Author Correction: Transcriptomic changes in the pre-implantation uterus highlight histotrophic nutrition of the developing marsupial embryo

An amendment to this paper has been published and can be accessed via a link at the top of the paper.


Discussion
Our transcriptomic analysis of dunnart uterus reveals differential expression of a range of genes putatively involved in the processes of early pregnancy, prior to implantation of the unshelled conceptus into the lining of the uterus. GO and pathway analyses indicate that there is significant differential regulation of groups of genes involved in metabolism and biosynthesis, and almost one third of the top 50 upregulated genes in pregnancy have these roles (Table 1), an unsurprising result that highlights the importance of these processes in the metabolically active uterus during pregnancy. Our results also point to a role for differential regulation of genes encoding nutrient transporters, cytoskeletal molecules, and immune factors in the uterus to support histotrophy, immunological protection and tissue remodelling required for early development of the embryo. Similar functions have been  identified using transcriptomic studies of species representing independent origins of viviparity, indicating that these processes are critical to maintaining pregnancy across taxa 15,32,34,35 .
Nutrient provisioning to the unimplanted embryo. In marsupials and eutherian mammals, the initial pre-attachment embryonic development is supported by histrotrophes secreted by uterine glands 36 . Following embryonic attachment, nutrient supply typically shifts to haemotrophy (i.e. secretion of material from the maternal blood circulation 4 ). Haemotrophic nutrient transfer either occurs through direct embryonic contact with maternal blood, or through diffusion or active transport of haemotrophes from maternal blood, followed by secretion by the uterine epithelium into the uterine lumen 37 . In marsupials, the shift from histotrophic to haemotrophic nutrient transfer typically occurs following rupture of the embryonic shell coat 38 . In S. crassicaudata, this shift is accompanied by structural changes to the uterus. Early in S. crassicaudata pregnancy (the period at which our pregnant transcriptome samples were collected), uterine stromal glands are abundant and actively secreting 12,24 . As pregnancy progresses, gland abundance decreases and glandular secretion is replaced by secretory activity in the luminal epithelium 12 . We identified a number of genes putatively responsible for nutrient transport to the early conceptus: Histotrophy. Almost one quarter of the top 50 upregulated genes in early S. crassicaudata pregnancy have putative transport-associated function, suggesting that nutrient transport underpins histotrophy in supporting the conceptus pre-implantation (Table 1), even before haemotrophic nutrient transport via the placenta. A number of secretion-related genes upregulated in early pregnancy may be associated with glandular secretion of histotrophe (e.g. AP4S1, HYOU1, SRPRA) ( Table 5). Early pregnancy involves significant upregulation of nutrient transporter genes, including APOL6, involved in cholesterol transport 39 , PLA2G10, involved in hydrolysis of fatty acids during pregnancy 40 , and a suite of solute carrier proteins (SLCs) involved in transport of nucleoside sugars, ions and anions, glucose, fatty acids, calcium and zinc ( Table 5). Upregulation of solute carrier proteins also occurs during pregnancy in the uterus of the viviparous skink Chalcides ocellatus 35,41 and the post-implantation uterus of the marsupial M. domestica 15 . Similarly, cathepsin L (CTSL), upregulated during pregnancy in C. ocellatus 35 and pigs 42,43 , is also significantly upregulated during pregnancy in S. crassicaudata (Table 5). Cathepsins are involved in remodelling of the uterine epithelium, which may enable transport of gases, macromolecules and micronutrients for embryonic development 43 . These molecules are also components of secreted uterine fluid in horses, pigs, sheep and cattle, along with phospholipases 44 . Additionally, cathepsins are present in the mouse and human yolk sac during early pregnancy, where they may degrade proteins to free amino acids for uptake by the fetus 20 , and we suggest that CTSL may play a similar role during early pregnancy in the dunnart uterus.
Macromolecule catabolism. Lysosomal activity is also one of the most significantly upregulated KEGG pathways during pregnancy in S. crassicaudata (Table 3). This result indicates that breakdown of macromolecules into small subunits for uterine secretion 41,45 occurs during the period of receptivity in dunnarts. Such catabolism is probably required during histotrophic nutrition to provide molecules small enough for uptake through the permeable shell coat of the conceptus. Lysosomes and lysosomal-associated genes are also upregulated during pregnancy in the uterine epithelium of both pigs 46 20 . Increased lysosomal activity is consistent with an increased protein content of luminal fluid in the marsupial uterus pre-implantation 24,47 . Lysosomal activity is also congruent with morphological observations of dark electron-dense vesicles in uterine glandular epithelial cells, which become electron-lucent pre-implantation in S. crassicaudata 12,26 . This morphological pattern also occurs during pregnancy in viviparous skinks 45 and pigs 48 . The lysosomal genes upregulated in pre-implantation S. crassicaudata uterus suggests that similar genetic mechanisms mediate nutrient breakdown for histotrophy in diverse viviparous groups.
Adenogenesis. Interestingly, both cadherins and the Wnt signaling pathway, involved in mammalian uterine adenogenesis (gland development, which is essential for histotrophy 49 ), are down-regulated in the pregnant S. crassicaudata uterus (Tables 4, 6). This finding suggests a cessation of gland development in the uterine stroma as pregnancy progresses, which is consistent with a morphological decrease in gland density in the uterine stroma of S. crassicaudata during the period of uterine receptivity 12 . Hence, the shift from histotrophic nutrient transfer may begin prior to implantation to allow a rapid shift to haemotrophic nutrient provisioning upon implantation.
Steroid biosynthesis. The steroid biosynthesis pathway is also significantly enriched in the list of upregulated genes during pregnancy (Table 3). CYP27A1 (sterol 27-hydroxylase P450) is involved in the conversion of cholesterol to its primary metabolite 27-hydroxycholesterol, after which 27-hydroxycholesterol is converted to bile salt precursors by HSD3B7 (3-beta-hydroxysteroid dehydrogenase-7); the conversion of the 5-beta-reduction of bile acid intermediates and steroid hormones carrying a delta (4)-3-one structure is effected by AKR1D1 (aldo-keto reductase family 1 member D1) 50 . All four of these genes are significantly upregulated during pregnancy, especially AKR1DA and HSD3B7, which are in the top 50 differentially expressed annotated genes ( Table 5). While deficiencies in this pathway cause adrenal dysfunction and bile acid reduction 51 , the reasons for their upregulation here is less clear. 27-hydroxycholesterol is a selective modulator of the estrogen receptors 52 , and bile acid intermediates are also nutrient signalling molecules 53 ; both functions may be important in the pre-implantation uterus. Linked with this pathway is the upregulation of steroid biosynthesis pathways ( Table 5). The production of 7-dehydrocholestrol is followed by a sequence of gene expressions culminating in the expression of 17-beta hydroxysteroid 7 (HSD17B7), which is involved in the conversion of steroid precursors to androgens 51 . The upregulation of these pathways may be linked to steroid recruitment mechanisms, but may also be important in other functions during pregnancy, including the transport and utilisation of fatty acids and electrolytes in the pre-attachment phase.
Immunity. The top five most significantly enriched GO categories in pregnancy downregulated genes are related to immune function (Supplementary Table 2), and 18% of the top 50 downregulated genes during pregnancy have putative immune function ( Table 2). Many of these downregulated genes are immunoglobulins that make up subunits of antibodies (Table 6), which may simply reflect a lower relative number of B cells in pregnant uterine tissue. Other genes involved in maternal-fetal tolerance are also downregulated, including IL34 54 . This result reflects an important role of the uterus in immunosuppression to prevent maternal rejection of the  semi-foreign embryo, even before the invasion of the embryo into the uterine epithelium. The dunnart embryonic shell membrane disintegrates prior to implantation, which in combination with remodelling may place maternal and embryonic tissues in close association 3,10 . The apposition of maternal and fetal tissues has likely driven the evolution of adaptations to 'hide' the embryo from the mother's immune system, despite a lack of tissue invasion at that point in pregnancy. A similar downregulation of some immune genes occurs in the uteri of other vertebrates that lack erosion of maternal epithelia throughout pregnancy e.g. 32,35,55 .
In S. crassicaudata, we also observe a large proportion of immune genes upregulated pre-implantation (14% of the top 50, Table 1). In contrast to other marsupial studies, we did not see a change in interleukin-6 gene expression 15,18 , even though interleukin-6 is expressed in other tissues in S. crassicaudata 56 . The differences may be because our study focussed on preimplantation pregnancy. In M. domestica, immune genes are upregulated at implantation, including a range of inflammatory and wound-healing markers 18 . There is increasing recognition of the importance of the presence of maternal immune factors in the eutherian uterus for embryo implantation and uterine remodelling; the maternal immune response must be precisely regulated for successful mammalian pregnancy 57,58 . Our results allow comparison of both major lineages of marsupials, Australididelphia (S. crassicaudata, here) and Didelphimorphia (M. domestica 15,18 ), and suggest that a delicate balance of up-and down-regulated immune factors was a feature of the pregnant uterus of the most recent common ancestor of therian mammals, exapted for the evolution of viviparity in this lineage. Immune genes of stable expression in M. domestica 18 across pregnancy display the same pattern in S. crassicaudata (CD3D, CD3D, CD3G, CD4, CD68, CD8B, IL4R). Further examination of gene expression at late stage pregnancy in S. crassicaudata is necessary to draw conclusions about the precise immunogenic changes that facilitate implantation and placentation in the dunnart, and whether these mirror the changes seen in the Didelphimorphia. Finally, immune factors prevent pathogenic infection in vertebrate gestational tissues 32,57 , and our dataset identifies several candidate genes responsible for immune defence in the pregnant dunnart uterus (BPI, BPIFB1, GZMA and PRF1) ( Table 5).
Remodelling of the pregnant uterus. Differentially regulated S. crassicaudata genes are significantly enriched for a number of GO categories related to tissue proliferation, tissue remodelling, and cell membrane components (Supplementary Table 1). The cell adhesion molecule pathway is significantly downregulated as identified by KEGG pathway analysis (Table 4), and more than one third of the top 50 downregulated genes have putative functions associated with cytoskeleton and remodelling ( Table 2). Alterations to both cell adhesion and remodelling are expected during the period of receptivity in preparation for implantation, and embryonic implantation in S. crassicaudata involves significant morphological and molecular remodelling 12,24,26 . Our findings demonstrate that, as for eutherian mammals 42,59 and viviparous skinks 35,41,60 , remodelling involves expression changes of cathepsins (CTSL), cadherins (e.g. CDH11, CDH20), and numerous protocadherins (Tables 5 and 6). Similar expression patterns of remodelling genes across diverse viviparous groups suggest a common suite of molecules is required in preparing the uterus for implantation in live-bearing taxa 60 . Down-regulation of cell adhesion molecules occurs in S. crassicaudata, including JAM2, which is associated with tight junctions 61,62 . Embryonic attachment in S. crassicaudata is invasive, yet unlike many eutherian mammal species with invasive placentation, the invasion involves embryonic erosion of an originally intact uterine epithelium, rather than a loss of cellular adhesion to facilitate invasion 12,24 . In viviparous skinks, reduced lateral cell adhesion makes the uterus more plastic and likely facilitates remodelling 63 . Down-regulation of the cell adhesion pathway may play a similar role in preparing the S. crassicaudata uterus for implantation of the embryo.
Several genes that function in angiogenesis and vascular morphogenesis are downregulated in the S. crassicaudata uterus during pregnancy (e. g. ADGRA2, ADGRB2, ANGPTL1, EPHB4, ISM1, PDZRN3 Table 6). This result was unexpected, given the upregulation of angiogenic genes such as EPAS1, HIF1A and VEGFA during pregnancy in skinks and rats e.g. 35,[64][65][66] ; however several of these genes are inhibitors, rather than promoters, of angiogenesis e.g. ISM1 67 . Their downregulation in S. crassicaudata uterus during pregnancy may simply reflect temporality of our sampling: the transcriptome comes from uteri prior to the development of extensive vascularisation during placental formation, and it is possible that embryos do not require much oxygen at this early developmental stage. Extracellular matrix molecules are down-regulated during early pregnancy in S. crassicaudata, including laminin (LAMA3), collagens (COL7A1, COL15A1), fibulin (FBLN7), fibronectins (FLRT2, FLRT3) and receptors (ITGA4), keratins (KRT22), and elastins (EMILIN1) ( Table 6). We suggest that uterine receptivity in S. crassicaudata involves significant remodelling of the extracellular matrix. Increased expression of laminins [68][69][70] , fibronectin 71 and fibronectin receptor ITGA4 72 is associated with uterine receptivity in eutherian mammals. The opposite trend for these molecules in S. crassicaudata is unexpected, yet could be explained by differences in alterations to the uterine stroma in marsupial and eutherian pregnancy. In eutherian mammals, increased expression of extracellular matrix molecules is related to cellular differentiation of uterine stromal fibroblasts to decidual cells (decidualisation) 73,74 . This cellular transformation does not occur in S. crassicaudata, as marsupials lack decidual cells 73 . In addition, the uterine stroma of S. crassicaudata and other marsupials is relatively cell-poor, and uterine receptivity involves a significant reduction in stromal cell abundance 12,27 . Thus, the specific markers of uterine receptivity may differ between viviparous amniotes, as they relate to species-specific uterine cellular processes. Additionally, reduction in extracellular matrix leading up to implantation may help to reduce the diffusion distance between maternal blood vessels and the uterine epithelium. In marsupials, reduction of this diffusion distance is a critical step in preparation for haemotrophic nutrient transfer 37 .
Uterine receptivity and quiescence. A number of genes differentially expressed in the dunnart uterus are similar to mediators of uterine receptivity in humans. Estrogen and progesterone are the key hormones controlling receptivity of the uterus to an implanting embryo 22 , and our data reveal differential expression of genes binding to and effecting action of these hormones (PAQR7; PRDM2) in the dunnart uterus just prior to implantation (Table 5). These hormones coordinate morphological and physiological changes in the uterus to promote receptivity, and a number of potential markers of uterine receptivity in eutherians 22 are differentially expressed in the S. crassicaudata uterus. Mucins, which are apically located glycoproteins in the epithelium of the uterus, have anti-adhesive properties, and must be removed from the site of attachment before implantation can take place; dysregulation of mucin expression affects eutherian fertility 22,75,76 . A similar situation is present in marsupials, given that the mucin MUC5AC is the most highly downregulated gene in pre-implantation dunnart pregnancy ( Table 2), and that MUC1 increases in the grey opossum uterus after breach of the shell coat 18 . Mucins are also downregulated in the uterus during pregnancy in a viviparous skink 34 . A number of other genes involved in uterine receptivity in humans and mice are also differentially expressed in the dunnart pre-implantation uterus, including the homeobox genes HOXA10 and HOXA11, and phospholipases (PLA2G10, PLA2G3) 22,77 .
Maintaining quiescence of the uterus (i.e. preventing uterine contraction) is another key requirement for progress of a successful pregnancy. Two of the most significantly downregulated genes in the pregnant dunnart uterus are the prostaglandin receptors PTGER3 and PTGFR ( Table 2). The products of these genes likely bind prostaglandins to stimulate myometrial contractions 78 . Similarities in early pregnancy between Australididelphia and Didelphimorphia. We identified 97% of the genes that were differentially expressed between non-pregnant and pre-implantation M. domestica uterus 18 in the S. crassicaudata uterine transcriptome. This result indicates a substantial overlap in the range of expressed genes between the two species, as expected given that these species derive from a single origin of viviparity. There are many shared genes that are differentially expressed in M. domestica and S. crassicaudata (at the same stages of pregnancy: non-pregnant uterus compared to pre-implantation uterus) (Supplementary  Tables 3 and 4). The overlap indicates that many of the uterine functions identified in S. crassicaudata are shared across both major marsupial lineages. For example, remodelling of the uterus is a shared characteristic, with genes involved in extracellular matrix (e.g. cadherin-related genes FAT4, CDH11, CDH19 and PCDH11X down in pregnancy; laminin-related genes EGFLAM, COL15A1 down in pregnancy), cellular motility (e.g. FGF1, NRG1, SEMA5B down in pregnancy; RAB25, FGFR1, HBEGF up in pregnancy) and cell adhesion (e.g. ITGA4, PTK7,  TRIP6 up in pregnancy) differentially regulated in both S. crassicaudata and M. domestica. Histotrophic function is also shared across early pregnancy in marsupials: genes involved in lysosomal transport are upregulated in pregnancy in both M. domestica and S. crassicaudata (e.g. ATP6V1B2, AP3D1, TMEM165, TMEM79), and pathway analysis indicates an overrepresentation of pregnancy-upregulated genes of protein processing and export, secretion, and lysosome function in the shared gene lists between the two species (Supplementary Table 7). Of the top 50 genes of M. domestica that are upregulated during pregnancy, 20% are also upregulated in S. crassicaudata early pregnancy. These genes include ELF5 (ESE2), an epithelium-specific transcription factor thought to regulate gene expression in glandular epithelium 79 and which we postulate may be important in supporting gene expression for glandular secretions; CTAGE5, involved in exporting collagen from the endoplasmic reticulum 80 , and therefore possibly important for remodelling of the extracellular matrix; FGFBP1, which mediates cellular proliferation and migration 81 ; and LVRN, which in humans is a trophoblast-specific factor 82 that may regulate molecules at the interface of maternal and embryonic tissue to facilitate the development of a placenta 83 . The expression of LVRN in uterine tissues during early pregnancy in both major marsupial lineages suggests that this molecule may also be involved in initiating placentation at the maternal tissue interface, although further research is required to explore this hypothesis. Of the top 50 M. domestica genes downregulated during early pregnancy, 14% are also downregulated in S. crassicaudata early pregnancy. These genes include transcription factors (CBX2, SOX4); the motor-protein encoding gene KIF26B; VTCN1 (B7-H4), which negatively regulates T-cell immune responses 84 ; and IGFBP5, which regulates the action of the insulin-like growth factors that mediate cell growth and also has apoptotic action 85 . Interestingly, transgenic mice that overexpress IGFBP5 display reduced female fertility 85 , suggesting that the downregulation of this gene may be essential to early pregnancy across mammals.

Conclusions
Genomic and transcriptomic methods are valuable tools for examining the physiology and evolution of marsupial pregnancy 15,17,18,86,87 . While the M. domestica transcriptome identified the importance of immune modulation for successful implantation and placentation in the marsupial uterus 18 , a range of other physiological changes is also required to support the internal incubation of the embryo prior to placentation. Our transcriptome study highlights the importance of such processes, including remodelling of the pre-implantation uterus, uterine quiescence, and nutrient provision via histotrophy prior to the development of the placenta; many of the genes underpinning these functions are shared across the dunnart and the opossum. The S. crassicaudata dataset is an ideal complement to the transcriptome of the opossum 15,18 , because these animals represent both major clades of marsupials (Australididelphia and Didelphimorphia, which diverged ~75 Mya 88 ), and the cladistic derivation of both groups is similar (within-clade divergence of Dasyuridomorphia and Didelphimorphia both ~30 Mya 88 ).
This transcriptome analysis reveals the importance of histotrophic nutrient transport prior to embryo implantation, before nutrient transport function is supplanted by the complex, nutritive placenta. Early pregnancy is a critical time for successful reproduction, and disruption to histotrophy could disrupt embryonic development. 40-50% Figure 1. Venn diagram indicating the differentially expressed genes between opossum pre-implantation pregnant and non-pregnant uterus that are also differentially expressed in dunnart pre-implantation pregnancy. EP = early/pre-implantation pregnancy.
SCienTifiC REPORTS | (2018) 8:2412 | DOI:10.1038/s41598-018-20744-z of human pregnancies fail in the first trimester 21 , most of which is prior to the development of the definitive chorioallantoic placenta 89 . The putative gene functions identified here are similar to those in the pregnant uterus in other amniotes 34,35,90 . The conservation of genes underpinning pre-placental nutrient transport, gestational tissue remodelling, and uterine quiscence in amniote pregnancy is remarkable given that mammals and reptiles represent multiple independent origins of viviparity. Conserved elements underpinning aspects of early eutherian and marsupial pregnancy may provide new information for understanding human pregnancy disorders 91,92 , which is important given the difficulties in studying the human uterus in vivo 22 . This work furthers our understanding of the mechanisms underlying the survival of early embryos in our earliest live bearing mammalian ancestors, and highlights the importance of histotrophic nutrition to the embryo prior to the development of the nutritive placenta.

Methods
Tissue collection. Animals were held at a temperature-controlled breeding colony at the University of Sydney (in accordance with approved University of Sydney Animal Ethics Committee Protocol 704). Animals were housed either singly or in pairs, in plastic cages, and were provided with nesting boxes, nesting material, and enrichment material. Animals were held under the natural photocycle for Sydney (33°52' S, 151°12' E) and fed commercial cat food daily; water was provided ad libitum. Vaginal epithelial cells in smears of the urogenital sinus were examined microscopically to monitor estrous cycling of females 93,94 . A large number of cornified epithelial cells in the urine and a sharp increase in body mass defined the peak of oestrous 93,95,96 . Females were then paired with males, and the first day that sperm were detected in urine of the female was designated day 1 after mating 25,95 . Paired females were monitored for signs of pregnancy, including an increase in pouch area and vascularisation, loss of the furred pouch lining, and increase in body mass 93,96 .
Early pregnant (n = 3) and non-pregnant (n = 3) females were euthanised by CO 2 inhalation, followed by immediate decapitation. The presence of embryos in excised uteri confirmed gestation, and the stage of pregnancy was determined by comparing size and morphology of embryos to the timetable of embryonic development 12 . We specifically targeted early-pregnant animals between days 6-8 of pregnancy, prior to implantation and placentation 12 , the stage of pregnancy where the shelled egg is present in the uterus.
Transcriptome sequencing and annotation. Uterine samples were homogenised using the 3 mm steel bead TissueLyser II system (Qiagen, Hilden Germany) and QiaShredder (Qiagen). Total RNA was extracted using an RNeasy Plus Mini Kit (Qiagen), which includes an in-built DNAse treatment. RNA concentration and integrity were assessed using a Bioanalyzer (Agilent, Santa Clara CA) and only high quality RNA (RIN > 8) was used for downstream analysis. Samples for transcriptomics were sequenced after Truseq RNA sample prep with on an Illumina HiSeq 2500 with 100 bp paired-end sequencing, at the Ramaciotti Centre for Genomics, Sydney, Australia. Reads from all samples were combined in a de novo assembly with Trinity v2.0.4 28 , using the default parameters and the-trimmomatic and-min_kmer_cov 2 options. To assess the assembly completeness we used BUSCO v2.0.1 29 with the default parameters in the transcriptome mode (-m tran), and searched against the tetrapod set of orthologs (tetrapoda_odb9). We used Kallisto 30 to estimate abundance and DESeq2 31 to call differential expression as implemented in the Trinity pipeline. We assessed correlation of gene expression between samples using the PtR script in Trinity. We annotated transcripts and assigned GO terms using the default parameters of the Trinotate pipeline v3.0.2 28 ; which allowed us to identify particular gene functions on which to focus our analyses. Graphical representation of enriched GO terms was carried out using the cateGOrizer tool 97 . KEGG pathway analysis of annotated genes was carried out using DAVID version 6.8 (available: http://david.abcc.ncifcrf. gov/home.jsp, last accessed June 2017) 98 , using EASE score of 0.1 and M. domestica as background. P-values were Benjamini-Hochberg corrected to account for multiple hypothesis testing.
Differentially expressed genes between non-pregnant and pre-implantation uterus in M. domestica were compared to the S. crassicaudata uterine gene expression data using discontiguous megablasts optimised for cross-species comparison, using the -task dc-megablast option and the default parameters. Monodelphis domestica transcripts 18 identified as differentially expressed between non-pregnant and mid-gravid (pre-implantation) uterus (adjusted P < 0.001) were searched against the S. crassicaudata uterine transcriptome assembly, and the results compared to the S. crassicaudata differential gene expression results from DESeq2. Differentially expressed genes shared between the two species were analysed using the DAVID functional annotation tool version 6.8 (available: http://david.abcc.ncifcrf.gov/home.jsp, last accessed November 2017) 33 , with GO_ALL biological process, cellular component and molecular function terms, using M. domestica as background. The Functional Annotation Clustering option was used to group significantly enriched GO terms using a modified Fisher's Exact Test by function and the DAVID Fuzzy clustering algorithm 33 . Grouping was performed using DAVID settings for highest stringency and P-values were Benjamini-Hochberg corrected to account for multiple hypothesis testing. KEGG pathway analysis using DAVID was carried out using an EASE score of 0.1 and Benjamini-Hochberg corrected P-values.
Data availability statement. All sequence data have been uploaded to GenBank (BioProject ID PRJNA399240).