Microenvironment-induced restoration of cohesive growth associated with focal activation of P-cadherin expression in lobular breast carcinoma metastatic to the colon.

This study was designed by MC, PWBD, HK and MG. LK, HC and MC performed the central immunohistochemical staining. LK, IG, SB and UL carried out mutational analyses. MC, MG and HK performed the central histology re‐review. All authors contributed to data collection, data analysis and interpretation. MC, MG and HK wrote the manuscript. All authors reviewed and approved the final manuscript version.

A 52‐year‐old female (referred to as index patient) presented with subileus symptoms. Colonoscopy showed an ill‐circumscribed stenosis of the colon transversum and colon descendens. The patient's medical record provided the following information: (i) systemic scleroderma with Raynaud syndrome and antinuclear autoantibodies (ANAs) of the anti‐RNA polymerase III type (diagnosed 16 months earlier), (ii) screening mammography without suspicious findings (13 months earlier), and (iii) lobular breast carcinoma (stage cT1b, cN0) diagnosed by ultrasound‐guided needle biopsy of the left breast (1 month earlier). The patient had not yet undergone lumpectomy when subileus symptoms developed. A colon mucosa biopsy was taken during colonoscopy. Histologic assessment revealed sparse infiltrates of an adenocarcinoma. Tumor cells formed cohesive tubules in the stroma between colon crypts (Figure  1A ). Nuclei were round and rather monomorphic. The cytoplasm was wide and pale to amphophilic. Tumor cells were negative for intestinal differentiation markers (CK20−, CDX2−, and SATB2−) but positive for mammary markers (CK7+, GATA3+, and ER+) (Figure  1B ). Autochthonous crypts showed the opposite immunophenotype (CK20+, CDX2+, SATB2+ and CK7−, GATA3−, ER−) (Figure  1B ). Crypts were E‐cadherin‐positive and P‐cadherin‐negative, but the tumor cells were E‐cadherin‐negative and P‐cadherin‐positive (Figure  1B , lower right). Accordingly, the colon biopsy presented a diagnostic dilemma: Although histomorphology argued against metastatic ILC because of cohesive growth in tubules, immunophenotypic features were consistent with those known for ILC. Index patient, endoscopic colon biopsy (ID HAN‐4770). (A) Photomicrographs of an H&E‐stained section. A submacroscopic view is presented on the left side. The scale bar corresponds to 3.0 mm. Note that the endoscopic biopsy included only mucosa and did not reach into the tela submucosa. Details of the H&E‐stained section are presented on the right side (×200 magnification). The scale bar corresponds to 200 μm. Note that tumor cells form cohesive tubular elements (pale to amphophilic) in the intercryptal stroma between autochthonous colon crypts (basophilic). (B) Photomicrographs of serial sections subjected to immunohistochemical stainings for intestinal differentiation markers (CK20, CDX2, SATB2) and mammary epithelial markers (CK7, GATA3, ER, E‐cadherin, P‐cadherin). Scale bars correspond to 200 μm. Note that autochthonous crypts are CK20+, CDX2+, SATB2+ and CK7−, GATA3−, ER−, E‐cadherin+, P‐cadherin−, while cancerous tubular elements are CK20−, CDX2−, SATB2− and CK7+, GATA3+, ER+, E‐cadherin−, P‐cadherin+. *P‐cadherin was stained ex post , after the colectomy specimen had already been evaluated. Next, the index patient's colon biopsy was compared with the breast needle biopsy obtained 1 month earlier. The breast biopsy displayed classic ILC with small tumor cells arranged in single files (supplementary material, Figure  S1 ). ILC cells were ER‐positive, PR‐positive, HER2‐negative, and E‐cadherin‐negative and showed minimal or no P‐cadherin expression (supplementary material, Figure  S1 ). Hence, histologic comparison did not clearly support an origin of the adenocarcinoma in the colon from the synchronous ILC in the breast. However, tumor cells were E‐cadherin‐negative in both sites. Two days after the onset of subileus symptoms, the index patient was treated by subtotal colectomy. Gross pathology examination showed diffuse thickening of the colon wall and a stenosis 7.5 cm in length. Histology revealed transmural adenocarcinoma infiltrates. The bulk of tumor cells was located in the tunica muscularis propria and in the tela submucosa and displayed a noncohesive growth pattern in dissociated single cells (Figure  2 ). This growth pattern left no doubts about metastatic ILC. The immunophenotype was also consistent with metastatic ILC (CK7+, GATA3+, ER+, and E‐cadherin−) (Figure  2 ). However, the histology differed completely from the colon biopsy seen before (Figure  1 versus Figure  2 ). Thorough examination of multiple FFPE blocks revealed different growth patterns in different anatomic layers of the colon wall. Non‐cohesive growth was evident in deep layers, namely in the tela submucosa (Figure  2 ) and in the tunica muscularis propria (supplementary material, Figure  S2 ). Strikingly, cohesive growth was evident wherever ILC cells invaded the mucosa (Figure  3 ). In fact, ILC cells formed cohesive aggregates and tubular elements in the stroma of the lamina propria mucosae (i.e. in the stroma between crypts) (Figure  3 ). Conversion from dissociated to cohesive growth was readily apparent on staining for CK7 (Figure  3 ). ILC cells also gained a much wider cytoplasm in this microenvironment (Figure  3 ). IHC demonstrated an EPS phenotype. ILC cells in deep layers were E‐cadherin‐negative and P‐cadherin‐negative (Figure  2 ), but intercryptal ILC cells switched to a P‐cadherin‐positive immunophenotype (Figure  3 ). Of note, multiple, separated foci of mucosa infiltration were evident along the thickened colon segment. ILC cells consistently switched to a P‐cadherin‐positive phenotype and formed tubular elements wherever mucosa infiltration was evident. Overview photomicrographs showing this phenomenon in multiple foci are included in the supplementary material, Figure  S3 . Index patient, colectomy, tela submucosa (ID HAN‐4947). (A) Photomicrographs of an H&E‐stained section. A submacroscopic view of the colon wall is presented on the left side. The scale bar corresponds to 5.0 mm. Numbers on the left margin annotate anatomic layers of the colon wall: 1, mucosa; 2, tela submucosa; 3, tunica muscularis propria; and 4, connective tissue of the mesenterium and subserosa. Details from the tela submucosa are presented on the right side (×200 magnification). The scale bar corresponds to 200 μm. Note the noncohesive, dissociated growth pattern. (B) Photomicrographs of serial sections subjected to immunohistochemical analyses for CK20, CK7, GATA3, ER, β‐catenin (beta‐catenin), p120‐catenin, E‐cadherin, and P‐cadherin. Scale bars correspond to 200 μm. Note that the tumor cells are E‐cadherin‐negative and P‐cadherin‐negative. Index patient, colectomy, mucosa (ID HAN‐4947). (A) Photomicrographs of an H&E‐stained section. A submacroscopic view of the colon wall is presented on the left side (same specimen as shown in Figure  2 ). The scale bar corresponds to 5.0 mm. Numbers on the left margin annotate anatomic layers of the colon wall: 1, mucosa; 2, tela submucosa; 3, tunica muscularis propria; and 4, connective tissue of the mesenterium and subserosa. Details from the mucosa are presented on the right side (×200 magnification). The scale bar corresponds to 200 μm. Note the growth in cohesive aggregates and tubular elements. (B) Photomicrographs of serial sections subjected to immunohistochemical analyses for CK20, CK7, GATA3, ER, β‐catenin (beta‐catenin), p120‐catenin, E‐cadherin, and P‐cadherin. Scale bars correspond to 200 μm. Note that E‐cadherin‐negative tumor cells switch to a P‐cadherin‐positive immunophenotype (EPS phenotype). Loss of E‐cadherin is typically accompanied by downregulation of β‐catenin protein expression and cytoplasmic or nuclear translocation of p120‐catenin [ 30 , 31 ]. EPS partially restores AJ formation and re‐establishes membranous β‐catenin and p120‐catenin expression [ 12 , 17 ]. Accordingly, β‐catenin and p120‐catenin were assessed by IHC. In deep layers, ILC cells were β‐catenin‐negative and showed aberrant cytoplasmic p120‐catenin (Figure  2 ). In the mucosa, however, ILC cells showed restored membranous β‐catenin immunoreactivity (Figure  3 ). Expression of p120‐catenin exhibited partially membranous, partially cytoplasmic immunoreactivity. Taken together, the histomorphology and immunohistochemical features confirmed a bona fide EPS phenotype of metastatic ILC cells in the intercryptal stroma of the colon mucosa. Histologic assessment of the index patient's colectomy suggested a metastasis from ILC. To verify this assumption, we employed the CDH1 /E‐cadherin mutation status as a marker of clonal relatedness [ 12 , 27 , 32 ]. As expected, identical somatic CDH1 mutations were detected in the ILC in the breast needle biopsy (p.Q610*, allelic frequency 42%) and in the adenocarcinoma in the colon (p.Q610*, allelic frequency 38%). Healthy tissue of the index patient showed a wild‐type CDH1 sequence. Cancerous tubular elements isolated from the lamina propria mucosae of the colon by laser‐assisted microdissection also harbored the same CDH1 p.Q610* mutation (supplementary material, Figure  S4 ). Together, this verified that the adenocarcinoma in the colon was indeed derived from the synchronous ILC in the breast. The index patient showed an EPS phenotype in metastatic ILC cells in the colon mucosa. Autochthonous crypts were P‐cadherin‐negative (Figure  3B ). For completeness, we characterized E‐cadherin and P‐cadherin expression in a control series ( n  = 41 cases) comprising normal colon mucosa specimens, inflammatory, hyperplastic, and neoplastic lesions (supplementary material, Table  S2 ). Normal colon mucosa specimens were E‐cadherin‐positive and P‐cadherin‐negative (10/10 cases), consistent with previous studies [ 33 ]. However, discrete P‐cadherin expression was noted in sporadic crypts above gut‐associated lymphatic tissue (GALT) (supplementary material, Figure  S5 ) [ 34 ]. Ulcerative colitis, hyperplastic colon polyps, and adenomatous polyps showed co‐expression of E‐cadherin and P‐cadherin (10/11, 10/10, and 10/10 cases, respectively) (supplementary material, Figure  S6 ). This is in line with the known up‐regulation of P‐cadherin in the colon mucosa in response to injury, inflammation, and neoplastic transformation [ 33 , 35 , 36 , 37 , 38 ]. GATA3 and CK7 expression were also evaluated in the colon control series. As expected, ulcerative colitis, hyperplastic colon polyps, and adenomatous polyps showed focal upregulation of CK7, but these lesions never showed any immunoreactivity for the mammary epithelial cell marker GATA3 (0/41 cases) (supplementary material, Table  S2 ). We wondered how often EPS occurs in colon metastases from ILC. Accordingly, we compiled a reference series of colon metastases from ILC (Table  1 ). Nine institutions from five countries contributed to the reference series. A total of 23 cases were submitted for central re‐evaluation. Five cases were excluded due to insufficient tissue size ( n  = 2) or due to re‐classification as BC of no special type ( n  = 3). The final reference series included 18 patients. Median patient age was 70 years. Colectomy or appendectomy specimens accounted for 13/18 (72%) cases and endoscopic biopsies accounted for 5/18 (28%) cases. Tumor burden differed interindividually. Mucosa infiltration was evident in 13/18 (72%) cases. Among patients with mucosa involvement, 9/13 (69%) cases showed sparse ILC infiltrates, and 4/13 (31%) cases showed bulky infiltrates or ulceration. All cases (18/18, 100%) were CK7‐positive, GATA3‐positive, and E‐cadherin‐negative (Table  2 ). A total of 15/18 (83%) cases expressed either ER or PR or both hormone receptors, and 3/18 (17%) cases were triple‐negative. Deleterious CDH1 mutations were detected in 9/14 (64%) cases (Table  2 ). No CDH1 sequencing results were obtained in 4/18 cases. Tumor specimens, immunophenotypic and molecular characteristics c/m, partially cytoplasmic and partially membranous immunoreactivity; c/n, partially cytoplasmic and partially nuclear immunoreactivity; FISH, fluorescence in situ hybridization; fpos, focally positive immunoreactivity; IHC, immunohistochemistry; na, not available; neg, negative; NGS, next‐generation sequencing; pos, positive; wt, wild type. Splice site mutation: c.1565+1G>T. Splice site mutation: c.48+1G>A. EPS was detected in 5/18 (28%) cases (Table  2 ). EPS occurred either in the subserosa (case 18, ID CEL‐9689), or in the tela submucosa (case 8, ID PEN‐12834), or in the intercryptal stroma of the mucosa (cases 9, 15, and 17; IDs PEN‐2914, LIS‐5751, and BAY‐61949). One case resembled the index patient (case 17, ID BAY‐61949). This large endoscopic biopsy covered mucosa and parts of the tela submucosa (Figure  4 ). In the submucosa, ILC cells showed noncohesive growth in dissociated single cells (Figure  4A , dashed arrows). In the mucosa, ILC cells formed cohesive aggregates or abortive tubules and gained a wider cytoplasm (Figure  4A , solid arrow). In the submucosa, ILC cells were E‐cadherin‐negative and P‐cadherin‐negative (Figure  4B , dashed arrows), but in the mucosa, ILC switched to a P‐cadherin‐positive immunophenotype (Figure  4B , solid arrow). Upregulation of P‐cadherin restored membranous β‐catenin expression (supplementary material, Figure  S7 ). In the other cases with EPS phenotypes, we observed less well‐defined aggregates admixed among noncohesive ILC cells (supplementary material, Figures  S8–S11 ). Taken together, EPS occurred in 28% of colon metastases from ILC. Among 13 patients with mucosa infiltration, one case resembled the index patient and showed EPS and conversion to cohesive growth in the intercryptal stroma of the lamina propria mucosae. Reference series, patient 17 (ID BAY‐61949), endoscopic colon biopsy. (A) Photomicrographs of an H&E‐stained section. A submacroscopic view is presented on the left side. The scale bar corresponds to 2.0 mm. Note that the biopsy included not only mucosa, but also fragments of the tela submucosa (lower gray arrow). Details from the mucosa and submucosa are presented on the right side (×200 magnification). Scale bars correspond to 200 μm. Note that tumor cells form cohesive aggregates in the intercryptal stroma between autochthonous colon crypts (solid arrows, but grow in a dissociated pattern in the tela submucosa (dashed arrows). (B) Photomicrographs of serial sections subjected to immunohistochemical analyses for CK7, GATA3, E‐cadherin and P‐cadherin. Scale bars correspond to 200 μm. Note that the tumor cells are E‐cadherin‐negative and P‐cadherin‐negative in the submucosa (dashed arrows) but switch to a P‐cadherin‐positive immunophenotype in the mucosa (solid arrow). Note that the tumor cell cluster that appears to be a cohesive aggregate on the H&E‐stained sections is a tubular element in the GATA3‐stained section level. Further immunohistochemical staining of this specimen (including β‐catenin) are presented in supplementary material, Figure S7 .

ILC is a special BC type and is defined by noncohesive growth and E‐cadherin loss [ 1 , 2 , 3 , 4 , 5 ]. In a subset of ILCs, EPS partially restores cell–cell adhesion and facilitates the formation of cohesive tubular elements [ 12 ]. We have recently provided a molecular characterization of such tumors, which we have named ILC with tubular elements [ 12 ]. We consider this as a morphological variant of ILC, which is related to EPS [ 1 , 12 , 13 ]. It remains unknown whether environmental conditions and which molecular mechanisms control EPS and whether EPS is a transient state of cellular differentiation [ 1 ]. EPS becomes clinically relevant if it occurs in metastases. In this constellation, cohesive growth argues against metastatic ILC and raises the possibility of a second malignancy. We have previously reported on an ovarian metastasis from ILC, which showed EPS and tubular elements harboring a CDH1 p.S9* mutation (case 3 in Christgen et al [ 12 ]). Originally, this case had been submitted for second opinion consultation to clarify the differential diagnosis between primary ovarian adenocarcinoma and metastatic ILC [ 12 ]. This case exemplified that unawareness of EPS may cause considerable diagnostic uncertainty [ 12 ]. Here we report a new series of ILC cases, which further underscore the relevance of EPS for clinical diagnostics. The colon biopsy of the index patient showed an adenocarcinoma, which formed cohesive tubules between colon crypts. Metastatic ILC seemed to be unlikely. A histomorphological comparison of the adenocarcinoma in the colon biopsy with the synchronous ILC in the breast did not support metastatic ILC. However, four lines of evidence established the diagnosis of metastatic ILC. First, the immunophenotype was consistent with metastatic ILC. Second, identical somatic CDH1 /E‐cadherin mutations (p.Q610*) were detected in the colon and the breast. This verified clonal relatedness [ 12 , 27 ]. Third, tubular elements in the colon biopsy showed an EPS phenotype. Fourth, the colectomy revealed that tumor growth was strictly noncohesive in deep layers of the colon wall. EPS and the unusual growth in cohesive tubular elements were limited to foci of mucosa infiltration. Together, this left no doubt about metastatic ILC. Findings in the index patient hold an interesting biological implication. Of note, EPS and tubular elements were seen in multiple foci of mucosa infiltration along the colon segment. This suggests an intrinsic competence for EPS in ILC cells. Furthermore, this implies that the intercryptal stroma of the lamina propria mucosae is a permissive microenvironment that can induce EPS in ILC cells. The reference series of ILC metastases partially supported this conclusion. EPS was mostly seen in foci of mucosa infiltration (3/5 cases), and one specimen (case 17) closely resembled the index patient's metastasis. In fact, case 17 featured conversion to cohesive growth and EPS in the intercryptal stroma. EPS, as described in the index patient, is not a singular finding. Cases presenting EPS reported in this study also hold a clinical implication. It is not uncommon that endoscopic forceps biopsies fail to grasp deep tissue layers. Accordingly, the tela submucosa is not always represented in colon biopsies [ 39 ]. Given that the colon mucosa appears to be a permissive microenvironment for EPS in some cases of ILC, endoscopic biopsies may be prone to sample ILC infiltrates with a growth pattern that is not representative for the bulk of the lesion buried beneath the mucosa. Awareness of EPS in ILC may thus be helpful for colon biopsy interpretation in diagnostic practice. Metastatic ILC must not be excluded when tumor cells grow in cohesive aggregates or tubular elements. This breaks with a longstanding paradigm, but it is necessary to appreciate the morphological plasticity of ILC [ 1 ]. Another interesting detail is the comorbidity of the index patient. As mentioned above, the index patient suffered from systemic scleroderma with ANAs of the anti‐RNA polymerase III type. This specific form of scleroderma is likely a paraneoplastic disease caused by a direct immune response against tumor neo‐antigens [ 40 ]. However, neither the primary tumor in the breast nor the metastasis in the colon showed an excess of tumor‐infiltrating lymphocytes. At the time of writing of this manuscript, the index patient was alive at 31 months after colectomy, without tumor progression. Limitations of the present study included the descriptive approach and the limited number of colectomy specimens with ILC metastases included in the reference series. However, this was related to the fact that subtotal colectomy is not always considered for palliation of metastatic BC. Experimental cell and/or animal models are needed to further dissect how the intercryptal microenvironment activates P‐cadherin expression in ILC cells. Taken together, the large bowel is a common metastatic site in lobular breast carcinoma. In endoscopic colon mucosa biopsies, the typical noncohesive growth of lobular breast carcinoma may be concealed by microenvironment‐induced EPS in classical ILC cells and conversion to cohesive growth.

Invasive lobular carcinoma (ILC) is a special histological breast cancer (BC) type [ 1 , 2 ]. ILC is defined by noncohesive growth patterns, such as dissociated growth and ‘single files’, as described by Foote, Stewart, Fechner, Azzopardi, and others [ 3 , 4 , 5 , 6 , 7 ]. In most cases, noncohesive growth is due to somatic mutation of the CDH1 gene, which encodes the epithelial cadherin (E‐cadherin) cell adhesion molecule [ 8 , 9 , 10 , 11 ]. The typical immunophenotype of ILC is E‐cadherin‐negative, estrogen receptor (ER)‐positive, progesterone receptor (PR)‐positive, HER2‐negative, and GATA3‐positive [ 1 ]. Focal activation of placental cadherin (P‐cadherin) expression in E‐cadherin‐negative tumor cells occurs in a subset of ILCs (<10% of cases) [ 12 ]. We have termed the switching from an E‐cadherin deficient to a P‐cadherin proficient status as EPS, which provides a short abbreviation. EPS partially restores cell–cell adhesion and facilitates the formation of focal cohesive tubules in ILC [ 12 ]. We have recently provided a molecular characterization of such tumors, which we have named ILC with tubular elements [ 12 ]. We consider this as a morphological variant of ILC, which is related to EPS [ 1 , 12 , 13 ]. P‐cadherin is an alternative classical cell adhesion molecule and is encoded by the CDH3 gene. From an evolutionary perspective, CDH3 arose by duplication of CDH1 [ 14 ]. CDH1 and CDH3 both map to chromosome 16q22.1 and share 66% homology [ 15 ]. P‐cadherin expression is rare in luminal (i.e. ER‐positive and/or PR‐positive) BC (5–23% of cases) [ 12 , 16 , 17 ]. In vitro cell models have shown that, in the absence of E‐cadherin, P‐cadherin alone partially rescues epithelial adherens junctions (AJs), relocates β‐catenin to the cell membrane, and partially re‐establishes cell–cell adhesion [ 17 ]. It is currently unknown what conditions control EPS, which appears to be a transient phenomenon in ILC [ 1 ]. ILC is associated with widespread metastatic dissemination [ 18 , 19 ]. Metastases may be detected at primary diagnosis or as late as >20 years after disease onset [ 20 , 21 , 22 , 23 , 24 ]. Common sites include bone, skin, orbit, ovary, stomach, and the large bowel [ 20 , 21 , 22 , 23 , 24 , 25 ]. Patients with intestinal ILC involvement are mostly treated with palliative endocrine and/or chemotherapy [ 21 ]. Palliative surgery is rarely considered for therapy, for instance, to prevent ileus [ 21 ]. The histology of ILC metastases to the colon has not been studied systematically, so far. The literature is limited to clinical reports [ 21 , 26 ]. Here, we provide a histologic study of 19 patients diagnosed with ILC metastases in the colon. We show that, in a subset of cases, ILC cells acquire an EPS phenotype and switch to cohesive growth, when they invade the lamina propria mucosae. Awareness of EPS may be helpful for the interpretation of colon biopsies from ILC patients.

Tumor tissues included formalin‐fixed paraffin‐embedded (FFPE) specimens from 19 patients with ILC metastatic to the colon (Table  1 ). The index patient was diagnosed with metastatic ILC at the Hannover Medical School (MHH) in 2021. For the index patient, FFPE specimens included (i) a core needle biopsy of the left mammary gland, (ii) an endoscopic colon mucosa biopsy, and (iii) a subtotal colectomy specimen. The reference series included 18 additional patients with ILC metastatic to the colon and/or appendix vermiformis. The reference series was compiled from tissues archives of nine institutions from five countries (Germany, Hungary, USA, The Netherlands, and Portugal). All non‐MHH archival specimens included in the reference series were transferred to the MHH tissue archive in accordance with local ethical guidelines. The reference series included cases diagnosed between 2004 and 2022. Colectomy specimens were preferred. However, colectomy specimens were rare because surgery is not regularly considered for palliation of intestinal metastases [ 21 ]. Biopsy specimens were admitted only if the biopsy size was sufficient for central immunohistochemical re‐assessment. Postmortem sections were excluded, due to common autolytic changes. Archival primary breast tumor specimens corresponding to the patients included in the reference series of colon metastases were no longer available at the time of this study. All available information is included in Table  1 . Control tissues included normal colon mucosa ( n  = 10, mostly from women treated with hemi‐colectomy for endometriosis), ulcerative colitis ( n  = 11), hyperplastic polyps ( n  = 10), and adenomatous polyps ( n  = 10). All specimens were made anonymous for scientific purposes. In addition, informed patient consent was obtained from the index patient. In addition, this retrospective study was approved by the local ethics committee (MHH, Hannover, ID 10166‐BO‐K‐2022). Tumor specimens, histologic characteristics ae, appendectomy; agg, aggregates; bx, biopsy; ce, colectomy; dis, dissociated; ILC, invasive lobular carcinoma; m, mucosa; meta, metachronous; mp, tunica muscularis propria; na, not available; sf, single files; sm, tela submucosa; sol, solid; syn, synchronous; tel, tubular elements; ulc, ulceration. Tubular elements and/or cohesive aggregates located in the intercryptal stroma of the colon mucosa. Aggregates located in the submucosa. Aggregates located in the subserosa. All cases were subjected to central immunohistochemistry (IHC) at the MHH. For IHC, 1‐μm thick sections of FFPE tissue blocks were mounted on superfrost slides (Thermo Fisher Scientific, Rockford, IL, USA). Slides were deparaffinized and rehydrated conventionally and were subjected to immunohistochemical staining on a Benchmark Ultra automated stainer (Ventana, Tucson, AZ, USA). The CC1 mild program was used for antigen retrieval and the ultraView DAB kit (Ventana) for signal detection. Antibodies used for IHC included the monoclonal anti‐E‐cadherin antibody ECH‐6 (1:100, Zytomed, Berlin, Germany) and the monoclonal anti‐P‐cadherin antibody clone 56 (1:100, BD Transduction Laboratories, Heidelberg, Germany). Further antibodies and scoring methods are included in the supplementary material, Table  S1 . Extraction of DNA and mutational analyses were performed as described previously [ 12 , 27 ]. In brief, genomic DNA was extracted with the Maxwell RSC DNA FFPE kit (Promega, Madison, WI, USA) on a Maxwell RSC instrument (Promega). DNA was quantified using a Qubit 2.0 fluorometer (Invitrogen, Darmstadt, Germany) and the Qubit dsDNA HS assay kit (Life Technologies, Carlsbad, CA, USA). Next‐generation sequencing (NGS) of the CDH1 gene was performed with genomic DNA on an Ion S5 system (Life Technologies) using a customized CDH1 NGS panel designed with Ion AmpliSeq™ designer software (version 5.6). This panel covered the complete protein‐coding sequence of the CDH1 gene (16 exons and 882 codons), the 5′‐UTR sequence of exon 1, and the 3′‐UTR sequence of exon 16 with 26 amplicons [ 27 ]. Mean ( n  = 15) mapped reads per sample was 244.414 (range, 52.438–852.359), and mean depths per base was 6.378 (range, 1.123–23.137). Amplicons passed QC parameters when the amplicon coverage was at least 500, with a minimum of 100 reads per sequencing direction. Variants were required to have an allelic frequency of at least 3% to be considered true positives. Variant annotation was performed with the ANNOVAR analysis software and database tools ( http://www.openbioinformatics.org/annovar ) [ 28 ]. BAM files were preanalyzed with the Ion Reporter software (version 5.10.1.0, Thermo Fisher Scientific). Laser‐assisted microdissection was performed as described previously [ 29 ]. In brief, serial sections (5‐μm‐thick FFPE tissue sections) were mounted on a poly‐L‐lysine‐coated membrane fixed on a metal frame. After deparaffinization and H&E‐staining, carcinoma infiltrates were purified from distinct anatomical regions (lamina propria mucosae and tela submucosa) using a CellCut Plus system (MMI Molecular Machines and Industries AG, Glattbrugg, Switzerland) [ 29 ].

Figure S1. Index patient, breast needle biopsy Figure S2. Index patient, colectomy, muscularis propria Figure S3. Index patient, colectomy, mucosa at multiple sites Figure S4. Index patient, laser microdissection and CDH1 sequencing Figure S5. P‐cadherin expression in the normal colon mucosa Figure S6. E‐cadherin and P‐cadherin expression in inflammatory, hyperplastic and neoplastic colon lesions Figure S7. Reference series, case 17, further IHC analyses Figure S8. Reference series, case 8 Figure S9. Reference series, case 9 Figure S10. Reference series, case 15 Figure S11. Reference series, case 18 Table S1. Antibodies and ICH scoring methods Table S2. GATA3, CK7, E‐cadherin and P‐cadherin expression in normal colon and colonic lesions Click here for additional data file.