Genetic testing in inherited endocrine disorders: joint position paper of the European reference network on rare endocrine conditions (Endo-ERN)

Background With the development of molecular high-throughput assays (i.e. next generation sequencing), the knowledge on the contribution of genetic and epigenetic alterations to the etiology of inherited endocrine disorders has massively expanded. However, the rapid implementation of these new molecular tools in the diagnostic settings makes the interpretation of diagnostic data increasingly complex. Main body This joint paper of the ENDO-ERN members aims to overview chances, challenges, limitations and relevance of comprehensive genetic diagnostic testing in rare endocrine conditions in order to achieve an early molecular diagnosis. This early diagnosis of a genetically based endocrine disorder contributes to a precise management and helps the patients and their families in their self-determined planning of life. Furthermore, the identification of a causative (epi)genetic alteration allows an accurate prognosis of recurrence risks for family planning as the basis of genetic counselling. Asymptomatic carriers of pathogenic variants can be identified, and prenatal testing might be offered, where appropriate. Conclusions The decision on genetic testing in the diagnostic workup of endocrine disorders should be based on their appropriateness to reliably detect the disease-causing and –modifying mutation, their informational value, and cost-effectiveness. The future assessment of data from different omic approaches should be embedded in interdisciplinary discussions using all available clinical and molecular data.


Background
Genetic disorders compose a substantial fraction of human diseases, and it is estimated that nearly 5% of live births have a genetically driven illness recognizable until the 25th year of life [1]. However, it is a challenge to diagnose these rare conditions by assessing clinical features and conventional diagnostic testing alone. As a result, many patients and their families undergo a longlasting diagnostic odyssey.
In clinical endocrine practice, genetic testing is primarily requested to confirm a suspected clinical and endocrine diagnosis, in particular in case the clinical features are ambiguous. Additionally, it also contributes to the identification of presymptomatic individuals. Thereby, their risk to develop an inherited endocrine disorder can be predicted, and prophylactic measures might be taken (i.e. thyroidectomy in carriers of specific MEN2 variants). This risk can also be determined in relatives of mutation carriers, and the knowledge on an inherited genetic variant is the basis to advice the patients´family in respect of family planning and prenatal testing. Finally, the precise determination of the molecular alteration causing the endocrine disorder allows to understand its pathophysiology and thereby to develop and apply an adapted therapy.
In recent years, high throughput genetic tests (i.e. next generation sequencing (NGS)) have become increasingly available for clinical use at reasonable costs, and significant progress has been achieved regarding the detection rate in human genetic diagnostic testing [2,3]. As a result, a diagnostic yield of up to 40% can currently been achieved in genetic heterogeneous disorders, depending on the precision of the clinical assessment and the disorder itself (e.g. [4], for general review: [5]). Molecular alterations also play a major role in tumor development, and NGS has turned out to be an appropriate tool (for review: [6]) for tumor profiling as the basis for treatment and prognosis. Accordingly, genetic testing has become an indispensable component of the comprehensive diagnostic workup in pediatric endocrinology, and increasingly also as part of adult endocrine diagnostics (Table 1), in addition to the common biochemical laboratory analysis. Consequently, the demand for genetic testing continues to increase, and the physician asking for genetic analysis should be aware of the indications for testing, of the used methods and their chances of success, but also of their limitations.
This position paper of the European reference networks on rare endocrine conditions (ENDO-ERN; www. https://endo-ern.eu) summarizes the current role of genetic testing in the diagnostic workup of (inherited) endocrine disorders and emphasizes the chances and advantages of modern genetic tests as well as the accompanying challenges and limitations. The paper will mainly refer to molecular germline variants and congenital disorders and will not cover genetic testing of somatic variants in tumors, which requires different considerations.
Due to the permanent improvement in diagnostic testing and increasing number of genetic factors associated with endocrine disorders, this review can only provide an overview on testing strategies and available tests. The authors therefore kindly ask the readers to visit curated and public databases like orphanet (https://www.orpha. net/consor/cgi-bin/Disease_Search.php?lng=EN) to get an up-dated overview on available diagnostic tests and therapies.

Relevance of genetic testing for the management of endocrine disorders
The decision on the application of genetic tests is not only based on considerations about the benefit for the patient and therapeutic options which can be inferred from the genetic test result, but should also be based on the feasibility of genetic tests, their availability, informational value, and cost-effectiveness. It should be mentioned that predictive genetic testing in children without therapeutic consequences is not indicated and even violates the law in some countries.
The prerequisite of a targeted and efficient genetic test is the comprehensive evaluation of phenotype (deep phenotyping) and recording of the medical history by using a standardized and curated terminology (e.g. Human Phenotype Ontology (HPO)), which helps to specify the order for genetic testing (Fig. 1). In many instances, specialized endocrine diagnostic approaches with baseline and dynamic tests are required and may be supplemented by dedicated functional imaging procedures and specific hormonal analyses. Furthermore, precise molecular diagnosis may direct laboratory evaluation to "condition-specific target ranges" rather than to comparison to usual reference ranges [8] (Table 1).
An example for the need of precise clinical data as the basis for targeted genetic testing is short stature [9] where the first assessment comprises the analysis of growth parameters like height, weight and head circumference, as well as growth velocity and the skeletal features (bone deformities, demineralization, etc.). Endocrine tests (e.g. serum IGF1 concentration) may point towards the basic endocrinopathy and narrow down the specific defect, such as IGF1 deficiency or IGF1 resistance. Another example is the determination of Anti-Müllerian hormone (AMH) concentration in serum of patients suspected of Persistent Müllerian Duct Syndrome (PMDS), which identifies relevant genes to be sequenced: If serum AMH is undetectable the AMH gene should be sequenced, whereas analysis of the AMH-R gene is indicated in case AMH is normal/high in PMDS. The central role of the precise molecular diagnosis as a decision aid for personalized clinical managements has meanwhile been shown for a broad range of endocrine disorders (Table 1), ranging from tumor predisposition syndromes [10] to disorders characterized by growth defects [9], glucose and insulin homeostasis (for review: [11]), obesity and lipodystrophy, hypogonadotropic hypogonadism [12], disorders of skeletal metabolism [13], and disorders of sexual development [14].
The therapeutic relevance of a precise genetic diagnosis can be illustrated for patients with growth disturbance disorders associated with molecular defects affecting the chromosomal region 11p15.5, i.e. Silver-Russell and Beckwith-Wiedemann syndrome (SRS, BWS). SRS is characterized by severe intrauterine and postnatal growth retardation, body asymmetry, feeding difficulties, relative macrocephaly at birth and characteristic facial features (for review: [15]). Among these features growth retardation is in the focus of the clinical management, and therapy is mainly based on recombinant growth hormone (rGH) treatment. The clinical heterogeneity results in an overlap with several other congenital growth retardation disorders and clinical misdiagnosis. These differential diagnoses comprise several tumor predisposition disorders (e.g. Bloom syndrome (OMIM #210900) or Mulibrey nanism (OMIM #253250)), for which rGH treatment is contraindicated [16]. In summary, the precise determination of the molecular cause of growth retardation in patients with SRS features is already nowadays required as the basis of a personalized therapeutic intervention (e.g. rGH treatment, tumor risk assessment). BWS is another example of an imprinting disorder associated with a broad spectrum of molecular alterations in 11p15.5. However, in contrast to SRS, BWS is characterized by overgrowth and an increased risk for embryonal tumors (for review: [17]). In fact, an association between specific molecular subtypes and tumor predisposition has been identified, and therefore the precise diagnosis of the molecular change in 11p15.5 has an impact on the tumor screening monitoring program [18]. Another example for the value of precise genetic subtyping for clinical decision-making is congenital hyperinsulinism due to K ATP channel mutations, where biallellic mutations in either ABCC8 or KCNJ11 predict a diffuse type, while a single paternally inherited recessive mutation is highly suggestive of a focal type and may entail specific imaging and curative treatment (e.g. [11]).
Medical history should be accomplished by documentation of the family history of at least three generations andif applicableconsanguinity and ethnicity. The family history might help to delineate the mode of inheritance and therefore provide hints at the disease-causing gene and mutation. However, the clinical manifestation and penetrance of genetic diseases can be highly variable even within the same family, therefore even minor   Table 1) clinical symptoms in apparently unaffected family members should also be checked thoroughly. Furthermore, there is a growing number of inherited disorders that do not exhibit the classical modes of inheritance, i.e. autosomal dominant, autosomal recessive or gonosomal inheritance ("Mendelian disorders"), but which differ from these rules. Mitochondrial inheritance is one example, as the respective disorders follow and exclusively maternal transmission pattern. Other examples of non-Mendelian inheritance are imprinting disorders like the Prader-Willi and Silver-Russell syndromes, in which the sex of the parent transmitting the molecular basic mechanism contributes to the phenotypic expression (for review: [19,20]), as well as some trinucleotide disorders like the FMR1-associated premature ovarian failure and fragile X syndrome [21].

Molecular alterations in endocrine disorders
Though the majority of pathogenic variants consist of pathogenic variants affecting only single nucleotides (single nucleotide variants, SNVs), there are further types of molecular alterations which can be associated with endocrine disorders (Table 1). SNVs as well as losses, gains or rearrangements (e.g. deletions, duplications, insertion-deletions / indel) of a small number of basepairs commonly have an impact only on a single gene, whereas larger copy number variants (CNVs) might have an effect on several genes. In addition to alterations of the DNA itself, modification of imprinted gene clusters can be altered. These epimutations can result in the disturbance of the fine-tuned monoallelic expression of imprinted genes which are expressed either from the maternal or the paternal gene copy.
In the majority of known inherited endocrine disorders, the variant is either inherited and follows a classical Mendelian trait (i.e. autosomal-recessive, autosomaldominant, X-linked) or arises de-novo (in case of autosomal-dominant mutations). However, in the latter case these variants as well as epimutations might arise after fertilization, and can therefore occur as somatic mosaicism, meaning that not all cells of an organism carry the variant. In case of mosaicism, the ratio of cells with different (epi)genotypes can differ considerably, as demonstrated for Silver-Russell syndrome and McCune Albright syndrome (e.g. [22,23])( Table 1). In some disorders, mosaicism is a well-known observation with a significant impact upon clinical manifestation and transmission risk (e.g. Neurofibromatosis type 1 [24]). As the presence of mosaicism can definitely not be excluded, testing of different tissues might be considered for every negative genetic test or particularly if an appropriate genetic test for a distinct phenotype comes back negative. Thus, the possibility of an undetected mosaicism should be discussed in a molecular genetic report if appropriate.

Genetic tests and their applications in endocrine disorders
Until recently, the detection of genomic variants of different sizes and nature often required the application of a step-wise process due to the limitations of the tests, accordingly this procedure was expensive and timeconsuming. The parallel analysis of several genes, or even the comprehensive analysis of the whole genome by NGS, is a quantum jump in routine molecular diagnostics. In heterogeneous disorders with hundreds of genes known to cause similar and overlapping phenotypes (Table 1), these factors can now be analyzed within the same diagnostic run and assessment pipeline. Additionally, in case of NGS formats addressing the whole exome or the whole genome (WES: whole exome sequencing; WGS: whole genome sequencing), new genetic causes of diseases can be identified. Thus, the capability of genomic NGS is enormous, but in a diagnostic context it should be applied after estimating the advantages and disadvantages ( Table 2). In fact, the estimation of the pathogenicity of genomic variants obtained by both WES and WGS even in proteincoding genomic regions is a major challenge. As illustrated in Fig. 2 for a patient with an unspecific growth retardation phenotype, WES results in a huge number (> 50, 000) of genomic variants. As WGS addresses hundredfold of base pairs as many as WES, the number of genomic variants grows exponentially, and accordingly their interpretation might be extremely laborious. Thus, the bioinformatic pipelines need to become further automated to facilitate the interpretation of data.
The decision on a genetic testing algorithm to confirm the clinical diagnosis of an endocrine disorder should consider both the range and types of pathogenic variants, as well as the appropriateness of the tests (Tables 1, 2): In endocrine disorders presenting with characteristic phenotypic expression and caused by pathogenic variants in only one gene, like multiple endocrine neoplasia type 1 (MEN1 gene) or von Hippel-Lindau syndrome (VHL gene), single gene testing is recommended. In some genetically heterogeneous disorders like Beckwith-Wiedemann syndrome and transient neonatal diabetes mellitus, a step-by-step molecular analysis might be meaningful, starting with the most frequent alterations. On the other hand, this stepwise strategy might not be useful in case of a suspected disorder with a genetically heterogenous background, variable expression and/or incomplete penetrance like permanent neonatal diabetes mellitus or hypothyroidism (for review: [25]). However, the causative genetic factors of several endocrine disorders can not be identified by these approaches, in these patients the application of WES and/or WGS should be Table 2 Currently applied methods in human genetic diagnostics of endocrine disorders: Applications, advantages and limitations. The methods can roughly be discriminated in respect to main type of molecular alteration they address, though some of them can also identify other changes. (*The currently used conventional diagnostic often address either copy number variants (CNVs, i.e. deletions and duplications) or single nucleotide variants (SNVs). In fact, CNVs represent a mutational burden in several genetic disorders. Therefore, parallel CNV assessment using alternate supplemental methods is normally required. For their identification, (semi)quantitative assays have been developed, and in human genetic testing multiplex ligation-dependent probe amplification (MLPA) is a broadly implemented diagnostic tool. However, the development of bioinformatics CNV pipelines for NGS data is in progress (e.g. [7]), and CNV detection by NGS is already in establishment. (*Multigene panels can either be based on targeted enrichment assays by which only the regions of interest are enriched in the wetlab, or they can be defined as a virtual WES dataset which has been filtered and analysed for the region of interest only. FISH: fluorescence in-situ hybridization, ASO: allele-specific oligonucleotide, MLPA: multiplex ligation-dependent probe amplification, SNP: single nucleotide polymorphism, CGH: comparative genome hybridization; WES: whole exome sequencing; WGS: whole genome sequencing; TGS: third generation sequencing; VUS: variant of unknown significance) Method considered. In fact, the deciphering of a disease-causing genetic change in these patients can be more successful if samples from the parents are also included in the analysis (trio-analysis).
In summary, specific tests targeting variants and genes will also be applied in the future in disorders with a small spectrum of mutations testing of which provides a high detection yield, whereas NGS based assays are used in genetically heterogeneous entities.

Technical aspects of genetic testing in the NGS era
The chosen assay also influences the source and amount of the patients´sample. For conventional cytogenetics, viable cells are required, whereas most molecular tests, including NGS, are based on genomic DNA which can be principally isolated from all freshly drawn or archived tissues. However, the molecular strategies also differ in the need of DNA.
The use of NGS gene panels allows to select and target specific groups of genes, and for some disorders it is therefore the first choice of testing. In contrast, NGSbased approaches covering hundreds of unselected genes or even the whole genome might lead to the inclusion of factors in the diagnostic setting that are functional candidates. The huge number of variants obtained by exome or genome-wide approaches reflect the variability and Table 2 Currently applied methods in human genetic diagnostics of endocrine disorders: Applications, advantages and limitations. The methods can roughly be discriminated in respect to main type of molecular alteration they address, though some of them can also identify other changes. (*The currently used conventional diagnostic often address either copy number variants (CNVs, i.e. deletions and duplications) or single nucleotide variants (SNVs). In fact, CNVs represent a mutational burden in several genetic disorders. Therefore, parallel CNV assessment using alternate supplemental methods is normally required. For their identification, (semi)quantitative assays have been developed, and in human genetic testing multiplex ligation-dependent probe amplification (MLPA) is a broadly implemented diagnostic tool. However, the development of bioinformatics CNV pipelines for NGS data is in progress (e.g. [7]), and CNV detection by NGS is already in establishment. (*Multigene panels can either be based on targeted enrichment assays by which only the regions of interest are enriched in the wetlab, or they can be defined as a virtual WES dataset which has been filtered and analysed for the region of interest only. FISH: fluorescence in-situ hybridization, ASO: allele-specific oligonucleotide, MLPA: multiplex ligation-dependent probe amplification, SNP: single nucleotide polymorphism, CGH: comparative genome hybridization; WES: whole exome sequencing; WGS: whole genome sequencing; TGS: third generation sequencing; VUS: variant of unknown significance) (Continued) complexity of the human genome. In fact, the majority of variants represent non-pathogenic polymorphisms which also occur in control cohorts (Fig. 2). Thus, the key for a successful and efficient NGS data analysis is the stringent filtering by bioinformatic pipelines which commonly refer to (a) databases of genetic variants in humans, (b) pathogenicity prediction tools, and (c) different modes of genetic inheritance.
a) With the increase of biological data ascertained by high-throughput omics technologies, the demand of databases on biological information has increased and the number of repositories is permanently growing (Nucleic Acids Research references 180 databases [26]). These curated databases daily exchange and update new data based on raw data from high-throughput laboratories. b) The freely or commercially available in-silico tools to predict the pathogenicity of a genomic variant mainly rely on its biochemical, structural, and functional properties, and its evolutionary conservation across species. Recent studies that compared the performance of the major prediction tools applied in genetic diagnostic testing have revealed a diverse picture of their reliability (for review: [27]). Therefore, the determination of pathogenicity should also consider additional information including variant frequencies obtainable from databases (see (a)), and segregation analyses in a family (see (c)), andif possiblefunctional analysis. c) Segregation analysis seeking for the association of a genetic variant with the phenotype in a family is an appropriate tool to corroborate its pathogenicity.
The combination of these information should support the laboratory to delineate the pathogenicity of a genomic variant. With the guidelines for interpretation of genomic variants suggested by the American College of Medical Genetics [28], a widely accepted system for variant classification has been developed (Table 3). Whereas the classification of a variant as benign/likely benign or pathogenic/likely pathogenic either excludes or confirms its pathogenicity, the prediction of a considerable number of variants remains ambiguous (so-called variants of unknown significance -VUS). With the application of WES or WGS, the number of detectable variants including VUS grows exponentially, and NGS data therefore requires a stringent variant filtering (Fig. 2). These aspects should clearly be addressed before the application of such a broad test to avoid misunderstandings and unrealistic expectations, and the patients and their families should be informed about these scenarios with the help of appropriate patient counseling, information and informed consent forms [29]. However, the physician should be aware of the slight possibility that genomic variants which are classified as pathogenic/likely pathogenic at the time of diagnosis might be downgraded to benign later due to an increase of knowledge and datasets.
Another challenge in the diagnostic use of WES/WGS is the handling of incidental findings, i.e. genetic alterations associated with conditions or diseases unrelated to the patient's present condition for which current tests are being performed but with important clinical ramifications. To circumvent these putative outcomes, which are difficult to manage in routine diagnostic workup and counseling, to reduce the costs for consumables and to avoid excessive amounts of data, targeted NGS panels have been established as an appropriate tool for NGS-based genetic testing (Table 1). However, multigene panels are not suitable for analysis of patients with unspecific phenotypes. In this situation the use of so-called "clinical exomes" might be discussed which target protein-coding regions of all genes for which disease-causing variants have been reported (e.g. "Kingsmore panel"). In fact, not all advantages and disadvantages of the different assay formats can be addressed in this paper, and every month new improvements of wet-lab and bioinformatic NGS tools as well as functional assays and suitable models to further characterize new variants are being published. For unusual phenotypes and challenging diagnostic scenarios, it is therefore recommended to contact laboratories experienced in NGS analysis in time for the up-to-date NGS testing strategies in connection with the pathology to be explored.
Laboratories offering genetic tests should implement a quality management system [30]. It should follow the national rules, but it should be leaned on the latest version of the ISO15189 standard, which specifies requirements for quality and competence in medical laboratories. Participation in external quality assessment schemes is a further key element of quality assurance in molecular genetic diagnostics, and these schemes help to test the laboratory workflow as well as the proficiency of data interpretation and reporting.
With the publication of guidelines for NGS testing, the European Society of Human Genetics has undertaken an essential step towards an international standard of NGSbased diagnostics [31].

Conclusions
The implementation of NGS assays in DNA testing has significantly increased its diagnostic yield [4,32], but it still leaves a considerable number of patients with an unusual clinical phenotype without molecular confirmation. With the rapid development of wet-lab assays and bioinformatic NGS pipelines it can be expected that the increasing application of NGS, as well as the improvements of databases and software tools underlying its data interpretation will significantly increase the rate of cases with a solved molecular basis. By complementing genomic NGS data with transcriptome (RNAseq) as well as methylome data and information from multiple omic sources, future diagnostic approaches will additionally become more comprehensive [33].
The growing knowledge on the contribution of genetic factors to endocrine disorders and the rapid implementation of new molecular tools in the diagnostic settings makes the interpretation of diagnostic data increasingly complex. Therefore, the data assessment should be embedded in interdisciplinary discussions using all available clinical and molecular information. Therefore, the metabolic and hormonal assessment remain fundamental. However, WES, WGS and further NGS formats are indispensable tools to identify new pathophysiological mechanisms of human disorders and to improve diagnostic algorithms. In the future, nearly all genetic alterations will be addressable by comprehensive NGS approaches.
Finally, the knowledge on the genetic cause of a disease does not only allow a precise clinical management, but it also helps to avoid invasive and expensive diagnostic tests which burden the patient, and lead to a faster diagnosis allowing an earlier and therefore more effective medical intervention (for review: [25]). The early diagnosis of a genetically based disorder supports the patients and their family in their self-determined planning of life as early as possible. Furthermore, it allows an accurate prognosis of recurrence risks for family planning as the basis of genetic counselling. Asymptomatic carriers of pathogenic variants can be identified, and prenatal testing might be offered, where appropriate.  -Not all parameters of pathogenicity are fulfilled.
-Bioinformatics prediction of pathogenicity but without final confirmation.

No clinical significance
Likely benign Benign -Epidemiological and bioinformatics data indicate that the variant is not pathogenic.
-These variants are commonly not reported but might be available on request.