Background Successful human reproduction requires normal spermatogenesis, oogenesis, fertilisation and early embryonic development, and abnormalities in any of these processes will result in infertility. Early embryonic arrest is commonly observed in infertile patients with recurrent failure of assisted reproductive technology (ART). However, the genetic basis for early embryonic arrest is largely unknown.
Objective We aim to identify genetic causes of infertile patients characterised by early embryonic arrest.
Methods We pursued exome sequencing in a proband with embryonic arrest from the consanguineous family. We further screened candidate genes in a cohort of 496 individuals diagnosed with early embryonic arrest by Sanger sequencing. Effects of mutations were investigated in HeLa cells, oocytes and embryos.
Results We identified five independent individuals carrying biallelic mutations in NLRP2. We also found three individuals from two families carrying biallelic mutations in NLRP5. These mutations in NLRP2 and NLRP5 caused decreased protein expression in vitro and in oocytes and embryos.
Conclusions NLRP2 and NLRP5 are novel mutant genes responsible for human early embryonic arrest. This finding provides additional potential diagnostic markers for patients with recurrent failure of ART and helps us to better understand the genetic basis of female infertility characterised by early embryonic arrest.
- female infertility
- embryonic arrest
- reproductive medicine
Statistics from Altmetric.com
Successful human reproduction requires normal oocyte maturation, fertilisation and embryonic development, and abnormalities in these processes will lead to infertility, which is estimated to affect more than 186 million people worldwide.1 Assisted reproductive technology (ART) is an effective treatment for infertility in which mature oocytes are fertilised with sperm through in vitro fertilisation (IVF) and intracytoplasmic sperm injection (ICSI). After fertilisation, the zygote goes through several steps of cleavage and differentiation and is then implanted to establish pregnancy.2 Despite the high success rate of ART, a number of couples experience recurrent failure of IVF/ICSI attempts. The recurrent failure IVF/ICSI were mainly caused by abnormality in spermatogenesis, oocyte maturation, fertilisation failure and early embryonic arrest. Genetic reasons of abnormal spermatogenesis have been widely investigated, and several mutant genes have been identified.3–5 However, genetics causes for abnormality in oocyte and early embryonic development remains to be elucidated. Until now, there are only few genes identified responsible for abnormal in oocytes and embryonic development.6–9
Maternal RNA and protein accumulate in the early stage of oocytes before ovulation. After fertilisation, a number of maternal RNAs and proteins are degraded, and the embryonic genome is activated, after which embryonic division begins and the embryo develops through the morula and blastocyst stages. Maternal factors play an essential role in maintaining normal early embryonic development through the degradation of maternal RNA and protein and the activation of the embryonic genome.10 11 A group of maternal factors called the subcortical maternal complex (SCMC) consists of at least six proteins, including transducin-like enhancer of split 6 (TLE6; MIM: 612399), KH domain containing 3 like (KHDC3L; MIM: 611687), protein-arginine deiminase type 6 (PADI6; MIM: 610363), oocyte-expressed protein homologue (MIM: 611689), NLR family pyrin domain containing 5 (NLRP5; MIM: 609658) and NLR family pyrin domain containing 7 (NLRP7; MIM: 609661). It has been shown in mice that knockout of any of the genes for these factors leads to infertility or subfertility caused by embryonic arrest.12–15 In humans, Murdoch identified mutations in NLRP7 responsible for recurrent hydatidiform moles and reproductive wastage.16 We previously identified mutations in PADI6 that cause embryonic arrest due to impaired zygotic genome activation in patients’ embryos.6 We and others have also identified novel mutations in two other members of the SCMC—TLE6 and KHDC3L—that are responsible for early embryonic arrest.17 18 However, mutations in these genes can only account for a small number of patients, and the genetic basis of early embryonic arrest remains largely unknown.
In this study, we investigated genetics causes of patients characterised by early embryonic arrest by whole exome sequencing. We identified different homozygous and compound heterozygous mutations in NLRP2 (MIM: 609364) in five patients and in NLRP5 in three patients, and all of these mutations were responsible for early embryonic arrest. All eight probands had biallelic mutations, suggesting that the mutations followed a Mendelian recessive inheritance pattern. We also investigated the effects of the mutations in Hela cells and in oocytes and embryos.
Infertility patients with early embryonic arrest and fertile controls were recruited from the Ninth Hospital affiliated with Shanghai Jiao Tong University, the Shanghai Ji Ai Genetics and IVF Institute affiliated with the Obstetrics and Gynecology Hospital of Fudan University, and Zhongshan Hospital of Fudan University. DNA was extracted from the peripheral blood based on a previously reported protocol.8 Oocytes and arrested embryos were donated by the patients. Control oocytes were matured in vitro from Germinal Vesicle (GV) or Metaphase I (MI) oocytes donated by other infertility patients, and control embryos were supernumerary embryos from donors undergoing successful IVF/ICSI attempts.
Genomic DNA was extracted from peripheral blood using the QIAamp DNA Blood Mini Kit (Qiagen). Whole-exome capture was performed using the SeqCap EZ Human Exome Kit (Roche), and sequencing was performed on the Illumina HiSeq 3000 platform (Illumina). Sequencing analysis was compared with the human reference sequence (NCBI Genome build GRCh37). Variants were annotated with the GRCh37, dbSNP (version 138), 1000 Genomes and Exome Aggregation Consortium (ExAC, version 0.3.1) databases and our in-house control exome database. Homozygosity mapping was performed with HomozygosityMapper.19 Because we recruited a proband who was from a consanguineous family and diagnosed with embryonic arrest, we preferentially considered homozygous variants as the pathogenic genes. The candidate variants were screened with following criteria: (1) homozygous variants with frequencies less than 0.1% in the East Asian population in the ExAC Browser and located within the homozygous regions, (2) variants predicted to be loss of function or damaging by SIFT or PolyPhen-2, (3) variants with high gene expression in human oocytes and early embryos according to our in-house RNA sequencing data and (4) an embryogenesis-related function for the candidate gene. Candidate variants were confirmed by Sanger sequencing.
Expression vector construction
Full-length coding sequences of NLRP2 and NLRP5 were amplified from control human oocyte cDNA and cloned into the Pcmv6-Entry vector with a FLAG-tag at the C-terminal end. Site-directed mutagenesis was performed to introduce mutations into the wild-type vector using the KOD-Plus Mutagenesis Kit (Toyobo) according to the manufacturer’s instructions. Wild-type and mutant clones were confirmed by Sanger sequencing.
Cell culture and transfection
Hela cells were cultured in Dulbecco’s Modified Eagle’s Medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin solution and incubated at 37°C and 5% CO2. NLRP2 and NLRP5 wild-type and mutant plasmids were transfected into Hela cells using the PolyJet In Vitro DNA Transfection Reagent (Signagene) according to the manufacturer’s instructions.
Hela cells were incubated for 36 hours after transfection with NLRP2 or NLRP5 wild-type and mutant plasmids. For collecting total protein, cells were lysed in RIPA lysis buffer (Shanghai Wei Ao Biological Technology) containing a protease inhibitor cocktail (B14001, Bimake) and centrifuged at 12 000 g for 30 min at 4°C. Supernatants were collected, mixed with 5× sodium dodecyl sulfate DS) loading buffer and heated at 100°C for 10 min. Proteins were separated on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE gel) followed by blotting onto a nitrocellulose membrane. The membrane was blocked in 5% milk in TBST for 1 hour and then incubated at 4°C overnight with anti-FLAG antibody (GNI4110-FG, GNI) at a 1:2000 dilution in 5% bovine serum albumin (BSA) and anti-GAPDH antibody (5174, Cell Signaling Technology) at a 1:2000 dilution in 5% BSA as the internal control. The membranes were washed with TBST three times and incubated with secondary antibodies (goat antimouse IgG or goat antirabbit IgG at a 1:3000 dilution, Abmart) for 1 hour at room temperature followed by washing again with TBST three times. Finally, the membranes were incubated with ECL Western Blotting Substrate (180–501, Tanon) and imaged on a Tanon 5200s Imaging Workstation.
The oocytes and embryos of the proband and healthy control subjects were fixed in 2% paraformaldehyde for immunofluorescence. The oocytes and embryos were blocked in blocking buffer at 4°C overnight and then incubated with anti-NLRP2 antibody (A8233, ABclonal) at a 1:100 dilution for 1 hour at room temperature. Then the oocytes and embryos were washed with washing buffer three times and incubated with an Alexa Fluor 488 secondary antibody (A-21206, ThermoFisher) at a 1:200 dilution for 1 hour at room temperature. DNA was labelled with Hoechst 33 342 solution (561908, BD) at 1:200 dilution for 1 hour at room temperature. All images were captured on a confocal laser-scanning microscope (Leica).
Clinical characteristics of the probands
All probands from seven families had been diagnosed with primary infertility with unknown cause for several years. The proband in family 1 was 29 years old at examination and had undergone a failed IVF and a failed ICSI attempt. In the IVF attempt, 19 oocytes, including 15 first polar body (PB1) oocytes, were retrieved. Fourteen of the oocytes were successfully fertilised, and 12 of them underwent normal cleavage. After cultivation, four viable embryos were frozen. The other embryos were all arrested at the two-cell to five-cell stages. The four viable embryos were subsequently implanted but failed to establish pregnancy. In the ICSI attempt, 10 PB1 oocytes were successfully fertilised, and 9 of these underwent normal cleavage. However, one embryo was arrested at the two-cell stage on day 3, and the remaining eight embryos failed to form blastocysts on day 5.
The proband in family 2 underwent two failed IVF attempts. In the first attempt, 12 oocytes were retrieved. Seven PB1 oocytes were successfully fertilised and six embryos underwent normal cleavage, but all of them were arrested at the two-cell to seven-cell stage with many small anucleate fragments on day 3. In her second IVF attempt, 12 oocytes were fertilised and 9 embryos underwent normal cleavage, but all of them were arrested at the two-cell to seven-cell stage with many small anucleate fragments on day 3.
The proband in family 3 was 34 years old at examination and had undergone six failed ICSI attempts. In the first ICSI attempt, two oocytes were retrieved and one was fertilised successfully. The only embryo was arrested at the four-cell stage on day 3. In the second ICSI attempt, two out of four oocytes were fertilised, but the only ensuing embryo failed to form a blastocyst on day 5. In the third ICSI attempt, two oocytes were successfully fertilised and underwent cleavage. The two viable embryos were frozen on day 3 but failed to establish pregnancy after implantation. In the fourth ICSI attempt, four out of six oocytes were fertilised and underwent cleavage. Two viable embryos were frozen, but they failed to establish pregnancy after implantation. In the fifth ICSI attempt, two out of four oocytes were fertilised and underwent cleavage, and one viable embryo was frozen on day 3. One was further cultured and formed a blastocyst on day 5 but failed to establish pregnancy after implantation. In the sixth ICSI attempt, three oocytes were fertilised and underwent cleavage. One six-cell stage embryo was further cultured but failed to form a blastocyst, while two viable embryos were frozen on day 3. The viable embryos failed to establish pregnancy after implantation.
The proband in family 4 was 27 years old at examination and had undergone three IVF/ICSI attempts. In the first IVF attempt, 14 oocytes, including six PB1 oocytes, were retrieved. Four oocytes were fertilised and underwent cleavage successfully. One viable embryo was frozen on day 3, while the others were further cultured, but only one could form a blastocyst on day 5. The two embryos were implanted on separate occasions, but both failed to establish pregnancy. In the second ICSI attempt, four out of five oocytes were fertilised and underwent cleavage. Only one was viable and frozen on day 3, while the others were arrested at the four-cell to five-cell stage. In the third ICSI attempt, six out of eight oocytes were fertilised and cleaved successfully. Three were arrested at the four-cell to six-cell stage on day 3, while the other three viable embryos were frozen. Two viable embryos were later implanted, and only one was successfully established pregnancy, and the patient gave birth to a full-term healthy baby.
The proband in family 5 was 27 years old at examination and had three IVF/ICSI attempts. In the first half-ICSI attempt, 10 out of 33 oocytes were fertilised and underwent cleavage. Only two embryos were viable. These embryos were implanted and established a pregnancy, but the fetuses were aborted spontaneously at 22 weeks because of cervical insufficiency of proband. In the second ICSI attempt, two out of three oocytes were fertilised and underwent cleavage, but the ensuing embryos were arrested on day 3. In the third ICSI attempt, 13 out of 21 oocytes were fertilised. The ensuing embryos were cultured on day 5, but only two formed blastocysts and were frozen. The two frozen embryos were implanted some weeks later, and the patient became pregnant and gave birth to two full-term healthy babies.
The proband II-1 in family 6 was 28 years old at examination and had undergone three failed IVF/ICSI attempts. In the first ICSI attempt, 26 oocytes, including 18 PB1 oocytes, were retrieved. Fourteen of them were successfully fertilised, and 11 of them underwent normal cleavage. All of embryos were arrested at the two-cell to four-cell stage. In the second IVF attempt, seven of nine oocytes were fertilised and underwent cleavage, but all were arrested at the four-cell to six-cell stage on day 3. In the third ICSI attempt, eight oocytes were fertilised and underwent normal cleavage, but all embryos were arrested at the two-cell to four-cell stage.
The proband II-2 in family 6 was 27 years old at examination and had undergone one failed ICSI attempt. In the ICSI attempt, 14 oocytes were retrieved, and nine of the PB1 oocytes were fertilised and underwent cleavage. However, all embryos were arrested at the two-cell to four-cell stage on day 3.
The proband in family 7 was 26 years old at examination and had undergone three IVF attempts. In the first IVF attempt, 16 oocytes were retrieved and 14 were fertilised successfully, but all embryos were arrested before the four-cell stage. In the second IVF attempt, 12 out of 16 oocytes were fertilised and underwent cleavage, but all embryos were arrested at the two-cell to six-cell stage. Similarly, in the third IVF attempt no viable embryos were frozen on day 3, and all were arrested at the three-cell stage (table 1).
Identification of mutations in NLRP2 and NLRP5
We first investigated the genetic cause for the infertility in the proband from the consanguineous family (family 1; figure 1A). After whole-exome sequencing, bioinformatics filtering analysis, homozygosity mapping (figure 1B) and genetic analysis using our in-house gene expression database, a homozygous truncation mutation c.1961C>A (p.Ser654*) in exon 6 in NLRP2 was found to be responsible for the phenotype. We confirmed the variant via Sanger sequencing. We then screened for homozygous and compound heterozygous mutations of NLRP2 in a cohort of 496 individuals diagnosed with early embryonic arrest. We found another four individuals (families 2–5) carrying different biallelic mutations in NLRP2 (figure 1A). The proband in family 2 carried the compound heterozygous mutation c.773T>C (p.Phe258Ser) and c.2254C>T (p.Arg752*). The proband in family 3 carried the compound heterozygous missense mutation c.525G>C (p.Trp175Cys) and c.2544A>T (p.Glu848Asp). The proband in family 4 carried the compound heterozygous missense mutation c.662C>T (p.Thr221Met) and c.1847A>T (p.Glu616Val). The proband in family 5 shared the same c.662C>T (p.Thr221Met) mutation with family 4 but carried a different mutation—c.1469C>T (p.Arg490Cys)—in the other allele. Because the parents and siblings of patients in family 4 and 5 were not available, we investigated whether the two mutations were on one or two alleles by constructing TA cloning and sequencing. As indicated in online supplementary figure S1, the two mutations were on different allele. Considering the similar functions of NLRP2 and NLRP5 in oocyte and early embryonic development,14 20 21 we further screened for homozygous and compound heterozygous mutations of NLRP5 in the cohort of patients. Expectedly, we found three individuals from two families carrying different biallelic mutations in NLRP5 (figure 1C). In family 6, two out of three sisters were infertile, and both infertile sisters carried the compound heterozygous mutation c.292C>T (p.Gln98*) and c.2081C>T (p.Thr694Ile), while the third sister was fertile and did not carry any mutations in NLRP5. The proband in family 7 carried the compound heterozygous mutation c.866G>A (p.Gly289Glu) and c.3320C>T (p.Thr1107Ile). All of the mutations were predicted to be loss of function or damaging (table 2). The positions of these mutations in NLRP2 and NLRP5 are highly conserved in different species (figure 2A,B). For the proband of family 6 II-2, in her only ICSI attempt, two out of nine fertilised oocytes were monitored by time-lapse microscopy. The embryos cleaved abnormally with many small anucleate fragments and were arrested in early stages (figure 1D). We measured the expression level of NLRP2 and NLRP5 using real-time PCR with specific primers (online supplementary table S1) and was normalised to the expression level of an internal GAPDH (MIM: 138400) control. We found that NLRP2 and NLRP5 were highly expressed in human oocytes and early embryos but were poorly expressed in somatic tissues (figure 2C,D).
Mutations in NLRP2 and NLRP5 impaired their expression in Hela cells, oocytes and embryos
To evaluate the functional effects of the identified NLRP2 mutations in vitro, we transfected wild-type and mutant NLRP2 vectors into Hela cells. Western blot showed that compared with wild-type, the c.1961C>A (p.Ser654*) and c.2254C>T (p.Arg752*) produced truncated protein and expressed at very low levels, while the c.773T>C (p.Phe258Ser), c.2544A>T (p.Glu848Asp) and c.1469C>T (p.Arg490Cys) mutations significantly decreased the protein expression level. For c.662C>T (p.Thr221Met) and c.1847A>T (p.Glu616Val), the reductions in protein expression were not statistically significant (figure 3A,B). For c.525G>C (p.Trp175Cys), there was no change in protein expression. These results suggest that most of mutations in NLRP2 lead to unstable protein.
Similarly, to evaluate the functional effects of the identified NLRP5 mutations in vitro, we transfected wild-type and mutant NLRP5 vectors into Hela cells. The amounts of p.Thr1107Ile proteins were significantly lower than wild-type, while the p.Gln98* proteins were nearly undetectable. The p.Gly289Glu and p.Thr694Ile proteins showed a tendency for reduced expression, but the reduction was not statistically significant (figure 3C,D). Immunofluorescence staining in oocytes and embryos from controls and the proband II-2 from family 6 suggested that the mutations caused NLRP5 protein degradation (figure 4). These results indicate that mutations in NLRP5 lead to protein degradation in Hela cells and in oocytes and embryos.
In this study, we identified homozygous and compound heterozygous mutations in NLRP2 and NLRP5 that are responsible for human early embryonic arrest. We found that mutations in NLRP2 impair the stability of the NLRP2 protein in vitro and that mutations in NLRP5 also impair the stability of the NLRP5 protein in Hela cells as well as in oocytes and embryos.
NLRP2 is a member of the NLRP family of proteins and is highly expressed in oocytes and early embryos, and two previous studies showed that Nlrp2 knockout female mice are subfertile. NLRP2 controls age-associated maternal fertility, and loss of NLRP2 results in a significant reduction in viable embryos, thus defining it as a new member of the SCMC.22–24 In our study, all probands with homozygous or compound heterozygous mutations in NRLP2 showed primary infertility and had histories of recurrent IVF/ICSI failure caused by early embryonic arrest. It has been shown in mice that severe reduction in the Nlrp2 protein level lead to lower blastocyst rates and that more embryos are arrested at the two-cell stage.20 The probands in families 1 and 2 who carried homozygous truncating mutations and compound heterozygous truncating and missense mutations in NLRP2 produced very few viable embryos. The proband in family 3 with the p.Trp175Cys and p.Glu848Asp mutations had a limited number of viable embryos. The proband in family 4 with compound heterozygous mutations p.Thr221Met and p.Glu616Val, and the proband in family 5 with the p.Thr221Met and p.Arg490Cys mutations also had limited numbers of viable embryos after several transplant attempts but were able to eventually give birth to live full-term infants. As indicated in figure 3A,B, different mutations impaired the function of the NLRP2 protein to differing extents. It has been known that the SCMC plays an important role during embryonic development. SCMC in oocytes regulate the RNA metabolism and zygotic genome activation, which is essential for normal embryonic development.25 In this study, mutations in NLRP2 and NLRP5 lead to unstable proteins. We therefore postulate that these unstable proteins may affect the stability and function of SCMC and consequently cause embryonic arrest.
Previously, a homozygous frameshift mutation located in exon 6 in NLRP2 was observed in the mother of two children with BWS, but because of the limited pedigree, a causal relationship between the mother’s mutation and the children’s phenotype could not be confirmed.26 Recently, additional rare variants in the maternal NLRP2 gene were shown to be associated with MLID in the offspring.27 In that study, six probands from five families diagnosed with multilocus imprinting disturbance (MLID) whose mothers carried NLRP2 mutations were identified. Most of these mothers had experience of miscarriage. The maternal reproductive history of a mother with the homozygous mutation c.1479_1480del (p.Arg493SerfsTer32) included one early abortion, two late miscarriages, two children carrying the heterozygous mutation and who were afflicted with MLID and one healthy child. The mothers of the children with MLID from the other four families had three different heterozygous mutation c.2237del (p.Asn746ThrfsTer4), c.2860_2861del (p.Cys954GlnfsTer18), c.314C>T (p.Pro105Leu) and one compound heterozygous mutations on one allele c.1885T>C (p.Ser629Pro) and c.2401G>A (p.Ala801Thr). The frequency of all of these mutations is less than 0.1% in the ExAC database. In our study, none of the probands’ mothers with heterozygous mutations in NLRP2 had a family history of miscarriage or of having children with MLID. Particularly, in family 4 and family 5, the patients carried compound heterozygous mutations had babies, and the three children are healthy until now. In addition, in our cohort of 800 fertile female controls, we identified 22 individuals harbouring rare heterozygous missense variants in NLRP2 with frequencies less than 0.1% in the ExAC East Asian population, and these mutations are predicted to be damaging. The fact that these fertile controls with rare missense variants in NLRP2 did not have family histories of miscarriage or of having children with MLID indicates that rare missense variants in NLRP2 might not be sufficient to produce MLID. This might be explained by the fact that mutations in a single gene can drive different diseases or phenotypes through different effects.28 Therefore, the relationship between maternal rare missense variants in NLRP2 and MLID is worthy of careful investigation in the future.
NLRP5 is a member of the SCMC, and it has been shown Nlrp5 knockout female mice are sterile14 and is also essential for early embryogenesis in sows.29 In our study, three probands with compound heterozygous mutations showed recurrent IVF/ICSI failure due to early embryonic arrest. Consistent with the result of the mutations’ effects in vitro, immunofluorescence showed low NLRP5 protein levels in these patients’ oocytes and embryos. These results suggest that unstable NLRP5 protein might be the reason for the failed IVF/ICSI attempts in the probands.
NLRP5 has also been shown to be associated with MLID. In a previous study,30 seven individuals with MLID were identified. Two mothers with compound heterozygous mutations—c.2320T>C (p.Cys774Arg), c.1664G>T (p.Gly555Val), c.2353C>T (p.Gln785*) and c.2840T>C (p.Leu947Pro)—had children with the heterozygous mutation and who were affected by MLID, and both mothers had experienced several pregnancy losses. One mother carried compound heterozygous mutations on one allele c.155T>C (p.Met52Thr) and c.226G>C (p.Glu76Gln), one carried heterozygous mutation c.1156_1158dupCCT (p.386dupP) and one carried c.1699A>G (p.Met567Val) and duplicate exon 7 in one allele and deletion in another. These mutations have frequencies of less than 0.1% in the ExAC East Asian population. In our cohort of 800 fertile female controls, we identified 13 individuals harbouring rare heterozygous missense variants in NLRP5 with frequencies less than 0.1% in the ExAC East Asian population, and these mutations are predicted to be damaging. Thus, the relationship between maternal rare missense variants in NLRP5 and MLID also needs to be considered.
The study has some limitations. First, because of scarcity of human oocytes and embryos, the exact molecular mechanism of early embryonic arrest is largely unknown. However, the fact that several mutations caused unstable protein implies that defected SCMC complex function entailed by mutations cause the early embryonic arrest. The molecular mechanism is worthy of being further investigated in the future by mutation knock-in mice. Second, parents and siblings of some probands were not available to target sequencing for NLRP2 and NLRP5. Future studies on genotyping these samples will help understand genetic inheritance pattern and penetrance of the mutant genes.
In conclusion, we identified mutations in NLRP2 and NLRP5 responsible for early embryonic arrest. The findings will provide potential biomarkers for evaluating the quality of embryos and will lay the foundation for the genetic diagnosis for clinical infertility patients.
We would like to thank the patients, their families and the volunteers of fertility female for their participation in this study.
JM and WW contributed equally.
Contributors JM and WW contributed equally to this work. QS and LW are cocorresponding authors. JM, WW, ZZ, QS and LW conceived and designed the study. JM, WW, BC, LW, XM, JF, YK and XS collected the samples. JM, BC, XM and JF organised the medical records. JM, WW, ZZ, BL, JF and QS performed the experiments. JM, WW, BC, QL, LJ, LH, QS and LW analysed the data. JM and LW wrote the manuscript. All authors reviewed and approved the manuscript.
Funding This work was supported by the National Key Research and Development Program of China (2018YFC1003800; 2017YFC1001500; and 2016YFC1000600), the National Basic Research Program of China (2015CB943300), the National Natural Science Foundation of China (81725006, 81822019,81771649, 81771581 and 81571501), the Shanghai Rising Star Program (17QA1400200), the Natural Science Foundation of Shanghai (17ZR1401900) and the Foundation of Shanghai Health and Family Planning Commission (20154Y0162).
Competing interests None declared.
Ethics approval All studies on human subjects were approved by ethnic committee of Shanghai Ninth Hospital, Shanghai Jiao Tong University (No 20161207).
Provenance and peer review Not commissioned; externally peer reviewed.
Patient consent for publication Not required.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.