This application claims priority from GB2116745.7, filed 19 November 2021, the contents and elements of which are herein incorporated by reference for all purposes.

[1]Field of the Invention

[2]The present invention relates to materials and methods, including biomarkers, for personalising therapy, including escalation and de-escalation strategies, to improve resistance to treatment in breast cancer, particularly patients having ER+/HER2+ breast cancer who are undergoing or will be treated with endocrine therapy, such as with an aromatase inhibitor.

[3]Background to the Invention

[4]Human epidermal growth factor receptor 2 positive (HER2+) breast cancer (BC) has been associated with an aggressive phenotype and poor patient outcome ¹ . However, the introduction of HER2-targeted therapies dramatically changed the prognosis of these patients and the natural history of the disease ^2,3 . Despite the improvement, long- term follow-up data indicate that approximately 15-23% of patients in early stage, still develop recurrent disease ⁴ .

[5]Twenty per cent of all BC overexpresses HER2 and approximately 50% of them are also classified as hormone receptor positive (HR+), which confers substantial differences in biology and clinical outcome from HR-/HER2+ disease. HR+/HER2+ disease is molecularly heterogeneous and around 30% are HER2-Enriched (HER2-E). This subtype is characterized by a high HER2/EGFR pathway activation, increased proliferation and an immune-activated stroma with elevated tumour infiltrating lymphocytes. It has a lower expression of luminal-related genes, than the Luminal A and B subtypes, potentially benefiting greatly from anti-HER2 therapies but poorly from endocrine therapy ^5-8 . [6]Resistance to endocrine therapies have been mainly studied in HR+/HER2- BC and include down regulation of estrogen receptor (ER) expression, altered expression of ER co-regulators, ER mutation and ligand-independent activation of ER and co-activators by growth factor receptor kinases ⁹ ' ¹⁰ . However, those mechanisms might differ between HER2 positive and negative tumours, in part due to the differential distribution of intrinsic subtypes. HER2-targeted therapies might be felt to negate the importance of resistance to Al but while anti-HER2 therapy is given for 1 year to primary BC patients, endocrine therapy is given for at least 5 years. Thus, any residual HER2+ disease after the HER2-targeted therapy remains at risk of an incomplete endocrine response.

[7]Despite advances in the treatment and management of breast cancer, there remains an unmet need for predictive signatures that address the clinical challenge of identifying patients who are likely to benefit from each of the specific therapies. The present invention seeks to meet this unmet need together with certain related advantages.

[8]Brief Description of the Invention

[9]The present inventors consider that the POETIC (PeriOperative Endocrine-Therapy for Individualised Care) trial ¹¹ is one of the best frameworks to study endocrine resistance mechanisms in a large set of ER+/HER2+ BC patients. In the context of POETIC, we hypothesized that resistance mechanisms to endocrine therapy are driven by baseline genomic features. Gene expression profiles at baseline were assessed and the key genomic characteristics were tested for association with response to Al measured by residual levels and changes of Ki67 after two weeks of treatment and clinical outcome. We sought to develop predictive signatures and address the clinical challenge of identifying patients who are likely to benefit from each of the specific therapies. The present inventors found that HER2-E subtype and ERBB2 play a crucial role in ER+/HER2+ BC, driving resistance to endocrine therapy and a higher risk of recurrence. Moreover, new molecular subgroups using signature expression enable the identification of patients at a higher risk of relapse. Altogether, the combination of these biomarkers could be essential for personalising therapy, including escalation and de- escalation strategies, to improve resistance to treatment in early breast cancer.

[10]Accordingly, in a first aspect the present invention provides method for predicting whether a human subject having breast cancer will respond to aromatase inhibitor (Al) therapy, the method comprising: a) measuring the gene expression in a sample obtained from the subject to obtain a sample gene expression profile of the breast tumour of at least the following genes: CHAD, NAT1, SLC39A6, BCL2, IGF1R, ESRI, GRB7 and ERBB2; b) assigning the sample to one of a plurality of predetermined clusters based on the similarity of the sample gene expression profile to the gene expression centroids of said clusters; and c) making a prediction of whether the subject will respond to said Al therapy based on the cluster to which the sample is assigned.

[11]In some embodiments, the plurality of predetermined clusters comprise: a first, poor response (PR), cluster corresponding to samples from breast cancer subjects having < 50% reduction in Ki67 expression from baseline to surgery; a second, intermediate response (IR), cluster corresponding to samples from breast cancer subjects having 50-75% reduction in Ki67 expression from baseline to surgery; and a third, good response (GR), cluster corresponding to samples from breast cancer subjects having > 75% reduction in Ki67 expression from baseline to surgery.

[12]In some embodiments, the plurality of predetermined clusters comprise: a residual Ki67 proliferation rate at 2 weeks after start of aromatase therapy (Ki67 _2wk ) of b 10%; and a residual Ki67 proliferation rate at 2 weeks after start of aromatase therapy (Ki6?2wk) of < 10%.

[13] In some embodiments the sample gene expression profile of the breast tumour comprises at least the genes: CHAD, SLC39A6, NAT1, DNAJC12, TCEAL1, MDM2, NPEPPS, BAG1, IFT140, HDAC5, ARID1A, TLE3, SIGIRR, MAPK3, EGLN2, CDKN1B, CCND1, SYTL4, FAM214A, ADCY9, BCL2, TBC1D9, IGF1R, GATA3, ELOVL2, LAMA3, SCUBE2, ZBTB16, FGFR2, CA12 and ESRI.

In some embodiments, the sample gene expression profile of the breast tumour comprises any set of at least 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650, or at least 700, or essentially at least all of the genes: ABCA8, ACTR3B, ACVR1B, ACVR1C, ACVRL1, ADAM12, ADCY9, ADD1, ADM, AGR2, AGT, AGTR1, AKT3, ALDH1A1, ALDOA, ANGPT1, ANLN, ANXA9, APH1B, APOD, APOE, AR, AREG, ARID1A, ARNT2, ASPM, ASPN, ATAD2, ATM, ATP10B, AURKA, AURKB, AXIN1, AXIN2, B3GNT3, BAD, BAG1, BAIAP2L1, BAIAP3, BAMBI, BAX, BBC3, BBOX1, BCAS1, BCL11A, BCL2, BCL2A1, BCL2L1, BCL6B, BDNF, BIRC5, BLM, BLVRA, BMP2, BMP4, BMP5, BMP6, BMP7, BMP8A, BMPR1A, BMPR1B, BMPR2, BNC2, BNIP3, BORCS7, BRCA1, BRCA2, BTG2, C5orf38, CA12, CACNA1D, CACNA1H, CACNA2D1, CACNA2D3, CACNG1, CACNG4, CACNG6, CALML5, CAMK2B, CAV1, CBLC, CCL2, CCL21, CCL3L1, CCL4, CCL5, CCL7, CCL8, CCNA1, CCNA2, CCNB1, CCND1, CCND2, CCNE1, CCNE2, CCR1, CCR2, CCR5, GDI63, CD19, CD1E, CD24, CD27, CD274, CD276, CD34, CD36, CD44, CD68, CD84, CD8A, CD8B, CDC14A, CDC14B, CDC20, CDC25A, CDC25B, CDC25C, CDC6, CDC7, CDCA5, CDCA7L, CDCA8, CDH1, CDH2, CDH3, CDH5, CDK1, CDK4, CDK6, CDKN1A, CDKN1B, CDKN1C, CDKN2A, CDKN2B, CDKN2C, CDKN2D, CDKN3, CEACAM5, CEACAM6, CENPF, CEP55, CETN2, CFD, CHAD, CHEK2, CHI3L1, CHIT1, CHRNA5, CKB, CKMT1A, CKS1B, CLDN1, CLDN3, CLDN4, CLDN7, CLEC14A, CLEC5A, CMKLR1, CNTFR, COL11A1, COL27A1, COL2A1, COL4A6, COL6A3, COL7A1, COL9A3, COLEC12, COMP, CPA3, CREBBP, CRYAB, CSF3R, CTSW, CXADR, CXCL10, CXCL12, CXCL13, CXCL5, CXCL8, CXCL9, CXCR6, CXorf36, CXXC5, CYBB, CYP4F3, DCN, DDB2, DDR2, DDX39A, DEPDC1, DHRS2, DKK1, DKK2, DLGAP5, DLL1, DLL3, DLL4, DNAJC12, DPT, DSC2, DTX1, DTX3, DTX4, DUSP4, DUSP6, E2F1, E2F5, ECM2, EDN1, EDNRB, EFNA3, EFNA5, EGF, EGFR, EGLN2, EGLN3, EIF2AK3, EIF3B, EIF4E2, EIF4EBP1, ELF3, ELK3, ELOVL2, EMCN, ENO1, ENPP2, EP300, EPAS1, ERBB2, ERBB4, ERCC1, EREG, ESPL1, ESRI, ETV4, ETV7, EXO1, EYA1, EYA2, EYA4, F3, FAM124B, FAM198B, FAM214A, FAM83D, FANCF, FAP, FBN1, FGF1, FGF10, FGF12, FGF13, FGF18, FGF2, FGF7, FGF9, FGFR2, FGFR3, FGFR4, FGL2, FHL1, FLU , FLNC, FLRT3, FLT3, FNBP1, FOS, FOSL1, FOXA1, FOXCI, FOXC2, FOXM1, FOXP3, FREM2, FST, FSTL1, FSTL3, FUT3, FXYD3, FZD10, FZD7, FZD8, FZD9, GABRP, GADD45A, GADD45B, GADD45G, GAS1, GATA3, GATA4, GDF15, GDF5, GGH, GHR, GJB2, GLI3, GNG4, GNLY, GPC4, GPR160, GPX3, GRB2, GRB7, GREM1, GRIA3, GRIN1, GRIN2A, GSK3B, GTF2H2, GZMA, GZMB, GZMH, GZMM, HAPLN1, HAS1, HBB, HDAC1, HDAC10, HDAC11, HDAC2, HDAC5, HDAC6, HDC, HEG1, HELLS, HEMK1, HES1, HGF, HIF1A, HIST1H1C, HIST1H2BH, HIST1H3H, HIST3H2BB, HK2, HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DMB, HLA-DOB, HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA- DQB1, HLA-DRA, HLA-DRB1, HLA-E, HMGA1, HNF1A, HOXA5, HOXA7, HOXA9, HOXB13, HOXB3, HSPA2, IBSP, ICAM1, ID1, ID2, ID4, IDO1, IFT140, IGF1, IGF1R, IKZF3, IL10RA, IL11RA, IL12RB2, IL13RA1, IL1B, IL1R2, IL1RN, IL20RA, IL20RB, IL22RA2, IL24, IL2RA, IL2RB, IL3RA, IL4R, IL6, IL6R, IL7R, INHBA, INHBB, IRF6, IRX1, ISG15, ISM1, ITGA6, ITGAV, ITGB1, ITGB3, ITGB6, ITPR1, JAG1, JAG2, JAK1, JAK2, JAK3, JAM2, JCAD, JUN, KAT2B, KCNB1, KDR, KIAA0040, KIF11, KIF14, KIF23, KIF2C, KIFC1, KIT, KLRK1, KRT14, KRT17, KRT5, KRT6B, KRT7, LAD1, LAG3, LAMA3, LAMB3, LAMC2, LEF1, LEFTY2, LEMD1, LEP, LEPR, LFNG, LIF, LIFR, LINC02381, LPL, LRP2, LRRC32, LTB, LTBP1, MAD2L1, MAE, MAML2, MAP2K4, MAP3K12, MAPK1, MAPK10, MAPK3, MAPK8IP2, MAPT, MARCO, MCM2, MCM3, MDM2, MED1, MELK, MEOX2, MET, MFNG, MIA, MIS18A, MKI67, MLH1, MLLT3, MLPH, MME, MMP11, MMP14, MMP3, MMP7, MMP9, MMRN2, MRE11, MS4A2, MSR1, MT1G, MTOR, MUC1, MUS81, MYBL2, MYC, MYCN, MYCT1, NASP, NAT1, NCAM1, NCAPH2, NDC80, NDP, NEIL1, NEIL2, NEIL3, NEO1, NETO2, NFATC1, NFKBIZ, NGFR, NKG7, NOD2, NOTCH1, NOTCH2, NOTCH3, NPEPPS, NPR1, NR4A1, NR4A3, NRCAM, NRXN1, NRXN3, NSD1, NSD3, NTRK2, NUDT1, NUF2, NUMBL, NUPR1, OAS3, OCLN, OGN, OLFML2B, ORC6, PALB2, PALMD, PARP1, PARP2, PARP4, PAX5, PAX8, PBX3, PCK1, PCNA, PDCD1, PDCD1LG2, PDE9A, PDGFB, PDGFRA, PDGFRB, PDK4, PECAM1, PFDN2, PGK1, PGR, PHGDH, PIK3CA, PIK3CD, PIK3CG, PIK3R1, PIK3R2, PIK3R3, PIK3R5, PIM1, PIP, PKMYT1, PLA1A, PLA2G2A, PLA2G3, PLA2G4A, PLA2G4F, PLAT, PLCB1, PLCB4, PLCE1, PLD1, PMS2, POLDI, POLQ, POPDC3, PPARG, PPARGC1A, PPP2CB, PPP2R1A, PPP2R2C, PRC1, PREP, PRF1, PRKAA2, PRKACA, PRKACB, PRKCA, PRKCB, PRKDC, PRKX, PRLR, PROMI, PSAT1, PSMB10, PSMB7, PSMB9, PTCHI, PTEN, PTGDS, PTGER3, PTGS2, PTTG1, PYCARD, RAC2, RAC3, RAD51, RAD51C, RAD52, RAD54L, RARRES3, RASAL1, RASGRF1, RASGRF2, RASGRP1, RBI, RBL1, RBL2, RBX1, RELN, RFC4, RNASE2, RNF103, ROBO4, ROCK1, ROCK2, RORA, RORB, RPS6KA5, RPS6KB1, RPS6KB2, RRM2, RUNX3, S100A14, S100A7, S1PR1, SCARA5, SCUBE2, SELE, SERBP1, SERPINB5, SERPINH1, SFN, SFRP1, SFRP2, SFRP4, SHC2, SHC4, SHE, SHMT2, SIDT1, SIGIRR, SIX1, SKA3, SKP1, SKP2, SLC2A1, SLC39A6, SLC44A4, SLPI, SMAD1, SMAD3, SMAD4, SMAD5, SMC1B, SMO, SMURF2, SNAI1, SNAI2, SOCS1, SOCS2, SOCS3, SOXIO, SOX17, SOX2, SOX9, SP1, SPC25, SPDEF, SPN, SPP1, SPRY1, SPRY2, SPRY4, SRPX, ST6GALNAC2, STAT1, STC1, SUV39H2, SYTL4, TAPI, TAP2, TAPBP, TBC1D9, TBX1, TCEAL1, TCF4, TCF7L1, TEK, TFDP1, TFF1, TFF3, TGFB1, TGFB2, TGFB3, TGFBR2, THBS1, THBS2, THBS4, THY1, TIE1, TIGIT, TIMP4, TLE3, TLR4, TLX1, TMEM45B, TMPRSS2, TMPRSS4, TNF, TNFAIP6, TNFSF10, TNKS, TNKS2, TNN, TOP2A, TP53, TPSAB1, TRIP13, TSPAN1, TSPAN7, TTK, TTYH1, TUBA4A, TWIST1, TWIST2, TYK2, TYMP, TYMS, UBE2C, UBE2T, VCAN, VEGFA, VEGFD, VIM, VIT, WDR77, WEE1, WIFI, WNT10A, WNT11, WNT2, WNT4, WNT5A, WNT5B, WNT6, WNT7B, WRN, WT1, XRCC2, XRCC3, ZBTB16, ZEB1, ZEB2, ZFPM2, ZFYVE9, ZIC2, and ZNF205.

[14] In some embodiments, the plurality of predetermined clusters comprise: single gene expression cluster 1 ("GC1") having high expression of immune and chemokine related genes; single gene expression cluster 2 ("GC2") having high expression of extracellular matrix organization (ECM) related genes and high expression of ERBB2; single gene expression cluster 3 ("GC3") having low gene expression with upregulation of gene involved in DNA-damage repair, including RAD51, BRCA, CCNE1, TRIP13, CDCA5, RFC4, KIF14 and BLM; single gene expression cluster 4 ("GC4") having high expression of endocrine receptor (ER) signalling related genes; and single gene expression cluster 5 ("GC5") having high expression of endocrine receptor (ER) signalling related genes and high expression of genes involved in MAPK/PI3K and RAS signalling. [15]In some embodiments, said predetermined clusters GC1 to GC5 are obtainable by carrying out consensus clustering by expression of the single genes of the NanoString BC360 panel on a population of samples comprising Al-treated and Al-untreated subjects having breast cancer, optionally wherein said population of samples comprise at least 100, 200 or at least 300 patients from the POETIC trial.

[16]In some embodiments said gene expression centroids of said clusters comprise the GC1, GC2, GC3, GC4 and GC5 centroids set forth in Table 4.

[17]In some embodiments, the sample is assigned to GC1, GC2 or GC3, and the subject is therefore predicted to respond poorly to Al therapy.

[18]In some embodiments, the sample is assigned to GC4 or GC5, and the subject is therefore predicted to respond well to Al therapy.

[19]In some embodiments said plurality of predetermined clusters comprise: signature expression cluster 1 ("SGC1") having high expression of immune features and low expression of ER signalling features; signature expression cluster 2 ("SGC2") having low expression of immune features and high expression of ERBB2; signature expression cluster 3 ("SGC3") having high expression of ESRI and low expression of PgR; signature expression cluster 4 ("SGC4") having high endocrine signalling expression and low ERBB2 expression.

[20]In some embodiments the predetermined clusters SGC1 to SGC4 are obtainable by carrying out consensus clustering by expression of the signatures of the NanoString BC360 panel on a population of samples comprising Al-treated and Al-untreated subjects having breast cancer, optionally wherein said population of samples comprise at least 100, 200 or at least 300 patients from the POETIC trial. [21]In some embodiments, the gene expression centroids of said clusters comprise the SGC1, SGC2, SGC3 and SGC4 centroids set forth in Table 5 and/or Table 6.

[22]In some embodiments, the sample is assigned to SGC2 or SGC4 and the subject is therefore predicted to respond poorly to Al therapy.

[23]In some embodiments, the sample is assigned to SGC1 or SGC3 and wherein the subject is therefore predicted to respond well to Al therapy.

[24]In some embodiments, the method further comprises providing a prognosis, e.g. of predicted time-to-recurrence (TTR), of the subject based on the cluster to which the sample is assigned.

[25]In particular, the predicted time-to-recurrence of the subject may be based on the combination of features as set forth in any one of models 3 to 8, as shown in Supplementary Table 4.

[26]In some embodiments, the method further comprises measuring the gene expression in the sample of one or more housekeeping genes.

[27]In particular, the housekeeping genes may comprise at least 2, 3, 4, 5, 6, 7, or at least 8 housekeeping genes selected from the group consisting of: ACTB, MRPL19, PSMC4, RPLPO, SF3A1, GUSB (alias GUS), PUM1 and TFRC.

[28]In some embodiments the subject is predicted to respond to said Al therapy, and the method further comprises administering, or recommending for administration, a therapeutically effective amount of an aromatase inhibitor to the subject, optionally wherein the aromatase inhibitor is selected from anastrozole and letrozole.

[29]In some embodiments, the gene expression level of one or more of said genes is measured using NanoString nCounter Analysis or RNA- seq analysis. [30]In some embodiments, the gene expression level of one or more of said genes is measured using RT-PCR, and wherein the RT-PCR gene expression measurements are adjusted to NanoString equivalent values by applying gene-wise linear conversion factors.

[31]In some embodiments, the gene-wise linear conversion factor for each gene has been determined by linear regression analysis of gene expression measurements made of the same sample by NanoString and

RT-PCR.

[32]In some embodiments, the gene expression measurements are normalised by reference to the expression of one or more housekeeping genes.

[33]In some embodiments, the subject (i) has ER+ and HER2+ breast cancer; and/or (ii) is a postmenopausal woman.

[34]In some embodiments, the subject has been treated with, is undergoing treatment with, or is planned to have treatment with, endocrine therapy, optionally wherein endocrine therapy comprises treatment with an aromatase inhibitor.

[35]In particular, the aromatase inhibitor may be selected from: anastrozole and letrozole.

[36]In some embodiments, the sample has been obtained from the subject prior to and/or following the commencement of treatment with an aromatase inhibitor, optionally wherein the aromatase inhibitor is selected from: anastrozole and letrozole. In some embodiments, the method of the present invention may further comprise assessment of a change or perturbation of a gene expression signature posttreatment as compared with pre-treatment to identify a change in gene expression (e.g. Al therapy-induced upregulation or downregulation of one of more genes). In some cases, a change in gene expression may indicate treatment response. In particular, upregulation of tumour-related immunity signatures such as TIGIT, CDS T-cells, inflammatory chemokines and/or IDO1, and/or downregulation of proliferation, HRD, TP53 and/or ER-signaling may indicate treatment response, including a desired response to the Al therapy.

[37]In some embodiments, the subject has had surgical removal of a breast tumour.

[38]In some embodiments, the method further comprises the step of administering a therapeutically effective amount of one or more agents selected from: a CDK inhibitor, optionally a CDK4/6 inhibitor, further optionally wherein the CDK4/6 inhibitor is selected from: palbociclib, abemaciclib and ribociclib; a HER2-targetting agent, optionally trastuzumab, trastuzumab- DM1 (T-DM1) and/or pertuzumab; and an endocrine therapy, optionally an aromatase inhibitor, further optionally wherein the aromatase inhibitor is selected from anastrozole and letrozole.

[39]In a second aspect, the present invention provides a computer- implemented method for predicting whether a human subject having breast cancer will respond to aromatase inhibitor (Al) therapy, the method comprising: a) obtaining gene expression data representing the gene expression profile of one or more samples obtained from the breast tumour of the subject of at least the following genes: CHAD, NAT1, SLC39A6, BCL2, IGF1R, ESRI, GRB7 and ERBB2; b) comparing the gene expression data obtained in a) with a plurality of gene expression centroids that define a plurality of predetermined clusters; c) assigning the sample to one of the plurality of predetermined clusters based on the similarity of the sample gene expression data to the gene expression centroids of said clusters; and d) making a prediction of whether the subject will respond to said Al therapy based on the cluster to which the sample is assigned. [40]In some embodiments, the respective genes, clusters and centroids are as defined in connection with the first aspect of the invention.

[41]In some embodiments, the gene expression data representing the gene expression profile is of one or more samples obtained from the subject prior to and/or following the commencement of treatment with an aromatase inhibitor, optionally wherein the aromatase inhibitor is selected from: anastrozole and letrozole.

[42]In a third aspect, the present invention provides a system for predicting whether a human subject having breast cancer will respond to aromatase inhibitor (Al) therapy, the system comprising:

A) a plurality of oligonucleotide probes for detection of gene transcripts of at least the following genes: CHAD, NAT1, SLC39A6, BCL2, IGF1R, ESRI, GRB7 and ERBB2;

B) a computer having at least one processor and at least one non-transitory computer readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:

(a) receiving gene expression data representing the gene expression profile of one or more samples obtained from the breast tumour of the subject of at least the following genes: CHAD, NAT1, SLC39A6, BCL2, IGF1R, ESRI, GRB7 and ERBB2;

(b) comparing the gene expression data obtained in a) with a plurality of gene expression centroids that define a plurality of predetermined clusters;

(c) assigning the sample to one of the plurality of predetermined clusters based on the similarity of the sample gene expression data to the gene expression centroids of said clusters; and

(d) making a prediction of whether the subject will respond to said Al therapy based on the cluster to which the sample is assigned.

[43]In some embodiments, the plurality of probes comprise NanoString nCounter probes. [44]In some embodiments, the gene expression data representing the gene expression profile is of one or more samples obtained from the subject prior to and/or following the commencement of treatment with an aromatase inhibitor, optionally wherein the aromatase inhibitor is selected from: anastrozole and letrozole.

[45]In some embodiments, the system of the third aspect of the invention is for use in the method of the first or second aspects of the invention.

[46]In a fourth aspect, the present invention provides an aromatase inhibitor for use in a method of treatment of breast cancer in a human subject, wherein the method of treatment comprises carrying out the method of the first aspect of the invention on a sample obtained from the subject and wherein the subject is predicted by the method of the first aspect of the invention to respond to said aromatase inhibitor therapy. Patients identified as likely to benefit from Al therapy constitute a novel patient subpopulation who can be expected to derive greatest benefit from such treatment.

[47]In some embodiments, the aromatase inhibitor is selected from: anastrozole and letrozole.

[48]In some embodiments in accordance with any aspect of the present invention, the gene expression level of one or more of said genes is measured using NanoString nCounter Analysis. In some embodiments the gene expression level of said genes is measured by measuring tumour derived RNA in a biological sample, e.g. a plasma or blood sample. Such non-invasive techniques may be preferred in certain clinical situations.

[49]In some embodiments in accordance with any aspect of the present invention, the gene expression level of one or more of said genes may be measured using a technique other than NanoString (e.g. RT-PCR) and then adjusted to NanoString equivalent values by applying gene-wise linear conversion factors. In particular embodiments the gene-wise linear conversion factor for each gene may be determined by linear regression analysis of gene expression measurements made of the same sample by NanoString and the alternative measurement method (e.g. RT-PCR).

[50]In some embodiments the gene expression measurements are normalised by reference to the expression of one or more housekeeping genes. Housekeeping genes are determined by selecting genes that minimize the pairwise variation statistic from a large dataset of ER+ postmenopausal patients.

[51]In some embodiments the subject is female, e.g. a postmenopausal woman.

[52]In some embodiments said breast cancer is ER+/HER2+ breast cancer.

[53]In some embodiments, the subject has been treated with, is undergoing treatment with, or is planned to have treatment with, endocrine therapy, particularly treatment with an aromatase inhibitor (e.g. anastrozole or letrozole).

[54]In some embodiments the sample has been obtained from the subject prior to or at the time of surgical treatment (e.g. breast tumour excision). In some embodiments the sample has been obtained from the subject prior to commencing treatment with an aromatase inhibitor (e.g. anastrozole or letrozole).

[55]In some embodiments, the breast tumour of the subject exhibits a marker of proliferation Ki-67 (MKI67) score of 8% or greater, meaning 8% or more tumour cells are positive for Ki-67 expression. As used herein Ki67 _B means the Ki67 measurement at baseline; Ki67 _2wk means the Ki67 measurement after 2 weeks of aromatase inhibitor treatment.

[56]In some embodiments the subject is predicted to be sensitive to said Al therapy, and wherein the method further comprises the step of administering, or recommending administration of, a therapeutically effective amount of an aromatase inhibitor, optionally anastrozole or letrozole. [57]In some embodiments the method comprises concurrent, sequential or separate administration of:

(i) trastuzumab-DMl (T-DM1) and/or pertuzumab; and

(ii) Al therapy, such as anastrozole and/or letrozole, to the subject in therapeutically effective amounts. In particular, tumours with higher levels of ERBB2 and lower associated immunity had a significantly higher risk of relapse, indicating a potential benefit from an intensified anti-HER2 treatment, using for example double anti-HER2 blockade or adjuvant TDM1.

[58]In some embodiments the method comprises concurrent, sequential or separate administration of:

(i) A CDK4/6 inhibitor such as palbociclib, abemaciclib or ribociclib; and

(ii) Al therapy, such as anastrozole and/or letrozole, to the subject in therapeutically effective amounts. Patients with tumours characterized by higher levels of ER-signaling and the lowest levels of ERBB2 also showed significant worse outcome, highlighting the need of additional treatments in this subgroup to enhance the effect of Al treatment (i.e. CDK4/6 inhibitors in combination with Al).

[59]In some embodiments the subject is predicted to be resistant to said Al therapy, and wherein the method further comprises administering more intensive HER2-targeted therapy (e.g. double anti-HER2 blockade or adjuvant TDM1) and/or CDK4/6 inhibitors (e.g. palbociclib, abemaciclib or ribociclib) to the subject in the absence of endocrine therapy, such as Al therapy. In this way subjects who are unlikely to benefit from a particular therapy may be spared such therapy and any related unwanted side effects.

[60]The present invention includes the combination of the aspects and preferred features described except where such a combination is clearly impermissible or is stated to be expressly avoided. These and further aspects and embodiments of the invention are described in further detail below and with reference to the accompanying examples and figures.

[61]Brief Description of the Figures

[62]Embodiments and experiments illustrating the principles of the invention will now be discussed with reference to the accompanying figures in which:

[63]Supplementary Figure 1. Consort Diagram

[64]Supplementary Table 1. Breast Cancer 360 Biological signatures

[65]Supplementary Table 2. Demography of the study population

[66]Table 1. Ki67 response categories by intrinsic subtype

[67]Figure 1. Spaghetti Plot Ki67 B/S

[68]Supplementary Table 3: Signature association with Ki67 categories

[69]Figure 2. Barplots showing the significant signatures associated with Ki67 categories.

[70]Supplementary Figure 2: Scatterplots of the association of signature expression with Ki67 categories by subtypes (HER2-E vs Luminals)

[71]Figure 3. SAM analysis by A. Multiclass for Ki67 response categories, B. Unpaired for Ki67 response categories, C. Unsupervised of the unpaired two classes for Ki672 weeks.

[72]Supplementary figure 3. ESRI gene expression levels amongst different levels of ERBB2 gene expression by PAM 50 subtype at baseline

[73]Table 2A. Table showing the coincidence/overlapped patients in each cluster. [74]Table 2B. Association analysis of the different clusters based on single gene expression and signature expression and response to Al by different endpoints.

[75]Figure 4. Consensus clustering by single gene expression at baseline. A. Supervised heatmap for the 5 new molecular subgroups from consensus clustering based on single gene expression at baseline. B. ERBB2 expression levels in the different single genebased clusters in the entire cohort and in. C. in HER2E and Luminals separately.

[76]Figure 5. Consensus clustering by signature expression at baseline. A. Supervised heatmap for the 4 new molecular subgroups from consensus clustering based on single gene expression at baseline. B. ERBB2 expression levels in the different signaturebased clusters in the entire cohort and in. C. in HER2E and Luminals separately.

[77]Figure 6. Survival Analysis for TTR. A. Kaplan Meier curves from the univariate analysis of the subtypes at baseline including all and HER2-E vs Luminals only. B. Kaplan Meier curves from the univariate analysis of the new molecular clusters based on C. single gene expression at baseline and D. signature expression.

[78]Supplementary table 4. Series of multivariable cox regression models for TTR for selected prognosis factors and new genomic-based prognostic variables adjusted separately by the basic clinic- pathological factors (age, nodal status, post tumour size, post tumor grade).

[79]Supplementary figure 4. Barplots comparing the multivariate chi square likelihood test for each model.

[80]Supplementary table 5. Series of multivariable analysis for TTR of baseline signature expression adjusted by the basic clinic- pathological factors (age, nodal status, post tumour size, post tumor grade).

[81]Supplementary figure 5. Spearman correlation tests of the significant signatures by nominal p-value in the series of multivariate analysis using each of the signatures in the baseline clinico-pathological model.

[82]Table 3. The list of genes making up the NanoString BC360 panel.

[83]Table 4. The centroids for each of the single gene clusters GC1, GC2, GC3, GC4 and GC5.

[84]Table 5. The centroids for each of the signature gene clusters SGC1, SGC2, SGC3 and SGC4.

[85]Table 6. The centroids for each of the signature gene clusters SGC1, SGC2, SGC3 and SGC4. Different from Table 5, in Table 6 the average gene expression values of the individual genes making up the multi-gene signatures of the signature gene clusters are broken out and shown in groups. For example, the genes of the DNA damage signature are shown in group 3.

[86]Figure 7. Differences of subtype classification from baseline to surgery in a. treated and b. controls. These figures show PAM50 intrinsic subtypes at baseline and surgery time points in both treatment arms separately.

[87]Figure 8. Supervised consensus clustering showing significant gene expression changes from baseline to surgery by multi-comparison t-test of log2FC gene expression (Surgery-Baseline) in a. treated and b. controls.

[88]Figure 9. a. Overall gene expression adjusted log2FC of DEGs in treated arm. Colours represent change directions (upregulated or downregulated in Surgery) of the significant DEGs. b. Gene expression adjusted log2FC of DEGs in Treated GR. Colours represent change directions (upregulated or downregulated in Surgery) of the significant DEGs. c. Gene expression adjusted log2FC of DEGs in Treated PR. Colours represent change directions (upregulated or downregulated in Surgery) of the significant DEGs. d. Gene expression adjusted log2FC in Treated GR and PR. Colours represent significant DEGs in GR and PR. [89]Figure 10. Consensus clustering of signature gene expression changes between surgery and baseline signatures in a. treated patients and b. controls.

[90]Figure 11. Kaplan Meier curves for TTR according to a. surgery subtypes (4 classes) and b. to subtype change (HL vs nonHL) in Treated arm in the overall population.

[91]Table 7. Gene expression adjusted log2FC in good responders (GR) and poor responders (PR).

[92]Figure 12. Heatmap showing changes in gene expression signatures of molecular clusters 1 to 5 following commencement of endocrine therapy. Red indicates increased gene expression whereas blue indicates reduced expression following endocrine therapy.

[93]Supplementary table 6. Definition of intrinsic subtype change categories.

[94]Supplementary table 7. Demography of the study population. Abbreviations: n: number; G: Grade; ILC: Invasive lobular carcinoma; IDG: Invasive ductal carcinoma; ER, Estrogen receptor; PgR, Progesterone receptor, ET, Endocrine Therapy.

[95]Supplementary figure 6. Differences of Ki67 according to intrinsic subtype, a. Ki67 measured by IHC at baseline and surgery within each of the intrinsic subtypes and coloured by the intrinsic subtype at baseline and at surgery in the Treated arm. b. and in the Control arm. c. Ki67 IHC at baseline and surgery in the different PAM 50 intrinsic subtype change categories in the treated arm. d. Ki67 IHC at baseline and surgery timepoints by subtype change category in control arm.

[96]Supplementary table 8. Intrinsic subtypes statistics at baseline and surgery time points in Control arm, where different cut-offs were applied to the difference between the 1 ^st and 2 ^nd highest centroids correlation scores for baseline subtype and surgery subtype. [97]Supplementary figure 7. Baseline and Surgery PAM50 intrinsic subtype centroids correlation coefficient scores and proliferation score in control arm (patients sorted by surgery subtype and PAM50 intrinsic subtype score reduction) in a. treated and b. control arm.

[98]Supplementary table 9. Association analysis of the subtype change and Ki67 change in Treated arm by Chi-square test.

[99]Supplementary figure 8. a. Supervised consensus clustering of single gene expression in Treated arm at baseline and at surgery, b. Heatmap of DEGs gene expression in Control arm.

[100]Supplementary figure 9. Signature change in Treated and Control arms. Colours represent the significances of paired T-test in Treated and/or Control arms.

[101]Supplementary table 10. Multivariable cox regression models for surgery subtype or subtype change vs TTR adjusted by standard clinicopathological variables. Abbreviations: HR, Hazard Ratio; CI: Confidence Interval; N, nodal; HER2-E, HER2-enriched; nonHL: subtype change not from high-risk to low-risk.

[102]Supplementary table 11. Series of multivariable Cox regression analysis for TTR of surgery signature adjusted by the basic clinicopathological factors (age, nodal status, post tumour size, post tumour grade).

[103]Supplementary figure 10. Kaplan Meier curves for TTR according to a. surgery intrinsic subtypes and b. intrinsic subtype risk changes of patients who did not receive adjuvant trastuzumab treatment and those who received Trastuzumab treatment.

[104]Detailed description of the invention

[105]Aspects and embodiments of the present invention will now be discussed with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference. [106]In describing the present invention, the following terms will be employed, and are intended to be defined as indicated below.

[107]Samples

[108]A "test sample" as used herein may be a cell or tissue sample (e.g. a biopsy), a biological fluid, an extract (e.g. a protein or DNA extract obtained from the subject). In particular, the sample may be a tumour sample, including a breast tumour (primary or secondary). The sample will generally be comprise nucleic acid (e.g. RNA or DNA) and/or protein. In some cases the sample may be a blood or plasma sample containing tumour-derived RNA. Measurement of gene expression may involving quantification of RNA from a sample, including a blood or plasma sample. The sample may be one which has been freshly obtained from the subject or may be one which has been processed and/or stored prior to making a determination (e.g. frozen, fixed or subjected to one or more purification, enrichment or extractions steps). In some embodiments, the sample is a fixed tumour tissue sample (such as e.g. a formalin-fixed paraffin- embedded (FFPE) tissue sample), or a frozen tumour tissue sample (such as e.g. a fresh frozen (FF) tissue sample). The preferred sample type according to the present invention is a FFPE tissue sample, as this type of samples is widely available. Indeed, FFPE tissue samples are commonly obtained in clinical settings, for example for histopathological diagnosis. Reference to "cancer cells" herein may refer to cancer cells present in a cell or tissue sample, such as e.g. cells in a tumour tissue from a biopsy.

[109] "and/or" where used herein is to be taken as specific disclosure of each of the two specified features or components with or without the other. For example "A and/or B" is to be taken as specific disclosure of each of (i) A, (ii) B and (iii) A and B, just as if each is set out individually herein.

[110]Gene expression

[111]Reference to determining the expression level refers to determination of the expression level of an expression product of the gene. Expression level may be determined at the nucleic acid level or the protein level. Within the context of the present invention, expression levels of genes of interest are preferably determined at the nucleic acid level, and in particular at the mRNA level.

[112]The gene expression levels determined may be considered to provide an expression profile. By "expression profile" what is meant is a set of data relating to the level of expression of one or more of the relevant genes in an individual, in a form which allows comparison with comparable expression profiles (e.g. from individuals for whom the prognosis is already known), in order to assist in the determination of prognosis and in the selection of suitable treatment for the individual patient.

[113]The determination of gene expression levels may involve determining the presence or amount of mRNA in a sample of cancer cells or a sample containing material derived from cancer cells (e.g. a blood, plasma, urine or other biological liquid comprising tumour-derived nucleic acids, such as circulating tumour RNA).

Gene expression levels may be determined in a sample of cancer cells using any conventional method, for example using nucleic acid microarrays or using nucleic acid synthesis (such as quantitative PCR). In some embodiments gene expression may be measured using a gene expression profiling technology and/or an RNA-transcript quantitating technology. For example, gene expression levels may be determined using a NanoString nCounter Analysis system (see, e.g., US7,473,767). In some cases, a blood sample may be analysed to measure tumour derived RNA in order to quantify gene expression of the genes of the clusters of the present invention (see, e.g., Xue et al., 2019, Nature Scientific Reports (2019) 9:12943 | https://doi.org/10.1038/s41598-019-49445-x, describing measurement of tumour gene expression by RNA sequencing of patient blood or plasma samples).

[114]Importantly, the order in which different genes making up the clusters and/or individual genes thereof are analysed to determine gene expression is not particularly limited. It is possible that gene expression for all genes of interest may be determined from a sample in parallel such as in a single assay or as multiple assays on the same day. However, it is specifically contemplated that the gene expression of any given gene may be determined separately from determination of one or more other genes. In particular, gene expression of the gene or genes as defined herein may be determined separately from other of the genes herein, such as being determined on different days, by different labs, and/or using different techniques.

[115]Gene expression measurements in accordance with the method of the present invention (e.g. the single gene expression clusters or the signature gene expression clusters) may be combined with other known predictive gene signatures, such as those having clinical relevance. In one particular embodiment contemplated herein, the genes as defined herein may be combined with selected prognostic factors and other variables as set forth in any one of Models 1-8, as depicted in Supplementary Table 4.

[116]Alternatively or additionally, the determination of gene expression levels may involve determining the protein levels expressed from the genes in a sample containing cancer cells obtained from an individual. Protein expression levels may be determined by any available means, including using immunological assays. For example, expression levels may be determined by immunohistochemistry (IHC), Western blotting, ELISA, immunoelectrophoresis, immunoprecipitation and immunostaining. Using any of these methods it is possible to determine the relative expression levels of the proteins expressed from the genes listed herein.

[117]Gene expression levels may be compared with the expression levels of the same genes in cancers from a group of patients whose survival time and/or treatment response is known. The patients to which the comparison is made may be referred to as the 'control group'. Accordingly, the determined gene expression levels may be compared to the expression levels in a control group of individuals having cancer. The comparison may be made to expression levels determined in cancer cells of the control group. The comparison may be made to expression levels determined in samples of cancer cells from the control group. The cancer in the control group may be the same type of cancer as in the individual. For example, if the expression is being determined for an individual with breast cancer, the expression levels may be compared to the expression levels in the cancer cells of patients also having breast cancer.

[118]Other factors may also be matched between the control group and the individual and cancer being tested. For example the stage of cancer may be the same, the subject and control group may be age- matched and/or gender matched.

[119]Additionally, the control group may have been treated with the same form of surgery and/or same therapeutic agent(s).

[120]Accordingly, an individual may be stratified or grouped according to their similarity of gene expression with the group previously identified as resistant to or sensitive to Al therapy.

[121]Methods for classification based on gene expression

[122]In some embodiments, the present invention provides methods for classifying or monitoring breast cancer in subjects. In particular, data obtained from analysis of gene expression may be evaluated using one or more pattern recognition algorithms. Such analysis methods may be used to form a predictive model, which can be used to classify test data. For example, one convenient and particularly effective method of classification employs multivariate statistical analysis modelling, first to form a model (a "predictive mathematical model") using data ("modelling data") from samples of known subgroup (e.g., from subjects known to have a particular breast cancer Al therapy response), and second to classify an unknown sample (e.g., "test sample") according to subgroup (likely responder or likely non-responder).

[123]Pattern recognition methods have been used widely to characterize many different types of problems ranging, for example, over linguistics, fingerprinting, chemistry and psychology. In the context of the methods described herein, pattern recognition is the use of multivariate statistics, both parametric and non-parametric, to analyse data, and hence to classify samples and to predict the value of some dependent variable based on a range of observed measurements. There are two main approaches. One set of methods is termed "unsupervised" and these simply reduce data complexity in a rational way and also produce display plots which can be interpreted by the human eye. However, this type of approach may not be suitable for developing a clinical assay that can be used to classify samples derived from subjects independent of the initial sample population used to train the prediction algorithm.

[124]The other approach is termed "supervised" whereby a training set of samples with known class or outcome is used to produce a mathematical model which is then evaluated with independent validation data sets. Here, a "training set" of gene expression data is used to construct a statistical model that predicts correctly the "subgroup" of each sample. This training set is then tested with independent data (referred to as a test or validation set) to determine the robustness of the computer-based model. These models are sometimes termed "expert systems", but may be based on a range of different mathematical procedures such as support vector machine, decision trees, k-nearest neighbour and naive Bayes. Supervised methods can use a data set with reduced dimensionality (for example, the first few principal components), but typically use unreduced data, with all dimensionality. In all cases the methods allow the quantitative description of the multivariate boundaries that characterize and separate each subtype in terms of its intrinsic gene expression profile. It is also possible to obtain confidence limits on any predictions, for example, a level of probability to be placed on the goodness of fit. The robustness of the predictive models can also be checked using cross-validation, by leaving out selected samples from the analysis.

[125]After stratifying the training samples according to subtype, a centroid-based prediction algorithm may be used to construct centroids based on the expression profile of the gene sets described herein, e.g. the genes of the NanoString BC360 panel or a compact signature as described herein. [126] "Translation" of the descriptor coordinate axes can be useful. Examples of such translation include normalization and meancentring. "Normalization" may be used to remove sample-to-sample variation. Some commonly used methods for calculating normalization factor include: (i) global normalization that uses all genes on the microarray or nanostring codeset; (ii) housekeeping genes normalization that uses constantly expressed housekeeping/invariant genes; and (iii) internal controls normalization that uses known amount of exogenous control genes added during hybridization (Quackenbush, 2002). In one embodiment, the genes as defined herein can be normalized to one or more control housekeeping genes. Exemplary housekeeping genes include ACTB, MRPL19, PSMC4, RPLPO, SF3A1, GUSB (alias GUS), PUM1 and TFR. It will be understood by one of skill in the art that the methods disclosed herein are not bound by normalization to any particular housekeeping genes, and that any suitable housekeeping gene(s) known in the art can be used. Many normalization approaches are possible, and they can often be applied at any of several points in the analysis. In one embodiment, microarray data is normalized using the LOWESS method, which is a global locally weighted scatterplot smoothing normalization function. In another embodiment, qPCR and NanoString nCounter analysis data is normalized to the geometric mean of a set of multiple housekeeping genes. Moreover, qPCR can be analysed using the fold-change method.

[127] "Mean-centering" may also be used to simplify interpretation for data visualisation and computation. Usually, for each descriptor, the average value of that descriptor for all samples is subtracted. In this way, the mean of a descriptor coincides with the origin, and all descriptors are "centered" at zero. In "unit variance scaling," data can be scaled to equal variance. Usually, the value of each descriptor is scaled by 1/StDev, where StDev is the standard deviation for that descriptor for all samples. "Pareto scaling" is, in some sense, intermediate between mean centring and unit variance scaling. In pareto scaling, the value of each descriptor is scaled by 1/sqrt(StDev), where StDev is the standard deviation for that descriptor for all samples. In this way, each descriptor has a variance numerically equal to its initial standard deviation. The pareto scaling may be performed, for example, on raw data or mean-centred data.

[128] "Logarithmic scaling" may be used to assist interpretation when data have a positive skew and/or when data spans a large range, e.g., several orders of magnitude. Usually, for each descriptor, the value is replaced by the logarithm of that value. In "equal range scaling, " each descriptor is divided by the range of that descriptor for all samples. In this way, all descriptors have the same range, that is, 1. However, this method is sensitive to presence of outlier points. In "autoscaling," each data vector is mean centred and unit variance scaled. This technique is a very useful because each descriptor is then weighted equally, and large and small values are treated with equal emphasis. This can be important for genes expressed at very low, but still detectable, levels.

[129]When comparing data from multiple analyses (e.g., comparing expression profiles for one or more test samples to the centroids constructed from samples collected and analysed in an independent study), it will be necessary to normalize data across these data sets. In one embodiment, Distance Weighted Discrimination (DWD) is used to combine these data sets together (Benito et al. (2004), incorporated by reference herein in its entirety). DWD is a multivariate analysis tool that is able to identify systematic biases present in separate data sets and then make a global adjustment to compensate for these biases; in essence, each separate data set is a multi-dimensional cloud of data points, and DWD takes two points clouds and shifts one such that it more optimally overlaps the other. Further methods for combining data sets include the "ComBat" method and others described in Lagani et al. 2016, the entire contents of which is expressly incorporated herein by reference. ComBat is a method specifically devised for removing batch effects in gene-expression data (Johnson WE, Li C, Rabinovic A. 2007, the entire contents of which is expressly incorporated herein by reference).

[130]In some embodiments described herein, the prognostic performance of the gene expression signature and/or other clinical parameters is assessed utilizing a Cox Proportional Hazards Model Analysis, which is a regression method for survival data that provides an estimate of the hazard ratio and its confidence interval. The Cox model is a well-recognized statistical technique for exploring the relationship between the survival of a patient and particular variables. This statistical method permits estimation of the hazard (i.e., risk) of individuals given their prognostic variables (e.g., gene expression profile with or without additional clinical factors, as described herein). The "hazard ratio" is the risk of death (or event such as a recurrence of the cancer) at any given time point for patients displaying particular prognostic variables.

[131]Genes making up the gene signature or gene expression profile

[132]In accordance with any aspect of the present invention, the genes that make up the gene expression profile may be selected from any set of at least 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650, or at least 700, or essentially all of the following genes: ABCA8, ACTR3B, ACVR1B, ACVR1C, ACVRL1, ADAMI2, ADCY9, ADD1, ADM, AGR2, AGT, AGTR1, AKT3, ALDH1A1, ALDOA, ANGPT1, ANLN, ANXA9, APH1B, APOD, APOE, AR, AREG, ARID1A, ARNT2, ASPM, ASPN, ATAD2, ATM, ATP10B, AURKA, AURKB, AXIN1, AXIN2, B3GNT3, BAD, BAG1, BAIAP2L1, BAIAP3, BAMBI, BAX, BBC3, BBOX1, BCAS1, BCL11A, BCL2, BCL2A1, BCL2L1, BCL6B, BDNF, BIRC5, BLM, BLVRA, BMP2, BMP4, BMP5, BMP6, BMP7, BMP8A, BMPR1A, BMPR1B, BMPR2, BNC2, BNIP3, BORCS7, BRCA1, BRCA2, BTG2, C5orf38, CA12, CACNA1D, CACNA1H, CACNA2D1, CACNA2D3, CACNG1, CACNG4, CACNG6, CALML5, CAMK2B, CAV1, CBLC, CCL2, CCL21, CCL3L1, CCL4, CCL5, CCL7, CCL8, CCNA1, CCNA2, CCNB1, CCND1, CCND2, CCNE1, CCNE2, CCR1, CCR2, CCR5, CD163, CD19, CD1E, CD24, CD27, CD274, CD276, CD34, CD36, CD44, CD68, CD84, CD8A, CD8B, CDC14A, CDC14B, CDC20, CDC25A, CDC25B, CDC25C, CDC6, CDC7, CDCA5, CDCA7L, CDCA8, CDH1, CDH2, CDH3, CDH5, CDK1, CDK4, CDK6, CDKN1A, CDKN1B, CDKN1C, CDKN2A, CDKN2B, CDKN2C, CDKN2D, CDKN3, CEACAM5, CEACAM6, CENPF, CEP55, CETN2, CFD, CHAD, CHEK2, CHI3L1, CHIT1, CHRNA5, CKB, CKMT1A, CKS1B, CLDN1, CLDN3, CLDN4, CLDN7, CLEC14A, CLEC5A, CMKLR1, CNTFR, COL11A1, COL27A1, COL2A1, COL4A6, COL6A3, COL7A1, COL9A3, C0LEC12, COMP, CPA3, CREBBP, CRYAB, CSF3R, CTSW, CXADR, CXCL10, CXCL12, CXCL13, CXCL5, CXCL8, CXCL9, CXCR6, CXorf36, CXXC5, CYBB, CYP4F3, DCN, DDB2, DDR2, DDX39A, DEPDC1, DHRS2, DKK1, DKK2, DLGAP5, DLL1, DLL3, DLL4, DNAJC12, DPT, DSC2, DTX1, DTX3, DTX4, DUSP4, DUSP6, E2F1, E2F5, ECM2, EDN1, EDNRB, EFNA3, EFNA5, EGF, EGFR, EGLN2, EGLN3, EIF2AK3, EIF3B, EIF4E2, EIF4EBP1, ELF3, ELK3, ELOVL2, EMCN, ENO1, ENPP2, EP300, EPAS1, ERBB2, ERBB4, ERCC1, EREG, ESPL1, ESRI, ETV4, ETV7, EXO1, EYA1, EYA2, EYA4, F3, FAM124B, FAM198B, FAM214A, FAM83D, FANCF, FAP, FBN1, FGF1, FGF10, FGF12, FGF13, FGF18, FGF2, FGF7, FGF9, FGFR2, FGFR3, FGFR4, FGL2, FHL1, FLU , FLNC, FLRT3, FLT3, FNBP1, FOS, FOSL1, FOXA1, FOXCI, FOXC2, FOXM1, FOXP3, FREM2, FST, FSTL1, FSTL3, FUT3, FXYD3, FZD10, FZD7, FZD8, FZD9, GABRP, GADD45A, GADD45B, GADD45G, GAS1, GATA3, GATA4, GDF15, GDF5, GGH, GHR, GJB2, GLI3, GNG4, GNLY, GPC4, GPR160, GPX3, GRB2, GRB7, GREM1, GRIA3, GRIN1, GRIN2A, GSK3B, GTF2H2, GZMA, GZMB, GZMH, GZMM, HAPLN1, HAS1, HBB, HDAC1, HDAC10, HDAC11, HDAC2, HDAC5, HDAC6, HDC, HEG1, HELLS, HEMK1, HES1, HGF, HIF1A, HIST1H1C, HIST1H2BH, HIST1H3H, HIST3H2BB, HK2, HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DMB, HLA-DOB, HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRA, HLA-DRB1, HLA-E, HMGA1, HNF1A, HOXA5, HOXA7, HOXA9, HOXB13, HOXB3, HSPA2, TBSP, ICAM1, ID1, ID2, ID4, IDO1, IFT140, IGF1, IGF1R, IKZF3, IL10RA, IL11RA, IL12RB2, IL13RA1, IL1B, IL1R2, IL1RN, IL20RA, IL20RB, IL22RA2, IL24, IL2RA, IL2RB, IL3RA, IL4R, IL6, IL6R, IL7R, INHBA, INHBB, IRF6, IRX1, ISG15, ISM1, ITGA6, ITGAV, ITGB1, ITGB3, ITGB6, ITPR1, JAG1, JAG2, JAK1, JAK2, JAK3, JAM2, JCAD, JUN, KAT2B, KCNB1, KDR, KIAA0040, KIF11, KIF14, KIF23, KIF2C, KIFC1, KIT, KLRK1, KRT14, KRT17, KRT5, KRT6B, KRT7, LAD1, LAG3, LAMA3, LAMB3, LAMC2, LEF1, LEFTY2, LEMD1, LEP, LEPR, LFNG, LIF, LIFR, LINC02381, LPL, LRP2, LRRC32, LTB, LTBP1, MAD2L1, MAF, MAML2, MAP2K4, MAP3K12, MAPK1, MAPK10, MAPK3, MAPK8IP2, MAPT, MARCO, MCM2, MCM3, MDM2, MED1, MELK, MEOX2, MET, MFNG, MIA, MIS18A, MKI67, MLH1, MLLT3, MLPH, MME, MMP11, MMP14, MMP3, MMP7, MMP9, MMRN2, MRE11, MS4A2, MSR1, MT1G, MTOR, MUC1, MUS81, MYBL2, MYC, MYCN, MYCT1, NASP, NAT1, NCAM1, NCAPH2, NDC80, NDP, NEIL1, NEIL2, NEIL3, NEO1, NETO2, NFATC1, NFKBIZ, NGFR, NKG7, NOD2, NOTCH1, NOTCH2, NOTCH3, NPEPPS, NPR1, NR4A1, NR4A3, NRCAM, NRXN1, NRXN3, NSD1, NSD3, NTRK2, NUDT1, NUF2, NUMBL, NUPR1, OAS3, OCLN, OGN, OLFML2B, ORC6, PALB2, PALMD, PARP1, PARP2, PARP4, PAX5, PAX8, PBX3, PCK1, PCNA, PDCD1, PDCD1LG2, PDE9A, PDGFB, PDGFRA, PDGFRB, PDK4, PECAM1, PFDN2, PGK1, PGR, PHGDH, PIK3CA, PIK3CD, PIK3CG, PIK3R1, PIK3R2, PIK3R3, PIK3R5, PIM1, PIP, PKMYT1, PLA1A, PLA2G2A, PLA2G3, PLA2G4A, PLA2G4F, PLAT, PLCB1, PLCB4, PLCE1, PLD1, PMS2, POLDI, POLQ, POPDC3, PPARG, PPARGC1A, PPP2CB, PPP2R1A, PPP2R2C, PRC1, PREP, PRF1, PRKAA2, PRKACA, PRKACB, PRKCA, PRKCB, PRKDC, PRKX, PRLR, PROMI, PSAT1, PSMB10, PSMB7, PSMB9, PTCHI, PTEN, PTGDS, PTGER3, PTGS2, PTTG1, PYCARD, RAC2, RAC3, RAD51, RAD51C, RAD52, RAD54L, RARRES3, RASAL1, RASGRF1, RASGRF2, RASGRP1, RBI, RBL1, RBL2, RBX1, RELN, RFC4, RNASE2, RNF103, ROBO4, ROCK1, ROCK2, RORA, RORB, RPS6KA5, RPS6KB1, RPS6KB2, RRM2, RUNX3, S100A14, S100A7, S1PR1, SCARA5, SCUBE2, SELE, SERBP1, SERPINB5, SERPINH1, SFN, SFRP1, SFRP2, SFRP4, SHC2, SHC4, SHE, SHMT2, SIDT1, SIGIRR, SIX1, SKA3, SKP1, SKP2, SLC2A1, SLC39A6, SLC44A4, SLPI, SMAD1, SMAD3, SMAD4, SMAD5, SMC1B, SMO, SMURF2, SNAI1, SNAI2, SOCS1, SOCS2, SOCS3, SOXIO, SOX17, SOX2, SOX9, SP1, SPC25, SPDEF, SPN, SPP1, SPRY1, SPRY2, SPRY4, SRPX, ST6GALNAC2, STAT1, STC1, SUV39H2, SYTL4, TAPI, TAP2, TAPBP, TBC1D9, TBX1, TCEAL1, TCF4, TCF7L1, TEK, TFDP1, TFF1, TFF3, TGFB1, TGFB2, TGFB3, TGFBR2, THBS1, THBS2, THBS4, THY1, TIE1, TIGIT, TIMP4, TLE3, TLR4, TLX1, TMEM45B, TMPRSS2, TMPRSS4, TNF, TNFAIP6, TNFSF10, TNKS, TNKS2, TNN, TOP2A, TP53, TPSAB1, TRIP13, TSPAN1, TSPAN7, TTK, TTYH1, TUBA4A, TWIST1, TWIST2, TYK2, TYMP, TYMS, UBE2C, UBE2T, VCAN, VEGFA, VEGFD, VIM, VIT, WDR77, WEE1, WIFI, WNT10A, WNT11, WNT2, WNT4, WNT5A, WNT5B, WNT6, WNT7B, WRN, WT1, XRCC2, XRCC3, ZBTB16, ZEB1, ZEB2, ZFPM2, ZFYVE9, ZIC2, and ZNF205. In some embodiments the genes are those that form the genes of the NanoString BC360 panel (see www.nanostring.com/support- documents/breast-cancer-360-v2-gene-list/ as at 19 November 2021). Subsets of the genes may be selected based on their significance of differential expression as between Al responders and Al-non- responders. Moreover, predictive power of clusters derived from the gene sub-sets can be determined using statistical techniques known in the art in order to select an appropriate number of genes to yield acceptable predictive power while in some cases minimising the number of genes that need to be interrogated/have their gene expression measured.

[133]Breast cancer [134]As used herein, "breast cancer" refers to any cancer of the breast, including, in particular, ER+ and HER2+ primary breast cancer. The breast cancer may in some cases be a HER2-enriched (HER2-E) subtype. A breast cancer patient may be undergoing or may be a candidate for surgery, medical therapy (including endocrine therapy, chemotherapy, CDK inhibitor therapy and/or monoclonal antibody therapy) and/or radiotherapy.

[135]Surgery

[136]As used herein, "breast cancer surgery" or similar terms refer to physical removal of a breast tumour, optionally together with removal of surrounding tissue and/or lymph nodes. A breast cancer patient as contemplated herein may have had or may be a candidate for breast cancer surgery.

[137]Endocrine therapy

[138]As used herein, "endocrine therapy" or "hormonal therapy" includes therapy with agents intended to block hormone receptors (e.g. tamoxifen) or to block production of oestrogen such as an aromatise inhibitor (Al), e.g. anastrozole or letrozole.

[139] CDK inhibitor

[140]As used herein, "CDK inhibitor" includes CDK4/6 inhibitors such as palbociclib, abemaciclib and ribociclib. Moreover, agents that are being or will be developed to inhibit CDK, particularly CDK4 and CDK6, such as trilaciclib, are specifically contemplated herein.

[141]Table 3 — Genes making up the NanoString BC360 gene panel BRCA2, BTG2, C5orf38, CA12, CACNA1D, CACNA1H, CACNA2D1, CACNA2D3, CACNG1, CACNG4, CACNG6, CALML5, CAMK2B, CAV1, CBLC, CCL2, CCL21, CCL3L1, CCL4, CCL5, CCL7, CCL8, CCNA1, CCNA2, CCNB1, CCND1, CCND2, CCNE1, CCNE2, CCR1, CCR2, CCR5, CD163, CD19, CD1E, CD24, CD27, CD274, CD276, CD34, CD36, CD44, CD68, CD84, CD8A, CD8B, CDC14A, CDC14B, CDC20, CDC25A, CDC25B, CDC25C, CDC6, CDC7, CDCA5, CDCA7L, CDCA8, CDH1, CDH2, CDH3, CDH5, CDK1, CDK4, CDK6, CDKN1A, CDKN1B, CDKN1C, CDKN2A, CDKN2B, CDKN2C, CDKN2D, CDKN3, CEACAM5, CEACAM6, CENPF, CEP55, CETN2, CFD, CHAD, CHEK2, CHI3L1, CHIT1, CHRNA5, CKB, CKMT1A, CKS1B, CLDN1, CLDN3, CLDN4, CLDN7, CLEC14A, CLEC5A, CMKLR1, CNTFR, COL11A1, COL27A1, COL2A1, COL4A6, COL6A3, COL7A1, COL9A3, COLEC12, COMP, CPA3, CREBBP, CRYAB, CSF3R, CTSW, CXADR, CXCL10, CXCL12, CXCL13, CXCL5, CXCL8, CXCL9, CXCR6, CXorf36, CXXC5, CYBB, CYP4F3, DCN, DDB2, DDR2, DDX39A, DEPDC1, DHRS2, DKK1, DKK2, DLGAP5, DLL1, DLL3, DLL4, DNAJC12, DPT, DSC2, DTX1, DTX3, DTX4, DUSP4, DUSP6, E2F1, E2F5, ECM2, EDN1, EDNRB, EFNA3, EFNA5, EGF, EGFR, EGLN2, EGLN3, EIF2AK3, EIF3B, EIF4E2, EIF4EBP1, ELF3, ELK3, ELOVL2, EMCN, ENO1, ENPP2, EP300, EPAS1, ERBB2, ERBB4, ERCC1, EREG, ESPL1, ESRI, ETV4, ETV7, EXO1, EYA1, EYA2, EYA4, F3, FAM124B, FAM198B, FAM214A, FAM83D, FANCF, FAP, FBN1, FGF1, FGF10, FGF12, FGF13, FGF18, FGF2, FGF7, FGF9, FGFR2, FGFR3, FGFR4, FGL2, FHL1, FLU , FLNC, FLRT3, FLT3, FNBP1, FOS, FOSL1, FOXA1, FOXCI, FOXC2, FOXM1, FOXP3, FREM2, FST, FSTL1, FSTL3, FUT3, FXYD3, FZD10, FZD7, FZD8, FZD9, GABRP, GADD45A, GADD45B, GADD45G, GAS1, GATA3, GATA4, GDF15, GDF5, GGH, GHR, GJB2, GLI3, GNG4, GNLY, GPC4, GPR160, GPX3, GRB2, GRB7, GREM1, GRIA3, GRIN1, GRIN2A, GSK3B, GTF2H2, GZMA, GZMB, GZMH, GZMM, HAPLN1, HAS1, HBB, HDAC1, HDAC10, HDAC11, HDAC2, HDAC5, HDAC6, HDC, HEG1, HELLS, HEMK1, HES1, HGF, HIF1A, HIST1H1C, HIST1H2BH, HIST1H3H, HIST3H2BB, HK2, HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DMB, HLA-DOB, HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRA, HLA-DRB1, HLA-E, HMGA1, HNF1A, HOXA5, HOXA7, HOXA9, HOXB13, HOXB3, HSPA2, IBSP, ICAM1, ID1, ID2, ID4, IDO1, IFT140, IGF1, IGF1R, IKZF3, IL10RA, IL11RA, IL12RB2, IL13RA1, IL1B, IL1R2, IL1RN, IL20RA, IL20RB, IL22RA2, IL24, IL2RA, IL2RB, IL3RA, IL4R, IL6, IL6R, IL7R, INHBA, INHBB, IRF6, IRX1, ISG15, ISM1, ITGA6, ITGAV, ITGB1, ITGB3, ITGB6, ITPR1, JAG1, JAG2, JAK1, JAK2, JAK3, JAM2, JCAD, JUN, KAT2B, KCNB1, KDR, KIAA0040, KIF11, KIF14, KIF23, KIF2C,

[142]Table 4 — Centroids of single gene clusters GC1-GC5

Table 5 — Centroids of Signature Clusters SGC1-SGC4

Table 6 — Centroids of Signature Clusters SGC1-SGC4 showing individual genes within multi-gene groups

[143]The features disclosed in the foregoing description, or in the following claims, or in the accompanying drawings, expressed in their specific forms or in terms of a means for performing the disclosed function, or a method or process for obtaining the disclosed results, as appropriate, may, separately, or in any combination of such features, be utilised for realising the invention in diverse forms thereof.

[144]While the invention has been described in conjunction with the exemplary embodiments described above, many equivalent modifications and variations will be apparent to those skilled in the art when given this disclosure. Accordingly, the exemplary embodiments of the invention set forth above are considered to be illustrative and not limiting. Various changes to the described embodiments may be made without departing from the spirit and scope of the invention.

[145]For the avoidance of any doubt, any theoretical explanations provided herein are provided for the purposes of improving the understanding of a reader. The inventors do not wish to be bound by any of these theoretical explanations.

[146]Any section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

[147]Throughout this specification, including the claims which follow, unless the context requires otherwise, the word "comprise" and "include", and variations such as "comprises", "comprising", and "including" will be understood to imply the inclusion of a stated integer or step or group of integers or steps but not the exclusion of any other integer or step or group of integers or steps. [148]It must be noted that, as used in the specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" one particular value, and/or to "about" another particular value. When such a range is expressed, another embodiment includes from the one particular value and/or to the other particular value. Similarly, when values are expressed as approximations, by the use of the antecedent "about," it will be understood that the particular value forms another embodiment. The term "about" in relation to a numerical value is optional and means for example +/- 10%.

[149]The following is presented by way of example and is not to be construed as a limitation to the scope of the claims.

[150]Examples

[151]Example 1 - Overview

[152] Background.

[153]Although new HER2-targeted therapies have improved the outcome of ER+/HER2+ breast cancer (BC), 20% of patients with early disease still progress. Several mechanisms of resistance to aromatase inhibitors (Al) have been identified but mainly in the context of ER+/HER2- BC. The POETIC trial allows the study of resistance mechanisms in a large number of ER+/HER2+ BC tumours with Ki67 expression data after two weeks of Al treatment to measure response. Our aim was to identify biomarkers of response and resistance to Al in this subgroup.

[154]Methods. 342 ER+/HER2+ baseline BC tumors (237 treated/105 controls) were gene expression profiled using BC360™ (NanoString) covering intrinsic subtypes and 46 key biological signatures. Early response to Al was assessed by changes in Ki67 expression and residual Ki67 at 2 weeks (Ki67 _2wk )• Time-To-Recurrence (TTR) was estimated using Kaplan-Meier methods and Cox models adjusted for post-surgery clinicopathological variables: grade, tumour size, nodal status and age. New molecular subgroups were identified using consensus clustering (CC). [155]Findings. 44.7% ER+/HER2+ samples were HER2-enriched (HER2-E) subtype, and it was significantly associated with poor response to Ki67 and high Ki67 _2wk (p<0.0001). Endocrine related signatures were associated with good response while high expression of ERBB2, Homologous Recombination Deficiency and TP53 with poor response. In addition, immune-related signatures associated with high Ki67 _2wk . Using CC, we identified 5 gene-based and 4 signature-based clusters predicting differential response to Al. HER2-E had significantly poorer TTR compared to Luminal tumours, remaining significant in the multivariable analysis. Two clusters characterised by their ERBB2, endocrine and immune features were independent predictors of shorter TTR, adding significant value beyond intrinsic subtypes. Several immune-related signatures such as CD8 T-cells, PD-L1, or PD1 were also associated with better TTR.

[156]Interpretation/Conclusions: Our results establish HER2-E subtype as the first standardized biomarker driving poor response to Al and worst outcome in ER+/HER2+. Signatures' expression of DNA damage response, TP53 mutational status and immune-tumour tolerance are also predictive biomarkers for early poor response to Al. Lastly, novel molecular subgroups identify additional non-HER2-E tumours not responding to Al with an increased risk of relapse. The appropriate additional treatment warrants further investigation.

[157]Example 2 - HER2-enriched subtype and novel molecular subgroups drive aromatase inhibitor resistance and an increased risk of relapse in early ER+/HER2+ breast cancer

[158]Methods

[159]Patients and samples

[160]All available ER+/HER2+ BC tumours from the POETIC trial in which patients were assigned to 2 weeks of peri-surgical Al (before and after surgery) or no Al (control) were included in this study ¹¹ . A consort diagram of the study is shown in Supplementary Figure 1. Ki67 staining of 2-week samples from the Control group was restricted to a randomly selected subset due to the minimal expected change from baseline to surgery ¹² . In summary, from 470 ER+/HER2+ patients included in POETIC we obtained successful results from 342 patients.

[161]RNA extraction

[162]RNA was extracted from three macrodissected 10pm FFPE sections, from the baseline and surgery sample of the patients included in the study, using the ROCHE High Pure miRNA isolation kit (Roche, Basel, Switzerland) following SOP M027 from The Cancer Genome Atlas (TCGA) Program developed by the Biospecimen Core Resource (BCR) at Nationwide Children's Hospital in Columbus, Ohio. Quantification was done using high sensitivity RNA Qubit assays (Thermo Fisher Scientific, Carlsbad, CA).

[163]Gene expression profiling

[164]Gene expression was assessed using the NanoString nCounter Platform (Nanostring Technologies, Seattle, WA) Breast Cancer 360™ codeset (BC360) (Supplementary Table 1). This panel evaluates the intrinsic molecular subtypes identified using the PAM50 classifier ¹³ and the intrinsic subtypes scores, defined as the correlation coefficient scores to each prototypical intrinsic subtype average gene expression profile (i.e. centroid). It also assesses the expression of 750 additional genes encompassing important biological variation from published genomic signatures and individual genes covering the most important aspects of BC including key biological signatures, immune-cell types/response and DNA-damage repair related genes. In summary, 150ng of RNA was run on a NanoString nCounter™ FLEX Analysis System. RNA was hybridised overnight at 65°C and the samples were processed using the NanoString nCounter Prep Station and Digital Analyzer according to manufacturer's instructions. NanoString raw data was extracted using nsolver and was normalized by NanoString according to the BC360 pipeline using 18 house-keeping genes and log2 base transformed.

[165]Immunohistochemistry

[166]ER status was measured locally and was centrally reviewed by IHC. HER2 status was measured locally using immunohistochemistry (IHC) and/or fluorescence in situ hybridisation (FISH). Ki67 proliferation rate was obtained by immunohistochemistry from staining on formalin-fixed samples using anti-MIB-1 (M7240, DAKO UK)n-14.

[167]Outcomes

[168]The primary endpoints in this study were based on Ki67 as a measure of tumour's resistance to Al. Two Ki67 endpoints were used in this study: 1) Ki67 change was calculated as the difference between Ki67 expression at surgery and baseline, and was categorised into Ki67 response categories defined as percentage-changes from baseline to surgery: poor response (PR) (reduction <50%), intermediate response (IR) (50-75%) and good response (GR) (>75%). This reflects the antiproliferative response to Al treatment which relates to the treatment benefit. 2) Residual Ki67 at 2 weeks timepoint (Ki67 _2wk ) High (b 10%) and Low (<10%) which correlates to the residual risk after Al treatment.

[169]The secondary endpoint in this study was Time to recurrence (TTR), which evaluated the prognostic significance of the molecular characteristics analyzed.

[170]Statistical analysis

[171]Statistical analysis was performed with R version 3.6.3 software. A two-tailed p-value of less than 0.05 was considered statistically significant. Wilcoxon tests were applied in unpaired comparisons and Kruskall-Wallis tests in all multiple comparisons in both treated and control tumours. Spearman Rank correlation was used to explore the correlation between particular genes or signatures. Logistic regression and ordinal regression models were performed to identify signatures significantly associated with the different Ki67 endpoints.

[172]Significance Analysis of Microarrays (SAM analysis) was performed in multiclass setting to compare gene expression values between the three Ki67 response categories (GR/IR/PR) and in the unpaired Two Class setting to compare extreme response classes (GR and PR) and also Ki67 _2wk High versus Low ¹⁵ . Hierarchical clustering of the expression profiles of the genes identified by SAM analysis were performed ¹⁶ .

[173]Consensus clustering was used to identify new molecular subgroups and then we tested the association of the new subgroups with Ki67 response categories and outcome. We included both controls and treated patients to obtain the clusters, but only the treated patients with Ki67 data were used to assess their predictive value for Al resistance.

[174]TTR was estimated using Kaplan-Meier methods and Cox models. Multivariable Cox-regression models adjusted for standard postsurgery clinicopathological variables: grade, tumour size, nodal status and for age as a main surrogate of adjuvant treatment choice, were performed. The independent prognostic value of those geneexpression based variables with differential survival in the univariate analysis were assessed. Both controls and treated patients were included in the survival analysis.

[175]Ethics statement:

[176]Ethical approval for POETIC (Trial Number CRUK/07/ 015) was provided by NRES Committee London-South East. Patients consented to molecular analysis of their samples for research purposes.

[177]Results

[178]Patient clinicopathological characteristics

[179]In this study, 342 ER+/HER2+ patients with baseline gene expression were included: 237 treated and 105 controls (Supplementary Figure 1). The demographics are shown in Supplementary Table 1. In summary, 93.3% of the tumours were ductal, 46.2% were grade 2 and 49.1% grade 3. At surgery 54.7% had a tumour diameter between 2 and 5 cm and 47.4% had positive nodal status. Sixty-nine-point five percent of patients received adjuvant chemotherapy and trastuzumab and 98.2% of patients were treated with adjuvant endocrine therapy. [180]PAM50 subtypes and K167 endpoints

[181]To evaluate whether intrinsic subtypes could predict response to endocrine therapy we analysed the association of PAM50 with Ki67 endpoints. At baseline, 44.7% of tumours classified as HER2-E, 53.5% Luminal A/B and 1.8% Basal-Like. The proportion of intrinsic subtypes was comparable between control and treated groups with controls slightly enriched with HER2-E (54% versus 41.4%) and reduced Luminal-A tumours (8.0% vs 21.1%) (Table 1). Within the subgroup of patients treated with peri-operative Al, 31% achieved GR, 22.5% were IR and 46.5% PR, while 52.0% had Ki67 _2wk High and 48.0% Low. As expected, most control tumours were classified as PR (96.0%) and Ki67 _2wk High (96.0%) (Table 1). In the treated group, there was a significant change in Ki67 values in all subtypes (Figure 1A), except for Basal-like, probably due to the small sample number. Overall, HER2-E was associated with poorer response to Al compared to non-HER2-E, evaluated as Ki67 response category and residual Ki67. In the control patients, no significant changes of Ki67 were observed amongst any of the subtypes (Figure IB). These findings suggest that HER2-E might be one of the main components driving poor early response to Al in ER+/HER2+ BC tumours.

[182]Signature expression and Ki67 endpoints

[183]We then evaluated the association of other biological molecular features beyond subtypes with Ki67 endpoints. The association of the expression of the 46 signatures for the two different Ki67 endpoints are shown in Supplementary Table 3. High expression of endocrine related signatures such as ESRI, ER- Signaling, FOXA1 and PgR as well as Luminal A and B correlation coefficient scores were associated with GR and Low Ki67 _2wk , while ERBB2, and basal-like and HER2-E correlation coefficient scores were associated with PR and High Ki67kwk (Figure 2). Noteworthy, the apoptosis signature was associated with GR whilst DNA-damage repair signature such as the homologous recombination deficiency (HDR), hypoxia, and the TP53 mutational status' surrogate signature were associated with PR. Additional signatures involved in immune- checkpoint component enrichment and tumour immunity such as IDO1, IFN Gamma, PD-L1 and TIS as well as the genomic risk score were associated with High Ki67 _2wk .

[184] To assess the above associations according to the different subtypes, we tested them in the HER2-E and Luminal subtypes separately. No significantly differently expressed signatures between response categories amongst HER2-E tumours for any of the two Ki67 endpoints were identified (Supplementary Figure 2A) . However, for the Luminal tumours only, proliferation, BC p53, genomic risk and the basal-like and HER2-E coefficient scores were associated with High Ki67 _2wk , while high expression of AR signature and the Luminal A coefficient scores were associated with Low Ki67kwk (Supplementary Figure 2B) .

[185] Single gene expression and Ki67 endpoints

[186] Multiclass SAM analysis of single gene expression for the three Ki67 response categories (GR/IR/PR) (Δ=0.26, FDR<0.05) identified 8 genes with significantly different expression amongst groups (p<0.05) . High expression of GRB7 and ERBB2, both involved in ERBB2 signaling were associated with PR while high expression of others such as IGF1R and ESRI involved in ER signaling or CHAD and BCL2, were associated with GR (Figure 3A) . Two unpaired class SAM analysis for GR versus PR (Δ=1, FDR<0.05) showed the association of 31 genes with GR (p <0.05) (Figure 3B) . Two class unpaired SAM analysis with Ki67kwk categories: High versus Low (Δ=1.63, median FDR=0) identified 128 genes associated to Low Ki67 _2wk including genes involved in PI3K/AKT, MAPK and estrogen signaling and 83 genes associated to High Ki67 _2wk , including genes involved in immune- checkpoint component, proliferation and cell-cycle regulation (Figure 3C) . The heatmap shown reflects the obvious patterns driving changes seen at the extreme samples on either sides: ER signaling in Luminal A and GR and ERBB2 /immune-related genes in HER2-E and PR. It is noteworthy that Luminal A tumors were divided into two different groups with differential expression of ER signaling and differential response to Al treatment.

[187] To further investigate the biological differences between these two groups we evaluated the ESRI gene expression levels amongst the four intrinsic subtypes. The highest levels of ESRI were seen in Luminal B tumours followed by Luminal A, HER2-E and Basal- like. Within the Luminal A tumours a group of patients showed lower levels of ESRI that seemed to be driven by higher ERBB2 signaling, especially when compared with other tumours such as Luminal B or HER2-E (Supplementary Figure 3).

[188]Identification of new molecular subgroups predicting Ki67 endpoints

[189]Using consensus clustering, we discovered 5 novel clusters of samples (C) based on single-gene (GC) and 4 clusters based on signature expression (SGC) predicting significantly different response to Al in treated samples. These clusters also associated with Ki67 at 2-week timepoint (Table 2A). Interestingly, both sets of clusters divided mainly HER2-E subtype samples with lower response to Al, into three groups with differential expression of molecular features such as ERBB2, ER signaling or immune-related pathways.

[190]Figure 4A shows the molecular features of the single gene expression clusters: GC1, GC2 and GC3 were characterized by poorer response to Al and an enrichment of HER2-E subtype (79/111; 71.2%) and Luminal B tumours (21/111, 18.9%). GC1 (29.2%) showed higher levels of immune and chemokine related genes; GC2 (14.2%) had higher levels of extracellular matrix organization (ECM) related genes and the highest levels of ERBB2 levels (Figure 4B), not observed when distributed by subtype (Figure 4C) and GC3 (5.7%) had low expression of most genes but a slight upregulation of a group of genes covering DNA-damage repair deficiency such as RAD51, BRCA, CCNE1, TRIP13, CDCA5, RFC4, KIF14 and BLM. GC4 (22.1%) and GC5 (28.8%) were characterized by good response to Al and an enrichment of Luminal subtype tumours (Luminal B 81/165; 49.1% and Luminal A 57/165; 34.5%). Both clusters overexpressed ER signaling related genes, while GC5 was also enriched with genes involved in MAPK/PI3K and RAS signaling.

[191]For the 4 new clusters based on signature expression (Figure 5): SGC1 (41.8%) was characterized by overexpression of immune features and lower ER signaling; SGC2 (15.5%) by low immune but significantly higher levels of ERBB2 expression levels (Figure 5B), independent of subtype (Figure 5C); SGC3 (4.1%) was characterized by ESRI high and PgR low and SGC4 (38.6%) by high endocrine signaling and lowest ERBB2 expression. PAM50 subtype distribution varied among these subgroups with an enrichment of: SGC1 HER2-E (62.2%) and Luminal B (26.6%), SGC2 HER2-E (73.6%), SGC3 Luminal B (64.3%) and SGC4 Luminal B (55.3%) and Luminal A (29.5%).

[192]The highest overlap between single gene and signature-based clusters was in GC1 (89.5%) and SGC1 (68.6%) (Table 2B), followed by SGC4 where 77.2% of patients are included in GC4 and GC5.

[193]Time to Recurrence Analysis

[194]Univariate analysis was performed to evaluate the prognostic value of each individual genomic feature analysed. In this analysis, HER2-E had significantly poorer time to recurrence (TTR) compared to Luminal tumours (Figure 6A and 6B). GC2 (the ECM enriched and highest ERBB2) and SGC2 (the immune low and highest ERBB2) - showed a significantly higher risk of recurrence compared to the rest of clusters (Figure 6C and 6D).

[195]To assess the independent prognostic value of those geneexpression based variables with differential survival in the univariate analysis, we performed a series of cox regression models for multivariable survival analysis for TTR and compared the changes of chi-square values between them to assess the added value of the different models (Supplementary Table 4). In the multivariable analysis adjusted for post-surgery clinicopathological variables: grade, tumour size, nodal status, and age as the main driver/surrogate for the adjuvant treatment choice, HER2-E remained as an independent predictor of higher risk of relapse. SGC2 and SGC4 and GC2 clusters were also independent predictors of shorter TTR compared to GC1 and SGC1 of each set, adding both models, significant value beyond intrinsic subtypes (Likelihood ratio test, p<0.001 and 0.0025, respectively). Two additional models: one including all CCs from both sets together and a second with all Cs and PAM50 did not add any additional value than just considering one set of clusters and PAM50 (Supplementary Figure 4).

[196]Finally, a series of multivariable analysis for TTR of signature expression adjusted for the clinicopathological factors showed that several immune related signatures as well as the apoptosis signature were independent predictors of better TTR. By contrast, claudin-low signature showed independent worse prognostic value (Supplementary Table 5). A final model including all identified features was not performed due to the high collinearity amongst the main significant signatures (Supplementary Figure 5).

[197]DISCUSSION

[198]Al treatment is the standard of care and most effective therapy for post-menopausal women with early ER+ BC. However, ER+ tumours that also over-express HER2 show limited response to endocrine therapy and thus, are at a higher risk of recurrence ¹⁴ . Most studies performed in ER+/HER2+ BC have focused on resistance to anti-HER2 targeted therapy while mechanisms of resistance to endocrine therapy are not well understood yet. Molecular characterization of gene expression profiles at baseline before neoadjuvant treatment are necessary to identify biomarkers of response and to improve treatment management. This study was designed to investigate the predictive and prognostic value of molecular features at baseline in ER+/HER2+ BC treated with perioperative Al. To our knowledge this is the largest cohort to date investigating response to pre-surgical Al in this BC subgroup.

[199]In this study, we demonstrate the importance of HER2-E subtype in resistance to endocrine therapy in ER+/HER2+ BC. The main issue whilst designing effective treatment approaches in BC is the heterogeneity of this tumor type. HER2+ BC is formed by all the transcriptional subtypes of BC, which are known to differ in terms of incidence, prognosis, and response to different therapies ¹⁷ -^. Prior studies have revealed that HER2-E subtype is a predictor of higher sensitivity to anti-HER2 targeted therapy but worse outcome than other subtypes such as Luminal A and Luminal B BC tumours {Cejalvo CCR 2018), however the role of the intrinsic subtypes in response to endocrine therapy has not been well stablished yet. Altogether, our results identify HER2-E as the first predictive biomarker of resistance to Al in ER+/HER2+ BC with an additional higher risk of relapse. These findings highlight two aspects. First, the potential need of treatment intensification for HER2-E ER+/HER2+ BC tumours with additional anti-HER2 targeted therapy. Second, the higher sensitivity to Al and good prognosis associated with luminal tumours, in particular with Luminal A, provides a rationale against de-escalation, which has been previously suggested ¹⁹ .

[200]Using consensus clustering, we identified new molecular subgroups based on either single gene expression or BC360 signatures at a higher risk of relapse beyond HER2-E subtype. Noteworthy, tumours characterised by an enrichment of immune features showed the lowest risk of recurrence despite being predictive markers of resistance to Al, suggesting an intrinsic good prognosis of those immune-related characteristics. By contrast, tumours with higher levels of ERBB2 and lower associated immunity had a significantly higher risk of relapse, indicating a potential benefit from an intensified anti-HER2 treatment, using for example double anti-HER2 blockade or adjuvant TDM1 ²⁰ ' ²¹ . Patients with tumours characterized by higher levels of ER-signaling and the lowest levels of ERBB2 also showed significant worse outcome, highlighting the need of additional treatments in this subgroup to enhance the effect of Al treatment (i.e. CDK4/6 inhibitors in combination with Al). Based on the high Chi-square Likelihood ratio and lower degrees of freedom in the multivariable model 4, the assessment of signature-based CCs in combination with intrinsic subtypes could serve as a key tool to select patients at a higher risk of relapse. Whether this approach would be cost-beneficial needs further investigation.

[201]Several trials in HER2+ BC have shown that tumors with higher baseline tumour infiltrating lymphocytes (TILs) and other immune features achieve (Roberto Salgado JAMA Oncol 2015). However, the role of tumour immunity on response to endocrine therapy has been mainly studied in ER+/HER2- BC disease, with higher expression of genes involved in immune cells and targetable immune checkpoint components being correlated with higher risk tumours and poor response to Al ²³ (Anurag JNCI 2020). Our study confirms the association of higher expression of immune-features and lower risk of recurrence in ER+/HER2+ BC. It also indicates that higher tumour immunity might also be a key driver of early resistance to endocrine therapy in this subgroup.

[202]Finally, our study shows other interesting molecular associations such as HDR and p53 signatures predicting poor response to Al, higher apoptosis signaling being associated with both good response and survival and claudin-low signature with higher risk of relapse. Inhibition of poly (ADP-ribose) polymerase-1 (PARP), a key enzyme in the repair of single-stranded DNA breaks, has shown antitumour effects with a strong synergism and good tolerance in combination with anti-HER2 targeted therapy or endocrine therapy in vitro and in phase I/II trials independently of DNA repair deficiency ²⁴ . P53 mutational status has been previously linked with poor survival ²⁵ and overexpression of HER2 ²⁶ , although its role in response to Al is still unclear. The apoptotic signature, formed by 5 genes, has an overall pro-apoptotic effect. Previous data has shown that some molecular features related to apoptosis can be predictive of adjuvant benefit from endocrine therapy ²⁷ . Finally, claudin-low has been recently re-defined as a complex additional phenotype distinguished by low genomic instability, mutational burden and proliferation levels, and high levels of immune and stromal cell infiltration, rather than just an additional subtype ²⁸ . The impact of these associations warrants further investigation.

[203]Our study has two main limitations. First, we only analysed gene expression profiles using the BC360 platform. However, this includes the intrinsic subtypes and the major molecular pathways involved in BC. Second, clinical practice is currently different to that from the recruited POETIC patients as high-risk tumours would be receiving additional pertuzumab and further anti-HER2 agents such as TDM1. It is noteworthy that one third of the patients in our cohort did not receive any adjuvant treatment apart from endocrine therapy due to their advanced age, but we adjusted the survival analysis for age as a main surrogate of treatment choice. Our major strengths are that to our knowledge this is the largest cohort investigating response to pre-surgical Al in ER+/HER2+ subgroup in a real-world cohort which has a unique value to assess global gene expression data at baseline as defined in the clinical practice. The predictive value of the molecular features is real because we included a control group that enables the distinction between real impact of Al therapy and artefactual effect derived from sample manipulation.

[204]Conclusions and clinical implications:

[205]HER2-E subtype and ERBB2 play a crucial role in ER+/HER2+ BC, driving resistance to endocrine therapy and a higher risk of recurrence. Beyond them, new molecular subgroups using signature expression enable the identification of patients at a higher risk of relapse. Altogether, the combination of these biomarkers could be essential for personalising therapy, including escalation and de- escalation strategies, to improve resistance to treatment in early

BC.

[206]Example 3 — Clinical Applications

[207]COHORT: Estrogen receptor positive (ER+) breast cancer (BC) accounts for about 80% of whole Breast Cancer (BC). Approximately 10-15% of them are also classified as positive and/or over-expressed for human epidermal growth factor receptor (HER2+), which confers distinct molecular biology and clinical behavior.

[208]Background: ER+/HER2+ BC is a very heterogenous disease: there are several treatment options but response to these different drugs varies within this clinical subgroup.

[209]To date the standard of care for patients is as follows:

[210]1.Trastuzumab + chemotherapy (anthracyclines + taxanes) in all early breast cancer without contraindications + 5-10 years of endocrine therapy (ET) depending on hormone receptor status; however, trastuzumab and chemotherapy might be avoided in cases identified as very low clinical risk (i.e.T1aN0). [211]2. For low clinical risk, such as small tumour size: <cTlc, anthracyclines could be avoided but treated according to Tolaney's scheme: 3 months taxanes + trastuzumab up to 1 year + ET.

[212]3. For high risk tumours (node positive), double blockade trastuzumab + pertuzumab + chemotherapy would be given. If given at the neoadjuvant setting, and patients did not achieve pathological complete response, TDM1 would be given at the adjuvant setting (based on the recent results from the Katherine trial)

[213]Clinical question being addressed

[214]In general, ER+/HER2+ BCs show a reduced antiproliferative effect of endocrine therapy and lower response rates to anti-HER2 targeted therapy compared to ER+/HER2- and ER-/HER2+ tumours, respectively. There is a lack of optimal biomarkers to pair with the optimal treatment for each patient within this clinical subgroup. As such, identifying robust molecular features and defining novel subgroups based on tumour biology would help to identify the most adequate treatment strategies for this particular BC subgroup.

[215]Using consensus clustering, we have identified (a) five new molecular groups based on individual gene expression and (b) 4 subgroups based on annotated gene signatures (BC360 code set defined signature), respectively.

[216]Without wishing to be bound by any particular theory, the present inventors presently believe that there would be more flexibility for developing a diagnostic or treatment response assay using single gene expression-based clusters.

[217]Our results suggest that identification of BC samples based on either (a) or (b) can identify patients with a higher risk of relapse and distinguish them from those at a lower risk, in addition to the HER2-E subtype

[218]Details of the molecular subgroups

[219]Gene expression defined clusters. [220]Characteristics: (Figure 4A). Clusters 1, 2 and 3 are characterized by poorer response to AT and enrichment of HER2-E subtype (71.2%) and Luminal B tumours (18.9%). Cluster 1 (29.2%) shows higher levels of immune and chemokine related genes, cluster 2 (14.2%) with higher levels of extracellular matrix organization (ECM) related genes and the highest levels of ERBB2 levels (Figure IB) and cluster 3 (5.7%) has low expression of most genes but a slight upregulation of a group of genes covering DNA-damage repair deficiency. Clusters 4 (22.1%) and 5 (28.8%) are characterized by good response to AT and an enrichment of Luminal subtype tumours (Luminal B: 49.1% and Luminal A: 34.5%). Both clusters overexpress ER signaling related genes, while cluster 5 is also enriched with genes involved in MAPK/PI3K and RAS signaling.

[221]Clinical value: In the multivariable Cox regression analysis adjusted for post-surgery clinicopathological variables: grade, tumour size, nodal status, and age (which was the main driver for the adjuvant treatment choice), cluster 2 is an independent predictor of shorter time to recurrence when compared to cluster 1 with a hazard ratio of 4.82 (95%CI1.93-12.05; p=0.00075). This molecular signature survival model also provides significantly more prognostic information to a model with intrinsic subtypes (Likelihood ratio test, p= 0.0025) (Table 1).

[222]Potential applicability of the single-gene clusters in clinics:

[223]These new molecular subgroups can help to select those patients at a higher risk of relapse to optimize their treatment: cluster 2 with the highest ERBB2 levels and low immune-features. These are the post-menopausal patients with the poorest early response to aromatase inhibitor (AT) (77.3% Ki67 _2wks high -Figure 4) and poorest outcome. These patients might not benefit from endocrine therapy due to the lack of early response but they might be good candidates for a more intensive anti-HER2 therapy scheme (i.e to complete at least adjuvant/neoadjuvant trastuzumab +/- pertuzumab or even adjuvant TDM-l/neratinib if pCR is not achieved in case of neoadjuvant treatment). [224]To select those patients that might not need Al or trastuzumab in the adjuvant setting due to their molecular features associated with intrinsically good prognostication: Cluster 1 is the largest subgroup of tumours showing low levels of ERBB2 and ER-signaling but highest levels of immune-related features. These tumours were also associated with the best survival outcome, thus patients with these tumours might not need additional treatment. This concept should be further investigated but may help to inform better future clinical study design. These tumours can also be candidates for new trials investigating de-escalating treatment such as less than the standard 1 year of adjuvant trastuzumab or 6 months of trastuzumab; these could also be explored with the ET treatment. These would be highly sought-after questions globally, particularly relevant to regions/countries where such cost of these standard of care treatments are not covered and patients to pay.

[225]Cluster 3 is a very small subgroup of tumours showing lower early responses to AT but not associated with higher risk of relapse at 5 years. Current standard of care (i.e. 5 year of AT) might be recommended for this group.

[226]Cluster 4 and 5 are good early responders to AT and all show good TTR, thus standard treatment with adjuvant AT +/- anti-HER2 treatment might be recommended.

[227]Annotated Gene Signatures defined clusters.

[228]Characteristics: (Figure 5A). For the 4 new novel molecular subgroups: Cluster 1 (41.8%) is characterized by overexpression of immune features and lower ER signaling; cluster 2 (15.5%) by low IM but significantly higher levels of ERBB2 expression levels than the rest of the clusters (Figure 5B); cluster 3 (4.1%) is characterized by ESRI high and PgR low and cluster 4 (C4) (38.6%) by high endocrine signaling and lowest ERBB2 expression. PAM50 subtypes distribution varies among these subgroups: Cluster 1 being mostly HER2-E (62.2%) and Luminal B (26.6%), Cluster 2 mostly HER2-E (73.6%), Cluster 3 mostly Luminal B (64.3%) and Cluster 4 mostly Luminal B (55.3%) and Luminal A (29.5%). [229]Clinical value: In the multivariable Cox regression analysis adjusted for post-surgery clinicopathological variables: grade, tumour size, nodal status, and age (which was the main driver for the adjuvant treatment choice), cluster 2 and cluster 4 are independent predictors of shorter TTR compared to cluster 1. This molecular signature survival model also provided significant more information in addition to one with the intrinsic subtypes (Likelihood ratio test, p= 0.0025) (Supplementary Table 4).

[230]Potential applicability of the signature-based clusters in clinics:

[231]Clusters 1 and 2 would probably be treated similarly. However, patients in Cluster 1 might be treated with a de-escalating approach while cluster 2 could receive a more intensive anti-HER2 therapy scheme such as complete at least adjuvant/neoadjuvant trastuzumab +/- pertuzumab or even adjuvant TDM1.

[232]Cluster 3 encompasses a small number of patients, who showed lower early responses to Al, yet not associated with higher risk of relapse at 5 years. Cluster 3 is characterized by tumours showing higher levels of ESRI with lower levels of PgR. Therefore, standard of care could be recommended, but longer follow-up would be needed to rule out late relapse.

[233]Cluster 4 of tumours show an increased risk of relapse. This cluster of tumours is characterized by high ER-signaling thus should respond to endocrine therapy but the low-level gene expression of ERBB2 might suggest anti-Her2- treatment less effective.

Alternative combination of standard adjuvant treatment Al with treatments enhancing those pathways such as CDK4/6 inhibitors might be needed to improve outcome, and could be explored further - this is important to explore the role of CDK4/6 inhibitors while the role of anti-HER2+ treatment in these patients might be less important.

[234]EXAMPLE 4 — Identification of molecular changes in early Al sensitive ER+/HER2+ BC and evaluation of the association of posttreatment features with recurrence: a POETIC sub-study. [235]Introduction:

[236]Based on our prior data, we hypothesize that molecular changes may vary according to response to Al and might predict differential outcome. Here we sought to investigate changes in gene expression profiles under 2 weeks (2wk) of Al and evaluate whether posttreatment characteristics provides more prognostic information than baseline amongst patients with ER+/HER2+ early BC in POETIC trial in which patients were randomized 2:1 to receive 2 weeks of Al vs no treatment.

[237]MATERIALS AND METHODS

[238]Patients and samples

[239]POETIC was a phase III trial including post-menopausal women with ER+ early BC receiving 2wk peri-operative Al followed by standard-of-care. All available ER+/HER2+ BC tumours with available baseline and surgical samples from the POETIC trial in which patients were assigned to 2 weeks of peri-surgical Al or no Al (control) were included in this study (I Smith and M Bergamino ebiomedicin). A consort diagram of the study is shown in Supplementary figure 1. Ki67 staining of 2-week samples from the control group was restricted to a randomly selected subset due to the minimal expected change on Ki67 from baseline to surgery. In summary, of 470 ER+/HER2+ patients included in POETIC, there were baseline and on-treatment formalin-fixed paraffin-embedded samples from 213 treated/101 controls ER+/HER2+ BC.

[240]RNA extraction

[241]RNA was extracted from three adjacent macro-dissected 10pm formalin-fixed paraffin-embedded (FFPE) sections from the baseline and the surgical block of the patients included in the study. The ROCHE High Pure miRNA isolation kit (Roche, Basel, Switzerland) was used following SOP M027 from The Cancer Genome Atlas (TCGA) Program developed by the Biospecimen Core Resource (BCR) at Nationwide Children's Hospital in Columbus, Ohio. Quantification was done using high sensitivity RNA Qubit assays (Thermo Fisher Scientific, Carlsbad, CA). [242]Gene expression profiling

[243]Gene expression of 758 genes was assessed at baseline and at surgery using the NanoString nCounter Platform (Nanostring Technologies, Seattle, WA) Breast Cancer 360™ codeset (BC360) covering intrinsic subtypes and 46 key biological signatures (Supplementary table 1). 150ng of RNA was run and processed on a NanoString nCounter™ FLEX Analysis System according to manufacturer's instructions. NanoString raw data was normalized by NanoString according to the BC360 pipeline using 18 house-keeping genes.

[244]Biomarker analysis

[245]ER status was measured locally and was centrally reviewed by immunohistochemistry (IHC). HER2 status was measured locally using IHC and/or fluorescence in situ hybridisation (FISH). Ki67 proliferation rate was obtained by IHC in FFPE tissues using the MIB-1 antibody (M7240, DAKO UK) (Smith et al Lancet Oncol, Dowsentt IMPACT JCO 2005).

[246]The primary endpoints of this study were based on Ki67 as a measure of tumour resistance to Al. Two Ki67 endpoints were used: 1) Ki67 change was calculated as the difference between Ki67 expression at surgery and baseline (relative change) and was categorised into Ki67 response categories defined as percentage-changes from baseline to surgery: poor response (PR) (reduction <50%), intermediate response (IR) (50-75%) and good response (GR) (>75%). This reflects the antiproliferative response to Al treatment which relates to the treatment benefit. 2) Residual Ki67 at 2-week timepoint (Ki67 _2wk ) High (≥ 10%) and Low (<10%) which correlated to the residual risk after Al treatment. We also classified tumours into four risk categories according to the intrinsic subtype change from baseline to surgery. Intrinsic subtypes were defined into two grades: low- risk (L) subtype grade that includes LumA. High-risk (H) subtype grade that includes LumB, HER2-E and Basal. Based on this definition, the intrinsic subtype change from baseline to surgery was assigned into 4 grades as show in supplementary table 6, including low-risk subtype to Low-risk subtype (LL), High-risk subtype to Low-risk subtype (HL), Low-risk subtype to High-risk subtype (LH) and High-risk subtype to High-risk subtype (HH).The secondary endpoint was time to recurrence (TTR) (local and metastatic recurrence) to evaluate the prognostic significance of the molecular characteristics analysed.

[247]Statistical analysis

[248]Statistical analysis was performed using the R software (version 3.6.3). Gene expression changes were calculated as Log2 Fold Change (Log2FC) using the Log 2 of the difference of the scores between surgery and at baseline. P values were considered significant if lower than 0-05. T-tests were applied in unpaired comparisons in both treated and control tumours. Multiple testing correction was undertaken by the Benjamini & Hochberg (FDR) method. For single gene and signature scores, a combined threshold of significance was defined as P adjusted <0.05 and log2FC (FC)> |0.5| to compare the Ki67 response categories (GR and PR) and Ki67 _2wk High versus Low. Spearman Rank correlation was used to explore the correlation between genes or signatures.

[249]Time to Recurrence (TTR) was measured as time from randomisation to local, regional, or distant tumour recurrence or death from breast cancer without previous notification of relapse. Second primary cancers and intercurrent deaths were censored.

[250]TTR was estimated using Kaplan-Meier methods and Cox models. Multivariable Cox-regression models were adjusted for standard postsurgery clinicopathological variables: grade, tumour size, nodal status and age. We included age as the main driver/surrogate for the adjuvant treatment choice as most patients ≥70 years old did not receive chemotherapy or trastuzumab (67.2%, 82/122) compared to patients <70 years (14%, 31/221). The independent prognostic value of those gene-expression changes with differential survival in the univariate analysis were assessed. Only gene expression changes from treated patients were included in the survival analysis.

[251]Ethics statement [252]POETIC trial was approved by the London-South East Research Ethics Committee (reference 08/H1102/37) and adopted by the Declaration of Helsinki. PPatients provided written informed consent to molecular analysis of their samples for research purposes.

[253]RESULTS:

[254]Patient clinicopathological characteristics

[255]In this study, 314 ER+/HER2+ patients with baseline and surgery biopsies were included: 213 Al-treated and 101 controls. The demographics were well balanced between both group (Supplementary table 7). In summary, 94.0% of the tumours were ductal, 46.0% were grade 2 and 50% grade 3. At surgery 54.5 had a tumour diameter between 2 and 5 cm and 49.4% had positive nodal status. Sixty-point two percent of patients received adjuvant chemotherapy and trastuzumab, and 7.3% chemotherapy without trastuzumab. Thirty-two point five per cent of patients did not receive adjuvant trastuzumab or chemotherapy mainly due to elderly age. Almost all patients received adjuvant ET (97.8%).

[256]Changes of intrinsic subtypes induced by short perioperative Aromatase Inhibitors

[257]In the treated samples most luminal B (59/75; 79%) and at a few HER2-E (13/91; 14%) BC tumours were reclassified as luminal A in early response to Al: 73% (82/111) GR or IR in H-L-IS compared to 28% in H-H-IS (28/99); Chi 48.56, p<0.0001). Basal-like intrinsic subtype tumours at baseline were not re-assigned to another subtype (Figure 7A, Supplementary figure 6A). Al sensitive tumours (defined by a higher reduction on the proliferation score from baseline to surgery) had correlation coefficient scores at surgery that were closer to Luminal B (Supplementary figure 7A). As expected, controls did not change significantly their Ki67 levels nor their intrinsic subtype from baseline to surgery (Figure 7B). Only 15 out of 101 control cases were re-classified to another intrinsic subtype - mostly due to borderline correlation coefficients at baseline or at surgery (Supplementary table 8 and supplementary figure 7B). [258]To investigate whether those patients with BC tumours being re-classified as lower risk subtype were associated with higher sensitivity to Al, we have assessed the correlation of changes from a 'highly proliferative' intrinsic subtype towards a lower proliferative intrinsic subtype with response to Al measured by Ki67 in treated patients. In treated samples, Ki67 was significantly different between baseline and surgery in all intrinsic subtype risk change categories, however in LL and HL almost all tumours decreased Ki67, while in nonHL there were both increase and decrease in Ki67 (Supplementary figure 6C). As expected, no significant changes in the Ki67 were found in the controls (Supplementary figure 6D). Just one patient with BC within the control arm was classified as L-H intrinsic subtype change.

[259]We aimed to correlate changes of intrinsic subtype towards a lower risk subtype with KI67 endpoint categories (Ki6?2wk high vs low, and Ki67 changes categories: GR, IR and PR). As expected, those H-L intrinsic subtype changing tumours showed more sensitivity to endocrine therapy (Supplementary table 9). These findings suggest that assessing the intrinsic subtype at both timepoints (baseline and surgery) might be useful as a predictor of response to Al.

[260]Changes of single gene expression profiles induced by short perioperative Aromatase Inhibitors.

[261]Significant changes on gene expression between baseline and surgery in treated samples were mainly seen in genes involved in ER- signaling and proliferation (Figure 8a and supplementary figure 8A).

[262]As expected, more and larger magnitude of gene expression changes was observed in treated samples compared to controls (Figure 8b and supplementary figure 8B). However, some genes such FOS, SPRY1, JUN and NR4A1, involved in cancer spread were also upregulated in both treated and control samples. However, some genes were exclusively downregulated in controls such as JAK2 and TNF. Both are associated with tumour infiltrated lymphocytes and the observed different expression might be due to the differences between core and excision sampling. Our group had previously identified the upregulation of EOS and JUN and related genes expression in both treated and control samples in ER+/HER2- BC patients as an artefactual effect resulted from preanalytical sample processing due to handling methodology ²⁹ ' ³⁰ . However, this effect had not been previously described in HER2+ BC. Based on our previous data, some of the changes observed in the control arm might be attributed to an artifactual effect (sample manipulation).

[263]A higher number of gene expression changes and with a higher magnitude was observed in GR compared to PR (figure 9), and also in Luminal tumors compared to HER2-enriched, probably due to the higher sensitivity to Al treatment leading to a higher impact on molecular characteristics. Overall, a higher decrease in proliferation and ER- related gene expression was associated with response to Al. We had also expected significantly different changes in immune related genes between responders and non-responders - not particularly seen in this analysis.

[264]Some genes showed significant changes in both GR and PR treated tumours, all involved in ER signalling and proliferation such as IGF1R, FGFR3, PGR, BIRC5. Only few genes were downregulated exclusively in PR such as EREG (epiregulin), SERPINB5 (serpin Family B member 5), CXCL5 (C-X-C Motif Chemokine Ligand 5) and KRT6B (keratin 6B). These molecular changes that happen after just 2 weeks of Al treatment might be key on resistance to endocrine therapy and might be a useful tool to predict early to treatment.

[265]We then analysed gene expression changes in treated samples adjusted by the changes observed within the controls.

[266]Changes of BC360 signatures expression induced by short perioperative Aromatase Inhibitors

[267]After 2wk-AI, there was a significant upregulation of tumour- related immunity signatures such as TIGIT, CD8 T-cells, inflammatory chemokines and IDO1, while proliferation, HRD, TP53 and ER-signaling were downregulated. In controls there was also a lower magnitude upregulation of TIGIT and CD8T-cells and an exclusive upregulation of hypoxia and downregulation of macrophages and PDL1 /Figure 10).

[268]Out of the treated patients, 62 (29.5%) were GR and 100

(48.6%) were PR. Similar to what was observed at a single gene expression level, the magnitude of changes in gene expression was significantly larger in GR compared to PR with particularly profound downregulation of ER-signalling and proliferation, and upregulation of CD8-T cells, stroma, mammary sternness and increased correlation to LumA (Table 7, Figure 10 and supplementary figure 9).

[269]The heatmap shown in Figure 12 shows changes in gene expression signatures of the novel molecular clusters 1-5 (as defined by the pre-treatment samples) following endocrine therapy. Previous results indicate that cluster 2 is associated with worse outcomes than cluster 1 despite both clusters being enriched for HER2-E subtype. However, the heatmap shows that the change in gene expression of cluster 1 and cluster 2 are comparable. Therefore, baseline samples can be used to predict survival outcomes of ER+HER2+ patients treated with 5 years endocrine therapy, even if these patients may have been treated with some length of endocrine therapy before their surgery. It is noted that surgical samples could be used, however it is often beneficial to use pre-treatment samples in a diagnostic setting to guide treatment. Molecular cluster 4 shows increased expression of immune related signatures and CDK6 expression but reduced expression of proliferation genes. Without wishing to be bound by any particular theory, the present inventors believe that the reduction in Ki67 in these samples is most likely due to Luminal B turning into temporary Luminal A. However, one third of cluster 4 already show early resistance to Al making cluster 4 a candidate for additional treatment such as CDK4/6 inhibitors because they have increased expression of CDK6 and stem cell signatures (mammary sternness). Cluster 3 is the most likely candidate for immunotherapy since the heatmap indicates cluster 3 shows increased expression of immune related signatures and macrophages following endocrine therapy. [270]Impact of the significant molecular changes on survival

[271]To assess the clinical impact of the observed findings, we tested the association of both the surgery expression and the significant features' changes described above with patient survival data.

[272]Lum B intrinsic subtype at surgery with worse TTR compared to Lum A (Figure 11), maintained in the multivariable analysis (Supplementary table 10). By contrast, high expression of TGF-beta in surgery signatures was associated with an independent better TTR (Supplementary table 11). The subgroup of H-L-IS tumours was also significantly associated with better TTR compared to H-H-IS in univariable analysis (supplementary figure 9), but not in the multivariable analysis.

[273]At a single gene level, none of the surgery gene expression were significantly correlated with OS before or after the adjustment of basic clinicopathological factors. None of the gene expression log2FC were significantly correlated with the survival (TTR and OS) before or after the adjustment of basic clinicopathological factors.

[274]To explore further the role of adjuvant HER2-targeted treatment and chemotherapy on survival, we replicated the same analysis but separately for those patients that had received both of these treatments and those that did not (Supplementary figure 10).

[275]In the group of patients treated with adjuvant trastuzumab treatment, surgery LumB was significantly associated with poor survival (TTR), compared with LumA. There was not significant difference in survival TTR among three categories of subtype assignment change (LL/HL/nonHL). However, within those that had not been treated with adjuvant trastuzumab treatment, surgery HER2-E was significantly associated with poor survival (TTR), compared with LumA or Luminals and the nonHL subtype assignment change category was significantly associated with poor survival TTR compared with the HL category. [276]DISCUSSION:

[277]Most LumA tumours remain LumA after 2 weeks of Al. Since Lum A at baseline shows better outcome it is probable there is no impact on re-assessing intrinsic subtype in these cases. However, there is also a decrease in their proliferation score highlighting the relevance of measuring that score in LumA.

[278]Most LumB tumours turn into LumA (78%) after 2 weeks of Al, similar figures to those of LumB turning into Lum A in the POETIC&NeoAI (90%). Thus, similar sensitivity to Al in Lum B tumour despite HER2 status.

[279]However, only 14% of HER2E tumours are re-classified to LumA after 2 weeks of Al in ER+/HER2+ tumours. This is a significantly lower percentage compared to the 50% of tumours changing from HER2E towards LumA observed in the POETIC ER+/HER2- cohort. This fact suggests a higher resistance to Al within HER2E according to HER2 status - thus, there might be molecular differences between HER2-E according the clinical subtype (it is the HER2-status) they belong to.

[280]The assessment of the correlation coefficient scores to the different intrinsic subtypes is necessary to understand the behaviour of changes of intrinsic subtype. On the one hand, AT sensitive tumours (defined by a higher reduction on the proliferation score from baseline to surgery had correlation coefficient scores at surgery that were closer to Luminal B. In addition, control cases that were re-classified to another intrinsic subtype were mostly due to borderline correlation coefficients at baseline or at surgery.

[281]Tumour transcriptional profiles undergo significant changes in response to just 2wk Al and less changes in PR. Identifying patients with PR after 2wk for future studies pairing them with appropriate additional treatments to improve sensitivity is needed. Patients with ER+/HER2+ BC with Al sensitive HER2-E or Luminal B (i.e. showing early changes in IS) are associated with better TTR, suggesting the need of assessing IS after 2-wk Al. The independent prognostic value of high levels of TGF-beta at surgery for better TTR and LumB IS for a very high risk of relapse deserve further investigation on their use to select patients for de-scalated and scalated therapies, respectively.

[282]Epiregulin (EREG) is a mediator of early stage breast tumorigenesis. Prior data using both cell lines and human samples suggest that EREG contributes to the formation of early stage cancer. Understanding tumour behaviour involving this complex pathway could ultimately lead to the development of novel biomarkers of neoplastic progression (M. Faroouqi)

[283]CXCL5 is a member of CXC-type chemokine family and plays an important role in tumorigenesis and cancer progression. Previous studies have shown that CXCL5 is associated with advanced tumour stages, local invasion and metastatic progression. A recent metaanalysis suggests that CXCL5 could be a potential prognostic biomarker for patients with cancer.

[284]Conclusions:

[285]Tumour transcriptional profiles undergo significant changes in response to just 2wk Al and less changes in PR. Identifying patients with PR after 2wk for future studies pairing them with appropriate additional treatments to improve sensitivity is needed. Patients with ER+/HER2+ BC with Al sensitive HER2-E or Luminal B (i.e. showing transcriptional changes in IS) are associated with better TTR, suggesting the need of assessing IS after Al. The independent prognostic value of high levels of TGF-beta at surgery for better TTR and LumB IS for worse TTR deserve further investigation on their use to select patients for de-scalated and scalated therapies, respectively.

-oOo-

[286]All references cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety.

[287]The specific embodiments described herein are offered by way of example, not by way of limitation. Any sub-titles herein are included for convenience only, and are not to be construed as limiting the disclosure in any way.

References

1. Slamon DJ, Clark GM, Wong SG, et al: Human breast cancer: correlation of relapse and survival with amplification of the HER- 2/neu oncogene. Science 235:177-82, 1987

2. Slamon DJ, Leyland-Jones B, Shak S, et al: Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med 344:783-92, 2001

3. Perez EA, Romond EH, Suman VJ, et al: Trastuzumab plus adjuvant chemotherapy for human epidermal growth factor receptor 2- positive breast cancer: planned joint analysis of overall survival from NSABP B-31 and NCCTG N9831. J Clin Oncol 32:3744-52, 2014

4. Cameron D, Piccart-Gebhart MJ, Gelber RD, et al: 11 years' follow-up of trastuzumab after adjuvant chemotherapy in HER2- positive early breast cancer: final analysis of the HERceptin Adjuvant (HERA) trial. The Lancet 389:1195-1205, 2017

5. Prat A, Pascual T, De Angelis C, et al: HER2-Enriched Subtype and ERBB2 Expression in HER2-Positive Breast Cancer Treated with Dual HER2 Blockade. J Natl Cancer Inst 112:46-54, 2020

6. Cejalvo JM, Pascual T, Fernandez-Martinez A, et al: Clinical implications of the non-luminal intrinsic subtypes in hormone receptor-positive breast cancer. Cancer Treat Rev 67:63-70, 2018

7. Prat A, Perou CM: Deconstructing the molecular portraits of breast cancer. Mol Oncol 5:5-23, 2011

8. Prat A, Baselga J: The role of hormonal therapy in the management of hormonal-receptor-positive breast cancer with coexpression of HER2. Nat Clin Pract Oncol 5:531-42, 2008

9. Martin LA, Ribas R, Simigdala N, et al: Discovery of naturally occurring ESRI mutations in breast cancer cell lines modelling endocrine resistance. Nat Commun 8:1865, 2017 10. Belachew EB, Sewasew DT: Molecular Mechanisms of Endocrine Resistance in Estrogen-Receptor-Positive Breast Cancer. Frontiers in Endocrinology 12, 2021

11. Smith I, Robertson J, Kilburn L, et al: Long-term outcome and prognostic value of Ki67 after perioperative endocrine therapy in postmenopausal women with hormone-sensitive early breast cancer (POETIC): an open-label, multicentre, parallel-group, randomised, phase 3 trial. Lancet Oncol 21:1443-1454, 2020

12. Pinhel IF, Macneill FA, Hills MJ, et al: Extreme loss of immunoreactive p-Akt and p-Erkl/2 during routine fixation of primary breast cancer. Breast Cancer Res 12:R76, 2010

13. Parker JS, Mullins M, Cheang MC, et al: Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 27:1160-7, 2009

14. Dowsett M, Ebbs SR, Dixon JM, et al: Biomarker changes during neoadjuvant anastrozole, tamoxifen, or the combination: influence of hormonal status and HER-2 in breast cancer— a study from the IMPACT trialists. J Clin Oncol 23:2477-92, 2005

15. Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98:5116-21, 2001

16. Gu Z, Eils R, Schlesner M: Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics 32:2847-9, 2016

17. Perou CM, Sorlie T, Eisen MB, et al: Molecular portraits of human breast tumours. Nature 406:747-52, 2000

18. Cancer Genome Atlas N: Comprehensive molecular portraits of human breast tumours. Nature 490:61-70, 2012

19. File D, Curigliano G, Carey LA: Escalating and De- escalating Therapy for Early-Stage HER2-Positive Breast Cancer. Am Soc Clin Oncol Educ Book 40:1-11, 2020 20. von Minckwitz G, Huang C-S, Mano MS, et al: Trastuzumab Emtansine for Residual Invasive HER2-Positive Breast Cancer. New England Journal of Medicine 380:617-628, 2018

21. von Minckwitz G, Procter M, de Azambuja E, et al: Adjuvant Pertuzumab and Trastuzumab in Early HER2-Positive Breast Cancer. New England Journal of Medicine 377:122-131, 2017

22. Gianni L, Pienkowski T, Im YH, et al: 5-year analysis of neoadjuvant pertuzumab and trastuzumab in patients with locally advanced, inflammatory, or early-stage HER2-positive breast cancer (NeoSphere): a multicentre, open-label, phase 2 randomised trial. Lancet Oncol 17:791-800, 2016

23. Dunbier AK, Ghazoui Z, Anderson H, et al: Molecular profiling of aromatase inhibitor-treated postmenopausal breast tumors identifies immune-related correlates of resistance. Clin Cancer Res 19:2775-86, 2013

24. Keung MY, Wu Y, Badar F, et al: Response of Breast Cancer Cells to PARP Inhibitors Is Independent of BRCA Status. Journal of Clinical Medicine 9:940, 2020

25. Baker L, Quinlan PR, Patten N, et al: p53 mutation, deprivation and poor prognosis in primary breast cancer. British Journal of Cancer 102:719-726, 2010

26. Roman-Rosales AA, Garcia-Villa E, Herrera LA, et al: Mutant p53 gain of function induces HER2 over-expression in cancer cells. BMC Cancer 18:709, 2018

27. Dowsett M, Smith IE, Ebbs SR, et al: Proliferation and apoptosis as markers of benefit in neoadjuvant endocrine therapy of breast cancer. Clin Cancer Res 12:1024s-1030s, 2006

28. Fougner C, Bergholtz H, Norum JH, et al: Re-definition of claudin-low as a breast cancer phenotype. Nat Commun 11:1787, 2020

29. Bergamino MA, Morani G, Parker J, et al: Impact of

Duration of neoadjuvant aromatase inhibitors on molecular expression profiles in estrogen receptor-positive breast cancers. CCR 28:1217- 1228, 2022.

30. Gao Q, Lopez-Knowles E, Cheang MCU, et al: Major impact of sampling methodology on gene expression in estrogen receptor- positive breast cancer. JNCI Cancer Spectr 2:pky005, 2018