β-Arrestin–biased agonism as the central mechanism of action for insulin-like growth factor 1 receptor–targeting antibodies in Ewing’s sarcoma

Abstract

Owing to its essential role in cancer, insulin-like growth factor type 1 receptor (IGF-1R)–targeted therapy is an exciting approach for cancer treatment. However, when translated into clinical trials, IGF-1R–specific antibodies did not fulfill expectations. Despite promising clinical responses in Ewing’s sarcoma (ES) phase I/II trials, phase III trials were discouraging, requiring bedside-to-bench translation and functional reevaluation of the drugs. The anti-IGF-1R antibody figitumumab (CP-751,871; CP) was designed as an antagonist to prevent ligand–receptor interaction but, as with all anti-IGF-1R antibodies, it induces agonist-like receptor down-regulation. We explored this paradox in a panel of ES cell lines and found their sensitivity to CP was unaffected by presence of IGF-1, countering a ligand blocking mechanism. CP induced IGF-1R/β-arrestin1 association with dual functional outcome: receptor ubiquitination and degradation and decrease in cell viability and β-arrestin1–dependent ERK signaling activation. Controlled β-arrestin1 suppression initially enhanced CP resistance. This effect was mitigated on further β-arrestin1 decrease, due to loss of CP-induced ERK activation. Confirming this, the ERK1/2 inhibitor U0126 increased sensitivity to CP. Combined, these results reveal the mechanism of CP-induced receptor down-regulation and characteristics that functionally qualify a prototypical antagonist as an IGF-1R–biased agonist: β-arrestin1 recruitment to IGF-1R as the underlying mechanism for ERK signaling activation and receptor down-regulation. We further confirmed the consequences of β-arrestin1 regulation on cell sensitivity to CP and demonstrated a therapeutic strategy to enhance response. Defining and suppressing such biased signaling represents a practical therapeutic strategy to enhance response to anti-IGF-1R therapies.

There is very strong experimental support to qualify the insulin-like growth factor receptor (IGF-1R) as a major therapeutic target in cancer: (i) the IGF-1R is essential for malignant transformation as its absence generates cells resistant to transformation by several oncogenes, viruses, or overexpression of other receptor tyrosine kinases (RTKs) (1); (ii) IGF-1R regulates multiple cellular functions essential to maintain the malignant phenotype, including cellular proliferation, survival, anchorage-independent growth, tumor neovascularization, migration, invasion, and metastasis (2, 3); and (iii) in preclinical settings, it has been clearly demonstrated that inhibition of IGF-1R would be beneficial for cancer treatment (4–6).

Accordingly, several strategies for IGF-1R molecular targeted therapy were developed, and those demonstrating significant activity in preclinical settings, mostly antibodies, were taken forward for clinical evaluation (7). Based on preclinical experimental findings, there was reason for optimism regarding the targeting of IGF-1R for cancer treatment, yet the outcomes of phase III clinical trials have been disappointing (8). In clinical settings, treatment with anti-IGF-1R antibodies induced clinical responses only in some cases of Ewing’s sarcoma (ES) (9–11) and selected cases of lung carcinoma (8, 12). The unsatisfactory results obtained in ES deserve special consideration as preclinical reports demonstrated IGF-1R activity to be sine qua non for this type of cancer. This childhood tumor is characterized by a chromosomal translocation t(11;22)(q24;q12) encoding the oncogenic EWS-Fli1 fusion protein. Early studies established that, in the absence of IGF-1R, the EWS-Fli1 oncogene cannot induce malignant transformation (13). Moreover, a feedforward mechanism between EWS-Fli1 and IGF-1R was revealed in experimental models of ES: IGF-1R regulates essential signaling pathways sustaining the ES malignant phenotype (14–16), whereas several microRNAs that target IGF signaling are repressed by the EWS-Fli1 fusion protein (17). In such experimental conditions, both in vivo and in vitro studies using anti-IGF-1R antibodies, small molecule inhibitors, and antisense technology have shown that the IGF-1R is functionally essential for ES cell growth and proliferation and its targeting prevents tumor growth and survival (6, 7, 18).

In the settings of the first clinical trial with the anti-IGF-1R antibody, figitumumab (CP-751,871; CP) (19) in ES, there were reports of major responses (10). However, initial phase III trials with CP-based therapies report a lack of efficacy (8). This gap between the promising in vitro results and the unsatisfactory clinical results might be explained by tumor heterogenicity, resistance mechanisms, or ligand/receptor switches (8, 20), although there is a contradiction of the mechanism of action of anti-IGF-1R antibodies to be considered. According to the classical paradigm, IGF-1R signaling and its down-regulation are exclusively dependent on the presence of the agonist (21–23). All anti-receptor antibodies, including CP, were designed as antagonists, aiming for signaling inhibition by blocking ligand binding to the receptor. In this way the anti-IGF-1R antibodies should prevent both receptor activation and down-regulation. However, all anti-IGF-1R antibodies enhance receptor down-regulation, a feature that contradicts the antagonist concept; therefore, the present study aims to investigate the paradox of CP-induced IGF-1R down-regulation.

β-Arrestins comprise a small family of multifunctional proteins that control G protein–coupled receptor (GPCR) signaling (24, 25). Initially described only as desensitizers of canonical GPCR signaling, the β-arrestins are now recognized as activators of G protein–independent signaling (24, 25). The discovery of β-arrestin–mediated signaling led to a major change in the understanding of GPCR physiology and a new appreciation of the biased agonism model. In this major paradigm shift, as opposed to balanced ligands that equally activate G protein– and β-arrestin–mediated pathways, biased agonists trigger only a subset of signaling effects, e.g., either G protein– (G protein–biased) or β-arrestin–mediated (β-arrestin–biased) signaling (26, 27).

Results

Effects of CP Treatment on Cell Viability in Ewing’s Sarcoma Cell Lines.

In the first set of experiments, we aimed to characterize our experimental model regarding the expression of IGF-1R, its capability to mediate IGF-1 signaling, and its sensitivity to CP. IGF-1R expression was investigated in a panel of validated ES cell lines: SKES, RDES, CADO, A673, and SKNMC. Breast cancer cell line SKBR3, expressing low IGF-1R levels (28), and IGF-1R KO mouse embryonic fibroblast (MEF) cells (R−) were used as negative controls. R− cells stably transfected with WT human IGF-1R (R+) were used as a positive control. As shown in Fig. 1A, the ES cell lines tested expressed variable levels of IGF-1R, with CADO and A673 displaying the highest and lowest levels, respectively. The R+ cells expressed approximately 10 times more receptors than CADO, whereas in the negative control cells, the IGF-1R was undetectable (R−) or expressed at very low levels (SKBR3). We also tested the function of IGF-1R by analyzing the IGF-1–induced receptor autophosphorylation of the activation loop and the subsequent activation of the two major IGF-1R downstream signaling pathways: MAPK/ERK and PI3K/protein kinase B (PKB, AKT). Serum-starved cells were stimulated with IGF-1 for 10 min, and levels of phosphorylated (P-)IGF-1R, P-ERK, and P-AKT were detected by Western blotting (WB) analysis (Fig. 1B). Total IGF-1R, AKT, ERK, and GAPDH were used as controls. In all ES cell lines, stimulation with 50 ng/mL IGF-1 induces receptor phosphorylation at tyrosine residues and subsequent signaling activation through the ERK and AKT pathways, demonstrating a functional IGF-1R response. In the negative control cells SKBR3, IGF-1 failed to induce any detectable level of P-IGF-1R or AKT phosphorylation, whereas ERK was constitutively phosphorylated and unresponsive to IGF-1 (Fig. 1B). Finally we investigated the sensitivity to antibody targeting IGF-1R by measuring the cell viability of ES cell lines following CP treatment. CP at 100 ng/mL (CP molar concentration 10-fold less compared with 50 ng/mL IGF-1) was added to the cells, and the cells were incubated in medium in the presence or absence of serum for 48 h. As shown in Fig. 1C, the negative control cells were unaffected, whereas all ES cell lines respond to CP treatment with decreased cell number, with the inhibition rate ranging from 15% in CADO to 30–40% in RDES, SKNMC, and A673. Surprisingly, CP alone consistently decreases cell proliferation in the absence of serum in all ES cell lines. Some of the ES cell lines tested were previously demonstrated to express IGF transcripts (29); thus, inhibition of the IGF-1 autocrine loop might partially explain this effect. However, the absence or very low levels of P-IGF-1R and P-AKT (Fig. 1B) in serum-free conditions do not support this mechanism and suggest that inhibitory CP effects on cell proliferation are essentially not dependent on competition with the ligand.

Fig. 1.

Effects of CP treatment on cell viability in ES cell lines. (A) Cell lysates were prepared from the indicated cell lines and analyzed by WB for total IGF-1R expression and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and displayed as a percentage of the total IGF-1R level in the CADO cell line. Data correspond to the mean ± SEM from three independent experiments. (B) IGF-1R activity was analyzed in the indicated cell lines by measuring IGF-1–induced receptor auto-phosphorylation and the subsequent activation of the IGF-1R downstream signaling pathways MAPK/ERK and PI3K/AKT. Cells serum starved for 12 h were stimulated or not with IGF-1 (50 ng/mL) for 10 min, and protein lysates were analyzed by WB for phosphorylated IGF-1R (P-IGF-1R), phosphorylated ERK (P-ERK), phosphorylated AKT (P-AKT), ERK, AKT, IGF-1R, and GAPDH. (C) Indicated cells were treated without and with 100 ng/mL CP for 48 h in the absence (SFM) or presence of 10% FBS (serum) and cell viability assayed by PrestoBlue reagent. Number of viable cells following CP treatment is displayed as percentage of untreated control. Data correspond to the mean ± SEM from three independent experiments. Statistical analysis compared with SKBR3: *P < 0.05, **P < 0.01, **P < 0.001.

Mechanism of CP-Induced IGF-1R Down-Regulation: β-Arrestin1 Recruitment and Receptor Ubiquitination.

The next experiments were designed to investigate in detail the CP effects on IGF-1R down-regulation. To avoid the competition between CP and IGF-1 normally present in serum, all experiments were performed in serum-free media (SFM). ES cell lines, serum starved for 12 h, were treated with CP concentrations of 100 ng/mL or 1 µg/mL for 24 h, and cell lysates were analyzed for IGF-1R expression using GAPDH as a loading control. As shown in Fig. 2A, all ES cell lines down-regulate IGF-1R in a dose-dependent manner in response to CP. When IGF-1 and CP were added together, the IGF-1R degradation pattern is similar with the one induced by CP alone, consistent with the reported higher affinity of antibodies for the IGF-1R (10) (Fig. S1). In a time-response experiment, comparing the CP and IGF-1 effects on receptor down-regulation (molar concentration CP ∼10-fold less than IGF-1), CP was proven to be more efficient at inducing receptor degradation in four ES cell lines: SKES, RDES, CADO, and SKNMC. Similar, very fast rates of receptor degradation by both CP and IGF-1 were observed in A673. These trends were confirmed by densitometric quantification of multiple experiments (Fig. 2B, graphs).

Fig. 2.

Mechanism of CP-induced IGF-1R down-regulation: β-arrestin1 recruitment and receptor ubiquitination. (A) Cells incubated in serum-free medium for 12 h were treated with 0, 0.1, or 1 µg/mL CP for 24 h. Protein lysates were analyzed by WB for IGF-1R and GAPDH as a loading control. (B) Cells were incubated in serum-free medium for 12 h and stimulated with 100 ng/mL CP or 50 ng/mL IGF-1 for indicated times. Protein lysates were analyzed by WB for IGF-1R and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and expressed as a percentage of the IGF-1R at 0 h. Data correspond to the mean ± SEM from three independent experiments. (C) Cells were incubated in serum-free medium for 12 h and were unstimulated or stimulated with either IGF-1 or CP for 10 min. IGF-1R was immunoprecipitated from lysates and analyzed by WB (IB) for ubiquitination (Ub) and IGF-1R as a loading control. Molecular weight markers in kDa are indicated to the left of the panels. (D) β-Arrestin1 (β-arr1) was immunoprecipitated from cell lysates prepared as for C. Immunoprecipitated proteins were analyzed by WB for IGF-1R. The whole lysates were analyzed by WB for GAPDH as a loading control.

Because the major outcome of ligand binding to the IGF-1R is receptor ubiquitination (21, 22, 30) “tagging” the receptor for degradation, to identify the mechanism underlying the effects of CP on IGF-1R degradation, we investigated receptor ubiquitination. The IGF-1R was immunoprecipitated from serum-starved ES cells, which had been stimulated or not stimulated for a short time (10 min) with CP or IGF-1, and the ubiquitinated receptors were detected by WB. Ligand-dependent ubiquitination of the IGF-1R was clearly detected, and CP was more potent in generating receptor ubiquitination in all investigated cell lines (Fig. 2C), consistent with receptor down-regulation experiments (Fig. 2B). Given that β-arrestin1 (β-arr1) is a key protein involved in IGF-1R ubiquitination (31), we also investigated the ligand-induced IGF-1R/β-arr1 association. As demonstrated in Fig. 2D, CP was much more potent than IGF-1 in recruiting β-arr1 to the receptor in all ES cell lines. Taken together, these experiments demonstrate that CP stimulates β-arr1 recruitment to the receptor with subsequent receptor ubiquitination and degradation.

β-Arrestin1 Dependence of CP-Induced IGF-1R Degradation.

Having shown that CP results in β-arr1/IGF-1R association, we next asked whether there is a causative relationship between this process and receptor degradation. To answer this, we used MEF cells expressing or not expressing β-arr1(β1KO), β-arr2, or both. Cells were serum starved, and the effects of CP or IGF-1 stimulation on IGF-1R degradation (Fig. 3A) and cell survival (Fig. 3B) were measured after 24 and 48 h, respectively. In control cells expressing both β-arr isoforms and in the cells expressing only the β-arr1 isoform (β2KO), CP or IGF-1 treatment drastically down-regulated the receptor, whereas in the MEF cells without β-arr1 (β1KO or double KO: β1/2 KO), this effect was almost completely abolished. This β-arr1 dependency was also displayed in the cell survival experiments: the β1KO and β1/2KO cells were insensitive to CP treatment, whereas the corresponding WT and β2KO cells containing β-arr1 responded with a 30–40% inhibition rate, regardless of the presence or absence of serum (Fig. 3B).

Fig. 3.

β-Arrestin1 dependence of CP-induced IGF-1R degradation. MEF and MEF KO for β-arrestin 1 (β1KO), β-arrestin 2 (β2KO), and both (β1/2KO) (A and B) and MEF and MEF expressing truncated IGF-1R, defective in binding β-arr1 (Δ1245) (C and D) were analyzed for effects of CP. (A and C) Cells were incubated in serum-free medium for 12 h and were unstimulated (SFM) or stimulated with 100 ng/mL CP or 50 ng/mL IGF-1 for 24 h. Protein lysates were analyzed by WB for IGF-1R and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and displayed as a percentage of the IGF-1R at 0 h. Data correspond to the mean ± SEM from three independent experiments. (B and D) Cell viability of the CP-treated cells for 48 h, in the absence (SFM) or presence of serum, was evaluated by PrestoBlue reagent. Numbers of viable cells following CP treatment are displayed as percentage of untreated controls. Data correspond to the mean ± SEM from three independent experiments.

Previous data indicate that an IGF-1R truncated at position 1245 (Δ1245) lacks the ability to bind β-arr (32). To fully validate β-arr1 as a key mediator of CP-induced IGF-1R down-regulation, we used an alternative experimental model of MEF cells expressing full-length, WT IGF-1R and MEF cells KO for IGF-1R (R−) stably transfected with the C-terminal–truncated Δ1245 IGF-1R (Fig. 3C). Over 48 h, the truncated IGF-1R, which is defective in binding β-arr1, was resistant to CP- or IGF-1–induced degradation, whereas full-length IGF-1R, expressed in the same cellular background, displayed a time-dependent degradation rate, with CP being more efficient than IGF-1, even at a 10-fold lower molar concentration. In line with the results described in the ES models, a decrease in cell number parallels the CP-induced IGF-1R down-regulation, with the MEF cells expressing truncated IGF-1R being essentially unresponsive (Fig. 3D). Taken together, these experiments validate β-arr1 as a key molecule controlling the CP-induced IGF-1R down-regulation.

β-Arrestin1 Enhances CP-Induced IGF-1R Down-Regulation and Inhibition of Cell Proliferation.

As β-arr1 plays an essential role in CP-induced IGF-1R down-regulation, we next explored whether β-arr1 overexpression could enhance CP effects on ES cells, with regards to IGF-1R down-regulation and overall cell survival. This experiment was done by CP treatment of cells transiently transfected with different amounts of β-arr1-flag plasmid. As demonstrated in Fig. 4A, and in line with previous studies reporting the β-arr1 involvement in ubiquitination and degradation of the IGF-1R (31), in the absence of the ligand, β-arr1 overexpression down-regulates IGF-1R expression in a dose-dependent manner. Nevertheless, increased β-arr1 expression potentiates CP-induced receptor degradation and enhances the CP-induced inhibition of cell proliferation/survival (Fig. 4B). Intriguingly, the clear β-arr1 dose-dependent decrease of IGF-1R expression and cell proliferation by CP was not observed in cells expressing the lowest amount of exogenous β-arr1, pointing to a possible increased proliferation by CP after small increases in β-arr1 level.

Fig. 4.

β-Arrestin1 enhances CP-induced IGF-1R down-regulation and inhibition of cell proliferation. (A) Cells transfected with different amounts of plasmid encoding β-arrestin1-flag (β1-flag) as indicated were treated without or with 100 ng/mL CP for 24 h. Protein lysates were analyzed by WB for IGF-1R, β-arrestin 1 (β-arr1), and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and expressed as percentage of mock transfected, unstimulated controls. Data correspond to the mean ± SEM from three independent experiments. (B) Cells transfected as in A were treated with 100 ng/mL CP for 48 h. Numbers of viable cells are displayed as percentage of mock transfected, unstimulated control. Data correspond to the mean ± SEM from three independent experiments.

CP-Induced β-Arrestin1–Mediated IGF-1R ERK Signaling Activation.

Previous reports demonstrated β-arr1 as a mediator of IGF-1R signaling and cell cycle progression (32); therefore, in the next experiments, we explored the possible agonistic properties of CP, secondary to β-arr1 recruitment. The roles of CP on IGF-1R signaling in ES cells were investigated by close monitoring of the dynamics of IGF-1– or CP-mediated activation of the two key downstream IGF-1R signaling pathways, the Ras/Raf/mitogen activated protein kinase kinase (MEK)/ERK pathway and the PI3K/AKT pathway, after short time stimulation. Serum-starved cells were stimulated with IGF-1 or CP (molar concentration of CP ~10-fold less than IGF-1), for up to 60 min before analyzing by WB. On IGF-1 stimulation, the IGF-1R activation loop was phosphorylated within 2 min, demonstrating an increase in its kinase activity. Consequently, both main downstream signaling pathways were activated as demonstrated by ERK and AKT phosphorylation (Fig. 5A). In the case of CP stimulation, IGF-1R and AKT phosphorylation were undetectable; however, clear ERK phosphorylation signals induced by CP were displayed in all ES cell lines. ERK activation levels were generally lower compared with IGF-1–mediated signaling activation, suggesting ERK phosphorylation independent of the IGF-1R kinase activity, possibly through a β-arr–mediated mechanism (32). To confirm this possibility, we again used the MEF cells expressing or not expressing the two β-arr isoforms and the MEF expressing the β-arr binding defective IGF-1R. As demonstrated in Fig. 5B, CP failed to activate the ERK phosphorylation in the absence of β-arr1, whereas the same pathway was clearly activated in the WT and the β2KO cells. Surprisingly, the ERK signaling was activated by CP in the double β1/2KO cells and greatly amplified in the cells expressing β-arr1 only (β2KO). The R− cells expressing truncated IGF-1R, unable to bind β-arr1, were insensitive to CP-induced ERK activation. Taken together, these experiments demonstrated the partial agonistic properties of CP, mediated by IGF-1R and β-arr1.

Fig. 5.

CP-induced β-arrestin1–mediated IGF-1R signaling activation. (A) Cells were incubated in serum-free medium for 12 h and then treated with either 50 ng/mL IGF-1 or 100 ng/mL CP for 0, 2, 5, 10, 30, or 60 min. Protein lysates were analyzed by WB for phosphorylated ERK (P-ERK), phosphorylated AKT (P-AKT), phosphorylated IGF-1R (P-IGF-1R), and GAPDH as a loading control. (B) MEF and MEF KO for β-arrestin1 (β1KO), β-arrestin2 (β2KO), and both (β1/2KO) or MEF with truncated IGF-1R (Δ1245) was incubated in serum-free medium for 12 h and then unstimulated (SFM) or stimulated with either 50 ng/mL IGF-1 or 100 ng/mL CP for 10 min. Protein lysates were analyzed by WB for phosphorylated ERK (P-ERK) and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and expressed as percentage of ERK activation relative to WT MEF. Data correspond to the mean ± SEM from three independent experiments.

β-Arrestin1 Dichotomy in Mediating CP Effects on Cell Proliferation/Survival: Therapeutic Implications.

From the above data, we hypothesized that lowering β-arr1 levels would protect against CP-induced IGF-1R down-regulation with effects on cell survival; therefore, we performed a rescue experiment by decreasing the β-arr1 levels in our panel of cell lines. The ES cell lines were stably transfected with a doxycycline-inducible shRNA β-arr1 knockdown system, incubated with varying doses of doxycycline for 4 d and then treated without and with 100 ng/mL CP. The experimental data were collected at different times as follows: T₀, before CP treatment, analyzing cell number and β-arr1 expression; T_10m, 10 min after CP addition for ERK activation; T_12h, 12 h after CP treatment for IGF-1R expression; T_48h, for cell viability. The data collected at T₀ validated the system and confirmed a doxycycline dose-dependent decrease in β-arr1 (Fig. 6A). In addition, characterization of the system at T₀ demonstrated that the reduced β-arr1 induces an overall dose-dependent loss of cell viability (Fig. 6A).

Fig. 6.

Effect of β-arrestin1 shRNA and MEK inhibitor on CP-induced cell viability reduction. (A) Indicated cells were stably transfected with doxycycline inducible β-arrestin1 shRNA and treated with doxycycline for 4 d. The cell viability was measured by PrestoBlue reagent and displayed as a percentage of doxycycline-untreated cells. Cell lysates were prepared from the same samples and analyzed by WB for β-arrestin1 (βarr1) and GAPDH as a loading control. Signals were quantified by densitometry, normalized to GAPDH, and presented as a percentage of the doxycycline-untreated cells. Data correspond to the mean ± SEM from three independent experiments. (B) Cells prepared as in A were treated without or with 100 ng/mL CP for 48 h, and the cell viability was assayed by PrestoBlue reagent. The inhibition ratio (quotient between CP-treated and CP-untreated cells) was calculated for each doxycycline dose and displayed as percentage of CP-untreated cells. Data correspond to the mean ± SEM from three independent experiments. During the experiment, cell lysates were collected at 10 min after CP stimulation and analyzed by WB for P-ERK and total ERK as a loading control, and at 12 h after CP stimulation and analyzed by WB for IGF-1R and GAPDH as a loading control. Signals were quantified by densitometry, normalized to the loading control, and presented as a percentage of the CP-untreated cells. Data correspond to the mean ± SEM from three independent experiments. (C) Indicated cells were pretreated for 60 min without or with the ERK inhibitor U0126, stimulated without or with 100 ng/mL CP. The cell viability assayed 48 h after CP treatment is displayed relative to untreated control. Data correspond to the mean ± SEM from three independent experiments. The percentage inhibition of CP-treated cells is displayed relative to CP-untreated cells.

The proportion of cells inhibited by CP treatment at T_48h, described by the CP ratio (CP treated versus untreated) confirmed our hypothesis, showing that down to a certain level of β-arr1, the cells are more resistant to CP (Fig. 6B). However, further decreases in β-arr1 reverse this protection, generating a bell-shaped curve for cell viability through decreasing β-arr1 levels in all five cell lines. The data collected at T_12h for IGF-1R expression (Fig. 6B) confirmed that β-arr1 down-regulation prevents CP-induced IGF-1R degradation, whereas results collected at T_10m verified lower CP-induced ERK activation following β-arr1 inhibition (Fig. 6B). The relative changes between P-ERK and IGF-1R levels appear to correlate with the peak of the CP cell viability inhibition rate (Fig. 6B), highlighting the dual role of β-arr1: receptor down-regulation and signaling activation. This experiment suggests that the biased activation of the ERK pathway by CP protects, to a certain extent, against reduction of IGF-1R expression and cell viability by CP. To verify this hypothesis, we selectively inhibited the ERK activation by adding MEK inhibitor U0126 in combination with CP stimulation to uncouple the protective effect of ERK activation from the detrimental IGF-1R down-regulation. The MEK inhibitor effectively prevented CP-induced ERK activation (Fig. S2). Combining treatment of CP with the MEK inhibitor demonstrated increased sensitivity over CP alone, confirming the protective role of CP-biased agonism and validating one approach to circumvent it (Fig. 6C).

Discussion

The classical paradigm describes the IGF-1R as active or inactive, with the natural ligand IGF-1 stabilizing the “on” state. Based on this paradigm, antagonist antibodies were designed to block the ligand binding, keeping the receptor in the “off” state. All IGF-1R targeting antibodies in clinical trials achieved the primary objective of preventing the ligand–receptor interaction. However, the experimental evidence showing that they also induce receptor down-regulation challenges the “switched off” model and identifies a contradiction with their intended antagonistic effect. Receptor down-regulation, a typical feature of agonists, has not been questioned despite having been described for all IGF-1R targeting antibodies (including CP) that made it to clinical trial (8, 33–36) and in all experimental settings including cell systems, animal models, and in clinical samples from patients receiving such therapy. Hence, as the first key finding, the present study elucidates the mechanism of CP-induced receptor down-regulation and identifies β-arr1 as the main mediator. Three lines of evidence support this conclusion: first, we demonstrated by coimmunoprecipitation that CP induced IGF-1R/β-arr1 association and subsequent receptor ubiquitination; second, we showed that CP-mediated IGF-1R degradation was enhanced by β-arr1 overexpression in a dose-dependent manner; and finally, we established that CP-induced IGF-1R degradation is prevented when the β-arr1/IGF-1R interaction is inhibited (C-truncated IGF-1R) or β-arr1 is decreased (KO cells, inducible shRNA system).

In addition to receptor degradation, β-arr1 binding to IGF-1R is sufficient to trigger activation of the MAPK/ERK signaling pathway, despite the absence of IGF-1R kinase activity. Accordingly, the second key finding of this report is identification of the agonistic properties of CP, with unambiguous activation of ERK signaling as an outcome of β-arr1 recruitment to the IGF-1R. These results were confirmed through a range of experimental models: ES cell lines expressing endogenous levels of β-arr1 and MEF cells lacking β-arr1 or expressing IGF-1R unable to bind β-arr1, all confirming CP-induced ERK activation relies on functional β-arr1 interaction with the IGF-1R. Additional support was provided by functional studies with β-arr1 shRNA under doxycycline control, revealing the dose-dependent effects of β-arr1 on CP-induced IGF-1R signaling activation.

Shifting the Paradigm for IGF-1R Signaling.

The corollary of this study is that the classical model of IGF-1R activation is both oversimplified and insufficient to explain the effects and therapeutic outcomes of IGF-1R targeting antibodies. During the drug development process, antibodies were designed to achieve the maximum binding to IGF-1R (to compete with the natural agonists) and the best specificity (to avoid binding to the insulin receptor). In subsequent drug screening for clonal selection, the assays were limited to detection of IGF-1R kinase activity (activation loop phosphorylation) and downstream kinase-dependent signaling such as PI3K/AKT or IRS phosphorylation. Such assays, however, do not take into consideration other IGF-1R functional responses such as β-arr1–mediated signaling. By systematically exploring these pathways, we confirmed that biased agonism (24–27, 37, 38), a model well known and largely accepted for the GPCR, can likewise occur for the RTK IGF-1R. In this paradigm shift, agonist binding to the IGF-1R can initiate several functional responses including signaling activation (both kinase- and β-arr1–dependent) and receptor internalization and degradation, all of them contributing to the biological response. The IGF-1 will activate these functional responses in a balanced manner, whereas a biased agonist (like CP) would favor one response over another (e.g., β-arr1 over kinase). This model accommodating all experimental data described for the IGF-1R targeting antibodies is also fully supported by our previous findings demonstrating that IGF-1R signaling is not exclusively dependent on its kinase activity and can be activated in a biased manner via β-arr1 by IGF-1R inhibitors or by natural biased agonists (39–42).

The theoretical background for this model was defined by our recent study demonstrating that the conformation of the IGF-1R that activates the kinase cascade can be distinct from that which interacts with β-arrs, as verified by IGF-1Rs mutated to constitutively bind β-arr1, which both trigger ERK signaling and are degraded in the absence of the ligand (43).

Therapeutic Implications for Cancer Treatment.

Despite big hopes regarding the use of anti-IGF-1R for cancer treatment and in particular for ES, the outcomes of phase III clinical trials have been rather disappointing (8). Several clinical trials were stopped due to futility, whereas some pharmaceutical companies are aborting their programs for developing IGF-1R inhibitors. Moreover, based on the clinical trial results with targeting antibodies, the value of IGF-1R as a target for cancer therapy is questioned. Possible reasons for the discouraging results with IGF-1R antibodies have been discussed in detail (8); yet, the present study highlights another potential cause while simultaneously validating a solution: agonistic properties of IGF-1R targeting antibodies counteracted by the use of MAPK inhibitors. The protective effect of the biased β-arr1 signaling was clearly demonstrated in all ES cells in which β-arr1 levels were modulated by inducible shRNA and further confirmed using MAPK inhibitors. In this context and taking into consideration that the IGF-1R C terminus has been proven essential for malignant transformation (18), whereas ectopic competitive expression of the C-terminal domain is inhibitory to tumor cell survival (18), identification of β-arr1 binding sites within this domain deserves particular consideration (43). These findings are remarkable for the fact that truncation of this domain, although preserving the RTK activity, abrogates receptor transforming abilities (18), clearly indicating that kinase activity is not enough to sustain the malignant phenotype. Conversely, our data suggest that inhibition of RTK signaling by CP, while preserving the C-terminal activity, might not be enough to reverse the malignant phenotype.

With the present study, we reveal β-arr1 as a critical regulator of antibodies targeting the IGF-1R. These results have led us to propose a paradigm shift in which targeting antibodies act as biased agonists for the IGF-1R. Defining and controlling such biased signaling represents a practical therapeutic strategy to enhance response to anti-IGF-1R therapies for all types of cancer relying on the IGF-1R.

Materials and Methods

Recombinant human IGF-1 and doxycycline were from Sigma. CP-751871 was a kind gift from Pfizer. The U0126 [1,4-diamino-2,3-dicyano-1,4-bis (2-aminophe-nylthio butadiene)] MEK1/2 inhibitor was from Calbiochem. Cell lines, other materials, and detailed procedures for Western blotting analysis, cell viability assay, the generation of constructs, doxycycline-inducible shRNA cell lines, and immunoprecipitation are described in SI Materials and Methods.