Tracking Chromosome Evolution in Southern African Gerbils Using Flow-Sorted Chromosome Paints

Desmodillus and Gerbilliscus (formerly Tatera) comprise a monophyletic group of gerbils (subfamily Gerbillinae) which last shared an ancestor approximately 8 million years ago; diploid chromosome number variation among the species ranges from 2n = 36 to 2n = 50. In an attempt to shed more light on chromosome evolution and speciation in these rodents, we compared the karyotypes of 7 species, representing 3 genera, based on homology data revealed by chromosome painting with probes derived from flow-sorted chromosomes of the hairy footed gerbil, Gerbillurus paeba (2n = 36). The fluorescent in situ hybridization data revealed remarkable genome conservation: these species share a high proportion of conserved chromosomes, and differences are due to 10 Robertsonian (Rb) rearrangements (3 autapomorphies, 3 synapomorphies and 4 hemiplasies/homoplasies). Our data suggest that chromosome evolution in Desmodillus occurred at a rate of ∼1.25 rearrangements per million years (Myr), and that the rate among Gerbilliscus over a time period spanning 8 Myr is also ∼1.25 rearrangements/Myr. The recently diverged Gerbillurus (G. tytonis and G. paeba) share an identical karyotype, while Gerbilliscus kempi, G. afra and G. leucogaster differ by 6 Rb rearrangements (a rate of ∼1 rearrangement/Myr). Thus, our data suggests a very slow rate of chromosomal evolution in Southern African gerbils.

drial and nuclear sequence data indicate that some of the lineages within Gerbilliscus cluster with Gerbillurus [Chevret and Dobigny, 2005;Colangelo et al., 2007;Granjon et al., 2012], suggesting a polyphyletic origin for the genera, which in turn prompted the recommendation for Gerbillurus to be considered a synonym of Gerbilliscus . Here, we adopt the name Gerbilliscus (subsuming Gerbillurus ) in recognition of the patterns suggested by the robust molecular phylogenies [Colangelo et al., 2007;Granjon et al., 2012]. In effect, this implies that Gerbilliscus contains 3 lineages: an eastern group (G. robusta, G. phillipsi, G. vicinus, G. nigricaudus) , a western group (G. gambianus, G. guinea, G. kempi) , and a southern group (G. brantsii, G. afra, G. leucogaster) , with G. paeba , G. setzeri and G. vallinus basal to the western + southern clade.
Here genome-wide comparisons of the Southern African D. auricularis, G. tytonis, G. leucogaster, G. afra , the West African G. kempi and the North African Psammomys obesus were undertaken in order to refine previous comparisons based on G-banding. In doing so, we aim at providing new insights into chromosomal rearrangements that distinguish these Southern African taxa. This was accomplished using G. paeba whole-chromosome painting probes (reported here for the first time) to identify regions of orthology, and to document genome repatterning among Desmodillus (D. auricularis), Gerbilliscus (G. paeba, G. tytonis, G. leucogaster, G. afra, G. kempi) and the outgroup species, P. obesus . Interpreting these differences in the context of the robust DNA sequencebased phylogeny [Chevret and Dobigny, 2005;Colangelo et al., 2005Colangelo et al., , 2007Granjon et al., 2012] allowed inferences on the mode and rate of chromosome evolution in Southern African gerbils.

Chromosome Banding and Karyotypes
Chromosome G-and C-banding followed the protocols of Seabright [1971] and Sumner [1972], respectively. Briefly, Gbanding of metaphase chromosome spreads required aging the slides for a minimum of 1 h at 65 ° C and treating aged slides with a trypsin solution (0.25% in 1× PBS) for a minimum of 30 s. Trypsin-treated slides were rinsed in calf serum buffer (containing 500 μl FCS and 50 ml 0.025 M KH 2 PO 4 buffer, pH 6.8) for 3 min followed by Giemsa staining (made in 0.025 M KH 2 PO 4 buffer; pH 6.8) to visualise chromosomes. C-banding entailed immersing slides in a 0.2 M HCl solution for 3 min, followed by incubation in a saturated Ba(OH) 2 solution for a minimum of 70 s at 55 ° C. A second incubation in 2× SSC for 30-60 min at 50-60 ° C (speciesdependent) followed. Chromosomes were visualised using 2% Giemsa staining and arranged according to karyotypes published by Qumsiyeh [1986a, b], Qumsiyeh and Chesser [1988] and Volobouev et al. [2007].
Flow-Sorting and Fluorescent in situ Hybridization G. paeba (2n = 36) was selected for flow-sorting primarily because it is a terminal taxon among the Southern African species in the molecular phylogeny [see Chevret and Dobigny, 2005]. Chromosomes were sorted on a MoFlo dual-laser cell sorter and isolated on size and base-pair composition ]. Biotin-16-dUTP-(Roche) or digoxigenin-11-dUTP-labelled painting probes were made by DOP-PCR amplification of flow-sorted chromosomes [Telenius et al., 1992]. Fluorescent in situ hybridization (FISH) experiments followed Rens et al. [2006] and Yang and Graphodatsky [2009] with minor modifications. Biotin-labelled probes were detected with the antibody Cy3-streptavidin (1: 500 dilution), and digoxigenin-labelled probes with FITC-conjugated anti-digoxigenin IgG made in sheep (1: 1,000). Slides were counterstained with 0.1 μg/ml DAPI (4 ′ ,6-diamidino-2-phenylindole) and mounted using Vectashield medium (Vector Labs). Images were captured using a CCD camera attached to an Olympus BX60 epifluorescence microscope equipped with DAPI, FITC and Cy3 fluorescence filters. Images were edited using the Genus software system (Applied Imaging Corp., Newcastle, UK).  Colangelo et al. [2005Colangelo et al. [ , 2007 and Granjon et al. [2012]. In effect, our tree emphasises the monophyly of Gerbilliscus + Gerbillurus (both referred to as Gerbilliscus ), the basal position of Desmodillus to the Southern African taxa, and the clustering of P. obesus outside the Southern African taxa, all well-established evolutionary associations. The latter species was used to polarize some of the chromosomal changes observed at the base of Southern African taxa.
Cross-species chromosome painting also revealed a striking resemblance between karyotypes of G. paeba and G. tytonis, confirming previous banding studies. All GPA chromosomes, including the Y chromosome, were retained in G. tytonis . In addition, the 2 species are very similar in other respects, notwithstanding their sympatric distribution in Namibia. Other aspects identical between the 2 species include quadrupedal saltation (adaptation for locomotion in sand), absence of sexual dimorphism, nocturnal activity and omnivorous diet [Griffin, 1990;Perrin et al., 1999a, b]. The overlap between the taxa is considerable. For example, G. paeba from the Namib Desert weighs on average 26.40 g (20-37 g), has a slightly tufted tail tip, and the hind foot is less than 30 mm (21-30 mm), while G. tytonis has an average mass of 24 g (maximum 35 g), a more pronounced, tufted tail, and narrow, hairy feet with hind foot length averaging 33.3 mm (28-36 mm) [Griffin, 1990;Perrin et al., 1999a, b;Chimimba and Bennett, 2005]. Such a pronounced overlap in morphology and habitat preference coupled to their invariant karyotypes suggests a recent divergence and further investigations may show them to be the same species.

Chromosomal Rearrangements and Homoplasy
Our chromosomal homology map established by FISH and banding comparison demonstrates that Rb rearrangements were important in chromosomal evolution and speciation in these gerbils ( figs. 2 , 6 ). Early G-band comparisons among gerbils have suggested that Rb rearrangements are homoplasic, and interpreting them in an evolutionary context may be problematic [Qumsiyeh et al., 1987] -a finding recently shown for Bovidae [Robinson and Ropiquet, 2011]. Using the gerbil sequence-based phylogeny as scaffold ( fig. 6 ), we show that of the 10 rearrangements ( table 2 ) identified by the GPA painting probes 1-6, 8, 10, 11, and 12, one is an asynapomorphy (GPA6) uniting G. paeba + G. tytonis , 2 are autapomorphies (GPA11 and 12)  (GPA1, 2, 3, 5) are potential homoplasies/hemiplasies . With respect to the first of these, GPA1, 2 possible hypotheses may be suggested for the observed patterns when the character is mapped to the tree in figure 6 . The first is that it is an example of hemiplasy, having undergone a fusion at ∼ 6 Mya (node C) and persisted as a polymorphism until the divergence of G. leucogaster and G. afra at ∼ 2.5 Mya. The alternate explanation (i.e. homoplasy) suggests that the fusion arose at node F (the common ancestor to G. tytonis and G. paeba ) and convergently so in G. leucogaster (i.e. requiring 2 changes vs. the one for hemiplasy). Although hemiplasy offers the most parsimonious solution in terms of rare genomic changes [Rokas and Holland, 2000], the required persistence time ( ∼ 3.5 Myr) is at the upper bound suggested by Robinson et al. [2008] and Robinson and Ropiquet [2011], and we regard homoplasy as the more likely of the 2 hypotheses. The same reasoning applies to the Rb rearrangements involving GPA2 and GPA3. A more convincing case for hemiplasy can be made with respect to GPA5 (arose at node C and persisted as a polymorphism for ∼ 2 Myr to D, at which point it was lost in the lineage leading to G. kempi ).
Consequently, the finding of Qumsiyeh et al. [1987] that homoplasy is present in the chromosomal evolution of gerbils is confirmed by the present study. Our results show that 4 of the 10 rearrangements identified are homoplasic when mapped to the tree (3 convergences/reversals and 1 probable hemiplasy, all of which contribute to the 'noise' evident in the species phylogeny). Qumsiyeh et al. [1987] on the other hand obtained 7 homoplasies in 15 Rb rearrangements among G. paeba, G. vallinus, G. robustus, G. nigricauda, G. leucogaster, G. afra , and G. brantsii . It should be noted, however, that the Qumsiyeh et al. [1987] study may have overestimated the number of homoplasic traits due to topological differences in the trees used (allozyme-based in Qumsiyeh et al. [1987], and mtDNA and nuclear sequences in our case) and, in some instances, to their misidentification of the chromosomes involved.

Rate of Chromosome Evolution
Interpretation of the rearrangements in the context of DNA sequence divergence times allows us to make inferences about the rate of evolution. Desmodillus accumulated 4 Rb rearrangements and 5 inversions since it separated from Gerbilliscus , reflecting a rate of 1.25 rearrangements/Myr (even taking the homoplasic rearrangement detected by GPA5 into account); G. kempi accumulated 3 Rb rearrangements following its split from the G. afra/G. leucogaster lineage 5.5 Mya, which translates into an evolutionary rate of less than 1 rearrangement/Myr. The 2 rearrangements separating G. leucogaster from G. afra accumulated over 2.5 Mya, representing a rate of approximately 1 rearrangement/Myr. On average, the rate of chromosomal evolution among Gerbilliscus over 8 Myr is ∼ 1.25 rearrangement/Myr (10 rearrangements over 8 Myr), which is considerably slower than values previously obtained in other gerbils (e.g. West African Taterillus have a rate of 45 rearrangements/Myr) [Dobigny et al., 2002. Some Gerbillus species show a rate of ∼ 12.3 rearrangements/Myr [Aniskin et al., 2006;Chevret and Dobigny, 2005]. Recently, it was shown that 6 representa-tive genera in the Rattini tribe displayed an evolutionary rate of 0.6-3.33 rearrangements/Myr, which was considered slow for murid rodents [Badenhorst et al., 2011]. Therefore, the rate of chromosome evolution in the Southern African gerbils studied here is unexpectedly slow.
In summary, we report the first genome-wide crossspecies chromosome painting in Southern African gerbilline rodents. The FISH data unequivocally establish homology among taxa from Desmodillus and Gerbilliscus which last shared an ancestor approximately 8 Mya. Our data also corrected some errors in published homeology maps that were based on banding results alone, and thus underscores the utility of FISH for inferring homology between species. Significantly, the present data demonstrates that the rate of chromosome evolution in Gerbilliscus is relatively slow in comparison to other gerbil lineages, which suggests evolutionary rate heterogeneity within these rodents. A more complete Zoo-FISH analysis including more representatives of the 15 genera comprising the subfamily Gerbillinae should provide more detailed assessment of chromosome evolution among these rodents.