In this study, 51 Eozapus specimens were collected from the Sichuan, Shaanxi, Gansu, and Ningxia provinces of China (Fig. 1, Table 1). Specimens were collected primarily using the snap trap method. Field surveys and collections of specimens both follow relevant laws and regulations in China. The field-collected samples were numbered, and morphological data (including body weight, tail length, ear height, and hindfoot length) were measured (Suppl. material 1: table S1). Liver and muscle tissues were obtained and preserved in analytically pure alcohol (99%). Upon return to the laboratory, tissues were stored in a -80 °C freezer for subsequent molecular analyses, while specimens were prepared as the study skins for morphological studies. All tissues and specimens used in this study were archived at the Sichuan Academy of Forestry (SAF) and the College of Life Sciences of Sichuan Normal University (SCNU) (Table 1).

Figure 1. 

Map of the collection sites of the genus Eozapus in this study.

The extraction of total DNA was carried out following the manufacturer’s protocols for the animal liver or muscle tissue DNA extraction kit (Chengdu Fukuji Biological Co.). In this study, three genes were amplified for molecular phylogenetic and related analyses: one mitochondrial gene (CYT B, 1059 bp) and two nuclear genes (growth hormone receptor [GHR, 569 bp] and inverted repeat-binding protein [IRBP, 590 bp]). PCR amplifications were performed in a reaction volume mixture of 25 μl, containing 3 mM MgCl₂, 0.2 U rTaq polymerase (Takara, Dalian, China), 1 × reaction buffer, 0.2 mM of each dNTP, 0.4 mM of each primer, and approximately 100–500 ng of genomic DNA. The primers used for PCR amplification are listed in Suppl. material 1: table S1. PCR amplification was performed with an initial denaturation step at 94 °C for 5 min, followed by 35 cycles of denaturation at 94 °C for 45 s, annealing at 49–60 °C for 45 s, extension at 72 °C for 90 s, and a final extension step at 72 °C for 10 min. PCR products were identified using 1% agarose gel electrophoresis and then sent to QingKe Biological Co., Ltd. for purification and bidirectional Sanger sequencing.

Sequencing results were aligned preliminarily using MEGA v5 (Tamura et al. 2013). All sequences were manual corrections edited to ensure the proper translation, thereby avoiding potential impacts on subsequent analyses. CYT B, IRBP, and GHR sequences from the selected Eozapus species and the outgroup Stylodipus sungorus, Stylodipus andrewsi, and Stylodipus telum were downloaded from GenBank (Suppl. material 1: table S2).

Three datasets were created for phylogenetic analysis: 1) a mitochondrial dataset (CYT B), 2) a nuclear gene dataset (nuDNA), and 3) an all-gene combined dataset (CYT B + nuDNA). The optimal base substitution model for each dataset was determined using jModelTest 2.1.7 (Posada 2008), and the Akaike information criterion (AIC) was used to assess the fitness of each model (Luo et al. 2010). The optimal substitution models for each gene and dataset are listed in Suppl. material 1: table S3.

Phylogenetic trees were constructed using Bayesian inference (BI) based on the three established datasets. The MrBayes 3 program was used to reconstruct the BI tree (Ronquist and Huelsenbeck 2003). Parameter settings for the BI tree were based on those described by Chen et al. (2012). Posterior distributions were obtained using the Markov chain Monte Carlo (MCMC) method with one cold chain and three heated chains for 100 million generations at a sampling interval of 10,000. The first 2,000 generations were treated as burn-ins. The Markov chain MCMC runs were repeated twice to confirm a consistent approximation of the posterior parameter distributions. Figtree v1.4.3 was used to visualize the BI tree output, and a posterior probability (PP) > 0.95 was considered strong support (Huelsenbeck and Rannala 2004).

Genetic distances between clades were calculated based on the CYT B gene using the Kimura 2-parameter model in MEGA v5.05 (Tamura et al. 2011).

BPP v3.0 was used to test the hypothesis of independent evolutionary lineages represented by populations (Yang and Rannala 2010; Yang et al. 2012; Yu et al. 2013). To verify the genetic differentiation detected in the genus Eozapus, the gene tree constructed in the previous step was used as a guide tree, and the validity of our assignment of the three putative species was tested in BPP v3.1. The analysis of Chen et al. (2022) was followed, which was defined using Algorithm 0 and Algorithm 1, with τ representing the root age of the species tree and θ representing the size parameter of the population. These two important parameters significantly affect the posterior probability of the BPP (Yang and Rannala 2010; Pan et al. 2019; Chen et al. 2022). Twelve analyses were conducted using three different prior combinations of τ and θ: (1) τ = G (2, 2000), θ = G (2, 2000); (2) τ = G (1, 10), θ = G (2, 2000); and (3) τ = G (1, 10), θ = G (1, 10) (Leaché and Fujita 2010). Each rjMCMC was run for 100,000 generations, with samples collected every 100 generations after discarding 10,000 generations as pre-burn-in. When the posterior probability exceeded 0.95, the analysis supports that each evolutionary branch is a separate species.

Since mitochondrial genes may lead to an earlier estimated differentiation time (Phillips 2009; Zheng et al. 2011), the nuclear gene dataset was used for estimation in this study. The time of differentiation of the genus Eozapus was analyzed and imputed using BEAST v1.7.5 (Drummond et al. 2012) based on the nuDNA dataset. Four fossil calibration points were used in this study, as described by Shenbrot et al. (2017). (1) Stylodipus and Chimaerodipus originated within the Dipodinae at approximately 3.84–5.49 Ma (mean = 4.85, standard deviation = 0.61, upper 95% = 3.84–5.85 Ma). (2) Eremodipus and Jaculus originated within the Dipodidae at approximately 3.75–5.74 Ma (mean = 4.73, standard deviation = 0.61, upper 95% = 3.73–5.73 Ma). (3) Differentiation of Jaculus occurs at approximately 1.18–2.31 Ma (mean = 1.73, standard deviation = 0.354, upper 95% = 1.15–2.31 Ma). (4) Dipodinae originated at approximately 9.46–11.47 Ma (mean = 10.53, standard deviation = 0.65, upper 95% = 9.46–11.60 Ma). The remaining parameters were set as described by Chen et al. (2022).

A total of 17 complete skull specimens of intact adult individuals were obtained and measured using digital Vernier calipers, with an accuracy of 0.01 mm, as described previously (Liu et al. 2007, 2012; Patton and Conroy 2017). The measurement metrics included body weight (BW) , head-body length (HBL) , tail length (TL) , hindfoot length (HFL) , ear length (EL) , profile length (PL) , skull basal length (SBL) , median palatal length (MPL) , zygomatic breadth (ZB) , least breadth between the orbits (LBO) , brainstem breadth (BB) , height of the braincase (HB) , auditory bulla length (ABL) , upper toothrow length (UTRL) , length of the upper molars (LUM) , upper molar row breadth (UMRB) , mandibular length (ML) , lower toothrow length (LTRL) , and length of the lower molar row (LLMR). Because external measurement data may display differences between different measurers, they were not used for the morphological analysis. SPSS v17.0 was used to analyze the measurement values of the indicators. When the data satisfied a normal distribution, principal component analysis (PCA) was used to evaluate the overall similarity between species, and discriminant analysis (DA) was used to determine whether the classification was correct.

We obtained 51 CYT B sequences of 1059 bp and 37 nuclear gene sequences of 1159 bp (GHR: 569 bp; IRBP: 590 bp). The new sequences were deposited in GenBank (Accession Numbers PQ723106–PQ723119, PQ75286–PQ752971, Table 1). MrBayes was used to reconstruct phylogenetic relationships based on the three datasets, and all three topological structures obtained were largely consistent (Fig. 2). BI results based on three different datasets supported the division of the genus Eozapus into three clades. E. vicinus was monophyletic and was consistently positioned as the most basal lineage across all three datasets (CYT B: PP = 1.00, nuDNA = 0.38, and CYT B + nuDNA = 1.00). Eozapus sp. was identified as a sister group of E. setchuanus. In the CYT B phylogenetic tree, evolutionary relationships within this taxon were highly supported. The CYT B + nuDNA gene tree was identical and well supported (PP = 1.00) by the topology of this taxon in the CYT B gene tree (Fig. 2).

Figure 2. 

Bayesian phylogenetic analyses based on the CYT B dataset, the nuDNA dataset, and the CYT B+ nuDNA dataset. The numbers above the branches refer to Bayesian posterior probabilities (PP).

The K2P genetic distances between the three clades of Eozapus based on the CYT B gene are shown in Table 2. The results revealed that the genetic distance values ranged from 10.9% to 17.8%. The genetic distance value between E. vicinus and Eozapus sp. was at 17.8%, followed by 16.6% between E. vicinus and E. setchuanus, and 10.9% between Eozapus sp. and E. setchuanus was the smallest. Genetic distances among all species showed good species-level variation.

The results from BPP based on the CYT B + nuDNA and nuDNA datasets produced 36 results, all of which strongly supported the division of Eozapus into three clades (PP = 1.00; Suppl. material 1: table S4).

The results indicated that the most recent common ancestor of Eozapus dates back to the Miocene (8.15 Ma, 95% CI = 5.23–10.62), marking the divergence of E. vicinus from the other two species. The divergence time of Eozapus sp. and E. setchuanus (6.12 Ma, 95% CI = 3.41–9.16) occurs in the Miocene (Fig. 3).

Figure 3. 

Divergence times estimated using BEAST based on the nuDNA dataset. Branch lengths represent time. Numbers above the nodes indicate posterior probabilities (PP). Numbers below the nodes represent the median divergence time. The four asterisks indicate fossil-calibrated nodes.

A total of 14 craniodental and 5 external indicators were measured (Suppl. material 1: table S5). External indicators were excluded from the morphological analysis because of potential interobserver variation in measurements (Jiang et al. 2003). Additionally, owing to damage to the zygomatic bones of several specimens during handling, ZB measurements were excluded from the analysis, leaving 13 skull measurements for morphological analysis (Table 3).

The results of the PCA based on 13 cranial measurements revealed that the two principal components explained 77.049% of the variance (Table 4). The first principal component (PC1) accounted for 59.533% of the variance and showed positive factor loadings. PL, SBL, and MPL had factor loadings > 0.9, indicating a strong correlation with PC1, Factor loadings > 0.8 included UTR, ML, and LTRL. The second principal component (PC2) explained 17.516% of the variance and had mostly negative factor loadings, with LUM and LLMR being the largest loading, correlating strongly with PC2. It also was negatively correlated with LBO (loadings < - 0.2) (Table 4). The PCA results indicated that most individuals were effectively classified into three species with minimal overlap (Fig. 4).

Figure 4. 

Result of the principal component analysis (PCA) and the discriminant analysis (DA) based on skull measurement data of the genus Eozapus.

Discriminant analysis (DA) showed that the three specimens within Eozapus were accurately classified (Fig. 4). Correct taxonomic determination was obtained for 100.0% of the original grouped samples, and cross-validation results indicated that 100% of the samples were correctly classified.

The skull of Eozapus sp. is the smallest (average PL = 21.58 mm), while that of E. setchuanus is 22.16 mm and E. vicinus is 22.37 mm. In fact, the average skull measurement of Eozapus sp. was less than that of the other two species. The SBL (average = 16.22 mm), UTRL (average = 9.51 mm), ML (average = 12.49 mm), and LTRL (average = 8.01 mm) of Eozapus sp. were the shortest, whereas the UTRL (average = 10.26 mm), ML (average = 13.21 mm), and LTRL (average = 8.65 mm) of E. setchuanus are the longest (Table 3). The tooth formula of this genus is 1.0.1.3/1.0.0.3. Depression of the occlusal surface of the upper molars in the genus Eozapus is variable; E. vicinus is slightly concave, Eozapus sp. is moderate, while E. setchuanus is severely concave (Fig. 5a2, b2, c2, a6, b6, c6). The lengths of the first and second lower molars are nearly equal, and the third lower molar (m₃) is reduced (approximately 30%). On the second lower molar (m₂), the metalophid of E. vicinus penetrated the entire surface of the second lower molar (m₂) from the middle, whereas for Eozapus sp. and E. setchuanus, the metalophid of the m₂ is discontinuous and concave inward medially (Fig. 5a6, b6, c6). In addition, Eozapus sp. has a distinctly depressed longitudinal groove in the middle of the third lower molar (m₃) and towards the lingual side (Fig. 5b6); the m₃ of E. vicinus is flat in the middle (Fig. 5a6), and E. setchuanus also has a depressed longitudinal groove in the middle of the third lower molar (m₃), but differs from E. vicinus in that it is directed to the buccal side (Fig. 5c6).

Figure 5. 

Skulls and tooth comparison among species of the genus Eozapus. a. E. vicinus a1–a6; b. Eozapus sp. b1–b6; c. E. setchuanus c1-c6 holotype; (1) ventral view; (2) upper tooth row; (3) dorsal view; (4) lateral view; (5) lower jaws; (6) lower tooth row; The red arrow indicates the position of the dental variation.

Hair characteristics of the genus Eozapus are as follows (Fig. 6d). The color of the dorsal and notal hairs was the same in all three species, brownish yellow, but E. vicinus was slightly lighter than Eozapus sp., while in E. setchuanus the extent of the yellow wash on the individual hairs extended further towards the root (Fig. 6). (1) E. vicinus exhibits pure white at the base and tip of hair in its abdomen. (2) Eozapus sp. has no brownish-yellow longitudinal stripes on its chest, and its hair base is white with yellow tips on its abdomen. (3) E. setchuanus has a brownish-yellow longitudinal stripe on its chest. The hair at the base was purely white with yellowish tips on its abdomen.

Figure 6. 

Pelage comparisons among species of the genus Eozapus: a. Eozapus vicinus (museum number SAF11163); b. Eozapus sp. (museum number SAF191248); c. Eozapus setchuanus (museum number SAF11706); d. Hair variation (below chest to hindfoot); (1) dorsal view; (2) ventral view; (3) lateral view.

Based on the distinct morphological characteristics and deep genetic divergence described in the preceding sections, we recognized the jumping mouse from the Wanglang National Nature Reserve as a distinct, undescribed species within Eozapus and formally described it as follows.