Efficacy of ERBB2-targeting monoclonal antibodies
Both trastuzumab and pertuzumab are expected to have significant, cell-autonomous efficacy against ERBB2+ breast cancer cell lines, based upon their respective action against either ligand-independent or ligand-dependent ERBB2 signalling. However, within a panel of ERBB2+ breast cancer cell lines (Fig. 1a, Additional file 2: Figure S1A), trastuzumab only efficiently inhibited the growth of the high ERBB2-expressing ZR-75-30 cell line (Fig. 1b, IC50 0.2 μg/mL, ~ 1.3 nM). Trastuzumab also partly inhibited the growth of the other high ERBB2-expressing lines, BT474, AU565 and SKBR3, but only at high concentrations for the latter two. Pertuzumab was even less effective, only inhibiting the growth of the ZR-75-30 line by ~ 30%, BT474 by ~ 20% and AU565 and SKBR3 by ~ 10% (Fig. 1b).
However, as these drugs are not often used as single agents in the clinical setting, we also investigated their combined effect on these high ERBB2-expressing cell lines (Fig. 1c). In line with their activity as single agents, trastuzumab and pertuzumab were both effective towards the ZR-75-30 line and also displayed an additive effect when present in combination (Fig. 1c, Additional file 2: Figure S1B). However, the relatively small effect of both drugs was not additive for either the BT474 or AU565 lines.
Because of their complementary inhibitory mechanism, it is commonly held that the combination of trastuzumab and pertuzumab will have increased efficacy by entirely blocking the signalling capacity of ERBB2 [3]. However, our findings demonstrate that combination therapy with these two drugs may not always be beneficial, even within the context of high ERBB2 expression. As would be expected, each of these high ERBB2-expressing cell lines also displayed elevated levels of phosphorylated Akt (Fig. 1a). We therefore investigated the ability of these drugs individually, and in combination, to inhibit ERBB3 and Akt activation, as this is considered the main signalling pathway through which ERBB2 mediates its oncogenic function [16]. Consistent with these expectations, trastuzumab inhibited ERBB3 phosphorylation by ~ 60% and Akt phosphorylation by 60–80% in all cell lines (Fig. 1d), while pertuzumab inhibited ERBB3 phosphorylation by 30–50% and Akt phosphorylation by 40–60%. The combination of trastuzumab and pertuzumab had a significantly additive effect upon ERBB3 phosphorylation in all three cell lines. However, this was not the case for downstream signalling to Akt, for which this combination only displayed an additive effect in the ZR-75-30 line.
As trastuzumab is proposed to inhibit ligand-independent dimerization between ERBB2 and ERBB3, this appears to be the predominant mechanism of ERBB2-mediated ERBB3 activation in these ERBB2+ cell lines. Although more importantly, these findings clearly demonstrate that there is a disconnection between the ability of ERBB2-targeting therapeutic antibodies to inhibit the ERBB2/ERBB3/PI3K signalling axis and their ability to exert an influence upon cell growth. This further suggests that our understanding of ERBB2 dimerization and signalling in this pathological context is incomplete and that targeting ERBB2 in this manner will not always be sufficient to completely inhibit its oncogenic function.
ERBB2 interacts with a wide array of RTKs
Therefore, to investigate the potential for ERBB2 to form heterodimers across the full spectrum of RTK families, we utilised a bimolecular fluorescence interaction technique that we and others have previously used to visualise the formation of EGFR family dimers [23, 31]. This technique uses the individual N-terminal (V1) and C-terminal (V2) fragments of the Venus fluorescent protein, which are non-fluorescent and associate with low affinity in the absence of an interaction between fusion partners. However, their close co-localisation upon interaction of bait and prey fusion proteins favours refolding of these split domains into a functional β-barrel structure containing the fluorophore [32] (Fig. 2a). Using this technique, a membranous fluorescent signal can be observed following co-transfection of ERBB2-V1 and ERBB2-V2 into HEK-293 T cells, but not for the individually tagged constructs in isolation (Fig. 2b, Additional file 2: Figure S2A). We have also previously observed this for the formation of ERBB2:EGFR and ERBB2:ERBB3 heterodimers [23], and here we have adapted this technique to screen for the formation of heterodimers between ERBB2 and a library of 46 full-length RTKs (Fig. 2c) (Additional file 1: Table S1).
Following co-transfection of the ERBB2-V1 plasmid with each of the 46 V2-tagged RTKs, we used high-content imaging to detect a fluorescent signal from the refolded Venus protein and filtered this for significance over the background signal generated by a non-interacting pair, ERBB2-V1 and Histone 2B-V2 (H2B-V2) (Additional file 2: Figure S2B). This interaction data was then overlaid onto an RTK dendrogram [33] in order to observe any phylogenetic patterns underlying heterodimer formation (Fig. 2c).
This library screen identified 37 RTKs that robustly dimerized with ERBB2 under these conditions, including all EGFR family members and other previously identified interacting partners, including IGF-1R, MET, AXL and NTRK1 [21]. However, many other novel interacting partners were also detected, including receptor families known to play a role in breast cancer, such as FGFRs, NTRKs and PDGFRs [34]. Interestingly, only a small number of non-interacting receptors clustered together, including VEGFR1 and VEGFR2, and EPHB4 and EPHA8.
Analysis of a subset of these unexpected interactions by confocal fluorescent microscopy confirmed that these non-canonical heterodimers are correctly localised at the plasma membrane (Fig. 2d). Therefore, these data suggest that ERBB2 may be able to exert its oncogenic activity through interaction with a broad array of RTKs, potentially diversifying the signalling response downstream of ERBB2 and also influencing the cellular response to ERBB2-targeting drugs.
Pertuzumab increases tyrosine phosphorylation of alternative RTKs
While pertuzumab has been shown to inhibit the ligand-induced interaction between transfected ERBB2 and ERBB3 in COS-7 cells [5] and endogenous receptors in MCF-7 and SKBR3 cells [35], the influence of pertuzumab or trastuzumab upon the promiscuous signalling activity of ERBB2 is not well established. To investigate this, we utilised antibody-based arrays (Fig. 3a) to measure the relative tyrosine phosphorylation of 49 RTKs in the ZR-75-30, BT474 and AU565 cell lines following treatment with pertuzumab or trastuzumab (Fig. 3b). Under control conditions, tyrosine phosphorylation of EGFR, ERBB2, ERBB3, INSR and IGF-1R could be observed in all lines, apart from EGFR which is not expressed in the ZR-75-30 line (Additional file 2: Figure S1A). Tyrosine phosphorylation of ERBB3, IGF-1R and INSR decreased in all lines following treatment with either trastuzumab or pertuzumab, while trastuzumab decreased EGFR phosphorylation in the AU565 cells. However, tyrosine phosphorylation of ERBB2 actually increased following pertuzumab treatment in all lines, as did phosphorylation of a subset of other RTKs, including MERTK, MET and NTRK1 in both AU565 and BT474 lines, and TIE-2 and RET in AU565 cells (Fig. 3b). Notably, the ZR-75-30 line expresses either low or undetectable levels of MET, MERTK and NTRK1 (Additional file 2: Figure S1A), which likely underlies the lack of phosphorylation of these receptors in this cell line.
The increased tyrosine phosphorylation of ERBB2 was validated by immuno-precipitation and blotting with PY100 (Additional file 2: Figure S3A). However, the decreased phosphorylation of ERBB2 following trastuzumab treatment on the antibody array is likely an artefact of antibody competition, as this decrease was not observed by immunoprecipitation and blotting with PY100 (Additional file 2: Figure S3B).
To confirm the results from the antibody array, we also undertook western blotting with phospho-specific antibodies (Fig. 3c, Additional file 2: Figure S4). This analysis also demonstrated that pertuzumab significantly decreased ERBB3Y1289 phosphorylation in both AU565 and BT474 lines, but also significantly increased the phosphorylation of ERBB2Y1221/1222, ERBB2Y1248, METY1234, NTRK1Y674/Y675, MERTKY681/Y749, TIE-2Y992 and RETY905 in both lines.
While pertuzumab increased the phosphorylation of this subset of non-canonical RTKs, it was not necessary to promote the physical interaction between ERBB2 and either MET or MERTK. Through co-immunoprecipitation, we could confirm that ERBB2 interacts with MET and MERTK in both BT474 and AU565 cell lines prior to exposure to pertuzumab (Fig. 4a). We also observed an interaction between the expected ERBB2 binding partners, EGFR and ERBB3, which decreased over time in the BT474 line. Interestingly, a robust interaction between TIE-2 and ERBB2 in the AU565 line was only observed following pertuzumab treatment (Fig. 4a), reflecting the absence of an interaction between these two receptors in our library screen (Fig. 2d).
While the interaction between ERBB2 and NTRK1 could not be observed by co-immunoprecipitation, the interaction between ERBB2 and EGFR, ERBB3, MET, MERTK and NTRK1 could all be observed by proximity-mediated ligation assay (PLA) in both the AU565 and BT474 cell lines (Fig. 4b, Additional file 2: Figure S5). The interaction between TIE-2 and ERBB2 could also be observed in the absence of pertuzumab using this method, albeit weakly.
Pertuzumab blocks Akt but promotes ERK activation
Pertuzumab is known to effectively block ligand-induced signalling from ERBB2 to both Akt and ERK [35]. Therefore, to determine the downstream consequences of the paradoxical actions of pertuzumab on these different subsets of ERBB2-containing heterodimers, we performed western blotting for activation of the main RTK-mediated signalling pathways (Fig. 5a, Additional file 2: Figure S6). This analysis confirmed that pertuzumab treatment significantly decreased Akt phosphorylation in both the AU565 and BT474 cell lines, along with significantly decreased STAT3 phosphorylation in the AU565 line and PLCγ phosphorylation in the BT474 line. However, ERK phosphorylation was significantly increased at all time points of pertuzumab treatment in both cell lines. Notably, the pertuzumab sensitive cell line, ZR-75-30, displayed a weaker, transient activation of ERK in response to pertuzumab (Additional file 2: Figure S7A), which was only significant at the 30-min time point.
This ERK activation mediated by pertuzumab was also confirmed with an orthogonal, live-cell readout of ERK activation, the ERK-KTR biosensor (Fig. 5b). This biosensor is based upon the fluorescent protein Clover, fused to a phosphorylatable nuclear localisation sequence [36]. Upon phosphorylation by ERK, the biosensor shuttles out of the nucleus, allowing a readout of ERK activity based upon the cytoplasmic-to-nuclear (C/N) ratio of the fluorescent signal. By treating AU565 cells stably expressing the ERK-KTR biosensor with pertuzumab, we were able to monitor the C/N ratio over time by high-content imaging. This analysis demonstrated that following pertuzumab treatment, ERK activity increased with similar dynamics to that observed by the western blotting analysis (Fig. 5a).
In both the BT474 and AU565 cell lines, this pertuzumab-induced ERK activation was accompanied by increased interaction between ERBB2 and the receptor proximal, MAPK activating adaptor proteins, GRB2 and SHC (Fig. 5b). Hence, while pertuzumab can at least partially inhibit PI3K signalling through ERBB2:ERBB3 dimers, the interaction of ERBB2 with a subset of non-canonical RTKs promotes ERBB2 phosphorylation, recruitment of GRB2 and SHC and the downstream activation of ERK.
We therefore investigated whether direct inhibition of ERK activation using the MEK inhibitor UO126 would attenuate pertuzumab resistance within these resistant cell lines. While UO126 did prevent pertuzumab-induced ERK activation, it also elevated Akt activity and prevented the efficient inhibition of Akt by pertuzumab (Fig. 5c), in line with previous observations that MEK inhibition relieves negative feedback on ERBB receptors and thus promotes PI3K/Akt activation [37]. Accordingly, UO126 also displayed a weak inhibitory effect upon AU565 cell growth as a single agent, but not in combination with pertuzumab (Fig. 5d). Whereas in BT474 cells, UO126 did not influence cell growth as a single agent and completely abrogated any effect of pertuzumab (Additional file 2: Figure S8A).
Targeting the network of non-canonical heterodimers
As the targeted inhibition of MEK/ERK was not a viable option, we next investigated whether targeting the network of non-canonical ERBB2 heterodimers would increase the efficacy of pertuzumab. We therefore utilised two broad-spectrum RTK inhibitors with nanomolar IC50 values for different combinations of these non-canonical RTKs. BMS-777607 is currently undergoing phase II clinical trials for a number of different cancers, including breast cancer, and has activity against MET (IC50 3.9 nM), MERTK (IC50 14 nM) and NTRK1 (IC50 290 nM) [38]. Cabozantinib is also undergoing phase II clinical trials for breast cancer and has activity against MET (IC50 1.3 nM) and TIE-2 (IC50 14.3 nM) [39]. However, neither of these inhibitors could prevent pertuzumab-induced ERBB2 phosphorylation or ERK activation in AU565 cells (Fig. 5e). While cabozantinib did appear to prevent the pertuzumab-induced phosphorylation of TIE-2 and NTRK1, this was not sufficient to prevent ERK activation. In line with these observations, neither of these RTK inhibitors could increase the efficacy of pertuzumab in AU565 cells (Fig. 5f) nor BT474 cells (Additional file 2: Figure S8B).
As inhibition of these non-canonical RTKs did not prevent pertuzumab-induced ERBB2 phosphorylation, it is likely that the kinase domains of the non-canonical RTKs within these pertuzumab-activated heterodimers are acting as allosteric activators of the ERBB2 kinase domain. In this scenario, targeting these RTKs directly will not influence the signalling capacity of the heterodimer. Instead, targeting the ERBB2 kinase domain should prevent activation of both ERBB2 and all heterodimer combinations.
Therefore, we investigated whether the TKI lapatinib was able to prevent the pertuzumab-induced activation of these non-canonical RTKs by ERBB2. In cell viability assays, lapatinib displayed strong activity as a single agent against all the high ERBB2-expressing ZR-75-30, BT474 and AU565 lines, with IC50 values between 50 and 100 nM (Fig. 6a). At higher concentrations, it was also effective against the lower ERBB2-expressing MDA-MB-361 and MDA-MB-453 lines, with IC50 values of ~ 1 μM and 10 μM, respectively.
In agreement with the hypothesis of allosteric activation, the treatment of BT474 cells with lapatinib prevented pertuzumab-induced recruitment of GRB2 and SHC to ERBB2 (Fig. 6b). Treatment of both BT474 and AU565 cells with lapatinib also decreased the activation of ERBB2, MET, NTRK1, MERTK, TIE2 and ERK under basal conditions and prevented their pertuzumab-induced activation (Fig. 6c, d). Lapatinib almost completely inhibited ERBB3 and AKT phosphorylation in the BT474 cell line (Fig. 6c), while the combination of lapatinib and pertuzumab had an additive effect upon ERBB3 and AKT phosphorylation in the AU565 cell line (Fig. 6d, Additional file 2: Figure S7B). Accordingly, lapatinib also increased the efficacy of pertuzumab in cell viability assays in both cell lines, resulting in a synergistic combination between these two ERBB2-targeting drugs (Fig. 6e, f, Additional file 2: Figure S8C, D).
Unlike the combination of pertuzumab and lapatinib, the combination of trastuzumab and lapatinib did not display any synergy in the AU565 cell line (Additional file 2: Figure S8E) nor did the combination of pertuzumab and trastuzumab (Fig. 1c). These findings suggest that within the setting of innate resistance driven by non-canonical ERBB2 heterodimers, only the combination of pertuzumab and lapatinib will give a beneficial combinatorial response.
It has been previously hypothesised that long-term treatment with lapatinib results in an adaptive response that rescues ERBB2 and ERBB3 activation, and consequently ERK and Akt signalling, through the increased expression of ERBB3 ligands [40]. After 24 h of lapatinib treatment in the AU565 cell line, we do observe a moderate recovery of ERBB3 phosphorylation and to a lesser extent Akt phosphorylation (Additional file 2: Figure S8F), but this is not influenced by the addition of pertuzumab. Instead, the effect of this drug combination on downstream signalling is already apparent at the earlier time points of drug treatment (Fig. 6c, d. Additional file 2: Figures S7B, S8F), and therefore, the synergy between these drugs does not result from long-term adaptation.
In vivo confirmation of synergy
To confirm the synergistic combination of pertuzumab and lapatinib within this resistant setting in vivo, we utilised the HCI-012 ERBB2+ PDX model, which originated from a patient who relapsed following treatment with trastuzumab and lapatinib [29] (Fig. 7). The expression of ERBB2 and the non-canonical RTKs MET, MERTK, NTRK1 and TIE-2 were observed by immuno-histochemistry within this PDX model (Fig. 7a, Additional file 2: Figure S9A). Furthermore, we also observed an interaction between ERBB2 and its canonical partner ERBB3, along with each of the non-canonical partners, using PLA (Fig. 7b, Additional file 2: Figure S9B).
Combination treatment with lapatinib and pertuzumab over 21 days significantly slowed tumour growth following sub-cutaneous implantation of this PDX within NOD-SCID-IL2γR−/− (NSG) mice (Fig. 7c, d). This response, albeit small, is remarkable given that this PDX originated from a trastuzumab and lapatinib resistant, relapsed tumour. In line with the patient treatment response, treatment with lapatinib alone did not result in a significantly slower growth rate, nor did single agent treatment with pertuzumab. Combination treatment also significantly decreased the expression of the proliferation maker Ki67 at the 14-day time point (Fig. 7e), indicating that only the combination of these two drugs reduced the growth rate of this otherwise single-agent-resistant ERBB2+ tumour.