Mouse Sox17 haploinsufficiency leads to female subfertility due to impaired implantation

Embryonic implantation comprises a dynamic and complicated series of events, which takes place only when the maternal uterine endometrium is in a receptive state. Blastocysts reaching the uterus communicate with the uterine endometrium to implant within a narrow time window. Interplay among various signalling molecules and transcription factors under the control of ovarian hormones is necessary for successful establishment of pregnancy. However, the molecular mechanisms that allow embryonic implantation in the receptive endometrium are still largely unknown. Here, we show that Sry-related HMG box gene-17 (Sox17) heterozygous mutant female mice exhibit subfertility due to implantation failure. Sox17 was expressed in the oviduct, uterine luminal epithelium, and blood vessels. Sox17 heterozygosity caused no appreciable defects in ovulation, fertilisation, blastocyst formation, and gross morphology of the oviduct and uterus. Another group F Sox transcription factor, Sox7, was also expressed in the uterine luminal and glandular epithelium relatively weakly. Despite uterine Sox7 expression, a significant reduction in the number of implantation sites was observed in Sox17 heterozygous mutant females due to haploinsufficiency. Our findings revealed a novel role of Sox17 in uterine receptivity to embryo implantation.


Results
Sox17 was expressed in female reproductive organs. To examine Sox17 expression in female reproductive organs, we first used Sox17-green fluorescent protein (GFP) knock-in mice, which express GFP under the control of the Sox17 promoter 9 . In nonpregnant, Sox17-GFP heterozygous mutant females (Sox17 +/GFP ), GFP fluorescence was observed in the oviduct and uterine endometrium (Fig. 1b-d), whereas wild-type (WT) female mice showed no discernible fluorescence in these organs under the same conditions of image acquisition (Fig. 1a). In the uterine endometrium, both the luminal and glandular epithelium expressed GFP (Fig. 1d,e). Blood vessels also expressed GFP (Fig. 1f). Uterine expression of GFP was stronger than that observed in the blood vessels.
To examine whether uterine expression of GFP in Sox17 +/GFP mice coincided with that of endogenous Sox17, we performed co-immunofluorescent staining using anti-Sox17 and anti-GFP antibodies. In the WT uterus, Sox17 staining was detected in the luminal and glandular epithelium (Fig. 1g-h"). In the Sox17 +/GFP uterus, GFP staining coincided with Sox17 staining (Fig. 1i-j"), confirming the reliability of Sox17-GFP for monitoring Sox17 expression. Fluorescence emitted from GFP without staining was also detectable ( Supplementary  Fig. S1). However, GFP staining was more distinct and clearly showed GFP-positive tubular structures in the stroma (Fig. 1j-j"). These Sox17-GFP-positive cells were likely to be capillary blood vessels. These microscopic analyses suggest that Sox17 was expressed in the oviduct and uterine luminal epithelium, with weak expression in the blood vessels of nonpregnant, female mice. Importantly, Sox17 +/GFP mice had no appreciable defects in uterine morphology, including the blood vessels. (a-f) Fluorescence emitted from GFP without staining in nonpregnant, WT and Sox17 +/GFP female reproductive organs. (g-j") Immunofluorescent staining of GFP (green) and Sox17 (red) in the WT (g-h") and Sox17 +/GFP (i-j") uterus. Nuclei are stained with Hoechst (blue). Enlarged images corresponding to dotted boxes in (g) and (i) are shown in (h-h") and (j-j"), respectively. Abbreviations: ov, ovary; od, oviduct; ut, uterus; le, luminal epithelium; ge, glandular epithelium; s, stroma; and bv, blood vessels. Scale bars, 500 μm for (a-c); 200 μm for (d-f); 100 μm for (g,i); and 50 μm for (h",j").
Next, we examined the expression of Sox17 together with other Sox-F group transcription factors, Sox7 and Sox18, in the oviduct, uterus, and uterine tissue fractions at day-of-pregnancy (DOP) 4 in natural pregnancy. Here, DOP 1 was the day at which the vaginal plug was formed, and DOP 4 was the day at which the uterus became receptive to implantation. First, we performed quantitative reverse transcription polymerase chain reaction (RT-PCR), in which Actb was used as an internal control, and the uterus was used as a calibrator (Fig. 2a-c). Similar results were also obtained using Gapdh as an internal control ( Supplementary Fig. S2). In the oviduct, Sox17, Sox7, and a lesser amount of Sox18 transcripts were detected; the expression levels were 1.7-, 1.9-, and 0.3-fold higher, respectively, than those detected in the uterus (n = 5; Fig. 2a,b). In the uterine luminal epithelium, Sox17 and Sox7 transcripts were abundant, exhibiting 6.3-and 5.8-fold higher expression, respectively, than that observed in the uterus (n = 5; Fig. 2a,b). In contrast, Sox18 transcript levels in the luminal epithelium were negligible (Fig. 2c). Moreover, Sox18 transcripts were undetectable in one of the five samples from the luminal epithelium.
Next, we performed western blotting to examine Sox17 and Sox7 protein levels. Proper preparation of the luminal epithelium was ensured by absence of smooth muscle actin (Fig. 2d). Similar to the PCR analysis, the blot was quantified by normalisation with β -actin and divided by the value from the uterus (i.e., uterine expression = 1.0). In the quantification, the amounts of Sox17 and Sox7 proteins were 6.1-and 1.8-fold higher than those observed in the uterus. These data implied that, among the Sox-F proteins, Sox17 was mainly expressed in the luminal epithelium. To examine the tissues expressing Sox17 and Sox7 in the oviduct and uterus, we conducted immunohistochemistry ( Fig. 2e-l'). In the oviduct, staining for both Sox17 and Sox7 was clearly detected at a comparable level (Fig. 2e,f). In the nonpregnant WT uterus, staining for Sox17 and Sox7 was found in the luminal and glandular epithelium (Fig. 2g,h). Consistent with their function as transcription factors, the staining for Sox17 and Sox7 was localised exclusively in the nuclei (insets in Fig. 2g,h). Similar patterns of Sox17 and Sox7 staining were also observed at the implantation site in the DOP 4 uteri after natural mating (Fig. 2i,j, and Supplementary  Fig. S1) and in a delayed implantation model (Fig. 2k,l). The latter was used to ensure the timing of implantation. Consistent with a previous report 13 , Sox17 showed moderately elevated expression at the embryo attachment site (Fig. 2k,k'). These data suggested that Sox17 and Sox7 were expressed in the oviduct and uterine luminal epithelium receptive to implantation.
Sox17 heterozygous mutant females exhibit subfertility. While breeding Sox17-GFP mice, we noticed that the number of pups from Sox17 +/GFP female and WT male pairs was fewer than that from WT female and Sox17 +/GFP male pairs. For quantitative analysis, we counted the cumulative number of pups while breeding Sox17-GFP mice for 12 months (Fig. 3a). WT female and Sox17 +/GFP male pairs produced an average of 105.0 ± 14.3 pups (n = 4), whereas Sox17 +/GFP female and WT male pairs produced an average of 16.4 ± 8.8 pups (n = 5). These numbers include the pups that died soon after birth. The difference was statistically significant as determined by t-tests (p = 0.0141). Notably, Sox17 +/GFP females had a smaller litter size and tended to become infertile early in adulthood.
Since Sox17 haploinsufficiency results in perinatal lethality in C57BL/6 background 14 , we also examined the genotype of the pups after weaning. WT to + /GFP ratio was 161:139 for the pups produced by WT females and 39:37 for the pups produced by Sox17 +/GFP females. Chi-square test revealed that WT and + /GFP pups were born approximately at the expected Mendelian ratio of 1:1 for both the mating groups (p = 0.204 and 0. 815 from WT and Sox17 +/GFP females, respectively), indicating that perinatal lethality of Sox17 +/GFP pups was not a major reason for the subfertility observed in Sox17 +/GFP females.
Subsequently, to examine the influence of Sox17 heterozygosity on the ovulation, fertilisation, and early development, we collected 2-cell stage embryos from superovulated, WT and Sox17 +/GFP females and found that the number of the embryos was comparable (Fig. 3b). These 2-cell stage embryos were further cultured in vitro and formed blastocysts at a similar rate (Fig. 3c). These data suggested that ovulation, fertilisation and embryonic development to the blastocyst stage were phenotypically normal and were not affected by Sox17 heterozygosity.
Sox17 heterozygous mutant females were defective in implantation. To investigate when the embryos were lost in Sox17 +/GFP females, we counted the number of implantation sites at DOP 5, which was the day blastocysts were implanted into the uterus. WT females showed an average of 8.3 ± 0.3 (n = 15) implantation sites (Fig. 4a,c). In contrast, Sox17 +/GFP females showed an average of 1.5 ± 1.2 (n = 4) implantation sites (Fig. 4b,c). The difference was statistically significant (p < 0.0001 by t-tests). Notably, unimplanted blastocysts were recovered from the Sox17 +/GFP uteri with no implantation sites (Fig. 4b'), clearly indicating the normal function of the oviduct and normal development of the blastocyst. Since GFP expressed in Sox17 +/GFP female mice may exert adverse effects on the number of implantation sites, we also counted the number of implantation sites using Sox17 point-mutant mice (SHIVA) 15 , with a point mutation that alters the 72 nd Met to Ala in the DNA-binding region of the HMG domain. Sox17 +/SHIVA female mice showed an average of 3.7 ± 3.7 implantation sites (n = 3; p = 0.0077 by t-tests; Fig. 4c), confirming that GFP expression was not responsible for the implantation failure observed in Sox17 +/GFP . These results indicated that implantation was defective in Sox17 heterozygous mutant females.
To directly demonstrate the failure in implantation, we performed embryo transfer and examined the implantation rate. In embryo transfer, we used WT embryos to exclude the possibility that the implantation failure was caused by problems with the embryo. In addition, the implantation site was observed at DOP 6 or 7 to examine whether the smaller number of implantation sites observed at DOP 5 in Sox17 heterozygous mutant females could be explained by factors other than a delay in implantation.
First, we transferred WT embryos into the oviduct. WT female mice showed an average implantation rate of 37.4% ± 11.3% (n = 9; Fig. 4d,h). In contrast, Sox17 +/GFP female mice showed a much lower average implantation rate (1.7% ± 1.7%, n = 6; Fig. 4e,h). The difference was statistically significant (p = 0.0273 by t-tests). Sox17 +/SHIVA female mice showed no implantation sites (n = 4; Fig. 4h). We also transferred approximately a 1:1 mixture of WT and Sox17 +/GFP embryos to WT female mice to examine whether Sox17 +/GFP embryos could be implanted successfully. Female mice with transferred embryos showed an average implantation rate of 68.8% ± 6.3% (n = 2). This number was comparable to the results of WT embryo transfer, suggesting that Sox17 heterozygosity in embryos was not responsible for implantation failure. This was also supported by the normal number of pups from the mating of WT female and Sox17 +/GFP male mice (Fig. 3a).
Secondly, we transferred WT embryos directly to the uterus. WT females showed an average implantation rate of 54.3% ± 10.9% (n = 11; Fig. 4f,i), whereas Sox17 +/GFP females showed a much lower average implantation rate (4.4% ± 4.4%, n = 3; Fig. 4g,i). This difference was statistically significant (p = 0.0394 by t-tests). These results suggested that the smaller number of implantation sites was not due to delays in implantation and further suggested that embryonic implantation into the uterus was defective in Sox17 heterozygous mutant females.

Discussion
In this study, we showed expression of Sox17 in the oviduct and uterine luminal and glandular epithelium, female subfertility, and decreased numbers of implantation sites in Sox17 heterozygous mutant females. In addition to Sox17, we also found expression of Sox7 in the uterine endometrium. However, the fact that most of the Sox17 heterozygous mutant females showed subfertility and implantation failure and that Sox17 protein was predominant in the luminal epithelium led to the conclusion that Sox17 was a major player in embryonic implantation among Sox-F proteins.
The observed implantation failure may have been caused by haploinsufficiency of the Sox17 gene. Sox17 is known to exhibit haploinsufficiency in bile duct formation, which results in biliary atresia and hepatitis in C57BL/6 background mice 14 . This Sox17 haploinsufficiency may result from lack of redundancy by other Sox-F proteins in the gallbladder and bile duct epithelia. It seems that Sox17 exhibits haploinsufficiency in implantation in a similar manner. However, we also observed that some Sox17 heterozygous mutant females had relatively normal litter sizes in mating and numbers of implantation sites. Functional redundancy among the Sox-F transcription factors has been reported in various events, namely, early cardiovascular development 6,7 , postnatal angiogenesis 5 , and haematopoiesis 9,10 in mice. In addition, Sox7 and Sox17 modify the Sox18-null phenotype in the lymphatic vasculature in a strain-specific manner 16 . In a similar way, it is possible that compensatory increase in Sox7 and/or Sox18 expression may occur in Sox17 heterozygous mutant females depending on the genetic background. Because we used a hybrid of C57BL/6 and ICR mice in this study, there may be individual differences in genetic background; this may be one reason for the presence of some outliers among the mutants.
Role of Sox17 in implantation remained elusive. However, there are some clues to how Sox17 may function during implantation. ChIP-Seq analysis revealed Sox17 as a direct transcriptional target of the progesterone receptor in the mouse uterus 17 , suggesting the role of Sox17 as a mediator of progesterone functions in early pregnancy. In addition to the progesterone, Leukaemia inhibitory factor (LIF) is an essential factor for implantation 18 . LIF regulates expression of a number of genes critical for implantation, such as the muscle segment homeobox (Msh) gene family member Msx1 and Indian hedgehog [19][20][21][22] . Sox17 is also reported as one of the genes that are upregulated after LIF treatment 21 . Therefore, progesterone together with LIF is likely to induce Sox17 expression at the timing of implantation. Intriguingly, Sox17 showed moderately enhanced expression at the blastocyst attachment site (Fig. 2i,k' and Wallingford et al. 13 ), implicating the direct role of Sox17 in blastocyst attachment. Of interest, Wnt/β -catenin signalling is also active at the implantation site 23 . Previous studies have reported that Wnt/β -catenin signalling induces Sox17 24,25 ; moreover, interaction of β -catenin with Sox17 may enhance the activity of Sox17 for induction of its target genes 26 . Therefore, expression of Sox17 together with Wnt/β -catenin signalling may be critical for embryonic implantation.
The mechanism of embryonic implantation is diverse across mammalian species, including humans and mice [27][28][29] . For example, mice produce many offspring in a single birth, whereas humans produce a single offspring in a single birth; additionally mouse implantation is eccentric (i.e., the blastocyst lies in a uterine crypt), whereas human implantation is interstitial (i.e., the blastocyst is completely embedded within the endometrium) 27 . However, availability of sophisticated genetic engineering has made the mouse an attractive model for human implantation and contributed to reveal the molecular basis of implantation, including elucidation of LIF as a critical factor for implantation 18,30 . Evidence from human studies suggests that Sox17 may also play a role in implantation. In a public database "The Human Protein Atlas" 31,32 , RNA expression of SOX17 in the endometrium is reported at a medium level (19 Fragments Per Kilobase gene model and Million reads [FPKM] in RNA-Seq), whereas RNA expressions of SOX7 and SOX18 in the endometrium are at low levels (4 and 2 FPKM, respectively). Therefore, it is suggested that SOX17 is the principal Sox-F protein expressed in the human uterus.
Another line of evidence implicates the involvement of SOX17 in successful pregnancy in humans. In exogenously administrated hormone-stimulated cycles aimed to conduct in vitro fertilisation and embryo transfer in humans, advanced endometrial maturation in histological dating compared to the expected chronological date is usually observed, in which, advanced maturation exceeding 3 days never results in successful pregnancy 33,34 . Microarray analysis revealed that SOX17 is one of the genes upregulated in the endometrium with 2-3 days advanced maturation (i.e., possible to achieve pregnancy) compared to those with 4 days advanced maturation (i.e., unable to achieve pregnancy) 35 , suggesting the relevance of SOX17 to pregnancy.
In this study, we found a novel role of uterine Sox17 in embryonic implantation by investigating both bulk knockout (Sox17 +/GFP ) and a point mutant (Sox17 +/SHIVA ). Further studies of changes in the localisation, expression, and molecular interactions of uterine Sox17 during the estrus cycle and pregnancy and the use of conditional knockout in the luminal epithelium will lead to clarification of the role of uterine Sox17 in implantation and ultimately contribute to improve infertility treatment in humans.

Methods
Animals and ethical statement. Sox17 GFP knock-in mice (Sox17 tm1Sjm ) 9 , and mutant mice carrying a point mutation in the Sox17 locus (SHIVA) 15 were used in this study. Sox17 +/GFP mice in the C57BL/6 background are also known to exhibit haploinsufficiency in bile duct formation, which leads to perinatal death due to biliary atresia and hepatitis 14 . For this reason, we used Sox17 GFP knock-in mice in a mixed background of C57BL/6 and ICR. Mice were housed in environmentally controlled, specific pathogen-free rooms in the Center for Experimental Animals of Tokyo Medical and Dental University (TMDU). All experiments were carried out in accordance with the approved guidelines by the institutional committees for animal and recombinant DNA experiments at TMDU. All experimental protocols were approved by the Institutional Animal Care and Use Committee of TMDU (Nos 0130082C, 0140007A, 0150259C2, and 0160024C2).
Superovulation and in vitro culture of embryos. Female mice of 8-to 12-week old were superovulated according to the standard procedure 36 . Pregnant mare serum gonadotropin (PMSG) and human chorionic gonadotropin (hCG) were obtained from ASKA pharmaceutical (Tokyo, Japan) and used at a concentration of 7.5 IU, each. Two-cell stage embryos were collected according to the standard procedure 36 , and cultured at 37 °C, 5% CO 2 in a drop of KSOM medium (ARK Resource, Kumamoto, Japan) covered with a layer of liquid paraffin (Nacalai Tesque, Kyoto, Japan). transferred to the oviduct 36 . Alternatively, embryos at the morula or blastocyst stage were transferred to the uterus at DOP 3 36 . The implantation rate was calculated as the number of implantation sites divided by the number of transferred embryos × 100 (%).

Implantation site counting.
Delayed implantation. Delayed implantation was performed as previously described by Ma et al. 37 . Embryo transfer and ovariectomy were conducted at DOP 2. Progesterone (P4) (Sigma-Aldrich) was administrated at DOP 5 and 6. At DOP 7, both P4 and estrogen (E2) (Sigma-Aldrich) was administrated to induce implantation. The uteri were fixed at 6 h after P4 + E2 administration.
Preparation of the luminal epithelium. The luminal epithelium at DOP 4 was prepared by incubating the uteri in Hank's balanced salt solution with pancreatin (Sigma-Aldrich) and dispase (Thermo Fisher, MA, USA) at 4 °C for 1 h and successively at room temperature for 1 h.
Quantitative RT-PCR. Total RNA was prepared from the naturally mated C57BL/6 female mice at 18:00-20:00 on DOP 4 using TRIzol ® in combination with a PureLink ® RNA mini kit and on-column DNase digestion (Thermo Fisher). Reverse transcription was performed using a SuperScript ® III first-strand synthesis system for RT-PCR (Thermo Fisher) with random hexamers. Real-time PCR was performed by StepOne ™ with TaqMan ® Fast Advanced Master Mix (Thermo Fisher). The TaqMan ® gene expression assays used in this study included Sox17 (Mm00488363_m1), Sox7 (Mm00776876_m1), Sox18 (Mm00656049_gH), Actb (Mm02619580_g1), and Gapdh (Mm99999915_g1). Data were analysed by the comparative C T (ΔΔC T ) method, in which Actb or Gapdh served as an internal control, and uterus data were used as a calibrator 38 . We confirmed that the target genes and the internal control genes had similar PCR efficiencies, which were close to 100% (Supplementary Fig. S2).
Statistical analysis. All statistical analyses were conducted using Prism 6 software (GraphPad Software, CA, USA). For all analyses, p < 0.05 was considered statistically significant. In implantation site counting, females with no implantation site due to failure in ovulation, fertilisation, and embryonic development were excluded from the statistical analysis.