Traditional models of twinning
Traditionally it has been thought that dizygotic twins result from fertilization of 2 distinct ova by 2 separate spermatozoa, whereas monozygotic twins are the product of a single ovum and sperm that subsequently divide to form 2 embryos.
Widely accepted models of monozygotic twinning are based on the unproven hypothesis of postzygotic division of the conceptus ( Figure 1 ). In this model, the number of fetuses, chorions, and amnions are determined by the timing of the embryo splitting ( Table 1 ).
| Zygosity | Twins | Time of split | Chorions | Amnions | Fetal mass |
|---|---|---|---|---|---|
| Dizygotic | DC DA | No split | 2 | 2 | 2 |
| Monozygotic | DC DA | Days 1–3 | 2 | 2 | 2 |
| Monozygotic | MC DA | Days 3–8 | 1 | 2 | 2 |
| Monozygotic | MC MA | Days 8–13 | 1 | 1 | 2 |
| Monozygotic | Conjoined | After day 13 | 1 | 1 | 1 |
Proposed triggers for splitting include postzygotic gene mutations, abnormalities in cell surface proteins, and abnormalities in the formation of the zona pellucida. The incidence of monozygotic twins is increased 2- to 5-fold in ART pregnancies, which might be predisposed to splitting because of handling, media, and microinjection or because of the intrinsic abnormalities associated with infertility.
It has conventionally been asserted that monochorionicity confirms monozygosity, opposite-sex twins confirm dizygosity, and same-sex dichorionic twins remain of uncertain zygosity until postnatal evaluation occurs.
New models of twinning
In 2013 Herranz argued that the hitherto-unchallenged hypothesis of postzygotic splitting lacked scientific proof. He argued that factors that initiate cleavage have not been specified, that coexistence of separate embryos within a single zona pellucida is unlikely, that postzygotic splitting becomes more unlikely with the passage of time, and that splitting has never been observed in vitro.
Herranz offered an alternative theory of twinning based on the following 2 principles: (1) monozygotic twinning occurs at the first cleavage division of the zygote and (2) subsequent chorionicity and amnionicity is determined by the degree of fusion of embryonic membranes within the zona pellucida ( Figure 2 ).
Denker opposed Herranz’s argument, emphasizing that a lack of evidence may stem from ethical limitations on scientific experimentation with human embryos. He highlighted that data regarding twinning mechanisms in animals, differences in the nature of the zona pellucida in vivo and in vitro, and the developmental potential of cell lineages after fertilization were not discussed. Denker concluded that both the traditional fission model and Herranz’s fusion model were unsubstantiated.
Further examination of twinning processes in typical and atypical twins might confirm which, if either, of these models is more accurate.
New models of twinning
In 2013 Herranz argued that the hitherto-unchallenged hypothesis of postzygotic splitting lacked scientific proof. He argued that factors that initiate cleavage have not been specified, that coexistence of separate embryos within a single zona pellucida is unlikely, that postzygotic splitting becomes more unlikely with the passage of time, and that splitting has never been observed in vitro.
Herranz offered an alternative theory of twinning based on the following 2 principles: (1) monozygotic twinning occurs at the first cleavage division of the zygote and (2) subsequent chorionicity and amnionicity is determined by the degree of fusion of embryonic membranes within the zona pellucida ( Figure 2 ).
Denker opposed Herranz’s argument, emphasizing that a lack of evidence may stem from ethical limitations on scientific experimentation with human embryos. He highlighted that data regarding twinning mechanisms in animals, differences in the nature of the zona pellucida in vivo and in vitro, and the developmental potential of cell lineages after fertilization were not discussed. Denker concluded that both the traditional fission model and Herranz’s fusion model were unsubstantiated.
Further examination of twinning processes in typical and atypical twins might confirm which, if either, of these models is more accurate.
Atypical twinning
A review of the evidence for atypical twinning provides insights into the mechanisms of twinning and challenges aspects of traditional models of twinning.
Chimeric twins
A chimera is a single organism containing 2 populations of genetically distinct cells originating from 2 different zygotes. Chimerism in humans was initially observed in studies of the ABO blood group. Blood group chimerism was demonstrated via genetic testing in 1976 and is now considered common and persistent.
Chimerism has since been described in twins with monochorionic dizygotic (MCDZ) placentation. Early reports of MCDZ twins were discounted because of the absence of formal placental histopathology or confirmatory genetic testing.
In 2003 Souter et al reported the first confirmed case of sex-discordant MCDZ twins born to a 48 year old woman following in vitro fertilization (IVF). Cytogenetic analyses demonstrated chimerism in peripheral blood leukocytes. A further 20 cases of MCDZ twins with confined hematological and/or tissue chimerism have been reported ( Appendix 1 ).
Concerns have been raised that chimeric twins might exhibit reproductive dysfunction analogous to that of the bovine freemartin. Early follow-up studies described normal genitalia, gonads and endocrinological function in gender-discordant chimeric twins to a maximum of 18 months of age. However, in 2013 Choi et al reported a case of MCDZ twins complicated by death in utero of the female twin and severe gonadal failure in the male cotwin. The authors concluded that close observation of chimeric infants is necessary to ensure that gonadal failure/dysfunction is identified and appropriately managed.
The mechanisms underlying human twin chimerism and monochorionic dizygotic twin pregnancies remain incompletely defined. Theories proposed are outlined in Table 2 and depicted in Figure 3 . Nevertheless, it is clear that the dogma of monochorionicity being synonymous with monozygosity is no longer appropriate.
| Hypothesis | Evidence | |
|---|---|---|
| 1 | Placental anastomoses allowing early transfer of genetic material |
|
| 2 | Fusion of elements of 2 genetically distinct zygotes |
|
| 3 | Fertilization of a binovular follicle |
|
Assumptions regarding the antenatal diagnosis of zygosity on the basis of sonographic features may be unreliable, with important implications for antenatal risk stratification, screening, and diagnosis. Failure to diagnose MCDZ twins might have long-term consequences. Individuals with blood or tissue chimerism might be at increased risk in the context of transfusion or transplantation, and modeling of epigenetic and genetic factors of disease in monochorionic twins may lead to erroneous conclusions.
Phenotypically discordant monozygotic twins
Phenotypic discordance in monozygotic twins commonly occurs as a consequence of epigenetic, mitochondrial, and genetic discordance. Epigenetics has been implicated as a mediator of stochastic and twin-specific environmental factors. Genetic differences within monozygotic pairs must arise de novo soon after zygotic cleavage if they are found in multiple somatic tissues and later in development if they are mosaic. Such differences can be single base pair mutations or copy number variation or involve whole chromosomes.
On a genome-scale, the frequency of epigenetic differences within monozygotic pairs is likely to be high. Less is known about the frequency of genetic discordance in monozygotic twins, although it is likely to be low. However, more genetic variation might occur outside coding regions. Little is known about the frequency of mitochondrial discordance in twins.
Mirror-image twins
Mirror-image twinning has been described in monozygotic twins with phenotypic features that are asymmetrical. As many as 25% of monozygotic twins may have mirror-image features. Mirror asymmetries observed include direction of occipital hair whorl, dental patterns, unilateral eye and ear defects, cleft lip and palate, bony abnormalities, and tumor patterns ( Appendix 2 ). Mirror-image central nervous system abnormalities, including optic glioma, colpocephaly, and arachnoid cysts, and cases of mirror-image organ laterality in heterotaxy syndromes have been described.
It has been suggested that higher-order cerebral functions including dominant handedness, eye dominance, and cerebral lateralization for language and mental rotation tasks also exhibit mirror asymmetries. However, Derom et al demonstrated that although left-handedness may be more common in twins, there is no evidence to suggest that discordant handedness represents mirror imaging. These conclusions were supported by large-scale data from Australia and The Netherlands.
According to traditional models of twinning, mirror-imaging results from late zygotic splitting at days 9-12, just prior to the formation of conjoined twins ( Table 1 ). Cases of heterotaxy syndrome have been likened to cases of conjoined twins in whom the close proximity of the body axes gives rise to organ laterality of one twin affecting that of the other. Nevertheless, little evidence exists to support these hypotheses.
Polar body twins
A polar body is a small, cellular by-product of the meiotic division of an oocyte. Apoptosis usually occurs within 17-24 hours of formation, and the resulting fragments remain within the zona pellucida.
It has been hypothesized that fertilization of an ovum and its first or second polar body by 2 distinct sperm may result in polar body twinning ( Figure 4 ). In 1981 Bieber et al described a monochorionic twin pregnancy with a normal male (XY karyotype) and an acardiac female (triploid XXX karyotype). Cytogenetic studies suggested a diploid contribution from the mother in the acardiac twin. Human leukocyte antigen (HLA) typing suggested dispermic fertilization. Thus, it was proposed that independent fertilizations of a haploid ovum and its diploid first polar body had occurred. The authors hypothesized that the proximity of the ovum and its first polar body allowed the development of distinct inner cell masses within a common trophoblast. A fusion mechanism for twinning was deemed unlikely.
In contrast, Fisk et al performed cytogenetic analyses on the tissues of 9 twin pregnancies affected by twin-reversed arterial perfusion sequence. All twin pairs were monochorionic and discordant for the acardiac anomaly. Deoxyribonucleic acid (DNA) fingerprinting revealed monozygosity. The calculated likelihood of fertilization of an ovum and its first or second polar body in any twin pair and in all twin pairs was less than 3.6% and 0.0003%, respectively. The authors disputed the existence of polar body twinning, ascribing previously reported cases to chimerism, and instead suggested embryonic fusion as an alternative but poorly understood possibility.
Complete hydatidiform mole with coexistent twin
A multiple pregnancy with a complete hydatidiform mole (CHM) and a coexisting live fetus (CLF) is characterized by the presence of a fetus with normal karyotype, anatomy, and placentation alongside a molar component with no identifiable fetal parts, a placenta with diploid paternal chromosomes, and the characteristic sonographic and histological features of a CHM ( Figure 5 ). CHM-CLF is rare, with a reported incidence of 1 in 22,000 to 1 in 100,000 pregnancies.
Three conditions require differentiation from CHM-CLF. First, a singleton pregnancy with a partial hydatidiform mole may occur in which the fetus has triploidy resulting from dispermic fertilization of a haploid normal oocyte. Second, a twin pregnancy may occur with a normal twin in one sac and a partial mole in the other sac. Third, mesenchymal dysplasia may occur and is associated with an enlarged cystic placenta, fetal growth restriction, and, occasionally, fetal death.
Historically, most CHM-CLF pregnancies have been terminated. More recently it has become apparent that the prognosis is not as poor as previously thought. The live birth rate varies from 21% to 40%. Complications include vaginal bleeding, hyperemesis gravidarum, thyrotoxicosis, early-onset severe preeclampsia, and fetal death. Outcomes in higher-order multiple pregnancies remain poor.
Conflicting evidence exists regarding rates of persistent gestational trophoblastic disease in CHM-CLF compared with CHM alone. Sebire et al showed a rate of 19% for CHM-CLF compared with 16% for CHM alone. Massardier et al showed a rate of 50% for CHM-CLF compared with 14% for CHM alone. The risk of gestational trophoblastic disease is independent of gestation and whether the pregnancy is terminated or is allowed to continue.
Although fetal loss is the most likely outcome for CHM-CLF, continuing the pregnancy is possible as long as maternal complications are manageable and the pregnancy is closely monitored.
Vanishing twins
Vanishing twin syndrome (VTS) refers to multiple pregnancies affected by the spontaneous loss of an embryo or fetus in the first trimester. In 1976 it was observed that twin pregnancies diagnosed on ultrasound prior to 15 weeks’ gestation frequently gave rise to the delivery of singleton infants. After the introduction of routine first-trimester ultrasound, VTS was increasingly observed.
VTS is thought to be underreported in spontaneous twin pregnancies. Fetal reduction may occur prior to recognition of pregnancy, with up to 80% occurring prior to 9 weeks’ gestation ( Figure 6 ). Most cases are asymptomatic but can be accompanied by vaginal bleeding. Alternatively, VTS is well characterized in ART pregnancies, with reported rates ranging from 10.4% to 18.8%.
Conflicting evidence exists regarding potential adverse effects of a vanishing twin on the remaining pregnancy. Pinborg et al described the outcomes of 642 survivors of VTS. When compared with singletons, survivors were found to be at an increased risk of small for gestational age, low and very low birthweight, and preterm birth. The degree of risk was inversely proportional to the timing of fetal loss. Further studies have demonstrated inconsistent results ( Appendix 3 ).
Concerns have been raised regarding the impact of VTS on neurodevelopmental outcomes. Cases of focal cortical sclerosis, microcephaly, and multicystic encephalomalacia have been reported. Initially it was suggested that VTS might increase the risk of cerebral palsy for survivors. Subsequent studies revealed no statistically significant increase in cerebral palsy or adverse neurological sequelae. Longer-term follow-up has suggested a possible increased risk of cerebral impairment at 1 year of age, but again, findings did not reach statistical significance. Furthermore, survivors of VTS are deemed less psychologically vulnerable by their parents to a maximum age of 11 years.
VTS may have an impact on prenatal screening and diagnosis. Studies evaluating serum markers in pregnancies affected by VTS have reported inconsistent results, perhaps because of differences in mean gestational age at sampling and the interval between sampling and fetal loss. Recent evidence suggests the presence of a vanishing twin is associated with a 21% increase in pregnancy-associated plasma protein A ( P = .0026), a 10% increase in alpha-fetoprotein ( P < .0001), and a 13% increase in dimeric inhibin A ( P = .0470).
VTS might also affect the interpretation of noninvasive prenatal testing (NIPT) using cell-free fetal DNA. Cases of misdiagnosis of fetal sex using NIPT may be due to VTS with subsequent persistence of sex chromosome sequences from the vanishing twin. In 2015 a large-scale evaluation of results of NIPT in 30,795 consecutive cases identified 130 cases with additional fetal haplotypes, 76 of which could be clinically correlated. VTS was evident in 42.1% of these 76 cases. Fetal haplotypes remained detectable via NIPT for up to 8 weeks after the fetal loss. It has been concluded that early ultrasound monitoring and careful pretest and posttest counseling regarding NIPT are essential.
Fetus papyraceus
Fetus papyraceus refers to a fetus in a multiple pregnancy that dies in utero and then appears as a compressed, mummified mass at delivery ( Figure 7 ). Fetal death is thought to occur between 12 and 20 weeks’ gestation. The incidence is 1 in 200 twin pregnancies and 1 in 12,000 pregnancies overall.
In most cases, fetus papyraceus is of no consequence to the surviving pregnancy. However, associations with aplasia cutis congenital intestinal atresia, gastroschisis, and cardiac and pulmonary anomalies have been described.
Fetus papyraceus has been reported in both monochorionic and dichorionic twin pregnancies and higher-order multiple pregnancies. There are no associations with age or parity, but there is a trend toward increased frequency with monochorionicity and velamentous cord insertion.
Many of the abnormalities associated with fetus papyraceus can be attributed to thromboembolic events following the death of a monochorionic twin. However, shared vascular anastomoses alone are not a complete explanation because the condition is seen in dichorionic twins. In dichorionic twins, fetal death might be due to placental ischemia leading to fetus papyraceus and consequences for the cotwin.
Fetus in fetu/parasitic twins
Fetus in fetu refers to 1 or more partially formed fetuses situated entirely within the body of another normally formed fetus ( Figure 8 ). First described in approximately 1800, this event has been estimated to occur once in every 500,000 births. Fetus in fetu remains rare despite the advent of ART, with fewer than 200 cases documented.
Historically, fetus in fetu was considered to represent a well-formed mature teratoma. However, in contrast to the disorganized tissues derived from uncontrolled pluripotent cell replication in teratomata, fetus in fetu is characterized by the presence of vertebrae with appropriately organized limbs and organs ( Figure 9 ). Serology and molecular genetic testing have indicated that fetus in fetu represents a monozygotic, monochorionic diamniotic twin gestation. Persistent anastomoses of the vitelline circulation lead to the absorption of one twin inside the other during the ventral folding of the trilaminar embryonic disc.
Nevertheless, an association between teratomata and fetus in fetu has been observed. Both are commonly located in the retroperitoneum and are histopathologically similar. Cases of teratoma and fetus in fetu occurring in the same individual have been reported. Fetus in fetu has been described as part of a parasitic continuum including conjoined twins, acardiac twins, and teratomata.
Fetuses in fetu have been identified in the mediastinum, scrotum, mouth, and skull. Usually there is 1 fetal mass, but cases of up to 11 fetuses in fetu have been reported. The diagnosis is commonly made following the incidental identification of an abdominal mass in a neonate or infant. Advances in fetal sonography have led to increased prenatal diagnosis. Unlike teratomata, fetuses in fetu do not demonstrate malignant potential but may cause significant mass effect, necessitating surgical removal ( Figure 10 ).
Superfetation
Superfetation refers to fertilization and implantation of a second conception during pregnancy ( Figure 11 ). Early cases of suspected superfetation were reported in the context of growth discordance. In 1989 Bhat et al reported a case of dichorionic diamniotic twins delivered at 36 weeks’ gestation with discordant birthweights and the subsequent death of the second twin. Given the significantly different Dubowitz scores, superfetation was presumed.
In 2003 Singhal et al reported a case of twins born to a mother with uterus pseudodidelphys. At the time of presentation, discordant estimated gestational age was determined. In the absence of Doppler parameters suggesting intrauterine growth restriction, superfetation was presumed. The demise of the second twin was confirmed at 32 weeks’ gestation, whereas the first twin was live born at 35 weeks’ gestation.
The plausibility of the phenomenon of superfetation has been questioned. Whereas ovulation has been reported to occur in the first trimester of pregnancy, the up-regulation of the hypothalamic-pituitary-ovarian axis by luteal and then placental progesterone typically suppresses ovulation. Progesterone-induced changes in cervical mucus may limit successful fertilization. The presence of a gestational sac in the uterus may limit a successful implantation. Therefore, it has been argued that spontaneous superfetation is unlikely.
Three alternative explanations for described cases of superfetation have been proposed. First, pregnancies complicated by growth discordance may give rise to the appearance of twins of differing gestational ages. Growth discordance because of placental insufficiency, infection, congenital anomalies, or twin-to-twin transfusion syndrome is not uncommon in twins. Second, interval delivery might contribute to the subsequent appearance of twins with different gestational ages. Third, in cases in which there is a misdiagnosis of a singleton pregnancy, multiple pregnancy on further imaging might be attributed to superfetation rather than to sonographic diagnostic error.
In contrast, with the advent of ART, it has been recognized that natural barriers to superfetation can be overcome. Cases of superfetation resulting from ART with or without additional spontaneous conception have been reported ( Appendix 4 ).
In 2005 Harrison et al attempted to confirm superfetation in a triplet pregnancy by estimating the gestational age postnatally using neurosonography and ophthalmic evaluation. The authors raised the possibility that superfetation can be determined conclusively. Further research is necessary to definitively prove or disprove this phenomenon.
Superfecundation
Superfecundation refers to the fertilization of 2 oocytes via separate instances of coital or artificial insemination during the polyovulatory period ( Figure 11 ). Heteropaternal superfecundation occurs after coitus with multiple partners. Monopaternal superfecundation occurs after coitus with one partner on multiple occasions. The latter is probably more common and not frequently detected.
In 1978 Terasaki et al described a case of suspected heteropaternal superfecundation on the basis of HLA typing. In 1982 heteropaternal superfecundation was again suspected in a pair of twins born with different skin colors. Additional cases have been described in the context of paternity disputes ( Appendix 5 ).
More recently, superfecundation has been associated with ART. In 2001 Amsalem et al reported spontaneous monopaternal superfecundation in a 25 year old woman undergoing IVF for secondary infertility. After an uncomplicated transfer of 2 embryos, 5 embryos were detected on ultrasound. Following multifetal pregnancy reduction, male twins were born at term. Subsequent cytogenetic testing confirmed monopaternity and suggested that cleavage/duplication was unlikely. In 2005 Peigné et al reported a case wherein monopaternal superfecundation was observed following multiple intrauterine inseminations and IVF cycles. Again, a selective fetal reduction resulted in the delivery of live-born twins.
Evaluation of large parentage databases has led to a reported overall frequency of heteropaternal superfecundation of 2.4%. It has been proposed that 1 in 400 twin pairs born to married white women in the United States are the result of heteropaternal superfecundation. Monopaternal superfecundation is thought to be more common, with an estimated prevalence of 1:12 dizygotic twins born to mothers in the United Kingdom. The rate of superfecundation depends on community rates of coital frequency, polyovulation, and subsequent DNA detection. The reported incidence may be increasing because of increased paternity testing.
Given that superfecundation may occur with ART, women should be advised to consider avoiding intercourse after embryo transfer to reduce the risk of subsequent higher-order multiple pregnancy and/or ectopic pregnancy.
Conclusion
Twinning is a complex and multifactorial phenomenon and elements of the twinning process remain incompletely understood. A conventional model of monozygotic twinning is based on fission events in the developing embryo. This model lacks definitive evidence and is challenged by cases of atypical twinning.
An alternative model proposes that embryonic fusion events underlie monozygotic twinning. However, supporting evidence is similarly limited. Elucidating the precise mechanisms by which twinning occurs will have significant implications for managing complications unique to multiple gestations; utilizing cell-free DNA for aneuploidy screening in multiple pregnancies; interpreting twin data to determine the relative contributions of genetic, epigenetic, and environmental factors to various phenotypic outcomes; and reducing the incidence of spontaneous and assisted multiple pregnancies. An examination of the anomalies of the placenta and umbilical cord in twin gestations has been recently published and provides an excellent review. Further research, including series of extended cytogenetic analyses, is needed to refine our understanding of early embryonic development in twin pregnancies.
Appendices
