Knockdown of CGRRF1 promotes the growth of breast cancer cell lines
Previous studies suggest a growth repressor function for CGRRF1; however, its role in breast cancer has not been determined. We first examined the expression of CGRRF1 by western blot analysis in a panel of breast cancer cell lines (Fig. 1a). Among the cell lines that we examined, estrogen receptor-positive MCF7 and T47D cells expressed a relatively high level of CGRRF1, whereas TNBC cell lines such as MDA-MB-468 and BT-549 and a HER2-positive cell line SKBR3 had relatively lower levels of CGRRF1. To study the effect of CGRRF1 on the growth of breast cancer cell lines, we used two shRNAs (shCGRRF1#1 and shCGRRF1#2) to generate stable CGRRF1-knockdown cell lines. In MCF7, both CGRRF1-knockdown cell lines grew faster than the control cells (shScr) (Fig. 1b). Similar results were obtained in SKBR3 and BT-549 cells (Fig. 1c, d). CGRRF1 depletion in BT-549 cells also promoted the ability of cells to form colonies (Fig. 1e and Additional file 1: Figure S1A). These data demonstrate a growth suppressor function for CGRRF1 in breast cancer cells.
Overexpression of CGRRF1 represses the growth of breast cancer cell lines
Since knockdown of CGRRF1 enhanced the proliferation of breast cancer cell lines, we generated stable CGRRF1-overexpressing cell lines to study whether overexpression of CGRRF1 could inhibit cell growth. To further understand whether the RING-finger domain of CGRRF1 is involved in its function in growth regulation, we also generated a mutant construct in which two cysteine residues were mutated to alanine (C274A/C277A) to disrupt the structure of the RING-finger domain (Fig. 2a). Indeed, overexpression of wild-type CGRRF1 suppressed the proliferation of BT-549, MDA-MB-231, and SKBR3 cell lines (Fig. 2b–d). However, mutant CGRRF1 partially lost its growth-inhibitory activity. We further performed colony formation assay in SKBR3 cells. Our result showed that overexpression of wild-type but not RING-finger mutant CGRRF1 inhibited their ability to form colonies, although the size of colonies of mutant CGRRF1-overexpressing cells was smaller than control cells (Fig. 2e and Additional file 1: Figure S1B).
Since the expression of mutant CGRRF1 is usually lower than the expression of wild-type CGRRF1 in our stable cell lines (Fig. 2b, c), we performed a cycloheximide chase assay to examine the protein stability of wild-type and mutant CGRRF1. As shown in Additional file 2: Figure S2A, mutant CGRRF1 is less stable than wild-type CGRRF1. Interestingly, we noticed that the serum concentration in the media affects the expression of CGRRF1. The level of exogenously expressed CGRRF1 increased upon serum starvation (Additional file 2: Figure S2B). While the expression of mutant CGRRF1 was less than that of wild-type CGRRF1 when these stable MDA-MB-231 cells were cultured under 5% serum-containing media, the expression of mutant CGRRF1 was in fact higher under serum starvation, consistent with a shorter half-life for mutant CGRRF1 (Additional file 2: Figure S2B). To further examine the growth suppressing effect of CGRRF1, we generated CGRRF1 doxycycline-inducible MDA-MB-231 cell lines in which the expression of both wild-type and mutant CGRRF1 could be induced by doxycycline to the same levels (Fig. 2f). Indeed, wild-type CGRRF1 repressed cell proliferation; however, we did not notice significant growth inhibition in cells that expressed mutant CGRRF1. These data suggest that the RING-finger domain is important for the growth-inhibitory activity of CGRRF1.
CGRRF1 inhibits breast cancer growth in vivo
To investigate the effect of CGRRF1 on breast tumor growth in vivo, we injected stable CGRRF1-overexpressing MDA-MB-231 cells into NOD scid IL2 receptor γ chain knockout (NSG) female mice. As shown in Fig. 3a, the tumor volumes and tumor weights were significantly decreased in the wild-type group (pLenti-CGRRF1) as compared to the control group (pLenti). Consistent with the data in Fig. 2c, the tumor sizes of the mutant (C274A/C277A) CGRRF1 group were between those in the control group and wild-type CGRRF1 group. The expression of wild-type and mutant CGRRF1 in these xenografts was verified by western blot analysis (Fig. 3b).
To identify the relevant target(s) of CGRRF1 in MDA-MB-231 breast cancer xenografts, the lysates of five tumors from each group were subjected to reverse phase protein array (RPPA) analysis which measured the expression of 236 proteins in triplicate from each tumor. The RPPA analysis identified 13 proteins which were expressed differentially between pLenti control and wild-type CGRRF1 at p < 0.05, and the changes are greater than 1.25-fold (a summary is presented as a heatmap in Additional file 3: Figure S3A). Among them, six were upregulated (14-3-3ζ/γ/ε, PDGFRβ, pS15-p53, pErk1/2 (T202/Y204), Integrin α4, and pSmad2(S465/467)), and seven were downregulated (epidermal growth factor receptor (EGFR), KLF4, p21, pBad (S136 and S155), Axl, pS1981-ATM, and IKKα) in wild-type CGRRF1-expressing xenografts. Their expressions in mutant CGRRF1-expressing xenografts fall between the pLenti control group and wild-type CGRRF1 group, but were more variable and not statistically different from the other groups. The RPPA quantitative analysis of EGFR expression in these three groups is shown in Fig. 3c. Consistent with the growth difference, the expression of Ki67 was also higher in the pLenti group and lowest in the wild-type CGRRF1 group (Additional file 3: Figure S3B), although the differences do not reach statistical significance due to small sample numbers. We also verified the RPPA result of EGFR expression using western blot analysis (Fig. 3d and Additional file 3: Figure S3C).
CGRRF1 has been demonstrated to be an E3 ubiquitin ligase of Evi and regulates its stability through ERAD [4]. Given the ER localization of CGRRF1, we suspected the proteins localized to the ER or plasma membrane would be more likely to be the direct targets of CGRRF1. Among the seven downregulated proteins in wild-type CGRRF1 xenografts, EGFR and Axl were the likely candidates. In the subsequent study, we concentrated on EGFR since we were not able to observe downregulation of Axl in cultured wild-type CGRRF1-overexpressing MDA-MB-231 cells.
CGRRF1 interacts with EGFR
To examine the interaction between CGRRF1 and EGFR, HEK293T cells were transiently transfected with a plasmid that expressed GFP or GFP-tagged EGFR, and then the cell lysates were subjected to GFP-pulldown/western blot analysis. The result showed that GFP-tagged EGFR interacted with endogenous CGRRF1 (Additional file 4: Figure S4). Conversely, when FLAG-tagged CGRRF1 was transiently transfected in BT-549 cells, it interacted with endogenous EGFR (Fig. 4a). We also performed reciprocal immunoprecipitation using EGFR and CGRRF1 antibodies in MDA-MB-468 cells. As shown in Fig. 4b, CGRRF1 interacted with EGFR in MDA-MB-468 cells at their endogenous protein levels.
EGFR is a receptor tyrosine kinase which upon ligand stimulation regulates cell proliferation, survival, differentiation, migration, and angiogenesis. EGF is one of the most common ligands to activate EGFR. To study whether the interaction between CGRRF1 and EGFR is regulated by ligands, MDA-MB-231 and BT-549 cells were transfected with FLAG-CGRRF1, starved for 24 h, and then stimulated with 100 ng/ml EGF for 15 min. Cells were then harvested for FLAG-pulldown/western blot analysis. Interestingly, EGF stimulation did not affect the interaction between CGRRF1 and EGFR (Fig. 4c). Taken together, these data demonstrate that CGRRF1 can interact with EGFR in a ligand-independent manner and suggest that EGFR might be a substrate for CGRRF1 E3 ligase activity.
CGRRF1 promotes K48-linked EGFR ubiquitination
Since we have identified EGFR as a CGRRF1-interacting protein, we next performed in vivo ubiquitination assay to investigate whether CGRRF1 regulates EGFR ubiquitination. Under normal cell growing condition, we found that knockdown of CGRRF1 decreased lysine 48 residue (K48) linkage-specific ubiquitination of EGFR (Fig. 5a). In contrast, overexpression of wild-type CGRRF1 promoted K48-linked ubiquitination of EGFR (Fig. 5b).
EGF-induced EGFR ubiquitination is important for controlling EGFR internalization, trafficking, and degradation. With EGF stimulation, we observed that knockdown of CGRRF1 diminished K48-linked EGFR ubiquitination in both starvation and EGF stimulation conditions compared to scramble control MDA-MB-231 cells (Fig. 5c), which is consistent with the interaction between CGRRF1 and EGFR in both starvation and EGF stimulation conditions (Fig. 4c). The effect of CGRRF1 knockdown on EGF-stimulated K48-linked EGFR ubiquitination was also seen in another cell line BT-549 (Additional file 5: Figure S5A). Conversely, overexpression of CGRRF1 enhanced EGFR K48-linked ubiquitination after EGF stimulation (Fig. 5d). These results indicate that CGRRF1 promotes K48 linkage-specific ubiquitination of EGFR.
In addition to K48-linked EGFR ubiquitination, we also examine the effect of CGRRF1 on K0- and K63-linked EGFR ubiquitination. In both normal growing and EGF stimulation conditions, knockdown of CGRRF1 inhibited K0-linked EGFR ubiquitination (Additional file 5: Figure S5B and S5C). This data is consistent with a role of CGRRF1 for EGFR ubiquitination. Surprisingly, we detected a smear pattern of K0-linked EGFR ubiquitination. HA-UbK0 functions as a monoubiquitin and as a ubiquitin chain terminator. The reason why we detected a smear pattern could be because that HA-UbK0 can be attached to the end of other polyubiquitination chains and/or there are multiple monoubiquitin conjugations. The CGRRF1 status did not affect K63-linked EGFR ubiquitination in normal growing condition (Additional file 5: Figure S5D and S5E). However, with EGF stimulation, overexpression of CGRRF1 enhanced K48-linked, but inhibited K63-linked, EGFR ubiquitination (Additional file 5: Figure S5F), supporting that CGRRF1 promotes K48 linkage-specific, but not K63 linkage-specific, ubiquitination of EGFR.
RING-domain mutant CGRRF1 interacts with EGFR but fails to promote K48-linked EGFR ubiquitination
To elucidate whether CGRRF1-induced K48-linked EGFR ubiquitination is due to the E3 ligase activity of CGRRF1, we first examined the interaction between EGFR and RING-domain mutant CGRRF1. From immunofluorescence staining, the subcellular localization patterns between wild-type and mutant CGRRF1 were essentially indistinguishable (Fig. 6a, b). Moreover, we observed colocalization of FLAG-tagged CGRRF1 (both wild-type and mutant) and Myc-tagged EGFR in both starvation and EGF stimulation conditions (Fig. 6a, b). We also performed co-immunoprecipitation assay to verify the interaction between mutant CGRRF1 and EGFR. In HEK293T cells, mutant CGRRF1 was able to interact with GFP-tagged EGFR as proficient as wild-type CGRRF1 (Additional file 6: Figure S6). A similar result was observed in MDA-MB-231 cells in which both FLAG-tagged wild-type and mutant CGRRF1 interacted with endogenous EGFR to a similar degree (Fig. 6c), suggesting that the mutations we made on the RING domain of CGRRF1 do not affect its ability to bind to EGFR.
We then performed in vivo ubiquitination assay to investigate whether our mutant CGRRF1 fails to ubiquitinate EGFR. Upon EGF stimulation, doxycycline-induced wild-type CGRRF1 was able to enhance K48-linked EGFR ubiquitination as compared to no doxycycline control. However, doxycycline-induced mutant CGRRF1 did not promote K48-linked ubiquitination of EGFR (Fig. 6d). Since these are different stable cell lines with endogenous CGRRF1, their EGF-induced EGFR ubiquitination in the absence of doxycycline may vary depending on many factors including the levels of endogenous E3 ligases and other undefined changes acquired during cell line establishment; therefore, we compared the samples without and with doxycycline within the same cell line. In this way, the only difference between both samples is the presence or absence of the exogenous CGRRF1 that was induced by doxycycline. The data from four independent experiments are very consistent and show that only induction of wild-type CGRRF1 caused an increase in EGFR ubiquitination (Fig. 6d, bottom panel). Together, these data support that CGRRF1 is an E3 ubiquitin ligase for EGFR.
Knockout of CGRRF1 increases EGFR protein level
K48-linked polyubiquitination has been known for regulating proteasome-mediated protein degradation. To determine a role for CGRRF1 in the regulation of EGFR, we generated CGRRF1-knockout clones in MCF7 cells and MDA-MB-231 cells using three different sgRNAs against CGRRF1. We also generated multiple sgVector control clones for each cell line at the same time for a more robust comparison with CGRRF1-knockout clones. Indeed, CGRRF1-knockout MCF7 cells (sgCGRRF1) had higher EGFR protein levels compared to the control cell lines (sgVector) (Fig. 7a). Although not as obvious as in MCF7, knockout of CGRRF1 in MDA-MB-231 also increased EGFR expression (Fig. 7b). We then performed growth assay in these knockout cells. Our results showed that knockout of CGRRF1 increased cell proliferation (Additional file 7: Figure S7A and S7B), which is consistent with Fig. 1. We also rescued CGRRF1 expression in one of the CGRRF1-knockout MCF7 cell lines and then performed proliferation assay. As shown in Fig. 7c, CGRRF1-rescued cells grew slower than CGRRF1-knockout cells. Besides, we noticed that the EGFR expression is decreased in CGRRF1-rescued cells (Fig. 7c, bottom panel). A similar result was observed in wild-type, but not mutant, CGRRF1-overexpressing MCF7 cell lines (Fig. 7d). We also checked the EGFR level in CGRRF1-knockdown cells. Just like the knockout cell lines, knockdown of CGRRF1 increased EGFR expression (Fig. 7d, bottom panel).
To further understand whether the increased EGFR in CGRRF1-knockout cell lines is due to stabilization of EGFR protein, we treated CGRRF1-knockout cells with a proteasome inhibitor, MG132, and compared the expression of EGFR. As shown in Fig. 7e, control cell lines had more induction of EGFR after MG132 treatment than CGRRF1-knockout cell lines. These data support that lower EGFR expression in control cell lines is due to CGRRF1-mediated EGFR proteasomal degradation. In the TCGA database, we also found that there is a negative correlation between CGRRF1 mRNA and EGFR protein levels (Fig. 7f). Although the sample number is small, there is a very significant negative correlation (r = −0.56) between CGRRF1 and EGFR protein levels in the breast cancer cell lines (Additional file 8: Figure S8). This further supports the idea that CGRRF1 regulates EGFR expression.
To examine whether CGRRF1 inhibits cell growth at least in part through EGFR, we established EGFR-overexpressing cell lines in wild-type CGRRF1 doxycycline-inducible MDA-MB-231 cells (Fig. 7g). As shown in Fig. 7h, the control cells (pcDNA6) grew slower after inducing the expression of CGRRF1 by doxycycline. However, the growth rate of the cells overexpressing EGFR was not affected by the induction of CGRRF1. This data suggests that CGRRF1 inhibits cell growth at least in part by decreasing the expression of EGFR.
CGRRF1 regulates AKT phosphorylation and EGFR nuclear translocation
EGFR contains a tyrosine kinase domain. Upon ligand stimulation, EGFR can form homo- or hetero-dimer and auto-phosphorylate itself. This will lead to the activation of RAS/MAPK, PI3K/AKT, PLC-PKC, and Jak/STAT pathways. To determine whether CGRRF1 regulates EGFR signaling, we checked the downstream signaling pathways of EGFR. Indeed, knockout of CGRRF1 increased AKT phosphorylation after EGF stimulation in both MCF7 cells (Fig. 8a) and MDA-MB-231 cells (Fig. 8b). Interestingly, while the enhanced AKT phosphorylation in CGRRF1-knockout MCF7 cells is at least in part due to higher EGFR protein levels in these cells, the EGFR levels were not significantly changed in CGRRF1-knockout MDA-MB-231 cells under this experiment. We also examine AKT phosphorylation in CGRRF1-overexpressing BT-549 cell lines. Overexpression of wild-type but not mutant CGRRF1 inhibited AKT phosphorylation (Fig. 8c). These data suggest that the ubiquitination of EGFR by CGRRF1 might also regulate EGFR signaling function besides its protein level.
EGFR also contains a nuclear localization sequence (NLS), which plays an important role in nuclear translocation [9]. Cell surface EGFR is translocated to the nucleus through COPI-mediated retrograde transport from the Golgi to the ER, then shuttled from the outer nuclear membrane (ONM) to the inner nuclear membrane (INM), and finally released from the INM to the nucleoplasm [10]. Since CGRRF1 has been reported to localize to the ER, CGRRF1 might be involved in EGFR nuclear translocation. To address this possibility, we made use of the TNBC cell lines BT-549 and MDA-MB-231, in which the total EGFR levels were only modestly affected by CGRRF1 expression or knockdown (Fig. 8d, right panel, and Fig. 8f, right panel). Indeed, the fraction of nuclear EGFR was significantly lower in wild-type CGRRF1-overexpressing BT-549 cells than in the control and mutant CGRRF1-overexpressing cells (Fig. 8d). In this experiment, cells were cultured in medium containing 10% serum; therefore, the expression level of mutant CGRRF1 was less than that of the wild type. To overcome this problem and to exclude the possibility that the lack of effect by mutant CGRRF1 was due to its lower expression, we cultured CGRRF1-overexpressing MDA-MB-231 cells in low serum condition (2% serum) for a few days to increase the expression of mutant CGRRF1, as we observed in Additional file 2: Figure S2B. Indeed, the nuclear EGFR was significantly decreased in wild-type but not in mutant CGRRF1-overexpressing MDA-MB-231 cell lines, despite mutant CGRRF1 being expressed at a higher level than the wild type (Fig. 8e). On the other hand, we detected higher nuclear EGFR fraction in CGRRF1-knockdown MDA-MB-231 cells (Fig. 8f). Correspondingly, the targets of nuclear EGFR, c-Myc [11] and Aurora A [12], were expressed at higher levels in CGRRF1-knockdown MDA-MB-231 cells (Fig. 8f, right panel). These results suggest that in addition to regulating EGFR stability, CGRRF1 might regulate the nuclear translocation of EGFR through its ubiquitin E3 ligase activity.
Low CGRRF1 expression in breast cancer is associated with poor patient survival
Given the growth suppressor role of CGRRF1 in breast cancer cell lines and the xenograft model shown above, we investigated publicly available breast cancer datasets to correlate CGRRF1 expression with patient survival. Kaplan-Meier survival analysis from the van de Vijver database (stages I and II breast cancer) [13] shows that breast cancer patients with low CGRRF1 expression in their breast tumors had poor survival (Fig. 9a). CGRRF1 expression is lower in estrogen receptor-negative breast cancers than in estrogen receptor-positive cancers (Oncomine). To avoid this confounding factor, we analyzed only estrogen receptor-positive breast cancers in the van de Vijver cohort. As shown in Fig. 8b, the lower the CGRRF1 levels are, the shorter the patient survived. The number of estrogen receptor-negative patients (n = 69) in the van de Vijver cohort is too small to perform analysis. We also found similar results in patients with kidney carcinoma or lung adenocarcinoma (Additional file 9: Figure S9A), suggesting a general role for CGRRF1 in suppressing tumor growth.
There are five major subtypes of breast cancer. We also checked the expression of CGRRF1 in each subtype. In the TCGA breast cancer cohort, both HER2-positive and basal-like breast cancer patients had significantly lower expression of CGRRF1 compared to other subtypes (Fig. 9c). Basal-like breast cancer patients have the worst survival, so the association of low CGRRF1 with poor survival could be in part due to the fact that some cancers with low CGRRF1 expression are basal-like subtype. To further investigate whether CGRRF1 expression bears prognostic significance within the same subtype of breast cancer, we used the KM Plotter server to perform analysis of each subtype from a large number of datasets. We found that lower expression of CGRRF1 is also associated with a shorter patient overall survival in the Luminal A subtype and HER2-positive subtype of breast cancer (Additional file 9: Figure S9B). We do not see this association in the basal-like cohort, probably because most basal-like breast cancers already express low levels of CGRRF1. The association between low CGRRF1 expression and shorter patient survival is also seen in the Luminal A subtype in the METRBRIC dataset (Fig. 8d). Thus, analyses from multiple datasets show an association between low CGRRF1 expression and poor patient survival.
The expression of CGRRF1 is often downregulated in breast carcinoma due to promoter hypermethylation
Comparing the transcript levels of CGRRF1 between matched normal breast tissues and breast tumors in the TCGA breast cancer cohort, we noticed that CGRRF1 expression is significantly lower in tumor tissues (Fig. 10a and Additional file 10: Figure S10A). Aberrant epigenetic modifications, such as changes in DNA methylation and histone modification, are associated with the development and progression of cancer. To understand whether the decrease in CGRRF1 in breast carcinoma is caused by alterations in epigenetic modifications, we examined the methylation status of the promoter of CGRRF1 in the TCGA breast cancer cohort. Indeed, there is a significant increase in CGRRF1 promoter methylation in breast tumors compared to that in normal breast tissues (Fig. 10b and Additional file 10: Figure S10B). We further analyzed the correlation between CGRRF1 transcript levels and its promoter methylation status in the TCGA breast cancer cohort. In normal breast tissues, there is no correlation between the expression of CGRRF1 and its promoter methylation (r = −0.07). However, there is a negative correlation in the matched 76 breast tumor samples (r = −0.32) (Fig. 10c) and in 888 TCGA breast cancer tissues (r = −0.325) (Fig. 10d). In addition, similar negative correlation between CGRRF1 expression and promoter methylation is found in other types of cancer (Additional file 10: Figure S10C). These data suggested the possibility that changes in CGRRF1 promoter methylation regulate the expression of CGRRF1 and are involved in the development of breast cancer.
To test the epigenetic mechanism for downregulation of CGRRF1 in breast cancer, we treated a panel of breast cancer cell lines with a demethylation agent, 5-azactidine. Indeed, 5-azacitidine treatment increased the CGRRF1 transcripts (Fig. 10e) and protein expression (Fig. 10f) in breast cancer cells. We also treated these cells with a histone deacetylase (HDAC) inhibitor, panobinostat. CGRRF1 expression was induced at both the mRNA level (Fig. 10g) and protein level (Fig. 10h) after the treatment. On the other hand, EGFR expression was decreased after the treatment (Additional file 11: Figure S11). Together, these results indicated that epigenetic alterations in the CGRRF1 promoter lead to its downregulation in breast cancer.