Degeneration of the Olfactory System in a Murid Rodent that Evolved Diurnalism

Abstract

In mammalian research, it has been debated what can initiate an evolutionary tradeoff between different senses, and the phenomenon of sensory tradeoff in rodents, the most abundant mammalian clade, is not evident. The Nile rat (Arvicanthis niloticus), a murid rodent, recently adapted to a diurnal niche through an evolutionary acquisition of daylight vision with enhanced visual acuity. As such, this model provides an opportunity for a cross-species investigation where comparative morphological and multi-omic analyses of the Nile rat are made with its closely related nocturnal species, e.g. the mouse (Mus musculus) and the rat (Rattus norvegicus). Thus, morphological examinations were performed, and evolutionary reductions in relative sizes of turbinal bone surfaces, the cribriform plate, and the olfactory bulb were discovered in Nile rats. Subsequently, we compared multiple murid genomes, and profiled olfactory epithelium transcriptomes of mice and Nile rats at various ages with RNA sequencing. The results further demonstrate that, in comparison with mouse olfactory receptor (OR) genes, Nile rat OR genes have experienced less frequent gain, more frequent loss, and more frequent expression reduction during their evolution. Furthermore, functional degeneration of coding sequences in the Nile rat lineage was found in OR genes, yet not in other genes. Taken together, these results suggest that acquisition of improved vision in the Nile rat has been accompanied by degeneration of both olfaction-related anatomical structures and OR gene repertoires, consistent with the hypothesis of an olfaction-vision tradeoff initiated by the switch from a nocturnal to a diurnal lifestyle in mammals.

Introduction

Each animal species has developed specific anatomical structures and gene repertoires for their senses that are required to perceive relevant environmental cues. However, owing to the finite energy budget and high energy demand required for sensation and perception (Niven and Laughlin 2008; Okawa et al. 2008), it has been suggested that increased acuity in a particular sensory system can result in tradeoffs with other systems during evolution (Barton et al. 1995; Wu et al. 2018; Keesey et al. 2019; Oteiza and Baldwin 2021). For example, shortwave vision has diminished with the origination of high duty cycle echolocation in some bats (Zhao et al. 2009), and vomeronasal sensitivity has been lost with the origin of trichromatic color vision in catarrhine primates (Zhang and Webb 2003).

To choose the right time for a given response or activity in face of anticipated ecological interactions in a daily manner, animal species may adaptively evolve a diel rhythm and specialize in a few particular senses (Kronfeld-Schor and Dayan 2003). Vision and olfaction are 2 anciently originated animal senses (Feinberg and Mallatt 2013). An evolutionary tradeoff between vision and olfaction triggered by a shift in diel rhythm (diurnal to nocturnal, or vice versa) has been reported in Lepidopteran insects (Balkenius et al. 2006; Montgomery and Ott 2015; Stockl et al. 2016) and possibly in birds (Le Duc et al. 2015). For mammals, such a tradeoff was initially proposed to have occurred in primates, based on comparisons of brain structures (Barton et al. 1995; Barton and Harvey 2000). However, an investigation of olfactory receptor (OR) gene repertoires encoded in various primate genomes showed that when phylogeny and nose structure were controlled, the acceleration of OR genes loss was unrelated to the transition between diel activity patterns (Niimura et al. 2018). Correspondingly, an independent study of >70,000 OR genes from 58 mammalian genomes reported no correlation between diel activity patterns and OR gene family size changes (Hughes et al. 2018). These studies that assigned a fixed diel pattern for each of the analyzed species did not consider that the analyzed species could exhibit within-population variations or flexibility in diel patterns as previously noted (Ankel-Simons and Rasmussen 2008). Aside from that, owing to the loss of evolutionary signatures and transition lineages over time, it may be difficult to precisely infer ancient sensory/behavioral phenotypic changes and their timing. Consequently, further studies are needed to verify or reject whether a shift in diel pattern during mammalian evolution has caused any olfaction-vision tradeoffs. A detailed investigation of recent evolutionary changes in the sensory systems of mammals that have just undergone a change in diel pattern may provide an answer to this question.

Olfaction plays important roles in shaping proper physiology and behaviors of murid rodents (Muridae, Rodentia) (Slotnick 2001; Glinka et al. 2012; Bhattarai et al. 2022). The majority of murid rodents are nocturnal, and these include the house mouse (Mus musculus, dubbed “mouse” hereafter) (Waterston et al. 2002) and the Norway rat (Rattus norvegicus, dubbed “rat” hereafter) (Gibbs et al. 2004), 2 most widely used laboratory mammalian models. According to molecular dating of murid rodents with fossil calibrations (Aghova et al. 2018), the Nile rat (Arvicanthis niloticus) is phylogenetically closer to the mouse than the rat, and the median estimated divergence times of Arvicanthis–Mus and Arvicanthis–Rattus are 12.4 million years ago (MYA) and 12.6 MYA, respectively (Fig. 1) (Kumar et al. 2017). In contrast to the nocturnal activities of the mouse and the rat, the Nile rat is strictly diurnal (Blanchong and Smale 2000). Based on the documented phenotypes of representative murid species (Fig. 1; supplementary tables S1 to S3, Supplementary Material online), we found that during the evolution of the Nile rat lineage, the transition from nocturnalism to diurnalism could have begun between 9.5 and 11.2 MYA and completed by 6.2 MYA. Meanwhile, no apparent direction of dietary or living habitat evolution was observed in either the Nile rat lineage or the mouse lineage. Experiments based on retinal anatomy and optokinetic response have revealed that, to adapt to a diurnal niche, the Nile rat has evolutionarily acquired daylight vision with enhanced visual acuity (Gaillard et al. 2008) (also see supplementary fig. S1, Supplementary Material online). In order to examine the possible existence of a vision-olfaction tradeoff during recent mammalian evolution, we propose to examine whether degeneration of olfaction has occurred and been accompanied by enhanced vision during the recent evolution of the Nile rat. This was accomplished by employing both morphological and genomic approaches.

Fig. 1.

Phylogenetic and phenotypic information of the representative murid species used to determine when the Nile rat lineage switched from being nocturnal to diurnal. The information of the phylogeny and divergence dating is from Aghova et al. (2018). The 3 species at the terminals of red branches (the rat, the mouse, and the Nile rat) are those with near-complete genome assemblies at the chromosome level. Genes in the genomes of the 2 species (M. pahari and G. surdaster; at the terminals of blue branches) that were assembled using WGS reads were also utilized to enhance the resolution of a part of analysis. The information of phenotypic assignment of diel pattern, diet, and habitat for each representative species were described in supplementary tables S1 to S3, Supplementary Material online, respectively. Habitat: GD, ground-dwelling; F, fossorial; A, arboreal.

To examine the potential olfaction-related morphological divergences between the mouse and the Nile rat, skull and brain structures of these 2 murid species were compared by examining their (i) relative sizes of the turbinal bone surfaces where the olfaction epithelium is distributed, (ii) relative sizes of the cribriform plate (a part of the ethmoid bone) which allows the olfactory nerves to perforate the brain from the nasal cavity, and (iii) relative weights of the olfactory bulb to the total brain. The skull structures of their outgroup species (i.e. the rat) were also examined. With well-annotated data of several murid chromosome-level genome assemblies that are highly contiguous, evolutionary dynamics of OR gene repertoires were compared as well. Both (i) gene gain/loss events and (ii) coding sequence constraint of mouse OR genes versus Nile rat OR genes were examined. In addition, RNA sequencing (RNA-seq) experiments were performed to investigate the expression status of OR genes in the olfactory epithelium of the mouse and the Nile rat to infer the functionality of OR genes encoded in their genomes. As many speciation events have occurred since the Arvicanthis–Mus divergence, we further incorporated the annotated genomes of Gairdner's shrewmouse (Mus pahari) (Thybert et al. 2018) and African woodland thicket rat (Grammomys surdaster) (see Methods), both of which were assembled based on reads generated from whole-genome shotgun (WGS) sequencing, to enhance the resolution of the analyses when applicable (see Fig. 1 and below). The results obtained from the abovementioned analyses consistently suggest that degeneration of the olfactory system has occurred in Nile rats, thereby supporting the existence of a vision-olfaction tradeoff that corresponds to a shift in the temporal niche during rodent evolution.

Results and Discussion

Reduced Turbinal Bone Surface Area and Cribriform Plate Size of the Nile Rat

Two features of cranial structures have been used to estimate the olfactory capability of mammalian species: (i) the surface of the olfactory turbinals (Green et al. 2012; Martinez et al. 2018), and (ii) the size of the cribriform plate. To collect data for these 2 morphological features of the mouse and the Nile rat, we performed high-resolution X-ray micro-computed tomography (micro-CT). Both males and females aged 8 and 17 wk were surveyed. The age of puberty is approximately 6 wk in the mouse (C57BL/6J strain) (Dutta and Sengupta 2016) and 7 wk in the Nile rat (Delany and Monro 1985). Therefore, for both species, the data examined at 8 and 17 wk could represent the morphological features of the young adult and mature adult, respectively. For each age and gender of each species, 5 animals were sampled and used for statistical analyses.

Micro-CT data were used to calculate the total surface area of the nasal cavity (A_Nasal) of each sample. The mammalian nasal cavity has a complicated structure consisting of a scaffold of paper-thin bones, termed turbinals. Turbinals enhance the efficiency of respiration and olfaction, and are traditionally divided into 3 regions: maxilloturbinals, nasoturbinals, and ethmoturbinals (Moore 1981). The olfactory turbinals are part of ethmo- and nasoturbinals which are covered by the olfactory epithelium. To define the distribution boundary of the olfactory epithelium of each species of a given age, histology of the cross-sectioned snout of 8-wk- and 17-wk-old female mice and Nile rats was examined (supplementary figs. S2 and S3, Supplementary Material online, for the example data of a 17-wk-old mouse and a 17-wk-old Nile rat, respectively). This defined distribution boundary was applied to all of the corresponding age and species samples by assuming no gender differences. The surface area of the olfactory epithelium (A_OE) was calculated and subsequently normalized with the corresponding A_Nasal for each sample. No significant difference in the relative size of the olfactory turbinal surfaces (measured as A_OE/A_Nasal) was found between genders or ages of the same species (supplementary fig. S4A, Supplementary Material online). However, the Nile rats consistently exhibited smaller A_OE/A_Nasal values than the mice in all comparisons (Fig. 2, A and B).

Fig. 2.

Comparisons of 2 olfaction-related cranial features between the mouse, the Nile rat, and the rat. (A to D) The relative size of olfactory turbinal surfaces, measured as A_OE/A_Nasal; (E to H) the relative size of cribriform plates, measured as A_CP/A_Cran. (A to C, E to G) Values of upper quartile, median, and lower quartile are indicated in each box for (A, E) all (N = 20 for each box), or (B, F) a subgroup of a given gender/age (N = 10 for each box), of the mouse and Nile rat samples (mouse: left; Nile rat: right). Bars indicate semi-quartile ranges. P-values were determined with the Mann–Whitney U test. (C, G) Comparing 8-wk-old rat samples with mouse and Nile rat samples of the same age (N = 10 for each box) (mouse: left; rat: middle; Nile rat: right). (D, H) Processed micro-CT images for measurement of A_OE/A_Nasal and A_CP/A_Cran, respectively. (D) Dark purple blue, olfactory turbinals; light blue, remainder of nasal cavity. (H) Yellow, cribriform plate; dark orange, remainder of cranial cavity. 8w, 8-wk-old; 17w, 17-wk-old.

The cribriform plate, a part of the ethmoid bone, is a sieve-like structure between the nasal cavity and the anterior cranial fossa that is perforated by olfactory nerves from the roof of the nasal cavity to the olfactory bulb of the brain. The relative size of the cribriform plate has been demonstrated as a readout of olfactory capacity across modern and extinct mammalian species (Bird et al. 2018, 2020). Therefore, based on the micro-CT image data collected, the surface area of the cribriform plate (A_CP) was measured within the cranial cavity and then normalized to the total surface area of the cranial cavity (A_Cran) to obtain A_CP/A_Cran for each of the samples examined. No significant differences in A_CP/A_Cran were observed between genders or ages of the same species (supplementary fig. S4B, Supplementary Material online). However, the Nile rats consistently exhibited smaller A_CP/A_Cran values than the mice in all comparisons (Fig. 2, E and F). Thus, the morphological data from micro-CT scans of skull morphology, specifically the relative surfaces of the olfactory turbinals and cribriform plate, consistently indicate that the Nile rat has an inferior olfactory capacity compared to the mouse.

We performed an additional analysis of the nasal and cranial structures of 8-wk-old rats (BN strain) (Fig. 2, D and H; the corresponding histology data are shown as supplementary fig. S5, Supplementary Material online). Our results showed that while rats do not differ from mice in A_CP/A_Cran or A_OE/A_Nasal, they do have larger values than Nile rats for both A_CP/A_Cran or A_OE/A_Nasal (Fig. 2, C and G). This indicates that the differences in A_CP/A_Cran and A_OE/A_Nasal values between the mouse and the Nile rat are due to a reduction in these values in the Nile rat lineage, rather than an increase in the mouse lineage, after the 2 species diverged.

Reduced Weight of Olfactory Bulbs in Nile Rats

The olfactory bulb is a forebrain structure that receives and processes the neural input about odors detected by neurons of the olfactory epithelium in the nasal cavity. The relative size (Healy and Guilford 1990; Zelenitsky et al. 2009) or weight (Ribeiro et al. 2014; Lundrigan et al. 2020) of the olfactory bulb has been used to infer the olfactory capability of vertebrate species. Accordingly, we measured the absolute weight of the olfactory bulb (W_OB) and the weight of the nonolfactory bulb area of the total brain (W_Br) in each of 211 Nile rats of various ages. For the mouse, published data of W_OB and W_Br were obtained from the authors of a previous study (Williams et al. 2001).

A correlation analysis of the Nile rat data showed that W_Br (and W_OB) are generally determined by body weight (supplementary fig. S6, A and B, Supplementary Material online), rather than by age (supplementary fig. S6, C and D, Supplementary Material online). Moreover, consistent with previously reported mouse data (Williams et al. 2001), W_Br and W_OB were observed to be strongly and linearly correlated (Pearson's correlation coefficient r = 0.933, P < 10⁻⁹³). Because no statistically significant difference in the relationship of W_Br versus W_OB was observed according to gender (supplementary fig. S7, Supplementary Material online), the data from both genders were pooled for subsequent analyses. When W_OB versus W_Br data for both the mouse and the Nile rat were compared, a smaller W_OB was observed for Nile rats than for mice, both before and after W_Br was controlled (Fig. 3, A and B). To determine whether this observed difference was due to technical issues resulting from the Nile rat data and mouse data being generated in independent studies, we further measured W_Br and W_OB of 53 mice (C57BL/6J strain). Larger W_OB values were observed for the mice than for the Nile rats both before and after controlling for W_Br (supplementary fig. S8, Supplementary Material online). Taken together, these data confirm that the Nile rat has a relatively smaller olfactory bulb weight than the mouse, suggesting that the Nile rat has an inferior olfactory capacity compared to the mouse.

Fig. 3.

Comparisons of olfactory bulb weights between the mouse and the Nile rat. (A) Linear regression of olfactory bulb weights (W_OB) versus total brain weights (W_Br): white circles and the upper regression line represent the mouse data; gray circles and the lower regression line represent the Nile rat data. The gray shaded area represents the 95% confidence interval of the corresponding regression line. (B) Comparison of W_OB/W_Br before (left) and after (right) control of W_Br. The mouse data used in (A) and (B) are from (Williams et al. 2001). (B) Bars indicate semi-quartile ranges. P-values were determined with the Mann–Whitney U test. (C) The sagittal plane illustrates placement of the dissection cuts for the Nile rat (left) and the mouse (right) to determine W_OB and W_Br. Positions of the dissection cuts are represented by black bars. Scale bars are indicated.

Increased Frequency of OR Gene Loss During Nile Rat Evolution

Animals detect environmental odor molecules using ORs, a type of G-protein-coupled receptors (GPCRs), expressed in the olfactory epithelium (Buck and Axel 1991; Touhara et al. 1999; Nei et al. 2008). Across mammalian species, the sizes of OR gene repertoires in genomes have been found to vary considerably, and they are positively correlated with cribriform plate sizes (Bird et al. 2018). In several previous studies, evolutionary dynamics of OR gene repertoire sizes have been used to estimate changes in olfactory capability over mammalian evolution (Niimura and Nei 2005; Hughes et al. 2018; Niimura et al. 2018).

An accurate assessment of the evolutionary dynamics of gene families relies heavily on the availability of high-quality, finished genome assemblies (Church et al. 2009). Murid genome assemblies that could be considered finished, or nearly finished with only few gaps, include those of the mouse (Church et al. 2009), the rat (Howe et al. 2021), and the Nile rat (Toh et al. 2022). The genome assemblies of the white-footed mouse (Peromyscus leucopus, Cricetidae, Rodentia; Long et al. 2019), and 3 primates including the common marmoset (Callithrix jacchus; Yang et al. 2021), the rhesus macaque (Macaca mulatta; Zimin et al. 2014), and the human (Homo sapiens; Nurk et al. 2022), have also been finished or nearly finished. The full sets of protein-coding genes in the genome assemblies of these species have been annotated (see Methods). Based on the available data, we identified 1,353, 1,143, 1,070, 1,122, 290, 310, and 401 OR genes from 23,139, 22,469, 22,330, 22,198, 22,078, 21,761, and 19,813 annotated protein-coding genes in the rat, mouse, Nile rat, white-footed mouse, common marmoset, rhesus macaque, and human genomes, respectively (see Methods). While the annotated genes of the Nile rat represent a 0.6% reduction compared with the mouse ([22,469 − 22,330]/22,469 = 0.6%), the reduced proportion of OR genes was significantly greater ([1,143 − 1,070]/1,143 = 6.4%) (P < 10⁻³⁰, 2 proportions z-test). We expected that a significant proportion of OR genes that have been recently lost have genomic relics that are present in the form of pseudogenes. Using PseudoPipe (Zhang et al. 2006), we identified 57 and 109 pseudogenes originating from frameshift mutations or premature stop codons in duplicated members of OR gene families in the mouse and Nile rat genomes, respectively (see supplementary table S4, Supplementary Material online, and Methods). An examination of syntenic chromosomal regions between the mouse and the Nile rat showed numerous regions in which the OR coding DNAs were intact in the mouse but pseudogenized in the Nile rat to confirm Nile rat specific OR gene losses (supplementary fig. S9, Supplementary Material online). The scaffold-level genome assembly of G. surdaster, on the other hand, has a similar number of OR genes (1,080) compared to the Nile rat genome (1,070); however, this WGS-based G. surdaster genome assembly has a shorter sequenced length, fewer total annotated genes, and fewer OR-derived pseudogenes (supplementary table S4, Supplementary Material online). This suggests that the number of intact OR genes identified from a WGS-based scaffold-level genome assembly may have been underestimated due to a lower genome quality. Therefore, we avoided the use of unfinished genomes in examining the patterns of OR gene gain and loss.

To compute gene loss/gain events for various gene families during murid evolution, we subsequently applied the OrthoFinder algorithm (Emms and Kelly 2015) to identify putative orthologous gene sets of the analyzed species (termed “ortholog sets” hereafter). The use of genomic data from species that diverged from the Nile rat before the Mus–Arvicanthis divergence allows tracing gene gain/loss events and sequence changes that have occurred specifically in the mouse and Nile rat lineages. As the full set of genes of a putative ortholog set may not be monophyletic, a method described previously (Niimura et al. 2014) was used to separate each ortholog set into subgroups of genes, with each subgroup forming a clade (see Methods). The Duplication-Transfer-Loss (DTL) reconciliation algorithm implemented in RANGER-DTL 2.0 (Bansal et al. 2018) was used to calculate gene gain and loss events after primate–rodent divergence for every subgroup (see Methods), and the events of subgroups belonging to the same ortholog set were then summed up for each subgroup of ortholog set. We focused our initial comparison on the 4 branches of the phylogenetic tree of the 3 murid species that were examined (the branches “a” and “b” represent the evolutionary durations of the Nile rat and the mouse, respectively, after the divergence of these 2 species; the branch “c” represents the duration of the outgroup lineage after its divergence from the common ancestor of the mouse and the Nile rat; the branch “d” represents the evolutionary duration of the common ancestor of the mouse and the Nile rat between the time of its divergence from rat and the time of mouse–Nile rat divergence; supplementary figs. S10 and S4A, Supplementary Material online). Although the ratio of the loss events to the gain events (G_Loss/G_Gain) of OR genes of the Nile rat (“a”) branch (218/161 = 1.354) was greater than G_Loss/G_Gain of the mouse (“b”) branch (169/185 = 0.914) (supplementary fig. S10, Supplementary Material online) (G_Loss/G_Gain = 1 when a gene family is neither expanding nor contracting), it was noticed that the OR genes in the genome of the common ancestor of murids in supplementary fig. S10, Supplementary Material online were estimated to be >1,100, a number greater than previously reported (∼940) (Niimura and Nei 2007; Niimura et al. 2014). This disparity in estimated OR gene numbers in ancestral murid genomes may be partly due to differences in the choice of taxa and the quality of their corresponding genomes analyzed between studies. However, the main cause could be the short time gap between the divergence of Arvicanthis–Mus (12.4 MYA) and Arvicanthis–Rattus (12.6 MYA) to generate stochastic errors in constructing an incorrect gene tree with high bootstrap values. This may have led to an overestimation of the number of ancestral OR genes and, as a result, an overestimation of gene loss events during tree reconciliation. To solve this issue, we recalculated gene gain/loss events by replacing the genomic data of the rat with that of the white-footed mouse, which diverged from Arvicanthis around 26.1 MYA (supplementary fig. S1, Supplementary Material online). The phenomenon of overestimated numbers of OR genes in the ancestral murid genome disappeared after this replacement of the outgroup species (Fig. 4A). More importantly, the regenerated result showed that G_Loss/G_Gain of the branch “a” (88/178 = 0.494) became at least 85% higher than G_Loss/G_Gain values of the branch “b” (59/222 = 0.266), branch “c” (86/332 = 0.259), and branch “d” (36/140 = 0.257) (Fig. 4A). These results suggested an increased rate of OR gene loss specifically in the Nile rat lineage after the divergence of the mouse and the Nile rat. Specifically, the Nile rat lineage has 44 fewer OR gain events and 29 more OR loss events than the mouse lineage according to Fig. 4A, which has caused a >5% difference in OR gene family size between the mouse and the Nile rat.

Fig. 4.

Gain/loss events and expression reduction of OR genes. Nile rat OR genes have experienced: (A) less frequent gain and more frequent loss, and (B) more frequent expression reduction, compared with mouse OR genes. (A) Numbers indicated in the black boxes represent the number of OR genes of the species indicated; the numbers in the gray boxes show the numbers of OR genes of the ancestral taxa that were estimated. The 4 focal branches (not drawn to scale) are labeled with “a” to “d” as indicated. The numbers after the “+” and “−” symbols represent the estimated numbers of branch-specific gene gain and loss events that have occurred, respectively. (C) According to transcriptomes of the olfactory epithelium that were profiled, the numbers of expression reduction events (to below the TPM threshold of 1), that were inferred to occur in branches “a”, “b”, and “d” according to the maximum parsimony principle (see Supplementary Dataset S1, Supplementary Material online), were calculated and are indicated. The white-footed mouse (P. leucopus) was used as the outgroup species of the Nile rat (A. niloticus) and the mouse (M. musculus) for the inference of gene gain/loss events.

In order to determine if the increased frequency of OR gene loss in the Nile rat could be explained by species-specific genomic background, we conducted an analysis similar to that in Fig. 4A on non-OR GPCR genes, excluding V1R (Miller et al. 2020) and V2R (Yang et al. 2005) genes known to duplicate highly during mouse evolution lineage-specifically. The patterns of gene gain/loss of the OR genes and non-OR GPCR genes were compared to control for the factor of molecular function. Our analysis revealed that the relative frequency of gene loss events for non-OR (and non-V1R, non-V2R) GPCR genes was actually slightly lower in the Nile rat (“a”) branch than in the mouse (“b”) branch, with G_Loss/G_Gain values of 40/70 = 0.571 and 38/57 = 0.667, respectively (supplementary fig. S11, Supplementary Material online). The inconsistent patterns between OR genes (Fig. 4A) and non-OR GPCR genes (supplementary fig. S11, Supplementary Material online) indicated that the observation of relatively fewer gain events and more loss events that is specific to Nile rat OR genes is not an artifact contributed by the variation of genomic background among species.

More Frequent Transcriptional Reduction of OR Genes During Nile Rat Evolution

A gene's transcriptional activity implies its function in expressed tissues (Liao and Weng 2015). To compare the expression status of OR genes, RNA-seq was performed to obtain the olfactory epithelium transcriptomes of 6 Nile rats and 6 mice (1 male and 1 female aged 8, 12, and 16 wk for each species) (see Methods and supplementary table S5, Supplementary Material online, for statistics regarding the RNA-seq data generated). Generating transcriptome data profiled with RNA-seq also facilitated accurate estimations of gene family size (Denton et al. 2014). The transcriptional abundances of genes were measured in TPM (transcript per million) and were normalized (see Methods).

Among all “rat–mouse–Nile rat” ortholog sets, 14,206 had a 1:1:1 relationship (for any of these 1:1:1 ortholog sets, no gene duplication or gene loss events are observed after the divergence of the 3 murids). Differences in normalized TPM were calculated between 1-to-1 orthologs of the mouse and the Nile rat as ΔTPM_An-Mm = TPM_An − TPM_Mm, where TPM_An and TPM_Mm represent the average normalized TPM measured for the 6 samples of the Nile rat and mouse genes, respectively. The OR genes exhibited significantly lower ΔTPM_An-Mm values than the non-OR genes (P < 10⁻², U test). In addition, the values of ΔTPM_An-Mm for the OR genes tended to be negative (P = 0.03), and this tendency was not observed for the non-OR genes (P = 0.85) (1-sided sign tests) (supplementary fig. S12, Supplementary Material online). We used 4 different thresholds of TPM to define presence of expression (>0, >0.5, >1, and >2). Regardless of the TPM threshold used, the mouse consistently showed a greater number of expressed ORs than the Nile rat across various conditions, and also when different criteria were applied to define the presence of expression (Table 1). We further observed that when a larger TPM threshold was used, a larger difference in the percentage of expressed ORs of mice over Nile rats was generally observed (Table 1). This result implies that the Nile rat has fewer genes encoding ORs, as well as a greater proportion of ORs that are expressed at a low level (or in few cells), compared with the mouse.

Table 1.

Numbers (and %) of OR genes expressed at an abundance level greater than a given normalized TPM threshold

TPM	Age and sex of the source individual						Meanb	∩c	∪d
	8 wk		12 wk		16 wk
	Ma	Fa	Ma	Fa	Ma	Fa
Mus musculus
>0	1,111 (97.2%)	1,105 (96.7%)	1,117 (97.7%)	1,118 (97.8%)	1,114 (97.5%)	1,110 (97.1%)	1,131 (99.0%)	1,088 (95.2%)	1,131 (99.0%)
>0.5	939 (82.2%)	945 (82.7%)	951 (83.2%)	962 (84.2%)	899 (78.7%)	916 (80.1%)	949 (83.0%)	862 (75.4%)	989 (86.5%)
>1.0	767 (67.1%)	828 (72.4%)	845 (73.9%)	821 (71.8%)	729 (63.8%)	761 (66.6%)	801 (70.1%)	678 (59.3%)	893 (78.1%)
>2.0	526 (46.0%)	607 (53.1%)	607 (53.1%)	591 (51.7%)	488 (42.7%)	520 (45.5%)	562 (49.2%)	434 (38.0%)	678 (59.3%)
Arvicanthis niloticus
>0	1,048 (97.9%)	1,030 (96.3%)	1,042 (97.4%)	1,022 (95.5%)	1,042 (97.4%)	1,031 (96.4%)	1,057 (98.8%)	985 (92.1%)	1,057 (98.8%)
>0.5	883 (82.5%)	915 (85.5%)	871 (81.4%)	875 (81.8%)	878 (82.1%)	845 (79.0%)	899 (84.0%)	755 (70.6%)	967 (90.4%)
>1.0	704 (65.8%)	785 (73.4%)	705 (65.9%)	759 (70.9%)	727 (67.9%)	674 (63.0%)	740 (69.2%)	574 (53.6%)	868 (81.1%)
>2.0	440 (41.1%)	530 (49.5%)	473 (44.2%)	532 (49.7%)	481 (45.0%)	454 (42.4%)	498 (46.5%)	352 (32.9%)	636 (59.4%)
Difference (M. musculus over A. niloticus)
>0	63 (−0.7%)	75 (0.4%)	75 (0.3%)	96 (2.3%)	72 (0.1%)	79 (0.8%)	74 (0.2%)	103 (3.1%)	74 (0.2%)
>0.5	56 (−0.4%)	30 (−2.8%)	80 (1.8%)	87 (2.4%)	21 (−3.4%)	71 (1.2%)	50 (−1.0%)	107 (4.9%)	22 (−3.9%)
>1.0	63 (1.3%)	43 (−0.9%)	140 (8.0%)	62 (0.9%)	2 (−4.2%)	87 (3.6%)	61 (0.9%)	104 (5.7%)	25 (−3.0%)
>2.0	86 (4.9%)	77 (3.6%)	134 (8.9%)	59 (2.0%)	7 (−2.3%)	66 (3.1%)	64 (2.6%)	82 (5.1%)	42 (−0.1%)

^aM, male; F, female.

^bExpression abundance determined by averaged TPM of 6 samples of the same species.

^cIntersection of OR genes expressed above the given threshold.

^dUnion of OR genes expressed above the given threshold.

Based on the olfactory epithelium transcriptome data profiled with RNA-seq, we considered weak expression of TPM < 1 as a signal of degeneration of OR gene function. We examined OR orthologous sets in which at least 1 mouse OR gene or Nile rat OR gene has a value of TPM < 1. By combining information from OR gene trees and this expression data, we assigned the number of expression reduction events (from TPM ≥ 1 to TPM < 1) of OR genes along branches after the divergence of Mus–Arvicanthis in each tree according to the principle of maximum parsimony (see supplementary Dataset S1, Supplementary Material online for the full record of expression reduction event assignment). Frequencies of OR gene expression reduction events at the branches “a”, “b”, and “d” were counted. Despite the mouse genome containing a greater number of OR genes than the Nile rat genome, 161 and 158 expression reduction events of OR genes were identified in the Nile rat (“a”) branch and the mouse (“b”) branch, respectively (Fig. 4B). Hence, in comparison with OR gene evolution in the mouse, the Nile rat OR genes not only experienced less frequent gains and more frequent losses, but also more frequent expression reductions during their evolution.

Less Constrained Protein Evolution for ORs in Nile Rats

Gene duplication events influence the sequence evolution of protein-coding genes (Kondrashov et al. 2002; Pegueroles et al. 2013). To investigate the selective constraint imposed on coding sequences of OR genes and other genes after removing potential confounding effects from gene duplications, we firstly analyzed the 14,206 “rat–mouse–Nile rat” 1:1:1 orthologs sets. Due to the occurrence of many speciation events after the Mus–Arvicanthis divergence (Fig. 1), genes in 2 WGS-based genomes of M. pahari and G. surdaster were also incorporated and analyzed by OrthoFinder algorithm to generated a second group of 7,286 1:1:1:1:1 ortholog sets of the 5 species (rat, mouse, M. pahari, Nile rat, and G. surdaster), in order to provide an enhanced resolution of the analysis. Each group of ortholog sets were classified into 2 categories: those exclusively composed of OR genes (3-species 1:1:1 OR ortholog sets, n = 365; 5-species 1:1:1:1:1 OR ortholog sets, n = 278) and others (3-species 1:1:1 non-OR ortholog sets, n = 13,841; 5-species 1:1:1:1:1 non-OR ortholog sets, n = 7,008) (see Methods). A maximum-likelihood method (Yang 2007) was used to estimate the branch-specific rate of nonsynonymous changes per nonsynonymous site (d_N), the rate of synonymous changes per synonymous site (d_S), and the d_N/d_S ratio of each of these ortholog sets. With the outgroup information provided by the rat genes, we compared d_N/d_S of the branch “a” (the Nile rat branch) versus the branch “b” (the moue branch) in the 3-species analysis for each ortholog set (Fig. 5, A to D). In the 5-species analysis, we compared d_N/d_S of the branch “a_A” (the ancient Nile rat branch, i.e. the branch after the Arvicanthis–Rattus divergence and before A. niloticus–G. surdaster divergence), the branch “a_R” (the recent Nile rat branch, i.e. the branch after A. niloticus–G. surdaster divergence), the branch “b_A” (the ancient mouse branch, i.e. the branch after Mus–Rattus divergence and before M. musculus–M. pahari divergence), and the branch “b_R” (the recent mouse branch, i.e. the branch after M. musculus–M. pahari divergence) for each ortholog set, as shown in Fig. 5. According to Fig. 1, the temporal niche switch from nocturnalism to diurnalism of the Nile rat could have been initiated in the branch “a_A”, and completed in the branch “a_R”. If degeneration of coding sequence of ORs has occurred following the temporal niche switch, we expect to see d_N/d_S values of ORs on branches after Mus–Arvicanthis divergence to be “a > b” in the 3-species analysis, and “a_R > a_A ≥ b_A ≈ b_R” in the 5-species analysis.

Fig. 5.

Relaxed selective constraint on OR genes during Nile rat evolution. (A) and (E) show the phylogeny of the 3 and 5 investigated murid species, respectively. The branch lengths of the phylogenetic tree are not drawn to scale. Violin plots of d_N/d_S for (A to D) the branches “a” versus. “b”, and (E to H) the branches “a_A” versus. “a_R” versus. “b_R” versus. “b_A”, were used for comparisons of (B, F) OR genes, (C, G) non-OR genes, and (D, H) non-OR GCPR genes. P-values were determined with the Mann–Whitney U test and are associated with arched lines that indicate the branches that were compared. P-values < 0.05 are marked in red.

The 3-species analysis identified a significantly higher d_N/d_S value for the branch “a” compared with the branch “b” for the 1:1:1 OR ortholog sets (Fig. 5B, P = 0.016; U test). In contrast, this difference was not observed for the non-OR ortholog sets (Fig. 5C, P = 0.934; U test). G-protein-coupled receptors (GPCRs) are reported to have undergone a more rapid evolution due to the action of positive selection (Kosiol et al. 2008; Daub et al. 2017). From non-OR ortholog sets, we further defined 304 non-OR GPCR ortholog sets, each of which was comprised of genes encoding GPCRs, yet not ORs. Focusing on these non-OR GPCR ortholog sets, there was no difference observed in d_N/d_S between the branch “a” and the branch “b” (Fig. 5D, P = 0.543; U test). The 5-species analysis based on 1:1:1:1:1 OR ortholog sets showed that the median d_N/d_S value of the branch “a_R” was the highest among the 4 branches that were compared, and was statistically greater than d_N/d_S values of both the branches of “b_A” (P = 0.018; U test) and “b_R” (P = 0.007; U test) (Fig. 5F). This pattern was neither observed in the analysis of non-OR ortholog sets (only “a_A vs. b_A” and “a_A vs. a_R” branches were statistically different, due to decreased median d_N/d_S of the branch “a_A”; Fig. 5G) nor the analysis of non-OR GPCR ortholog sets (no statistical difference was detected between any pair of branches; Fig. 5H), indicating that the elevated d_N/d_S value of Nile rat OR genes was associated with the transition from being diurnal to nocturnal during the evolution of this lineage. The results of non-OR and non-OR GPCRs, respectively, showed that the higher d_N/d_S value for the Nile rat OR genes was neither due to systematic errors associated with sequencing and assembly of the Nile rat genome nor due to intrinsic evolutionary properties of GPCRs. We repeated the analysis of Fig. 5 by replacing the rat genomic data with the white-footed mouse genomic data, and obtained a result that was consistent with Fig. 5 to reach the same conclusion (supplementary fig. S13, Supplementary Material online).

Both positive selection and relaxed purifying selection can lead to an elevation of d_N/d_S. According to our data, there was no d_N/d_S value greater than 0.6 observed for any branch of “a”, “a_A”, or “a_R” in any OR ortholog set. Moreover, the median value of the Nile rat branch-specific d_N/d_S for the orthologs with ΔTPM_An-Mm < 0 (median d_N/d_S = 0.141) was found to be slightly greater than that of the orthologs with ΔTPM_An-Mm > 0 (median d_N/d_S = 0.133), although this difference was not statistically significant (P = 0.51, U test) (supplementary fig. S14, Supplementary Material online). Taken together, although the possible involvement of positive selection on a few Nile rat OR genes could not be completely excluded, these results suggest that the elevated d_N/d_S value for Nile rat ORs tends to be associated with reduced expression. Although that lowly expressed genes tend to evolve rapidly (have high d_N/d_S values) is a widely observed phenomenon whose causes have been reviewed elsewhere (Zhang and Yang 2015), positively selected OR genes are not expected to exhibit reduced expression in the presumed functioning tissue (i.e. olfactory epithelium) because of their increased ecological importance to the organism during evolution. Hence, it appears that the coding sequences of Nile rat OR genes have experienced relaxed selective constraints in the past largely due to degeneration of gene functions.

Concluding Remarks

In this study, we tested whether a vision-olfaction tradeoff is associated with a shift in diel pattern during mammalian evolution. Although this tradeoff was previously proposed in a study that compared brain substructures of rodents (Lundrigan et al. 2020), its experimental design did not incorporate information from any outgroup species that is phylogenetically closed to the 2 species that were compared, making lineage-specific changes difficult to discern. In the present study, our morphological investigations demonstrate that the Nile rat has relatively smaller surface areas of olfaction turbinals (A_OE/A_Nasal; Fig. 2, A to C) and cribriform plates (A_CP/A_Cran; Fig. 2, D to F), as well as a smaller size of the olfactory bulb (Fig. 3), compared with its closely related nocturnal species. Behaviors (van der Staay and Blokland 1996) and potentially anatomy can vary between strains of laboratory mice and rats. Our analysis found no difference in A_OE/A_Nasal and A_CP/A_Cran between C57BL/6 strain mice and BN strain rats. However, an previous study, using Fischer-344 strain rats and B6C3F1 strain mice and a different methodology, found that the percentage of the nasal cavity covered by olfactory epithelium is greater in rats than in mice (Gross et al. 1982). Hence, the trend in nasal structure, along with the lowest number of OR gene counts among all murids studied here, suggests that the smaller sizes of olfaction-related anatomical structures observed in the Nile rats than the mice are derived rather than ancestral.

It is important to note that even though the Nile rat lineage has experienced a greater loss and lesser gain of OR genes compared to the mouse lineage, the difference in OR gene counts between these 2 species may be smaller than the variation within nocturnal murid species, such as rats and mice (Fig. 4A). Additionally, despite all 3 primate species used in the analysis (human, macaque, and marmoset) being diurnal, their OR gene counts vary substantially (as shown in Fig. 4A). There are 2 potential explanations for this, and they are not mutually exclusive. Firstly, the number of OR genes in mammals can fluctuate due to various reasons, with the switch of temporal niche being only 1 of many causes. Second, the loss of OR genes in the Nile rat is an ongoing process and it may take more time for their OR gene repertoire to decrease to a more appreciable level. Considering that decreased expression and relaxed constraint of the coding sequence are signs of functional degeneration that may eventually lead to gene loss, the results of OR gene expression and coding sequence evolution are consistent with the second explanation, as we found that the Nile rat exhibits reduced expression of OR genes more frequently than the mouse after their divergence. Additionally, OR genes in the Nile rat lineage have experienced relaxed selective constraints, which was not observed for other genes. The combination of morphological data and molecular data thus confirms the degeneration of the olfactory system in the Nile rat.

Dietary transition from frugivory to folivory has been linked to a radical loss of OR genes in the ancestor of colobine primates (Niimura et al. 2018), which predominantly consume young leaves or unripe fruit/seeds (Davies 1995). Although Nile rats are more herbivorous than mice and rats (Blanchong and Smale 2000), seeds and insects constitute up to ∼48% and ∼38% of the food content of Nile rats captured in natural fields during the wet and dry seasons, respectively (Rabiu and Rose 1997). In addition, it has been shown that Nile rats exhibit food preferences similar to those of the Rattus species (Suliman et al. 1984). Therefore, it appears that the senses to detect food types are nearly equivalent between Nile rats and their closely related nocturnal species. If there is a contribution of dietary shift to olfaction degeneration in Nile rats, this contribution should be minor. Acquired daylight vision (Gilmour et al. 2008) and improved visual acuity (Gaillard et al. 2008) to adapt to a diurnal niche represent 2 major evolutionary changes that have occurred in the Nile rat, and these likely account for the olfaction degeneration observed in this species. Thus, the results of our present study are consistent with the hypothesis that a tradeoff between the sensory systems of the Nile rat has accompanied the evolution of a diel pattern during murid evolution.

Methods

Animals

All animal breeding and experimental procedures were approved by the Institutional Animal Care and Use Committee of the National Health Research Institutes (NHRI) (Approval No. IACUC-109074M2-A and IACUC-109118-AE). The Nile rats used were descendants of 2 breeding pairs of Nile rats derived from the colonies maintained by the KC Hayes Laboratory at the Brandeis University. The mice (C57BL/6; wild-type) and rats (BN/SsNNarl) used were obtained from the National Laboratory Animal Center, Taiwan. The animals were maintained in the Laboratory Animal Center of the NHRI under 12 h/12 h (mice and rats) and 11 h/13 h (Nile rats) light/dark cycles and were provided food (mice and rats: MFG, Oriental Yeast Co. Ltd., Tokyo, Japan; Nile rats: LabDiet 5326, LabDiet Ltd., St. Louis, MO, USA) and water ad libitum. The animals were euthanized by CO₂ inhalation prior to sacrifice for the following experiments.

Histological Analysis of Nasal Cavities

Whole heads were obtained from 8-wk- and 17-wk-old female mice and Nile rats. Those of 8-wk-old female rats were also obtained. After the lower jaw was removed, the heads were fixed in 4% formaldehyde (pH 6.9 to 7.1), were decalcified with Decalcifier II (Surgipath Europe Ltd. Peterborough, UK), and embedded in paraffin wax. Sections (5-µm thick) were then deparaffinized and stained with hematoxylin and eosin. Digital images were obtained with Pannoramic MIDI II (3dHisTech Ltd., Budapest, Hungary) and were examined with Pannoramic Viewer 1.15.4. Olfactory epithelium was characterized by the presence of Bowman's glands and increased thickness due to the presence of specialized intraepithelial neurons (supplementary figs. S2, S3, and S5, C and D, Supplementary Material online).

Micro-CT Experiments and Image Processing

Animal skulls were scanned using a SkyScan 1276 CMOS Edition high-resolution micro-CT (Bruker, Kontich, Belgium) equipped with a microfocus X-ray source, L10321 (Hamamatsu Photonics, Hamamatsu, Japan) and a MH110XC-KK-TP camera (XIMEA, Munster, Germany). The following parameter settings were used: isotropic voxel size, 9 µm per pixel; source voltage, 70 kV, with an aluminum filter (0.5 mm); source current, 200 µA; rotation, 180° with 0.2° per step. To calculate the surface sizes of turbinal bones, nasal cavities, cribriform plates, and cranial cavities of the skulls examined, NRecon software (nReconServer64bit, Bruker) was used to reconstruct primary images. DataViewer (Bruker) was used to obtain series of cross-sectional images. Based on the abovementioned histology data, each region of interest (ROI) was manually selected in CTAn (Bruker). After being transferred as a solid object, the corresponding surface area of each ROI (i.e. A_Nasal, A_OE, A_Cran, and A_CP) of each sample was also calculated by CTAn.

Brain Dissection

The area of the foramen magnum is considered to be at the junction of the brain and the spinal cord. The olfactory bulb was separated from the remainder of the forebrain with a razor under a dissecting microscope. Briefly, the brain of each animal was placed downward on its dorsal surface, with the edge of the blade aligned with the position marked as a broad black line in Fig. 3C. Once aligned, the blade was used to cut the olfactory bulb. The brain and olfactory bulb of each animal were then immediately weighed individually to the nearest 0.1 mg on a digital balance. The weight data for the mouse and the Nile rat are presented in supplementary Datasets S2 and S3, Supplementary Material online, respectively.

Genome Assembly, Annotation, Orthologs

Genome assemblies and corresponding annotations for the mouse (M. musculus, GRCm39; v.105), rat (R. norvegicus, mRatNB7.2; v.105), human (H. sapiens, GRCh38.p13; v.107), macaque (M. mulatta, Mmul_10; v.107), and marmoset (C. jacchus, mCalJac1.pat.X; v.107) were obtained from Ensembl (http://www.ensembl.org, last accessed July 2022); those for the Nile rat (A. niloticus; mArvNil1.pat.X, GCA_011762505.1) were obtained from Ensembl Rapid Release (v.105; https://rapid.ensembl.org/, last accessed March 2022) (Cunningham et al. 2022); those of the Gairdner's shrewmouse (M. pahari) (genome assembly: PAHARI_EIJ_v1.1, NCBI Annotation Release 101), African woodland thicket rat (G. surdaster) (genome assembly: NIH_TR_1.0, NCBI Annotation Release 100), and white-footed mouse (P. leucopus) (genome assembly: UCI_PerLeu_2.1, NCBI Annotation Release 101) were obtained from Genome data package of NCBI Datasets (https://www.ncbi.nlm.nih.gov/data-hub/genome/, last assessed December 2022). Pseudogene identification was conducted with the pipeline of PseudoPipe (http://pseudogene.org/pseudopipe/, last accessed March 2022) (Zhang et al. 2006) with default settings. The pseudogenes for each genome were then classified as processed pseudogenes (PSSD), duplicated pseudogenes (DUP), pseudogenic fragments (FRAG), and etc. Orthology relationships of rodent and primate genes were inferred by OrthoFinder (v.2.5.2) (Emms and Kelly 2015) with default parameters. The longest transcription variant was used as the representative transcript for each annotated gene.

RNA-seq Experiments and Analyses

RNA-seq experiments were performed to profile transcriptomes of olfactory epithelium samples isolated from the nasal cavity of 8-, 12-, and 16-wk-old male and female mice and Nile rats. Briefly, total RNA was extracted with TRIzol Reagent and subsequently purified with a RNeasy Mini Kit (Qiagen, Hilden, Germany). The quantity and purity of the isolated RNA samples were measured with a NanoDrop ND-1000 Spectrophotometer prior to cDNA library construction. OD_260/280 values for all of the samples were >1.70. RNA integrity number (RIN) scores were determined with an Agilent 2100 Bioanalyzer and ranged from 7.4 to 8.6. A sequencing library was prepared from purified RNA by using a TruSeq Stranded mRNA Library Prep Kit (Illumina, San Diego, CA, USA), according to the manufacturer's recommendations, for each sample. Briefly, mRNA was purified from total RNA (1 μg) by using oligo(dT)-coupled magnetic beads and then fragmented into small pieces under elevated temperature. First-strand cDNA was synthesized using reverse transcriptase and random primers. After generating double-stranded cDNAs and adenylating the 3′ ends of the DNA fragments, adaptors were ligated and samples were purified with an AMPure XP system (Beckman Coulter, Danvers, USA). The library quality was assessed by using Agilent Bioanalyzer 2100 and Real-Time PCR systems. Qualified libraries were subsequently sequenced on an Illumina NovaSeq 6000 platform with 150-bp paired-end (PE) reads generated by Genomics, BioSci & Tech Co. (New Taipei City, Taiwan). More than 120 million raw reads were obtained for each sample (see supplementary table S5, Supplementary Material online).

To process the raw transcriptome data, low quality bases and bases from adapters of the raw reads were removed using fastp (v.0.22.0) (Chen et al. 2018). The filtered reads were aligned to the reference genomes using HISAT2 (v.2.1.0) (Kim et al. 2019) to measure overall alignment rate and to count uniquely aligned reads. Kallisto was reported to have better performance in dealing with paralogs amongst different quantification programs (v.0.46.1) (Bray et al. 2016; Ma et al. 2021), therefore was used to estimate TPM for each annotated transcript with default parameters and the option of “–rf-stranded”. The TPM of a gene is represented by the TPM of the gene's longest transcript. For each transcriptome, log₂ of the TPM + 1 (pseudocount) per gene (log₂E) was used to calculate a Z-score using the formula: Z = (log₂E − T_M)/T_SD, where T_M is the mean of all log₂E values within a transcriptome sample and T_SD is the standard deviation of all log₂E values within a sample. Averaged values of T_M and T_SD (T′_M and T′_SD, respectively) were calculated for all 12 transcriptome samples, respectively. The Z-score of each gene of each sample was subsequently back-calculated to a normalized TPM value as: $[2^{(Z\times T_{SD}^{\mathrm{\prime }}+T_M^{\mathrm{\prime }})}]-1$. The normalized TPM values of genes with an original TPM = 0 were set back to 0. The raw and processed RNA-seq results have been deposited in the GEO database with accession number, GSE201761.

Defining OR and GPCR Genes

The program, hmmsearch, of the software package, HMMER (v.3.3.2; http://hmmer.org, last accessed March 2022) (Eddy 2011), was used to search proteins that matched with Pfam profile 7tm_4 (PF13853, olfactory receptors) in order to define putative OR genes. A primary set of genes encoding putative OR proteins were detected from the genomes that are analyzed. As described in a previous study (Niimura 2013), further refinement was performed to exclude non-OR genes, OR pseudogenes, and truncated OR genes. This refinement excluded: (i) genes encoding proteins shorter than 250 amino acids, (ii) genes producing proteins that were placed outside the clade of ORs on the phylogenetic tree constructed by LINTRE (Takezaki et al. 1995) using 11 rhodopsin-like GPCR protein sequences as the outgroup (see supplementary Dataset S4, Supplementary Material online), and (iii) based on alignment with the reference 7tm_4 profile protein, O51B2_HUMAN, genes encoding proteins showing any gap with ≥5 amino acids in size within any of the conserved 7 transmembrane regions (amino acid positions: 30 to 47, 59 to 85, 97 to 118, 139 to 159, 199 to 227, 239 to 263, and 269 to 289 of O51B2_HUMAN). The numbers of OR genes that were identified for each species following this procedure are shown in Fig. 4A and supplementary Fig. S10, Supplementary Material online.

Similarly, GPCR genes were derived from hmmsearch on protein sequences of each species that matched with any of the Pfam profiles: 7tm_1 (PF00001; rhodopsin-like receptors), 7tm_2 (PF00002; secretin receptors), 7tm_3 (PF00003; metabotropic glutamate receptors), 7tm_4 (PF13853; olfactory receptors), V1R (PF03402; vomeronasal 1 receptors), Frizzled (PF01534; Frizzled receptors), TAS2R (PF05296; taste 2 receptors), COX1 (PF00115; Cytochrome C oxidase subunit I), FA_desaturase (PF00487; fatty acid desaturase), and Ocular_alb (PF02101; ocular albinism type 1 protein). The mouse or rat V2R (vomeronasal 2 receptors) genes were identified by their Ensembl-annotated names beginning with “Vmn2” or “Vom2”, respectively, whereas the V2R genes of the other species were defined by those that were orthologs of mouse or rat V2R genes. The total numbers of non-OR GPCR genes that were identified for each species are shown in supplementary fig. S11, Supplementary Material online.

Loss and Gain of OR Genes

The ete-build gene-tree workflow of ETE v3 toolkit (Huerta-Cepas et al. 2016) was utilized to generate a gene tree of OR (or non-OR GPCR) putative orthologous sets there were defined by OrthoFinder, using the option “soft_modeltest_bootstrap” which tests 5 protein evolutionary models and infers tree with PhyML with 100 bootstraps. Following Niimura et al. (2014), a clade of genes that met the following criteria was separated from the putative ortholog set where they originally belong to: (i) This clade contained a lineage of primate gene(s) and a lineage of rodent gene(s), (ii) this clade was supported with a >90% bootstrap value, and (iii) this clade and its sister clade contained gene(s) from at least 1 common species. The process continued until no further separation could be made. The gene tree of each separated unit of ortholog sets was then converted from a rooted tree to an unrooted tree, and the unrooted tree was further separated if it met the following criteria: (i) It contained 2 clades, each of which contained a lineage of primate genes and a lineage of rodent genes, and (ii) the separation of the 2 clades was supported with a >70% bootstrap value. This process was carried out for all ortholog sets until no further divisions could be made within each unit. Ortholog sets that contain either only primate genes or only rodent genes before the separation process, or both primate and rodent genes before or after the separation process, were considered to have originated from an ancestral gene before the divergence of rodents and primates, and were numbered 693 and 729 in Fig. 4A and supplementary Fig. S10, Supplementary Material online, respectively. According to the solved phylogeny of the examined species, the DTL reconciliation algorithm implemented in RANGER-DTL 2.0 (Bansal et al. 2018) was used to infer all of the necessary gene gains and losses that explain the gene tree topology of each subgroup (clade) of genes separated from the OR (or non-OR GPCR) orthogroups, with the parameters of “-D 2 -L 1 -T 100 -B 70”. The module OptResolutions of RANGER-DTL 2.0 and Notung v2.9 (Chen et al. 2000) were employed to resolve low bootstrap support branches (<70%) in order to reduce the computational cost. A customized script was used to parse the outputs and calculate the numbers of gene loss and gain events along different branches of the species tree. The inferred events of subgroups within the same ortholog set were totaled and summarized as supplementary Dataset S5, Supplementary Material online.

Estimating Branch-specific d_N/d_S Values

Branch-specific d_N/d_S values were computed for all 3-species 1:1:1 orthologous genes or 5-species 1:1:1:1:1 orthologous genes. Coding sequences of these genes were obtained from the FTP site of Ensembl (https://ftp.ensembl.org/pub/, last accessed July 2022) and NCBI Datasets (https://www.ncbi.nlm.nih.gov/data-hub/genome/, last assessed December 2022). If multiple transcripts were annotated for 1 gene, the longest coding sequence was chosen. Translated protein sequences of ortholog coding sequences were aligned using ClustalW (Thompson et al. 1994), and then were back-translated into DNA after removing alignment gaps. With phylogeny of the 3 rodents known, the program “codeml” in PAML (Yang 2007) was used to estimate d_N/d_S for the orthologs in each branch of the tree, with the options, “runmode = 0, seqtype = 1, CodonFreq = 0, model = 1, Nssites = 0”, chosen in the control file. Next, d_N/d_S values of the branches of interests were compared. Repeating the analysis based on the branch-site model, “model = 2, Nssites = 0”, generated consistent results (supplementary figs. S15 and S16, Supplementary Material online).

Supplementary Material

Supplementary material is available at Molecular Biology and Evolution online.

Author Contributions

B.-Y.L. conceived and designed the study; B.-Y.L., M.-P.W., T.-Y.C., and C.-H.W. performed animal experiments; B.-Y.L. and Y.-H.C. supervised animal experiments; B.-Y.L., M.-P.W., T.-Y.C., and Y.-H.C. performed RNA-seq experiments; B.-Y.L., M.-P.W., and T.-Y.C. performed morphological analyses; B.-Y.L., M.-P.W., and A.Y.-F.C. conducted computational and statistical analyses; B.-Y.L. wrote the manuscript.

Data Availability

Raw sequences and values of processed signals of olfactory epithelium transcriptomes of mouse and Nile rat were deposited in NCBI GEO as series GSE201761 under NCBI BioProject, PRJNA832952.