Molecular Pathways and Mechanisms of HER2 in Cancer Therapy

ERBB2, commonly referred to as HER2, is a proto-oncogene (chromosome 17q21) that encodes a transmembrane receptor tyrosine kinase involved in cell growth and differentiation (Fig. 1). It was one of the first genes identified in the 1980′s efforts to discover oncogenes and was quickly shown to have relevance to diverse human cancers (1, 2). What followed was an explosion of scientific inquiries that validated the role of HER2 in carcinogenesis and led to development of a variety of targeting approaches, which have paved the way in transforming drug development. The oncogenic potential of HER2 was evident in the early days of mouse genetic engineering with several models of HER2-driven mammary tumorigenesis showing an aggressive and metastatic disease biology (3). As more elegant inducible mouse genetic models were developed, HER2 was shown to be continuously required throughout the entire disease process from initiation through advanced metastatic disease, nicely illustrating the concept of oncogene addiction and the potential of this oncogene as a target for cancer therapeutics (4). Advances in sequencing technologies brought about the era of widespread tumor genome sequencing with findings that extended the scope of HER2 amplification to include substantial subsets of gastric and esophageal cancers as well as smaller subsets of colon, bladder, endometrial, lung, salivary gland, and other cancers (Fig. 2; refs. 5–7). These surveys also discovered much smaller subsets of cancers harboring activating HER2 mutations without amplification (8). Whether these HER2 mutants are dominant tumor-drivers like HER2 amplifications remains to be determined.

The technology to generate mAbs and to humanize them for repeated administration as therapeutic agents evolved in the late 1980s and HER2 was one of the first targets explored by this novel pharmaceutical technology (9). Trastuzumab, the anti-HER2 mAb, entered clinical phase in 1992, eventually leading to a pivotal phase III study in HER2-amplified metastatic breast cancer, and landmark studies in HER2-amplified early-stage breast cancers, paving the way and establishing the role of mAbs in treatment of human cancers (10–13). As understanding of structure and function among the HER family receptors evolved in 1990s, these insights were applied to development of structure-guided mAbs such as pertuzumab to disrupt HER2 dimerization more effectively with additional, albeit incremental benefits demonstrated in clinical studies (14, 15). Technologies to arm these mAbs with cytotoxic agents evolved in late 1990s with improvements in linker chemistries and HER2 was among the early targets pursued leading to development of the antibody–drug conjugate (ADC) ado-trastuzumab emtansine (T-DM1; ref. 16). While the naked HER2 mAbs, trastuzumab or pertuzumab, showed only limited efficacies and their clinical benefits were mainly manifest in combination regimens with chemotherapeutics, T-DM1 was found to be highly efficacious in monotherapy with activities surpassing chemotherapies in treatment refractory tumors with a much more favorable toxicity profile (17). This remarkable breakthrough fueled efforts to further explore ADCs leading to development of trastuzumab deruxtecan (DS-8201 or T-Dxd; ref. 18). Exploring T-Dxd revealed unforeseen findings of non–cross-resistance with T-DM1 and much broader clinical activity in other HER2-amplified cancers and in cancers with lower expression of HER2 (19–22). The high efficacy of HER2 ADCs has led to development of a plethora of ADCs, now in trials.

As technologies to develop selective tyrosine kinase inhibitors (TKI) evolved, the HER family of kinases including EGFR and HER2 were among the first targets pursued leading to many compounds proceeding to clinical testing in HER2-amplified breast cancers. Of these, the pan-HER reversible TKI lapatinib, pan-HER irreversible TKI neratinib, and HER2-selective reversible TKI tucatinib have made it to clinical use with many others in development. TKIs have only modest activities in monotherapy of HER2-amplified cancers and their clinical use has been predominantly focused on combination therapies (23, 24). In contrast, the HER2-mutated subtypes of breast and other cancers appear amenable to TKI monotherapy as shown in the basket neratinib study (25).

We now have a good understanding of mechanisms by which HER family of receptors function. Signals are generated when two receptors from the family come together in a homo- or hetero-dimerization configuration. The receptor extracellular domains (ECD) are generally restrained by a closed conformation that is prohibitive to dimerization and it is the binding of extracellular ligands that exposes an interface that promotes dimerization (26). HER2 is unique in that it lacks this self-restraint or a physiologic ligand, and its ECD is always poised for dimerization (27). As such, the only physiologic restraint built into HER2 is its low expression, which is overwhelmed in cancer cells through amplification and overexpression. Dimerization leads to phosphorylation of C-terminal tails and consequent initiation of second messenger signaling pathways (28). A particularly strong relationship between HER2 and HER3 as dimerization partners is apparent from in vitro signaling assays (29). Many lines of investigation also confirm that HER3 is an important HER2 partner in HER2 overexpressing tumors (30–32). These structural insights provide ample mechanistic rationale to inactivate HER2 signaling through ECD targeting antibodies that disrupt dimerization or through TKIs that inactivate catalytic activity and eliminate phosphorylation. Yet, the story has turned out to be much more complicated.

The translational research in pursuit of mutationally activated oncogene-driven cancers such as EGFR- or ALK-mutant lung cancer among others has followed a somewhat direct path with the rational expectation that the catalytic inactivation of these disease drivers should be a highly effective treatment strategy. However, the pursuit of amplified HER2 as a target has followed a much more tortuous and unexpected path with many mechanistic complexities and ambiguities, new insights learned, and new opportunities discovered. It is evident now that in most cases, HER2 is activated through massive overexpression rather than mutation and this makes for a mechanistically more challenging target, something that was not readily anticipated in 1990s. While in the early days following development of trastuzumab and pertuzumab it was thought that these mAbs could interfere with HER2 signaling by eliminating their expression or disrupting dimerization, this was clearly not the case as was evident in simple in vitro studies in HER2-amplified cancer cells (33–37). Pertuzumab was specifically developed based on a structure-based design to bind the HER2 dimerization interface and clearly does inhibit signaling in physiologic cell systems with low HER2 expression but fails to do the same in HER2-overexpressing cancer cells (35–38). Similarly, a variety of rationally designed HER2 or HER3-targeting mAbs, bispecific mAbs, DARPins, and other biotherapeutics have shown only limited ability to interfere with HER2-HER2 or HER2–HER3 signaling in HER2-amplified cancer cells. Experimental models now more clearly demonstrate the mechanistic futility of these approaches.

The kinase domain interactions in these cancers are driven entirely by massive HER2-overexpression with no promoting or restraining functions from the ECDs essentially uncoupling intracellular signaling from the ECDs (Fig. 1; ref. 37). Even strategies to add bulk to the ECDs and prevent receptor proximation leave kinase domain signaling intact due to conformational flexibilities across the span of these receptors and the curvatures in the plasma membrane (Fig. 1; ref. 37). Although mechanistic studies with trastuzumab failed to show effective inhibition of HER2 signaling, preclinical and clinical studies do show measurable antitumor effects (11, 39). This has fueled efforts to understand the mechanistic basis for the observed in vivo effects and a large body of work now demonstrates this to have a substantial immunologic basis. This was initially shown through elegant mouse studies showing the critical role of host Fc receptor in mediating the in vivo antitumor activity of trastuzumab and subsequently shown to encompass a wider repertoire of immunologic activities including innate and adaptive immune functions, memory, and cytokines (40–43). Capitalizing on these insights, enhancements in the Fc portion of trastuzumab to optimize its immunologic activities has produced margetuximab with superior clinical activity compared with trastuzumab (44).

In contrast to mAbs, HER2 targeting TKIs do effectively inactivate HER2 signaling in cell-based studies in HER2-amplified cancer cells, leading to apoptotic cell death (45, 46). Yet these agents fail to show substantial activities in patients as monotherapy (47, 48). Mechanistic studies highlight limitations in target inactivation, in particular with regards to inactivation of HER2–HER3 heterodimer signaling. In HER2-amplified cancers, HER3 is tightly linked with downstream PI3K–AKT signaling in a pathway that involves robust compensatory feedback regulation. As such, inhibition of HER2 with TKIs leads to a compensatory upregulation of HER3 unleashing a 100-fold reserve in signaling capacity that overpowers an incomplete inhibition of HER2 kinase and restores HER2–HER3 signaling (49–51). Only the complete inactivation of HER2 kinase can durably inactivate HER2–HER3 signaling in these cancers, and although this is feasible in cell culture models at higher concentrations of TKIs, it is beyond the therapeutic index of all current TKIs (50). High dosing of oral TKIs in patients has been studied but limited by a bioavailability ceiling (52). Irreversible TKIs, such as neratinib, that covalently bind their target have much higher molar potency, but this comes at a cost of substantial off-target activities, within and outside of the kinome, significantly limiting therapeutic index (53–56). A rational strategy for increasing potency would be co-targeting HER2 and HER3, but HER3 is a challenging target for pharmaceutical inactivation. Its function in HER2-amplified cancers is engaged in a ligand and ECD-independent manner limiting the efficacy of ECD-targeting mAbs in this disease, and its kinase domain functions in allostery, not catalysis, and conventional TKIs binding within its ATP pocket have no effects on its signaling function (37, 57). Novel approaches to target HER3 through degradation are being pursued and may provide breakthroughs in this arena (58, 59).

Although the massive overexpression of HER2 makes it a challenging target for inactivation, its surface expression provides significant opportunities for targeted delivery of cytotoxic molecules, radioisotopes, liposomes, nucleic acids, and other moieties designed to kill cancer cells. In this regard, the pursuit of HER2-targeting ADCs has proven particularly fruitful. In contrast to HER2-targeting mAbs and TKIs, which underperformed expectations as oncogene inhibitors when they entered the clinical arena, HER2 ADCs have overperformed expectations, and their success has spawned many lines of study to better understand the mechanistic basis for their activities and fueled substantial investments in the pharmaceutical sector with many new agents in development. The science underlying exact mechanisms of activity of ADCs continues to evolve. The key variables appear to be the linker chemistry, the activity, cell permeability and potency of the cytotoxic payload molecule, the payload-to-antibody ratio, and the characteristics of the mAb, all of which contribute to the observed clinical activities of these agents (60, 61). Although the initial and purest vision for the development of ADCs was for the most precise and protected delivery of cytotoxic molecules intracellularly to cancer cells, it is now apparent that reducing the stringency of this vision can potentially increase, not decrease, their therapeutic index. Using cleavable linkers that more readily release their payloads or cell permeable cytotoxic agents that diffuse out of cells can lead to exposure beyond the target cells including non-targeted surrounding cells, a so-called bystander effect, and a measurable low exposure in the systemic circulation (Fig. 1; ref. 62). Clinical exploration of the cleavable HER2-targeting ADC, T-Dxd, has revealed an unexpected range of activity. The highest efficacy is still seen in patients with classic “HER2-high” (HER2-overexpressing/amplified) subtypes of cancers, consistent with the original and simplest concept of ADCs as agents for precise tumor-targeted cytotoxic delivery (19–21). However, pursuing an assumption that cancers with lower expression of HER2 may also afford a therapeutic index by mechanisms involving heterogeneous expression and bystander effects, significant clinical efficacy was also observed in breast cancers with “HER2-low” expression (IHC 1+ or 2+; ref. 22). Further exploration has shown no lower cutoff for this quantitative biomarker with efficacy spanning the entire spectrum from 0 to 2+ expression and across the hormone receptor status in breast cancer (19, 63). This wide range of activity is also seen in HER2-mutant lung cancers wherein, there is no correlation of activity with tumor HER2 expression (64). The observed broad and marker-independent clinical activities of T-Dxd have upended the mechanistic hypotheses regarding the mode of action of ADCs. Tumor-specific targeted delivery of cytotoxic payloads related to high expressing and bystander effect in lower expressing cancers can only partially account for this observed broad range of activity and the mechanistic basis remains to be determined. Traditionally, mechanism of action forms the starting point for development of rationally designed therapeutics. But clinical development of T-Dxd and similar ADCs has clearly exited this mechanistic orbit and their clinical exploration continues in an open space until a ceiling is encountered, while mechanistic studies will have to follow to fill this ever-enlarging knowledge gap.

The newly discovered broad clinical activity is not limited to T-Dxd and appears to be a class effect. Preliminary evidence from other HER2-targeting ADCs appears to show similar broad range of efficacies (65–67). This property of ADCs is also not unique to the target HER2 and appears to apply to other targets as well, as is evident with the experience with Trop-2 targeting ADCs. While, initially explored in triple-negative breast cancer which has the highest expression of Trop-2, its efficacy has poor correlation with Trop-2 expression levels and shows similar clinical activity in hormone receptor–positive breast cancers with lower expression of Trop-2 (68–71). A plethora of other ADCs in the investigational pipelines exploring other types of linkers, payloads, release properties, and antibodies and drug–antibody ratios will further expand the body of data available and allow us to formulate more informed hypotheses regarding potential mechanisms of action. Although initial pharmacokinetic studies of T-Dxd showed low levels of free deruxtecan in circulation, the role of systemic release products must be revisited (72, 73). Much of the mechanistic foundation for the broad clinical activities of these ADCs remains unknown and eagerly awaits new hypotheses and experimental studies.

Distinct from cancers driven by amplification and overexpression of HER2, there are also rarer cancers that harbor somatic mutations in HER2 (74, 75). Most, but not all, are within the kinase domain and generally result in increased catalytic activity of the HER2 kinase. Individual mutations appear to have different characteristics with respect to catalytic activities and partner preferences and exhibit cell-context dependent characteristics (74). Most experimental studies have used overexpression systems and in these artificial systems, many are more potent oncogenes compared with wild-type HER2. But in human cancers they are typically mutated in copy-neutral fashion without overexpression and whether they are primary disease drivers in human cancers remains speculative. The irreversible TKI neratinib has modest but short-lived activity in HER2-mutant cancers, confirming a limited biologic role for at least some of these mutants in the human disease. The diverse nature of these HER2 mutants makes for a complex arena for analysis and it remains difficult to know whether nonresponsive mutants are not biologically relevant or whether they harbor intrinsic resistance to TKIs. There is much more to be learned about the biology of HER2-mutant cancer.

Distinct from HER2-amplified and HER2-mutant cancers are exceedingly rare cancers driven by ligand activation of HER2. These appear to be driven by genetic fusions of the NRG1 gene leading to neuregulin fusion proteins expressed at the plasma membrane (76, 77). Anecdotal reports of these rare cases suggest that these tumors are responsive to HER2 TKIs or HER2 or HER3 targeting mAbs that block dimerization or ligand binding (78, 79).

The field of HER2-amplified breast cancer has benefited greatly from decades of translational cancer research through the development of numerous targeted therapies that have made substantial impact in the clinical realm. Addition of trastuzumab and pertuzumab to chemotherapy significantly improves outcomes in patients with metastatic or early-stage breast cancer and is now the standard of care for the treatment of HER2-amplified breast cancer at all stages of disease (11–13, 15, 80). T-DM1 has shown significant clinical activity in trastuzumab-resistant disease and has become a standard second-line salvage option in patients with metastatic disease or in early-stage patients who have significant residual disease following neoadjuvant therapy (17, 81). The second generation ADC, T-Dxd, has shown clinical activity superior to T-DM1 and activity in disease that is resistant to T-DM1 (82). As discussed previously, T-Dxd has a much broader range of clinical activity that encompasses HER2-negative breast cancers, including cancers traditionally labeled HER2-low (22, 63). This broad range of activity appears to somewhat uncouple T-Dxd from its target biomarker and it may be best considered a broadly active breast cancer agent without a biomarker association, exhibiting clinical activity in breast cancers spanning HER2-positive and -negative and hormone receptor–positive and –negative subtypes, albeit with superior activity against HER2-amplified cancers. This has sparked substantial interest and numerous other investigational HER2 targeting ADCs are exploring this newly discovered terrain (83).

The HER2 TKIs lapatinib or tucatinib in combination with capecitabine are active regimens for the treatment of metastatic HER2-amplified breast cancers refractory to prior HER2 therapy, providing additional options for later lines of therapy (23, 24). The combination of lapatinib and trastuzumab has clinical activity that provides a non-chemo option for patients with low disease burden, indolent disease, or not suitable candidates for chemotherapy (84).

The HER2-amplified subtype of breast cancer has a higher predilection for brain metastases and numerous studies have attempted to define the activities of HER2-targeted therapies in the treatment of brain metastases (85). Small-molecule HER2 TKIs have better central nervous system (CNS) penetration than mAbs and these and have been actively pursued for CNS activity. The CNS activity with lapatinib is minimal, but more significant activity is evident with neratinib in the treatment of brain metastases (86). The highest CNS activity is seen with the brain-penetrant TKI tucatinib and the tucatinib, capecitabine, trastuzumab combination has become a standard in the management of patients with brain metastases (87). Although the activity of ADCs in the CNS compartment were thought to be restricted by the large size of their mAb component, hints from subset analyses of their clinical studies suggested these agents may have activity in the CNS and this was followed by specific clinical studies showing modest activity with T-DM1 (88–90) and substantial intracranial activity with T-Dxd (91, 92).

HER2-overexpression/amplification, to date, remains the only clinically usable biomarker for selection of targeted therapy in metastatic gastroesophageal adenocarcinoma (GEC) and is seen in about 12% to 20% of cases (Fig. 2; refs. 6, 7, 93). HER2 amplification is seen in a greater proportion of patients with liver metastasis, gastroesophageal junction (GEJ) adenocarcinomas and intestinal subtype tumors, although its prognostic significance is unclear (93, 94). Compared with circumferential staining seen in breast cancer, HER2 expression in GEC is predominantly basolateral/lateral and has more notable intratumoral heterogeneity, making testing (ToGA criteria) and patient selection, more challenging (93, 94).

Anti-HER2 therapy is mainstay of therapy for advanced HER2-overexpressed/amplified GEC. Addition of trastuzumab to chemotherapy for first-line treatment of HER2-positive advanced GEC demonstrated a clear survival benefit over chemotherapy alone (94). However further efforts, as in the case of breast cancer, such as addition of dual anti-HER2 therapy with pertuzumab and T-DM1, showed limited activity and failed to significantly improve overall survival in patients with HER2-positive metastatic GEC (95, 96). More recently, addition of pembrolizumab to trastuzumab and chemotherapy significantly improved objective response rate (74% vs. 52%) highlighting the immune interactions of the HER2 pathway (97). T-Dxd has also shown significant activity in treatment of HER2-positive metastatic GEC after progression of first-line anti-HER2 therapy with a significantly higher response rate (51%) compared with physician's choice of chemotherapy (14%) and longer overall survival (median: 12.5 vs. 8.4 months; ref. 21).

Unlike breast and gastric cancer, HER2 amplification are rare in lung cancer (0.88% cases) and HER2 mutations are the driver events in 3.5% cases, comprising 80% of all HER2 alterations (Fig. 2; refs. 6, 7). These HER2 mutations, specifically exon 20 mutations are analogous to in-frame EGFR deletions and dysregulate the HER2 pathway due to constitutive tyrosine kinase activity, are enriched in nonsmokers, adenocarcinomas, and are mutually exclusive of other oncogenic mutations in non–small cell lung cancer (NSCLC; ref. 98). HER2 mutations, barring amplifications and overexpression, appear to be key determinant of response to HER2-targeted therapies. No clinical benefit was observed with addition of trastuzumab to chemotherapy in HER2-overexpressed/amplified NSCLC (99). This aspect of aberrant HER2 pathway is unique to lung cancer. Although, HER2 kinase domain mutations are also seen in other tumor types, such as breast (4.3%), gastric (5.0%), and colorectal (2.9%) carcinomas, they are often missense mutations (not deletions/insertions) and involve exons other than exon 20, indicating tissue-dependent oncogenic mechanisms (8). In addition, HER2 mutation are not associated with HER2 overexpression/amplification (100).

HER2-mutant lung cancer responds to HER2 inhibitors, especially small-molecule TKIs and ADCs (101). Although, selective HER2 TKIs have yielded response rates of 20% to 35% in single-arm phase II studies, many of these have shown limited overall clinical benefit. Recent evidence has brought HER2 ADCs to the forefront in treatment of HER2-mutant NSCLC. T-DM1 demonstrated a response rate of 44% (N = 18) in treatment refractory HER2-mutant lung cancer, across a variety of HER2 exon 20 insertions and point mutations (in kinase, transmembrane, and ECDs) and regardless of HER2 expression or amplification status (102). Similarly, T-Dxd showed a confirmed objective response in 55% patients (N = 91) with refractory metastatic HER2-mutant NSCLC, again notwithstanding HER2 expression or amplification (64). Currently there is no mechanistic paradigm to account for this activity of ADCs and this is an active area of pursuit as discussed previously.

HER2 overexpression/amplification is seen in 2% to 3% of all colorectal cancers and represents a very distinct subset of colorectal cancer (Fig. 2; refs. 6, 7). Preclinical and retrospective clinical evidence shows that HER2 amplification is enriched in RAS/BRAF wild-type tumors (5%–6%) compared with RAS-mutant colorectal cancer (1%–2%; refs. 103, 104). In RAS//BRAF wild-type metastatic colorectal cancer (mCRC), HER2 amplification appears to be a negative predictive biomarker of response to the anti-EGFR mAb-based therapy, which is the current standard of care in these patients with mCRC (104). Although a diverse array of HER2 somatic mutations are seen in a very small subset of colorectal cancer, only a small number appear to have oncogenic potential and show response to irreversible TKIs (such as neratinib and afatinib), and as such they have not been exploited clinically (105).

As opposed to gastric cancer, single agent anti-HER2 therapy shows limited efficacy in HER2-amplified mCRC (103, 106). However, dual-HER2 targeting with trastuzumab combined with lapatinib, pertuzumab, and tucatinib has shown robust activity in treatment refractory mCRC with response rates ranging from 28% to 40% in RAS wild-type tumors (106–109). Notably, dual anti-HER2 therapy benefits only patients with RAS wild-type tumors (response rate of 40% vs. 8% for RAS-mutant cases; ref. 108). Along similar lines although the HER2 ADC, T-DM1 showed very limited activity (response rate 9.7% combined with pertuzumab) in HER2-amplified mCRC, trastuzumab deruxtecan (DS-8201) showed very promising and durable activity in HER2-amplified mCRC refractory to standard treatment with response rate of 45%, including those who had received prior anti-HER2 therapies (20, 110).

In addition to tumor types mentioned above, HER2 alterations can be seen in a variety of tumor types such as salivary gland tumors (particularly non-adenoid cystic carcinoma, nonsecretory type), hepatobiliary tumors (principally gallbladder cancer) and others (Fig. 2). Dual HER2 targeting with trastuzumab and pertuzumab showed a response rate of 23% (N = 39) in HER2-amplified metastatic biliary tract cancer (111). A phase II trial of trastuzumab and docetaxel in HER2-amplified salivary duct carcinoma (N = 57) showed an overall response rate of 70.2% and is the preferred treatment option for these rare cancers (112). Basket trials of HER2-targeted therapies to engage tumor types with lower prevalence of HER2 alterations are ongoing.

Numerous basket studies and disease-specific phase II studies of TKIs have been conducted in cancers harboring somatic mutation of HER2 (Fig. 2; refs. 6, 7). As a class, TKIs do have activity, albeit modest, in these cancers although there is variation in activity according to disease type, specific mutations, and specific drug. The irreversible class of TKIs have clinically meaningful activity and neratinib has been studied the most with modest activity in breast and lung cancers but evidence of activity in other cancers (25, 101, 113–116). Pyrotinib and poziotinib also have similar efficacies while afatinib appears to lack efficacy (117–119).

The field of HER2 targeting has been an archetype for pan-cancer developmental therapeutics with promising clinical results, continuously evolving new mechanistic insights, profound translational efforts, and tremendous hope for patients suffering with cancer. While we have achieved unprecedented success, a substantial subset of patients derive limited benefits from current approaches and resistance develops eventually. Through a barrage of ongoing clinical trials (Table 1), three major themes are evolving: (i) therapeutics strategies involving combining anti-HER2 therapies with immunotherapy, inhibitors of DNA damage pathway, receptor cycling modulators among others which may work synergistically to increase the responses, (ii) targeting a large subset of patients who have tumors that were traditionally considered HER2-low and not amenable to past generation anti-HER2 therapies, and (iii) immune targeting of HER2 using novel immunotherapy approaches, such as CAR-T cells, and vaccines. No matter the strategy, one thing is for sure, HER2 targeting remains the epitome of precision cancer medicine in oncology.

K.P.S. Raghav reports personal fees and other support from Daiichi and other support from Genentech outside the submitted work. No disclosures were reported by the other author.

This work was supported by the NCI (P30 CA016672 to K.P.S. Raghav).