Background
Human epidermal growth factor receptor 2 (HER2) is the clinically validated biomarker for HER2-targeted therapies, several of which are approved in the neoadjuvant, adjuvant, and metastatic settings for HER2-positive breast cancer. Pertuzumab (PERJETA
®, F. Hoffmann-La Roche Ltd., Basel, Switzerland) is directed at the dimerization domain of HER2 and inhibits dimerization with other HER receptors while stimulating antibody-dependent cell-mediated cytotoxicity (ADCC) [
1‐
3]. Trastuzumab (Herceptin
®, F. Hoffmann-La Roche Ltd.) binds to the transmembrane domain to inhibit mitogenic signaling, block HER2 cleavage, and stimulate ADCC [
4‐
6]. Due to their different binding modalities, trastuzumab and pertuzumab have complementary proposed mechanisms of action [
7].
In the NeoSphere study, patients treated with neoadjuvant pertuzumab, trastuzumab, and docetaxel (Taxotere
®, Sanofi-Aventis, Paris, France) (group B) had a significantly higher pathologic complete response (pCR) rate compared with those treated with trastuzumab plus docetaxel (group A), pertuzumab plus trastuzumab (group C), or pertuzumab plus docetaxel (group D) (45.8% versus 29.0%, 16.8%, and 24.0%, respectively) [
8]. The combination of pertuzumab, trastuzumab, and docetaxel has also been shown to be superior to trastuzumab plus docetaxel in the first-line treatment of metastatic breast cancer [
9,
10]. The combination is now approved by the European Medicines Agency (EMA) and the US Food and Drug Administration (FDA) in both settings.
HER2-positive/estrogen receptor (ER)-positive and HER2-positive/ER-negative breast cancer are known to have different patterns of gene expression, and treatment outcome seems to be driven by different biologic pathways [
8,
11‐
14]. Hormone-receptor-negative status was associated with an increase in treatment benefit with pertuzumab plus trastuzumab when groups A and B were compared in NeoSphere [
8]. It is known that HER2 expression in ER-negative, HER2-positive tumors is higher than that in ER-positive, HER2-positive tumors [
15].
We assessed a panel of biomarkers, in tumor specimens and in sera, to characterize the molecular profile of the biomarker population of NeoSphere, and to explore biomarkers correlated with different treatments and/or different subsets of patients in this trial. The objective of this exploratory analysis was to identify those biomarkers (or combinations of biomarkers) with the best association (positive or negative) with pCR in response to the treatment regimens used in the NeoSphere study. In particular, we attempted to identify biomarkers that would predict a benefit from the addition of pertuzumab to a trastuzumab-based regimen. The markers tested included those that are involved in downstream signaling of HER2, belong to a group of related receptor tyrosine kinases that could serve as salvage routes for an inhibited HER2 pathway, or are ligands of HER family proteins that induce activation of the HER2 pathway. We also assessed the presence of truncated forms of HER2, including p95
HER2, with regards to pCR, as p95 was previously reported to predict worse outcomes in patients with HER2-positive breast cancer [
16].
Methods
Study design and patients
The study design and patient characteristics have been reported previously [
8]. Briefly, NeoSphere (NCT00545688) [
17] was a multicenter, open-label, phase II study. Treatment-naive women with HER2-positive breast cancer were randomized (1:1:1:1) centrally and stratified by operable, locally advanced, and inflammatory breast cancer, and by hormone receptor expression, to receive four neoadjuvant cycles of study treatment. The clinical endpoint was pCR, defined as the absence of invasive neoplastic cells at microscopic examination of the primary tumor at surgery. Remaining
in situ lesions were allowed [
8].
Specimen characteristics
Collection of core biopsies, sera, and whole blood from all patients was mandatory at baseline. Tumor samples were obtained as formalin-fixed, paraffin-embedded tissue. Tissue obtained after the neoadjuvant treatment period was derived from resection specimens.
Assay methods
Tissue processing, immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), RNA extraction, quantitative reverse transcription polymerase chain reaction (qRT-PCR), and DNA isolation were performed centrally by Targos Molecular Pathology GmbH, Kassel, Germany. Targos followed the protocols and processing instructions developed by Roche Diagnostics GmbH, Penzberg, Germany. Commercially available assays or kits were used where specified, and performed according to the manufacturer’s instructions. All other assays were developed by Roche Diagnostics for exploratory research purposes only.
Expression of HER2 (HercepTest, Dako, Glostrup, Denmark), HER3 (HER3 M7297, Dako), insulin-like growth factor 1 receptor (IGF1R, Clon 1.004.168, Roche Diagnostics), phosphatase and tensin homolog (PTEN, AF847, R&D Systems, Minneapolis, MN, USA), and pAKT (#3787, Cell Signaling Technology, Danvers, MA, USA) were assessed by IHC, and a modified H-score [
18] was derived for each marker. The modified H-score was calculated as the percentage of cells stained per intensity level, multiplied by a factor composed of the intensity category plus 1:
$$ \mathrm{Modified}\ \mathrm{H}\hbox{-} \mathrm{score} = \left(1 + 1\right) \times \mathrm{P}1 + \left(2 + 1\right) \times \mathrm{P}2 + \left(3 + 1\right) \times \mathrm{P}3 $$
Therefore the modified H-score has a maximum value of 400 instead of the standard H-score of 300. The percentage of cells stained with an intensity of 0 was used only for quality control. Cases with no staining on the tissue section were assigned a score of 0. Modified H-scores were calculated for subcellular compartments for which specific staining was identified and for which there was a biologic rationale for the subcellular location of the respective marker (e.g. for PTEN and AKT nuclear staining).
Image acquisition and analysis of HER2 staining intensity took place at the Royal Marsden Hospital (London, UK) using the Ariol image analysis system (Leica Microsystems (Gateshead) Ltd., UK) equipped with a BX61 microscope (Olympus, Southend-on-Sea, UK) and a black and white MegaPlus ES 4.0/E camera (Redlake MASD, Inc., San Diego, CA, USA). Slides were scanned and analyzed as previously described [
19], except that five representative invasive breast cancer areas in each image were selected. The mean membrane intensity of all five representative areas selected for analysis was used as a measurement of HER2 staining intensity.
HER1, HER2, HER3, Amphiregulin (AREG) and Betacellulin mRNA levels in tumor tissue were assessed relative to the G6PD gene by qRT-PCR (Roche Diagnostics, research-only assay). c-Myc amplification was assessed by FISH (MYC/CEN-8 FISH Probe Mix, Dako).
Mutational analyses of eight mutations at four hot spots in exons 7, 9, and 20 within the gene encoding the catalytic subunit of phosphoinositide 3-kinase (PIK3CA) on tumor DNA were performed using TaqMan-PCR (Roche Diagnostics, research-only assay) at Roche Translational Research Sciences (TRS, Basel, Switzerland). This led to a technical success rate of approximately 60%. To increase the sample size, a different method requiring less DNA input was applied at Genentech Laboratories, San Francisco, CA, USA. DNA was extracted from IHC-stained slides, amplified, and subsequently assessed using the DxS assay (DxS Ltd., Manchester, UK), which detects mutations E542K, E545K/D, and H1047R in exons 9 and 20. Mutational analysis per exon is presented to allow for pooling of the data on 328 cases overall. PIK3CA subgroups were defined based on the presence or absence of any mutation (mutant versus wild-type). “Wild-type” was only assigned to samples in which all reactions gave a valid result for “no mutation.” If one or more reactions failed, that sample was classified as “not assessable.”
Enzyme-linked immunosorbent assay (ELISA) was used to detect AREG, EGF, serum HER2 extracellular domain (sHER2), and transforming growth factor alpha (TGFα) in serum using an immunologic multiparametric chip technique (Roche Diagnostics). Levels of each serum marker were compared at baseline and at surgery in matched pairs from the same patients. Truncated forms of HER2 (the intracellular domain (ICD) and HER2 extracellular domain (ECD), where a ratio <1 indicates the presence of truncated forms), were measured in formalin-fixed, paraffin-embedded breast cancer tissue as described by Krüger et al. [
20].
Statistical and analytical methods
The patient population used (the biomarker population) was a subset of the intent-to-treat (ITT) population that contributed meaningful data to the analysis (Additional file
1: Table S1). In this analysis, median values were used as cutoffs for high and low biomarker levels for each of the continuous biomarkers, except for
PIK3CA mutation and
c-Myc amplification, where cutoffs were applied following a biologic rationale (wild-type versus mutant
PIK3CA and
c-Myc:centromere ratio ≥2). Instead of a data-driven cutoff, we used a prespecified median cutoff to reduce the number of false-positive findings due to the concomitant multiple testing and optimal cutoff selection. Box and whisker plots and summary statistics were used for initial data exploration and selection of relevant biomarkers. Spearman’s correlation was performed to identify the correlated biomarkers. Biomarkers with measurements near the limit of quantification were excluded from analyses.
False discovery rate was implemented to adjust for multiple comparisons at α = 0.05. Only groups A and B were considered for determining the additional benefit of adding pertuzumab to trastuzumab plus docetaxel.
Association between clinical variables (ER-positive versus ER-negative, local/inflammatory versus operable breast cancer, node-positive versus node-negative disease, and age ≥50 versus <50 years) and biomarkers were assessed in the overall biomarker population using the Chi-square test.
The association between biomarkers and pCR by treatment group were assessed by the Cochran–Mantel–Haenszel Chi-square test, stratified by hormone receptor status and breast cancer type. The treatment interaction test was performed for groups A and B to analyze the interaction between biomarkers and treatment in the association with pCR.
Analyses of the association between biomarkers and clinical variables, and of the association between biomarkers and pCR by treatment group, were repeated in ER-positive and ER-negative subgroups. These analyses were exploratory and secondary in nature, and therefore were not taken into account when adjusting for multiple comparisons.
Discussion
NeoSphere demonstrated that mandatory biomarker sampling is feasible in a clinical trial setting. Samples were available from nearly all patients and the planned panel of biomarkers could be assayed reliably in 99.8% (HER2) to 65.9% (c-Myc) of cases. While no new single marker could be confirmed as predictive of treatment benefit, these analyses revealed several important observations that will impact future trial designs.
Higher levels of HER2 membrane protein expression were associated with higher pCR rates with the combination of pertuzumab, trastuzumab, and docetaxel (group B) using both the modified H-score and digital image analyses. The digital readout resulted in a wider dynamic range compared with the manual scoring, showing a more pronounced effect in predicting pCR. While digital image analyses allowed us to better discern the pCR and non-pCR subgroups, HER2 levels remained highly overlapping between patients experiencing pCR versus patients not experiencing pCR and did not allow us to define clearly distinct populations that were either clinically or biologically meaningful.
The
PIK3CA mutational status appears to have a relevant role in defining the likelihood of pCR with all treatment regimens tested in NeoSphere. However, while a decreased benefit was observed, the magnitude of the treatment effect with the addition of pertuzumab was maintained (a difference in pCR rates of approximately 15% in the pertuzumab, trastuzumab, and docetaxel group between patients with wild-type and mutant genes). This was similar to associations reported in the CLEOPATRA trial [
21]. CLEOPATRA demonstrated shorter median progression-free survival with
PIK3CA mutant tumors, regardless of treatment regimen [
21]. Interestingly, in NeoSphere, only
PIK3CA mutations in exon 9 were linked to a lack of sensitivity to HER2-directed monoclonal antibodies. Although independent confirmation is needed, due to the lower frequency and lack of statistical power for exon 9 mutations, this finding might be explained by the different mechanisms and downstream effects that are proposed based on preclinical observations for the gain of function mutations in helical (exon 9) versus kinase (exon 20) domain [
22].
In CLEOPATRA, there was an association between poorer prognosis and pooled
PIK3CA mutations at exons 9 and 20, although the benefit from pertuzumab was maintained regardless of mutation status [
21]. Similar analyses from the TRYPHAENA study showed that pCR rates were lower, although not significantly, in patients with
PIK3CA mutant tumors regardless of treatment regimen [
23]. In addition, data from the EMILIA trial showed that patients with
PIK3CA mutations had worse outcomes compared with those with wild-type tumors when treated with lapatinib plus capecitabine [
24]. This was not the case with ado-trastuzumab emtansine, where the treatment benefit was similar for patients with wild-type and mutant tumors. This is an interesting finding and future trials should ascertain whether ado-trastuzumab emtansine can overcome the impact of
PIK3CA mutation on outcome, either when given as monotherapy or in combination with other therapeutics such as PI3K inhibitors, with the goal of improving the poor prognosis of patients carrying
PIK3CA mutations. The data from NeoSphere suggest that such an approach could be useful only in tumors carrying the activating exon 9 mutation.
Our results also provide clinical support for the blockade of HER2 cleavage by trastuzumab [
5]. Trastuzumab is thought to inhibit HER2 shedding, as through binding to HER2 it may sterically hinder the metalloproteinases from accessing the cleavage sites [
5]. By comparing serum marker levels at baseline and surgery, we showed that sHER2 levels decreased in all study groups except group D regardless of pCR status. Although serum HER2 extracellular domain did not predict pCR, this is a salient finding, as group D was the only group that did not include trastuzumab. To our knowledge this is the first report of clinical data to support this component of the proposed mechanism of action of trastuzumab. Although this is a reasonable mechanism based on our data, others cannot be ruled out.
Another relevant aspect that emerged from our analyses is the distinct difference between ER-positive and ER-negative tumors [
15]. HER2-positive/ER-positive and HER2-positive/ER-negative breast cancer have different patterns of gene expression, and treatment outcome seems to be driven by different biologic pathways [
11]. This effect was also seen in NeoSphere, where ER negativity was associated with increased likelihood of pCR with pertuzumab and trastuzumab [
8], and in NOAH, where ER-/progesterone receptor- negativity was associated with increased benefit from trastuzumab [
25,
26]. The different treatment effects according to ER status are reflected in a different molecular biomarker profile in ER-positive versus ER-negative tumors. At this point our data do not provide information that would result in patients receiving altered treatment, but future trial designs should take into account the intrinsic differences in marker profiles and related sensitivity to treatment.
A limitation of the biomarker analysis was that there was a narrow range of HER2 protein levels with little numerical difference between cutoffs, hindering an appropriate exploration of HER2 protein-related impact on treatment outcome. In addition, there was no significant difference in the pCR rates of patients in the subgroups with higher and lower HER2 levels, who received pertuzumab plus trastuzumab without docetaxel (group C). As the chemotherapy backbone is not known to impact HER2-related effects, the association between HER2 protein levels and pCR rates should have been detectable in both of the groups receiving trastuzumab plus pertuzumab.
Given that we were limited by small numbers of patients in some subpopulations, our analyses did not provide any additional predictive markers supporting a refinement of the HER2-positive target population treated with trastuzumab, pertuzumab, and docetaxel. It should be acknowledged that the use of the median cutoff for biomarker assessment could have overlooked a predictive value of some of these biomarkers at different, biologically meaningful, cutoffs that have not been explored in this study. However, our analysis did highlight the different biology of ER-positive and ER-negative HER2-positive tumors. The observation of a lower likelihood of pCR with trastuzumab, pertuzumab, and docetaxel in tumors harboring PIK3CA mutation at exon 9 is limited by the small sample size and warrants further exploration in larger datasets.
Acknowledgements
We would like to acknowledge all patients and investigators participating in this trial and also the central testing laboratories: Targos Molecular Pathology GmbH, Kassel, Germany and Roche laboratories, Penzberg, Germany. Finally, we would like to thank Lori Steiner from Roche Diagnostics for technical and scientific advice on the molecular tests for PI3K and HER2/3. Support for third-party writing assistance for this manuscript, furnished by Daniel Clyde, PhD, of Health Interactions, was provided by F. Hoffmann-La Roche Ltd.
Authors’ contributions
All authors critically reviewed drafts of the manuscript and interpreted the data, read and approved the final manuscript, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Conception: GB, AK, TP, LCA, PV, LG. Design: GB, AK, GVB, YI, TP, PV, LG. Conduct: GB, AK, GVB, YI, TP, MCL, LMT, MD, SK, TS, LG. Analysis: GB, GVB, TP, MCL, LMT, LZ (HER2 staining intensity), LG. Assembly of data: GB, YI, TP, MCL, LMT, LZ (image analysis data), SK, JEW, PV, LG. Data analysis: GB, AK, GVB, TP, MCL, MD, LZ (Ariol HER2 membrane staining intensity versus outcome in four groups), SK, LCA, PV, LG. Acquisition of biomarker samples: TP, LMT, LG.
Ethics approval and consent to participate
The study was undertaken in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki. All patients provided written informed consent. Approvals for the study protocol and modifications were obtained from independent ethics committees or institutional review boards (reference numbers not applicable): Australia: Barwon Health Research & Ethics Advisory Committee, Geelong. Brazil: Comitê de Ética em Pesquisa; Universidade Regional do Noroeste do Estado do Rio Grande do Sul UNIJUÍ, Ijuí; Comitê de Ética em Pesquisa em Seres Humanos da Universidade do Vale do Itajaí - UNIVALI/SC, Itajaí; Comitê de Ética em Pesquisa da Fundação Hospital Amaral Carvalho, Jau; Comitê de Ética em Pesquisa do Grupo Hospitalar Conceição, Porto Alegre; Centro de Referência de Saúde da Mulher; Comitê de Ética em Pesquisa em Seres Humanos, São Paulo; CEP ICAVC; Instituto do Câncer Dr. Arnaldo Vieira de Carvalho, São Paulo; Comitê de Ética em Pesquisa/FMABC, São Paulo; CEP do Centro de Ciencas Médicas e Biológicas - PUC/SP, Sorocaba;. Canada: Horizon Health Network Research Ethics Board, Saint John; Comite d’ethique de la recherche, Quebec; Ontario Cancer Research Ethics Board, Toronto. Israel: Helsinki Committee – Meir, Kfar Saba; Sourasky Medical Centre Ethics Committee, Tel Aviv. Italy: Comitato Etico - A.O. Universitaria di Bologna Policlinico S.Orsola-Malpighi di Bologna, Bologna; Comitato Etico A.O. Ospedale Civile di Legnano, Legnano; CEI Ist. Naz. per lo Studio e la Cura dei Tumori, Milan; Comitato Etico Indipendente della Fondazione IRCCS Istituto Nazionale dei Tumori di Milano, Milan; Comitato Etico IRCCS Ospedale San Raffaele, Milan; Comitato Etico per Piacenza e Parma, Parma; Comitato Etico Regionale Unico CERU, Udine; CESC della Provincia di Venezia e IRCCS San Camillo: Azienda ULSS 12 Venezia, Venice Mestre; CE per la Sperim. e Clinica della Prov. di Vicenza, Vicenza. Korea: Samsung Medical Center EC, Seoul; Seoul National Univ. Ethics Committee, Seoul. Mexico: Subcomité de Investigación del Centenario Hospital Miguel Hidalgo, Aguascalientes. Peru: Comité de Ética en Investigación - Arequipa; Hospital Nacional Carlos Alberto Seguín Escobedo, Arequipa; Comité de Etica del Hospital Rebagliati, Lima. Poland: Komisja Bioetyczna przy Uniwesytecie Medycznym im. Karola Marcinkowskiego w Poznaniu, Poznan. Russian Federation: E. C. of SI of Healthcare Kazan Oncology Dispensary, Kazan; E. C. of FSBI “Russian Oncology Research Center named after N.N. Blokhin” of RAMS, Moscow; E.C. of NSI of Healthcare Central Clinical Hospital #2 n.a. N. A. Semashko of the Russian Railways, Moscow; Ethics Board of the Ministry of Health Care and Social Development of the Russian Federation, Moscow; Ethics Committee at Roszdravnadzor, Moscow; E. C. of SBI of Healthcare of Stavropol region Pyatigorsk Oncology Dispensary, Pyatigorsk; E. C. of SBI of Healthcare Regional Clinical Oncology Dispensary, Ryazan; E.C of SBI of Healthcare Samara Regional Clinical Oncology Dispensary, Samara; E. C. of Research Oncology Institute n.a. N.N. Petrov of Ministry of Health and Social Development, St. Petersburg; E. C. of SBI of Healthcare Leningrad Regional Oncology Dispensary, St. Petersburg. Spain: CEIC Área 5 - Hospital Universitario La Paz, Madrid; Comité de Investigación Clínica, Madrid; Hospital Ramon y Cajal ;Comité Etico de Investigación Clínica, Madrid; CEIC Corporacio Sanitaria Parc Tauli, Sabadell (Barcelona); CAEC-CEIC Autonomico de Ensayos Clinicos Andalucia, Seville; Hospital Clinico Universitario de Valencia; CEIC, Valencia; CEIC de Euskadi; Dep. de Sanidad del Gobierno Vasco, Vitoria. Comité Ético de Investigación Clínica de Aragón, Zaragoza. Sweden: Regionala Etikprövningsnämnden i Stockholm, Stockholm. United Kingdom: University Hospitals Coventry & Warwickshire NHS Trust; R&D Department, Coventry; The Christie NHS Foundation Trust; Research & Development Office, Manchester.