Prognostic impact of total and tyrosine phosphorylated GIV/Girdin in breast cancers

Abstract

Gα-interacting vesicle-associated protein (GIV, aka Girdin) is a guanine exchange factor (GEF) for the trimeric G protein Gαi and a bona fide metastasis-related gene that serves as a platform for amplification of tyrosine-based signals via G-protein intermediates. Here we present the first exploratory biomarker study conducted on a cohort of 187 patients with breast cancer to evaluate the prognostic role of total GIV (tGIV) and tyrosine phosphorylated GIV (pYGIV) across the various molecular subtypes. A Kaplan-Meier analysis of recurrence-free survival showed that the presence of tGIV, either cytoplasmic or nuclear, carried poor prognosis, but that nuclear tGIV had a greater prognostic impact (P = 0.007 in early and P = 0.0048 in late clinical stages). Activated pYGIV in the cytoplasm had the greatest prognostic impact in late clinical stages (P = 0.006). Furthermore, we found that the prognostic impacts of cytoplasmic pYGIV and nuclear tGIV were additive (hazard ratio 19.0548; P = 0.0002). Surprisingly, this additive effect of nuclear tGIV/cytoplasmic pYGIV was observed in human epidermal growth factor receptor 2–positive tumors (hazard ratio 16.918; P = 0.0005) but not in triple-negative breast cancers. In triple-negative breast cancers, tGIV and cytoplasmic pYGIV had no prognostic impact; however, membrane-association of pYGIV carried a poor prognosis (P = 0.026). Both tGIV and pYGIV showed no correlation with clinical stage, tumor size, pathologic type, lymph node involvement, and BRCA1/2 status. We conclude that immunocytochemical detection of pYGIV and tGIV can serve as an effective prognosticator. On the basis of the differential prognostic impact of tGIV/pYGIV within each molecular subtype, we propose a diagnostic algorithm. Further studies on larger cohorts are essential to rigorously assess the effectiveness and robustness of this algorithm in prognosticating outcome among patients with breast cancer.—Dunkel, Y., Diao, K., Aznar, N., Swanson, L., Liu, L., Zhu, W., Mi, X.-Y., Ghosh, P. Prognostic impact of total and tyrosine phosphorylated GIV/Girdin in breast cancers.

Gα-interacting vesicle-associated protein (GIV; also known as Girdin) is a bona fide metastasis-related protein that modulates multiple signaling pathways triggered by diverse classes of receptors (reviewed in refs. 1–4). Among the numerous pathways that GIV affects are the prometastatic PI3K-Akt and STAT3 signaling pathways (2, 4, 5). Mechanistically, GIV modulates multireceptor signaling via its fundamental ability to serve as a guanine exchange factor (GEF) for the heterotrimeric (henceforth trimeric) G protein, Gαi. GIV couples Gαi proteins to various types of ligand-activated receptors, e.g., growth factor receptor tyrosine kinases (RTKs), integrins, GPCRs, and Toll-like receptors, many of which are known to engage in tyrosine-based signaling [reviewed in Aznar et al. (6) and Ghosh (7)]. By virtue of its ability to link G proteins to multiple receptor classes, GIV facilitates the transactivation of Gαi proteins in response to tyrosine-based signals that are initiated by a variety of external cues. During the process of such transactivation, GIV is phosphorylated by multiple RTKs and non-RTKs at 2 key tyrosines (Y1764 and Y1798) (5). These phosphosites directly bind and activate class I PI3Ks (5), and therefore the phosphoevents mark a key step of activation in GIV-dependent signaling.

Consistent with its ability to serve as a central hub for modulation of multireceptor/pathway signaling, GIV is involved in a wide range of biologic processes such as cancer cell migration, tumor angiogenesis, tumor–stroma interaction during cancer progression, cancer invasion, epithelial wound healing, organ fibrosis, neuronal migration, memory formation, macrophage chemotaxis, and vascular repair (2, 4, 8–11). The finding that GIV and its GEF function are essential for signal enhancement (such as PI3K-Akt) and actin remodeling during cancer cell migration and invasion in vitro (10, 12) led to the discoveries that GIV is essential for tumor invasion and metastasis in murine models (12–17). Depletion of GIV impairs metastasis and inhibits VEGF-mediated neoangiogenesis (12, 17), further supporting GIV’s role in tumor progression. Consistent with its ability to signal at the hub of multiple upstream receptors and multiple downstream pathways (2, 4) and its ubiquitous expression pattern (18), GIV is found to be expressed in a wide range of cancer tissues, including breast, colorectal, esophageal, gastric, glioblastoma multiforme, lung, and hepatocellular carcinomas (reviewed in ref. 4). Its expression at high levels invariably correlates with aggressiveness across the spectrum of tumors (19–23).

In the context of breast cancer, full-length GIV has been found to be significantly overexpressed in tumor cells with high metastatic potential; it is virtually undetectable in those with poor metastatic potential (9, 24). While high GIV expression levels correlate with tumor aggressiveness and disease progression (5, 9, 17, 20, 25–30), depletion of GIV by short hairpin RNA reduces lung metastases in nude mice (17). Although multiple prior exploratory studies have revealed the potential for GIV to serve as a biomarker for tumor aggressiveness across various molecular subtypes of breast cancers (9, 20, 25–30) (Supplemental Table 1), the relevance of those findings in the clinical setting has not been rigorously examined. Additionally, the prognostic impact of GIV in certain molecular subtypes, such as triple-negative breast cancers (TNBCs), remains unclear. More importantly, no attempt has been made to study the prognostic impact of the active pool of tyrosine phosphorylated GIV (pYGIV), nor has it been determined whether such analysis may add to the prognostic value of detecting the total pool of GIV (tGIV).

Here we analyzed the tGIV and pYGIV status and their patterns of staining (subcellular distribution) by immunohistochemistry (IHC) in a cohort of 187 patients with breast cancer. We found that the presence or absence of tGIV in the nucleus had the greatest prognostic impact: its presence both in early (carcinoma in situ stage I, II) and late (stage III, IV) stages of disease was associated with poorer outcome (P = 0.007 and 0.0006, respectively). In late stage disease, tumor recurrence at the 5 yr mark was 0% in the GIV-negative group and 47% in the GIV-positive group. Analysis of pYGIV added significant prognostic value when combined with tGIV; tumors with nuclear tGIV and cytoplasmic pYGIV had the highest recurrence rates. Findings also revealed that the patterns of tGIV and pYGIV staining, as well as their individual or combined prognostic impact, show striking differences between TNBCs and non-TNBCs. On the basis of these results, a diagnostic/prognostic algorithm is proposed that integrates GIV status with molecular subtyping of breast cancers.

MATERIALS AND METHODS

Patient cohort

In this study, 187 patients who had histologically confirmed breast cancer and who had undergone radical surgery at the First Affiliated Hospital of China Medical University were recruited between January 2009 and December 2012. Informed written consent was obtained. This study was approved by the ethics committee of China Medical University (Shenyang, China). Detailed patient cohort information is listed in Table 1.

TABLE 1.

Demographic, clinicopathologic, and molecular makeup of study cohort

Prognostic factor	Category	Patients (n)
Age	≤35 yr	15
	36–45 yr	44
	46–55 yr	55
	≥56 yr	71
	Unknown	2
Tumor stage	0	2
	I	34
	II	94
	III	44
	IV	6
	Unknown	7
Receptor status	ER (+)	118
	ER (−)	68
	PR (+)	117
	PR (−)	69
	HER-2 (+)	144
	HER-2 (−)	42
	TNBCs	34
	Non-TNBCs	152
	Unknown	1
Tumor grade	Ductal carcinoma in situ	28
	Invasive ductal carcinoma	159
Follow-up status	Known	168
	Unknown	19

Immunohistochemistry

A total of 187 breast cancer specimens of known histologic type and grade were analyzed by IHC using anti-GIV (1:100; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and anti-p-GIV (pY17654) (1:100; Spring Bioscience, Pleasanton, CA, USA) rabbit monoclonal antibodies. Briefly, formalin-fixed, paraffin-embedded tissue sections of 4 µm thickness were cut and placed on glass slides coated with 3-aminopropyl triethoxysilane, followed by deparaffinization and hydration. Heat-induced epitope retrieval was performed using citrate buffer (pH 6) in a pressure cooker. Tissue sections were incubated with 3% hydrogen peroxidase for 40 min to block endogenous peroxidase activity, followed by incubation with primary antibodies overnight in a humidified chamber at 4°C. Immunostaining was visualized with a labeled streptavidin-biotin using 3,3′-diaminobenzidine as a chromogen and counterstained with hematoxylin. All the samples were first quantitatively analyzed and scored on the basis of 2 independent criteria. First, the intensity of staining was scored on a scale of 0 to 3, where 0 = no staining, 1 = light brown, 2 = brown, and 3 = dark brown. Second, the percentage of the cells that stained positive in the tumor area was scored on a scale of 0 to 4, where 0 = 0, 1 = ≤10%, 2 = 11–50%, 3 = 51–75%, and 4 = >75%. Subsequently, each tumor sample was assigned a final score, which is the product of its (intensity of staining) × (% cells that stained positive). Tumors were categorized as negative when their final score was <3 and as positive when their final score was ≥3.

Statistical analysis

The relationship between the level of the pYGIV membrane expression and receptor status of patients was investigated by Fisher’s exact test. Nonparametric Spearman’s correlation analysis was used to assess the association between GIV expression and the following: tumor size; patient age; pathologic type; estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER-2) status; BRCA1/2 status; lymphovascular invasion; tumor stage; and lymph node involvement. Kaplan-Meier and multivariate Cox regression survival analyses were used to study associations between GIV/pYGIV expression status and recurrence-free survival (RFS). Kaplan-Meier analysis was performed to determine independent predictors of RFS for all patients studied. The Cox model included the following variables: age, size, clinical stage, and tGIV/pYGIV marker status. Kaplan-Meier analyses were performed by GraphPad Prism 5 software (GraphPad Software, La Jolla, CA, USA), whereas Cox regression analyses were performed by R software (R Foundation for Statistical Computing, Vienna, Austria). Likelihood ratio tests were used to compare nested models. The prognostic strength of the Cox models was evaluated using Harrell’s concordance index (c index) and compared using the dependent-sample Student’s t test. All statistical tests were 2-sided, and values of P < 0.05 were considered to be statistically significant.

In vitro kinase assay

These assays were performed using purified His-tagged carboxyl terminal aa 1660–1870 of GIV (His-GIV-CT) proteins (∼3 μg) and commercially obtained active HER-2 kinase (SignalChem, Richmond, BC, Canada) as previously (5). Briefly, reactions were started by addition of 1000 μM ATP and carried out at 25°C for 60 min in 25 μl kinase buffer [60 mM HEPES (pH 7.5), 5 mM MgCl₂, 5 mM MnCl₂, 3 μM Na₃OV₄]. Reactions were stopped by addition of Laemmli sample buffer and boiling. His-GIV-CT proteins phosphorylated were detected by immunoblotting.

Immunofluorescence

Cos7 cells expressing HER-2–green fluorescent protein (GFP), a gift from Hisataka Sabe (Hokkaido University, Sapporo, Japan) (31), were fixed at room temperature with 3% paraformaldehyde in PBS for 25 min, treated with 0.1 M glycine for 10 min, and subsequently permeabilized for 20 min (0.2% Triton X-100 in PBS) and blocked in PBS containing 1% bovine serum albumin and 0.1% Triton X-100 as described previously (11). Primary and secondary antibodies where incubated for 1 h at room temperature in blocking buffer. ProLong (Thermo Fisher Scientific, Waltham, MA, USA) was used as mounting medium.

Dilutions of antibodies used were as follows: phospho-Tyr1764-GIV (1:300), GFP (1:400), DAPI (1:2000), secondary goat anti-rabbit (488), and goat anti-mouse (594) Alexa conjugated antibodies (1:500).

Images were acquired with a TCS SPE-II w. DMI4000 microscope equipped with a Leica Hamamatsu 9100-02 camera and the LAS AF SPE software (Leica, Wetzlar, Germany) using a ×63 oil-immersion objective using 488, 561, and 405 laser lines for excitation. The settings were optimized and the final images scanned with line averaging of 3 scans. All images were processed by ImageJ software (Image Processing and Analysis in Java; National Institutes of Health, Bethesda, MD, USA) and assembled for presentation using Photoshop and Illustrator software (Adobe Systems, San Jose, CA, USA). Images shown are representative of cells that were evaluated across 3 independent experiments.

Immunoprecipitation

Lysates of Cos7 cells (∼2 mg of protein) expressing GFP-tagged HER-2 or vector control were incubated for 3 h at 4°C with 2 µg anti-GFP mAb. Protein G Sepharose beads (GE Healthcare, Waukesha, WI, USA) were added and incubated at 4°C for an additional 60 min. Beads were washed, and bound immune complexes were eluted by boiling in nonreducing Laemmli sample buffer.

RESULTS AND DISCUSSION

tGIV and pYGIV display distinct patterns of staining in breast cancers

To determine the levels and patterns of GIV expression and its activation downstream of tyrosine-based signaling pathways that drive breast cancers, we analyzed tGIV and pYGIV in formalin-fixed, paraffin-embedded breast tumors. The tumor tissues were obtained from 187 women who were diagnosed with breast cancer and who had undergone radical surgery at the First Affiliated Hospital of China Medical University between January 2009 and December 2012. As shown in Table 1, histopathologic, clinical staging, and receptor status information were available on all 187 tumors in this cohort, and detailed follow-up on clinical outcome (death or disease recurrence) was available for 168 patients.

tGIV was detected with a rabbit polyclonal antibody raised against the last 18 aa epitope of GIV’s C terminus (Fig. 1A); this antibody has previously been validated in several IHC studies on breast cancers by others and us (Supplemental Table 1). pYGIV was detected by a custom-designed pYGIV rabbit monoclonal antibody (Fig. 1A); although this antibody has been extensively validated in multiple applications [e.g., immunoblotting, immunofluorescence studies (32–34)] and IHC on non-breast tissues (11) by our group, it has never been validated in IHC studies on breast cancers. We tested both antibodies in breast cancer cells for use in IHC by analyzing tGIV and pYGIV in formalin-fixed, paraffin-embedded MDA-MB231 cell pellets (Supplemental Fig. 1). Both antibodies showed a cytoplasmic signal in control cells but not in GIV-depleted cells, indicating that the observed signals detected by both antibodies were specific.

Figure 1.

Patterns of total (tGIV) and tyrosine phosphorylated (pY1764) GIV staining in breast cancers. A) Schematic display of domain arrangement of GIV (top) and C-terminally located 2 epitopes against which antibodies used in this work were raised (bottom). GBD, Gα-binding domain. B) Paraffin-embedded human breast tumor samples were analyzed for tGIV and pY1764 GIV by IHC. Shown are representative images of stained tumor tissues with different patterns of staining for tGIV (top) and pY1764 GIV (bottom). Panels on right show magnified versions of boxed areas on left.

When the 187 tumors were analyzed for tGIV and pYGIV, 2 predominant patterns of staining were observed with each antibody: cytoplasmic and nuclear for tGIV and cytoplasmic and membrane for pYGIV (Fig. 1B and Table 2). When staining patterns were compared within the same tumor, there were several tumors in which tGIV was positive but pYGIV was not (Fig. 2A, patient 1); the converse was never observed. In many tumors, both were positive (Fig. 2A, patients 2, 3, and 4). The patterns of staining with tGIV and pYGIV were similar in some tumors (cytoplasmic pattern; Fig. 2A, patient 4) and dissimilar in others. For example, although pYGIV was frequently seen at the plasma membrane in several tumors (Fig. 2A, patient 3), tGIV was cytoplasmic. In these tumors, it is possible that the plasma membrane–associated pool of GIV is undetectable as a result of limited accessibility to the 18 aa epitope C-terminal antigen that was used to raise this the tGIV antibody. However, when tGIV was detected at the plasma membrane (Fig. 2B, patient 5), a nonparametric Spearman’s correlation analysis showed that such pattern of staining showed a negative correlation with the presence of ER, PR, and HER-2 in the tumors and positive correlation with the presence of lymphovascular invasion (Supplemental Table 1). Neither tGIV nor pYGIV showed any correlation with tumor size, histopathologic types (ductal carcinoma in situ or invasive ductal carcinoma), BRCA1/2, tumor stage, or lymph node involvement.

TABLE 2.

Staining pattern of tGIV and pYGIV across various clinical stages

Stage	IHC staining	tGIV			Phospho-Y1764-GIV			n
		Membrane	Cytoplasm	Nucleus	Membrane	Cytoplasm	Nucleus
0	Positive	0	0	0	1	2	0	2
	Negative	2	2	2	1	0	0
	Percentage positive	0%	0%	0%	50%	100%	0%
I	Positive	2	22	16	11	15	0	34
	Negative	32	12	18	23	19	34
	Percentage positive	6%	65%	47%	32%	44%	0%
II	Positive	8	52	27	46	58	5	94
	Negative	86	42	67	48	36	89
	Percentage positive	9%	55%	29%	49%	62%	5%
III	Positive	3	28	23	18	25	1	44
	Negative	41	16	21	26	19	43
	Percentage positive	7%	64%	52%	41%	57%	2%
IV	Positive	1	6	3	3	5	0	6
	Negative	5	0	3	3	1	6
	Percentage positive	17%	100%	50%	50%	83%	0%
Unknown	Positive	0	7	5	2	6	0	7
	Negative	7	0	2	5	1	7
	Percentage positive	0%	100%	71%	29%	86%	0%

Figure 2.

Correlation between total (tGIV) and tyrosine-phosphorylated (pY1764) GIV patterns in breast tumors. Paraffin-embedded breast tumors from patients 1–4 (A) and patient 5 (B) were analyzed for both tGIV and pY1764 GIV by IHC. Representative images of stained tumor tissues are shown. Panels on right show magnified versions of boxed areas on left.

Prognostic impact of tGIV and pYGIV staining patterns in early vs. late stage breast cancers

Next we asked whether tGIV or pYGIV status (positive vs. negative) and their subcellular distribution had any significant prognostic impact on the RFS. For this, we plotted Kaplan-Meier survival curves for tumors positive or negative for staining on 168 patients (out of 187 total) in whom detailed follow-up information was available for a period of 5 yr (Table 1) and compared survival between the 2 groups using the log rank test. In the case of tGIV, we found that RFS for patients (all clinical stages combined) whose tumor epithelium stained positive in any compartment, either nuclear or cytoplasmic, was significantly reduced compared to the patients with tumors negative for tGIV (P = 0.0032; Fig. 3A). Nuclear tGIV (any tumor with nuclear signal for tGIV, irrespective of the presence or absence of cytoplasmic signal) performed better than cytoplasmic tGIV (Fig. 3B, C), which is in keeping with prior reports that nuclear GIV correlates with worse prognosis (26). Nuclear tGIV maintained its prognostic edge over cytoplasmic tGIV when the analyses were carried out exclusively among early (clinical stages 0, I, and II; P = 0.007; Fig. 3D–F) or late (clinical stages III and IV; P = 0.0006; Fig. 3G–I) stage tumors.

Figure 3.

Prognostic impact of total (tGIV) and tyrosine-phosphorylated (pYGIV) GIV in early vs. late stage breast cancers. Kaplan-Meier plots were generated using survival data (y axis) of 168 patients with breast cancers stratified according to presence (+) or absence (−) of either tGIV (A–I) or pYGIV (J–L) plotted against duration of follow-up (x axis). Nuc, nuclear; Cyto, cytoplasmic; Nuc + Cyto, tumors that stained positive in either location.

In the case of pYGIV, we found that RFS for patients of all clinical stages combined whose tumor epithelium stained positive, either membrane or cytoplasmic, was significantly reduced compared to those that were negative (P = 0.0433; Fig. 3J). Cytoplasmic pYGIV alone accounted for most of the observed significance (P = 0.0061; Fig. 3K) because a Kaplan-Meier analysis with membrane pYGIV alone failed to reach significance (Supplemental Fig. 2A). An early vs. late stage analysis revealed that the prognostic impact of cytoplasmic pYGIV was limited only to late stage disease (P = 0.0048; Fig. 3L and Supplemental Fig. 2B–D). Taken together, the data imply that while both tGIV and pYGIV may have independent prognostic significance in differentiating those at highest risk for disease recurrence, there were significant differences: tGIV performed well in prognosticating RFS at both early and late stages of disease, but pYGIV had a prognostic value only in the late stages; and presence of tGIV in the nucleus or pYGIV in the cytosol were the overriding determinants within each independent assessment.

Prognostic impact of tGIV and pYGIV are additive

Next we asked whether assessment of cytoplasmic pYGIV in tumors that are tGIV positive can refine and further improve the prognostic impact of tGIV alone. We found that among the tumors that were tGIV positive, RFS was significantly reduced among those that were positive for cytoplasmic pYGIV than those that were not (P = 0.016; Fig. 4A). RFS among patients whose tumors were tGIV and pYGIV positive was significantly reduced than in those whose tumors were tGIV negative (P = 0.0003; Fig. 4B), indicating that cytoplasmic pYGIV improves the prognostic value of tGIV alone. Such improvement was also observed when tGIV-positive patients were first stratified into cytoplasmic (Fig. 4C, D) or nuclear (Fig. 4E, F) tGIV and then analyzed for the presence or absence of cytoplasmic pYGIV. Consistent with the superior prognostic ability of nuclear tGIV we observed earlier, the addition of cytoplasmic pYGIV to nuclear tGIV gave the best risk stratification; RFS among patients with tumors that were positive for both nuclear tGIV and cytoplasmic pYGIV was significantly lower than those negative for nuclear tGIV (P < 0.0001; Fig. 4F).

Figure 4.

Prognostic impact of adding tyrosine-phosphorylated (pYGIV) GIV-status to GIV-positive breast cancers. A) Kaplan-Meier plot was generated using survival data (y axis) of 110 patients with breast cancers positive for tGIV in any location (either nuclear or cytoplasmic) stratified according to presence (+) or absence (−) of cytoplasmic pYGIV plotted against duration of follow-up (x axis). B) Overlay of Kaplan-Meier plots in A and Fig. 3A. P value (0.0003) was achieved between 69 patients with positive cytoplasmic pYGIV and positive tGIV in any location (either nuclear or cytoplasmic), and 58 patients with negative tGIV in any location (either nuclear or cytoplasmic). C) Kaplan-Meier plots were generated as above using survival data from subset of patients in A and B (100 patients) whose tumors were positive for cytoplasmic tGIV. D) Overlay of Kaplan-Meier plots in C and Fig. 3C. P value (0.0003) was achieved between 36 patients with positive cytoplasmic pYGIV and tGIV and 68 patients with negative cytoplasmic tGIV. E) Kaplan-Meier plots were generated as above using survival data from subset of patients in A and B (62 patients) whose tumors were positive for nuclear tGIV. Nuc, nuclear; Cyto, cytosolic; Nuc + Cyto, tumors that stained positive in either location. F) Overlay of Kaplan-Meier plots in E and Fig. 3E. P < 0.0001 was achieved between 37 patients with positive cytoplasmic pYGIV and positive nulear tGIV and 106 patients with negative nuclear tGIV. G) Schematic summarizing progressive increase in risk for tumor recurrence. From left to right: tumors that are negative for both tGIV and pYGIV have lowest risk, followed by those that have cytoplasmic tGIV, followed by nuclear tGIV. Presence of cytoplasmic pYGIV within each category increases risk for recurrence, such that tumors positive for nuclear tGIV and cytoplasmic pYGIV have highest risk.

Next we carried out Cox proportional hazard regression analysis to further evaluate the effect of tGIV/pYGIV status on survival rates after adjustment for clinical covariates including age, tumor size, and tumor stage. The analyses revealed that cytoplasmic pYGIV enhanced the prognostic impact of tGIV compared to tumors without tGIV. RFS was significantly lower when cytoplasmic pYGIV was present in addition to nuclear tGIV [hazard ratio (HR) 19; P = 0.0002; 95% confidence interval [CI] 4.14–87.7], cytoplasmic tGIV (HR 16; P = 0.008; 95% CI 2.0–124.3), or tGIV at either of those locations (HR 11; P = 0.02; 95% CI 1.43–84.7). Although statistical significance was reached in each case, it is noteworthy that the 95% CIs are wide, and further studies with larger cohort sizes are warranted. Regardless, both log rank and Cox proportional hazard regression analyses are in agreement that risk stratification for tumor recurrence improves by adding cytoplasmic pYGIV status to tGIV at either the nuclear or cytoplasmic location. Risk was lowest when tumors are negative for both tGIV and pYGIV, and higher when they are positive for both; risk was highest when tumors were positive for cytoplasmic pYGIV and nuclear tGIV (Fig. 4G).

Prognostic impact of tGIV and pYGIV staining patterns differ on the basis of the molecular subtype of breast cancers

Because molecular classification has become the reference standard for complete characterization of breast cancers (35), next we asked whether and how tGIV and pYGIV staining differed among different molecular subtypes of breast cancers classified using receptor expression status as a surrogate. Of the 187 tumors analyzed in this study, 34 were basal-like TNBCs characterized by a lack of expression of ER, PR, and HER-2 (Table 1). Of the remaining 152 tumors that were non-TNBCs, 144 were HER-2 positive, and in most cases positive also for ER, PR, or both.

Fisher’s exact test and correlation analysis revealed that HER-2-positive tumors were more likely to have an increased presence of pYGIV on membranes (Fig. 5A and Supplemental Table 2). Because HER-2 is a receptor tyrosine kinase (RTK) that belongs to the epidermal growth factor receptor (EGFR) family (HER, EGFR, ERBB), and because GIV is phosphorylated by multiple receptor and nonreceptor tyrosine kinases, such as EGFR (5), we asked whether HER-2 phosphorylates GIV. In vitro kinase assays confirmed that recombinant HER-2 can phosphorylate His-GIV-CT directly, but not a nonphosphorylatable YF mutant in which Y1764 is replaced by Phe(F) (Fig. 5B). Overexpression of HER-2-GFP (31) in Cos7 cells was associated with increased pYGIV, as determined by both immunofluorescence (Fig. 5C) and immunoblotting (Fig. 5D, left). Consistent with the fact that GIV directly binds cytoplasmic tails of multiple ligand-activated RTKs, e.g., EGFR, InsR, VEGFR, and platelet-derived growth factor receptor (reviewed in ref. 4), coimmunoprecipitation studies confirmed that pYGIV also interacts with HER-2 (Fig. 5D, right). These findings suggest that HER-2, like other RTKs, may bind and phosphorylate GIV and subsequently utilize pYGIV to relay tyrosine-based signals (5).

Figure 5.

Prognostic impact of GIV in HER-2-positive breast cancers. A) Bar graph showing frequency of HER-2 positivity (percentage of tumors; y axis) among those with (+) or without (−) membrane staining with pYGIV. Larger fraction of tumors that stained positive for pYGIV expressed HER-2. mem, membrane. B) In vitro kinase assays were carried out by incubating recombinant His-GIV-CT wild-type or mutant proteins in presence (+) or absence (−) of HER-2 kinase and ATP. Reactions were analyzed for phosphoproteins by immunoblotting. Dual color imaging was carried out to visualize substrate (red; His-GIV-CT) and pYGIV (green). Yellow pixels, phosphorylated substrate. C) Cos7 cells transiently transfected with GFP-HER-2 were fixed and stained for pYGIV (red) and HER-2 (GFP; green) and DAPI (nucleus; blue) and analyzed by confocal microscopy. Representative image is shown. Transfected (star) cells showed increased cytoplasmic pYGIV compared to untransfected cells. D) Immunoprecipitation assays were carried out using equal aliquots of lysates of Cos7 cells expressing HER-2-GFP or vector control (left) and anti-GFP mouse IgG and protein G beads. Bound immune complexes were analyzed for HER-2 (GFP) and pYGIV by immunoblotting. E–J) Kaplan-Meier plots were generated using survival data (y axis) of 128 patients with HER-2-positive breast cancers stratified according to presence (+) or absence (−) of either tGIV (E–G) or pYGIV (H) or combination of both (I, J) plotted against duration of follow-up (x axis). Nuc, nuclear; Cyto, cytoplasmic; Nuc + Cyto, tumors that stained positive in either location.

As for the prognostic impact of tGIV/pYGIV, the previously observed patterns were retained in HER-2-positive group, as follows: 1) the prognostic impact of nuclear tGIV was higher than cytoplasmic tGIV (Fig. 5E–G); 2) cytoplasmic pYGIV alone (Fig. 5H), but not membrane pYGIV, was effective in stratifying recurrence risk (Supplemental Fig. 3A, B); and 3) the addition of cytoplasmic pYGIV to nuclear tGIV improved the prognostic impact of nuclear tGIV alone (Fig. 5H–J). Cox proportional hazard regression analysis was performed (after accounting for variables such as age, tumor size, and tumor stage) to further evaluate the additive effect of cytoplasmic pYGIV on the nuclear tGIV-positive group in HER-2-positive patients. Such analysis revealed that with the addition of cytoplasmic pYGIV to nuclear tGIV, the RFS between tumors negative or positive for both was significantly different (HR 16.9, P = 0.0005; 95% CI 3.5–81.8). We conclude that the combined prognostic impact of tGIV and pYGIV (both intensity and subcellular patterns) that we observed in the entire cohort (Fig. 4) is also retained in the HER-2-positive subgroup.

Parameters that had significant prognostic impact among HER-2-positive tumors, i.e., nuclear or cytoplasmic tGIV or cytoplasmic pYGIV, lost their prognostic impact when analyzed in the TNBC cohort (Supplemental Fig. 4A–E). Instead, in the case of TNBCs, there was a significant negative correlation between TNBC status and membrane localization of pYGIV (Fig. 6A), and it was presence or absence of pYGIV on membrane that successfully stratified the risk of tumor recurrence in this molecular subtype (P = 0.026; Fig. 6B). These results indicate that membrane pYGIV alone, but not tGIV or cytoplasmic pYGIV, is a useful prognostic tool among patients with TNBCs.

Figure 6.

Comparing and contrasting prognostic impact of GIV in TNBCs vs. non-TNBCs. A) Bar graph displaying frequency of pYGIV staining on membrane (percentage of tumors; y axis) among TNBCs and non-TNBCs analyzed in this study. Tumors with pYGIV on membranes are significantly overrepresented among non-TNBCs than among TNBCs. B) Kaplan-Meier plots were generated using survival data (y axis) of 34 patients with TNBCs stratified according to presence (+) or absence (−) of pYGIV on membranes. C) Schematic summarizing proposed diagnostic algorithm for stratifying risk for tumor recurrence among patients with diverse molecular subtypes of breast cancers. Neg, negative; pos, positive.

CONCLUSIONS

Early (node-negative) breast cancer accounts for two-thirds of newly diagnosed cancer identified by screening. Of these, 70% of these patients would remain cured with surgery and radiation alone, whereas ∼30% would experience relapse by 10 yr. Treating all patients with chemotherapy improves survival in a few (36, 37) while subjecting all to toxic adverse effects (38). Despite the development of multigene signatures to classify tumors or single receptor-based (HER-2) markers, there is still an urgent need for differentiating aggressive tumors from indolent ones (39, 40). More importantly, because breast cancer is a heterogeneous disease composed of various molecular subtypes, the need for a biomarker that permits prognostication across various subtypes remains unmet. The major finding of this work is demonstration of the individual and additive prognostic impact of tGIV and pYGIV. While the combined presence of nuclear tGIV and cytoplasmic pYGIV was associated with worse outcome among patients with the HER-2-positive tumor subtype, the presence of membrane pYGIV alone was associated with worse outcome among patients with the TNBC subtype. The discovery of such a strikingly different tGIV/pYGIV signature among the HER-2-positive and TNBC tumor subtypes is the second major finding of this work. We have no insight into why nuclear tGIV is an effective prognosticator in HER-2-positive tumors but not in TNBCs, and it remains unclear why cytoplasmic pYGIV is a key determinant of poor outcome in HER-2-positive tumors. However, membrane pYGIV appears to be the critical determinant in the case of TNBCs. These findings expose our incomplete understanding of the biology of GIV in breast cancer progression and underscore the urgent need for mechanistic insights into what role GIV—and more specifically tyrosine-based signaling via GIV—may play in the context of each molecular subtype of breast cancer.

This study has several limitations. First, the wide CIs in the Cox proportional hazard regression analysis indicate that the sample size was small and that the power of this study was low for all but large effects. It is noteworthy that despite the low-power nature of this study, nuclear tGIV and cytoplasmic pYGIV emerged as prognostic markers that can effectively stratify risk for disease recurrence. This is particularly important because only a large effect in an exploratory study like this would likely have any use comparable to or better than current clinical practice and be robust to the noise and potential biases of retrospective analysis. Second, we did not examine the predictive effect of adjuvant chemotherapy with the tGIV/pYGIV algorithm. The retrospective nature of our study and the population-based cohorts (from a single country) impaired our ability to draw reliable conclusions about the ability of this tGIV/pYGIV algorithm to predict which patients are more or less likely to benefit from adjuvant chemotherapy. This question is best addressed with samples and data from a randomized, multicenter clinical trial.

In summary, by demonstrating that the prognostic significance expression levels and subcellular localization of tGIV or pYGIV varies from one molecular subtype to another, this work accomplished 2 goals. First, it introduced a new set of tools (tGIV and pYGIV antibodies) in the armamentarium for prognosticating RFS in patients with breast cancer. Second, it helps us propose an algorithm (Fig. 6C) for stratification of risk of recurrence among patients with different molecular subtypes of breast cancers. Further studies on larger cohorts are warranted to determine whether such an algorithm is effective and robust in prognosticating outcome and predicting chemoresponsiveness in the clinical setting.

Glossary

EGFR

epidermal growth factor receptor

ER

estrogen receptor

GEF

guanine exchange factor

GIV

Gα-interacting vesicle-associated protein

HER-2

human epidermal growth factor receptor 2

His-GIV-CT

His-tagged carboxyl terminal aa 1660–1870 of GIV

HR

hazard ratio

IHC

immunohistochemistry

PR

progesterone receptor

pYGIV

tyrosine phosphorylated GIV

RFS

recurrence-free survival

RTK

receptor tyrosine kinase

tGIV

total GIV

AUTHOR CONTRIBUTIONS

Y. Dunkel, K. Diao, X.-Y. Mi, and P. Ghosh designed the study; Y. Dunkel, K. Diao, and P. Ghosh analyzed data; M. Y. Dunkel, K. Diao, N. Aznar, L. Swanson, and L. Liu performed research; Y. Dunkel and P. Ghosh wrote the paper; X.-Y. Mi contributed the patient cohort (i.e., the tumor samples analyzed and the demographic and clinical information associated with each tumor specimen); and W. Zhu and Y. Dunkel performed the statistical analysis using appropriate software.