Convergent reduction of olfactory genes and olfactory bulb size in mammalian species at altitude

Summary

The invasion of specialized ecological niches can cause drastic changes to selection regimes, resulting in genomic and phenotypic transformation ¹. High-altitude habitats offer an excellent opportunity to investigate the genetic basis of local adaptation ^{2, 3}, as the repeated specialization of multiple lineages for high altitude has produced striking examples of convergent evolution, adaptation, and changes in their underlying genes ^{4,5 6}. Although enlightening, this focus on adaptation has left aspects of evolution in high-altitude locations understudied – including the role of gene loss and pseudogenization, maladaptation and trait loss, and physiological aspects outside of respiration and gas exchange. To characterize how mammals responded to high-altitude in a new, unbiased way, we screened the genomes of 27 species living exclusively at high altitude (>1,000–1,500 m) and their lowland relatives (Figure 1) for inactivated pseudogenes or lost genes ⁷. Genes that convergently lost function in high-altitude species were highly enriched for olfactory receptor genes (ORs), with an average reduction of ~23% of OR repertoire in high-altitude species. No such trend was found for genes involved in pheromone detection and taste perception. In addition to OR loss, cranial endocasts show the brains of high-altitude species have on average ~18% smaller olfactory bulbs relative to lowland relatives. Together, these repeated evolutionary outcomes suggest a general relaxation of constraint on olfaction at altitude, perhaps due to reduced odorant diversity in high-altitude environments or reduced effectiveness of mammalian olfactory physiology in thin, dry, or cold air.

Results/Discussion

Repeated pseudogenization of olfactory receptors in high-altitude lineages

Based on the initial identification of pseudogenes across species, we found evidence of a consistent and repeated loss of olfactory receptor (OR) genes in high-altitude lineages. Specifically, analysis of likelihood models of trait evolution found the pseudogenization of many genes to be correlated with life at high altitude (Table S1), and those genes were highly enriched for olfactory receptors and annotations of olfactory function (Table 1). However, since this initial screen was dependent on orthology to human genes, it did not sufficiently capture lineage-specific OR repertoires ⁸. Thus, using the full proteomes of each species, we quantified OR repertoire using a domain search, which was more sensitive to lineage-specific expansions and contractions. This analysis further supported the conclusion that species living at high altitude have significantly fewer olfactory receptor (OR) genes when compared to their low-altitude relatives (Mean = −22.7%, −228 OR genes; Median = −17.8%, −138 OR genes; Wilcoxon Paired P-value = 0.00007). Excluding the genomes of domesticated breeds from the comparisons yielded the same significant result (Wilcoxon Paired P-value = 0.0003; Table S2). Moreover, using a phylogenetic generalized linear model, we showed that OR counts were significantly negatively correlated with altitude (de novo OR counts: P-value = 0.025). These results also showed that the same pattern holds true for all group comparisons within artiodactyls, rodents, lagomorphs, carnivores and primates (Figure 2). To assure that the observed pattern was not simply due to high-altitude genomes being more poorly assembled or annotated, we also analyzed phylogenetic models with OR counts, lowest altitude, and assembly metrics. They did not show a relationship between altitude and genome quality metrics for completeness (de novo OR counts: Contig N50 P-value = 0.144, BUSCO P-value = 0.536; Alt Status: Contig N50 P-value=0.521, BUSCO P-value =0.681), as would be expected if gene counts were generally decreasing due to lower quality genomes.

Table 1.

Enriched categories associated with preferential gene loss in HA lineages significantly enriched

Gene Set	P val	P adj	Num genes
KEGG Olfactory Transduction	6.01E-39	7.74E-36	384
REACTOME Olfactory Signaling Pathway	1.68E-37	1.08E-34	319
REACTOME GPCR Downstream Signaling	1.04E-14	4.45E-12	781
REACTOME Signaling by GPCR	3.73E-12	1.20E-09	881
REACTOME Generic Transcription Pathway	2.82E-10	7.25E-08	333
NABA ECM Affiliated	2.89E-05	0.00619816	167

Figure 2:

Comparison of de novo olfactory-receptor (OR) counts from available proteomes, in Primates, Artiodactyls, Rodents, Lagomorphs and Carnivores. Silhouette images are from Phylopic (CC BY-SA 3.0). Data within Table S5.

We also tested if this pattern of gene loss extended to other families of 7-transmembrane G-protein–coupled receptors (7tm-GPCRs) that recognize odorant and pheromone ligands. These GPCR families include trace-amine associated receptors (TAARs) in the main olfactory epithelium and several families expressed in the vomeronasal organ (V1Rs, V2Rs) ^9–11. They showed no consistent pseudogenization relationship with lineages at altitude (Table S3, Figure S1, S2). This may be due to the chemical profiles sensed by those receptors– for example, TAARs perceive volatile and highly aversive amines like those linked to predators or decay, while V1Rs and V2Rs are associated with male-female pheromone detection ^12,13. Our results suggest that 7tm-GPCRs as a whole were not affected, and that life at altitude has changed the selection landscape of odorant receptors specifically.

Characterization of individual OR genes preferentially pseudogenized in high-altitude lineages

There is clear evidence that the OR gene family underwent widespread losses in general, with only certain OR genes being consistently lost across different high-altitude species. Based on the human ortholog dataset, there were a total of 21 OR genes individually significantly associated with being pseudogenized in high-altitude lineages at an FDR of 1%. At the same time, there were no OR genes that were preferentially retained. There was a significantly higher frequency of pseudogenization occurring in Class I ORs compared to Class II ORs (χ² = 6.643, p < .01; Table S3). Mammalian class I ORs are thought to be specialized in detecting water-soluble and/or polar odorants ^14,15, while class II ORs tend to respond to acids, especially carboxylic acids ^16,17 - chemicals associated with microbial action and body odors ¹⁸. Presumably, the potential advantages of being able to sense water-soluble and/or polar odorants might not apply in environments where water availability and humidity is lower.

Most OR genes are expressed in multiple tissues, as well as nasal tissue, though their extra-nasal function is largely undescribed, but there are instances of them having different effects on cellular functions ^19,20. In addition, the 21 individually significantly associated genes were composed of ORs with expression in both nasal (olfactory mucosae) and extra-nasal tissues ²¹; however, there were not significantly more extra-nasally expressed ORs associated with high-altitude status (χ² = 1.92, p > 0.05). There is not widely available expression profile/location information for ORs outside of humans, so whether these same genes are more or less extra-nasal in their expression in other mammals is unknown. In our results, the OR gene whose pseudogenization is most significantly correlated with high-altitude living is OR10AD1; this gene is known to be expressed in the nose, but also eye and brain in humans (Table S4) with its role being linked to regulation of circadian rhythm through the OR-dopaminergic system ²². Hypoxia-mediated changes in circadian rhythms are thought to be a key driver of sleep quality differences in lowlanders at altitude compared to native high-altitude populations ^23,24. Though how a non-functional OR10AD1 might affect those circadian rhythm is unclear. Other OR genes of interest include OR6N2, OR10Z1 which have enriched expression in bone marrow, OR6V1 enriched in the brain, and OR1J1, OR2J3 in lung/airway ^25,26.

High-altitude lineages have reduced olfactory-bulb size

The pattern of olfactory receptor loss is clear, though the question now becomes how this loss may affect olfactory capability in high-altitude lineages. It is known that total OR gene repertoire is positively correlated with olfactory bulb size in the brain, thus OR subgenomes are frequently used as a proxy for olfactory ability ^27–31. To probe this relationship and to further characterize how altitude affected olfactory ability over evolutionary timescales, we studied measurements of olfactory bulb volume (OFB) and whole brain/endocast volumes (TB) from 21 mammal species (Placentals and Marsupials; Table S5). When comparing normalized olfactory bulb size, the OFB/TB ratio, between high-low altitude pairs, we found that olfactory bulb size in high-altitude species is reduced by ~18% on average (Wilcoxon Paired, P-value = 0.001, z = −3.076; Median = −15.3%). In addition, using a phylogenetic generalized linear model, we were able to show that OFB/TB were significantly negatively correlated with altitude (P-value = 0.036; 8 species pairs). This corroborates our finding of reduced OR repertoires and suggests that olfactory ability has been affected in a profound way during the evolutionary path to life at altitude (Figure 3).

Figure 3:

High-altitude species have on on-average smaller olfactory-bulb to total brain ratio (OFB/TB). [Left panel] An illustrated example of size difference between Ochotona princeps and low-altitude Lagomorphs [Right panel] Difference in olfactory bulb to total brain ratio. The line denotes 0% difference, and black dotted line shows the mean (−17.8%) from 21 species pairs. Data within Table S4 and Table S9.

Reduced olfactory bulb is likely not a hypoxia-induced plastic trait

Ultimately, our results suggest that life in high-altitude environments leads to a significant loss of functional olfactory receptors and reduction in olfactory bulb size. This phenomenon could be the result of OR repertoire size and olfactory bulb size being intrinsically linked, although it is also possible that these reductions occurred independently of one another. During development, the pathing of olfactory sensory neurons (OSNs) from the olfactory epithelium/placode to the forming olfactory bulb is known to be changed by olfactory receptor expression ^32,33. This in turn alters olfactory bulb morphology and neural map formation ^34,35; thus, reductions in OR repertoire size and olfactory bulb size are likely linked. However, there is still the question of whether the OR repertoire or the olfactory bulb size was affected “first” during their evolutionary history. The possible scenarios include, but are not limited to: one, some aspect of being at high-altitude released selective pressures on olfactory receptors writ large, which then affected olfactory bulb development; and two, that during development in utero hypoxia directly affects olfactory bulb development (i.e., environmentally plastic), which then over time released constraint on olfactory receptors.

One way of testing these scenarios would be to measure OFB/TB ratios in individuals which had developed in utero at altitude versus lowland. Thus, cranial endocasts were obtained from an experiment using high-altitude (HA) and low-altitude (LA) deer mice (P. maniculatus) ³⁶. The OFB/TB ratio of wild-caught mice, as well as those raised in a common-garden environment were measured. If HA wild-caught OFB/TB were different than those of HA ancestry raised at low-altitude, then it would suggest that OFB/TB ratio is induced by the environment (hypoxia); however, in this case, we found that there was no significant difference, which instead suggests that there is a strong genetic component to this trait. At the same time, we found that wild-caught high- and low-altitude populations did not differ. The later result also strengthens the notion that whatever the mechanism is causing decreased OFB/TB, it is happening over a longer period that may not be observed within a species and could depend on population connectivity. Overall, this suggests that olfactory bulb size is likely not an environmentally induced plastic trait and potentially points to the scenario where pseudogenization of ORs is affecting olfactory-bulb development, though the proximate mechanism for a relaxation of selection on the ORs is still unclear. Thus, alterations in the transcriptional machinery specifically associated with the olfactory bulb development or OSN pathing may potentially be the main driver, the latter of which could potentially affect olfactory bulb development without having pleiotropic effects on overall development of the brain or central nervous system.

Olfactory receptor loss typically occurs over longer timescales

We wanted to understand the timescale over which lineages lose their ORs - mainly through contrasting groups having arrived at altitude over a shorter (thousands of years; kya) versus longer time frame (millions of years; mya). Our results were primarily based on inter-species comparisons that have divergence estimates on the order of millions of years (6.1 – 8.6 mya; Table S6, column “MRCA MYA”); however, we also see a significant reduction on relatively short timescales in the Tibetan breeds of pig, sheep and even yak (2.5 – 5.0 kya), suggesting the potential for the pattern to arise over much shorter time periods ³⁷. These breeds rapidly evolved phenotypic features advantageous to high-altitude life presumably under strong selection, including changes to skin color, hair density, agility, heart size, vascular density, and hemoglobin concentration ^38–40. It is possible that domesticated breeds are an exception, since their parallel loss of ORs could well have resulted from relaxed constraint due to domestication, in addition to high altitude life; however, this seems unlikely since the other domesticated non-high-altitude lineages did not have significantly fewer ORs compared to their wild ancestors. In addition, comparisons between domesticated breeds of both low- and high-altitude showed a profound reduction in functional OR number compared to breeds of non-high-altitude origin. Ultimately, our assessment of the “timing” of loss is based on the available genomes, which constrains our ability to specifically target this question, and thus leans heavily on prior calculations of MRCA.

To further probe the question of timescale limits associated with this pattern of loss, we examined a high-altitude human population, the Tibetans, who are estimated to have established populations at above 3,000 meters sometime between 9.0 – 30 kya ^41–43. However, according to our analysis, pseudogenizing or deleterious polymorphisms in ORs are not found at higher frequency in the Tibetans compared to Han Chinese (27 Tibetan, 62 Han; Table S7). Therefore, it seems that a reduction in OR repertoire has not occurred in Tibetans, perhaps due to continued admixture between low and highland populations or because there have not been sufficient generations. Tibetans are not genetically secluded and have significant evidence of gene-flow in and out ^44,45. Moreover, humans already have a degraded sense of smell compared to most mammals, so the effect of altitude might cause little to no further reduction. Olfactory receptors have been known to largely be under relaxed selection ^46,47, which has increased the rates of pseudogenization, especially for primates and hominids ^48,49. Ultimately, it also signals that loss of functionality in ORs occurs primarily over longer time periods, unless there are additional factors (i.e. artificial selection, limited introgression).

Currently, there are no published results regarding any potential difference in olfactory capability to discern odor variety, or intensity, for any high-altitude human populations (Tibetans, Sherpas, Andeans, Ethiopians); however, there is evidence to suggest that there is substantial functional variation in both noncoding regions and coding regions of ORs ^50–53, as well as differences in olfactory ability between populations in general ⁵⁴, though the latter is hard to decouple from cultural effects.

Potential selective pressures at altitude

Our results show that there is a clear and consistent pattern of preferential pseudogenization of OR genes, as well as phenotypic change in olfactory bulb size; the next logical question is to the reason “why”.

First, hypoxia is the hallmark hazard associated with life at high-altitude, though there are additional abiotic environmental factors like low temperatures, increased UV radiation and aridity. Foremost, the change in partial pressure of oxygen availability can result in cerebral and/or pulmonary edema at extremes. Most of these issues are due to hypoxia-induced chronic systemic inflammation. Organisms that are adapted to high-altitude have a blunted inflammatory response, since such hyperactivation of this response is detrimental long-term ^55,56. Inflammation of the olfactory epithelium is connected to exposure to intermittent hypoxia, which affects overall olfactory capability ^57–61. Thus, it seems possible that the reduction in functional olfactory receptors, and then downstream reduction in OFB ratio might be a direct result. The lack of humidity at altitude, or aridity, is also known to cause nasal congestion, impaired nasal mucocilliary transport rate and increased nasal resistance ⁵⁸. Thus, it seems possible that nasal inflammation and congestion brought on by a combination of hypoxia and aridity may have released ORs from selective constraint. However, if aridity and the resulting inflammation were the cause, then you might expect to see a relaxation of constraint on TAAR genes, which are also expressed in the olfactory epithelium. Yet, TAARs are not significantly pseudogenized in high-altitude lineages, like OR genes are, despite also being expressed in the olfactory bulb.

Second, it is also important to note the potential influence of biotic factors in shaping the OR selection landscape, either within or between species at altitude. The fact that V1Rs or TAARs are similarly functional in high-altitude species could be suggestive of that the driver is affecting something other than intraspecies communication; but at the same time, receptors associated with intraspecies communication are being actively protected. Typically, at altitude, there are fewer resources, precipitating small population sizes of high-altitude exclusive species or groups. A lower density of conspecifics may mean it is that much more crucial to help identify mates or predators, through increased importance of vomeronasal or trace-amine receptors. This also matches with the finding that Class II ORs (carboxylic acids) were largely kept intact compared to Class I ORs (water-soluble). Other than conspecifics there is the possibility that changes to food availability may be involved, especially for herbivores. In plants grown at altitude, volatile organic compound (VOC) profiles are known to be different, potentially due to differences in types of stressors ^62–65. Moreover, there is generally less biotic diversity at high altitude, and it is known that a combination of lower barometric pressure, humidity and temperature, actively impacts the ability of scents to diffuse - specifically there is a decrease in the amount of odorous molecules per volume, and cold, dry air has a reduced capacity to carry odorants bound to vapor molecules ⁶⁶.

Thus, we postulate several hypotheses explaining the reduction of ORs at altitdue: [i] relaxation of constraint due to inflammation and olfactory malfunction at altitude, [ii] relaxation of constraint due to reduced importance of general smell at altitude (i.e., fewer odorants), [iii] relaxation of constraint, due to increased investment/positive selection in other senses (taste, vision), or other olfaction-related genes (i.e., V1R/V2R, TAAR). Of course, these possibilities are not mutually exclusive, and high-altitude environments possibly provide a perfect confluence of conditions for certain olfactory functions to be affected from multiple directions. Although it will require future experimentation to narrow down the possibilities, it is worth exploring what we know from prior work.

Conclusions

There is no doubt that life in high elevation environments is affected by a wide array of selective pressures. Altogether, these stressors impose a significant selective pressure, with most work on organisms at high-altitude focused on adaptive traits resulting from positive selection; however, invasion of new environments often results in the removal or weakening of a source of selection/constraint that was previously important for the maintenance of a particular trait (i.e. relaxed selection) ⁶⁷. Our work discovers that olfaction is profoundly affected on multiple levels (genomic, morphological), and that it is also a highly convergent pattern across diverse mammalian lineages. This surprising phenomenon raises questions about which element of living exclusively at high-altitude that is repeatedly altering the selection landscape, and what genetic mechanisms are driving the downstream changes seen in olfactory bulb development over time. In addition, it is also unknown whether this loss reflects a reduction of utility of ORs specifically, or perhaps is an indication of a more pronounced shift in need for other olfactory-related gene groups (V1R, V2R, TAAR) instead.

Resource Availability

Lead Contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Allie M Graham (graham.allie@gmail.com or graham.allie@ku.edu).

Materials availability

This study did not generate new unique reagents

Data and code Availability

All data is available in either the original papers as cited, and/or in the supplemental data of this paper. Any additional data generated in this study which are not already available in the supplementary material are available upon request. No original code has been developed in this study.

STAR Methods

Experimental model, subject and method details

High-altitude classification

Of the mammalian genomes available, we chose to create a dataset enriched in high-altitude exclusive species. Designating a species as “high-altitude” involved multiple criteria: first, species limits were obtained from a variety of sources including primary literature, the IUCN red list, the University of Michigan Museum of Zoology’s “Animal Diversity Web” resource ⁶⁸ and government species records (state, federal, foreign) (Table S8). We classified a species as “high-altitude” if their lower range bound was no lower than 1,000 m (3,280 ft) which resulted in a list of 24 exclusively high-altitude species and included 49 low-altitude related species as comparison. The altitude limits that can be designated as “high-altitude” are technically arbitrary, with sources claiming moderate or high-altitude environment refer to any elevation from 1,500 – 3,00 meters ^69–72. For the purposes of our study, we include species which have long diverged from their lower-altitude sister-species counterparts and utilize a rough altitude cutoff, to maximize species inclusion: lowest recorded altitude no lower than 800, but whose average altitude between highest and lowest recorded was greater than 1,000 meters (27 total). Thus, we refer to these species as “high-altitude exclusives”, though we fully acknowledge the semantic nature of this decision; in addition, for ease of discussion “non-high-altitude exclusives” will be called “low-altitude” henceforth in this study. We also performed the analyses with a stricter 1,500 m cutoff (16 total).

Method details

Pseudogene calling pipeline

To compare gene completeness and to screen for gene losses, we used TOGA - a method that uses pairwise genome alignment chains between an annotated reference genome (here human hg38 assembly) and other query species. Briefly, TOGA uses machine learning to infer orthologous loci for each reference transcript, capitalizing on the observation that orthologous genes typically have more alignments between intronic and flanking intergenic regions. TOGA then projects each reference transcript to its orthologous query locus using CESAR 2.0, a Hidden Markov model method that takes reading frame and splice site annotation of the reference exons into account. CESAR avoids spurious frameshifts, can detect evolutionary splice site shifts and precise intron deletions. Using the CESAR alignment, TOGA determines whether the transcript has inactivating mutations (frameshifting mutations, premature stop codons, splice site disrupting mutations, deletions of entire coding exons).

The output involved 7 different categories of potential losses including [1] Intact: no inactivating mutations and missing sequence in the middle 80% of CDS, [2] Partial, Intact: no inactivating mutations in the middle 80% of CDS, less than 50% of sequence is missing, [3] Uncertain Loss: there are intact mutations in middle 80% of CDS, but not enough evidence to call it lost, [4] Lost: inact mutations in middle 80% of CDS and enough evidence to call lost (intact mutations in >=2 exons), [5] Missing: 50% of CDS is missing; a copy of the gene is likely present in the query genome, but we don’t know whether it’s functional or not, [6] Partial, Missing: subset of missing, alignment chain covers less than 50% of the gene and [7] ParaloG: TOGA could not find any orthologous chain, therefore it used paralogous ones. For analyses in BayesTraits (next section), we designated categories 1–3 as “retained”, category 4 as “lost/pseudogene”, whereas categories 5–7 were “NA / -”. The final pseudogene matrix, and a phylogenetic tree was used to analyze correlated evolution between pairs of discrete binary traits, exclusively high-altitude and low-altitude versus pseudogene and non-pseudogene status (Table S9).

Assessment of loss associated with altitude status

The predicted pseudogene status results formed a gene-by-species matrix of gene presence (functional), absence (pseudogene), and excluded (not assigned). We ran two nested likelihood models in BayesTraits ⁷³ using 19,305 gene vectors and a vector indicating which species are ‘high-altitude exclusive’ and ‘non-high-altitude exclusive’. The independent model contained two parameters – a gene loss rate (the rate at which a functional gene becomes a predicted pseudogene) and a rate for transition from not-exclusive to exclusive high-altitude status. Because our study was focused on gene loss, gene gain was not allowed; its rate was constrained to zero. This independent model contained no relationship between gene loss and altitude exclusivity state, and so it served as the null hypothesis. The dependent model, on the other hand, added another free parameter by dividing the gene loss rate into two parameters – loss rate on each group branches, separately. We compared these two nested models using a likelihood ratio test (LRT). Since we were interested in the evidence for higher loss on high-altitude branches, we reversed the sign of the LRT statistic for all genes inferred to have a higher loss rate on terrestrial branches in the independent model (Table S1).

Gene ontology analysis of pseudogenized genes

The gene list and corresponding significance values produced from the BayesTraits correlation analyses were used to calculate pathway enrichments for all canonical pathways from mSigDB ⁷⁴, Mouse Genome Informatics (MGI) functional annotations ^75,76, Gene Ontology terms ⁷⁷ and canonical datasets which includes annotations from Reactome, KEGG, BIOCARTA and PID ^78–81. This analysis was performed for results from the 75 species, as well as the ungulates, rodents and primates, separately.

Assessment of olfactory-receptor genes across the genome

To interrogate the number of olfactory receptors present, a count matrix was generated based on the human complement for the 75 species genomes, which includes olfactory-receptor genes, vomeronasal receptors (V1R) and Trace amine-associated receptors (TAAR). The counts were tested for significance using a Mann Whitney U test of the 28 high-altitude lineages and 43 non high-altitude lineages; a paired Wilcox Rank Sum test was performed between 24 high-altitude species and their closest relative. The same statistics were performed including Tibetan domesticated breeds, and without domesticated breed comparisons. In addition, de novo annotation was performed on 52 publicly available transcriptomes/proteomes (21 highalt, 31 non-highalt; Table S6). A number of the species used in the previous analyses have genomes available on NCBI, though there are no annotated transcripts or proteins.

As olfactory-receptor gene families are known to actively expand and contract across lineages, hmmsearch command from the HMMER 3.0 program ^82,83 was used to identify proteins that contained either a 7tm_4 domain (OR; PF13853.8), or V1Ra (PF03402) ⁸⁴. The TAAR genes do not have a canonical domain available, thus counts were performed based on annotated identity. Sequences with significant hits were then parsed by identifier, any duplicate ID’s removed, and then total counts tallied. The transcriptomes/proteomes used were assessed for completeness using BUSCO against the Mammalian dataset (mammalia_odb10)^85,86, and any assembly with a completeness score less than 80% was not included in further analyses. A paired Wilcox Rank Sum test was performed for 14 species pairs, which included a high-altitude species to their closest relative.

Allele frequency shifts in population level human data

We used previously published dataset of human population variation to ascertain whether a similar pattern of potential pseudogenization of olfactory receptor genes was present ⁸⁷. Details concerning sample collection, sequencing parameters and filtering steps can be found within the respective publication. We will therefore detail any additional steps taken from the data obtained through various repositories.

A dataset with both Tibetan and Han individuals was obtained from a previously published study and based on the hg19 version of the human genome. Complete Genomics (v2.0, v2.5) performed sequencing and variant-calling for the Tibetan samples, with quality thresholds of less than Q30 converted to missing genotype calls, and sites with more than 5% missing genotype rate removed. Genotyping errors were checked by comparing the dataset to 62 genomes that had been sequenced in both CG public genome data and 1KG Project Phase I data. Monomorphic sites were removed using Samtools - BCFtools ⁸⁸, and the input file annotated using the vcf-annotate-polyphen package in combination with PolyPhen ⁸⁹. After annotation, the file was split between sites that were labeled as “benign” (HVar/Hdiv- benign/benign) or “damaging” (HVar/Hdiv- possibly/probably, possibly/possibly, probably/probably damaging), and sites with other combinations were excluded. Locations of known olfactory genes were identified based on HGNC database (genenames.org) and pulled from using the human genome (hg19). In addition, a series of comparative lists were created of other olfactory G-coupled Receptor genes including GPCR families (A, 2/B, 3C; 553 genes) ⁹⁰, as well as 710 known GPCR downstream signaling genes ⁹¹. The “benign” or “damaging” sets were further split into those intersecting with 875 olfactory receptor genes, and allele frequency for sites with only 2 alleles was calculated using VCFtools ⁹².

Olfactory bulb measurements from cranial endocasts across eutherians (placentals, marsupials)

The olfactory bulb ratio is positively correlated with the total OR gene repertoire suggesting that the OR subgenomes can be used as a proxy for olfactory ability; thus we used data from previously published olfactory bulb volume (OFB) and whole brain/endocast volumes (TB), including bears, rodents/squirrels, lagomorphs, primates, marsupials and other various eutherians ^93–102. Overall, these studies used either physical or virtual endocasts from crania, certain landmark anatomical features annotated, and measurements made using various programs – specifics regarding methods are contained within the corresponding publications. There was a total of 300 individual data points across 235 species with data available to calculate the OFB/TB ratio, the extant species altitude status was determined via lowest recorded elevation, resulting in 120 individuals from 21 species classified as “high altitude”, with their lowest recorded elevation being above 1,500 m (Tables S5, S10). A phylogenetic tree was created based on full mitochondrial genomes available; to do so, a multiple sequence alignment was built with MAFFT ¹⁰³ and a tree created using maximum-likelihood FastTree ¹⁰⁴, using default settings.

In order to assess the order of operations concerning OR repertoire and OFB/TB size, microCT scans of 27 P. maniculatus skulls (Table S11) were obtained through MorphoSource ³⁶. These skull measurements were taken from mice that were both wild-caught and from lab-reared populations. The wild-caught adults were trapped from a high-altitude location (Colorado at 4,350 m above sea level) and at low altitude (Nebraska at 430 m above sea level) - additional information about animal husbandry, specimen collection, euthanasia and endocast imaging is described in previously published work ¹⁰⁵. The cast of the olfacto ry bulb, defined as the region of the endocranial cast anterior to the circular fissure, was reconstructed for each specimen using Avizo^™ 8.1. The “Edit Labelfield” module was used to digitally segment the anterior region of the endocranium. The “Slice” and “Resample Transformed Image” modules were used to reorient and resample the segmented dataset along the plane of the circular fissure. For each specimen, the resampled segments anterior to the circular fissure (representing the olfactory bulb) were reloaded into Avizo^™ and a surface model was generated using the “Generate Surface” module. Volumes were measured using the “Surface Area Volume” module.

Quantification and statistical analysis

Assessment of loss associated with altitude status

In order to estimate empirical study-wide false discovery rates (FDR), we ran the analysis for 500 iterations of the phylogenetic tree with high-attitude status randomly assigned ¹⁰⁶. We then used the distribution of test statistics from simulated genes to estimate the FDR in an approach similar to empirical permutation-based FDR calculations. Studies that perform permutation-based FDR calculations commonly use a modification of the Benjamini-Hochberg procedure wherein they compare observed test statistics with empirically defined null distributions obtained from repeated permutations of the data and labels, in place of the procedure’s traditional comparison of observed P-values to a null distribution based on uniform quantiles. However, in our case permuting tip labels would frequently change the branch lengths on which functional losses could occur and modify the well-supported relationships among foreground species. In our analysis, we thus use the same modified Benjamini-Hochberg procedure to compare the observed modified LRT statistic distributions to the distribution of modified LRT statistics for simulated genes (the empirical null distribution), in place of the distribution of a permutation-based test statistic. This approach results in FDR calculations based on test statistic distributions from a null dataset more closely matching the true dataset, commonly preferred in genomic data analysis.

Gene ontology analysis of pseudogenized genes

For each pathway, a Wilcoxon Rank-Sum statistic was calculated to compare the sign of Rho times the negative log p-value for correlations of genes in the pathway to the same measure for all genes included in a pathway annotation using getStat() from RERconverge ¹⁰⁷ to calculate the sign of Rho times the negative log p-value for correlations for each gene and fastwilcoxGMTall() to quickly calculate an approximation of the Wilcoxon Rank-Sum statistic.

Assessment of olfactory-receptor genes across the genome

The functional OR counts from this analysis and previous estimates of functional OR counts ¹⁰⁸ were also subjected to a regression analysis (R²). Using an Ornstein-Uhlenbeck Motion model in Ape and Phytools ^109,110, a phylogenetic general linearized model was used to test significance of OR counts to various metrics, including [1] high-altitude status (1 or 0), [2] lowest recorded elevation in meters, [3] BUSCO completeness score, and [4] genome contig N50. Lastly, olfactory receptor genes lost significantly in high-altitude lineages were characterized by their known association with human phenotypes ¹¹¹, expression patterns using tissue-specific and single-cell RNA-seq from the Human Protein Atlas (https://www.proteinatlas.org/) ^112–114, and RNA-seq results from human olfactory mucosae ²¹.

Allele frequency shifts in population level human data

The minor-allele-frequencies (MAF) were calculated for each SNP in each population, graphed in R using tidyverse ¹¹⁵ and ggplot ¹¹⁶, and tested for significance with Mann-Whitney U Tests.

Olfactory bulb measurements from cranial endocasts across eutherians (placentals, marsupials)

Each of the “high” altitude species were then paired with their closest related species or group, and a Wilcoxon Signed-Rank Test performed on the OFB/TB proportions. A one-tailed phylogenetic general linearized model for Brownian Motion was used to test significance of OFB/TB to high-altitude status (1 or 0), and lowest recorded elevation in meters.

Supplementary Material

Table_S4

Table S4: Comprehensive information on each known member of the Human Olfactory Receptor repertoire. Including expression exclusivity (tissue, cell, blood, cell line), and any known GWAS correlation.

Table_S7

Table S7: Comprehensive information for each of the 104 mammal genomes used, including assembly metrics, count information (OR, V1R, TAAR) and time to most recent common ancestor (TMCRA).

Table_S1

Table S1: Bayesian assessment of loss associated with altitude status, using a likelihood ratio test (LRT), across human orthologs. Significance was based on whether the chi-squared pvalue from the independent vs independent model assessment were greater than the gene-by-gene false discovery rate (FDR).

Table_S8

Table S8: Information associated with classification of species as “high-altitude” based on known altitudinal records.

Table_S10

Table S10: Data used to calculate olfactory bulb to total brain ratio (OFB/TB) for 123 individuals across 75 mammalian species, and their associated original locations. Individual measurements can be found in Table S10. Additional columns include which mitochondrial genome used for the PGLS analysis for pairs (see Table S5)

Table_S11

Table S11: Endocranial cast measurements associated with members of a common-garden experiment (High v Low / Lab v Wild) using Peromyscus maniculatus.

Tables_S3_S6

Table S3: Overview of instances of pseudogenization in human orthologs of olfactory receptor (OR) genes.

Table S6: Comparisons of frequencies of damaging or benign variation found within olfactory receptors (OR), other GPCR genes, or downstream genes of olfactory receptors between high-altitude human population (Tibetan) and their closest non-high-altitude population relatives (Han); statistical p-values yielded from Mann-Whitney U tests.

Table_S2

Table S2: Data used for a phylogenetic least squares analysis (PGLS) to test correlations between olfactory receptor (OR) counts and Altitudinal Status. The column titled “MWU Pair” refers to when a paired Mann-Whitney U test was used, which values were used to compare the high-atitude species to.

Table_S5

Table S5: Pairwise species comparisons of olfactory bulb to total brain ratio (OFB/TB). Individual measurements can be found in Table S10. Additional columns include which mitochondrial genome used for the PGLS analysis, if available.

Table_S9

Table S9: Characterization of each H.sapiens ortholog for 72 species. Those species highlighted in yellow were considered “high altitude”. Explanation of each catatory is detailed in the methods.

Figure 1:

Summary figure showing genomes of species used in initial pseudogene analysis of high-altitude exclusive (blue) and non high-altitude exclusive related lineages. Based on the phylogenetic trees of Silhouette images are from Phylopic (CC BY-SA 3.0)