Intellectual disability (ID) and multiple congenital anomalies (MCA) affect 1–3% of the population [1] and represent a large and heterogeneous group of disorders. The molecular bases still remain unresolved in a large proportion of cases due to this high heterogeneity, which makes their diagnosis challenging. Improvements in genomic investigation techniques in recent decades have allowed the identification of numerous genes in ID and MCA, first with array-comparative genomic hybridization (CGH), which gives a diagnostic yield of around 14% in ID/MCA [2], and more recently with exome sequencing (ES), which raised the diagnostic yield to around 28.8% [3]. For example, ES allowed the identification of about 555 genes between 2010 and 2015, including ZBTB20 [4].

The first description of Primrose syndrome was made by Primrose in 1982, in a 33-year-old male with ID, muscle weakness of the lower limbs, calcified ear flaps, bone abnormalities, and a torus palatinus. Today, it is characterized by a possibly recognizable but largely underdiagnosed entity that associates ID, autism spectrum disorders, facial features, ectopic calcifications, and distal muscle wasting. Other findings, such as corpus callosum anomalies, tall stature, bilateral cataract, hearing loss, hypothyroidism, and diabetes mellitus have been reported, in particular in older patients [5,6,7,8]. The implication of the ZBTB20 gene in this syndrome was established in [7], through the analysis of eight patients, all carrying heterozygous de novo missense variants in ZBTB20, detected by trio ES. Functional studies showed a dominant negative effect of missense variants affecting the DNA binding domain of this transcription factor. This gene plays a role in glucose metabolism, postnatal growth, and neurogenesis. It contains an N-terminal domain involved in protein–protein interactions and five zinc finger C2H2 domains binding the regulatory sites of α-fetoprotein promoters. It has been suggested that the ZBTB20 gene was strongly implicated in the phenotype of 3q13.31 deletion syndrome, which associates ID, corpus callosum agenesis, skeletal malformations, and facial features [9].

Here, we report on 14 new patients: 6 patients with intragenic ZBTB20 missense SNVs, 8 patients with either 3q13.31 microdeletions encompassing ZBTB20, and 1 with a ZBTB20 truncating indel, collected through a national collaboration based on the French AnDDI-Rares network (Figs. 1 and 2). The aim of this study was to collect and compare the clinical features observed in patients carrying ZBTB20 missense SNVs one the one hand, and ZBTB20 microdeletions or truncating variant on the other hand, in our series and a review of 37 patients in the literature, and to determine whether patients with haploinsufficiency of ZBTB20 are at risk of a progressive phenotype, as described in patients with missense variants.

Fig. 1: Pictures of the patients with ZBTB20 missense variants.
figure 1

a Patient 1. b Patient 2. c Patient 3. d Patient 5.

Fig. 2: Pictures of the patients with ZBTB20 microdeletions.
figure 2

a Patient 7. b Patient 8. c Patient 9. d Patient 11.

Patients and methods


Two groups were defined according to whether the proband carried a ZBTB20 missense SNV (dominant negative group) or a deletion involving ZBTB20/ZBTB20 truncating variant (haploinsufficiency group). The medical history and clinical features of each patient are detailed in Supplemental data.

Molecular and cytogenetic analyses

Written and informed consent were obtained from all patients or legal guardians. Peripheral blood samples were provided in a diagnosis context for the proband and both parents, when available.

Panels of different sizes were used for the diagnosis of patients 1, 3, 4, 5, 6, and 14. For patients 1, 3, 6, and 14, a panel comprising 456 genes involved in cognitive disorders was used (Strasbourg University Hospital). Sequence libraries preparation and coding region capture were performed [10] with in-solution enrichment methodology (Agilent XT2 or QXT SureSelect custom panel) and sequenced with an Illumina NextSeq 550 instrument (paired-end sequencing 2 × 150 bases). Reads were aligned on the hg19 human genome reference sequence, and SNVs and indels were called with the Genome Analysis Toolkit v.3.4.46 thanks to an in-house (Strasbourg Hospital University) pipeline (STARK) and following the Genome Analysis Toolkit (GATK) best practice. Annotation and analysis of the variants were performed using Varank [11]. For patients 4 and 5, the diagnostic panel test included 19 genes implicated in syndromes associating overgrowth and ID (Hôpital Necker, Paris). Sequencing was performed on a MiSeq instrument (Illumina), after multiplex PCR using the TruSeq Custom Amplicon (Illumina). Reads were aligned on the hg19 human genome reference sequence with the Burrows–Wheeler Aligner. Variant calling was done using the GATK. Annotation and analysis of the variants were performed via a local interface PolyWeb.

Exome capture and sequencing were performed for patient 2, in a research framework by the GAD team aiming to identify the molecular bases of marfanoid syndrome with ID. Sequence libraries preparation, coding region capture, and the in-house bioinformatics pipeline are detailed elsewhere [12].

In all cases, the presence of the ZBTB20 variant in the proband and in the parents was validated by Sanger sequencing.

Array-CGH was performed using Agilent microarray 44 K for patients 7 and 9, 60 K for patients 8, 12, and 13, 180 K for patient 10 and 11. In all cases, the deletion was confirmed by fluorescence in situ hybridization experiments in the proband using bacterial artificial chromosome (BAC) clones containing chromosome 3 specific sequences, in accordance with publicly available genome resources (NCBI Map Viewer, Santa Cruz Human Genome Browser). The BACs were obtained from the RP library (BACPAC Resources Center, CHORI, Oakland, CA, USA) and selected according to their positions on chromosome 3. The same method was used for segregation in the parental samples.

Identified variants have been submitted to freely accessible public databases: SNVs and indel to ClinVar (IDs SCV000930642 to SCV000930648) and CNVs to Decipher (IDs 388596, 388597, 388611, 388612 to 388615).

Statistical analysis

Clinical features were compared within the missense SNV group, with an expected dominant negative effect, and the group comprising CNVs and the truncating ZBTB20 variant, with an expected haploinsufficient effect, in patients gathered in this study and from the literature. χ2 tests and Fisher’s exact tests were used when appropriate. A p value below 0.05 was considered significant.


Molecular and cytogenetic results

Molecular results of the SNV group are summarized in Table 1 and Fig. 3. All missense SNVs occurred de novo and impacted the zinc finger domains of the protein, at its C-terminal extremity. Of note, the variant of patient 2 (c.1862T>C; p.(Leu621Pro)), presented an amino-acid substitution in the same position as that in subject 11D5028 reported in 2014 [7] (c.1861C>T; p.(Leu621Phe)).

Table 1 Molecular data of patients with ZBTB20 SNVs/indel reported in this series.
Fig. 3
figure 3

Position of the ZBTB20 intragenic SNVs and indel of the patients 1–6 and 14.

The extents of the CNVs are summarized in Table 2 and Fig. 4. The size of the 3q13.31 microdeletions ranged from 2.78 to 20.35 Mb, and contained 15 to 56 RefSeq genes, including 3–11 OMIM genes. The CNVs start and end positions listed in Table 2 only represent the minimal extent of the deletions detected by using CGH-array. All CNVs were found to be de novo.

Table 2 Cytogenetic and molecular data of patients with ZBTB20 CNVs.
Fig. 4
figure 4

ZBTB20 deletions of patients 7–13 showed in UCSC Genome Browser.

Genotype–phenotype correlations

The clinical features of all patients of this study are presented in Table 3.

Table 3 Clinical features of the 14 reported patients.

The comparison between the missense SNV and CNVs/truncating variant groups was extended to patients reported in the literature (Table 4). The early truncating variant, located in the first exon above the BTB domain, prevents the synthesis of a functional protein. Therefore, this variant is expected to lead to haploinsufficiency. The median age was 15.1 years in patients with missense SNVs and 9.4 years in patients with CNVs/truncating variant. All the patients presented ID and/or mild-to-severe psychomotor delay. Behavioral features, stereotypies, and anxiety were found in both groups. One patient of each group developed schizophrenia, 2/23 with a missense SNV, and 5/34 patients in the CNVs/truncating variant group were diagnosed with autism. ADHD was found only in the CNVs/truncating variant group. Facial features of the two groups overlapped, with macrocephaly, prominent forehead, downslanting palpebral fissures, ptosis, and large ears. Ectopic calcifications, progressive muscle wasting, and contractures were observed only in patients with missense SNVs although clinical significance is reached for hearing loss (p = 0.002), observed in 18/23 in missense SNV group versus 1/8 in CNV/truncating variant group. In addition, corpus callosum anomalies, diabetes, and hypothyroidism were more frequent in this group. Again, clinical significance is reached only for corpus callosum anomalies (p = 0.00004) and hypothyroidism (p = 0.047). Skeletal malformation, mainly deformations of the knees, feet, and spine were present with no significant difference between the two groups (65.2% in SNVs and 56.3% in CNVs). However, the differential median age between the two groups requires caution in the interpretation of results.

Table 4 Genotype–phenotype correlations in this series and patients of the literature.


Primrose syndrome is a rare syndrome, usually unrecognized in childhood. Indeed, patients display a specific association in late childhood or adulthood, with the emergence of progressive distinctive features, such as hearing loss, pinnae calcification, endocrine manifestations, and muscle wasting. Previous authors have mentioned that patients with missense SNVs may be more severely affected than patients with CNVs [7], but a large genotype–phenotype correlation was needed.

To date, the progression of the symptoms has not been described in patients with CNVs, but this finding has to be interpreted with caution since the mean age at the time of reports in patients with missense SNVs is significantly higher than the mean age of patients with CNVs (15.1 years for SNV patients and 9.4 years for CNVs/early truncating variant patients, including this series, and data from the literature). Indeed, patients with CNVs are now rapidly diagnosed since array-CGH is widely available and considered as a first-intention genetic test in developmental disorders. In the future, easier accessibility to large panels and ES should decrease the time to diagnosis for patients with SNVs.

In our study, ID and/or psychomotor delay were observed in all patients and ranged from mild to severe in both groups. Patients in the two groups shared the same facial features. Birth weight was generally normal in both groups. Overgrowth was a rare feature, but observed in both groups. The most distinctive difference was the presence of perceptive hearing loss, diagnosed in the majority of affected patients from the missense SNVs group, whereas in patients with CNVs only one patient developed hearing loss at the age of 14 months. The mean age at the diagnosis of the hearing loss cannot be determined from the literature data, but the youngest patient of the missense SNVs group, aged 2 years, had not yet been diagnosed with hearing loss. However, not all CNV patients had been screened, and this could make it difficult to draw definitive conclusions. Similarly, cerebral and pinnae calcifications have never been reported in ZBTB20-CNV patients and they therefore appear to be specific to patients carrying ZBTB20 missense SNVs. The age at onset of ear calcifications is unknown. In our study, it has been reported only in a 40-year-old female patient. An 11-year-old patient had been reported in 2018 with pinnae calcification [13]. However, this feature had not been explored by specific imaging in all cases. A higher rate of corpus callosum anomalies was found in the missense SNVs group, even though anomalies were observed in both groups [9]. Concerning endocrine complications, hypothyroidism was also more frequent in the missense SNVs group (21.7% vs 3.2%), with variable age of onset, from 2 years and 7 months in the present series to 43 years in the literature [6]. In addition, diabetes mellitus was mostly reported in patients with missense SNVs (10% vs 3.2% in the CNVs group). Diabetes mellitus was usually reported in adulthood (in a 43 and 23-year-old patients with missense SNV and in a 41-year-old patient with CNV), although an 8-year-old patient with glucose intolerance has been described [14]. Muscle wasting is often reported in association with missense SNVs (6/23), but never reported in patients with CNVs.

Among patients with ZBTB20 CNVs, the phenotype could also be explained by the deletion of other adjacent genes in the 3q13.31 region. Patients 9, 10, and 11 also had a deletion comprising the LSAMP gene. This gene encodes a limbic system-associated membrane protein. Significant associations between the LSAMP gene and schizophrenia or neuropsychiatric features have been found [15]. DRD3 is close to ZBTB20 and encodes a dopamine receptor. Polymorphisms in this gene are involved in susceptibility to schizophrenia and essential tremor [16, 17]. Altogether, 32 patients from the CNV group had a DRD3 deletion, and only one 41-year-old patient with a ZBTB20 and DRD3 complete deletion and partial deletion of LSAMP had schizophrenia [18]. Since one 40-year-old patient was also diagnosed with schizophrenia in the missense SNVs group, we cannot exclude that ZBTB20 could contribute to schizophrenia.

In data from the study and the literature, there were no findings present in patients with CNVs and absent in patients with missense SNVs, in favor of the major role of ZBTB20 in the 3q13.31 CNV phenotype. Functional studies performed by transfection experiments were in favor of a dominant negative impact of missense ZBTB20 SNVs [7], compatible with a more severe phenotype in patients with a missense ZBTB20 SNV than in patients with a ZBTB20 CNV. The difference in the mean age of patients remains an obstacle and makes it difficult to draw definitive conclusions since the majority of the differential features is progressive. The long-term follow-up of audiometric and endocrine parameters in patients with ZBTB20 CNVs is needed in order to determine risk and monitoring recommendations for these patients.

In conclusion, we compared the clinical features of six patients with missense ZBTB20 SNVs and eight patients with ZBTB20 CNVs or ZBTB20 early truncating variant, associated with a review of the literature. We highlighted the existence of a genotype–phenotype correlation between ZBTB20 missense SNVs and haploinsufficient variants, with a tendency toward a more severe and progressive clinical presentation in missense SNVs, although a longer follow-up of patients with CNVs is needed to draw further conclusions.