Brain malformations in diprosopia observed in clinical cases, museum specimens and artistic representations

The term ‘diprosopus’ describes a symmetrically and laterally double faced (πρόσωπο = face), monocephalic and single-trunk individual. It is defined by the duplication of at least two facial structures [1]. Diprosopus is considered a subtype of conjoined twins; in this context, it is also referred to as ‘parapagus diprosopus’. The Janus type of opposing facial duplication, occurring in ventrally fused ‘cephalothoracopagous’ twins, may be regarded as a monocephalic diprosopus in a broader sense of the term.

The oldest preserved representations of a diprosopus originate from the pre-Columbian village Tlatilco in central Mexico at 1200–400 B.C.E. [2‐4] (Fig. 1d + e) and from Schedel ‘s world chronicle 1493 C.E. [5] (Fig. 1h). Also in modern arts, we found a depiction of an apparent diprosopus by Paul Klee from 1933 (Fig. 1f). A total of 81 scientific reports on human facial duplications have been documented since 1642, 38 early ones cited by Barr [6], 33 more recent ones cited by Bidondo et al., [1] and 10 additional cases [7‐16]. The prevalence rate of Diprosopus has been calculated to be 2 per 1,000,000 births in Argentinia, accounting for 10% of all conjoined twins [1]. The prevalence of conjoined twins worldwide is 14.7 per 1,000,000, with a rate of < 3% of diprosopus, of < 1% of Janiceps twins and of 3.9% of parasitic twins [17]. There is a female preponderance with a ratio of m:f = 1:1,4. In approximately half of the cases, diprosopus is associated with craniorachischisis. Other frequent anomalies are cleft lip and palate (CLP), bifid tongue, cerebral duplication, cardiac septal defect, diaphragmatic hernia and laterality defects [1]. Prenatal diagnosis of diprosopia by 2D- and 3D colour Doppler ultrasonography is possible at 12 weeks gestation [15]. Diprosopia is incompatible with prolonged extrauterine life.

The aetiology of diprosopus is unknown. Increasing numbers of conjoined twins (not including diprosopia) born after the Chernobyl accident (5 in 96,438 births), the lack of familial aggregation, and the 21 singletons fathered by the Siamese twins Chang and Eng Bunker support causative nongenetic factors [1, 18, 19]. The presumably higher incidence of diprosopia in Latin America and the induction of a facial double malformation by a higher expression of Sonic Hedgehog (Shh) in animal models would be compatible with the involvement of causative genetic factors [20]. Proposed pathogenic mechanisms are as follows: Lateral fusion of two monozygotic embryonic discs [21]; fission of a single embryonic disc with the formation of two notochords, and thus of two cephalic neural plates and of an extra medially located cranial neural crest [22]; duplication of the early primitive node as the signal emission centre for the neural plate, thereby causing cranial bifurcation of the notochord, the formation of two vertebral axes and of two neural plates [1].

We added five more patients with true diprosopus (clinical cases 1–3 and exhibits cases 4 and 5) and two double-faced monocephalic homopagous and heteropagous conjoined twins (clinical case 6 + exhibit case 7) to these observations under consideration of the accompanying central nervous system (CNS)-abnormalities and with reference to corresponding representations in art.. In case 3 singleton whole-exome sequencing (WES) was performed.

Clinical and morphological data are summarized in Table 1.

A variety of major CNS malformations were observed in seven monocephalic double-faced foetuses, five presenting with true single-trunk diprosopus and two presenting as Janus-type and parasitic cephalothoracopagous twins, respectively (cases 6 and 7). CNS malformations comprised craniorachischisis with anencephaly and exencephaly (cases 1, 4), spina bifida (cases 5, 7), secondary hydrocephaly due to HPE and to spina bifida (cases 2, 5), cerebral duplication (cases 2, 3, 6), a Dandy-Walker cyst due to hypoplasia of the vermis (case 3), anterior encephalocele (cases 2, 3), HPE (cases 2, 7) and cerebral di-symmetry (case 6).

With respect to the prevalence rate of NTDs of 1:538 worldwide with 38.7% of NTDs being anencephalies [21, 23], the high rate of almost 50% of true diprosopus cases showing ex- or anencephaly among cases from our study and from the literature (22 anencephalies out of a total of 48 accessible diprosopus cases) makes a random association unlikely. However, the observed CNS malformations should be seen in the context of the accompanying craniospinal abnormalities. Duplication of the forebrain and part of the midbrain, pons, medulla, and spinal cord, thought to result from duplication of the early primitive node, notochord, and cephalic neural plate [1, 21, 22] are associated with duplications of bones particularly of the frontal skull base, the cervical spine, and of the thoracic and lumbar vertebral bodies. Corresponding bone anomalies were also seen in double-faced foetuses with craniorachischisis, suggesting a similar pathogenesis, regardless of whether the duplicated CNS tissues are preserved or perished in the absence of a sheltering skull and spinal canal. This makes cerebral duplication an obligatory feature of diprosopia. The high risk for cranial NTDs in diprosopia could be explained by the problems with neural tube closure caused by incomplete duplication of still-fused neural plates in the presence of a single extra medial neural crest [22] at an initiation site of the bidirectionally proceeding neural tube closure [24]. This is underlined by a low risk for anencephaly in dicephalic single trunk foetuses showing duplicated and separated spines and spinal cords.

Congenital anterior meningoencephaloceles (MECs) are rare in Western countries (1:35 000 live births). The majority of cases are not attributable to a classical NTD but rather represent a herniation through a defect or focal structural instability of the skull base [25]. The site of the larger intrapharyngeal MEC in case 2 represented fused and at the bottom cleaved pituitary fossae between duplicated sphenoid bones, and pituitary fossae fusion was facilitated by frontal angling of the medial sphenoid wings (Fig. 3c). The MEC contained cerebral tissue of the left holoprosencephalic brain and presumably irritated the development of the left face, causing displacement of its right eye socket and ipsilateral anophthalmia (see below; Fig. 1c). In the foetus in case 3, a smaller retropharyngeal MEC was associated with a midline defect in the lower ‘occipital’ clivus, and the upper ‘sphenoidal’ clivus was duplicated, thereby leading to osseous instability through widening of the sphenooccipital junction (Fig. 2d). In both of these foetuses the anterior intracranial MEC was thus closely related to the duplication of cranial bones, suggesting an association of diprosopus with a higher risk of anterior MECs. However, an anterior MEC has only been reported once in a patient with diprosopus. This patient had a large extracranial MEC extruding from a midline aperture between duplicated frontal bones and anterior fossae. It was associated with an extended interfacial cleft and–similar to our case 2—with discordant medial anophthalmia. Notably, a clival encephalocele has been observed in association with duplications of the pituitary, sella, ventricular system, basilary artery, tongue, frontal teeth and cervical vertebral bodies in one case, [26]. Other examples of small anterior MECs associated with diprosopia may have been missed in the past.

HPE is the most common forebrain malformation accounting for 1:7,500 livebirths and 1:250 early embryos [27, 28]. According to the degree of noncleavage into the two hemispheres, HPE is classified into an alobar, semilobar and lobar type, not considering HPE-microforms. HPE is associated with facial midline defects, such as cyclopia, ethmocephaly, cebocephaly or just hypotelorism, hypoplastic nose and CLP, and the severity of these features largely correlate with the HPE type. Following DeMyer’s statement ‘the face predicts the brain’, we may attribute concordant cyclopia in our case 7 to concordant alobar HPE, even though the brain was not exploitable [29, 30]. However, in case 2, this statement has only limited validity. Discordant cebocephaly of the right face was associated with lobar HPE, while left-sided alobar HPE did not result in cebocephaly or any other typical facial midline defect, possibly due to interference with an intracranial MEC. Three cases of severe facial midline defects in association with true diprosopus have been observed thus far—in addition to our case 2 resulting in 4 HPEs out of 48 accessible diprosopus cases. In one of two cases presenting with discordant cyclopia, autopsy had been declined [13]. In the second case in which the individual showed cyclopia on the left face and a normal mouth, nose and right eye, but anophthalmia with a depression at the site of a lacking left palpebral fissure on the right face, the brain had been described as having 'two hemispheres'—although two 'holospheres' would have been the more likely diagnosis as in cas 2 [31]. In a diprosopus with concordant cyclopia, HPE was confirmed but not specified [12]. However, in this case, the first HPE panel analyses of a diprosopus had been performed with sequencing of SHH, ZIC2, SIX3 and TGIF1 genes. Causative variants were excluded. This might suggest that HPE in diprosopia developed independently from an addtional pathogenic HPE gene variant.

The first WES in diprosopus, as performed in case 3, concerning a nonholoprosencephalic foetus of nonconsanguineous parents, did not reveal any causative variants, either in known disease-causing genes, or in known genes involved in the Shh pathway. A single ‘singleton WES ‘ does not, of course, allow the exclusion of a monogenic disorder, a genomic rearrangement or of somatic mosaicism. However, along the way, it may be possible that future technologies will reveal a genetic contribution to this developmental disorder. A careful molecular testing including ‘trio WES ‘ and ‘long-read whole-genome sequencing' is therefore warranted, even though the likelihood of finding a monogenic cause is very low. However, in our case 3, pregnancy was achieved after intracytoplasmic sperm injection (ICSI). It has been shown that artificial reproductive techniques are associated not only with a twofold increase in congenital malformations [32] but also with a two- to threefold increase in monozygotic twin pregnancies after single embryo transfer. The use of frozen and thawed embryos—not applicable to our case 3—and zona pellucida manipulations, such as artificial handling or ICSI have been discussed as possible risk factors [33]. This may favour a mechanically conditioned non-genetic pathogenesis of diprosopus.

The cephalothoracopagous Janus-type diprosopus in case 6 differs from the parapagous diprosopus in presenting two bodies and ’frontal’ fusion. The two anterior fossae of the fused cranial vaults, the two faces and the two chest fronts were split into halves; each half turned outward at right angles and fused to the contralateral half of the co-twin's anterior fossa, face and chest. This created cranial and thoracic di-symmetry with crossing median axes. The two opposing faces thus located lateral to the vertebral axes showed symmetrical midline defects with marked hypotelorism, hypoplastic nose, microstomia and microgenia. These features are quite common in di-symmetric Janiceps twins [34, 35]. More common, however, is facial asymmetry, ranging from discordant cebocephaly, cyclopia or synotia up to mono-symmetrical single-faced cephalothoracopagous twins [34‐39]. Consistent with the di-symmetry of the cranial vaults in the foetus in case 6, there was cerebral di-symmetry with the two frontal lobes exhibiting right-angled outward rotation and pairing with the co-twin’s frontal lobes (Fig. 5c). Cerebral di-symmetry has not been described previously and can as best be guessed from published ultrasound and MRI images [35‐37]. Facial midline defects are generally related to HPE due to failure of forebrain cleavage into two hemispheres. However, this explanation does not hold for the Janiceps twins, since each set of frontal lobes originates from two different forebrains. This may indicate a different and non-genetic pathogenesis of midline defects in Janiceps twins being due to forebrain fusion instead of failure of forebrain fission. Frontal fusion of the brains in Janiceps twins is frequently mentioned in the literature, but we did not find illustrations showing secondary fusion of the two heterogenic frontal lobes in cases of Janiceps midline defects with the exception of a graphic representation [35].

Parasitic heteropagous twins represent asymmetric conjoined twins with a defective smaller and nonviable 'parasite' attached to and supplied by a near normal viable 'autosite'. In many cases, the site of attachment is the epigastric region, as in case 7, concerning the parasite’s hypoplastic pelvis and lower limbs, eventually including defective upper limbs and with rudimentary internal organs enclosed in the host’s thoracic cavity [41, 42]. Most parasitic twins are a-cephalic and a-cardiac. This finding is reminiscent of the a-cephalic and a-cardiac umbilicopagous cotwin in the twin-to-twin reversed transfusion syndrome suggesting that similar circulatory mechanisms may be responsible for defective development, preferably of the upper body parts [41, 42]. The peculiarity of our case 7 lies in the simultaneity of frontal thoracic fusion and lateral cephalic fusion with the two faces displaying two eyes, no nose, two mouths and four ears (parapagous diprosopus). Fusion of two orbits and eyes each, a bilaterally median lower lid notch, arrhinia and microstomia indicated concordant cyclopia (Fig. 5b) and strongly suggested HPE. This combination has not been described before. It raises the question of whether different pathogenic mechanisms are involved that lead to thoracic frontal and facial lateral fusion and to ‘selective ischaemic damage’ of the parasite’s upper body parts.

The ceramic figurines of tri-ophthalmic double-faced women (Fig. 1d + e) from the early and middle Formative Mesoamerican period, found in the open cornfields of the Tlatilco village near what is now Mexico City, represent images of an early lethal congenital malformation projected onto a so-called ‘pretty lady’ image, provided with a spiritual exaltation and associated with fertility beliefs. From the accuracy of the portrayals, we can conclude that they are based on real observations [2‐5]. The braided hairstyle of the figurine in Fig. 1d (1200–900 B.C.E.) resembles the brain shape in case 4, showing exencephaly with two parallel bars of bulging brain tissue (Fig. 1a). The question arises of whether the seemingly braided coiffure does not in fact refer to an exencephaly, especially since cranial NTDs occur in nearly half of the diprosopus cases. The same applies to the double-faced ceramic figurine in Fig. 1e (500–400 B.C.E.), but with the difference that her head shape is reminiscent of anencephaly with a narrow hair ring between the forehead and the membrana cerebrovasculosa, typically covering the exposed skull base (Fig. 1b). The date of origin of numerous other double-faced Tlatilco figurines found in different geologic strata is given as the early (1800–1200 B.C.E.) and middle (1200–400 B.C.E.) preclassical periods [2‐4]. They show various degrees of facial duplication and different hairstyles, referring to different observations. With respect to the higher prevalence of diprosopia in Latin America, it seems worth noting that these early observations concern a pre-Hispanic Mesoamerican population and cultural environment.

In Schedel's world chronicle, published in 1493, we found another early illustration of a diprosopus showing a man with four eyes but only one nose and one mouth in the interfacial midline [5] (Fig. 1h). This is certainly not an accurate rendering of 'partial' diprosopus. When two eye pairs are present, one would expect duplication of the forebrain, each developing a set of telencephalic vesicles with a rhinencephalon, that are responsible for the formation of two sets of olfactory placodes in between two lens placodes each, and finally of the localization of two noses in between two eyes each [6] according to the graphics of Foerster [43], Bendersky [2] and Bidondo et al. [1]. Thus, we may assume that the artist had not personally witnessed this event, but rather gained knowledge of it by hearsay, when collecting whimsical phenotypes for his world chronicle.

More difficult is the interpretation of Paul Klee's (* 1879) water colour painting of a diprosopus from 1933, which marked the beginning of his 'late work’ in the year of his escape from Germany into his Swiss exile (Fig. 1f). This creative period was overshadowed by a severe sclerodermia that led to a first respiratory crisis in 1935 and to his early death in 1940. The illness gave him the feeling of constriction and disintegration of body parts, known to have influenced his late works. We may thus consider this painting named 'parent's image' (in literal translation 'parent's mirror') as a perceived symbolic facial duplication with hopefully intact central nervous system.

Conclusion Brain malformations in patients with diprosopus may not be regarded as an independent mutagenic or teratogenic event but rather as a sequel closely related to the duplication of the notochord and neural plate and as a consequence of the cerebral and associated craniospinal structures. Thus, neural tube closure may be hampered by still-fused incompletely duplicated neural plates, thus explaining the high rate of cranial NTDs. Focal structural instability caused by duplication of bones of the predominantly anterior skull base may give rise to anterior encephaloceles at osseous junctions. HPEs may develop independently from an additional causative HPE gene variant and in Janiceps twins by secondary fusion of primarily cleaved frontal lobes. The aetiology of diprosopus seems complex. The exclusion of a causative gene variant by WES and the occurrence of diprosopia after ICSI in case 3 may be understood as mechanically induced, supporting a non-genetic aetiology, but cannot rule out with certainty a monogenic or polygenic cause by an underlying germline or somatic mutation. The association of lateral facial duplication with frontal exoparasitic twinning may indicate the involvement of different pathogenic pathways, including ischaemic damage. With regard to the double-faced Tlatilco figurines, DeMyer's statement ‘the face predicts the brain' could be modified to state ‘the hairstyle predicts the brain’.

High resolution computed tomography with 3D-reconstructions (3D-CT) was performed on the foetuses in cases 4 and 5 (HRi). In the foetus in case 7, CT could not visualize the skeleton, indicating postmortem removal in accordance with the extended dorsal skin sutures. A corresponding skeletal museum specimen was not found. CT imaging was performed with a third generation Dual Source scanner (Somatom Force, Siemens Healthineers, Erlangen, Germany) in spiral mode with the foetus in the supine position. Imaging parameters were 150 kVp, fixed tube current of 50 mAs, 0.4 mm section thickness, 0.3 mm section interval, ultra-high-resolution kernel (Ur69), iterative reconstruction (Sapphire) strength 5, and transverse orientation of the sections. Multiplanar images of the bones were reconstructed using the postprocessing platform Syngo Via (VB40, Siemens Healthineers, Erlangen, Germany). For 3D visualization, cinematic rendering was applied using environmental lightning for the optimal delineation of all three-dimensional structures.

Molecular analysis by WES in case 3 was performed on DNA derived from foetal muscle tissue (M.S.). Foetal DNA samples were prepared following the workflow of the Twist Library Preparation with Enzymatic Fragmentation and Twist Universal Adapter System (Twist Bioscience; South San Francisco, California, USA). For enrichment of exonic regions the “Twist Human Comprehensive Exome” with standard hybridization was used. The final library was paired-end sequenced on an Illumina NextSeq500 sequencer. The obtained sequencing reads were aligned to GRCh37/hg19 using Burrows Wheeler Aligner (BWA-MEM) and further processed in house according to the GATKs best practice protocol for calling single nucleotide variants, insertions and deletions. Evaluation of the called variants was performed using the program VarSeq from Golden Helix Inc®. Variants were classified according to the American College of Medical Genetics guidelines [44].