Functional effects of EpCAM N-glycosylation in MDA-MB-468 breast cancer cells.

Introduction
Breast cancer is the most common malignancy in American women and the second leading cause of cancer-related death in American women1. While advancements in breast cancer treatments have resulted in a 5-year survival rate for localized breast cancer of > 99%, most breast cancer deaths are due to metastatic disease, which has a 5-year survival rate of only 32%2. Effective treatment for metastatic breast cancer is the final frontier in breast cancer research, but developing these treatments requires a deeper understanding of the molecular mechanisms of breast cancer metastasis. Cancer cell behaviors which confer breast cancer metastasis include cancer cell migration, invasion, and adhesion3–5.
One interesting therapeutic target in breast cancer metastasis is the epithelial cell adhesion molecule (EpCAM). EpCAM is a transmembrane protein that facilitates homotypic cell–cell adhesion in epithelial cells, and has been linked to several epithelial cancers, including breast, colorectal, and prostate cancers6,7. Higher EpCAM expression in cancer tissues has been linked to worse prognosis, but its role in cancer and metastasis is still incompletely understood6,7. EpCAM is not structurally related to any other CAMs nor any cell junctions but is rather thought to modulate adhesion through transcriptional regulation6.
As shown in Fig. 1, EpCAM interferes with E-cadherin mediated adhesion by disrupting the association of E-cadherin with the actin cytoskeleton6,8. Additionally, EpCAM undergoes regulated intramembrane proteolysis. First, the extracellular domain of EpCAM is cleaved by ADAM17 near the plasma membrane, then the intracellular domain is cleaved by presenilin 2 (PSEN2)7,9. The intracellular domain of EpCAM (EpIC or EpICD), when cleaved, is thought to form a complex with β-catenin and FHL2, which is translocated to the nucleus and interacts with transcription factor LEF16,10,11. This leads to upregulation of c-myc (MYC), cyclin A (CCNA2), and cyclin E (CCNE1), which is supported by findings that EpCAM expression leads to upregulation of c-myc and cyclins A, E, and D (CCND1) in vitro6,11–14. When cleaved, the extracellular portion of EpCAM (EpEx or EpEC) activates the MAPK signaling pathway through binding to the epidermal growth factor receptor (EGFR) in colon cancer cells7,11,15.
In breast cancer metastasis specifically, Cimino et al. reported that while EpCAM protein expression was increased in metastatic breast cancer cells compared to primary breast tumor cells, EpCAM mRNA expression remained unchanged16. This suggested that post-translational modification of EpCAM may be playing a role in the regulation of EpCAM in breast cancer metastasis. A highly conserved modification of EpCAM protein is its N-glycosylation. EpCAM is N-glycosylated at three sites, which are Asn74, Asn111, and Asn198, although neither the N-glycosylation profile nor the function of EpCAM N-glycosylation is known6,11. Another study of metastatic breast cancer patient tissues showed that EpCAM expression not only increased in metastasis, but was also correlated with the elevated abundance of 7 distinct N-glycans, further suggesting that EpCAM N-glycosylation may play a role in breast cancer metastasis17.
The role of N-glycosylation in a specific protein cannot be as easily probed as other modifications such as phosphorylation where many proteins are modified by a small subset of kinases because all N-glycosylated proteins are processed by the same set of enzymes18. It is thus not feasible to inhibit N-glycosylation of a single protein without potentially inhibiting N-glycosylation of all proteins and causing significant off-target effects. Two previous studies mutated some or all of the three EpCAM N-glycosylation sites to prevent N-glycosylation and observed decreases in EpCAM stability and overall cell–matrix adhesion as a result19,20.
In this report, we build on these prior results by conducting a comprehensive study of the effects of N-glycosylation on EpCAM and critical cancer cell properties in breast cancer cells. With EpCAM N-glycosylation sites mutated to glutamine in mutant EpCAM (mutEpCAM) in MDA-MB-468 human breast cancer cells (Fig. 1), we examined the effects on EpCAM stability and subcellular localization, the expression of downstream targets of EpCAM, as well as effects on cell viability, migration, invasion, and adhesion. We hypothesized that the lack of N-glycosylation in mutant unglycosylated EpCAM would decrease EpCAM protein stability, decrease the metastasis-driving cell properties of migration and invasion, and alter EpCAM localization, homotypic cell–cell adhesion and expression of downstream targets of EpCAM in triple-negative MDA-MB-468 breast cancer cells. We provide novel evidence that our hypotheses regarding the role of EpCAM N-glycosylation in protein stability and localization proved true, while the functional outcomes of migration, invasion, homotypic cell–cell adhesion, and downstream targets were not affected by EpCAM N-glycosylation in MDA-MB-468 breast cancer cells.

Results

EpCAM sequence is largely conserved in human breast cancer cell lines
Before beginning to mutate EpCAM N-glycosylation sites in human breast cancer cells, we first sought to confirm that EpCAM is not already mutated in breast cancer cells at any of the sites we planned to target. To this end, genomic DNA from EpCAM was isolated from two non-malignant human breast cell lines (MCF10A and MCF12A) and four breast cancer cell lines (MCF7, BT474, MDA-MB-468, and SUM149). EpCAM is a transmembrane protein with six domains (including a signal peptide), three N-glycosylation sites, i.e., Asn74, Asn111, and Asn198 (Fig. 1), and nine coding DNA sequences (CDS)6,7,11,21. In each cell line, the nine EpCAM CDSs were amplified, sequenced by Sanger sequencing, and aligned to the EpCAM reference sequence (Supplemental Fig. S1 in Supporting Information, NM_002354). At the genomic level, the EpCAM gene in MCF12A, BT474, MDA-MB-468, and SUM149 cells contains a missense mutation of T344C, which leads to a polymorphism of M115T at the protein level. This polymorphism has been previously reported and is associated with higher risk for, and worse prognosis of several epithelial cancers including breast, lung, cervical, and other cancers, though the downstream mechanisms of this mutation remain unknown22–24. A silent mutation was also found in MCF7 cells of G501A. As the M115T polymorphism had been previously reported and no gene regions we sought to target for mutation were affected, we proceeded with the remainder of this study.

Mutation of Asn74, Asn111, and Asn198 to Glns successfully abrogates N-glycosylation
To study the effects of EpCAM N-glycosylation, we sought to generate cells which produced only non-glycosylated mutant EpCAM (mutEpCAM) by mutating the glycosylated asparagine residues (Asn74, Asn111, and Asn198) to glutamine (Fig. 1). As the mutation from N to Q requires two base changes between purines and pyrimidines at each site (A to C, C to A, or T to A), single nucleotide editing of the native gene is not possible with currently available technologies25. Therefore, we first needed to knock out native, glycosylated EpCAM in the selected breast cancer cell line, MDA-MB-468. While MDA-MB-231 cells are commonly used to study EpCAM in breast cancer, they did not express EpCAM in significant quantities in our hands when screened for this study26–29. Thus, the MDA-MB-468 cell line was chosen instead, as it is a triple-negative breast cancer cell line with metastatic potential and consistently expresses high levels of EpCAM30,31. To knock out native EpCAM, we infected MDA-MB-468 cells with a pLentiCRISPRv2 lentivirus with three different EpCAM guide RNAs (Genscript) to target the EpCAM gene for cutting by Cas932,33. After puromycin selection for infected cells and single cell clonal selection, a stable EpCAM knockout (KO) cell line was established (Fig. 2A–C).
To re-express mutEpCAM protein in EpCAM KO cells, a custom gene was designed with missense mutations at each of the three N-glycosylation sites (N74Q, N111Q, and N198Q) and silent mutations in the protospacer adjacent motif (PAM) sites that were recognized by the Cas9 (Supplemental Fig. S2 in the Supporting Information). Additionally, we designed a glycosylated EpCAM re-expression vector to act as a control for differing EpCAM expression at the non-native promoter. This EpCAM overexpression (OV) gene also contained the silent mutations to the PAM sequences used in the mutEpCAM gene but did not have mutations to the N-glycosylation sites (Supplemental Fig. S2 in the Supporting Information). Both gene sequences were otherwise identical to the reference sequence (Supplemental Fig. S2 in the Supporting Information). These genes were inserted into a pLenti-CMV-blast empty vector (Supplemental Fig. S3 in the Supporting Information)34. EpCAM KO cells were infected with either the pLenti-CMV-blast empty vector (EpCAM KO), pLenti-CMV-blast-EpCAMOV (EpCAM OV cells), or pLenti-CMV-blast-mutEpCAM (mutEpCAM cells), with parent MDA-MB-468 cells being used as a control for native EpCAM expression (wild-type or WT cells).
qRT-PCR confirmed expected EpCAM mRNA overexpression (6–10x) in EpCAM OV cells and EpCAM knockout in EpCAM KO cells compared to WT, which was stable for up to 10 passages (Fig. 2A). mutEpCAM mRNA expression in mutEpCAM cells was not significantly different than native EpCAM mRNA expression in WT cells (Fig. 2A). To determine N-glycosylation of EpCAM protein, protein lysates were digested with PNGaseF, an enzyme which cleaves all mammalian N-glycans. Western blots of digested and undigested protein lysates confirmed that while EpCAM is N-glycosylated in WT and EpCAM OV cells, it is not in mutEpCAM cells (Fig. 2B, C). Western blots also confirmed complete EpCAM knockout in EpCAM KO cells (Fig. 2B, C).
Pavšič et al. published the crystal structure of the extracellular domain of mutated human EpCAM (mutEpCAM), which contains the same three mutations of N74Q, N111Q and N198Q as in our study, and which is publicly available under 4MZV at the RCSB (Research Collaboratory for Structural Bioinformatics Protein Data Bank) protein data bank (PDB) at https://www.rcsb.org/structure/4MZV21. The N to Q mutations in mutEpCAM were introduced by Pavšič et al. to remove N-glycosylation of human EpCAM to avoid potentially high variability of N-glycans in human EpCAM, which would have resulted in a large variety of different protein structures for crystallization. Based on the measured crystal structure of mutEpCAM, Pavšič et al. computationally modeled the wild-type EpCAM protein’s extracellular domain with high-mannose N-glycans at the three N-glycosylation sites Asn74, Asn111, and Asn19821. Using the existing crystal structure of the human mutEpCAM’s extracellular domain, we reverse-modeled the three mutations to glutamine at Q74, Q111, and Q198 to place back in three asparagines at N74, N111, and N198, thereby producing a model of wild-type EpCAM, to calcuate the difference between wild-type EpCAM and mutEpCAM as shown in Supplemental Fig. S4 in the Supporting Information. This reverse modeling was carried out in MutationExplorer (https://proteinformatics.uni-leipzig.de/mutation_explorer/), a webserver for mutation of proteins and 3D visualization of energetic impacts recently introduced by Phillip et al.35. Our modeling showed that reverse mutating Q74, Q111, and Q198 back to N74, N111, and N198 had a small stabilizing effect of − 7.44 in Rosetta Energy Units (REU) on the energy of EpCAM (with the total energy of wild-type EpCAM recorded at − 860.51 REU), and no effect on the hydrophobicity of EpCAM (see Supplemental Fig. S4). The absolute and differential energies and hydrophobicities of reverse-modeled wild-type EpCAM versus the published crystal structure of mutEpCAM are shown in Supplemental Fig. S4 in the Supporting Information.

EpCAM N-glycosylation increases protein stability
Due to the role of N-glycosylation in protein folding and stabilization36, we hypothesized that ablation of EpCAM N-glycosylation would decrease EpCAM protein stability. This hypothesis was supported by a previous study that reported that mutating asparagine to alanine at N198A significantly reduced EpCAM protein half-life20. We wanted to verify this finding in our cells, as we hypothesized that mutation of a polar residue such as Asn on the protein surface11,21,37,38 to a hydrophobic residue such as Ala may have contributed to decreasing protein stability20. To avoid the possibility that hydrophobic Ala residues contributed to EpCAM instability independent of ablation of glycosylation, in this study we replaced Asn with a second polar amino acid, Gln. By using this strategy, we reasoned we could more precisely evaluate the impact of N-glycosylation on EpCAM protein stability, which we tested experimentally by treating WT and mutEpCAM cells with the protein synthesis inhibitor cycloheximide (CHX) and harvesting protein at four-hour intervals. Western blots of these samples revealed that although initial amounts of mature EpCAM protein were comparable between native and mutEpCAM, mutEpCAM degraded much more quickly (Fig. 3A). These results demonstrate a significant decrease in protein stability of mutEpCAM compared to native EpCAM at all timepoints (Fig. 3A), which was confirmed by densitometry analysis (Fig. 3B). Based on these findings, we concluded that N-glycosylation significantly impacts EpCAM protein stability in MDA-MB-468 cells.

EpCAM N-glycosylation affects subcellular protein localization
After protein synthesis in the endoplasmic reticulum (ER) and glycan modification in the Golgi apparatus are completed, EpCAM must be translocated to the cell membrane6,11. Disruption of N-glycosylation is known to impair protein folding and disrupt subcellular localization of different proteins39–43, or cause protein aggregation39,44–46. We therefore hypothesized that ablation of EpCAM N-glycosylation would affect the subcellular localization of EpCAM. To test this hypothesis, immunofluorescence staining and confocal laser scanning microscopy at 63 × magnification was performed to visualize EpCAM protein in subcellular compartments. WT, EpCAM KO (shown in supplemental information, Supplemental Fig. S5 in the Supporting Information), EpCAM OV, and mutEpCAM cells were grown in chamber slides and co-stained for EpCAM (red), plasma membrane marker wheat germ agglutinin (WGA, cyan), endoplasmic reticulum (ER) membrane marker calnexin (green), and nuclear marker 4′,6-diamidino-2-phenylindole (DAPI, blue) (Fig. 4A, B, and Supplemental Fig. S5 in the Supporting Information).
EpCAM is a transmembrane protein and typically a cell surface marker, so we hypothesized that native EpCAM in WT cells and EpCAM OV would localize to the plasma membrane to a larger extent than mutEpCAM. While EpCAM was primarily concentrated in the plasma membrane in WT cells, it was distributed in a vesicular pattern throughout the cell in both EpCAM OV and mutEpCAM cells (Fig. 4A). Colocalization analysis of EpCAM and WGA confirmed that membrane colocalization of both EpCAM OV and mutEpCAM was significantly decreased compared to native EpCAM (Fig. 4C).
Similarly, we expected that mutEpCAM may accumulate in the ER due to improper protein folding and translocation. While very little ER localization of native EpCAM was observed in WT cells, significantly increased colocalization between EpCAM and the ER marker calnexin was observed in EpCAM OV and mutEpCAM cells (Fig. 4B). Colocalization analysis confirmed a significant increase in ER colocalization of mutEpCAM compared to native EpCAM, while increased ER colocalization of EpCAM OV was not statistically significant (Fig. 4D). Taken together, these findings suggest that both overexpression of EpCAM and loss of EpCAM N-glycosylation affect subcellular localization, with loss of N-glycosylation in particular leading to a significant shift in EpCAM subcellular localization from the cell membrane to the ER in MDA-MB-468 cells.

N-glycosylation does not affect EpCAM signaling
EpCAM has been shown to regulate various signaling pathways with downstream targets including E-cadherin, c-myc, β-catenin, cyclins A and E, and the epidermal growth factor receptor (EGFR)6,11,15,47. To determine if loss of N-glycosylation affected the expression of these downstream targets, we performed qRT-PCR analysis of total mRNA and Western blot analysis of total protein lysates from WT, EpCAM KO, EpCAM OV, and mutEpCAM cells. No effect of mutEpCAM was observed on any of the probed proteins in Western blot (Fig. 5A, B). While decreases in cyclin A and E mRNA expression (CCNA and CCNE) were observed by qRT-PCR, this was likely caused by lentiviral transduction rather than mutEpCAM expression, as it was present in all three cell lines transduced with lentiviruses, i.e., EpCAM KO, EpCAM OV, and mutEpCAM, as compared to untransduced WT cells (Fig. 5C). EGFR mRNA did show a significant increase in expression in mutEpCAM cells (Fig. 5C), but this was not reflected in total EGFR protein expression (Fig. 5A, B). We therefore concluded that N-glycosylation of EpCAM has minimal effects on EpCAM regulation of these downstream targets in MDA-MB-468 cells.

EpCAM N-glycosylation does not affect cell migration, invasion, or adhesion
Finally, we sought to determine if EpCAM N-glycosylation had any effect on cancer cell properties related to cancer progression and metastasis. Our results clearly show that no significant differences were detected between WT cells and EpCAM KO, EpCAM OV, or mutEpCAM cells in assays of cell viability, migration, or invasion in MDA-MB-468 cells (Fig. 6A, B). EpCAM overexpression was found to increase homotypic cell–cell adhesion, with a significant increase at 3 h; however, no effects of mutEpCAM expression on homotypic cell–cell adhesion were observed in MDA-MB-468 cells (Fig. 6C).

Discussion
The role of EpCAM in cancer progression has been difficult to define because of a large body of contradictory findings of EpCAM knockout, knockdown, and overexpression in different in vitro studies26–29. An early study by Osta et al. reported that EpCAM knockdown reduced cell migration and invasion by > 90% in MDA-MB-231 breast cancer cells26. Similarly, Gao et al. showed that EpCAM knockdown decreased cell proliferation, migration, and invasion in MCF7 breast cancer cells, while EpCAM overexpression increased cell migration and invasion > sixfold in MDA-MB-231 breast cancer cells27. By contrast, Gostner et al. found no effect of EpCAM overexpression on migration, invasion, or proliferation in MDA-MB-231 cells28. Martowicz et al. even noted paradoxical results between different cell lines in their own study, with EpCAM knockdown decreasing in vivo invasion of MCF7 and T47D cells in xenograft models on chorioallantoic membranes of fertilized chicken embryos, but EpCAM overexpression also decreasing in vivo invasion of MDA-MB-231 cells in the same xenograft models29. This study also observed no effect on in vitro proliferation or migration and a decrease in in vitro invasion in MDA-MB-231 cells, in contrast to prior findings29. It is also important to note in this discussion that the reported baseline expression of EpCAM varies significantly in MDA-MB-231 cells, which are one of the most commonly used breast cancer cell lines in in vitro studies, with some authors finding high baseline EpCAM expression and others finding little to no baseline EpCAM expression26–29. These studies are only a select few examples of the vast literature on the role of EpCAM in cancer cell behaviors, but they illustrate why EpCAM’s role remains uncertain despite significant study.
To investigate these cell line-based differences in the effects of EpCAM and to prepare for targeting the EpCAM gene with CRISPR Cas9 guide RNAs, we performed Sanger sequencing of EpCAM coding DNA sequences in six immortalized breast cell lines. The only mutation observed resulting in any changes in the protein sequence was the M115T single nucleotide polymorphism (rs1126497) in MCF12A, BT474, MDA-MB-468, and SUM149 cells22. While the AlphaFold AlphaMissense Pathogenicity score of this mutation was 0.040, indicating a likely benign polymorphism (UniProt ID: P16422), previous studies have found this polymorphism to be associated with worse prognosis in several cancers, suggesting that the M115T single nucleotide polymorphism may have an impact on secondary protein structure22–24,37,38,48. The finding of this same polymorphism in some, but not all breast cancer cells lines suggests that this polymorphism or other mutations in the EpCAM gene may play a role in the cell line-specific findings related to EpCAM. Though outside the scope of this study, protein crystallization would likely be required to conclusively determine the impact of this polymorphism and other mutations on protein secondary structure and investigate the mechanism of its correlation with poor prognosis in patients.
Loss of EpCAM N-glycosylation was previously reported to decrease protein stability by Munz et al.20. We sought to confirm this finding in our model, as there were significant differences to the model used in the prior study20. While our study mutated all three N-glycosylated asparagine residues to glutamine, the prior study mutated these residues to alanine. Alanine is a smaller amino acid and therefore less likely to cause steric hinderance in the protein structure, whereas glutamine is more similar to asparagine, as they are both polar amino acids and differ by only one methylene group. According to the protein structures shown by Pavšič et al. and AlphaFold, Asn74 and Asn198 are external residues, while Asn111 is an internal residue21,37,38. Changes in either size or polarity could affect protein structure. However, the consistent finding of our study and the study by Munz et al.20 that N-glycosylation stabilizes EpCAM, suggests that this is due to EpCAM N-glycosylation irrespective of the specific mutations introduced.
In addition to the effects of N-glycosylation on EpCAM stability, we also found that N-glycosylation affects EpCAM subcellular localization in human triple-negative MDA-MB-468 breast cancer cells. While EpCAM membrane colocalization decreased for both overexpressed EpCAM and mutEpCAM, it is important to note that mutEpCAM was not overexpressed to the same degree as EpCAM in OV cells. In EpCAM OV cells, EpCAM was expressed at > fivefold the level of WT, whereas EpCAM expression levels did not significantly differ between WT and mutEpCAM cells. It is likely that the high degree of EpCAM overexpression in EpCAM OV cells exceeds the capacity of the cells to localize all EpCAM to the cell membrane. It is also possible that overexpression of EpCAM in the EpCAM OV cells increases misfolding and mistrafficking of EpCAM as previously described for other proteins49,50. However, this cannot explain the decreased membrane colocalization in mutEpCAM. N-glycosylation assists in protein folding and as previously discussed, stability, so the decreased membrane colocalization and increased ER colocalization in mutEpCAM may be due faster turnover of the protein, which is being degraded before it can localize to the membrane.
The large extent of EpCAM overexpression in EpCAM OV cells may also explain the increased homotypic cell–cell adhesion observed in these cells. While one would expect to see a corresponding decrease in homotypic cell–cell adhesion in EpCAM KO cells, the degree of EpCAM overexpression is great and the effect on adhesion is small in EpCAM OV cells. Very large changes in EpCAM expression may be necessary to produce a small effect on homotypic cell–cell adhesion, such that the difference between WT and KO EpCAM expression is too small to produce a corresponding decrease in adhesion in EpCAM KO cells. It is also possible that there is a threshold of EpCAM expression required to affect adhesion, which is reached only in EpCAM OV cells, but not WT, EpCAM KO, or mutEpCAM cells. Although EpCAM can affect adhesion both through direct adhesion to other EpCAM molecules and through regulation of other pathways, the increased adhesion in EpCAM OV cells is likely a result of direct EpCAM adhesion. This is because the mechanism by which EpCAM indirectly affects adhesion is by interfering with E-cadherin mediated adhesion. If the effects of EpCAM on adhesion in these cells were due to this pathway, we would expect cell–cell adhesion to decrease with increasing EpCAM, not increase6,8.
Although N-glycosylation clearly affected EpCAM protein stability and subcellular localization in our study in the human triple-negative MDA-MB-468 breast cancer cell line, we did not observe any effects on cell viability, migration, or invasion in vitro. These results were consistent with our results that EpCAM knockout and EpCAM overexpression also did not affect these cellular processes in vitro. As previously discussed, there are considerable differences reported between the effects of EpCAM knockdown and overexpression between different cell lines, with some studies finding significant effects of EpCAM knockdown and overexpression on these cellular processes, and other studies finding no or opposite or even paradoxical effects26–29. While the findings of our study in MDA-MB-468 cells are clear and consistent, future studies in other cell lines and disease models may be appropriate to determine if the effects of EpCAM N-glycosylation reported in this study hold true in other cell types.
Overall, altering EpCAM expression and N-glycosylation did not significantly affect EpCAM signaling and downstream targets in MDA-MB-468 breast cancer cells as far as total mRNA and total protein levels of these downstream targets is concerned. While activity of some downstream targets of EpCAM, such as β-catenin and E-cadherin, could be affected by EpCAM without a change in total expression, this would be detected through the downstream effects of those proteins6,8,10,13. If E-cadherin activity was impacted by EpCAM in these cells, we would see an increase in cell–cell adhesion in EpCAM KO cells and/or a decrease in homotypic adhesion in EpCAM OV cells6,8. Direct EpCAM adhesion is much weaker than E-cadherin-mediated adhesion, so that the effects of E-cadherin would be expected to outweigh direct adhesion by EpCAM7,11. In the case of β-catenin, increased β-catenin complex formation with EpCAM would lead to increased LEF1 activation, and subsequently, increased transcription of targets such as c-myc and cyclin E6. Since no increased transcription of these targets was observed, it is unlikely that changes in EpCAM expression significantly affect β-catenin activity in MDA-MB-468 cells. We did observe a substantial rise in total EGFR mRNA levels in mutEpCAM cells, without a corresponding increase in total EGFR protein levels. This discrepancy could be caused by post-transcriptional processes such as changes in translation efficiency or protein stability.
A previous study by Liu et al. reported that mutations of Asn to Gln at all three EpCAM N-glycosylation sites decreased cell-fibronectin-matrix adhesion, decreased the expression of fibronectin and integrin β1, activated the FAK/PI3K/Akt signaling pathway, and decreased the expression of β-catenin and matrix metalloproteinases-2 and -919. In this study, the researchers performed site-directed mutagenesis to generate an Asn-to-Gln-mutated EpCAM plasmid, which was then transfected into wild-type MCF-7 and MDA-MB-231 breast cancer cells, resulting in the co-existence of glycosylated wild-type EpCAM and unglycosylated mutated EpCAM at the same time in these two breast cancer cell lines19. The same research group recently presented new mechanistic insights showing that adding the expression of deglycosylated EpCAM promoted apoptosis, inhibited proliferation, activated autophagy, and suppressed the Akt/mTOR signaling pathway51. In this 2022 study, they again used the same breast cancer cell lines as in their previous study, which are co-expressing glycosylated wild-type and deglycosyated mutant EpCAM at the same time19,51. In contrast, in our study we expressed Asn-to-Gln mutated EpCAM, i.e., mutEpCAM, in MDA-MB-468 breast cancer cells, in which we had knocked out wild-type EpCAM first, to solely express unglycosylated mutEpCAM without any glycosylated wild-type EpCAM present, as evident by our clean and reproducible Western blots. This approach allowed us to study clear cause-and-effect relationships based on mutEpCAM expression only. Moreover, our study used a different breast cancer cell line and tested for a different type of adhesion, namely homotypic cell–cell adhesion, while Liu et al. examined cell–matrix adhesion19.
In conclusion, our study comprehensively examined the role of EpCAM N-glycosylation on EpCAM protein stability, localization, and critical breast cancer cell properties in MDA-MB-468 cells. N-glycosylation was found to stabilize EpCAM and increase cell membrane localization of EpCAM in this cell line. However, N-glycosylation of EpCAM had little downstream consequences. While a significant increase in homotypic cell–cell adhesion was found in the context of EpCAM overexpression, no effect of EpCAM N-glycosylation was observed on cell migration, invasion, or proliferation, homotypic cell–cell adhesion, or the expression of downstream targets of EpCAM in MDA-MB-468 cells. Due to the variability of EpCAM effects between different cell lines and models noted in previous studies, future studies of EpCAM N-glycosylation in other cell lines or cancers are needed to confirm the generalizability of these findings. Nonetheless, this study represents a significant contribution to the understanding of EpCAM biology and function.

Methods

Cell culture
Two non-malignant immortalized human breast cell lines, MCF10A (female, American Type Culture Collection/ATCC #CRL-10317, Manassas, VA, USA) and MCF12A (female, ATCC #CRL-3598), four immortalized human breast cancer cell lines, MDA-MB-468 (female, ATCC #HTB-132), MCF7 (female, ATCC #HTB-22), BT474 (female, ATCC #HTB-20), and SUM149 (female, BioIVT #HUMANSUM-0003004, Hicksville, NY, USA), and HEK293T packaging cells (female, ATCC #CRL-3216) were used in this study. These cells were purchased from the ATCC and BioIVT and maintained in a humidified incubator with 5% CO2 at 37 °C. MDA-MB-468, SUM149, and HEK293T cells were cultured in high-glucose Dulbecco’s Modified Eagle Medium (DMEM) (Thermo Fisher Scientific #11965118, Waltham, MA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Corning, #35-011-CV, Corning, NY, USA). MCF10A and MCF12A cells were cultured in DMEM/F12 50:50 Mix Media (Corning #15-090-CV), supplemented with cholera toxin (Sigma Aldrich #227036, Burlington, MA, USA), human epidermal growth factor (Sigma Aldrich #E9644), hydrocortisone (Sigma Aldrich #H0888), insulin (Sigma Aldrich #11376497001), and 10% horse serum (Sigma Aldrich #H1138). MCF7 cells were cultured in Minimum Essential Medium (MEM) (Thermo Fisher Scientific #11095098) supplemented with 10% FBS. BT474 cells were cultured in Roswell Park Memorial Institute Medium (RPMI) 1640 with L-glutamine (Corning #10-040-CV) supplemented with 10% FBS. Cells were used for no more than ten passages after thawing and were passaged using 0.25% trypsin–EDTA solution (Sigma Aldrich #T4049). Cell lines were routinely tested for mycoplasma using a PCR-based (polymerase chain reaction) MycoDtect kit (Greiner Bio-One, Monroe, NC, USA) and authenticated by STR (short tandem repeat) profiling at the Genetic Resources Core Facility of the Johns Hopkins University School of Medicine.

Genomic DNA isolation and coding DNA sequencing
Cell pellets were washed with PBS and genomic DNA was isolated using the QIAamp DNA Mini kit (Qiagen #51304, Germantown, MD, USA) according to the manufacturer’s instructions (Appendix B: Protocol for Cultured Cells). DNA concentration was measured using a Biotek Epoch microplate spectrophotometer (Agilent, Lexington, MA, USA) and a Take3 microvolume plate (Agilent). Each of the nine coding DNA sequences (CDS) were amplified separately using iProof High-Fidelity DNA Polymerase (Bio-Rad #172-5302, Hercules, CA, USA) and Promega dNTP mix (Promega #U1518, Madison, WI, USA) according to the manufacturer’s instructions. Amplification conditions were optimized for each coding DNA sequence with the following alterations to the standard manufacturer’s protocol: 100 ng genomic DNA was used per well for all codons except CDS1 (125 ng/well), CDS6 (200 ng/well) and CDS9 (200 ng/well); GC buffer was used for CDS1, with 10% formamide (Sigma Aldrich #F9037) and 0.1 µg/µL bovine serum albumin (Thermo Fisher Scientific #23209). Annealing temperatures were also optimized for each CDS, and optimized annealing temperatures were as follows: CDS1, CDS2, and CDS8: 62.5 °C; CDS3, CDS5, and CDS7: 65.0 °C; CDS4: 70.5 °C; CDS6 and CDS9: 64.1 °C. Primer sequences are available in Supplemental Table S1 in the Supporting Information.
Following successful amplification of all EpCAM coding DNA sequences, amplification products were purified by agarose gel electrophoresis. Briefly, 2% agarose (VWR #490001-580, Radnor, PA, USA) gel was prepared in 1 × TAE buffer (Tris–acetate-EDTA, VWR #75800-922) with SYBR Safe DNA Gel Stain (Thermo Fisher Scientific #S33102). Gel was loaded with GeneRuler DNA Ladder Mix (Thermo Fisher Scientific #SM0333) and CDS amplification products mixed with NEB Gel Loading Dye (New England Biolabs #B7024S, Ipswich, MA, USA) and run at 150 mA for 45 min. Bands were excised using a blue light LED transilluminator (VWR #76151-834) and purified using QIAquick PCR & Gel Cleanup kit (Qiagen #28506) according to the manufacturer’s instructions. Sanger sequencing of products was performed at the Johns Hopkins University Genetic Resources Core Facility using an Applied Biosystems 3730xl DNA Analyzer (Thermo Fisher Scientific), and sequence analysis and alignment were performed using Snapgene (San Diego, CA, USA).

Generation of stable cell lines
HEK293T cells were transfected with pLentiCRISPRv229 with EpCAM guide RNA30 (Genscript, Piscataway, NJ, USA), lentiviral packaging plasmid psPAX2 (courtesy of Didier Trono, Addgene #12260, Watertown, MA, USA), and envelope plasmid PMD2.G (courtesy of Didier Trono, Addgene #12259) using Lipofectamine 3000 (Thermo Fisher Scientific #L3000-008) according to the manufacturer’s instructions. Three pLentiCRISPRv2 plasmids with three different EpCAM guide RNAs were used. The supernatant containing virions was harvested after 48 h, centrifuged, filtered with 0.45 µm PES (polyethersulfone) membrane filters, and snap frozen in liquid nitrogen. Lentiviral transduction of MDA-MB-468 cells with the pLentiCRISPRv2-containing viruses (an even mixture of all three EpCAM guide RNAs was used) was performed with 6 µg/mL polybrene (Sigma Aldrich #TR-1003-G) for 48 h, after which infected cells were selected using 4 µg/mL puromycin (Sigma Aldrich #P8833). After puromycin selection and a round of cell sorting with an Aria IIu Cell Sorter the Johns Hopkins Ross Flow Cytometry Core Facility using an anti-EpCAM antibody conjugated to Alexa Fluor 647 (1:200 dilution, Santa Cruz Biotechnology #sc25308-AF647, Dallas, TX, USA), the EpCAM knockout (KO) remained unstable, showing EpCAM expression to slowly increase again over time, indicating a mixed EpCAM KO and WT cell population. Single cell clonal selection was then performed, which yielded stable EpCAM KO cells.
To generate the EpCAM overexpressing (OV) and mutEpCAM expressing plasmids, custom genes were produced by Thermo Fisher Scientific GeneArt and inserted into pLentiCMV-blast31 (courtesy of Eric Campeau & Paul Kaufman, Addgene #17486) using restriction enzyme cloning with XbaI (New England Biolabs #R0145S) and SalI-HF (New England Biolabs #R3138S) using NEB rCutSmart Buffer (New England Biolabs #B6004S). Fragments were isolated by agarose (VWR #490001-580) gel electrophoresis as described above and purified using QIAquick PCR & Gel Cleanup kit (Qiagen #28506). Ligation was performed with NEB Quick Ligation kit (New England Biolabs #M2200S) and plasmids were transformed into NEB Stable Competent E. coli (New England Biolabs #C3040H) using heat shock at 42 °C for 30 s. Outgrowth was performed with 10-beta/ Stable Outgrowth medium (New England Biolabs # B9035S), after which cells were streaked on LB Agar plates with 100 µg/mL ampicillin (Quality Biological #340-108-231, Gaithersburg, MD, USA) and incubated overnight at 37 °C. Colonies were picked the following morning, and liquid culture was inoculated in LB broth (Thermo Fisher Scientific #10855-021) with 100 µg/mL ampicillin (Quality Biological #351-344-731) and incubated at 37 °C with shaking at 200 RPM overnight. Plasmids were then isolated from liquid culture the following day using the Plasmid Maxi kit (Qiagen #12163). Plasmid sequences were confirmed using Plasmidsaurus whole plasmid sequencing (Eugene, OR, USA).
HEK293T cells were transfected with empty vector pLentiCMV-blast, pLentiCMV-blast-EpCAMOV, or pLentiCMV-blast-mutEpCAM, lentiviral packaging plasmid psPAX2, and envelope plasmid PMD2.G using Lipofectamine 3000 according to the manufacturer’s instructions. The supernatant containing virions was harvested after 48 h, centrifuged, filtered with 0.45 µm PES (polyethersulfone) membrane filters, and snap frozen in liquid nitrogen. Lentiviral transduction of EpCAM KO MDA-MB-468 cells with the empty vector pLentiCMV-blast, pLentiCMV-blast-EpCAMOV, or pLentiCMV-blast-mutEpCAM containing virions was performed with 6 µg/mL polybrene for 48 h, after which infected cells were selected using 2 µg/mL blasticidin (Thermo Fisher Scientific #A11139-03). After blasticidin selection, the expression of EpCAM was stable in all four cell lines: WT MDA-MB-468 (as purchased from ATCC), EpCAM KO (EpCAM KO cells infected with pLentiCMV-blast empty vector), EpCAM OV (EpCAM KO cells infected with pLentiCMV-blast-EpCAMOV), and mutEpCAM (EpCAM KO cells infected with pLentiCMV-blast-mutEpCAM).

Protein lysate preparation
Cells were washed with PBS (phosphate buffered saline, Quality Biological #119-069-131) prior to addition of RIPA buffer (Sigma Aldrich #R0278), a strong ionic and non-ionic detergent-based buffer, with Halt protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific #1861281). Cells were collected and homogenized by vortexing for 10 s followed by 20 min rest on ice, repeated three times. Samples were centrifuged at 16,100 × g for 20 min at 4 °C. Protein lysates were isolated from pellets and total protein concentration was measured using a Pierce BCA Protein Assay kit (Thermo Scientific #23227) and an Epoch microplate spectrophotometer (Agilent) according to the manufacturer’s instructions, using an absorbance wavelength of 562 nm.

PNGaseF digestion
4 × SDS-PAGE (sodium dodecyl-sulfate polyacrylamide gel electrophoresis) loading buffer was prepared by mixing 4 × Laemmli sample buffer (Bio-Rad #1610747) with 2-mercaptoethanol (Thermo Fisher Scientific #21985023) according to the manufacturer’s instructions. For each sample, 32 µg protein and 4 µL 4 × SDS-PAGE loading buffer was added to DI H2O to reach a final volume of 17 µL and this mixture was heated at 95 °C for 10 min to denature soluble proteins. The samples were cooled on ice, and 2 µL of the non-ionic detergent 10% NP-40 (VWR #M158) and 1 µL of PNGaseF prime (Bulldog Bio #NZPP010LY, Portsmouth, NH, USA) were added to each sample and incubated at 37 °C for 30 min to cleave N-glycans.

Western blot
For samples not digested with PNGaseF, protein was diluted in water and mixed 3:1 with 4 × SDS-PAGE loading buffer for a final protein concentration of 2 mg/mL, and this mixture was heated at 95 °C for 10 min to denature soluble proteins. Samples digested with PNGaseF were prepared as described above. For each sample, 10 µL was loaded onto a 10% PAGE gel (Bio-Rad #4561033, #4561036) for separation via electrophoresis for 50 min at 120 V and transferred to an Immobilon-P PVDF (polyvinylidene fluoride) membrane (Sigma Aldrich #IPVH00010) for 45 min at 100 V. Membrane was washed for 5 min in PBS, then incubated for 45 min at room temperature in blocking buffer of 5% skim milk (Sigma Aldrich #1153630500) in PBS. The membrane was then washed three times in PBST (PBS with 0.1% tween), followed by overnight incubation at 4 °C with primary antibody in blocking buffer. The membrane was then washed three times with PBST followed by incubation for 45 min at room temperature with horseradish peroxidase conjugated secondary antibody, against the species in which the primary antibody was raised, diluted 1:2,000 in blocking buffer. The membrane was then washed three times in PBST. Chemiluminescence was detected using a ChemiDoc imaging system (Bio-Rad) following incubation using the SuperSignal West Pico chemiluminescent substrate kit (Thermo Fisher Scientific #34577) according to the manufacturer’s protocol. Antibodies used were anti-EpCAM antibody (1:500 dilution; Santa Cruz Biotechnology #sc25308); anti-α-tubulin antibody (1:500 dilution, Sigma Aldrich #T5168); anti-Ku70 antibody (1:500 dilution, Cell Signaling Technology #4588S, Danvers, MA, USA); anti-EGFR antibody (1:2,000 dilution, Proteintech #66455-1-Ig, Rosemont, IL, USA); anti-E-cadherin antibody (1:1,000 dilution, Proteintech #20874-1-AP); anti-β-catenin antibody (1:250 dilution, Cell Signaling Technology #8814S); anti-cyclin E antibody (1:100 dilution, Cell Signaling Technology #4129S); anti-cyclin A antibody (1:500 dilution, Santa Cruz Biotechnology #sc271882); anti-GAPDH antibody (1:3,000 dilution, Sigma Aldrich #G8795); anti-mouse IgG, HRP-linked antibody (1:2,000 dilution, Cell Signaling Technology #7076S); anti-rabbit IgG, HRP-linked antibody (1:2,000 dilution, Cell Signaling Technology #7074S). All densitometry analysis was performed using Image Lab software (Bio-Rad).

RNA extraction, cDNA synthesis, and qRT-PCR
Cell pellets were washed with PBS and lysed in RLT lysis buffer (Qiagen #74106), vortexed, and homogenized using QIAshredder columns (Qiagen #79654). RNA was extracted using the RNAeasy mini kit (Qiagen #74106) according to the manufacturer’s instructions, using the RNase-Free DNase Set (Qiagen #79256) for DNA digestion. RNA concentration was measured using an Epoch microplate spectrophotometer and a Take3 microvolume plate (Agilent). iScript Reverse Transcription Supermix (Bio-Rad #1708841) was used to synthesize cDNA, for quantitative real-time PCR by iQ SYBR green supermix (Bio-Rad #1708882). Biorad CFX384 (Bio-Rad) was used for thermal cycling. The DDCT method was used to quantify results, with β-actin used as the housekeeping gene. Primers used for these experiments are listed in Supplemental Table S1 in the Supporting Information. Experiments were performed using three technical replicates and four biological replicates.

Cycloheximide protein stability assay
Cells were counted and plated in DMEM containing 10% FBS at 1 × 106 cells/well in 6 well plates, with one well each of WT and mutEpCAM cells for each time point (starting with the 0-h timepoint) and left to adhere overnight at 37 °C and 5% CO2. The following day, 200 µg/mL cycloheximide (Selleck Chemicals #S7418, Houston, TX, USA) was added to all wells except the 0-h wells, which were harvested using 0.25% trypsin–EDTA solution. Cells were washed with PBS and 30 µL RIPA buffer with Halt protease/phosphatase inhibitor cocktail was added to each sample. Cells were collected and homogenized by vortexing for 10 s followed by 20 min rest on ice, repeated three times. Samples were centrifuged at 16,100 × g for 20 min at 4 °C. Supernatants were removed, and 10 µL of 4 × SDS-PAGE loading buffer was added to each sample. This mixture was heated at 95 °C for 10 min to denature soluble proteins. After cooling, 15 µL of each sample was loaded onto a 10% PAGE gel and Western blot proceeded as described above, using Ku70 as a loading control due to its long half-life.

Immunofluorescence staining and confocal microscopy of cells
Cells were counted and 4 × 105 cells in 1 mL of DMEM containing 10% FBS were seeded in each chamber of a 4-chamber slide (Thermo Fisher Scientific #154917) and left to adhere overnight at 37 °C and 5% CO2. The following day, cells were washed briefly twice with 500 µL/well PBS. Cells were then fixed for 10 min in 500 µL/well with 4% paraformaldehyde in PBS at room temperature. Fixative was removed and cells were washed twice with 500 µL/well PBS for 5 min on a rocker. Wheat germ agglutinin (WGA) conjugated to Alexa Fluor Plus 568 (Thermo Fisher Scientific #W56133) was diluted 1:2000 in PBS, and cells were incubated with 500 µL/well of dilute WGA stain at room temperature for 10 min on a rocker. WGA solution was removed, and cells were washed twice with 500 µL/well PBS for 5 min on a rocker. Cells were then permeabilized with 500 µL/well ice-cold methanol at − 20 °C for 10 min. Methanol was removed and cells were washed twice with 500 µL/well PBS for 5 min on a rocker. Cells were then incubated with 500 µL/well blocking serum of 10% normal mouse serum (Thermo Fisher Scientific #10,410) in PBS for 30 min at room temperature on a rocker. Blocking serum was removed and cells were washed twice with 500 µL/well PBS for 5 min on a rocker. Cells were then incubated with 500 µL/well primary antibody solution overnight at 4° C on a rocker. Primary antibodies used were anti-calnexin conjugated to Alexa Fluor 488 (1:200 dilution, Santa Cruz Biotechnology #sc23954-AF488) and anti-EpCAM conjugated to Alexa Fluor 647 (1:100 dilution, Santa Cruz Biotechnology #sc25308-AF647) diluted in blocking serum. The following morning, primary antibody solution was removed, and cells were washed with three changes of 500 µL/well PBS for 5 min on a rocker. Chambers were removed from slides, and the coverslip was mounted with Prolong Gold Antifade Reagent with DAPI (Thermo Fisher Scientific #P36931) and sealed. After curing, slides were examined using a Zeiss LSM800 Confocal Laser Scanning microscope in the Johns Hopkins School of Medicine Microscope Facility and images were acquired using Zen Blue software and analyzed using ImageJ software (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.net/ij/, 1997–2018.)52. Colocalization analysis was performed using the JACoP plugin (https://imagej.net/plugins/jacop)52,53. Briefly, five individual cells in each image (63 × magnification) were selected as regions of interest, the JACoP plugin was applied to these regions, and the Pearson’s coefficient and Mander’s overlap coefficients were recorded. Of cells which had clear cell membrane and endoplasmic reticulum staining, cells were chosen such that variation in EpCAM staining within the image was represented. This was repeated with five biological replicates, and the Pearson’s coefficients were analyzed with GraphPad Prism using a one-way ANOVA with Tukey’s multiple comparisons test.

Viability assay
Cells were counted and 4 × 103 cells/well in 100 µL of DMEM containing 10% FBS were seeded in a 96 well plate and left to adhere overnight at 37 °C and 5% CO2. The following day, 10 µL of WST-1 cell proliferation reagent (Sigma Aldrich #11644807001) was added to each well and incubated at 37 °C and 5% CO2 for three hours. Absorbance was measured at 440 nm using an Epoch microplate spectrophotometer. Experiments were performed with five technical replicates and three biological replicates.

Migration and invasion assays
Matrigel basement membrane aliquots were thawed overnight at 4 °C prior to the experiment. (Corning #35434) To prepare transwell inserts (Corning #3422), inserts were added to sterile 24-well plates, and 700 µL of serum-free DMEM was added to the well outside of the insert. For inserts used for migration assays, 100 µL of serum-free DMEM was added to the insert. For inserts used for invasion assays, 80 µL of Matrigel (diluted 1:10 in serum-free DMEM) was added to each insert. Plates containing inserts were then incubated at 37 °C and 5% CO2. Media was then removed from well outside of insert and from inside inserts used for migration assays. Cells were counted and 4 × 105 cells/well in 100 µL serum-free DMEM were added to each insert. In the well outside of each insert, 100 µL of DMEM containing 10% FBS was added, and plates were incubated for 48 h at 37 °C and 5% CO2. Media was then removed from both inside and outside inserts, and cells were fixed with cold methanol at room temperature for ten minutes. Methanol was removed and cells were stained with 0.1% crystal violet (Sigma Aldrich #C0775) at room temperature for 20 min. Inserts were then cleaned with DI H2O and cotton tipped applicators and 100 µL DI H2O was added to each insert. Images of stained cells were acquired at four fields per insert at 20 × magnification using a Nikon TS100 inverted microscope with a charge-coupled device camera (Tokyo, Japan). Migrating/invading cells were quantified using ImageJ52. Migration assays were performed with five biological replicates and invasion assays were performed with four biological replicates.

Homotypic cell–cell adhesion assay
Prior to experiment, cells were seeded in four wells/ cell line/ timepoint of a 24-well plate and allowed to reach confluence. The day of the experiment, a separate population of each cell line were washed twice with PBS, counted, and 2 × 106 cells each were resuspended in 400 µL serum-free DMEM (5 × 106 cells/mL). Calcein AM dye (Vybrant Cell adhesion Assay Kit, Thermo #V13181) was diluted 2:5 to 400 nM in serum-free DMEM. 2 µL of the dilute calcein AM stock solution was added to each 400 µL cell suspension to achieve a final concentration of 2 µM and mixed well. Suspensions were incubated at 37 °C for 30 min. After 30 min, cell suspensions were mixed well and 100 µL (5 × 105 cells) was removed from each cell suspension and added to a separate control tube. Cells were washed twice with prewarmed (37 °C) serum-free DMEM and control cells were washed twice with sterile PBS. The cells were resuspended in serum-free DMEM for a final concentration of 1.5 × 105 cells/mL, and the control cells were resuspended in sterile PBS for a final concentration of 1.5 × 105 cells/mL. Media was removed from confluent plates and washed once with serum-free DMEM. 1 mL/well of the calcein-labeled cell suspension (1.5 × 105 cells) was added to prepared microplate wells containing confluent cells (medium removed), and 1 mL/well of each control cell suspension was added to 2 wells of separate 24-well plate and incubated at 37 °C for 30 min-6 h. Nonadherent calcein-labeled cells were removed by carefully adding prewarmed serum-free DMEM to each well and gently swirling, inverting the plate, and blotting the excess liquid onto paper towels (control wells were not washed). This washing procedure was repeated four times, and 1 mL of PBS was added to each well. The fluorescence was measured using a Molecular Devices SpectraMax iD3 Multi-Mode Microplate Reader (San Jose, CA, USA) with an excitation wavelength of 494 nm and an emission wavelength of 535 nm. The percentage of adhesion was determined by dividing the corrected (background subtracted) fluorescence of adherent cells by the total corrected fluorescence of cells added to each microplate well. Experiments were performed with three technical replicates and 3–6 biological replicates at each timepoint.

Supplementary Information
Below is the link to the electronic supplementary material.