Sampling

Between 2003 and 2011 we collected 97 Lemniscomys specimens in Guinea from 25 localities (Fig. 1). Superficially at least, these specimens divided into two morphological groups: 19 specimens bearing a single dorsal black stripe and 78 specimens with multiple dorsal stripes. Rodents were sacrificed following the recommendation of Sikes et al. (2016) and standard measurements were taken (weight, body length, tail length, hind foot length, ear length), as well as notes on their reproductive state. Liver and spleen samples were fixed in ethanol for molecular analyses. Bone marrow was extracted from femurs for preparation of standard karyotypes. Carcasses of whole animals were preserved in 5% formalin and skulls were cleaned and prepared for measurement.

Molecular analyses

Total genomic DNA was extracted following the CTAB method as described in Winnepenninckx et al. (1993). For the 28 new specimens sequenced, a large part of the cytochrome b gene (cytb) was amplified using PCR primers L7 (5′-ACCAATGACATGAAAAATCATCGTT-3′) and H14896 (5′-TAGTTGTCGGGGTCTCCTA-3′) or H15915 (5′-TCTCCATTTCTGGTTTACAAGAC-3′). The PCR consisted of an initial denaturation for 3 min at 94 °C, followed by 38 cycles: 30 s at 94 °C, 40 s at 52-55 °C and 90 s at 72 °C. The reaction was completed by a final extension for 10 min at 72 °C. PCR Products were purified and commercially Sanger-sequenced at EUROFINS (by Automatic Sequenceur ABI 3730).

Numerous cytb sequences are available in GenBank for L. striatus. Nicolas et al. (2008) showed a strong phylogeographic structure for this species with only one lineage (IV-1) present in Guinea. We included in the present study all specimens from Guinea available in GenBank. For other lineages we retained only one specimen per lineage. For other Lemniscomys species all sequences available in GenBank were included in our study, as well as several newly sequenced specimens, giving a total of 62 individuals (Table S1).

For phylogenetic analyses, we retained a fragment on 846 bp available for most specimens. Evolutionary relationships among sequences were estimated by constructing a phylogenetic tree using Bayesian inference. We used several Arvicanthini to root our tree: Lamottemys okuensis, Desmomys harringtoni, Rhabdomys pumilio, Mylomys dybowski, Pelomys campanae, Dasymys incomtus and Arvicanthis niloticus. The computer program MRMODELTEST ver. 2 (Nylander et al. 2004) was used to evaluate the fit of 24 nested models of nucleotide substitution in the data. The model chosen by MRMODELTEST according to the Akaike information criterion (GTR + I + G) was subsequently used for Bayesian analysis. Bayesian inference was performed with MRBAYES ver. 3.2.5 (Ronquist et al. 2012) using 5 million generations (burnin = 25%), sampling every 1000 steps. Convergence between chains was checked by manual inspection using the ‘sump’ command and by using the mean SD of split frequencies, which was below 0.01 in our runs, as an indicator of convergence.

Cytogenetics

Standard karyotype analysis of eight individuals (four multi-striped specimens and four single-striped individuals) from two localities (Tanganya and Bantou) was performed under field conditions using the colchicine method of Lee & Elder (1980) detailed in Denys & Aniskin (2012) (Table S1).

External and skull morphometry

Morphological observations of skulls were performed using microscopy. Molars were drawn with the use of a camera lucida and skulls photographed with a Canon EOS reflex camera. The dental nomenclature follows Monadjem et al. (2015). A classical morphometrics analysis was performed using new specimens and comparisons with holotypes from the NHM London, the MRAC Tervuren, the ZFMK Bonn, the SMNS Stuttgart and previously karyotyped individuals of the MNHN Paris collections (Table S1). Skull measurements were taken with a MITUTOYO calliper using the protocol of Denys et al. (2012). Following a repeatability test, 10 skull and three mandibular measurements were retained: LGT (greatest length of the skull), WNAS (width of nasals across the points where maxillofrontal sutures contact nasals), LNAS (maximal length of nasals), CIO (interorbital constriction width), WZYG (bizygomatic width), WPAR (maximal posterior width of the skull at the level of the interparietal), LPAL (length of incisor foramen), OCCPAL (post-palatal length), LS13 (upper tooth row length), LBT (tympanic bulla length), HMDB (maximal height of the mandible), LMDB (length of the mandible) and LI13 (lower tooth row length). Molars are abbreviated as follows: M1/, M2/, M3/ upper M123, M/1, M/2, M/3 lower M123.

Univariate and multivariate morphometric analyses were performed using external and skull measurements with XLSTAT version 10.1 software (Addinsoft). T-tests were employed to compare distributions of each variable for each species. We used discriminant function analysis (DFA) on log-transformed external measurements and skull distances based only on genetically-typed specimens and holotypes and topotypes in good condition (Table S1). Because not all specimens are integrally preserved (broken tails, broken skull bones, etc.) there was some variability in the number of individuals (N) for each variable. In total we observed the skins and measured the skulls of 31 L. linulus, including the holotype and a paratype, 58 L. striatus including the holotypes of L. s. venustus and L. s. pulchella, 14 L. bellieri including the holotype, five paratypes and five topotypes. We performed comparisons with L. zebra and L. macculus holotypes.

Fig. 2.

Cytochrome b phylogeny of Lemniscomys individuals recovered by Bayesian analysis (GTR + I + G model). To improve clarity outgroup taxa were removed. Nodes with posterior probabilities greater than 0.96 are indicated with a black dot. Specimens from Guinea are in red. For L. striatus, Roman numbers indicate the genetic lineage following Nicolas et al. (2008). For newly sequenced specimens only field numbers are indicated. For sequences retrieved from GenBank both field numbers (when known) and GenBank numbers are indicated. Three-letter codes indicate country of origin: BEN = Benin, CAM = Cameroon, CAR = Central African Republic, DRC = Democratic Republic of Congo, GAB = Gabon, GUC = Guinea, KEN = Kenya, MAL = Mali, MOR = Morocco, NIG = Nigeria, RCI = Ivory Coast, RWA = Rwanda, SEN = Senegal, TAN = Tanzania, SAF = South Africa.

Mitochondrial phylogeny

Specimens from Guinea fit within two mitochondrial clades (Fig. 2): the first clade includes L. striatus specimens from Franfina, Tanganya, Bamba, Bantou, Benti, Bowé, Gania, Gayebombo, Gbaolé, Kaali, Samedou, Yerende, Ziama and Zogota, and corresponds to the clade IV-1 of Nicolas et al. (2008). The second clade includes those specimens from Guinea morphologically identified as L. linulus and derived from Fressoudou, Kodoko, Bantou and Tanganya. This second clade also includes one specimen of L. linulus from Mali, specimens of L. bellieri from RCI and Benin, and one sequence of L. macculus from the Central African Republic (CAR).

Fig. 3.

Standard karyotype of new Guinean specimens of Lemniscomys. A) Lemniscomys striatus with 2n = 44, FNa = 66 (MNHN-ZM-2008-8), B) Lemniscomys linulus with 2n = 56, FNa = 66 (MNHN-ZM-2012-1028).

Fig. 4.

Dorsal pelage of Lemniscomys holotypes. 1) L. linulus NHM Holotype NHM 11.6.10.89 Gamon, Senegal; 2) L. venustus Holotype NHM 11.3.24.13 Panyam, northern Nigeria; 3) L. macculus Holotype NHM 6.12.4.57 Mokia, SE Ruwenzori, Uganda; 4) L. oweni Holotype NHM 11.6.10.61 Gemenjulla, Senegal; 5) L. bellieri Holotype RMNH 75-74-M-1 1446 Lamto, RCI Credit Royal Museum for Central Africa; 6) L. zebra Holotype S.M.N.S. 5422-1100 Bongo, Sudan, credit H. Turni/SMNS.

Cytogenetic analysis

Among the eight new karyotyped specimens, we found only two discrete karyotypes. The first was obtained from specimens associated with the first molecular clade. It corresponds well with the standard L. striatus karyotype. It is characterized by 2N = 44, NFa = 66 and comprises seven metacentric, nine acrocentric and five subtelocentric autosomes. Two females and two males from central Guinea (localities of Tanganya and Bantou) display this karyotype (MNHN-ZM-2008-7, MNHN-ZM-2008-8, MNHN-ZM-2008-9 and MNHN-ZM-2008-10; Fig. 3A). The second karyotype was obtained in specimens associated with the second molecular clade. The karyotype is 2N = 56, NFa = 66 (Fig. 3B). It is characterized by 4 metacentric, 21 acrocentric and 2 telocentric autosomes. One female and three males from Tanganya and Bantou show this karyotype (MNHN-ZM-2012-1026, MNHN-ZM-2012-1030, MNHN-ZM-2012-1028 and MNHN-ZM-2012-1031). This karyotype is novel for the genus based on data from the literature (Table 1).

Fig. 5.

Dorsal pelage of Guinean specimens collected in this study. 1) L. linulus (MNHN-ZM-2012-1029, Tanganya); 2) L. striatus (MNHN-ZM-2012-1040, Bantou).

Pelage colouration of types and Guinean specimens

Lemniscomys linulus is characterized by a single black median stripe and no other light stripes while L. striatus, L. bellieri, L. macculus and L. zebra display supplementary stripes in addition to the median stripe. Compared to the holotype of L. bellieri and L. macculus, the L. linulus holotype is characterized by the absence of punctuated lines and by a browner colouration, including the belly (Fig. 4). In L. bellieri there is one thin median black stripe bounded by two thin lines of white-cream spots. The punctuated line of white-cream spots increase in size towards the flanks and five lines are visible on each side of the dorsal region, separated by light grey-brown pelage. This pattern differs clearly from the L. macculus holotype, which displays one large black line, bordered on each side by four nearly continuous rows of small white-cream spots alternating with four dark or brown zones that do not form clear lines, on its dorsal pelage. It was not possible to see the Linnaeus holotype of L. striatus, but we compared our specimens with other west central African type specimens, which were considered junior synonyms, such as L. venustus and L. pulchella. L. venustus was re-described in detail by Van der Straeten (1981) who noticed slight differences with L. striatus but considered it as a valid sub-species of striatus. The holotype of L. striatus, re-described by Carleton & Van der Straeten (1997), corresponds to an immature juvenile. It bears seven pairs of dark and light punctuated stripes and a median brown dorsal line (not illustrated here). In L. venustus there are three dark brown bands on each side of the black median dorsal line and 3 to 4 yellow-cream lines of spots. The same number of dark lines is observed in L. pulchella and the yellow-cream spots are organized in more continuous and visible lines than in L. venustus. In the L. zebra holotype and in L. oweni (a junior synonym of L. zebra) 3 to 4 continuous yellow-white-cream lines and 3 to 4 brown lines can be seen on both sides of the black median line (Fig. 4).

We found two different types of dorsal pelage in our Lemniscomys from Guinea. 78 specimens display the typical L. striatus dorsal colour pattern, while in 19 specimens only one dorsal stripe was visible (Fig. 5). Because, only one species with a single dark line and no yellow lines is known to occur in West Africa, L. linulus, we attributed to the specimens displaying this morphological characteristic. Like the holotype of L. linulus, a clear median-dorsal black stripe and no lateral yellow stripes or punctuated lines was visible in our specimens. The pelage is brown with some yellow hairs and spots visible on the flanks, but not organized geometrically (Fig. 5). Our Guinean specimens appear to have slight differences compared with the holotype of L. linulus, including a slightly larger black median stripe that is framed by two conspicuous thin and nearly continuous yellow lines of spots. Several supplementary specimens from the Guinean locality of Dinguiraye (not genotyped) display the same pattern and were consequently attributed to L. linulus. We did not observe any variability in this dorsal pelage pattern among our new Guinean specimens. The new specimens attributed to L. striatus (following the molecular work of Nicolas et al. 2008) have one median black line with 3-4 lateral lines on each flank, which corresponds to the original description. In our specimens, the upper stripes are organized in eight nearly continuous lines of spots and the lateral ones are less continuous (Fig. 5). The median black stripe is relatively large. The new Guinean specimens attributed here to L. striatus differ from specimens of L. zebra by not having continuous lines on both sides of the median black line (Fig. 5). They have four lateral lines of brown and yellow-white spots on both sides of a median black line like in L. striatus.

Table 2.

Standard measurements of West African Lemniscomys species using Types, MNHN karyotyped and new genotyped Guinean specimens. *after Tranier & Gautun‘s work. **measured on wet and dry mounted specimens by S.M. ***after Carleton & Van der Straeten (1997).

Specimens attributed to L. bellieri by Ziegler et al. (2002) from Faranah (Guinea) housed at the ZFMK were examined. The dorsal skin displays a narrow black median line surrounded by two cream-white thin lines of nearly continuous spots. From each side of the thin white spot lines four dark-brown-grey lines alternating with four white cream spots lines are visible. This pattern is similar to the holotype of L. bellieri.

Fig. 6.

Skulls of Lemniscomys type specimens. 1-4: dorsal view, 5-8: ventral view. 1 & 5: L. linulus paratype NHM 11.6.10.70; 2 & 6: L. bellieri Holotype RMNH 75-74-M-1 1446, Credit Royal Museum for Central Africa 3 & 7: L. macculus Holotype NHM 6.12.67.57; 4 & 8: L. zebra Holotype S.M.N.S. 5422-1100, credit C. Leidenroth/SMNS.

Fig. 7.

Dorsal and ventral skulls of L. striatus – 2 & 4 (MNHN-ZM-2012-1044) and L. linulus – 1 & 3 (MNHN-ZM-2012-1031) from Tanganya, Guinea.

Some West African specimens housed at the MNHN,andalreadykaryotypedbyseveralauthors, were re-examined here (not illustrated). The striped South Mali specimens from Balamansala and Tombane (Table S1), already attributed to L. linulus, display the same dorsal pattern as the holotype with a relatively thin median black line, brown pelage and yellow-brown hairs on the hind feet and tail. The tail length is equal or smaller to the head and body length. On several specimens the pelage is speckled, with yellow and cream spots, but never organized into a continuous line. Two RCI specimens from Bouna are deposited in the MNHN collection as L. griselda. These specimens were karyotyped by Matthey (1954) under this specific name but are clearly similar to L. linulus and display the same dorsal pattern and colouration. Thus, we re-attribute them to the species L. linulus. In these two specimens the tail length is equal to the head and body length.

Variability in external measurements

Standard statistics for external morphological measurements show large overlap among Lemniscomys species from West Africa (Table 2), as noted by Rosevear (1969). Lemniscomys striatus looks slightly larger than L. linulus, L. bellieri and L. zebra, and seems to have a proportionally smaller hind foot. The tail is longer than the head and body length for all species (Table 2). A comparison of genotyped L. bellieri versus L. linulus, shows only a significant difference in tail length, while between L. linulus and L. striatus only ear length is significantly different. Between L. bellieri and L. striatus all the external measurements showed no significant differences. The holotype of L. striatus, re-described by Carleton & Van der Straeten (1997), has a tail shorter than head and body length. The L. linulus from Guinea and Mali have similar external measurements to the holotype and topotype specimens. The tail length is equal or slightly longer than the head and body length in all the specimens examined here (Table 2).

Table 3.

Skull measurements of Guinean Lemniscomys and surrounding countries (southern Mali, Senegal, and RCI). In bold: highest mean value. In italic lowest mean value. Min-max = minimum and maximum values.

Skull characters

The skulls of L. striatus and L. linulus display short and wide nasals, a wide interorbital constriction, large braincase, long incisor foramen, large molars and small tympanic bullae. The holotype of L. linulus (not illustrated) is old and with posterior damage, but it displays a short and wide nasal whose anterior end is oblique on both sides of the median suture, with well-marked fronto-parietal crests. This conformation is also observed on the L. linulus paratype (Fig. 6).

The holotype of L. striatus is a juvenile and its skull has never been illustrated (Carleton & Van der Straeten 1997). The holotypes of L. venustus and L. pulchella (syn. L. striatus) are damaged and few morphological skull characters are visible. On the L. bellieri, L. macculus and L. zebra holotypes the nasals are short and narrow and with a round anterior end on both sides of the median suture. It is more oblique in L. zebra and L. macculus than in L. bellieri (Fig. 6).

In ventral view (Fig. 6), the incisor foramen of L. linulus looks slightly longer and narrower at its distal part compared to those of L. striatus; it stops at the level of the t1 of the upper M1/ in L. linulus while in our L. striatus and the L. pulchellus holotype it stops before the t1. In L. bellieri the incisor foramen is rectilinear and relatively large, ending at the level of the pre-lobe of the upper M1/. In L. pulchella, despite its poor state of preservation, the incisor foramen stops at the level of the pre-lobe of the M1/. In L. zebra and L. macculus the incisor foramen is also narrower at its distal end and stops before the t1 of the upper M1/. The pterygoid fossae looks larger in L. striatus than in L. linulus, the foramen magnum is wider and more open ventrally in L. striatus than in L. linulus. In lateral view (not illustrated), the incisors are orthodont in L. linulus and opistodont in L. striatus. The mastoid processes in L. linulus is well developed. The mandible is high and short in L. linulus and the angular process less developed than in L. striatus.

In dorsal view, our Guinean specimens assigned to L. striatus also display short nasals, narrower than in L. linulus and ending with a round anterior end on both sides of the median suture (Fig. 7). Our Guinean L. linulus specimens show the anterior ends of the nasals to be oblique, like in the type specimen, and well-marked frontoparietal crests. In ventral view, our new specimens show either a short large incisor foramen ending at the level of the M1/ prelobe (like L. striatus or L. bellieri) or a longer one that is narrower at its distal part, ending at the level of the t1 of the M1/, like in L. linulus. The new specimens attributed to L. linulus also show orthodont incisors and a well-developed mastoid process (not illustrated).

L. linulus has similar skull dimensions to specimens attributed to L. striatus but differs in having significantly longer molar rows and smaller tympanic bullae (Table 3). There were significant differences between L. bellieri and L. linulus in WNAS, LMDB, HMDB, LS13 and LI13. Between L. bellieri and L. striatus the mean values of CIO width, WNAS and LBT differ significantly.

Dental morphology

The upper molars of west African Lemniscomys all display the typical Arvicanthini dental pattern with large molars and well-aligned transversal, bunodont cusps t2 and t3. On the holotype of L. linulus, there is a small t9 on the M1/ and the M2/. The t3 and the t4 of the M1/ displays a small trace of stephanodont crest. The upper M3/ has a large t1 and no t3, two lophs and the second lobe of the M3/ displays one cusp. The lower M/1 does not show a supplementary median anterior cusp (tma) nor a posterior cingulum or labial Cv5. The antero-labial cusp of the prelobe is smaller than the antero-lingual cusp and the prelobe is fused to the first lobe by a longitudinal crest, the latter also visible on the second lobe whose cusps are oblique and of equal size. On the M/2 there is a small Cp and a round antero-labial cusp. The M/3 is smaller than the M/2 and has two lobes of two cusps (not illustrated). In the paratypes and type of L. bellieri a similar disposition of the molar cusps can be seen to that in L. linulus except for the prelobe of the lower M/1 (Fig. 8). The two anterior cusps are approximately equal in size and nearly longitudinal in their orientation, which differs from L. linulus where the two anterior cusps of the prelobe of M/1 have a more transverse orientation, with the labial cusp smaller than the lingual. The holotype of L. zebra corresponds to a stage 3 specimen and displays bunodont cusps. The t9 is small. On the upper M3 only one small cusp on the second lobe is visible. On the lower M/1 the two anterior cusps of the prelobe are small, of similar size and oblique with a 90° angle between them. There are no stephanodont crests, and the molars do not display any cingular supplementary cusps. The posterior cingulum is small on M/1 and slightly larger on M/2. The M/3 display one large cusp and a tiny external cingular cusp.

Fig. 8.

Upper molar rows of several West African Lemniscomys. 1-4: L. linulus (MNHN-ZM-2012-1030 Tanganya, Guinea, 2 – MNHN-ZM-2011-615 Kodoko, Guinea, 3 – MNHN-ZM-2013-2778 Dinguiraye, Guinea, 4 – MNHN-ZM-MO-2004-728 Tombané, Mali); 5-6: L. striatus (5 – MNHN-ZM-2008-10 Tanganya, Guinea, 6 – 207 = MNHN-ZM-2011-619 Bantou, Guinea); 7: L. bellieri (MNHN-ZM-1979-156 Lamto, RCI); 8: L. zebra (MNHN-ZM-1971-756, Foro RCI); 9: L. macculus (MNHN-ZM-1992-1542, Bossangoa, CAR).

On our Guinean L. linulus specimens a similar organisation of the cusps to those in the holotype can be seen, accompanied by some variability in the numbers of accessory tubercles. The t3 of the upper M1/ display a small t3 bis. The upper M3/ has two small cusps on its second lobe. On the lower molars a tma, Cp, Cv1, cv5 are visible on the lower M/1. Cusps are oblique and the prelobe is fused with the first lobe. The same orientation and smaller size of the antero-labial cusp as in the holotype can be seen at all stages of wear. The lower M/3 has one cusp on the second lobe (Figs. 8, 9).

Fig. 9.

Lower molar rows of several West African Lemniscomys. 1-4: L. linulus (1 – MNHN-ZM-2012-1030 Tanganya, Guinea, 2 – MNHN-ZM-2011-615 Kodoko, Guinea, 3 – MNHN-ZM-2013-2778 Dinguiraye, Guinea, 4 – MNHN-ZM-2004-728 Tombané, Mali), 5-6: L. striatus (5 – MNHN-ZM-2008-10 Tanganya, Guinea, 6 – MNHN-ZM-2011-619 Bantou, Guinea), 7: L. bellieri (MNHN-ZM-1979-156 Lamto RCI), 8: L. zebra (MNHN-ZM-1971-756, Foro, RCI), 9: L. macculus (MNHN-ZM-1992-1542, Bossangoa, CAR).

In our Guinean L. striatus material, the cusps are more bunodont and less fused than in L. linulus. A small stephanodont crest is seen on the t4. The t9 is small. On the upper M3/ there is always a trace of two cusps on the second lobe. On the lower M/1 the cusps of the prelobe and the first lobe are well separated until an advanced stage of wear, the two anterior cusps are about equal size and nearly aligned transversally. A tma and a cv5, as well as a small Cp are present on the M/1. The second lobe of the M/1 display a transverse lamina with well aligned cusps (Figs. 8, 9).

Fig. 10.

Discriminant function analysis on 13 skull distances of 36 genetically typed Guinean and Malian specimens of L. linulus (L), Guinean L. striatus (S), and karyotyped L. bellieri from the type locality of Lamto in Ivory Coast (B). Circles represent 95% confidence ellipses and yellow points barycentres of each species.

Malian and karyotyped RCI L. linulus display the same slightly stephanodont upper molars, the well-fused cusps of the prelobe and first lobe of the M/1 with the labial anterior cusp smaller than the lingual in comparison with the holotype. The upper M3/ has a large t1 and no t3, the second lobe of the upper M3/ displays two cusps on the second lobe of the M3/ in two Malian specimens and only one cusp on the RCI in specimens from the locality of Bouna (Fig. 8). The MNHN karyotyped specimen attributed to L. macculus from CAR (locality of Bossangoa) have bunodont cusps, a small upper M3/ with one cusp on the second lobe, slightly stephanodont t4 on M12/, well fused cusps on the pre-lobe of M/1, a cv5 and a tma (Figs. 8, 9).

Multivariate morphometric analyses

The discriminant function analysis performed on four external measurements of 47 genetically typed specimens (not shown) indicates that there is substantial overlap among species and high intra-specific variability. With the exception of ear length, no other variable is well correlated with the two first discriminant axes. As a result, only 64% of the specimens are reliably classified.

A second discriminant function analysis performed on the 13 skull measurements for 36 genetically typed specimens (Fig. 10) allowed a better discrimination among species, with an 86% correct classification. On the 1-2 axis plane there is still overlap among the three species. The variable which best explained the separation between L. linulus and L. bellieri compared to L. striatus are LI13, LS13, HMDB (positively correlated with axis 1), CIO, FO, WOCC and LGT (negatively correlated with axis 1). LBT and WNAS separate L. linulus from L. bellieri along axis 2. The best classification score in this case is obtained for L. striatus (93%), then L. linulus (87%) and finally L. bellieri (60%). The two specimens incorrectly classified as L. bellieri were reattributed to L. linulus and not to L. striatus.

Number of species of Lemniscomys in Guinea

Based on published data, there are potentially four species of Lemniscomys in Guinea: L. striatus, L. linulus, L. bellieri and L. zebra. Our integrative study shows that the new Guinean specimens belong to two different mitochondrial clades: one assigned to L. striatus clade IV-1 by Nicolas et al. (2008), and the other corresponding with the L. linulus-bellieri-macculus clade. Our specimens also cluster in two morphological groups (multi-striped and mono-striped pelage, respectively) and with two different karyotypes (2N = 44 and NFa = 66, vs. 2N = 56 and NFa = 66, respectively), reinforcing the hypothesis that our sample comprises two species. However, assigning specific attributes to our specimens requires further analyses especially in the case of the second clade.

In our phylogenetic tree, L. linulus and L. bellieri are not recovered as monophyletic, and the genetic distance is higher between specimens of L. linulus from Guinea and Mali, than between specimens of L. linulus from Guinea and L. bellieri. Only one specimen of L. macculus was sequenced and its cytb sequence is closer to sequences of L. linulus from Guinea and L. bellieri than to those of L. linulus from Mali. In the Arvicanthini molecular phylogeny based on three mitochondrial genes (cytb, 16S, 12S), Ducroz et al. (2001) also found a robust clade including L. bellieri + L. macculus, and suggested that these two taxa are conspecific. Further studies including nuclear markers at the genome scale may better clarify the number of species in this complex (de la Harpe et al. 2017, Pedraza-Marrón et al. 2019). It is important to note that within the L. linulus-bellieri-macculus clade, different karyotypic formulae were recorded. Van der Straeten & Verheyen (1978) described a karyotype for L. bellieri from RCI of 2N = 56 and NF = 78, while in the type locality of Lamto a different karyotype (2N = 56, NFa = 60) was found (Volobouev et al. unpublished data, Table 1). The single known karyotype of L. linulus comes from south Mali (Mandingue Mountains, Baffing, 2N = 56, NFa = 64, Granjon & Duplantier 2009) and is slightly different from that observed in our study at two localities (Tanganya, Bantou: 2N = 56, NFa = 66). For L. macculus a specimen from CAR (Bossangoa) and others from Ethiopia displayed a karyotype of 2N = 56 and NFa = 64 (Ducroz et al. 2001, Bulatova et al. 2002). According to Ducroz (1998), L. bellieri and L. macculus differ only by one pericentric inversion. More detailed cytogenetic analyses are needed to draw conclusions on the biological consequences of these differences, particularly whether they represent karyotypically distinct sibling species or intraspecific polymorphism. Our discriminant analyses on skull distances showed poor discrimination between L. bellieri and L. linulus. For sibling species, it is recognised that rates of reliable classification are often relatively low. For example, for sibling species of the African murid genus Mastomys, the discrimination rate varies between 75 and 92 % (Denys et al. 2012). Even if these two species cannot be readily discriminated based on skull measurements they can be distinguished on external and skull morphology: the dorsal pelage has only one black stripe and no yellow lines in L. linulus while it has eight bands of dark brown/ grey alternating with obvious yellow spot lines in L. bellieri. The incisor foramen is narrower posteriorly in L. linulus while it is of equal width posteriorly and anteriorly in L. bellieri. Given these morphological differences between the two species we will consider them as distinct species and confirm the presence of L. linulus and L. striatus in Guinea.

The Faranah specimens attributed to L. bellieri by Ziegler et al. (2002) were not genetically typed but their skin, skull characters and dimensions fit well with the L. bellieri type specimen. This observation implies the presence of a third species of Lemniscomys in Guinea. Despite previous reports, we failed to collect L. zebra in savannah regions of Guinea during this study but it remains possible that the species is present in the north of the country.

Patterns of variability in West African Lemniscomys

Like for other African murid rodents, the west African Lemniscomys may present a complex of cryptic species that arises from complex biogeographical and historical patterns (Taylor 2000, Dobigny et al. 2002a, Castiglia 2013). Rosevear (1969) hypothesised that multi-striped species are derived compared to mono-striped taxa. In our phylogenetic tree the mono-stripe pattern may appear in different lineages (clade L. linulusbellieri-macculus and clade L. rosalia), a conclusion also supported by Castiglia et al. (2002).

High cytogenetic variability appears to be a feature of L. striatus in West Africa (Table 1). If in Guinea all specimens display the 2n = 44 and NFa = 66 karyotype (Denys et al. 2009, this study), the fundamental number of autosomes can vary for this species from 58 in RCI and Burkina Faso (Van der Straeten & Verheyen 1978, 1985, Ducroz 1998) to 72-74 or 68 in Benin (Capanna et al. 1997, Castiglia et al. 2002). Such polymorphism may reflect geographical structuring in this species accompanied by gene flow restrictions among populations, as already proposed based on cytb sequencing results by Nicolas et al. (2008).

Based upon our morphological and morphometrical study, we observe that Lemniscomys in West Africa is characterized by substantial variability in skull and dental morphology. However, in Guinea it appears that L. striatus has a larger skull and smaller molars compared to L. linulus. In his description of L. bellieri, Van der Straeten (1975) mentioned its large tympanic bulla and small size compared to L. striatus. These differences were not recovered in our measurements of the paratypes of L. bellieri or the karyotyped specimens from RCI. We found a large overlap in both external and skull measurements between the two species. In L. bellieri the tympanic bullae length varied between 4.8-5.9 mm against 4.5-5.9 mm in Guinean L. striatus. We did not identify any diagnostic characters for L. bellieri, such as a smaller hindfoot length and the Sm (or tma) on M/1. Thus, we cannot confirm here Van der Straeten & Verheyen's (1978) conclusion that L. striatus has a larger skull, on average, than L. linulus and L. bellieri.

West African Lemniscomys – diagnosis and identification key

Using new data from the present study, we provide a revised diagnosis of the poorly known L. linulus and confirm the extension of its distribution to Guinea. This new information allows us to present a new identification key for the West African species of the genus.

Following the initial description of Thomas (1910), we confirm his diagnosis of the species' external characters of: “grizzled greyish – buff general colour of the dorsal pelage becoming ochraceous posteriorly. Under surface is white, edge on each side with a pinkish-buff line. The fur is coarse, hairs of back have about 7 mm length. The ears and a spot above and below each eye are ochraceous buff. Dorsal black line commencing between the ears and running to the root of the tail. Hairs on each side of it lighter than the general colour. Hands and feet light pinkish buff. Tail dark above dull ochraceous laterally, dull buffy below”.

To this description we can add an amended diagnosis: tail equal or slightly longer than HB (%TL/HB: 98-132, N = 31, Mean: 114, Sd = 10.4). Nasals short and large ending obliquely from the median suture. Fronto-parietal crests well developed and rectilinear. Large braincase. Incisor foramen long and narrower posteriorly than anteriorly, ending between the t1 of the upper M1/. Lower M/1 with the antero-labial cusps small and transverse.

Identification key for West African Lemniscomys species:

Distribution and ecology of the Guinean Lemniscomys species

Based on molecular, cytogenetic and morphological analysis, we have been able to identify the specimens of Lemniscomys collected in Guinea. Table S1 provides specimen identifications for the Guinean localities we sampled between 2003 and 2011. We found L. linulus in six localities and L. striatus in 25 localities (Fig. 1). There are only three localities where both species were found together (Bantou, Ganya and Tanganya). Ganya and Tanganya are two localities situated in the Faranah region close to the location were Ziegler et al. (2002) recovered L. bellieri (Fig. 1).

Lemniscomys striatus was the only species present in the coastal lowland Guineo-Congolian forest block of Guinea and in upland deciduous forest and grassland regions. It was also trapped in bush and secondary forest enclosing grasslands in the Guinean savannah region. It can occur at an elevation of 1000 m on Mount Ziama and Nimba. According to Granjon & Duplantier (2009), this species occupies Guinean savannah, which we confirm here. We also encountered it in degraded forest habitats and anthropogenic habitats (cultivated areas), and the margins of enclosed grassland (Denys et al. 2009). It was collected by Ziegler et al. (2002) in Samaria grassland at the edge of the Mafu forest and on the eastern bank of the River Niger, corresponding with a Guinean savannah-type environment.

Lemniscomys linulus was the only species of the genus found in north Guinea in Sudanian savannah or at the ecotone between Sudanian and Guinean savannah at around 500 m altitude (Kodoko, Dinguiraye, Fressoudou; Fig. 1). We encountered it in Sudanian savannah and in cultivated areas close to villages. We did not collect it close to Senegal in the Guinean Niokolo-Koba buffer zone, which is the closest region to the type locality. According to Granjon & Duplantier (2011), L. linulus is also known in south Mali (Mandingues Mountains, Bafing) in the same type of habitat as in Guinea.

In central Guinea, L. linulus and L. striatus occurred in sympatry at three localities at about 500 m altitude in the transition zone between Guinean and Sudanian savannah (Fig. 1). It is in the same region (Faranah) that Ziegler et al. (2002) also found L. bellieri in sympatry with L. striatus. Despite regular ecological surveys in that region, we have been unable to trap L. bellieri. We unsuccessfully investigated essentially anthropogenic habitats but did encounter L. bellieri in the High Niger National Park, which supports more native vegetation and mosaic environments that are preferred habitats. The recent discovery of L. linulus in north Guinea may reflect a lack of previous studies in that region, though it is also possible that L. linulus has only recently arrived in central Guinea. As in the Sahelo-Sudanian countries of West Africa, global change (aridification) and deforestation may have caused a southward contraction of the Sudanian, Guinean and evergreen forest zones. No data are available for Guinea, but such changes in the biogeographic vegetation zones of west Africa has already been reported in Senegal (Bâ et al. 2006) and appear rapid. Rodents may represent informative indicators for monitoring these changes, especially in West Africa.