3D Chromatin Architecture Re-Wiring at the CDH3/CDH1 Loci Contributes to E-Cadherin to P-Cadherin Expression Switch in Gastric Cancer

Simple Summary The cell architecture is created and maintained via adhesion molecules, called cadherins, that keep similar cells bound to each other. E-cadherin is a cadherin encoded by the CDH1 gene, essential in normal epithelia. P-cadherin, encoded by the CDH3 gene, exerts a similar function but is mainly present in tumours. E- to P-cadherin switch is a common event in several epithelial tumours. We unveiled a mechanism explaining the E- to P-cadherin switch (CDH1 to CDH3 switch), which we found to occur in many gastric cancers. We induced this switch by depleting CDH1 in gastric cancer cells, and, as they started to produce CDH3, this rescued cell-adhesion. Despite this rescue, cells bearing the switch showed migratory and proliferative features, commonly observed in aggressive tumours. We observed that CDH1 depletion leads to novel genomic interactions between the CDH1 and CDH3 gene sequences. We identified the sequences needed for these interactions to occur, which seem also responsible for controlling the amount of E- or P-cadherin molecules in the cell. Our data provide evidence that loss of E-cadherin is able to compromise the normal interactions between the CDH1 to CDH3 gene sequences, allowing the expression of P-cadherin in gastric cancers. Abstract Cadherins are cell–cell adhesion molecules, fundamental for cell architecture and polarity. E-cadherin to P-cadherin switch can rescue adherens junctions in epithelial tumours. Herein, we disclose a mechanism for E-cadherin to P-cadherin switch in gastric cancers. CDH1 and CDH3 mRNA expression was obtained from 42 gastric tumours’ RNA-seq data. CRISPR-Cas9 was used to knock out CDH1 and a putative regulatory element. CDH1-depleted and parental cells were submitted to proteomics and enrichment GO terms analysis; ATAC-seq/4C-seq with a CDH1 promoter viewpoint to assess chromatin accessibility and conformation; and RT-PCR/flow cytometry to assess CDH1/E-cadherin and CDH3/P-cadherin expression. In 42% of gastric tumours analysed, CDH1 to CDH3 switch was observed. CDH1 knockout triggered CDH1/E-cadherin complete loss and CDH3/P-cadherin expression increase at plasma membrane. This switch, likely rescuing adherens junctions, increased cell migration/proliferation, commonly observed in aggressive tumours. E- to P-cadherin switch accompanied increased CDH1 promoter interactions with CDH3–eQTL, absent in normal stomach and parental cells. CDH3–eQTL deletion promotes CDH3/CDH1 reduced expression. These data provide evidence that loss of CDH1/E-cadherin expression alters the CDH3 locus chromatin conformation, allowing a CDH1 promoter interaction with a CDH3-eQTL, and promoting CDH3/P-cadherin expression. These data highlight a novel mechanism triggering E- to P-cadherin switch in gastric cancer.


Introduction
Cadherins are tissue-specific transmembrane glycoproteins that play a key role in stable cell-cell adhesion, cellular polarization, maintenance of tissue architecture, and embryogenesis [1,2]. E-cadherin and P-cadherin are encoded by CDH1 and CDH3 genes, respectively, and present a great degree of sequence and gene structure homology [3]. Despite their homology, both proteins display tissue-specific expression patterns and functions, being often mutually exclusive at the cellular level [3]. While E-cadherin is widely expressed in epithelial tissues [4], P-cadherin expression is restricted to organ regions, such as myoepithelial cells of the breast or mesothelia [5]. Whereas E-cadherin acts as a tumour suppressor in most epithelial tissues, P-cadherin de novo expression or upregulation has been linked to mammary cells' proliferation and migration in mice [6], tumour-promotion in gastric [7,8] and breast cancer [9,10], and worse prognosis in breast cancer [11][12][13].
Aberrant cadherin isoform switching occurs in tumours with implications for patients' survival [14,15]. E-cadherin to N-cadherin switch is a mark of epithelial to mesenchymal transition, in which differentiated epithelial cells, expressing E-cadherin, transition to dedifferentiated mesenchymal-like cells, expressing N-cadherin, and are prone to invade and metastasize [16]. Evidence of E-cadherin to P-cadherin switching has been observed in both gastric [17] and breast cancer, with implications for survival in the latter [11].
Mechanisms for E-cadherin to P-cadherin switching have been suggested, mainly through MiR-34a/WNT1 cascade [18], but fail to explain all observed switching phenotypes. As CDH1 and CDH3 are juxtaposed genes, derived from gene duplication, and both belong to the same topological-associating domain (TAD) [19], their expression is most likely co-regulated and dependent on the tissue-specific 3D chromatin architecture. The 3D chromatin architecture tightly regulates gene expression in time and space through physical interactions involving promoters and regulatory elements located in non-coding regions in their vicinity [20]. These regulatory networks are generally tissue-specific [21], and their impairment may interfere with the normal 3D chromatin architecture, leading to genes' misexpression patterns, causing or aggravating disease [22,23].
Herein, we aimed to characterize a potential 3D chromatin architecture re-wiring process, occurring at the CDH3-CDH1 loci, that may contribute to E-cadherin to P-cadherin switch in gastric cancer.

Samples
Gastric cancers from 43 patients were selected for the study, displaying equivalent male to female ratio, range of age at diagnosis, distribution of tumour stage, and histological type. Clinicopathological features were collected from clinical records. The present project was approved by the Ethical Committee of Centro Hospitalar Universitário de São João on 16 March 2017, with internal reference CES072017.

RNA Extraction
Tumour areas with at least 75% tumour cells were selected for RNA extraction using MagMax FFPE DNA/RNA Ultra Kit (Applied Biosystems, Life technologies, Carlsbad, CA, USA), according to the manufacturer's instructions. Briefly, sections of paraffin-embedded tumour samples were de-paraffinized and protease K-digested. RNA was bonded to magnetic beads, pelleted against a magnetic stand, and the supernatant containing RNA washed and eluted. Concentrations were determined using Qubit ® 2.0 Fluorometer (Invitrogen, Life Technologies Europe BV, Porto, Portugal) with RNA high sensitivity assay, and integrity was assessed with a 2100 Bioanalyzer Instrument (Agilent Technologies, Instruments Soquimica, Lisboa, Portugal).

RNA-Seq and Whole Transcriptome Sequencing Bioinformatics Analysis
RNA was sequenced with Ion AmpliSeq™ Transcriptome Human Gene Expression Panel kit in an Ion Chef™ 550 NGS instrument. We obtained sequencing reads aligned to the GRCh37 human genome using TMAP [24] and annotated with USCS GRCh37. Only canonical transcripts were selected. CDH1 and CDH3 mRNA expression was considered positive for a threshold of 10 TPM or higher. Processed total RNA-seq from normal stomach tissue samples ENCSR853WOM, ENCSR471RUK, and ENCSR000AFI was collected from ENCODE [25][26][27].

Cell Lines and Culture
MKN74 human cell line was acquired from the JCRB Cell Bank, cultured in RPMI medium (GIBCO, Fisher Scientific, Porto Salvo, Portugal) supplemented with 10% fetal bovine serum (GIBCO, Fisher Scientific, Porto Salvo, Portugal) and maintained at 37 • C and 5% CO 2 in a high-humidity atmosphere. STR analysis was used to confirm cell identification. Cells were free from mycoplasma contamination.

Generation of CRISPR-Cas9 E-Cadherin Deletion
Single guided RNAs (sgRNAs) were designed to target the CDH1 exon 2 and impair E-cadherin function, resourced using the Zhang Lab tool, currently available at Benchling online platform. Individual sgRNAs (Sigma-Aldrich, Merck Life Science, Darmstadt, Germany) (Supplementary Table S1) were cloned into pSpCas9(BB)-2A-Puro (PX459) V2.0 plasmid (Addgene 62988), in BbsI restriction site (New England Biolabs), according to the method of Ran and colleagues [28]. Each sgRNA-plasmid was transformed into stbl3 cells (Invitrogen), and colonies were screened by PCR and Sanger sequenced (Supplementary  Table S1). sgRNAs were transfected in MKN74 cell line. Briefly, cells were seeded in 12-well plates and grown for 24 h, and pairs of plasmids were transfected with lipofectamine 3000 (Invitrogen), according to the manufacturer's instructions. Puromycin (Merck, Merck Life Science, Darmstadt, Germany) treatment started at 48 h and was renewed every 72 h, until non-transfected cells were dead.

mRNA and Protein Expression Analysis
CDH1/3 mRNA expression was assessed by qPCR in triplicate. Briefly, RNA was extracted using mirVana RNA Isolation Kit (Invitrogen), according to manufacturer's protocol. cDNA was synthesized using 1 µg of template RNA, and SuperScriptII reverse transcriptase (Invitrogen) was used to synthesize cDNA, according to the manufacturer's instructions. CDH1 mRNA expression was analysed by qPCR with KAPA PROBE FAST qPCR Master Mix (2X) Kit (Sigma-Aldrich) and probes for CDH1 (Hs.PT.58.3324071, IDT), CDH3 (Hs.PT.51.5028751, IDT), and 18S (custom assay, IDT) as endogenous control. Reactions were sequenced on a 7500 Real-Time PCR System (Applied Biosystems). Relative expression was normalized for the endogenous 18S control and quantified using the 2 −∆∆Ct method [29].
E-and P-cadherin expression was assessed by flow cytometry in triplicate. Cells were blocked for 30 min with 3% bovine serum albumin-phosphatase buffer saline (BSA, NZYTech) and incubated with primary mouse antibodies HECD-1 or P-cadherin (1:100 dilution; 1 h at 4 • C; Invitrogen), followed by secondary anti-mouse Alexa Fluor 647 antibody (1:500; 45 min at 4 • C; Invitrogen). FACS ARIA (BD Biosciences, Franklin Lakes, NJ, USA) was used to measure mean fluorescence, and Flow Jo version 10 software was used to analyse the data.

Statistical Analysis
GraphPad Prism version 7.00 software (GraphPad Software Inc, San Diego, CA, USA) was used to perform statistical analysis. Student's t-test was used for comparison analysis, assuming equal variance between clones and wild-type samples. Differences were considered significant for p-value <0.05.

Protein Extraction
Cells were lysed with lysis buffer containing 1% (v/v) Triton X-100 and 1% (v/v) IGEPAL in PBS, supplemented with phosphatase (Sigma-Aldrich) and protease inhibitor cocktails (Roche, Mannheim, Germany) at 4 • C. Modified Bradford Assay (Bio-Rad, Hercules, CA, USA) was used to quantify total protein content of the samples.

Protein Identification and Quantification, and Enrichment GO Terms Analysis
Proteome Discoverer 2.1 software was used for identification (UniProt/SwissProt human database using Mascot (Version 2.5.1, Matrix Sciences)) and/or quantification of proteins. Carbamidomethylation of cysteines was set as fixed modification and methionine oxidation as variable modification, with trypsin as selected enzyme. High confidence master proteins with at least two unique peptides and without contaminants were considered. Differentially expressed proteins were selected considering adjusted p-value ≤0.05, abundance ratio ≤0.67 (downregulated) or ≥1.5 (upregulated), and concordance between replicates. ClusterProfiler (v.4.4.4) R package was used for enrichment GO terms analysis (q-value < 0.05). Statistics were performed using R (v.4.2.0).

ATAC-Seq
ATAC-seq data from normal stomach tissue were collected from phase 3 ENCODE project [31]. ATAC-seq libraries were generated from MKN74 cell line and CDH1 deletion Biology 2023, 12, 803 5 of 13 clone, as described previously [32]. A total of 5 × 10 4 cells were lysed and chromatin was tagmented using Tn5 transposase and purified with MiniElute PCR purification kit (Qiagen), following the manufacturer's protocol. The appropriate number of PCR cycles was determined using SYBR Green qPCR. Transposed DNA fragments were amplified by PCR and purified using QIAquick PCR purification kit (Qiagen), following the manufacturer's protocol. Quality control of ATAC-seq libraries was accessed by tapestation, and samples were sequenced in Illumina HiSeqX technology according to standard protocols. ATAC-seq peaks were mapped resourcing to an ENCODE-DCC ATAC-seq bioinformatics pipeline with default parameters for adaptors trimming (cutadapt: -e 0.1 -m 5), alignment (bowtie), quality filtering for mapq < 30, removal of duplicates and mitochondrial reads (piccard), peak calling (macs2), and replicate statistical analysis (idr, p-value < 0.05).

CDH1 to CDH3 mRNA Expression Switch Is a Frequent Event in Gastric Cancer
To understand the frequency of and explore potential mechanisms which underly E-to P-cadherin switching in gastric cancer, we started by analysing RNA-seq data from 43 gastric cancers to evaluate CDH1 and CDH3 mRNA expression. We found that CDH1 downregulation and CDH3 upregulation are the most frequent combination of events in this cohort, occurring in 42% of all cases, followed by expression of both cadherins (33%) ( Figure 1A). Downregulation of both cadherins occurred in 19% of tumours, while only 7% of all tumours expressed CDH1 and lacked CDH3 expression ( Figure 1A). In this cohort, 75% of the tumours expressed CDH3 ( Figure 1A, upper panels), a feature that is acquired in tumours, as normal stomach epithelial cells do not express CDH3 ( Figure 1B). In 42% of all cases, there was a CDH1 to CDH3 expression switch ( Figure 1A, upper left panel).
In 42% of all cases, there was a CDH1 to CDH3 expression switch ( Figure 1A, upper left panel).

E-cadherin Loss of Function Triggers E-cadherin to P-cadherin Switch in Gastric Cancer
Given the abundance of gastric cancers bearing simultaneous loss of CDH1 expression and CDH3 upregulation, we tested whether depleting CDH1 would trigger CDH3 upregulation.
We used sgRNAs flanking CDH1 exon 2 (Supplementary Table S1), and successful removal of exon 2 was confirmed by Sanger sequencing in the MKN74 cell line (Supplementary Table S1). This clone carried a homozygous deletion encompassing the CDH1 exon 2 ( Figure 2A) and presented complete loss of both CDH1 mRNA and Ecadherin protein expression ( Figure 2B). As a consequence of this deletion, both CDH3 mRNA and P-cadherin protein expression increased by 60% (p-value = 0.05 and p-value = 0.037, respectively) ( Figure 2C). The immunofluorescence data highlight the P-cadherin expression increase at the plasma membrane upon E-cadherin expression loss, likely rescuing adherens junctions ( Figure 2D). Indeed, P-cadherin co-localized with adhesion partner β-catenin at the cell membrane in the CDH1 deletion clone, while the parental cell line displayed lower (and heterogeneous) P-cadherin, as was the case in the cytoplasm ( Figure 2D). Altogether, these results support that depletion of E-cadherin, due to gene deletions, can trigger E-to P-cadherin switch.

E-Cadherin Loss of Function Triggers E-Cadherin to P-Cadherin Switch in Gastric Cancer
Given the abundance of gastric cancers bearing simultaneous loss of CDH1 expression and CDH3 upregulation, we tested whether depleting CDH1 would trigger CDH3 upregulation.
We used sgRNAs flanking CDH1 exon 2 (Supplementary Table S1), and successful removal of exon 2 was confirmed by Sanger sequencing in the MKN74 cell line (Supplementary Table S1). This clone carried a homozygous deletion encompassing the CDH1 exon 2 ( Figure 2A) and presented complete loss of both CDH1 mRNA and E-cadherin protein expression ( Figure 2B). As a consequence of this deletion, both CDH3 mRNA and P-cadherin protein expression increased by 60% (p-value = 0.05 and p-value = 0.037, respectively) ( Figure 2C). The immunofluorescence data highlight the P-cadherin expression increase at the plasma membrane upon E-cadherin expression loss, likely rescuing adherens junctions ( Figure 2D). Indeed, P-cadherin co-localized with adhesion partner β-catenin at the cell membrane in the CDH1 deletion clone, while the parental cell line displayed lower (and heterogeneous) P-cadherin, as was the case in the cytoplasm ( Figure 2D). Altogether, these results support that depletion of E-cadherin, due to gene deletions, can trigger E-to P-cadherin switch.

E-cadherin Expression Loss Triggers Downregulation of Adhesion Complex Partners and Adhesion-Related Pathways
To further dissect the impact of E-cadherin loss, we used proteomics to evaluate the differential proteome between the parental cell line and the CDH1 deletion clone. A total of 4579 proteins were identified, from which 41 were downregulated and 39 were upregulated in the CDH1 deletion clone (Supplementary Table S3). The E-to P-cadherin switch was confirmed in the CDH1 deletion clone, with downregulation of the former (0.37-fold, p-value = 0.0002) and upregulation of the latter (1.99-fold, p-value = 0.018)

E-Cadherin Expression Loss Triggers Downregulation of Adhesion Complex Partners and Adhesion-Related Pathways
To further dissect the impact of E-cadherin loss, we used proteomics to evaluate the differential proteome between the parental cell line and the CDH1 deletion clone. A total of 4579 proteins were identified, from which 41 were downregulated and 39 were upregulated in the CDH1 deletion clone (Supplementary Table S3). The E-to P-cadherin switch was confirmed in the CDH1 deletion clone, with downregulation of the former (0.37-fold, p-value = 0.0002) and upregulation of the latter (1.99-fold, p-value = 0.018) ( Figure 2D). In addition to E-and P-cadherin expression, other adhesion complex partners became impaired in the CDH1 deletion clone, namely, CTNNA1 (αE-catenin) and CTNNB1 Biology 2023, 12, 803 8 of 13 (β-catenin), while others remain unchanged, namely, CTNND1 (p120) ( Figure 2D). Gene ontology analysis revealed several adhesion-associated pathways downregulated upon CDH1 deletion ( Figure 2E, Supplementary Table S4), encompassing 'adherens junction', 'cadherin binding', 'focal adhesion assembly', and catenin-and actin-associated pathways ( Figure 2E, Supplementary Table S4). In contrast, the 'establishment and maintenance of cell polarity' pathway became upregulated, likely revealing the rescue by P-cadherin. Events associated with migration, such as 'cell projection membrane' and 'wound healing', were found upregulated in cells with E-to P-cadherin switch ( Figure 2E, Supplementary  Table S4). Downregulation of immune-associated pathways 'macrophage chemotaxis' and 'regulation of leucocyte migration' and upregulation of oncogenic pathway 'Ras protein signal transduction' were also found upon CDH1 deletion ( Figure 2E, Supplementary  Table S4).

D Chromatin Architecture Re-Wiring at the CDH3-CDH1 Loci Contributes to E-Cadherin to P-Cadherin Switch in Gastric Cancer
In an attempt to identify the mechanism triggering the E-to P-cadherin switch, upon CDH1 depletion, we analysed regulatory elements within the CDH3-CDH1 loci that may control expression of cadherins in the stomach. We mined the Ensembl database and found an eQTL located within the CDH3 intron 2 (chr16: 68661867-68662578 and chr16: 68663306-68670017, hg38) ( Figure 3A). Normal epithelial cells from bariatric surgeries, MKN74 parental cells as well as CDH1 deletion cells, were profiled for CDH1-promoter interactions with the eQTL using 4C-seq, and for accessibility of chromatin using ATAC-seq. Normal stomach accessible chromatin was also collected from the ENCODE project and analysed [31]. In normal stomach, which expresses CDH1 but not CDH3, the chromatin was not accessible at the CDH3 promoter nor at the CDH3-eQTL, but was accessible at the CDH1 promoter ( Figure 3A, ATAC-seq normal stomach). Neither the CDH3 promoter nor the CDH3-eQTL interacted with the CDH1 promoter ( Figure 3A, 4C-seq normal stomach). In contrast, upon CDH1 deletion, availability of the chromatin increased around the eQTL sequence and was maintained as available at the CDH1 and CDH3 promoters ( Figure 3A, ATAC-seq parental vs. ATAC-seq CDH1 deletion). Additionally, new interactions were established between the CDH1 promoter and the CDH3-eQTL region ( Figure 3A, 4C-seq parental vs. 4C-seq CDH1 deletion). These chromatin architecture changes seem to indicate that, upon CDH1/E-cadherin depletion, the CDH1 promoter interacts with the CDH3 eQTL, potentially leading to CDH3/P-cadherin expression increase ( Figure 2B-D). If this is the case, deleting the CDH3-eQTL in a cell line expressing both CDH3 and CDH1 could provide information on the expression regulatory role of the CDH3-eQTL. To address this, we CRISPR-Cas9 deleted the eQTL in parental MKN74 cells. This led to a 50% decrease in both CDH3 and CDH1 mRNA expression (p-value = 0.0196 and p-value = 0.0084, respectively), and a 30% decrease in P-cadherin and E-cadherin protein expression (p-value = 0.035 and p-value = 0.0486, respectively) as compared to parental MKN74 cells ( Figure 3B). These results pinpoint CDH3-eQTL as a potential enhancer of both CDH3 and CDH1 in this cell line. Altogether, these data provide evidence that loss of CDH1/E-cadherin expression alters the chromatin conformation at the CDH3 locus, allowing interaction of the CDH1 promoter with the CDH3-eQTL, which seems to potentiate CDH3/P-cadherin expression.

Discussion
Herein, we aimed to characterize a potential 3D chromatin architecture re-wiring process, occurring at the CDH3-CDH1 loci, that may explain the CDH1/E-cadherin to

Discussion
Herein, we aimed to characterize a potential 3D chromatin architecture re-wiring process, occurring at the CDH3-CDH1 loci, that may explain the CDH1/E-cadherin to CDH3/P-cadherin switch observed in gastric cancer. First, we demonstrated that 75% of the gastric cancers analysed overexpressed CDH3. Additionally, we observed that, in 42% of all cases, CDH3 overexpression occurred concomitantly with CDH1 downregulation, representing an E-to P-cadherin switch. We also found that an eQTL within the CDH3 intron 2, acting as a regulatory element, was able to modulate both CDH1 and CDH3 expression, and that it interacted with the CDH1 promoter upon CDH1 loss of expression. To the best of our knowledge, this body of results highlights a novel mechanism which triggers E-to P-cadherin switch in gastric cancer.
Our data suggest that E-to P-cadherin switch is a common event in gastric cancers. The impact of P-cadherin overexpression in cancer is still controversial, as this molecule seems to act as tumour-promoting and a marker of a bad prognosis in several cancer types-namely, in breast, glioblastoma, and gastric cancer [36]-but may also act as a marker of a good prognosis-namely, in colon cancer [37]. Breast cancer cells presenting E-to P-cadherin switch, and those expressing both cadherins, reveal increased migration capacity in vitro [38], while breast cancer epithelial cells expressing E-cadherin and lacking P-cadherin expression, resembling the normal breast, and cells lacking both cadherins display reduced migration [38].
Herein, we created a model of E-to P-cadherin switch by using CRISPR-Cas9 editing to deplete CDH1. This model presented reduced expression of adhesion complex partners, but maintained β-catenin expression at adherens junctions, which co-localized with P-cadherin at the cell membrane. This likely represents an adherens junctions rescue by P-cadherin in E-cadherin-depleted cells, as previously reported [39]. Evidence for E-to P-cadherin switch has been linked to invasive lobular breast cancer with tubular elements, which show reduced E-cadherin expression, normal β-catenin expression, and P-cadherin upregulation, highlighting rescue of adherens junctions in breast tubular elements when E-cadherin is lost [39].
Our data on differential protein expression and gene ontology analysis in the CDH1 deletion clone, in comparison to an isogenic parental CDH1 cell line, highlighted downregulation of adhesion-, catenin-and actin-related pathways, as well as upregulation of cell polarity, all associated with E-cadherin main functions. Despite the fact that E-cadherin deficiency weakens cell-adhesion, cell polarity regulation further supports P-cadherin rescue of adherens junctions (Christgen et al. 2020). Our gene ontology data also support increased migration potential of cells with E-to P-cadherin switch, namely, through activation of the 'wound healing' pathway, also described for breast cancer cells [38]. Our findings also suggest that cells with E-to P-cadherin switch may have increased proliferation through the activation of Ras oncogenic signalling pathways. As in many published reports, the abovementioned findings support that E-to P-cadherin switch is associated with more aggressive tumour phenotypes.
In the present study, we used the abovementioned cell model to identify the mechanism responsible for the E-to P-cadherin switch. We raised the hypothesis that regulatory elements in the CDH3/CDH1 loci could play a role in the observed E-to P-cadherin expression switch, as gene expression is tightly regulated through the physical proximity enabled by the 3D-chromatin architecture (Andrey and Mundlos 2017) [20]. We demonstrated that, upon CDH1 depletion triggered by a gene deletion, the chromatin re-shuffles at the CDH3/CDH1 loci, and its promoter becomes physically bound to an eQTL overlapping accessible chromatin located in CDH3 intron 2; this associates with increased CDH3/Pcadherin expression. The absence of interactions between the CDH1 promoter and this eQTL in normal stomach epithelial tissue may explain why CDH3 is not expressed in this particular tissue. While these data provide a rationale for E-to P-cadherin switch in gastric cancer, they do not inform whether the eQTL promotes CDH3 expression. To address this question, we deleted the eQTL in a tumour cell line expressing wild-type Eand P-cadherins, preventing it from interacting with the CDH1 promoter. This resulted in downregulation of both P-and E-cadherin expression and demonstrated that the CDH3-eQTL controlled P-cadherin expression. Taken together, these data provide evidence that loss of CDH1/E-cadherin expression alters the chromatin conformation at the CDH3 locus, allowing interaction of the CDH1 promoter with the CDH3-eQTL, which seems to potentiate CDH3/P-cadherin expression. Our findings also suggest that the CDH3-eQTL acts as an enhancer regulating both P-and E-cadherin protein expression, thus regulating P-cadherin expression in the switch mechanism. Similar findings have been observed for the ZRS enhancer located within the Lmbr1 gene, also controlling Shh gene expression [40].
Single-cell sorting of CRISPR-Cas9 clones herein described only yielded one homozygous clone for each set of sgRNAs harbouring the intended deletions, which we acknowledge as a study limitation.

Conclusions
Overall, this study provides insights into a noncoding regulatory element that controls E-and P-cadherin expression and triggers E-to P-cadherin switch in the presence of a CDH1 depletion, through a re-shuffle of the CDH1/CDH3 loci.