4.1. Genetic Counseling
A multidisciplinary approach can ensure optimal management of the CDH1 germline mutation carriers. Genetic counseling by clinical geneticists with expertise in the field is a critical component of the risk assessment. To identify individuals with HDGC, the clinical evaluation should include collection and review of personal and family history with detailed three-generation family pedigree and confirmation of DGC or LBC diagnoses. The latest IGCLC consensus guidelines [6
] suggest CDH1 germline testing in a proband with one or more of the following features—(a) personal history of GC with one or more relatives with GC, regardless of age, in which there is at least one confirmed DGC; (b) personal history of DGC before 40 years; (c) personal or family history of DGC and LBC, one diagnosed before 50 years; (d) personal history of bilateral LBC or family history of two or more cases of LBC diagnoses before 50 years; (e) personal history of DGC and personal or family history of cleft/lip palate; (f) in situ SRC or pagetoid spread of SRC. Criteria will be updated in the new IGCLC guidelines.
Family history should be considered positive in presence of one or more first- or second-degree relatives. It is to note that the suspicion of HDGC could be underestimated in individuals with unknown or limited family history. Initial testing should be considered in an affected proband and when more than one family member is affected with cancer, consider starting from individual with a confirmed diagnosis of DGC or with youngest age at diagnosis. Unaffected individuals should be tested only when affected family member is not available and informed of possible limitations in interpreting test results. An individual (healthy or affected) with a known familial pathogenic (or likely pathogenic) variant could be offered gene testing for the specific familial variant.
Genetic testing in individuals younger than age 18 years could be considered in families with cases of early-onset [96
]. Of note, diagnosis of DGC before 20 years is rare [97
]. Gene testing of individuals younger than age 18 years requires counseling for both the parents and the child.
Genetic testing can be performed on DNA extracted from blood or buccal samples, except for patients who have received allogeneic bone marrow transplant or with a recent diagnosis of hematologic malignancy (DNA from a fibroblast culture is a preferable sample in these patients).
CDH1 gene analysis should include search of point mutations and large rearrangements by Sanger sequencing and multiplex ligation-dependent probe amplification (MLPA) or by next generation sequencing (NGS). Point mutations (about 30–50% of all variants) might include small intragenic deletions/insertions and missense, nonsense, and splice site variants [4
]. Exon or whole-gene deletions/duplications have been detected in 6.5% of individuals with HDGC and without pathogenic variants on sequence analysis [99
]. CDH1 germline variants are classified according to the IARC 5-tiered classification system in five classes (Class 5: Pathogenic, Class 4: Likely Pathogenic; Class 3: Variant of Uncertain Significance or VUS; Class 2: Likely Benign; Class 1: Benign) [100
]. The ACMG variant classification guidelines published in 2015 [101
] have recently been revised for the analysis of germline CDH1 sequence variants [102
During genetic counseling, the likelihood of a positive test, technical aspects, inheritance pattern, and significance of the possible outcomes of testing, such as positive (i.e., pathogenic or likely pathogenic variant), inconclusive or uncertain (i.e., VUS), or uninformative (i.e., no mutations detected), or true-negative (i.e., absence of the known familial mutation) should be discussed. A discussion on life-time risks of DGC or LBC, and PTG, or options of surveillance, should be provided. The counselee should be informed about potential significance of the test results for the family, and about reproductive options, such as the availability of genetic testing through prenatal and preimplantation genetic diagnosis. Genetic counseling should include discussion on the need of a continuous update of VUS. Testing a family member for a VUS should not be offered in clinical setting but could be considered for research purposes. Patients and families suggestive of HDGC with uninformative results (about 50–70% of cases), might have an undetectable defect in CDH1 gene (i.e., mosaicism, sequence variant in an intron or regulatory region, or others not covered for technical limits) or pathogenic variants in other cancer predisposition genes. Therefore, intensive endoscopic surveillance is recommendable for first-degree relatives of patients meeting criteria for CDH1 germline testing with uninformative results. Nowadays, genetic testing includes two main clinical approaches—single-gene testing or MGPT [103
]. The spread of MGPT has led to the identification of CDH1 germline pathogenic and likely pathogenic variants in individuals without a personal and family history, suggestive of HDGC. This kind of results poses major clinical management challenges [106
]. Secondary findings could be explained by the presence of families or CDH1 germline mutations with reduced penetrance, and there is a strong need for specific studies in order to obtain more data and drive specific clinical management guidelines.
4.2. Psychological Counseling
CDH1 carriers face several challenges, including the burden of cancer treatment, if already affected by DGC or LBC, and the need to decide upon prophylactic surgery if still unaffected. As a result, individuals might experience depressive symptoms [107
], general distress [108
], and anxiety (109), which can impact their decision-making abilities [111
]. Psychological support is, therefore, a key component of the multidisciplinary management of these patients [114
]. To help manage the psychological and medical burden, improve coping skills and the decision-making process, counseling should include a strong psycho-decisional support component [115
]. The approach should be personalized by taking into consideration the patient’s personality, age, lifestyle, psychological well-being, social support, and self-efficacy [115
]. The final aim is for patients to better communicate with their healthcare providers, in order to determine the most adequate preventative, and therapeutic options [120
]. Psychological counselling and decision-making support are important tools through which counsellors and health care providers can strengthen patient autonomy and provide informed self-care aligned to the patient’s own deliberative decisions [121
]. Empowered patients [122
] express higher satisfaction towards the care received, higher trust in healthcare providers [123
], better adherence, and less conflict with healthcare providers [124
This approach is very recent and should be addressed by HDGC and HLBC investigators in the future.
4.3. CDH1 Missense Variants: Challenging Routine Laboratory Tests
The implementation of MGPT has led to a dramatic increase in the identification of missense variants. Determining the clinical relevance of such variants is currently a major goal in genomic medicine [125
In HDGC, CDH1 missense mutations occur in 22% of cases [128
] and a large proportion of variants remain unclassified [102
]. We, therefore, recommend that a complementary set of analyses encompassing familial and population data, as well as in silico and in vitro tests should be carried out to determine putative variant pathogenicity [130
]. The assessment of variant effects cannot be achieved by standalone evidence in a single category nor by population, in silico or in vitro tests, since all approaches have limitations [101
Genetic parameters such as mutation frequency in healthy control population, co-segregation of mutation with the disease within pedigrees, and mutation recurrence in unrelated families should be considered as a first approach [128
]. Variant frequency in different ethnic groups can be assessed through genomic databases, which compile data from large sequencing consortiums with a few to several thousand participants, namely the 1000 Genomes Project (http://browser.1000genomes.org
), the Trans-Omics for Precision Medicine Program (TOPMed; https://www.nhlbiwgs.org/
), or The Genome Aggregation Database (gnomAD; https://gnomad.broadinstitute.org/
). Nevertheless, databases have limitations, including low-quality data, lack of details on the origin of studies or absence of information regarding possible associated phenotypes [101
]. Moreover, the segregation of a variant within a population at low (<1%) or very low (<0.1%) frequencies per se cannot exclude pathogenicity, especially in the presence of clinical and experimental evidence supportive of variant deleterious effects. As HDGC caused by missense mutations has low penetrance [128
], other host genetic and environmental factors are expected to play a role in the presentation of clinical phenotypes and disease onset. A comprehensive analysis of the pedigree is, thus, crucial to evaluate variant significance and disease risk in germline carriers. Nevertheless, this can be challenging, given the small size of the families and lack of information from patient relatives [128
In silico predictions are valuable tools to estimate the degree of conservation of mutated aminoacids within species, their impact on splicing and, ultimately, on the protein structure [128
]. However, the use of multiple programs is mandatory, as different outputs can be obtained, depending on the underlying algorithm [101
]. Predictions are based on the principle that aminoacids conserved across species are functionally relevant and their substitution is likely to affect protein function [128
]. The limitation of this approach is that the degree of conservation of each aminoacid is considered separately and, as such, possible compensatory effects of neighboring positions are not contemplated [128
]. Structural modeling is currently suitable to predict the impact of most missense mutations in E-cadherin native-state stability, by covering the major part of protein, including the prodomain, the extracellular, and the catenin-binding domains. A correlation between variants that induce higher energetic penalties and their in vitro loss of function has been clearly demonstrated [133
In the last few years, a panel of in vitro assays has been developed specifically to assess CDH1 sequence variants [131
]. The workflow starts by transfecting cell lines with vectors encoding the variant and the wild-type protein. Subsequently, protein expression level, protein localization, and main E-cadherin functions (cell-cell adhesion and invasion suppression) are evaluated [131
]. The CHO (Chinese Hamster Ovary) cell line is the conventionally accepted model to perform all tests, as this cell line is completely negative for cadherin protein expression and displays in vitro invasive properties. Upon transfection with the wild-type E-cadherin, CHO cells acquire the capacity to form cellular aggregates on soft agar and become non-invasive through artificial extracellular matrices [130
]. The effect of missense variants on E-cadherin expression level is assessed by western blot. Low E-cadherin expression strongly indicates structural destabilization and premature degradation of the protein through mechanisms of Protein Quality Control [133
]. Immunostaining and its detailed quantitative analysis is then used to address whether the variant induces abnormal patterns of E-cadherin localization [136
]. Occasionally, a band mobility shift can also be observed, suggesting aberrant glycosylation of the protein [138
]. Contrary to wild-type E-cadherin, which is normally present at the plasma membrane, deleterious variants can be diffusely distributed throughout the cell or abnormally accumulated in cytoplasmic regions/organelles [135
]. Remarkably, some deleterious variants are expressed at the plasma membrane, affecting the stability of the interaction between cadherin molecules (on adjacent cells) and increasing protein turnover rates [135
]. Cell–cell adhesive abilities of variants are evaluated by slow aggregation assays. In this technique, a single-cell suspension is seeded on a semi-solid agar substrate and cells with a competent adhesion complex spontaneously aggregate [142
]. It is well-established that cells transfected with the wild-type protein form compact cellular aggregates, while cells with dysfunctional E-cadherin form small cellular aggregates or present an isolated phenotype [134
]. Topological features in cell meshes should also be examined to determine morphological and structural consequences upon adhesion loss. Indeed, it was demonstrated that triangles within cell networks of dysfunctional mutants have bigger areas and edges, when compared with those formed by wild-type cells, which indicates that E-cadherin defective cells are loosely attached and display increased protrusion formation [137
The invasive suppressive potential and interaction of E-cadherin mutant cells with the surrounding extracellular matrix (ECM) are currently being studied using matrigel invasion chambers. Matrigel is the most frequently used ECM in invasion assays and its major advantage lies on its heterogeneous composition [144
]. This matrix contains not only structural proteins, such as collagens, fibronectin, laminin, and proteoglycans, but also a panel of growth factors, mimicking the basement membrane composition in vitro [144
A major limitation of in vitro experiments is the time pressure surrounding genetic counseling. Nonetheless, and despite the fact that these assays are low throughput and technically challenging, their results reflect a broad approach evaluating alterations in protein structure, trafficking, cellular signaling, and function, which would be impossible to predict through in silico analysis [130
]. By following this analytical pipeline, it is possible to determine the functional impact of 85% of E-cadherin missense variants. However, in 15% of cases, these assays are not sufficient to assure a confident result, and their functional significance remains inconclusive. In such cases, other approaches including cell migration analysis, assessment of the interplay between E-cadherin and its binding partners (by proximity ligation assays), as well as evaluation of downstream targets activation can be applied [135
]. In vivo models are also being developed to study variants that still lack a clear functional classification.
Overall, a comprehensive approach aggregating multiple lines of evidence will be crucial to correlate variable effects of missense alterations with disease penetrance and phenotypes. Moreover, given the pleiotropic nature of CDH1 variants, a careful interpretation of data should take into account the disease context, including diffuse GC, LBC, or congenital malformations, so a proper management can be offered to CDH1 variant carriers.
4.4. Bioimaging Strategies to Identify Aberrant E-Cadherin Expression Signatures
For most cancer types, to support the diagnosis and orientation of therapeutic strategies, the analysis of specific proteins is often required, as alteration in expression or localization of proteins is commonly observed. Immunofluorescence is currently performed to identify aberrant patterns of E-cadherin expression, indicative of protein deregulation and functional impairment at the cellular level, while maintaining tissue architecture [131
]. In parallel with a proper staining method, the adequate acquisition and quantification of E-cadherin expression is mandatory. Several high-resolution microscope imaging modalities, such as time-lapse, confocal laser scanning microscopy (CLSM) and spinning disk microscopy can be used in the acquisition of immunofluorescence images for subsequent examination. However, image analysis is mostly qualitative and strongly operator-dependent [149
], which is likely to influence the biological and clinical evaluation of the data. Quantitative approaches are, thus, essential and required to be implemented for an accurate data interpretation.
Over the last decade, a number of software systems has emerged that aims at the quantification of specific features in microscopy images. Nevertheless, these tools often measure total fluorochrome intensity in defined areas, neglecting the expression profile of the target protein in distinct subcellular/cellular/tissue compartments and possibly failing to recognize events of protein delocalization [152
]. Additional limitations include the discrepancy of parameters during image acquisition and quantification among different experiments and the lack of tools to take into account cell heterogeneity, with respect to morphology (size and shape). Therefore, in order to overcome the current limitations and in view of the urgent need to implement automatic quantitative tools, a novel bioimaging strategy was recently developed [136
]. This new pipeline allows for a detailed characterization of protein expression with regard of its level and distribution in intra and intercellular spaces. More specifically, the developed algorithm is able to map specific protein signatures, in images of single cells or cell populations, in a multistep process encompassing automatic selection of cells, networking of cells upon nuclei segmentation and recognition of their geometric centroids [136
]. Fluorescence signals detected between the centroids of two contiguous cells (internuclear profiles) or in radial profiles from a single cell/centroid are then evaluated. Notably, internuclear profiles are of particular relevance to the study proteins located at the plasma membrane or in specific cellular organelles. In contrast, radial profiles should be employed to capture signals throughout the cytoplasm of single cells and are very useful to investigate cytoplasmic proteins. Overall, and as expected, the compilation of all profiles generates a complex and highly heterogeneous protein map, due to the morphological variability of the cells. Therefore, to solve this drawback, geometric compensation techniques are applied, allowing the extraction of a representative profile of protein distribution in the whole cell population [154
]. Geometric compensation is indeed a common procedure used in several image modalities and consists of the estimation of rigid or non-rigid transformations to bring the objects under alignment, as similar as possible in terms of shape and size [155
]. Ultimately, a comprehensive dataset of fluorescence intensities and their respective locations is generated. With this innovative approach, a new window of intervention has emerged, not only to identify protein patterns associated with cancer but also to help disclose their related molecular mechanisms.
The value of this pipeline was demonstrated in different studies addressing the functional impact of E-cadherin variants [136
]. Immunofluorescence images of cells transfected with wild-type E-cadherin and with a panel of cancer-related variants were subjected to the described bioimaging processing. As verified by Sanches and collaborators, cells expressing E-cadherin dysfunctional variants present a distinct profile from that displayed by the wild-type cells [136
]. More specifically, the variants induce a significant decrease in fluorescence intensity at the plasma membrane or aberrant intensity peaks in the cytoplasm. Remarkable differences were also detected between the typical virtual cells produced by the wild-type and variant expressing cells, suggesting abnormal E-cadherin trafficking and protein accumulation in distinct cell compartments [136
]. Based on these results, bioimaging approaches have been proposed as a complementary method to assess the functional relevance of novel E-cadherin missense variants, found in the context of HDGC [6
]. Interestingly, the same algorithm proved to be efficient in the quantification and mapping of a panel of molecules, including P- and E-cadherin, tubulin, and a mitochondria dye, which are distributed in distinct cellular compartments [154
]. This could be of particular importance in diagnostic and research laboratories to unravel protein regulation mechanisms or identify predictive biomarkers of the disease.
In the future, a combined approach of quantitative distribution of biomolecules with morphological parameters such as cell area and cytoskeletal organization, would be of great value to determine molecular pathways involved in disease progression, but also in the analysis of drug-screening strategies, pinpointing compounds that better rescue protein scores and, simultaneously, normal cellular phenotypes. Modern machine-learning and data mining techniques could be explored to develop more sensitive and efficient methods for signal detection and quantitative molecular analysis in tissue samples.