A growing family

Abstract

As the upward spiral of novel cancer gene discoveries and novel molecular compounds continues to accelerate, a repetitive theme in molecular drug development remains the lack of activity of initially promising agents when given to patients in clinical trials. It is however invigorating that a few targeted agents directed against a select group of a few ‘cancer gene superfamilies’ have escaped this all to common fate, and have evolved into novel, clinically meaningful molecular therapy strategies. Targeting dysregulated signaling of the epidermal growth factor family of transmembrane receptors (Erbb family) has encompassed over the last decade an ever increasing role in personalized treatment approaches in an increasing number of human malignancies. Erbbs are receptor tyrosine kinases that are important regulators of several signaling pathways. Two of its family members (Erbb1/EGFR and Erbb2/HER2) have previously been shown to be somatically mutated of human cancers. To determine if this family is somatically mutated in melanoma, its sequences were recently analyzed and one of its members, Erbb4, was found to be somatically mutated in 19% of melanoma cases. Functional analysis of seven of its mutations was shown to increase its catalytic and transformation abilities as well as providing essential survival signals. Similar to other Erbb family members, mutant Erbb4 seems to confer “oncogene addiction” on melanoma cells, making it an attractive therapeutic target. Gaining further understanding into the oncogenic mechanism of Erbb4 may not only help in the development of targeted therapy in melanoma patients but might accelerate the acceptance of a novel taxonomy of cancer which is based on the genomic perturbations in cancer genes and cancer gene families and their response to targeted agents.

General Introduction

Despite an ever increasing number of novel cancer targets and an explosion of new molecular and biological agents currently emanating from preclinical studies, genotype-directed therapy targeting the epidermal growth factor family of transmembrane receptors (Erbb family) remains one of the few prime models of successful personalized medicine. The journey of targeting EGFR mutations acting as driving oncogenic signals in lung cancer has now reached population-based, large scale screening efforts treating patients with non-small cell lung cancer (NSCLC) with the ‘classical’ EGFR L858R and exon 19 deletion mutations with the small molecule tyrosine kinase (TKI) inhibitors erlotinib or gefitinib.^1,2 Multiple studies have shown that the presence of activating EGFR mutations in NSCLC is associated with response to erlotinib and gefitinib treatment, improved progression-free, and median overall survival.^3–8 If the favorable outcome of NSCLC harboring EGFR mutations treated with erlotinib and gefitinib is solely due to the impact of TKI treatment and if EGFR mutations are solely a predictive factor for TKI response, or if EGFR mutations are also a prognostic factor for NSCLC outcome and reflect a more favorable tumor biology remains, however, to be determined.^9,10 In a recent large randomized study patients with EGFR mutations not treated with erlotinib or gefitinib had significantly longer survival rates than patients with wild-type EGFR tumors.¹¹ The predictive, and possibly prognostic, impact of activating EGFR mutations on cancer outcome has now been further dissected by demonstrating a substantial heterogeneity of these mutations in respect to TKI responsiveness and clinical outcome: of the two most common activating EGFR kinase-domain mutations, in-frame exon 19 deletions (Class I; 44% prevalence of EGFR mutation harboring NSCLC), and EGFR L858R mutations (Class II; 41%) response rates to erlotinib and gefitinib are twice as high in tumors with exon 19 deletions (70–100%) compared to tumors harboring L858R mutations in exon 21 (30–67%).^10,12,13 Similarly, patients with exon 19 deletion mutations treated with TKI had a median overall survival twice as long as patients with L858R mutations treated with gefitinib.^12,13 The biochemical and structural correlates of this different clinical behavior of the two mutants have now also been elucidated: for example, EGFR L858R mutations have a 10-fold higher K_M (μmol/L) and a two-fold lower Ki (nmol/L) for ATP than exon 19 deletion mutants,¹⁴ and structural data suggest that exon 19 deletions causing a greater shift of the αC helix narrowing the ATP-binding cleft of the kinase domain to a greater degree which increases the affinity of the mutated receptor for TKIs.¹⁵ Now, novel discoveries on the role of non-kinase domain mutations in the Erbb family and the role of Erb somatic mutations in other histologies is raising hopes for a true ‘dent’ on cancer mortality even higher.

Why are mutated Erbb receptor family members such a prime cancer target? All members of the Erbb receptor family, EGFR/Erbb1/HER1, Erbb2/HER2/neu, Erbb3/HER3, Erbb4/HER4, are known to play a pivotal role in cell-cell signaling and signal transduction regulating cell growth and development.^16,17 Members of the Erbb family are structurally very similar membrane-spanning tyrosine kinase receptors composing of an extracellular domain subdivided into four subdomains, an α-helical transmembrane segment, and an intracellular protein kinase domain (Fig. 1A).¹⁸ Erbb receptor activation requires dimerization of two Erbb molecules which normally exist as inactive monomers.^17,18 Ligand binding induces the release of intramolecular bonds (‘autoinhibition’) between the extracellular domains II and IV leading to a conformational twist which exposes the dimerization domain in subunit II of the receptor.^18,19 Pairing with another receptor leads to activation of the kinase domain, phosphorylation of the other receptor, and consequently binding and activation of downstream proteins (Fig. 1A).^18,19 There are, however, important differences between Erb receptors which are relevant in regard to their functional roles. Some of these are summarized in Table 1.^19,20 Among these differences are the lack of an intrinsic kinase activity of Erbb3 and the ligand-independent activation of Erbb2/HER2 which normally exists in the ‘open’ form being constitutively available for pairing and signal transduction.^17,21 In general, the nature of the activating ligand and the different homo- and heterodimer pairings are the major determinants which of the various downstream targets are activated.^17,22 Regulation of Erb receptor mediated signaling also occurs at the post-receptor level via ubiquitination, mediation of endocytosis, or neddylation.^17,22 Therefore, taking the multilayered regulatory complexity of the Erbb signaling network into account, it is not surprising that dysregulated Erbb signaling is causally linked with the development and progression of cancer, and cells afflicted by dysregulated Erbb signaling display at least four of the six hallmarks of cancer (metastasis, resistance to apoptosis, independence of growth signals, resistance to growth inhibitory signals).¹⁴⁸ To date, well established mechanisms of Erb dys-regulation include the amplification of the Erbb2/HER2 gene locus, SNPs within the promoter region or intronic sequences of ERBB4,^23–25 abnormal proteolytic cleavage of ligand and Erb isoforms,²⁵ and most notably, activating somatic mutations in the EGFR receptor kinase domain.³ Now, large-scale, high-throughput genotyping efforts, including studies conducted by the ‘Tumor Sequencing Project’ (TSP) initiative or ‘The Cancer Genome Atlas’ (TCGA),²⁶ have shifted the focus towards the role of gene perturbations in the non-kinase domain as well as mutations in other histologies.

Figure 1.

Somatic mutations of the extracellular, non-kinase domain region of Erb receptors cause activated Erb signal transduction via different mechanisms. (A) Ligand activation of Erb family receptors. Ligand-binding to subregions I and III of the ectodomain of the closed, ‘thethered’ from of the receptor prompts a conformational ‘switch’ of the molecule characterized by the formation of ligand-binding pocket formed by subregions I, II and III and a 180° twist and exposure of the dimerization region of subdomain II as a prerequisite for dimerization and signal transduction activation. (B) Mutations in highly conserved regions of domains II and IV disrupt intersurface, intramolecular bonds between domains II and IV which keep the receptor monomer in its ‘thethered’ and inactive state. The inability to form these interactions results in the loss of domain II and IV binding, and a shift in the equilibrium to the open form of the receptor which exposes the dimerization region of domain II leading to constitutive pairing and abnormal signal transduction. Mutations disrupting these interactions include e.g. EGFR A597T, Erbb3 G284R and Q298, and Erbb4 Y285C, R306S and D595V mutations in lung cancer, EGFR T263P, A289D/T/V and P596L mutations in glioblastoma multiforme, and Erbb4 D609N mutations in melanoma. (C) Activating somatic mutations might affect the ligand-binding domains increasing the affinity of the receptor for the ligand and resulting in abnormal receptor activation. Examples of somatic mutations potentially involved involved in increased ligand-binding are the EGFR domain I mutations R108K in glioblastoma and Erbb4 mutation Y111H in melanoma, and the domain III mutation R451F and S442F in the EGFR receptors of lung cancer and glioblastoma and the Erbb4 mutants R393W, P409L and E452K in melanoma.

Table 1.

Structural and functional differences between Erb receptor family members in respect to signal transduction

	Ligand	Homo-/ Heterodimer	Intrinsic kinase activity	Downstream signaling	Alternative splice sites
EGFR	EGF, TGFα, EPR, HB-EGF, AR, BTC	Both	Yes	Adaptor proteins GRB2 and GAB1, STAT5, Shc	No
ERBB2	No direct ligand identified; exists in active confirmation state	Both	Yes	Adaptor proteins GRB2 and GAB1	No
ERBB3	NRG1α, NRG1β, NRG2α, NRG2β, NGC	Heterodimer only	No	p85 subunit PI3K	No
ERBB4	Heregulin, NRG1α, NRG1β, NRG2α, NRG2β, NRG3, NRG4, AR, HB-EGF, TR1, TR2	Both	Yes	p85 subunit PI3K, GRB2, STAT5, Shc	JM-a/JM-b CYT-1/CYT-2

Abbreviations: AR, amphiregulin; BTC, netacellulin; EGF, epidermal growth factor; EPR, epiregulin; HB-EGF, heparin-binding epidermal growth factor; TGFα, transforming growth factor; NRG, neuregulin; NGC, neuroglycan; TR, tomoregulin.

Mechanisms of non-kinase domain mutations in the Erb receptor family and their role in the development and evolution of cancer

For Erb receptors to remain in the inactive, ‘tethered’ configuration requires the formation of intramolecular bonds between domains II and IV of the ectodomain, which buries the dimerization interface of domain II within a deep pocket (Fig. 1A).^27–29 Residues 266 to 283 of domain II in the EGFR receptor form a small β-hairpin loop that interacts with the C-terminal portion of domain IV. This contact is mediated via a main-chain hydrogen bond between tyrosine residue Y270 in domain II and an aspartic acid residue at position 586 of domain IV, and three side-chain interactions between residues Y270 and K610, F275 and G587, and Q276 and H589 of domains II and IV.^27,29 The integrity of these interactions mediated by these amino acid residues is essential for maintaining the ‘folded’, inactive form of Erb receptors as shown by the ‘tether-mutant’ EGFR forms having an up to 12-fold increased affinity for the ligand binding to domains I and III.^18,27,30 These directly contacting residues in domains II and IV are highly conserved among all four receptors and various species with the exception of G587 and H589 in the Erbb2 receptor which are substituted with proline and phenylalanine residues in this receptor. As the substitution of the glycin at position 587 with a proline residue and the histidine at 589 with a phenylalanine residue prevents the formation of these critical hydrogen bonds, the receptor loses the close intramolecular contact between domains II and IV contact and its the ability to maintain in the “closed” and inactive confirmation (Fig. 1B).¹⁸ Thus, this loss of intramolecular bonds also explains the unique feature of the Erbb2 monomer which is constitutively present in the “open,” active form, available for dimerization and signal transduction in a ligand-independent fashion. The biological importance of these residues for controlled Erb signaling has now been further re-affirmed by the finding of somatic mutations occurring at or in close proximity of these critical positions across the Erb receptor family and across different cancers. Examples of mutations with similar positioning disrupting interactions between the interdomain surfaces of domains II and IV include EGFR A597T, Erbb3 G284R and Q298, and Erbb4 Y285C, R306S and D595V mutations in lung cancer,²⁶ EGFR T263P, A289D/T/V and P596L mutations in glioblastoma multiforme,³¹ and Erbb4 M313I, E317K, E542K, R544W, E563K and D609N mutations in melanoma (Fig. 1B and Table 2).³² Recent crystallographic data from the Erbb4 receptor seem to support the activating mechanism of some of the above mutations in the Erb receptor molecule: for example, D609N which lies on position one of the tight β turn of domain IV opposite the juxtramembrane part, and which is conserved as Asp in EGFR, Erbb2 and Erbb3, forms the base of the ‘tether’ pocket. Disruption of this critical local structure might destabilize the ‘thether’. Both, R544W and E563K form salt bridges with other residues of domain IV (Glu 538 and R 519), and disruption of these bridges will likely affect the stability and local structure of domain IV favoring a shift towards the open, and active, configuration.³³

Table 2.

Somatic mutations including missense, nonsense, deletions, insertions and duplications previously identified in the Erb receptors of various epithelial cancers

	EGFR	Erbb2	Erbb3	Erbb4
Lung	P136T	S310F	G69R	N181S	26, 126–135
	R451F	D769H	G284R	T244R
	A597T	774insAYVM	Q298	Y285C
	V651M	L755P		R306S
	E709delT710	775insYVMA		V348L
	G719S	G776V		D595V
	G719C	G776L		H618P
	G719A	G776V, Cins		E810K
	E746delA750	776insYVMA		T926M
	E746delT751insA	779insVGS		D931Y
	E746delS752insV	781insGSP		K935I
	L747delT751	W906
	L747delS752
	L747delP753insS
	L747delA750insP
	L747delT751insP
	P753F
	774insNPH
	S768I
	770insASV
	771insGL
	771insSVD
	773insTHP
	T790M
	L833V
	P848L
	L858R
	L861R
	L861Q
Breast		L755delT759		E872K	57, 136, 137
		L755S
		S760A
		R896C
Gastric		G776S		A773S	57, 126, 136, 138
		K724N		Intron 19 2302-8del-7insT
		T733I		Intron 19 2302-8delT
		L755S
		L755S
		D769H
		Q799P
		V777L
		V777L
		T862A
		L869Q
GBM	D46N	E914K			31, 126
	G63R
	R108K
	T263P
	A289D
	A289T
	A289V
	R324L
	E330K
	S442F
	P596L
	G598V
	L861Q
Ovarian	E746delA750	N857S			126, 139, 140
	E746delA750	A771insM774dupAYVM
		Y772insA775dupYVMA
HCC		H878Y			141
CRC	E749K	V777L		V721I	57, 66, 136, 142
	E762G	V777M		P854G
	A767T	V842I		D861Y
				I1030M
Head and Neck	K745R	V773A			143–146
	E746delA750
	E746delT751insA
	L747delE749
	L747delE749, A750P
	S784Y
	T790M
	F795S
	P848L
	L862P
Esophageal	E746delA750				147
	T790M
	L858R
Pancreas	DelE746-A750				138, 147
	DelE746-A750
	C818Y
Melanoma				L39F	32
				Y111H
				M313I
				E317K
				S341L
				R393W
				P409L
				E452K
				R491K
				E542K
				R544W
				E563K
				D609N
				P700S
				E836K
				E872K
				G936R
				P1033S
				S1246N
				R1174Q

‘Hotspots’ (mutations occurring in more than 10% of cases) are shaded in gray (Del, deletion; ins, insertion; dupl, duplication).

Mutations affecting the extracellular subdomains I and III, which are involved in ligand-binding, can result in increased affinity of the receptor to the ligand and abnormal Erb signaling.^20,34 Ligand binding occurs via interactions with domain III, and to a lesser degree via domain I.^19,27,28,30 Essential residues in the EGFR receptor are the highly conserved residues R108 and Y113 in domain I, and the conserved residues D379, Q408, H418, Q432 and H433 of domain III.^27,28 Mutations observed in the Erb family assumed to increase ligand binding leading to abnormal receptor activation are the EGFR domain I R108K mutations in glioblastoma,³¹ EGFR P136T in lung cancer,²⁶ Erbb4 Y111H in melanoma,³² and for domain III EGFR R451F in lung cancer,²⁶ EGFR S442F in glioblastoma,^20,31 and Erbb4 H374Q, R393W, P409L and E452K in malignant melanoma (Fig. 1C and Table 2).^32,35 While at this point crystallographic structure data from the Erbb4 molecule do not suggest a direct involvement of these residues in ligand binding, all domain III mutations in this region including R393W, P409L and E452K involve residues which are exposed on the surface impacting on receptor solubility and domain arrangement.³³ If altered domain arrangement is the cause for such enhanced ligand binding remains at this point speculative.

In general, the observed clustering of activating somatic gene mutations in the Erb family around functional subdomains and structurally critical ‘hotspots’ of multiple Erb receptors and across different tumor histologies are proof of the biologic importance of these mutations in the development and progression of cancer. The recent observation in NSCLC that mutations in the Erb receptors are mutually exclusive with mutations in other Erb pathway genes (e.g., PI3KCA, BRAF, KRAS) further affirms their ‘tumor- driving’ potential.

Downstream targets

The most commonly affected signaling pathways by aberrant Erb signal transduction are the PI3K-Akt and MAP kinase pathways. Following receptor dimerization, the amino terminal part of one of the tyrosine kinase domains interacts with the C-terminal lobe of the other in an asymmetrical fashion which leads to phosphorylation of the respective C-terminal tyrosine residues of the other, interacting receptor molecule. This event recruits a variety of adaptor and cell signaling proteins and initiates downstream signaling.

Erb dimer composition and ligand-binding determines which pathway will be activated, a process which is still only incompletely understood. EGFR and Erbb2 bind indirectly via the Erb growth factor bound adaptor proteins GRB2-Sos (sons of sevenless)-RAS complex or through an additional mediator GAB1 (GRB2-associated binding protein 1) to PI3K, whereas Erbb3 and 4 have direct binding sites for the p85 subunit of class IA PI3K. EGFR and Erbb4 have also been shown to bind to STAT5 and Shc proteins.³⁶ Erbb3 contains six binding sites for the p85 subunit of PI3K, and it is suggested that Erbb3 and 4-mediated signaling occurs preferentially via the PI3K-Akt pathway, an observation supported by activation of the PI3K-Akt pathway in cancers with unregulated Erbb3 signaling, and the unfavorable clinical outcome of cancers with abnormal Erbb2/Erbb3 heterodimer formation and deregulated PI3K-Akt signaling.^21,37 It has also been shown that inhibition of EGFR or Erbb2 signaling with tyrosine kinase inhibitors leads to frequent compensatory upregulation of Erbb3 because of negative Akt-mediated feedback signaling, and that the upregulation of Erbb2 and Erbb3 dimers following cetuximab treatment, a monoclonal antibody against EGFR, induces PI3K-Akt activation.^37,38

The elucidation of Erb-mediated downstream signal transduction is of substantial clinical relevance: to date, with the exception of imatinib mesylate in the treatment of BCR-ABL-positive chronic myeloid leukemia and gefitinib and erlotinib in lung cancer and to a lesser degree in glioblastoma, no single agent targeted therapy has been shown to impact on survival rates in a clinically meaningful way. The future success of molecular therapy will largely depend on the ability of developing either more effective compounds or combinational approaches which effectively target escape pathways and the extensive crosstalk of cancer-driving pathways. Figure 3 shows the potential therapeutic targets and their respective, clinically advanced compounds which are currently available. Based on the current knowledge of Erb-mediated signaling such an approach would either pursue:

Figure 3.

Somatic mutations cluster around structurally and functionally unique residues and subdomains in the Erb receptor family. Somatic mutations in all members of the Erb family (EGFR, Erbb2/HER2, Erbb3, Erbb4) discovered in epithelial cancers are shown. The locations of the mutations are shown in reference to their amino acid position and the structural domains of the involved receptors. Triangles indicating missense or nonsense mutations, diamonds are indicating deletions, insertions or duplications. Colors of the different cancers are depicted on the right. Mutations involved in the disruption of the formation and function of the highly conserved Y270 residue of domain II and G587 and H589 of domain IV mediating the formation of intramolecular bonds and stabilization of the ‘tethered’ form are highlighted in grey.

1. Double inhibition of the upstream, mutated/dysregulated Erb receptor targeting different domains of the receptor (e.g., the combination of the monoclonal antibody trastuzumab and the tyrosine kinase inhibitor lapatinib or combination of trastuzumab and pertuzumab two anti-Erbb2 directed monoclonal antibodies targeting different functions of the receptor).^39–42

2. Vertical pathway inhibition by targeting the upstream receptor and downstream targets of its effector pathway (e.g., combining trastuzumab or lapatinib with molecular compounds targeting the PI3K-Akt pathway as recently shown in preclinical models using the dual PI3K-mTOR inhibitor NVP-BEZ235).^43,44

3. Horizontal pathway inhibition by targeting parallel, via intracellular crosstalk activated pathways. In the case of Erb-mediated signal transduction targeting the MAP kinase pathway would likely yield the highest reduction of proliferative Erb-mediated cell signaling (e.g., while up to date no direct inhibitors of RAS are available, targeting BRAF in combination with other agents has gained greater attention; e.g., phase I/II studies currently exploring the combination of lapatinib and sorafinib or erlotinib and sorafinib) (Phase I NCT00984425, Phase II NCT00696696).

4. Targeting two concomitant oncogenic mutations that are not interconnected. Examples are mutations/dys-regulation of the Erb receptor family and other major oncogenic events as e.g., controlling cell cycle progression with CDK inhibitors like flavoperidol, PD-002991, or roscivitine after loss of CDKN2A.⁴⁵

Need for additional targets in malignant melanoma

Melanoma is the sixth leading cancer in both men and women. Metastatic melanoma has a poor prognosis with five-year survival of less than 5%. Melanoma ranks second in terms of loss of years of potential life.⁴⁶ FDA-approved treatments for metastatic melanoma include aldesleukin (IL-2) and dacarbazine chemotherapy.⁴⁷ Aldesleukin as the most efficacious therapy has an objective clinical response rate of about 16% and a complete response rate of 6%.⁴⁸ It is the only treatment known to cure some patients with metastatic melanoma. Dacabarzine or temazolamide-based chemotherapy has a clinical response rate of up to 20% but almost no complete responders or long-term survivors.⁴⁸ With the incidence of melanoma increasing in the United States, and the fact that the last time a novel treatment for melanoma was approved by the FDA dates back to 1996, novel and more effective treatment strategies for metastatic melanoma are clearly needed. To date, none of the few targeted single-agents evaluated without additional patients’ selection in melanoma in phase II trials has shown any clinically meaningful activity.⁴⁹ This rather bleak outlook has only very recently changed with novel molecular agents specifically targeting tumors harboring the BRAFV600 mutation or the initial success of imatinib in melanoma harboring c-KIT mutations.^50,51 While the uneven distribution of c-KIT harboring melanoma might hamper a more widespread use of c-KIT targeted therapy,⁵² the completion of the ongoing phase II trials testing compound PLX-4032, which inhibits mutant BRAFV600, the most common oncogene in melanoma, with a 1,000-fold greater specificity than wild type BRAF,⁵³ is currently eagerly awaited. It is evident that for molecular therapy to succeed a much larger pool of both potential targets and effective compounds is necessary.

Mutational Analysis of the Tyrosine Kinome in Melanoma

Protein Tyrosine Kinases (PTKs) are ideal drug targets in that they have an important role in cell proliferation and apoptosis, they are amenable to pharmacological manipulation, and have been implicated in the pathogenesis of several cancer types.^54,55 This gene family is therefore a particularly good source to derive novel potential targets in melanoma. One way to conclusively associate a gene in cancer pathogenesis is to identify somatic mutations in the particular gene and to test whether the discovered mutations alter the function of the derived protein. Identification of somatic mutations in the PTK gene family in melanoma is therefore highly warranted.

In order to do this, several reagents and tools are necessary. For example, a high quality melanoma tumor bank, established through the generation of early passage tumor cell lines, allows the detection of somatic mutations that are otherwise masked by surrounding normal tissue. Additionally, automated methods for high throughput amplification of regions of interest coupled with their sequencing permit a robust analysis of large areas in the genome. Finally, the comparisons of these tumor derived sequences to their normal counterparts results in the identification of somatic mutations.

These methods have recently been harnessed to the high throughput sequencing of the PTK family in melanoma.³² In order to evaluate whether any PTKs were genetically altered in melanoma, the kinase domain coding exons of all 86 PTKs were assessed in a panel of 29 melanoma samples. 593 exons were PCR amplified from melanoma derived genomic DNA and sequenced. All sequences were compared to the reference human genome and whenever a sequence alteration was identified, the same sequence was examined in genomic DNA derived from the normal tissue from the same patient. From this it was determined that 19 genes contained a total of 30 somatic (i.e. tumor-specific) mutations affecting the kinase domains of the protein.

The next step in this evaluation was to expand the analysis beyond the kinase domains of the 19 genes that were found to harbor somatic mutations in a larger melanoma tumor panel. To do this, all coding exons of each of the 19 genes were sequenced in a total of 80 samples leading to the identification of 99 non-synonymous somatic mutations. One commonly used method to determine whether somatic alterations in particular proteins have a role to play in tumorigenesis is to calculate the ratio of non-synonymous (NS) to synonymous (S) alterations, as it is expected that synonymous changes would have a neutral effect on structure and function of the respective protein. The NS/S ratio of the whole PTK screen was 5.8:1 which is significantly higher than the N/S ratio of 2.5:1 predicted for nonselected passenger mutations (p < 1 × 10⁻⁵) leading to the conclusion that a large panel of these alterations are driver mutations.⁵⁶ Another way to determine whether mutations have a role in tumorgenesis is to functionally evaluate how these alterations affect the activity of the wild type version of the protein.

Out of the 19 genes that were found to harbor somatic mutations in the above screen, the most highly mutated was Erbb4, containing 24 non-synonymous somatic mutations affecting 19% of melanoma patients making Erbb4 the most highly mutated protein tyrosine kinase in melanoma. All but three of the alterations were heterozygous and no truncating or nonsense mutations were identified. One of the mutations identified in the kinase domain (E872K), was previously discovered in breast cancer,⁵⁷ and two melanoma cases harbored mutations at the same position (hotspot E452K). Similarly to the kinome screen, Erbb4 had a NS:S ratio of 24:3 which is also significantly higher than the NS/S ratio predicted for non selected passenger mutations, thus suggesting that at least a portion of the discovered mutations are driver mutations and that Erbb4 is playing a role in tumorigenesis. A recurring theme in the identified mutations is a substitution of an acidic residue with a basic residue which might be due to the high frequency of C:G>T:A transitions, known as a UV signature found in melanomas.⁵⁸ Interestingly, most of the tumors with Erbb4 mutations also contained mutations in either NRAS or BRAF, suggesting the Erbb4 operates via different pathways to the ones of NRAS and BRAF. A similar scenario was described for the well known oncogene PIK3CA in colorectal cancer where the majority of tumors with PIK3CA mutations also contained mutations in KRAS or BRAF.⁵⁹ Importantly, at least 50% of the discovered mutations in Erbb4 were found in melanoma samples derived from patients who have not been treated with chemo or radiotherapy and two of the Erbb4 mutations (E452K and E872K) found in chemotherapy treated patients wre proven functionally to be activating.³² These alterations are summarized in Table 2.

Functional Evaluation of Erbb4 Mutations

To determine the biological effects of the identified Erbb4 mutations seven of the identified mutations were cloned and investigated in biochemical and biological assays. Since a total of 24 somatic mutations were discovered in the mutation screen several criteria were used to determine which of these to test. These included the proximity of the mutations to previously identified mutations in EGFR and Her2 on the Erbb4 protein crystal structure as well as the clustering of two of the mutations into two mini-hotspots (E452K and E872K)³² (Table 2).

To assess the biochemical effects of these alterations on Erbb4 kinase activity, wild type Erbb4 or the seven mutations were transiently transfected into HEK-293T cells and their autophosphorylation as well as kinase activity were evaluated. Expression of mutant Erbb4 resulted in significantly more kinase activity than wild-type protein.³² While the exact mechanism that lies behind the increased activity of mutations located in the extracellular domain remains in the majority of the mutations to be determined, crystallographic, structural and function data from mutations in similar positions in EGFR and Her2 suggest that the evaluated extracellular domain mutations may increase signaling activity due to a shift in the conformational equilibrium, increased ligand binding or by constitutive dimerization (Fig. 1B and C).

Functional assessment of the above Erbb4 mutations revealed them to show oncogenic activity: They transformed primary fibroblasts in culture, to a similar potency as oncogenic K-RasG12V. In addition, they induced increased foci formation and anchorage independent growth of melanoma cells.³² The seven evaluated mutations can therefore be regarded as ‘driver’ mutations and Erbb4 as an oncogene.

The dependence of melanoma cells on mutant Erbb4 signaling was initially established by performing shRNA knockdown of endogenous Erbb4 in melanoma cells that harbor either its wild type or mutant version. In each case in which Erbb4 was knocked down in cells containing mutant Erbb4, melanoma cells halted their proliferation, whereas cells maintained their growth rate when similar knockdowns were performed in cells containing wild type Erbb4. This ‘addiction’ of melanoma cell lines harboring mutated variants of the Erbb4 receptor to Erbb4 signal transduction established Erbb4 as an essential signal and dominant oncogene that allows for the growth of melanomas harboring these mutations. Indeed, to institute the potential relevance of mutant Erbb4 signaling to the clinic, the broadly acting Erbb inhibitor, lapatinib, was tested for its inhibition of cell proliferation of melanoma cells with or without the Erbb4 mutations. Strikingly, despite the fact that lapatinib is not Erbb4-specific, a clear growth inhibition of cells with Erbb4 mutations versus cells expressing wild type Erbb4 was observed,³² thus suggesting that cells with Erbb4 mutations are subject to ‘oncogene addiction’. The fact that response to lapatinib occurs despite the existence of other somatic mutations, strengthens the point that melanoma cells harboring Erbb4 mutations are dependent on its signaling. It is reasonable to expect that inhibition of co-existing oncogenes, such as Braf, will further increase the response and increased growth inhibition will be observed.

As several different Erbb4 isoforms exist which differ in their functions and their ability to promote growth,²⁵ one of the interesting follow-up questions to this study is to test whether any of the splice site mutations that were discovered affect which Erbb4 isoform is expressed. Using a subset of the melanoma samples that were used in the mutational analysis study, it was shown that the melanoma samples exclusively express Erbb4 JM-a coupled to both types of its cytoplasmic isoforms, CYT-1 and CYT-2. No signal was observed for the Erbb4 JM-b isoform in any of the samples.⁶⁰ This corroborates well with the studies described above for two reasons: (a) Prickett et al. reported that the Erbb4 mutations activate the AKT pathway specifically and the CYT-1 Erbb4 isoform is the isoform that is known to directly activate this pathway and (b) when the kinase activity of the Erbb4 mutants was evaluated, the JM-a CYT-1 Erbb4 isoform was used as the plasmid backbone to which the mutations were cloned. As this represents the Erbb4 isoform that is expressed by the melanoma cell lines, this makes the kinase results physiologically relevant.⁶¹

Notably, numerous different somatic mutations were identified that are widely distributed over the coding sequence and occur in all domains of Erbb4. Cleary, the seven mutations that were evaluated show gain of function phenotypes. However, since the identified mutations do not cluster into hot spot regions as seen in oncogenes such as PIK3CA, Kras, Braf and GNAQ,^59,62–64 the oncogenic activity is probably due to the removal of negative regulatory domains. This is reminiscent of the mutations identified in FLT3.^1,35 Clearly many questions remain unanswered regarding the effects of these mutations. For example, what effect do these mutations have on the binding of ligand, on Erbb4 structure, on its interaction with regulatory proteins. Do all the observed Erbb4 mutations results in increased kinase activity? Further biochemical and biological analysis will be required to fully investigate the effects of the identified alterations on Erbb4 function.

As mentioned above, Erbb4 has previously been described to contain mutations in various cancer types, including, gastric carcinomas (1.7%), colorectal carcinomas (0.68–2.9%), NSCLC (2.3–4.7%) and breast carcinomas (1.1%),^26,57,65,66 and their biological impact has not always been consistent across these cancers. For example, analysis of some alterations suggested that Erbb4 may function as a tumor suppressor in breast cancer and prostate cancer.^67,68 Another report describing the functional effects of Erbb4 mutations revealed two mutations to cause the loss of kinase activity when Erbb4 homodimerizes, but maintain Erbb4’s ability to heterodimerize and promote signaling via the ERK and AKT pathways. The findings of some Erbb4 mutations causing a loss of function is, however, not necessarily contradictory to the noted oncogenic effects of Erbb4 mutations. One important explanation for this is that as indicated above, Erbb4 encodes four different isoforms that promote various cellular responses,^25,69 and different isoforms were used in the various cited studies. For example, in recent reports in which endogenous Erbb4 levels was downregulated in tumor cells, a growth promoting role for Erbb4 was observed strengthening the notion that Erbb4 acts as an oncogene.^25,70

Somatic Mutations of the Erb Receptor Family in Other Cancers

Activating gene mutations in the Erbb tyrosine kinase receptor family have first been described in the EGFR/Erbb1 receptor which harbor these mutations in 10–30% of cases.^4,5 The most common, biologically and clinically relevant, EGFR L858R and the exon 19 746 and 747 deletion variants, have now been found in no less than six different tumor types including lung, glioblastoma, ovarian, head and neck, esophageal and pancreas cancer (Fig. 2; Table 2),⁵⁸ and their clinical value in NSCLC and GBM tumors is now firmly established as tumors harboring these mutations are commonly sensitive to the small molecule TKIs erlotinib and gefitinb.^3,71–73 Similar to the activation mechanism of cyclin-dependent kinases, the A-loop of the intracellular kinase domain is in the active state extended and directed away from the catalytic cleft of the kinase domain in order to (1) enable and facilitate the interaction and transient binding of the phospharylation sites of the peptide substrate (commonly the C-terminal domain of the other, paired receptor) and (2) to stabilize and enhance ATP binding via an interaction of a hydrophilic glutamate residue from the C-loop of the kinase domain, which forms the backbone of the catalytic cleft, and a hydrophilic lysine residue.^74,75 In the ‘closed’, inactive state, the A loop is ‘covering’ the cleft preventing docking and binding of peptide substrate, and the glutamate residue essential for ATP binding is not available as through an outward displacement the C-loop is rotated away separating the glutamate residue from the lysine. This conformational change is associated with a dramatic drop in the ATP affinity of the receptor.^76,77 Mutations in the kinase domain, and possibly in the extracellular domain, disrupt this tightly regulated and balanced equilibrium. In the best studied example of EGFR L858R mutations, the substitution of the hydrophobic leucine with the hydrophilic arginine favors interactions with the surrounding ‘hydrophilic’ solvent inducing a conformational change which moves the A loop away from protecting the catalytic cleft from binding of a peptide substrate, destabilizes the inactive form, and facilitates formation of the ‘open’, active status.^77,78 Since the degree, mechanism and pattern of these induced conformational alterations is different for each individual mutant, it is not surprising that different responses to allosteric inhibitors, like an increased sensitivity of cells with the L858 mutations to gefitinib over erlotinib, have been observed.⁷⁹ Translation of this knowledge into a truly comprehensive molecular approach will ultimately require considering each distinct mutant as a unique cancer target and the design of a distinct and unique molecular agent for each distinct mutation.

Figure 2.

Post-receptor signal transduction pathways of Erb-receptor mediated signal transductions occurs mainly via the PI3K-Akt and MAPK kinase pathways (Erbb-mediated STAT signaling not shown). Potential targets and their inhibitors for combinational molecular and biological therapy strategies are shown.

Extensive searches have been performed to identify similar hotspots of activating mutations in other members of the Erbb family. Several independent studies have identified mutations in the kinase domains of the Erbb2/HER2 and Erbb4 receptors.^31,71,80,81 The majority of kinase domain mutations in Erbb2 and Erbb4 are similar to the mutations seen in EGFR. They cluster around the two hotspots (‘classical’ EGFR mutations) and include exon 19 missense mutations, deletions, and insertions at positions 746 and 7474, and missense mutations at or in close proximity of L858R. Mutations in the kinase domain in Erbb2 and Erbb4 have been identified in lung, gastric, ovarian, colorectal, and head and neck cancers and are overall the most common type of activating somatic mutations affecting the Erbb family (Table 2). It is interesting that Erbb2 and Erbb4 receptor mutations affecting the kinase domain have a higher rate of missense mutations in exon 19 compared to the more common deletion variants at this location in the EGFR receptor (Table 2). This is particularly the case for gastric and colorectal cancer where no deletion or insertion mutations in the kinase domain of Erbb2 and 4 receptors have been identified. When comparing the predominantly affected Erb receptor within cancer types, it is striking that with the exception of colorectal cancer predominantly only one of the Erb receptors is affected in a near exclusive fashion. This is the EGFR receptor in lung cancer, glioblastoma, and head and neck cancers, the Erbb2 receptor in breast, gastric, and hepatocellular cancer, and Erbb4 in melanoma (Table 2). This receptor specific pattern of somatic mutation involvement points to a receptor preference for Erb signaling in individual tumors. If this receptor preference is due to an endogenous dominance of the affected receptor, or if oncogenic mutations transformed the affected Erb receptor into the driving Erb signaling molecule by providing a selection and growth advantage remains unresolved at this point. For a large number of mutations which do not involve one of the functional 4 subdomains of the ectodomain, the juxta- or transmembrane or kinase domains, or regulatory elements within the C-terminus harboring sites for effector proteins, adaptor molecules, or binding sites for proteins mediating receptor turnover, the exact mechanisms and impact onto Erb signaling remains to be determined. Also, it is important to note that there is great clinical, environmental, and ethnic variation of both the prevalence of activating mutations in Erb receptors and the response to TKI treatment of these EGFR mutations.^5,82 For example, mutations are more commonly found in women, patients of Asian descend, never smokers and patients with adenocarcinomas of the lung. African Americans have a substantially lower frequency of EGFR mutations, and response to TKI treatment, and their mutation profile is significantly different from that of Asian or Caucasian cancer populations.⁸²

Current and future strategies targeting aberrant Erb signaling

Current efforts on Erb-directed therapy exploiting the novel understanding of normal and aberrant Erb signaling focuses on (1) the development of improved, kinase-directed therapy targeting molecular subtypes of dysregulated Erb signaling, and (2) the development of second generation inhibitors to overcome resistance to Erb-directed targeted therapy.

The heterogeneity of activating Erb mutations, across the different domains of the individual Erb molecules, across the Erb receptor family itself, and across different tumor types will likely require a variety of different molecular interventions designed against the most prevalent Erb mutations to achieve maximum suppression of Erb signaling. To date, the monoclonal antibodies trastuzumab and cetuximab, and the TKIs gefitinib, erlotinib and lapatinib as regulatory approved agents form the backbone of Erb-directed targeted therapy (reviewed in ref. 83). The investigational monoclonal antibody pertuzumab, which exclusively binds to the dimerization region of subdomain II, aims to specifically target the dimerization requirement of Erb-mediated signal transduction,^84,85 and might be of particular value in tumor harboring mutations in subdomains II and IV which lead to ‘unfolding’ of the closed receptor monomer, exposure of the dimerization domain, and enhanced signal transduction. Preclinical data with pertuzumab confirm the effective inhibition of Erbb2 dimerization,^86,87 in particular Erbb-2-Erbb3 dimerization, and inhibition of Erb-mediated signal transduction that mediates tumorigenesis via activation of the PI3K-Akt pathway.^16,86 While pertuzumab has shown substantial activity in human xenograft models derived from ovarian,⁸⁸ prostate,⁸⁹ lung⁹⁰ and colorectal cancer⁹¹ cell lines most of them overexpressing Erbb2, clinical activity of this biological agent in an initial trial of non-selected patients with meta-static breast cancer has been disappointing.⁹² More recent trials, however, have confirmed activity and clinical benefit of pertuzumab in combination with trastuzumab in HER2 amplified metastatic breast cancer.^39,41,42 These findings point not only to the need of selecting patients with activated Erbb2 signaling, but make pertuzumab potentially one of the first anti-Erb agents targeting a specific molecular feature of dysregulated Erb signaling. Improved ‘pharmacogenomic matching’ of the underlying genomic aberration and resulting structural alteration with the chosen anti-Erb targeted agent will ensure maximum reduction of oncogenic signaling and success of targeted therapy. This concept of exploiting molecular subtypes of an altered genotype of a receptor super family is in line with ongoing studies which combine trustuzumab and pertuzumab in breast cancer targeting different receptor domains of the Erbb2 molecule (Phase III; NCT00567190) or the addition of the TKI lapatinib to rescue trastuzumab-resistant breast cancer in patients with HER2-overexpressing tumors.⁴⁰ The ideal agent against mutated Erb receptors, in particular for mutations like EGFR L858R and del747-P754insS, are TKIs with increased affinity to the mutated molecule compared to the wild type receptor. While such compounds have recently been described for the BRAFV600 mutation, a common oncogene in thyroid cancer and melanoma, where the novel compound PLX 4302 has a more than 1,000-fold increased sensitivity to mutant BRAFV600 than wild type BRAF kinase,⁵³ similar compounds are currently not available for mutated Erb receptors. Current efforts on improved direct targeting of Erb kinase activity, which have completed phase I trials, include the recent development of TKIs like neratinib (HKI-272) and BIBW-2992,⁹³ which bind irreversibly to Erb receptors or ARRY-334543 and TAK-285 which concomitantly inhibit multiple members of the Erb family.^94,95

The other approach of improved targeting of mutant Erb signaling focuses on the evolving understanding of the resistance mechanisms of Erb-directed targeted therapy. Further development along this avenue is insofar more essential in respect to the durability of Erb-directed therapy as the clinical experience with single-agent targeted therapy uniformally tells that eventually all patients initially responding to TKIs will relapse and progress in their disease.⁹⁶ Multiple trials supporting this notion have shown that initial differences in progression-free survival between the investigational arm of single-agent molecular therapy and the control cohort vanish with longer follow-up and no difference in overall survival was detected. The latest example hereto is the large, prospective randomized trial on adjuvant imatinib in patients with gastrointestinal stromal tumors (GIST).⁹⁷

Resistance mechanisms against Erb-directed therapy include both the presence of an a priori resistance mechanism, most notably the presence of additional oncogenic mutations, or the acquaintance of de novo resistance during molecular therapy. The prime example of concomitant resistance-mediating genomic alterations in Erb-targeted therapy which has found entrance into clinical practice is the presence of KRAS mutations in NSCLS and colorectal cancer.^11,98,99 For example, clinical benefit to cetuximab, an EGFR-directed monoclonal antibody, in colorectal cancer is not observed in tumors which harbor KRAS mutations.⁹⁹ It is therefore recommended to test KRAS mutation status in patients considered for this biological agent prior to start of treatment, and cetuximab administration is discouraged in patients harboring KRAS mutations.¹⁰⁰ Other examples of concurrent genomic alterations mediating a negative impact on the response to Erb-directed TKI therapy include absence of wild type BRAF in KRAS wild type colorectal cancer,¹⁰¹ concurrent PI3KCA mutations,¹⁰² amplification of wild type EGFR,¹⁰³ PTEN deletions activating the PI3K-Akt pathway,^100,104 and concomitant HER2 mutations.¹⁰⁵

Equally clinically relevant, but frequently more common, are acquired resistance mechanisms during anti-Erb therapy.⁹⁶ EGFR-mutant tumors in NSCLC previously treated with either gefitinib or erlotinib and initially responsive to these TKIs develop a secondary EGFR T790M mutation in more than half of the cases.^106–108 EGFR T790M mutations increase the affinity for ATP to the receptor overcoming the competitive binding of the reversible allosteric inhibitors erlotinib and gefitinib.^106–108 The development of similar mutations at analogous positions in the kinase domains have been observed in CML and GIST treated with imatinib (T315I in ABL, T674I in PDGFRα, and T670I in c-KIT).^109–111 It is hoped that second-generation TKIs like neratinib (HKI-272) and BIBW-2992, which bind irreversibly to Erb receptors, including receptors which harbor T790M mutations, are able to overcome TKI resistance.^93,108 Another strategy to overcome TKI resistance in patients treated previously with TKIs is the administration of heat shock protein inhibitors.¹¹² By disrupting the heat shock protein function, which stabilizes Erb receptors, Erb molecules are more prone to ubiquitylation and degradation thereby reducing Erb signaling. The compound with the most advanced preclinical and clinical development is the heat shock protein 90 (Hsp90) inhibitor tanespimycin (17-AAG) which has now completed phase I trials.¹¹³ Another important, recently elucidated mechanism of TKI resistance is the amplification of the proto-oncogene MET which has been observed in 20% of NSCLCs treated with erlotinib or gefitinib.^114–116 MET-mediated activation of downstream signaling, in particular activation of the PI3K-Akt, STAT3 and five pathways through activation of Erbb3 substitutes for the loss of EGFR-mediated signaling abrogated by TKI inhibition,¹¹⁷ and MET kinase inhibitors, like SU-11274, might be of particular clinical value in this scenario. Finally, it recently has been recognized that cancers with Erb mutations treated with TKI might usurp physiologic regulatory and feedback mechanism to develop resistance.^37,118 Preclinical models of lung, colorectal and pancreatic cancer cell lines treated with erlotinib found compensatory upregulation of the insulin growth factor receptor 1 insulin receptor substrate (IGF-IR IRS-I) and a compensatory upregulation of Akt activity via IGF-IR.¹¹⁸ Addition of the IGF-IR inhibitor OSI-906 to erlotinib has shown promising activity in preclinical studies,¹¹⁸ a finding which is currently translated into the clinic.¹¹⁸

Personalized Therapy

Cancer treatment is currently undergoing a major paradigm shift in the U.S. Up to now, the majority of administered systemic cancer treatment, in particular cytotoxic therapy, is currently administered without any clinical benefit to patients disregarding the vast molecular heterogeneity of cancers. This shockingly high number will with the inevitable emerge of personalized medicine undoubtedly decrease, a goal now embraced by many lead organization like ASCO and the ACS.

There are two fundamental advantages of personalized cancer medicine: (1) to determine prior to treatment which patients will likely benefit from the planned therapeutic intervention, and preventing toxicities in patients who would derive no benefit from treatment, and (2) to further modulate additional therapies should the first-line therapy fail, the tumor become resistant or is limited by unacceptable toxicities.

The improved diagnostic and predictive value of genomic information enabling improved patient selection is expected to have one of the most profound impacts on the quality of cancer care in the forthcoming decade.¹¹⁹ To treat only patients which will likely benefit from treatment has prime examples, like the estrogen-receptor status for anti-estrogen therapy in breast cancer,¹²⁰ SNP analysis of the UDP-glucoronosyltransferase 1 A1 gene for irinotecan toxicity,^121,122 and Bcr-Abl fusion and c-KIT mutation analysis for imatinib therapy in CML and GIST to name a few. The advent of genomic technology for the analysis of human tumor samples has now added an unprecedended source of information to aid prognosis and clinical decisions.¹²³ The dimensions and complexity which such data can capture provides an opportunity to predict clinically valid trends, like risk of recurrence or response to treatment, that can distinguish phenotypes in ways that traditional methods cannot.^124,125 It is this variability in outcome not captured by current predictive and prognostic models based on classical clinicopathological data which currently deprives patients of personalized treatment allocations which optimally tailor selected treatment decisions to their individual tumor biology.

Therefore, a true commitment to personalized medicine will require us to abandon a so far histologically defined approach to cancer care, and embrace molecularly defined cancer subtypes and their different treatment algorithms for the selection of effective treatment strategies. It was initially thought that increased fragmentation and compartmentalization of cancer subtypes will hamper further development and progress of personalized cancer care. It was assumed that an ever increasing pool of cancer subgroups and subtypes with an ever decreasing size through further and further split-ups into novel molecular subgroups which arise from new molecular discoveries would make it impossible to conduct adequately powered clinical trials, obtain interest by pharmaceutical companies for very small markets, or even obtain enough clinical expertise in these rare phenotypes. Only recently has the pendulum shifted into the opposite direction now acknowledging the huge potential of a truly personalized approach. This includes not only the initial recognition of the huge savings of unnecessarily administered inefficient therapies including the saved costs of their associated toxicities but more so the potential of combining molecular subtypes, like Erb mutated cancers, across the boundaries of different histologies into a novel subclass of cancer negating some of the concerns raised above. To go a step further, such a classification should ideally not only be based on tumor genotype but also on the corresponding chemical signature associated with it. For example, the response to the TKI inhibitor lapatinib (currently approved for treatment of advanced breast cancer in combination with capacitabine or after previously failed treatment with either a taxane, anthracycline or trastuzumab in metastatic breast cancer) could define a class of dual-TKI responsive tumors comprising of Erbb2-amplified metastatic breast cancers, metastatic melanoma patients harboring somatic mutations of the Erbb4 gene which renders these tumors sensitive to lapatinib, and possibly—in the future—Erbb2-amplified gastric cancer or glioblastoma harboring the EGFRvIII deletion mutation (Fig. 4). As genotype-directed therapy is currently the fastest moving and dominating frontier in cancer medicine, the question will already not be if personalized medicine will make the conventional “one fits all” approach obsolete but when.

Figure 4.

A novel taxonomy of cancer based on genotype and the ‘chemotype’ response to the tyrosine kinase inhibitor lapatinib. Malignancies characterized by genetic dysregulation of the Erb receptor family (e.g., amplification of the Erbb2/HER2 gene in breast cancer, activating somatic gene mutations of the EGFR receptor in lung cancer and glioblastoma, or of Erbb4 in melanoma) and by response to the dual tyrosine kinase inhibitor lapatinib may form a novel genetically and clinically defined subgroup of cancer. Currently, Erbb2/HER2 positive metastatic breast cancer and malignant melanoma harboring Erbb4 mutations are malignancies that can be defined by genetically dysregulated Erb signaling and response to lapatinib. As genomic data and functional tests in other cancers emerge, glioblastoma multiforme harboring the EGFR deletion variant EGFRvIII (GBM EGFRvIII) and gastric cancer with Erbb2/HER2 amplification might be added to this subgroup in the near future (shaded background). Seize of rectangles reflect both prevalence of malignancy and proportion with dysregulated Erb signaling.