The Evolution of the Discrete Multirenculate Kidney in Mammals from Ecological and Molecular Perspectives

Abstract Mammals have developed different kinds of renal structures during evolution, yet the origin of the renal structural phenotypes and the molecular mechanisms underlying their adaptive evolution remains unclear. Here, we reconstructed the ancestral state of the renal structures across mammals and found that the unilobar kidney was the ancestral character in mammals. The subsequent correlation analyses between renal phenotypes and life history traits revealed that species with a larger body or in aquatic habitats tend to have evolved discrete multirenculate kidneys (DMKs). To explore the molecular convergent mechanisms among mammals with this most distinct renal structure, the DMK, we used 45 genes related to duplex/multiplex kidney diseases to compare the evolutions of species with DMKs and with other renal phenotypes. Twelve rapidly evolving genes that were functionally enriched in cilium assembly and centrosome were identified in species with DMKs, suggesting that these genes played key roles in the evolution of DMKs. In addition, positive selection was detected in six crucial genes which are mainly involved in epithelial tube morphogenesis and the regulation of neurogenesis. Finally, 12 convergent amino acid substitutions, 6 of which are in crucial domain of proteins, were shared by 2 or more lineages with DMKs. These findings could provide some novel insights into the origin and evolution of renal structures across mammals and the pathogenesis of renal diseases in humans.


Introduction
The kidney is an essential metabolic organ in mammals and looks like a reddish-brown bean. Its basic function is to produce urine and clear the system of metabolic waste and poisons (Finco 1997). A whole kidney consists of renal parenchyma, which is a combination of the outer renal cortex and the inner renal medulla and renal pelvis (Rouiller 1969). The renal cortex is composed of glomerulus, partial renal tubules, blood vessels, and cortical collecting ducts (Kriz and Bankir 1988). Ultrafiltration occurs in the renal cortex, which also produces erythropoietin. Another part of the renal parenchyma, the renal medulla, contains the loop of Henle, medullary capillary plexus, vasa rectae, and more (Kriz 1981). Its function is to maintain the balance of salt and water in blood and reabsorb water. In addition, the renal pelvis acts as the funnel through which urine flows into the ureter (Schmidt-Nielsen 1987;Dwyer and Schmidt-Nielsen 2003;Marieb and Hoehn 2006).
So far, two basic renal structural types have been found in mammals: 1) the unilobar kidney (UK) with continuous cortex and medulla and 2) the multilobar kidney with continuous/discrete cortex and discrete medulla (Oliver 1968). The unipapillary kidney, the simpler renal type in mammals, is made up of renculi, which consist of the renal cortex, renal medulla, and renal pelvis (Hodson 1972). There are actually two other types of renal structures-a crest kidney and a kidney with tubi maximi-that are simply enlargements of the renculus unit (Kriz et al. 2008). What is more, multilobar kidneys, which are made of multiple renculus units, can be divided into CMKs and discrete multirenculate kidneys (DMKs) according to whether the cortex is compound or discrete (Jamison and Kriz 1982;Dantzler 1989). Interestingly, the unipapillary kidney is found mainly in small species, like mice and cats; the crest kidney in, for example, monkeys and camels; and the kidney with tubi maximi in, for example, horses and hippopotamuses. Finally, the CMK is found in humans, pigs, beavers, and manatees and the DMK in cetaceans, pinnipeds, bears, elephants, and more (Beuchat 1996).
Over the past few decades, numerous hypotheses or conjectures about the origin and evolution of renal structures in mammals have been proposed. The UK may be the original renal structure in mammals, and then, the complex multilobar kidney appears to be derived from it (Verhaegen 1993;Beuchat 2002). Almost all marine mammals have multilobar kidneys, and the majority of nonmarine aquatic mammals may have inherited the multilobar kidney phenotype from their ancestors that lived in marine environments (Ortiz 2001). Nevertheless, terrestrial mammals with lobed kidneys (e.g., bears, elephants, and rhinos) may also have inherited multilobar kidneys from their semiaquatic ancestors (Lavergne et al. 1996;Gaeth et al. 1999). For external or internal structures, the DMK is clearly the most distinct one. The mammalian DMK was thought to have originated as an adaptive response to a large body size, deep and prolonged diving in aquatic environments, and a hypertonic marine diet (Williams 2006). Because marine mammals live in a hypertonic environment, which increases intracellular dehydration if saline water is ingested (Williams 1997), the lobulated state of the marine mammal kidney seems to be an adaptation, because the enlarged surface area between the renal cortex and renal medulla can enhance its ability to rapidly handle hypertonic fluids, excrete excess salt and nitrogenous wastes, and relieve saline-induced intracellular dehydration (Williams 2006). The high filtration rate of the glomeruli allows the kidneys of marine mammals to operate at a high energy rate between dives. In addition, the kidneys of marine mammals can resist the effects of reduced blood supply during diving (Pfeiffer 1997). To accommodate the absolute increase in metabolism and the corresponding increase in excretion of end products with increasing body size, the number of nephrons required increases. However, as the number of nephrons required exceeds the maximum number that can be contained in an unipapillary kidney, the unipapillary kidney may give way to the multilobed kidney. Although these suppositions and reports provide us with considerable insights into the origin and evolution of multilobar kidneys in mammals, more evidence is needed to support them. Furthermore, the molecular evidence underlying the phenotypic convergence among species with DMKs remains unknown.
In this study, we reconstructed the ancestral state of renal structures in mammals and explored the factors that drove the evolution of DMKs in mammals. Then, we analyzed genes involved in duplex/multiplex kidney formation to reveal the molecular convergent mechanism among mammals with DMKs from aspects of rapid evolution, positive selection, and convergent amino acid substitution. Our study may also provide implications for the etiology and genetic mechanisms underlying duplex/multiplex kidney diseases in human.

UK as the Ancestral State of the Mammalian Renal Structure
We mapped the phenotypic states of renal structural types onto the species tree from TimeTree (Kumar et al. 2017) (supplementary fig. S1, Supplementary Material online). The evolution of the renal structure across the mammalian phylogeny is summarized graphically in figure 1. Intuitively, there are five finely sorted renal structures in mammals ( fig. 1b). The ancestral state reconstruction showed that the renal structural phenotypes are diverse among mammals ( fig. 1a and supplementary table S1, Supplementary Material online). By comparing the results generated from six models applied for reconstruction, we found that the best-fit model was the DEC + J one, which is derived from the DEC model and specifies the weight of each jump dispersal event in the cladogenesis matrix by assigning a parameter-j (Matzke 2014) (supplementary table S2, Supplementary Material online). The reconstruction with the DEC + J model supported that the UK, whose fit probability is 0.9919 in the ancestral node of whole mammals, is the ancestral state for mammals and revealed that the multilobar kidney phenotype is derived from it multiple times in several independent lineages ( fig. 1a and supplementary table S3, Supplementary Material online). What is more, the result suggested that GBE renal structural phenotypes shifted from the UK to the DMK in the ancestral node of Proboscidea and some internal nodes within Cetartiodactyla, Carnivora, and Perissodactyla. In addition, the UK evolved into the CMK in the ancestral node of Sirenia, and some internal nodes within Primates, Rodentia, and Cetartiodactyla ( fig. 1a).

Relationships between Life History Traits and Renal Structural Types
Because life history traits here are classified into continuous and discrete types (supplementary table S1, Supplementary Material online), we used different tools with specific and best-fitting models to detect the correlation between life history traits and renal structural types in mammals. First, we utilized Caper, an R package, to analyze the relationship between body size and renal structure types with the phylogenetic generalized least squares (PGLS) method, which showed a significant association between two characters, as expected (R 2 = 0.1055, P < 0.05, n = 62) (table 1). Then, we assessed the correlations between the renal structural types and two discrete traits in the phylolm package with a phylogenetic logistic regression method, which revealed a significant correlation between the renal structure type and habitat (P < 0.05, n = 62) but did not detect any association between the renal structure type and diet (table 1).
Given that habitat was significantly correlated with the evolution of renal structures, we then searched for dependent relationships between these variables. The results showed that habitat was dependent on renal structures or both traits were dependent on each other, but renal structures were not shown to be dependent on habitat (table 2).

Selection Tests of Genes Related to Duplex Kidney Formation across Mammals
Based on the consensus tree from TimeTree (supplementary fig. S2, Supplementary Material online), we used the branch model in Codeml of PAML to identify rapidly evolving genes (REGs) in 41 mammals, which were divided into a DMK group and a nondiscrete multirenculate kidney (non-DMK) group. The results revealed that divergent selective pressure might have acted on mammals with different kinds of renal structures.
After correcting for multiple testing by false discovery rate (FDR, adjusted P < 0.05), we identified 14 REGs in DMK species, 11 REGs in non-DMK species, and 2 genes (i.e., FAT4 and TBC1D32) that overlapped in both groups ( fig. 2a, supplementary table S4 and S5, Supplementary Material online). The result also suggested that the evolutionary rates of these DMK-specific REGs are significantly greater than those of non-DMK species, as high as 24.1-fold ( fig. 2b and c). In addition, in light of gene counts and significance, the Gene Ontology (GO) enrichment analysis showed that these DMK species-specific REGs are significantly enriched for several biological processes, such as cilium assembly, kidney development, urogenital system development, and the cellular components cilia and centrosome ( fig. 2d and supplementary table S6, Supplementary Material online). The Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis suggested that these REGs were overrepresented in several KEGG pathways associated with neurodegeneration and regulating the pluripotency of stem cells (supplementary table S7, Supplementary Material online). However, enrichment analysis of REGs unique to non-DMK species showed no significant enrichment results, which may be due to unconcentrated functions of these genes.
With the branch-site model, fifteen positively selected sites in six key genes were identified in DMK species, and functional enrichment showed that these genes are associated with epithelial tube morphogenesis and regulation of neurogenesis (biological process [BP]), proteoglycan binding (molecular function [MF]), and axon guidance (KEGG) (table 3 and supplementary table S8
Further exploration on how these convergent amino acid changes affect protein function found that six convergent amino acid changes were in the key domain of proteins (e.g., DNAH5 . 3). Among them, the site 3991 of the DNAH5 protein changed from an acidic, negatively charged aspartic acid (D) to a polar, uncharged glycine (G); the site 38 of SLIT2 protein, too, changed from a nonpolar, hydrophobic alanine (A) to polar, uncharged, hydroxyl-containing threonine (T). The likelihood ratio analysis was based on best-fitting model "fitMk" with R package "phytools." Three dependent models were tested: (a) renal structures depend on habitats, (b) habitats depend on renal structures, and (c) both traits depend on each other.

Discussion
The Multilobar Kidney Independently Evolved Multiple Times in Mammals Until now, researchers offered many hypotheses about the origin of renal structures in mammals. The lobulation of the kidney in marine mammals was thought to excrete excess salt and nitrogen waste by expanding the surface area between the medulla and the outer cortex (Williams 2006). However, several lineages of freshwater aquatic mammals and terrestrial mammals shared similar renal structures with marine mammals, and they were considered to have inherited the multilobar kidney from ancestors who originally inhabited marine environments and semiaquatic marine ancestors (Domning 2001;Wursig and Perrin 2009). Previous studies also suggested that the simple UK may be primitive for mammals although the complex multilobar kidney may have derived from it (Straus and Arcadia 1958;Verhaegen 1993). Inferring ancestral states can suggest the evolutionary trajectory of a trait by mapping various phenotypes of living taxa onto phylogenies (Bokma 2008).
Our ancestral state reconstruction suggested that the UK existed in several deep nodes of the mammalian phylogeny, which confirms that the UK is the primordial phenotype of mammals from which the complex multilobar kidney evolved ( fig. 1a). Multiple lineages were inferred to have derived multilobar kidneys (e.g., Cetartiodactyla, Carnivora, Perissodactyla, Rodentia, Primates, Proboscidea, and Sirenia). In more details, CMKs were derived in some lineages (e.g., Cetartiodactyla, Rodentia, Primates, and Sirenia), although DMKs were inferred to occur in several lineages (e.g., Cetartiodactyla, Carnivora, Perissodactyla, and Proboscidea). In summary, our reconstruction of the evolutionary trajectory of the renal structure 1) was based on a widely accepted phylogeny and covered nearly all orders of mammals and 2) adds more reliable information about the origin and evolution of the renal structure. What is more, this finding on the independent evolution of the multilobar kidney may have intriguing implications for research on convergent phenomena.

The Evolution of Renal Structures Was Driven by Body Size and Habitats in Mammals
Species have evolved various life histories based on their environments, and some individual species have evolved adjustable life histories to physiologically respond to different environments (Fisher 1958;Scheiner 1992). Over the past few decades, a lot of correlation analyses have been conducted on renal phenotypes and other traits in mammals and vertebrates. For example, body weight was found to be positively correlated with kidney weight and  (Prothero 1984), whereas body mass was positively correlated with cortical thickness and total/outer/ inner medullary thickness and negatively correlated with relative medullary thickness and maximum urine osmolality, and maximum urine osmolality was positively correlated with relative medullary thickness (Beuchat 1996). Accordingly, body mass seems important to various renal phenotypes.
Here, we found significant correlations between renal structure and two traits (body mass and habitats) based on the consensus phylogeny, which confirmed the two previous hypotheses about the evolution of renal structures in mammals (King 1991) (table 1). The inference mentioned above explains the correlation between renal structure and body mass: to adapt to the absolute increase in metabolism needed for an increase of body size and the number of needed nephrons increased beyond the optimal nephron number of UKs, and thus, the multilobar kidney emerged (Dantzler 1989). Thus, mammals with larger body sizes tend to evolved DMKs.
The change from UK to multilobar kidney may be related to the physical force required to move the fluid along the nephron and the need to reabsorb filtered solutes and water (Dantzler and Braun 1980;Calder and Braun 1983). Previous studies revealed a correlation between glomerular development and habitat in vertebrates; meanwhile, glomeruli are structurally located in the renal cortex (Marshall and Smith 1930). In short, these signs support the finding in our study (table 2) that there is a mutual dependent relationship between the renal structure and habitats. Therefore, aquatic mammals tended to derive DMKs. Understanding the origin and evolution of renal structures across mammals is vital and can make it easier to explore the molecular mechanisms underlying the adaptive evolution of mammals.

Potential Molecular Mechanisms of Convergent Evolution among Mammals with DMKs
Convergent evolution occurs when species that live in the same or similar environments for a long time evolve similar phenotypes. Convergent phenotypic evolution provides a unique perspective on how the genome encodes phenotypes. Recently, many studies have shown that the existence of phenotypic convergence can be seen at different molecular levels, such as convergent amino acid substitutions, convergent REGs, convergent positively selected genes, and convergent changes in amino acid bias (Pascoal et al. 2014;Foote et al. 2015;Hao et al. 2019).
Physiological adaptation or adaptive characters are a homeostatic mechanism in response to instant environmental stimuli, whereas evolutionary adaptation is the result of reproductive success reflecting natural selection (Bennett 1997). The mammalian kidney is critical for homeostasis because it plays a major role in the excretion of metabolic waste, the regulation of extracellular fluid volume, electrolyte balance, and acid-base balance (Berglund et al. 2020). Crucial factors in the evolution of the renal

R F P c F R a H t P
The key functional domains with positively selected sites are shown in bold font in parentheses. Twenty-one amino acid properties in TreeSAAP: P α , α-helical tendencies; N s , average number of surrounding residues; P β , β-structure tendencies; B l , bulkiness; B r , buriedness; R F , chromatographic index; P c , coil tendencies; c, composition; K 0 , compressibility; pK', equilibrium constant for ionization of COOH; C a , helical contact energy; h, hydropathy; pHi, isoelectric point; E l , long-range nonbonded energy; F, Mean r.m.s. fluctuational displacement; M v , molecular volume; M w , molecular weight; H nc , normalized consensus hydrophobicity; V 0 , partial specific volume; P r , polar requirement; P, polarity; α c , power to be-C-term of the α-helix; α m , power to be-middle of the α-helix; α n , power to be-N-term of the α-helix; μ, refractive index; E sm , short-and medium-range nonbonded energy; R a , solvent accessible reduction ratio; H p , surrounding hydrophobicity; H t , thermodynamic transfer hydrophobicity; E t , total nonbonded energy; P, turn tendencies. a Posterior probabilities (pp) of BEB > 95%. b pp > 99%. structure and function of the vertebrate appear to be related to body fluid regulation, including the maintenance of constant water and salt levels in the body (Mahasen 2016). In the present study, we found several possible molecular mechanisms that account for the phenotypic convergence of DMK mammals.
First, 12 genes involved in duplex/multiplex kidney formation were found to have evolved rapidly only in DMK mammals, and the functional enrichment analyses of these genes were mainly about cilium assembly, kidney development, urogenital system development, and two cellular components (cilia and centrosome) (fig. 2). The results of enrichment analyses suggested the potential impact of cilium organization and centrosome amplification on the evolutionary development of kidneys in DMK mammals, given that the primary cilium is an organelle that plays a key role in cell signaling and is mainly associated with SHH signaling, which has been proposed to be involved in duplex kidney formation (San Agustin et al. 2016;Elliott and Brugmann 2019), and amplification of the centrosome could disrupt renal development (Dionne et al. 2018). The etiology of most duplex kidneys can be traced back to ureteral induction, although both the ureter and kidney belong to the urogenital system (Uetani et al. 2009). Thus, the evolutionary development of the DMK is thought to benefit from the evolution of the entire urogenital system. Additionally, two genes (FAT4 and TBC1D32) were found evolved rapidly either in DMK mammals or in non-DMK mammals, FAT4 is a key gene in vertebrate planar cell polarity and its loss can overactivate GDNF-RET signaling, causing premature branching with incomplete duplication (Zhang et al. 2019). TBC1D32 was found to account for the duplex kidney formation as a part of a ciliopathy phenotype (San Agustin et al. 2016). Two convergent amino acid substitutions, which were in crucial functional domains of protein TBC1D32, were detected among species with DMKs. And the detail of amino acid changes is discussed later. This result suggests the same gene may underlie the evolution of both species with DMKs and with other structural types of kidneys for resisting to stress ( fig.  2a) (Valenzano et al. 2015).
Second, comparative analysis of the DMK and non-DMK species revealed that six genes are under positive selection in DMK mammals (table 3). Among them, six positively selected sites in four genes were found in key domains of corresponding proteins. For example, the ARM (Armadillo repeat) domain is a repetitive sequence about 40 residues   FIG. 3.-Convergent amino acid substitutions among DMK mammals. The figure showed the DMK group, which are classified into five taxa. The convergent amino acid substitutions are clearly shown in the picture. The number of the sites in key functional domains is marked clearly too, combined with underlines and bold font, although the one with a box means the mutation in this site of the corresponding protein may be deleterious or damaging.
Genome Biol. Evol. 15(5) https://doi.org/10.1093/gbe/evad075 Advance Access publication 9 May 2023 long, multiple copies of which form an alpha solenoid structure; the domain can control act branching and bundling and is pivotal for transducing WNT signals during embryonic development (protein CTNNB1, site 167) (Nusse and Clevers 2017). In addition, the LRRNT (leucine-rich repeat N-terminal) domain is rich in the hydrophobic amino acid leucine and related to the formation of protein-protein interactions (protein SLIT2, site 279) (Kobe and Kajava 2001). The EGF (epidermal growth factor) domain occurs in many tandem copies in protein and fold together into a functional unit (protein SLIT2, site 984) (Rao et al. 1995). The Fn3 (fibronectin type III) domain consists of many proteins related to ligand binding (protein ROBO2, site 597) (Koide et al. 1998); the B9-C2 (ciliary basal body-associated, B9 protein) domain exists in proteins involved in the ciliary basal body (protein MKS1, site 261 and 364) (Dawe et al. 2007). In addition, 10 out of 15 positively selected sites with radical amino acid property changes were shown to impact the structure and function of their corresponding proteins. These positively selected sites in the crucial domains of proteins may be evidence for the molecular convergence of mammals with DMKs. Furthermore, the enrichment suggested that the biological processes involving epithelial tube morphogenesis, the regulation of neurogenesis, and axon guidance-all of which positively selected genes are enriched in-may be closely related to the adaptive evolution of DMKs in mammals. Among them, Roundabout Guidance Receptor 2 (ROBO2) plays a role in axon guidance and cell migration as a transmembrane receptor for Slit Guidance Ligand 2 (SLIT2). SLIT proteins function in axon guidance and neuronal migration. Knock out of ROBO2 or SLIT2 in mice can lead to an abnormal GDNF expression domain, which causes duplex kidney formation (Grieshammer et al. 2004). MKS Transition Zone Complex Subunit 1 (MKS1) is necessary for ciliated epithelial cells to form primary cilia, and MKS1 mutant can also form the duplex kidney, which is a part of a ciliopathy phenotype (San Agustin et al. 2016). Inactivation of Catenin Beta 1 (CTNNB1) in the nephric duct or hypoxia-induced reduction of it can result in duplex kidneys (Bridgewater et al. 2008;Wilkinson et al. 2015).
Finally, in this study, 6 out of 12 unique convergent amino acid changes were identified in key domains of their respective proteins ( fig. 3). Of them, the dynein heavy chain region D6 P-loop domain comes from a chain involved in ATPase activity, microtubule binding ability, cilia and flagella movement (Mocz and Gibbons 2001), and the ROMI (broad-minded) domain's interaction with cell cycle-related kinase (CCRK) proteins, which together regulate ciliary membrane and axonal growth (Ko et al. 2010); the AAA+ lid domain represents the C terminus of AAA domains from dynein heavy chain D3 (Mocz and Gibbons 2001). Moreover, two unique convergent amino acid changes (e.g., D3991G in protein DNAH5 and A38T in protein SLIT2) were inferred to be deleterious or damaging to protein properties and potentially affect their functions. DYNEIN, AXONEMAL, HEAVY CHAIN 5 (DNAH5) encodes a kinesin protein, and mutation in it can cause primary ciliary dyskinesia type 3 (Olbrich et al. 2002). Mutants in mice lacking SLIT2 or ROBO2 form additional ureteral buds, which can lead to phenotypes such as duplex kidney (Grieshammer et al. 2004). Taken together, these DMK species-specific convergent amino acid substitutions in genes associated with duplex kidney formation may account for the phenotypical convergence among them.
However, renal structural development is an intricate process; future studies based on multiomics analysis are needed to shed light on the deeper and more comprehensive molecular mechanisms underlying the evolution of DMKs in mammals. Experimental verification is also essential, if conditions permit.

Conclusion
With regard to the origin and macroevolution of renal structures across mammals, we found that the UK was the ancestral state in mammals, and that multilobar kidneys independently evolved multiple times. In addition, mammals with larger body sizes or living in aquatic environments tended to evolve DMKs. With regard to molecular evolution, comparative genomic analyses of 45 genes related to duplex/multiplex kidney formation in 41 mammals revealed that the convergent evolution of mammals with DMKs may be resulted from the REGs associated with cilium assembly and centrosome amplification, the positively selected genes involved in epithelial tube morphogenesis, regulation of neurogenesis and axon guidance, and six key convergent amino acid substitutions in functional domains of proteins. Taken together, our study provides novel insights into the origin and evolution of renal structures, especially DMKs, and offers a better understanding of the pathology of duplex/multiplex kidneys in human from macroecological and micromolecular perspectives.

Species Coverage and Sequence Acquisition
The ancestral state reconstruction and correlation analyses covered a total of 62 mammals from 13 orders: Cetartiodactyla ( The gene set related to duplex/multiplex kidney formation in humans was collected from a review article published in 2020 (Kozlov and Schedl 2020) and a database using the keyword "duplex kidney" (http://www. informatics.jax.org/mp/annotations/MP:0004017) (supplementary table S9, Supplementary Material online). The protein-coding sequences (CDS) were then downloaded from the NCBI database (https://www.ncbi.nlm. nih.gov/). In addition, the partial or unannotated CDS was further verified using BlastN searches with custom perl scripts. The longest transcript was retained for each gene in this analysis. Finally, 45 one-to-one orthologous genes among 41 species were used for the analysis described below. To obtain better quality sequence alignments, we performed multiple sequence alignments for each orthologous gene using PRANK v.170427 (Löytynoja 2014) combined with MACSE v2 (Ranwez et al. 2018) in the codon mode. The aligned sequences were then trimmed using Gblocks v0.91 (Talavera and Castresana 2007) with default settings.

Ancestral State Reconstruction
To reconstruct the ancestral state of mammals, we took the phenotype data from a 1996 research article (Beuchat 1996) and downloaded the corresponding phylogenetic tree file from TimeTree (http://www.timetree.org/) (Kumar et al. 2017) (supplementary fig. S1, Supplementary Material online). Then, we carried out an ancestral state reconstruction using Bayesian Binary MCMC (BBM) analysis implemented in RASP v4.2 (Yu et al. 2020). The program calculated several probable results using corresponding models (e.g., DEC, DEC + J, DIVALIKE, DIVALIKE + J, BAYAREALIKE, and BAYAREALIKE + J), and we chose the most reliable one, given that the reconstruction with chosen model possessed the lowest AIC value and the highest AICc_wt value (i.e., the fixed model DEC + J).

Association Analysis of Life History Traits and Renal Structural Phenotypes
We collected the life history traits from different reliable resources, such as the body mass of 62 mammals from the PanTHERIA database (Jones et al. 2009) and feeding and habitat data from the ADW database (https:// animaldiversity.org/).
We used several approaches in this analysis. Regarding the relationship between body mass and renal structures in mammals, PGLS regression was employed in the R package Caper (Orme et al. 2013). We transferred the renal structural phenotype into a binary state: discrete and non-DMKs. In terms of discrete variables (habitats and diets), we conducted the function phyloglm in the R package phylolm version 2.6 with the IG10 phylogenetic generalized linear model (Tung Ho and Ané 2014). To further explore the independent and dependent evolution between habitats and renal structures in mammals, we implemented the analysis using the R package phytools with maximum likelihood models to test how one trait affects the transition rates of another trait (Singh et al. 2012).

Molecular Evolution Analyses
To test the selective pressure on genes, we estimated the ratio of nonsynonymous (d N )/synonymous (d S ) substitution rates (d N /d S ) implemented in the CodeML program of the PAML software package v4.9 (Yang 2007). The ratio is also called ω value, and ω < 1, ω = 1, and ω > 1 mean purified selection, neutral evolution, and positive selection, respectively. The phylogenetic tree used here was downloaded from TimeTree (supplementary fig. S2, Supplementary Material online). The ancestral and terminal branches of all species with DMKs were labelled as foreground branches, other branches as background branches.
We detected the rapid evolution of genes using the branch model in 41 mammals. The one-ratio model assumes that all branches on the phylogenetic tree have the same ω ratio against an alternative hypothesis (two-ratio model), which allows the ω ratio of the foreground branch to differ from that of the background branch. Then, we executed a likelihood ratio test (LRT) with a chi-square distribution and applied the FDR correction for multiple testing.
The branch-site model was then used to detect the positively selected genes and amino acids in mammals. All positively selected sites in this analysis were identified using a Bayes Empirical Bayes (BEB) analysis with posterior probabilities ≥ 0.80 (Yang et al. 2005). Furthermore, we estimated the amino acid physicochemical properties of crucial sites GBE using TreeSAAP software, and sites with values of 6-8 were considered to have radical amino acid property changes (Woolley et al. 2003).

Convergent Amino Acid Substitution Detection
Convergent phenotypic characteristics can result from specific substitutions that independently evolved in different species (Natarajan et al. 2015). We detected the molecular basis of convergent evolution in species with DMK by identifying the unique convergent amino acid substitutions based on sequence alignments using FasParser2 with the "Segregate" function (Sun 2018). If we identified the same amino acid changes in at least two lineages with DMKs, then these amino acid changes were considered to be group-specific convergent amino acid changes.

Functional Domain Searching and Key Site Labeling on 3D Structures of Proteins
To verify whether the positively selected sites or unique convergent amino acid sites were located in the key domains of proteins, we searched the UniProt database (https://www.uniprot.org/) (UniProt Consortium 2019) and mapped the sites onto the 3D protein structures predicted from SWISS-MODEL (https://swissmodel.expasy. org/) by EzMol (http://www.sbg.bio.ic.ac.uk/ezmol/) (Reynolds et al. 2018;Waterhouse et al. 2018).

Gene Functions and Signaling Pathway Annotation and Enrichment
The functional enrichment analyses in GO for biological process (BP), cellular component (CC), molecular function (MF), and Kyoto Encyclopedia of Genes and Genomes (KEGG) were performed in R package clusterProfiler (Yu et al. 2012). FDR was performed using the Benjamini and Hochberg (BH) method (Benjamini and Hochberg 1995).

Functional Effects Estimating of Convergent Amino Acid Substitutions
We tested the potential effects of these convergent amino acid substitutions on the corresponding proteins using PolyPhen-2 (Adzhubei et al. 2010) and PROVEAN (Choi and Chan 2015). The output of PolyPhen-2 is classified as benign, possibly damaging (low confidence), and probably damaging (high confidence), along with score ranging from 0 (benign) to 1 (damaging). In order to obtain high balanced accuracy, the cut-off value of the PROVEAN score is set to −2.5. The protein variant is predicted to be deleterious when the score is less than or equal to the predefined threshold.

Supplementary Material
Supplementary data are available at Genome Biology and Evolution online (http://www.gbe.oxfordjournals.org/).