Human Rhomboid Family-1 Suppresses Oxygen-independent Degradation of Hypoxia-inducible Factor-1a in Breast Cancer

Intermittent oxygen deficiency in cancers promotes prolonged inflammation, continuous angiogenesis, and increased drug resistance. Hypoxia-inducible factor-1 (HIF1) has a pivotal role in the regulation of cellular responses to oxygen deficiency. The a-subunit of HIF1 (HIF1a) is degraded in normoxia but stabilized in hypoxia. However, the molecular mechanism that controls oxygen-independent degradation of HIF1a has remained elusive. Human rhomboid family-1 (RHBDF1) is a member of a large family of nonprotease rhomboids whose function is basically unknown. We report here that RHBDF1 expression in breast cancer is highly elevated and is strongly correlated with escalated disease progression, metastasis, poor prognosis, and poor response to chemotherapy. We show that RHBDF1 interaction with the receptor of activated protein-C kinase-1 (RACK1) in breast cancer cells prevents RACK1-assisted, oxygen-independent HIF1a degradation. In addition, we show that the HIF1a-stabilizing activity of RHBDF1 diminishes when the phosphorylation of a tyrosine residue on the RHBDF1 molecule is inhibited. These findings are consistent with the view that RHBDF1 is a critical component of a molecular switch that regulates HIF1a stability in cancer cells in hypoxia and that RHBDF1 is of potential value as a new target for cancer treatment.


Introduction
Intermittent oxygen deficiency in cancer microenvironment promotes prolonged inflammation, continuous neovascularization, and escalated drug resistance.The transcription factor hypoxia-inducible factor-1 (HIF1) predominantly regulates cellular responses to oxygen deficiency, including the expression of genes required for tissue oxygen delivery and energy metabolism in developmental, physiologic, and pathologic conditions such as ischemic cardiovascular disease, stroke, and cancer (1).HIF1 is a heterodimer composed of aand b-subunits.HIF1b level is maintained constitutively, whereas HIF1a under nor-moxic conditions is removed by oxygen-dependent prolyl hydroxylation, ubiquitination, and proteasomal degradation (2).In hypoxia, however, HIF1a level increases markedly as it is continuously synthesized (3), whereas its oxygen-dependent degradation is downregulated (2).An oxygen-independent mechanism has been proposed to account for the modulation of HIF1a in hypoxia (4), involving the receptor of activated protein C kinase-1 (RACK1) and HSP90 (5).HSP90 is a molecular chaperone that protects client proteins from misfolding and degradation (6).HSP90 binding stabilizes HIF1a.RACK1 competes with HSP90 for binding to HIF1a.RACK1 also binds to Elongin C and recruits other components of E3 ubiquitin ligases, thus facilitating HIF1a ubiquitination and degradation in an oxygen-independent manner (5).However, what this "molecular switch" consists of and how it operates remain unclear.
The human rhomboid family-1 gene product RHBDF1 is a protein found mainly within the endoplasmic reticulum and Golgi complex (7).Rhomboids are 6 or 7 transmembrane proteins that may be divided into two categories.Many rhomboids are serine proteases, conserved across all kingdoms of life, and regulate biologic processes as diverse as growth factor signaling, mitochondrial morphology, parasitic invasion, and bacterial protein translocation (8).The other category of rhomboid proteins, present in all sequenced metazoans, lack the known catalytic residues essential to the function of serine proteases (9).Few functions of these so called "inactive rhomboids," including RHBDF1, are known.In Drosophila, noncatalytic rhomboids were shown to prevent the cleavage of the substrates of rhomboid proteases by promoting their destabilization by endoplasmic reticulum-associated degradation (10).In human, RHBDF1 was shown to have a pivotal role in sustaining growth signals in epithelial cancers (11,12).RHBDF1 mRNA level is significantly elevated in clinical specimens of invasive ductal carcinoma of the breast, and RHBDF1 gene silencing results in apoptosis or autophagy in breast cancer or head and neck cancer cells and inhibition of xenograft tumor growth (11).RHBDF1 was found to participate in the modulation of Gprotein-coupled receptor-mediated transactivation of EGF receptor (12).These findings indicate that RHBDF1 may function as a regulatory protein involved in growth signal transduction.
In this study, we discovered that RHBDF1 function is critical for the maintenance of HIF1a stability in breast cancer cells in hypoxia.We report here that elevated RHBDF1 expression in breast cancer strongly correlates with escalated disease progression, poor prognosis, and poor responses to chemotherapy.We show that RHBDF1 is an essential component of a "molecular switch," which also consists of RACK1 and HSP90 that modulate oxygen-independent degradation of HIF1a.In addition, we demonstrate that disrupting the phosphorylation of a tyrosine residue on the RHBDF1 molecule leads to diminished HIF1a stability.These findings define a new function for a nonprotease rhomboid gene and provide new insights into the mechanism underlying the modulation of HIF1 activity in hypoxia.

Cells
Human breast cancer cell lines MCF7, MDA-MB-231, T47D, and human kidney 293 cell (all from American Type Culture Collection) were maintained in Dulbecco's Modified Eagle Medium (Lonza), 10% FBS, L-glutamate, penicillin and streptomycin.For hypoxic conditions, the cells were cultured in a modular incubator chamber flushed with mixed gas consisting of 1% O 2 , 5% CO 2 , and 94% N 2 at 37 C.For normoxic conditions, mixed gas consisting of 20% O 2 , 5% CO 2 , and 75% N 2 at 37 C was used.

Clinicopathological analysis of breast cancer specimens
The Ethics Committee of the Cancer Hospital of Tianjin Medical University (Tianjin, China) approved the use of human tissues for this study.Each patient signed an informed consent form for participation.Paraffin-embedded blocks of normal breast and breast cancer tissues diagnosed in 2003 were retrieved randomly from the archives of Cancer Hospital of Tianjin Medical University.Patients were women 24 to 83 years of age (mean age 52.6 years).Specimens were collected before radiation or chemotherapy, or both.Histopathology was reviewed and diagnosis confirmed independently by two pathologists (LF and FFL) using the WHO criteria (14).For pathologic analysis of patient responses to chemotherapy, breast cancer specimens were retrieved from archives diagnosed during 2007 to 2008.All invasive breast cancer (IBC) specimens were obtained from patients who had completed preoperative neoadjuvant chemotherapy consisted of 4 to 6 cycles of anthracycline-based or anthracycline and taxanebased regimen before surgery.Response to chemotherapy was assessed according to Miller and Payne histologic grading system (15): grade 1, no change or some alteration to individual malignant cells but no reduction in overall cellularity; grade 2, minor loss (up to 30%) of cancer cells but overall cellularity remains high; grade 3, reduction of 30% to 90% of cancer cells; grade 4, more than 90% loss of cancer cells but small clusters or widely dispersed individual cancer cells remain; grade 5, no malignant cells identifiable in sections from the site of the tumor consisting of vascular fibroblastic stroma, often containing macrophages; however, ductal carcinoma in situ (DCIS) may be present.In this study, grades 3 to 5 were regarded as having a good response to chemotherapy, whereas grades 1 and 2 were regarded as having a poor response.

Immunohistochemistry
Tissue sections (4 mm thickness) were deparaffinized and rehydrated.Antigen retrieval was performed at 121 C for 2 minutes, using citrate buffer (pH 6.0) for RHBDF1, ER, PR, C-erbB-2, or EDTA solution (pH 8.0) for HIF1a.After blocking with hydrogen peroxide and normal goat serum, the sections were incubated with primary monoclonal antibody against RHBDF1 (Abcam, cat# ab81342, 1:250 dilution) or HIF1a (Abcam, cat# ab8366, 1:250 dilution) for 16 hours at 4 C.The sections were sequentially incubated with biotinylated goat anti-mouse immunoglobulin and peroxidase-conjugated streptavidin (DAKO), and the substrate 3, 3 0 -diaminobenzidine tetrahydrochloride.Sections incubated with only PBS served as negative controls.RHBDF1 staining levels were classified using a modified scoring method (16) based on staining intensity (Supplementary Fig. S1A).Nuclear positivity of HIF1a was defined as the presence of perinecrotic or diffuse stained nuclei (Supplementary Fig. S1B).A specimen was considered to contain HIF1a-positive nuclei when more than 0% of nuclei were positive (median value cut-off: 0%; ref. 17).There was a positive correlation between cytoplasmic and nuclear positivity of HIF1a in the 263 IBC specimens we analyzed (Supplementary Fig. S1C).Interpretation and scoring of ER, PR, and C-erbB-2 staining were described previously (18,19).

Immunoprecipitation and Western blotting
Cell lysis buffer contained 50 mmol/L Tris, pH 8, 150 mmol/L NaCl, 0.1% SDS, 0.5% NaDoc, 1% NP-40, and protease inhibitors (Roche).The lysates were centrifuged at 14,000 rpm for 15 minutes at 4 C.For coimmunoprecipitation (co-IP) studies, in which the extracts were incubated overnight at 4 C, then with protein G or protein L-Sepharose beads (Santa Cruz Biotechnology) for 2 hours at 4 C.The beads were washed exhaustively with the lysis buffer.Immobilized proteins were eluted with 2Â Laemmli sample buffer and subjected to SDS-PAGE and Western blotting analysis.

Statistical analysis
Student two-tailed t test was used to analyze the statistical significance of differences in continuous variables between two groups.The Spearman correlation rank for nonparametric variables was used to assess the relationships between the categorical variables.Survival curves were calculated using the Kaplan-Meier method, and the differences were estimated by using the log-rank (Mantel-Cox) test.All statistical tests used a two-tailed significance level of P < 0.05.

Elevated RHBDF1 expression in breast cancer correlates with escalated disease progression
We compared RHBDF1 expression patterns in 343 clinical specimens of breast cancer of various stages and normal breast tissues by immunohistochemistry, using a four point grading scale (À, þ, 2þ, and 3þ for negative, weak, medium, and strong, respectively; Supplementary Fig. S1).RHBDF1 levels were mostly undetectable in normal breast tissues (n ¼ 20), but steadily increased in the order of atypical ductal hyperplasia (ADH; n ¼ 10; Z ¼ À2.019; P ¼ 0.043 by Mann-Whitney U test), DCIS (n ¼ 30; Z ¼ À2.032; P ¼ 0.042), and IBC (n ¼ 263; Z ¼ À2.443; P ¼ 0.015; Fig. 1A).RHBDF1 expression level in normal tissue adjacent to cancer (n ¼ 20) was also higher compared with that in normal breast tissues (Z ¼ À3.106; P ¼ 0.002; Fig. 1A).The percentages of medium and strong RHBDF1 expression (2þ/3þ) in normal, ADH, DCIS, and IBC specimens were 0, 20, 24, and 55%, respectively (Fig. 1B).In addition, IBC patients with high RHBDF1 expression (2þ/3þ; n ¼ 146) exhibited a poorer overall survival rate compared with patients with low RHBDF1 expression (À/þ; n ¼ 117; P ¼ 0.0021; Fig. 1C).Progression-free survival was also markedly worse for patients with IBC with high RHBDF1 expression than for those with low RHBDF1 expression (P ¼ 0.0003; Fig. 1D).Moreover, we examined the correlations between RHBDF1 expression levels and a range of clinicopathologic parameters in the IBC cases.The percentages of local recurrence in low and high RHBDF1 groups were 2.6% and 9%, respectively (P ¼ 0.029; Fig. 1E).The percentages of cases in RHBDF1 low and high groups with metastasis to at least one positive lymph node were 53% and 66%, and those with 10 or more were 15% and 24%, respectively (P ¼ 0.027; Fig. 1F).The percentages of distant metastasis in low and high RHBDF1 groups were 9.3% and 32.7%, respectively (P ¼ 0.012; Fig. 1G).There were no statistically significant differences between RHBDF1 low and high groups with regard to age, tumor size, histologic grade, TNM stage, estrogen receptor expression, progesterone receptor expression, and C-erbB-2 levels (Supplementary Table S1).Furthermore, we studied RHBDF1 expression levels in the tumor specimens of 67 patients with breast cancer who had completed a full protocol of neoadjuvant chemotherapy before surgical excision of the tumor (refer to Supplemental Materials and Methods for treatment detail).We found that more than 80% of the cases in low RHBDF1 group had good responses in comparison with about 40% of good responders in high RHBDF1 group (Fig. 1H).These findings indicate that elevated RHBDF1 expression in breast cancer is strongly correlated with facilitated disease progression, local recurrence, lymph node and distant metastasis, poor prognosis, and poor responses to chemotherapy.

Raising RHBDF1 expression levels in cancer cells leads to enhanced HIF1a protein stability
To begin to investigate RHBDF1's role in breast cancer progression, we compared RHBDF1 and HIF1a protein levels in the IBC specimens, and found that the expression patterns of the two are strikingly similar (rs ¼ 0.593, P < 0.0001; Fig. 2A).Less than 30% of the patients in RHBDF1 low group exhibited high HIF1a (2þ, 3þ), whereas in sharp contrast, about 75% of the patients in RHBDF1-high group showed high HIF1a levels (Fig. 2B).In addition, we found that the percentage of poor responders to neoadjuvant chemotherapy was less than 20% when both RHBDF1 and HIF1a levels are low, whereas it reached about 70% when both RHBDF1 and HIF1a are high (Fig. 2C).We thus overexpressed RHBDF1 by cDNA transfection in breast cancer T47D cells, which exhibit constitutively low expression of RHBDF1 (11).We found that raising RHBDF1 levels in these cells resulted in a marked increase in HIF1a levels in normoxia (20% O 2 ; Fig. 2D) as well as in hypoxia (1% O 2 ; Fig. 2E; the first two lanes).We cultured the cells in hypoxia for 4 hours, and then placed them in normoxia, and found a significant delay in HIF1a protein degradation in RHBDF1-overexpressing cells compared with vector-transfected control cells (Fig. 2E); the slower mobility species above the main HIF1a band were later identified as ubiquitinated HIF1a (see below).These findings indicate that not only high RHBDF1 levels are strongly correlated with markedly enhanced HIF1a stability in breast cancer in clinical settings as well as in cancer cell cultures under either normoxic or hypoxic conditions, but RHBDF1 may have a role in the modulation of HIF1a degradation.

RHBDF1 gene silencing leads to HIF1a destabilization
We determined the effect of RHBDF1 gene silencing on HIF1a stability, using breast cancer cell line MCF7 cells, which express high level of RHBDF1 (11).Transient transfection with a shRNA against RHBDF1 (shRHB) under hypoxic conditions caused a marked decline of HIF1a protein (Fig. 3A).To confirm this finding, we treated the cells with a mixture of 4 specific shRNA plasmids against RHBDF1.Western blotting analysis results indicated that silencing the RHBDF1 gene led to marked decline of HIF1a protein level in the cells under hypoxic conditions (Supplementary Fig. S5).We also found much faster HIF1a degradation during reoxygenation in shRHB-treated cells than that in shScr-treated cells (Fig. 3B).Treatment of the cells with shRHB in hypoxia led to a decreased nuclear accumulation of the HIF1a protein, which is consistent with a reduction of HIF1a stability (Fig. 3C).To find out whether shRHB inhibited HIF1 activity, we analyzed the secretion of VEGF, a target gene of HIF1a (20), by the cells.Treatment with Figure 2. Positive correlation between RHBDF1 and HIF1a levels in breast cancer specimens and RHBDF1-transfected T47D cells.A, RHBDF1 and HIF1a immunostaining of serial sections of IBC cases.Magnification, Â200; inset magnification, Â400.B, percentages of IBC specimens exhibiting various HIF1a levels (white, À; stripes, þ; crosses, 2þ; black, 3þ) with regard to low (À/þ) or high (2þ/3þ) RHBDF1 levels.C, responses of patients to chemotherapy with regard to levels of RHBDF1 and HIF1a; white, poor; stripes, good.D, impact of RHBDF1 (RHB) overexpression on HIF1a stabilization in T47D cells in normoxia compared with empty vector-transfection.E, impact of RHBDF1 overexpression on HIF1a protein stability in T47D cells during reoxygenation; cells were cultured in 1% O2 in the absence or presence of MG132 (10 mmol/L) for 4 hours, then reexposed to 20% O2 in the absence of MG132 for the indicated period of time (minutes).ÃÃ , P < 0.01; ÃÃÃ , P < 0.001, c 2 test, Kruskal-Wallis test, and Mann-Whitney U test.
shRHB for 4 to 24 hours in hypoxia led to an approximately 50% decrease in VEGF secretion into the culture media determined by ELISA (Fig. 3D).A similar inhibition of hypoxia-induced VEGF production by shRHB treatment was observed when the experiment was repeated with human breast cancer cell line MDA-MB-231 (Fig. 3E).We found that HIF1a mRNA levels were not affected by RHBDF1 shRNA treatment of the cells; however, the transcription of GLUT1, a target gene of HIF1, diminished in RHBDF1-treated cells, consistent with decreased HIF1a activities (Supplementary Fig. S2).These findings indicate that RHBDF1 function is essential to the maintenance of HIF1a protein stability and HIF1 activity in breast cancer cells under hypoxic conditions.

RHBDF1 protects HIF1a from RACK1-facilitated oxygenindependent degradation
To investigate whether RHBDF1 modulation of HIF1a stability involved prolyl hydroxylase activity, we transfected MCF7 cells with a HA-tagged mutant HIF1a that had two proline-toalanine substitutions (P402A/P564A), which confers irresponsiveness to prolyl hydroxylase (21).We found that shRHB treatment facilitated the degradation of the mutant HIF1a in hypoxia, indicating an independence on prolyl hydroxylation (Fig. 4A).We then cotransfected T47D cells with RHBDF1 cDNA and the mutant HIF1a, and found that raising RHBDF1 levels led to increased HIF1a stability in hypoxia (Fig. 4B).We then examined whether RACK1 and HSP90 are involved in RHBDF1-facilitated HIF1a stabilization.We first treated MCF7 cells with both RHBDF1 shRNA and RACK1 siRNA, and found that RACK1 siRNA treatment alone caused an accumulation of HIF1a in the cells; however, HIF1a diminished when the RHBDF1 gene was silenced at the same time (Fig. 4C and Supplementary Fig. S6A).In addition, we cotransfected T47D cells with RHBDF1 and the mutant HA-HIF1a, and treated the cells with RACK1 siRNA.RACK1 siRNA treatment of the RHBDF1-overexpressing cells significantly enhanced HIF1a stability (Fig. 4D and Supplementary Fig. S6B).To determine the involvement of HSP90, we simultaneously treated MCF7 cells with RHBDF1 shRNA and 17-AAG, which inhibit HSP90 activity (6), and found that RHBDF1 gene silencing further accelerated HIF1a degradation caused by 17-AAG treatment (Fig. 4E).Moreover, we treated RHBDF1-overexpressing T47D cells with 17-AAG, and found that RHBDF1-facilitated HIF1a stabilization was significantly abrogated by 17-AAG treatment (Fig. 4F).These findings support the view that RHBDF1 activity has an impact on the actions of RACK1 and HSP90 on HIF1a stability.
To determine whether RHBDF1 physically interacts with HIF1a or RACK1, we transfected human kidney 293 cell with FLAG-tagged RHBDF1 either alone or together with HA-tagged wild-type HIF1a.Co-IP analysis indicated that RHBDF1 was able to bind to RACK1 (Fig. 4G) and HIF1a (Fig. 4H), but not to HSP90 (data not shown).To determine whether RHBDF1interacting HIF1a was ubiquitinated, we treated MCF7 cells with or without MG132 and analyzed co-IP of RHBDF1 with HIF1a and RACK1.The results indicated that RHBDF1, RACK1, and ubiquitinated HIF1a formed a complex (Fig. 4I).To determine the effect of oxygen on these interactions, we transfected T47D cells with FLAG-RHBDF1 and the P402A/ P564A mutant HA-HIF1a, then treated the cells with either siRACK1 or 17-AAG and carried out co-IP analysis.We found that siRACK1 treatment enhanced HIF1a binding to HSP90, whereas 17-AAG treatment led to more HIF1a binding to RACK1 (Fig. 4J).These findings suggest that RHBDF1 overexpression lead to diminished RACK1-HIF1a interaction, thus protecting HIF1a from RACK1-induced, oxygen-independent degradation.

RHBDF1 inhibits RACK1 binding to HIF1a
To determine the effect of RHBDF1 gene silencing on RACK1 and HSP90 binding to HIF1a, we stably expressed shRHB in MCF7 cells, and treated the cells with MG132.Co-IP analysis revealed that shRHB inhibited HIF1a binding to HSP90 but, in sharp contrast, enhanced HIF1a binding to RACK1 (Fig. 5A).We then transfected the HA-tagged P402A/P564A mutant HIF1a into these cells and treated them with either 17-AAG or MG132, or both, in hypoxia.Ubiquitination of the HA-tagged mutant HIF1a and HIF1a or Elongin C binding to RACK1 were readily observed in shRHB-expressing cells when protein degradation was blocked with MG132, whereas HSP90 binding to HIF1a diminished; 17-AAG treatment caused a further decline of HSP90 binding to HIF1a; it is not immediately apparent to us, however, as to why there appeared to be a reduction in the ubiquitination of the mutant HIF1a upon treatment with 17-AAG and shRHB together compared with shRHB alone (Fig. 5B).To determine the effect of elevated RHBDF1 levels on the interaction between RACK1 and HIF1a, we transfected RHBDF1 cDNA into T47D cells that were engineered to overexpress the P402A/P564A mutant HA-HIF1a, cultured the cells in hypoxia, and carried out co-IP analysis (Fig. 5C).We found that the mutant HIF1a became more stable and its binding to RACK1 declined when RHBDF1 level was raised.Inhibition of HSP90 activity with 17-AAG led to increased RACK1 or Elongin C binding to HIF1a; however, these effects diminished in RHBDF1-overexpressing cells.In addition, we carried out a "pull-down" experiment to determine RHBDF1 impact on RACK1 dimerization required for RACK1 binding to HIF1a (22).We overexpressed FLAG-RACK1 in human kidney 293 cell, isolated the protein by using an anti-FLAG antibody immobilized on agarose resin, and used the FLAG-RACK1-enriched resin to isolate endogenous RACK1 from MCF7 cell homogenates.The MCF7 cells were engineered to express either shRHB or shScr.We found that RACK1 dimerization became more prominent in RHBDF1 gene-silenced cells in normoxia and also in hypoxia in the presence of MG132 (Fig. 5D).Moreover, consistent with a previous report that RACK1 dimerization gives rise to phosphorylation of serine-146 on the RACK1 protein (22), we found that RHBDF1 gene silencing led to enhanced serine phosphorylation of RACK1 (Fig. 5E).These results indicate that RHBDF1 interaction with RACK1 prevents RACK1 dimerization, which in turn inhibits HIF1a binding to RACK1.

Inhibition of RHBDF1 tyrosine phosphorylation causes disruption of RHBDF1-RACK1 interaction and facilitates HIF1a degradation
We carried out bioinformatics analysis on the primary sequence of RHBDF1 and discovered a number of potential phosphorylation sites (Supplementary Fig. S3).We synthesized a series of peptides with sequences corresponding to these sites.A poly-arginine tag (R 11 ) was added to each of the peptides at the C-terminus to ensure entrance of the peptide into the cells (23) in case some of these potential phosphorylation sites were located inside the cells (Supplementary Table S2).We treated MDA-MB-231 cells with each of the peptides, and found that peptide VR56, which mimics the amino acid sequence flanking tyrosine-848 of RHBDF1, was able to substantially inhibit VEGF production by the cells in hypoxia (Fig. 6A) as well as the transcription of a number of HIF1targeted genes, including VEGF, GLUT1, CA9, and ET1 (Supplementary Fig. S4).VR56 treatment also facilitated HIF1a degradation (Fig. 6B).A peptide identical to VR56 but without the tag (VH36), or the R 11 tag itself (ER20), did not exhibit these activities.In addition, we found that VR56 treatment of RHBDF1-overexpressing MCF7 cells in hypoxia led to a marked inhibition of RHBDF1-enhanced HIF1a stabilization (Fig. 6C).VR56 inhibition of RHBDF1-enhanced HIF1a stabilization was dose dependent (Fig. 6D).VR56 also inhibited RHBDF1 tyrosine phosphorylation (Fig. 6E).Treatment of MCF7 cells with VR56 resulted in a marked decrease of cell viability (Supplementary Fig. S4).We then treated RHBDF1-and RACK1-overexpressing MCF7 cells with VR56 in hypoxia, and found that the treatment significantly inhibited RHBDF1 binding to RACK1 (Fig. 6F).When these experiments were carried out in the presence of MG132, VR56 treatment inhibited RHBDF1 binding to RACK1 but enhanced HIF1a binding to RACK1 (Fig. 6G).These findings indicate that disruption of RHBDF1 tyrosine phosphorylation with VR56 leads to inhibition of RHBDF1 binding to RACK1, enhancement of HIF1a binding to RACK1, and destabilization of HIF1a.

Discussion
On the basis of these findings, we propose a "molecular switch" to explain RHBDF1 suppression of HIF1a degradation in hypoxia (Fig. 7).In this mechanism, RHBDF1 is in a position to control the competition of RACK1 and HSP90 for binding to HIF1a.RHBDF1 binding to RACK1 either takes the latter away from RACK1-HIF1a complex, or prevents RACK1 from binding to HIF1a.The latter scenario appears to be more likely, as RHBDF1 interaction with RACK1 results in an inhibition of the phosphorylation of the RACK1 protein and a disruption of RACK1 dimerization, which are necessary for RACK1 binding to HIF1a.Disruption of RACK1 binding to HIF1a by the action of RHBDF1 allows HIF1a to bind to HSP90 more readily instead of binding to RACK1 and being taken into the ubiquitin-Elongin C pathway for degradation.The action of RHBDF1 thus shifts the balance toward HIF1a stabilization.
It has been suggested that competition between HSP90 and RACK1 for binding to HIF1a may contribute to the establishment of the HIF1a "set point," which is a given HIF1a protein level in a given type of cells (5).Influence of RHBDF1 on the competition may have a significant impact on such "set point".Especially, considering that RHBDF1, RACK1, HSP90, and HIF1a protein levels are all significantly elevated in many cancer cells and tumor tissues (11,24,25), RHBDF1 influence on the equilibrium between HIF1a-RACK1 and HIF1a-HSP90 may have a pivotal role in the maintenance of HIF1a stability under hypoxic conditions.It is plausible that this mechanism underlies the association of RHBDF1 with clinicopathologic parameters critical for the progression of the disease.As we have shown, RHBDF1 protein is nearly absent or present at low levels in normal mammary gland tissues, then changes to moderate levels in atypical ductal hyperplasia as well as in normal tissues adjacent to tumors, then increases markedly to high and very high levels in DCIS and IBC, respectively.Facilitation of HIF1 activity by RHBDF1 may explain the close correlation of high RHBDF1 protein levels with lymph node and distant metastasis of the cancer cells, local recurrence, and poorer overall and disease-free patient survival.
Hypoxia-associated drug resistance is a major clinical issue.Consistent with RHBDF1-facilitation of HIF1 activity, we found that patients with low RHBDF1 protein levels responded better to the new adjuvant chemotherapy regimen.Because HIF1 directly promotes hypoxia-associated drug resistance by enhancing the expression of antiapoptotic proteins and diminishing the expression of proapoptotic proteins or by nonapoptotic mechanisms (26,27), inhibition of RHBDF1-facilitated HIF1a stabilization may potentially be beneficial not only to suppress tumor angiogenesis, but also to curb cancer cell resistance to apoptosis-inducing chemotherapies.In this regard, our study has demonstrated that by using the peptide mimic VR56, we may intervene RHBDF1-facilitated stabilization of HIF1a by inhibiting RHBDF1 tyrosine phosphorylation.This suggests that RHBDF1 may serve as a target for cancer drug development.It is of interest that noncatalytic rhomboids, such as human RHBDF1, may function as regulatory proteins.Noncatalytic rhomboids are considered to have evolved from rhomboid proteases that lost their catalytic activity but retained their location in the protein synthesis apparatus (10), and plausibly maintained their abilities to bind to what were once their substrates.New functions as regulatory proteins may be acquired by taking advantages of the expression pattern, subcellular location, and substrate-binding capacity.Consistent with this notion is the wide range of inactive cognates of many enzymes (28).That RHBDF1 may have evolved from a membrane-bound protease into a regulatory protein may exemplify this evolutionary route.
In summary, our experimental data are consistent with the view that RHBDF1, RACK1, and HSP90 form a "molecular switch" that controls oxygen-independent degradation of HIF1a.In this mechanism, RHBDF1 facilitates HIF1a stability by preventing RACK1 binding to HIF1a, thus attenuating ubiquitin-mediated HIF1a proteasomal degradation and shifting HIF1a binding toward HSP90.RHBDF1 is therefore an essential component of cell survival mechanism underlying cellular responses to oxygen deficiency.In addition, our findings illustrate that intervention of RHBDF1 activity by gene silencing or by specific inhibition of tyrosine phosphorylation of the RHBDF1 protein presents a potentially new approach to develop anticancer therapeutics.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.

Figure 1 .
Figure 1.Correlations between RHBDF1 expression, breast cancer disease progression, patient survival, and responses to chemotherapy.A, RHBDF1 immunostaining of specimens of various disease states, including normal breast tissue, adjacent normal tissue, ADH, DCIS, and IBC.Magnification, Â200.B, percentage of specimens of various disease stages exhibiting different RHBDF1 levels (P < 0.05, Kruskal-Wallis test and Mann-Whitney U test); white, À; stripes, þ; crosses, 2þ; black, 3þ.C, Kaplan-Meier plots of overall survival rate of patients with IBC exhibiting high (green, 2þ/3þ) or low (blue, À/þ) RHBDF1 levels (P ¼ 0.002, log-rank test).D, Kaplan-Meier plots of progression-free survival of patients with IBC exhibiting high (green, 2þ/3þ) or low (blue, À/þ) RHBDF1 levels (P < 0.001, log-rank test).E, percentages of recurrence in RHBDF1 low or high groups.F, percentages of lymph node metastasis in RHBDF1 low or high groups; marks on bars indicate the number of cancer cell-positive lymph nodes per case: white, 0; stripes, 1-3; crosses, 4-9; black, 10 or more.G, percentages of distant metastasis in RHBDF1 low or high groups.H, percentages of poor (white) or good (stripes) responses to chemotherapy with regard to RHBDF1 levels.Numbers on bars indicate the number of cases.Spearman rank correlation test.Ã , P < 0.05; ÃÃ , P < 0.01.

Figure 3 .
Figure 3. Inhibitory effect of RHBDF1 gene silencing in MCF7 cells on HIF1a stability and VEGF secretion.A, effect of RHBDF1 shRNA (shRHB) or scrambled shRNA (shScr) on HIF1a degradation in MCF7 cells in hypoxia.B, effect of shRHB or shScr on HIF1a protein stability in MCF7 cells during reoxygenation.C, nuclear translocation of HIF1a in shSrc-or shRHB-treated MCF7 cells; a-tubulin and Myc are cytoplasmic and nuclear protein controls, respectively.D, changes of VEGF concentrations in the condition media of shRHB (white bars)-or shScr (black bars)-treated MCF7 cells under normoxic or hypoxic conditions as a function of time (hours).E, changes of VEGF concentrations in the condition media of shRHB (white bars)-or shScr (black bars)-treated MDA-MB-231 cells under normoxic or hypoxic conditions as a function of time (hours).Each experiment was repeated two times.Ã , P < 0.05, Student t test.

Figure 4 .
Figure 4. Involvement of RHBDF1, RACK1, and HSP90 in oxygen-independent HIF1a degradation.A, effect of RHBDF1 gene silencing in MCF7 cells on the stability of HA-tagged P402A/P564A mutant HIF1a (HA-HIF1a) in hypoxia.B, effect of RHBDF1 overexpression in T47D cells on the stability of the mutant HIF1a in hypoxia.C, effect of RACK1 gene silencing on HIF1a stability in MCF7 cells in hypoxia.D, effect of FLAG-RHBDF1 overexpression in T47D cells on the stability of the mutant HIF1a in hypoxia.E, effect of 17-AAG treatment (0.5 mmol/L, 24 hours) of RHBDF1 gene-silenced MCF7 cells on HIF1a stability in hypoxia.F, effect of 17-AAG treatment (0.5 mmol/L, 24 hours) of RHBDF1-overexpressing T47D cells on HIF1a in hypoxia.G, co-IP of RHBDF1 and RACK1 from MCF7 cells transfected with FLAG-RHBDF1.H, co-IP of RHBDF1 and HIF1a from MCF7 cells cotransfected with HA-tagged wild-type HIF1a.I, co-IP of RHBDF1 with HIF1a and RACK1 from MCF7 cells in the absence or presence of MG132 (10 mmol/L, 4 hours) in normoxia or hypoxia.J, co-IP of HIF1a with HSP90 and of RHBDF1 with RACK1 from MCF7 cells cotransfected with FLAG-RHBDF1 and the mutant HIF1a upon treatment with either control siScr, RACK1 siRNA (siRACK1), or 17-AAG (0.5 mmol/L).Each experiment was repeated two times and representative Western blotting analysis or co-IP results from one experiment are shown.

Figure 5 .
Figure 5. Involvement of RHBDF1, RACK1, HSP90, and Elongin C in oxygen-independent HIF1a ubiquitination and degradation.A, co-IP of HIF1a, HSP90, and RACK1 from MCF7 cells stably expressing shRHB or ShScr.B, co-IP of HA-P402A/P564A-HIF1a (HA-HIF1a) and RACK1 or RACK1 and Elongin C from MCF7 cells stably expressing HA-HIF1a and shRHB in the presence or absence of 17-AAG (0.5 mmol/L, 24 hours) or MG132 (10 mmol/L, 4 hours), or both, in hypoxia.C, co-IP of HA-HIF1a, RACK1, and Elongin C from T47D cells overexpressing HA-HIF1a and FLAG-RHBDF1 and treated with 17-AAG in hypoxia.D, "pull-down" analysis of RACK1 dimerization in MCF7 cells stably expressing shRHB or shSrc, transfected with FLAG-RACK1, and cultured in normoxia or hypoxia.E, analysis of RACK1 serine-phosphorylation and Elongin C-binding in MCF7 cells stably expressing shRHB or shScr and cultured in normoxia or hypoxia.Each experiment was repeated two times and representative results from one experiment are shown.

Figure 6 .
Figure 6.Interruption of RHBDF1 tyrosine phosphorylation causes an inhibition of RHBDF1-facilitated HIF1a stabilization and disruption of RHBDF1 with RACK1.A, ELISA measurement of VEGF secretion by MDA-MB-231 cells in hypoxia treated with various peptides (20 mmol/L, 24 hours).B, HIF1a levels in MDA-MB-231 cells treated with the indicated peptides for 24 hours, followed by 6-hour culture in hypoxia.C, HIF1a stability in RHBDF1-or empty vector-transfected MCF7 cells treated with VR56 or ER20 in hypoxia.D, HIF1a stability in RHBDF1-overexpressing MCF7 cells in response to various VR56 doses.E, inhibition of RHBDF1 tyrosine phosphorylation by VR56 in RHBDF1-overexpressing MCF7 cells.F, disruption of RHBDF1-RACK1 interaction by VR56 in MCF7 cells cotransfected with RHBDF1 and RACK1; the cells were treated with vehicle, ER20, or VR56 for 48 hours in normoxia, followed by 6-hour culture in hypoxia.G, RACK1 binding to RHBDF1 or HIF1a in MCF7 cells cotransfected with RHBDF1 and RACK1 in response to VR56 or ER20 treatment; the cells were treated with the peptides for 24 hours in normoxia, followed by 6-hour culture in hypoxia in the presence of MG132.Each experiment was repeated two times and representative results from one experiment are shown.