Bisphosphonates inactivate human EGFRs to exert antitumor actions

Connectivity Mapping Suggests That Bisphosphonates Act on the HER RTK Family.

Although it is known that bisphosphonates inhibit FPPS activity to decrease bone resorption, there are many FPPS-independent actions for which there is no clear explanation. To discover new mechanisms through which bisphosphonates act, we used a novel bisphosphonate signature to interrogate the Connectivity Map (CMAP; www.broad.mit.edu/cmap), a compendium of connections between genes, diseases, and drugs (Fig. S1A) (23). For this process, we generated osteoclasts from peripheral blood mononuclear cells obtained from three donors, exposed these cells to alendronate or risedronate, which are two of the most commonly used bisphosphonates, and performed microarrays on the isolated mRNA. The microarray dataset was used to develop a combined bisphosphonate gene signature by identifying genes that displayed the greatest statistical significance in being up- or down-regulated across both alendronate- and risedronate-treated samples (Fig. S1B). This approach allowed us to create a signature free from changes attributable solely to either agent.

The bisphosphonate signature consisted of six up-regulated (P ≤ 0.006) and seven down-regulated (P ≤ 0.0005) genes (Fig. S1B). This signature, validated by quantitative PCR (qPCR) (Fig. S1C), was used to interrogate CMAP to identify chemicals with similar genomic mechanisms to bisphosphonates. Of the chemicals displaying the highest enrichment scores, the top two were anticancer agents, namely the poly(ADP-ribose) polymerase (PARP) inhibitor 1,5-isoquinolinediol, and the first-generation EGFR kinase inhibitor tyrphostin AG-1478 (Fig. S1D). We confirmed that these compounds were “true” bisphosphonate mimetics using osteoclast formation assays in vitro. This finding was intriguing because parallel analysis on the same microarray dataset by the Kyoto Encyclopedia of Genes and Genomes (KEGG) showed that 11 of the 30 bisphosphonate-associated pathways contained HER family signaling molecules, including HER1, HER2/neu, EGF, phosphoinositide 3-kinase, protein kinase B (AKT), phosphatase and tensin homolog, and rapidly accelerated fibrosarcoma 1 (Fig. S1E). Together, the CMAP and KEGG outputs suggest that a hitherto unappreciated action of bisphosphonates may be to regulate EGFR/HER2.

Bisphosphonates Bind the HER1/2 Kinase Domain to Inhibit Activation.

We used a thermal shift assay to determine whether bisphosphonates bind directly to recombinant HER proteins, namely HER1^wt, HER1^L858R, and HER2^wt. Heat-induced protein unfolding unmasks hydrophobic regions that are captured by Sypro Orange to yield a melting temperature (Tm). The binding of a ligand stabilizes protein structure and causes a shift in Tm (ΔTm) (Fig. 1A). Incubation of the bisphosphonate residronate with HER1^wt or HER1^L858R caused a highly significant ∼2 °C thermal shift, indicating direct protein–ligand interaction (Fig. 1A). A significant thermal shift was also noted, as expected, with erlotinib, but was not seen with tiludronate (Fig. 1A), a bisphosphonate that contains a p-cholorophenol ring instead of an N-atom in the R1 side chain (Fig. S2). With the recombinant HER2^wt construct, risedronate induced a peak at ∼48.2 °C that was not seen with vehicle (Fig. 1A, table). Taken together, the studies not only establish bisphosphonate-HER1/2 binding in vitro, but also demonstrate that an N atom is required for binding.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Bisphosphonates bind to the kinase domain of the HER family. (A) Protein thermal shift uses Sypro Orange to capture hydrophobic domains during heat-induced protein unfolding to yield a Tm. The Tm increases when ligand binds to protein resulting in a thermal shift (ΔTm = Tm_ligand – Tm_veh)._. Risedronate (Ris) or erlotinib (Ert) binding to recombinant HER1^wt or constitutively active HER1^L858R protein causes a significant thermal shift, whereas tiludronate (Til) does not. HER1^{K745A/N842A/D855A}, in which the predicted interacting amino acids are mutated, does not display a bisphosphonate-induced thermal shift: this constitutes biophysical proof for the binding of bisphosphonates to the HER1 kinase domain. With recombinant HER2^wt protein, a second peak is induced upon bisphosphonate binding, again indicative of direct protein–ligand interaction. Tm (°C) is shown on representative traces; ΔTm, Tm, P values (determined by pairwise comparisons using two-tailed Student t test), and number of replicate traces (in parentheses) are shown in the table (Inset). (B) Molecular docking using the X-ray crystal structure of HER1^wt (PDB ID code 2GS7) confirms binding of Ris, but not Til to the kinase pocket. Bisphosphonates containing one or more N-atoms interact with T790 (gatekeeper residue) via a water (WAT) molecule, whereas Til, which has a p-chlorophenyl group, does not (also see Fig. S2). Similarly, the N6 atom of the adenine group of AMP-PNP interacts with T790 via WAT. Superimposition of crystallized AMP-PNP and docked Ris thus shows an overlap of the purine ring of adenine and the pyridine ring of Ris. Activated HER1^L858R (PDB ID code 2ITV) resembles the active state of HER1 (PDB ID code 2GS6) and binds Ris, Ert, and AMP-PNP. Docking also confirms loss of bisphosphonate binding to HER1^{K745A/N842A/D855A}, where substituted Ala residues are unable to coordinate with Mg²⁺. Similar interactions between Ris and zoledronic acid (ZA) occur with HER2; here WAT bridges the bisphosphonates to the gatekeeper residue T798. (C) Ris causes a concentration-dependent inhibition of the kinase activity (in triplicate) of recombinant HER1 and HER2. The inhibition is similar to that with recombinant FGFR, considerably less marked with VEGFR, and virtually absent with the INSR. Correspondingly, the structure of HER1 and FGFR kinase domains superimpose well and residues that coordinate Mg²⁺ are well positioned for bisphosphonate binding. In contrast, the VEGFR structure is in a more open conformation, where D1046 and N1033 are positioned further away from the binding pocket. This shifts Mg²⁺ coordination outwards (arrow) and reduces the bisphosphonate fit. The INSR kinase domain is structurally distinct from that of HER1 and is unable to bind bisphosphonates.

Computational modeling was used not only to confirm HER-bisphosphonate binding in silico, but also to predict interacting amino acid residues in HER1 that could be mutated. Docking studies using HER1^wt (PDB ID code 2GS7) (24) revealed that only N-containing bisphosphonates dock into the ATP-binding site of the kinase domain (Fig. 2B and Fig. S1). Their phosphate groups coordinate an Mg²⁺ together with amino acids D855 and N842 and a water molecule, mimicking the ATP analog AMP-PNP (Fig. 1B). Additionally, the pyridine ring of risedronate interacts with the gatekeeper residue T790 and residue T854 via a structural water molecule (Fig. 1B), an interaction that is equally essential for the binding of erlotinib and gefitinib (PDB ID codes 1M17, 2ITY, and 4HJO) (25⇓–27). Importantly, the non–N-containing bisphosphonate tiludronate failed to dock (Fig. 1B). A slightly modified docking mode was identified for HER2^wt, in which the pyridine ring of risedronate and imidazole ring of zoledronic acid interact with the gatekeeper residue T798 (Fig. 1B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Bisphosphonates inhibit cancer cell viability through a HER-dependent mechanism. Effects of bisphosphonates or erlotinib (Ert) (as shown) on cell viability assessed by the MTT assay (mean % ± SEM) in HER1^ΔE746-A750 (HCC827), HER1^wt (H1666 and H1703), and HER1^wt:RAS^G12S (A549) lung cancer cells, as well as in HER1^wt breast cancer (MB231) and HER^low colon cancer cells (SW620) (triplicate wells; repeated three times; statistics by ANOVA with Bonferroni’s Correction against zero-dose; P values versus zero-dose; mean ± SEM; *P < 0.05, **P < 0.01). qPCR was used to determine mRNA expression of all four HER isoforms HER1 to -4 (transcript copy number per cell ± SEM; three biological replicates, with three technical replicates each; estimated using 2,500 copies of β-actin per cell). To study whether the bisphosphonate effect was mediated by one or more HER isoforms, siRNAs for each of the four isoforms (siHER) were applied all at once, with scrambled siRNAs (siScr) as controls. Knockdown of individual HER proteins was confirmed by Western blotting using isoform-specific antibodies SI Methods). MTT assay assessed the viability of siRNA- or siScr-transfected cells in response to zoledronic acid (ZA, 40 µM). The extent of reduction in HER corresponded to the extent of reversal of the ZA effect on cell viability (four biological replicates transfected separately; expressed as mean percent reduction with bisphosphonate ± SEM; statistics: two-tailed Student t test; P values from siScr vs. siHER transfectants; *P < 0.05, **P < 0.01). Of note is that siHER itself reduced viability in all but SW620 cells; results are therefore expressed as a percentage of vehicle-treated siScr and siHER transfectants, respectively. A partial down-regulation of HER1 in MB231 cells corresponds with a partial attenuation of bisphosphonate action. With HCC827 cells, the oncogenic stimulation is because of mutated HER1^ΔE746-A750; hence, knockdown almost abolishes bisphosphonate action, despite a relatively lower extent of HER2 knockdown. Note: Only relevant bands from Western blots are shown, with gaps introduced where irrelevant lanes are excised.

To specifically establish bisphosphonate binding to the kinase domain of HER1, we created a construct in which the predicted interacting residues N842 and D855, as noted above, were mutated to alanine residues. We also mutated K745, a residue involved in stabilizing the Cα-helix. This triple mutant HER1^{K745A/N842A/D855A} yielded a more open conformation into which risedronate failed to dock (Fig. 1B). Consequently, and in contrast to HER1^wt, recombinant HER1^{K745A/N842A/D855A} also failed to display a thermal shift with risedronate (Fig. 1A). Consistent with docking predictions, these data established precisely the amino acid residues required for bisphosphonate binding to the HER1 kinase domain.

The most common driver mutations for nonsmall-cell lung cancer, namely HER1^ΔE746-A750 and HER1^L858R, reside in the HER1 kinase domain (28, 29). Having shown that risedronate and erlotinib bind to HER1^L858R in the thermal shift assay (Fig. 1A), we sought to investigate the precise conformational changes that are induced by the drugs using dynamic simulations. Activation of HER1^wt changes the conformation of the Cα-helix from an outward to a collapsed position resembling the HER1^L858R conformation (Fig. S3) (30). Molecular dynamics (50 ns) and anisotropic network modeling (10 µs) showed that the Cα-helix remains collapsed upon erlotinib binding (Fig. S3). Similarly, both risedronate and zoledronic acid docked into the adenine-binding domain of HER1^L858R without affecting the position of the Cα-helix (Fig. 1B and Fig. S3).

Downstream inhibition of Tyr-kinase signaling by bisphosphonates was examined in a cell-free system by assessing the phosphorylation of a poly-(Glu-Tyr) substrate with recombinant HER1^wt, HER1^L858R, and HER2^wt proteins. There was a marked concentration-dependent inhibition of Tyr-kinase activity with both risedronate and zoledronic acid (Fig. 1C and Fig. S4). Importantly, this inhibition was not reversed in high [Mg²⁺] (8 mM), reinforcing direct binding of bisphosphonates, as opposed to an indirect effect via Mg²⁺ chelation (Fig. S4).

To evaluate the potential effects of bisphosphonates on other RTKs, we examined substrate phosphorylation similarly using recombinant fibroblast growth factor receptor (FGFR), vascular endothelial growth factor receptor (VEGFR), and insulin receptor (INSR). We independently docked risedronate into the crystal structures of the three receptors (PDB ID codes 3KY2, 3VID, and 3ETA, respectively). Risedronate inhibited FGFR phosphorylation somewhat similarly to HER1. The kinase domains of the two RTKs superimpose, with residues that coordinate Mg²⁺ being positioned for bisphosphonate binding (Fig. 1C). In contrast, inhibition was significantly less marked with VEGFR. Unlike the HER1 structure, residues D1046 and N1033 in VEGFR are positioned further away from the binding pocket; this configuration shifts Mg²⁺ coordination outwards reducing the bisphosphonate fit (Fig. 1C). Risedronate failed to inhibit Tyr-phosphorylation of INSR, the kinase domain of which, being structurally distinct, is unable to bind bisphosphonates in silico (Fig. 1C). Thus, subtle differences in kinase domain structure determine the extent of RTK–bisphosphonate interactions. With that said, in contrast to a marked reduction in HER1 phosphorylation in vivo by bisphosphonates, there was no such reduction noted in pFGFR, pVEGFR, or pINSR (Fig. S5).

Bisphosphonates Inhibit HER-Driven Cancer Growth.

We studied the effect of several Food and Drug Administration-approved bisphosphonates on HER1^ΔE746-A750- (HCC827) and HER1^L858R- (H3255) driven lung cancer, as well as HER1^wt-expressing lung (H1666 and H1703), breast (MB231), and colon (SW480) cancer cells. Cell viability (MTT assay) was inhibited in a concentration-dependent manner in all cell lines with N-containing bisphosphonates, notably risedronate, alendronate, zoledronic acid, and minodronic acid (Fig. 2 and Figs. S2 and S6). In contrast, non–N-containing bisphosphonates, such as tiludronate, failed to affect cell viability (Fig. 2 and Fig. S6).

Interestingly, zoledronic acid and minodronic acid have almost identical actions on H3255 and HCC827 cell viability, similar imidazole rings, and near-identical binding energies (−10.48 and −9.47 kcal/mol) (Fig. S2). Docking studies confirm that the phosphate groups of zoledronic acid, minodronic acid, and tiludronate coordinate with the Mg²⁺ ion. In contrast, only the imidazole rings of zoledronic acid and minodronic acid), but not the p-chlorophenyl ring of tiludronate, interact with T790 (Fig. 1B and Fig. S2). This finding suggests that, to act as an HER1 kinase inhibitor, a bisphosphonate must interact with T790 and, at the same time, coordinate with the Mg²⁺ ion.

Zoledronic acid failed to inhibit cell viability in SW620 cells, a colon cancer line that expresses low, <5 mRNA transcripts per cell for all HER isoforms (Fig. 2). This finding suggested that cellular inhibition by bisphosphonates was HER-mediated. We quantitated the expression of HER isoforms, HER1 to -4, by qPCR, and used siRNA to simultaneously knock down all four HER isoforms in select lines (Fig. 2). Western blots on whole-cell extracts confirmed siRNA-induced knockdown of individual HER isoforms compared with scrambled siRNAs.

Knocking down the predominant, mutated HER1^ΔE746-A750 isoform and HER2^wt in HCC827 cells dramatically attenuated bisphosphonate-induced viability inhibition (Fig. 2). Similar effects were noted in H1666 and H1703 lung cancer cells, which predominantly expressed HER1/2 (Fig. 2). Bisphosphonate inhibition was also dampened by knocking down the predominant isoform HER1 in MB231 breast cancer cells. Expectedly, bisphosphonate-resistant HER^low SW620 colon cancer cells showed no response to siRNA-induced HER knockdown. In contrast, A549 cells, which express HER1/2, but are driven by a RAS^G12S mutation, showed modest viability inhibition with bisphosphonate. However, this small effect was completely lost by knocking down HER1/2, indicating that bisphosphonate action was HER- and not RAS-mediated (Fig. 2). Taken together, these data firmly establish that bisphosphonate action on tumor cell viability is mediated via HER.

Bisphosphonates Inhibit Global HER Signaling to Cause Apoptosis.

We explored first whether the anticancer action of bisphosphonates was, to any extent, mediated by their known inhibitory effects on FPPS, a ubiquitous enzyme in the mevalonate pathway. As expected, zoledronic acid induced the accumulation of the substrate unprenylated Rap-1α (u-Rap-1α) to varying extents in cancer cells (Fig. S7A). However, there was no correlation between u-Rap-1α accumulation and reduced cell viability (MTT assay) (Fig. S7A). Toward obtaining definitive independent evidence, we knocked down FPPS expression completely using siRNA (Fig. S7B). We found that the inhibitory effect of zoledronic acid on the viability of HCC827 and MB231 cells persisted, despite the total absence of FPPS protein (Fig. S7B). This finding is in stark contrast to the dramatic reversal of bisphosphonate inhibition of cell viability when the HER isoforms were knocked down (Fig. 2).

Exploration of HER signaling indicated a global reduction of downstream phosphorylation, cell-cycle arrest, and apoptosis induced by bisphosphonates. FACS showed concentration-dependent G₁ arrest in HCC827 cells, accompanied by a profound reduction in expression of the cell-cycle genes cyclin B1, cyclin D1, and proliferating cell nuclear antigen (PCNA) (except for cyclin B1 in H1703 cells) (Fig. 3A). In addition to causing cell-cycle arrest, zoledronic acid also induced apoptosis (sub-G₁ phase on FACS, and PARP cleavage on Western blotting) to varying extents in HER1-expressing H1666, H1703, H3255, and MB231 cells (Fig. 3). The drug displayed limited effects in HER1^wt-expressing, RAS^G12S-driven A549 cells, and minimal effects on HER^low SW620 cells; this again reaffirms that the anticancer actions of bisphosphonates were HER-mediated.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Bisphosphonates inhibit global HER signaling to cause cell-cycle arrest and apoptosis. (A) Flow cytometry showing the effect of the bisphosphonate zoledronic acid (ZA) on the cell-cycle profile of HER1^wt (H1666), HER1^ΔE746-A750 (HCC827), HER1^L858R (H3255), and HER1^wt:RAS^G12S (A549) expressing lung cancer cells, as well as HER1^wt-expressing breast cancer (MB231) and HER^low colon cancer (SW620) cells (repeated three times on single samples; colored bars include mean ± SEM). Representative Western blots show PARP cleavage and cyclin D1, cyclin B1, and PCNA protein expression. (B) Downstream signaling pathways triggered by HER1 activation, of which the ones shown in red were examined. ZA inhibited EGF-induced phosphorylation of Y845, Y992 (Western blots), Y1045, Y1068, and Y1173 (ELISA) (two-tailed Student t test; in duplicate; mean fold-change over no-EGF ± SD; *P < 0.05, **P < 0.01). Western blots also show reduced pAKT and pERK in whole cell extracts of HER1^wt-expressing H1666 cells, whereas there was no inhibition in HER^low SW620 cells. Western blots of cytosolic (C) and nuclear (N) subfractions of H1666 and A549 cells show a marked reduction by ZA of pSTAT3, pSTAT5, and p50 in the nuclear subcompartment (normalized to histone-H3). Note: Only relevant bands from Western blots are shown, with gaps introduced where irrelevant lanes are excised.

HER1 activates downstream signaling through the phosphorylation of seven Tyr residues distal to its kinase domain (Fig. 3B). Zoledronic acid significantly reduced EGF-induced autophosphorylation of residues Y1045, Y1068, Y1173 (ELISA), Y845, and Y992 (Western blots) (Fig. 3B). We then explored downstream pathways, notably those arising from Y1086 and Y1146 that were not tested for autophosphorylation directly. Phosphorylation of ERK and AKT was strongly inhibited in HER1^wt H1666 cells, but not in HER^low SW620 cells (Fig. 3B). STAT3/5 and NF-κB signaling was evaluated by Western blotting of cytosolic and nuclear fractions. In H1666 and A549 cells, zoledronic acid induced a reduction in the nuclear localization of pSTAT3, pSTAT5, p50, and p65 (normalized to histone-H3) (Fig. 3B). The data suggest that bisphosphonates reduce HER signaling globally, consistent with our evidence for their direct action on the kinase domain.