1. Developmental Biology
Download icon

DNA methylation and gene expression changes derived from assisted reproductive technologies can be decreased by reproductive fluids

  1. Sebastian Canovas
  2. Elena Ivanova
  3. Raquel Romar
  4. Soledad García-Martínez
  5. Cristina Soriano-Úbeda
  6. Francisco A García-Vázquez
  7. Heba Saadeh
  8. Simon Andrews
  9. Gavin Kelsey
  10. Pilar Coy  Is a corresponding author
  1. Universidad de Murcia-Campus Mare Nostrum, Spain
  2. Instituto Murciano de Investigación Biosanitaria, Spain
  3. The Babraham Institute, United Kingdom
  4. University of Cambridge, United Kingdom
Research Article
  • Cited 12
  • Views 2,113
  • Annotations
Cite this article as: eLife 2017;6:e23670 doi: 10.7554/eLife.23670

Abstract

The number of children born since the origin of Assisted Reproductive Technologies (ART) exceeds 5 million. The majority seem healthy, but a higher frequency of defects has been reported among ART-conceived infants, suggesting an epigenetic cost. We report the first whole-genome DNA methylation datasets from single pig blastocysts showing differences between in vivo and in vitro produced embryos. Blastocysts were produced in vitro either without (C-IVF) or in the presence of natural reproductive fluids (Natur-IVF). Natur-IVF embryos were of higher quality than C-IVF in terms of cell number and hatching ability. RNA-Seq and DNA methylation analyses showed that Natur-IVF embryos have expression and methylation patterns closer to in vivo blastocysts. Genes involved in reprogramming, imprinting and development were affected by culture, with fewer aberrations in Natur-IVF embryos. Methylation analysis detected methylated changes in C-IVF, but not in Natur-IVF, at genes whose methylation could be critical, such as IGF2R and NNAT.

https://doi.org/10.7554/eLife.23670.001

eLife digest

Infertility has become more common in many countries, particularly those where many people delay having children until later in life. To help individuals experiencing infertility conceive a child, scientists have developed treatments called assisted reproductive technologies (or ARTs for short). So far, more than 5 million children have been born with the help of these treatments. Most of the children seem healthy; however, birth defects are more common in ART-conceived babies than those conceived without treatment.

The cause of these birth defects is not known, though scientists suspect it may have something to do with techniques used in ART. One possible culprit is the liquid that is used in the laboratory to help the parents’ sperm and egg come together for fertilization. This same liquid is also used to bathe the developing embryo for the first few days after fertilization before it is implanted into its mother’s womb. Some scientists wonder whether adding the fluids normally found in the reproductive tract of their mother to this liquid could reduce defects in children conceived via ART.

Now, Canovas et al. have shown that fertilizing and growing pig embryos in liquids supplemented with fluid from the wombs of female pigs results in embryos that are closer to naturally conceived pig embryos than in non-supplemented liquids. In the experiments, naturally conceived embryos were compared to ART embryos exposed to the usual liquids and with ART embryos grown in liquids with fluid collected from the pig’s reproductive tract added. Cutting edge technologies were used to sequence the entire genomes of all of the embryos and compare which genes were active in each case. Canovas et al. also looked at chemical markers on the DNA – called epigenetic changes – that turn on or off the expression of genes without changing the DNA code itself.

The analysis showed that ART-conceived embryos grown in the usual liquid had different patterns of gene expression and epigenetic changes compared to naturally conceived embryos. Gene expression and epigenetic changes in the ART embryos grown with the pig reproductive fluid was more similar to the naturally conceived embryos.

These findings suggest that abnormal gene expression in the ART-liquid exposed embryos may lead to birth defects, and that using natural reproductive fluids may be safer. To confirm this, scientists will have to implant embryos conceived in these three different conditions into mother pigs and assess the health and gene expression patterns of the resulting piglets. If successful, these new insights might one day lead to improvements in ART techniques used to treat infertility in people.

https://doi.org/10.7554/eLife.23670.002

Introduction

“Most fertility researchers are trying to improve Assisted Reproductive Technologies (ARTs) success as measured by a single, clear standard: the birth of an apparently healthy baby. Only a few are trying to discern whether in vitro fertilization (IVF) leaves a subtle legacy in children. What will happen to these kids when they are middle-aged?” (Servick, 2014). In humans, according to a study by the World Health Organisation (WHO) in 190 countries, infertility affects 20% of couples and it was estimated that at least 40.5 million women were seeking infertility medical care in 2007 (Mascarenhas et al., 2012). ARTs provide a helpful alternative for a high proportion of infertility cases and the number of children born to date using these methods exceeds 5 million (International Committee for Monitoring Assisted Reproductive Technology I, 2012). Although the majority of them seem healthy, studies have reported higher rates of preterm births (Rubens et al., 2014), non-chromosomal birth defects and adverse perinatal effects in ART pregnancies (El Hajj and Haaf, 2013), with long-term effects being under study in humans (Kissin et al., 2014). Epidemiological data suggest that perturbed epigenetic gene regulation by the application of ART could be a contributory factor in these adverse outcomes (El Hajj and Haaf, 2013; Whitelaw et al., 2014), although such alterations could also be considered as consequences of parental characteristics, gamete quality or other non-epigenetic technique-derived effects (Simpson, 2014). To clarify the impact of each of these factors, the use of an animal model that avoids, as much as possible, the effect of parental circumstances and the use of protocols minimizing the technique-derived effects would help to attain the goal of offering safer ART for patients.

For modeling ART-related disorders in human, swine could be a good candidate for several reasons: their genetic, anatomical and physiological similarities with human (Swindle et al., 2012); their size and length of gestation; and the availability of individuals genetically selected by their excellent reproductive performance in artificial insemination centres. Importantly, this last trait could be useful to remove the paternal factor (low-quality male gametes) from studies as a possible reason for any epigenetic alterations found. However, most protocols for processing boar spermatozoa for IVF include their selection by density gradient centrifugations and just a few used the swim-up procedure to isolate highly motile spermatozoa which is the routine selection in human infertility clinics. Since it was observed that spermatozoa selected by swim-up show higher rates of normal morphology and motility, and decreased DNA fragmentation and methylation levels (Kim et al., 2015), it would be necessary to adapt the sperm selection protocols in pig before using them to model ART-derived epigenetic alterations.

In both mouse and human, accumulating evidence indicates that the embryo is sensitive to its very early environment and that culture media used in ART (as factors involved in technique-derived effects) may have long-lasting consequences (Kleijkers et al., 2014; Fernandez-Gonzalez et al., 2004). Several imprinting disorders and abnormal phenotypes have been linked to ART, but of special significance is the relationship between the presence of serum in culture media and the incidence of Large Offspring syndrome (LOS) in ruminants (Young et al., 1998), which includes diverse pathologic alterations and shows phenotypic and epigenetic similarities with the imprinted disorder Beckwith-Wiedemann syndrome (BWS) in humans (Chen et al., 2013). Since it was proposed that serum in the culture medium could be a crucial factor in LOS incidence, the tendency in the procedures for both human and livestock was to move toward the use of chemically defined media, limiting the presence of proteins in the culture medium to serum albumin. Although practical, this approach may have unpredictable consequences, because it ignores the fact that the reproductive fluids have a different composition to serum and are extremely rich in proteins other than serum albumin (more than 150 have been described in the oviductal fluid [Avilés et al., 2010]). If these proteins are physiologically present, they must play a variety of roles supporting the normal development of the embryo, roles that serum albumin alone cannot properly provide and serum cannot fully mimic. In addition, although ART in species such as cattle and sheep usually results in foetal overgrowth (Young et al., 2001; Chen et al., 2015), opposing phenotypes such as low birth weights (excluding BWS) are often seen in humans (Schieve et al., 2002) and pigs (García-Vázquez et al., 2010). A study showing the relationship between child birth weight and the protein source in embryo culture media (Zhu et al., 2014) reinforces the hypothesis that the protein composition of the culture media plays a role in the correct regulation of epigenetic marks in the growing embryo. A similar conclusion can be reached from a clinical trial showing that protein enrichment of media compared with addition of serum albumin alone improved the blastocyst implantation rate and may increase human births by more than 8% (Meintjes et al., 2009). Therefore, as with breast milk, which is so complex and so rich in bioactive factors that cannot be easily replaced with any artificial composition (Hennet and Borsig, 2016), the idea that reproductive secretions could be necessary in the culture media should not be underrated. At least, it should be explored under experimental conditions to unveil the relevance of these secretions.

DNA and RNA sequencing have become affordable cutting-edge technologies that could help to understand the mechanisms underlying abnormalities observed in ART-derived offspring. However, so far, single blastocyst whole-genome DNA methylation profiles comparing in vivo and in vitro produced embryos have not been published for any mammalian species and we therefore aim to produce these in this study.

We report here that modified swim-up protocols for the selection of spermatozoa in pigs and the use of reproductive secretions as additives in the culture media significantly increase the yield and quality of the blastocysts produced from a morphological, epigenetic and gene expression point of view. Using genome-wide analyses of gene expression by RNA-Seq and DNA methylation by Bisulfite-Seq in single blastocysts, we provide datasets of pig blastocysts produced in vitro with and without reproductive secretions as additives in the culture medium and show that the former are more similar to the in vivo specimens than the later. This suggests an alternative approach for conceiving healthier ART-derived children.

Results

Swim-up method improves the yield of pig IVF

In order to select spermatozoa before IVF, a swim-up protocol was set up and compared with a conventional selection system by density gradient centrifugations. To do this, it was necessary to design a suitable washing and sperm selection medium imitating, as far as possible, in vivo conditions (NaturARTs PIG sperm washing medium and NaturARTs PIG sperm swim-up medium, EmbryoCloud, Murcia, Spain). The swim-up medium was supplemented either with bovine serum albumin (BSA) (Swim-up BSA group) or porcine oviductal fluid (POF, Swim-up fluid group) collected at the late follicular (LF) phase of the estrous cycle (NaturARTs POF-LF, EmbryoCloud, Murcia, Spain) (Figure 1). All the fluids used in this study were directly aspirated from the lumen of ovarian follicles, oviducts or uterus and processed according to the information described in the Materials and methods section, at http://embryocloud.com, and in previous references (Coy et al., 2008).

Schematic representation of three different sperm processing protocols used for in vitro fertilization.

Swim-up-BSA: NaturARTs PIG medium + BSA; Swim-up-Fluid: NaturARTs PIG medium + POF-LF*. Density gradient centrifugation: centrifugation through a discontinuous Percoll: gradient (45% and 90% v/v). *POF-LF: porcine oviductal fluid collected at the late follicular phase of the estrous cycle. Red box represents the portion of the reproductive tract whose conditions we tried to resemble in vitro. IVF results after using these three different sperm processing protocols are included in Table 1.

https://doi.org/10.7554/eLife.23670.003

Polyspermy after IVF is a major issue in the pig (Coy and Avilés, 2010). With these new protocols, we obtained significantly higher rates of monospermy than with conventional ones (49.6 ± 4.5 vs 17.4 ± 4.1, Table 1) and the final percentage of putative zygotes (evaluated at 24 hours post insemination, hpi) was significantly higher (35.2 ± 0.2 vs 14.6 ± 0.1, Table 1). Moreover, the addition of POF-LF to the Swim-up media instead of BSA increased the final yield of the system (35.2 ± 0.2 vs 29.7 ± 0.2, Table 1).

Table 1

IVF results after using three different sperm processing protocols (Density gradient, Swim-up-BSA and Swim-up-Fluid) as represented in Figure 1. a,b: Different letters in the same column indicate values statistically different (p<0.05). Penetration: proportion of oocytes penetrated by one or more spermatozoa. Monospermy: Monospermy percentage, calculated from penetrated oocytes, represents the proportion of penetrated oocytes with only one spermatozoon inside the ooplasm. Spermatozoa/Oocyte: Mean number of sperm per penetrated oocyte. Spermatozoa/ZP: Mean number of spermatozoa attached to ZP per oocyte. Yield: Percentage of putative zygotes per oocyte.

https://doi.org/10.7554/eLife.23670.004
Sperm processing methodNPenetration (%)Monospermy (%)Spermatozoa/OocyteSpermatozoa/ZPZygote yield (%)
Density gradient centrifugation10584.3 ± 3.6a17.4 ± 4.1a8.4 ± 0.7a17.3 ± 2.3a14.6 ± 0.1a
Swim-up-BSA18069.6 ± 3.5b42.7 ± 4.6ab2.1 ± 0.1b7.2 ± 0.5b29.7 ± 0.2b
Swim-up-Fluid18371.1 ± 3.4b49.6 ± 4.5b2.7 ± 0.1b8.6 ± 0.5b35.2 ± 0.2c

Reproductive fluids added to the culture media increase blastocyst quality

In a second experiment, and using the Swim-up protocol for sperm selection, a new IVF/Embryo culture (EC) system (Natur-IVF) was developed, which included preincubation of oocytes in oviductal fluid (NaturARTs PIG OF-LF) and the presence of reproductive fluids as additives in the IVF and EC media (0–8 hr: NaturARTs POF-LF; 8–48 hr: oviductal fluid from the early luteal–EL- phase of the estrous cycle, NaturARTs POF-EL; 48–180 hr: uterine fluid -UF-from this same phase, NaturARTs PUF-EL) (Figure 2). Corresponding controls with BSA instead of OF/UF for each step (referred as C-IVF group) were analyzed (Figure 2). Evaluation at 24 hpi revealed higher penetration rate (66.6 ± 0.1 vs 43.7 ± 0.1, p<0.05) and similar monospermy rate (78.6 ± 0.1 vs 72.7 ± 0.1, p<0.05) for the Natur-IVF and C-IVF groups, respectively. Regarding embryo development, more than 40% of cleaved embryos reached the blastocyst stage in both groups (Table 2A). However, the Natur-IVF blastocysts showed a significant increase in their mean number of cells (81.8 ± 7.2, Table 2A) compared to the C-IVF ones (49.9 ± 3.7), and this number was similar to that observed in the in vivo samples (In-vivo group, 87.0 ± 7.2). Moreover, at day 7.5, embryos reaching the hatching or hatched stages were observed only in the Natur-IVF group (Table 2B). Taken together, these data indicate a higher quality, in terms of cell number and ability to hatch, in the ART-derived blastocysts when reproductive fluids were added to the culture medium.

Schematic representation of the different steps of the new IVF/EC system.

Swim-up-BSA or Swim-up-Fluid protocols were used for IVF. Previously, oocytes were preincubated in OF-LF for 30 min. Then, each group of putative zygotes were incubated in different media (0–8 hr, 8–48 hr and 48 hr-7days) as indicated in the diagram. O*: ovary with hemorrhagic corpus luteum; O**: early corpus luteum; OF-LF: oviductal fluid-late follicular phase of the estrous cycle; OF-EL: oviductal fluid-early luteal phase of the estrous cycle; UF-EL: uterine fluid-early luteal phase of the estrous cycle. Swim-up-BSA: NaturARTs PIG medium + BSA; Swim-up-Fluid: NaturARTs PIG medium + POF-LF. TALP: culture medium used for IVF. NCSU23: culture medium used for embryo development in vitro supplemented with sodium lactate, pyruvate and non-essential amino acids (NCSU23a) or with glucose and essential and non-essential amino acids (NCSU23b).

https://doi.org/10.7554/eLife.23670.005
Table 2

(A) Comparative results of IVF yield by using BSA (C- IVF) or reproductive fluids (Natur-IVF) as additives in the culture medium for 7.5 days. (B) Results of blastocyst development (for each type) using BSA (C- IVF) or reproductive fluids (Natur-IVF) as additives in the culture medium for 7.5 days. Columns from ‘Early blastocyst’ to ‘Hatched blastocyst’ indicate the percentage of each type of blastocyst from Total blastocyst (Table 2A), classified according to Bo and Mapletoft25. a,b: Different letters in the same column indicate values statistically different (p<0.05). Cleavage: Cleavage percentage from N. Total Blastocysts: Percentage of blastocysts calculated from cleaved embryos. Yield: Percentage of putative blastocysts from N. Cell/blastocyst: mean number of cells per blastocyst.

https://doi.org/10.7554/eLife.23670.006
A)
GroupNPenetration (%)Monospermy (%)Cleavage (%)Total blastocysts (%)Blastocyst Yield (%)Cell/ blastocyst
In vivo41   87.0 ± 7.2b
C- IVF903395
(43.7 ± 0.1a)
656
(72.7 ± 0.1)
429
(47.5 ± 1.6a)
178
(41.4 ± 2.4)
19.6 ± 1.349.9 ± 3.7a
Natur-IVF961640
(66.6 ± 0.1b)
755
(78.6 ± 0.1)
405
(42.1 ± 1.6b)
180
(44.5 ± 2.5)
18.7 ± 1.281.8 ± 7.2b
  
B)
GroupNEarly blastocyst (%)Blastocyts (%)Expanded blastocyst (%)Hatching blastocyst (%)Hatched blastocyst (%)
C- IVF17857
(31.7 ± 6.1)a
50
(28.3 ± 5.9)
71
(40.0 ± 6.4)
0
(0)a
0
(0)a
Natur -IVF18023
(12.8 ± 5.4)b
55
(30.8 ± 7.5)
65
(35.9 ± 7.8)
28
(15.4 ± 5.9)b
9
(5.1 ± 3.6)b

The blastocyst transcriptome can be modulated in vitro by reproductive fluids

In vitro culture systems significantly alter embryonic gene expression as previously observed in pooled pig blastocysts (Bauer et al., 2010). Here, the transcriptomes from three individual day 7.5 blastocysts produced by C-IVF or Natur-IVF were compared with their in vivo counterparts (Figure 3A–B). RNA libraries showed acceptable quality in all nine blastocysts. Mean number of raw reads was 14.24 ± 2.23 (±SD) millions, and transcripts from 13,309 to 14,512 different genes (from a total of 20,789 annotated pig mRNAs) were detected in each individual. Principal Component Analysis (PCA) showed that, despite expected individual variability, the three embryos from each group clustered together (Figure 3B), with the C-IVF embryos showing higher variability, which could represent high embryo plasticity in response to suboptimal culture conditions. Therefore, after combining the triplicates, data from both in vitro groups showed high correlation (Pearson correlation coefficient [r] = 0.964), but Natur-IVF was closer to the In vivo group ([r] = 0.95) than C-IVF ([r] = 0.938). RNA-Seq data analysis (DESeq2 p<0.05 after multiple testing correction) identified 787 differentially expressed genes (DEG) between the C-IVF and In-vivo, and 621 DEGs between Natur-IVF and In vivo (Figure 3—source data 1, including also all the expression values for all the genes). Of the genes that were significantly different (adjusted p-value < 0.05, fold change > 1.5) in the pair-wise comparisons, there was a higher number of up-regulated (534/787–68%- in C-IVF embryos and 431/621–69%- in Natur-IVF) than down-regulated (253 and 190, respectively) (Figure 3C, Figure 3—source data 1).

Gene expressed analysis in blastocysts obtained in vivo, by the Natur-IVF system or by C-IVF system.

(A) Heatmap of global gene expression (with log2 fold change >1.5 and adjusted B-H p-value < 0.05). Numbers denote ID of a specific embryo. (B) Principal Component Analysis (PCA) of the RNA-Seq samples: In vivo embryos (IV, red), Natur-IVF (N, green) and C-IVF (C, blue). Numbers denote ID of specific embryos. (C) Venn diagram with DEGs (Figure 3—source data 1). *, #, § denotes DEGs exclusive for C-IVF, Natur-IVF and In vivo, respectively (Figure 3—source data 2). (D) Heat map of gene expression of key genes associated with embryo development/differentiation, epigenetic reprogramming, cell cycle/cell growth, gene expression and imprinting.

https://doi.org/10.7554/eLife.23670.007

Top Canonical Pathways, Physiological Systems and Molecular and Cellular Functions related to DEGs were identified (summarized in Supplementary file 1) using the Ingenuity Pathway Analysis (IPA) software. Globally, down-regulated genes in C-IVF and in Natur-IVF were linked to similar Top-cellular functions (Supplementary file 1). Equally, top Canonical Pathways affected by up-regulated genes were similar for both groups. In contrast, two pathways were identified in down-regulated DEGs in C-IVF embryos, but not in Natur-IVF DEGs (Supplementary file 1). Increased pathways in Natur-IVF and C-IVF included cholesterol, mevalonate, serine and glycine biosynthesis and p53 signaling. Decreased pathways (protein ubiquitination and 14-3-3 mediated signaling) were detected only in C-IVF. Similarly, Physiological Systems and Functions over-represented by up-regulated or down-regulated DEGs were different between C-IVF or Natur-IVF. These results show that, in spite of similarity, there were differences that could influence specific pathways and affect key molecular and cellular functions in the embryos from each group.

Natur-IVF blastocysts show fewer aberrantly expressed genes than C-IVF blastocysts

Natur-IVF and C-IVF blastocysts shared 334 genes that were aberrantly expressed in both groups vs In vivo (Exclusive DEGs, Figure 3C- Figure 3—source data 2). However, there were 440 genes (from the 784 DEGs in C-IVF) that showed aberrant expression only in C-IVF vs In vivo (DEGs only in C-IVF, Figure 3C), while 40% fewer genes (n = 281 from the 620 DEGs in Natur-IVF) showed aberrant expression only in the Natur-IVF group vs In vivo (DEGs only in Natur-IVF, Figure 3C). Importantly, several genes related to epigenetic reprogramming (down: DNMT3B, DNMT1; up: HDAC5, KDM5A), embryo development (down: CTGF, ING2, KIT, EZH2; up: BMP4, TLN1, ADAR), cell growth (down: CDCA5, SMC1A; up: RB1, SMARCA2) or imprinting (up: IGF2BP2, GNAS; down: DIRAS3) were amongst the C-IVF-specific DEGs (Figure 3D).

Direct comparison between Natur-IVF vs In vivo and C-IVF vs In vivo DEGs revealed that only 29 genes reached significant expression differences between the two in vitro groups after DESeq2 analysis (Figure 3—source data 1). Interestingly, of these 29 DEGs, 13 were similarly expressed in Natur-IVF and In vivo, and only seven showed similar expression between C-IVF and In vivo groups (Figure 3C, Figure 3—source data 2). Although the number of these genes was low, they could be critical because among the 13 genes exclusively different in the C-IVF blastocysts (Figure 3—source data 2), those down-regulated (n = 6) were KIT, MPPA6, MTA3, KIF4A, UBR2 and ISOC1 (Log Fold Change from −5.9 to −54.18). For all six genes data were available for the corresponding knock-out mice or knock-down studies, which showed phenotypes of altered/abnormal growth/size, reproduction/fertility, mortality/aging, hematopoietic system, homeostasis/metabolism and other abnormalities (Supplementary file 2).

These data suggest that in vitro culture significantly alters embryonic gene expression to a lesser extent than previously proposed (Bauer et al., 2010), and a better modulation of the blastocyst transcriptome was achieved by mimicking physiological conditions of fertilization and early embryo development by the addition of reproductive fluids (Natur-IVF).

Genome-wide DNA methylation of the pig blastocyst is affected by the in vitro culture system

In this study, for the first time, whole-genome DNA methylation profiles on individual porcine blastocysts were generated by a low-cell adaptation of the post-bisulphite adaptor-tagging (PBAT) method (Miura et al., 2012; Peat et al., 2014). Three blastocysts from each group were analyzed. The number of unique alignments in the samples ranged from 13,150,508 to 42,208,651 and the coverage of CpGs (≥1 read) from 52% to 59.2%. The global methylation percentages of CpGs were 15.02 ± 3.3, 11.09 ± 2.6 and 12.33 ± 3.6 for the C-IVF, Natur-IVF and In-vivo groups, respectively. The distribution of methylation levels in windows of 150 CpGs across the genome and a general view of the methylation profiles of the nine individual blastocysts are shown in Figure 4A–B. The generally low level of methylation suggests that the genome has experienced substantial loss of methylation from the gametes, analogous to that observed in other mammals (Guo et al., 2014; Kobayashi et al., 2012). The landscape of methylated cytosines suggests some structure across the genome, with regions with more methylation consistent between the individual blastocysts (Figure 4B). What contributes to this structure, for example, the regions of relatively higher methylation, is not immediately obvious, as methylation was similar in different genomic contexts with no marked enrichment in repetitive elements, for example (Table 3). Regarding the different classes of blastocysts, methylation over specific genomic features followed the same tendency as the global differences, with higher values for C-IVF (Table 3).

Distribution of methylation levels and general view of the methylation profiles of 9 individual pig blastocysts.

(A) Distribution of methylation percentages across tiles of 150 CpGs on the pig genome for three groups of blastocysts (In-vivo, C-IVF and Natur-IVF). (B) Random browser shot as example of methylation landscape of the nine individual blastocysts analysed (Chr8:37027152–118458156). The two first rows in the picture represent the genes and CpG islands annotated (Ensembl, RRID:SCR_006773 Sus scrofa 10.2) in the pig genome, respectively. Color scale represents methylation levels from red (highest methylation, up to 25%) to blue (lowest methylation-0%).

https://doi.org/10.7554/eLife.23670.010
Table 3

Percentages of methylation over genome features in porcine blastocysts produced in vitro (C-IVF and Natur-IVF) or collected in vivo (In vivo).

https://doi.org/10.7554/eLife.23670.011
% Methylation
In vivoC-IVFNatur-IVF
CpG islands9.6911.8010.11
Promoters9.2611.619.11
TU12.8415.4712.36
Intergenic11.7514.4811.37
LINE112.6315.4312.02
LTR12.7715.5312.06
SINE12.4515.3011.94
GLOBAL12.3315.0211.09

PCA revealed a good level of clustering for In-vivo and Natur-IVF embryos but not for C-IVF embryos (Figure 5A). In particular, embryos C34 and C36 were far from the other seven embryos analyzed.

DNA-methylation analysis in blastocysts obtained in vivo, by the Natur-IVF system or by C-IVF system.

(A) Principal Component Analysis (PCA) of the DNA methylation samples: In vivo embryos (red), Natur-IVF (green) and C-IVF (blue). Numbers denote ID of specific embryo. (B) Venn diagram of DMRs by pair-wise comparison (adjusted-p <0.05). Number of DMRs with higher (↑) or lower (↓) methylation in each pair-wise comparison are indicated (Figure 5—source data 1). (C) Heatmap of the 417 DMRs between the C-IVF group and the other two groups (In vivo and Natur-IVF). (D) Heatmap of the 324 DMRs between Natur-IVF group and the other two groups (In vivo and C-IVF). (E) Heatmap of the 448 DMRs between the In vivo group and the other two groups (Natur-IVF and C-IVF). For C, D and E (Figure 5—source data 2): Relative methylation measure as the difference in percent of methylation from the median methylation across all samples.

https://doi.org/10.7554/eLife.23670.012

The low level of global methylation suggested that few differentially methylated regions (DMRs) could be found. For this reason, and to obtain an unbiased measure of differences in genome methylation, a fixed size of 150 CpGs was used for analysis, as this was found to give a modal tile size of around 3 kb with about 150 reads per tile for most individuals. To make the data comparable to enable the detection of DMRs, separately from the global changes, the tiles informative in all samples (258,885) were extracted and quantile normalized. To identify DMRs, the comparison was filtered to require a consistent ≥5% absolute methylation change between all replicates of the first and second condition, followed by a T-test (B-H adjusted p<0.05). Differences between the groups were observed with fewer than 4000 DMRs for each pair-wise comparison (Figure 5—source data 1). Globally, fewer DMRs showed higher methylation in In vivo vs Natur-IVF (n = 1,660) than in In vivo vs C-IVF (n = 2244) (Figure 5B).

To better characterize the changes in methylation exclusively affecting one of the groups (p<0.05 for both comparisons), the corresponding subsets of DMRs (‘exclusive’ DMRs for each group) were obtained by combining the previous lists (Figure 5B,C,D and E; Figure 5—source data 2), and the enrichment in specific features in those DMRs was evaluated (Supplementary file 3). For the three subsets of DMRs, there was a lower proportion of promoters compared to the global average (p<0.001). A lower proportion of LINE1s (p<0.05) was also found for the C-IVF group, while the Natur-IVF blastocyst group showed a higher proportion of DMRs in transcription units (defined over the annotated genes from 500 bp downstream of the annotated TSS, p<0.05). Both C-IVF and Natur-IVF DMRs were less enriched in intergenic regions (p<0.001) and at LTRs (p<0.05) than In vivo blastocysts. These departures from the methylation state might reflect global differences in the DNA methylation and/or demethylation capacity of the different groups at a developmental time when DNA methylation is rather dynamic.

Exclusive DMRs for each group were linked to Canonical Pathways (p<0.01) and Diseases and Bio Functions (adjusted p-value < 0.05; Figure 6) by IPA software. Representative genes for specific DMRs in each group are listed in Supplementary file 4. A DMR overlapping IGF2R, a gene directly related with the LOS in ruminants and mouse, was found in the subset of exclusive C-IVF DMRs (Figure 5—source data 2). The methylation percentages for this region (Chr1: 9,199,522–9,201,143) were 12.45%, 28.3% and 35.5% for C-IVF, Natur-IVF and In vivo, respectively (Figure 7A). In addition, a CpG island (oe = 0.89, Chr1:9,200,658–9,202,276) that overlapped the DMR showed significant differences in methylation (p<0.05): 14.1%, 27.8% and 29.4% for C-IVF, Natur-IVF and In vivo groups, respectively (Figure 7B), although we should be cautious about their significance since the CpG island distribution in the pig genome is very different to the human or mouse genome.

Top Diseases and Bio Functions linked by Ingenuity Pathways Analysis to DMRs exclusive for each group with low or high methylation.
https://doi.org/10.7554/eLife.23670.015
Methylation differences at IGF2R.

(A) Methylation quantitation at IGF2R from the unbiased analysis of genome methylation in SeqMonk with a fixed size of 150 CpG windows. Mean percentages of methylation are shown by the bars for each group. Blue (unmethylated) and red (methylated) dots represent methylation reads. Asterisks indicate that methylation at the indicated region showed significantly different values (p<0.05) in Natur-IVF (*) and In vivo (**) vs C-IVF. TSS: transcription starting site. (B) Detailed view and methylation quantitation of the CpGi at the identified IGF2R DMR. Red rectangles represent, as indicated, CpG islands of the genes. Black boxes indicate the position of the targeted features, whose mean percentages of methylation are shown by the bars for each group. Blue (unmethylated) and red (methylated) dots represent methylation reads.

https://doi.org/10.7554/eLife.23670.016

Top Diseases and Bio Functions linked by IPA to DMRs exclusive for each group with low or high methylation are represented in Figure 6. Top Molecular and Cellular Functions and representative genes related to DMRs with higher or lower methylation in each group (C-IVF, Natur-IVF and In vivo) are listed in Supplementary file 4.

Three imprinted genes were differentially methylated in C-IVF, but not in Natur-IVF blastocysts, compared to in vivo blastocysts

Following the finding of a DMR at IGF2R, targeted analysis of candidate imprinted genes was done, as the differentially methylated regions of imprinted gene (igDMRs) are expected to maintain constant methylation in preimplantation embryos to ensure faithful imprinted expression of the associated genes throughout development. Therefore, they represent sites of methylation in preimplantation of clear biological significance. To identify putative igDMRs in the pig genome, all mouse igDMRs were lifted-over onto the pig genome. Where this was not possible, a gene-by-gene approach was taken to find the best possible fit for a candidate igDMR based on the known organization of the corresponding mouse imprinted gene. All the genomic regions were then inspected manually to confirm that the correct regions had been found (Table 4A). It is not possible to conclude that all regions were actually igDMRs (as this would require methylation information from oocyte and sperm) and, indeed, the methylation values indicated that for some of the genes there was no conserved DMR (i.e. methylation in blastocysts was far below the theoretical 50%) and the associated locus was unlikely to be imprinted. This would seem to be the case, for example, for the genes IMPACT, ZFP787 and ZFP777. For some, there was difficulty in finding possible homologous igDMRs, probably because of gaps in the porcine genome assembly (such as SNRPN, KCNQ1 and GRB10), and there were a number of others that were excluded because the homologous pig region had no suggestion of a CpG island in the region equivalent to the igDMR in mouse (e.g. U2AF1-RS1, MCTS2/H13). Comparison of methylation in the three groups of blastocysts for the resulting 14 candidate igDMRs (with sufficient read coverage) revealed differences for ZAC1 and PEG10, which were more methylated (p<0.05) in the C-IVF than in In vivo group, and PEG10 and NNAT, which were more methylated (p<0.05) in the C-IVF than in Natur-IVF and In vivo groups (Table 4B). No statistical differences were found between Natur-IVF and In vivo groups. Of these three igDMRs, the one at NNAT coincides with the promoter CpG island (Kobayashi et al., 2012) and, in addition, one 150 CpG tile overlapping NNAT had methylation higher than 50% in C-IVF in the unbiased analysis (Figure 8).

Methylation quantitation at NNAT from the unbiased analysis of genome methylation in SeqMonk with a fixed size of 150 CpG windows.

Black boxes indicate the position of the selected 150 CpG windows, whose mean percentages of methylation are shown by the bars for each group. Blue (unmethylated) and red (methylated) dots represent methylation reads. Asterisks indicate that methylation at the indicated region (black box) showed significantly different values (p<0.05) in Natur-IVF (*) and In-vivo (**) vs C-IVF.

https://doi.org/10.7554/eLife.23670.017
Table 4

Targeted analysis of candidate imprinted genes. (A) Predicted imprinted regions in the pig genome by lifted-over mouse igDMRs the pig genome and manually inspected. (B) Pair-wise comparison of methylation by Analysis Chi-Square in the three groups of blastocysts for the resulting 14 candidate igDMRs. *C-IVF vs In vivo: p<0.05 with 20 minimum observations and 10 minimum percentage of difference % methylation. ** C-IVF vsNatur-IVF: Analysis Chi- Square p<0.05 with 20 minimum observations and 10 minimum percentage of difference % methylation. Natur-IVF vs In vivo: no statistical differences.

https://doi.org/10.7554/eLife.23670.018
A)
TileChromosomeStartEnd
IGF2R/AIR19,244,2399,248,054
ZAC1123,638,88723,643,228
SOCS5399,885,36099,887,132
ZFP787655,574,08055,575,926
ZIM2656,641,19056,644,823
IMPACT6102,001,929102,002,533
NAT1l58139,773,830139,775,461
PEG10981,642,95781,644,146
INPP5FV214141,186,219141,188,231
NNAT1746,041,84346,045,629
NESPAS1766,313,67366,320,932
GNAS-exon1a1766,348,00966,352,062
MEST1819,340,33519,345,549
ZFP7771860,941,42160,943,096

B)
TileChromosomeC-IVFNatur-IVFIn-vivo
ZAC1142.4133.5523.87*
PEG10947.7536.91**30.90*
NNAT1734.6319.22**23.28*

Discussion

The milieu in which fertilization and embryo development takes place is crucial for healthy fetal and offspring growth, as revealed by developmental and epigenetic alterations as a consequence of in vitro culture and ART (Fernández-Gonzalez et al., 2004Kleijkers et al., 2014Lazaraviciute et al., 2014; Song et al., 2015). However, the progress made by ART during the past two decades make a future without their use inconceivable, thus it is necessary i) to characterize the real epigenetic cost of ART, separated from other factors and ii) to develop new protocols to safeguard against possible negative impacts in offspring. Our study evaluated, by single blastocyst profiling, the genetic and epigenetic impacts of modified protocols to produce embryos in vitro that mimic, as far as possible, the physiological conditions of fertilization and early embryo development. This imitation of the natural environment was first approached in both gametes separately: in the male gamete, by using sperm selection procedures that avoided centrifugations, and sperm washing and processing media containing oviductal fluid from the pre-ovulatory phase of the cycle; and, in the female gamete, by preincubating the oocytes within the precise fluid they encounter when, after ovulation, they are transported through the ampulla of the oviduct to the fertilization site, at the ampullar-isthmic junction (Halbert et al., 1988). Secondly, two experimental groups were established for a comparison with the in vivo specimens, where either BSA or reproductive fluids (obtained sequentially at the corresponding phases of the cycle) were added at every step of the IVF and EC procedures.

The results showed that reproductive fluids improve the outcome of IVF and the quality of pig blastocysts produced in vitro. The approach used, with spermatozoa coming from boars selected by their excellent reproductive performance, avoids the possibility of aberrations due to a paternal factor, which cannot be avoided in the human model, and helps to elucidate the epigenetic cost of ART independently of any paternal pathology. The figure of >40% progression of the cleaved embryos to blastocysts in vitro means an improvement over the best previous results (Redel et al., 2016). Nonetheless, the most remarkable findings were that Natur-IVF blastocysts attained a more advanced developmental stage and that the mean number of cells per blastocyst was the same as In-vivo embryos and 61% higher than C-IVF ones, which it is also above some of the best data previously reported in pigs (Redel et al., 2016). These results indicate that the use of reproductive fluids as additives, even at the low dose used in this study (1%) is beneficial for in vitro development of pig embryos so that it is now possible to obtain similar or even higher yields in the pig (45%) than in the bovine species. Although the possibility of transferring these methods to the human clinic might seem far off, the fact that nowadays other natural fluids such as breast milk for baby feeding or blood serum for transfusions are collected and stored at biobanks, make it possible to predict the future availability of human reproductive fluids obtained from oocyte donors during interventions at human infertility clinics (Coy and Yanagimachi, 2015). In fact, the first samples of these fluids are already stored at Biobanc-Mur in Spain (National Register of Biobanks N° B.0000859).

Our study also showed that Natur-ART blastocysts are closer to the gene expression profile of the In vivo blastocysts than C-IVF blastocysts. Amongst the most striking differences found was the expression of genes related to epigenetic reprogramming. It has been shown in mice and human that during the transition from zygote to blastocyst there is a massive loss of DNA methylation, with the exception of imprinted genes and some repetitive elements (Guo et al., 2014; Reik and Kelsey, 2014). In agreement with this observation, the global methylation level in the three groups of pig blastocysts analyzed was below 15%, suggesting that they had largely undergone a reprogramming event. This globally low methylation level compared to somatic cells or gametes, made it difficult to find high quantitative differences between embryos. Despite this, methylation percentage was higher in C-IVF embryos than in the other two groups, in agreement with previous studies indicating that ART-derived blastocysts displayed higher levels of methylation than in vivo derived ones (Deshmukh et al., 2011). This difference appeared to be global, with all features affected and, no evidence of multiple sub-groups over different genomic regions; therefore, there was no indication of specific regions resisting reprogramming. At the same time, genes for DNMT1 and the binding protein of its crucial cofactor UHRF1, which are considered responsible for maintenance of methylation patterns in replicating DNA and for maintaining imprints during preimplantation embryonic stages, were less expressed in C-IVF blastocysts, as was DNMT3B, required for de novo remethylation from this stage onwards. Differences in cell numbers, as a result of a probable additional round of cell division in In vivo and Natur-IVF embryos compared to C-IVF, is unlikely to explain a shift from ~11–12% to ~15% global methylation. All together, these data suggest an impaired demethylation in the C-IVF group. Analysis of hemimethylated CpG dyads by deep hairpin bisulfite sequencing, as recently reported in mouse (Arand et al., 2015), could help to clarify this issue.

A second key finding in this study was that the methylation levels in the samples analyzed showed much lower overall methylation levels (mean across all samples was 13.1%) than would be expected from somatic tissues. Furthermore, there were differences in the global mean methylation levels between different samples, ranging from 8.9% to 18.5%. Taken together, these observations suggest that the samples were collected during a time of global methylation reprogramming. The variability in global methylation levels would have confounded a direct comparison focussing on locus-specific methylation differences, so to account for this a quantile normalization was required to allow for a direct quantitative comparison of methylation levels.

Given that these samples are undergoing active reprogramming, it is also not unreasonable to think that some previously reported DMRs may not be established yet, or that the strength of the DMRs would be reduced. Despite this, we were able to find candidate DMRs between the groups with a reasonable statistical significance, although the magnitude of the methylation differences was low. Considering that previous studies have shown extremely close correlations between qPCR and RNA-seq data (Asmann et al., 2009; Griffith et al., 2010; Wu et al., 2014) and that validation by qPCR has its own probe-bias based on what region of the cDNA is amplified, we deem, in contrast to microarrays data, that there is not solid evidence that validation of the RNA-Seq and DNA methylation results by qPCR will provide extra significance to our results. For this reason, we did not perform qPCR validation in this study.

Another key observation in this study was that the in vitro culture affects imprinted gene expression and methylation. Plasticity of the preimplantation embryo could enable a recovery of alterations in methylation and further expression of non-imprinted genes during development, but any erosion of methylation marks at imprinted genes are unlikely to be corrected. In our data, from the 10 candidate imprinted regions retaining more than 30% of methylation in the pig blastocysts, we found three in C-IVF (ZAC1, PEG10 and NNAT) with significantly different methylation compared to In vivo blastocysts, and two (PEG10 and NNAT) compared to Natur-IVF. Knock-out mice lacking PEG10 showed early embryonic lethality with placental defects, indicating the importance of this gene in embryonic development (Ono et al., 2006). The protein encoded by NNAT, on the other hand, may be involved in the regulation of ion channels during brain development and may also play a role in forming and maintaining the structure of the nervous system. Defects in methylation at ZAC1 and IGF2R have been found in patients with the imprinted disorders transient neonatal diabetes mellitus (TNDM) or Silver-Russell syndrome (SRS), respectively, including those born following the use of ART (Le Bouc et al., 2010). In addition, genes related to the IGF axis, IGF2BP2 and IGF2BP2-IMP2, were up-regulated in C-IVF, and IGF2R in both C-IVF and Natur-IVF embryos. Altered IGF2BP2 expression in C-IVF is of interest, since reduced abundance of IGF2 has been associated with lower fetal weight after in vitro culture (El Hajj and Haaf, 2013). The imprinting status of IGF2R in the pig is unclear (Killian et al., 2001; Braunschweig, 2012) but, independently of this uncertainty, our data indicated higher expression of this gene in the two in vitro groups of blastocysts, which would be in agreement with previous reports in other species and could indicate a possibility of LOS-related alterations observed in abnormal in vitro and cloned embryos (Young et al., 2001). At the same time, the reduced methylation in IGF2R specifically in the C-IVF group could suggest that this group is more likely to be susceptible to sustained deregulation of IGF2R expression and a greater probability of LOS-like syndromes.

Altered expression in both groups of blastocysts produced under in vitro conditions was observed in some genes related to embryonic development, but some aberrations were absent in Natur-IVF embryos. In human blastocysts, it has been observed that those with higher implantation rate and higher number of cells per embryo showed up-regulation of DNMT3A (Kleijkers et al., 2015). In our data, the In vivo and Natur-IVF blastocysts showed a higher number of cells than those from the C-IVF group, in which expression of DNMT3A was decreased. We also observed higher expression of CDKN1A in the two in vitro groups, with an intermediate value in Natur-IVF. CDKN1A inhibits embryonic cell proliferation in response to DNA damage and it is considered one of the key genes responsible for the abnormalities in ART embryos since an aberrant increase of CDKN1A expression might be related to the growth-defect phenotype (Ishimura et al., 2016). Methylation of the CDKN1A gene, however, was similar in all three groups, between 5 and 7%. Other genes involved in DNA repair and cell cycle regulation were found to be altered, such as MDM2 (in C-IVF) and TP53INP (up-regulated in Natur-IVF and C-IVF) and HSPA4L, HSP40B1, HSPH1, HSP90 (down-regulated only in C-IVF). Altered expression of these genes may limit the ability of the embryo to respond to DNA damage, such that in vitro culture may lead to dysregulation of such genes, thus affecting long-term embryo viability (Zheng et al., 2005). The same situation was found for SLC2A3 (Glut-3) and SLC2A2, which have been related to LOS (Wrenzycki et al., 2004) and were highly up-regulated in the two in vitro groups. Again, no differences at the methylation level were found for any of these genes. Although DNA methylation at the promoter/gene bodies is directly/indirectly correlated with gene expression, this is not strictly true during the periods of dramatic loss of DNA methylation, as occurs during early embryo development or primordial germ cells (PGC) formation. For example, Gkountela et al. (2015) showed a general uncoupling between DNA methylation and gene expression during demethylation of PGCs, commenting ‘Our data reveal a remarkable and pervasive loss of DNA methylation in human PGCs and AGCs during prenatal life that has almost no relationship to changes in gene expression’. Comparative analyses between our methylation and gene expression data also showed this lack of correlation. In our opinion, at this stage of development and with this low level of methylation, this was an expected result.

Finally, the exclusive alteration in C-IVF of genes such as KIT, whose knock-out in mouse results in multiple alterations including embryonic lethality (Ro et al., 2010), UBR2, whose deletion results in female embryonic lethality and growth arrest (Kwon et al., 2003), or ISOC1, whose mutation produces phenotypes with body weight loss (Rainger et al., 2013), support the hypothesis that offspring produced with Natur-IVF conditions would be healthier than those produced with C-IVF, although additional studies are necessary to confirm this finding.

In conclusion, we report here the first time genome-wide DNA methylation and transcription analysis in single blastocysts (in vivo and in vitro) of a mammalian species and propose a new strategy for prevention of aberrant epigenetic and gene expression profiles induced by ART. This strategy, based on the addition of reproductive fluids in the culture media used during the ART procedures, can be applied in other animals as well as in humans, after safety concerns of transmission of diseases have been properly addressed. The design of new culture media containing all the proteins that are naturally present in the original biological fluid, represents not only a technical challenge but a biomedical responsibility that must be addressed to prevent future pathologies both in animals and humans. In addition, we offer a new protocol for the in vitro production of pig embryos with a significant improvement over the previous data published. Our study represents a new form of thinking in the field, far from the chemically defined culture media, and could help to face one of the biggest milestones of the current reproductive medicine: safer ART.

Materials and methods

Culture media

Unless otherwise indicated, all chemicals and reagents were purchased from Sigma-Aldrich Quimica S.A. (Madrid, Spain).

Collection and processing of follicular, oviductal and uterine fluids

Fluids were obtained from animals raised at a commercial farm (CEFU, S.A., Murcia, Spain) and slaughtered in an abattoir belonging to a food industry (El Pozo, S.A) near the University of Murcia. For the collection of follicular fluid, ovaries from 6-month-old Large White animals weighing 100–110 kg were transported to the laboratory in saline containing 100 μg/ml kanamycin sulfate, washed once in 0.04% cetrimide solution (alkyltrimethylammoniumbromide) and twice in saline within 30 min of slaughter. The content of follicles between 3 and 6 mm diameter, from at least 50 ovaries (25 females), was quickly aspirated, centrifuged at 1800 g for 30 min at 4°C and the supernatant filtered through 0.22 µm diameter filter (Naito et al., 1988). One ml follicular fluid (FF) aliquots were stored at −80°C until their use as additives for the IVM medium.

For the collection of oviductal (OF) and uterine (UF) fluids, genital tracts from cyclic Large White sows (2–4 years old) were obtained at the abattoir and transported to the laboratory on ice within 30 min of slaughter. The cyclic stage of animals was assessed once in the laboratory, on the basis of ovarian morphology on both ovaries from the same female. Oviducts and uteri were classified as early follicular, late follicular, early luteal or late luteal phase (Carrasco et al., 2008). Both oviducts and uteri coming from the same genital tract were classified as in the same stage of the cycle. Once classified, oviducts and uteri were separated and quickly washed once with 0.4% v/v cetrimide solution and twice in saline. Oviducts and uteri were dissected on Petri dishes or trays, respectively, sitting on ice. Once dissected, OF were collected by aspiration with an automatic pipette by introducing a 200 µl pipette tip into the ampulla and manually making an increasing pressure gradient from the isthmus to the ampulla. The UF was collected by making a manual increasing pressure gradient from the proximal end to the distal end (utero-tubal junction) of the uterine horn and letting the fluid drop into a sterile 50ml Falcon tube. Once recovered, samples (OF and UF) were centrifuged twice at 7000 g for 10 min at 4°C to remove cellular debris. Then the supernatant was immediately stored at −80°C until use. Oviducts from animals at the late follicular phase (POF-LF) and at the early luteal phase (POF-EL) gave a mean volume of around 50 µl and 40 µl, respectively per oviduct. At the early luteal phase, approximately 10 ml of UF per uterine horn were collected each time. Aliquots of 50 µl OF and 50 ml UF of pooled samples from at least 20 animals for OF and five animals for UF were used. Only samples that passed quality controls (pH 7.0–7.6, osmolality 280–320 mOsm/kg, endotoxin <0.10 EU/mL, a minimum 90% of Metaphase II oocytes after IVM with FF and ZP hardening for oocyte preincubation in POF-LF >1 hr) were used for experiments.

Oocyte collection and in vitro maturation

Ovaries from 6 months old animals weighing 100–110 kg were transported to the laboratory in saline containing 100 µg/ml kanamycin sulfate at 38°C, washed once in 0.04% cetrimide solution and twice in saline within 30 min of slaughter. Cumulus–oocyte complexes (COCs) were collected from antral follicles (3–6 mm diameter), washed twice with Dulbecco’s PBS (DPBS) supplemented with 1 mg/ml polyvinyl alcohol (PVA) and 0.005 mg/ml red phenol, and twice more in maturation medium previously equilibrated for a minimum of 3 hr at 38.5°C under 5% CO2 in air. Maturation medium was NCSU37 supplemented with 0.57 mM cysteine, 1 mM dibutyryl cAMP, 5 mg/ml insulin, 50 µM β-mercaptoethanol, 10 IU/ml equine chorionic gonadotropin (eCG; Foligon; Intervet International BV, Boxmeer, Holland), 10 IU/ml human chorionic gonadotropin (hCG; Veterin Corion; Divasa Farmavic, Barcelona, Spain), and 10% porcine follicular fluid (v/v). Only COCs with complete and dense cumulus oophorus were used for the experiments. Groups of 50 COCs were cultured in 500 µl maturation medium for 22 hr at 38.5°C under 5% CO2 in air. After culture, oocytes were washed twice in fresh maturation medium without dibutyryl cAMP, eCG and hCG and cultured for an additional period of 20–22 hr.

In vitro fertilization

Before IVF, mature oocytes were preincubated in 100% porcine oviductal fluid (POF) from the late follicular (LF) phase (NaturARTs POF-LF) for 30 min (Coy et al., 2008) and then washed three times in TALP medium. TALP medium consisted of 114.06 mM NaCl, 3.2 mM KCl, 8 mM Ca-lactate.5H2O, 0.5 mM MgCl2.6H2O, 0.35 mM NaH2PO4, 25.07 mM NaHCO3, 1.85 mM Na-lactate, 0.11 mM Na-pyruvate, 5 mM glucose, 2 mM caffeine, 1 mg/ml PVA and 0.17 mM kanamycin sulfate. Either 3 mg/ml BSA-FAF (A-6003) or 1% of NaturARTs POF-LF was included as additives in the IVF medium for the first 8 hr of coincubation (C-IVF and Natur-IVF groups, respectively). Ejaculated spermatozoa from boars of proven fertility (1–2 years old) were transported to the laboratory and 1 ml of semen was lay below 1 ml of NaturARTs PIG sperm swim up medium (http://embryocloud.com) at the bottom of a conical tube. After 20 min of incubation at 37°C (with the tube at a 45° angle), 0.75 ml from the top of the tube were aspirated and used for insemination of the IVF dishes (105 cells/ml) with the oocytes. For the density gradient group, aliquots of the semen samples (0.5 ml) were centrifuged (700 g, 30 min) through a discontinuous Percoll (Pharmacia, Uppsala, Sweden) gradient (45% and 90% v/v) and the resultant sperm pellets were diluted in TALP medium and centrifuged again for 10 min at 100 g. Finally, the pellet was diluted in TALP and 250 μl of this suspension were added to the wells containing the oocytes, giving a final concentration of 105 cells/ml.

Spermatozoa and oocytes were incubated at 38.5°C under 5% CO2 for 8 hours. Later on, the putative zygotes were transferred to embryo culture medium. At this point, a sample of the putative zygotes from each group was collected, fixed and stained as previously described (Coy et al., 2008) to assess the fertilization rates (percentage of penetrated oocytes, percentage of monospermy, mean number of spermatozoa penetrating each oocyte and mean number of spermatozoa attached to the zona pellucida). Penetration rate was defined as the proportion of oocytes penetrated by one or more spermatozoa.

In vitro culture of putative zygotes

Media for embryo culture were NCSU23 supplemented with sodium lactate (5 mM), pyruvate (0.5 mM) and non-essential amino acids (NCSU23a, for the first 48 hr) or NCSU23 supplemented with glucose (5.5 mM) and essential and non-essential amino acids (NCSU23b, 48–180 hr). At 8 hr post insemination (hpi), putative zygotes were transferred to culture dishes containing NCSU23a medium and 0.4% BSA in the C-IVF group or 1% POF from the early luteal (EL) phase of the estrous cycle (NaturARTs POF-EL) in the Natur-IVF group. At 48 hpi, the cleavage was assessed under the stereomicroscope and the 2–4 cell stage embryos were transferred to NCSU23b with 0.4% BSA (C-IVF group) or 1% of porcine uterine fluid (PUF) from early luteal phase (NaturARTs PUF-EL, Natur-IVF group). On day 7.5 (180 hpi), blastocyst stage morphology was assessed under the stereomicroscope and later on a sample was fixed and stained (Coy et al., 2008) and the remaining blastocyst were washed in PBS and frozen in PCR tubes in the minimum volume of medium. The parameters assessed in the stained blastocysts were development stage (2–4 cells, 8–16 cells, morula or blastocyst), mean number of cells per blastocyst, and ability for hatching (rhythmic movements of expansion and contraction before going out of the zona pellucida). The blastocysts frozen for genetic and epigenetic study were passed through liquid nitrogen vapours for 5 s and immediately introduced in the freezer at -80°C until the day of use for RNA extraction or bisulphite treatment.

Statistical analysis of IVF data

Data are presented as mean ± SEM, and all percentages were modeled according to the binomial model of variables and arcsin transformation to achieve normal distribution. The variables in all the experiments were analyzed by one-way or two-way ANOVA. When ANOVAs revealed a significant effect, values were compared by the Tukey test. A pvalue < 0.05 was taken to denote statistical significance.

Collection of blastocysts In vivo

Ten sows 18-month old were weaned 21 days after second parturition and five days later showed signs of standing estrous. Animals were inseminated in the collaborative farm and slaughterhoused 7.5 days after. Genital tracts were collected and transported to the laboratory where uterine horns were briefly dissected and washed with PBS within 2 hr from slaughtering. Blastocysts were identified under the stereomicroscope, collected and immediately frozen as described for the in vitro produced embryos. A portion of these blastocysts was fixed in glutaraldehyde and stained with Hoechst for cell counting.

Experimental groups

C-IVF group (C-IVF): six blastocysts classified as 7A according to Bo and Mapletoft (Bo and Mapletoft, 2013) (#34, 35, 36, 93, 94 and 96) were produced in vitro with BSA as the only protein source. Sperm were processed by swim up in NaturARTs sperm medium with BSA (Swim-up-BSA). IVF medium consisted of TALP (0–8 hr) and embryo culture medium consisted of NCSU23a (8–48 hr) and NCSU23b (48–180 hr). Natur-IVF group: six blastocysts classified as 7A (#55, 85, 86, 27, 54 and 60) were produced in vitro with NaturARTs POF and PUF as the protein source. Sperm were processed by swim up in NaturARTs sperm medium with NaturARTs POF-LF (Swim-up-Fluid). IVF medium consisted of TALP +1% NaturARTs POF-LF (0–8 hr) and embryo culture medium consisted of NCSU23a + 1% NaturARTs POF-EL (8–48 hr) and NCSU23b + 1% NaturARTs PUF-LL (48–180 hr). For both groups, before IVF oocytes were pre-incubated for 30 min in preovulatory oviductal fluid (NaturARTs POF-LF). In vivo group: six blastocysts classified as 7A (#186, 193, 197, 189, 190 and 191) were collected by flushing the uteri of animals within 2 hr of slaughtering. The animals were under natural heat after weaning and insemination was performed 7 days before slaughtering.

RNA isolation and RNA-Seq

ARCTURUS PicoPure RNA Isolation Kit (KIT0204, Life Technologies) was used to extract the RNA from individual blastocysts. RNA-Seq libraries were generated using Ovation RNA-Seq System V2 (NuGEN, Cat. 7102–08) for low amount of starting material and further amplified with NEB Next DNA Library Prep Master Mix for Illumina (NEB, Cat. E6040S). All steps were performed according to manufacture guidelines. iPCRTag reverse primer with individual index was used to generate three independent biological replicates from each condition. 100 bp single end reads were sequenced on Illumina HiSeq 1000. Sequencing data were processed. For RNA-Seq libraries, raw sequence reads were trimmed using Trim Galore to remove adapter contamination and reads with poor quality defined by low PHRED score. Mapping was performed using Tophat software (http://tophat.cbcb.umd.edu/) and data were visualized with Seqmonk (RRID:SCR_001913, http://www.bioinformatics.babraham.ac.uk/projects/seqmonk/). RNA quality was assayed by Bioanalyzer and even though each sample came from a single blastocyst, RIN score was between 6.1–8.2.

Analysis of RNA-Seq data

Annotated pig mRNA features were quantitated with raw read counts in SeqMonk and these were fed into DESeq2 for differential expression analysis using a p-value cutoff of 0.05 and not applying independent filtering. Reads were subsequently re-quantitated as log2RPM (reads per million reads of library) and globally normalized to the 75th percentile of the data. Significant effect sizes were selected using the Seqmonk intensity difference filter where the difference in expression in each gene was compared to the set of differences in the 1% of the data with the most similar average expression level as the gene being tested. Only genes with significantly higher changes (p<0.05 after Benjamini and Hochberg correction) were kept.

Bisulfite sequencing based on post-bisulfite adapter tagging

An adaptation of whole genome bisulfite sequencing that involves post-bisulfite adapter tagging (PBAT) (Miura et al., 2012) was used to analyze the methylome of individual pig blastocysts at single-base resolution on a genome-wide scale. Further modification of the method described in Smallwood et al. (Smallwood et al., 2014) was used to generate BS-seq libraries. Briefly, an individual blastocyst was lysed for 1 hr in 1% SDS with proteinase K and treated with bisulfite reagent using Imprint DNA modification kit (Sigma, MOD50). DNA was eluted in EB buffer and one round of first strand synthesis was performed using a biotinylated oligo 1 (5-[Btn]CTACACGACGCTCTTCCGATCTNNNNNNNNN-3). Samples were further treated with Exonuclease I, washed and eluted in 10 mM Tris-Cl and incubated with washed M-280 Streptavidin Dynabeads (Life Technologies) to pull down the biotinilated fraction of DNA. Second strand synthesis was performed using oligo 2 (5’-TGCTGAACCGCTCTTCCGATCTNNNNNNNNN −3’) and samples were amplified for 15 PCR cycles using indexed iPCRTag reverse primer (Smallwood et al., 2014) with KAPA HiFi HotStart DNA Polymerase (KAPA Biosystems) and purified using 0.8× Agencourt Ampure XP beads (Beckman Coulter). Libraries were assessed for quality and quantity using High-Sensitivity DNA chips on the Agilent Bioanalyser, and the KAPA Library Quantification Kit for Illumina (KAPA Biosystems). Three libraries generated from individual blastocysts for each experimental condition were prepared for 100 bp single-end sequencing on Illumina HiSeq 1000 and sequenced at three samples per lane.

Analysis of methylation data

For the unbiased analysis, tiles were defined in SeqMonk using the read position tile generator tool and selecting one read count per position and 150 valid positions per window, in all the nine individual data sets (286,136 tiles). Then, the bisulphite quantitation pipeline was run over existing tiles, one minimum count to include position and 20 minimum observations to include feature. To remove the tiles without data, the filter on values for individual tiles was applied, where values had to be between 0 and 100 for exactly 9 of the nine selected data stores. Then, tiles with data for all the samples were obtained (N = 258,885 tiles). Bisulphite quantitation pipeline was run again over the new tiles and data were normalized by the match distribution quantile normalization tool. Finally, every pair-wise comparison was filtered to require a consistent 5% change between all replicates of the first and second condition, and then replicate sets stats was applied where every comparison had a significance below 0.05 after Benjamini and Hochberg correction. For the targeted analysis of the candidate imprinted regions a Chi-Square test (p<0.05) was applied for every comparison.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
    Evaluation and classification of bovine embryos
    1. G Bo
    2. R Mapletoft
    (2013)
    Animal Reproduction 10:344–348.
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
    5 million babies
    1. International Committee for Monitoring Assisted Reproductive Technology I
    (2012)
    Annual Meeting European Society of Human Reproduction And Embryology.
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58

Decision letter

  1. Jessica K Tyler
    Reviewing Editor; Weill Cornell Medicine, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Epigenetic and gene expression changes derived from assisted reproduction can be mitigated by reproductive secretions" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that we cannot accept your present manuscript for publication in eLife. We felt that you have undertaken a very important subject matter and there are many new and potentially significant advances in the study. However, all three reviewers expressed significant technical concerns regarding the quality of the material pre and post amplification, characteristics and quality of the natural fluid obtained from pigs, and data analysis. They also felt that the conclusions have not been justified by the data. We hope you will find the comments helpful as you revise your work. Our usual policy is that we reject papers unless technical concerns could be addressed within 1-2 months. We feel that the amount of revision work necessary for your study to be competitive would require more work than one could reasonably do in that time frame. If you do gather additional evidence down the road that could satisfy in full the reviewer concerns, we could potentially welcome the submission of a new manuscript for consideration.

Reviewer #1:

The authors have undertaken a very important and difficult study. They are correct in emphasizing the need for data of this type. An enormous amount of work went into carrying out this work. There are some very impressive results, but in some areas, the authors may have attempted some interpretations that are beyond what the data can support.

Figure 1 is instructive and essential, although the legend should be expanded. For all of the figures, there are items that are not explained in the legend. Perhaps these are intuitive for the authors. What is the meaning of the red boxes toward the top of the figure? There are abbreviations that are not explained such as OF-EL and UF-EL. Abbreviations are frequently not explained in the text. Some of the font is illegible.

The data in Tables 1, 2, and 3 are quite useful. Particularly the fact that some blastocysts achieve hatching for the Natur-IVF is impressive. Perhaps it would be helpful to show some images of hatching blastocysts. The data in Figures 25 are legitimate and can be made available as supplements, but interpretation now becomes quite speculative beyond the overall trend that changes from the in vivo control are greater for the C-IVF than for the Natur-IVF. The amount of speculation about the possible significance of specific gene changes should be restricted.

The tables and figures related to DMRs are much weaker. All of the qualifications and problems described in the subsection “Three imprinted genes were differentially methylated in C-IVF, but not in Natur-IVF blastocysts, compared to in vivo blastocysts” raise questions about whether it is feasible to interpret the data in the way that is presented. Given concerns about gaps in the pig genome, manual inspection of all genome regions, and failure to detect prominent DMRs such as SNRPN, which is widely conserved across species, raise questions as to whether these analyses and the associated interpretations are meaningful. The legend for Figure 7 is particularly lacking. There is no explanation of the color significance of the blue and red read data. Is red methylated and blue unmethylated?

An attempt to reduce the manuscript to those data and interpretations that are most solid would be useful.

Philosophically, this reviewer is not convinced that use of human reproductive fluids is the way of the future. The analogy to breast milk is valid in some ways but in most ways not. Reproductive fluids are overwhelmingly less accessible than breast milk, and are likely to vary from patient to patient especially in pathologic circumstances. Even for breast milk, which is far more accessible, babies whose mothers are unable to breastfeed do not usually get fed breast milk from other women, but rather receive artificial formulas of greater and greater sophistication. Wet nursing is unlikely to make a big comeback in the future. Rather proteomic analysis of various fluids should make it possible to prepare reproducibly better and better artificial reproductive fluids.

Reviewer #2:

The work by Canovas et al. titled: "Epigenetic and gene expression changes derived from assisted reproduction can be mitigated by reproductive secretions" explores the developmental, transcriptomic (RNA seq) and epigenetics (WGBS by using post-bisulfite adapter tagging- PBAT) effects of IVF. The authors tested single pig blastocysts generated by IVF using conventional media (C-IVF) or media with the addition of physiologic secretions obtained from the fallopian tubes (late follicular phase, early luteal phase) or the uterus (called Nature IVF). The control is provided by in vivo blastocysts flashed out of the uterus. Further they tested performance of a new swim up technique as opposed to the more conventional gradient centrifugation methods to select sperm for IVF.

The authors found that:

1) Swim up technique is superior to standard fertilization, resulting in higher blastocyst rate

Nature IVF compared to C-IVF resulted in

2) higher blastocyst state with faster development (higher number of hatched/ hatching blastocysts

3) higher total cell number, that was similar to in vivo blasts

4) less gene expression changes (789 DEG vs 623)

5) Less DNA methylation changes in CpGs (11% vs 15%) while in vivo embryos had an average of 12%. Further they indicate that more imprinted genes or genes involved in epigenetic controls are altered in C-IVF.

The work has important translational value, given that more than 5 million children have been conceived by IVF and that epigenetic and developmental changes have been postulated in children and described in IVF offspring using animal models. Testing novel fertilization and culture techniques that could decrease the epigenetic and developmental problems in offspring is therefore valuable and important. Further, performing single blastocyst analysis is an important technical achievement.

While the study appears to be well designed, there are few but significant technical concerns.

1) First, it is not described in the paper (there is a reference to a website) the standard operating procedure utilized to obtain the physiologic follicular and uterine fluid from pigs and how the media was stored and controlled for contamination. This is a critical aspect of the paper that is missing. For example: is there evidence that late follicular fluid collected from live pigs is always the same, given that changes in nutrition/ temperature/ wellbeing of the animal could change the composition of the fluid? Further: is the fluid from a single animal or pooled from how many animals?

2) Quality controls and validation of RNA seq and WGBS are missing. This is critical given that authors worked with extremely low amount of material. (How was the quality of RNA or RNA amplification assessed? How many reads were obtained per sample? Further, validation of some key genes in independent samples, by RT PCR and bisulfite sequencing is missing).

3) Unsupervised clustering of the 9 samples for both RNA seq and WGBS of all genes (not only the statistically different genes) should be provided to truly confirm that samples cluster based on fertilization and culture methods.

Reviewer #3:

In the manuscript entitled "Epigenetic and gene expression changes derived from assisted reproduction can be mitigated by reproductive secretions", Canovas and colleagues investigate the transcriptome and methylome of in vivo-derived and in vitro produced pig blastocysts. To do this, the group first developed a swim-up method for isolating sperm. Following fertilization, in vitro-produced embryo was cultured in conventional conditions that include BSA, or conventional culture augmented with 1% oviductal fluid and uterine fluid. RNA seq and whole genome bisulfite seq were performed on three individual blastocysts per treatment group then data were pooled for analysis. Differential expressed genes and differentially methylated regions were identified between the various experimental groups. In this manuscript, the authors have produced a vast amount of data using novel approaches, including the production of pig embryos with reproductive secretions and single blastocyst methylation analyses. Overall, the paper is well written.

1) Overinterpretation/misrepresentation of data.

A) The manuscript concluded that "reproductive fluids improve the outcome of IVF and the quality of pig blastocysts produced in vitro", and that "remarkable findings were that Natur-IVF blastocysts attained a more advanced developmental stage". These conclusions are not readily discernible from the data. In Tables 2 and 3, C-IVF embryos produced a great percent of cleavage stage embryos and early blastocysts than Natur-IVF. Additionally, there was no significant difference between C-IVF and Natur-IVF in percent of blastocysts and expanded blastocyst formed or the yield. How does this show an improved outcome or advanced developmental stage? While there was a significant difference in hatching rate, it would appear that embryos were not left long enough to truly assess this, since 0 out of 903 C-IVF embryos and 48 out of 961 (5%) Natur-IVF had hatched.

B) The authors bias the presentation of the transcriptome and methylome data towards finding less difference in the Natur-IVF group. The major questions that should be asked are whether there are expression/methylation changes between the three experimental groups and then where do these differences exist. Furthermore, what is the variation between embryos in the same group? The authors should present PCA graphs to visualize the three embryos and three groups. This should then be followed by pairwise comparisons for all groups, with VENN diagrams shown for all comparisons, including Natur-IVF and C-IVF. In fact, when the comparison is made (subsection “Natur-IVF blastocysts show fewer aberrantly expressed genes than C-IVF blastocysts”, second paragraph), only 29 differentially expressed genes were identified between Natur-IVF and C-IVF, demonstrating little difference in the cultured groups. Instead, as seen in Figure 1B, greater differences were found between the in vivo group and the two cultured groups. For DNA methylation, there does not appear to be a significant difference between the three groups for total methylation levels or methylation levels over specific genomic features (Table 4). Here again, C-IVF and Natur-IVF groups need to be compared to determine how much they differ from each other. Finally, while comparisons between the in vivo and each culture group produced a differ subset of genes affected, there does not appear to great differences in the percent of genes misregulated (gene expression: C-IVF 68% up-, 32% down-regulated; Natur-IVF 69% up- and 31% down-regulated; DNA methylation: C-IVF 57% more and 43% less methylated; Natur-IVF 66% more and 34% less methylated). Thus, from the present analysis, it does not appear that "Epigenetic and gene expression changes derived from assisted reproduction can be mitigated by reproductive secretions"(title).

2) In this manuscript, the authors produced single blastocyst profiling then pooled the data. This suggests that only genes affected in all three embryos would be analyzed. Previous DNA methylation analyses have shown that losses and gains of DNA methylation appear to occur stochastically. Thus, pooling of the data may mask gene differences in response to culture. Is there enough coverage to perform analyses on single embryos? If not, can differences be assessed with 2/3 embryos showing the same differences? If not, the authors need to acknowledge that they have analyzed only the most common changes.

3) The authors performed both transcriptome and methyome analyzes. Have they mapped the two data sets to determine whether DNA methylation changes are correlated with gene expression changed in neighboring genes?

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your article "DNA methylation and gene expression changes derived from assisted reproduction can be decreased by reproductive fluids" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Mellissa Mann (Reviewer #3), and the evaluation has been overseen Jessica Tyler as the Senior Editor and Reviewing Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

In summary we feel that this is a valuable contribution to the field. There are numerous concerns but they can be addressed by changes to the text. One of the key concerns that needs to be addressed is the nature of the natural fluid used in the study, because if the same lot were used for all the experiments, this would be a problem. There are also a lot of major grammatical and editorial improvements that are essential, and this work must be edited by a scientist who speaks English as a first language before being sent back to us. I have included the full reviews here, so that you can make all of the necessary changes, as we consider them all essential and very easy to incorporate.

Reviewer #1:

The resubmission of the work by Canovas et al. has responded to the majority of comments. However there still remain several concerns. Overall the style of the new version is more difficult to read and will benefit from significant editing both from a language and spelling point of view.

1) In particular the standard operating procedure utilized to obtain the physiologic follicular and uterine fluid from pigs and how the media was stored and controlled for contamination needs to be described in Materials and methods. Referring to other manuscripts is not helpful for readers. For example, the Carrasco 2008 manuscript is not even mentioned in the manuscript. The Coy 2008 paper includes only a short Methods section with a short description of the methods to collect samples from pigs, with no indication on how many pigs were pooled or the range of their age. As an example of lack of clarity, based on the Coy paper: for how many hours are the fallopian tubes at room temperature from sacrificing the animal to collecting the fluid? This is critical to the reader, as quality of the fluid could be compromised or lost during manipulation/transport at room temperature. More Details, including some included in the rebuttal to authors need to be included in Materials and methods section. Given the novelty of the method, this is absolutely needed.

2) The fact that adding only 1% of oviduct fluid can obtain such remarkable results is impressive and this finding is surprisingly missing in the Discussion. If the authors were to add 5% or 10% of OF, would the result change? How did the authors come to the 1% dose? This needs to be commented on in the Discussion.

3) A statement in the Discussion that validation of the RNA and DNA methylation results in independent samples was not performed need to be added as an important limitation of the paper.

4) Please confirm and describe in the Methods section that the 3 biological replicates in the nature IVF group were generated from the addition of 3 independent sets of uterine and oviductal fluids (POF-LF, OF-EL; UF-EL), collected from different animals. This is a key experimental factor that is not described in the Methods section. In fact, adding the same OF to 3 different blastocysts would be expected to lead to similar/ consistent transcriptomic and epigenetic results in blastocysts. If this was not done, data from independent fluid donors should be provided.

Reviewer #2:

It was not possible to review the manuscript in much depth as was the case for the original submission. However, in general, the authors appear to have responded well to lengthy criticisms of three reviewers. The interpretations are more restricted. This reviewer continues to doubt that the future of the field of human ART lies in collecting biological fluids from humans or animals but rather that artificial fluids mimicking natural fluids more and more closely are the answer. However, the authors are entitled to their opinion. In general, the manuscript is an excellent contribution to the knowledge in this field.

Reviewer #3:

In the manuscript, Canovas and colleagues have investigated an important question in the ART field, for which they produced an impressive amount of data and novel findings. They have clarified and addressed all my concerns.

https://doi.org/10.7554/eLife.23670.026

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

[…]Reviewer #1:

The authors have undertaken a very important and difficult study. They are correct in emphasizing the need for data of this type. An enormous amount of work went into carrying out this work. There are some very impressive results, but in some areas, the authors may have attempted some interpretations that are beyond what the data can support.

We acknowledge the encouraging comments from the reviewer and are glad to see he/she agrees with the need of this type of study. We have modified the manuscript according to his/her suggestions and those from the other reviewers and think that interpretations are now more adjusted to the data presented. So, we hope this version of our manuscript can be now considered acceptable for publication.

Figure 1 is instructive and essential, although the legend should be expanded. For all of the figures, there are items that are not explained in the legend. Perhaps these are intuitive for the authors. What is the meaning of the red boxes toward the top of the figure? There are abbreviations that are not explained such as OF-EL and UF-EL. Abbreviations are frequently not explained in the text. Some of the font is illegible.

Figure 1 and the others have been modified according to the reviewer´s comments. Legends have been expanded, abbreviations have been included and font is now higher. Also, it has been split into Figure 1 and 2 for better clarification.

The data in Tables 1, 2, and 3 are quite useful. Particularly the fact that some blastocysts achieve hatching for the Natur-IVF is impressive. Perhaps it would be helpful to show some images of hatching blastocysts.

Thank you for the comment. We have included some images of blastocysts in Table 2.

The data in Figures 25 are legitimate and can be made available as supplements, but interpretation now becomes quite speculative beyond the overall trend that changes from the in vivo control are greater for the C-IVF than for the Natur-IVF. The amount of speculation about the possible significance of specific gene changes should be restricted.

The amount of speculation has been restricted. In addition, a PCA analysis has been included showing how the three individuals of each group cluster together, thus legitimising the three pairwise comparisons made for exploring the gene expression data. Also, some figures have been modified and moved to Supplementary files.

The tables and figures related to DMRs are much weaker. All of the qualifications and problems described in the subsection “Three imprinted genes were differentially methylated in C-IVF, but not in Natur-IVF blastocysts, compared to in vivo blastocysts” raise questions about whether it is feasible to interpret the data in the way that is presented. Given concerns about gaps in the pig genome, manual inspection of all genome regions, and failure to detect prominent DMRs such as SNRPN, which is widely conserved across species, raise questions as to whether these analyses and the associated interpretations are meaningful.

The qualifications described in relation to the imprinted gene analysis are a legitimateand candid account of how we went about trying to identify candidate imprinted genes in the pig genome, based on the more extensive knowledge of imprinting in mouse and human, and in the face of some deficiencies in the pig genome assembly and the generally poor level of gene annotation. It is recognised that not all imprinted genes are conserved in their imprinted status between species and, indeed, our analysis of the genome organisation of some of these genessuggested that they are unlikely to be imprinted in pig. In the case of SNPRN, in human and mouse this is a very complex imprinted locus in which there are multiple alternative promoters and associated CpG islands that have arisen through a complex series of genomic duplication events (for example, see Smith et al. 2011PLoSGenet. 7:e1002422) so, in the absence of further evidence, we did not feel it justified to attribute to any of possible CpG islands in the pig SNRPN locus the candidate igDMR. The fact that some imprinted genes could not be confidently identified based on comparative sequence organisation does not detract from the fact that we were able to identify 14 candidate igDMRs, which corresponds to more than half of the corresponding regions known from the mouse, and represents a very substantial subset and legitimises the analysis. Amongst these, we did detect significant differences in methylation at 3 igDMRs, and this finding is detailed in the text.

Nonetheless, we would like to thank reviewer 1 for his/her comment because, while preparingthis answer, we have detected some mistakes in the writing that could have been confusing for the reader. We have re-written these paragraphs and hope now the methylation data and the associated interpretations in the manuscript are meaningful.

The legend for Figure 7 is particularly lacking. There is no explanation of the color significance of the blue and red read data. Is red methylated and blue unmethylated?

An attempt to reduce the manuscript to those data and interpretations that are most solid would be useful.

Legend for Figure 7 has been expanded and corrected because we found some mistakes. We thank the reviewer for this. Red represents methylated cytosines and blue unmethylated. Among the numbers of DMRs found in the unbiased analysis, we identified one overlapping IGF2R in C- IVF embryos compared to in vivo or to Natur-IVF. Also, when the analysis was targeted to CpG islands, we identified one differentially methylated CpGisland in the IGF2R gene. For this reason, and due to the importance of IGF2R in the anomalies observed after ART in animals, we thought it was important and legitimate to include it in this figure as an example of consistent methylation difference in our groups. Similarly, we thought it important to present the result for another igDMR (NNAT) that showed methylation differences between groups, given the importance of correct maintenance of methylation at imprinted genes.

Philosophically, this reviewer is not convinced that use of human reproductive fluids is the way of the future. The analogy to breast milk is valid in some ways but in most ways not. Reproductive fluids are overwhelmingly less accessible than breast milk, and are likely to vary from patient to patient especially in pathologic circumstances. Even for breast milk, which is far more accessible, babies whose mothers are unable to breastfeed do not usually get fed breast milk from other women, but rather receive artificial formulas of greater and greater sophistication. Wet nursing is unlikely to make a big comeback in the future. Rather proteomic analysis of various fluids should make it possible to prepare reproducibly better and better artificial reproductive fluids.

We acknowledge this comment and agree partially. Of course, human reproductive fluids are not easy to collect, although we have developed a system to do it and are now getting samples of around 50-150 μL per woman. The trick is to take the samples from healthy ovum donors attending the fertility centers. With the appropriate informed consents and legal permits, we are collecting these “health” samples simply by introducing a special catheter throughthe vagina to the uterus in these donors and characterizing the proteomic profiles as well as other parameters. Bearing in mind that the amount of fluids needed is very small (we have used in the pig model 1% in the final volume of 500 μL) and that the volumes of culture medium in the human clinics are around 50 μL per embryo (or smaller microdrops), each sample collected from only one woman could be potentially used as additive for around 200-400 drops of 50 μL. So, if the beneficial effect were proved, it would not be so difficult to replace the current rHBSA used as protein additive in the human culture media by the corresponding reproductive fluid (previously tested, characterized and endotoxin-free).

Regarding human milk, to use breastfeeding or artificial formula is a personal decision, butmedical benefits of the first one are very clear: “it contains nutrients necessary for infant's growth but also numerous bioactive factors contributing to beneficial effects on gastrointestinal maturation, host defence, infection, cardiovascular risks, metabolic disease, neurodevelopmental outcome as well as in infant's and mother's psychological well-being” (reviewed recently by de Halleux V et al., Semin Fetal neonatal Med 2016). Human milk is the gold standard to provide nutritional support for all healthy and sick newborn infants including the very low birth weight (VLBW) infant (<1500 g) (Johnston M et al. Pediatrics 2012) When own mother's milk it is unavailable, donor human milk is recommended as the first alternative, which could be obtained through human milk biobanks that are growing worldwide (Updegrove KH. Breastfeed Med 2013). Similarly, the idea of creating reproductive fluids biobanks with fully characterized samples could provide some benefits and may be important in the future for treatment of specific pathologies.

Reviewer #2:

[…] While the study appears to be well designed, there are few but significant technical concerns.

1) First, it is not described in the paper (there is a reference to a website) the standard operating procedure utilized to obtain the physiologic follicular and uterine fluid from pigs and how the media was stored and controlled for contamination. This is a critical aspect of the paper that is missing. For example: is there evidence that late follicular fluid collected from live pigs is always the same, given that changes in nutrition/ temperature/ wellbeing of the animal could change the composition of the fluid? Further: is the fluid from a single animal or pooled from how many animals?

All the samples were pooled from commercial animals raised for human consumption (Large White x Landrace hybrids) from collaborative farms, fulfilling all the requirements for wellbeing, health and nutrition. The animals were slaughtered in an abattoir near the University of Murcia belonging to a big food industry (El Pozo, S.A). We have not included more details because all the procedures for fluid collection and storing were first described in 2008 (Carrasco et al.) and later on we have published different articles using these fluids and testing their effect or the effect of specific oviductal proteins on IVF, zona pellucida hardening, embryo development, embryo survival after cryopreservation, etc. These are the references published until now:

Algarra B, Han L, Soriano-Úbeda C, Avilés M, Coy P, Jovine L, Jiménez-Movilla M. The C-Terminal Region Of Ovgp1 Remodelsthe Zona Pellucida1 And Modifies fertility parameters. Scientific Reports 2016. | 6:32556 | Doi: 10.1038/Srep32556

Lopera-Vásquez R, Hamdi M, Maillo V, Lloreda V, Coy P, Gutiérrez-Adán A, Bermejo-Álvarez P, Rizos D..Effect Of Bovine Oviductal Fluid On Development And Quality Of Bovine Embryos in Vitro. Reproduction Fertility And Development 2015.Doi: 10.1071/Rd15238.

Coy P, Yanagimachi R. Common And Species-Specific Roles Of Oviductal Proteins In Mammalian Fertilization And Embryo Development. Bioscience65:973-984.2015. (Doi:10.1093/Biosci/Biv119).

Ballester L, Romero-Aguirregomezcorta J, Soriano-Úbeda C, Matás C, Romar R, Coy P. Timing Of Oviductal Fluid Collection, Steroid Concentrations And Sperm Preservation Method Affect in Vitro Fertilization Efficiency. Fertility And Sterility. 102:1762-1768. 2014 Http://Dx.Doi.Org/10.1016/J.Fertnstert.2014.08.009..

Mondéjar I, Martínez I,Avilés M, Coy P. Identification Of Potential Oviductal Factors Responsible Of The Zona Pellucida Hardening And Monospermy During Fertilization In Mammals. Biology Of Reproduction 89 (3): 67, 1-8. 2013.

Mondéjar I, Avilés M, Coy P. The Human Is An Exception To The Evolutionarily-Conserved Phenomenon Of Pre-Fertilization Zonapellucida Resistance To Proteolysis Induced By Oviductal Fluid. Human Reproduction28:718-728. 2013.

Grullón La, Gadea J, Mondéjar I, Matás C, Romar R, Coy P. How Is Plasminogen/Plasmin System Contributing To Regulate Sperm Entry Into The Oocyte? Reproductive Sciences 20:1075-1082. 2013doi:10.1177/1933719112473657.

Cebrián-Serrano A, Salvador I, Garcia-Roselló E, Pericuesta E, Pérez-Cerezales S, Gutierrez- Adán A, Coy P, Silvestre Ma. Influence Of Oviductal Fluid On in Vitro Fertilisation, Development And Gene Expression Of in Vitro Produced Bovine Blastocysts. Reproduction In Domestic Animals 48(2): 331-338. 2013

Coy P, García-Vázquez Fa, Visconti P, Avilés M. Roles Of The Oviduct In Mammalian Fertilization. Reproduction144 649–660. 2012.

Coy P, Jiménez-Movilla M, García-Vázquez Fa, Mondéjar I, Grullón L, Romar R. Oocytes Use Plasminogen- Plasmin System To Remove Supernumerary Spermatozoa. Human Reproduction 27(7):1985-1993. 2012.

Mondéjar I, Acuña S, Izquierdo Rico Mj, Coy P, Avilés M. The Oviduct: Functional Genomic And Proteomic Approach. Reproduction In Domestic Animals 47(3): 22-29.2012

Mondéjar I, Grullón La, García-Vázquez Fa, Romar R, Coy P. Fertilization Outcome Could Be Regulated By Binding Of Oviductal Plasminogen To Oocytes And By Releasing Of Plasminogen-Activators During Interplay Between Gametes. Fertility And Sterility2012 97(2):453-461.

Avilés M, Gutiérrez-Adan A, Coy P. Oviductal Secretions: Will They Be Key Factors For The Future Arts?.Molecular Human Reproduction. 2010 16 (12): 896-906.

Coy P, Avilés M. What Controls Polyspermy In Mammals, The Oviduct Or The Oocyte? Biological Reviews Cambphilos Soc. 2010;85(3):593-605.

Coy P, Lloyd R, Romar R, Satake N, Matas C, Gadea J, Holt Wv. Effects Of Porcine Pre-Ovulatory Oviductal Fluid On Boar Sperm Function. Theriogenology74(4):632-642. 2010.

Lloyd R, Romar R, Matas C, Gutiérrez-Adán A, Holt Wv, Coy P. Effects Of Oviductal Fluid On The Generation, Quality And Gene Expression Of Porcine Blastocyst Produced in Vitro.Reproduction 137:679-687. 2009.

Coy P, Cánovas S, Romar R, Saavedra Md, Grullón L, Mondéjar I, Matás C, Avilés M. Oviduct-Specific Glycoprotein And Heparin Modulate Sperm-Zona Pellucida Interaction During Mammalian Fertilization. Proc Nat Acadsci Usa (Pnas), 105:15809–15814. 2008.

Carrasco Lc, Coy P, Avilés M, Gadea J, Romar R.Glycosidase Determination In Bovine Oviduct Fluid At Follicular And Luteal Phases Of The Estrous Cycle. Reproduction, Fertility And Development20 1-10. 2008

Carrasco Lc, Romar R, Avilés M, Gadea J, Coy P. Determination Of Glycosidase Activity In Porcine Oviduct Fluid At The Different Phases Of The Estrous Cycle. Reproduction136: 833–842. 2008.

In the present study, apart from the website mentioned, we have cited in the Methods section our first article describing the beneficial effects of oviductal fluid on the IVF results in pig and cow and the procedures for collection of fluids (Coy and Avilés, 2010, Coy et al., PNAS 2008), as well as two reviews (Coy and Avilés, 2010 and Coy and Yanagimachi, 2015) that summarize the information from the other articles and the procedures for the collection, standardization and quality control of the samples. However, parts of the procedures are patented by University of Murcia (ES-2532659B2) and for this reason we cannot give more details in the paper. Because we are directly involved in Embryocloud, we know that the biological activity of all the samples of fluids used for research purposes, as it was the case, was tested by their ability to induce zona pellucida resistance to proteolytic digestion (see Coy et al., 2008). We have included part of this information in Materials and methods and could include the one that this reviewer considers necessary to avoid any technical concern.

2) Quality controls and validation of RNA seq and WGBS are missing. This is critical given that authors worked with extremely low amount of material. (How was the quality of RNA or RNA amplification assessed?

RNA quality was assayed by Bioanalyzer and even though each sample came from asingle blastocyst, RIN score was acceptable, with values between 6.1-8.2. According with a recent report (Sigurgeirsson et al., 2014), “there does not seem to be any justification to set a threshold at any specific RIN but rather it is important to be aware of the effects of low RIN and all samples should preferably be in close range in terms of quality.” In fact, this report showed the biggest effect in differential expression by comparing samples with RIN 10 vs RIN8 (affecting 36% of all the expressed genes), while only 1% genes showed as differentially expressed when comparison was between samples with RIN 8 vs RIN6.

In addition, quality control plots were obtained from SeqMonk, which indicated similar output inall samples and low presence of ribosomal RNA (Author response image 1). Although high proportion of ribosomal RNA (rRNA) compared to mRNA represents a typical limitation for RNA-seq studies, the use in different species of Ovation RNA-Seqamplification system(Nugen), which was used for this experiment, have resulted in the low percentages of rRNA fragments (Tariq et al. NucleicAcids Res. 2011; Chitwood et al. BMC Genomics 2013).

(Where C means C-IVF embryos, N means Natur-IVF embryos and IV means in vivo embryos).

How many reads were obtained per sample?

These are the numbers for RNASeq and PBAT,respectively:

ID sampleRNA-seqMapped READS
C937155950
C946765974
C966781941
N276008873
N544615700
N604043748
IV1896431837
IV1906750663
IV1916278365
ID SamplePBATUnique alignments
C3413,150,508 (59.2%)
C3542,208,651 (54.6%)
C3634,002,302 (58.9%)
N8533,153,208 (58.5%)
N8640,676,107 (52.0%)
N5528,310,723 (54.5%)
IV18635,990,635 (53.3%)
IV19731,379,752 (56.9%)
IV19332,812,689 (53.5%)

Regarding the quality of the PBAT data, we are using a low-cell protocol that is now wellestablished in our lab and was optimised for the analysis of a similar number of oocytes (~100) as the number of cells in the individual pig blastocysts (see Stewart et al. 2015 Genes Dev.29:2449) and which has proven to be very robust (the highly stereotypical oocyte methylation landscape provides an excellent reference for the reproducibility of the low-cell PBAT method). The pig blastocyst PBAT libraries had slightly higher sequence duplication rates and lower overall complexity, which probably reflects a slightly lower number of cells the single embryos. The unique sequence alignment rate (table above) is somewhat lower than we typically obtain by mapping PBAT data from mouse or human samples (65-70%), and this likely reflects the gaps in the pig genome assembly.

Further, validation of some key genes in independent samples, by RT PCR and bisulfite sequencing is missing).

We had considered performing qPCR studies to ‘re-validate’ some of our gene- expression findings but there is little evidence that qPCR analyses from the same samples will add any extra utility to our data. Previous studies have shown extremely close correlations between qPCR and RNAseq data (Griffith M et al. (2010) Nat Methods 7: 843–847, Asmann YW, et al. (2009). BMC Genomics 10: 531, Wu AR, et al. (2014). Nat Methods 11: 41–46, Shi Y and He M (2014). Gene 538: 313–322). In contrast to microarrays, which always require qPCR validation, in RNAseq experiments probe bias, poor sensitivity and reduced linear range are not as problematic, since the entire transcript is assessed in a more or less unbiased manner (Wang Z, Gerstein M, Snyder M (2009) Nat Rev Genet 10: 57–63). In cases where there are discrepancies between RNAseq and qPCR, it would be most likely due to bias in the qPCR experiment (which has its own probe-bias based on what region of the cDNA is amplified). Ideally, we could re- validate our findings (potentially by qPCR) in a separate cohort of samples collected for a new experiment including new animals for the in vivo embryos (without guarantee of similarity with the ones already used), but we would consider to perform it only if reviewers consider it critical to publish our results.

The same reason would apply for PBAT DNA methylation analysis by sequencing.

3) Unsupervised clustering of the 9 samples for both RNA seq and WGBS of all genes (not only the statistically different genes) should be provided to truly confirm that samples cluster based on fertilization and culture methods.

For RNASeq, the PCA analysis gave the plot which has been included in the manuscript (Figure 3B).

For DNA methylation, we got the PCA plot which explains (32%) how in vivo and N-IVF are close each other and also shows the difference among the three C-IVF embryos (Figure 5A).

Reviewer #3:

[…] 1) Overinterpretation/misrepresentation of data.

A) The manuscript concluded that "reproductive fluids improve the outcome of IVF and the quality of pig blastocysts produced in vitro", and that "remarkable findings were that Natur-IVF blastocysts attained a more advanced developmental stage". These conclusions are not readily discernible from the data. In Table 2 and 3, C-IVF embryos produced a great percent of cleavage stage embryos and early blastocysts than Natur-IVF. Additionally, there was no significant difference between C-IVF and Natur-IVF in percent of blastocysts and expanded blastocyst formed or the yield. How does this show an improved outcome or advanced developmental stage? While there was a significant difference in hatching rate, it would appear that embryos were not left long enough to truly assess this, since 0 out of 903 C-IVF embryos and 48 out of 961 (5%) Natur-IVF had hatched.

We have modified the tables to make them clearer, because from the reviewer´scomments we extracted that we were not able to explain it properly. In Table 2, C-IVF embryos cleaved 5% more than Natur-IVF at day 2 (47.5 ± 1.6 vs 42.1 ± 1.6). However, this difference was not maintained through the development since the proportion of blastocysts in both groups was similar on day 7.5 (41.4 ± 2.4 vs 44.5 ± 2.5). Nonetheless, both percentages were higher than the best results previously described (Redel et al., 2016. Mol Reprod Dev), and it could be explained because oocytes in the C-IVF group also were incubated with oviductal fluid before IVF, as shown in the Figure 2. In addition, Natur-IVF group (whose embryos were in addition exposed to reproductive fluids during embryo culture) showed higher number of cell per blastocysts than C- IVF, (49.9 ± 3.7 vs 81.8 ± 7.2) and it was similar to in vivo embryos (87.0 ± 7.2). With these arguments we think that we can state that reproductive fluids improve the outcome of IVF and the quality of pig blastocysts produced in vitro.

In Table 3 (now Table 2.B) the higher percentage of early stage blastocysts in C-IVF group (57 out of 178) is due to the delay in their developmental kineticscompared to Natur-IVF group. The different blastocyst stages assessed in Table 3 (now Table 2.B) correspond with the total of blastocysts produced with each system (table 2.A) and, on day 7.5, only 12.8% of them (23 out of 180) were early blastocysts in the Natur-ART group (compared to 31.7% in C-IVF, 57 out of 178) because most of them had progressed to more advanced stages, such as blastocyst (55 embryos) expanded (65 embryos), hatching (28 embryos) or hatched (9 embryos). We thank you for thiscomment and have modified this Table adding information in order to clarify the data provided.

B) The authors bias the presentation of the transcriptome and methylome data towards finding less difference in the Natur-IVF group. The major questions that should be asked are whether there are expression/methylation changes between the three experimental groups and then where do these differences exist. Furthermore, what is the variation between embryos in the same group? The authors should present PCA graphs to visualize the three embryos and three groups. This should then be followed by pairwise comparisons for all groups, with VENN diagrams shown for all comparisons, including Natur-IVF and C-IVF. In fact, when the comparison is made (subsection “Natur-IVF blastocysts show fewer aberrantly expressed genes than C-IVF blastocysts”, second paragraph), only 29 differentially expressed genes were identified between Natur-IVF and C-IVF, demonstrating little difference in the cultured groups. Instead, as seen in Figure 1B, greater differences were found between the in vivo group and the two cultured groups.

Thank you for this comment. We have included PCA graphs to show how the three embryos from each group cluster together. As reviewer points, "the major questions that should be asked are whether there are expression/methylation changes between the three experimental groups and then where do these differences exist". That is exactly what we tried to do here. After checking the quality of the data and the clustering of the embryos (now we have added the PCA plot), we did pairwise comparisons and found that, for gene expression analysis, and despite of individual variability, both in vitro groups were more similar between them than any of them compared to in vivo. For this reason, we started analyzing differences between each in vitro group compared to the in vivo one (789 and 623 genes out of 19,638) but continued looking at differences between the two in vitro groups (29 genes out of 19,638). First conclusion was that all the groups were similar, having only 4%, 3.17% and 0.15% DEGs, respectively, for each pairwise comparison. However, we also found that the general tendency for the Natur-IVF DEGs was to show intermediate values between the in vivo and C-IVF ones. And, finally, despite the scarce differences, we found that some DEGs between C-IVF and Natur-IVF could be critical because the corresponding knock out or knock-down studies in mice showed phenotypes of altered/abnormal growth/size, reproduction/fertility, mortality/aging, hematopoietic system, homeostasis/metabolism and other abnormalities. Bearing in mind that the alteration of just one gene could have consequences in the normal development of an embryo, we thought it was important to highlight some of the most representative genes. Altogether, we think that our data "confirm that in vitro culture significantly alters embryonic gene expression”.

For DNA methylation, there does not appear to be a significant difference between the three groups for total methylation levels or methylation levels over specific genomic features (Table 4). Here again, C-IVF and Natur-IVF groups need to be compared to determine how much they differ from each other. Finally, while comparisons between the in vivo and each culture group produced a differ subset of genes affected, there does not appear to great differences in the percent of genes misregulated (gene expression: C-IVF 68% up-, 32% down-regulated; Natur-IVF 69% up- and 31% down-regulated; DNA methylation: C-IVF 57% more and 43% less methylated; Natur-IVF 66% more and 34% less methylated). Thus, from the present analysis, it does not appear that "Epigenetic and gene expression changes derived from assisted reproduction can be mitigated by reproductive secretions"(title).

In the context of demethylation that happens during early embryo development, withmean values of DNA methylation around 12% in vivo (12.33 ± 3.6) the observed differences (Natur-IVF: 11.09 ± 2.6 and C-IVF: 15.02 ± 3.3) could have a significant biological effect. In addition, specific DMRs, especially in the context of imprinting regions can also impact embryo development. We never hypothesized to find huge DNA methylation differences, because during the blastocyst stage there is an almost complete resetting of the methylome.

For DNA methylation, the differences were, indeed, subtle, as shown in Table 4, andless than4,000 DMR were found from the total of 258,885 probes created in the unbiased analysis, which represents 1.5%. Here again, C-IVF vs Natur-IVF groups were compared as well as C-IVF vs in vivo and Natur-IVF vs in vivo. We apologise for not having included the comparison between the two in vitro groups in Figure 5. Now it is included, with 3112 DMRs between both groups. Recognising these rare differences, we tried then to search for specific DMRs of obvious potential biological significance, thinking again of the impact that altered regulation of just a single gene could have for correct embryo development. For this reason, we first focused on the specific DMRs characterizing exclusively in each group (Figure 5C, D and E) and second on the imprinted igDMRs. We were aware that by extracting only the DMRs with a consistent 5% significant change in all the three embryos of one group compared to the other we could be missing other differences occurring only in one individual but we decided this approach as an attempt of collecting at least a first list of potential DMRs having the highest probability to occur. Since we are providing here the first datasets of single pig blastocysts methylome and transcriptome, there are now multiple options of different valid analyses in future studies.

2) In this manuscript, the authors produced single blastocyst profiling then pooled the data. This suggests that only genes affected in all three embryos would be analyzed. Previous DNA methylation analyses have shown that losses and gains of DNA methylation appear to occur stochastically. Thus, pooling of the data may mask gene differences in response to culture. Is there enough coverage to perform analyses on single embryos? If not, can differences be assessed with 2/3 embryos showing the same differences? If not, the authors need to acknowledge that they have analyzed only the most common changes.

Thank you for the comment, we agree with you. We have indeed indicated that wechose the option of analyzing only the most common changes in this study, as above explained.

3) The authors performed both transcriptome and methyome analyzes. Have they mapped the two data sets to determine whether DNA methylation changes are correlated with gene expression changed in neighboring genes?

Yes, we did it and we did not found significant correlation between both changes.Although DNA methylation at the promoter/gene bodies is directly/indirectly correlated with gene expression, this is not strictly true during the periods of dramatic loss of DNA methylation, as during early embryo development or primordial germ cells (PGC) formation. For example, Gounktela et al., 2015 showed that during demethylation of PGCs there is a general uncoupling between DNA methylation and gene expression at some stages; literally they say: “Our data reveal a remarkable and pervasive loss of DNA methylation in human PGCs and AGCs during prenatal life that has almost no relationship to changes in gene expression”. We did correlation analysis and saw this lack of correlation between our methylation and gene expression data. In our opinion, at this stage of development and with this low level of methylation, this is not a surprising result.

[Editors' note: the author responses to the re-review follow.]

[…] Reviewer #1:

The resubmission of the work by Canova et al. has responded to the majority of comments. However there still remain several concerns. Overall the style of the new version is more difficult to read and will benefit from significant editing both from a language and spelling point of view.

We acknowledge the encouraging comments from the reviewer and they have been very useful to improve the manuscript. We have modified it according to his/her suggestions and those from the other reviewers. An extensive review of the grammar and spelling has been done by Dr. Gavin Kelsey.

1) In particular the standard operating procedure utilized to obtain the physiologic follicular and uterine fluid from pigs and how the media was stored and controlled for contamination needs to be described in Materials and methods. Referring to other manuscripts is not helpful for readers. For example, the Carrasco 2008 manuscript is not even mentioned in the manuscript. The Coy 2008 paper includes only a short Methods section with a short description of the methods to collect samples from pigs, with no indication on how many pigs were pooled or the range of their age. As an example of lack of clarity, based on the Coy paper: for how many hours are the fallopian tubes at room temperature from sacrificing the animal to collecting the fluid? This is critical to the reader, as quality of the fluid could be compromised or lost during manipulation/transport at room temperature. More Details, including some included in the rebuttal to authors need to be included in Materials and methods section. Given the novelty of the method, this is absolutely needed.

We apologize for the lack of clarity about the operating procedure used to obtain the follicular, oviductal and uterine fluids from pigs. We tried to summarize the more relevant information and to avoid an extensive Materials and methods section. However, following reviewer´s comment we have now included detailed information in this section:

“Collection and processing of follicular, oviductal and uterine fluids

Fluids were obtained from animals raised at a commercial farm (CEFU, S.A., Murcia, Spain) and slaughtered in an abattoir belonging to a food industry (El Pozo, S.A) near the University of Murcia. […] One ml follicular fluid (FF) aliquots were stored at -80°C until their use as additives for the IVM medium.”

This is the standard protocol used for porcine IVM in the last 25 years in our laboratory following Naito et al. indications (Naito et al., Gamete Res. 1988 21(3):289-95) and has been proven to provide a beneficial microenvironment for further development of the immature oocytes (e.g., Ito et al., 2008 and Bijttebier et al., 2008).

For the collection of oviductal (OF) and uterine (UF) fluids, genital tracts from cyclic Large White sows (2-4 years old) were obtained at the abattoir and transported to the laboratory on ice within 30 min of slaughter. The cyclic stage of animals was assessed once in the laboratory, on the basis of ovarian morphology on both ovaries from the same female. Oviducts and uteri were classified as early follicular, late follicular, early luteal or late luteal phase (Carrasco et al., 2008). Both oviducts and uteri coming from the same genital tract were classified as in the same stage of the cycle. Once classified, oviducts and uteri were separated and quickly washed once with 0.4% v/v cetrimide solution and twice in saline. Oviducts and uteri were dissected on Petri dishes or trays, respectively, sitting on ice. Once dissected, OF were collected by aspiration with an automatic pipette by introducing a 200µl pipette tip into the ampulla and manually making an increasing pressure gradient from the isthmus to the ampulla. The UF was collected by making a manual increasing pressure gradient from the proximal end to the distal end (utero-tubal junction) of the uterine horn and letting the fluid drop into a sterile 50 ml Falcon tube. Once recovered, samples (OF and UF) were centrifuged twice at 7000 g for 10 min at 4°C to remove cellular debris. Then the supernatant was immediately stored at -80°C until use. Oviducts from animals at the late follicular phase (POF-LF) and at the early luteal phase (POF-EL) gave a mean volume of around 50µl and 40µl, respectively per oviduct. At the early luteal phase, approximately 10 ml of UF per uterine horn were collected each time. Aliquots of 50µl OF and 50ml UF of pooled samples from at least 20 animals for OF and 5 animals for UF were used. Only samples that passed quality controls (pH 7.0-7.6, osmolality 280-320 mOsm/kg, endotoxin <0.10 EU/mL, a minimum 90% of Metaphase II oocytes after IVM with FF and ZP hardening for oocyte preincubation in POF-LF > 1 hour) were used for experiments.

2) The fact that adding only 1% of oviduct fluid can obtain such remarkable results is impressive and this finding is surprisingly missing in the Discussion. If the authors were to add 5% or 10% of OF, would the result change? How did the authors come to the 1% dose? This needs to be commented on in the Discussion.

We acknowledge the encouraging comment from the reviewer.

The use of 1% of fluid was established after testing a large number of experimental conditions in our laboratory where cleavage rate, monospermy, morphology and blastocyst development were assayed (for example, in Ballester et al., 2014 we published the first IVF results adding 1% OF to the IVF medium). Similarly, our collaborator Dr. Rizos, after testing different OF concentrations in the cow model, published that OF concentrations (1.25% and 0.625%) supported embryo development until Day 9 (27.5%) and produced higher-quality blastocysts (Lopera-Vasquez et al., 2015).

We agree that this finding is significant and it has been included in the Discussion.

3) A statement in the Discussion that validation of the RNA and DNA methylation results in independent samples was not performed need to be added as an important limitation of the paper.

Following this request, we have introduced the following paragraph in the Discussion section:

“Considering that previous studies have shown extremely close correlations between qPCR and RNA-seq data (Griffith M et al. (2010) Nat Methods 7: 843–847; Asmann YW et al. (2009) BMC Genomics 10: 531; Wu AR, et al. (2014). Nat Methods 11: 41–46) and that validation by qPCR has its own probe-bias based on what region of the cDNA is amplified, we deem, in contrast to microarrays data, that there is not solid evidence that validation of the RNA-Seq and DNA methylation results by qPCR will provide extra significance to our results. For this reason, we did not perform qPCR validation in this study”.

4) Please confirm and describe in the Methods section that the 3 biological replicates in the nature IVF group were generated from the addition of 3 independent sets of uterine and oviductal fluids (POF_LF, OF-EL; UF-EL), collected from different animals. This is a key experimental factor that is not described in the Methods section. In fact, adding the same OF to 3 different blastocysts would be expected to lead to similar/ consistent transcriptomic and epigenetic results in blastocysts. If this was not done, data from independent fluid donors should be provided.

Thank you for the comment. It has helped us to clarify in the manuscript that the observed effect in the Natur-IVF group is not the result of adding reproductive fluid from one specific animal, which would be not representative. As described in the new Materials and methods section reproductive fluids were pooled from different animals (at least 25 females for FF, 20 animals for OF and 5 animals for UF), collected simultaneously to form a large batch/lot of fluid that was homogenous for the different experiments. This experimental design aims to follow the very well established and routine practice for cell cultures or IVF-EC assays that promotes the use of serum from the same lot/batch for a set of experiments. Otherwise, assuming that some variability could exist between animals (although very high similarity exists in age, health status, weight, nutrition, etc.) individual effect and treatment effect could be masked or confused.

We had to use different batches of fluids for the oocyte preincubation step during the in vitro production of embryos because of the volume of fluid needed (see Table 2 where 961 oocytes were used for the Natur-IVF group which means 961 microliters of OF just for the preincubation step). But we used the same batch of fluids for the remaining steps and for the 6 blastocysts used for the DNA and gene expression analysis because to have done otherwise would not have been correct in our opinion. It would have been the same as using different batches of BSA for the C-IVF group or different ages or breeds of animals for the in vivo group. In all the ART laboratories one of the main objectives is the use of standardized protocols and, for example when serum is added, the same batch is used for long periods of time.

We do not consider that data from independent fluid donors can be useful and representative for these experiments. The use of big homogenous batches of follicular fluids is the standard protocol used for porcine IVM in the last 25 years in our laboratory and other groups around the world, following Naito et al. indications (Naito et al., Gamete Res. 1988 21(3):289-95) and it has been demonstrated to provide a beneficial microenvironment for further development of the immature oocytes (e.g., Ito et al., 2008 and Bijttebier et al., 2008). Similarly, the use of large and homogenous batches of OF has been proved to be consistently beneficial by our group when added during IVF (Ballester et al., 2014, Fertility & Sterility 102(6): 1762-1768). Finally, the addition of UF, despite being used in this article for the first time, has been checked in our lab with different batches and always has shown an improvement compared with conventional protocols (e.g. BSA).

Reviewer #2:

It was not possible to review the manuscript in much depth as was the case for the original submission. However, in general, the authors appear to have responded well to lengthy criticisms of three reviewers. The interpretations are more restricted. This reviewer continues to doubt that the future of the field of human ART lies in collecting biological fluids from humans or animals but rather that artificial fluids mimicking natural fluids more and more closely are the answer. However, the authors are entitled to their opinion. In general, the manuscript is an excellent contribution to the knowledge in this field.

We really appreciate the reviewer´s comments, including the scepticism about the future of the field of human ART. Only the future will reveal the right answer. We are glad to see he/she recognizes our work as “an excellent contribution to the knowledge in this field”.

https://doi.org/10.7554/eLife.23670.027

Article and author information

Author details

  1. Sebastian Canovas

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    SC, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4190-6258
  2. Elena Ivanova

    Epigenetics Programme, The Babraham Institute, Cambridge, United Kingdom
    Contribution
    EI, Data curation, Validation, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  3. Raquel Romar

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    RR, Data curation, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  4. Soledad García-Martínez

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    SG-M, Data curation, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  5. Cristina Soriano-Úbeda

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    CS-Ú, Data curation, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  6. Francisco A García-Vázquez

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    FAG-V, Data curation, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  7. Heba Saadeh

    1. Epigenetics Programme, The Babraham Institute, Cambridge, United Kingdom
    2. Bioinformatics Group, The Babraham Institute, Cambridge, United Kingdom
    Present address
    Computer Science Department,KASIT, The University of Jordan, Amman, Jordan
    Contribution
    HS, Data curation, Methodology
    Competing interests
    The authors declare that no competing interests exist.
  8. Simon Andrews

    Bioinformatics Group, The Babraham Institute, Cambridge, United Kingdom
    Contribution
    SA, Software, Formal analysis, Validation, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  9. Gavin Kelsey

    1. Epigenetics Programme, The Babraham Institute, Cambridge, United Kingdom
    2. Centre for Trophoblast Research, University of Cambridge, Cambridge, United Kingdom
    Contribution
    GK, Supervision, Funding acquisition, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9762-5634
  10. Pilar Coy

    1. Physiology of Reproduction Group, Departamento de Fisiología, Facultad de Veterinaria, Universidad de Murcia-Campus Mare Nostrum, Murcia, Spain
    2. Instituto Murciano de Investigación Biosanitaria, Murcia, Spain
    Contribution
    PC, Conceptualization, Data curation, Supervision, Funding acquisition, Investigation, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    pcoy@um.es
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3943-1890

Funding

Research Councils UK

  • Gavin Kelsey

Ministerio de Economía y Competitividad (AGL2012-40180-C03-01)

  • Pilar Coy

Ministerio de Educación, Cultura y Deporte (PRX14/00348)

  • Pilar Coy

Fundación Séneca. Región de Murcia. Spain (20040/GERM/16)

  • Pilar Coy

Ministerio de Economía y Competitividad (AGL2015-66341-R)

  • Pilar Coy

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank CEFU, SA and El Pozo, SA for providing the biological material; Juan Antonio Carvajal and Soledad Rodriguez for collecting the oviducts, uteri and ovaries at the slaughterhouse; Carmen Matás for technical support with IVF and Kristina Tabbada for sequencing RNA-Seq and BS-seq libraries. Funding: Work in GK’s laboratory was supported by the UK Biotechnology and Biological Sciences Research Council and Medical Research Council. Work in PC’s laboratory was supported by grants AGL2012–40180 C03-01 and AGL2015–66341-R from the Ministry of Economy and Competitiveness (Spain), and 20040/GERM/16 from Fundación Séneca. PC stay at The Babraham Institute was funded by a mobility grant of the Spanish Ministry of Education, Culture and Sports (PRX14/00348).

Ethics

Animal experimentation: This study was carried out in strict accordance with the recommendations in the Guiding Principles for the Care and Use of Animals (DHEW Publication, NIH, 80-23). The protocol was approved by the Ethical Committee for Experimentation with Animals of the University of Murcia, Spain (Project Code: 192/2015).

Reviewing Editor

  1. Jessica K Tyler, Weill Cornell Medicine, United States

Publication history

  1. Received: November 26, 2016
  2. Accepted: January 28, 2017
  3. Accepted Manuscript published: January 30, 2017 (version 1)
  4. Accepted Manuscript updated: February 1, 2017 (version 2)
  5. Version of Record published: March 7, 2017 (version 3)
  6. Version of Record updated: April 11, 2017 (version 4)

Copyright

© 2017, Canovas et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,113
    Page views
  • 624
    Downloads
  • 12
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Developmental Biology
    2. Stem Cells and Regenerative Medicine
    Anna Yoney et al.
    Research Article Updated
    1. Developmental Biology
    2. Evolutionary Biology
    Francesca Caroti et al.
    Research Article Updated