Adaptive Evolution of the Insulin Gene in Caviomorph Rodents
http://www.100md.com
分子生物学进展 2005年第5期
* Center for Advanced Studies in Ecology and Biodiversity, Departamento de Ecología, Facultad de Ciencias Biológicas, and Departamento de Genética Molecular y Microbiología, Pontificia Universidad Católica de Chile, Santiago, Chile; and Laboratorio de Evolución, Facultad de Ciencias, Universidad de la República, Montevideo, Uruguay
Correspondence: E-mail: jopazo@genetics.wayne.edu.
Abstract
Insulin is a conservative molecule among mammals, maintaining both its structure and function. Rodents that belong to the Suborder Hystricognathi represent an exception, having a very divergent molecule with unusual physiological properties. In this work, we analyzed the evolutionary pattern of the insulin gene in caviomorph rodents (South American hystricomorph rodents). We found that these rodents have higher rates of nonsynonymous:synonymous substitutions (dN/dS) than nonhystricomorph rodents and that values are heterogeneous inside the group. We estimated codons under positive selection, specifically the second binding site (A13 and B17) and others related with hexamerization (B18, B20, and B22). In the monomer structure, all selected sites formed a single patch around the second binding site. In the hexamer structure, these amino acids were grouped into three major patches. In this structure, contacts between B chains involved all selected sites (except B18), and between faces in the center of the molecule, all contacts were among selected sites. While there is no clear hypothesis regarding the cause of this drastic change, experimental evidence does show that this group of rodents has some peculiarities in growth function, and, whether coincidental or not, these changes appeared together with important changes in life-history traits.
Key Words: insulin ? caviomorph ? adaptive evolution ? life-history traits ? Hystricognathi
Introduction
Insulin is a hormone secreted by beta cells in the Langerhans islets of the pancreas, which is released mainly in response to blood glucose levels. Among mammals the structure and function of insulin is well conserved (Chan and Steiner 2000; Conlon 2001). In fact, the regulatory properties of insulin are interchangeable between species. Nevertheless, hystricomorph rodents are an exception among mammals (Neville, Weir, and Lazarus 1973; Conlon 2001), possessing an insulin molecule with many amino acid substitutions, which has important effects on several physiological properties of the molecule. Among these peculiarities is the observation that hystricomorph insulin is not neutralized by the antibovine insulin antibody (Davidson, Zeigler, and Haist 1968, 1969), indicating that amino acid substitutions have changed the immunological properties of the molecule. Additionally, insulin from this group cannot self-associate in hexamers, a structure which is used by nonhystricomorph mammals to store the hormone in secretion vesicles (Wood et al. 1975; Bajaj et al. 1986). It is possible that hystricomorph rodents are unable to store their insulin or that storage is accomplished via another unknown way. An aspect which has been better characterized is insulin's diminished biological activity in this group of rodents. Among hystricomorphs, there are variable degrees of change in biological activity; the lowest activity change is seen in rodents of the superfamily Chinchilloidea (chinchillas and vizcachas), and the highest activity change is observed for rodents of the superfamily Octodontoidea (degus, tucu-tucus, spiny rats, and allies) (Zimmerman, Moule, and Yip 1974; Horuk et al. 1979; Bajaj et al. 1986). Despite this variation in biological activity, hystricomorphs can reach the same maximal response as nonhystricomorph mammals, although at greater insulin concentrations (Zimmerman, Moule, and Yip 1974; Bajaj et al. 1986). Despite these atypical physiological properties, hystricomorph rodents are able to regulate glycemia like other nonhystricomorph mammals (Opazo, Soto-Gamboa, and Bozinovic 2004). Finally, hystricomorph insulin possesses greater growth-promoting activity than other mammals (King and Kahn 1981) and is able to bind to additional growth factor receptors (King, Kahn, and Heldin 1983).
The unique insulin molecules of hystricomorph rodents have been an enigma for many years. Understanding how changes at the molecular level produce changes in the biological function of this molecule is of great interest to researchers, especially considering insulin's central role in physiology and the serious health consequences that occur when the molecule malfunctions. The task of evaluating the structure-function relationship of this molecule is difficult due to the lack of natural variants of insulin among most mammals. In this sense, an analysis of the insulin gene in this group of rodents will have an important impact on future studies.
The objective of this study is to describe the evolutionary pattern of the insulin molecule in caviomorph rodents (South American hystricomorph rodents). To this end, we sequenced the insulin gene in representatives of the caviomorph rodents and analyzed these sequences at two distinct levels. First, we analyzed the variation in the ratio of nonsynonymous substitutions to synonymous substitutions among lineages of caviomorphs in a phylogenetic framework. Second, we carried out analyses directed at identifying which amino acids were under positive selection, mapping these sites onto the established monomer and hexamer structures and determining their location in relation to the molecule function.
Materials and Methods
Rodent Species
We obtained complete sequences of both insulin peptides for 21 caviomorph rodent species, representing the superfamilies Cavioidea, Chinchilloidea, and Octodontoidea (supplementary table 1).
DNA Extraction, Amplification, and Sequencing
Alcohol-preserved samples were subjected to proteinase K digestion, NaCl precipitation of proteins, and DNA precipitation with ethanol (Miller, Dykes, and Polesky 1988). Each chain was amplified separately by the polymerase chain reaction (PCR). PCR conditions were as follows: 30 cycles of alternating denaturation at 94°C for 30 s, annealing at 50–58°C for 30 s, and extension at 72°C for 30 s. An initial denaturation at 94°C for 2 min and final extension at 72°C for 2 min were also performed. All experiments included negative controls. Primers used to obtain chain B were Vaca-3 CGCCATGGCCCCGTGGATGC (forward), Vaca-4 CTGGCCCTGCTGGCCCTCTG (forward), and Láctica-3 AGGGGCTCACCCKKTGGGTCCTC (reverse). To obtain chain A, we used Pata-10 CTGCTGCTCCTGACAGCRTCT (forward), Pata-13 GGYCTTCTGCTGCTCCTGAC (forward), Pata-17 GGCGCTGTGCTGCTCCTRACA (forward), Cabra-5 TAGACACCTGCCTTGGGCCTGG (reverse), and Cabra-6 CATTCAAGGGGTTTATTGGTTGCCA (reverse).
PCR products were visualized in 5% nondenaturing minigels, and the remaining product was cleaned in Sephadex columns and used as templates for sequencing both strands.
Statistical Analysis
Variable (=dN/dS) Rates Among Lineages
To detect the possible role of positive selection among specific lineages in the phylogeny (fig. 1), we used the maximum likelihood codon substitution model of Goldman and Yang (1994) implemented in the PAML 3.13 package (Yang 1997). This model takes into account the structure of the genetic code, transition/transversion ratios, and base frequencies at the three codon positions.
FIG. 1.— The tree topology used to conduct the analyses of variable (=dN/dS) among lineages and sites. Branches labeled from 1 to 4 and a to f were used to test hypotheses mentioned in Materials and Methods. In the case of variable among sites, we excluded noncaviomorph mammalian species. Tree topology is based on published literature (Murphy et al. 2001; Leite and Patton 2002; Rowe and Honeycutt 2002; Honeycutt, Dowe, and Gallardo 2003; Spotorno et al. 2004; Springer et al. 2004).
Several models of variable (=dN/dS) rates among lineages were implemented using the 21 aligned sequences of caviomorph rodents and 7 sequences of noncaviomorph mammals (supplementary table 1), based on figure 1. The simplest model assigns the same for all branches, while the free ratio model assumes an independent for each branch. Intermediate models with phylogenetic sense were also implemented. Each of these models makes different assumptions about the value of for branches labeled 1 to 4 and the branches that descend from them (fig. 1). For instance, the one-ratio model assumes the same for ancestral branches of the noncaviomorph rodents and caviomorph species belonging to the superfamilies Cavioidea, Chinchilloidea, and Octodontoidea and also for the branches that descend from them (fig. 2A). The two-ratio model assumes 1 for the ancestral branch of noncaviomorph rodents and their descendants and another for the ancestral branches of the caviomorph superfamilies and their descendants (fig. 2B). The 4 model assumes that each ancestral branch of the caviomorph superfamilies and their descendants has an independent rate (fig. 2C). The 8 model assumes that the descendant branches have independent rates from the ancestral branches (fig. 2D). The last model implemented examined variation at the family level among members of the superfamily Octodontoidea. This model was implemented because members of this group have more amino acid substitutions than the members of the other superfamilies and also present insertions and deletions, a characteristic that is not shared with other caviomorph rodents. In this vein, we reproduced the 8 model; however, we assigned specific parameters for branches a, b, c, d, e, and f (fig. 1). In this model, the descendant branches from e (corresponding to the ancestor of the family Octodontidae) and f (the ancestor of the family Ctenomyidae) had independent values, and the total model included 15 parameters. Three starting values (0.5, 1, and 2) were used to check the existence of multiple local optima. Nested models were compared using the likelihood ratio test.
FIG. 2.— Graphical representation of nested models implemented to test the possible role of positive selection over lineages. For details see Materials and Methods.
Variable (=dN/dS) Rates Among Sites
Models of variable among sites were used to test for the existence of particular sites under positive selection and to identify these sites. We used the models recommended by Yang et al. (2000) implemented in the CODEML module of the PAML 3.13 package (Yang 1997); for this analysis, we used tree topology of figure 1 excluding the noncaviomorph mammals. Model M0 assumes constant rates across sites. Model M1 (neutral) assumes two site classes with values of 0 (purifying selection) and 1 (neutral). In addition to the classes mentioned for M1, the M2 Model (selection) assumes a third category of sites with an value estimated from the data. Model M3 (discrete) has three classes of sites with proportions p0, p1, and p2 and 0, 1, and 2 values estimated from the data. Model M7 assumes a beta distribution between 0 and 1 depending on the parameters p and q. Finally, Model M8 adds an extra class of sites with values and proportions estimated from the data. Only Models M2, M3, and M8 can detect sites under positive selection. Because some models, especially M2 and M8, are prone to multiple local optima, we used three starting values (0.5, 1, and 2) to check for the existence of multiple local optima. When the estimation of the parameters was finished, Bayes theorem was used to calculate the posterior probability that each site belonged to one site class (Nielsen and Yang 1998; Yang et al. 2000).
Obtaining reliable estimations of sites under positive selection demands that the data meet certain requirements. Anisimova, Bielawski, and Yang (2002) used simulations to evaluate the accuracy and power of the Bayesian approximation for estimating sites under positive selection. They concluded that the best way to improve accuracy and power is to use sequences that are both divergent and numerous and to apply multiple models to the data. In another study, the same authors (Anisimova, Bielawski, and Yang 2001) analyzed the power of the likelihood ratio test to detect adaptive molecular evolution. They showed that the power was nearly 100% when they had at least 17 sequences in the data set. Other data set characteristics such as length, divergence, and the strength of positive selection affect the power of the likelihood ratio test. All of these requirements are fulfilled by our data; therefore, we believe that our estimations will be reliable.
Location of Selected Codons in Insulin Monomer and Hexamer and Contact Maps
To determine where selected amino acids are located in the insulin molecule, we obtained the monomer structure of pig insulin (Badger et al. 1991) and the hexamer structure of human insulin (Chang et al. 1997) from Protein Data Bank (Berman et al. 2002). Contact maps were calculated for the hexamer structure using software that we developed, which is available for free upon request. All chain-chain contacts were calculated, including the homomeric contacts. In each case, three independent files were generated. The first considers only side chain–side chain atomic contacts, the second considers only main chain–main chain atomic contacts, and the third considers all atomic contacts between two given residues. In all cases where at least one pair of residues exhibited more contacts than those specified in the threshold of contacts (based on the program's input), a bidimensional density plot was generated. The plots were created automatically as postscript files using the free ASGL software. This program requires several parameters as input, including the distance cutoff for defining a contact between two atoms in three-dimensional space, a minimum graphical threshold for the number of atomic contacts between two residues that are to be plotted, and the minimum sequence separation between two residues to be considered a contact (Melo and Feytmans 1997). In our work, the distance cutoff was fixed at 4.0 ?, the minimum number of atomic contacts was set to 5, and the minimum sequence separation was defined with a value of three residues in the case of contacts within the same chain (in the case of contacts between two different chains, this value was not used). These parameters will only select pairs of nonlocal residues that have a substantial fraction of their atoms in close proximity in three-dimensional space (Melo and Feytmans 1998; Melo, Sánchez, and Sali 2002). We used the contact maps to evaluate whether selected sites are involved in contacts in the hexamer structure.
Results
Variable (=dN/dS) Rates Among Lineages
The model assuming the same value for all branches in the phylogeny leads to l1 = –1,368.81, with an average value over all sites and lineages of 0.23. The log likelihood value of the model that assumes an independent for each branch (free model) was l0= –1,297.49. The free ratio model involved 53 parameters for 53 branches, while the model assuming 1 for all branches involved 1; then, 2l = 2(l1 – l0) = 142.62 was compared with a 2 distribution with df = 52 to test whether the free ratio model provided a significantly better fit to the data. The difference between models was significant (P < 10–5), indicating that rates are different among lineages.
Values of estimated from the different models are shown in table 1. The model distinguishing between noncaviomorph rodents and caviomorph rodents (fig. 2B) had a significantly better fit (P < 10–5) than the model that did not (fig. 2A). The value estimated for ancestral and descendant branches of the caviomorph rodents was 0.48, more than 40-fold the value estimated for noncaviomorph rodents (table 1). The model segregating between noncaviomorph rodents and within the caviomorph group (fig. 2C) did not have a better fit than the model that only segregated between noncaviomorph rodents and caviomorph rodents (fig. 2B) (P = 0.34). To test whether ancestral and descendant branches had their own values, we compared the model that assumed an independent value for each branch (fig. 2D) with a 2 model (fig. 2B). These models were not significantly different; the 8 model did not have a better fit than the 2 model (P = 0.72). Finally, the model that discriminates between rates within the superfamily Octodontoidea (fig. 1) was compared with the 2 model. The first model had an almost significantly better fit than the 2 model (P = 0.07). This model indicates that each labeled branch has its own parameter (fig. 1) in addition to the descendants of branches 1, 2, and 3. Among the ancestral branches of families within the superfamily Octodontoidea, we observed the following decreasing pattern: the ancestral branch of the family Echimyidae had the highest value (branch c = 0.69), followed by Abrocomidae (branch a = 0.53), and then the Octodontidae family (branch e = 0.25), and the lowest value was observed for the ancestral branch of the family Ctenomyidae (branch f = 0.03). The external branches for the family Octodontidae maintained their value (0.27) in relation to the ancestral branch, while the external branches of the family Ctenomyidae increased (0.08). The ancestral branch of the families Ctenomyidae and Octodontidae had the highest values among the entire superfamily, having only nonsynonymous substitutions (branch d = 10.9/0), while branch b had a lower value (0.06).
Table 1 Log Likelihood Values and Parameter Estimates Under Different Lineage Models
Variable (=dN/dS) Rates Among Sites and Identification of Amino Acids Under Positive Selection
Table 2 shows the parameters estimated under variable selective pressure among sites using the unrooted tree topology of figure 1 without the out-group (noncaviomorph mammals). Models designed to detect positively selected sites (M2, M3, and M8) were significantly better, and all models converged to the same sites. Sites B17, B22, and A13 had a posterior probability of being in the selected class of 99%. Sites B18 and B22 had smaller posterior probability values (table 2). The value for the selected site class varied from 3.38 to 4.15, indicating strong positive selection acting on these sites. The likelihood ratio tests were always in favor of models with the ability to detect sites under positive selection.
Table 2 Parameter Estimates and Log Likelihood Values Under Models of Variable Rates Among Codons
Location of Selected Sites in the Monomer Structure, Hexamer Structure, and Contact Maps
Because the crystal structure of caviomorph insulin has not yet been determined, we used a noncaviomorph insulin structure to evaluate the spatial location of selected sites. In the monomer structure, the selected sites are grouped together in a single patch (fig. 3A), compromising the second binding region (fig. 3B). In the hexamer structure, the selected sites are grouped (fig. 3C) in three discrete patches (fig. 3D). This phenomenon is produced because two monomer patches are contiguous in the hexamer structure.
FIG. 3.— Space-filling structure of the insulin monomer and hexamer obtained from Protein Data Bank. Chain A is shown in blue and chain B in red. (A) Insulin monomer structure showing estimated sites under positive selection (yellow). (B) Insulin monomer structure showing residues corresponding to the putative second binding region (B17 in yellow and A13 in light blue). (C) Lateral view of the hexamer insulin structure showing the estimated sites under positive selection (yellow). (D) Frontal view of the hexamer insulin structure showing the estimated sites under positive selection (yellow).
Considering the main chain contact maps between B chains in the hexamer structure, residue B20 is involved in contacts. When we consider main and lateral chains, all selected sites (except B18) are involved in contacts. At the center of the molecule, where the zinc is coordinated, all contacts between hexamer sides are among selected sites.
Discussion
Insulin Structural Patterns
According to the data presented here and previously published, it is possible to differentiate, based on the length of the insulin, two groups of insulin molecules among caviomorph rodents. The first group includes the molecule belonging to the species of the superfamilies Cavioidea and Chinchilloidea, which conserves the number of residues (chain A with 21 and chain B with 30 residues). The second type of insulin molecule is found in caviomorph species belonging to the superfamily Octodontoidea. In this group, the insulin molecule presents a phenylalanine deletion at position B24 and an insertion of two amino acids at the carboxy-terminal end of chain A. We can infer that these changes were produced gradually because the species with intermediate structures are currently living. Concordant with the phylogeny of the group (Honeycutt, Dowe, and Gallardo 2003), Abrocoma bennetti (representing the species of the family Abrocomidae) is the most basal clade of the superfamily and presents only the phenylalanine deletion in chain B (fig. 4). In turn, Myocastor coypus (representing the family Myocatoridae) insulin sequence obtained from GenBank presents the phenylalanine deletion as well as an insertion of 1 amino acid in the carboxy-terminal end of the A chain (Bajaj et al. 1986). While the phylogenetic position of this species in not well resolved (Leite and Patton 2002), the structure of the insulin molecule strongly suggests that M. coypus is the sister group of the species of the families Echimyidae, Ctenomyidae, and Octodontidae (fig. 4). The other families (Ctenomyidae, Echimyidae, and Octodontidae) have a chain B deletion and two amino acid insertions at the carboxy-terminal end of chain A.
FIG. 4.— Cladogram of the superfamily Octodontoidea showing when insulin changes occurred.
Chain B partial sequences obtained from members of the families Bathyergidae (Cryptomys damarensis), Hystridae (Hystrix africaeaustralis and Atherurus macrourus), and Erethizontidae (Coendu bicolor and Erethizon dorsatum) suggest that these clades have insulins of standard length (unpublished data). For the Hystricidae family, this pattern is confirmed by protein sequence data of the porcupine (Hystrix cristata) (Horuk et al. 1980).
Protein sequence of the Echimyidae species, Proechimys guairae (Horuk et al. 1979), shows a B chain deletion; however, the A chain does not have any insertions. This result contrasts with our nucleotide sequence data in Hoplomys gymnurus, which presents a deletion in chain B and two amino acid insertions in chain A. Additionally, the genera Proechimys and Hoplomys are grouped together in the phylogeny (Leite and Patton 2002).
Rate Variation Among Lineages
We have drawn two major conclusions from maximum likelihood analysis of rate variation among lineages. First, that the ancestral branches of caviomorph superfamilies have higher (=dN/dS) rate ratios than noncaviomorph rodents, and second, that these values are heterogeneous. Omega values for the ancestral branches of each caviomorph superfamily were always higher than the ancestral branch for noncaviomorph rodents (table 1). The ancestral branches of the caviomorph superfamilies were not demonstrably affected by positive selection (using dN/dS >1 as an indicator) (Yang 2001). Nevertheless, for the ancestral branch of the superfamily Cavioidea were estimated 6.4 nonsynonymous substitutions, 12.8 in the stem of the superfamily Octodontoidea, and 2.4 for Chinchilloidea. Furthermore, the values for the ancestral branches of these superfamilies were between 3,000 and 8,300 times higher than the ancestral branch of noncaviomorph rodents (table 1).Moreover, physiological evidence demonstrated that the insulin molecule in species of these superfamilies also has divergent physiological properties (Horuk et al. 1979; King and Kahn 1981; King, Kahn, and Heldin 1983). Additionally, it has been documented that important changes in function are not necessarily the result of numerous amino acid changes; rather, a few changes in key positions are enough to change the properties of the molecule (Golding and Dean 1998). Although these few changes could not elevate the overall ratio over 1, it is possible to identify which sites are under positive selection (Yang et al. 2002; Civetta 2003). It is important to note that in spite of heterogeneous values among lineages, there is a relationship with physiological divergence (defined as the decrease in biological activity in comparison with noncaviomorph mammals). The lowest value (0.30) and the least physiological divergence (Horuk et al. 1979) were seen for the ancestral branch of the superfamily Chinchilloidea, and the highest values of were observed for the superfamily Cavioidea (Zimmerman, Moule, and Yip 1974). The superfamily Octodontoidea is a special case because its divergence is underestimated as PAML package does not take indels into account (Yang 1997). Nevertheless, it has been documented that the physiological divergence of this group is greater than that for members of the superfamily Cavioidea (Bajaj et al. 1986). In this sense, if the relationship between the value and the physiological divergence is correct, the intensity of selection in this group should be greatest among caviomorph rodents. Moreover, if this relationship is indeed genuine, it would be an important fact because, until now, the value has been used as a discrete indicator, where values greater than 1 are indicative of positive selection events and values below 1 indicate a lack of positive selection. Our data indicate that differences resulting from different values during evolution could possibly be measured functionally. On the other hand, it is important to comment on the importance of experimental evidence, when the lineage analysis could not detect positive selection events with this criterion. In this sense, the ratio between the control group (i.e., noncaviomorph rodents) and the group of interest could possibly be an alternative view for estimating positive selection.
Test of Positive Selection Among Sites
Analyses of variation among sites provide clear evidence of positive selection. The results show that four of the five sites estimated under positive selection are in chain B. Interestingly, these sites are not in the classical receptor binding domain or the domain determined by alanine-scanning mutagenesis, in concordance with the previous analysis based on amino acid sequences (Conlon 2001). Among noncaviomorph mammals, amino acidic variation is confined to the hypervariable region A8 to A10 and the extreme N- and C-terminals of chain B. These regions are not considered to be important in determining the biological activity of the molecule (Kristensen et al. 1997). The positively selected site in chain A is not in the hypervariable region described for mammals. Rather, some sites of chain B are in the middle of the peptide (B17, B18, and B20), while one is in the C-terminal portion of the molecule (B22). Among these sites under positive selection, we believe that sites B17 and A13, which had the highest posterior probabilities of being in the proportion of sites under positive selection, may be the most important in causing the decreased biological activity of insulin in caviomorph rodents because they were described as the putative second binding region (De Meyts 1994; Sch?ffer 1994). De Meyts (1994) described a binding model in which he established the relationship between insulin molecules with an altered hexamer-forming surface (e.g., A13, B17, B18, and B20) and the rate of association and dissociation from the insulin receptor, negative cooperativity, and metabolic and mitogenic activity (for more details, see De Meyts 1994). In this model, De Meyts (1994) explained the higher mitogenic and lower metabolic activities of this kind of insulin (i.e., hystricomorph insulin) as an asymmetrical phosphorylation of the receptor when the first insulin does not bind through its hexamer-forming surface (Lee et al. 1993). Alanine-scanning mutagenesis experiments give independent information regarding the importance of the second binding region. Replacement of the residues A13 and B17 by alanine results in a decrease of 70% and 38% in binding affinity, respectively, for the low-affinity binding site (Kristensen et al. 1997) and recently tested, with similar results, for the high-affinity binding site (De Meyts and Whittaker 2002). Nevertheless, this evidence should be taken with caution because none of the caviomorph rodents sequences presented in our study has an alanine residue in position B17 or A13, except for Dolichotis patagonum. Additionally, these measurements represent a change of just one amino acid in the molecule; the effect of multiple ones (like in caviomorph rodents) other than alanine replacements is unknown.
Residue B22 (arginine) has been described as being trypsin sensitive in nonhystricomorph mammals, so the change in hystricomorph rodents due to the inability to form hexamers would make the insulin of these rodents most stable in a monomeric state (Blundell and Wood 1975). Nevertheless, this change would produce a decrease in the biological activity of the molecule (Horuk et al. 1979, 1980). This explanation is based on a Cavia porcellus sequence that has an aspartic acid residue in position B22. In this sense, the interaction of the negatively charged B22 and A21 residues would produce a change in the structure, which would have functional consequences (Horuk et al. 1979, 1980). For the sequences obtained in this work, all A21 residues were asparagine. Nevertheless, among chain B sequences, only two species (C. porcellus, Cavia aperea) had an aspartic acid in position B22. Furthermore, some species have both positions like noncaviomorph mammals (A21 asparagine and B22 arginine) (Chinchilla lanigera, Chinchilla brevicaudata, D. patagonum, Hoplomys gimnurus). Of these species, C. lanigera shows diminished biological activity (Horuk et al. 1979). Thus, the simple replacement of the B22 position by a negatively charged residue cannot alone explain reduction in activity. We believe that deviations in physiological activity are the result of more than just the sum of effects due to each individual substitution.
Spatial Location of Selected Sites in the Monomer and Hexamer Structures and the Contact Maps
It is interesting that the spatial location of the sites estimated under positive selection formed a single patch in the monomer structure. This patch is located near the second binding site (fig. 3A and B), in agreement with the binding model of De Meyts (1994) and his explanation of the atypical physiological properties. Consistent with this, the hexamer structure also presents a nonrandom pattern, where we were able to differentiate three discrete patches (fig. 3D), each consisting of two monomer patches (fig. 3C). This spatial distribution of the sites estimated to have evolved under positive selection in the hexamer structure might explain, in part, the inability of hystricomorph rodent insulin to aggregate.
The main molecular interactions established in the hexamer structure are between the B chains of the two insulin monomers and among B chains at the center of the structure, where the zinc atom is coordinated. Our results showed that in the main chain interactions among B chains, sites under positive selection represented 33% of total contacts. When we considered main and lateral chains, all sites estimated under positive selection were involved, except site B18. At the center of the structure the amino-terminal ends of the B chains established contacts on both sides, in which no selected sites were involved. Nevertheless, B chains established few contacts between the sides of the structure, in which all selected sites were involved. We know that contact maps are not absolute proof of the cause of the inability of caviomorph insulin to form hexamers; however, we believe that this analysis is a good initial approximation to the problem. It is important to highlight the fact that there are other contacts inside the hexamer insulin structure; however, it is not our purpose to discuss all of them here.
Concluding Remarks
When positive selection events occur, we are at least able to distinguish between events that increase the established function of a molecule (Grossman et al. 2001) and those in which the molecule acquires a new function (Messier and Stewart 1997). Considering the limited information available regarding insulin physiology in hystricomorph rodents, we believe that this is a case in which the increase in nonsynonymous substitutions is associated with the acquisition or expansion of the function. There is not enough evidence to conclude what this new function is; however, we suspect that it may be related to the unusual patterns of pre- and postnatal growth of caviomorph rodents.
King, Kahn, and Heldin (1983) showed that the insulin of C. porcellus can bind to the receptor of the platelet-derived growth factors and that this binding was not competed either by the insulin of other mammals or by growth factors, where insulin growth factor (IGF)-I was tested. Additionally, when researchers measured thymidine incorporation (as a measurement of growth-promoting function), noncaviomorph insulin did not show additivity. However, when pig and caviomorph insulins were tested together, researchers obtained additive response. Hystricomorph insulin shares some physiological properties with IGF-I: neither are able to aggregate (De Meyts 1994), both present a lack of self-antagonism in a dose-response curve for negative cooperativity (Christoffersen et al. 1994), and both have a growth-promoting function (King and Kahn 1981). At the systemic level, it has been demonstrated that C. porcellus is insensitive to growth hormone, although the structures of the hormone and its receptor are conserved (Adkins, Vandeberg, and Li 2000) and the gland produces and releases the hormone normally (Gabrielson, Fairhall, and Robinson 1990). Additionally, when the effector of the growth hormone is blocked (IGF-I), there are no effects on the growth pattern of individuals (Kerr, Laarveld, and Manns 1990). Finally, in contrast to other rodents, IGF-II levels do not drop after birth, instead hystricomorph rodents maintain IGF-II levels throughout their adult life (Lenovinovitz et al. 1992). This collection of evidence indicates that pre- and postnatal growth regulations are different for this group of rodents.
In sum, we believe that the divergence of caviomorphs (relative to rodents of the Suborder Sciurognathi) and the drastic changes in life-history traits and growth pattern are more than just a coincidence.
Supplementary Material
Supplementary table 1 is available at Molecular Biology and Evolution online (www.mbe.oupjournals.org).
Acknowledgements
We acknowledge I. Tomasco, G. Wlasiuk, A. Castillo, and G. D'Elía for their help with laboratory work. We also thank M. Soto-Gamboa and M. A. Lardies for their assistance in the field and Z. Yang for his assistance with PAML. This work would not have been possible without tissue loans from the following institutions: the Museum of Vertebrate Zoology at the University of California at Berkeley, the Museum of Southwestern Biology at the University of New Mexico, the Museum of Zoology at the University of Michigan, Colección de Flora y Fauna Profesor Patricio Sánchez Reyes at Pontificia Universidad Católica de Chile, Laboratorio de Citogenética de Mamíferos at Universidad de Chile, Conrad Matthee at the Zoology Department of Stellenbosch University, Guillermo D'Elía from the Laboratorio de Evolución at the Universidad de la República, and Hynek Burda from the Department of General Zoology at the University of Essen. Derek E. Wildman and Lawrence I. Grossman made useful suggestions on early versions of the manuscript. This work was supported by a doctoral thesis fellowship (2002) from Comisión Nacional de Desarrollo Investigación Y Tecnológico to J.C.O and the Center for Advanced Studies in Ecology & Biodiversity program 1 (F. Bozinovic) and 2 (F. Jaksic) and Fondo Nacional de Desarrollo Cientifico Y Tecnológico grant 1010959 to F.M.
References
Adkins, M., J. Vandeberg, and W.-H. Li. 2000. Molecular evolution of the growth hormone and receptor in the guinea-pig, a mammal unresponsive to growth hormone. Gene 246:357–363.
Anisimova, M., J. P. Bielawski, and Z. Yang. 2001. Accuracy and power of the likelihood ratio test in detecting adaptive molecular evolution. Mol. Biol. Evol. 18:1585–1592.
———. 2002. Accuracy and power of Bayes estimation of amino acids sites under positive selection. Mol. Biol. Evol. 19:950–958.
Badger, J., M. R. Harris, C. D. Reynolds, A. C. Evans, E. J. Dodson, G. G. Dodson, and A. C. T. North. 1991. Structure of the pig insulin dimer in the cubic crystal. Acta Crystallogr. B 47:127–136.
Bajaj, M., T. L. Blundell, R. Horuk, J. E. Pitts, S. P. Wood, L. K. Gowan, C. Schwabe, A. Wollmer, J. Gliemann, and S. Gammeltoft. 1986. Coypu insulin. Biochem. J. 238:345–351.
Berman, H. M., T. Battistuz, T. N. Bhat et al. (16 co-authors).2002. The protein data bank. Acta Crystallogr. D 58:899–907.
Blundell, T. L., and S. P. Wood. 1975. Is the evolution of insulin Darwinian or due to selectively neutral mutation? Nature 257:197–203.
Chan, S. J., and D. F. Steiner. 2000. Insulin through the ages: phylogeny of a growth promoting and metabolic regulatory hormone. Am. Zool. 40:213–222.
Chang, X., A. M. Jorgensen, P. Bardrum, and J. J. Led. 1997. Solution structures of the R6 human insulin hexamer. Biochemistry 36:9409–9422.
Christoffersen, C. T., K. E. Bornfelt, C. M. Rotella, N. Gonzales, H. Vissing, R. M. Shymco, J. ten Hoeve, J. Groffen, N. Heisterkamp, and P. De Meyts.1994. Negative cooperativity in the insulin growth factor-I receptor and chimeric IGF-I/insulin receptor. Endocrinology 135:472–475.
Civetta, A. 2003. Positive selection within sperm-egg adhesion domains of fertilin: an ADAM gene with potential role in fertilization. Mol. Biol. Evol. 20:21–29.
Conlon, J. M. 2001. Evolution of the insulin molecule: insight into structure-activity and phylogenetic relations hips. Peptides 22:1183–1193.
Davidson, J. K., M. Zeigler, and R. E. Haist. 1968. Failure of guinea pig antibody to beef insulin to neutralize coypu (nutria) insulin. Diabetes 17:8–12.
———. 1969. Failure of guinea pig antibodies to beef insulin, chicken insulin, and cod insulin to neutralize capybara insulin. Diabetes 18:212–215.
De Meyts, P. 1994. The structural basis of insulin and insulin growth factor-I receptor binding and negative cooperativity, and its relevance to mitogenic versus metabolic signaling. Diabetologia 37:S135–S148.
De Meyts, P., and J. Whittaker. 2002. Structural biology of insulin and IGF1 receptors: implications for drug design. Nat. Rev. Drug Discov. 1:769–783.
Gabrielson, B., K. M. Fairhall, and C. A. F. Robinson. 1990. Growth hormone secretion in the guinea pig. J. Endocrinol. 124:371–380.
Golding, G. B., and A. M. Dean. 1998. The structural basis of molecular adaptation. Mol. Biol. Evol. 15:355–369.
Goldman, N., and Z. Yang. 1994. A codon-based model of nucleotide substitution for a protein coding DNA sequences. Mol. Biol. Evol. 11:725–736.
Grossman, L. I., T. R. Schmidt, D. E. Wildman, and M. Goodman. 2001. Molecular evolution of aerobic energy metabolism in primates. Mol. Phylogenet. Evol. 18:26–36.
Honeycutt, R. L., D. L. Dowe, and M. H. Gallardo. 2003. Molecular systematics of the South American caviomorph rodents: relationships among species and genera in the family Octodontidae. Mol. Phylogenet. Evol. 26:476–489.
Horuk, R., T. L. Blundell, N. R. Lazarus, R. W. Neville, D. Stone, and A. Wollmer. 1980. A monomeric insulin from the porcupine (Hystrix cristata), and old world hystricomorph. Nature 286:822–824.
Horuk, R., P. Goodwin, K. O'Connor, R. J. W. Neville, N. R. Lazarus, and D. Stone.1979. Evolutionary change in the insulin receptors of hystricomorph rodents. Nature 279:439–440.
Kerr, D. E., B. Laarveld, and J. G. Manns. 1990. Effects of passive immunization of growing guinea pigs with an insulin growth factor-I monoclonal antibody. J. Endocrinol. 124:403–415.
King, G. L., and R. Kahn. 1981. Non-parallel evolution of metabolic and growth-promoting functions of insulin. Nature 292:644–646.
King, G. L., C. R. Kahn, and C.-H. Heldin. 1983. Sharing of biological effect and receptors between guinea pig insulin and platelet-derived growth factor. Proc. Natl. Acad. Sci. USA 80:1308–1312.
Kristensen, C., T. Kjeldsen, F. C. Wiberg, L. Schaffer, L. Hach, S. Havelund, J. Bass, D. F. Steiner, and A. S. Andersen. 1997. Alanine scanning mutagenesis of insulin. J. Biol. Chem. 271:12978–12983.
Lee, J., T. O'Hare, P. F. Pilch, and S. E. Shoelson. 1993. Insulin receptor autophosphorylation occurs asymmetrically. J. Biol. Chem. 268:4092–4098.
Leite, Y. L. R., and J. L. Patton. 2002. Evolution of South American spiny rats (Rodentia, Echimyidae): the star phylogeny hypothesis revisited. Mol. Phylogenet. Evol. 25:455–464.
Lenovinovitz, A., G. Norstedt, S. van der Berg, I. C. A. F. Robinson, and T. J. Ekstrom. 1992. Isolation of an insulin-like growth factor II cDNA from guinea pig liver: expression and developmental regulation. Mol. Cell. Endocrinol. 89:105–110.
Melo, F., and E. Feytmans. 1997. Novel mean force potential at atomic level. J. Mol. Biol. 267:207–222.
———. 1998. Assessing protein structures with a non-local atomic interaction energy. J. Mol. Biol. 277:1141–1152.
Melo, F., R. Sánchez, and A. Sali. 2002. Statistical potentials for fold assessment. Protein Sci. 11:430–448.
Messier, W., and C.-B. Stewart. 1997. Episodic adaptive evolution of primates lysozymes. Nature 385:151–154.
Miller, S. A., D. D. Dykes, and H. F. Polesky. 1988. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res. 16:1215.
Murphy, W. J., E. Eizirik, W. E. Johnson, Y. P. Zhang, O. A. Ryder, and S. J. O'Brien. 2001. Molecular phylogenetics and the origins of placental mammals. Nature 409:614–618.
Neville, R. W. J., B. J. Weir, and N. R. Lazarus. 1973. Insulin of hystricomorph rodents. Diabetes 22:851–853.
Nielsen, R., and Z. Yang. 1998. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene. Genetics 148:929–936.
Opazo, J. C., M. Soto-Gamboa, and F. Bozinovic. 2004. Blood glucose concentration in caviomorph rodents. Comp. Biochem. Physiol. 137A:57–64.
Rowe, D. L., and R. L. Honeycutt. 2002. Phylogenetic relationships, ecological correlates, and molecular evolution within the Cavioidea (Mammalia, Rodentia). Mol. Biol. Evol. 19:263–277.
Sch?ffer, L. 1994. A model for insulin binding to the insulin receptor. Eur. J. Biochem. 221:1127–1132.
Spotorno, A. E., J. P. Valladares, J. C. Marín, R. E. Palma, and C. Zulueta. 2004. Molecular divergence and phylogenetic relationships of chinchillids (Rodentia: Chinchillidae). J. Mammal. 85:384–388.
Springer, M., M. J. Stanhope, O. Madsen, and W. W. de Jong. 2004. Molecules consolidate the placental mammal tree. Trends Ecol. Evol. 19:430–438.
Wood, A., T. L. Blundell, A. Wollmer, N. R. Lazarus, and R. W. J. Neville. 1975. The relation of conformation and association of insulin to receptor binding; X-ray and circular dichromism studies on bovine and hystricomorph insulins. Eur. J. Biochem. 55:531–542.
Yang, J., J. Huang, H. Gu, Y. Zhong, and Z. Yang. 2002. Duplication and adaptive evolution of chalcone synthase genes in the genus Dendranthema (Asteraceae). Mol. Biol. Evol. 19:1752–1759.
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput. Appl. Biosci. 13:555–556.
———. 2001. Adaptive molecular evolution. Pp. 327–350 in D. Balding, M. Bishop, and C. Cannings, eds. Handbook of statistical genetics. John Wiley & Sons Ltd., London.
Yang, Z., R. Nielsen, N. Goldman, and A.-M. K. Pedersen. 2000. Codon substitution models for heterogeneous selection pressure at amino acids sites. Genetics 155:431–449.
Zimmerman, A. E., M. L. Moule, and C. C. Yip. 1974. Guinea pig insulin. II. Biological activity. J. Biol. Chem. 249:4026–4029.(Juan C. Opazo*,,1, R. Edu)
Correspondence: E-mail: jopazo@genetics.wayne.edu.
Abstract
Insulin is a conservative molecule among mammals, maintaining both its structure and function. Rodents that belong to the Suborder Hystricognathi represent an exception, having a very divergent molecule with unusual physiological properties. In this work, we analyzed the evolutionary pattern of the insulin gene in caviomorph rodents (South American hystricomorph rodents). We found that these rodents have higher rates of nonsynonymous:synonymous substitutions (dN/dS) than nonhystricomorph rodents and that values are heterogeneous inside the group. We estimated codons under positive selection, specifically the second binding site (A13 and B17) and others related with hexamerization (B18, B20, and B22). In the monomer structure, all selected sites formed a single patch around the second binding site. In the hexamer structure, these amino acids were grouped into three major patches. In this structure, contacts between B chains involved all selected sites (except B18), and between faces in the center of the molecule, all contacts were among selected sites. While there is no clear hypothesis regarding the cause of this drastic change, experimental evidence does show that this group of rodents has some peculiarities in growth function, and, whether coincidental or not, these changes appeared together with important changes in life-history traits.
Key Words: insulin ? caviomorph ? adaptive evolution ? life-history traits ? Hystricognathi
Introduction
Insulin is a hormone secreted by beta cells in the Langerhans islets of the pancreas, which is released mainly in response to blood glucose levels. Among mammals the structure and function of insulin is well conserved (Chan and Steiner 2000; Conlon 2001). In fact, the regulatory properties of insulin are interchangeable between species. Nevertheless, hystricomorph rodents are an exception among mammals (Neville, Weir, and Lazarus 1973; Conlon 2001), possessing an insulin molecule with many amino acid substitutions, which has important effects on several physiological properties of the molecule. Among these peculiarities is the observation that hystricomorph insulin is not neutralized by the antibovine insulin antibody (Davidson, Zeigler, and Haist 1968, 1969), indicating that amino acid substitutions have changed the immunological properties of the molecule. Additionally, insulin from this group cannot self-associate in hexamers, a structure which is used by nonhystricomorph mammals to store the hormone in secretion vesicles (Wood et al. 1975; Bajaj et al. 1986). It is possible that hystricomorph rodents are unable to store their insulin or that storage is accomplished via another unknown way. An aspect which has been better characterized is insulin's diminished biological activity in this group of rodents. Among hystricomorphs, there are variable degrees of change in biological activity; the lowest activity change is seen in rodents of the superfamily Chinchilloidea (chinchillas and vizcachas), and the highest activity change is observed for rodents of the superfamily Octodontoidea (degus, tucu-tucus, spiny rats, and allies) (Zimmerman, Moule, and Yip 1974; Horuk et al. 1979; Bajaj et al. 1986). Despite this variation in biological activity, hystricomorphs can reach the same maximal response as nonhystricomorph mammals, although at greater insulin concentrations (Zimmerman, Moule, and Yip 1974; Bajaj et al. 1986). Despite these atypical physiological properties, hystricomorph rodents are able to regulate glycemia like other nonhystricomorph mammals (Opazo, Soto-Gamboa, and Bozinovic 2004). Finally, hystricomorph insulin possesses greater growth-promoting activity than other mammals (King and Kahn 1981) and is able to bind to additional growth factor receptors (King, Kahn, and Heldin 1983).
The unique insulin molecules of hystricomorph rodents have been an enigma for many years. Understanding how changes at the molecular level produce changes in the biological function of this molecule is of great interest to researchers, especially considering insulin's central role in physiology and the serious health consequences that occur when the molecule malfunctions. The task of evaluating the structure-function relationship of this molecule is difficult due to the lack of natural variants of insulin among most mammals. In this sense, an analysis of the insulin gene in this group of rodents will have an important impact on future studies.
The objective of this study is to describe the evolutionary pattern of the insulin molecule in caviomorph rodents (South American hystricomorph rodents). To this end, we sequenced the insulin gene in representatives of the caviomorph rodents and analyzed these sequences at two distinct levels. First, we analyzed the variation in the ratio of nonsynonymous substitutions to synonymous substitutions among lineages of caviomorphs in a phylogenetic framework. Second, we carried out analyses directed at identifying which amino acids were under positive selection, mapping these sites onto the established monomer and hexamer structures and determining their location in relation to the molecule function.
Materials and Methods
Rodent Species
We obtained complete sequences of both insulin peptides for 21 caviomorph rodent species, representing the superfamilies Cavioidea, Chinchilloidea, and Octodontoidea (supplementary table 1).
DNA Extraction, Amplification, and Sequencing
Alcohol-preserved samples were subjected to proteinase K digestion, NaCl precipitation of proteins, and DNA precipitation with ethanol (Miller, Dykes, and Polesky 1988). Each chain was amplified separately by the polymerase chain reaction (PCR). PCR conditions were as follows: 30 cycles of alternating denaturation at 94°C for 30 s, annealing at 50–58°C for 30 s, and extension at 72°C for 30 s. An initial denaturation at 94°C for 2 min and final extension at 72°C for 2 min were also performed. All experiments included negative controls. Primers used to obtain chain B were Vaca-3 CGCCATGGCCCCGTGGATGC (forward), Vaca-4 CTGGCCCTGCTGGCCCTCTG (forward), and Láctica-3 AGGGGCTCACCCKKTGGGTCCTC (reverse). To obtain chain A, we used Pata-10 CTGCTGCTCCTGACAGCRTCT (forward), Pata-13 GGYCTTCTGCTGCTCCTGAC (forward), Pata-17 GGCGCTGTGCTGCTCCTRACA (forward), Cabra-5 TAGACACCTGCCTTGGGCCTGG (reverse), and Cabra-6 CATTCAAGGGGTTTATTGGTTGCCA (reverse).
PCR products were visualized in 5% nondenaturing minigels, and the remaining product was cleaned in Sephadex columns and used as templates for sequencing both strands.
Statistical Analysis
Variable (=dN/dS) Rates Among Lineages
To detect the possible role of positive selection among specific lineages in the phylogeny (fig. 1), we used the maximum likelihood codon substitution model of Goldman and Yang (1994) implemented in the PAML 3.13 package (Yang 1997). This model takes into account the structure of the genetic code, transition/transversion ratios, and base frequencies at the three codon positions.
FIG. 1.— The tree topology used to conduct the analyses of variable (=dN/dS) among lineages and sites. Branches labeled from 1 to 4 and a to f were used to test hypotheses mentioned in Materials and Methods. In the case of variable among sites, we excluded noncaviomorph mammalian species. Tree topology is based on published literature (Murphy et al. 2001; Leite and Patton 2002; Rowe and Honeycutt 2002; Honeycutt, Dowe, and Gallardo 2003; Spotorno et al. 2004; Springer et al. 2004).
Several models of variable (=dN/dS) rates among lineages were implemented using the 21 aligned sequences of caviomorph rodents and 7 sequences of noncaviomorph mammals (supplementary table 1), based on figure 1. The simplest model assigns the same for all branches, while the free ratio model assumes an independent for each branch. Intermediate models with phylogenetic sense were also implemented. Each of these models makes different assumptions about the value of for branches labeled 1 to 4 and the branches that descend from them (fig. 1). For instance, the one-ratio model assumes the same for ancestral branches of the noncaviomorph rodents and caviomorph species belonging to the superfamilies Cavioidea, Chinchilloidea, and Octodontoidea and also for the branches that descend from them (fig. 2A). The two-ratio model assumes 1 for the ancestral branch of noncaviomorph rodents and their descendants and another for the ancestral branches of the caviomorph superfamilies and their descendants (fig. 2B). The 4 model assumes that each ancestral branch of the caviomorph superfamilies and their descendants has an independent rate (fig. 2C). The 8 model assumes that the descendant branches have independent rates from the ancestral branches (fig. 2D). The last model implemented examined variation at the family level among members of the superfamily Octodontoidea. This model was implemented because members of this group have more amino acid substitutions than the members of the other superfamilies and also present insertions and deletions, a characteristic that is not shared with other caviomorph rodents. In this vein, we reproduced the 8 model; however, we assigned specific parameters for branches a, b, c, d, e, and f (fig. 1). In this model, the descendant branches from e (corresponding to the ancestor of the family Octodontidae) and f (the ancestor of the family Ctenomyidae) had independent values, and the total model included 15 parameters. Three starting values (0.5, 1, and 2) were used to check the existence of multiple local optima. Nested models were compared using the likelihood ratio test.
FIG. 2.— Graphical representation of nested models implemented to test the possible role of positive selection over lineages. For details see Materials and Methods.
Variable (=dN/dS) Rates Among Sites
Models of variable among sites were used to test for the existence of particular sites under positive selection and to identify these sites. We used the models recommended by Yang et al. (2000) implemented in the CODEML module of the PAML 3.13 package (Yang 1997); for this analysis, we used tree topology of figure 1 excluding the noncaviomorph mammals. Model M0 assumes constant rates across sites. Model M1 (neutral) assumes two site classes with values of 0 (purifying selection) and 1 (neutral). In addition to the classes mentioned for M1, the M2 Model (selection) assumes a third category of sites with an value estimated from the data. Model M3 (discrete) has three classes of sites with proportions p0, p1, and p2 and 0, 1, and 2 values estimated from the data. Model M7 assumes a beta distribution between 0 and 1 depending on the parameters p and q. Finally, Model M8 adds an extra class of sites with values and proportions estimated from the data. Only Models M2, M3, and M8 can detect sites under positive selection. Because some models, especially M2 and M8, are prone to multiple local optima, we used three starting values (0.5, 1, and 2) to check for the existence of multiple local optima. When the estimation of the parameters was finished, Bayes theorem was used to calculate the posterior probability that each site belonged to one site class (Nielsen and Yang 1998; Yang et al. 2000).
Obtaining reliable estimations of sites under positive selection demands that the data meet certain requirements. Anisimova, Bielawski, and Yang (2002) used simulations to evaluate the accuracy and power of the Bayesian approximation for estimating sites under positive selection. They concluded that the best way to improve accuracy and power is to use sequences that are both divergent and numerous and to apply multiple models to the data. In another study, the same authors (Anisimova, Bielawski, and Yang 2001) analyzed the power of the likelihood ratio test to detect adaptive molecular evolution. They showed that the power was nearly 100% when they had at least 17 sequences in the data set. Other data set characteristics such as length, divergence, and the strength of positive selection affect the power of the likelihood ratio test. All of these requirements are fulfilled by our data; therefore, we believe that our estimations will be reliable.
Location of Selected Codons in Insulin Monomer and Hexamer and Contact Maps
To determine where selected amino acids are located in the insulin molecule, we obtained the monomer structure of pig insulin (Badger et al. 1991) and the hexamer structure of human insulin (Chang et al. 1997) from Protein Data Bank (Berman et al. 2002). Contact maps were calculated for the hexamer structure using software that we developed, which is available for free upon request. All chain-chain contacts were calculated, including the homomeric contacts. In each case, three independent files were generated. The first considers only side chain–side chain atomic contacts, the second considers only main chain–main chain atomic contacts, and the third considers all atomic contacts between two given residues. In all cases where at least one pair of residues exhibited more contacts than those specified in the threshold of contacts (based on the program's input), a bidimensional density plot was generated. The plots were created automatically as postscript files using the free ASGL software. This program requires several parameters as input, including the distance cutoff for defining a contact between two atoms in three-dimensional space, a minimum graphical threshold for the number of atomic contacts between two residues that are to be plotted, and the minimum sequence separation between two residues to be considered a contact (Melo and Feytmans 1997). In our work, the distance cutoff was fixed at 4.0 ?, the minimum number of atomic contacts was set to 5, and the minimum sequence separation was defined with a value of three residues in the case of contacts within the same chain (in the case of contacts between two different chains, this value was not used). These parameters will only select pairs of nonlocal residues that have a substantial fraction of their atoms in close proximity in three-dimensional space (Melo and Feytmans 1998; Melo, Sánchez, and Sali 2002). We used the contact maps to evaluate whether selected sites are involved in contacts in the hexamer structure.
Results
Variable (=dN/dS) Rates Among Lineages
The model assuming the same value for all branches in the phylogeny leads to l1 = –1,368.81, with an average value over all sites and lineages of 0.23. The log likelihood value of the model that assumes an independent for each branch (free model) was l0= –1,297.49. The free ratio model involved 53 parameters for 53 branches, while the model assuming 1 for all branches involved 1; then, 2l = 2(l1 – l0) = 142.62 was compared with a 2 distribution with df = 52 to test whether the free ratio model provided a significantly better fit to the data. The difference between models was significant (P < 10–5), indicating that rates are different among lineages.
Values of estimated from the different models are shown in table 1. The model distinguishing between noncaviomorph rodents and caviomorph rodents (fig. 2B) had a significantly better fit (P < 10–5) than the model that did not (fig. 2A). The value estimated for ancestral and descendant branches of the caviomorph rodents was 0.48, more than 40-fold the value estimated for noncaviomorph rodents (table 1). The model segregating between noncaviomorph rodents and within the caviomorph group (fig. 2C) did not have a better fit than the model that only segregated between noncaviomorph rodents and caviomorph rodents (fig. 2B) (P = 0.34). To test whether ancestral and descendant branches had their own values, we compared the model that assumed an independent value for each branch (fig. 2D) with a 2 model (fig. 2B). These models were not significantly different; the 8 model did not have a better fit than the 2 model (P = 0.72). Finally, the model that discriminates between rates within the superfamily Octodontoidea (fig. 1) was compared with the 2 model. The first model had an almost significantly better fit than the 2 model (P = 0.07). This model indicates that each labeled branch has its own parameter (fig. 1) in addition to the descendants of branches 1, 2, and 3. Among the ancestral branches of families within the superfamily Octodontoidea, we observed the following decreasing pattern: the ancestral branch of the family Echimyidae had the highest value (branch c = 0.69), followed by Abrocomidae (branch a = 0.53), and then the Octodontidae family (branch e = 0.25), and the lowest value was observed for the ancestral branch of the family Ctenomyidae (branch f = 0.03). The external branches for the family Octodontidae maintained their value (0.27) in relation to the ancestral branch, while the external branches of the family Ctenomyidae increased (0.08). The ancestral branch of the families Ctenomyidae and Octodontidae had the highest values among the entire superfamily, having only nonsynonymous substitutions (branch d = 10.9/0), while branch b had a lower value (0.06).
Table 1 Log Likelihood Values and Parameter Estimates Under Different Lineage Models
Variable (=dN/dS) Rates Among Sites and Identification of Amino Acids Under Positive Selection
Table 2 shows the parameters estimated under variable selective pressure among sites using the unrooted tree topology of figure 1 without the out-group (noncaviomorph mammals). Models designed to detect positively selected sites (M2, M3, and M8) were significantly better, and all models converged to the same sites. Sites B17, B22, and A13 had a posterior probability of being in the selected class of 99%. Sites B18 and B22 had smaller posterior probability values (table 2). The value for the selected site class varied from 3.38 to 4.15, indicating strong positive selection acting on these sites. The likelihood ratio tests were always in favor of models with the ability to detect sites under positive selection.
Table 2 Parameter Estimates and Log Likelihood Values Under Models of Variable Rates Among Codons
Location of Selected Sites in the Monomer Structure, Hexamer Structure, and Contact Maps
Because the crystal structure of caviomorph insulin has not yet been determined, we used a noncaviomorph insulin structure to evaluate the spatial location of selected sites. In the monomer structure, the selected sites are grouped together in a single patch (fig. 3A), compromising the second binding region (fig. 3B). In the hexamer structure, the selected sites are grouped (fig. 3C) in three discrete patches (fig. 3D). This phenomenon is produced because two monomer patches are contiguous in the hexamer structure.
FIG. 3.— Space-filling structure of the insulin monomer and hexamer obtained from Protein Data Bank. Chain A is shown in blue and chain B in red. (A) Insulin monomer structure showing estimated sites under positive selection (yellow). (B) Insulin monomer structure showing residues corresponding to the putative second binding region (B17 in yellow and A13 in light blue). (C) Lateral view of the hexamer insulin structure showing the estimated sites under positive selection (yellow). (D) Frontal view of the hexamer insulin structure showing the estimated sites under positive selection (yellow).
Considering the main chain contact maps between B chains in the hexamer structure, residue B20 is involved in contacts. When we consider main and lateral chains, all selected sites (except B18) are involved in contacts. At the center of the molecule, where the zinc is coordinated, all contacts between hexamer sides are among selected sites.
Discussion
Insulin Structural Patterns
According to the data presented here and previously published, it is possible to differentiate, based on the length of the insulin, two groups of insulin molecules among caviomorph rodents. The first group includes the molecule belonging to the species of the superfamilies Cavioidea and Chinchilloidea, which conserves the number of residues (chain A with 21 and chain B with 30 residues). The second type of insulin molecule is found in caviomorph species belonging to the superfamily Octodontoidea. In this group, the insulin molecule presents a phenylalanine deletion at position B24 and an insertion of two amino acids at the carboxy-terminal end of chain A. We can infer that these changes were produced gradually because the species with intermediate structures are currently living. Concordant with the phylogeny of the group (Honeycutt, Dowe, and Gallardo 2003), Abrocoma bennetti (representing the species of the family Abrocomidae) is the most basal clade of the superfamily and presents only the phenylalanine deletion in chain B (fig. 4). In turn, Myocastor coypus (representing the family Myocatoridae) insulin sequence obtained from GenBank presents the phenylalanine deletion as well as an insertion of 1 amino acid in the carboxy-terminal end of the A chain (Bajaj et al. 1986). While the phylogenetic position of this species in not well resolved (Leite and Patton 2002), the structure of the insulin molecule strongly suggests that M. coypus is the sister group of the species of the families Echimyidae, Ctenomyidae, and Octodontidae (fig. 4). The other families (Ctenomyidae, Echimyidae, and Octodontidae) have a chain B deletion and two amino acid insertions at the carboxy-terminal end of chain A.
FIG. 4.— Cladogram of the superfamily Octodontoidea showing when insulin changes occurred.
Chain B partial sequences obtained from members of the families Bathyergidae (Cryptomys damarensis), Hystridae (Hystrix africaeaustralis and Atherurus macrourus), and Erethizontidae (Coendu bicolor and Erethizon dorsatum) suggest that these clades have insulins of standard length (unpublished data). For the Hystricidae family, this pattern is confirmed by protein sequence data of the porcupine (Hystrix cristata) (Horuk et al. 1980).
Protein sequence of the Echimyidae species, Proechimys guairae (Horuk et al. 1979), shows a B chain deletion; however, the A chain does not have any insertions. This result contrasts with our nucleotide sequence data in Hoplomys gymnurus, which presents a deletion in chain B and two amino acid insertions in chain A. Additionally, the genera Proechimys and Hoplomys are grouped together in the phylogeny (Leite and Patton 2002).
Rate Variation Among Lineages
We have drawn two major conclusions from maximum likelihood analysis of rate variation among lineages. First, that the ancestral branches of caviomorph superfamilies have higher (=dN/dS) rate ratios than noncaviomorph rodents, and second, that these values are heterogeneous. Omega values for the ancestral branches of each caviomorph superfamily were always higher than the ancestral branch for noncaviomorph rodents (table 1). The ancestral branches of the caviomorph superfamilies were not demonstrably affected by positive selection (using dN/dS >1 as an indicator) (Yang 2001). Nevertheless, for the ancestral branch of the superfamily Cavioidea were estimated 6.4 nonsynonymous substitutions, 12.8 in the stem of the superfamily Octodontoidea, and 2.4 for Chinchilloidea. Furthermore, the values for the ancestral branches of these superfamilies were between 3,000 and 8,300 times higher than the ancestral branch of noncaviomorph rodents (table 1).Moreover, physiological evidence demonstrated that the insulin molecule in species of these superfamilies also has divergent physiological properties (Horuk et al. 1979; King and Kahn 1981; King, Kahn, and Heldin 1983). Additionally, it has been documented that important changes in function are not necessarily the result of numerous amino acid changes; rather, a few changes in key positions are enough to change the properties of the molecule (Golding and Dean 1998). Although these few changes could not elevate the overall ratio over 1, it is possible to identify which sites are under positive selection (Yang et al. 2002; Civetta 2003). It is important to note that in spite of heterogeneous values among lineages, there is a relationship with physiological divergence (defined as the decrease in biological activity in comparison with noncaviomorph mammals). The lowest value (0.30) and the least physiological divergence (Horuk et al. 1979) were seen for the ancestral branch of the superfamily Chinchilloidea, and the highest values of were observed for the superfamily Cavioidea (Zimmerman, Moule, and Yip 1974). The superfamily Octodontoidea is a special case because its divergence is underestimated as PAML package does not take indels into account (Yang 1997). Nevertheless, it has been documented that the physiological divergence of this group is greater than that for members of the superfamily Cavioidea (Bajaj et al. 1986). In this sense, if the relationship between the value and the physiological divergence is correct, the intensity of selection in this group should be greatest among caviomorph rodents. Moreover, if this relationship is indeed genuine, it would be an important fact because, until now, the value has been used as a discrete indicator, where values greater than 1 are indicative of positive selection events and values below 1 indicate a lack of positive selection. Our data indicate that differences resulting from different values during evolution could possibly be measured functionally. On the other hand, it is important to comment on the importance of experimental evidence, when the lineage analysis could not detect positive selection events with this criterion. In this sense, the ratio between the control group (i.e., noncaviomorph rodents) and the group of interest could possibly be an alternative view for estimating positive selection.
Test of Positive Selection Among Sites
Analyses of variation among sites provide clear evidence of positive selection. The results show that four of the five sites estimated under positive selection are in chain B. Interestingly, these sites are not in the classical receptor binding domain or the domain determined by alanine-scanning mutagenesis, in concordance with the previous analysis based on amino acid sequences (Conlon 2001). Among noncaviomorph mammals, amino acidic variation is confined to the hypervariable region A8 to A10 and the extreme N- and C-terminals of chain B. These regions are not considered to be important in determining the biological activity of the molecule (Kristensen et al. 1997). The positively selected site in chain A is not in the hypervariable region described for mammals. Rather, some sites of chain B are in the middle of the peptide (B17, B18, and B20), while one is in the C-terminal portion of the molecule (B22). Among these sites under positive selection, we believe that sites B17 and A13, which had the highest posterior probabilities of being in the proportion of sites under positive selection, may be the most important in causing the decreased biological activity of insulin in caviomorph rodents because they were described as the putative second binding region (De Meyts 1994; Sch?ffer 1994). De Meyts (1994) described a binding model in which he established the relationship between insulin molecules with an altered hexamer-forming surface (e.g., A13, B17, B18, and B20) and the rate of association and dissociation from the insulin receptor, negative cooperativity, and metabolic and mitogenic activity (for more details, see De Meyts 1994). In this model, De Meyts (1994) explained the higher mitogenic and lower metabolic activities of this kind of insulin (i.e., hystricomorph insulin) as an asymmetrical phosphorylation of the receptor when the first insulin does not bind through its hexamer-forming surface (Lee et al. 1993). Alanine-scanning mutagenesis experiments give independent information regarding the importance of the second binding region. Replacement of the residues A13 and B17 by alanine results in a decrease of 70% and 38% in binding affinity, respectively, for the low-affinity binding site (Kristensen et al. 1997) and recently tested, with similar results, for the high-affinity binding site (De Meyts and Whittaker 2002). Nevertheless, this evidence should be taken with caution because none of the caviomorph rodents sequences presented in our study has an alanine residue in position B17 or A13, except for Dolichotis patagonum. Additionally, these measurements represent a change of just one amino acid in the molecule; the effect of multiple ones (like in caviomorph rodents) other than alanine replacements is unknown.
Residue B22 (arginine) has been described as being trypsin sensitive in nonhystricomorph mammals, so the change in hystricomorph rodents due to the inability to form hexamers would make the insulin of these rodents most stable in a monomeric state (Blundell and Wood 1975). Nevertheless, this change would produce a decrease in the biological activity of the molecule (Horuk et al. 1979, 1980). This explanation is based on a Cavia porcellus sequence that has an aspartic acid residue in position B22. In this sense, the interaction of the negatively charged B22 and A21 residues would produce a change in the structure, which would have functional consequences (Horuk et al. 1979, 1980). For the sequences obtained in this work, all A21 residues were asparagine. Nevertheless, among chain B sequences, only two species (C. porcellus, Cavia aperea) had an aspartic acid in position B22. Furthermore, some species have both positions like noncaviomorph mammals (A21 asparagine and B22 arginine) (Chinchilla lanigera, Chinchilla brevicaudata, D. patagonum, Hoplomys gimnurus). Of these species, C. lanigera shows diminished biological activity (Horuk et al. 1979). Thus, the simple replacement of the B22 position by a negatively charged residue cannot alone explain reduction in activity. We believe that deviations in physiological activity are the result of more than just the sum of effects due to each individual substitution.
Spatial Location of Selected Sites in the Monomer and Hexamer Structures and the Contact Maps
It is interesting that the spatial location of the sites estimated under positive selection formed a single patch in the monomer structure. This patch is located near the second binding site (fig. 3A and B), in agreement with the binding model of De Meyts (1994) and his explanation of the atypical physiological properties. Consistent with this, the hexamer structure also presents a nonrandom pattern, where we were able to differentiate three discrete patches (fig. 3D), each consisting of two monomer patches (fig. 3C). This spatial distribution of the sites estimated to have evolved under positive selection in the hexamer structure might explain, in part, the inability of hystricomorph rodent insulin to aggregate.
The main molecular interactions established in the hexamer structure are between the B chains of the two insulin monomers and among B chains at the center of the structure, where the zinc atom is coordinated. Our results showed that in the main chain interactions among B chains, sites under positive selection represented 33% of total contacts. When we considered main and lateral chains, all sites estimated under positive selection were involved, except site B18. At the center of the structure the amino-terminal ends of the B chains established contacts on both sides, in which no selected sites were involved. Nevertheless, B chains established few contacts between the sides of the structure, in which all selected sites were involved. We know that contact maps are not absolute proof of the cause of the inability of caviomorph insulin to form hexamers; however, we believe that this analysis is a good initial approximation to the problem. It is important to highlight the fact that there are other contacts inside the hexamer insulin structure; however, it is not our purpose to discuss all of them here.
Concluding Remarks
When positive selection events occur, we are at least able to distinguish between events that increase the established function of a molecule (Grossman et al. 2001) and those in which the molecule acquires a new function (Messier and Stewart 1997). Considering the limited information available regarding insulin physiology in hystricomorph rodents, we believe that this is a case in which the increase in nonsynonymous substitutions is associated with the acquisition or expansion of the function. There is not enough evidence to conclude what this new function is; however, we suspect that it may be related to the unusual patterns of pre- and postnatal growth of caviomorph rodents.
King, Kahn, and Heldin (1983) showed that the insulin of C. porcellus can bind to the receptor of the platelet-derived growth factors and that this binding was not competed either by the insulin of other mammals or by growth factors, where insulin growth factor (IGF)-I was tested. Additionally, when researchers measured thymidine incorporation (as a measurement of growth-promoting function), noncaviomorph insulin did not show additivity. However, when pig and caviomorph insulins were tested together, researchers obtained additive response. Hystricomorph insulin shares some physiological properties with IGF-I: neither are able to aggregate (De Meyts 1994), both present a lack of self-antagonism in a dose-response curve for negative cooperativity (Christoffersen et al. 1994), and both have a growth-promoting function (King and Kahn 1981). At the systemic level, it has been demonstrated that C. porcellus is insensitive to growth hormone, although the structures of the hormone and its receptor are conserved (Adkins, Vandeberg, and Li 2000) and the gland produces and releases the hormone normally (Gabrielson, Fairhall, and Robinson 1990). Additionally, when the effector of the growth hormone is blocked (IGF-I), there are no effects on the growth pattern of individuals (Kerr, Laarveld, and Manns 1990). Finally, in contrast to other rodents, IGF-II levels do not drop after birth, instead hystricomorph rodents maintain IGF-II levels throughout their adult life (Lenovinovitz et al. 1992). This collection of evidence indicates that pre- and postnatal growth regulations are different for this group of rodents.
In sum, we believe that the divergence of caviomorphs (relative to rodents of the Suborder Sciurognathi) and the drastic changes in life-history traits and growth pattern are more than just a coincidence.
Supplementary Material
Supplementary table 1 is available at Molecular Biology and Evolution online (www.mbe.oupjournals.org).
Acknowledgements
We acknowledge I. Tomasco, G. Wlasiuk, A. Castillo, and G. D'Elía for their help with laboratory work. We also thank M. Soto-Gamboa and M. A. Lardies for their assistance in the field and Z. Yang for his assistance with PAML. This work would not have been possible without tissue loans from the following institutions: the Museum of Vertebrate Zoology at the University of California at Berkeley, the Museum of Southwestern Biology at the University of New Mexico, the Museum of Zoology at the University of Michigan, Colección de Flora y Fauna Profesor Patricio Sánchez Reyes at Pontificia Universidad Católica de Chile, Laboratorio de Citogenética de Mamíferos at Universidad de Chile, Conrad Matthee at the Zoology Department of Stellenbosch University, Guillermo D'Elía from the Laboratorio de Evolución at the Universidad de la República, and Hynek Burda from the Department of General Zoology at the University of Essen. Derek E. Wildman and Lawrence I. Grossman made useful suggestions on early versions of the manuscript. This work was supported by a doctoral thesis fellowship (2002) from Comisión Nacional de Desarrollo Investigación Y Tecnológico to J.C.O and the Center for Advanced Studies in Ecology & Biodiversity program 1 (F. Bozinovic) and 2 (F. Jaksic) and Fondo Nacional de Desarrollo Cientifico Y Tecnológico grant 1010959 to F.M.
References
Adkins, M., J. Vandeberg, and W.-H. Li. 2000. Molecular evolution of the growth hormone and receptor in the guinea-pig, a mammal unresponsive to growth hormone. Gene 246:357–363.
Anisimova, M., J. P. Bielawski, and Z. Yang. 2001. Accuracy and power of the likelihood ratio test in detecting adaptive molecular evolution. Mol. Biol. Evol. 18:1585–1592.
———. 2002. Accuracy and power of Bayes estimation of amino acids sites under positive selection. Mol. Biol. Evol. 19:950–958.
Badger, J., M. R. Harris, C. D. Reynolds, A. C. Evans, E. J. Dodson, G. G. Dodson, and A. C. T. North. 1991. Structure of the pig insulin dimer in the cubic crystal. Acta Crystallogr. B 47:127–136.
Bajaj, M., T. L. Blundell, R. Horuk, J. E. Pitts, S. P. Wood, L. K. Gowan, C. Schwabe, A. Wollmer, J. Gliemann, and S. Gammeltoft. 1986. Coypu insulin. Biochem. J. 238:345–351.
Berman, H. M., T. Battistuz, T. N. Bhat et al. (16 co-authors).2002. The protein data bank. Acta Crystallogr. D 58:899–907.
Blundell, T. L., and S. P. Wood. 1975. Is the evolution of insulin Darwinian or due to selectively neutral mutation? Nature 257:197–203.
Chan, S. J., and D. F. Steiner. 2000. Insulin through the ages: phylogeny of a growth promoting and metabolic regulatory hormone. Am. Zool. 40:213–222.
Chang, X., A. M. Jorgensen, P. Bardrum, and J. J. Led. 1997. Solution structures of the R6 human insulin hexamer. Biochemistry 36:9409–9422.
Christoffersen, C. T., K. E. Bornfelt, C. M. Rotella, N. Gonzales, H. Vissing, R. M. Shymco, J. ten Hoeve, J. Groffen, N. Heisterkamp, and P. De Meyts.1994. Negative cooperativity in the insulin growth factor-I receptor and chimeric IGF-I/insulin receptor. Endocrinology 135:472–475.
Civetta, A. 2003. Positive selection within sperm-egg adhesion domains of fertilin: an ADAM gene with potential role in fertilization. Mol. Biol. Evol. 20:21–29.
Conlon, J. M. 2001. Evolution of the insulin molecule: insight into structure-activity and phylogenetic relations hips. Peptides 22:1183–1193.
Davidson, J. K., M. Zeigler, and R. E. Haist. 1968. Failure of guinea pig antibody to beef insulin to neutralize coypu (nutria) insulin. Diabetes 17:8–12.
———. 1969. Failure of guinea pig antibodies to beef insulin, chicken insulin, and cod insulin to neutralize capybara insulin. Diabetes 18:212–215.
De Meyts, P. 1994. The structural basis of insulin and insulin growth factor-I receptor binding and negative cooperativity, and its relevance to mitogenic versus metabolic signaling. Diabetologia 37:S135–S148.
De Meyts, P., and J. Whittaker. 2002. Structural biology of insulin and IGF1 receptors: implications for drug design. Nat. Rev. Drug Discov. 1:769–783.
Gabrielson, B., K. M. Fairhall, and C. A. F. Robinson. 1990. Growth hormone secretion in the guinea pig. J. Endocrinol. 124:371–380.
Golding, G. B., and A. M. Dean. 1998. The structural basis of molecular adaptation. Mol. Biol. Evol. 15:355–369.
Goldman, N., and Z. Yang. 1994. A codon-based model of nucleotide substitution for a protein coding DNA sequences. Mol. Biol. Evol. 11:725–736.
Grossman, L. I., T. R. Schmidt, D. E. Wildman, and M. Goodman. 2001. Molecular evolution of aerobic energy metabolism in primates. Mol. Phylogenet. Evol. 18:26–36.
Honeycutt, R. L., D. L. Dowe, and M. H. Gallardo. 2003. Molecular systematics of the South American caviomorph rodents: relationships among species and genera in the family Octodontidae. Mol. Phylogenet. Evol. 26:476–489.
Horuk, R., T. L. Blundell, N. R. Lazarus, R. W. Neville, D. Stone, and A. Wollmer. 1980. A monomeric insulin from the porcupine (Hystrix cristata), and old world hystricomorph. Nature 286:822–824.
Horuk, R., P. Goodwin, K. O'Connor, R. J. W. Neville, N. R. Lazarus, and D. Stone.1979. Evolutionary change in the insulin receptors of hystricomorph rodents. Nature 279:439–440.
Kerr, D. E., B. Laarveld, and J. G. Manns. 1990. Effects of passive immunization of growing guinea pigs with an insulin growth factor-I monoclonal antibody. J. Endocrinol. 124:403–415.
King, G. L., and R. Kahn. 1981. Non-parallel evolution of metabolic and growth-promoting functions of insulin. Nature 292:644–646.
King, G. L., C. R. Kahn, and C.-H. Heldin. 1983. Sharing of biological effect and receptors between guinea pig insulin and platelet-derived growth factor. Proc. Natl. Acad. Sci. USA 80:1308–1312.
Kristensen, C., T. Kjeldsen, F. C. Wiberg, L. Schaffer, L. Hach, S. Havelund, J. Bass, D. F. Steiner, and A. S. Andersen. 1997. Alanine scanning mutagenesis of insulin. J. Biol. Chem. 271:12978–12983.
Lee, J., T. O'Hare, P. F. Pilch, and S. E. Shoelson. 1993. Insulin receptor autophosphorylation occurs asymmetrically. J. Biol. Chem. 268:4092–4098.
Leite, Y. L. R., and J. L. Patton. 2002. Evolution of South American spiny rats (Rodentia, Echimyidae): the star phylogeny hypothesis revisited. Mol. Phylogenet. Evol. 25:455–464.
Lenovinovitz, A., G. Norstedt, S. van der Berg, I. C. A. F. Robinson, and T. J. Ekstrom. 1992. Isolation of an insulin-like growth factor II cDNA from guinea pig liver: expression and developmental regulation. Mol. Cell. Endocrinol. 89:105–110.
Melo, F., and E. Feytmans. 1997. Novel mean force potential at atomic level. J. Mol. Biol. 267:207–222.
———. 1998. Assessing protein structures with a non-local atomic interaction energy. J. Mol. Biol. 277:1141–1152.
Melo, F., R. Sánchez, and A. Sali. 2002. Statistical potentials for fold assessment. Protein Sci. 11:430–448.
Messier, W., and C.-B. Stewart. 1997. Episodic adaptive evolution of primates lysozymes. Nature 385:151–154.
Miller, S. A., D. D. Dykes, and H. F. Polesky. 1988. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res. 16:1215.
Murphy, W. J., E. Eizirik, W. E. Johnson, Y. P. Zhang, O. A. Ryder, and S. J. O'Brien. 2001. Molecular phylogenetics and the origins of placental mammals. Nature 409:614–618.
Neville, R. W. J., B. J. Weir, and N. R. Lazarus. 1973. Insulin of hystricomorph rodents. Diabetes 22:851–853.
Nielsen, R., and Z. Yang. 1998. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene. Genetics 148:929–936.
Opazo, J. C., M. Soto-Gamboa, and F. Bozinovic. 2004. Blood glucose concentration in caviomorph rodents. Comp. Biochem. Physiol. 137A:57–64.
Rowe, D. L., and R. L. Honeycutt. 2002. Phylogenetic relationships, ecological correlates, and molecular evolution within the Cavioidea (Mammalia, Rodentia). Mol. Biol. Evol. 19:263–277.
Sch?ffer, L. 1994. A model for insulin binding to the insulin receptor. Eur. J. Biochem. 221:1127–1132.
Spotorno, A. E., J. P. Valladares, J. C. Marín, R. E. Palma, and C. Zulueta. 2004. Molecular divergence and phylogenetic relationships of chinchillids (Rodentia: Chinchillidae). J. Mammal. 85:384–388.
Springer, M., M. J. Stanhope, O. Madsen, and W. W. de Jong. 2004. Molecules consolidate the placental mammal tree. Trends Ecol. Evol. 19:430–438.
Wood, A., T. L. Blundell, A. Wollmer, N. R. Lazarus, and R. W. J. Neville. 1975. The relation of conformation and association of insulin to receptor binding; X-ray and circular dichromism studies on bovine and hystricomorph insulins. Eur. J. Biochem. 55:531–542.
Yang, J., J. Huang, H. Gu, Y. Zhong, and Z. Yang. 2002. Duplication and adaptive evolution of chalcone synthase genes in the genus Dendranthema (Asteraceae). Mol. Biol. Evol. 19:1752–1759.
Yang, Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput. Appl. Biosci. 13:555–556.
———. 2001. Adaptive molecular evolution. Pp. 327–350 in D. Balding, M. Bishop, and C. Cannings, eds. Handbook of statistical genetics. John Wiley & Sons Ltd., London.
Yang, Z., R. Nielsen, N. Goldman, and A.-M. K. Pedersen. 2000. Codon substitution models for heterogeneous selection pressure at amino acids sites. Genetics 155:431–449.
Zimmerman, A. E., M. L. Moule, and C. C. Yip. 1974. Guinea pig insulin. II. Biological activity. J. Biol. Chem. 249:4026–4029.(Juan C. Opazo*,,1, R. Edu)