Inhibition of G Protein–Coupled Receptor Kinase 2 Promotes Unbiased Downregulation of IGF1 Receptor and Restrains Malignant Cell Growth

When activated by their natural, balanced agonist, G protein–coupled receptors (GPCR) modify their conformations to engage various signaling proteins, including the heterotrimeric G proteins and the multifunctional adapter proteins β-arrestins (β-arrs). Subsequently, a plethora of functional responses are generated in a balanced manner, including receptor trafficking and modulation of diverse signaling pathways. The concept of unbalanced, biased signaling (sometimes referred to as allosteric modulation or functional selectivity) recognizes the ability of distinct ligand–receptor pairings to stabilize particular receptor conformations that selectively engage only a certain subgroup of signaling proteins, preferentially activating a subset of biological events (1–4). The functional selectivity can be achieved by biased ligands (e.g., agonists that preferentially activate certain signaling pathways), biased receptors (e.g., specific mutations preventing β-arr/receptor interaction), or system bias (relative expression levels of transducers; refs. 5, 6). Notably, biased responses can unbalance any particular task of the receptor such as degradation versus recycling, or G protein versus β-arr–dependent signaling (7, 8).

Originally described for the largest and most “druggable” class of plasma membrane receptors, GPCRs, the biased-ligands/receptors paradigm is now recognized in the context of the smaller family of cell surface receptor tyrosine kinases (RTK; refs. 9–12). Among them, the insulin-like growth factor 1 receptor (IGF1R) is known to be involved in all cancer hallmarks and altered in the vast majority of cancers types (11, 13–16). As any other RTK, the IGF1R was classically described as a binary ON/OFF system, whereby the ligand stabilizes the “ON” state, followed by exclusively kinase-dependent, balanced downstream signaling activation. Yet, over the last decades, essential GPCR effectors such as G proteins and β-arrestin1, ARRB1 (β-arr1) or β-arrestin2, ARRB2 (β-arr2) were shown to be key mediators of IGF1R internalization, trafficking, and signaling (11, 17–20). β-arrs serve as adaptors to bring the E3 ubiquitin ligase Mdm2 to the ligand-activated IGF1R, triggering receptor ubiquitination and degradation (21–24). Of critical importance, the β-arr–engaged receptor conformation not only activates IGF1R downregulation but simultaneously initiates new signaling waves, independent of the kinase domain of the receptor (23). In the case of receptor activation by its natural ligands, kinase and β-arr-signaling occurred in a balanced manner (25), whereas some therapeutic agents aiming to inhibit IGF1R display a striking bias toward β-arr1-signaling (β-arr-BS; refs. 26, 27). Such behavior was demonstrated for the IGF1R–targeting monoclonal antibody CP-751871 (CP), as a potential cause for the failure of anti–IGF1R therapy in Ewing's sarcoma (ES; ref. 27). Although preventing the ligand/receptor interaction and kinase signaling as planned, CP acts as a “biased” IGF1R agonist, preferentially activating cancer-protective β-arr-BS. As is the case with GPCRs, the IGF1R's ability to recruit and activate β-arrs is dependent on a specific serine phosphorylation pattern or “barcode” (28, 29), generated by G protein–coupled receptor kinases (GRK; ref. 30). Although GRK6 promotes a stable β-arr1/IGF1R association, sustained β-arr1–mediated ERK signaling, and receptor degradation, GRK2 endows a weaker β-arr1/IGF1R association, transient ERK signaling with less receptor downregulation (20, 30).

Anticancer strategies aiming to inhibit the kinase activity of the receptor have been long recognized to be unsuccessful due to their inefficiency in removing IGF1R from the cell surface. Most antibodies targeting IGF1R prevent kinase activation and downregulate the receptor, yet an essential bottleneck is to untangle downregulation from associated biased signal cascades. Therefore, in this study, we aimed to investigate the contrasting abilities between GRK2 and GRK6 isoforms in controlling IGF1R trafficking as a potential target to achieve unbiased IGF1R downregulation in cancer cells.

Paroxetine hydrochloride hemihydrate (PX) from Sigma Aldrich was dissolved in DMSO at 100 mmol/L, stored at −20°C, and prepared to required concentration in serum-free media (SFM) prior to each experiment. IGF1R–targeting monoclonal antibody αIR3 was from Sigma Aldrich. IGF1R–targeting antibody CP 751871 (CP) was a kind gift from Pfizer and was previously described (27). All other reagents were purchased from Sigma Aldrich unless otherwise stated.

The human embryonic kidney cell line HEK293T; ES cell lines A673, CADO, RDES, SKES, and SKNMC; and the osteosarcoma cell lines U2OS and SAOS2 were obtained from the ATCC (via LGC Promochem). HEK293T and A673 were grown in DMEM, CADO and RDES in RPMI, SKES in McCoy's 5A, and SKNMC, U2OS, and SAOS2 in IMDM medium. Mouse embryonic fibroblasts (MEF) wild-type (WT) and knockout for β-arrestin1 (KOβ1) were obtained from professor R.J. Lefkowitz (Duke University, Durham, NC; refs. 31, 32) and cultured in DMEM. MEF with an IGF1R–null (Igf1r-null) background cells (R-) stably transfected with IGF1R with C-terminal tail truncation at residue 1245 (MEF IGF1RΔ1245/ΔCT) were obtained from professor R. Baserga (Thomas Jefferson University, Philadelphia, PA; ref. 33) and cultured in DMEM, under constant antibiotic selective pressure of 100 μg/mL G418. All media were supplemented with 10% (vol/vol) FBS and 1% penicillin/streptomycin. The human cell lines were validated by short tandem repeat profiling of extracted genomic DNA (Uppsala Genome Centre, Uppsala, Sweden or Eurofins Genomics) once per year, prior to freezing down stocks, for the entire duration of the project. Mouse cell lines were authenticated by expression of β-arrestin1/truncated IGF1R. All cell lines were tested for Mycoplasma contamination once per year (Eurofins Genomics) for the entire duration of the project. For all in vitro experiments, cells from these frozen stocks were used within 25 passages. For the animal experiments, a new batch of low passage A673 was obtained from the ATCC and used directly to exclude any extensive passaging effects.

All siRNAs were purchased from Thermo Fisher Scientific. The Silencer Select validated siRNA ID s1127 (set a) and s1128 (set b) were used for targeting human GRK2, whereas ID s6091 (set a) and s6092 (set b) were used to deplete human GRK6. β-Arrestin1 (ARRB1) downregulation was achieved using the ON-TARGET plus pool of four sequences (assay ID 408) and validated using a single sequence from the pool (J-011971–05 5′UGGAUAAGGAGAUCUAUUAtt-3′). Cells were siRNA transfected at 70% confluency in 6-well plates using Lipofectamine RNAiMAX (Invitrogen, Thermo Fisher Scientific) according to the manufacturer's instructions, using nontarget siRNA transfections as mock controls. GRK2 and GRK6 in pcDNA3 were a gift from professor V.V. Gurevich (Vanderbilt University, Nashville, TN). Cells were plasmid transfected at 50% confluency in 6-well plates using Turbofect (Thermo Fisher Scientific) according to the manufacturer's instructions, using empty vector (pcDNA3) as mock controls. All transfection experiments were verified for efficiency by Western blot (WB).

Prior to IGF1 stimulation experiments (long-term receptor degradation or short-term signaling), adherent cells were washed twice with PBS, changed to SFM, and incubated at 37°C for 4 to 8 hours. Stimulation was performed using recombinant human IGF1 (Sigma Aldrich) at 50 ng/mL for indicated times.

Protein samples were dissolved in lithium dodecyl sulfate sample buffer (Invitrogen, via Thermo Fisher Scientific) and analyzed by SDS-PAGE with 4% to 12% Bis-Tris gel (Invitrogen). Upon separation, proteins were transferred to nitrocellulose membranes at appropriate voltage for 1 hour. Membranes were then blocked for 1 hour in BSA and 0.1% Tween 20 in tris-buffer saline (TBS). Primary antibody in BSA was incubated overnight at 4°C. Following 3 × 10 minutes of washing (TBST), membranes were incubated with secondary antibody, either with fluorescent-conjugated IRDye (LI-COR Biosciences) and detection with LI-COR Odyssey, or horseradish peroxidase conjugation and chemiluminescence detection with ECL substrate (Pierce via Thermo Fisher Scientific) and exposure to X-ray film. All antibodies used in the study are presented in Supplementary Table S1.

Densitometry calculations were carried out using Image Studio (LI-COR Biosciences) or ImageJ (NIH, USA) and displayed relative to their respective loading control across multiple experiments.

RNA was extracted from cells using Ambion PureLink RNA Mini Kit (Life Technologies) and eluted in Nuclease Free Water (Ambion, Life Technologies). Twenty-five ng of RNA was used per qPCR reaction, and qRT-PCR was carried out using TaqMan RNA-to-CT 1-Step Kit (Life Technologies) and TaqMan Gene Expression Assays for GRK2, 6 and GAPDH. qRT-PCR was performed using Ambion STEPONEPLUS real time PCR system (Applied Biosystems). Relative RNA expression was evaluated by the ΔΔC_T method using GAPDH and RNA expression in untransfected samples as reference.

Cell viability was assayed with PrestoBlue cell viability reagent (Life Technologies, Thermo Fisher Scientific). Fluorescence was measured with excitation at 560 nm and emission at 590 nm using a TECAN Infinite 1000 plate reader. Fluorescent signal was converted to cell number using a standard curve for every cell line. IGF1 proliferation was calculated as change in the number of viable cells between stimulated and unstimulated groups. Cell viability after transfection was calculated as the number of viable cells in transfected groups compared with mock-transfected controls (empty vector/nontarget siRNA).

Tissue culture plates were coated with 200 μg/cm² of poly-2-hydrozyethylmethacrylate (poly-2-HEMA) solution dissolved in 95% ethanol, allowed to dry, washed twice with PBS, and UV-sterilized before use. Cells were plated into coated wells and cell number assayed using PrestoBlue (see above).

Cells cultured during their exponential growing phase were detached, counted, and plated at low confluency (10³ cells/plate) in 10 cm cell culture–treated petri dishes. Once attached, drugs were added to serum-containing media, and dishes were incubated until control groups (solvent only) had adequate colony formation (8–14 days). Following media aspiration, colonies were washed with PBS, fixed with methanol (3 mL/dish), stained with crystal violet for 15 minutes, washed, and colonies manually counted.

After indicated treatments/transfections, cells in 6-well plates at 80% to 90% confluency were lysed with 600 μL Pierce IP Lysis buffer (Thermo Fisher Scientific) containing protease inhibitor cocktail (Thermo Fisher Scientific). When detecting IGF1R ubiquitination, 10 mmol/L N-ethylmaleimide was added. Protein concentration was determined by bicinchoninic acid assay (Thermo Fisher Scientific). Equivalent protein amounts were incubated with 10 μL of anti-FLAG agarose beads overnight at 4°C, or 1 μg of IGF1R antibodies overnight at 4°C followed by 10 μL of Protein G agarose beads (GE Healthcare) for 2 hours at 4°C. Immunoprecipitates were collected by centrifugation, washed 4 times with lysis buffer, dissolved in SDS/PAGE sample buffer, and analyzed by WB. The FLAG-tagged protein and IGF1R levels in the lysates prior to precipitation, as well as the amounts of FLAG protein captured by the anti-FLAG beads were used as loading and transfection efficiency controls.

Cellular fractionation was carried out using the Qproteome Cell Compartment Kit (QIAGEN) according to the manufacturer's protocol.

Xenograft studies were approved by the MD Anderson Cancer Center (MDACC) Institutional Animal Care and Use Committee, and all animal care was in accordance with institutional guidelines. ES cells (A673) were cultured to a confluence of 75%, harvested with trypsin/EDTA, washed twice, and resuspended in PBS. Xenografts were inoculated in 5- to 8-week-old male nude mice by s.c. injection of 5 × 10⁶ cells in 0.1 mL sterile saline. Tumors were calliper measured every 4 days, and tumor volume was calculated using the formula V = π/6 × (larger diameter) × (smaller diameter)². At tumor mean volume of 65 mm³, mice were randomly assigned to receive i.p. either 10 mg/kg one dose daily for 8 days followed by 15 mg/kg one dose daily for 24 days (PX in 10% DMSO, 150 μL) or vehicle alone group, once daily for 32 days (10% DMSO in PBS, 150 μL). Injections were split between two sites to minimize local inflammation. Mice were monitored for side effects and sacrificed at 32 days after commencement of treatment, when control-treated tumors reached 1,000 mm³. Collected tumors were measured and split in two halves that were further processed either for histology (paraffin) or frozen for protein/RNA extraction.

Total protein samples were isolated from frozen tumor tissues using Tissue Protein Extraction Reagent (Thermo Fisher Scientific) containing protease inhibitor cocktail (Thermo Fisher Scientific) according to the manufacturer's protocol. Equal amounts of total protein were analyzed by WB.

Tumor tissue was immediately fixed in formalin and processed using standard histologic procedures for hematoxylin and eosin staining. IGF1R IHC was performed on freshly cut 4 μm sections, as described (34), using the IGF1R β (D23H3) antibodies (Supplementary Table S1). IHC staining was independently scored by two readers according to the University of Colorado IHC H-score criteria, with assessment of staining intensity (0–4) multiplied by the percentage of positive cells (0%–100%) for each category for a final IHC score of 0–400 (35). For each tumor, results from six different areas (three from each reader) were averaged to obtain the final IGF1R IHC score (35).

Where indicated, data from three independent experimental replicates of two conditions were compared using a two-tailed unpaired t test, assuming equal variance, or ANOVA plus post hoc analysis (Dunnett test) where appropriate, using GraphPad Prism (version 8.2.1). Experimental design included a threshold value of P = 0.05 for testing any null hypothesis. Data expressed with error bars indicate mean ± SEM from three independent experiments. Significance is given as *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

All data associated with this study are present in the paper or the Supplementary Materials.

For a cell population whose proliferation/survival relies on the IGF1R, any alterations in its expression or signaling are likely to manifest effects on overall cell viability. Thus, we evaluated the outcomes of GRK2/6 modulation on IGF1–induced proliferation. As an experimental model, we selected a panel of malignant ES cell lines, previously demonstrated to be dependent on the IGF1R (27), including as a reference model HEK293T (HEK), a well-characterized system for transgenic expression of GRKs (30). All tested cell lines expressed IGF1R, GRK2, and GRK6 (Supplementary Fig. S1A), and they were functionally responsive to IGF1 in terms of activation of the downstream pathways MAPK/ERK and PI3K/AKT, as well as resultant cell proliferation/survival (Supplementary Fig. S1B and S1C). Optimal experimental windows of 24- to 72-hour posttransfection were defined in the cell lines expressing the highest levels for GRK2 and GRK6 (HEK and RDES) by monitoring GRK expression kinetics after plasmid overexpression or siRNA (Supplementary Fig. S1D and S1E) and used to interrogate the effects on IGF1–induced proliferation.

The GRK2/6 status quo in all cell lines was unbalanced by transfecting either GRK2/6 plasmid or siRNA, and confirmed by WB 24 hours after transfection (Supplementary Fig. S2A). Cells in SFM were stimulated with IGF1 and evaluated for total cell viability after another 24 hours. Depletion of GRK2 (GRK2⁻) or overexpression of GRK6 (GRK6⁺) was systematically detrimental to IGF1 response in all cell lines. GRK2⁻ brought cell number to near, or in some instances even below, unstimulated control levels (Fig. 1A). The opposing transfections, GRK2⁺ and GRK6⁻, showed overall less impact when compared with mock-transfected controls (Fig. 1A). When controlling for siRNA off-target effects in the cell lines expressing the highest GRK2/6 levels, different targeting sequences conveyed similar downregulation efficiency (Supplementary Fig. S2B) and near-identical effects on IGF1–induced proliferation (Supplementary Fig. S2C).

As the β-arr/GRK system plays an essential role in ligand-induced IGF1R downregulation, we next questioned whether GRK2/6 imbalance modifies the expression of the receptor. Ligand-induced IGF1R degradation, as monitored by WB, was increased in GRK2⁻/GRK6⁺ while GRK2⁺/GRK6⁻ have protective effects (Fig. 1B, transfection efficiency; Supplementary Fig. S2A). These trends were confirmed by densitometry quantification of multiple experiments (Fig. 1C) and validated with different siRNA sequences (Supplementary Fig. S2D). Notably, in all mock-transfected cells, the IGF1R levels decline in a linear manner in response to IGF1. However, in conditions that unbalanced the GRK2/6 ratio toward the latter, the degradation process is commonly initiated before experimental addition of the ligand, evident by decreased IGF1R levels at time 0 and confirmed across all cell lines by quantification of multiple experiments (Fig. 1D) and validated with two different siRNA sequences (Supplementary Fig. S2D). Consistent with the suppressed growth response to IGF1 stimulation, these results demonstrate that an imbalance toward GRK6 (GRK6⁺ or GRK2⁻) enhances receptor degradation, thus these two expression conditions were further explored for their impact on the IGF1R signaling. Given that alternative siRNA sequences infer no different effects, one set per target was used for the next experiments.

A key role of the β-arr/GRK system in regulating IGF1R function is to promote signaling by connecting the receptor with downstream cytoplasmic signaling complexes (18, 20, 23). For the larger class of GPCRs, as well as for the IGF1R, the stability of the receptor–arrestin interaction directs the fate of receptor throughout the endocytic pathway, with a clear relationship between early effects (0–60 minutes) on ligand-induced signaling and late effects (12–24 hours) on degradation (21, 22, 30). Enhanced stability of the receptor–arrestin complex initiates β-arr-BS, evident as sustained ERK activity (36). Thus, we next explored the consequences of GRK2 inhibition and GRK6 overexpression on the temporal dynamics of IGF1 signaling. Transfected cells (efficiency shown in Supplementary Fig. S2A) were stimulated with IGF1 for 0 to 60 minutes, and phosphorylated (p) levels of the receptor, AKT, and ERK1/2 were measured by WB as indicators of IGF1R–dependent signaling activation (Fig. 2A). When compared with mock-transfected cells, GRK6⁺ and GRK2⁻ result in decreased pIGF1R and pAKT in all cell lines, suggesting receptor internalization before stimulation. On the other hand, the pERK signal, while showing a moderate overall decrease, demonstrated sustained activity at late time points (30/60 minutes) in GRK6⁺/GRK2⁻ cells (Fig. 2A). Densitometry analysis of late (60 minutes) pERK signal relative to its maximal activation confirmed sustained ERK activity following GRK6⁺/GRK2⁻ in all cell lines (Fig. 2A, graph), indicative of an additional signaling wave initiation, likely through β-arr mechanisms (23, 30). To follow the functional impact of increased receptor degradation with sustained pERK signaling, we evaluated cell proliferation/survival. Although not an absolute requirement for many cells' growth in monolayer cultures, the IGF1R has been described to be quasiabsolute for growth in anchorage-independent conditions (15, 37); thus, we used poly-2-hydroxyethylmethacrylate (poly-2-HEMA)–coated plates to prevent adhesion in a serum-free supplemented with only IGF1 assay. Under these stringent conditions, four out of six mock-transfected ES cell lines were able to proliferate in the presence of IGF1 (Fig. 2B). GRK6⁺ and GRK2⁻ consistently reduced their proliferation/survival, bringing cell number below mock-unstimulated level (Fig. 2B). This loss of proliferation/viability suggests that the sustained (possibly β-arr biased) ERK signaling is not sufficient to compensate for the receptor downregulation.

Both GRK6⁺ and GRK2⁻ transfections initiate IGF1R downregulation, even in conditions with low ligand concentrations. As protein overexpression is not a feasible clinical approach, we streamlined our investigation to pharmacologic inhibition of GRK2. The selective serotonin reuptake inhibitor (SSRI) PX is used widely for the treatment of mood and anxiety disorders, and its mechanism of action was recently recognized as GRK2 kinase inhibition with an in vitro IC₅₀ of 1.4 μmol/L (38, 39). In animal models, PX treatment reduces heart failure by preventing GRK2-dependent β-adrenergic receptor desensitization (40, 41). PX was also tested as anticancer therapy in models of colon cancer and found to be effective on cell viability and anchorage-independent colony formation with an IC₅₀ in a range of 13 to 26 μmol/L and 5 μmol/L, respectively (42). For the ES cell lines in our panel, micromolar concentrations of PX decreased cell viability in a time- and dose-dependent manner, with IC₅₀s in a range of 15 to 27 μmol/L at 24 hours (Supplementary Fig. S3A). We selected a high concentration of 15 μmol/L, preserving at least 50% viable cells, and a low concentration of 2 μmol/L, high enough to inhibit GRK2 yet with negligible effects on cell viability at 24 hours, to verify the molecular mechanism in subsequent experiments. The low dose is also clinically relevant as toxicological tests demonstrate acceptable plasma concentrations of PX up to 3 μmol/L (43, 44). We investigated the consequences of PX treatment on IGF1R downregulation, signaling, and overall cell survival, alongside their dependency on β-arr1. HEK cells, expressing endogenous or siRNA-decreased β-arr1 levels (transfection efficiency; Supplementary Fig. S4A, left plot), were stimulated with IGF1 in SFM, in the presence or absence of PX. In untreated cells, β-arr1 silencing lessened IGF1–dependent degradation of the receptor, without affecting IGF1R expression at time 0 (Fig. 3A). On the other hand, PX treatment recapitulated siGRK2 effects by diminishing receptor levels, most prominently in low-ligand conditions, i.e., time 0 (Fig. 3A, graphs). This decrease, more evident at the high dose (15 μmol/L), was further enhanced following ligand stimulation. However, these effects were mitigated in conditions with siRNA-decreased β-arr1 (Fig. 3A). All these effects were consistent between a pool of siRNA sequences and one independent sequence (Supplementary Fig. S4B and S4C). Using the same experimental system and short (0–60 minutes) IGF1 stimulation times, we questioned whether IGF1R downregulation initiated by PX promotes biased activation of the ERK signaling pathway. Unlike GRK2⁻, PX treatment not only failed to sustain IGF1–induced ERK phosphorylation, but in fact decreased overall ERK activity (Fig. 3B). In the absence of β-arr1 (siRNA-transfected cells), 1 hour of PX treatment had no effects on IGF1–mediated ERK signaling (Fig. 3B). Remarkably, 1-hour PX treatment was enough to make the cells less responsive to IGF1 as verified by a reduction in IGF1R activation (phosphorylation) and pAKT levels (Supplementary Fig. S4D), indicative of a reduction of cell surface receptor expression. Once more, these effects were prevented in HEK cells with low β-arr1 expression (Supplementary Fig. S4D). The biological consequences of PX-induced IGF1R downregulation, evaluated using a cell viability assay, again indicate β-arr1 dependency, as siRNA-transfected cells were less affected by PX treatment then mock-transfected cells (Supplementary Fig. S4E).

To further substantiate the β-arr1 dependency of PX effects on IGF1R expression and signaling, we employed an alternative experimental system: MEFs expressing full-length IGF1R and β-arr1 (WT), alongside equivalent MEFs derived from mice carrying genetic knockout for β-arr1 (Arrb1, β1KO), or Igf1r KO cells stably transfected with a C-terminal truncated IGF1R (IGF1RΔCT; ref. 33). The IGF1RΔCT mutant is unable to bind β-arr1, the adaptor protein directing Mdm2-mediated IGF1R ubiquitination and degradation (21–23). Following validation of the β-arr1/IGF1R phenotype by WB (Supplementary Fig. S4), the effects of PX on IGF1R expression, ERK-biased signaling, and cell survival were measured after short (1 hour) or long (24 hours) IGF1 stimulation in SFM. In WT MEFs expressing both β-arr1 and IGF1R, 24 hours of PX treatment downregulated the receptor at time 0 and enhanced ligand-induced receptor degradation (Fig. 3C), whereas in the MEF β1KO or MEF IGF1RΔCT (deficient in binding β-arr1), these effects were almost completely abolished. Confirming the data obtained in HEK, PX treatment initiated IGF1R downregulation and overall dampened IGF1–induced ERK activation with no sign of biased signaling in WT cells (Fig. 3D), whereas these patterns were absent in β1KO or IGF1RΔCT MEFs (Fig. 3D). The β-arr1 dependency was also recapitulated in the cell survival experiments: the β1KO and IGF1RΔCT MEFs were less sensitive than WT to PX treatment (Supplementary Fig. S4E). Taken together, these experiments further substantiate the findings that PX-induced IGF1R–unbiased downregulation is dependent on a β-arr1/IGF1R interaction.

Previous studies have established β-arr recruitment to the IGF1R as a critical step controlling receptor trafficking and signaling through Mdm2-dependent ubiquitination (21–23). Therefore, the next experiments investigated in detail the PX effects on IGF1R ubiquitination and its dependency on Mdm2. As Mdm2 overexpression or silencing could modulate several other p53 and/or GRK2-dependent pathways, indirectly interfering with IGF1R signaling and trafficking (45, 46), we instead selected an experimental system of two human osteosarcoma cell lines, U2OS and SAOS2, endogenously expressing high or low Mdm2 levels, respectively (20, 45). Initial characterization of cell viability verified that 24-hour treatment with up to 20 μmol/L PX had no significant effect on SAOS2, whereas the U2OS were slightly affected, decreasing their number by 20% at the highest dose (Supplementary Fig. S5A). We used the same experimental conditions as with HEK/MEF, investigating the effects of low (2 μmol/L) or high (15 μmol/L) PX concentrations on IGF1R degradation and signaling. PX was more efficient at inducing receptor downregulation in high Mdm2-expressing U2OS, demonstrated by decreased IGF1R at time 0 after 2 μmol/L, whereas in SAOS2, similar effects were evident only after 15 μmol/L (Fig. 4A). In both osteosarcoma cell lines, IGF1–generated ERK signaling was not sustained longer in PX treatment cells as compared with control-treated cells (Fig. 4B), confirming the data obtained in HEK/MEF cells that PX-initiated IGF1R depletion does not convey biased MAPK signaling. Intriguingly, unlike HEK/MEF, 15 μmol/L of PX treatment in U2OS cells caused an early increased phosphorylation of ERK observed at time 0 (absence of exogenous ligand) when compared with untreated cells (Fig. 4B). Therefore, we explored the possible PX agonistic properties by monitoring dynamics of MAPK/ERK in cells treated with PX alone in SFM. In contrast to control-treated cells, PX clearly activates MAPK/ERK pathway in U2OS but not in low Mdm2-level SAOS2 (Supplementary Fig. S5B). To interrogate whether this ERK signaling is sustained (biased) or transient, we compared the pattern of pERK following stimulation with IGF1, PX, the neutral IGF1R blocking antibody “antagonist” clone αIR3 (47), or the prototypical IGF1R “biased agonist” antibody CP (27). The effects on IGF1R and AKT phosphorylation confirmed that only the natural (balanced) ligand IGF1 activates canonical kinase-dependent signaling (Supplementary Fig. S5C). In parallel, IGF1 strongly activates the MAPK pathway (pERK), whereas lower but clear kinase-independent (β-arr–dependent) pERK signals are evident after PX, αIR3, and CP treatment (Fig. 4C). As the ultimate biological effects of ERK activation depend on its membranar, cytoplasmic, or nuclear distribution (1, 7, 20, 48, 49), we used cell fractionation to follow the compartmentalization of sustained pERK. One hour after stimulation, sustained ERK signals (localized in the nucleus, cytoplasm, and membrane fraction) were displayed only in the case of IGF1 and CP, ruling out PX-induced biased signaling (Fig. 4C).

Previous studies demonstrate that IGF1R-kinase–independent, β-arr–biased-MAPK signaling depends on receptor ubiquitination (22, 23, 26). Considering the absence of ERK activation in cells expressing low levels of the E3 ligase Mdm2, the next logical step was to evaluate whether PX treatment itself leads to IGF1R ubiquitination. Untreated U2OS cells cultured in SFM had very low levels of receptor ubiquitination, which increased dramatically after 1-hour PX treatment. Consistent with the pattern observed for IGF1R degradation, PX increases IGF1R ubiquitination in SAOS2 cells at the highest dose only (Fig. 4D).

β-arr recruitment to the IGF1R is a critical step controlling Mdm2-dependent receptor ubiquitination. Thus, we next questioned how the two β-arr isoforms interact with the IGF1R following PX treatment and the ligand dependency of these interactions. We used U2OS, expressing functional Mdm2, sequentially transfected with the IGF1R and FLAG-tagged versions of either β-arr isoform. FLAG-tagged β-arr immunoprecipitates obtained from cells treated in SFM with and without IGF1/PX were analyzed by WB for the IGF1R (Fig. 4E). In the absence of IGF1, both β-arr isoform immunoprecipitates contained clearly detectable levels of IGF1R; however, the β-arr2/IGF1R interaction predominates. Following IGF1 stimulation, β-arr2 disengages while β-arr1 is recruited to the activated receptor. One-hour treatment with low-dose PX switches the receptor affinity to the other isoform: an increase in the association of the unstimulated receptor with β-arr1, which rapidly disengages from the receptor 2 minutes after ligand stimulation. The β-arr2/IGF1R interaction, decreased by PX treatment in the absence of the ligand, remains low after IGF1 stimulation (Fig. 4E).

Altogether, these results reveal that in conditions with low ligand, PX prevents basal β-arr2 recruitment while promoting a β-arr1/IGF1R interaction, which initiates Mdm2-dependent ubiquitination and subsequent IGF1R downregulation. In instances where Mdm2 is overexpressed, this IGF1R ubiquitination coincides with transient ERK activation.

So far, our results demonstrate that PX affects the expression, degradation, and signaling of the IGF1R: promoting receptor downregulation without activating β-arr–biased MAPK signaling. To this end, we verified whether ES cells exhibit a similar response, while also asking whether unbiased IGF1R downregulation would ultimately be a better therapeutic strategy than β-arr–biased. ES cell lines were treated with low (2 μmol/L) or high (15 μmol/L) PX, and the effects on IGF1R expression and signaling were measured with or without IGF1 stimulation in SFM. All ES cell lines downregulate IGF1R in a PX dose- and time-dependent manner (Fig. 5A). Comparing receptor levels in the absence of exogenous ligand (time 0), ES cell lines display very fast PX-induced receptor degradation (Fig. 5A, graph). One-hour treatment with the lowest and clinically relevant dose (2 μmol/L PX) was sufficient to initiate removal of the receptor from the cell surface, as demonstrated by considerably lower levels of pIGF1R in response to ligand (Fig. 5B). Importantly, IGF1–induced ERK signaling was not enhanced in PX-treated cells (Fig. 5B), endorsing the unbiased mechanism of IGF1R downregulation. Finally, we explored the impact of IGF1R targeting through GRK2 inhibition versus strategies associated with biased signaling on ES viability potential. We performed a clonogenic assay (50), comparing the long-term effects of PX versus CP, the anti–IGF1R antibody with biased agonist properties (27). Cells receiving a single dose of PX and incubated for up to 14 days to test the “unlimited division” of every cell in the population demonstrated a dose-dependent inhibition of their ability to produce progeny (Fig. 5C). Of note, in every ES cell line that produced viable and countable colonies in untreated controls, the PX inhibition was significant at the lowest tested and clinically relevant dose (2 μmol/L), whereas CP failed to produce effects at any tested dose (Fig. 5C). Taken together, these results demonstrate a clear benefit of targeting IGF1R through GRK2 inhibition over biased IGF1R downregulation.

Finally, we aimed to verify the mechanism of PX-induced inhibition of ES cell growth in animal models, and for this purpose, A673 xenografts were established in nude mice. When the tumors reached a volume of 65 mm³, mice were treated daily with PX (i.p. 10 mg/kg for 8 days followed by 15 mg/kg for 24 days) or drug-free vehicle (i.p. 150 μL 10% DMSO PBS solution for 32 days; Fig. 6A). The dosing schedule was designed to diminish unspecific toxicity, and whole blood samples and body weight were monitored as indicators of off-target effects of GRK2 inhibition. The tested PX treatment regimen had no effect on body weight (Supplementary Fig. S6A) and did not alter total white blood cell counts, segmented white blood cells, lymphocytes, large unstained cells, red blood cells count nor hemoglobin concentration; only platelet counts were increased by PX (P = 0.0215, Supplementary Fig. S6A). On the other hand and in line with the in vitro data, tumors immediately responded with growth inhibition, and by the end of the experiment, PX treatment led to greater than 50% reduction in tumor volume (438 ± 78.4 vs. 891.9 ± 183.2 mm³, P = 0.0429; Fig. 6B) and over 70% decreased tumor weight (165.6 ± 49.56 vs. 555 ± 146.9 mg, P = 0.0278; Fig. 6B; Supplementary Fig. S6B). The growth pattern of engrafted tumors was confirmed histologically. Routine hematoxylin–eosin staining verified the presence of sheets of small, round, uniform blue cells with scant and clear cytoplasm, divided into irregular lobules by thin amounts of stroma, consistent with typical ES growth (Fig. 6C). IHC analysis of IGF1R expression showed an extensive decrease of IGF1R staining after PX treatment (Fig. 6C), quantified as >50% reduction in IHC intensity score. A similar decline in IGF1R expression following PX treatment was confirmed by WB analysis of protein samples extracted from multiple xenografts (Fig. 6D). The same analysis demonstrates that PX treatment did not affect the insulin receptor (InsR), further verified by densitometry quantification displaying a similar decrease of IGF1R levels relative to either GAPDH or InsR (Fig. 6D). Taken together, these results confirm in vivo that GRK2 inhibition downregulates IGF1R through a mechanism that is selective over the InsR.

The IGF1R system is intricately linked with most signaling pathways promoting and sustaining the malignant phenotype, thus a strong rationale for targeting IGF1R in anticancer therapies exists. It is now generally accepted that downregulation of the receptor, and not only restraining its tyrosine kinase activity, is necessary to produce clinically relevant responses (51, 52). Despite this framework, all known anti–IGF1R strategies have been exclusively developed to prevent the classical kinase signaling cascade (e.g., blocking antibodies or kinase inhibitors). Hence, as the first main finding, we report a novel approach for targeting IGF1R in cancer. Unlike all other strategies developed so far, our paradigm of inducing system bias exploits the opposing roles of GRK2 and 6 in directing receptor trafficking, to initiate IGF1R downregulation from “inside” the cell. Experiments using controlled expression of either isoform unambiguously demonstrate the proof of concept that system bias toward GRK6 (GRK6⁺ or GRK2⁻) enhances IGF1R degradation, even in conditions with low ligand concentration. The opposing behavior of the two isoforms also manifests in relation to IGF1R signaling: receptor downregulation triggered by GRK6⁺ or GRK2⁻ generates sustained biased MAPK/ERK activity in response to IGF1 stimulation that could conceivably counteract receptor loss in terms of cell viability. In this context, another key finding of this study is that pharmacologic inhibition of GRK2 by PX promotes IGF1R downregulation without detectable protective β-arr biased signaling (β-arr-BS). Clinically relevant micromolar concentrations of PX decrease IGF1R expression in a time- and dose-dependent manner, effects that were critically dependent on an IGF1R/β-arr1/Mdm2 interaction. The downstream manipulation via GRK2 of the IGF1R trafficking, uncouples downregulation from its usual ligand requirements; in conditions with low/absent ligand, PX prevents the normal β-arr2 recruitment and instead supports a β-arr1/IGF1R interaction, which initiates Mdm2-dependent ubiquitination and subsequent IGF1R degradation, usually exclusive to ligand stimulation. PX-initiated IGF1R downregulation not only failed to sustain IGF1–induced ERK phosphorylation, but in fact decreased overall IGF1R activity. Although an overarching concern of oncological IGF1R targeting is concurrent inhibition of the closely related InsR and hence detrimental insulin resistance, PX depletes the IGF1R without affecting InsR levels demonstrating a high degree of selectivity. This is in line with the differential handling of the IGF1R/InsR by the four widely expressed GRK isoforms (53). In the case of the IGF1R, GRK2/3 oppose GRK5/6 on degradation, with the former preventing receptor downregulation (30), whereas for the InsR, GRK2 plays inhibitory roles on insulin-mediated glucose uptake (12) without interfering with receptor trafficking (23, 53).

Over the last decades, GRKs are increasingly acknowledged as important regulators of signal transduction pathways downstream of both GPCRs and RTKs. Among them, GRK2 has emerged as a critical signaling hub, orchestrating most—if not all—cancer hallmarks (54, 55). Preclinical data clearly demonstrate that GRK2 inhibition impairs tumor progression, and modified GRK2 levels are reported in various tumors (46, 47). However, clinical exploration of GRK2 as a therapeutic target is in its initial stages. In support, several studies describe the cancer-protective effects of PX not only in experimental models of lung, breast, or colon cancers (42, 56) but also in clinical settings. For instance, its extensive use as antidepressant raises questions surrounding cancer incidence associated with long-term use. As such it has been concluded that individuals prescribed PX had a reduced risk of ovarian, colorectal, kidney, or bladder cancer (57–60). Surprisingly, all of these studies are not considering the GRK2-inhibitory properties of PX but only the SSRI mechanism of action.

Based on the strength of the receptor/β-arr interaction (controlling intracellular trafficking and signaling), GPCRs are classified into two major classes (61): “Class A” receptors, with low-affinity receptor–β-arr complexes, are rapidly recycled and induce weak/transient ERK activity; “Class B” receptors engage β-arr in stable complexes that sustain ERK activity and limit the recycling process by directing the receptor toward a degradation route. For the ligand-activated IGF1R, clear correspondence between the effects of β-arr1/2 and GRK2/6 modulation advocates for a GRK2/β-arr2 or GRK6/β-arr1 functional partnership (Fig. 7). The GRK2/β-arr2^{Class A-like} configuration occurs at a basal level in absent/low ligand environments to preserve receptor homeostasis and limit inconsistent signals by counteracting the action of the GRK6/β-arr1^{Class B-like} alignment (Fig. 7A). This model would explain the equivalent behavior triggered by GRK6⁺/GRK2⁻ or β-arr1⁺/β-arr2⁻ (Fig. 7B; refs. 20, 22, 23, 30). The GRK2/β-arr1 configuration may also be possible, yet likely still results in Class A behavior (30). This indicates that the cancer-protective, Class B–like biased signaling downstream of IGF1R is exclusively mediated by GRK6/β-arr1 and only in conditions of abundant ligand (Fig. 7C), e.g., biased receptors in response to the natural ligand (30) or normal receptors stimulated by therapeutic antibodies (27).

An important finding of the present study is that GRK6⁺/GRK2⁻ (siRNA or PX) initiates IGF1R downregulation in conditions with no exogenous ligand in cell models (Fig. 7B) as well as in animal models, raising questions of the mechanisms ligand-dependency. At GPCRs, GRK5/6 can phosphorylate inactive receptors, whereas GRK2/3 phosphorylate only active receptors (62), and hence it is unlikely that GRK2 is involved in ligand unoccupied conformations. Therefore, the most conceivable model is to consider the serum-free setup as conditions with low IGF concentration, the available ligand being generated through an autocrine loop, demonstrated for all cell lines in our panel (63, 64). In this model, in conditions with low ligand, GRK2/β-arr2^ClassA-like recycling competes and opposes GRK6/β-arr1^ClassB-like degradation (Fig. 7A). The ligand or PX/GRK2⁻/GRK6⁺ shifts the equilibrium toward the GRK6/β-arr1^ClassB-like, and hence receptor levels are depleted faster (Fig. 7B–D). Such a model offers a reasonable explanation for the IGF1R decreased at time 0 as well as the switch between the β-arr isoforms recruited to the IGF1R following PX treatment (Fig. 7B and D). The most intriguing finding is the absence of β-arr-BS in conditions with PX-induced IGF1R downregulation (Fig. 7D). This was surprising as PX replicates the effects of GRK2⁻/GRK6⁺ indicative of a Class B–like degradation, but displayed Class A–like signaling. The immunoprecipitation experiments provide mechanistic insight for this paradox: PX treatment favors initial β-arr1 recruitment to the IGF1R, but this cannot be sustained following ligand stimulation. For the GPCRs, a switch from “Class B” to “Class A” is caused by stability of the receptor/β-arr complex, critically controlled by the levels of β-arr ubiquitination (48, 49). Notably, Mdm2 is the essential E3 ligase directing the Class A/B behavior by controlling β-arr1/2 ubiquitination, as well as being the ligase for IGF1R (21) and GRK2 (46). Thus, a possible scenario is that the IGF1R signaling switch from “Class B” to “Class A” following PX treatment is caused by distinct patterns of β-arr ubiquitination (Fig. 7). In this model, substrate competition modified by siRNA decreased GRK2 levels as compared with pharmacologic GRK2 inhibition could differentially modulate the β-arr ubiquitination and stability of the IGF1R/β-arr1 complex (Fig. 7).

This study provides the proof of concept for targeting IGF1R intracellularly through manipulation of the GRK/β-arr system. We validate a widely used drug PX in this context, warranting the rational design of more potent/specific system bias targeting of the IGF1R, to uncouple downregulation from β-arr-BS. In cancers dependent upon IGF1R for survival, as exemplified by ES, unbiased PX-mediated IGF1R downregulation effectively restrains cell growth, superior to the biased IGF1R downregulation characteristic of targeting antibodies. Equally important, as the GRK/β-arr system is increasingly recognized to orchestrate the activities of several other RTKs, this concept could be used as a starting point for system bias therapeutics to selectively cross-target multiple RTKs.

C. Crudden reports grants from European Commission outside the submitted work. No disclosures were reported by the other authors.

C. Crudden: Data curation, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. T. Shibano: Validation, investigation, visualization, methodology. D. Song: Validation, investigation, visualization, methodology. M.P. Dragomir: Formal analysis, investigation, visualization, methodology, writing-review and editing. S. Cismas: Validation, visualization, methodology. J. Serly: Formal analysis, investigation, methodology. D. Nedelcu: Validation, investigation, methodology. E. Fuentes-Mattei: Validation, investigation, methodology. A. Tica: Formal analysis, methodology, writing-review and editing. G.A. Calin: Resources, formal analysis, supervision, funding acquisition, writing-review and editing. A. Girnita: Resources, supervision, funding acquisition, validation, methodology, writing-review and editing. L. Girnita: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.

We thank Drs. R.J. Lefkowitz, R. Baserga, and V. Gurivech for generously providing reagents and cell lines. We thank Anna Malmerfelt at the histology core facility, Karolinska Institutet, for technical assistance. Research support was received from the Swedish Research Council, Swedish Cancer Society, The Swedish Childhood Cancer Foundation, Crown Princess Margareta's Foundation for the Visually Impaired, Welander Finsen Foundation, King Gustaf V Jubilee Foundation, SINF StratCan, Stockholm Cancer Society, Stockholm County, and Karolinska Institute.

Dr. G.A. Calin is the Felix L. Haas Endowed Professor in Basic Science. Work in Dr. G.A. Calin's laboratory is supported by NIH (NIH/NCATS) grant UH3TR00943–01 through the NIH Common Fund, Office of Strategic Coordination (OSC), the NCI grants 1R01 CA182905–01 and 1R01CA222007–01A1, an NIGMS 1R01GM122775–01 grant, a U54 grant #CA096297/CA096300—UPR/MDACC Partnership for Excellence in Cancer Research 2016 Pilot Project, a Team DOD (CA160445P1) grant, a Chronic Lymphocytic Leukemia Moonshot Flagship project, a CLL Global Research Foundation 2019 grant, a CLL Global Research Foundation 2020 grant, and the Estate of C.G. Johnson, Jr.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.