|Home | About | Journals | Submit | Contact Us | Français|
Inactivation of the HER2 tyrosine kinase holds significant promise as a cancer treatment hypothesis, making it a high value target for drug discovery. Screening and structure-based efforts have led to the development of several classes of ATP analog inhibitors of HER2 tyrosine kinase. These efforts have been further enhanced by detailed structural information regarding its sibling kinase EGFR and structural properties that can be exploited to confer activity and even selectivity towards HER2 kinase. Signaling and structural studies also suggest the critical involvement of the kinase inactive HER3 in the regulation of HER2 creating unique challenges in the efforts to inactivate HER2.
The ErbB proteins are a four-member family of highly homologous receptor tyrosine kinases comprised of ErbB1 (EGFR, HER1), ErbB2 (HER2), ErbB3 (HER3), and ErbB4 (HER4). These proteins consist of a ligand-binding extracellular domain (ECD), a transmembrane domain, an intracellular tyrosine kinase (TK) domain, and a c-terminal signaling tail. An intracellular signal is generated through receptor dimerization and transphosphorylation of their c-terminal tails  (Figure 1). The differentiation of this family from a primordial ErbB gene has been associated with functional complimentarity and a necessity for cooperative activity in some of its members. Such cooperativity is exemplified by HER2 and HER3. HER2 has evolved into a catalytic driver, with robust kinase activity but no ligand-binding ability and little ability for self-regulation. On the other hand, HER3 has no significant kinase activity but is an optimal dimerization and regulatory partner for HER2 [2,3]. In fact, in the presence of ligand stimulation, the HER2-HER3 heterodimer is the most active signaling unit in this family . EGFR, on the other hand, has maintained its bifunctional attributes and performs equally well as a catalytic or ligand-activated regulatory partner  .
The reduced ability of HER2 to self-regulate suggests a potent oncogenic potential, and indeed overexpression of HER2 is seen in a number of human cancers, mostly breast cancers. The etiologic role of HER2 in tumorigenesis has been extensively studied in mouse transgenic models, confirming unequivocally the potent tranforming potential of the mouse HER2 homolog Neu when overexpressed or overactive . The driving role of HER2 in tumorigenesis and the large number of cancer patients affected by this cancer subtype have made HER2 a high priority target for drug development for the past two decades.
Initial attempts to target HER2 in the 1980s focused on the development of monoclonal antibodies to interfere with functions residing within its ECD. These efforts have produced clinically active drugs, but they do not appear to effectively inactivate HER2 signaling and the molecular basis for their clinical activities remains undefined (reviewed in ). Furthermore, the HER2 ECD is redundant for its oncogenic function and is often proteolytically cleaved in tumors, indicating a potential limitation of ECD-targeting approaches [8,9]. However, the HER2 TK domain is essential for its transforming function  and targeting the catalytic TK function of HER2 presents the most compelling approach for the development of highly effective anti-cancer drugs.
Numerous HER-family selective TK inhibitors (TKIs) have been synthesized over the past decade and are listed in Table 1. The ongoing evolution of these TKIs started with investigations of EGFR, followed by pan-HER family inhibitors, followed only recently by HER-2 selective compounds. There is still much to be learned about the structure-activity relationships for these new HER-2 selective drugs. This review focuses on the key structural classes and discovery strategies for the HER family kinase inhibitors; a detailed analysis of preclinical models or clinical activity is outside the current scope. For clinical evaluation, the reader is directed to the references in Table 1 and to other recent reviews [11–14].
The kinase domains of HER1, 2, and 4 are structurally similar to other kinases . As shown schematically in Figure 2, the kinase domains contain an N-lobe comprised mostly of anti-parallel B-strands and a C-lobe comprised mostly of alpha-helices. The active site sits in the cleft between the N- and C-lobes, called the hinge region. Common features of the kinase active site include an ATP-binding pocket which is homologous among kinases, a more variable substrate binding site, and two regulatory regions called the Activation loop (located on the C-lobe) and the C-Helix (on the N-lobe). In the inactive conformation of the kinase domain, the C-helix, containing a catalytic glutamate residue, is pointed away from the active site. In addition, the Activation loop occludes the substrate binding site. Upon activation of the kinase, the C-helix rotates ~90 degrees to position the glutamate residue, and the Activation loop extends away from the C-helix, thereby exposing the substrate binding site. The small-molecule inhibitors described in this review contain a heterocyclic core that mimics the shape and hydrogen-bonding of ATP (Figure 3). Most TKIs bind to the active conformation, though there are therapeutically important examples of kinase inhibitors that bind to the inactive conformation [16,17] and/or gain selectivity through contacts with the substrate binding site .
The effort to identify small molecule inhibitors of HER family kinases began in the early 1990s with the identification of natural compounds, such as erbstatin, with activity against HER kinases. One of the first classes of synthetic compounds, called “tyrphostins,” was based on the structure of erbstatin and was designed to compete with the tyrosine substrate . Synthesis of hundreds of these benzylidene malononitrile compounds yielded micromolar inhibitors with relative selectivity for HER kinases, including EGFR and HER2. Further studies identified compounds that even showed selectivity between EGFR and HER2 in vitro . This is despite 80% homology in the kinase domains of EGFR and HER2. These EGFR and HER2 selective compounds led to the first observation that compounds showing EGFR or HER2 selectivity in several different in vitro assays did not appear to show such selectivity in cell based assays . This paradoxical finding has been reproduced with all subsequent generations of HER TKIs (see below). Ultimately this class did not yield compounds with the potency or selectivity suitable for clinical development.
The field was revolutionized in the mid 1990s with the identification of a new generation of potent and selective classes of compounds. The best described of these classes are the 4-anilino quinazolines (Figure 4), which were simultaneously reported by Zeneca Pharmaceuticals and Parke-Davis Pharmaceuticals. Enzymological studies of the EGFR kinase suggested a ternary complex intermediate, in which ATP and the protein substrate bound simultaneously to the kinase, and in which the ATP γ-phosphate, tyrosyl hydroxyl, and the tyrosyl aromatic ring all interacted with the protein during catalysis . Querying a three-dimensional structure database for compounds that mimic these three interactions identified 4-anilino-quinazolines as low nanomolar, ATP-competitive inhibitors of EGFR kinase . Interestingly, while the aniline group was intended to mimic tyrosine, these compounds are noncompetitive with peptide substrate. High-throughput screening for inhibitors of EGFR kinase also identified 4-substituted quinazolines as highly potent and selective inhibitors of EGFR kinase . Strategic substitutions of these bicyclic compounds increased potency to the picomolar range while maintaining selectivity . A number of 4-anilinoquinazolines have been developed for clinical use including gefitinib , erlotinib , and lapatinib [26,27] (see Table 1).
The structure-activity relationship between 4-anilinoquinazolines and HER kinases has been described (eg ). The quinazoline bicycle binds in the ATP binding site; N1 hydrogen bonds to the main chain NH of methionine in the hinge region, and N3 forms a water mediated hydrogen bond with the side chain of threonine 766 (in the active conformation of EGFR, see below) . The 4-anilino group nestles in a hydrophobic pocket behind the ATP site, and substitutions on this ring play a significant role in kinase selectivity. Early studies suggested that small, hydrophobic substitutions at the 3 position increased affinity for EGFR [23,28], but large substitutions are also tolerated and are correlated with increased affinity for HER-2 [17,27,30,31]. The HER kinases prefer electron-rich substituents at the 6 and 7 position of the quinazoline ring, and ether substitutions are often found at these positions . However, the SAR is quite flexible at this edge of the quinazoline, and these are common sites for manipulating the compound’s physical chemical properties and, ultimately, their activity in vivo (e.g. ).
The structural features of quinazoline binding to the EGFR kinase domain have been determined thus far for erlotinib , gefitinib , and lapatinib . These compounds inhibit EGFR similarly, with IC50 values of 27 nM, 2 nM, and 11 nM for erlotinib, gefitinib, and lapatinib, respectfully   . In all three structures, the anilino-quinazolines bind at the ATP site, with N1 of the quinazoline bonding with the backbone carbonyl of a methionine residue in the hinge (Figure 3, ,4).4). As predicted , N3 forms a water-mediated hydrogen bond to a threonine side-chain, and the anilino group binds within a hydrophobic pocket . The structures in complex with erlotinib and gefinitib show the kinase in the active conformation [29,33]. By contrast, the structure in complex with lapatinib shows EGFR kinase in the inactive conformation . The bulky anilino substituent of lapatinib reaches deep into a back-pocket that is seen only in the inactive conformation (Figure 5). The compound appears enclosed by the protein, and the c-terminal tail of EGFR blocks the opening of the inhibitor binding site. As such, dissociation of lapatinib from EGFR likely requires a conformational change in the kinase; consistent with this prediction, lapatinib has a markedly slow off-rate in vitro and shows long-lived suppression of EGFR autophosphorylation in cells after washout .
In addition to the quinazolines, at least four other bicyclic compound classes have been identified as potent and selective inhibitors of HER kinase (Table 1, Figure 4). Though there is less published information on the chemical development of these classes compared to quinazoline, they appear to follow similar structure-activity relationships and to bind to EGFR analogously to the quinazolines. Pyridopyrimidines  and pyrrolopyrimidines  were both reported in the mid-1990s. Novartis has advanced the pyrrolopyrimide AEE-788 to clinical trials; this compound is described as an EGFR/VEGFR dual family inhibitor . The crystal structure of AEE-788 bound to EGFR, shown in Figure 6, indicates that it binds analogously to gefitinib and erlotinib . More recently, compounds with a pyrrolotriazine core have also been described; BMS-599626 is a clinical-stage example of this class . Finally, expanding on the idea that the N3 of quinazoline makes a water-mediated hydrogen bond with EGFR kinase, investigators at Wyeth-Ayerst Research replaced this nitrogen with a nitrile group that could hydrogen-bond directly to the threonine side chain . These cyanoquinolines have been developed as covalent inhibitors of HER kinases (see below), and the most advanced compound, HKI-272, is currently in clinical trials [39,40].
Irreversible inhibitors offer greater potency and durability of target inhibition compared to their reversible counterparts, and may also show a different pattern of disease resistance . Investigators at Parke-Davis generated irreversible inhibitors by adding an alkylating group to the 6- or 7- position of quinazoline-based compounds . These compounds were found to permanently and selectively inactivate HER kinases by covalently binding to a cysteine residue within the ATP pocket . Solubilizing side chain modifications at the 7 position of the quinazoline yielded orally bioavailable compounds such as PD183805 (CI-1033) and CL-387,785 [43,44]. These irreversible inhibitors showed highly promising anti-tumor activity in mouse xenograft models, and potent and prolonged inactivation of their HER targets, promoting them into clinical development. Aveo Pharmaceuticals and Mitsubishi have an a related compound, MP-412 (AV-412) in early clinical development . Analogously, alkylating substitutions at the 6 position in the pyridopyrimidine  and cyanoquinoline [34,40] scaffolds have yielded more potent and irreversible inhibitors of EGFR and HER2 without compromising selectivity. EKB-569 and HKI-272, from Wyeth-Ayerst, are irreversible cyanoquinolines currently in clinical development [30,39,41,47,48]. The recently published structure of HKI-272 bound to EGFR (Figure 6) shows the kinase in the inactive conformation, analogous to the structure of EGFR bound to lapatinib [17,49]. The structure suggests that binding to the inactive conformation might be a general feature of compounds with a bulky group on the aniline; furthermore, the structure indicates that alkylation of the cysteine forces modest changes in the orientation of the compound in the active site. Although the potential for nonselectivity raises safety concerns with the clinical use of irreversible inhibitors, such TKIs have thus far shown acceptable toxicity profiles in clinical studies [47,50,51]. The irreversible inhibitors have activities in clinical studies that are more modest compared to preclinical models, similar to many other classes of drugs.
Although nearly all of the compounds listed in Table 1 inhibit EGFR and HER2 kinase activity, most of them favor EGFR over HER2. Designing HER2 selective inhibitors is made more challenging by the absence of a crystal structure of the HER2 kinase domain. Investigators at GlaxoSmithKline pursued a dual screening program to identify compounds equally active against EGFR and HER2 . They found that addition of a bulky substituent, such as benzyl ether, to the 3-position of aniline increases potency against HER2 while maintaining activity against EGFR . One such quinazoline derivative GW572016 (lapatinib) has been clinically developed for the treatment of HER2-amplified breast cancer . In fact, addition of a bulky substituent consistently increases activity towards HER2. For instance, evolution of EKB-569 in the cyanoquinoline series yielded HKI-272 and HKI-357, which are equipotent for EGFR and HER2 and are currently in clinical testing . Additionally, HER-2 activity in the pyrrolopyrimidine series is associated with addition of phenethylamine to an analogous exocyclic amine. While it is not yet known why bulky substitutions on the aniline increase binding to HER-2 kinase, it is noteworthy that the structures of lapatinib and HKI-272 both show EGFR in the inactive conformation [17,49]. It is possible that HER-2 prefers to bind inhibitors in the inactive over the active conformation, or that the HER-2 binding site has a larger hydrophobic pocket even in the active conformation. Detailed analysis awaits determination of the HER2 kinase domain structure.
Common toxicities of HER TKIs include skin rashes and diarrhea which could be mediated through EGFR . Pfizer and OSI therefore pursued inhibitors of HER2 that were inactive against EGFR [31,53]. As with the GlaxoSmithKline and Wyeth-Ayerst studies, these investigators found that bulky substitutions at the 4-anilino position afforded HER2-selective inhibitors, including CP-654577 and the clinical candidate CP-724714. In a phase I study of CP-724714 diarrhea was not reported, but skin rashes and liver toxicities were observed . So far, these are the only disclosed compounds that prefer HER2 over EGFR. It is not yet known why CP-724714 shows low activity towards EGFR, though it does lack a hydrogen-bond acceptor at the 7-position, which has previously been found to be important for binding to EGFR. Detailed structural studies such as x-ray crystallography and computational modeling will help to address the selectivity of compounds for HER-2 kinase over EGFR.
Since the earliest days of the development of HER family TKIs, the concept of family member selectivity has been complicated by discrepancies between in vitro and cell-based data. Although certain compounds clearly show much higher potency against purified EGFR kinase compared with purified HER2 kinase, these differences are much less apparent in cell based assays. The EGFR-selective quinazoline gefitinib shows equipotent growth inhibitory activity against EGFR-overexpressing and HER2 overexpressing tumor cells . Both EGFR-mediated (EGF-stimulated) and HER2-HER3 mediated (neuregulin-stimulated) signal transduction is inhibited by gefitinib in cells . These non-discriminatory cellular observations have been extensively reproduced with several of the quinazoline HER inhibitors including gefitinib, erlotinib, and AG1478 [55–58]. This effect is due to the direct inhibition of HER2 kinase by these compounds, as demonstrated by elegant models established in EGFR-negative cells confirming the direct inhibition of HER2 kinase by erlotinib . Virtually all HER TKIs, regardless of their in vitro selectivities, show growth inhibitory efficacy against HER2-driven tumor cell lines in xenograft models. It remains unclear why selectivity in cells and tumor models is much less than observed in vitro. It is possible that accumulation of compounds in cells raises the intracellular concentration above concentrations in biochemical assays. It is even more likely that purified kinase domains in in vitro reactions do not faithfully reflect the biochemical properties of their cellular counterparts, and protein orientation, position of the carboxy termini, and dimerization events in the cellular context may all be pharmacologically relevant parameters.
HER TKIs are mostly cytostatic in cell culture models. In the specific case of HER2-amplified breast cancers, this falls short of expectations, since these tumors are known to be highly HER2-dependent, and transgene-inducible models show apoptotic tumor cell death when the HER2 transgene expression is withdrawn . The reasons underlying this discrepancy are only now becoming apparent. Recent structural and biological insights into HER2 function suggest that the tumorigenic function of HER2 is critically dependent on its kinase-inactive dimerization partner HER3 . While in the most simplistic model, HER3 can be thought of as merely a substrate for the kinase reaction, the HER2-HER3 interaction appears to be much more intricate. The recent analysis of the EGFR kinase structure reveals a unique mode of activation in this family whereby one partner in the kinase dimer performs a stimulatory function through an asymmetric c-lobe to n-lobe interaction with the other partner . Asymmetric dimerization results in activation of the recipient partner, which in turn performs a catalytic function, transferring phosphates onto the carboxy tail of the stimulatory partner. While only the structure of the EGFR dimer has thus far been described, the highly homologous dimerization interfaces suggest that this mechanism applies to the whole family. As such, it is likely that HER2 and HER3 have evolved to optimize their catalytic and stimulatory partners, suggesting a previously unexplained evolutionary advantage to the loss of catalytic function in HER3.
The power of this highly evolved activation function became apparent when it was discovered that HER family TKIs are much less potent at inactivating the HER2-HER3 signaling complex compared with EGFR or HER2 homo- or hetero-dimeric signaling activities, significantly undermining their anti-tumor effects . Feedback signaling can significantly induce HER3 expression and membrane localization, thereby buffering HER2-HER3 signaling against incomplete inactivation of HER2 catalytic function . Therefore it appears that inactivation of the HER2-HER3 oncogenic signaling complex requires much more potent inhibitors that can completely inactivate HER2 catalytic function. Therefore, the modest clinical anti-tumor activities of current TKIs is entirely consistent with the fact that within their therapeutic index they can only partially inhibit HER2-HER3 signaling. While irreversible inhibitors can completely inactivate catalytic function in cell culture models, their off-target effects may limit their therapeutic indices, and it remains to be determined whether they can be administered to patients in high enough doses to fully inactivate tumor HER2.
Several reversible and irreversible small molecule HER2 inhibitors from structurally distinct classes have been developed and we will know the clinical activities of these classes of compounds within the next few years. But the current structural, biological, and preclinical studies have already provided the necessary hints that there is more to be done to fully reverse the robust tumorigenic powers of HER2. Combination therapies are one avenue that will be pursued in the near future. Allosteric inhibitors of HER2-HER3 transactivation represent another new strategy to target this resilient oncoprotein complex. Antibody-based strategies targeting the extracellular functions of HER2-HER3 signaling also continue to be pursued. The increasing interest in and the broad scope of endeavors to target HER2 reflect an ever increasing understanding of the value of this target in cancer therapeutics.
The authors would like to acknowledge Katherine Augustyn for Figure 2 and Daniel Gray for helpful discussion.