1 Gene expression phylogenies and ancestral transcriptome

2 reconstruction resolves major transitions in the origins of pregnancy

4 Katie Mika1,2, Camilla M. Whittington3, Bronwyn M. McAllan3,4, and Vincent J. Lynch5,*

1
6 Department of Human Genetics, The University of Chicago, 920 East 58th Street, CLSC 319C,
7 Chicago, IL 60637, USA.

2
8 Current address: Department of Organismal Biology and Anatomy, University of Chicago,
9 1025 E 57th Street, Chicago, IL, 60637, USA
3
10 School of Life and Environmental Sciences, University of Sydney, Sydney, NSW 2006,
11 Australia
4
12 Charles Perkins Centre, University of Sydney, Sydney, NSW 2006, Australia
5
13 Department of Biological Sciences, University at Buffalo, SUNY, 551 Cooke Hall, Buffalo, NY,
14 14260, USA.
*
15 Correspondence: [email protected]

16
17

18

19

20

21

22

23

24

25

1
26 Abstract

27 Structural and physiological changes in the female reproductive system underlie the origins of
28 pregnancy in multiple vertebrate lineages. In mammals, the glandular portion of the lower
29 reproductive tract has transformed into a structure specialized for supporting fetal development.
30 These specializations range from relatively simple maternal nutrient provisioning in egg-laying
31 monotremes to an elaborate suite of traits that support intimate maternal-fetal interactions in
32 Eutherians. Among these traits are the maternal decidua and fetal component of the placenta,
33 but there is considerable uncertainty about how these structures evolved. Previously we showed
34 that changes in uterine gene expression contributes to several evolutionary innovations during
35 the origins of pregnancy (Marinic, Mika, and Lynch 2021). Here we reconstruct the evolution of
36 entire transcriptomes ("ancestral transcriptome reconstruction") and show that maternal gene
37 expression profiles are correlated with degree of placental invasion. These results indicate that
38 an epitheliochorial-like placenta evolved early in the mammalian stem-lineage and that the
39 ancestor of Eutherians had a hemochorial placenta, and suggest maternal control of placental
40 invasiveness. These data resolve major transitions in the evolution of pregnancy and indicate
41 that ancestral transcriptome reconstruction can be used to study the function of ancestral cell,
42 tissue, and organ systems.

43 Introduction

44 Studies of the fossil record have revealed in fine detail major stages in the origin and
45 diversification of evolutionary novelties (structures) and innovations (functions) such as the
46 vertebrate limb and skull (Abzhanov, 2015; Hirasawa and Kuratani, 2015), the turtle shell (Lyson
47 and Bever, 2020), feathers (Chen et al., 2015), and flowers (Chanderbali et al., 2016). Many
48 features of soft tissues and macromolecular structures, however, are lost during the fossilization
49 process and thus leave little to no trace in the fossil record. To reconstruct the history of these
50 characters, including DNA and amino acid sequences, morphology, physiology, and behavior,
51 among others, evolutionary studies have traditionally relied on comparative (phylogenetic)
52 methods such as parsimony or model-based maximum likelihood or Bayesian inference (Lewis
53 and Olmstead, 2001). These methods infer ancestral characters from the distribution of
54 character states among extant species, the latter of which also allows increased model
55 complexity including unequal character state transitions (Lewis and Olmstead, 2001; Pagel,
56 1999), variable rates among sites and branches (Galtier, 2001; Yang, 1994, 1993), and
57 character dependent diversification (Maddison et al., 2007).

2
58 Extant mammals span crucial transitions in the origins of pregnancy (Figure 1A) and are
59 an excellent system in which to explore the origins of evolutionary novelties. The platypus and
60 echidna (monotremes), for example, are oviparous (egg-laying) but retain the egg in the
61 glandular portion of uterus for about two weeks. During this period the developing embryo is
62 nourished by uterine secretions (matrotrophy) delivered through a simple yolk sac placenta
63 (Hughes and Hall, 1998; Renfree and Shaw, 2013). Viviparity (live-birth) evolved in the stem-
64 lineage of therian mammals, but marsupials and eutherians have very different reproductive
65 strategies, particularly in the ontogenetic origins of the definitive placenta and the arrangement
66 of the maternal-fetal interface (Freyer and Renfree, 2009; Renfree and Shaw, 2013; Renfree,
67 2010). In most marsupials, the embryonic portion of the placenta is derived from the yolk sac,
68 which may come into direct contact with but does not invade the maternal endometrium. While
69 the yolk sac has essential functions during early pregnancy in eutherians, the definitive placenta
70 is derived from chorion and allantois (chorioallantois) and varies in its degree of endometrial
71 invasion (Swanson and Skinner, 2018)(Figure 1B). Thus, matrotrophy and a yolk-sac derived
72 placenta were present in the mammalian stem-lineage and transitioned to a chorioallantoic
73 placenta in eutherians.

74 Unfortunately, most characters related to pregnancy and the nature of the maternal-fetal
75 interface leave little to no trace in the fossil record. Thus, studies exploring the evolution of
76 pregnancy have relied on comparing morphological differences between extant mammals to
77 reconstruct the steps in the origins of pregnancy. These comparative analyses have used
78 multiple methods to reconstruct the arrangement of the mammalian maternal-fetal interface and
79 reached contradictory conclusions about the degree of placental invasiveness in the eutherian
80 ancestor, a debate which has persisted since at least 1876 (Table 1). Here, we use comparative
81 transcriptomics and maximum likelihood to infer ancestral gene expression states (ancestral
82 transcriptome reconstruction) from 23 amniotes with different parity modes and degrees of
83 placental invasion to identify the evolutionary history of the maternal-fetal interface. We found
84 strong evidence that the last common ancestor of eutherian mammals had an invasive
85 hemochorial placenta, as well as convergence in gene expression profiles during the
86 independent evolution of non-invasive epitheliochorial placentas in marsupials and some
87 eutherian lineages. These data indicate that the degree of placental invasion can be inferred
88 from endometrial gene expression profiles, and suggest that placental invasiveness is regulated
89 by gene expression profiles in the maternal endometrium rather than the fetal portion of the
90 placenta.

3
91 Table 1. Inferences of placental invasiveness in the eutherian ancestor 1880-2021. Note
92 that we follow Mossman (Mossman, 1937, p. 156) and define a placenta as “any intimate
93 apposition or fusion of the fetal organs to the maternal (or paternal) tissues for physiological
94 exchange.”

Study Data Method(s) State

(Turner, 1876) – Ontogeny Epitheliochorial

(Haeckel 1883) – Ontogeny “Non-Invasive”

(Minot, 1891) Ontogeny Hemochorial

(Wislocki, 1929) – Implicit Parsimony? Hemochorial

(Hill, 1932) – Ontogeny Epitheliochorial

(Mossman, 1937) Ontogeny Endotheliochorial

(Portmann, 1938) – Ontogeny Hemochorial

(Martin, 1969) – Ontogeny / Implicit Parsimony Endotheliochorial

(Kihlström, 1972) – Gestation length / Placenta- (Endothelio/hemo)chorial

type Correlation

(Luckett, 1976, 1975, – Ontogeny Epitheliochorial

1974)

(Carter and Enders, 2004) 14 taxa Ontogeny / Implicit Parsimony Endotheliochorial

(Vogel, 2005) 22 taxa Implicit Parsimony Hemochorial

(Elliot and Crespi, 2006) 88 taxa Maximum Likelihood Hemochorial

(Mess and Carter, 2006) 36 taxa Maximum Parsimony (Endothelio/hemo)chorial

(Wildman et al., 2006) 44 taxa Maximum Parsimony & Hemochorial

Maximum Likelihood

(Martin, 2008) 18 taxa Parsimony Endotheliochorial

(ordered) or
(Endothelio/hemo)chorial
(unordered)

(Elliot and Crespi, 2009) 334 taxa Maximum Parsimony & (Endothelio/hemo)chorial
Maximum Likelihood (MP) or Hemochorial
(ML)

(McGowen et al., 2014) 66 taxa Maximum Likelihood (Superficial)

Hemochorial

95

4
96 Results

97 Endometrial Gene Expression Profiling

98 We previously assembled a collection of transcriptomes from the pregnant or gravid

99 uterine endometrium of therian mammals with varying degrees of placental invasiveness, as
100 well as a monotreme (platypus), a bird, as well as viviparous, oviparous, and reproductively bi-
101 modal lizards (Marinić et al., 2021). The complete dataset includes expression information for
102 21,750 genes from 23 species (Figure 2–source data 1). Principal Component Analysis (PCA)
103 based on gene expression levels in transcripts per million (TPMs), using a variant of PCA that
104 extends PCA to binary data, generally grouped species randomly (Figure 2A), consistent with
105 noise in gene expression data overwhelming phylogenetic signal. Therefore, we transformed
106 quantitative gene expression values in transcripts per million (TPM) into discrete character
107 states – Genes with TPM ≥ 2.0 were coded as expressed (state=1), genes with TPM < 2.0 were
108 coded as not expressed (state=0), and genes without data in specific species coded as missing
109 (?) (Marinić et al., 2021; Mika et al., 2021). In contrast to PCA based on TPM values, PCA of the
110 binary encoded endometrial transcriptome dataset grouped species by phylogenetic relatedness
111 (Figure 2B), indicating significant noise reduction in the binary encoded dataset.

112 Phylogenetic analyses of endometrial transcriptomes

113 We used IQ-TREE to infer the best fitting model of character evolution and the
114 maximum-likelihood (ML) phylogeny. The best-fit model of character evolution was a general
115 time reversible model for binary data (GTR2) with character state frequencies optimized by
116 maximum-likelihood (FO) and rate heterogeneity across sites accommodated with the FreeRate
117 model that had three rate categories (R3). Next, we assessed the tree topology and branch
118 support metrics for the ML phylogeny inferred using the binary encoded endometrial gene
119 expression dataset and the GTR2+FO+R3 model. The ML phylogeny (Figure 3A) generally
120 followed taxonomic relationships (Figure 3B), however, four particularly discordant relationships
121 within therian mammals were inferred with high support: 1) Rather than grouping marsupials
122 into a monophyletic sister-clade to the eutherians, opossum and wallaby were placed as sister-
123 species within the Boreoeutheria; 2) Armadillo groups within the Euarchontoglires rather than as
124 sister to all other eutherians; 3) Bat groups within the Euarchontoglires rather than within
125 Laurasiatheria; and 4) Dog groups within the Euarchontoglires rather than with Laurasiatheria
126 (Figure 3A/B). We also used multiple non-parametric topology tests to directly compare the ML
127 tree to alternative trees with the correct phylogenetic placement of these four discordant

5
128 lineages, all of which rejected alternative “corrected” trees in favor of the (Figure 3C).
129 Remarkably while these discordant relationships are incorrect with respect to the species
130 phylogeny, they are correct with respect to placenta-type, i.e., species form well-supported
131 clades based on their degree of placental invasiveness: Wallaby and opossum, which have
132 epitheliochorial placentas similar to ungulates forms a clade with ungulates, Armadillo, which
133 has an invasive placenta (Carter and Enders, 2004; Chavan and Wagner, 2016), forms a clade
134 with the other species with hemochorial placentas, and dog, which has an invasive
135 endotheliochorial placenta, forms a clade with Euarchontoglires that have invasive hemochorial
136 placentas.

137 Ancestral transcriptome reconstruction and fuzzy C-Means clustering

138 We also used IQ-TREE and the species phylogeny (Figure 3B) to reconstruct ancestral
139 gene expression states for each gene (ancestral transcriptome reconstruction). To explore the
140 similarity of extant and reconstructed transcriptomes we used Fuzzy C-Means (FCM) clustering,
141 a “soft” clustering method that allows each sample to have membership in multiple clusters and
142 assigns samples to clusters based on their degree of cluster membership. FCM with two to four
143 clusters (K=2-4) had a clear biological interpretation (Figure 4): 1) K=2 clustered eutherians and
144 non-eutherians; 2) K=3 clustered most therians with non-invasive (epitheliochorial) placentas,
145 eutherians with invasive (endotheliochorial or hemochorial) placentas, platypus and sauropsids
146 (i.e. viviparous and oviparous lizards, and birds); 3) K=4 clustered eutherians with invasive
147 placentas, eutherians with epitheliochorial placentas, opossum/wallaby, and sauropsids. A
148 notable exception with K=3-4 is the cluster membership of dunnart, which is discussed in
149 greater detail below. FCM clusters with K=5 and K=6 were similar to K=4, but divided
150 Eutherians with hemochorial placentas into two clusters, and clustered dog, dunnart, and the
151 viviparous skink Chalcides ocellatus (Figure 4); Beyond K=6 clusters had no clear biological
152 interpretation.

153 Ancestral transcriptome reconstructions generally clustered with extant species having
154 similar parity modes and degrees of placenta invasiveness (Figure 4). For example, FCM with
155 K=2 grouped extant eutherians and their ancestral lineages as well as extant non-eutherians
156 and their ancestral lineages. Similarly, FCM with K=4 grouped extant eutherians with invasive
157 placentas and their ancestral lineages with extant therians with non-invasive placentas and their
158 ancestral lineages. FCM with K=3–5 clustered the ancestral eutherian (AncEutheria)
159 transcriptome with extant species that have invasive hemochorial placentas and clustered the

6
160 ancestral therian (AncTheria) and mammalian (AncMammalia) transcriptomes with extant
161 mammals that have non-invasive epitheliochorial placentas. These data suggest that FCM
162 clustering of extant and ancestral reconstructions of endometrial transcriptomes can predict
163 ancestral placenta invasiveness, implying that an epitheliochorial placenta evolved early in the
164 development of mammalian pregnancy and that a hemochorial placenta is ancestral for
165 eutherians.

166 While FCM clustering generally groups extant and ancestral transcriptomes by
167 phylogenetic relatedness and degree of placental invasiveness, a notable exception is the
168 marsupial fat-tailed dunnart (Sminthopsis crassicaudata). FCM clusters with K=2–5 grouped
169 dunnart with non-therians while FCM K=6 clustered dog, dunnart, and the skink, C. ocellatus
170 (Figure 4). FCM cluster membership coefficients of all three species with K=2–5 were mixed,
171 with significant membership across clusters. In contrast, dog, dunnart, and C. ocellatus formed
172 a distinct cluster from all other species at K=6 with nearly 100 % FCM cluster membership in
173 group 6 (Figure 4). Remarkably, both dog and dunnart have endotheliochorial placentas
174 whereas the Chalcides maternal-fetal interface has been described as either endotheliochorial
175 or epitheliochorial with extensive vascularization and interdigitating folds of hypertrophied
176 uterine and chorioallantoic tissue (Blackburn, 1993; Blackburn and Callard, 1997; Corso et al.,
177 2000). These data suggest that species with endotheliochorial placentas have a gene
178 expression profile that is intermediate between epitheliochorial or hemochorial placentas, yet is
179 also distinct, and that gene expression at the Chalcides maternal-fetal interface are converging
180 on a therian-like endotheliochorial pattern.

181 Convergent loss of RORA in species with epitheliochorial placentas

182 Among the genes with discordant species- and gene-expression phylogenies is RAR-
183 related orphan receptor alpha (RORA), an orphan nuclear hormone receptor that regulates the
184 development and function of type 2 and type 3 innate lymphoid cells (ILC2 and ILC3), which
185 negatively regulates the immune system in local microenvironments especially during
186 inflammation (Haim-Vilmovsky et al., 2020), the establishment of tolerance to intestinal
187 microbiota (Lyu et al., 2022), and the regulation of inflammatory responses and vascular
188 remodeling during placentation and pregnancy (Balmas et al., 2018; Mendes et al., 2020; Miller
189 et al., 2018). Remarkably RORA independently lost endometrial expression at least four times,
190 each loss coincident with the independent evolution of non-invasive epitheliochorial placentas
191 (Figure 5A). The restricted expression of RORA to immune cells in the human first trimester

7
192 decidua (Vento-Tormo et al., 2018) (Figure 4B) suggests that the repeated loss of RORA
193 expression reflects changes in the composition of immune cells at the maternal-fetal interface in
194 species with epitheliochorial placentas. To explore this possibility, we used CIBERSORT to
195 deconvolve endometrial bulk RNA-Seq datasets from each species into cell-type abundance
196 estimates with a signature gene expression matrix composed of cell-types from the human first
197 trimester maternal-fetal interface. We found that CIBERSORT inferred that nearly all species
198 that lacked RORA expression also lacked ILC2/3 cells at the maternal-fetal interface with the
199 exception of armadillo and chicken which had RORA expression but were inferred to lack
200 endometrial ILC2/3 cells, and the viviparous Saiphos population which has an epitheliochorial
201 placenta and RORA expression and was inferred to have a population of ILC2/3 cells. Thus, we
202 conclude that there is convergent loss of RORA in species with epitheliochorial placentas which
203 in many species correlates with the convergent loss of ILC2/3 cells at the maternal-fetal
204 interface.

205 Discussion

206 One of the central of goals of DevoEvo is a mechanistic explanation for the origin and
207 evolution of evolutionary novelties (Amundson, 2005; Brigandt and Love, 2010; Lynch, 2022;
208 Wagner, 2001, 2000; Wagner and Larsson, 2003). A major challenge for reconstructing the
209 origins of evolutionary novelties, however, is a lack of transitional forms among living species.
210 For example, while feathers and the turtle’s shell are excellent examples of morphological
211 novelties, there are no living species with transitional forms between scale and feather (Chen et
212 al., 2015) or with protoshells (Lyson and Bever, 2020). Despite the lack of transitional forms
213 among extant taxa, an abundance of fossil data has reconstructed the major steps in the
214 evolution of these structures that when combined with molecular studies of their development
215 provides a rich explanation for their developmental evolution. Unfortunately, many features of
216 soft tissues leave little to no trace in the fossil record, thus the steps in their evolution have
217 remained elusive. Here we used ancestral transcriptome reconstruction to trace gene
218 expression changes during the origins of mammalian pregnancy, in which extant species
219 preserve intermediate stages in the evolution of both pregnancy and a diversity of placenta
220 types and reconstruct the evolution of placental invasiveness of ancestral species.

221 Binary encoding uncovers hidden biological signal

222 While transforming gene expression count data into binary categories likely loses
223 information about evolutionarily relevant variation in expression levels, it may not be possible to

8
224 infer meaningful gene expression levels from bulk RNA-Seq. For example rather than
225 evolutionary differences between individuals or species, variation in transcript abundance
226 between samples can result from various sources of experimental noise such as technical
227 variation in library preparation, sequencing, or batch effects (Gilad and Mizrahi-Man, 2015; Tung
228 et al., 2017), variation in cell-type composition of a tissue (Price et al., 2022), sampling different
229 timepoints in development or only a few individuals that do not capture the variance properties
230 of gene expression levels within a population or species (Pal et al., 2021; Thompson et al.,
231 2020). Thus, by transforming gene expression data into not/expressed states we may reduce
232 the potential for these and other biases to influence our ancestral transcriptome reconstructions,
233 revealing biological signal. Our binary encoded endometrial gene expression dataset, for
234 example, cluster species with similar placenta types and parity mode (Figure 2B) indicating
235 binary encoding preserves functional signal in gene expression data that is otherwise masked
236 by variation in gene expression levels (Figure 2A).

237 Phylogenetic analyses of endometrial transcriptomes

238 Previous studies of molecular phylogenies have shown that gene-species tree
239 discordance can result from convergent or parallel amino acid changes in lineages that
240 independently evolved morphological traits. For example, parallel amino acid changes in prestin
241 (SLC26A5), a motor protein expressed in outer hair cells in the cochlea that essential for
242 hearing, have occurred in different lineages of echolocating mammals, including multiple
243 lineages bats and whales, leading a strongly monophyletic clades in gene trees that do not
244 reflect taxonomic relationships (Li et al., 2008, 2010; Liu et al., 2010; Teeling, 2009). Gene-
245 species tree discordance is also associated coadaptation of amino acid substitutions in Na+K+–
246 ATPase of frogs that prey on toxic toads (Mohammadi et al., 2021), the evolution of C4
247 phosphoenolpyruvate carboxykinases (PCK) in grasses (Christin et al., 2009), and between
248 snake and agamid lizard mitochondrial genomes (Castoe et al., 2009). These examples
249 highlight how phylogenetic discordance can be a signal of convergent evolution, rather than just
250 a sign of biased phylogenetic inference that can arise from processes like long branch attraction
251 (Bergsten, 2005; Felsenstein, 1978), incomplete lineage sorting and introgression (Guerrero
252 and Hahn, 2018; Hibbins et al., 2020; Maddison and Knowles, 2006), biased character state
253 frequencies and gene conversion (Figuet et al., 2014; Kostka et al., 2012; Lartillot, 2013;
254 Romiguier et al., 2013), and heterotachy (Kolaczkowski and Thornton, 2008; Philippe et al.,
255 2005), among others.

9
256 Similar to the signals of convergence from gene-species tree discordance, our
257 transcriptomic data showed significant phylogenetic support for uterine transcriptomes grouping
258 by parity mode and degree of placental invasiveness rather than phylogenetic relationships.
259 This transcriptome-species tree discordance is particularly striking for opossum and wallaby,
260 which are deeply nested within eutherians with epitheliochorial placentas based on uterine
261 transcriptome data, because the eutherian and marsupial placenta is derived from different
262 extra-embryonic tissues – the chorion and allantois in eutherians and the yolk-sac in marsupials.
263 Thus, there must be significant convergence in endometrial gene expression profiles between
264 eutherians with chorioallantois-derived epitheliochorial placentas and marsupials with yolk-sac
265 epitheliochorial placentas because our transcriptome phylogeny is based on gene expression in
266 the endometrium during pregnancy rather than gene expression in the placenta. This
267 convergence in gene expression between species with non-invasive epitheliochorial placentas
268 may be related to the expression of genes that limit the ability of the trophoblast to invade into
269 maternal tissues (see below).

270 Reconstruction of ancestral (endometrial) transcriptomes

271 Previous studies have explored transcriptome evolution with the goal of developing
272 methods of ancestral transcriptome inference (Price et al., 2022), characterizing the general
273 tempo and mode of gene expression evolution (Bauernfeind et al., 2021), implicating
274 transcriptome evolution in the origin and evolution of morphological traits (Church et al., 2022;
275 Katelyn et al., 2021; Lynch et al., 2015; Marinić et al., 2021; Munro et al., 2021), and identifying
276 specific genes with derived expression levels (Brawand et al., 2011; Necsulea et al., 2014),
277 usually in the context of characterizing genes whose expression levels have changed because
278 of the action of positive selection, i.e., directional selection on transcript abundance (Gu, 2004;
279 Price et al., 2022; Yang et al., 2020). In contrast, only a few studies have treated the expression
280 of individual genes as a discrete character, transforming quantitative gene expression levels into
281 binary not/expressed states, and used ancestral state reconstruction to determine the
282 expression state of all genes in an ancestral transcriptome rather than changes in the
283 expression levels of specific genes (Katelyn et al., 2021; Kin et al., 2015; Lynch et al., 2015;
284 Marinić et al., 2021). Our ancestral transcriptome reconstructions of gene expression state
285 indicate that an epitheliochorial-like placenta evolved early in the mammalian stem-lineage,
286 before the loss of the egg-shell in Therians, and that the ancestor of Eutherians had a
287 hemochorial placenta, which resolves a longstanding debate about the nature of ancestral
288 mammalian placentas. These data suggest that ancestral transcriptome reconstruction can be

10
289 used to infer the function of ancestral cell, tissue, and organ systems which leave little to no
290 trace in the fossil record even if soft tissues might by chance fossilize.

291 Convergent loss of RORA in species with epitheliochorial placentas

292 Our observation that RORA expression has been lost multiple times in species with non-
293 invasive epitheliochorial placentas, which is coincident with the inferred absence of ILC2/ILC3
294 cells from the endometrium during at least one stage in pregnancy. This result suggests a
295 mechanistic connection between absence of decidual RORA expression, the absence of
296 decidual ILC2/ILC3 cells, and the loss of placental invasion. Indeed, the major decidual ILC
297 produce signaling factors such as GM-CSF, XCL1, MIP1α, and MIP1β, whose receptors are
298 expressed by EVT likely regulate placental invasion (Huhn et al., 2020). ILC2 also contributes to
299 the maintenance of a type-2 anti-inflammatory immune environment in the uterus during
300 pregnancy (Balmas et al., 2018), which may be particularly important in species in which the
301 endotheliochorial or haemochorial placenta invades maternal tissues. ILC2s, for example, are
302 increased in the decidua basalis of women with spontaneous preterm labor compared to those
303 who delivered preterm without labor (Mendes et al., 2021; Xu et al., 2018), suggesting that ILCs
304 may participate in the chronic inflammatory process that occurs during pregnancy. Thus, loss of
305 placental invasion may be associated with a reduced need to limit local inflammation by ILCs,
306 leading to an evolutionary loss of ILCs in the endometrium during pregnancy. However, while
307 our data are consistent this role for decidual ILC2 and its association with placental invasion,
308 more detailed studies specifically aimed at characterizing cell-type composition differences
309 across species are necessary to determine if there are correlations between cell-types in the
310 endometrium during pregnancy and placenta type.

311 Ideas and Speculation: Maternal control of placental invasion

312 Our observation that endometrial gene expression patterns are correlated with degree of
313 placental invasiveness might seem surprising, however, rather than acting as a passive
314 substrate into which the trophoblast invades, the endometrium directly controls trophoblast
315 invasion (Cui et al., 2012; Graham and Lala, 1991). For example, the trophoblast of mammals
316 with hemochorial placentas, such as humans and rodents, is only permissive to invasion when
317 the “window of implantation” is opened by the endometrium. Similarly, while the trophoblast of
318 mammals with non-invasive endotheliochorial and epitheliochorial placentas, such as cats,
319 dogs, horses, cows, pigs, and sheep cannot invade into the endometrium, they can invade into
320 ectopic sites (reviewed in (Corpa, 2006)). The trophoblasts of guinea pig (Loeb, 1914), mouse

11
321 (Billington, 1965), rat (Jollie, 1961), and pig (Samuel and Perry, 1972) also invade ectopic sites
322 in experimentally induced ectopic pregnancy. While there is no similar ectopic pregnancy data
323 for marsupials, some marsupial lineages, including dunnart, have evolved invasive placentation.
324 In contrast, there is no invasion of maternal tissues during ectopic pregnancy in viviparous
325 reptiles with epitheliochorial placentation such as Pseudemoia entrecasteauxii (Griffith et al.,
326 2013). Thus, the invasive ability of trophoblasts most likely either evolved in the stem-lineage of
327 therian mammals or multiple times, including in the stem-lineage of eutherians and some
328 lineages of marsupials. Regardless, ancestral maternal control of placental invasion likely
329 allowed us to infer ancestral placental invasiveness from ancestral endometrial transcripomes.

330 Ideas and Speculation: Evolution of placental invasion and cancer metastasis

331 Numerous authors have noted the similarity between placental invasion and cancer
332 metastasis (Costanzo et al., 2018; Ferretti et al., 2006; Kozlov, 2022; Lala et al., 2021; Manzo,
333 2019; Murray and Lessey, 1999; Perry et al., 2009; Piechowski, 2019), which was first proposed
334 in 1902 by Scottish embryologist John Beard (1858–1924) who hypothesized that ectopic
335 trophoblasts gave rise cancer (Gurchot, 1975; Ross, 2014). While Beard’s hypothesis is
336 incorrect, at least for cancers not derived from the placenta such as choriocarcinoma, there are
337 numerous mechanistic similarities between implantation, placental invasion, and tumor
338 progression to malignancy (Nordor et al., 2017; Wagner et al., 2021). These data suggest that
339 the evolution of maternal mechanisms that prevent endometrial invasion through expression
340 gain and loss of genes that restrain and promote, respectively, trophoblast invasion, may be
341 related to resistance to metastasis in eutherian lineages with non-invasive placentas (Boddy et
342 al., 2020; Kshitiz et al., 2019; Wagner et al., 2020). For example, pleiotropy can lead to
343 correlated patterns of gene expression between the transcriptomes of different tissue and organ
344 systems (Liang et al., 2018). Thus, the evolution of a gene regulatory module that restricts
345 placental invasion into the endometrium can be coopted to restrict implantation and spread of
346 cancer cells into metastatic locations.

347

348

349 Caveats and limitations

350 A limitation of this study not directly addressed thus far is that we have only sampled a
351 small number of species. For example, we lack pregnant endometrial samples from most

12
352 mammals, particularly those with endotheliochorial placentas, as well as a diversity of oviparous
353 and viviparous squamates (there are at least 115 origins of viviparity in squamates (Blackburn,
354 2015; Blackburn and Brandley, 2015; Blackburn and Starck, 2015). Thus, our inferences from
355 phylogenetic, ancestral reconstruction, and clustering analyses may be biased by small sample
356 sizes and non-random sampling. We also assume that models of evolution designed for
357 phylogenetic inference and ancestral reconstruction of morphological and molecular data are
358 appropriate for gene expression data or binary encoded gene expression data, which may affect
359 our results; for example, we have not directly accommodated incomplete lineage sorting which
360 can mislead phylogenetic inference (Guerrero and Hahn, 2018; Hibbins et al., 2020). Similarly,
361 while Fuzzy C-Means clustering is conceptually similar to topic ("grade of membership") models
362 used in population genetics, its underlying assumptions may be violated for gene expression
363 and binary encoded gene expression data. More detailed studies are necessary to determine if
364 our results are robust to potential sources of error such as model mis-specification, small
365 sample sizes, and non-random taxon sampling, as well as incomplete lineage sorting, the latter
366 of which we were unable to directly test.

367 Conclusions

368 Previous critiques of statistical methods to infer ancestral states, particularly in the
369 context of parity mode evolution in squamates, have suggested that ancestral state
370 reconstructions of morphological characters must be supported by additional kinds of biological
371 support such as anatomical, physiological, and ecological evidence, to be persuasive (Griffith et
372 al., 2015). Here we explored the evolution of parity mode and placental invasiveness in
373 amniotes utilizing comparative gene expression data. While our study also relies on statistical
374 methods to infer ancestral (gene expression) states, this approach is orthogonal to traditional
375 methods that infer ancestral states from morphological characters among extant species.
376 Indeed, gene expression ultimately underlies the development, evolution, and function of
377 anatomical systems. Thus, by reconstructing the evolution of entire transcriptomes we may be
378 able to infer function of ancestral cell, tissue, and organ systems. Our results resolve several
379 evolutionary transformations during the origins of pregnancy, including the early evolution of an
380 epitheliochorial-like placenta in the mammalian stem-lineage, a hemochorial placenta in the
381 ancestor of eutherians, multiple reversions to non-invasive epitheliochorial placentas within
382 some eutherian lineages, convergent evolution of gene expression profiles among species with
383 different ontogenetic origins of epitheliochorial placentas, and maternal control of placental
384 invasiveness.

13
385

386 Materials and Methods

387 Endometrial Gene Expression Profiling

388 We previously published a dataset of uterine endometrial transcriptomes, which is also

389 used in this study. Interested readers are referred to Mika, Marinic, and Lynch (2021) for
390 specific details. Briefly, we searched the NCBI BioSample, Short Read Archive (SRA), and
391 Gene Expression Omnibus (GEO) databases using the search terms “uterus”, “endometrium”,
392 “decidua”, “oviduct”, and “shell gland”. These anatomical terms refer to the glandular portion of
393 the female reproductive tract, which is specialized for maternal-fetal interactions or shell
394 formation. We then manually curated transcriptomes and excluded those that did not indicate
395 whether tissue samples were from pregnant or gravid tissues and datasets composed of pooled
396 tissues. Gene expression data were analyzed with Kallisto (Bray et al., 2016) version 0.42.4 to
397 pseudo-align the raw RNA-Seq reads to reference transcriptomes and to generate transcript
398 abundance estimates (see Figure 2–source data 1 for accession numbers and reference
399 genome assemblies); Kallisto was run using default parameters, bias correction, and 100
400 bootstrap replicates.

401 Gene expression phylogeny and ancestral transcriptome reconstruction

402 We used the binary encoded endometrial transcriptome dataset for phylogenetic
403 analyses and to reconstruct ancestral gene expression states. Gene expression phylogenies
404 were inferred with IQ-TREE2 (Nguyen et al., 2015) using the best-fitting model of character
405 evolution determined by ModelFinder (Kalyaanamoorthy et al., 2017). The best fitting model
406 was inferred to be the General Time Reversible model for binary data (GTR2), with character
407 state frequencies optimized by maximum-likelihood (FO), and a FreeRate model of among site
408 rate heterogeneity with three categories (R3) (Soubrier et al., 2012). The rate at which
409 characters evolve may vary over time, with the same character evolving rapidly or slowly in
410 different lineages. This phenomenon, known as heterotachy, can bias phylogenetic trees using
411 models of evolution that assume the rates of character evolution are constant such as the
412 GTR+FO+R3 model. Therefore we compared the GTR2+FO+R3 model to the General
413 Heterogeneous evolution On a Single Topology (GHOST) model; The GHOST model
414 accommodates heterotachy by combining features of mixed substitution rate models (Foster
415 2004; Lartillot and Philippe 2004; Pagel and Meade 2004), whereby each class of characters

14
416 has its own substitution rate matrix, and mixed branch length models (Kolaczkowski and
417 Thornton 2004; Meade and Pagel 2008), whereby each class of characters has its own set of
418 branch lengths.

419 We found the best-fit GHOST model was GTR2+FO*H4 (AICc = 215681.71), which
420 accommodated rate heterotachy with 4 categories, described the binary encoded gene
421 expression dataset better than the standard GTR2+FO+R3 model (AICc = 220930.65)
422 indicating that there is extensive variation in the rate of character evolution over time. However,
423 while there was a significant likelihood difference between GTR2+FO+R3 and GTR2+FO*H4
424 (AICc difference = 5248.94), as well as other heterotachy with models with less than or more
425 than 4 categories, the topology of the trees was the same. Thus, while accommodating
426 heterotachy improves estimation of parameters of the substitution model it had no affect on the
427 tree topology and we therefore use the computationally simpler GTR2+FO+R3 model for
428 downstream analyses; This is particularly important for non-parametric topology tests and
429 ancestral state reconstructions (discussed below), which as currently implemented in IQ-TREE2
430 cannot accommodate heterotachy models with unlinked model parameters.

431 Ancestral gene expression states for each gene were inferred using the empirical
432 Bayesian method implemented in IQ-TREE2, the GTR2+FO+R3 model of character evolution,
433 and the species phylogeny as a constraint tree (Figure 3B). Branch support was assessed
434 using the standard (StdBoot) and ultrafast (UFBoot) bootstraps, which assess the effects of
435 sampling bias on branch support (Hoang et al., 2018; Minh et al., 2020). We also used several
436 single branch tests, including the SH-like aLRT and the parametric aLRT (Anisimova et al.,
437 2006; Guindon et al., 2010), aBayes (Anisimova et al., 2011), and the local (LBoot) bootstrap
438 tests (Minh et al., 2020); single branch tests assess whether a branch provides a significant
439 likelihood improvement compared to a null hypothesis that collapses the branch to a polytomy
440 but leaves the rest of the tree topology unaltered. We considered a clade to be highly-supported
441 if its StdBoot support ≥ 80%, UFboot ≥ 95%, SH-aLRT ≥ 80%, aBayes ≥ 0.90, parametric aLRT
442 ≥ 0.95, and LBoot ≥ 90% (Anisimova et al., 2011).

443 The bootstrap and single branch tests assess the robustness of individual branch
444 bipartitions and cannot directly compare complex alternate tree topologies. Therefore we used
445 non-parametric topology tests to directly compare the inferred ML tree to alternative trees with
446 the correct phylogenetic placement of platypus, armadillo, dog, marsupials (opossum and
447 wallaby), and bat, as well as the correct species phylogeny (Figure 2B); tests included the BP-

15
448 RELL, KH-test (Kishino et al., 1990; Kishino and Hasegawa, 1989), SH-test (Anisimova et al.,
449 2011; Guindon et al., 2010), c-ELW (Strimmer and Rambaut, 2002), weighted KH- and SH-
450 tests, and the AU-test (Shimodaira, 2002). We note that the KH-test compares two a priori
451 defined trees rather than the ML and alternative trees (Goldman et al., 2000) and does not
452 correct for multiple hypothesis tests, it is included solely for comparison to other methods and
453 previous studies. The SH-test can be used to compare the ML tree to multiple alternative trees
454 selected a priori (i.e., is dataset independent) and corrects for multiple hypothesis tests, but is
455 too conservative when many trees are tested. The AU-test, in contrast, resolves the
456 conservative nature of the SH-test and thus is the preferred test. All tests performed 100,000
457 resamplings using the RELL method.

458 Clustering methods

459 We evaluated multiple methods to summarize and visualize the binary encoded extant
460 and ancestral reconstructed transcriptomes, including: 1) Logistic Principal Component Analysis
461 (LPCA), a version of principal component analysis for dimensionality reduction of binary data
462 (https://cran.r-project.org/web/packages/logisticPCA/vignettes/logisticPCA.html); 2) classical
463 Multi-Dimensional Scaling (MDS); 3) Uniform Manifold Approximation and Projection (UMAP);
464 4) tSNE; and 5) Fuzzy C-Means (FCM) clustering. All clustering analyses were conducted in R
465 after removing columns (genes) with missing data (coded as ?/NA) or that were invariant (all 0
466 or all 1). LPCA was performed using the logisticPCA R package (Landgraf and Lee, 2015),
467 which implements three methods: exponential family PCA applied to Bernoulli data, logisitic
468 PCA, and the convex relaxation of logistic PCA. For each of the methods, we fit the parameters
469 assuming two-dimensional representation, returning four principal components (ks=4), and
470 selecting the best m value to approximate the saturated model for cross validation. MDS was
471 performed using the vegan R package (Oksanen et al., 2008) with four reduced dimensions.
472 UMAP was performed using the umap R package. tSNE was performed using the Rtsne R
473 package.

474 To explore the data in greater detail we focused on FCM clustering because the results
475 were qualitatively similar to the other methods, and it has several desirable properties including
476 providing a statistically sound way to identify clusters rather than an ad hoc approach that might
477 be applied to the other methods. FCM also allows each sample to have membership in multiple
478 clusters and is conceptually similar to topic ("grade of membership") models used in population
479 genetics to visualize private and shared genetic structure across populations. FCM membership

16
480 coefficients can thereby account for multiple sources of similarity including noise, phylogenetic
481 signal, and convergence of gene expression. FCM was performed in R using the cluster R
482 package, using Manhattan distances (cluster membership was not altered by using other
483 distance metrics), and an estimated fuzzifier (m=1.034978). FCM clustering requires a priori
484 knowledge of the number of clusters (K) to include, therefore we evaluated FCM with K=2–9
485 following the suggestions given in https://www.r-bloggers.com/2019/01/10-tips-for-choosing-the-
486 optimal-number-of-clusters/. First, we used the “elbow” method, in which the sum of squares of
487 each cluster number is calculated and graphed and the optimal number of clusters estimated by
488 a change of slope from steep to shallow (the elbow). We also assessed the optimal number of
489 clusters using the clustree R package, which assess the optimal number of clusters by
490 considering how samples change groupings as the number of clusters increases; clustree is
491 useful for estimating which clusters are distinct and which are unstable but cannot determine the
492 optimal number of clusters (K).

493 CIBERSORT analyses

494 We inferred the proportion of different cell-types in the endometrium during pregnancy
495 across our comparative gene expression datasets using CIBERSORT (Newman et al., 2019),
496 which takes as input a file of gene expression levels from a mixed cell population and a gene
497 expression signature file with expression levels of marker genes in specific cell-types.
498 CIBERSORT was run on the bulk RNA-Seq data from each species using a signature gene
499 expression file based on the Vento-Tormo et al. scRNA-Seq dataset. The signature gene
500 expression file included gene expression data (TPM-like) for genes with an expression level
501 greater than or equal to the expression threshold that also have at least five-fold higher
502 expression levels in a particular cell compared to all other cells (Jain and Tuteja, 2021).

503 Acknowledgements

504 The authors would like to thank MB Thompson (University of Sydney) for constructive
505 comments on this manuscript. This study was supported by a grant from the March of Dimes
506 (March of Dimes Chicago-Northwestern-Duke Prematurity Research Center) and a Burroughs
507 Welcome Fund Preterm Birth Initiative grant (1013760) to principal investigator VJL. The
508 funders had no role in study design, data collection and analysis, decision to publish, or
509 preparation of the manuscript.

510

17
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536 References

537

538 Abzhanov A. 2015. Evolution of the Animal Face: from Principles to Mechanisms. Faseb J 29.
539 doi:10.1096/fasebj.29.1_supplement.867.1

18
540 Amundson R. 2005. The Changing Role of the Embryo in Evolutionary Thought: Roots of Evo-
541 Devo. Cambridge University Press.

542 Anisimova M, Gascuel O, Sullivan J. 2006. Approximate likelihood-ratio test for branches: A
543 fast, accurate, and powerful alternative. Systematic Biol 55:539–552.
544 doi:10.1080/10635150600755453

545 Anisimova M, Gil M, Dufayard J-F, Dessimoz C, Gascuel O. 2011. Survey of branch support
546 methods demonstrates accuracy, power, and robustness of fast likelihood-based
547 approximation schemes. Systematic Biol 60:685–699. doi:10.1093/sysbio/syr041

548 Balmas E, Rana BM, Hamilton R, Shreeve N, Kieckbusch J, Aye I, Hawkes D, Trotter S, López-
549 Tello J, Yong HE, Valenti S, Sferruzi-Perri A, Gaccioli F, McKenzie AN, Colucci F. 2018.
550 Maternal group 2 innate lymphoid cells control fetal growth and protect from endotoxin-
551 induced abortion in mice. bioRxiv.

552 Bauernfeind AL, Zintel TM, Pizzollo J, Ely JJ, Raghanti MA, Hopkins WD, Hof PR, Sherwood
553 CC, Babbitt CC. 2021. Tempo and mode of gene expression evolution in the brain across
554 Primates. Biorxiv 2021.04.21.440670. doi:10.1101/2021.04.21.440670

555 Bergsten J. 2005. A review of long‐branch attraction. Cladistics 21:163–193.

556 doi:10.1111/j.1096-0031.2005.00059.x

557 Billington WD. 1965. THE INVASIVENESS OF TRANSPLANTED MOUSE TROPHOBLAST

558 AND THE INFLUENCE OF IMMUNOLOGICAL FACTORS. Reproduction 10:343–352.
559 doi:10.1530/jrf.0.0100343

560 Blackburn DG. 2015. Evolution of vertebrate viviparity and specializations for fetal nutrition: A
561 quantitative and qualitative analysis. J Morphol 276:961–990. doi:10.1002/jmor.20272

562 Blackburn DG. 1993. Histology of the late‐stage placentae in the matrotrophic skink Chalcides
563 chalcides (Lacertilia; Scincidae). J Morphol 216:179–195. doi:10.1002/jmor.1052160206

564 Blackburn DG, Brandley M. 2015. Viviparous placentotrophy in reptiles and the parent-offspring
565 conflict. J Exp Zoology Part B Mol Dev Evol 324:532–548. doi:10.1002/jez.b.22624

566 Blackburn DG, Callard IP. 1997. Morphogenesis of placental membranes in the viviparous,
567 placentotrophic lizard Chalcides chalcides (Squamata: Scincidae). J Morphol 232:35–55.
568 doi:10.1002/(sici)1097-4687(199704)232:1<35::aid-jmor2>3.0.co;2-n

569 Blackburn DG, Starck JM. 2015. Morphological specializations for fetal maintenance in
570 viviparous vertebrates: An introduction and historical retrospective. J Morphol 276:E1–E16.
571 doi:10.1002/jmor.20410

572 Boddy AM, Abegglen LM, Pessier AP, Schiffman JD, Maley CC, Witte C. 2020. {Lifetime cancer
573 prevalence and life history traits in mammals}. Evol Medicine Public Heal.
574 doi:10.1093/emph/eoaa015

19
575 Brawand D, Soumillon M, Necsulea A, Julien P, Csardi G, Harrigan P, Weier M, Liechti A,
576 Aximu-Petri A, Kircher M, Albert FW, Zeller U, Khaitovich P, Grutzner F, Bergmann S,
577 Nielsen R, Paabo S, Kaessmann H. 2011. The evolution of gene expression levels in
578 mammalian organs. Nature 478:343–348. doi:10.1038/nature10532

579 Brigandt I, Love AC. 2010. Evolutionary Novelty and the Evo-Devo Synthesis: Field Notes. Evol
580 Biol 37:93–99. doi:10.1007/s11692-010-9083-6

581 Carter A M, Enders AC. 2004. Comparative aspects of trophoplast development and
582 placentation. Reprod Biol Endocrin 2:46.

583 Carter Anthony M, Enders AC. 2004. Comparative aspects of trophoblast development and
584 placentation. Reprod Biol Endocrin 2:46. doi:10.1186/1477-7827-2-46

585 Castoe TA, Koning APJ de, Kim H-M, Gu W, Noonan BP, Naylor G, Jiang ZJ, Parkinson CL,
586 Pollock DD. 2009. Evidence for an ancient adaptive episode of convergent molecular
587 evolution. Proc National Acad Sci 106:8986–8991. doi:10.1073/pnas.0900233106

588 Chanderbali AS, Berger BA, Howarth DG, Soltis PS, Soltis DE. 2016. Evolving Ideas on the
589 Origin and Evolution of Flowers: New Perspectives in the Genomic Era. Genetics 202:1255–
590 1265. doi:10.1534/genetics.115.182964

591 Chavan AR, Wagner GP. 2016. The fetal-maternal interface of the nine-banded armadillo:
592 endothelial cells of maternal sinus are partially replaced by trophoblast. Zoological Lett 2:11.
593 doi:10.1186/s40851-016-0048-1

594 Chen C-F, Foley J, Tang P-C, Li A, Jiang TX, Wu P, Widelitz RB, Chuong CM. 2015.
595 Development, Regeneration, and Evolution of Feathers. Annu Rev Anim Biosci 3:1–27.
596 doi:10.1146/annurev-animal-022513-114127

597 Christin P-A, Petitpierre B, Salamin N, Büchi L, Besnard G. 2009. Evolution of C4

598 Phosphoenolpyruvate Carboxykinase in Grasses, from Genotype to Phenotype. Mol Biol
599 Evol 26:357–365. doi:10.1093/molbev/msn255

600 Church SH, Munro C, Dunn CW, Extavour CG. 2022. Identifying evolutionary divergence in
601 gene expression across species and organs: a case-study using Hawaiian Drosophila.
602 Biorxiv 2021.11.30.470652. doi:10.1101/2021.11.30.470652

603 Corpa JM. 2006. Ectopic pregnancy in animals and humans. Reproduction 131:631–640.
604 doi:10.1530/rep.1.00606

605 Corso G, Delitala GM, Carcupino M. 2000. Uterine morphology during the annual cycle in
606 Chalcides ocellatus tiligugu (Gmelin) (Squamata: Scincidae). J Morphol 243:153–165.
607 doi:10.1002/(sici)1097-4687(200002)243:2<153::aid-jmor4>3.0.co;2-q

608 Costanzo V, Bardelli A, Siena S, Abrignani S. 2018. Exploring the links between cancer and
609 placenta development. Open Biol 8:180081. doi:10.1098/rsob.180081

20
610 Cui Y, Wang W, Dong N, Lou J, Srinivasan DK, Cheng W, Huang X, Liu M, Fang C, Peng J,
611 Chen S, Wu S, Liu Z, Dong L, Zhou Y, Wu Q. 2012. Role of corin in trophoblast invasion and
612 uterine spiral artery remodelling in pregnancy. Nature 484:246–250.
613 doi:10.1038/nature10897

614 Elliot MG, Crespi BJ. 2009. Phylogenetic Evidence for Early Hemochorial Placentation in
615 Eutheria. Placenta 30:949–967. doi:10.1016/j.placenta.2009.08.004

616 Elliot MG, Crespi BJ. 2006. Placental Invasiveness Mediates the Evolution of Hybrid Inviability
617 in Mammals. Am Nat 168:114–120. doi:10.1086/505162

618 Felsenstein J. 1978. Cases in which Parsimony or Compatibility Methods will be Positively
619 Misleading. Systematic Biol 27:401–410. doi:10.1093/sysbio/27.4.401

620 Ferretti C, Bruni L, Dangles-Marie V, Pecking AP, Bellet D. 2006. Molecular circuits shared by
621 placental and cancer cells, and their implications in the proliferative, invasive and migratory
622 capacities of trophoblasts. Hum Reprod Update 13:121–141. doi:10.1093/humupd/dml048

623 Figuet E, Ballenghien M, Romiguier J, Galtier N. 2014. Biased Gene Conversion and GC-
624 Content Evolution in the Coding Sequences of Reptiles and Vertebrates. Genome Biol Evol
625 7:240–250. doi:10.1093/gbe/evu277

626 Freyer C, Renfree MB. 2009. The mammalian yolk sac placenta. J Exp Zoology Part B Mol Dev
627 Evol 312B:545–554. doi:10.1002/jez.b.21239

628 Galtier N. 2001. Maximum-Likelihood Phylogenetic Analysis Under a Covarion-like Model. Mol
629 Biol Evol 18:866–873. doi:10.1093/oxfordjournals.molbev.a003868

630 Gilad Y, Mizrahi-Man O. 2015. A reanalysis of mouse ENCODE comparative gene expression
631 data. F1000research 4:121. doi:10.12688/f1000research.6536.1

632 Goldman N, Anderson JP, Rodrigo AG, Olmstead R. 2000. Likelihood-based tests of topologies
633 in phylogenetics. Systematic Biol 49:652–670. doi:10.1080/106351500750049752

634 Graham CH, Lala PK. 1991. Mechanism of control of trophoblast invasion in situ. J Cell Physiol
635 148:228–234. doi:10.1002/jcp.1041480207

636 Griffith OW, Blackburn DG, Brandley MC, Dyke JUV, Whittington CM, Thompson MB. 2015.
637 Ancestral state reconstructions require biological evidence to test evolutionary hypotheses: A
638 case study examining the evolution of reproductive mode in squamate reptiles. J Exp
639 Zoology Part B Mol Dev Evol 324:493–503. doi:10.1002/jez.b.22614

640 Griffith OW, Dyke JUV, Thompson MB. 2013. No implantation in an extra-uterine pregnancy of a
641 placentotrophic reptile. Placenta 34:510–511. doi:10.1016/j.placenta.2013.03.002

642 Gu X. 2004. Statistical Framework for Phylogenomic Analysis of Gene Family Expression
643 Profiles. Genetics 167:531–542. doi:10.1534/genetics.167.1.531

21
644 Guerrero RF, Hahn MW. 2018. Quantifying the risk of hemiplasy in phylogenetic inference. Proc
645 National Acad Sci 115:12787–12792. doi:10.1073/pnas.1811268115

646 Guindon S, Dufayard J-F, Lefort V, Anisimova M, Hordijk W, Gascuel O. 2010. New algorithms
647 and methods to estimate maximum-likelihood phylogenies: assessing the performance of
648 PhyML 3.0. Systematic Biol 59:307–321. doi:10.1093/sysbio/syq010

649 Gurchot C. 1975. The Trophoblast Theory of Cancer (John Beard, 1857–1924) Revisited.
650 Oncology 31:310–333. doi:10.1159/000225037

651 Haim-Vilmovsky L, Henriksson J, Walker JA, Miao Z, Natan E, Kar G, Clare S, Barlow JL,
652 Charidemou E, Mamanova L, Chen X, Proserpio V, Pramanik J, Woodhouse S, Protasio AV,
653 Efremova M, Griffin JL, Berriman M, Dougan G, Fisher J, Marioni JC, McKenzie AN,
654 Teichmann SA. 2020. Rora regulates activated T helper cells during inflammation. Biorxiv
655 709998. doi:10.1101/709998

656 Hibbins MS, Gibson MJ, Hahn MW. 2020. Determining the probability of hemiplasy in the
657 presence of incomplete lineage sorting and introgression. Elife 9:e63753.
658 doi:10.7554/elife.63753

659 Hill JP. 1932. II. Croonian lecture. - the developmental history of the primates. Philosophical
660 Transactions Royal Soc Lond Ser B Contain Pap Biological Character 221:45–178.
661 doi:10.1098/rstb.1932.0002

662 Hirasawa T, Kuratani S. 2015. Evolution of the vertebrate skeleton: morphology, embryology,
663 and development. Zoological Lett 1:2. doi:10.1186/s40851-014-0007-7

664 Hoang DT, Chernomor O, Haeseler A von, Minh BQ, Vinh LS. 2018. UFBoot2: Improving the
665 Ultrafast Bootstrap Approximation. Mol Biol Evol 35:518–522. doi:10.1093/molbev/msx281

666 Hughes RL, Hall LS. 1998. Early development and embryology of the platypus. Philosophical
667 Transactions Royal Soc Lond Ser B Biological Sci 353:1101–1114.
668 doi:10.1098/rstb.1998.0269

669 Huhn O, Ivarsson MA, Gardner L, Hollinshead M, Stinchcombe JC, Chen P, Shreeve N,
670 Chazara O, Farrell LE, Theorell J, Ghadially H, Parham P, Griffiths G, Horowitz A, Moffett A,
671 Sharkey AM, Colucci F. 2020. Distinctive phenotypes and functions of innate lymphoid cells
672 in human decidua during early pregnancy. Nat Commun 11:381. doi:10.1038/s41467-019-
673 14123-z

674 Jain A, Tuteja G. 2021. PlacentaCellEnrich: A tool to characterize gene sets using placenta cell-
675 specific gene enrichment analysis. Placenta 103:164–171.
676 doi:10.1016/j.placenta.2020.10.029

677 Jollie WP. 1961. The incidence of experimentally produced abdominal implantations in the rat.
678 Anatomical Rec 141:159–167. doi:10.1002/ar.1091410209

22
679 Kalyaanamoorthy S, Minh BQ, Wong TKF, Haeseler A von, Jermiin LS. 2017. ModelFinder: fast
680 model selection for accurate phylogenetic estimates. Nat Methods 14:587–589.
681 doi:10.1038/nmeth.4285

682 Katelyn M, Mirna M, Manvendra S, Joanne M, Joris B Jan, J L Vincent. 2021. Evolutionary
683 transcriptomics implicates new genes and pathways in human pregnancy and adverse
684 pregnancy outcomes. Elife 10:e69584. doi:10.7554/elife.69584

685 Kihlström JE. 1972. Period of gestation and body weight in some placental mammals. Comp
686 Biochem Physiology Part Physiology 43:673–679. doi:10.1016/0300-9629(72)90254-x

687 Kin K, Nnamani MC, Lynch VJ, Michaelides E, Wagner GP. 2015. Cell-type Phylogenetics and
688 the Origin of Endometrial Stromal Cells. Cell reports.

689 Kishino H, Hasegawa M. 1989. Evaluation of the maximum likelihood estimate of the
690 evolutionary tree topologies from DNA sequence data, and the branching order in
691 hominoidea. J Mol Evol 29:170–179. doi:10.1007/bf02100115

692 Kishino H, Miyata T, Hasegawa M. 1990. Maximum likelihood inference of protein phylogeny
693 and the origin of chloroplasts. J Mol Evol 31:151–160. doi:10.1007/bf02109483

694 Kolaczkowski B, Thornton JW. 2008. A Mixed Branch Length Model of Heterotachy Improves
695 Phylogenetic Accuracy. Mol Biol Evol 25:1054–1066. doi:10.1093/molbev/msn042

696 Kostka D, Hubisz MJ, Siepel A, Pollard KS. 2012. The role of GC-biased gene conversion in
697 shaping the fastest evolving regions of the human genome. Mol Biol Evol 29:1047–1057.
698 doi:10.1093/molbev/msr279

699 Kozlov AP. 2022. Mammalian tumor-like organs. 1. The role of tumor-like normal organs and
700 atypical tumor organs in the evolution of development (carcino-evo-devo). Infect Agents
701 Cancer 17:2. doi:10.1186/s13027-021-00412-0

702 Kshitiz, Afzal J, Maziarz JD, Hamidzadeh A, Liang C, Erkenbrack EM, Kim HN, Haeger J-D,
703 Pfarrer C, Hoang T, Ott T, Spencer T, Pavličev M, Antczak DF, Levchenko A, Wagner GP.
704 2019. Evolution of placental invasion and cancer metastasis are causally linked. Nat Ecol
705 Evol 3:1743–1753. doi:10.1038/s41559-019-1046-4

706 Lala PK, Nandi P, Hadi A, Halari C. 2021. A crossroad between placental and tumor biology:
707 What have we learnt? Placenta 116:12–30. doi:10.1016/j.placenta.2021.03.003

708 Landgraf AJ, Lee Y. 2015. Dimensionality Reduction for Binary Data through the Projection of
709 Natural Parameters. arXiv.org stat.ML.

710 Lartillot N. 2013. Phylogenetic Patterns of GC-Biased Gene Conversion in Placental Mammals
711 and the Evolutionary Dynamics of Recombination Landscapes. Mol Biol Evol 30:489–502.
712 doi:10.1093/molbev/mss239

23
713 Lewis PO, Olmstead R. 2001. A Likelihood Approach to Estimating Phylogeny from Discrete
714 Morphological Character Data. Systematic Biol 50:913–925.
715 doi:10.1080/106351501753462876

716 Li G, Wang J, Rossiter SJ, Jones G, Cotton JA, Zhang S. 2008. The hearing gene Prestin
717 reunites echolocating bats. Proc National Acad Sci 105:13959–13964.
718 doi:10.1073/pnas.0802097105

719 Li Y, Liu Z, Shi P, Zhang J. 2010. The hearing gene Prestin unites echolocating bats and
720 whales. Curr Biol 20:R55–R56. doi:10.1016/j.cub.2009.11.042

721 Liang C, Musser JM, Cloutier A, Prum RO, Wagner GP. 2018. Pervasive Correlated Evolution in
722 Gene Expression Shapes Cell and Tissue Type Transcriptomes. Genome Biol Evol 10:538–
723 552. doi:10.1093/gbe/evy016

724 Liu Y, Cotton JA, Shen B, Han X, Rossiter SJ, Zhang S. 2010. Convergent sequence evolution
725 between echolocating bats and dolphins. Curr Biol 20:R53–R54.
726 doi:10.1016/j.cub.2009.11.058

727 Loeb L. 1914. The experimental production of an early stage of extrauterine pregnancy. P Soc
728 Exp Biol Med 11:103–106. doi:10.3181/00379727-11-64

729 Luckett P. 1976. Cladistic Relationships among Primate Higher Categories: Evidence of the
730 Fetal Membranes and Placenta. Folia Primatol 25:245–276. doi:10.1159/000155719

731 Luckett WP. 1975. Phylogeny of the Primates, A Multidisciplinary Approach 157–182.
732 doi:10.1007/978-1-4684-2166-8_7

733 Luckett WP. 1974. Comparative development and evolution of the placenta in primates. Contrib
734 Primatol 3:142–234.

735 Lynch VJ. 2022. Evolution: Stress fans the flames of innovation. Curr Biol 32:R158–R160.
736 doi:10.1016/j.cub.2022.01.030

737 Lynch VJ, Nnamani MC, Kapusta A, Brayer K, Plaza SL, Mazur EC, Emera D, Sheikh SZ,
738 Grutzner F, Bauersachs S. 2015. Ancient transposable elements transformed the uterine
739 regulatory landscape and transcriptome during the evolution of mammalian pregnancy. Cell
740 reports 10:551–561.

741 Lyson TR, Bever GS. 2020. Origin and Evolution of the Turtle Body Plan. Annu Rev Ecol Evol
742 Syst 51:1–24. doi:10.1146/annurev-ecolsys-110218-024746

743 Lyu M, Suzuki H, Kang L, Gaspal F, Zhou W, Goc J, Zhou L, Zhou J, Zhang W, Bank JLC, Shen
744 Z, Fox JG, Sockolow RE, Laufer TM, Fan Y, Eberl G, Withers DR, Sonnenberg GF. 2022.
745 ILC3s select for RORγt+ Tregs and establish tolerance to intestinal microbiota. Biorxiv
746 2022.04.25.489463. doi:10.1101/2022.04.25.489463

747 Maddison WP, Knowles LL. 2006. Inferring Phylogeny Despite Incomplete Lineage Sorting.
748 Systematic Biol 55:21–30. doi:10.1080/10635150500354928

24
749 Maddison WP, Midford PE, Otto SP. 2007. Estimating a Binary Character’s Effect on Speciation
750 and Extinction. Systematic Biol 56:701–710. doi:10.1080/10635150701607033

751 Manzo G. 2019. Similarities Between Embryo Development and Cancer Process Suggest New
752 Strategies for Research and Therapy of Tumors: A New Point of View. Frontiers Cell Dev
753 Biology 7:20. doi:10.3389/fcell.2019.00020

754 Marinić M, Mika K, Chigurupati S, Lynch VJ. 2021. Evolutionary transcriptomics implicates
755 HAND2 in the origins of implantation and regulation of gestation length. Elife 10:e61257.
756 doi:10.7554/elife.61257

757 Martin RD. 2008. Evolution of Placentation in Primates: Implications of Mammalian Phylogeny.
758 Evol Biol 35:125–145. doi:10.1007/s11692-008-9016-9

759 Martin RD. 1969. The evolution of reproductive mechanisms in primates. J Reprod Fertil Suppl
760 49–66.

761 McGowen MR, Erez O, Romero R, Wildman DE. 2014. The evolution of embryo implantation.
762 Int J Dev Biol 58:155–161. doi:10.1387/ijdb.140020dw

763 Mendes J, Areia AL, Rodrigues-Santos P, Santos-Rosa M, Mota-Pinto A. 2020. Innate

764 Lymphoid Cells in Human Pregnancy. Front Immunol 11:551707.
765 doi:10.3389/fimmu.2020.551707

766 Mendes J, Rodrigues-Santos P, Areia AL, Almeida J-S, Alves V, Santos-Rosa M, Mota-Pinto A.
767 2021. Type 2 and type 3 innate lymphoid cells at the maternal-fetal interface: implications in
768 preterm birth. Bmc Immunol 22:28. doi:10.1186/s12865-021-00423-x

769 Mess A, Carter AM. 2006. Evolutionary transformations of fetal membrane characters in
770 Eutheria with special reference to Afrotheria. J Exp Zoology Part B Mol Dev Evol 306B:140–
771 163. doi:10.1002/jez.b.21079

772 Mika K, Marinić M, Singh M, Lynch VJ. 2021. Evolutionary transcriptomics implicates new genes
773 and pathways in human pregnancy and adverse pregnancy outcomes. Biorxiv
774 2021.04.19.440514. doi:10.1101/2021.04.19.440514

775 Miller D, Motomura K, Garcia-Flores V, Romero R, Gomez-Lopez N. 2018. Innate Lymphoid

776 Cells in the Maternal and Fetal Compartments. Front Immunol 9:2396.
777 doi:10.3389/fimmu.2018.02396

778 Minh BQ, Schmidt HA, Chernomor O, Schrempf D, Woodhams MD, von HA. 2020. IQ-TREE 2:
779 New Models and Efficient Methods for Phylogenetic Inference in the Genomic Era. Mol Biol
780 Evol 37:1530–1534. doi:10.1093/molbev/msaa015

781 Minot CS. 1891. A Theory of the Structure of the Placenta. Anatomischer Anzeiger Bd 6:125–
782 131.

783 Mohammadi S, Yang L, Harpak A, Herrera-Álvarez S, Rodríguez-Ordoñez M del P, Peng J,

784 Zhang K, Storz JF, Dobler S, Crawford AJ, Andolfatto P. 2021. Concerted evolution reveals

25
785 co-adapted amino acid substitutions in Na+K+-ATPase of frogs that prey on toxic toads. Curr
786 Biol 31:2530-2538.e10. doi:10.1016/j.cub.2021.03.089

787 Mossman 1898- Harland W. (Harland Winfield). 1937. Comparative morphogenesis of the fetal
788 membranes and accessory uterine structures, by Harland W. Mossman ... With twenty-four
789 plates and twelve text-figures. Washington.

790 Munro C, Zapata F, Howison M, Siebert S, Dunn CW. 2021. Evolution of gene expression
791 across species and specialized zooids in Siphonophora. Biorxiv 2021.07.30.454354.
792 doi:10.1101/2021.07.30.454354

793 Murray M, Lessey B. 1999. Embryo Implantation and Tumor Metastasis: Common Pathways of
794 Invasion and Angiogenesis. Semin Reprod Endocr 17:275–290. doi:10.1055/s-2007-
795 1016235

796 Necsulea A, Soumillon M, Warnefors M, Liechti A, Daish T, Zeller U, Baker JC, Grutzner F,
797 Kaessmann H. 2014. The evolution of lncRNA repertoires and expression patterns in
798 tetrapods. Nature 505:635–640.

799 Newman AM, Steen CB, Liu CL, Gentles AJ, Chaudhuri AA, Scherer F, Khodadoust MS,
800 Esfahani MS, Luca BA, Steiner D, Diehn M, Alizadeh AA. 2019. Determining cell type
801 abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol 37:773–
802 782. doi:10.1038/s41587-019-0114-2

803 Nguyen L-T, Schmidt HA, Haeseler A von, Minh BQ. 2015. IQ-TREE: A Fast and Effective
804 Stochastic Algorithm for Estimating Maximum-Likelihood Phylogenies. Mol Biol Evol 32:268–
805 274. doi:10.1093/molbev/msu300

806 Nordor AV, Nehar-Belaid D, Richon S, Klatzmann D, Bellet D, Dangles-Marie V, Fournier T,

807 Aryee MJ. 2017. The early pregnancy placenta foreshadows DNA methylation alterations of
808 solid tumors. Epigenetics 12:793–803. doi:10.1080/15592294.2017.1342912

809 Oksanen J, Kindt R, Legendre P, O’Hara B, Simpson GL, Solymos P, Stevens MHH, Wagner H.
810 2008. vegan: Community Ecology Package.

811 Pagel M. 1999. The Maximum Likelihood Approach to Reconstructing Ancestral Character
812 States of Discrete Characters on Phylogenies. Systematic Biol 48:612–622.
813 doi:10.1080/106351599260184

814 Pal S, Oliver B, Przytycka TM. 2021. Modeling gene expression evolution with EvoGeneX
815 uncovers differences in evolution of species, organs and sexes. Biorxiv 2020.01.06.895615.
816 doi:10.1101/2020.01.06.895615

817 Perry JK, Lins RJ, Lobie PE, Mitchell MD. 2009. Regulation of invasive growth: similar
818 epigenetic mechanisms underpin tumour progression and implantation in human pregnancy.
819 Clin Sci 118:451–457. doi:10.1042/cs20090503

820 Philippe H, Zhou Y, Brinkmann H, Rodrigue N, Delsuc F. 2005. Heterotachy and long-branch
821 attraction in phylogenetics. Bmc Evol Biol 5:50. doi:10.1186/1471-2148-5-50

26
822 Piechowski J. 2019. Plausibility of trophoblastic-like regulation of cancer tissue. Cancer
823 Management Res 11:5033–5046. doi:10.2147/cmar.s190932

824 Portmann A. 1938. Die Ontogenese der Säugetiere als Evolutionsproblem. S. Karger.

825 Price PD, Droguett DHP, Taylor JA, Kim DW, Place ES, Rogers TF, Mank JE, Cooney CR,
826 Wright AE. 2022. Detecting signatures of selection on gene expression. Nat Ecol Evol 1–11.
827 doi:10.1038/s41559-022-01761-8

828 Renfree M, Shaw G. 2013. eLS. doi:10.1038/npg.els.0001856

829 Renfree MB. 2010. Review: Marsupials: Placental Mammals with a Difference. Placenta
830 31:S21–S26. doi:10.1016/j.placenta.2009.12.023

831 Romiguier J, Ranwez V, Delsuc F, Galtier N, Douzery EJP. 2013. Less Is More in Mammalian
832 Phylogenomics: AT-Rich Genes Minimize Tree Conflicts and Unravel the Root of Placental
833 Mammals. Mol Biol Evol 30:2134–2144. doi:10.1093/molbev/mst116

834 Ross CA. 2014. The Trophoblast Model of Cancer. Nutrition Cancer 67:61–67.
835 doi:10.1080/01635581.2014.956257

836 Samuel CA, Perry JS. 1972. The ultrastructure of pig trophoblast rransplanted to an ectopic site
837 in the uterine wall. J Anat 113:139–49.

838 Shimodaira H. 2002. An Approximately Unbiased Test of Phylogenetic Tree Selection.

839 Systematic Biol 51:492–508. doi:10.1080/10635150290069913

840 Soubrier J, Steel M, Lee MSY, Sarkissian CD, Guindon S, Ho SYW, Cooper A. 2012. The
841 Influence of Rate Heterogeneity among Sites on the Time Dependence of Molecular Rates.
842 Mol Biol Evol 29:3345–3358. doi:10.1093/molbev/mss140

843 Strimmer K, Rambaut A. 2002. Inferring confidence sets of possibly misspecified gene trees.
844 Proc Royal Soc Lond Ser B Biological Sci 269:137–142. doi:10.1098/rspb.2001.1862

845 Swanson P, Skinner MK. 2018. Encyclopedia of Reproduction 1–2. doi:10.1016/b978-0-12-

846 811899-3.20528-4

847 Teeling EC. 2009. Hear, hear: the convergent evolution of echolocation in bats? Trends Ecol
848 Evol 24:351–354. doi:10.1016/j.tree.2009.02.012

849 Thompson A, May MR, Moore BR, Kopp A. 2020. A hierarchical Bayesian mixture model for
850 inferring the expression state of genes in transcriptomes. P Natl Acad Sci Usa 117:19339–
851 19346. doi:10.1073/pnas.1919748117

852 Tung P-Y, Blischak JD, Hsiao CJ, Knowles DA, Burnett JE, Pritchard JK, Gilad Y. 2017. Batch
853 effects and the effective design of single-cell gene expression studies. Sci Rep-uk 7:39921.
854 doi:10.1038/srep39921

27
855 Turner W. 1876. Some general observations on the placenta, with
856 special reference to the theory of evolution. J Anat Physiol 33–53.

857 Vento-Tormo R, Efremova M, Botting RA, Turco MY, Vento-Tormo M, Meyer KB, Park J-E,
858 Stephenson E, Polański K, Goncalves A, Gardner L, Holmqvist S, Henriksson J, Zou A,
859 Sharkey AM, Millar B, Innes B, Wood L, Wilbrey-Clark A, Payne RP, Ivarsson MA, Lisgo S,
860 Filby A, Rowitch DH, Bulmer JN, Wright GJ, Stubbington MJT, Haniffa M, Moffett A,
861 Teichmann SA. 2018. Single-cell reconstruction of the early maternal-fetal interface in
862 humans. Nature 563:347–353.

863 Vogel P. 2005. The current molecular phylogeny of Eutherian mammals challenges previous
864 interpretations of placental evolution. Placenta 26:591–596.
865 doi:10.1016/j.placenta.2004.11.005

866 Wagner GP. 2001. What is the promise of developmental evolution? Part II: A causal
867 explanation of evolutionary innovations may be impossible. Journal of Experimental Zoology.

868 Wagner GP. 2000. What is the promise of developmental evolution. Part I. Why is
869 developmental biology necessary to explain evolutionary innovations? J Exp Zool (Mol Dev
870 Evol) 288:95–98.

871 Wagner GP, Kshitiz, Dighe A, Levchenko A. 2021. The Coevolution of Placentation and Cancer.
872 Annu Rev Anim Biosci 10:1–21. doi:10.1146/annurev-animal-020420-031544

873 Wagner GP, Kshitiz K, Levchenko A. 2020. Comments on Boddy et al. 2020: Available data
874 suggest positive relationship between placental invasion and malignancy. Evol Medicine
875 Public Heal 2020:211–214. doi:10.1093/emph/eoaa024

876 Wagner GP, Larsson HCE. 2003. What is the promise of developmental evolution? III. The
877 crucible of developmental evolution. J Exp Zoology Part B Mol Dev Evol 300B:1–4.
878 doi:10.1002/jez.b.41

879 Wildman DE, Chen C, Erez O, Grossman LI, Goodman M, Romero R. 2006. Evolution of the
880 mammalian placenta revealed by phylogenetic analysis. P Natl Acad Sci Usa 103:3203–
881 3208. doi:10.1073/pnas.0511344103

882 Wislocki GB. 1929. On the placentation of primates, with a consideration of the phylogeny of the
883 placenta. Carnegie Institution.

884 Xu Y, Romero R, Miller D, Silva P, Panaitescu B, Theis KR, Arif A, Hassan SS, Gomez‐Lopez
885 N. 2018. Innate lymphoid cells at the human maternal‐fetal interface in spontaneous
886 preterm labor. Am J Reprod Immunol 79:e12820. doi:10.1111/aji.12820

887 Yang J, Ruan H, Zou Y, Su Z, Gu X. 2020. Ancestral transcriptome inference based on RNA-
888 Seq and ChIP-seq data. Methods 176:99–105. doi:10.1016/j.ymeth.2018.11.010

889 Yang Z. 1994. Maximum likelihood phylogenetic estimation from DNA sequences with variable
890 rates over sites: Approximate methods. J Mol Evol 39:306–314. doi:10.1007/bf00160154

28
891 Yang Z. 1993. Maximum-likelihood estimation of phylogeny from DNA sequences when
892 substitution rates differ over sites. Mol Biol Evol 10:1396–401.
893 doi:10.1093/oxfordjournals.molbev.a040082

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908 Figure 1. Evolution and arrangement of the maternal-fetal interface and degree of
909 placental invasion in viviparous Amniotes.

910 (A) Phylogenetic relationships among amniote lineages, mammalian lineages labeled. Major
911 evolutionary steps in the evolution of pregnancy are shown for Mammalia, Theria, and
912 Eutheria.
913 (B) Placenta classification based on the arraignment of the maternal-fetal interface and
914 degree of invasiveness. Epitheliochorial placenta, in which placental invasion does not
915 occur and the barrier between maternal blood and the chorion (tan cells) consists of
916 maternal vascular endothelium (red cells) and uterine epithelium (blue cells).

29
917 Endotheliochorial placenta, in which the placenta invades through the uterine epithelium
918 and the barrier between maternal blood and the chorion consists of the maternal
919 vascular endothelium. Hemochorial placenta, in which the placenta invades through both
920 the uterine epithelium and vascular endothelium, and maternal blood directly bathes the
921 chorionic villi. Examples of species with each type of placenta are also shown. While the
922 egg passes through the glandular uterine portion of the oviduct in oviparous species, the
923 immunologically inert shell (black) prevents direct contact between maternal and fetal
924 tissues.

925 Figure 2. Binary encoding uncovers phylogenetic signal in transcriptome data.

926 (A) Upper, Principal Component Analysis (PCA) of gene expression levels (TPMs) grouped
927 species randomly, consistent with significant noise. Lower, scree plot showing percent of
928 variance explained by dimensions 1-10. Note that the flat curve of variance explained
929 indicates there are no significant dimensions to the PCA.
930 (B) Upper, logistic PCA of the binary encoded endometrial transcriptome dataset groups
931 species by phylogenetic relatedness, indicating significant noise reduction in the binary
932 encoded dataset. Lower, scree plot showing percent of variance explained by
933 dimensions 1-10. Note that the "elbow" of the graph is around dimension 4, suggesting
934 that dimensions 1-4 of the PCA are significant.

935 Figure 2–Source data file 1. Species names (common and binomial), genome annotations
936 used for RNA-Seq analysis and parity mode.

937 Figure 2–Source data file 2. Binary encoded endometrial transcriptome dataset.

938 Figure 3. Maximum-likelihood (ML) phylogeny of binary encoded endometrial

939 transcriptome data.

940 (A) ML phylogeny of binary encoded endometrial transcriptome data inferred by IQ-
941 TREE under the GTR2+FO+R3 model. Highly-supported branch splits, i.e., those
942 with SH-aLRT ≥ 80%, LBoot ≥ 90%, aBayes ≥ 0.90, UFboot ≥ 95%, StdBoot support
943 ≥ 80%, and parametric aLRT ≥ 0.95, are shown with black circles. Branch splits that
944 are highly-supported by at least 4 of 6 methods are shown with gray circles.
945 Particularly discordant phylogenetic relationships are shown in bold. Oviparous
946 species are in black, viviparous species with epitheliochorial placentas in blue,
947 endotheliochorial placentas in orange, and hemochorial placentas in red. Lizard

30
948 species are shown with genus names in italics.
949 (B) Phylogenetic relationships of species in (A). Discordant phylogenetic relationships
950 compared to the ML tree are shown in bold.
951 (C) Tree topology tests comparing the maximum likelihood tree (GTR2+FO+H4) shown
952 in panel A to alternate trees that correct the phylogenetic placement of specific
953 species or the species tree shown in panel B. Delta Log L: log L difference from the
954 maximal Log L in the set (A). bp-RELL: bootstrap proportion using RELL method. p-
955 KH: p-value of one sided Kishino-Hasegawa test. p-SH: p-value of Shimodaira-
956 Hasegawa test. p-WKH: p-value of weighted KH test. p-WSH: p-value of weighted
957 SH test. c-ELW: Expected Likelihood Weight. p-AU: p-value of approximately
958 unbiased (AU) test. + indicates tree is within the 95% confidence sets, – significant
959 exclusion of tree from the 95% confidence sets. All tests performed 100000
960 resamplings using the RELL method.

961

962

963

964

965

966

967 Figure 3–figure supplement 1. ML phylogeny of binary encoded endometrial transcriptome

968 data inferred by IQ-TREE under the best-fitting GTR2+FO+R3 model. Branch supports are
969 shown for SH-aLRT (%) / LBoot (%) / aBayes / UFBoot (%) / StdBoot (%) / parametric aLRT.
970 Clades are considered highly-supported if its SH-aLRT ≥ 80%, LBoot ≥ 90%, aBayes ≥ 0.90,
971 UFboot ≥ 95%, StdBoot support ≥ 80%, and parametric aLRT ≥ 0.95. Particularly discordant
972 phylogenetic relationships are shown in bold. Oviparous species are in black, viviparous species
973 with epitheliochorial placentas in blue, endotheliochorial placentas in orange, and hemochorial
974 placentas in red. Lizard species are shown with genus names in italics.

975 Figure 4. Fuzzy C-Means clustering of extant and ancestral transcriptomes.

976 (A) Species phylogeny with extant species and ancestral lineages colored according to

31
977 Fuzzy C-Means cluster membership (K=2–6) shown in (B). Note that ancestral
978 nodes are numbered. Lizard species are shown with genus names in italics.
979 (B) Fuzzy C-Means clustering of extant and ancestral transcriptomes (K=2–6). Extant
980 species and ancestral lineages colored according to maximum degree of cluster
981 membership shown in (C). Lizard species are shown with genus names in italics.
982 (C) Fuzzy C-Means cluster membership of extant and ancestral transcriptomes (K=2–6).
983 Degree of cluster membership is shown as a 100% stacked bar and colored
984 according to proportion of membership in each cluster. Lizard species are shown
985 with genus names.

986 Figure 4–figure supplement 1. Exploratory clustering of extant and ancestral binary encoded
987 transcriptomes, including: 1) Exponential Family Principal Component Analysis (PCA); 2)
988 Logistic Principal Component Analysis (LPCA); 3) Uniform Manifold Approximation and
989 Projection (UMAP); 4) t-distributed stochastic neighbor embedding (t-SNE); and 5) Hierarchical
990 clustering (Manhattan distances).

991 Figure 4–figure supplement 2. Exploratory analyses of Fuzzy C-Means cluster number.

992 (A) FCM cluster number estimate using the elbow plot method with K=1–20, which was
993 uninformative in choosing the optimal number of clusters.
994 (B) Exploratory analyses of Fuzzy C-Means cluster number using the clustree method with
995 K=1–11. In this figure the size of each node corresponds to the number of samples in
996 each cluster, and the arrows colored according to the number of samples in each
997 cluster. Transparent arrows (the incoming node proportion), show how samples from
998 move between lower and higher cluster numbers and can be used as an indicator of
999 cluster instability. In this graph samples switch clusters between K=1–5, are stable
1000 between K=6–9, and unstable with more than 9 clusters. These data indicate that 6
1001 clusters is optimal for resolution, i.e., more than 6 clusters no longer provides
1002 informative cluster memberships, while more than K≥9 is over-clustered.

1003 Figure 5. Convergent loss of RORA expression in species with epitheliochorial placentas.

1004 (A) Ancestral reconstruction of RORA expression. Pie charts at each internal node
1005 indicate the Bayesian Posterior Probability (BPP) that RORA is expressed (state 1,
1006 black) or not expressed (state 0, gray) at that node. Extant species are colored
1007 according to RORA expression (TPM≥2 = expressed, black; TPM<2 = not

32
1008 expressed, gray); note that the RORA is not annotated in the dunnart genome and is
1009 therefore indicated as “?”. Degree of placental invasiveness is shown for extant
1010 species: Hemochorial (red), endotheliochorial (orange), epitheliochorial (blue),
1011 oviparous (black). Inference of an endometrial population of ILC2/3 cells from
1012 deconvolution of bulk RNA-Seq datasets from each species is shown for extant
1013 species: Species with ILC2/3 cells and RORA expression (blue dots), species
1014 without ILC2/3 cells and RORA expression (yellow dots), and species without ILC2/3
1015 cells and without RORA expression (white).
1016 (B) RORA expression in single-cell RNA-Seq data from cell-types at the first trimester
1017 human maternal-fetal interface. Expression data is shown as a violin plot of log
1018 transformed counts for each cell-type. Cell-types are color coded by their location at
1019 the maternal-fetal interface: Blood (grey), decidua (pink), placenta (embryonic
1020 chorion, green).

1021 Figure 5 – source data 1. Nexus file of the binary encoded endometrial transcriptome dataset
1022 with gene names and species phylogeny.

1023

1024

33
 Placental Barrier
A B
Maternal Side Fetal Side
Eutheria
Maternal Recognition Of Pregnancy Hemochorial
Maternal-fetal Immunotolerance
Maternal-fetal Communication
Interrupted Estrous Cycle
Decidualization
Implantation

Theria Endotheliochorial
Placenta Elaboration
Viviparity
Mammalia
Yolk-sac Placenta
Egg-retention
Matrotrophy
Marsupials
Epitheliochorial

Monotremata

Oviparous

300 250 200 150 100 50 0

Millions of Years Ago (MYA)
Endometrial Epithelium
A PCA (TPMs) B PCA (Binary Encoded)
30 Wallaby Lampropholis
20 Wallaby
Armadillo Opossum
20 Pseudemoia
Chalcides
10 Lerista Platypus
Horse Human SaiphosVivip Chicken
Pig SaiphosOvip Horse
10 Bat Platypus Dunnart
Dog Chalcides
Pig Cow
PC2

PC2
0
Goat
Baboon Squirrel Hamster SaiphosVivip
0 Opossum Cow Dunnart Rabbit Dog
Lerista HumanBaboon
-10 Squirrel Armadillo
Hamster
-10 Pseudemoia
Chicken SaiphosOvip
Rabbit
-20
-20 Goat
Lampropholis Bat
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20
PC1 PC1

Scree plot Scree plot

Percentage of explained variances

Percentage of explained variances

20 20

16%
15 15

11.6%
10 10 8.6%
7.8% 7.3%
6.5% 6.9%
5.8% 5.5% 6.1% 5.8% 5.6%
5.3% 5.2% 5% 4.8% 4.8% 5% 4.8%
5 5 4.3%

0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Dimensions Dimensions
A Human Human B
Baboon Baboon
Rabbit Rabbit
Armadillo Mouse
Squirrel Hamster
Hamster Squirrel
Mouse Dog
Bat Horse
Dog Cow
Cow Goat
Goat Pig
Pig Bat
Horse Armadillo
Wallaby Opossum
Opossum Wallaby
Dunnart Dunnart
Platypus Platypus
Chicken Chicken
Chalcides Chalcides
Lerista Lerista
Saiphos Ovip Saiphos Ovip
Saiphos Vivip SaiphosVivip
Lampropholis Lampropholis
Pseudemoia Pseudemoia
0.06 0 50 100 150 200 250 300
Millions of Years Ago (MYA)

C Tree Log L ΔLog L bp-RELL p-KH p-SH p-WKH p-WSH c-ELW p-AU
Maximum Likelihood (A) -118348.54 0.00 1.00 + 1.00 + 1.00 + 1.00 + 1.00 + 1.00 + 1.00 +

Correct Dog -118395.01 46.47 0.003 - 0.004 - 0.22 + 0.004 - 0.016 - 0.004 - 0.003 -

Correct Bat -118449.02 100.47 0- 0- 0.008 - 0- 0- 2.5e -21 - 4.9e -56 -

Correct Armadillo -118514.13 165.59 0- 0- 1e -05 - 0- 0- 4.8e -30 - 1.4e -66 -

Correct Marsupials -118568.37 219.83 0- 0- 0- 0- 0- 5. e -50 - 4.3e -07 -

Species Tree (B) -118978.23 629.69 0- 0- 0- 0- 0- 1.1e -196 - 1.4e -130 -

 Platypus
Dunnart
98/98/1/98/94/1
100/99/1/85/72/1 Wallaby
97/93/1/82/66/1 Opossum
100/98/1/99/94/1 Horse
99/99/1/100/100/1
Goat
64/70/1/70/48/1 Cow
100/100/1/97/89/1
Pig
100/1/1/100/97/1 Dog
100/100/1/100/100/1
Bat
100/100/1/100/100/1 Mouse
99/1/1/100/97/1
Hamster
71/73/11/53/41/1
Squirrel
21/48/0.51/46/31/0.6
33/47/0.77/28/22/1 Armadillo
Rabbit
82/76/1/50/34/1
100/100/1/80/71/1
Baboon
Human
Chalcides
100/100/1/100/100/1
Lerista
93/89/1/93/85/1
100/99/1/99/98/1 Saiphos Ovip
100/100/1/100/100/1 Saiphos Vivip
98/98/1/98/94/1 100/100/1/100/100/1 Lampropholis
Pseudemoia
Chicken
SH-aLRT (%) / LBoot (%) / aBayes / UFBoot (%) / StdBoot (%) / parametric aLRT
 32 Human 44

A 31 Baboon B Bat
C 45
46
48
26
34 Rabbit 47

K=2 35
36
Squirrel
Hamster
AncMammalia
AncTheria
42
43
30 AncEutheria
Mouse 30
37
38 Bat 40
41
39 Dog 38
35 39
AncEutheria 37 Horse 34 36 31
34
Pig Hamster Armadillo 43 35
40 36
AncEutheria Wallaby 32
41 Cow 31 Mouse 37 AncTheria Mouse
AncTheria 30 Wallaby
Goat Squirrel Squirrel
Armadillo 40 38 SaiphosVivip
32 Baboon Cow 41 AncMammalia SaiphosOvip
Rabbit
AncMammalia Opossum Goat Pig Pseudemoia
42 Rabbit Dog Horse
42 Platypus
43 Dunnart Human
Pig
39 Opossum Opossum
AncAmniota Wallaby Lerista
Dunnart Lampropholis
Platypus Horse
26 SaiphosVivipPlatypus Chalcides Hamster
Goat
Chicken SaiphosOvip44 Chicken Dunnart
44 Chalcides 48 Lerista 46 45 Dog
Cow
Pseudemoia Chicken
45 47 Lampropholis Chalcides
AncAmniota Bat
46 Pseudemoia Baboon
Armadillo
Lerista Lampropholis Human
48 AncAmniote
26 Saiphos Ovip
47 0.0 0.2 0.4 0.6 0.8 1.0
Saiphos Vivip

32 Human 44
45
31 Baboon 46
Bat 48
26
Rabbit 47
34
K=3 35
36
Squirrel
Hamster
AncMammalia
AncTheria
42
43
30 AncEutheria
Mouse 30
37
38 Bat 40
41
39 Dog 38
35 39
AncEutheria 37 Horse 34 36 31
34
Pig Hamster Armadillo 43 35
40 AncEutheria
36
32
41 Cow Wallaby
AncTheria 31 Mouse 37 AncTheria Mouse
Wallaby
Goat 30
Squirrel Squirrel
Armadillo 40 38 SaiphosVivip
32 Baboon Cow 41 AncMammalia SaiphosOvip
Rabbit
AncMammalia Opossum Goat Pig Pseudemoia
42 Rabbit Dog Horse
42 Platypus
43 Dunnart Human Opossum
Pig
Opossum
Wallaby 39 Lerista
AncAmniota Dunnart Lampropholis
Platypus Horse
26 SaiphosVivipPlatypus Chalcides Hamster
Goat
Chicken SaiphosOvip44 Chicken Dunnart
44 Chalcides 48 Lerista 46 45 Dog
Cow
Pseudemoia Chicken
45 47 Lampropholis Chalcides
AncAmniota Bat
46 Pseudemoia Baboon
Armadillo
Lerista Lampropholis Human
48 AncAmniote
26 Saiphos Ovip
47 0.0 0.2 0.4 0.6 0.8 1.0
Saiphos Vivip

32 Human 44
45
31 Baboon 46
Bat 48
26
Rabbit
34 47

K=4 35
36
Squirrel
Hamster
AncMammalia
AncTheria
42
43
30 AncEutheria
Mouse 30
37
38 Bat 40
41
39 Dog 38
35 39
AncEutheria 37 Horse 34 36 31
34
Pig Hamster Armadillo 43 35
40 AncEutheria
36
32
41 Cow Wallaby
AncTheria 31 Mouse 37 AncTheria Mouse
Wallaby
Goat 30
Squirrel Squirrel
Armadillo 40 38 SaiphosVivip
32 Baboon Cow 41 AncMammalia SaiphosOvip
Rabbit
AncMammalia Opossum Goat Pig Pseudemoia
42 Rabbit Dog Horse
42 Platypus
43 Dunnart Human Opossum
Pig
Opossum
Wallaby 39 Lerista
AncAmniota Dunnart Lampropholis
Platypus Horse
26 SaiphosVivipPlatypus Chalcides Hamster
Goat
Chicken SaiphosOvip44 Chicken Dunnart
44 Chalcides 48 Lerista 46 45 Dog
Cow
Pseudemoia Chicken
45 47 Lampropholis Chalcides
AncAmniota Bat
46 Pseudemoia Baboon
Armadillo
Lerista Lampropholis Human
48 AncAmniote
26 Saiphos Ovip
47 0.0 0.2 0.4 0.6 0.8 1.0
Saiphos Vivip

32 Human 44
45
31 Baboon 46
Bat 48
26
Rabbit
34 47

K=5 35
36
Squirrel
Hamster
AncMammalia
AncTheria
42
43
30 AncEutheria
Mouse 30
37
38 Bat 40
41
39 Dog 38
35 39
AncEutheria 37 Horse 34 36 31
34
Pig Hamster Armadillo 43 35
40 AncEutheria
36
32
41 Cow Wallaby
AncTheria 31 Mouse 37 AncTheria Mouse
Wallaby
Goat 30
Squirrel Squirrel
Armadillo 40 38 SaiphosVivip
32 Baboon Cow 41 AncMammalia SaiphosOvip
Rabbit
AncMammalia Opossum Goat Pig Pseudemoia
42 Rabbit Dog Horse
42 Platypus
43 Dunnart Human
Pig
39 Opossum Opossum
AncAmniota Wallaby Lerista
Dunnart Lampropholis
Platypus Horse
26 SaiphosVivipPlatypus Chalcides Hamster
Goat
Chicken SaiphosOvip44 Chicken Dunnart
44 Dog
Chalcides 48 Lerista 46 45 Cow
Pseudemoia Chicken
45 47 Lampropholis Chalcides
AncAmniota Bat
46 Pseudemoia Baboon
Armadillo
Lerista Lampropholis Human
48 AncAmniote
26 Saiphos Ovip
47 0.0 0.2 0.4 0.6 0.8 1.0
Saiphos Vivip

32 Human 44
45
31 Baboon 46
Bat 48
26
Rabbit
34 47

K=6 35
36
Squirrel
Hamster
AncMammalia
AncTheria
42
43
30 AncEutheria
Mouse 30
37
38 Bat 40
41
39 Dog 38
35 39
AncEutheria 37 Horse 34 36 31
34
Pig Hamster Armadillo 43 35
40 AncEutheria
36
32
41 Cow Wallaby
AncTheria 31 Mouse 37 AncTheria Mouse
Wallaby
Goat 30
Squirrel Squirrel
Armadillo 40 38 SaiphosVivip
32 Baboon Cow 41 AncMammalia SaiphosOvip
Rabbit
AncMammalia Opossum Goat Pig Pseudemoia
42 Rabbit Dog Horse
42 Platypus
43 Dunnart Human Opossum
Pig
Opossum
Wallaby 39 Lerista
AncAmniota Dunnart Lampropholis
Platypus Horse
26 SaiphosVivipPlatypus Chalcides Hamster
Goat
Chicken SaiphosOvip44 Chicken Dunnart
44 Chalcides 48 Lerista 46 45 Dog
Cow
Pseudemoia Chicken
45 47 Lampropholis Chalcides
AncAmniota Bat
46 Pseudemoia Baboon
Armadillo
Lerista Lampropholis Human
48 AncAmniote
26 Saiphos Ovip
47 0.0 0.2 0.4 0.6 0.8 1.0
Saiphos Vivip
 Exponential Family PCA Logistic PCA

Lerista 35
100
SaiphosOvip 34
SaiphosVivip 36 Bat

Pseudemoia
Lampropholis Platypus 20
50 45 Chicken 31
30 AncTheria
Dunnart 40
Human Squirrel Armadillo AncMammalia
AncAmniote Chalcides Dog Baboon
Rabbit Cow 38
0 Baboon Pig
0 Human Dog Wallaby
PC2

PC2
31 Horse
Pig
Horse Cow
38 Armadillo
Hamster Opossum
40 Dunnart
35
-50
AncTheria Bat
-20 26 Chicken AncAmniote
Opossum Platypus
Lerista
-100 Pseudemoia

-40
Wallaby Lampropholis

-150 -100 -50 0 50 -25 0 25 50

PC1 PC1

UMAP tSNE

42 47
2

Opossum Opossum
42
Wallaby Lampropholis
Pseudemoia
45 46
50

Lampropholis Wallaby
47 Dunna
AncAmniote
1

44 SaiphosVivip
SaiphosOvip
Human
Horse Chalcides
38 Platypus Lerista
Pig Cow 32 Platypus
SaiphosVivip Chicken
UMAP2

tSNE2

SaiphosOvip Dunnart 40 Baboon

0

Chicken
Lerista
0

40 37 Dog 44 Chalcide
AncMammalia
AncTheria
AncTheria AncAmniote
Cow Horse 37 45
30 46
AncMammalia Rabbit Squirrel
30 38 AncEutheria
Rabbit
-1

Armadillo Pig Pseudemoia

AncEutheria Baboon Human 31
Dog
31 Armadillo
-50

32
Bat 34
Squirrel 35
Hamster 34
-2

Bat Hamster
35
-100 -50 0 50 100
-2 -1 0 1 2
UMAP1 tSNE1

Cluster Dendrogram
200
150
Height
100

Bat
50

Dog

Dunnart
Human
Armadillo
Pig

Platypus
Chalcides
0

Rabbit
37

47
31
Squirrel

AncAmniote
Pseudemoia
AncTheria

35
38

32
AncMammalia
40

46
45
34

Lampropholis

44
Chicken
30
AncEutheria

SaiphosOvip
Cow

Opossum
42

Lerista
41

39

33

36

43

48

26
Horse

Hamster

Wallaby
Baboon

SaiphosVivip
A B K
1
1 1
2
700

3
4
2
600

1 2
5
6
7
500

3 1 3 2
8
wss

9
400

10
4 3 4 1 2
11
300

in_prop
5 4 5 2 3 1
0.25
200

0.50

6 6 2 1 3 5 4 0.75

k
5 10 15 20 1.00

Number of Clusters
7 5 3 6 4 7 1 2 size

10

8 4 7 1 3 8 5 2 6
20

30
9 2 7 8 1 4 6 5 9 3

count
10 4 2 3 8 9 1 7 5 10 6

15

10
11 8 10 5 4 6 2 3 9 7 1 11
5
 C ta
A

IL cen
3
Pl RA
B

2/
Blood

O
a
R
Human RORA expression Decidua
BPP Not Expressed (0) Baboon Expressed
Rabbit Not Expressed Placenta

log transformed normalized counts

Squirrel
2
BPP Expressed (1) Hamster Placenta Type/Parity
Mouse Hemochorial
Bat Endotheliochorial
Dog Epitheliochorial
Horse Oviparous
Pig
Cow Endometrial ILC2/3
Goat Present & match RORA expression
Armadillo Abcent & mismatch RORA expresion
Opossum Abcent & match RORA expression
1
? Dunnart
Wallaby
Platypus
Chicken
Chalcides
Lampropholis
Pseudemoia
Lerista
SaiphosOvip 0

19.2 Endo L

1 VCT
9 VCT
20 SCT
13 EVT
16 EVT
5 dNK1
2 dNK2
11 dNK3
24 dNKp
32 ILC2/3
17 NK CD16
31 NK
7 dM1
6 dM2
22 M3
18 HB
23 MO
29 MO
27 DC1
28 DC2
30 Gran
4 TCells
8 TCells
10 TCells
15 TCells
26 Plasma
19.1 Endo (m)

19.3 Endo (f)

25.1 Epi1
25.2 Epi2
14.1 PV1
14.2 PV2
0 DSC1
3 DSC2
21 DSC3
12 F1
33 F2
SaiphosVivip