forefeet with a poorly-developed claw on finger V and reduced lateral fringe on the fore- and hindfeet (Figs 9B, 9G, S8B); narrower skull; and more globose auditory bullae. In contrast, specimens from the Andean highlands near Santiago (mainly phylogroup SCL1) exhibit a duller grayish-yellow dorsal fur coloration that contrast with the whitish ventral fur (Fig. S7A) and are unique in having a well-developed claw on finger V of the forefeet (Fig. 9A), densely hairy fore- and hindfeet with a well-developed lateral fringe of hairs (Figs 9F, S8A); shorter, densely haired and strongly bicolored tail (Fig. S7A; Table 1), larger molars, and more robust skulls with elongate bullae and a conspicuous tubular eustachian tube (Fig. 7A). Specimens from north-central Chile, associated with phylogroup SCL2, are generally intermediate in morphology, exhibiting grayish-yellow to grayish-orangish coloration (Fig. S7E, F); moderately hairy fore- and hindfeet with a short claw in the finger V of forefeet and reduced fringe of hairs on the lateral sides of the fore- and hindfeet (Figs 9C, 9H, S8C), medium to long tails (Table 1), and variable development of the auditory bullae and length of the anteroloph on M1. A much more detail comparison among these three groups within SCL is showed in Table 2.

Morphological variation within the High Andean lineage. This lineage, which comprises the smallest specimens of Angelomys, exhibit a silky, slightly uniform dorsal fur coloration consistent of a pale orangish-brown (Fig. S7G, H) with tricolored hairs that have a lead-gray bases (50% or less of the length of the hairs); a pale yellowish middle band, and orangish tips (Fig. 8E); a pale yellowish or pale orangish ventral fur with lead- to slate-gray base (50 to 60% of the length of the hairs) that extends to the half of the muzzle (Fig. 8J); flanks slightly paler than the dorsum (Fig. S7G, H), a conspicuous pinna but covered more than the half of its length by the dorsal fur when laid back (Fig. 8J) and with a whitish patch of hair; a large claw on the finger V of the forefeet (Fig. 9D); densely hairy fore- and hind feet with a well-developed fringe of hairs on their lateral side (Figs 9I, S8D); and a short ( < 70% of HBL; Table 1), densely hairy and slightly or strongly bicolored tail (Fig. S7G, H). The skull is the smallest within the subgenus Angelomys in all measurements (Table 1), with the anterior tip of the rostrum nearly squared; the shortest incisive foramina that posteriorly extends no more than the posterior border or M1 anterocone; the smallest upper molars without anteroloph on M1 in most cases and when present it is noticed just like a small lump; with or without alisphenoid strut; a squarish or rounded mesopterygoid fossa; and a comparatively much globose and rounded auditory bullae with a short or very reduced eustachian tube (Fig. 7D, E). Specimens from the northern fraction of the range exhibit some differences from those from the SL range (Table 2).

Morphological variation in Southern lineage. The specimens examined from this lineage exhibit a dorsal coloration pattern similar to that observed in the first and fourth morphotypes of the Santiago-Coastal lineage (SCL) but in some cases specimens of SL show darker hair tips. The ventral fur is more grayish due to a longer lead- to slate-gray base (> 60% of the length of the hairs) and only extends to the gula region and in few cases reaches but gradually to the lower lips. The lateral sides are like the dorsal fur. The pinna is evident and is covered by the dorsal fur less than the half of its length and does not exhibit a patch of whitish hairs behind or at its base. The claw on digit V of the forefeet is short (Fig. 9E); and the fore- and hindfeet are sparsely hairy lacking a fringe of hairs on their lateral sides (Figs 9J, S8E). The tail is medium in length (> 70% of HBL; Table 1), slightly hairy and varies from monocolored to slightly bicolored tail. The skull is large in size with a wider rostrum and mostly rounded; longer and wider incisive foramina that posteriorly extends to the level of the half of the M1 paracone; medium-size upper molars (Table 1), with long or short anteroloph on M1 in most cases; with a shallow anteromedium flexus on M1 in some juvenile specimens; with or without alisphenoid strut; with squarish or inverted M-shape mesopterygoid fossa, and with globose and slightly elongate auditory bullae but with an evident, short or long, slightly narrow eustachian tube (Fig. 7F).

We identified five specimens exhibiting discordance between their cyt b phylogroup assignment and morphological affinity. Cyt b haplotypes 102 and 106, corresponding to specimens UACH8972 and UACH8976 from Alcohuaz, Chile (locality 45), were recovered within phylogroup HAL2 (Figs 3, 4). However, both quantitatively and qualitatively morphological evidence assigns these specimens to the Santiago-Coastal lineage. In both PCA and DAPC analyses, these specimens fall within the morphometric space occupied by the Santiago-Coastal lineage and are classified as part of the morphotype of phylogroup SCL2 with a probability of 1.00 (Figs 5, S6). Consistent with their morphometric placement, these specimens also match the external and craniodental features characteristic of the Santiago-Coastal lineage (Fig. S7F). UCE data are unavailable for these specimens.

Similarly, cyt b haplotypes of four specimens from the Highlands near Santiago, Metropolitana Region (UACH 9241–UACH9244; haplotypes 155, 156, 159, and 160; localities 69–71) were recovered within phylogroup SL1 (Figs 3D, 5). However, in the morphometric space, specimens bearing those haplotypes fall in the area of overlap between SCL1 and SL1 (green diamonds in Fig. 6) and were classified as part of SCL1 in the DAPC with more than 0.9 of probability (Fig. S6). These specimens also share discrete external traits with specimens of SCL, including densely haired fore- and hindfeet with a well-developed fringe of hairs along the external margins and a short, densely haired tail (Figs 8A, S7A). Although UCE data are unavailable for three of these specimens, UCE data from specimen UACH9241 (cyt b haplotype 155) recovered it as sister to UACH8303 (Fig. 2A), a specimen carrying SCL1 mitochondrial haplotype (haplotype 157; Fig. 3).

Analyses based on mitochondrial DNA sequences (mostly cyt b sequences for mammals) have provided the foundation of phylogeographic and molecular taxonomic studies, driving major advances in our understanding of species limits and diversification processes (see also Wüster 2025). In fact, taxonomic studies of sigmodontine rodents over the past three decades have been predominantly guided by mitochondrial phylogenies (e.g., Pearson and Smith 1999; Hoffman et al. 2002; Hurtado and D’Elía 2018). However, the recognition that the topologies of individual gene trees may not to be congruent with the species tree (see Pamilo and Nei 1988), highlights the need for multilocus data to reliably infer evolutionary relationships and species boundaries. In this context, the present study represents a substantive step forward in the much-needed integration of multilocus evidence and morphological data for evaluating species boundaries within Sigmodontinae.

Our results show that the mitochondrial cyt b genealogy of Angelomys does not reflect either the pattern of phenotypic variation or the species tree inferred with UCE data. Given that the UCE dataset comprises 1839 loci and was analyzed using a coalescent-based approach, it is reasonable to interpret the resulting topology as a good proxy of the species tree. Therefore, we conclude that the cyt b genealogy does not accurately trace the species tree and is an unreliable proxy for species boundaries in Angelomys.

Existence of mito-nuclear discordances have been previously documented in Patagonian and Fueguian populations of Abrothrix olivacea (Giorello et al. 2021; Quiroga-Carmona et al. 2022), and our results indicate that such discordance also extends to central and northern populations of Angelomys. For example, phylogroups HAL1 and HAL2 are deeply divergent in the mitochondrial genealogy and are not recovered as sister groups, yet they form a cohesive and well-supported nuclear lineage that is phenotypically distinct from the other two main lineages of Angelomys. The processes driving this pattern remain unclear, and different mechanisms may underlie the observed discordance, including scenarios involving deep coalescence or incomplete lineage sorting, as documented in other mammals such as the European bison (Wang et al. 2018). Alternatively, this pattern could reflect mitochondrial capture following events of historical introgression, a process widely reported across animal taxa (Good et al. 2008; Singhal and Moritz 2012; Toews and Brelsford 2012; Melo-Ferreira et al. 2014), or “ghost introgression” from an extinct or unsampled lineage, which can generate deep mitochondrial splits without similar level of nuclear divergence (Zhang et al. 2019).

In contrast, the mito-nuclear discordance involving specimens of Santiago-Coast lineage bearing HAL2 or SL1 cyt b haplotypes seems to be more consistent with geographically localized and relatively recent introgression in zones of sympatry or parapatry. Two regions are particularly notable in this regard. The first corresponds to Alcohuaz (locality 45), where specimens morphologically assigned to SCL carry HAL2 cyt b haplotypes. The second involves specimens from highlands near Santiago (localities 69–71), where they carry SL1 cyt b haplotypes but cluster morphologically and genomically with SCL. In both cases, the observed pattern of discordance could be compatible with introgressed mitochondrial genomes, which often track contact zones more predictably than patterns generated by ILS alone (Toews and Brelsford 2012). Therefore, denser sampling at and near the areas where lineages HAL2 or SL1 reach each other is still needed to test this hypothesis. Comparable cases of mito-nuclear discordance associated with contemporary hybridization have recently been documented in other rodents, including the sigmodontines Phyllotis limatus–P. vaccarum (Quiroga-Carmona et al. 2025), which are mostly co-distributed with Angelomys. More generally, climatic oscillation and associated demographic changes may have facilitated the emergence and persistence of the gene tree discordance uncovered here (see Quiroga-Carmona et al. 2022).

Together, these findings highlight the need for future studies incorporating denser geographic sampling across putative contact zones between species of Angelomys, as well as genome-wide tests of admixture to disentangle the relative roles of incomplete lineage sorting and introgression, and to better constrain the timing and directionality of gene flow. Such approaches will be critical for understanding the evolutionary history of the species of the subgenus Angelomys.

The taxonomic proposal presented here prioritizes that pattern emerging from genome-wide nuclear relationships recovered from UCE data and the patterns of morphologic variation. Species names are assigned following the principle of priority and are anchored to specimens collected at or near type localities to ensure nomenclature stability. Within this framework, we recognize three species within the subgenus Angelomys of the genus Abrothrix: Ab. andina (= Santiago-Coastal lineage), Ab. dolichonyx (= High Andean lineage), and Ab. olivacea (= Southern lineage). This arrangement departs both from the previous scheme recognizing two species (Ab. andina and Ab. olivacea) and a more recent proposal that treats Ab. andina as a junior synonym of Ab. olivacea and recognized Ab. gossei as a distinct species (Fig. 1B; Tammone et al. 2025, 2026). It also entails a substantial reconfiguration of the geographic ranges for each species (Fig. 1C). Moreover, we acknowledge that additional genome-wide sampling and more detailed morphological analyses will be necessary to further evaluate intraspecific variation and assess the potential recognition of infraspecific units, especially within Ab. andina. We discuss each species in detail below.

Historically, the name Abrothrix andina has been applied broadly to high Andean populations ranging from the highlands of Santiago (central Chile) and adjacent areas of Mendoza (central Argentina) northward through the highlands of northern Chile and Argentina, western Bolivia, and south and central Peru (e.g., Patterson et al. 2015). The type locality of andina is imprecise (“andibus elevatis prov. Santiago”) and the specimens studied by Philippi are presumably lost (e.g., our search in the MNHN in Santiago was not successful; but see below regarding the existence at the BMNH of a specimen potentially studied by Philippi when describing the species). Several nominal forms (i.e., Akodon jucundus, Ak. gossei, Ak. andinus polius, Hesperomys dolichonyx, H. d. cinnamomea) were subsequently synonymized under Ab. andina. Two subspecies have been traditionally recognized: Ab. a. andina Ab. a. dolichonyx (Osgood 1943, 1944; Musser and Carleton 1993; Patterson et al. 2015). However, recent studies showed that cyt b sequences recovered from specimens collected in the general area of the type locality of Ab. andina as well as others assignable to dolichonyx, are nested within a clade containing cyt b sequences of Ab. olivacea (Tammone et al. 2024, 2025, 2026; see also Quiroga-Carmona et al. 2022). In contrast, cyt b sequences from Mendoza province (Argentina), assignable to the form gossei, were recovered as sister to all other cyt b sequences of the subgenus Angelomys. Based on the topology of the cyt b gene tree and genetic divergence values, but without conducting direct comparative assessment of topotypic specimens of A. andina and A. olivacea, and without considering evidence from nuclear markers, Tammone et al. (2025, 2026) assigned populations of Ab. andina sensu stricto and those of the subspecies dolichonyx to Ab. olivacea; as such, these authors proposed the synonymy of andina and dolichonyx (and its associated forms polius and jucundus) under Ab. olivacea; meanwhile, gossei was recognized as a distinct species (Tammone et al. 2026: 231). Given the vagaries of gene tree evolution, reliance on a single gene tree to advance taxonomic decisions is prone to error, a scenario overlooked by Tammone et al. (2025, 2026). Indeed, our results strongly indicate that overreliance on mitochondrial gene tree topology led these authors to propose a taxonomic arrangement that does not accurately reflect species boundaries.

Our results strongly support a different taxonomic scheme. Across cyt b genealogy, UCEs, and morphology, specimens from the general area of the type locality of Ab. andina are neither closely related nor morphologically similar to the high Andean populations historically included within Ab. andina. In fact, these specimens are recovered within a clade composed of specimens distributed along the Pacific lowlands and mid-elevation areas from southern Peru to central Chile, which have traditionally been assigned to the subspecies Ab. o. tarapacensis under a broad geographic concept of Ab. olivacea (e.g., Quiroga-Carmona et al. 2022; see also Pacheco et al. 2024). This clade, referred as the Santiago-Coastal lineage throughout our Results sections, is here recognized as a distinct species of Abrothrix (Angelomys). Pending further evaluation of historical material, we provisionally apply the name Abrothrix andina to this species, as it has nomenclatural priority over Ab. o. tarapacensis and its type locality lies within the general distributional area of the Santiago-Coastal lineage. Accordingly, we treat Ab. o. tarapacensis as a junior synonym of Ab. andina. Under this revised concept, Abrothrix andina encompasses populations from the Highlands of central Chile and extends northward through coastal and mid-elevation areas of northern Chile into southern Peru (Figs 1C, 3). Therefore, Ab. andina does not appear to be distributed in high Andean Puna environments as traditionally assumed. In central Chile, Ab. andina occurs in the Andean steppe ecoregion and may also occur at lower elevation as suggested by Iriarte and Simonetti (1986), who reported specimens of Ab. andina from San Carlos de Apoquindo (Metropolitana Region) at only 950 m a.s.l.; however, no voucher specimen appears to support this mention.

As shown for Abrothrix olivacea (e.g., Quiroga-Carmona et al. 2023), Abrothrix andina exhibits substantial geographic variation in morphology. Specimens from highlands near Santiago, corresponding to Ab. andina sensu stricto, include the largest specimens of the species and are distinctive externally in having a short, densely haired tail and a well-developed lateral fringe of long hairs on both fore- and hind feet (Figs 9A, S7A, S8A). In contrast, specimens from its northernmost portion of the distribution (southern Peru-northernmost Chile; 17°S–20°S) form a distinctive group characterized by smaller cranial size and lighter orangish dorsal fur coloration, characteristics that match the original description of tarapacensis as a small, pale-golden form (see Rodríguez-Serrano et al. 2006). Populations distributed along the Chilean coast (21°S–32°S) and adjacent mid-elevation areas (0–3100 m a.s.l.) are generally intermediate between these two extremes, suggesting potential clinal variation (Figs 5, 6, 7, S5, S7E, F).

At present, we interpret the pattern of phenotypic variation as pronounced geographic structuring within a single species, Abrothrix andina. This scenario is plausible given the fragmented and environmentally heterogeneous landscapes of central and northern Chile. In such settings, complex topography and discontinuous habitats can reduce connectivity and promote local differentiation, as highlighted by Quiroga-Carmona and D’Elía (2022) and Quiroga-Carmona et al. (2023). Cases of morphological variation correlated with environmental variation are well documented in species of Abrothrix (e.g., Naya et al. 2014; Teta et al. 2022) and in other mammals inhabiting arid and semi-arid environments (e.g., Webster and Webster 1980; Al-Kahtani et al. 2004; Taylor et al. 2022). Although the observed differences may eventually support the recognition of subspecific units, it remains unclear whether this variation reflects continuous clinal or ecogeographic variation, localized adaptation, or partially independent evolutionary histories (see Patton and Conroy 2017). Consequently, we prefer to treat these population as part of a single species with no interna taxonomic subdivision pending additional analyses based on a denser geographic sampling.

Finally, clarifying species limits for the subgenus Angelomys also requires stabilizing the application of the name Abrothrix andina, a taxon with an imprecise type locality. Historical sources indicate that one specimen associated with Philippi’s original material was deposited at the Museo Nacional de Historia Natural, Santiago (MNHN). During our examination of material at the MNHN, however, we did not locate any specimen that could be confidently identified as original material of Ab. andina, contrary to Osgood (1943), who reported examining a specimen identified as Mus andinus. Similarly, also of interest is another specimen housed at the British Museum of Natural History, London (BMNH), which may have been or not part of the original series studied by Philippi (see Thomas 1920; Osgood 1943). The identity and status of the BMNH specimen also remain uncertain since we were unable to access it. Thomas (1920: 418) considered the BMNH specimen morphologically inconsistent with Philippi’s original description of Mus andinus and referred it to Akodon gossei, whereas Osgood (1943) regarded it as consistent with the rather vague description concept of andinus. Given this uncertainty, and pending direct examination of the BMNH specimen, we refrain from either selecting it as lectotype or designating a neotype for Mus andinus.

Nevertheless, to contribute towards nomenclature stability, we herein restrict the type locality of Mus andinus (= Ab. andina) to Las Melosas, Fundo el Ingenio, Ingreso Fundo Cruz de Piedra, Region Metropolitana, Chile, 1694 m a.s.l., 33°55’04.4”S 70°12’21.5”W. This locality lies within the general area originally indicated by Philippi (“Highlands of Santiago”) and corresponds to a region where the species has been documented and genetically characterized in the present study, including specimens such as UACH 8303, which was included in the cyt b and UCE datasets. Although this area, Las Melosas, currently belongs to the Cordillera Province of the Metropolitan Region, during the time of the work of Philippi it belonged to the Santiago Province (see maps in Sagredo Baeza et al. 2017).

Under our revised taxonomy, this species corresponds broadly to the traditional concept of Abrothrix andina after excluding specimens from the highlands of Santiago (i.e., typical andina). Historically, dolichonyx (including cinnamomea and jucundus) was treated as a subspecies of Ab. andina, whereas the nominal forms polius and gossei were regarded as junior synonyms of Ab. andina s.s. (e.g., Osgood 1943; Musser and Carleton 1993; Patterson et al. 2015). Recent studies, based solely on cyt b DNA sequences indicated that these names are not related to Ab. andina and, on that basis, proposed major taxonomic changes, including treating dolichonyx and allied names (here corresponding to phylogroup HAL1) as part of Ab. olivacea, while recognizing gossei (here corresponding to phylogroup HAL2 due to cyt b haplotypes from specimens from Mendoza can be regarded as topotypes of this form) as a distinct species of Angelomys (Tammone et al. 2024, 2025, 2026). Our analyses, based on distinct lines of evidence, support a different interpretation. Although our UCE dataset includes a limited number of specimens of this lineage (n = 5; Fig. 2A), these samples span much of its geographic range and include representatives of both cyt b phylogroups HAL1 and HAL2. Nuclear and morphological data consistently support the existence of a single species-level lineage adopted here correspond to the High Andean lineage, that is distinct from Ab. olivacea and Ab. andina. Importantly, the cyt b genealogy links this lineage with specimens collected at or near the type localities of H. dolichonyx, Ak. a. polius, and Ak. gossei, and these specimens are morphologically consistent with the High Andean lineage. Accordingly, evidence support recognizing this lineage as Abrothrix dolichonyx (the oldest available name) and treating gossei as part of this species (contra Tammone et al. 2025, 2026). Under this scheme, Ab. dolichonyx encompasses specimens from highlands of Peru, Bolivia, Chile, and Argentina up to 5837 m a.s.l. (Storz et al. 2024). In addition, it occurs in sympatry with Ab. andina at Vallecitos, Atacama Region, Chile (Fig. 2C).

Within this revised concept of Abrothrix dolichonyx, two geographically structured mitochondrial groups are found. Specimens from the northern Puna, bearing cyt b haplotypes of the HAL1 phylogroup, are geographically disjunct from specimens from the southern Andean steppe of Chile and Argentina, which carry HAL2 haplotypes. Although these two groups exhibit qualitative morphological differences, these are not accompanied by statistically significant distinction in the multivariate morphometric space. As such, the northern and southern groups may be regarded a subspecies; due to HAL2 includes haplotypes of specimens collected close to the type locality of gossei and fits the original description of this form (Thomas 1920), this name should be applied for this subspecies. However, we refrain from formally recognizing infraspecific taxa pending denser geographic sampling and explicit tests of diagnosability and geographic coherence.

Philippi (1896) indicated that two specimens formed the basis of the original description of Hesperomys dolichonyx but did not designate a holotype or provide catalogue information on those specimens. Therefore, these specimens constitute a syntype series. At the Museo Nacional de Historia Natural, Santiago (MNHN), a mounted specimen (MNHN No. 23) labeled “No. 23” was located and bearing a handwritten annotation reading “Hesperomys dolichonyx Ph., W.H. Osgood, April 29, 1928” (Fig. S7). This specimen (Fig. S8A) strongly resembles the original description of Philippi and the illustration of Hesperomys dolichonyx and H. d. cinnamomea (Figs S8C, D), as well as specimens examined here and assigned Ab. dolichonyx, particularly specimens from phylogroup HAL1 collected near San Pedro de Atacama, the type locality of dolichonyx (e.g., locality 20–22; Figs 7F, S8B). All these specimens exhibit a marked contrast between the dorsal and ventral fur coloration, a short and densely haired tail, and a densely hired forefeet. To stabilize the application of the name dolichonyx, we herein designate MNHN No. 23 as the lectotype of Hesperomys dolichonyx Philippi, 1896 and restrict its type locality, as specified below.

Abrothrix dolichonyx (Philippi, 1896)

Hesperomys dolichonyx Philippi, 1896: 21.

Hesperomys dolichonyx cinnamomea Philippi, 1896: 22.

Akodon jucundus Thomas, 1913: 140.

Akodon gossei Thomas, 1920: 418

Akodon andinus polius Osgood, 1944: 196.

Lectotype.

MNHN N°23, mounted skin without skull, tail detached (Figs 10, S9A).

Figure 10. 

Specimen of Hesperomys dolichonyx housed at the Museo Nacional de Historia Natural (MNHN, Santiago, Chile), herein designated as the lectotype of the species. The specimen is shown with its associated museum label (lower left), which bears the name Akodon andinus dolichonyx, locality information, and catalog number (No. 23). The inscription attached to the wooden base (lower right) preserves the original name Hesperomys dolichonyx and includes a handwritten note attributed to W.H. Osgood, dated April 29, 1928.

Restricted type locality.

The type locality of this species was referred as “Vecindad de Atacama” and was equaled to “San Pedro de Atacama” in Osgood (1943) and Paynter (1988). Here we restrict the type locality of this nominal form to Ruta B-241 km 177, San Pedro de Atacama, Antofagasta, Chile (23°1.095' S, 68° 8.175' W; locality 21 in Fig. 4). This locality is in the vicinity of the city of San Pedro de Atacama and we are certain that A. dolichonyx is present there (specimen UACH 8672).

Description.

The lectotype fits the original morphological description provided by Philippi (1896), including the presence of dorsal hairs with yellowish tips, ventral and lateral pelage paler than the dorsum, ears largely concealed by dorsal fur, whitish lips, a brownish nose, a densely haired tail with limited dorsal-ventral contrast, and feet covered by long hairs with short, dense hairs along the outer margins (Figs 10, S9A).

Remarks.

Osgood (1943: 179–180) mentioned: “Philippi’s second specimen, which perhaps should be regard as a cotype, is now in the British Museum (N°11.11.17.7). The skin has been remade, and the skull is somewhat crushed”. Following Osgood’s account, specimen BM 11.11.17.7 could be considered as a paralectotype; however, it was not examined firsthand in the present study, and we do not formalize it as a paralectotype.

Under our revised taxonomy, Abrothrix olivacea does not occur along the Pacific coast or in the mid-elevation valleys of northern Chile and southernmost Peru, as specimens from these areas are recovered as part of the new definition of Ab. andina (Figs 1C, 2C). Accordingly, the species concept of Ab. olivacea adopted here corresponds to the Southern lineage recovered in the UCE-based phylogeny (Fig. 2A). Although our UCE dataset includes only three specimens of Ab. olivacea, one of these clusters in the mitochondrial genealogy with a haplotype from the general area of the type locality of Ab. olivacea (Valparaíso; haplotype 133, locality 60; Fig. 5), which can be regarded as a topotype, forming part of phylogroup SL1 and thus linking the nuclear-defined lineage with topotypical material. Notably, this specimen is recovered in the UCE phylogeny within the same clade as UACH8467 from Puerto Hambre (Magallanes; locality 115), which carries an SL2 mitochondrial haplotype (haplotype 212), indicating that although SL1 and SL2 do not form a single clade in the mitochondrial genealogy, they do so in the nuclear tree (Fig. 2A). This pattern is concordant with our morphological results, as SL1 and SL2 show extensive overlap in multivariate space, lack statistical or discrete differences in external characters among the examined specimens (Tables 2, S9). Moreover, previous genome-wide analyses demonstrated genetic cohesion and evidence of gene flow among populations from Central Chile, Patagonia, and Tierra del Fuego (Giorello et al. 2021). Taken together, these lines of evidence support treating all these populations, referred to as olivacea sensu stricto, brachiotis, and xanthorhina (see Quiroga-Carmona et al. 2022) as part of a single species, Ab. olivacea, which extends from lowlands of central Chile (Valparaíso and Metropolitana Regions) and western Argentina (Mendoza Province) southward to Cape Horn, encompassing Mediterranean scrubs, forests, steppes, and grasslands of Central Chile, Patagonia, Tierra del Fuego, and adjacent islands. Interestingly, in the Metropolitana Region of Chile and Mendoza province of Argentina, Ab. olivacea appears restricted to lowland areas and does not occur in sympatry with Ab. andina (Figs 1C, 3D, 5).

Nevertheless, geographic variation within Abrothrix olivacea is likely more complex than suggested by our sampling. Quiroga-Carmona et al. (2023) and Sánchez et al. (2022) documented morphometric differentiation across its range, including larger body size in insular populations relative to mainland populations and differentiation among populations from the Patagonian steppe, Chilean shrublands, and the Magellanic and Valdivian forests. Similarly, Yañez et al. (1979) identified latitudinal clinal variation in multiple cranial and external traits across population traditionally assigned to olivacea and xanthorhina, arguing that much of the observed phenotypic variation represent geographically structured continuous variation rather than discrete taxonomic units, and consequently proposed treating xanthorhina as subspecies of Ab. olivacea. Comparable ecogeographic patterns have also been documented in other species of the genus, such as Abrothrix hirta (see Teta et al. 2022), a species with a broadly similar geographic distribution of the new concept of Ab. olivacea. Taken together, these studies suggest that phenotypic differentiation across southern South America may largely reflect geographically structured and environmentally mediated variation rather than sharply delimited evolutionary units. A comprehensive assessment of geographic structure within Ab. olivacea will therefore require expanded genome-wide and phenotypic sampling across its range, particularly for lineages underrepresented in our dataset (e.g., populations corresponding to hershkovitzi, markhami, llanoi, xanthorhina, and those from Mendoza, see Quiroga-Carmona et al. 2022 and Sánchez et al. 2022). Such efforts will be essential to evaluate whether named geographic variants are best interpreted as subspecies within a widespread species or whether Ab. olivacea lacks meaningful infraspecific differentiation.