Endocrine responsiveness in estrogen receptor-positive breast cancer
Kruger, D.T.
2020
document version
Publisher's PDF, also known as Version of record
Link to publication in VU Research Portal
citation for published version (APA)
Kruger, D. T. (2020). Endocrine responsiveness in estrogen receptor-positive breast cancer: Search for
biomarkers associated with treatment failure.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal ? Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
E-mail address:
in the PI3K and MAPK pathways in ER-positive,
HER2-negative breast cancer with potential
therapeutic consequences
Dinja T Kruger Karin J Beelen Mark Opdam Joyce Sanders Vincent van der Noort Epie Boven Sabine C Linn
Abstract
Background
The phosphatidylinositol-3-kinase (PI3K) and/or mitogen-activated protein kinase (MAPK) pathways are frequently activated in breast cancer which can result in anti-oestrogen resistance. Single protein markers failed to be introduced into clinical practice. We therefore aimed to find a better read-out of activation of the PI3K and MAPK pathways in ER+/HER2- breast cancer. Assessment of seven PI3K/MAPK proteins might better reflect pathway activation and distinguish patients without adjuvant tamoxifen benefit.
Methods
Tumour blocks were recollected from 293 primary postmenopausal ER+/HER2- breast cancer patients randomized between tamoxifen and no adjuvant therapy. PTEN, p-AKT(Thr308), p-AKT(Ser473), p-p70S6K, p-4EBP1, p-ERK1/2 and p-S6RP expression was assessed by immunohistochemistry followed by unsupervised hierarchical clustering. The primary endpoint was recurrence-free interval. Multivariate Cox models were used to assess tamoxifen benefit. A classification tool was developed based on protein expression profile.
Results
Subgroups were identified with preferentially activated (A) and preferentially not activated (N) proteins. Patients in group N derived significant benefit from tamoxifen (multivariate HR 0.23, p = 0.000101), while patients from group A did not (multivariate HR 1.37, p = 0.64), p for interaction 0.020. Our generated classification tool confirmed these results (p for interaction 0.024).
Conclusions
Hierarchical clustering of seven PI3K/MAPK proteins reflects pathway activation and can guide treatment decisions in primary ER+/HER2- postmenopausal breast cancer patients.
Background
Both the phosphatidylinositol-3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) and the mitogen-activated protein kinase (MAPK) pathways are important in normal cell function and cancer cell biology. By the transmission of cell signals in response to extracellular events, they promote RNA translation, proliferation, growth and cell survival (Figure S1)(1, 2). In breast cancer, cross-talk between both pathways and the oestrogen receptor (ER) signalling pathway has been described(2-5).
The PI3K pathway is the most frequently altered pathway in breast cancer(5, 6). Alterations in genes encoding proteins in this pathway occur in over 75% of primary breast cancer cases(6), but the numbers vary among breast cancer subtypes. For example, activating mutations in PIK3CA, which encodes the catalytic subunit of PI3K, occur most frequently in ER-positive (29-45%) and human epidermal growth factor receptor 2 (HER2)-positive tumours (39%) and less in triple-negative breast cancer (TNBC) cases (9%)(7). Alterations in protein expression of the tumour regulator PTEN are most often seen in TNBC (35-67%) compared to 29-44% in ER-positive breast tumours and 19-22% of HER2-positive breast cancer cases(7). High levels of phosphorylated Akt are common in all breast cancer subtypes of which p-Akt(Ser473) has most often been studied(8). Overexpression of p-Akt(Ser473) can be detected in 78% of luminal-like breast cancer, in 80% of HER2-positive breast cancer and in 58-62% of TNBC cases(9).
The MAPK pathway is less frequently altered in breast cancer. Mutations in genes encoding this pathway range from 2% to 10%(10). Activation seems most important in the tumour biology of TNBC(1). Umemura et al(11) have demonstrated more phosphorylation of ERK1/2 in TNBC than in other subtypes, although this result was not significant possibly due to the small sample size. Analysis of an ERK1/2 miRNA signature has indicated that a high activation status was significantly associated with ER-negativity, high tumour grade, increased proliferation, basal and HER2 molecular subtypes, and poor clinical outcomes(12).
Increasing evidence exists that activation of the PI3K pathway or the MAPK pathway in breast cancer is associated with a higher chance to develop resistance to anti-oestrogens or oestrogen deprivation(4, 13, 14). Pathway activation may also influence prognosis(8, 15-17). Therefore, research has focused on the identification of predictive and prognostic biomarkers associated with pathway activation that might be relevant for patient selection. Consequently, both total levels and phosphorylation of proteins downstream in the PI3K and MAPK pathways have been studied. Phosphorylated proteins, however, are considered as typical markers for pathway activation.
Previously, our group has demonstrated that high levels of phosphorylated p70S6K (p-p70S6K) in primary breast cancer were associated with less adjuvant
tamoxifen benefit(18). Similarly, tamoxifen was less effective in patients with tumours that expressed high p-mTOR and positive p-ERK1/2(18). Analysis of the tumour material for changes in the single proteins PTEN, p-AKT(Ser473), p-AKT(Thr308), p-mTOR, p-ERK1/2 as well as the presence of PIK3CA mutations and a possible association with tamoxifen benefit was negative(18, 19). The phosphorylation status of two other proteins downstream in the PI3K/MAPK pathways, p-4EBP1 and p-S6RP, is often used as a measure of activation of the PI3K pathway(20). These proteins act separately downstream of mTOR (Figure S1) and might also serve as markers for endocrine therapy benefit. We now analysed the predictive and prognostic potential of p-4EBP1 and p-S6RP expression in the same primary breast cancer samples as examined before(18, 19).
Since various feedback loops and cross-talks between the pathways exist, a single marker is at risk to produce false positive or false negative results when used as readout for pathway activation. Therefore, the aim of this study was to search for a better read-out of activated proteins of the PI3K and MAPK pathways in ER-positive(ER+)/human epidermal growth factor receptor 2 (HER2)-negative(HER2–) breast cancer patients, predictive for endocrine resistance or associated with prognosis. Unsupervised hierarchical clustering was performed to show for the first time how seven proteins downstream in both pathways are expressed in ER+/HER2– breast cancer cases. We assessed the prognostic value of panels of protein expression levels reflecting differences in activation status of the PI3K/ MAPK pathways in ER+/HER2– breast cancer patients as well as of a potential tool to predict the usefulness of adjuvant tamoxifen.
Methods
Patients and material
Primary tumour tissue blocks were recollected from postmenopausal patients with stage I-III breast cancer who participated in the IKA trial and were randomised between adjuvant tamoxifen versus no adjuvant endocrine therapy(21). The patient characteristics and clinical outcome of the original study population have been presented elsewhere(18, 21) and are part of the Oxford patient-level meta-analysis(22). A summary of the IKA trial can be found in the Supplemental Methods. Tumour material was available from 739 patients. These patients did not differ in clinico-pathological characteristics from the total group (Table 1)(18, 19). From the formalin-fixed paraffin-embedded tumour blocks, tissue microarrays (TMAs) were constructed using three 0.6 mm cores. Each TMA was stained for ERα, progesterone receptor (PR), HER2 and Ki67 (Supplementary Methods). The mitotic count was assessed per 2 mm2 as before(23). Tumour grade was scored on a hematoxylin-eosin stained slide according to the modified Bloom & Richardson scoring system(24). For
this retrospective translational study, no additional consent was required according to Dutch legislation(25), since the use of anonymized archival pathology left-over material does not interfere with patient care. Tumour tissue was handled according to the Dutch code of conduct for responsible use of human tissue in the context of health research(26).
Immunohistochemistry
TMAs have previously been analysed for PTEN, p-AKT(Thr308), p-AKT(Ser473), p-p70S6K and p-ERK1/2 of which methodology for scoring is summarized in Table S1(18, 19). The age of the tumour samples and different fixation procedures did not affect the phospho-protein staining procedure(18, 19). For p-4EBP1 and p-S6RP, staining was performed using a standardized protocol on the Ventana Benchmark® Ultra system (Ventana Medical Systems, Tucson, Arizona, USA). To ensure phospho-specificity for all phospho-antibodies, a test TMA was treated with λ-phosphatase before staining, resulting in disappearance of the positive staining. For p-4EBP1, the percentage of tumour cells with positive nuclear staining was scored. For p-S6RP, the percentage of tumour cells with positive cytoplasmic and membranous staining was scored. Similar as before, the highest score out of three cores from the same tumour tissue block was used for the statistical analyses. The inter observer variability was determined by independent assessment of one stained TMA for each antibody by a second blinded observer. To that end, the scorings were analysed as binary factor using the median as cut-off to calculate the inter-observer variability expressed as kappa coefficient (Table S1)(27). For further analyses we used the scores generated by one of the two observers (MO).
Hierarchical clustering
Hierarchical clustering was performed using the continuous scoring values of all seven proteins to create a heatmap. The dendrogram of the heatmap was created by unsupervised clustering, meaning that the clusters formed were based purely on the similarities and dissimilarities among the patients by the expressions of the seven proteins. No information about the patients’ survival or any other information about the patient or arm of randomization was used generating the heatmap. Concretely, the dendrogram was computed as follows. In the subset of 293 ER+/HER2- tumours for which scorings of the seven proteins of interest were available, we first normalised the scores by dividing each number by the standard deviation of the scores for that protein. On the resulting rescaled dataset, hierarchical clustering was carried out by the ‘heatmap.2’ function of the R-package ‘gplot’, using the Euclidean distance to determine the similarity among the tumours and a complete linkage function to iteratively build up the clusters. These latter two choices are the default settings in R’s ‘heatmap.2’ function. The heatmap cluster showing the highest levels of activation of the proteins was denoted subgroup A and the cluster showing less activation was denoted subgroup N.
Classification tool
A practical disadvantage of hierarchical clustering is that it cannot be used to classify a new patient into one of the two clusters. As a solution, we devised a classification tool for ER+/HER2- patients. With this rule, a new patient would be assigned to either the A or the N subgroup based on a particular expression profile of the seven proteins. The classification tool was optimised for reproducing the clusters detected by unsupervised hierarchical clustering without information on treatment or outcome.
Statistics
Recurrence-free interval (RFI) was defined as the time from the first randomisation to the occurrence of a local, regional or distant recurrence, or breast cancer-specific death(28). Patients with a secondary contralateral breast tumour were censored at the time of the contralateral diagnosis, since it was not possible to link breast cancer-specific events to the first or to the contralateral malignancy. Patients who died from other causes or who were lost to follow up, for instance because follow up was no longer required according to local practice, due to emigration or by patient’s request, were censored at the time of this occurrence. The Kaplan-Meier method was used to construct survival curves. The primary end-point of all analyses, tamoxifen benefit, was defined as the Hazard Ratio (HR) for RFI (as estimated by a Cox proportional hazard model) between the two arms of randomization: adjuvant tamoxifen for 1-3 years (TAM) and no adjuvant treatment (CON).
p-4EBP1 and p-S6RP were dichotomized before entering further analyses. An exploratory analysis to test which binary score would give the best prediction of tamoxifen benefit with p-4EBP1 and p-S6RP was performed by using multiple cut-off points for p-4EBP1 and p-S6RP. The binary score which reached the lowest p-value for interaction (i.e. showing the clearest difference in tamoxifen benefit between the two resulting subgroups) was used in further analyses. Fisher’s exact test was used to test the association between expression of p-4EBP1, p-S6RP and clinico-pathological characteristics or the other proteins associated with activation of the PI3K/MAPK pathway.
Tamoxifen benefit of high vs low p-4EBP1, high vs low p-S6RP, subgroup A vs N identified by hierarchical clustering, and subgroup FALSE and TRUE identified by our classification tool was estimated in bivariate and multivariate Cox proportional hazard regression analyses with a term for interaction between arm of randomisation and the variable of interest. Known prognostic variables were included in the multivariate analyses as covariates: age (<65 vs ≥65), histological grade (grade 3 vs grade 1 to 2), tumour size (T3 to T4 vs T1 to T2), histological subtype (lobular vs ductal), PR status (negative vs positive), Ki67 score (<10% vs ≥10%) and mitotic count (<8 mitotic counts/2 mm2 vs ≥8 mitotic counts/2 mm2). All Cox models were stratified for nodal status.
To test whether p-4EBP1 or p-S6RP expression levels, heatmap or classification tool subgroups were associated with prognosis of ER+ patients, univariate and multivariate Cox proportional hazard regression analyses were performed and stratified for lymph node status. Covariates included in the multivariate tests were treatment arm, age (<65 vs ≥65), histological grade (grade 3 vs grade 1 to 2), tumour size (T3 to T4 vs T1 to T2), histological subtype (lobular vs ductal), PR status (negative vs positive), Ki67 score (<10% vs ≥10%) and mitotic count (<8 mitotic counts per 2 mm2 vs ≥8 mitotic counts per 2 mm2). Prognostic analyses were carried out both in the entire study cohort using treatment arm as a covariate, as well as in the tamoxifen and control arm separately.
This study complied with reporting recommendations for tumour marker prognostic studies (REMARK) criteria(29). The analyses and heatmaps were generated using R for statistics (Windows version 3.3.1).
Results
Hierarchical clustering of proteins associated with the PI3K and MAPK pathways and differences in tamoxifen benefit and/or prognosis
The results regarding p-4EBP1 and p-S6RP as single markers can be found in the Supplemental Material (Results, Table S2 and Figure S2). In summary, neither p-4EBP1 nor p-S6RP appeared to have prognostic value or were useful to predict tamoxifen failure. We, therefore, hypothesized that discriminating tumours differing in general activation of the PI3K/MAPK pathways based on seven downstream protein expression levels would enable identifying patients without tamoxifen benefit.
A total of 489 tumours were ER+/HER2– and of these, 293 samples had with scorings for all seven proteins. In this cohort, a total of 63 RFI events were observed. The median follow-up of patients without a recurrence event was 8.2 years (95% CI 7.6 – 8.7). Unsupervised hierarchical clustering of the proteins in the PI3K/MAPK pathways yielded two patient subgroups (A: preferentially activated and N: preferentially not activated) (Figure 1). There were no significant differences in clinico-pathological characteristics between the two heatmap subgroups (Table 1 and 2). More proteins were activated in tumour tissue from patients in group A, while in tumours of patients from group N activation was relatively absent. Patients in group N derived significant benefit from tamoxifen (multivariate HR = 0.23, 95% CI 0.11 − 0.49, p = 0.000101)(Table 3 and S3A), while patients in group A did not (multivariate HR = 1.37, 95% CI 0.38 4.97, p = 0.636)(Table 3 and S3A). Figure 2 shows the Kaplan-Meier curves for the heatmap subgroups by treatment arm. When stratified for lymph node status, a significant interaction for heatmap group with tamoxifen efficacy was observed (multivariate p for interaction = 0.020).
The prognostic potential of heatmap cluster was analysed only in patients randomised to the control arm to rule out bias from treatment. Interestingly, control patients in group A demonstrated a better prognosis than control patients in group N when stratified for lymph node status (multivariate HR = 0.061, 95% CI 0.0079 − 0.47, p = 0.0073)(Table S3B) as shown in Figure 2 (CON groups).
Figure 1. Hierarchical clustering of seven proteins in ER-positive, HER2-negative tumours visu-alized in a heatmap. Red bars indicate a higher score in activation. Blue bars indicate a lower score and, therefore, less activation of the corresponding protein. The red box classifies the patients in group A with a more activated PI3K and/or MAPK pathway, while the black box iden-tifies patients in group N in whom these pathways are grossly not activated. Abbreviations: p-ERK1/2, phosphorylated extracellular signal-regulated kinase 1 and 2; p-S6RP, phosphory-lated 40S ribosomal protein S6; p-p70S6K, phosphoryphosphory-lated p70 ribosomal protein S6 kinase, p-AKT: phosphorylated AKT at phospho-sites Thr308 and Ser473; p-4EBP1, initiation factor eukaryotic initiation factor 4E binding protein 1; PTEN: phosphatase and tensin homolog.
Ta bl e 1 . D is tr ib ut io n o f c lin ic o-pa th ol og ic al c ha ra ct er is tic s o f p at ie nt s p re se nt ed i n t he E R-po si tiv e, H ER 2-ne ga tiv e h ea tm ap a nd s ub gr ou ps ( A a nd N ) a s w el l a s i n t he o ri gi na l d at a s et o f p at ie nt s w ith t um or m at er ia l a va ila bl e a nd i n t he t ot al I K A t ri al p op ul at io n H eat map grou p A H eat map grou p N To ta l h eat map po pu la tio n To ta l E R+ /H ER 2- po pu la tio n Pa tie nt s w ith t umor mat er ia l a vail ab le To ta l I K A s tu dy po pu la tio n n ( %) n ( %) n ( %) n ( %) n ( %) n ( %) To ta l 111 (3 8) 18 2 ( 62 ) 29 3 ( 10 0) 48 9 ( 10 0) 73 9 ( 10 0) 16 62 (1 00 ) A ge <65 59 (5 3) 84 (4 6) 14 3 ( 49) 23 3 (4 8) 37 8 ( 51) 86 9 ( 52) ≥65 52 (4 7) 98 ( 54 ) 15 0 ( 51 ) 25 6 ( 52) 36 1 ( 49) 79 3 (4 8) Ly m ph n od e s ta tu s N ega ti ve 64 ( 58 ) 95 (5 2) 15 9 ( 54 ) 276 (5 6) 39 3 ( 53) 901 ( 54 ) Po si ti ve 47 (4 2) 87 (4 8) 13 4 ( 46 ) 213 (4 3) 346 (4 7) 761 (4 6) T s ta ge T1-2 101 ( 91 ) 16 0 ( 88 ) 26 1 ( 89) 43 7 (8 9) 65 9 ( 89) 14 82 ( 89) T3 -4 10 ( 9) 22 (12 ) 32 (11 ) 52 (11 ) 80 (11 ) 18 0 ( 11 ) G ra de G ra de 1-2 72 (6 5) 11 4 ( 63 ) 18 6 ( 63 ) 34 2 ( 70 ) 43 5 ( 59) 43 5 ( 59) * G ra de 3 39 (35) 68 ( 37 ) 10 7 ( 37 ) 14 7 ( 30 ) 30 4 ( 41) 30 4 ( 41) * Hi st ol og ic al s ubt yp e D uc tal 79 (9 3) 15 4 ( 94 ) 23 3 ( 94 ) 347 (8 7) 54 0 (8 9) 54 0 (8 9) * Lo bula r 6 ( 7) 10 (6 ) 16 (6 ) 50 (13 ) 66 (11 ) 66 (11 )* HE R 2 s ta tus N ega ti ve 111 (1 00 ) 18 2 ( 10 0) 29 3 ( 10 0) 48 9 ( 10 0) 59 4 ( 88 ) 59 4 ( 88 )* Po si ti ve 0 ( 0) 0 ( 0) 0 ( 0) 0 ( 0) 85 (12 ) 85 (12 )* PR s ta tu s N ega ti ve 44 (4 0) 84 (4 6) 12 8 ( 44 ) 222 (4 6) 414 (5 7) 34 6 (4 0) † Po si ti ve 67 (6 0) 97 ( 54 ) 16 4 ( 56 ) 261 (5 4) 30 4 (4 3) 513 (6 0) † ER s ta tu s N ega ti ve 0 ( 0) 0 ( 0) 0 ( 0) 0 ( 0) 15 9 ( 23 ) 311 (2 3) ‡ Po si ti ve 111 (1 00 ) 18 2 ( 10 0) 29 3 ( 10 0) 48 9 ( 10 0) 56 3 ( 77 ) 101 4 ( 77 )‡ * o nl y r ev is ed s co ri ng s f ro m 7 39 p at ie nt s o f t he I K A t ri al p op ul at io n f ro m w ho m t um or t is su e c ou ld b e o bt ai ne d a re s ho w n † d et er m in ed b y p ro ge st er on e r ec ep to r ( PR ) l ig an d b in di ng a ss ay i n o ri gi na l t ri al , m is si ng d at a o f 8 03 p at ie nt s ‡ d et er m in ed b y e st ro ge n r ec ep to r ( ER ) l ig an d b in di ng a ss ay i n o ri gi na l t ri al , m is si ng d at a o f 3 37 p at ie nt s
2.
Ta bl e 2 . A ss oc ia tio n b et w ee n c lin ic o-pa th ol og ic al f ac to rs i n h ea tm ap c lu st er A v s N a nd c la ss ifi ca tio n t oo l T RU E v s F A LS E ( n ( % )). H ea tm ap ( n = 2 93 ) Cl as si fi ca ti on t oo l ( n = 2 93 ) A N TRU E FA LS E 111 (3 8) 18 2 ( 62 ) p-val ue * 12 4 ( 42 ) 16 9 ( 58 ) p-val ue * A ge <65 59 (5 3) 84 (4 6) 0. 28 63 (5 1) 80 (4 7) 0.6 4 ≥65 52 ( 75 ) 98 ( 54 ) 61 ( 49) 89 (5 3) Ly m ph n od e s ta tu s N ega ti ve 64 ( 58 ) 95 (5 2) 0. 4 73 ( 59) 86 (5 1) 0.1 9 Po si ti ve 47 (4 2) 87 (4 8) 51 (4 1) 83 ( 49) T s ta ge T1-2 101 ( 91 ) 16 0 ( 88 ) 0. 45 111 (9 0) 15 0 ( 89) 1 T3 -4 10 ( 9) 22 (12 ) 13 (1 0) 19 (11 ) G ra de G ra de 1-2 72 (6 5) 11 4 ( 63 ) 0.7 1 82 (6 6) 10 4 ( 62 ) 0. 46 G ra de 3 39 (35) 68 ( 37 ) 42 ( 34 ) 65 ( 38 ) Hi st ol og ic al s ubt yp e D uc tal 79 (9 3) 15 4 ( 94 ) 0.7 9 88 ( 94 ) 14 5 ( 94 ) 1 Lo bula r 6 ( 7) 10 (6 ) 6 ( 6) 10 (6 ) PR s ta tu s N ega ti ve 44 (4 0) 84 (4 6) 0. 28 50 (4 0) 78 (4 6) 0. 34 Po si ti ve 67 (6 0) 97 ( 54 ) 74 (6 0) 90 ( 54 ) Ki 67 <1 0% 66 (6 0) 15 9 ( 88 ) < 0 .0 001 77 (6 3) 14 8 ( 88 ) < 0 .0 001 ≥1 0% 44 (4 0) 22 (12 ) 46 (3 7) 20 (12 ) M itot ic c oun t <8 / 2 m m 2 51 (4 6) 84 (4 7) 1 55 (4 5) 80 (4 8) 0.7 2 ≥8 / 2m m 2 59 (5 4) 96 (5 3) 67 ( 55 ) 88 (5 2) * F is he r’s e xa ct t es t b as ed o n c as es w ith ou t m is si ng v al ue s
Table 3.Multivariate Cox proportional hazard model of recurrence free interval (RFI) including heatmap group and treatment interaction
Variable HR 95% CI p
Interactie Cluster vs treatment 0.012
Tamoxifen vs CON (ref) In cluster N 0.257 0.125 – 0.529 0.00023 In cluster A 1.670 0.469 – 5.943 0.429 Cluster A vs N (ref) CON patients 0.229 0.061 – 0.856 0.0283
TAM patients 1.49 0.74 – 3.00 0.261 Age ≥65 vs <65 (ref) 1.091 0.630 – 1.890 0.757 T stage T3-4 vs T1-2 (ref) 1.965 0.977 – 3.951 0.058 Grade Grade 3 vs grade 1-2 (ref) 1.813 0.805 – 4.080 0.151 PR status Positive vs negative (ref) 1.489 0.850 – 2.606 0.164 Histological subtype Lobular vs ductal (ref) 2.630 0.966 – 7.156 0.058 Ki67 ≥10% vs <10% (ref) 1.258 0.619 – 2.556 0.525 Mitotic count ≥8 / 2mm² vs <8 / 2mm² (ref) 0.734 0.316 – 1.703 0.471
Classification tool
A decision rule was developed based on the expression profile of the seven proteins by hierarchical clustering, blinded to outcome, to classify future postmenopausal patients with primary ER+/HER2– breast cancer into one of the heatmap subgroups. When a patient fulfilled at least one of the following four criteria, the rule would be TRUE and the patient would be assigned to the activated A group: [p-AKT(Thr308) > 1]; or [p-S6RP > 70%]; or [p-4EBP1 > 50% and p-ERK1/2 > 70%]; or [p-p70S6K > 0 and PTEN > 1 and p-AKT(Ser437) > 1]. If none of these criteria was met, the rule would be FALSE and the corresponding patient would be categorized in the not activated N group. After application of this classification tool in the 293 ER+/ HER2– patients, 88% were categorized in the same heatmap subgroup as observed with unsupervised hierarchical clustering. The clinico-pathological characteristics of the classification tool subgroups are shown in Table 2. The classification tool was equally successful in differentiating ER+/HER2– patients with or without tamoxifen benefit (p for interaction = 0.024). When the rule was TRUE, no benefit of tamoxifen was observed (multivariate HR = 1.30, 95% CI 0.36−4.68, p = 0.69)(Table S3C and S3D). When the rule was FALSE, patients derived significant benefit from tamoxifen (multivariate HR = 0.25, 95% CI 0.07−0.91, p = 0.000124)(Table S3C and S3D). The Kaplan-Meier curves are visualized in Figure 3.
Similar to the heatmap cluster analyses, the prognostic potential of the classification tool was analysed only in patients randomised to the control arm and stratified for lymph node status. When the rule was TRUE, control patients demonstrated a better prognosis than when the rule was FALSE (multivariate HR = 0.065, 95% CI 0.0091 − 0.469, p = 0.0067)(Table S3E) as shown in Figure 3 (CON groups).
Fi gu re 2 . K ap la n– M ei er a na ly se s f or r ec ur re nc e-fr ee i nt er va l i n E R-po si tiv e, H ER 2-ne ga tiv e b re as t c an ce r p at ie nt s a cc or di ng t o t am ox ife n tr ea tm en t a nd a hi er ar ch ic al cl us te r N , i nd ic at in g pa tie nt s as si gn ed to gr ou p N (p at hw ay s gr os sl y no t a ct iv at ed )(M ul ti va ri at e H R: 0. 26 , p = 0. 00 02 28 ); b h ie ra rc hi ca l c lu st er A , i nd ic at in g p at ie nt s a ss ig ne d t o g ro up A ( pr ef er en tia lly a ct iv at ed P I3 K a nd /o r M A PK p at hw ay ) (M ul ti va ri at e H R: 1. 67 , p = 0. 42 9) ; T he in te rr up te d lin e sp ec ifi es pa tie nt s tr ea te d w ith ta m ox ife n. Th e co nt in uo us lin e sh ow s pa tie nt s ra nd om iz ed t o t he c on tr ol a rm .
Fi gu re 3. K ap la n –M ei er an al ys es fo r r ec ur re nc e-fr ee i nt er va l in ER -p os it iv e, H ER 2-ne ga ti ve b re as t c an ce r pa ti en ts a cc or di ng t o ta m ox ifen tr ea tm en t a nd a p at ie nt s as si gn ed to gr ou p FA LS E (c or re sp on di ng to th e pr ef er en tia lly no t a ct iv at ed he at m ap gr ou p N ) b y us e of ou r cl as si fic at io n to ol (M ul ti va ri at e H R : 0. 25 , p = 0 .0 00 18 3) ; an d b pa ti en ts a ss ig ne d to g rou p TR UE ( co rr es po nd in g to t he pr ef er en ti al ly ac ti va te d he at m ap g ro up A ) b y us e of o ur c la ss ifi ca tio n to ol (M ul ti va ri at e H R: 1 .6 9, p = 0 .4 11 ). Th e in te rr up te d lin e sp ec ifi es p at ie nt s tr ea te d w ith t am ox ife n. T he c on tin uo us l in e s ho w s p at ie nt s r an do m iz ed t o t he c on tr ol a rm .
2.
Discussion
We demonstrate by unsupervised hierarchical clustering of seven proteins that ER+/HER2– breast cancer patients with a tumour containing preferentially activated PI3K/MAPK pathways derived no benefit from adjuvant tamoxifen, while those without preferential activation had an improved RFI on tamoxifen. Our classification tool successfully categorized patients in these two subgroups based on their immunohistochemistry scorings. Further, patients in the more activated PI3K/MAPK pathway group had a better prognosis.
Within the ER+/HER2– breast cancer samples, we show for the first time that unsupervised hierarchical clustering resulted in the distinction of tumours with a preferentially activated (group A) and not activated (group N) profile categorizing patients with differential adjuvant tamoxifen benefit. Further, we successfully generated a tool to classify future ER+/HER2– breast cancer patients in subgroups A or N based on particular protein expression profiles. The clustering method has been used before by Horii et al(17) to explore biomarkers for clinico-pathological relevance. They examined 337 unselected primary breast cancer cases for staining scores of p-AKT(Ser473), cyclin D1, P27, p-p70S6K, p-4EBP1 and p-ERK1/2, and cluster classification showed significant relationships with subgroups expressed by hormone receptor, HER2, grade and histological subtype, but not with prognosis. Their study design did not allow for insight in treatment benefit.
Activation of the PI3K/MAPK pathways in the ER+/HER2– breast cancer samples was correlated with a better prognosis. Although this might seem a counterintuitive finding, it is in line with previous research. We have previously shown that patients with ER+ tumours harbouring high p-AKT(Ser473), p-AKT(Thr308), p-mTOR or p-p70S6K expression who did not receive adjuvant tamoxifen also had a decreased risk of breast cancer recurrence(18). Other groups have analysed the prognostic significance of PIK3CA mutations or the expression of individual proteins. In a recent meta-analysis(30), PIK3CA mutation status was not associated with relapse-free or overall survival in hormone receptor-positive breast cancer patients. Bostner et al(31) have not detected prognostic value for either p-mTOR or p-Akt(Ser473), but high expression of S6K1 predicted a worse prognosis(32). In the same cohort of patients, Karlsson et al(33) have described that strong nuclear 4EBP1 correlated with good prognosis, while strong cytoplasmic p-4EBP1 was a poor prognostic factor. In all, future research should reveal whether a panel of (phosphorylated) proteins related to the PI3K/MAPK pathways in hormone receptor-positive breast cancer might better reflect the true activation status of these pathways, and therefore give more precise prognostic information than single markers. For instance, earlier clinical research already
demonstrated the lack of correlation between the presence of PIK3CA mutations and the protein activation status of these same pathways(6, 19). Furthermore, considering it is unclear how an activated PI3K pathway can provide a better prognosis, it would be very useful if this becomes the subject of future research as well.
High levels of cytoplasmic PTEN were positively associated with PI3K pathway activation in our patient cohort. PTEN is generally known as a tumour suppressor gene and inactivation has been shown to be involved in heritable and sporadic cancer types(34). The incidence of mutations or deletions in PTEN is low in primary tumours(34), being 4% in primary breast cancer(6). PTEN staining results were only found to be weakly correlated with PTEN gene expression(35). Localization of PTEN appears to be cell cycle dependent; quiescent tissues exhibit predominantly nuclear localization, whereas cancerous tissues have a higher percentage of cells in S phase and exhibit increased levels of cytoplasmic PTEN(34). Cytoplasmic PTEN has several proteoforms varying in functionality(36). Tumours with a high degree of immunoreactivity with a PTEN antibody may in fact reflect functionally PTEN-deficient samples(36). These findings could be an explanation that cytoplasmic
PTEN expression is positively associated with PI3K pathway activation in our study. Whether our classification tool can be used to distinguish high-risk postmenopausal patients with primary ER+/HER2– breast cancer that do not benefit from adjuvant tamoxifen should be validated in a retrospective independent trial cohort before it can be implemented in the clinic. The strength of the IKA trial of which we recollected the tumour tissue from the participating patients, is that it included a control group not receiving any systemic treatment to discern a predictive from a prognostic effect of a biomarker(37). This kind of retrospective cohorts are scarce and it would not be ethically justified to withhold a control group adjuvant hormonal therapy today when setting up a new trial. Two retrospective trials that could be used to validate our results are the Stockholm trial(32, 33) and/or the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-14 trial(38). Before using such valuable patient material, more evidence should be generated on less precious material. For this, a matched case-control study could be set up using the Netherlands Cancer Registry database coupled with the nationwide digital pathology archive, through which tumour material can be recollected for research purposes, e.g. the PARADIGM initiative(39).
Current adjuvant endocrine therapy in postmenopausal patients generally consists of tamoxifen given in sequence with an aromatase inhibitor (AI) for at least 5 years(40). In our study, patients received only 1 to 3 years of tamoxifen without the addition of an AI. Since PI3K pathway activation is associated with endocrine resistance irrespective of the given treatment(5), we anticipate that our findings
might also be applicable to indicate a subgroup of patients without advantage from adjuvant AIs. Although we have no information on the meaning of our results when applied to longer treatment duration with tamoxifen, we think that it is highly unlikely that a tumour with intrinsic resistance against tamoxifen via activated PI3K/ MAPK pathways, will lose this resistance mechanism after longer treatment with the same drug. We therefore assume that our findings can also be applied to patients who receive current standard care and this could in the future be addressed by analysing studies like the NSABP B-14 trial(38). For premenopausal patients, the most recent ASCO guideline recommends offering adjuvant tamoxifen for at least 5 or even 10 years(40). We demonstrated that the absence of activated PI3K/MAPK pathways in the primary tumour was predictive for substantial tamoxifen benefit in postmenopausal patients. However, whether these findings will also be applicable in the premenopausal setting is currently unknown and requires further research.
In our study a substantial proportion of patients had a PI3K/MAPK pathway activation (111/293=38%) and did not derive benefit from adjuvant tamoxifen. Nevertheless, not all PI3K/MAPK pathway-negative patients continued to be recurrence-free after tamoxifen, indicating that additional resistance mechanisms remain to be identified(41).
Currently, a number of multigene tests are available, such as the MammaPrint, PAM50-based risk of recurrence, Oncotype DX, IHC4 score and Breast Cancer Index(42, 43). These tests have been investigated for the prediction of outcome after endocrine treatment and/or endocrine therapy benefit in ER+/HER2- breast cancer patients. Although they added valuable prognostic information, the most recent ASCO guidelines recommend using them for clinical decisions on adjuvant chemotherapy only and not for decisions regarding adjuvant endocrine therapy due to lack of evidence for that specific goal(42, 43). To decide whether or not adjuvant endocrine (mono) therapy should be used, our classification tool could be of added value.
In conclusion, our study demonstrates the value of determining the expression of multiple proteins as readout for PI3K/MAPK pathway activation to predict adjuvant tamoxifen benefit in an ER+/HER2– breast cancer cohort. Furthermore, we developed a tool to classify patients into two categories that either derive substantial benefit or not from adjuvant tamoxifen on the basis of seven proteins at predefined cut-off expression levels. Before introduction into daily clinical practice, these results require validation in an independent study.
Additional information
Ethics approval and consent to participate
The IKA trial was approved by the central ethics committee of the Netherlands Cancer Institute. Informed consent was obtained from all participating patients. The trial was performed in accordance with the Declaration of Helsinki. For this retrospective translational study, no additional consent was required according to Dutch legislation, since the use of anonymized archival pathology left-over material does not interfere with patient care. Tumour tissue was handled according to the Dutch code of conduct for responsible use of human tissue in the context of health research.
References
1. Saini KS, Loi S, de Azambuja E, Metzger-Filho O, Saini ML, Ignatiadis M, et al. Targeting the PI3K/AKT/mTOR and Raf/MEK/ERK pathways in the treatment of breast cancer. Cancer treatment reviews. 2013;39(8):935-46.
2. Ciruelos Gil EM. Targeting the PI3K/AKT/mTOR pathway in estrogen receptor-positive breast cancer. Cancer treatment reviews. 2014;40(7):862-71.
3. Osborne CK, Schiff R. Mechanisms of endocrine resistance in breast cancer. Annual review of medicine. 2011;62:233-47.
4. Nardone A, De Angelis C, Trivedi MV, Osborne CK, Schiff R. The changing role of ER in endocrine resistance. Breast. 2015;24 Suppl 2:S60-6.
5. Miller TW, Balko JM, Arteaga CL. Phosphatidylinositol 3-kinase and antiestrogen resistance in breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2011;29(33):4452-61.
6. Cancer Genome Atlas N. Comprehensive molecular portraits of human breast tumours. Nature. 2012;490(7418):61-70.
7. Hosford SR, Miller TW. Clinical potential of novel therapeutic targets in breast cancer: CDK4/6, Src, JAK/STAT, PARP, HDAC, and PI3K/AKT/mTOR pathways. Pharmacogenomics and personalized medicine. 2014;7:203-15.
8. Yang ZY, Di MY, Yuan JQ, Shen WX, Zheng DY, Chen JZ, et al. The prognostic value of phosphorylated Akt in breast cancer: a systematic review. Scientific reports. 2015;5:7758.
9. Aleskandarany MA, Rakha EA, Ahmed MA, Powe DG, Ellis IO, Green AR. Clinicopathologic and molecular significance of phospho-Akt expression in early invasive breast cancer. Breast cancer research and treatment. 2011;127(2):407-16. 10. Kourea HP, Zolota V, Scopa CD. Targeted pathways in breast cancer: molecular and
protein markers guiding therapeutic decisions. Current molecular pharmacology. 2014;7(1):4-21.
11. Umemura S, Yoshida S, Ohta Y, Naito K, Osamura RY, Tokuda Y. Increased phosphorylation of Akt in triple-negative breast cancers. Cancer science. 2007;98(12):1889-92.
12. Miller PC, Clarke J, Koru-Sengul T, Brinkman J, El-Ashry D. A novel MAPK-microRNA signature is predictive of hormone-therapy resistance and poor outcome in ER-positive breast cancer. Clinical cancer research : an official journal of the American Association for Cancer Research. 2015;21(2):373-85.
13. Tryfonidis K, Zardavas D, Katzenellenbogen BS, Piccart M. Endocrine treatment in breast cancer: Cure, resistance and beyond. Cancer treatment reviews. 2016;50:68-81. 14. Miller TW, Rexer BN, Garrett JT, Arteaga CL. Mutations in the phosphatidylinositol
3-kinase pathway: role in tumor progression and therapeutic implications in breast cancer. Breast cancer research : BCR. 2011;13(6):224.
15. Yang SX, Polley E, Lipkowitz S. New insights on PI3K/AKT pathway alterations and clinical outcomes in breast cancer. Cancer treatment reviews. 2016;45:87-96.
16. Mayer IA. Clinical Implications of Mutations in the PI3K Pathway in HER2+ Breast Cancer: Prognostic or Predictive? Current breast cancer reports. 2015;7(4):210-4. 17. Horii R, Matsuura M, Dan S, Ushijima M, Uehiro N, Ogiya A, et al. Extensive analysis
of signaling pathway molecules in breast cancer: association with clinicopathological characteristics. International journal of clinical oncology. 2015;20(3):490-8. 18. Beelen K, Opdam M, Severson TM, Koornstra RH, Vincent AD, Wesseling J, et al.
Phosphorylated p-70S6K predicts tamoxifen resistance in postmenopausal breast cancer patients randomized between adjuvant tamoxifen versus no systemic treatment. Breast cancer research : BCR. 2014;16(1):R6.
19. Beelen K, Opdam M, Severson TM, Koornstra RH, Vincent AD, Wesseling J, et al. PIK3CA mutations, phosphatase and tensin homolog, human epidermal growth factor receptor 2, and insulin-like growth factor 1 receptor and adjuvant tamoxifen resistance in postmenopausal breast cancer patients. Breast cancer research : BCR. 2014;16(1):R13.
20. Li S, Kong Y, Si L, Chi Z, Cui C, Sheng X, et al. Phosphorylation of mTOR and S6RP predicts the efficacy of everolimus in patients with metastatic renal cell carcinoma. BMC cancer. 2014;14:376.
21. 21st Annual San Antonio Breast Cancer Symposium — December 12–15, 1998; Abstracts. Abstract 329, poster session III: Vermorken JB, et al. Adjuvant tamoxifen in breast cancer: interim results of a comprehensive cancer center Amsterdam trial Breast Cancer Res Treat. 1998;50(3):203-335.
22. Early Breast Cancer Trialists’ Collaborative G. Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet. 2005;365(9472):1687-717.
23. van Diest PJ, Baak JP, Matze-Cok P, Wisse-Brekelmans EC, van Galen CM, Kurver PH, et al. Reproducibility of mitosis counting in 2,469 breast cancer specimens: results from the Multicenter Morphometric Mammary Carcinoma Project. Human pathology. 1992;23(6):603-7.
24. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology. 1991;19(5):403-10.
25. Non-WMO research, Central Committee on Research involving Human Subjects: (Centrale Commissie Mensgebonden Onderzoek) [Web page]. http://www.ccmo.nl/ en/non-wmo-research. Available from: http://www.ccmo.nl/en/non-wmo-research. 26. Dutch Federation of Biomedical Scientific Societies: (Federatie van medisch
wetenschappelijke verenigingen) gebruik-van-lichaamsmateriaal-2011. Available from: https://www.federa.org/code-goed-gebruik-van-lichaamsmateriaal-2011.
27. McHugh ML. Interrater reliability: the kappa statistic. Biochemia medica. 2012;22(3):276-82.
28. Hudis CA, Barlow WE, Costantino JP, Gray RJ, Pritchard KI, Chapman JA, et al. Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: the STEEP system. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2007;25(15):2127-32.
29. McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM, et al. Reporting recommendations for tumor marker prognostic studies (REMARK). Journal of the National Cancer Institute. 2005;97(16):1180-4.
30. Pang B, Cheng S, Sun SP, An C, Liu ZY, Feng X, et al. Prognostic role of PIK3CA mutations and their association with hormone receptor expression in breast cancer: a meta-analysis. Scientific reports. 2014;4:6255.
31. Bostner J, Karlsson E, Pandiyan MJ, Westman H, Skoog L, Fornander T, et al. Activation of Akt, mTOR, and the estrogen receptor as a signature to predict tamoxifen treatment benefit. Breast cancer research and treatment. 2013;137(2):397-406. 32. Bostner J, Karlsson E, Eding CB, Perez-Tenorio G, Franzen H, Konstantinell A, et al.
S6 kinase signaling: tamoxifen response and prognostic indication in two breast cancer cohorts. Endocrine-related cancer. 2015;22(3):331-43.
33. Karlsson E, Perez-Tenorio G, Amin R, Bostner J, Skoog L, Fornander T, et al. The mTOR effectors 4EBP1 and S6K2 are frequently coexpressed, and associated with a poor prognosis and endocrine resistance in breast cancer: a retrospective study including patients from the randomised Stockholm tamoxifen trials. Breast cancer research : BCR. 2013;15(5):R96.
34. Planchon SM, Waite KA, Eng C. The nuclear affairs of PTEN. Journal of cell science. 2008;121(Pt 3):249-53.
35. Yndestad S, Austreid E, Knappskog S, Chrisanthar R, Lilleng PK, Lonning PE, et al. High PTEN gene expression is a negative prognostic marker in human primary breast cancers with preserved p53 function. Breast cancer research and treatment. 2017;163(1):177-90.
36. Malaney P, Uversky VN, Dave V. PTEN proteoforms in biology and disease. Cellular and molecular life sciences : CMLS. 2017.
37. Beelen K, Zwart W, Linn SC. Can predictive biomarkers in breast cancer guide adjuvant endocrine therapy? Nature reviews Clinical oncology. 2012;9(9):529-41. 38. Fisher B, Jeong JH, Dignam J, Anderson S, Mamounas E, Wickerham DL, et al. Findings
from recent National Surgical Adjuvant Breast and Bowel Project adjuvant studies in stage I breast cancer. J Natl Cancer Inst Monogr. 2001(30):62-6.
39. Dackus GM, Ter Hoeve ND, Opdam M, Vreuls W, Varga Z, Koop E, et al. Long-term prognosis of young breast cancer patients (</=40 years) who did not receive adjuvant systemic treatment: protocol for the PARADIGM initiative cohort study. BMJ open. 2017;7(11):e017842.
40. Burstein HJ, Temin S, Anderson H, Buchholz TA, Davidson NE, Gelmon KE, et al. Adjuvant endocrine therapy for women with hormone receptor-positive breast cancer: american society of clinical oncology clinical practice guideline focused update. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2014;32(21):2255-69.
41. Clarke R, Tyson JJ, Dixon JM. Endocrine resistance in breast cancer--An overview and update. Mol Cell Endocrinol. 2015;418 Pt 3:220-34.
42. Harris LN, Ismaila N, McShane LM, Andre F, Collyar DE, Gonzalez-Angulo AM, et al. Use of Biomarkers to Guide Decisions on Adjuvant Systemic Therapy for Women With Early-Stage Invasive Breast Cancer: American Society of Clinical Oncology Clinical Practice Guideline. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2016;34(10):1134-50.
43. Krop I, Ismaila N, Andre F, Bast RC, Barlow W, Collyar DE, et al. Use of Biomarkers to Guide Decisions on Adjuvant Systemic Therapy for Women With Early-Stage Invasive Breast Cancer: American Society of Clinical Oncology Clinical Practice Guideline Focused Update. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2017;35(24):2838-47.
Supplements
Supplementary Methods
Patients and material
The IKA trial (July 1982 until September 1993) included postmenopausal patients with stage I-III breast cancer who were randomised in a 2:1 ratio between adjuvant tamoxifen (30 mg/day) for 1 year versus no adjuvant endocrine therapy. After 1 year, patients on tamoxifen were randomised a second time to continue another 2 years of tamoxifen or to stop further treatment. In 1988, based on two interim analyses demonstrating a significant improvement in recurrence-free survival (RFS) in lymph node-positive patients, all node-positive patients were given 1 year of tamoxifen before they participated in the second randomization. None of the patients received chemotherapy. The IKA trial was approved by the central ethics committee of the Netherlands Cancer Institute. All patients gave informed consent. A total of 1662 patients were included.
Tissue microarrays (TMAs) were constructed from the 739 patients for which tumour material was available and were stained for ERα, progesterone receptor (PR), HER2 and the various PI3K/MAPK proteins. ER and PR were considered positive if ≥10% of tumour cells showed nuclear staining. HER2 was considered positive if membranous staining was DAKO score 3. If the DAKO score was 2, chromogenic in situ hybridization was performed and considered positive if the sample showed HER2 amplification. Immunohistochemistry for Ki67 was performed using the monoclonal mouse anti-human Ki67 antigen clone MIB-1 (DAKO, Agilent Technologies, Santa Clara, California, USA) and a standard staining protocol on the Ventana Benchmark® Ultra system (Ventana Medical Systems, Tucson, USA). The proportion of invasive tumour cells with nuclear staining was assessed by MO, a cut-off for binary score of <10% vs ≥10% was used in the analyses. During the scoring of the PI3K/MAPK proteins, cases with ductal carcinoma in situ or without residual invasive tumour and cores with technical errors were excluded. The number of tumour samples with staining results for all proteins was further diminished due to competing studies requiring slides from the same TMAs.
Supplementary Results
Expression of p-4EBP1 and p-S6RP, tamoxifen benefit and prognosis in ER-positive patients
In the group of ER-positive patients (N = 563), a total of 132 RFI events occurred. The median follow-up of patients without a recurrence event was 8.2 years. There were 430 and 442 ER-positive breast cancer cases that could be evaluated for p-4EBP1 and p-S6RP expression, respectively.
After evaluating multiple cut-off points in patients evaluable for staining results, we found that dichotomization of ≤20% vs >20% for p-4EBP1 and of negative vs positive for p-S6RP provided the best cut-off. Tumours with low and high p-4EBP1 (Table S2) and those with absent or present p-S6RP (Table S2) were analysed for a possible association with a series of patient and tumour characteristics. Both proteins were significantly associated with the other six proteins of the PI3K and/ or MAPK pathways. Tumours with a positive PR preferentially stained high for p-4EBP1 (p-4EBP1 high 66% and low 34%), which was also the case for tumours with a lower grade (p-4EBP1 high 64% and low 36%). Tumours with a positive PR were preferentially positive for p-S6RP (p-S6RP positive 68% and negative 32%), while those negative for p-S6RP were preferentially of lower grade (grade 1-2 70% and grade 3 30%). Patients with tumours harbouring high levels of p-4EBP1 experienced benefit from adjuvant tamoxifen (multivariate Hazard Ratio (HR) 0.39, 95% Confidence Interval (CI) 0.20 – 0.76, p = 0.006), while those with low tumour levels of p-4EBP1 did not (multivariate HR 0.74, 95% CI 0.32 – 1.74, p = 0.49) (Figures S2A and S2B). Patients with tumours without p-S6RP expression derived significant tamoxifen benefit (multivariate HR 0.28, 95% CI 0.12 – 0.67, p = 0.004), while those with positive tumour staining had no benefit (multivariate HR 0.64, 95% CI 0.33 – 1.21, p = 0.17) (Figure S2C and S2D). The tests for interaction between p-4EBP1 or p-S6RP and treatment were not significant (p = 0.23 and p = 0.12, respectively) indicating no significant difference in tamoxifen benefit according to the protein subgroups.
We explored tamoxifen efficacy in lymph positive and lymph node-negative patients separately. As expected from the interim analyses in the original trial, all lymph node-positive patients derived significant benefit from tamoxifen regardless of various p-4EBP1 or p-S6RP cut-off points. However, dichotomization of both p-4EBP1 and p-S6RP by ≤10% vs >10% in lymph node-negative patients revealed differences in tamoxifen outcome. In the Cox models, the test for interaction between treatment arm and protein expression was significant, although only in the univariate analysis (p-4EBP1: univariate p = 0.049 and multivariate p = 0.10; p-S6RP:
univariate p = 0.016 and multivariate p = 0.075). Patients with low tumour levels of p-4EBP1 had no tamoxifen benefit (multivariate HR 2.67, 95% CI 0.70 – 10.3, p = 0.15), while in patients with high levels more benefit was seen (multivariate HR 0.64, 95% CI 0.22 – 1.88, p = 0.42). On the contrary, a tamoxifen benefit was observed in patients with p-S6RP tumour levels ≤10% (multivariate HR 0.45, 95% CI 0.14 – 1.43, p = 0.18), while no benefit was observed in patients with p-S6RP levels of >10% (multivariate HR 2.17, 95% CI 0.60 – 7.87, p = 0.24).
The prognostic potential of high vs low p-4EBP1 and positive vs negative p-S6RP was analysed only in patients randomized to the control arm to rule out bias from treatment. There was no significant association of either p-4EBP1 or p-S6RP expression with prognosis (multivariate HR = 1.47, 95% CI 0.54 – 3.97, p = 0.45 and multivariate HR = 0.59, 95% CI 0.23 – 1.57, p = 0.24, respectively).
Sup
pl
em
ent
ar
y t
ab
le
s
Ta bl e S 1. A nt ib od ie s u se d f or i m m un oh is to ch em is tr y, s co ri ng p ro ce du re s a nd k ap pa c oe ffi ci en ts t o d et er m in e t he i nt er ob se rv er v ar ia bi lit y A nt ib od y A rt . N o. (cl one ) 1 Sc ori n g s ys te m Sc or in g f or hie ra rc hic al clu st er in g Co m par ab le cor es from N p at ie nt s fo r k ap pa ca lc ula tio n Cu t-off f or bi na ry s co re fo r k ap pa ca lc ula tio n K ap pa f or bi nar y s co re PT EN 95 59 (1 38 G 6) cy to pla sm ic in te nsi ty 0 – 3 12 6 0 vs 1-3 0.7 5 p-A K T ( Th r3 08 ) 29 65 (C 31E 5E ) cy to pla sm ic in te nsi ty 0 – 3 13 3 0 vs 1-3 0. 53 p-A K T ( Se r4 73 ) 40 60 (D 9E) cy to pla sm ic in te nsi ty 0 – 3 94 0-1 vs 2-3 0.7 0 p-4E B P1 (S er 65 ) 94 56 ( 17 4A 9) pe rc en ta ge of tum our ce lls w ith n uc le ar st ain in g 0 – 1 00 % 11 2 0-50 vs 5 1-10 0 0. 59 p-p7 0S 6K ( Th r3 89) 92 06 (1A 5) cy to pla sm ic in te nsi ty 0 – 3 97 0 vs 1-3 0. 44 p-ER K1/ 2 ( Th r2 02 / Ty r2 04 ) 43 70 (D 13. 14 .4E ) pro po rt io n of tum our ce lls w ith n uc le ar st ain in g 0 – 1 00 % 69 ne ga ti ve vs po si ti ve 0. 85 p-S6R P ( Se r2 35 /2 36) 22 11 pe rc en ta ge of tum our ce lls w ith c yt op la sm ic an d m em br an ous st ain in g 0 – 1 00 % 85 0-19 v s 2 0-10 0 0. 55 1al l f ro m C el l S ig na lli ng T ec hn ol og y, D an ve rs , M A , U S2.
Ta bl e S 2. A ss oc ia tio n b et w ee n c lin ic o-pa th ol og ic al f ac to rs a nd d ow ns tr ea m p ro te in s i n l ow v s h ig h p -4 EB P1 a nd n eg at iv e v s p os iti ve p -S 6R P p -4 EB P1 ( N = 4 30 ) p -S 6R P ( N = 4 42 ) ≤20 % >2 0% N eg at iv e ( 161 ) Po si ti ve ( 281 ) N (%) N (%) p-val ue * N (%) N (%) p-val ue * A ge <65 72 (4 2%) 13 0 ( 50 % ) 0. 11 72 (4 5) 13 8 ( 49) 0.4 3 ≥65 99 (5 8%) 12 9 ( 50 % ) 89 ( 55 ) 14 3 ( 51) Ly m ph n od e s ta tu s N ega ti ve 94 (5 5%) 13 8 ( 53 % ) 0.7 7 86 (5 3) 15 1 ( 54 ) 1 Po si ti ve 77 (4 5%) 12 1 ( 47 % ) 75 (4 7) 13 0 ( 46 ) T s ta ge T1-2 14 5 (8 5% ) 23 4 ( 90 %) 0.0 9 139 (8 6) 25 2 (9 0) 0. 35 T3 -4 26 (1 5% ) 25 (1 0% ) 22 (14 ) 29 (1 0) G ra de G ra de 1-2 97 (5 7%) 17 6 ( 68 % ) 0. 019 11 3 ( 70 ) 16 7 ( 59) 0. 02 5 G ra de 3 74 (4 3% ) 83 (3 2% ) 48 ( 30 ) 11 4 ( 41 ) Hi st ol og ic al s ubt yp e D uc tal 13 2 ( 89 % ) 19 3 ( 92 % ) 0. 46 11 9 ( 88 ) 211 (9 2) 0. 2 Lo bula r 16 (11 % ) 17 (8 % ) 17 (12 ) 18 (8 ) HE R 2 s ta tus N ega ti ve 14 9 ( 89 % ) 23 4 ( 93 %) 0. 11 14 8 ( 95 ) 24 3 (8 9) 0.0 5 Po si ti ve 19 (11 % ) 17 (7 % ) 8 ( 5) 30 (11 ) PR s ta tu s N ega ti ve 97 (5 7%) 11 4 ( 44 % ) 0. 01 89 ( 56 ) 12 8 ( 46 ) 0. 047 Po si ti ve 73 (4 3%) 14 4 ( 56 % ) 70 (4 4) 15 1 ( 54 ) PT EN N ega ti ve 51 (3 2% ) 20 (9 %) <0 .000 1 40 ( 29) 34 (13 ) 0. 000 23 Po si ti ve 10 8 ( 68 % ) 212 (9 1% ) 99 (7 1) 22 5 ( 87 ) p-A K T( Th r3 08 ) 0 11 4 ( 68 % ) 12 0 ( 48 % ) <0 .000 1 10 7 ( 71) 13 2 ( 49) <0 .000 1 1-3 53 (3 2% ) 12 9 ( 52% ) 43 ( 29) 14 0 ( 51) p-A K T( Th r4 73 ) 0-1 91 (6 3%) 65 (3 0%) <0 .000 1 73 (6 0) 87 (3 6) <0 .000 1 2-3 54 (3 7%) 15 3 ( 70 % ) 49 (4 0) 15 8 ( 64 ) p-p7 0S 6K 0 91 (5 6%) 77 (3 2% ) <0 .000 1 86 (61 ) 86 (3 2) <0 .000 1 1-3 71 (4 4%) 16 3 ( 68 % ) 55 ( 39) 18 0 ( 68 ) p-S 6R P N ega ti ve 76 (4 6%) 71 (2 9%) 0. 000 8 NA NA Po si ti ve 91 (5 4%) 17 3 ( 71 % ) NA NA p-4EB P1 ≤20 % NA NA 76 (5 2) 91 ( 34 ) 0. 000 8 >2 0% NA NA 71 (4 8) 17 3 ( 66 ) p-ER K1/ 2 N ega ti ve 10 0 ( 62 % ) 72 (2 9%) <0 .000 1 97 (6 7) 79 ( 30 ) <0 .000 1 Po si ti ve 62 (3 8%) 17 4 ( 71 % ) 47 (3 3) 18 6 ( 70 ) * F is he r’s e xa ct t es t b as ed o n c as es w ith ou t m is si ng v al ue s
Table S3A. Univariate Cox proportional hazard model of recurrence free interval (RFI) including heatmap group and treatment interaction, stratified for lymph node status.
Variable HR 95% CI p
Interaction Cluster vs treatment 0.007
Tamoxifen vs CON (ref) In cluster N 0.23 0.12 – 0.44 0.0000081 In cluster A 1.53 0.44 – 5.25 0.50
Table S3B. Multivariate Cox proportional hazard model for prognosis, stratified for lymph node status, including heatmap group in control patients only
Variable HR 95% CI p
Cluster A vs N (ref) CON patients 0.061 0.0079 – 0.47 0.0073
Age ≥65 vs <65 (ref) 0.80 0.27 – 2.4 0.69
T stage T3-4 vs T1-2 (ref) 2.0 0.44 – 9.5 0.37
Grade Grade 3 vs grade 1-2 (ref) 1.3 0.12 – 14 0.82 PR status Positive vs negative (ref) 2.0 0.66 – 6.1 0.22 Histological subtype Lobular vs ductal (ref) 5.3 0.75 – 38 0.095
Ki67 ≥10% vs <10% (ref) 5.9 0.58 – 60 0.13
Mitotic count ≥8 / 2mm² vs <8 / 2mm² (ref) 0.26 0.027 – 2.5 0.24
Table S3C. Multivariate Cox proportional hazard model of recurrence free interval (RFI) including classification tool groups and treatment interaction
Variable HR 95% CI p
Interaction Cluster vs treatment 0.024
Tamoxifen vs CON (ref) In rule FALSE 0.24 0.11 – 0.49 0.00012 In rule TRUE 1.3 0.36 – 4.7 0.69 Rule TRUE vs FALSE (ref) CON patients 0.24 0.066 – 0.91 0.036
TAM patients 1.3 0.64 – 2.8 0.43
Age ≥65 vs <65 (ref) 1.0 0.59 – 1.8 0.89
T stage T3-4 vs T1-2 (ref) 2.0 0.92 – 4.2 0.082
Grade Grade 3 vs grade 1-2 (ref) 1.9 0.81 – 4.6 0.14 PR status Positive vs negative (ref) 1.4 0.79 – 2.5 0.24 Histological subtype Lobular vs ductal (ref) 2.7 0.99 – 7.5 0.053
Ki67 ≥10% vs <10% (ref) 1.2 0.56 – 2.5 0.65
Mitotic count ≥8 / 2mm² vs <8 / 2mm² (ref) 0.72 0.29 – 1.8 0.47
Table S3D. Univariate Cox proportional hazard model of recurrence free interval (RFI) including classification tool groups and treatment interaction
Variable HR 95% CI p
Interaction Cluster vs treatment 0.004
Tamoxifen vs CON (ref) In rule FALSE 0.21 0.11 – 0.41 0.0000050 In rule TRUE 1.6 0.47 – 5.4 0.46
Table S3E. Multivariate Cox proportional hazard model for prognosis including classification tool groups in control patients only
Variable HR 95% CI p
Rule TRUE vs FALSE (ref) CON patients 0.065 0.0091 – 0.47 0.0067
Age ≥65 vs <65 (ref) 0.71 0.23 – 2.2 0.55
T stage T3-4 vs T1-2 (ref) 3.1 0.57 – 17 0.19
Grade Grade 3 vs grade 1-2 (ref) 1.2 0.11 – 13 0.88 PR status Positive vs negative (ref) 1.8 0.60 – 5.3 0.30 Histological subtype Lobular vs ductal (ref) 4.6 0.62 – 33 0.14
Ki67 ≥10% vs <10% (ref) 6.0 0.61 – 58 0.12
Supplementary figures
Figure S1. Canonical phosphatidylinositol 3-kinase (PI3K) (in dark grey) and mitogen-acti-vated protein kinase (MAPK) (in light grey) signalling network, showing examples of cross-talk and feedback loops. Both pathways function downstream of receptor tyrosine kinas-es (RTKs) and G protein-coupled receptors. When PI3K is activated, it triggers a cascade of downstream proteins. PI3K is responsible for the phosphorylation of phosphatidylinositol diphosphate (PIP2) to phosphatidylinositol triphosphate (PIP3), which in turn activates AKT. AKT activates mammalian target of rapamycin complex (mTOR) complex 1 (mTORC1) by in-hibiting the tuberous sclerosis (TSC) complex (TSC1 combined with TSC2) (1). Activation of mTORC1 phosphorylates p70 ribosomal protein S6 kinase (p70S6K) and eukaryotic trans-lation initiation factor eukaryotic initiation factor 4E (eIF4E)-binding protein 1 (4EBP1) (2). When S6K is activated, it can phosphorylate 40S ribosomal protein S6 (S6RP), while phos-phorylation of 4EBP1 leads to its dissociation from eIF4E, both processes initiating RNA translation (3). Phosphatase and tensin homolog (PTEN) is known as a tumour suppressor of the PI3K pathway as it directly dephosphorylates PIP3 to PIP2. When an RTK or G pro-tein-coupled receptor becomes activated, it is also able to activate the MAPK pathway by phosphorylating RAS. Phosphorylated RAS activates RAF which leads to the activation of mitogen-activated and extracellular signal-regulated kinase kinase (MEK)1/2. Activation of MEK 1/2 in turn can phosphorylate extracellular signal-regulated kinase (ERK)1/2 (1). Phosphorylated ERK1/2 has many targets, including p90 ribosomal six kinase-1 (p90RSK1), which leads to cross-talk with the PI3K pathway through mTORC1, S6K, S6RP and eIF4B (4). However, phosphorylated ERK1/2 can also translocate to the nucleus to influence gene ex-pression and regulate various transcription factors by itself (5). Neurofibromatosis type 1 (NF1) deactivates RAS and therefore functions as a tumour suppressor for this pathway (6). Other abbreviations: PDK1: phosphoinositide-dependent protein kinase 1; SGK3: serum- and glucocorticoid-regulated kinase 3;
11
Figure S2. Kaplan-Meier curves for recurrence-free interval according to tamoxifen treatment in patients
with low expression of 4EBP1 in tumour tissue (A), high expression of 4EPB1 (B), no expression of p-S6RP (C) and any expression of p-p-S6RP (D). The interrupted line specifies patients treated with tamoxifen. The continuous line shows patients randomized to the control arm. The light grey line specifies patients with lymph positive disease. The black line shows patients with lymph node-negative disease.
A
B
C
D
Figure S2. Kaplan-Meier curves for recurrence-free interval according to tamoxifen treatment in patients with low expression of p-4EBP1 in tumour tissue (A), high expression of p-4EPB1 (B), no expression of p-S6RP (C) and any expression of p-S6RP (D). The interrupted line spec-ifies patients treated with tamoxifen. The continuous line shows patients randomized to the control arm. The light grey line specifies patients with lymph node-positive disease. The black line shows patients with lymph node-negative disease.
Supplementary references
1. Saini KS, Loi S, de Azambuja E, Metzger-Filho O, Saini ML, Ignatiadis M, et al. Targeting the PI3K/AKT/mTOR and Raf/MEK/ERK pathways in the treatment of breast cancer. Cancer treatment reviews. 2013;39(8):935-46.
2. Dancey J. mTOR signaling and drug development in cancer. Nature reviews Clinical oncology. 2010;7(4):209-19.
3. Li S, Kong Y, Si L, Chi Z, Cui C, Sheng X, et al. Phosphorylation of mTOR and S6RP predicts the efficacy of everolimus in patients with metastatic renal cell carcinoma. BMC cancer. 2014;14:376.
4. McCubrey JA, Steelman LS, Chappell WH, Abrams SL, Montalto G, Cervello M, et al. Mutations and deregulation of Ras/Raf/MEK/ERK and PI3K/PTEN/Akt/mTOR cascades which alter therapy response. Oncotarget. 2012;3(9):954-87.
5. Asati V, Mahapatra DK, Bharti SK. PI3K/Akt/mTOR and Ras/Raf/MEK/ERK signaling pathways inhibitors as anticancer agents: Structural and pharmacological perspectives. European journal of medicinal chemistry. 2016;109:314-41.
6. Kourea HP, Zolota V, Scopa CD. Targeted pathways in breast cancer: molecular and protein markers guiding therapeutic decisions. Current molecular pharmacology. 2014;7(1):4-21.
