Abstract
Background/Aim: Breast cancers constitute heterogeneous tumor groups and their categorization in subtypes based on the expression of the estrogen (ER), progesterone (PR) and HER2 receptors has advanced therapeutics. Claudin-low breast cancer has been proposed as an additional subtype which is mostly ER, PR and HER2 negative, but its identification has not led to corresponding specific treatments yet. Materials and Methods: Breast cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) were assessed for mRNA suppression of claudins and mRNA expression of ER and ERBB2 (the gene encoding HER2). The set of identified claudin-low cell lines were compared with representative ER−/ERBB2− cell lines for associated molecular alterations, gene dependencies through CRISPR and microRNA arrays and in vitro drug sensitivities using the Genomics of Drug Sensitivity in Cancer (GDSC) project. Results: Claudin-low cell lines display up-regulation of mRNA expression of epithelial to mesenchymal transition (EMT) regulators. Methylation sensitive genes are down-regulated in claudin-low lines compared with other cell lines, without associated up-regulation of DNA methyltransferases. Dependency screen microarrays reveal dependencies of claudin-low cell lines on components of the cytoskeleton but no consistent dependencies in known oncogenes or tumor suppressors. Potential drug sensitivities revealed in the drug screens included sensitivities to WNT pathway modulators, tyrosine kinase cascade inhibitors and BET inhibitors. On the other hand, claudin-low cell lines showed resistance to deacetylase inhibitors. Conclusion: Claudin-low cell line models duplicate features of claudin-low breast cancers and may serve as guides for identification of drugs worth exploring for further development.
Triple negative breast cancers characterized clinically by negativity for the estrogen receptor (ER), the progesterone receptor (PR) and the human epidermal growth factor receptor 2 (HER2) represent about 10% to 15% of all breast cancers but produce a higher burden of morbidity and mortality among breast cancer patients due to their aggressive behavior and worse prognosis (1). In addition, triple negative breast cancers have fewer therapeutic options, lacking the targeting opportunities of hormonal agents and anti-HER2 drugs that ER positive and HER2 positive cancers benefit from (1, 2). Although, triple negative cancers are overwhelmingly characterized by deleterious mutations in the gene TP53 encoding for tumor suppressor p53, they are otherwise a heterogeneous disease group with no unifying molecular alterations (3). Diverse molecular classifications disclose at least three distinct subsets. The two most extensive subsets include basal-like cancers, with two sub-categories that are immune activated and immune suppressed, respectively, and a mesenchymal or claudin-low subset (4, 5). A third smaller group of triple negative breast cancers consists of luminal-like cases, which by definition lack expression of ER and PR but express another steroid nuclear receptor, the androgen receptor (AR) (6, 7). Claudin-low breast cancers constitute a subtype of the disease with special molecular characteristics that include, as the name implies, suppressed expression of the tight junction proteins claudins and other adhesion proteins (8, 9). Moreover, claudin-low cancers display features of the epithelial to mesenchymal transition (EMT), a process associated with metastasis (10). EMT as a process in cancer is interwoven with cell plasticity and the stem cell phenotype (11, 12). Most claudin-low breast cancers are triple negative, but smaller percentages of ER positive and HER2 positive breast cancers display suppression of claudins and other adhesion proteins and the claudin-low phenotype. For example, in the METABRIC cohort, 6% of ER+/HER2− cancers with low proliferation index are claudin-low (13).
Currently, the heterogeneity of triple negative breast cancers is only translated to clinical therapeutic decisions in the minority of cases that are targetable with PARP inhibitors or immune checkpoint inhibitors, due to BRCA1/BRCA2 mutations or high (>10%) expression of the PD-L1 receptor (14-17). In contrast, triple negative breast cancers without these molecular characteristics, representing the majority of cases, are treated with chemotherapies as their only systemic therapeutic option (18). Discovery of new targeted treatments are thus a clinical need. These discoveries could be aided by cell line models capturing the heterogeneity of triple negative breast cancer phenotypes. Claudin-low cell lines in particular may be a valuable instrument for in vitro and in vivo preclinical studies of this subset. In the current investigation, breast cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE) with claudin-low characteristics were evaluated for molecular attributes and sensitivities to various drugs. Dependencies to gene knock-out using gene knock-out arrays were also examined to identify potential synthetic lethalities with the aim to guide rational future targeted therapeutics.
Materials and Methods
Cell lines. Cancer cell lines included in the current investigation constitute part of the CCLE collection (19, 20). Molecular attributes of breast cancer cell lines of interest were extracted from the CCLE data as presented in the cBioportal for Cancer Genomics (http://www.cbioportal.org) site (21, 22). Claudin-low breast cancer cell lines were identified in the CCLE collection through analysis of mRNA expression of key adhesion molecules, including claudin 3, claudin 4, claudin 7, and E-cadherin. Breast cancers with suppressed expression of these adhesion molecules were categorized in the claudin-low group and were compared with cell lines without suppressed expression. This latter group was subdivided according to the expression of ER (encoded by ESR1) and HER2 mRNA (encoded by ERBB2) into 4 groups (ER−/ERBB2−, ER+/ERBB2−, ER+/ERBB2+, and ER−/ERBB2+).
Data extraction and analysis from online databases. Molecular alterations, such as mutations, copy number alterations, and fusions were identified in the CCLE collection of cell lines through whole-exome sequencing (20). Copy number alteration measurement was performed with the GISTIC (Genomic Identification of Significant Targets in Cancer) algorithm, according to which a gene was considered putatively amplified if it had a score of 2 or above, while genes with a score of −2 or below were considered deleted (23). Genes with scores between −2 and 2 were considered not copy number altered. For the mRNA expression analysis, the RSEM algorithm was used for normalization of mRNA expression (24). mRNA grids were generated with the online tool provided by cBioportal.
The functional implications of identified mutations in cell lines were assessed with the use of the OncoKB tool. OncoKB knowledgebase is a database of cancer-related genes and classifies cancer-related genes as oncogenes or tumor suppressors (25). Drug sensitivity of claudin-low and other breast cancer cell lines were assessed in the Genomics of Drug Sensitivity in Cancer (GDSC) dataset (www.cancerrxgene.org). The dataset was interrogated using the cell line tool available in the online portal (22). The GDSC project includes two datasets of drug sensitivity arrays, GDSC1 and GDSC2, which differ in the experimental conditions employed and thus results are not completely overlapping. Dependencies on knocking out of specific genes of cell lines of interest were obtained from the Cancer Dependency Map Project (depmap.org) portal that contains data from both CRISPR arrays and RNA-interference arrays for CCLE cell lines (26, 27). These gene arrays examine the included cell lines for essential genes that are important for their survival and, as a result, when they are knocked out, produce significant reduction in cell survival and proliferation in vitro (28-30).
Statistical analysis. Statistical comparisons of categorical data were performed using the Fisher’s exact test or the χ2 test. The Mann–Whitney U-test was used to compare medians. Corrections for multiple comparisons were performed with the Bonferroni adjustment. All statistical comparisons were considered significant if p<0.05.
Results
The breast cancer cohort of CCLE consists of 71 cell lines and contains 8 cell lines (11.3%) with down-regulation of claudins 3, 4 and 7 and E-cadherin at the mRNA level (Table I, Figure 1). All 8 claudin-low cell lines are ER and ERBB2 negative. No cell lines with increased ER or ERBB2 expression display claudin suppression. Representative cell lines with non-suppressed claudin expression are also shown in Figure 1A. A sub-set of luminal A clinical breast cancers are claudin-low, and these were not represented in the breast cancer cell line collection (8, 13, 31). EMT core transcription factors including ZEB1, ZEB2, SNAIL1, SNAIL2 (Slug), FOXC2 and TWIST1 were up-regulated, although not homogeneously, in claudin-low cell lines, while they were suppressed in cell lines without claudin suppression (Figure 1B). All claudin-low breast cancer cell lines with available data were microsatellite stable and polyploid (Table I). Their tumor mutation burden ranged from 22.58 mutations/Mb to 51.77 mutations/Mb.
ER−/HER2− claudin-low breast cancer cell lines included in CCLE.
mRNA expression calculated as z-score relative to all samples (logarithmic normalization of RNA Sequencing as reads per kilobase million, RPKM) in breast cancer cell lines of the claudin-low phenotype, representative of other ER−/ERBB2+ cell lines and cell lines with ER or ERBB2 positivity. (A) Expression of ESR1, ERBB2, claudins 3, 4 and 7 and E-cadherin. (B) Expression of core EMT transcription regulators. Red color denotes overexpression and blue denotes suppression. Data were extracted from the Cancer Cell Line Encyclopedia. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B; ESR1: estrogen receptor 1 gene; EMT: epithelial mesenchymal transition.
The only common molecular lesions observed in all 8 claudin-low cell lines were TP53 mutations. These mutations were diverse, but they were all considered oncogenic or likely oncogenic in the OncoKB database. However, TP53 mutations are also frequently observed in ER−/HER2− breast cancers and corresponding cell lines. For example, 5 of 6 (83.3%) representative ER−/HER2− cell lines without claudin suppression possessed oncogenic or likely oncogenic TP53 mutations (Figure 2A). Among the 7 claudin-low cell lines with copy number alteration data available, 6 cell lines possessed amplifications of MYC at chromosome locus 8q24.21. MYC amplifications were also observed in 3 of the 6 representative ER−/HER2− cell lines without claudin suppression (Figure 2B).
Percentage of molecular alterations in claudin-low breast cancer cell lines and representative basal ER−/ERBB2− cell lines and ER+ or ERBB2+ cell lines. (A) TP53 mutations. (B) MYC amplifications. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B.
Alterations in core EMT regulators were present in 3 of the 8 claudin-low cell lines. Cell line MDA-MB-157 possessed a frameshift mutation in ZEB1 as well as amplification of SNAI1. Cell line BT-549 possessed amplification of SNAI2 and MDA-MB-436 had a point mutation of unknown significance as well as amplification of TWIST1. As a comparison, two of the 6 representative ER−/HER2− cell lines without claudin suppression (HCC-70 and HCC-1143) possessed alterations (amplifications in ZEB1 and TWIST1 the former and in SNAI2 the latter) in core EMT regulators (data not shown).
Epigenetic modifications include DNA methylation and histone methylation and acetylation; they provide the opportunity of altering the expression of genes without changes in the genome during epithelial to mesenchymal transition (32). A survey of DNA methylation sensitive genes, including CEACAM6, CDH1, CST6, LCN2, MYB, SCNN1A, TFF3, and CCND2, in claudin-low breast cancer cell lines showed that they were down-regulated, suggestive of active promoter methylation, compared with cell lines without claudin suppression (Figure 3A). The mechanism was not up-regulation of DNA methyltransferases DNMT1, DNMT3A and DNMT3B, as the 3 enzymes were not up-regulated in claudin-low cell lines compared with other cell lines (Figure 3B). The median mRNA expression of DNMT1 in claudin-low cell lines was not significantly different from each of the other groups of cell lines (Mann-Whitney U-test, p>0.05, Figure 3C). DNMT3A and DNMT3B median mRNA expression of claudin-low cell lines was down-regulated compared with ER+ERBB2− and ER−/ERBB2+ cell lines (Mann-Whitney U-test, p<0.05) but not significantly different from ER−/ERBB2− non-claudin suppressed and ER positive/ERBB2 positive cell lines.
mRNA expression calculated as z-score relative to all samples (logarithmic normalization of RNA Sequencing as reads per kilobase million, RPKM) in breast cancer cell lines of the claudin-low phenotype, representative other ER−/ERBB2− cell lines and cell lines with ER or ERBB2 positivity. (A) Expression of methylation sensitive genes. (B) Expression of DNA methyltransferases DNMT1, DNMT3A and DNMT3B. (C) Median values presented in B for each cell line group mRNA expression. Red color denotes overexpression and blue denotes suppression. Data are from the Cancer Cell Line Encyclopedia. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B; DNMT1: DNA methyltransferase 1; DNMT3A: DNA methyltransferase 3A; DNMT3B: DNA methyltransferase 3B.
Methylation of histone H3 is also a key epigenetic modification and is regulated by methyltransferases and demethylases adding or removing methyl groups at lysines located in positions 4, 9, 27, 36 and 79 of this histone. Compared with ER+ or ERBB2+ breast cancer cell lines, demethylases of lysine 27, and particularly KDM7A and PHF8, showed lower mRNA expression in claudin-low cell lines. However, claudin non-suppressed ER−/ERBB2− breast cancer cell lines displayed also down-regulation of these demethylases of lysine 27, suggesting that this is a phenomenon associated with all triple negative cell lines rather than the claudin-low subset exclusively (Figure 4). In addition, partially due to small numbers, differences were not statistically significant, except for the comparison of mRNA expression of PHF8 in claudin-low and ER+/ERBB2+ cell lines. H3K4 methyltransferases KMT2B, KMT2C and KMT2D, and also SETD1B were suppressed in several claudin-low cell lines and were not suppressed or were even slightly up-regulated in most of other cell lines (not shown). H3K4 demethylases KDM5B and KDM5C were up-regulated in non-claudin-low cell lines but were not consistently suppressed in claudin-low cell lines. Expression of methyltransferases and demethylases of H3 lysines at positions 9, 36 and 79 showed higher variability in claudin-low cell lines (data not shown). Overall, although exceptions existed in the cell line landscape, a combinatorial epigenetic deregulation with DNA hypermethylation and histone H3 hypermethylation at lysine 27 combined with hypomethylation at lysine 4, may provide a favorable epigenetic state for claudins suppression and the claudin-low phenotype development.
mRNA expression calculated as z-score relative to all samples (logarithmic normalization of RNA Sequencing as reads per kilobase million, RPKM) in breast cancer cell lines of the claudin-low phenotype, representative other ER−/ERBB2− cell lines and cell lines with ER or ERBB2 positivity. (A) Expression of histone 3 lysine 27 (H3K27) methyltransferase and demethylase. (B) Median values of mRNAs for the demethylases KDM7A and PHF8 in each cell line group presented in (A). Red color denotes overexpression and blue denotes suppression. Data were extracted from the Cancer Cell Line Encyclopedia. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B; KDM7A: lysine demethylase 7A; PHF8: PHD finger protein 8.
Regarding acetylation related enzymes, claudin-low cell lines displayed suppressed expression of several acetyltransferases, while other cell lines tended to have higher expressions of these enzymes, with the exception of KAT2A and KAT2B, which were more globally suppressed in most breast cancer cell lines (Figure 5). Acetyltransferase KAT7 in particular was up-regulated in many, but not all, ER+ cell lines and in many ERBB2 over-expressing cell lines, as the KAT7 gene is located close to ERBB2 at chromosome 17q21 and is often co-amplified (Figure 5). Among deacetylases, HDAC11 was up-regulated, while HDAC9 was down-regulated in ER+ cell lines compared with both claudin-low and other ER− cell lines (Figure 6).
(A) mRNA expression of histone acetyltransferases calculated as z-score relative to all samples (logarithmic normalization of RNA Sequencing as reads per kilobase million, RPKM) in breast cancer cell lines of the claudin-low phenotype, representative other ER−/ERBB2− cell lines and cell lines with ER or ERBB2 positivity. (B) Median values for acetyltransferase KAT7 in each group of cell lines. Red color denotes overexpression and blue denotes suppression. Data are from the Cancer Cell Line Encyclopedia. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B; KAT7: lysine acetyltransferase 7.
(A) mRNA expression of histone deacetylases calculated as z-score relative to all samples (logarithmic normalization of RNA Sequencing as reads per kilobase million, RPKM) in breast cancer cell lines of the claudin-low phenotype, representative other ER−/ERBB2− cell lines and cell lines with ER or ERBB2 positivity. (B) Median values for deacetylases HDAC9 and HDAC11 in each group of cell lines. Red color denotes overexpression and blue denotes suppression. Data are from the Cancer Cell Line Encyclopedia. ER: Estrogen receptor; ERBB2: erythroblastic leukemia oncogene B; HDAC9: histone deacetylase 9; HDAC11: histone deacetylase 11.
In the CRISPR and RNAi screens for gene vulnerabilities in the 8 claudin-low breast cancer cell lines, there were no classic oncogenes or tumor suppressors in the top 10 list in more than one cell line, with the exception of cyclin dependent kinase CDK1, which was in the RNAi screen top list of 2 cell lines, Hs-578-T and MDA-MB-157 (Table II). KRAS, MCL1 and RAD51 were in the top 10 list of cell line HMC-1-8. SWI/SNF components. SMARCA4 and SMARCB1 were in the top 10 list of cell line BT-549. Cyclin dependent kinase CDK2 was in the top 10 list of cell line MDA-MB-157. Potentially relevant additional dependencies observed in more than one claudin-low cell line included protein phosphatase PPP1R12A (in cell lines Hs-578-T and MDA-MB-231), the alpha V integrin subunit gene ITGAV (in cell lines BT-549, MDA-MB-231 and SUM159PT), the dynactin subunit gene ACTR1A (in cell lines BT-549 and HMC-1-8), and the candidate tumor suppressor helicase INTS6 (in cell lines MDA-MB-436, MDA-MB-157, SUM157PT and HCC1395).
Six of the 8 claudin-low breast cancer cell lines were assayed for in vitro drug sensitivities and resistance in GDSC1 and GDSC2 collection of microarrays. Inhibitors of several key cancer pathways and processes were among the drugs with the highest sensitivity in the six cell lines. WNT pathway modulators, including porcupine inhibitors LGK974 and Wnt-C59, and GSK3B inhibitors KIN001-042 and CHIR-99021, figured in the top sensitivity lists of four cell lines (Table III). Inhibitors of receptor tyrosine kinases, such as IGF1R, IR, PDGFRA/B, and FGFR1-3, and of down-stream signal transduction targets, such as kinases PI3K and ERK1/2, were also in the top sensitivity list of three claudin-low cell lines. Inhibitors of proteins involved in DNA damage response and repair, such as PARP, ATM, DNAPK and CHEK1/2, were among the top sensitivities in four cell lines. Bromodomain inhibitors were present among the top sensitivity drugs in three cell lines. The clinically used chemotherapeutic fludarabine was also in the top sensitivity lists of two cell lines (Table III). Several receptor tyrosine kinases and down-stream signal transduction inhibitors as well as DNA damage response inhibitors were also in the top sensitivity lists of ER−/PR-non-claudin-low breast cancer cell lines (not shown). Two of six basal-like cell lines (HCC38 and HCC1143) displayed WNT pathway modulators, including porcupine, GSK3 kinase and tankyrase inhibitors, in their top sensitivities list. One cell line (CAL851) was sensitive to bromodomain inhibitors.
Most sensitive drugs in the 6 claudin-low breast cancer cell lines. Data are from the Genomics of Drug Sensitivity in Cancer (GDSC) dataset. Question marks denote that the drug target is not known or not confirmed.
On the other hand, one or more HDAC inhibitors were included in the top lists of drug resistance in four of the six claudin-low breast cancer cell lines (Table IV). However, IC50s of several HDAC inhibitors, including belinostat, dacinostat, entinostat, panobinostat, romidepsin and vorinostat, for claudin-low breast cancer cell lines were not significantly different from IC50s for ER−/PR− non-claudin-low cell lines (Mann-Whitney U-test, p>0.05 for the comparison of all HDAC inhibitors; data not shown). Various inhibitors of receptor tyrosine kinases and down-stream pathways were also among the drugs with highest resistance in both claudin-low and non-claudin-low cell lines.
Drugs to which the 6 assayed claudin-low breast cancer cell lines were most resistant in the GDSC1 and GDSC2 arrays. Data are from Genomics of Drug Sensitivity in Cancer (GDSC) dataset.
Discussion
The triple negative subset of breast cancers characterized clinically by negativity for the ER, the PR and the HER2 receptors is in fact a heterogeneous group with diverse molecular attributes and prognosis (5, 33). At least 3 distinct groups of triple negative breast cancers include the most extensive basal-like group, the claudin-low group and a luminal-like group which is positive for the AR (4, 34). Potential therapeutic implications of this heterogeneity are suggested but it has not been translated to treatment diversification, and all triple negative cancers are currently treated with the same protocols, based on cytotoxic chemotherapy (35). Subcategorization of triple negative breast cancers, although extensively characterized in research is not sought in the clinical arena, partially due to lack of evidence for distinct sensitivity profiles of these subsets to specific therapies. The claudin-low sub-group is characterized by suppression of adhesion molecules claudins, including family members claudin 3, claudin 4 and claudin 7, as well as suppression of the expression of other adhesion molecules, such as occludin and E cadherin (9). The claudin-low breast cancer phenotype expands beyond the triple negative subgroup and a minority of luminal cancers, mostly luminal A, display a claudin-low phenotype (8, 31). Thus, some claudin-low cancers are non-triple negative.
In the current investigation, 8 claudin-low breast cancer cell lines were identified in CCLE by suppressed mRNA expression of adhesion molecules. These cell lines included six of the nine original cell lines (MDA-MB-436, Hs-578-T, BT-549, MDA-MB-231, MDA-MB-157, SUM159PT) that were used to define the claudin-low centroid by the group that proposed the claudin-low feature as an intrinsic breast cancer sub-type (9, 36). The 3 other cell lines of the original study (MDA-MB-435, SUM1315MO2 and HBL-100) were excluded because either they are not of breast cancer origin or are not in the CCLE collection. Two additional breast cancer cell lines (HCC1395 and HMC-1-8) were determined to have low claudins expression and were included in the claudin-low group. Claudin-low cell lines were shown to display up-regulated expression of core EMT transcription regulators confirming an inverse relationship with claudin expression. Expression of EMT core regulators appears to be an innate characteristic of the claudin-low phenotype and a prerequisite for acquisition of claudin-low characteristics. Consistent with this, luminal MCF-7 breast cancer cells stably transduced with the transcription factor Snail acquire mesenchymal features and a genomic profile akin to the claudin low phenotype (37). In addition, they become resistant to ionizing radiation and to some chemotherapeutics, such as gemcitabine and mitomycin C, although they remain sensitive to others, such as doxorubicin, cisplatin, and 5-fluorouracil. In another study, adenovirus-mediated expression of Snail or Slug in MCF-7 cells up-regulated TGF-β signaling and induced claudin-low phenotype features (38). Up-regulation of TGF-β receptor TGFBR2 gene was associated with increased histone acetylation in the region of its promoter. Conversely, claudin-low MDA-MB-231 cells transfected with a lentivirus to obtain miR-200 family microRNAs re-expression, which are suppressors of core EMT transcription factors, acquired an epithelial phenotype (39). A role for SUZ12, a component of the H3K27 methyltransferase polycomb repressor complex 2 (PRC2), was proposed as the mediator of the observed miR-200 family forced expression effect. Thus, chromatin epigenetic regulations are involved in the molecular perturbations, forming the basis of the EMT transition states.
The miR-200 family members are intricately affected by hormone receptors expression in breast cancer cells (40, 41). A study that examined breast cancer cell lines clustered in an ER positive and an ER− group disclosed that miR-200 family microRNAs are down-regulated in the ER− cluster of cell lines and up-regulated in ER positive cell lines (42). However, ER expression is neither absolutely required nor sufficient by itself to up-regulate EMT suppressive microRNAs and abort EMT programs, as witnessed by the fact that not all ER− cancers are claudin-low, and some ER+ cancers display claudin-low features (8).
Epigenetic marks, such as DNA and histone methylation and histone acetylation, may contribute to the development of different epithelial and mesenchymal states across the EMT spectrum without permanent DNA alterations. As discussed above, epigenetic modifications can regulate expression of key EMT players, such as miRs and the TGF-β signaling cascade (38). Claudin-low cell lines have also been shown to down-regulate methylation sensitive genes, suggestive of active DNA promoter hypermethylation. The down-regulation of methylation sensitive genes is not explained by alterations of DNA methyltransferases, which, in fact showed suppressed expression compared with other cell lines (statistically significant for the de novo methyltransferases DNMT3A and DNMT3B when claudin-low cell lines were compared with ER+/ERBB2− and ER−/ERBB2+ lines). Thus, alternative perturbations must be at play in down-regulation of methylation sensitive genes in claudin-low cell lines.
DNA methylation affects gene suppression partially through recruiting histone methyltransferases and deacetylases and promoting methylation and deacetylation of neighboring chromatin (43-45). Thus, alterations in these enzymes or the demethylases and acetyltransferases that perform the reverse enzymatic functions on histones could mediate the permissive epigenetic state for the claudin-low phenotype. The most significant differences between ER−/ERBB2− (both claudin-low and claudin non-suppressed) and other cell lines in the expression of acetylation enzymes concern acetyltransferase KAT7 and deacetylases HDAC9 and HDAC11. Both KAT7 and HDAC11 enzymes show higher expression in the claudin non-suppressed group of breast cell lines, while HDAC9 is suppressed. Acetyltransferase KAT7, also known as HBO1 or MYST2, targets lysine 14 of H3 (H3K14ac) and may also perform other short fatty acid modifications at the same site, such as propionylation, butyrylation and crotonylation in conjunction with scaffold protein BRPF2 (46). KAT7 has a key role in the maintenance, quiescence, and renewal of hematopoietic stem cells (47). KAT7 knock-out mice were moribund due to marrow failure and bone marrow cells with deletion of KAT7 were ineffective in repopulating the bone marrow of ablated recipients. In contrast to this stem cell maintenance role, KAT7 inactivation alleviated senescence and repressed the cell cycle inhibitor p15 in human mesenchymal precursor cells derived from embryonic stem cells with mutations in the progeria genes causing Werner or Hutchinson-Gilford syndrome (48). In leukemogenesis, KAT7 associates with fusion proteins that contain methyltransferase KMT2A (also known as MLL) to promote trimethylation of H3K4 (49). Loss of KAT7 in leukemia cells with KMT2A fusions leads to proliferation arrest and apoptosis (50). H3K14 acetylation is critical for expression of HOXA9 and HOXA10 factors that participate in leukemia stem cell maintenance (51). In HeLa cervical cancer cells, which do not possess MLL fusions, KAT7 protects against chromosomal instability by participating in the assembly of histone CENP-A in the centromere, antagonizing methyltransferase KMT1A (SUV39H1), whose activity would lead to heterochromatin invasion into the centromere (52). KAT7 has also a role in pathologic proliferation of retinal vessel endothelium, promoting vessel sprouting (53).
Histone deacetylases remove acetyl groups from the ε amino moiety of lysines in histones, releasing the positive charge of these proteins, which promotes the closed DNA configuration (54). Acetyl groups are also docking sites for bromodomain proteins that initiate the process of transcription machinery assembly (55). HDAC11, the only member of class IV histone deacetylases has the shorter half-life among the HDAC super-family, being regulated by proteasomal degradation (56, 57). The gene encoding HDAC11 is located at the chromosomal locus 3p25.1 that is not commonly amplified in breast cancers (13). HDAC11 appears to possess weak deacetylase activity in vitro but it has a higher deacylase enzymatic activity for longer acyl chains (58). Epigenetic regulators mediated by HADC11 in vivo may require co-operation with other deacetylases, such as HDAC6 and HDAC2 (54, 59). HDAC11 is tethered in vitamin D receptor binding sites of the MYC promoter together with HDAC6, and HDAC2 and with the co-repressor complex. The complex suppresses the expression of MYC in normal immortalized prostate cells exposed to the ligand 1,25 dihydroxy-vitamin D (59). Relevant for its role in claudin suppression, in human intestinal epithelial cells, HDAC11 was reported to be increased in the promoters of tight junction proteins ZO-1, claudin 5 and occludin, in 1,25 dihydroxy-vitamin D deficient conditions (60). In these conditions, HDAC11 was bound to promoters of tight junction genes and inhibited their transcription, leading to a compromise of tight junctions between cells cultured in monolayers. In breast cancer, vitamin D deficiency is associated with the triple negative subtype (61). Together, these data suggest that in ER− breast cancer cells, low levels of HDAC11 may be sufficient to suppress tight junction protein expression in vitamin D depleted conditions.
HDAC9 belongs to the class IIa histone deacetylases and was associated with estrogen resistance in MCF7 cells (62). Transfection of HDAC9 in MCF7 cells led to suppression of ER expression and repression of ER signaling as measured by activity in promoters possessing estrogen receptor elements. Another study showed that HDAC9 displayed a higher mRNA and protein expression in basal breast cancer cell lines compared with luminal cell lines, consistent with the current report (63). In triple negative cell lines, HDAC9 suppresses expression of microRNA miR-206, a negative regulator of MAPK3 and VEGF (64). miR-206 suppression results in MAPK3 and VEGF up-regulation, promoting invasion and angiogenesis.
Dependencies of claudin-low cell lines revealed in the current evaluation include a few components and regulators of the cytoskeleton, consistent with the importance of this apparatus in cell shape, polarity, and motility during EMT. One such dependency, the myosin light chain phosphatase (MLCP) regulatory subunit PPP1R12A (also called myosin phosphatase target subunit 1, MYPT1) regulates the interaction of myosin with actin in non-muscle cells and is regulated by GTPase RhoA (65). RhoA receives signals from the non-canonical WNT pathway that regulates planar cell polarity (66). Induction of EMT in MCF7 breast cancer cells was associated with dramatic changes in the cytoskeleton with decreased actin cross-linking and increased contraction (67). Mutations in the mouse homologue of PPP1R12A was identified together with other genes to be involved in the regulation of actin cytoskeleton and to promote development of lobular breast cancers in a mouse model of insertional mutagenesis (68). PPP1R12A over-expression was associated with decreased proliferation and decreased invasion and motility potential of gastric cancer cells (69). Another recurrent dependency identified in claudin-low cell lines concerned the alphaV integrin sub-unit ITGAV, which together with beta3 sub-unit, constitutes the classical vitronectin receptor, and may also pair with subunits beta5 and beta1 (70). The receptor interacts with E cadherin and the cytoskeleton in breast cancer cells and has roles in migration of breast cancer cells (71). The ITGAV gene was among the most differentially up-regulated genes in a comparison of claudin-low breast cancers with other subtypes (9). It is evident that both dependencies, PPP1R12A and alpha V integrin are integral parts of the network of adhesion, polarity, and cell motility, providing input but also been regulated by the canonical/β-catenin dependent and the non canonical WNT pathways. Other dependencies of claudin-low breast cancer cells to proteins involved in crucial cellular processes, such as the transporter SLC20A1, the main provider of cells with inorganic phosphate, have been discovered (72). However, these vulnerabilities are not peculiar to claudin-low cells but are shared by basal-like and ER+ breast cancer cells. In another example, the immune receptor CD58 (also known as LFA3), which interacts with CD2 in T lymphocytes, appears to be enriched in claudin-low but also basal breast cancers (73). When co-expressed with stem cell marker ALDH1A3, CD58 may predict a poor prognosis in these sub-types, in contrast to luminal cancers, where CD58 is associated with improved prognosis (73).
The in vitro drug sensitivity screen of claudin-low cell lines revealed that four of the six cell lines were most sensitive to several pharmacologic modulators of the WNT pathway. Interestingly, representatives of both inhibitors (porcupine inhibitors) and activators (kinase GSK3 inhibitors) of the WNT pathway were included among the drugs with top sensitivities in the claudin-low breast cancer cell lines, suggesting that perturbations of the pathway in both directions (activation or suppression) has the potential to affect survival of cancer cells of this phenotype. This drug screen results are also consistent with the dependency of claudin-low cells to components of the cytoskeleton, which have close ties with the WNT pathway, as discussed above. Amplifications of genes encoding for the non-canonical WNT pathway regulating planar cell polarity are present in a subset of triple negative breast cancers, although these tend to be more of the basal like rather than the claudin-low subtype (74). Epigenetic modifications leading to altered expression of these genes and resulting proteins may be present without genetic lesions, such as mutations or copy number alterations in claudin-low cancers.
Drugs targeting the epigenetic apparatus, such as deacetylase inhibitors (HDACs) have an intuitive rational as candidates for treatment of cancers with the claudin-low phenotype associated with the EMT state. Consistently, a morphological screen for reversal of mesenchymal phenotype using mammary organoids revealed HDACs and bromodomain inhibitors as classes of drugs that were successful in morphology reversal (75). In vivo mice experiments with mesenchymal breast xenografts showed that the combination of the histone deacetylase inhibitor mocetinostat and the methyltransferase inhibitor azacytidine with carboplatin extended survival, while the epigenetic combination alone had minimal effect. The HDAC inhibitor givinostat (ITF2357) when combined with azacytidine suppressed the expression of the MYC oncogene and showed strong anti-proliferative synergism in a mouse model of lung cancer (76). In this model, the epigenetic combination also had a positive effect in antitumor immunity by reversing MYC-induced immune cell exhaustion. MYC-induced immune exhaustion reversal may be relevant for claudin-low cancers, given the high prevalence of MYC amplification. Another HDAC inhibitor, panobinostat suppressed the mesenchymal phenotype of MDA-MB-231 breast cancer cells in vitro and suppressed expression of ZEB1 and ZEB2 and increased expression of E-cadherin in these cells (77). Panobinostat also inhibited metastases of these cells in vivo. The drug had a suppressive effect on human claudin-low breast cancer stem cell compartment, as shown in mammosphere formation experiments (78). Despite differences in used models, HDACs appear to have more pronounced antitumor effects when combined with other epigenetic drugs or chemotherapy. Devising rational combinatorial treatments may be particularly important for the success of these therapies in claudin-low cancers, where epigenetic plasticity is an embedded feature enabling the EMT states. As suggested from the in vitro drug sensitivity data presented here, individual HDAC inhibitors in monotherapy are unlikely to be effective.
There are some limitations in the current analysis of claudin-low breast cancer cell line models. First, all claudin-low cell lines available for analysis are ER− and HER2− as no established claudin-low cell lines with positivity in these markers exist. Thus, results may not be representative for ER+ or HER2+ claudin-low breast cancers, which represent a sizable minority of the claudin-low phenotype (8). Another limitation is that the drug sensitivity screen includes various drugs as single drug exposures and no combinations are tested, limiting the value of this screen for evaluating combination therapies with potentially higher probabilities for clinical efficacy. Lastly, cell lines represent in vitro models that cannot recapitulate the in vivo tumor environment which includes a variety of other non-neoplastic cells with both pro-tumorigenic and anti-tumorigenic activities. The influence of the tumor microenvironment, including the immune system, although hinted, is not fully captured in in vitro experiments. Despite these limitations, cell line models of claudin-low breast cancers are valuable in confirming and expanding knowledge of the molecular landscape of these cancers and generating hypotheses for rational targeted treatments.
Conclusion
Claudin-low breast cancer cell lines constitute useful models for the study of the corresponding subtype of triple negative breast cancers. Claudin-low cell lines display features of an active EMT program which include, besides suppressed expression of claudins, up-regulation of transcription factors ZEB1, ZEB2, SNAIL1, SNAIL2 and TWIST through epigenetic modifications. Prevalent modifications include DNA hypermethylation and histone H3 hypermethylation at lysine 27 combined with hypomethylation at lysine 4. Genomic vulnerability screens disclosed the key role of cytoskeleton and regulators of cytoskeletal apparatus in claudin-low cells. A drug sensitivity screen suggested WNT pathway modulators, tyrosine kinase inhibitors and DNA damage response inhibitors as candidate drugs for development in claudin-low breast cancers. Targeted drug development will allow claudin-low breast cancer patients to benefit from these modalities, similarly to patients with other sub-sets of breast cancers, in which targeted drugs play a crucial role in the therapeutic armamentarium.
- Received June 25, 2023.
- Revision received August 16, 2023.
- Accepted August 28, 2023.
- Copyright © 2023, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).