|Home | About | Journals | Submit | Contact Us | Français|
The lysyl oxidase (LOX) gene reverted Ras transformation of NIH 3T3 fibroblasts and tumor formation by gastric cancer cells, which frequently carry mutant RAS genes. The secreted lysyl oxidase pro-enzyme is processed to a propeptide (LOX-PP) and a functional enzyme (LOX). Unexpectedly, the tumor suppressor activity mapped to the LOX-PP domain, which inhibited tumor formation and invasive phenotype of NF639 breast cancer cells driven by Her-2/neu, which signals via Ras. A single nucleotide polymorphism G473A (rs1800449) resulting in an Arg158Gln substitution in a highly conserved region within LOX-PP occurs with an average 473A allele carrier frequency of 24.6% in the HapMap database, but was present in many breast cancer cell lines examined. Here we show that the Arg-to-Gln substitution profoundly impairs the ability of LOX-PP to inhibit invasive phenotype and tumor formation by NF639 cells in a xenograft model. LOX-PP Gln displayed attenuated ability to oppose the effects of LOX, which promoted a more invasive phenotype. In a case-control study of African-American women, a potential association of the Gln-encoding A allele was seen with increased risk of estrogen receptor α (ER) negative invasive breast cancer in African-American women. Consistently, LOX gene expression was higher in ER negative vs positive primary breast cancers, and LOX-PP Gln was unable to inhibit invasion by ER negative cell lines. Thus, these findings identify for the first time genetic polymorphism as a mechanism of impaired tumor suppressor function of LOX-PP, and suggest it may play an etiological role in ER negative breast cancer.
The copper amine enzyme lysyl oxidase is required for the maturation of collagen and elastin precursors in the biosynthesis of a functional extracellular matrix (1, 2). The LOX gene inhibited the transforming activity of the Ras oncogene in NIH 3T3 fibroblasts and was named the “ras recision” gene (rrg) (3, 4). Reduced LOX expression has been reported in many carcinomas (5-10). Ectopic LOX gene expression in gastric cancer cells reduced tumor formation in nude mice (7). LOX gene expression in Ras-transformed NIH 3T3 fibroblasts inhibited the activities of the Akt and Erk1/2 kinases and NF-κB transcription factors (11). Lysyl oxidase is secreted as a 50 kDa inactive pro-enzyme (Pro-LOX), which is processed by proteolytic cleavage to a functional 32 kDa enzyme (LOX) and an 18 kDa propeptide (LOX-PP) (12). The LOX-PP domain was identified as the inhibitor of transformed phenotype of Ras-NIH 3T3 fibroblasts (13), lung and pancreatic cancer cells with mutant RAS genes (10), and of NF639 breast cancer cells, driven by Her-2/neu, which signals via Ras (14). Specifically, LOX-PP decreased Her-2/neu-mediated signaling and mesenchymal phenotype in vitro, and NF639-derived tumor xenograft formation in a nude mouse model (14). Furthermore, LOX-PP attenuated fibronectin-stimulated integrin signaling and migration in breast cancer cells (15). Thus, LOX-PP can inhibit the invasive phenotype of carcinomas.
High-penetrance germline mutations account for less than 25% of the familial risk of breast cancer. It has been hypothesized that the remaining susceptibility to breast cancer is polygenic in nature, involving a relatively large number of germline, genetic variations with low to moderate penetrance (16). We observed a single nucleotide polymorphism (SNP) accompanied by a non-synonymous amino acid substitution G473A/Arg158Gln (rs1800449) in a highly conserved region within LOX-PP (Fig. 1A) in 6 of 9 breast cancer cell lines examined (Supplemental Fig. 1). The frequency of the 473A allele of the LOX gene in European, Asian, Sub-Saharan African and African-American populations in the International HapMap Project averaged 24.6%. Here, the Gln variant, encoded by this minor A allele was shown to display impaired tumor suppressor ability compared with LOX-PP wild type (WT) encoded by the major G allele. As LOX-PP WT appeared to prevent a more invasive phenotype, the association of the LOX rs1800449 polymorphism with breast cancer risk was examined in a study of African-American women who tend to have more aggressive breast cancer and higher mortality rates than Caucasian women (17, 18). In a nested case-control study within a cohort of participants of the Black Women's Health Study (BWHS), the LOX G473A polymorphism appeared associated with increased risk of ER negative invasive breast cancer. Thus, our findings identify genetic polymorphism as a mechanism of impaired tumor suppressor function of LOX-PP and suggest that further analysis of the potential association of the LOX rs1800449 polymorphism with increased risk of ER negative breast cancer is warranted.
V5/His-tagged wild type murine Pro-LOX and LOX-PP vectors were described (14). The amino acid Arg152 corresponding to the rs1800449 polymorphism was mutated to Gln152 in the murine LOX-PP and Pro-LOX constructs. These DNA fragments were cloned into the retroviral vector pC4bsrR(TO) containing a doxycycline (Dox) inducible promoter and vector pCXbsr under the control of constitutive cytomegalovirus (CMV) promoter. Wild type human LOX-PP was amplified by PCR and cloned into pcDNA4-V5/His vector. Human LOX-PP Gln was generated by site-directed mutagenesis.
Mouse NF639 cells and human breast cancer and epithelial cells were cultured as described previously (14, 19). Retrovirus stocks were made as described (14). Stable NF639 infectants carrying the Dox inducible constructs of empty vector (EV), Pro-LOX WT, LOX-PP WT, Pro-LOX Gln and LOX-PP Gln were generated by retroviral infection (14). Recombinant rat LOX-PP WT and LOX-PP Gln proteins were expressed and purified as published (20).
Whole cell extracts (WCEs) were prepared and subjected to immunoblotting (14). Antibodies were as follows: phospho-Akt (Ser473), Akt, phospho-Erk1/2 and Erk1/2, Cell Signaling; vimentin, NeoMarker; fibronectin and E-cadherin, BD Transduction Laboratories; cyclin D1, Santa Cruz Biotechnology; β-actin, Sigma; V5 epitope, Invitrogen. The results from a minimum of three independent experiments were subjected to densitometry and, after normalizing to β-actin, the mean values relative to control EV cells (set to 1.0) ± SD given. The detection of recombinant proteins in conditioned medium and WCEs were performed as described (14).
Matrigel outgrowth and invasion assays assays were performed as published (14, 19). Cells that had migrated to the lower side of the filter were quantified as described (15). All invasion assays were performed three times in triplicate.
Cells (6 × 104 cells/well) were plated in 6-well plates. After overnight incubation, cells were co-transfected with 2 μg of the appropriate cDNA in pCXbsr vector and 1 μg of a green fluorescent protein (GFP) expression vector in DMEM-0.5% FBS using Fugene 6 reagent. The number of GFP positive cells was determined 24 h and 72 h post-transfection by fluorescence microscopy, as described in the legend.
NCrnu/nu nude mice, purchased from Taconic Laboratories at 7 to 9 weeks of age, received 2% sucrose plus 2 mg/mL Dox solution in lieu of water 3 days before tumor cell inoculation. Dox inducible stable cells were pre-treated with 2 μg/mL Dox for 24 h and then 4 × 106 cells injected subcutaneously in both flanks of the same mouse (n = 6) with NF639-EV, left vs -LOX-PP WT, right or NF639-EV, left vs -LOX-PP Gln, right. Tumor size was measured, as described (14). Mice were sacrificed 30 days after inoculation and tumors dissected and weighed. For statistical analysis, tumor volumes and tumor weights of NF639-LOX-PP WT vs -EV cells and NF639-LOX-PP Gln vs -EV cells were compared using paired Student's t test.
The BWHS is an ongoing prospective follow-up study of cancer and others illnesses among black women in the United States, involving 59,000 African-American women 21-69 years of age from across the U.S at baseline in 1995. In follow-up questionnaires, sent every two years, participants report the occurrence of incident cancer and information on risk factors such as age at first birth and family history of breast cancer. Information has been abstracted from medical records on tumor characteristics such as ER and HER2 status. To date, among all self-reported cases for which records have been obtained, 99% have been confirmed as breast cancer.
DNA was isolated from mouthwash/saliva samples, obtained from participants (21), using the QIAAMP DNA Mini Kit (Qiagen) (21) and quality tested by a real-time PCR SNP-genotyping assay for the Paraoxonase gene (rs854565). A >97% call rate was achieved and the resulting data met Hardy-Weinberg expectations for genotype distribution. Whole genome amplification of the DNA samples was performed using the Qiagen REPLI-g Kit. Study protocols were approved by the Institutional Review Board of Boston University.
For the present analyses, 336 cases of incident invasive breast cancer identified during follow-up were matched on age and geographic region (Northeast, South, Midwest, West) with 465 women unaffected by breast cancer (controls), selected at random from among unaffected women. Because of our interest in ER negative breast cancer, which is more common in African-American women than in other ethnic groups, we included all ER negative cases and a sample of ER positive cases.
Using multivariable logistic regression (22), we calculated odds ratio (OR) and 95% confidence interval (CI) for the heterozygous (GA) and homozygous (AA) genotypes in one model, relative to the wildtype (GG) genotype, adjusted for age, geographic region, family history of breast cancer, age at menarche, body mass index at age 18, parity, age at first birth, oral contraceptive use, menopausal female hormone use, and years of education. There was little evidence of confounding, as OR estimates from these multivariable models were closely similar to estimates from models adjusting for age only. We carried out a test for trend by the inclusion of a term in the logistic regression, in which the number of variant alleles was entered as 0, 1, or 2 (23). We combined the GA and AA genotypes in a second model, assuming a dominant genetic model comparing the presence of either one or two variant alleles (GA or AA) to none (GG). Analyses were conducted separately for all invasive, ER positive, ER negative, and HER2 positive breast cancers.
We first asked whether the G473A/Arg158Gln polymorphism is present in human breast cancer cells. The results indicated that the untransformed MCF-10A line and breast cancer lines MDA-MB-231, MCF7, and ZR75 were homozygous for the major G allele. Five lines [BT474, BT549, SKBR3, T47D, and BT20] were heterozygous and one line [Hs578T] was homozygous for the minor A allele (Supplemental Fig. 1). Thus, of the 9 human breast cancer cell lines examined, 6 carry the minor 473A allele, substantially above the 24.6% seen in the HapMap database.
This finding led us to question whether the Gln variant has altered ability to impair Ras signaling in breast cancer cells. NF639 breast cancer cells express low levels of ER and E-cadherin, and have a highly invasive phenotype (19). Given that the substitution falls within the carboxy-terminal region of the propeptide domain which is required for Pro-LOX secretion (14, 24), we first compared the expression, secretion and proteolytic cleavage of Pro-LOX WT and Pro-LOX Gln in stable populations of NF639 cells expressing Dox inducible DNAs. No substantial differences were seen in expression or secretion of Pro-LOX Gln relative to Pro-LOX WT (Fig. 1B). Furthermore, a comparable or even higher level of LOX-PP Gln vs LOX-PP WT was detected in the culture medium (Fig. 1C). Similar data have been obtained in Hs578T human breast cancer cells (not shown). Thus, synthesis and secretion are not reduced by the Gln variant.
The ability of LOX-PP Gln vs LOX-PP WT to inhibit Her-2/neu signaling was then compared. The LOX-PP Gln variant appeared unable to suppress phosphorylation of Akt in contrast to LOX-PP WT, while it retained partial ability to suppress Erk1/2 activation (Fig. 2A). Quantification of data from this and two duplicate experiments, given in the legend for Fig. 2, confirmed these observations. Akt activation in breast cancer cells with high levels of Her-2/neu is required for D-type cyclin expression (25), which promotes their proliferation. Previously, we demonstrated that LOX-PP reduced cyclin D1 (14). The Gln variant showed a markedly reduced ability to inhibit cyclin D1 expression (see Fig. 2A legend). Next, we compared the effects of LOX-PP Gln vs LOX-PP WT on NF639 cell numbers under serum deprivation condition using a GFP expression vector as a marker of transfected cells. Inhibitory effects of LOX-PP WT were much more robust than those of the Gln variant at both 24 h and 72 h post-transfection (Fig. 2B). Interestingly, we failed to detect the induction of PARP cleavage by LOX-PP WT (not shown), suggesting it did not induce substantial levels of apoptosis of NF639 cells. Together, these findings suggest that LOX-PP reduces proliferation of NF639 cells under serum deprivation conditions, whereas LOX-PP Gln has lost this ability.
Low expression of E-cadherin is frequently associated with a more migratory, invasive phenotype, and the induction of its levels can lead to enhanced cell-cell contacts. The undetectable levels of E-cadherin normally found in the highly invasive NF639 cells were robustly induced by LOX-PP WT, whereas a much more modest effect was seen with LOX-PP Gln (Fig. 2A). Quantification of three independent experiments indicated that E-cadherin induction by LOX-PP WT was 4.8 ± 1.9-fold higher than LOX-PP Gln. Consistently, the Gln variant also displayed a profoundly reduced ability to inhibit formation of branching structures in Matrigel (Fig. 2C), as well as invasion through Matrigel (Fig. 2D) compared to LOX-PP WT. Thus, the LOX-PP Gln variant has a substantially impaired ability to inhibit transformed phenotype in these Her-2 high/ER low breast cancer cells.
Previously, we demonstrated that LOX-PP WT expression, driven constitutively by the CMV promoter, reduced the average weight of tumors formed by NF639 cells in nude mice by ~60% (14). We next compared the effects of the two genotypes on xenograft tumor formation by NF639 cells using the inducible vector system. Tumors resulting from NF639-LOX-PP WT cells began to grow at a noticeably slower rate than NF639-EV cells by day 17, and a significant difference was reached on days 27 and 30 (Fig. 3A). Consistently, the average tumor weight for LOX-PP WT cell xenografts was 52% of those for the EV group on day 30 (P = 0.03; Fig. 3B). Expression of LOX-PP Gln in NF639 cells had no significant effect on either tumor growth rate or tumor weight (Fig. 3C and 3D). The average tumor weight for NF639 LOX-PP Gln xenografts was 126% of those for the EV group (P = 0.38; Fig. 3D). Thus, the LOX-PP Gln variant has substantially reduced tumor suppressor activity.
While LOX-PP suppresses Ras-mediated transformation, the LOX enzyme has recently been implicated in promoting tumor progression (14, 15, 26-28). The effects of the two genotypes within the full precursor protein (Pro-LOX Gln vs Pro-LOX WT) on signaling and transformed phenotype were next compared. Previously, we observed Pro-LOX WT attenuated activation of Akt and Erk1/2 and had some modest effects reducing cyclin D1 while leaving vimentin unaffected (14). Expression of the Pro-LOX Gln variant increased Erk1/2 phosphorylation, while a small decrease was noted with Pro-LOX WT (Fig. 4A). Furthermore, the Pro-LOX Gln variant was unable to substantially reduce phosphorylation of Akt in contrast to the Pro-LOX WT (see Fig. 4A legend for quantitation). A slight decrease in cyclin D1 was seen with Pro-LOX WT, whereas no substantial changes were seen in vimentin or fibronectin (Fig. 4A). Interestingly, Pro-LOX Gln caused a slight induction in the levels of vimentin, as well as of fibronectin, while having little effect on cyclin D1 (Fig. 4A). As E-cadherin expression was not induced by Pro-LOX (14), the effects of the Pro-LOX Gln variant on this gene were not pursued. Notably, while Pro-LOX WT decreased branching structure formation and invasion through Matrigel, Pro-LOX Gln robustly increased branching formation (Fig. 4B) and Matrigel invasion (Fig. 4C). These findings are consistent with a combined effect of loss of functional tumor suppressor activity of LOX-PP Gln, and maintenance of the pro-tumorigenic role of LOX (see below).
African-American women tend to have more aggressive breast cancer than Caucasian women, such as ER negative tumors (17, 18). We next tested the hypothesis that the LOX rs1800449 SNP is associated with risk of breast cancer in participants in the BWHS. DNA from 311 incident cases of invasive breast cancer and 446 controls, matched to the cases on age and geographic region, were assayed for the rs1800449 SNP in the LOX-PP domain. No association of LOX rs1800449 polymorphism with overall invasive breast cancer risk was seen (OR, 1.14; 95% CI, 0.82-1.58) (Table 1). The OR for the homozygous variant (AA) relative to the homozygous wild type (GG) was 1.99 (95% CI, 0.86-4.61), and this increase was largely accounted for by ER negative breast cancer cases. When we analyzed ER negative cases specifically, there was a significant dose-dependent association (Ptrend = 0.045) of the A allele with increased risk of ER negative breast cancer in African-American women, with the OR increasing from 1.40 (95% CI, 0.89-2.19) for heterozygous GA genotype to 2.34 (95% CI, 0.81-6.74) for homozygous AA genotype (Table 1). In the dominant model, the OR was 1.48 (95% CI, 0.96-2.28). Associations of LOX rs1800449 with risk of ER positive breast cancer were weaker than those with ER negative breast cancer with no significant trend (Ptrend = 0.44). Also, there was no significant association of LOX rs1800449 with risk of HER2 positive breast cancer, but there were only 56 HER2 positive cases (Table 1). A small proportion (4%) of the women had parents who were born outside of the United States. After controlling for parental birthplace, the results were unchanged. Thus, a dose-dependent association of the Gln-encoding A allele was seen with increased risk of ER negative breast cancer in African-American women in this hypothesis-generating initial study.
Based on the preliminary findings that LOX rs1800449 polymorphism has a stronger association with risk of ER negative than positive breast cancers, we hypothesized that LOX gene expression is higher in these ER negative breast cancers. Microarray gene expression datasets available at Oncomine.org were analyzed. Figure 5A shows the box plots of three datasets [Van de Vijver (29), Sotiriou (30), and Hess (31)]. Levels of LOX mRNA were significantly higher in ER negative vs ER positive breast cancers in all 3 studies (see Fig. 5A legend), consistent with the findings of Erler et al. (26). Similar data were seen with 8 other datasets at Oncomine.org (Supplemental Fig. 2). Thus, ER negative breast cancers express significantly higher LOX levels than ER positive ones.
ER negative breast cancers have a highly invasive phenotype. The functional consequences of Arg-to-Gln substitution on the ability of ER negative human breast cancer cell lines MDA-MB-231 and Hs578T to invade through Matrigel were next examined (Fig. 5B-D). Consistent with the assays of invasive phenotype of NF639 cells above, purified LOX-PP WT protein significantly reduced invasion of MDA-MB-231 and Hs578T cells at 55.5 nM and 111 nM, whereas the LOX-PP Gln had no significant effect at either dose (Fig. 5B and 5C). Furthermore, invasion of Hs578T cells was significantly reduced by ectopic expression of LOX-PP WT, but not LOX-PP Gln (Fig. 5D). Together, these data indicate that compared to LOX-PP WT, LOX-PP Gln has impaired ability to suppress invasive phenotype of ER negative cells.
Here for the first time, the minor A allele of the rs1800449 SNP within the LOX gene region encoding the propeptide domain is shown to profoundly impair its ability to inhibit growth, reverse invasive phenotype of breast cancer cells in culture and function as a tumor suppressor in a mouse xenograft model. Importantly, a dose-dependent association of the Gln-encoding A allele with increased risk of ER negative breast cancer was suggested in a nested case-control study within a cohort of participants in the BWHS. While our case-control study has a relatively small sample size, the association between LOX Gln-encoding allele with ER negative breast cancer risk in African-American women is strongly supported by functional evidence and the microarray analysis. The Arg-to-Gln substitution in LOX-PP impaired its ability to reduce transformed phenotype or invasive properties of ER negative or low breast cancer cells or to effectively oppose the pro-tumorigenic effects of LOX. ER negative tumors expressed significantly higher levels of LOX mRNA than ER positive ones further supporting the findings as a polymorphism within a gene tends to have stronger association with risk of the disease displaying higher levels of its expression. Although we can not rule out selection bias due to lack of control for population stratification, because cases and controls came from the same cohort of women being followed, it is unlikely that population stratification would have influenced the results. Overall, this first epidemiologic study of LOX 473A allele and breast cancer suggests that this variant is a potential risk allele for ER negative breast cancer in African-American women, but further studies are needed to confirm this conclusion.
Tumor suppressor genes are frequently subjected to genetic or epigenetic variations that reduce their activity (32, 33). The human LOX gene maps to chromosome 5q23.2 (34, 35), which is a frequent site of loss of heterozygosity (LOH) in many cancers, e.g., colorectal and gastric cancers (34, 36). In breast cancers, LOH within chromosomal region 5q21-32 marked “basal-like” breast cancers (37). Of note, LOH has been shown to coincide with promoter methylation resulting in a profound suppression of gene expression (9, 38). Consistently, LOX promoter methylation has been reported in several cancers (7, 9), e.g., LOX was one of the most frequently and specifically methylated genes in non-small cell lung cancers (95%) and breast cancer. Here, polymorphic genetic variation is identified as a new mechanism whereby the tumor suppressor activity of LOX-PP can be attenuated.
The biology of lysyl oxidase in tumor progression is complex. Studies from several laboratories have shown that the LOX gene and specifically LOX-PP functions as a Ras tumor suppressor (3, 4, 10, 11, 13-15, 39). Notably, we observed that LOX-PP reverts the highly mesenchymal breast cancer cells to a more epithelial phenotype. Additional work is needed to determine whether LOX-PP can functionally activate a process of mesenchymal to epithelial transition (MET). In contrast, the LOX enzyme was found to facilitate a more migratory and invasive phenotype during breast cancer progression (14, 15, 26-28). Here, LOX-PP Gln was observed to have impaired ability to inhibit transformed phenotype of breast cancer cells, whereas Pro-LOX Gln induced Erk activity, levels of fibronectin and vimentin and more invasive properties in Matrigel. These observations suggest that the facilitating effect of the Pro-LOX Gln variant on invasive phenotype is due to a profound reduction in the tumor suppressor function of LOX-PP, while the ability of the LOX enzyme to promote transformed phenotype is retained. The functional consequences of the Arg-to-Gln substitution provide an important biological rationale for the potential association of LOX Gln-encoding allele with increased risk of ER negative breast cancer observed in our hypothesis-raising study of African-American women.
An estimated 10-25% of breast cancer cases cluster in families and are believed to have a genetic component or basis (40). High-penetrance germline mutations have been found in the BRCA1, BRCA2, TP53, PTEN, and CDH1 genes. These mutations account for approximately 25% of the familial risk (41). The remaining factors fall into two categories: very rare, moderate-penetrance and common low-penetrance breast cancer susceptibility genes. Notably, the very rare, moderate-penetrance susceptibility genes ATM, BRIP1, CHEK2 and PALB2 interact or intersect with the BRCA1 and BRCA2 protein DNA repair pathways, suggesting their involvement in common mechanisms. Genome-wide association studies have identified a small number of common low-penetrance susceptibility genes, including MAP3K1 (42), FGFR2 (42, 43), and TWIST1 (44). Interestingly, the rs1800449 polymorphism, which would fit in the low-penetrance, high abundance category, can be connected via involvement in Ras signaling and regulation of NF-κB with the previously identified MAP3K1, PTEN, and TWIST1 genes.
Breast cancer is a highly heterogeneous disease. Accumulating epidemiologic data suggest that different subtypes of breast cancers have different risk factor profiles (45, 46). Associations between breast cancer risk and common genetic variants are often modified by tumor characteristics such as ER status (47). Here, LOX rs1800449 polymorphism has a stronger association with the risk of ER negative than ER positive breast cancer, supporting the recent hypothesis that ER negative and positive tumors arise from different etiologic pathways, rather than representing different stages of tumor progression (47, 48).
We thank Martin Minns and Danielle N. Stephens for assistance with preparing the LOX-PP Gln vector, and recombinant protein, respectively. These studies were supported by grants from the NIH CA082742, CA106468, CA058420, CA098663, and DE14066, and the DOD BC050957.