E-cadherin signal sequence disruption: a novel mechanism underlying hereditary cancer

The heterozygous germline mutation c.38_46del, leading to the amino acid deletion p.L13_L15del, was identified by direct sequencing in a 33-year old woman from New Zealand (patient A). Histological examination of gastric specimens revealed that the proband was affected by signet ring cell (diffuse) carcinoma. One paternal aunt and a cousin were diagnosed with the same type of neoplasia and died at 40 and 30 years of age (Fig. 1a). The affected cousin, as well as patient A’s father were carriers of the same genetic alteration. Of note, this mutation was not found in the largest database of human genetic variation to date (gnomAD: The Genome Aggregation Database) comprising several thousand of unrelated individuals [6].

To determine the pathogenicity of the L13_L15del CDH1 mutation, we first studied the conservation of the signal peptide. We performed a multiple sequence alignment of the E-cadherin amino acid sequence from different species and of P-cadherin, given the similarity of both cadherins with respect to their cell-to-cell adhesive function and epithelial expression [3]. Although most of the sequence is variable, the signal hydrophobic core is highly conserved across different cadherins (Fig. 1b). The p.L13_L15del mutation affects this specific region by removing three amino acids from the six that comprise the hydrophobic region, which suggests its possible functional relevance. Accordingly, PROVEAN software predicts a deleterious effect for this mutation with a score of − 6.102 (score ≤ − 2.5 is considered deleterious, Additional file 1: Table S1). A putative effect on the signal peptide cleavage was also tested and, while the predicted cleavage site of the wild-type and the variant sequences remains unchanged, the probability of the variant to generate a functional signal peptide is greatly decreased (Additional file 2: Figure S1). In addition, we could confirm p.L13_L15del mutation as “likely affecting signal peptide quality” through the inhibition of protein translocation to the endoplasmic reticulum (ER) membrane (min(∆C)= − 0.321 > min(∆S)= − 0.430) [7].

To assess the impact of the p.L13_L15del alteration in vitro, we transfected E-cadherin negative cells with vectors encoding the wild-type protein and the variant, as well as the empty vector, as a control (Additional file 3: Methods and Materials). Despite similar transfection efficiencies in all conditions, we verified that total protein levels were significantly reduced in the mutant cells, when compared with the wild-type expressing cells (Fig. 1c). An abnormal pattern of E-cadherin localization was also detected by immunofluorescence in most of the cells. In contrast to the strong membrane staining presented by the great majority of the wild-type cells, the p.L13_L15del cells displayed very low levels of E-cadherin at the membrane and, occasionally, aberrant cytoplasmic accumulation of the protein (Fig. 1d). Quantitative evaluation of the staining showed more intense E-cadherin in the wild-type cells when compared with those expressing the p.L13_L15del (Fig. 1e). This difference is also reflected in fluorescence intensity exhibited at the plasma membrane.

Taken together, these results indicate that the p.L13_L15del variant affects a conserved region of the E-cadherin signal peptide and impacts protein expression (the total level and the membrane fraction).

To investigate the functional significance of p.L13_L15del variant, we tested cell invasive properties and cell-to-cell adhesiveness (Fig. 1f-h). In contrast to the transfection of wild-type E-cadherin, which significantly decreases the number of invasive cells, the variant is not able to efficiently suppress invasion through a matrigel matrix. Regarding adhesion, wild-type cells form large and compact aggregates with an average area of 20,277 pixels², contrasting with p.L13_L15del expressing cells, which exhibit an isolated appearance with cellular structures of 5662 pixels². Overall, our findings strongly support the pathogenic nature of the p.L13_L15del variant.

Finally, we determined the molecular mechanism underlying the deleterious effects of p.L13_L15del at the signal peptide of E-cadherin. Possible alterations at the RNA level were first evaluated by real-time PCR. We verified that CDH1 mRNA levels were not significantly changed both in wild-type and in the variant conditions, despite the huge difference at the protein level (Fig. 2a). To exclude protein trafficking deregulation or premature degradation by quality control mechanisms, we treated the cells with DMSO chemical chaperone and MG132 proteasome inhibitor. Upon DMSO treatment, no effect was detected in either E-cadherin total expression, number of E-cadherin positive cells, or staining intensity of mutant cells (Fig. 2b-f). Proteasome inhibition induced a significant increase in p.L13_L15del protein levels and a slight increase in the number of cells expressing E-cadherin, as well as in the number of molecules present at the membrane. However, these increased levels were significantly different from those observed in the wild-type cells.

With this set of experiments, we demonstrated that the deleterious effect of p.L13_L15del is not dependent on protein trafficking deregulation or on its early degradation.

To test the involvement of post-translational machinery in the regulation of the p.L13_L15del mutant, we set up a cell-free system for in vitro protein translation (Fig. 3b). The approach allows coupled transcription and translation of a specific DNA sequence without the action of intermediate and regulatory cellular moieties. Vectors encoding the wild-type and the mutant CDH1 cDNAs were used as templates for the production of E-cadherin molecules. E-cadherin synthesis was subsequently detected by immunoblot and it was verified that the wild-type and the p.L13_L15del plasmids produce similar amounts of the protein, in contrast to the clear difference observed in the cell model system (Fig. 3c-g).

To determine the impact of the specific leucine residues from the hydrophobic core in protein translation, a series of mutants was engineered. The sequential deletion of one, two or three leucine residues produces a decreasing effect on the quantity of E-cadherin molecules that are translated by the cells and that are exposed at the cell surface (Fig. 3f-g and Additional file 4: Figure S2). Concordant results were obtained by in silico analysis of these mutants, where the p.L13_L15del always displayed stronger deleterious effect than the p.L14_L15del, and the p.L15del behaved as a nearly neutral variant (Additional file 1: Table S1). Remarkably, in the cell-free system, no decreasing effects were observed in protein levels of any of the mutants when compared with the wild-type condition. These findings demonstrated that p.L13_L15del impairs the interaction of E-cadherin with post-translational machinery, decreasing protein synthesis, ER import and membrane activity. Further, we verified that the impairment of protein translation is dependent on the extension of the signal peptide core disruption.

Interestingly, two different germline mutations affecting the signal peptide core and which do not induce truncated CDH1 forms were previously identified in diffuse gastric cancer cases: c.44_46del (p.L15del) and c.46insTGC (p.L15dup) [8‐10]. Taking into account our results, both mutations are unlikely to cause HDGC. In fact, we demonstrated that p.L15del does not affect the function of the signal peptide, and in the case of p.L15dup, the integrity of the signal peptide core is preserved, which is indicative of normal translation and translocation into the ER.