Breast cancer classification and treatment

Abstract

The present invention relates to methods of obtaining an indication of the prognosis of a breast cancer subject, for classifying breast tumours, and related methods. The methods are based on the production of a signature score which is derived from normalised expression levels of a plurality of specific breast tumour stromal protein or tumour stromal RNA biomarkers.

Description

[0001] 489.167562 / 01

[0002] BREAST CANCER CLASSIFICATION AND TREATMENT

[0003] The present invention relates to methods of obtaining an indication of the prognosis of a breast cancer subject, for classifying breast tumours, and related methods. The methods are based on the production of a signature score which is derived from normalised expression levels of a plurality of specific breast tumour stromal protein or tumour stromal RNA biomarkers.

[0004] Breast cancers are complex systems of tumour cells and microenvironment (TME) components. During recent years, large-scale gene expression and genetic analyses with follow-up information have provided a basis for molecular classification. The following main subtypes, reflecting tumour cell phenotypes, are used in clinical practice: luminal-A, luminal-B, HER2 enriched, and triple negative tumours [1], It is well known that heterogeneity is also present within these established categories [2, 3], Lately, a few groups have investigated human breast cancer using global proteomics data [4-7], These reports have indicated that the correlation between proteome expression profiles and standard molecular subtypes is incomplete.

[0005] Studies of human tumours, such as breast cancer, have mostly used whole tissues for genetic or transcriptomic interrogation, with no separation of the various compartments [1, 2, 8, 9], Here, we studied the breast cancer proteome including cellular and secreted proteins by using separate samples of laser micro-dissected epithelial and stromal tumour tissues. We aimed to perform a stromal proteomic mapping with particular attention to differences between low-grade (luminal-like) and high-grade (basal-like) tumours. We asked whether TME-based protein profiles might capture additional clinical and prognostic information independent of the current epithelial-based classification. Our data suggest that the stromal proteome can improve breast cancer stratification, in particular among low-grade (luminal-like) tumours.

[0006] In this study, we aimed to explore the global proteome of the breast cancer microenvironment. We initially focused on differences between low-grade (luminal-like) and

[0007] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxhigh-grade (basal-like) tumours in terms of clinical correlates and prognostic impact, and we wanted to examine whether stromal-based proteomic information could improve patient stratification beyond current epithelial-based tumour classification. By laser capture microdissection and separation of the epithelial and stromal tumour compartments combined with MS-based proteomics, we found that a stromal protein signature (35P) provided independent prognostic impact and thus appears to be needed for precise tumour stratification. This indicates that molecular classification should also incorporate TME-based information, not only tumour cell characteristics. By extension, our findings support the concept of combined epithelial and stromal directed therapy following integrated diagnostic and prognostic profiling of these two compartments.

[0008] Only a very few studies have previously combined microdissection of human breast cancer with MS-based proteomics [30, 31], To our knowledge, this is the first such study of the stromal proteome to focus on differences between high-grade and low-grade tumours linked with clinically-relevant information. Notably, our data indicate that the epithelial proteome clustered in accordance with the current breast cancer classification. This was in contrast to the stromal samples, which showed three subgroups by unsupervised clustering, across epithelial-based tumour subtypes. This indicates that additional information might be extracted from the stromal compartment.

[0009] By using our approach, we identified the 35P stromal-based signature that could separate patients based on survival in several independent cohorts. The 35P signature showed independent prognostic influence by multivariate survival analysis, in contrast to the molecular subtype of the samples (basal-like versus luminal like). Notably, the most low-grade breast cancers, the luminal A subgroup, could be split prognostically by the 35P stromal signature. These findings support the clinical significance of TME information in tumour stratification.

[0010] The present data indicates that the epithelial and stromal compartments do not always parallel each other. Some discordant luminal A cases, assumed to be low-grade with an excellent prognosis, might develop a more “high-grade” stroma and show worse prognosis than expected, as revealed by our survival analyses. It is known that some

[0011] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxluminal A tumours have a worse prognosis than the rest, and our data indicate that the 35P stromal signature is able to identify such cases with independent value. Notably, luminal A patients were represented in all four quartiles (Q1-4) of the 35P score, demonstrating stromal diversity at the proteomic level among luminal A tumours.

[0012] As part of the 35P signature validation, we included mRNA data from the METABRIC Discovery cohort. This may present as a limitation as mRNA and protein expression show variable correlation depending on e.g. location and function. However, as most of the 35P showed significant positive protein-mRNA correlations [6], we consider it appropriate to use these mRNAs as proxies for protein expression. Importantly, 35P showed a significant impact on disease progress when using corresponding mRNA values derived from whole tissue samples. This would support the robustness of the signature, and it also indicates that regular tissues could be applied in a routine setting, without the microdissection step used in our discovery study.

[0013] In a large randomized clinical trial of tamoxifen treatment, the 35P signature was found to be prognostically different in the control arm, although no significant separation was found in the treatment arm, and there was no interaction between the 35P level and effect of tamoxifen treatment. Thus, there was no evidence that 35P, and by that the stromal protein composition, was able to predict the effect of endocrine treatment in this cohort of hormone receptor positive tumours.

[0014] Further characterization of 35P revealed a link to ECM biology, and processes like EMT, hypoxia and angiogenesis were highlighted. These findings indicate a relation between prognostically “high-grade stroma” and processes reflecting increased interaction between tumour cells and their supporting microenvironment (EMT), along with stromal programs associated with tumour progress (hypoxia, angiogenesis). Our data might be relevant for improved identification of more high-grade luminal cases, which could provide clues for better treatment of this subgroup.

[0015] In summary, this study of compartment-based breast cancer proteomes indicates that stromal information can improve tumour classification and prediction of prognosis

[0016] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxbeyond current criteria, in particular among the frequent luminal cases. The 35P signature reflects processes indicating a more permissive microenvironment supporting tumour progress. Thus, the stromal proteome may provide a basis for better patient stratification and treatment opportunities.

[0017] There have been previous attempts to characterize breast cancers at both morphologic and molecular levels. Previously, Oncotype DX (van de Vijver MJ et al. “A geneexpression signature as a predictor of survival in breast cancer”. N Engl J

[0018] Med 2002;347:1999-2009) and PAM50 (Parker JS et al., “Supervised risk predictor of breast cancer based on intrinsic subtypes”. J Clin Oncol 2009; 27:1160-7) have been used to classify breast tumours to inform prognosis and guide treatment. Oncotype DX is based on a panel of 16 cancer-related genes. PAM50 is a 50-gene signature that classifies breast cancer into five molecular intrinsic subtypes: Luminal A, Luminal B, HER2-enriched, Basal-like, and Normal-like. Each of the five molecular subtypes varies by their biological properties and prognoses. Luminal A generally has the best prognosis; HER2-enriched and Basal-like are considered more aggressive diseases.

[0019] However, the PAM50 and Oncotype DX expression signatures focus on the tumour cell compartment. They do not focus the tumour stroma.

[0020] The invention aims to overcome one or more of the above-mentioned problems or limitations by providing prognostic and diagnostic methods based on proteomic patterns of biomarkers which have been obtained from tumour stromal proteomes. It is an object of the invention to provide methods of obtaining indications of the prognosis of breast cancer subjects and for classifying breast tumours.

[0021] In one embodiment, the invention provides a method of obtaining an indication of the prognosis of breast cancer in a subject, the method comprising the step:

[0022] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:

[0023] AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (hereinafter “5P”),

[0024] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxwherein the biomarkers were obtained from a biological sample which was obtained from the subject; and wherein the produced signature score is indicative of the prognosis of breast cancer in the subject.

[0025] In another embodiment, the invention provides a method of classifying breast tumours, the method comprising the steps:

[0026] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; and (b) classifying the breast tumour based on the signature score produced.

[0027] In another embodiment, the invention provides a method of obtaining an indication of the efficacy of a drug which is being used to treat breast cancer in a subject, the method comprising the steps:

[0028] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a first biological sample which was obtained from the subject at a first time point; and

[0029] (b) producing a second signature score using corresponding levels of the corresponding biomarkers in a corresponding second biological sample obtained from the subject at a second (later) time point;

[0030] wherein the drug has been administered to the subject in the interval between the first and second time points, wherein a decrease in the second signature score compared to the first signature score is indicative of the efficacy of the drug, and wherein an increase in the second signature score compared to the first signature score is indicative of the lack of efficacy of the drug.

[0031] In another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0032] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;

[0033] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(b) comparing the signature score with a reference signature score; and

[0034] (c) administering a treatment appropriate for treating breast cancer to the subject if the signature score is above the reference signature score, thereby treating the breast cancer in the subject.

[0035] In another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0036] (a) administering a treatment appropriate for treating breast cancer to the subject, wherein, prior to administration, a signature score which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject, had been determined to be above a reference signature score.

[0037] In another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0038] (a) receiving a signature scope which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; and

[0039] (b) identifying the subject as having a signature score above a reference signature score, thereby providing an indication of the potential efficacy of administering treatment appropriate for treating the breast cancer to the subject; and

[0040] (c) administering treatment appropriate for treating the breast cancer to the subject.

[0041] In another embodiment, the invention provides a method of detecting biomarkers in a breast tissue sample obtained from a human subject, the method comprising measuring:

[0042] (i) a protein expression level for every protein in a group of classifier proteins; or (ii) a mRNA expression level for every gene in a group of classifier genes; wherein the group of classifier proteins or classifier genes consists of only:

[0043] (a) 5P;

[0044] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(b) 18P (as defined herein); or

[0045] (c) 35P (as defined herein).

[0046] In another embodiment, the invention provides an ex vivo method of screening for agents for treating breast cancer, the method comprising the steps:

[0047] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein biomarkers were obtained from a breast cancer sample which has been treated with an agent; (b) producing a second signature score from corresponding biomarkers obtained from the breast cancer sample which has not been treated with the agent; and (c) comparing the first and second signature scores;

[0048] wherein a decrease in the first signature score compared to the second signature score is indicative of an agent which is capable of treating breast cancer.

[0049] In another embodiment, the invention provides a method of predicting the risk of recurrence of breast cancer in a subject who has previously had breast cancer but who is currently in remission, the method comprising the step:

[0050] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein biomarkers were obtained from a biological sample which was obtained from the subject; wherein the signature score is predictive of the risk of recurrence of breast cancer in the subject.

[0051] In another embodiment, the invention provides a method of predicting the therapeutic efficacy of treatment (preferably radiotherapy treatment) on a subject with breast cancer, the method comprising the step:

[0052] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; wherein the produced signature score is predictive of the therapeutic efficacy of the treatment (preferably the radiotherapy treatment) on the breast cancer.

[0053] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxAs used herein, the term “5P” refers to the following biomarkers: AGRN, CTSZ, LSM1, MYL9 and TPSAB1. Of the 5P biomarkers, AGRN, CTSZ, LSM1, MYL9 are upregulated biomarkers; and TPSAB1 is a down-regulated biomarker.

[0054] As used herein, the term “18P” refers to the following biomarkers: AGRN, ATP13A1, TMEM258, CTSZ, EFEMP2, EIF3F, FBF1, IDH2, P3H3, LSM1, MYL9, NID1, PIP, PXDN, TOP1, TPSAB1, VAT1 and VWA1. Of the 18P biomarkers, AGRN, ATP13A1, TMEM258, CTSZ, EFEMP2, EIF3F, IDH2, LSM1, MYL9, NID1, PXDN, TOP1 and VWA1 are upregulated biomarkers; and FBF1, P3H3, PIP, TPSAB1 and VAT1 are down-regulated biomarkers.

[0055] As used herein, the term “35P” refers to the following biomarkers: NID1, VWA1, PXDN, IGHG4, MYL9, LSM1, TMEM258, CTSZ, AGRN, TP53I3, NNMT, IGKV3D-15, HLA-H, ATP13A1, CDV3, SEC24A, EIF3F, TMSB4X, LSM7, TOP1, IGKV3D-20, EFEMP2, CBR1, IDH2, WIPF1, PSMC4, P3H3, FBF1, VAT1, MMP2, ITIH1, CAVIN1, TPSAB1, CMA1 and PIP. Of the 35P biomarkers, NID1, VWA1, PXDN, IGHG4, MYL9, LSM1, TMEM258, CTSZ, AGRN, TP53I3, NNMT, IGKV3D-15, HLA-H, ATP13A1, CDV3, SEC24A, EIF3F, TMSB4X, LSM7, TOP1, IGKV3D-20, EFEMP2, CBR1, IDH2, WIPF1 and PSMC4 are upregulated biomarkers; and P3H3, FBF1, VAT1, MMP2, ITIH1, CAVIN1, TPSAB1, CMA1 and PIP are down-regulated biomarkers.

[0056] In some embodiments of the invention, the method is carried out in vitro or ex vivo.

[0057] In one embodiment, the invention provides a method of obtaining an indication of the prognosis of breast cancer in a subject, the method comprising the step:

[0058] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; wherein the produced signature score is indicative of the prognosis of breast cancer in the subject.

[0059] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxThe subject is preferably a human subject. The human may, for example, be 0-10, 10-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, 80-90, 90-100 or above 100 years old. The human may be one who is suffering from or at risk from a particular disease or disorder, e.g. cancer, preferably breast cancer. In some preferred embodiments, the subject is one who is suffering from or who has previously suffered from cancer, e.g. breast cancer. In some embodiments, the subject is one who has previously been treated for breast cancer, e.g. by surgery and / or chemotherapy and / or radiotherapy.

[0060] A control subject may be defined as a non-diseased subject, a subject without breast cancer, a typically-developed subject or a healthy-aged subject.

[0061] In some embodiments, at least 5 of the biomarkers are from a first group consisting of 5P. These 5 biomarkers have been found to have a significant association value in a METABRIC analysis.

[0062] In some embodiments, the plurality of biomarkers are selected from a second group consisting of 18P. This second group includes all of the first group of biomarkers. In some embodiments, the plurality of biomarkers are selected from at least 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 or 18 biomarkers in this second group. Preferably, the second group includes all of 5P. Preferably, at least 10 biomarkers (in addition to 5P) are selected from this second group. In some embodiments, all 18 biomarkers are selected from this second group.

[0063] In some embodiments, the plurality of biomarkers are selected from a third group consisting of 35P. This third group includes all of the first and second groups of biomarkers. At least 18 biomarkers are selected from this third group. In some embodiments, at least 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34 or 35 biomarkers are selected from this third group. Preferably, the third group includes all of 18P. Preferably, at least 10 biomarkers in addition to 18P are selected from this third group. In some preferred embodiments, all 35 biomarkers are selected from this third group.

[0064] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxThe biomarkers may be protein biomarkers or RNA (preferably mRNA) biomarkers.

[0065] The corresponding protein identifiers of the biomarker genes are given in Table 2.

[0066] References to these biomarkers (whether protein or RNA) include references to naturally-occurring human variants of these biomarkers. In some embodiments of the invention, the method includes the step of selecting the biomarkers.

[0067] In some embodiments, the biological sample is a sample of blood, serum or plasma. Serum and plasma may be obtained from a blood sample from the subject, wherein the blood cells have been removed. In such cases, the biomarkers are proteins.

[0068] In other embodiments, the biological sample is a whole tissue breast tumour sample. This contains a mixture of tumour cells and tumour stroma.

[0069] In some preferred embodiments, the biological sample is obtained by needle biopsy of the breast cancer or from a surgical specimen (resection) from the breast cancer. In such cases, the biomarkers are proteins or mRNA, preferably mRNA.

[0070] In some embodiments, the biological sample is tumour stroma. The term “tumour stroma” relates to the non-malignant cells and the extracellular matrix which are present in the tumour microenvironment. The stroma comprises a variable portion of the entire tumour: up to 90% of a tumour may be stroma, with the remaining 10% as cancer cells. Many types of cells are present in the stroma, but four abundant types are fibroblasts, T-cells, macrophages and endothelial cells. Tumour stroma may be obtained, for example, by laser capture microdissection.

[0071] In yet other embodiments, the biological sample may be saliva or urine.

[0072] The biological sample is a sample which is obtained or which has previously been obtained from the subject. In some embodiments, the method additionally comprises the step of obtaining one or more biological samples from the subject.

[0073] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxIn some embodiments, the biomarkers are either all proteins or all RNA (e.g. mRNA). In other embodiments, the biomarkers are a mixture of protein and RNA (e.g. mRNA). Preferably, the biomarkers are either all protein or all RNA (e.g. mRNA).

[0074] In some embodiments, the biomarkers are proteins. The proteins may be obtained from the biological sample by any suitable method. The protein biomarkers may, for example, be identified by MS analysis or shotgun proteomics analysis.

[0075] In other embodiments, the biomarkers are RNA, preferably mRNA. RNA may be extracted by any suitable method. The RNA biomarkers may, for example, be identified by RNASeq or qRT-PCR.

[0076] The signature score is a numerical value which is representative of the overall (normalised) expression levels (either protein expression levels or (m)RNA expression levels) of the selected biomarkers in the biological sample. Normalisation of each of the biomarker levels is generally necessary in order to obtain an accurate interpretation of the signature score. The normalisation step may be performed using any suitable method. Numerous such methods are known in the art.

[0077] In some embodiments, the levels of each selected biomarkers are normalised against the levels of one or more reference genes / proteins which are expressed in the selected biological sample, preferably control or housekeeping genes / proteins which have low variability in their expression levels in the selected biological sample (e.g. in blood or in breast cancer tissue).

[0078] The levels of the one or more reference genes / proteins are levels which are or have been obtained from the selected biological sample (i.e. the actual same sample from which levels of the biomarkers are obtained). The levels of the one or more reference genes may be determined by measuring the level of the corresponding expressed RNA, preferably mRNA.

[0079] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxFor example, Tilli et al. (BMC Genomics (2016) 17:639) identified a set of control genes including CCSER2, SYMPK, ANKRD17 and PUM1. These were found to be usable in the clinical analyses of breast cell lines and tissue samples. The levels of each of the biomarkers used in the invention may be normalised against the levels of one or more of the latter control genes in the biological sample.

[0080] In another example, the Oncotype DX 21 -gene test uses 5 housekeeping genes to normalize their 16 cancer-related genes. These 5 housekeeping genes are ACTB, GAPDH, GUS, RPLPO and TFRC. The levels of each of the biomarkers used in the invention may be normalised against the levels of one or more of the latter control genes / proteins in the biological sample.

[0081] In another example, the Prosigna® Breast Cancer Prognostic Gene Signature Assay uses 8 housekeeping genes to normalize their 50 (PAM50) cancer-related genes. These 8 housekeeping genes are ACTB, MRPL19, PUM1, SF3A1, GUSB, PSMC4, RPLPO and TFRC. The levels of each of the biomarkers used in the invention may be normalised against the levels of one or more of these control genes / proteins in the biological sample.

[0082] In some embodiments, the normalisation step involves subtracting the level of one or more reference genes or proteins from the obtained level of each selected biomarker.

[0083] In other embodiments, the normalisation step involves dividing the obtained level of each selected biomarker by the level of one or more reference genes or proteins.

[0084] The signature score is preferably produced by calculating the sum of the normalised expression levels of the up-regulated biomarkers (e.g. proteins or genes), and then subtracting the sum of the normalised expression levels for the down-regulated biomarkers (e.g. proteins or genes) in accordance with Equation 1:

[0085] Equation 1. Signature score =

[0086]

[0087] iEU jED

[0088] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxwherein

[0089] - Xirepresents the normalized expression level of an up-regulated biomarker in the set of up-regulated biomarkers, U.

[0090] - X represents the normalized expression level of a down-regulated biomarker j in the set of down-regulated biomarkers, D.

[0091] The first summation computes the total expression of all up-regulated biomarkers. The second summation computes the total expression of all down-regulated biomarkers. The up-regulation and down-regulation of the biomarkers are defined herein in accordance with Table 2. In some embodiments of the invention, the expression level of each protein / gene does not have to be zero-mean normalized.

[0092] A weighting may be added to or multiplied to one or more of the normalised biomarker levels before those levels are summed (e.g. (k1 x BM1) + (k2 x BM2) +..., where k1 and k2 are independently numbers which may be the same or different, and BM1 and BM2 are the determined levels of two of the biomarkers).

[0093] In embodiments of the invention wherein more than one signature scores are compared, the signature scores are all produced by the same method.

[0094] In embodiments of the invention which refer to “corresponding biomarkers”, this is referring to the same biomarkers as the previously-mentioned biomarkers. For example, if the first signature score is produced using ten of the 18P biomarkers, then the second signature score is also produced using the same ten 18P biomarkers.

[0095] In embodiments of the invention which refer to “corresponding biological samples”, this is referring to the same biological samples as the previously-mentioned biological samples. For example, if the first signature score is produced from mRNA, then the second signature score is also produced using mRNA.

[0096] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxAs used herein, the term “reference signature score” or “corresponding reference signature score” refers to a signature score which has been produced using the same (i.e. corresponding) parameters as the signature score to which it is being compared, (e.g. the same type of biomarkers (e.g. protein or RNA), the same number of biomarkers (e.g. 35), the same set of biomarkers (e.g. 35P) from the same type of biological sample (e.g. breast tumour stroma) and using the same normalisation steps) wherein the biomarkers for the reference signature score were obtained from control (e.g. healthy) subjects (and not from the subject with breast cancer). Thus the reference signature score provides a baseline from a control subject against which to compare the subject with breast cancer’s signature score.

[0097] In some embodiments of the invention, the signature score is indicative of the prognosis of breast cancer in the subject. A comparison of the signature score from the subject to that of a corresponding reference signature score provides an indication of the prognosis of breast cancer in the subject, the likely outcome or course of the breast cancer in the subject or the chance of recovery of the subject.

[0098] As used herein, the term "is indicative of the prognosis of breast cancer in the subject” means that there is a negative correlation between the signature score and a good prognosis of breast cancer in that subject. Consequently, a signature score from the subject which is higher than a corresponding reference signature score (e.g. from a healthy control subject or a control subject without breast cancer) is indicative of an increased likelihood or statistically-significant chance (where the difference is significant) of the subject having a poor prognosis for breast cancer. The reference signature score may also be one which has been obtained from a subject having breast cancer but with a good prognosis. The reference signature score may also be one which has been obtained from a cohort of subjects having low-grade breast cancers.

[0099] In this case, the extent of the difference between the signature score from the subject and the reference signature score (e.g. from a healthy control subject or a control subject without breast cancer) provides an indication of the degree of the poor prognosis of the subject.

[0100] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxFurthermore, a signature score from the subject which is lower than a corresponding reference signature score (e.g. from a breast tumour stroma sample from a breast cancer subject) is indicative of an increased likelihood or statistically-significant chance (where the difference is significant) of the subject having a good prognosis for breast cancer.

[0101] The method may comprise the additional step of administering a treatment appropriate for treating the breast cancer to the subject if the produced signature score is indicative of the subject having a poor prognosis for breast cancer.

[0102] In another embodiment, the invention provides a method of classifying breast tumours, the method comprising the steps:

[0103] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; and (b) classifying the breast tumour based on the signature score produced.

[0104] The biomarkers are selected and the signature score is produced as disclosed herein. The biological sample is one as disclosed herein. The levels of the selected biomarkers are normalised as disclosed herein.

[0105] In this embodiment of the invention, the obtained signature score is compared against a corresponding panel of reference signature scores or set of ranges of references signature scores which have (previously) been obtained from tissues which are representative of different breast tumours having different phenotypes or genotypes or other physical properties; and classifying the breast tumour based on which reference signature score is closest to the obtained signature score, or into which range of reference scores the obtained signature score falls.

[0106] In some embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour on the basis of breast tumour type. The breast tumour may be any type of breast tumour, e.g. Luminal A (LumA), Luminal B

[0107] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(LumB), Basal-like (Basal), HER2-enriched (HER2) breast tumour or Normal-like.

[0108] Preferably, the breast tumour is a Luminal A breast tumour.

[0109] In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of different breast tumour types; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0110] In some embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the breast tumour as having a basal-like phenotype or not. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of breast tumours having a basal-like phenotype or not; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0111] In other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the breast tumour as having a luminal-like phenotype or not, e.g. Luminal-A or Luminal B, or not. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of breast tumours having a luminal-like phenotype or not; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0112] In other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour on the basis of the tumour’s size. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of different breast tumour sizes; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0113] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxIn other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour on the basis of histologic grade. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of different breast tumour histologic grades; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0114] In other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour on the basis of its likelihood of having lymph node metastases. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of breast tumours having lymph node metastases or not; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0115] In other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour as being ER negative or not. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of breast tumours which are ER negative or not; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0116] In other embodiments, the term “classifying the breast tumour based on the signature score obtained” refers to classifying the tumour on the basis of as having high levels of tumour cell proliferation. In this embodiment, the obtained signature score is compared against a corresponding panel of reference signature scores which have (previously) been obtained from tissues which are representative of different breast tumours having high levels of tumour cell proliferation or not; and classifying the tumour based on which reference signature score is closest to the obtained signature score.

[0117] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxIn another embodiment, the invention provides a method of obtaining an indication of the efficacy of a drug which is being used to treat breast cancer in a subject, the method comprising the steps:

[0118] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a first biological sample which was obtained from the subject at a first time point; and

[0119] (b) producing a second signature score using corresponding levels of the corresponding biomarkers in a corresponding second biological sample obtained from the subject at a second (later) time point;

[0120] wherein the drug has been administered to the subject in the interval between the first and second time points, wherein a decrease in the second signature score compared to the first signature score is indicative of the efficacy of the drug, and wherein an increase in the second signature score compared to the first signature score is indicative of the lack of efficacy of the drug.

[0121] In all methods of the invention, the increase and / or the decrease is preferably a significant one. Significance may be measured, for example, using Student’s t-test, with a p-value significance threshold set to 0.05.

[0122] In another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0123] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;

[0124] (b) comparing the signature score with a reference signature score (e.g. one from a healthy control subject or from a subject without breast cancer); and

[0125] (c) administering a treatment appropriate for treating breast cancer to the subject if the signature score is above the reference signature score, thereby treating the breast cancer in the subject.

[0126] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxIn another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0127] (a) administering a treatment appropriate for treating breast cancer to the subject, wherein, prior to administration, a signature score which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers were 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject, had been determined to be above a reference signature score (e.g. one from a healthy control subject or from a subject without breast cancer).

[0128] In yet another embodiment, the invention provides a method of treating breast cancer in a subject, the method comprising the steps of:

[0129] (a) receiving a signature scope which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; and

[0130] (b) identifying the subject as having a signature score above a reference signature score (e.g. one from a healthy control subject or from a subject without breast cancer), thereby providing an indication of the potential efficacy of administering treatment appropriate for treating the breast cancer to the subject; and

[0131] (c) administering treatment appropriate for treating the breast cancer to the subject.

[0132] In all embodiments of the invention, the method may comprise the additional step of administering a treatment appropriate for treating the breast cancer to the subject.

[0133] Treatments for breast cancer are well known in the art, including treatment with surgery, which may be followed by chemotherapy or radiation therapy, or both. For example, the following list includes some of the commonly-used adjuvant chemotherapy for breast cancer:

[0134] CMF: cyclophosphamide, methotrexate, and 5-fluorouracil.

[0135] FAC (or CAF): 5-fluorouracil, doxorubicin, cyclophosphamide.

[0136] AC (or CA): Adriamycin (doxorubicin) and cyclophosphamide.

[0137] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxAC-Taxol: AC followed by paclitaxel (Taxol).

[0138] TAC: Taxotere (docetaxel), Adriamycin (doxorubicin), and cyclophosphamide.

[0139] FEC: 5-fluorouracil, epirubicin and cyclophosphamide.

[0140] AT: Adriamycin (doxorubicin) and Taxotere (docetaxel).

[0141] In yet a further embodiment, the invention provides a method of detecting biomarkers in a breast tissue sample obtained from a human subject, the method comprising measuring:

[0142] (i) a protein expression level for every protein in a group of classifier proteins; or (ii) a mRNA expression level for every gene in a group of classifier genes; wherein the group of classifier proteins or classifier genes consists of only:

[0143] (a) 5P;

[0144] (b) 18P; or

[0145] (c) 35P.

[0146] In yet another embodiment, the invention provides an ex vivo method of screening for agents for treating breast cancer, the method comprising the steps:

[0147] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein biomarkers were obtained from a breast cancer sample which has been treated with an agent; (b) producing a second signature score from corresponding biomarkers obtained from the breast cancer sample which has not been treated with the agent; and (c) comparing the first and second signature scores;

[0148] wherein a decrease in the first signature score compared to the second signature score is indicative of an agent which is capable of treating breast cancer.

[0149] The breast cancer sample may be a breast cancer cell line, e.g. MCF-7 or MDA-MB-231; or a sample (e.g. tissues or cells) of a breast cancer from a subject (which may be used in the form of a cell line, spheroid or organoid).

[0150] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxAgents which are identified as being capable of treating breast cancer on the basis of samples of breast cancer from a subject may then be formulated for administration to the subject, and then optionally administered to the subject.

[0151] In yet another embodiment, the invention provides a method of predicting the risk of recurrence of breast cancer in a subject who has previously had breast cancer but who is currently in remission, the method comprising the step:

[0152] (a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein biomarkers were obtained from a biological sample which was obtained from the subject; wherein the signature score is predictive of the risk of recurrence of breast cancer in the subject.

[0153] The biomarkers are selected and the signature score is produced as disclosed herein. The biological sample is one as disclosed herein. The levels of the selected biomarkers are normalised as disclosed herein.

[0154] As used herein, the term "is predictive of the risk of recurrence of breast cancer in the subject” means that there is a positive correlation between the signature score and the risk of recurrence of breast cancer in the subject. In particular, a signature score from the subject which is higher than a corresponding reference signature score (e.g. from a healthy control subject or a control subject without breast cancer) is indicative of an increased likelihood or statistically-significant chance (where the difference is significant) of recurrence of breast cancer in the subject. In particular, a signature score from the subject which is lower than a corresponding reference signature score (e.g. from a breast tumour sample from a breast cancer subject) is indicative of a decreased likelihood or statistically-significant chance (where the difference is significant) of recurrence of breast cancer in the subject.

[0155] In yet another embodiment, the invention provides a method of predicting the therapeutic efficacy of treatment (preferably radiotherapy treatment) on a subject with breast cancer, the method comprising the step:

[0156] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are 5P, and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; wherein the produced signature score is predictive of the therapeutic efficacy of the treatment (preferably the radiotherapy treatment) on the breast cancer.

[0157] The biomarkers are selected and the signature score is produced as disclosed herein. The biological sample is one as disclosed herein. The levels of the selected biomarkers are normalised as disclosed herein. A produced signature score from the subject which is higher than a corresponding reference signature score is indicative of an increased likelihood or statistically-significant chance of the subject having a shorter survival time after radiotherapy treatment.

[0158] In yet another embodiment, the invention provides a kit comprising reagents sufficient for the detection and / or quantitation of expression of each the following biomarker genes:

[0159] (i) 5P

[0160] (ii) 18P; or

[0161] (iii) 35P,

[0162] characterised in that said reagents comprise a plurality of forward and reverse primers pairs, wherein said forward and reverse primers pairs are selected from forward and reverse primer pairs which are capable of identifying expression of each the following genes:

[0163] (i) 5P

[0164] (ii) 18P; or

[0165] (iii) 35P.

[0166] Preferably, in all methods of the invention, the method steps are carried out (one after the other) in the order specified.

[0167] The disclosure of each reference set forth herein is specifically incorporated herein by reference in its entirety.

[0168] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxBRIEF DESCRIPTION OF THE FIGURES

[0169] Figure 1. Survival analysis of CPTAC-TCGA and METABRIC-Discovery cohorts stratified by 35P stromal signature. (A-B) Distribution of 35P scores (quartiles) across PAM50 subtypes in TCGA CPTAC and METABRIC. (C) High 35P signature score (Q4) predicted poorer survival in the CPTAC-TCGA cohort (log-rank<0.001; overall survival; luminal-like and basal-like patient samples included, n=87). (D-F) High 35P score was associated with worse breast cancer specific survival in the METABRIC Discovery cohort (p<0.001; all patients, n=852), and luminal A subtype (p=0.001; n=466; Q4 n=40; Q1-3 n=426), but not in luminal B (n=268; Q4 n=69; Q1-Q3 n=199). P-values were calculated using log-rank test. CPTAC - Clinical Proteomic Tumor Analysis Consortium; METABRIC - Molecular Taxonomy of Breast Cancer International Consortium; TCGA -The Cancer Genome Atlas.

[0170] Figure 2. The 35P signature correlates to features (by gene expression signatures) associated with high-grade tumours. The 35P stromal signature scores were plotted against scores from various gene expression signatures associated with high grade tumours. Significant positive correlations were seen for EMT (A), angiogenesis (B-C), sternness (D), hypoxia (E-G) and stromal inflammation (H).

[0171] Figure 3. Survival analyses of the 35P stromal signature. Kaplan-Meier plots were generated by KMPIotter with 35P-stromal signature as input, and by separating patients by upper quartile (Q4; dashed line) versus the rest (Q1-Q3; solid line). Recurrence-free survival in all patients (no restrictions) included 2032 patients (A). The patients were stratified into luminal A (B), luminal B (C), and basal-like (D) subtypes (by PAM50). The analyses were truncated at 10-years follow-up.

[0172] Figure 4. Long-term survival of luminal-like breast cancer patients from the STO trials control group (A, B) and tamoxifen treatment group (C, D), grouped by high (solid line) versus low (dashed line) 35P score. The plot shows reduced survival, both DRFI and BCS, in patients with high 35P score from the control group, but not in the Tamoxifen group. DRFI, Distant Recurrence-Free Interval; BCS, Breast Cancer-Specific.

[0173] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxFigure 5. Reduced subsets of PROCAST-35 show predictive power in univariate analyses of the METABRIC Discovery cohort. A) We used a variant of recursive feature elimination where we iteratively removed the protein that contributed the least to the parent signature, as determined by the lowest chi-square value. The x-axis shows, from left to right, which protein that were left out of the parent signature for the next iteration. Two interesting subsets of PROCAST-35 were identified: 18P (PROCAST-18), which showed the strongest predictive power (Chi-square=37.2; Log-rank test p=1E-09), and 5P (PROCAST-5) which consisted of only five proteins with high predictive power (Chi-square=26.4; Log-rank test p=2E-07). B-l) Univariate survival analysis (Kaplan-Meier method) showing the probability of survival for patients from the METABRIC Discovery cohort ranked by the PROCAST-18 and PROCAST-5 signatures. The upper quartile (Q4; high PROCAST-18 score) shows reduced probability of survival in all patients (B) and in patients diagnosed with luminal A breast cancer (C). No difference was seen in luminal B (D) or basal-like (E). The upper quartile (Q4; high PROCAST-5 score) shows reduced probability of survival in all patients (F) and in patients diagnosed with luminal A breast cancer (G). No difference was seen in luminal B (H) or basal-like (I).

[0174] EXAMPLES

[0175] The present invention is further illustrated by the following Examples, in which parts and percentages are by weight and degrees are Celsius, unless otherwise stated. It should be understood that these Examples, while indicating preferred embodiments of the invention, are given by way of illustration only. From the above discussion and these Examples, one skilled in the art can ascertain the essential characteristics of this invention, and without departing from the spirit and scope thereof, can make various changes and modifications of the invention to adapt it to various usages and conditions. Thus, various modifications of the invention in addition to those shown and described herein will be apparent to those skilled in the art from the foregoing description. Such modifications are also intended to fall within the scope of the appended claims.

[0176] The following Materials and Methods were used in one or more of the Examples.

[0177] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxPatient series. Breast cancer tissue samples (FFPE) were collected from patients (aged 50-69 years) diagnosed with invasive breast carcinoma during 1996-2003 as part of the prospective and population-based Norwegian Breast Cancer Screening Program (NBCSP) (n=534)

[0010] . For proteomics studies, primary tumours with a diameter larger than 5 mm were randomly selected from this series although with a balance between luminal-like and basal-like tumours. Initially, 50 cases were included for whole tissue proteomics; four cases had insufficient tissue, and 46 cases were finally included.

[0178] Subsequently, 24 cases (12 basal-like, 6 luminal A, 6 luminal B) were selected for laser capture microdissection with separation of epithelial and stromal compartments.

[0179] Notably, all basal-like tumours were triple-negative, and all luminal-like samples were oestrogen and progesterone receptor positive and HER2-negative. The luminal B tumours displayed more than 15% Ki67-positive nuclei

[0010] .

[0180] For initial validation of the identified proteomic signature (35P; described below), publicly available resources were explored for survival analysis (TCGA CPTAC proteomics subset, n=76 [7]; the METABRIC Discovery cohort (n=852)

[0011] , and the KMplotter database (n=2032)

[0012] ,

[0181] To validate the 35P signature in the context of a randomized clinical trial, we explored the Stockholm Tamoxifen trials (STO) of ER+ / HER2- breast cancer patients (n=1046) [13, 14], These studies were performed during 1976-1997 and designed to investigate the long-term (20-year) benefits from tamoxifen therapy by clinically applied tumour characteristics. Patients were randomized to at least 2 years of tamoxifen therapy versus no endocrine therapy (control). Distant recurrence-free interval (DRFI) was assessed by Kaplan-Meier analysis, multivariable Cox’ proportional hazard analysis, and time-varying flexible parametric modelling. Long-term follow up was included, and last date of information was December 31, 2016. Informed consent was obtained before random assignment, and the trial was approved by the Karolinska Institute Regional Ethics Committee. Gene expression information was available.

[0182] Tissue micro-dissection and sample preparation. 10μm thick FFPE sections were deparaffinized, rehydrated and stained with haematoxylin. Tumour epithelium and

[0183] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxstromal compartments were laser micro-dissected (PALM MicroBeam, Zeiss) and pressure catapulted into tube caps (AdhesiveCap 500 opaque, Zeiss). Depending on the available tissue, 0.5-1.9×10⁷ μm³ were micro-dissected. The samples were prepared using the FFPE-FASP protocol, which is described in detail elsewhere

[0015] . After enzymatic digestion of proteins, the resulting peptides were eluted from the FASP filters and desalted using Oasis HLB pElution plates (Waters, Milford, MA, USA).

[0184] Mass spectrometry and raw data analysis. The samples were analysed in their entirety during a 180 min reverse-phase gradient on a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) connected to a Dionex Ultimate NCR.3500RS LC system (for detailed description of the HPLC-MS / MS settings, see Supplementary Methods). Raw mass spectrometry (MS) data was processed with MaxQuant (v1.6.0.16)

[0016] , using recommended settings for label-free quantification

[0017] . Identified features were cross-checked against the “reference proteome” database from UniProt.org (full proteome analysis; fasta file downloaded October-2017) or the core matrisome database

[0018] (ECM protein analysis).

[0185] The processed MS data was analysed with Perseus (v1.6.0.7)[5]. Protein intensity data was log2-transformed and grouped according to breast cancer subtype. Identified proteins were filtered to retain only proteins with intensity data (valid values) in at least 50% of samples within one group.

[0186] Protein expression correlation analysis. Protein intensity values were correlated between samples, and the correlation coefficients (rs) were visualized by unsupervised hierarchical clustering (distance: Euclidean; linkage: average).

[0187] Protein and gene expression signature scoring. Each protein / gene expression level was zero-mean normalized by subtracting the average protein / gene expression value of that protein / gene (from all patient samples) from the expression value of that protein / gene of each patient sample. The signature score was the sum of the normalized expression values from the upregulated proteins / genes minus the sum of the normalised expression values from the downregulated proteins / genes.

[0188] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxGene ontology (GO) analysis. GO analyses were performed using Panther Classification System

[0019] (PANTHER 16.0, Overrepresentation Test, GO Ontology database DOI: 10.5281 / zenodo.4735677 Released 2021-05-01).

[0189] Gene sets enrichment analysis (GSEA). GSEA was used to identify enriched Hallmark gene sets (MSigDB)

[0020] between the Q4 and Q1 groups, using Qlucore Omics Explorer 3.7 (Qlucore AB, Lund, Sweden). The genes were ranked using a two-sided Student’s t-test.

[0190] Network analysis. The protein-protein interaction network was made using the StringDB [21, 22] (v11.5) and Cytoscape software

[0023] (v3.8.2) and core app NetworkAnalyzer

[0024] (v4.4.6) for computing basic network properties. For identifying subclusters of proteins, the Cytoscape add-on MCODE

[0025] (v2.0.0) was used with the following settings: network scoring: include loops: false, degree cutoff: 2; cluster finding node score cutoff: 0.2, haircut: true, fluff: false, K-Core: 2, max. depth from seed: 100.

[0191] Statistics. Fisher's exact test was used to test categorical clinical variables, and the Mann-Whitney U-test was used to test continuous clinical variables. Differential abundance of proteins was tested using the Student’s t-test and multiple-sample ANOVA test. Spearman’s rank correlation was used to test nonparametric correlations. The p-value significance threshold was set to 0.05.

[0192] Study approval. The study protocol was approved by the Western Norway Regional Committees for Medical and Health Research Ethics (REK 2014 / 1984).

[0193] Example 1: Proteomic mapping of breast cancer cells and tumour stroma indicates separate patterns

[0194] Tumour epithelial cells and tumour stroma from 24 FFPE breast cancers (12 basal-like and 12 luminal-like) were separated by laser capture microdissection, followed by analysis of the extracted proteomes using label-free MS-based shotgun proteomics. We were able to quantify 4,157 and 2,150 proteins from micro-dissected tumour epithelium and tumour stroma, respectively.

[0195] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxProtein expression correlation analysis was performed to identify similarities between samples. By plotting the global correlations as a heatmap with unsupervised clustering, the tumour epithelium samples formed two distinct clusters that perfectly grouped the luminal-like and basal-like subtypes. The tumour stroma samples, however, generated three clusters: a “basal-like” cluster (n=4), a “luminal-like” cluster (n=8), and a mixed cluster (n=12), the latter consisting of both luminal-like and basal-like samples). For comparison, whole tissue proteomics (n=46) gave a mixed pattern with no distinct subgroups by unsupervised clustering.

[0196] As the correlation analysis of micro-dissected tumour stroma did not simply reflect the epithelial-based subtypes, we further explored the clustering of stromal samples. The three stromal clusters were compared (by ANOVA), and we found a significant difference in the ratio between extracellular matrix proteins and intracellular proteins between the basal-like and luminal-like clusters, with more intracellular proteins in the basal-like samples, suggesting a higher stromal cel I ularity. This difference appeared to contribute to the stromal clustering. In contrast, the mixed cluster showed an intermediate stromal profile, with moderate levels of both intra- and extracellular proteins. Taken together, these patterns may reflect differences in cell-to-ECM ratios between the three stromal clusters.

[0197] The mixed stromal cluster consisted of both basal-like and luminal-like samples. The high correlation between these samples suggests that the stromal composition was more similar between the samples in this cluster. Therefore, stromal tissues from the mixed cluster were selected for further analyses, assuming that this approach would be suitable for detecting novel stromal-based traits and not only reflect known variations in immune cell content between these subtypes.

[0198] Since we focused on case differences within the stromal compartment, we limited our investigation to proteins that were differentially abundant only in the stromal compartment (i.e., not also in the tumour epithelium compartment). Therefore, proteins that were differentially abundant in the tumour epithelium (by Student’s t-test, p<0.05,

[0199] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxno fold change cut-off) were subtracted from the list of proteins being differentially abundant in stromal samples (Student’s t-test, p<0.05, fold change>1.5). Subsequently, we identified 35 proteins that were significantly different between basal-like and luminallike tumours in the stromal tissue fraction from the mixed patient cluster (Table 2).

[0200] Example 2: A stromal proteome signature represents both extracellular matrix and intracellular proteins reflecting major tumour programs

[0201] To investigate the 35P stromal protein signature in more detail, we performed a proteinprotein interaction analysis (STRING) and found a few intra-signature protein connections. 10 of the 35 proteins were components of the ECM, and the signature was overrepresented by proteins associated with “extracellular matrix” (such as CTSZ, MMP2 and PXDN; FDR<0.001), “basement membrane” (such as AGRN, NID1 and VWA1; FDR=0.008) and “extracellular exosomes” (such as IDH2, PIP and VAT1;

[0202] FDR=0.025). Cell-type affiliation and specificity (Human Protein Atlas) indicate that the signature proteins associate with several different cell types, including fibroblasts, endothelial cells and immune cells. Taken together, our findings suggest that the 35P stromal proteome signature is composed of both intracellular proteins and extracellular matrix proteins, originating from a range of cell types in the tumour microenvironment.

[0203] Using an additional approach to bypass the proteome differences introduced by varying levels of immune cells, we re-analysed the 24 micro-dissected stroma samples using an ECM database (Core Matrisome

[0018] ). The differences in core ECM proteins between basal-like and luminal-like subtypes, within the mixed stromal cluster, were only moderate (8 significantly different proteins; higher abundance in basal-like samples: NID1, VWA1, COL12A1, AEBP1, and LAMB2; lower abundance in basal-like samples: COL1A1, VCAN, COL3A1; ranked by p-value). Two proteins overlapped with the 35P signature (NID1 and VWA1). Of note, NID1 and VWA1 had the highest positive fold change when comparing basal-like and luminal-like cases in both the Core Matrisome analysis (4.4 and 3.7, respectively) and in 35P (full proteome analysis).

[0204] We then used k-means clustering (k=3) to group the stroma samples into three clusters to investigate how these clusters aligned with the stomal sample clusters and found that

[0205] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxthe clusters from these two analyses were highly similar. This indicates that the matrisome proteins are relevant to the clustering from the full-proteome analysis.

[0206] Example 3: The 35P signature adds clinical and prognostic information independent of epithelial-based tumour subtypes

[0207] We wanted to explore the stromal proteomic diversity among breast tumours based on the expression of the 35P proteins. We used the METABRIC Discovery cohort (microarray data; n=852; HER2 and normal-like excluded), and each patient-identifier was ranked from low to high 35P score and grouped into quartiles (Q1-4); 32 of the 35 proteins in the signature could be matched with genes (the non-matching proteins were immunoglobulins IGHG4, IGKV3-15 and IGKV3D-20). The 35P protein signature was positively associated with histological grade, ER and PR status, and proliferation as measured by Ki67 expression, but not with tumour size and lymph-node status (Table 1). Notably, a marked diversity between epithelial-based tumour subtypes and level of stromal 35P was observed. Both luminal A and luminal B cases were represented in all four quartiles of 35P (Q1-4), whereas basal-like cases were linked to three quartiles (Q2-4) (Figure 1A). We then compared the expression differences between the upper and lower quartiles (Q4 versus Q1) of 35P, overall, and separately for each subtype. The most striking results were found within the luminal A subtype, where we found by GSEA that the Q4 group was significantly enriched in features associated with “highgrade” tumours, such as inflammatory response, hypoxia features, angiogenesis, and epithelial-to-mesenchymal transition (all FDR<0.05, Table 4). Notably, established signatures representing these processes correlated significantly (Figure 2) with 35P.

[0208] Next, we investigated the differentially-expressed genes between Q4 and Q1 in the luminal A group (FDR<0.05, FC>1.5; 287 genes, with special emphasis on markers used for breast cancer classification. Estrogen receptor (ESR1) was downregulated in the Q4 group (FC=0.5). Notably, none of the commonly used basal and luminal cytokeratin markers (CK5, CK 14, CK8 / 18) were different between the two groups. The most up- and down-regulated genes were MMP9 and PIP, respectively, both of which are linked to breast cancer invasion [26-29],

[0209] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxBy network analysis, we found that differentially-expressed genes between Q4 and Q1 reflect a highly connected set of proteins (PPI enrichment p-value<1.0x10-16. This network contained three sub-networks consisting of genes with higher expression in Q4 compared with Q1. The proteins with the highest fold change in each sub-cluster were CXCL9 and CXCL10 (subcluster-1), MMP9 and CXCR4 (subcluster-2) and GBP1 and EPSTI1 (subcluster-3). In the breast, CXCL9, CXCL10, MMP9 and EPSTI1 show highest expression in macrophages (ProteinAtlas). CXCR4 is most frequently associated with T-cells, and GBP1 with adipocytes and endothelial cells. These data support that various TME-cells are associated with breast cancer aggressiveness, as reflected by the 35P signature.

[0210] We hypothesized that the 35P stromal proteome signature, based on differences between basal-like and luminal-like tumour stroma within the mixed stromal patient cluster and reflecting stromal diversity, might be used to improve stratification of breast cancer based on clinical correlates and prognostic impact.

[0211] To test the hypothesis of differences in patient survival, we first looked at 35P in the TCGA proteomics dataset analysed by the Clinical Proteomic Tumour Analysis Consortium (CPTAC) [7], In this dataset of 76 patients (luminal-like and basal-like), we were able to match 27 of the 35 proteins in the 35P signature. Each patient-identifier was ranked from low to high 35P score and grouped into quartiles (Q1-4). Overall, we found a significantly lower overall survival in the 35P high (Q4) group compared to 35P low (Q1 ) (log-rank = 0.001 ); (Figure 1 B).

[0212] Next, to further validate the 35P in a dataset with long-term follow-up, we applied the METABRIC Discovery cohort (n=852; HER2 and normal-like excluded). Given that extracellular matrix proteins correlate poorly, and sometimes negatively, with mRNA expression, as reported by Johansson et al. [6], we first examined the mRNA-protein correlation for 35P. Of the 28 proteins that could be matched with protein-mRNA correlation data, 18 (64%) showed significant positive correlation with mRNA expression, and none showed significant negative correlation (Table 6). This suggested

[0213] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxthat mRNA expression levels corresponding to the 35P signature reflects the protein phenotype.

[0214] Overall, we found significantly lower probability of breast cancer specific survival in the 35P-high group (cut-off upper quartile, Q4; basal-like and luminal-like subtypes included; log-rank test, p<0.001; Figure 1C). We then stratified for subtype and found significantly lower probability of breast cancer specific survival for the 35P-high subgroup within luminal A tumours (p=0.001; Figure 1 D), but not in luminal B (p=0.46; Figure 1 E) or basal-like subtypes (p=0.92; Figure 1 F).

[0215] Further, lower probability of breast cancer specific survival among all cases and luminal A patients with high 35P-score was found in the cohort from KMplotter (n=2032) (Figure 3).

[0216] In the METABRIC cohort, 35P predicted lower breast cancer specific survival independent of basic prognostic factors tumour size, histologic grade and lymph node metastases, as well as the PAM50-based molecular subtypes (Cox’s regression, I ratio test, p=0.009, Table 3). When stratifying the cohort into luminal A, luminal B and basal-like subtypes, the signature was still significantly independent of basic prognostic factors within the luminal A subtype (p=0.002) (Table 5).

[0217] Taken together, the 35P-high group showed worse outcome, which indicates that the addition of stromal proteomic information may be necessary to classify breast cancer as precisely as possible, especially in patients with assumed low-grade (luminal A) tumours based on current criteria. The fact that 35P was independently significant in the METABRIC cohort, based on mRNA values from whole tissue samples, supports the robustness of the 35P signature.

[0218] Example 4: The 35P signature shows significant prognostic impact in a randomized clinical trial of breast cancer

[0219] We then asked whether the 35P stromal proteome signature could predict patient survival in the setting of a randomized clinical trial (RCT). In the STO-cohort (n=1046),

[0220] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx35P (by upper tertile) was significantly associated with shorter long term survival over 20 years in the control arm and for both end points (DRFI and BCSS: p=0.033 and p=0.023, respectively) (Figure 4), whereas no difference between high and low 35P was observed in the treatment arm (tamoxifen). There was no significant interaction between level of 35P and treatment effect.

[0221] Example 5: Reduced subsets of PROCAST-35 show predictive power in univariate analyses of the METABRIC Discovery cohort

[0222] To refine and reduce the number of proteins in 35P, we applied a variant of recursive feature elimination, where we iteratively removed the protein that contributed the least to the parent signature, as determined by the lowest chi-square values (Figure 5A). By this method we found a subset of 18-proteins (AGRN, ATP13A1, TMEM258, CTSZ, EFEMP2, EIF3F, FBF1, IDH2, P3H3, LSM1, MYL9, NID1, PIP, PXDN, TOP1, TPSAB1, VAT1, VWA1) that showed the strongest predictive power overall and for patients diagnosed with luminal A breast cancer(METABRIC Discovery; Log-rank test p=1E-09; Figure 5 B-E). In Addition, we smaller subset of only 5-proteins (AGRN, CTSZ, LSM1, MYL9 and TPSAB1) that also showed strong predictive power (Figure 5 F-l).

[0223] Example 6: Production of signature score and use in prognosis

[0224] A sample of breast tumour is obtained from a breast cancer patient. The tumour is fixed, cryopreserved, and sectioned for downstream analysis. Total RNA is extracted, and the expression levels of 5P (AGRN, CTSZ, LSM1, MYL9 and TPSAB1) and a set of reference proteins are quantified using NanoString RNA assay. The reference proteins are UBA1, PLEC and SPTAN1. The expression levels of the 5P proteins are normalised against the set of reference proteins.

[0225] The signature score for the breast tumour is produced by calculating the sum of the normalised expression levels of the up-regulated proteins (AGRN, CTSZ, LSM1, MYL9), then subtracting the sum of the normalised expression levels from the down-regulated proteins (TPSAB1) in accordance with Equation 1:

[0226] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxEquation 1. Signature score =

[0227]

[0228] iEU jED

[0229] wherein

[0230] - Xirepresents the normalized expression level of an up-regulated protein i in the set of up-regulated proteins, U. (In this Example, U is all proteins with a positive fold change.)

[0231] - X represents the normalized expression level of a down-regulated protein j in the set of down-regulated proteins D. (In this Example, D is all proteins with a negative fold change.

[0232] The first summation computes the total expression of all up-regulated proteins. The second summation computes the total expression of all down-regulated proteins. The final score is obtained by subtracting the total down-regulated expression from the total up-regulated expression. The signature score is shown to be higher than a set cut-off (reference value) and hence the breast cancer prognosis is poor.

[0233] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxTABLES

[0234] Table 1. Associations between 35P and clinical variables

[0235] 35P Q1-Q3 Q4 p-value* Tumour size

[0236] < 20 mm 195 68 0.70 >= 20 mm 444 145

[0237] Histologic grade

[0238] 1 -2 398 45 < 0.001 3 241 168

[0239] Lymph node status

[0240] Negative 354 99 0.24 Positive 285 114

[0241] ER status

[0242] Negative 20 100 < 0.001 Positive 619 113

[0243] PR status

[0244] Negative 200 150 < 0.001 Positive 439 63

[0245] Proliferation (Ki67 expression)

[0246] Q1-Q3 533 112 < 0.001 Q4 106 101

[0247] *Pearson’s chi-square test

[0248] P:\1675\167562-01\Speos\167562-01 2026-02-17 - PCT des cls abs.docxTable 2. 35P stromal signature proteins

[0249] UniProtlD Gene names

[0250]

[0251] .I,.

[0252] (basal-hke / lummal-hke) P14543 NID1 5.56 Q6PCB0 VWA1 3.36 Q92626 PXDN 2.67 P01861 IGHG4(1)2.52 P24844 MYL9 2.40 015116 LSM1 2.12 P61165 TMEM258 2.03 Q9UBR2 CTSZ 2.03 000468 AGRN 2.02 Q53FA7 TP53I3 1.99 P40261 NNMT 1.98 P01624 IGKV3D-15(1)1.95 P01893 HLA-H 1.82 Q9HD20 ATP13A1 1.81 Q9UKY7 CDV3 1.77 095486 SEC24A 1.75 000303 EIF3F 1.72 P62328 TMSB4X 1.69 Q9UK45 LSM7 1.67 P11387 TOP1 1.66 A0A0C4DH25 IGKV3D-20<1) 1.60 095967 EFEMP2 1.59 P16152 CBR1 1.58 P48735 IDH2 1.55 043516 WIPF1 1.53 P43686 PSMC4 1.52 Q8IVL6 P3H3 -1.53 Q8TES7 FBF1 -1.57 Q99536 VAT1 -1.69 P08253 MMP2 -1.86 P19827 ITIH1 -1.95 Q6NZI2 CAVIN1 -2.01 Q15661 TPSAB1 -3.78 P23946 CMA1 -6.47

[0253]

[0254] P12273 PIP -13.5(1)Protein did not have a corresponding gene in the METABRIC discovery cohort.

[0255] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxTable 3. Multivariate survival analysis (proportional hazards regression model) of breast cancer patients (METABRIC discovery cohort; n=852).

[0256] Univariate analysis Multivariate analysis Variable nHR HR

[0257] (95 % Cl) P-value(95o / o C|j p-value Tumour size < 20 mm 263 1.00 1.00 > 20 mm 589 1.90 <0.001 1.60 0.001

[0258] (1.44-2.52) (1.12-2.14) Histologic grade 1-2 443 1.00 1.00 3 409 1.83 <0.001 1.48 0.003

[0259] (1.44-2.32) (1.14-1.92) Lymph node status Negative 453 1.00 1.00 Positive 399 2.04 <0.001 1.74 <0.001

[0260] (1.61-2.59) (1.36-2.22) PAM50 subtype Luminal-like(2)734 1.00 1.00

[0261]

[0262] Basal-like 118 1.45 0.02 0.80

[0263] (1.06-1.98) (0.54-1.19)

[0264]

[0265] 35P Stromal signature

[0266] Q123 639 1.00 1.00

[0267] Q4 213 1.67 <0.001 1.53 0.009

[0268]

[0269] (1.29-2.13) (1.11-2.09)(1)Only patients with luminal A, luminal B and basal-like breast cancers were included (n=852, METABRIC discovery cohort).

[0270] (2)Luminal-like: patients with luminal A and luminal B breast cancer subtypes.

[0271] Cl: Confidence interval. HR: hazard ratio, n: number of patients. NS: not significant.

[0272] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxTable 4. Gene set enrichment analysis of 35P-hiqh (Q4) versus 35P-low (Q1) among luminal A patients in the METABRIC discovery cohort (FDR<0.05)

[0273] i. iv,.... Enrichment. ___.

[0274] Hallmark g

[0275] aene set(1)Matches p-value FDR score

[0276] Interferon gamma response 193 / 200 0.74 0.004 0.005 Complement 190 / 200 0.62 0.000 0.005 Apoptosis 157 / 161 0.53 0.000 0.005 E2F targets 181 / 200 0.66 0.002 0.006 Inflammatory response 194 / 200 0.62 0.004 0.006 Allograft rejection 192 / 200 0.74 0.002 0.006 PI3K Akt mTOR signaling 102 / 105 0.51 0.000 0.007 I L2 STAT5 signaling 189 / 199 0.54 0.004 0.008 G2m checkpoint 180 / 200 0.61 0.006 0.008 Myc targets v1 182 / 200 0.61 0.002 0.011 Epithelial mesenchymal transition 195 / 200 0.66 0.006 0.012 Hypoxia 187 / 200 0.47 0.000 0.012 Coagulation 135 / 128 0.50 0.006 0.013 KRAS signaling up 192 / 200 0.52 0.010 0.013 IL6 JAK STAT3 signaling 87 / 87 0.59 0.012 0.013 TNFA signaling via NFKB 193 / 200 0.55 0.021 0.015 Interferon alpha response 89 / 97 0.76 0.006 0.016 Unfolded protein response 106 / 113 0.47 0.002 0.018 MTORC1 signaling 189 / 200 0.52 0.012 0.020 Mitotic spindle 190 / 199 0.46 0.010 0.021 UV response up 150 / 158 0.40 0.000 0.027 Apical junction 192 / 200 0.44 0.004 0.029 Angiogenesis 36 / 36 0.57 0.020 0.037 <1> Hallmark gene sets; MSigDB, www.broadinstitute.org / gsea / msigdb

[0277] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxTable 5. Multivariate survival analysis (proportional hazards regression model) stratified by molecular subtype.

[0278] Univariate analysis Multivariate analysis Variable

[0279]

[0280] HR HR

[0281] p-value p-value (95 % Cl) (95 % Cl)

[0282]

[0283] Luminal A subtype (n=466) Tumour size

[0284] < 20 mm 164 1.00 1.00

[0285] > 20 mm 2.22 <0.001 2.01 0.001

[0286]

[0287] (1.44-3.42) (1.36-3.25)

[0288] Histologic grade

[0289] 1-2 333 1.00 1.00

[0290] 3 1.63 0.010 1.39 0.090

[0291] (1.12-2.37) (0.95-2.03) Lymph node status Negative 273 1.00 1.00 Positive 1.58 0.014 1.44 0.054i yd(1.10-2.27) (0.95-2.08) 35P Stromal signature Q123 426 1.00 1.00 Q4 2.25 0.002 2.27 0.002

[0292] (1.36-3.72) (1.36-3.77) Luminal B subtype (n=268) Tumour size < 20 mm 64 1.00 1.00 > 20 mm 204 1.99 0.005 1.57 0.083

[0293] (1.24-3.20) (0.94-2.60) Histologic grade 1-2 102 1 00 1 00 3 166 C38 113

[0294] (0.94-2.03)u(0.76-1.69)uLymph node status Negative 127 1.00 1.00 Positive 141 2.39 <0.001 2.10 <0.001

[0295]

[0296] (1.63-3.51) (1.41-3.14) 35P Stromal signature

[0297] Q123 199 1.00 0.464 1.00

[0298] P:\1675\167562-01\Speos\167562-01 2026-02-17 - PCT des cls abs.docxQ4 1.17 1.21 0.384

[0299] Basal-like subtype (n=118) Tumour size < 20 mm 35 1 00 _n 7So 1 00 _nnon > 20 mm 83 0.91 0.753 Q Ji 0.293

[0300] (0.49-1.68)(NS)(0.38-1.34)(NS)Histologic grade _ 1-2 _ 8 3 110 -0 32444n - 0.519

[0301] (0.50-8.42)(NS)(0.38-6.71)(NS)Lymph node status Negative _ 53 1.00 _ 1.00 Positive 65 2.39 0.006 2.62 0.005

[0302] (1.28-4.48) (1.35-5.11) 35P Stromal signature Q123 _ 14

[0303]

[0304] Q4 104 - 0.992 44? -0 373

[0305] (041-2.26) <NS> (027-1.63) <NS> Cl: Confidence interval. HR: hazard ratio, n: number of patients. NS: not significant.

[0306] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxTable 6. mRNA-protein correlation of 35P in the OSLO2 cohort (Johansson et al. 2019) 35P Correlation p-value

[0307] TP53I3 0.688 < 0.001

[0308] P3H3 0.652 < 0.001

[0309] LSM1 0.633 < 0.001

[0310] NNMT 0.633 < 0.001

[0311] CBR1 0.630 < 0.001

[0312] PIP 0.603 < 0.001

[0313] PXDN 0.599 < 0.001

[0314] IDH2 0.598 < 0.001

[0315] WIPF1 0.551 < 0.001

[0316] CTSZ 0.540 < 0.001

[0317] MMP2 0.528 < 0.001

[0318] MYL9 0.504 < 0.001

[0319] ATP13A1 0.485 0.001

[0320] SEC24A 0.471 0.001

[0321] TOP1 0.395 0.008

[0322] TMSB4X 0.390 0.008

[0323] PSMC4 0.351 0.018

[0324] EFEMP2 0.308 0.040

[0325] VAT1 0.273 0.070

[0326] CDV3 0.261 0.083

[0327] AGRN 0.199 0.190

[0328] CMA1 0.192 0.205

[0329] EIF3F 0.130 0.395

[0330] CAVIN1 0.116 0.447

[0331] VWA1 0.109 0.473

[0332] ITIH1 0.064 0.676

[0333] LSM7 0.038 0.803

[0334] P:\1675\167562-01\Speos\167562-01 2026-02-17 - PCT des cls abs.docxNID1 -0.228 0.133

[0335] FBF1 ND ND

[0336] HLA-H ND ND

[0337] IGHG4 ND ND

[0338] IGKV3D-15 ND ND

[0339] IGKV3D-20 ND ND

[0340] TMEM258 ND ND

[0341] TPSAB1 ND ND

[0342] REFERENCES

[0343] 1. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature 2000;406(6797):747-52.

[0344] 2. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumour subclasses with clinical implications. Proc Natl Acad Sci U S A 2001;98(19): 10869-74.

[0345] 3. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumour subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A 2003;100(14):8418-23.

[0346] 4. Yanovich G, Agmon H, Harel M, et al. Clinical Proteomics of Breast Cancer Reveals a Novel Layer of Breast Cancer Classification. Cancer Res 2018;78(20):6001-6010.

[0347] 5. Tyanova S, Temu T, Sinitcyn P, et al. The Perseus computational platform for comprehensive analysis of (prote)omics data. Nat Methods 2016;13(9):731-40.

[0348] 6. Johansson HJ, Socciarelli F, Vacanti NM, et al. Breast cancer quantitative proteome and proteogenomic landscape. Nat Commun 2019; 10(1): 1600.

[0349] 7. Mertins P, Mani DR, Ruggles KV, et al. Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 2016;534(7605):55-62.

[0350] 8. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005;365(9460):671- 9.

[0351] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx9. van 't Veer L J, Dai H, van de Vijver MJ, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415(6871 ):530-6.

[0352] 10. Knutsvik G, Stefansson IM, Aziz S, et al. Evaluation of Ki67 expression across distinct categories of breast cancer specimens: a population-based study of matched surgical specimens, core needle biopsies and tissue microarrays. PLoS One 2014;9(11):e112121.

[0353] 11. Curtis C, Shah SP, Chin SF, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012;486(7403):346-52. 12. Gyorffy B, Lanczky A, Eklund AC, et al. An online survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer prognosis using microarray data of 1,809 patients. Breast Cancer Res Treat 2010;123(3):725-31.

[0354] 13. Dar H, Johansson A, Nordenskjold A, et al. Assessment of 25-Year Survival of Women With Estrogen Receptor-Positive / ERBB2-Negative Breast Cancer Treated With and Without Tamoxifen Therapy: A Secondary Analysis of Data From the Stockholm Tamoxifen Randomized Clinical Trial. JAMA Netw Open 2021;4(6):e2114904.

[0355] 14. Johansson A, Dar H, van 't Veer LJ, et al. Twenty-Year Benefit From Adjuvant Goserelin and Tamoxifen in Premenopausal Patients With Breast Cancer in a Controlled Randomized Clinical Trial. J Clin Oncol 2022;40(35):4071-4082.

[0356] 15. Wisniewski JR. Proteomic sample preparation from formalin fixed and paraffin embedded tissue. J Vis Exp 2013; 10.3791 / 50589(79).

[0357] 16. Cox J, Mann M. MaxQuant enables high peptide identification rates, individualized p.p.b. -range mass accuracies and proteome-wide protein quantification. Nat Biotechnol 2008;26(12): 1367-72.

[0358] 17. Cox J, Hein MY, Luber CA, et al. Accurate Proteome-wide Label-free Quantification by Delayed Normalization and Maximal Peptide Ratio Extraction, Termed MaxLFQ. Mol Cell Proteomics 2014; 13(9):2513-26.

[0359] 18. Naba A, Clauser KR, Ding H, et al. The extracellular matrix: Tools and insights for the "omics" era. Matrix Biol 2016;49:10-24.

[0360] 19. Mi H, Ebert D, Muruganujan A, et al. PANTHER version 16: a revised family classification, tree-based classification tool, enhancer regions and extensive API.

[0361] Nucleic Acids Res 2021;49(D1): D394-D403.

[0362] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx20. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005; 102(43): 15545-50.

[0363] 21. Snel B, Lehmann G, Bork P, et al. STRING: a web-server to retrieve and display the repeatedly occurring neighbourhood of a gene. Nucleic Acids Res 2000;28(18):3442-4.

[0364] 22. Szklarczyk D, Gable AL, Lyon D, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res 2019;47(D1): D607-D613.

[0365] 23. Shannon P, Markiel A, Ozier O, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res

[0366] 2003; 13(11):2498-504.

[0367] 24. Assenov Y, Ramirez F, Schelhorn SE, et al. Computing topological parameters of biological networks. Bioinformatics 2008;24(2):282-4.

[0368] 25. Bader GD, Hogue CW. An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 2003;4:2.

[0369] 26. Jiang H, Li H. Prognostic values of tumoural MMP2 and MMP9 overexpression in breast cancer: a systematic review and meta-analysis. BMC Cancer 2021;21 (1 ): 149. 27. Joseph C, Alsaleem M, Orah N, et al. Elevated MMP9 expression in breast cancer is a predictor of shorter patient survival. Breast Cancer Res Treat 2020;182(2):267-282.

[0370] 28. Baniwal SK, Chimge NO, Jordan VC, et al. Prolactin-induced protein (PIP) regulates proliferation of luminal A type breast cancer cells in an estrogen-independent manner. PLoS One 2014;8(6):e62361.

[0371] 29. Urbaniak A, Jablonska K, Podhorska-Okolow M, et al. Prolactin-induced protein (PIP)-characterization and role in breast cancer progression. Am J Cancer Res 2018;8(11):2150-2164.

[0372] 30. Braakman RB, Stingl C, Tilanus-Linthorst MM, et al. Proteomic characterization of microdissected breast tissue environment provides a protein-level overview of malignant transformation. Proteomics 2017; 17(5).

[0373] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx31. Reddy LA, Mikesh L, Moskulak C, et al. Host response to human breast Invasive Ductal Carcinoma (IDC) as observed by changes in the stromal proteome. J Proteome Res 2014;13(11):4739-51.

[0374] 32. Vizcaino JA, Deutsch EW, Wang R, et al. ProteomeXchange provides globally coordinated proteomics data submission and dissemination. Nat Biotechnol 2014;32(3):223-6.

[0375] 33. Vizcaino JA, Cote RG, Csordas A, et al. The PRoteomics IDEntifications (PRIDE) database and associated tools: status in 2013. Nucleic Acids Res 2013;41 (Database issue): D1063-9.

[0376] Supplementary References

[0377] 51. Parker JS, Mullins M, Cheang MC, et al. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 2009;27(8):1160-7.

[0378] 52. Mertins P, Mani DR, Ruggles KV, et al. Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 2016;534(7605):55-62.

[0379] 53. Curtis C, Shah SP, Chin SF, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012;486(7403):346-52.

[0380] 54. Gyorffy B, Lanczky A, Eklund AC, et al. An online survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer prognosis using microarray data of 1,809 patients. Breast Cancer Res Treat 2010;123(3):725-31.

[0381] P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx

Claims

CLAIMS1. A method of obtaining an indication of the prognosis of breast cancer in a subject, the method comprising the step:(a) producing a signature score from normalized levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;wherein the produced signature score is indicative of the prognosis of breast cancer in the subject.

2. A method as claimed in claim 1, wherein the breast cancer is a luminal-A breast cancer.

3. A method as claimed in claim 1 or claim 2, wherein a produced signature score which is higher than a corresponding reference signature score is indicative of an increased likelihood or statistically-significant chance of the subject having a poor prognosis for breast cancer.

4. A method of classifying breast tumours, the method comprising the steps:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject; and(b) classifying the breast tumour based on the signature score produced.

5. A method as claimed in claim 4, wherein the produced signature score is compared against a corresponding panel of reference signature scores or set of ranges of references signature scores which have (previously) been obtained from tissues which are representative of different breast tumours having different phenotypes orP:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxgenotypes or other physical properties; and classifying the breast tumour based on which reference signature score is closest to the obtained signature score, or into which range of reference scores the obtained signature score falls.

6. A method as claimed in claim 4, wherein the different breast tumours having different phenotypes or genotypes or other physical properties are selected from:(i) breast tumours having a basal-like phenotype or not;(ii) breast tumours having a luminal-like phenotype or not;(iii) breast tumours having different tumour sizes;(iv) breast tumour having different histologic grades;(v) breast tumours having lymph node metastases or not;(vi) breast tumours which are ER negative or not; and(vii) breast tumours having high levels of tumour cell proliferation or not.

7. A method as claimed in claim 4, wherein the breast tumour is classified as being a basal-like tumour or a luminal-like tumour based on the signature score produced.

8. A method of obtaining an indication of the efficacy of a drug which is being used to treat breast cancer in a subject, the method comprising the steps:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a first biological sample which was obtained from the subject at a first time point; and(b) producing a second signature score using corresponding levels of the corresponding biomarkers in a corresponding second biological sample obtained from the subject at a second (later) time point;wherein the drug has been administered to the subject in the interval between the first and second time points, wherein a decrease in the second signature score compared to the first signature score is indicative of the efficacy of the drug, and wherein an increase in the second signature score compared to the first signature score is indicative of the lack of efficacy of the drug.P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx9. A method of treating breast cancer in a subject, the method comprising the steps of:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;(b) comparing the signature score with a reference signature score; and(c) administering a treatment appropriate for treating breast cancer to the subject if the signature score is above the reference signature score, thereby treating the breast cancer in the subject.

10. A method of treating breast cancer in a subject, the method comprising the step of:(a) administering a treatment appropriate for treating breast cancer to the subject, wherein, prior to administration, a signature score which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject, had been determined to be above a reference signature score.

11. A method of treating breast cancer in a subject, the method comprising the steps of:(a) receiving a signature scope which was produced from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers were:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(b) identifying the subject as having a signature score above a reference signature score, thereby providing an indication of the potential efficacy of administering treatment appropriate for treating the breast cancer to the subject; and(c) administering a treatment appropriate for treating the breast cancer to the subject.

12. A method of detecting biomarkers in a breast tissue sample obtained from a human subject, the method comprising measuring:(i) a protein expression level for every protein in a group of classifier proteins; or (ii) a mRNA expression level for every gene in a group of classifier genes; wherein the group of classifier proteins or classifier genes consists of only:(a) AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P);(b) AGRN, ATP13A1, TMEM258, CTSZ, EFEMP2, EIF3F, FBF1, IDH2, P3H3, LSM1, MYL9, NID1, PIP, PXDN, TOP1, TPSAB1, VAT1 and VWA1 (18P); or (c) NID1, VWA1, PXDN, IGHG4, MYL9, LSM1, TMEM258, CTSZ, AGRN, TP53I3, NNMT, IGKV3D-15, HLA-H, ATP13A1, CDV3, SEC24A, EIF3F, TMSB4X, LSM7, TOP1, IGKV3D-20, EFEMP2, CBR1, IDH2, WIPF1, PSMC4, P3H3, FBF1, VAT1, MMP2, ITIH1, CAVIN1, TPSAB1, CMA1 and PIP (35P).

13. An ex vivo method of screening for agents for treating breast cancer, the method comprising the steps:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein biomarkers were obtained from a breast cancer sample which has been treated with an agent;(b) producing a second signature score from corresponding biomarkers obtained from the breast cancer sample which has not been treated with the agent; and (c) comparing the first and second signature scores;wherein a decrease in the first signature score compared to the second signature score is indicative of an agent which is capable of treating breast cancer.P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx14. A method of predicting the risk of recurrence of breast cancer in a subject who has previously had breast cancer but who is currently in remission, the method comprising the step:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein biomarkers were obtained from a biological sample which was obtained from the subject;wherein the signature score is predictive of the risk of recurrence of breast cancer in the subject.

15. A method as claimed in claim 14, wherein:(i) a signature score from the subject which is higher than a corresponding reference signature score is indicative of an increased likelihood or statistically-significant chance of recurrence of breast cancer in the subject; and(ii) a signature score from the subject which is lower than a corresponding reference signature score is indicative of a decreased likelihood or statistically-significant chance of recurrence of breast cancer in the subject.

16. A method of predicting the therapeutic efficacy of treatment (preferably radiotherapy treatment) on a subject with breast cancer, the method comprising the step:(a) producing a signature score from normalised levels of a plurality of biomarkers, wherein at least 5 of the biomarkers are:AGRN, CTSZ, LSM1, MYL9 and TPSAB1 (5P)and wherein the biomarkers were obtained from a biological sample which was obtained from the subject;wherein the produced signature score is predictive of the therapeutic efficacy of the treatment (preferably the radiotherapy treatment) on the breast cancer.

17. A method as claimed in claim 16, wherein a produced signature score from the subject which is higher than a corresponding reference signature score is indicative ofP:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxan increased likelihood or statistically-significant chance of the subject having a shorter survival time after treatment.

18. A method as claimed in any one of the preceding claims, wherein the plurality of biomarkers comprises:(i) 5P and one or more other biomarkers selected from the group consisting of 18P; (ii) 5P and at least 10 other biomarkers selected from the group consisting of 18P; or(iii) all of the biomarkers in 18P.

19. A method as claimed in any one of the preceding claims, wherein the plurality of biomarkers comprises:(i) 18P and one or more other biomarkers selected from the group consisting of 35P; (ii) 18P and at least 10 other biomarkers selected from the group consisting of 35P; or(iii) all of the biomarkers in 35P.

20. A method as claimed in any one of the preceding claims, wherein the biological sample is:(i) blood, serum or plasma;(ii) whole tissue breast tumour;(iii) tumour stroma; or(iv) saliva or urine;preferably tumour stroma.

21. A method as claimed in any one of the preceding claims, wherein the biomarkers are:(i) proteins; or(ii) RNA, preferably mRNA.

22. A method as claimed in any one of the preceding claims, wherein the levels ofP:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docxthe biomarkers are normalised against the levels of one or more control or housekeeping genes or proteins.

23. A method as claimed in claim 22, wherein the levels of each of the selected biomarkers are normalised by:(i) subtracting the level of one or more control or house-keeping genes or proteins from the obtained levels of each selected biomarker; or(ii) dividing the obtained levels of each of the selected biomarkers by the level of one or more control or house-keeping genes or proteins.

24. A method as claimed in any one of the preceding claims, wherein the signature score is produced by calculating the sum of the normalised expression levels of the up-regulated biomarkers, and then subtracting the sum of the normalised expression levels for the down-regulated biomarkers in accordance with Equation 1:Equation 1. Signature score =iEU jEDwherein- Xi represents the normalized expression level of an up-regulated biomarker i in the set of up-regulated biomarkers, U; and- X represents the normalized expression level of a down-regulated biomarker j in the set of down-regulated biomarkers, D;and wherein NID1, VWA1, PXDN, IGHG4, MYL9, LSM1, TMEM258, CTSZ, AGRN, TP53I3, NNMT, IGKV3D-15, HLA-H, ATP13A1, CDV3, SEC24A, EIF3F, TMSB4X, LSM7, TOP1, IGKV3D-20, EFEMP2, CBR1, IDH2, WIPF1 and PSMC4 are upregulated biomarkers; and P3H3, FBF1, VAT1, MMP2, ITIH1, CAVIN1, TPSAB1, CMA1 and PIP are down-regulated biomarkers.

25. A kit comprising reagents sufficient for the detection and / or quantitation of expression of each of the following biomarker genes:(i) 5P;(ii) 18P; orP:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx(iii) 35P;characterised in that said reagents comprise a plurality of forward and reverse primers pairs, wherein said forward and reverse primers pairs are selected from forward and reverse primer pairs which are capable of identifying expression of each of the following genes:(i) 5P;(ii) 18P; or(iii) 35P.P:\1675\167562-01\Specs\167562-01 2026-02-17 - PCT des cls abs.docx

Images