Premature fusion of calvarial sutures, i.e., craniosynostosis (CS), represents a highly heterogeneous neurocranium malformation (Morriss-Kay and Wilkie, 2005). The disease can be classified following at least three criteria: the origin, the number of affected sutures, and the occurrence of additional clinical features. Regarding the etiology, CS is distinguished as a primary condition, i.e., genetic; or secondary, if it arises from mechanical, metabolic, or hormonal defects. The assessment of affected sutures’ numbers allows recognising single or compound CS, and finally, a syndromic CS is described when additional features other than secondary to premature suture fusion occur. Otherwise, CS should be classified as an isolated condition (Wilkie et al., 2007; Johnson and Wilkie, 2011; Lattanzi et al., 2017).

The neonatal skull comprises six calvarial sutures—one metopic, one sagittal, two coronal, and two lambdoid, closing physiologically from 3 months to late 50 years of postnatal life. Calvarial sutures are fibrous junctions, which allow the skull to grow and develop during the expansion of the brain and permit skull compression during delivery (Baer, 1954; Opperman, 2000; Rice, 2008). Thus, the lack of physiological sutural patency impedes the allometric cranial growth, resulting in cranial deformities and, frequently, increased intracranial pressure, i.e., craniostenosis. Consequently, affected patients present with facial dysmorphism, cortex lesion, seizures, intellectual disability, visual and hearing impairments, or breathing difficulties that are all secondary to CS (Renier et al., 1982; Thompson et al., 1995; Tubbs et al., 2001; Gupta et al., 2003; Mathijssen and Arnaud, 2007; Chieffo et al., 2010).

CS affects approximately one in 2,500 births and burdens public health due to the requirement of extensive surgical treatment in the first year of life and multi-level specialist medical care in the subsequent postnatal periods (Wilkie et al., 2010; Lattanzi et al., 2017). Despite recent advancements in genetic diagnostics, the pathogenesis of CS remains still unknown or partially understood. The large cohort screenings reveal genetic etiology in barely 21–62% of all recruited cases, depending on the size of the study, ethnicity of the population, and range of the molecular analysis. Conversely, about 40–80% of CS cases remain molecularly unresolved (Roscioli et al., 2013; Paumard-Hernandez et al., 2015; Timberlake et al., 2016; Lee et al., 2018; Topa et al., 2020). In this paper, we have presented the study results encompassing 166 individuals in whom we had applied a three-step diagnostic algorithm to identify different mutation types. In addition, we have also performed a statistical analysis of the clinical data we had obtained.

All procedures involving human participants were performed under the Helsinki Declaration. The Institutional Review Board of Poznan University of Medical Sciences granted ethics approval (no. 742/17). All patients agreed to participate in this study, and written informed consent for participation and publishing the information and images in an online open-access publication was obtained from all participants and the parents of minors before genetic testing.

We summarized all differentiating phenotypic features in our cohort in Table 1 and Supplementary Table S3. Patients involved in this study were more frequently sporadic (81%) than familial cases (19%)–χ² (1;166) = 62.67; p < 0.001. We have not revealed any differences between the occurrence of isolated (56%) and syndromic forms of CS (44%)–χ² (1; 146) = 2.22; p = 0.14. However, we have reported more frequently single CS (66%) than multiple CS (34%)–χ² (1;144) = 14.69; p < 0.001. Next, we have also shown that the most often affected suture in isolated CS was coronal (53%), followed by metopic (33%), sagittal (15%), and lambdoid (0%)–χ² (3;95) = 51.54; p < 0.001.

Our diagnostic success was 43% (n = 72). The following paragraphs described the detailed results, entitled Tier 1–3, which were also summarized in Tables 2–4. Mutations located within the FGFR1, FGFR2, FGFR3 genes, and in the TWIST1 gene constitute the most often occurring alterations (73%)–χ² (1;71) = 14.43; p < 0.001. This diagnostic indicator occurred significantly statistic (χ2 (1;52) = 4.92; p < 0.05) more often in a female group of patients (48%, FGFR2 was the most frequently affected–19%) than male–25% (the most common pathogenic variants were located in the TWIST1 gene–12%).

Regarding sex, we have not revealed any relevant diagnostic success rate changes, which were made in 62% among female patients and 38% among male patients–χ² (1;71) = 3.61; p = 0.06. Next, we have shown that more often, we could diagnose the patients affected with multiple CS (65%) than single CS (35%)–χ² (1;49) = 4.50; p = 0.03. Detailed results were summarized in Supplementary Table S4.

Craniosynostosis represents a highly heterogeneous medical condition whose etiology has not been yet fully elucidated. The results obtained from cohorts screened worldwide showed that the molecular background could be indicated in merely 21%, to propitiously 62% of patients (Roscioli et al., 2013; Paumard-Hernandez et al., 2015; Timberlake et al., 2016; Lee et al., 2018; Topa et al., 2020). Positive genetic testing is mainly achieved among subgroups with syndromic CS. The reported germline mutations are usually classified as point mutations, however, chromosomal aberrations, copy number variations, minor exonic deletions/duplication, or biallelic inheritance were also reported in the medical literature (Timberlake et al., 2016; Goos and Mathijssen, 2019; Yilmaz et al., 2019).

In this study, we have reported 166 individuals affected with different forms of CS in whom we had applied a three-step diagnostic algorithm (Tier 1–3) (Figure 1). To our best knowledge, this is the first large CS patients’ screening in which multi-leveled methods, including chromosomal aberrations and CNVs detection, were applied. The proposed approach allowed us to identify an exact genetic cause in around 43% of all CS patients. Since our multi-leveled molecular diagnostic strategy of CS patients is unique and previously unreported, we were unable to directly compare all the results obtained here with previous similar research studies. This is because all reports that have been submitted so far, were aimed at identifying only point mutations via targeted Sanger sequencing, targeted gene panel NGS or ES. Similar to other researchers analyzing these mutations type occurrence, we have shown that causative variants in the FGFR1, FGFR2, FGFR3, TWIST1, and TCF12 genes account for most common cause of CS (Roscioli et al., 2013; Paumard-Hernandez et al., 2015; Lee et al., 2018; Topa et al., 2020).

Notably, FGFR1, FGFR2, FGFR3 variants mainly occur in hot-spot positions and, along with TWIST1 gene mutations, were analyzed in Tier 1. Undeniably, this step was crucial in the CS testing algorithm and cost-effective compared to other genetic methods (73% of all diagnoses;–χ² (1;71) = 14.43; p < 0.001). On the other hand, four novel variants in the TCF12 gene and additional FGFR2, FGFR3 alterations (other than hot-spots) were found using custom gene panel NGS (Table 2). No other gene included in the applied custom gene panel housed more than one pathogenic variant. Interestingly, based on the medical literature, the EFNB1 gene is usually reported as the sixth most commonly affected gene in CS (Paumard-Hernandez et al., 2015; Miller et al., 2017; Lee et al., 2018). However, we postulate that a very characteristic disease, i.e., craniofrontonasal dysplasia (CFND) resulting from pathogenic variants of the EFNB1, should be considered a standalone genetic disorder in which features other than CS guide the proper diagnosis (Bukowska-Olech et al., 2021). Moreover, CFND represents a classic viscerocranium defect, whereas CS is a neurocranium abnormality. Hence, we have excluded all individuals suggestive of CFND and consequently have not reported EFNB1 mutations in this study.

Next, we have revealed that six patients with syndromic CS carried pathogenic variants in genes involved in epigenetic regulations such as ARID1A (P134), KMT2A (P137), KMT2D (P114), NSD1 (P58, P99, P166) (Table 2), resulting in Coffin-Siris syndrome type 2, Wiedemann-Steiner syndrome, Kabuki syndrome type 1, and Sotos syndrome type 1, respectively. Such Mendelian disorders, i.e., those resulting from disruptions of epigenetic processes, were termed chromatinopathies. They are all characterized by intellectual disability, immune deficiencies, or skeletal anomalies. However, CS was rarely described among them (Zollino et al., 2017). Occasionally, CS has been reported only in Kabuki syndrome and Sotos syndrome thus far (Tatton-Brown et al., 2005; Martínez-Lage et al., 2010; Topa et al., 2017). The second group of Mendelian diseases in which CS has been noted are RASopathies resulting from the dysregulation of RAS/MAPK pathway. Retrospectively, we have described here one patient carrying CNVs in which BRAF gene deletion occurred (P141) (Bukowska-Olech et al., 2020b). Similar to our finding, other researchers have also highlighted the co-occurrence of both CS and RASopathy, resulting from mutations in other components of RAS/MAPK pathway (Kratz et al., 2009; Takenouchi et al., 2014; Ueda et al., 2017).

Custom genes panel allowed us to detect novel pathogenic variants–p.Arg132* in the ERF gene (P129), and p.Ser391* in the ZIC1 gene (P131), resulting in Craniosynostosis 4 and Craniosynostosis 6, respectively (Twigg et al., 2013, 2015). Both ERF and ZIC1 are newly recognized CS-related genes, however, only a few cases carrying variants in those two have been reported (Twigg et al., 2015; Miller et al., 2017; Glass et al., 2019). In addition, we have evaluated one variant in the ZIC1 gene as VUS p.Ser404Pro (P130) since it was present in the patient’s healthy father. However, this alteration was absent from the gnomAD v3.1.2 database (accession date: 3 December 2021). Finally, ES revealed two additional alterations in individuals with syndromic CS–p.Arg1295* in the MN1 (P63), which was recently discovered, and subsequently linked to CS, and p.Arg305Gly in FAM111A (P136), resulting in Kenny-Caffey syndrome type 2, in which CS represents an unseen clinical feature (Table 2) (Mak et al., 2020).

Importantly, we have noted an apparent gap regarding screening for chromosomal aberrations or CNVs among CS patients (Lattanzi et al., 2012; Poot, 2019). To our knowledge, no major CS groups were analyzed via karyotyping or CMA, therefore most research data describing microscopic chromosomal changes or submicroscopic CNVs causative for CS were published as single case reports (Villa et al., 2007; Marques et al., 2015). Besides, only a few chromosomal aberrations and CNVs known thus far represent recurrent changes underlying CS (e.g., deletions in 7p21, 9p22-p24, and 11q23-q24 or duplication in 5q33.3), however none of them was present in our cohort (Reardon et al., 1993; Shiihara et al., 2004; Poot, 2019). All genomic losses or gains reported in this research were not commonly associated with CS (Budisteanu et al., 2010; Dilzell et al., 2015). Hence, we could not recommend additional loci to be screened regarding the cohort of syndromic CS, especially those associated with intellectual disability. Here, karyotyping followed by GTG banding allowed us to detect two intrachromosomal deletions 7q32.3-q35, and 18q21.32-q23, both resized to chr7:131837067-144607071, and chr18:620405559-80247644, respectively. Next, using CMA, we have detected four CNVs including three duplications–1q22-q23.1 (chr1:155961428-157217426), 2p21 (chr2:44990857-45008348), 17p13.3 (chr17:847955-1641601), and one deletion 5q35.3 (chr5:177277901-177283748) from which the largest encompassed 1.3 Mb, whereas the smallest 5.8 kb (Sowińska-Seidler et al., 2018; Bukowska-Olech et al., 2020a). The detailed mutations description following the International System for Human Cytogenomic Nomenclature (ISCN) was listed in Table 3; for a list of genomic mutations’ content, see Supplementary Table S5. Notably, in most chromosomal aberrations or CNVs identified here, CS occurred as an additional phenotype. The above findings suggest that karyotyping and CMA cannot be replaced by targeted chromosomal testing when applied in a cohort of syndromic CS associated with an intellectual disability or developmental delay.

Regarding the results presented in this study, we would like to point to the possible underestimated co-occurrence of CS and intellectual disability. Our clinical experience suggests that CS may be overlooked when intellectual disability constitutes a primary clinical complaint. Hence, we recommend calvarial sutures’ evaluation in patients with intellectual disability. On the other hand, in any case of already detected syndromic CS and intellectual disability, the possible occurrence of clinical features suggestive for either chromatinopathies or RASopathies should also be considered (Kratz et al., 2009; Cao et al., 2017; Zollino et al., 2017; Davis et al., 2019).

Undeniably, the molecular diagnosis of CS should distinguish its isolated or syndromic form, which presence determines the subsequent diagnostic steps. In addition, syndromic CS should be classified as a disorder associated with intellectual disability or a disorder without intellectual disability (Figure 2). Targeted PCR and Sanger sequencing of FGFR1, FGFR2, FGFR3, TWIST1, and TCF12 genes resulted in the highest diagnostic rate in our cohort of craniosynostosis patients strongly recommend analysing those genes first (isolated CS and syndromic CS without intellectual disability). Because of many advantages of NGS-based methods, including mosaicism detection, screening those genes using targeted genes panel via NGS would be optimal. However, based on our results, we suggest applying a custom genes panel limited to the fewer genes, such as FGFR1, FGFR2, FGFR3, TWIST1, TCF12, ERF, ZIC1, RECQL4, and NSD1 (Siitonen et al., 2009; Sharma et al., 2013; Twigg et al., 2013, Twigg et al., 2015; Twigg and Wilkie, 2015; Miller et al., 2017; Kutkowska-Kaźmierczak et al., 2018; Lee et al., 2018; Bukowska-Olech et al., 2020c; Topa et al., 2020). In the case of syndromic CS and intellectual disability, the genetic investigation should start from chromosomal aberrations or CNVs detection. Lastly, when ES data bioinformatic analysis is performed, genes associated with RASopathies and chromatinopathies should be considered.

The heterogeneity of CS is enormous, resulting from either anatomical variability of the disorder, in which different types and number of sutures can be affected, or epidemiological aspects, including isolated presentation of CS or occurrence of various accompanying symptoms (Kutkowska-Kazmierczak et al., 2018). It has also been shown that CS may be detected in at least 180 different syndromes, which often are rare and in which CS is not a pathognomic feature. Besides, some researchers have also documented or postulated the association between CS and two-locus inheritance. Consequently, molecular causes of the disease seem to be complex in some CS individuals and, as presented in this study, the pathogenic variant or affected gene may be restricted to only one individual (Sharma et al., 2013; Flaherty et al., 2016; Timberlake et al., 2016, Timberlake et al., 2018; Wilkie et al., 2007; Timberlake and Persing, 2018). However, in many CS patients, including our cohort, genetic causes of observed phenotypes remain unrevealed. One may suspect the presence of deep-intronic and regulatory variants, polygenic inheritance, or even epigenetic influences. Another explanation may be the technical limitations of currently available diagnostic methods. It has been shown, for example, that the application of long-read sequencing in NGS-based methods may clarify the genetic background in many unresolved cases (Fujimoto et al., 2021; Hiatt et al., 2021; Miller et al., 2021; Rastegar and Yasui, 2021). Considering the above, the next steps that we should consider to implement for diagnosis of our unsolved cases include whole-genome sequencing, RNA-seq, whole-genome bisulfite sequencing, or long-read ES.

To conclude, our research may constitute a significant source of epidemiological information as we have presented precise phenotypic and genetic data derived from 166 consecutive CS patients of Caucasian origin. We yielded a 43% diagnostic success rate using the presented approach, highlighting the high occurrence of pathogenic variants within “classic” CS genes, i.e., FGFR1, FGFR2, FGFR3, TWIST1, and TCF12. Moreover, we have critically summarized the applied diagnostic methods (Figure 1) and proposed the optimized, cost-effective diagnostic algorithm, which may be helpful in a daily diagnostic routine of various CS’ types (Figure 2).