Examining the molecular basis of coat color in a nocturnal primate family (Lorisidae)

Abstract Organisms use color for camouflage, sexual signaling, or as a warning sign of danger. Primates are one of the most vibrantly colored Orders of mammals. However, the genetics underlying their coat color are poorly known, limiting our ability to study molecular aspects of its evolution. The role of the melanocortin 1 receptor (MC1R) in color evolution has been implicated in studies on rocket pocket mice (Chaetodipus intermediusi), toucans (Ramphastidae), and many domesticated animals. From these studies, we know that changes in MC1R result in a yellow/red or a brown/black morphology. Here, we investigate the evolution of MC1R in Lorisidae, a monophyletic nocturnal primate family, with some genera displaying high contrast variation in color patterns and other genera being monochromatic. Even more unique, the Lorisidae family has the only venomous primate: the slow loris (Nycticebus). Research has suggested that the contrasting coat patterns of slow lorises are aposematic signals for their venom. If so, we predict the MC1R in slow lorises will be under positive selection. In our study, we found that Lorisidae MC1R is under purifying selection (ω = 0.0912). In Lorisidae MC1R, there were a total of 75 variable nucleotides, 18 of which were nonsynonymous. Six of these nonsynonymous substitutions were found on the Perodicticus branch, which our reconstructions found to be the only member of Lorisidae that has predominantly lighter coat color; no substitutions were associated with Nycticebus. Our findings generate new insight into the genetics of pelage color and evolution among a unique group of nocturnal mammals and suggest putative underpinnings of monochromatic color evolution in the Perodicticus lineage.


| INTRODUC TI ON
The evolution of color variation in organisms highlights the selective pressures that impact survivability. Research on color variation has demonstrated coloration contributes to the ability of animals to conceal themselves from predators, to regulate body temperatures, to advertise their toxicity, or to attract mates (c.f. Caro, 2005). For taxonomists, striking color differences and patterns have been used | 4443 MUNDS et al. for distinguishing species (Bradley & Mundy, 2008;Caro, 2009;Nekaris & Jaffe, 2007). Several genetic loci have been identified that influence coat color and pattern variation in numerous organisms, but the most well-studied locus in vertebrates is the melanocortin 1 receptor (MC1R).
The MC1R is involved in the regulation of the melanin production of two common pigments: eumelanin (brown and black) and pheomelanin (yellow and red) (Barsh, 1996;. Studies on MC1R variation have focused on differences between melanistic, or dark-colored (brown/black) morphs, and amelanistic, or light-colored morphs (yellow/red). For ease of understanding, we will refer to those with a melanistic phenotype as "dark" and those with an amelanistic phenotype as "light." Changes in MC1R have been associated with color changes in insects, birds, mice, domesticated livestock and pets, and other mammals (Theron). Many of these studies examined the synonymous (dS: a substitution that does not change the amino acid structure) and nonsynonymous (dN: a substitution that changes the amino acid structure) substitution ratio in MC1R. The dS/dN ratio is used as an indicator of the strength and mode of natural selection in a coding gene over an extended timescale. A dS/dN ratio of less than 1 is associated with purifying selection, equal to 1 is neutral selection, and greater than 1 is positive selection (Yang & Bielawski, 2000). Generally, the dS/dN ratio is expected to be less than 1 in coding regions as changes in the amino acid structure in coding genes can be detrimental, thus resulting in fewer nonsynonymous substitutions (Jeffares et al., 2015). But positive selection can occur on coding genes if changes are beneficial.
As stated, MC1R influences vertebrate pigmentation (Barsh, 1996;Hoekstra, 2006;Mundy, 2005;Protas & Patel, 2008;Rees, 2003;Theron et al., 2001), but it is not the only gene to influence coat or hair color differences, as the agouti signaling protein, encoded by the ASIP gene, is another well-studied locus that regulates the distribution of melanin in mammals. Changes in ASIP result in agouti coloration: single hair bands with alternating yellow and black sections of varying size (Fontanesi et al., 2009;Furumura et al., 1996).
Other key pigmentation genes have been identified and research on this topic is growing, but much of it is focused on domesticated animals, like livestock and dogs, and mice-predominantly laboratory mice (Daverio et al., 2016;Mallarino et al., 2016;Neves et al., 2017;Protas & Patel, 2008;Rees, 2003;Saif et al., 2020;Silvers, 2012;Yang et al., 2019;Zhao et al., 2018). Although these studies provide invaluable information, their methods and results can be difficult to transfer to wild, less-studied populations. For example, the ALX homeobox 3 (ALX3) in African striped mice (Rhabomys pumilo) impacts dorsal stripe formation, and ASIP contributes to the shade of brown produced in these stripes (Mallarino et al., 2016). Mutations of the KIT proto-oncogene, receptor tyrosine kinase (KIT), have been noted in mice, horses, cats, and other vertebrates and are attributed to white morphs as well as white spotting that can result in "masking" patterns on the face and spotting patterns throughout the body (Brookes & Bailey, 2005;David et al., 2014;De Sepulveda et al., 1995;Wong et al., 2013). In dogs, melanocyte-inducing transcription factor (MITF) is known to contribute to irregular white spotting, and numerous other canid genes have been identified that result in their color and pattern differences (Saif et al., 2020). Yet, few studies have examined these genes in wild populations, or explored the role of MC1R in color regulation in nondomesticated species, such as primates (c.f. Mundy & Kelly, 2003). Even our own study attempted to examine KIT and ASIP in nonhuman primates, but we were unable to design successful primers for amplification based on available primers from closely related species (Mundy & Kelly, 2006) or published sequences. These regions of the genome could be different enough in Lorisidae that it made it challenging to successfully amplify them. Establishing a baseline of color evolution in a group of organisms will aid in the development of future projects to examine color variation, which is easiest to do by using the most well-studied and understood pigmentation gene: MC1R.
For example, Nachman et al. (2003) found four adaptive chargechanging amino acid polymorphisms in the MC1R of rock pocket mice (Chaetodipus intermediusi) in Arizona. These MC1R amino acid changes were linked to aiding the mice in blending in with their environment. Mice living in light-colored environments had substitutions associated with light-color phenotypes and vice versa for dark-colored mice . Corso et al. (2016) examined variation in MC1R sequences in toucans (Ramphastidae). By comparing the dS/dN ratio in MC1R and relating it to the observed phenotypic changes, they were able to demonstrate that molecular genetic variants were related to the evolution of feather coloration in Ramphastidae. For toucans, they suggested that darker plumage resulted from either positive selection or a relaxation of selection on MC1R (Corso et al., 2016). In humans (Homo sapiens), changes in MC1R are linked to hair and skin color variation. Lighter skin pigmentation, red/yellow hair color, freckles, and sensitivity to UV are the result of more than 30 nonsynonymous mutations in the MC1R gene (Sulem et al., 2007). The changes in hair color production are due to a relationship between MC1R and ASIP (Beaumont et al., 2007;Harding et al., 2000;Hoekstra, 2006;Moro et al., 1999;Mundy & Kelly, 2003Nasti & Timares, 2015;Valverde et al., 1995). The transition to the production of pheomelanin is considered to be derived, whereas the production of eumelanin is ancestral (Han et al., 2008;Jackson, 1997;Sturm et al., 2003;Sulem et al., 2007). These studies and the extensive body of literature on MC1R demonstrate that just one amino acid coding change in MC1R can result in stark color differences among closely related taxa and may be the product of selection (Corso et al., 2016;Gompel & Prud'homme, 2009;Hoekstra, 2006;Majerus & Mundy, 2003;Nunes et al., 2011).
Less studied is the impact of MC1R mutations on nonhuman primates, even though nonhuman primates are considered to be some of the most colorful and patterned mammals (Bradley & Mundy, 2008;Moreira et al., 2019). One of the first and seminal studies of MC1R in nonhuman primates found that nonsynonymous mutations were not correlated to coat color and appeared to be more likely influenced by phylogeny (Mundy & Kelly, 2003). This study also found that purifying selection is the primary mode of evolution for the MC1R gene in nonhuman primates. There were a few exceptions, especially with respect to Leontopithecus rosalia (golden lion tamarin), where results indicated a higher-than-expected dN/dS ratio (0.91) that was not significantly different from 1 and revealed several substitutions and deletions at functionally important sites. Their finding suggests the red hair phenotype observed in L. rosalia could be from a loss of function in MC1R, as found in humans (Mundy & Kelly, 2003). Although their study is one of the most comprehensive studies of MC1R evolution in nonhuman primates, Mundy and Kelly (2003) focused mainly on monkeys and apes (catarrhines). Another study examined functional changes within MC1R in nine phylogenetically distant primate species with varying coat colors (Haitina et al., 2007). Although this study found an amino acid substitution (E94K) that corresponds to color change in various vertebrates, this particular substitution does not seem to affect color change in the studied primates. Other substitutions were noted that possibly impact ligand binding sites, such as D117G and D121N. Additionally, their examination of the red/orange orangutans found no MC1R mutations associated with a pure pheomelanin coat production; not even substitutions similar to ones found in humans (Haitina et al., 2007). As for studies of MC1R on a finer scale, there have been two studies that examined MC1R variation in and among macaque species (Macaca) (Bradley et al., 2013;Nakayama et al., 2008). Neither study was able to find support for MC1R to be influencing coat color changes within these species, suggesting MC1R may not be the ideal gene to use when trying to infer coat color genetics or evolution in catarrhine primates.
Neither study examined MC1R evolution and its association with skin color. Furthermore, no studies have solely focused on MC1R mutations in lemurs or lorises (strepsirrhines). Studies that did include strepsirrhines either treated them as an outgroup to catarrhines or found little support for an effect of change within MC1R on coat color (Mundy & Kelly, 2003).
However, coat coloration might not play the same role in catarrhine and diurnal species as it does in nocturnal species. For nocturnal primates, the contrasting light and dark coloration is essential for visual recognition at night, and aids in intraspecific identification and communication, as well as camouflage (Bearder, 1999;Caro, 2005;Ford, 1994;Munds et al., 2013;Nekaris & Bearder, 2007). The dark coloration can result in different contrasting values, which further reinforces species diversification (Ford, 1994;Munds et al., 2013).
Within the nocturnal primate family Lorisidae (Nycticebus, Loris, Perodicticus, Arctocebus), variation might not just be reflective of conspecific recognition, it might also be aposematic signaling. The slow loris (Nycticebus) is the only venomous primate .
This has led some to propose that their coat patterns and coloration are aposematic signals, warning predators, and conspecifics of their dangerous venom, as well as a form of mimicry to the sympatric cobras (c.f. Nekaris et al., 2013).
Aposematism is as a type of interspecific signal that warns a predator this particular prey is noxious. These signals can be bright colors or, as typically observed in mammals, distinct contrasting markings such as the stripes found on skunks (Caro, 2009(Caro, , 2013. Slow lorises possess distinct contrasting dorsal stripes and facemasks, and when threatened, they apply a toxin secreted from glands in the armpits to their mouth. The mixing of this toxin with saliva makes the venom. After application, the slow loris will then connect the hands above its head, in an appearance that is like the fanning of a cobra-hood. The ocular patches of the loris with the raised arms appear to replicate the open hood of an Indian cobra (Naja naja) with its two distinct spots . Thus, the slow loris could be a Mullerian mimic (when a toxic animal imitates or appears like another toxic animal) through behavior and replication of the spots (Mallet & Joron, 1999;Pasteur, 1982). Of even more interest is that slender lorises are not venomous but they display contrasting ocular patches and when threatened will raise their arms over their heads like slow lorises (Gursky-Doyen & Nekaris, 2007). Such behavior and markings suggest slender lorises are Batesian mimics (a nontoxic animal that takes on the appearance and or behavior of a toxic animal) of cobras and/or slow lorises (Mallet & Joron, 1999;Pasteur, 1982). The evolution of aposematic coloration and patterning is speculated to be selected upon and seems to result from a few loci changes, such as one or two changes within MC1R, that result in major phenotypic changes (Mallet & Joron, 1999). If this is the case for Lorisinae, then we would expect them to have MC1R nonsynonymous mutations on their branch in comparison with the closely related African subfamily Perodictinae (Perodicticus, Arctocebus), who are monochromatic, lack venom, or mimetic. Such dichotomous coat colors and patterns coupled with one subfamily evolving rare defense strategies (venom and mimicry) not seen in any other primate present a unique opportunity to examine the evolution of coat color and its association with aposematic/mimicry and phylogenetic signals.
We predict that Perodictinae, more specifically Perodicticus, retain the ancestral form of Lorisidae coloration due to their monochromatic phenotype and lack of specialized defenses. We expect the overall Lorisidae dN/dS ratio of MC1R to indicate purifying selection (e.g., <1), as has been found with numerous other studies, including the few that examined primates (Majerus & Mundy, 2003;Makova & Norton, 2005;Mundy & Kelly, 2003;Nunes et al., 2011).
Such a finding would not suggest MC1R lacks a functional role in coat or skin morphology in Lorisidae, but instead would indicate it is conserved as changes in MC1R affect survival (Beaumont et al., 2007;Harding et al., 2000;Hoekstra, 2006;Makova & Norton, 2005;Moro et al., 1999;Nasti & Timares, 2015;Valverde et al., 1995). Finally, we predict there are more nonsynonymous mutations along the Lorisinae (Loris and Nycticebus) branches than along the Perodictinae branches (Arctocebus and Perodicticus). Our predictions are based on the idea that the color and patterning found in Lorisinae are adaptive ( Figure 1).

| Study collection
Due to the difficulty of trapping Lorisidae from the wild (Munds et al., 2018;Nekaris et al., 2020), we obtained samples (already extracted DNA from various tissue sources, or whole tissue samples we extracted from such as hair) and photographs from American Zoological Association (AZA) institutions. As the majority of our tissue samples are from hair follicles which are a poor source of genomic DNA, we focused our project on a few key genes (MC1R, mitochondrially encoded cytochrome b (MT-CYB)), rather than taking a more genome-wide approach. Additionally, collecting from captive institutions limits the species studied as only two to three of the recognized eight slow loris species are found in captivity (N. coucang, N. bengalensis, N. pygmaeus), possibly two species of slender loris (L. tardigradus, L. lydekkerianus), and only one member of the subfamily Perodictinae, but the specific species or subspecies is unknown (Perodicticus). Currently, there are three recognized species of Perodicticus, but it is uncertain which species are represented in captivity. There is a good probability that captive individuals are hybrids, as the new species are recently recognized and all species are highly similar in appearance (Mittermeier et al., 2013). The genus Arctocebus is not found in captivity and thus could not be included in our study. As there are multiple species of Nycticebus species in captivity, we will refer to it as Nycticebus spp. or

| Photography collection
We requested photographs of all genetically sampled individuals from participating AZA institutions, with four photographs per animal. Specifically, we requested each animal have left side and full frontal of the face, top of the head, and the entire dorsal photographed ( Figure 2). Some of the genetic samples used were acquired from deceased Lorisidae (Frozen Zoo ® and Duke Lemur Center (DLC)), which limited the number of photographs we had for these individuals. With Nycticebus from the Frozen Zoo ® collection, we were able to get all four photographs requested from historical records. DLC shared a few photographs of the two N. pygmaeus individuals we examined from their institution. We were not able to obtain photographs of the slender lorises (Loris) used in this study but acquired photographs from other zoo websites (Memphis Zoo, San Diego Zoo, London Zoo, Antwerp Zoo, and Biblical Gardens).
We cannot confirm the relationship of these captive individuals to our genetics samples, but many captive individuals are related to each other, so using photographs from these other slender lorises still provides an accurate depiction of slender loris coat and skin colors. Additionally, although there are species differences between the gray and red slender lorises, individual differences are considered to be minimal within species (Nekaris & Bearder, 2007). The photographs shared by AZA institutions are not our property and cannot be made publicly available. We are happy to share images privately, with approval from the AZA institutions.

| Ancestral state reconstruction
We analyzed one to six images from faces and bodies of 28 individual Lorisidae (4 Perodicticus, 11 N. pygmaeus, 4 Nycticebus spp., and 7 Loris). Additionally, we included to our analyses nonloris outgroups:  and facial areas that have skin completely exposed (nose, ear, hand, and around the eyes). It is important to note that some of these areas (i.e., circumocular patch, crown, ear forks, and dorsal stripe) are not present in all species examined. Nevertheless, we carefully selected areas that are functionally like the ones in Lorisidae. The analyzed features were previously shown to provide meaningful discrimination between slow loris species (Munds et al., 2013;Nekaris & Jaffe, 2007;Nekaris & Munds, 2010). The skin area predominantly covered with hair (hereafter, hair) and completely exposed (hereafter, skin) were subjectively categorized as white, yellow, red, orange, brown, black, gray, and pink (Supplemental Material S2). The colors of facial and body areas were independently coded by three raters. Because coding colors is subjective, we measured the interrater agreement by using the Fleiss' kappa test (Fleiss, 1971;Fleiss et al., 2003) and our results indicated that the strength of agreement between observers was good (K = 0.62; p < .005). We should point out that while the strength of agreement was found to be sufficient (Altmann, 1999;Arstein & Poesio, 2008), kappa values lower than 0.8 have been found to introduce some degree of noise to the data.
Our kappa values are below this threshold, but we feel confident in our results, given the general agreement between individuals in assigning color values. The answers were compared and the most common answer to a trait was used in the analysis.
We performed ancestral state reconstructions by mapping the hair and skin colors using a phylogenetic tree drawn from a published Lorisidae phylogeny (Arnold et al., 2010;Munds et al., 2018). We inferred the evolutionary history of hair and skin color (discrete traits) using a stochastic mapping approach implemented in the R phytools package (Revell, 2013). The ancestral states at each node were estimated under three basic models: equal rates (ER), all rates different (ARD), and symmetrical transition rate (SYM). The ER was the best fitting model. From this model, we simulated 1,000 character histories across the phylogeny using the phytools make.simmap function in R version 3.6.3 (Bollback, 2006).

| Tissue collection and sequencing
The collection protocol adhered to the humane animal handling   (Tavare, 1986). We used a Bayesian partitioned analysis for phylogenetic reconstruction using MT-CYB with program BEAST ver.2.4.5 (Drummond et al., 2012). The program BEAST permits the user to analyze different gene sequences together, while maintaining the optimal substitution models for each gene (Drummond et al., 2012). We used the uncorrelated lognormal relaxed-clock model and a Yule process of speciation on the tree prior, with birth rate as a gamma distribution (α = 0.001, β = 1,000). Gamma shape was exponential with a mean of 1. Calibration points for diver- Dates used are based on well-supported phylogenetic studies and the fossil record (Perelman et al., 2011;Seiffert, 2007;Yoder et al., 2001).

| Phylogenetic reconstruction and test of selection
We ran four independent Markov chain Monte Carlo (MCMC) runs with 40 million generations with an initial 50,000 burn-in where we sampled every 1,000, which was done for both log and tree files. logenies resulting from a species-tree analysis and found the concatenated tree to be more reliable (Munds et al., 2018;Pozzi et al., 2014).
A maximum-likelihood estimate of the rate of nonsynonymous substitutions per nonsynonymous site to synonymous substitutions per synonymous site (ω = dN/dS) under a codon-based substitution model (codeML) was done using paML v.4 (Yang, 2007). We ran a series of tests to see the effects of natural selection on MC1R. We used the MT-CYB Bayesian partitioned analysis tree from BEAST as our input tree.
Our phylogeny from the MT-CYB tree agrees with past studies that found Lorisidae to represent a monophyletic family (Roos et al., 2004).
The aligned MC1R codon FASTA file was used for the sequence file. A

729-bp alignment of MC1R was used for the three analyzed genera of
Lorisidae (Loris, Perodicticus, Nycticebus). The N. spp. and N. pygmaeus were analyzed separately as previous research found N. pygmaeus to have a deep divergence from N. spp., and even suggested they be their own genera (Munds et al., 2018;Pozzi et al., 2015), giving a total of five analyzed groups, with the outgroup (non-Lorisidae species) as one set. We ran a total of nine models to estimate ω (Yang, 2007) (Table 3).
As demonstrated, there are color differences among the genera of Lorisidae (Nekaris & Bearder, 2007) which prompted us to run clade and branch-site models. These three models permit the researcher to select specific branches or clades to test for comparison to other groups on the phylogeny (Corso et al., 2016;Yang, 2007 TCC AAC ATC TCA GCA TGA  TGA AA   CCC TCA GAA TGA TAT TTG  TCC TCA   55°C   2   Rag2/716  GAT TCC TGC TAY CTY CCT CCT CT  CCC ATG TTG CTT CCA AAC  Likelihood ratio tests (LRTs) were used to test for significant differences among models (Yang, 2007). Using the null model (M0), we examined the specific substitutions that occurred within Lorisiformes. The nonsynonymous mutations were checked for deleterious effects using PROVEAN (Protein Variation Effect Analyzer) v1.1.3 (Choi et al., 2012). This program lets us infer the functional effects of protein differences in Lorisidae MC1R sequences. We used PROVEAN's default threshold of −2.5 for these analyses to determine whether the several analyzed MC1R amino acid mutations were deleterious. To do so, PROVEAN gets amino acid mutation scores by averaging within and across clusters to generate a final score. If that score is above −2.5, then it is predicted to be a deleterious mutation (Choi et al., 2012).

Finally, to better understand amino acid substitutions in
Lorisidae MC1R, we aligned our 315 MC1R amino acid sites to humans. This helped us determine whether mutations along Lorisidae evolution correspond to known MC1R mutations in other organisms.
Particularly, we focused on sites known to influence color variation.
Also, we examined the general location on MC1R (i.e., specific transmembrane, intra-or extracellular loop) and amino acid substitution types (i.e., aliphatic to an acid) ( Figure 3).

| Ancestral state reconstructions
Ancestral state reconstructions of hair color were unable to find a consensus for the hair coloration of the last common ancestor
Of the nine models analyzed in PAML, in no cases were the more complex ones (i.e., branch or clade models) found to be significant over the simpler models (M0 and M1, Table 3). Our finding was further supported by our transition (ts)-to-transversion (tv) ratios, which is expected to be higher in coding genes. This is because there are fewer transversions as they are a passage of purine to pyrimidine or vice versa which result in a change of the amino acid structure.
Transversions are more detrimental than transitions; transitions rarely change the amino acid as they are marked as a change from a purine to a purine or a pyrimidine to a pyrimidine. Hence, a low ts/tv ratio would imply false positives with our data, whereas a high ratio conserves the amino acid structures (Lyons & Lauring, 2017;Yang & Yoder, 1999).
Although our ts/tv results do not vary much, the clade model had the lowest ratio (6.87), and the branch model the highest (8.32) (Table 3).  (Table 4). There were 8 nonsynonymous and 18 synonymous substitutions that led to Galago (Figure 6).

| D ISCUSS I ON
We found MC1R in Lorisidae is under purifying selection (ω = 0.0912), as predicted. Ancestral state reconstructions were inconclusive F I G U R E 3 2D structure of Lorisidae MC1R amino acid sequence with amino acid changes indicated (blue). Mutations affecting coat color in other mammals are also indicated with either a notation for light color (yellow) or dark color (black). Circles that are green indicate both a light coloration in a vertebrate, as well as a mutation found within Lorisidae (figure modified from Buades et al., 2013) on the coloration of the LCA of Lorisidae. Phylogenetic analysis of MC1R found it to be semicorrelated to phylogeny based on MT-CYB and previous more in-depth phylogenetic analyses (Munds et al., 2018;Pozzi et al., 2015). The only exception was when examining the phylogenetic relationship between Nycticebus spp. and N. pygmaeus; here, MC1R results found the separation between the two to be shorter in depth: 21 MYO versus 12 MYO, respectively.
Further examination of MC1R nucleotide differences found 75 variable sites associated with Lorisiformes (Galago and Lorisidae), but only 18 were nonsynonymous, and only 3 of these were found to be deleterious. We predicted most nucleotide site changes would be associated with the more colorful Lorisinae, but our results found that the monochromatic Perodicticus had the most site changes associated with their branch: 6 in total, with one being deleterious.
While our MC1R tree corresponds to Lorisidae phylogeny, the correspondence is not perfect. If MC1R were purely indicative of phylogeny, we would expect distinctions between Lorisidae genera and even species to be discernible, especially as coat colors and patterns are used to discriminate Lorisinae species (Munds et al., 2013;Nekaris & Jaffe, 2007 I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I  II  III  IV  V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III   IV   V   I   II   III Nunes et al., 2011;Pointer & Mundy, 2008;Shimada et al., 2009).

One explanation as to why purifying selection is common with
MC1R studies is that single nonsynonymous substitutions in MC1R, and similar to many other coding genes, cause large phenotypic changes that can be detrimental to the survival of an organism (Cvijovic et al., 2018;Hoekstra, 2006;Mundy & Kelly, 2003;Nunes et al., 2011;Theron et al., 2001). Purifying selection acts on coding, nuclear, and even mitochondrial genes as a way to rid the genome of deleterious mutations, particularly when such changes could have pleiotropic effects (Cvijovic et al., 2018;Hamosh et al., 2005;Popadin et al., 2013). For example, in male tawny owls (Strix aluco) there are two morphs (light and dark) due to melanocortin system mutations, with darker males not only having a higher survival rate but also producing higher quality, dark morph offspring. Studies We expected MC1R nucleotide changes to be associated with the more patterned and colorful Lorisinae (Loris and Nycticebus), but that was not the case. We predicted this as researchers have speculated Lorisinae patterning and coloration are adaptive (Nekaris & Bearder, 2007;Nekaris et al., 2013Nekaris et al., , 2019. Our study was unable to truly test the relationship between venom and coloration, as only Nycticebus possess venom, and the genetics of venom production in this species is still being explored. We hope in the future to determine the correlation between venom and coloration in these mammals. But the results from our study do setup exploring the proximate molecular mechanisms regarding the adaptive purposes of coat colors and patterns; future research should bridge these gaps as has been done in beach and deer mice (c.f., Barrett et al., 2019;Hoekstra, 2006;Nachman et al., 2003). Only four nonsynonymous mutations occurred on Lorisinae branches, with one resulting in a charge change (Ala61Asp). The majority of nonsynonymous charge changes in Lorisidae (n = 6) were found on the Perodicticus branch.  I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV   I   II   III   IV  I   II   III   IV  I   II   III   IV   I   II   III   IV   I   II   III  Nonfilled circles indicate non-charge-changing amino acid substitutions. Filled circles are charge-change amino acid substitutions. Site numbers correspond to specific sites found within this study's MC1R sequence. Those marked with a * are ones that were found to be deleterious. Node numbers are posterior probabilities F I G U R E 7 Bayesian MC1R tree. Node labels are posterior probabilities. Note N. coucang and some N. pygmaeus have low probability of separation using the MC1R phylogeny. Scale in millions of years ago third extracellular loop (Ala191Thr) (Figure 3). Changes in the first intracellular loop of MC1R impact signaling activity, often as a loss of function (Wolf Horrell & D'Orazio, 2016). Yet, the most significant substitutions for Perodicticus occurred in the fourth transmembrane. Transmembrane regions are noted to be highly conserved and changes in these areas are often damaging (Peters et al., 2016). Furthermore, only Perodicticus possessed a major deleterious mutation. Although our findings are based on captive Lorisidae, of the 9 genera and 11 primate species examined only Perodicticus had the Ser to Phe deleterious mutation (Ser169Phe, −4.393). Other studies on MC1R in primates did not find the Ser to Phe mutation, but these studies did not include Perodicticus (Mundy & Kelly, 2003). This Ser to Phe nonconservative substitution in MC1R has been noted to result in a blanched, paling, or lighter phenotype in horses and fur seals (Marklund et al., 1996;Peters et al., 2016). Not only is this particular mutation associated with a lighter phenotype, but also studies have found substitutions in the fourth transmembrane result in lighter phenotypes too. Fourth transmembrane mutations in humans (R160W; R163Q; I180H) result in red hair and fair skin (Dessinioti et al., 2011;Gerstenblith et al., 2007;Pavan & Sturm, 2019;Raimondi et al., 2008). For lizards, a fourth transmembrane of Thr to Iso at 170 results in a blanched phenotype (Rosenblum et al., 2010); Perodicticus possess an identical Thr to Iso substitution at 171. trait between unrelated species (Zhang & Kumar, 1997). Initially, parallel phenotypes were thought to arise from differing genetic mechanisms, but more studies are finding these phenotypes are from changes in the same genes, even across diverse taxa (Hoekstra, 2006;Miller et al., 2007). In fact, our MC1R study found that 5 Perodicticus substitutions along with 1 mutation in Galago (Ala183Val) are known to cause lighter coloration in other vertebrates (Table 4; Figure 3). Two of these substitutions (Ala to Val and Ala to Thr) are attributed to causing red coloration in pigs (Peters et al., 2016). The other two mutations of Perodicticus (Ser169Phe, Thr171Iso) could be contributing to their monochromatic, blanched phenotype (Peters et al., 2016;Rosenblum et al., 2010). Not only is the specific type of amino acid substitutions noted to cause lighter coloration in other vertebrates, but also the location of these sub-  (Caro, 2005;Nekaris et al., 2013). Many semiarboreal and arboreal nocturnal mammals still possess striking patterns and contrasting facemasks and dorsal stripes (i.e., felines, sugar gliders, civets) for signaling in the dim moonlight (Allen et al., 2010;Ancillotto & Mori, 2016;Caro, 2013). Losing these signals could impact survival for Lorisinae.
Yet, others have noted, MC1R is not strongly correlated to coat color in primates (Bradley et al., 2013;Bradley & Mundy, 2008;Haitina et al., 2007;Mundy & Kelly, 2003). It is likely that several genes are responsible for the color variation observed in Lorisidae.
Ours is not the first study to come to this conclusion, and other proposed promoter genes, such as ASIP, TYRP1, solute carrier family 24 member 5 (SLC24A5), and dopachrome tautomerase (DCT) may influence coat colors but research that has examined their impact was not conclusive (Bradley et al., 2013;Haitina et al., 2007;Mundy & Kelly, 2006). More research is needed to better understand the coat color and pattern genetics of primates. Incorporating genomewide studies and transcriptomics will be invaluable in understanding why, of all mammals, primates are so vibrantly colored and patterned (Bradley & Mundy, 2008).
Our study found several MC1R amino acid substitutions that appear to influence the hair color, and possibly skin color, of Lorisidae primates. In particular, the substitutions on the Perodicticus branch are strongly correlated to other vertebrate studies that found similar MC1R substitutions to be associated with lighter hair and/or skin color. Although putative, such a finding is supported by our phenotypic analyses which demonstrate that Perodicticus possess a monochromatic, lighter phenotype in comparison with Lorisinae. We recognize the limitations of our study as it is difficult to definitively answer evolutionary questions based on captive populations. Yet, such research provides a window into understanding and developing future projects that can further explore this topic in-depth. Lorisidae species are challenging to catch in the wild (Nekaris et al., 2020), but they are a fascinating family to study the evolution of coat color and patterns for aposematism (Nekaris et al., , 2020. Future research on these primates and on these topics should examine the genetics of wild populations and incorporate other genes that are known to influence patterning and coloration. Furthermore, linking the genetic analyses directly to phenotypes will help determine whether the patterning and coloration are adaptive. Such work will help us better understand the colorful palette displayed by primates and ascertain the adaptive purposes behind the colors.

ACK N OWLED G M ENTS
We thank the American Zoological Association for their assistance with this project, and the participatory zoos and institutions that Dr. G. Duytschaever for helping with hair and skin coding, as well as providing advice on ancestral state reconstruction analysis; and K. Budd for her assistance in uploading sequences to GenBank.
We also thank Dr. A Melin, our review editor, and two anonymous reviewers for providing invaluable suggestions to the improvement of this manuscript. Finally, we thank Sigma Xi for their financial support.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest. writing-original draft (equal); writing-review & editing (equal).

E TH I C A L A PPROVA L
The authors adhered to appropriate animal care guidelines maintained by the University of Missouri. Although this study was noninvasive, we followed The Animal Behavior Society's Guidelines.