Prognostic method for non-metastatic female breast cancer

Abstract

The present invention relates to methods based on miRNA biomarkers and clinical data for the risk stratification of non-metastatic breast cancer of any molecular subgroup according to the risk of developing distant metastases, preferably in fresh- frozen or formalin fixed paraffin embedded (FFPE) specimens, or in a liquid biopsy (serum or plasma sample). More in particular, the invention refers to the use of a (2) miRNA signature within a prognostic model that includes the (2) miRNAs together with biomarkers already implemented in the clinical management of breast cancer patients.

Description

[0001] Title

[0002] Prognostic method for non-metastatic female breast cancer

[0003] Description

[0004] The present invention relates to methods based on miRNA biomarkers and clinical data for the risk stratification of non-metastatic breast cancer of any molecular subgroup according to the risk of developing distant metastases, preferably in fresh- frozen or formalin fixed paraffin embedded (FFPE) specimens, or in a liquid biopsy (serum or plasma sample). More in particular, the invention refers to the use of a 2- miRNA signature within a prognostic model that includes the 2 miRNAs together with biomarkers already implemented in the clinical management of breast cancer patients.

[0005] Background of the invention

[0006] Breast cancer is the most frequently diagnosed cancer and the leading cause of cancer death among women, accounting for 12% of the total cancer cases (2.26 million) and 7% of the cancer deaths (684,996) worldwide (1).

[0007] Although mortality rates have declined thanks to screening programs and more effective adjuvant therapies, there remains the need of identifying metastatic risk early, in order to minimize overtreatment or under treatment, and developing new approaches to prevent metastasis. Traditionally, risk stratification in operable breast cancer has been based on prognostic histopathological variables, including primary tumour size, lymph node metastases (2) and Nottingham tumour histological grade (Elston-Ellis modification of Scarff-Bloom-Richardson grading system) (3, 4). Such parameters with individual prognostic significance have been integrated into the tumour, node and metastasis AJCC staging system (2). In addition, immunohistochemical (IHC) markers, oestrogen receptor alpha (ERa), progesterone receptor (PgR), the human epidermal growth factor receptor 2 (HER2) and Ki67 / Mib1 also represent mandatory elements for decision-making (5).

[0008] High-throughput global gene expression profiling have helped unravel genotypic / phenotypic heterogeneity, diversity of outcome and therapeutic response in Breast Cancer (6). A major milestone in this path has been the definition of five biologically and clinically relevant intrinsic breast cancer subtypes, based on genome-wide expression: two oestrogen receptor (ER) positive subtypes characterized by a relatively low (luminal A) and high (luminal B) expression of proliferation-related genes; a subtype enriched for HER2-amplified tumours (HER2- enriched); a subtype characterized by the absence of ER, PgR expression and HER2 amplification (Basal Like), and a subtype with an expression profile similar to that of normal breast tissue (normal-like) (6).

[0009] The attempt to exploit molecular subtyping for risk stratification has been only partially successful. Currently, four risk prediction methods have been implemented into clinical practice (e.g. Prosigna, Endopredict, Mammaprint and Oncotype DX) (5), but all of them have limited application within a subset of ER positive / HER2 negative breast cancer patients.

[0010] Thus, a large proportion of breast cancer patients, and whose risk of disease progression cannot be predicted by the clinic-pathological criteria, still suffer from the lack of reliable prognostic tests.

[0011] Alternative molecular markers that could be used to predict the risk of disease progression in breast cancer are microRNAs (microRNAs or miRNAs), that are endogenous RNAs of approximately 23 nucleotides, capable of negatively controlling gene expression through the complex RNA-induced silencing (RISC). A single miRNA can control several mRNAs to promote or suppress specific cellular processes (7). The aberrant expression of miRNAs is associated with different pathophysiological conditions including cancer, where they seem to play a pivotal role in various mechanisms and signaling pathways associated with disease progression, metastases and resistance to treatment (8).

[0012] Some prior art discloses the use of detecting miRNAs as diagnostic or prognostic biomarkers in cancer, including breast cancer. In this regard, patent application US20140147495 describes methods for the diagnosis of breast cancer and identifies subgroups associated with already known prognostic factors by determining the level of one or more miRNAs, however it only describes the association of some miRNAS with others, but it does not report whether the miRNAs increase the prognostic power of known prognostic factors. The patent application WO2013110053 describes useful methods for determining the survival prognosis of an individual with breast cancer, by determining the levels of a signature of miRNAs / mRNAs, however this signature contains more than 30 markers, which makes the analysis and interpretation of the results relatively complex. The patent application US20140162888 describes methods to characterize different types of cancer, among which breast cancer is mentioned. This technology is mainly focused on prostate cancer, so it does not describe a particular signature for breast cancer, nor does it describe a prognostic value for a specific signature. The patent application US20110003704 describes methods of diagnosis or prognosis of a disease by detecting at least one miRNA in a sample containing microvescicles of the subject; however, it requires the previous isolation of microvescicles and although it reports a specific signature for breast cancer, it does not assign a particular prognostic value for such signature. The patent application US20100298151 refers to methods of diagnosis, evaluation of the efficacy of treatment and prognosis in different types of cancer, among which breast cancer is only mentioned, by determining one or more miRNAs in exosomes derived from cancer; however said methods require isolation from exosomes and they do not describe a specific signature for breast cancer for prognostic purposes.

[0013] Lastly, the patent application WO2017095213A1 refer to a method that is useful for determining the prognosis of an individual diagnosed with breast cancer, by means of determining the level of expression of a miRNA signatures comprising hsa-miR-224, hsa-miR-372, hsa-miR-10b, hsa-miR-548a, hsa-miR-589, hsa-miR-127, hsa-miR- 133a, hsa-miR-135b and hsa-miR-189 / hsa-miR-24-1 . However, this signature is not specific to define the risk of distant metastases development in patients and the ability to improve the risk prediction of already known breast cancer biomarker was not described.

[0014] There is therefore the need of developing a new diagnostic method able to identify at the first stage of the diagnosis the risk of developing distant metastasis in breast cancer patients, in order to minimize the possibility of an overtreatment or un undertreatment of the patient and to choose the right therapeutic approach since the beginning.

[0015] Brief description of the figures

[0016] Figure 1. Global views of miRNA expression by using the Principal Component Analysis (PCA) among Normal Breast Tissue (NBT), Ductal Carcinoma in situ (CIS), Non-metastatic breast cancer (BC-MO) and metastatic breast cancer (BC-M1 ).

[0017] Figure 2. A) Receiver Operating Characteristic (ROC) curve of predicted individual probabilities of developing metastases, estimated from the multivariable logistic model that included only the two miR-3916 and miR-3613-5p. B) ROC curves of predicted probabilities from two nested multivariable logistic models which included: stage (AJCC 7th edition), histological grade and Ki67 (i.e. the “base model”), and also the two miRNAs (i.e. the “extended model”) as covariates. The Area under the ROC curve (AUC) was reported along with its 95% confidence interval (Cl), estimated using DeLong method. The AUC of the two nested models were compared using DeLong test.

[0018] Figure 3. Calibration curve of predicted vs. actual (i.e. observed) probabilities of developing metastases. A) Individual predicted probabilities were estimated from the multivariable logistic model that included stage (AJCC 7th edition), histological grade and Ki67 (i.e. the “base model”) only; B) Individual predicted probabilities were estimated from the multivariable logistic model that included stage, histological grade, Ki67, miR-3916 and miR-3613-5p (i.e. the “extended model”).

[0019] Figure 4. Results of the Recursive Partitioning And Regression Tree (RPART) algorithm to identify the predicted probability cut-off for classifying patients into low, intermediate and high risk classes .

[0020] Definitions

[0021] Unless otherwise defined, all terms of art, notations and other scientific terminology used herein are intended to have the meanings commonly understood by those persons skilled in the art to which this disclosure pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and / or for ready reference; thus, the inclusion of such definitions herein should not be construed to represent a substantial difference over what is generally understood in the art.

[0022] The term "microRNA" or "miRNA" used herein refers to a small non-coding RNA molecule of about 22 nucleotides found in plants, animals and some viruses, that has role in RNA silencing and post-transcriptional regulation of gene expression. miRNA exerts its functions via base-pairing with complementary sequences within mRNA molecules.

[0023] The term "signature" herein refers to an expression pattern derived from combination of several miRNA (i.e. transcripts) used as biomarkers.

[0024] The term “LP” herein refers to the Linear Predictor estimated from a multivariable logistic model. This measure is defined as the weighted sum of the values of the individual covariates, using the estimated regression coefficients as weights. The inverse formula of the logistic function applied to the LP allows the individual predicted probabilities to be estimated.

[0025] The term "AUC" herein refers to the Area Under the ROC (Receiver Operating Characteristic) Curve, drawn by plotting the true positive rate (i.e. sensitivity) against the false positive rate (i.e. 1 -specificity) achieved from the predicted probabilities, derived from a binary classification model, for all its possible threshold values.

[0026] The term "prognostic" herein refers to the ability to discriminate patients with good / poor prognosis.

[0027] The term "biomarkers" (short for biological markers) herein refers to biological indicators (for example a transcript, i.e. miRNA) and / or measures of some biological state or condition.

[0028] The term “distant metastasis” herein refers to cancer that has spread from the original (primary) tumour to distant organs or distant lymph nodes.

[0029] The term “stage of the tumour” herein refers to the extent of the disease evaluated according with the AJCC staging system.

[0030] The term “histological grade of the tumour” herein refers to the morphological characteristics of the tumour as determined by a pathologist according to the

[0031] Nottingham histological grade system. The terms “approximately” and “about” herein refer to the range of the experimental error, which may occur in a measurement.

[0032] The terms “comprising”, “having”, “including” and “containing” are to be construed open-ended terms (i.e. meaning “including, but not limited to”) and are to be considered as providing support also for terms as “consist essentially of”, “consisting essentially of”, “consist of” or “consisting of”.

[0033] The terms “consists of”, “consisting of” are to be construed as closed terms.

[0034] Description of the invention

[0035] The inventors have now identified a risk stratification model based on the detection of miR3916 and miR3613-5p expression on fresh frozen or FFPE specimen, measured by quantitative real-time PCR (qRT-PCR) with a standard curve (9-11 ), together with clinical parameters typical of the breast cancer patients, such as identification of the tumour stage (2), the evaluation of ki67 / Mib1 percentage of positive cells (5) and of tumour histological grade (3,4).

[0036] As reported above, no prior art document discloses a method based on a prognostic model including the evaluation of miR3916 and miR3613-5p expression together with other clinical parameters, such as the tumour stage, the ki67 / Mib1 percentage of positive cells and the histological grade.

[0037] In particular, the inventors initially applied a global miRNA analysis approach on a selected cohort of subjects (training cohort n=34) and then validated the results on an validation cohort (n=223), which enabled them to develop a new prognostic model prediction of breast cancer distant metastases, to determine individual patient’s risk to develop metastases and stratify breast cancer patients according to their risk of developing metastasis in distant organs or distant lymph nodes. More in details, the inventors identified on the training cohort 8-miRNA differentially expressed in BC-MO patients has compared to BC-M1 patients.

[0038] In the validation cohort two of those miRNAs, miR3916 and miR3613-5p, showed the highest discriminatory power between BC-MO e BC-M1 patients (AUC: 0.68, 95%CI: 0.59-0.77).

[0039] BC-MO group of breast cancer patients are individuals who were free from distant metastases after 15 years from diagnosis.

[0040] BC-M1 group of breast cancer patients are individuals who developed distant metastases within 15 years from diagnoses.

[0041] As disclosed in the experimental section, the inventors tested on the same patients a logistic model (i.e., the “base model”), which includes prognostic factors currently used in clinical practice to make decisions on treatments and follow-up management of breast cancer patients. This model included the identification of the tumour stage, the Nottingham histological grade and the Ki67 / Mib1 percentage of positive cells, showing an AUC of 0.76 (95%CI: 0.68-0.84).

[0042] Surprisingly, by adding the expression levels of miR-3916 and miR-3613-5p to the “base model” it was possible to develop a new prognostic method, herein referred to as the “extended model”, that significantly improved the discriminatory power to predict the risk of developing distant metastasis in a patient: AUC=0.85 (95%CI: 0.79-0.91) (delta-AUC p value = 0.001).

[0043] In particular, according to a rule of thumb (16), it is possible to attribute a label to the strength of the discriminatory power obtained from the AUC as follows:

[0044] 0.5 < AUC < 0.6 No discrimination

[0045] 0.6 < AUC < 0.7 Poor discrimination

[0046] 0.7 < AUC < 0.8 Acceptable discrimination 0.8 < AUC < 0.9 Excellent discrimination

[0047] 0.9 < AUC Outstanding discrimination

[0048] In the present case, by moving from the “basic” to the “extended” model respectively, the discriminatory power goes from being classified as “acceptable” to “excellent”.

[0049] This evidence highlights the reliability of miRNA-based biomarkers for breast cancer prognosis, provided they are evaluated together with other clinical parameters, and boost their implementation in patient clinical management.

[0050] The advantages of the new method identified, with respect to the methods known in the art, are that of being a reliable prognostic method for identify, at the time of diagnoses, breast cancer patients at high risk to develop distant metastases. In particular, said method displays several characteristics that are desirable in a routine clinical setting:

[0051] - it analyses specific transcript sequences which augment the specificity of the said method.

[0052] - it can be implemented in a point-of-care testing (POCT) for miRNA-based diagnostics.

[0053] - it can measure quantities of transcripts using small amount of biological material (up to10 nanograms of purified nucleic acids).

[0054] - combining the data obtained on the two miRNAs expression and routine clinical information increases the ability to identify at early stage the risk of breast cancer patients to develop distant metastasis.

[0055] - the method can be applied to all breast cancer subtypes in order to choose the best therapeutic approach since the beginning, avoiding overtreatment or undertreatment of the patient. It is therefore an embodiment of the present invention a method in-vitro or ex-vivo for calculating the risk score of developing metastases in a patient affected by breast cancer, comprising the following steps: a) detecting the expression level of miR-3916 and miR3613-5p in a biological sample obtained from a patient; b) determining the stage of the tumour; c) determining the percentage of KI67 / Mib1 positive cells; d) determining the histological grade of the tumour; e) calculating the risk score of said patient.

[0056] According to a preferred embodiment, said metastasis are distant metastasis, that develop in distant organs or distant lymph nodes.

[0057] Preferably the biological sample is a fresh tissue sample, a frozen tissue sample, a formalin fixed paraffin embedded sample (FFPE) or a liquid biopsy (serum or plasma sample).

[0058] According to a further preferred embodiment the risk score of the patient is calculated in step e) by a multivariate logistic regression model according to the following formula:

[0059] 1 risk score (%) = 100 x -

[0060] 1 + e ’LPwherein

[0061] LP = = -2.498 + (bi x Xi) + (b2x X2) + (0.018 x X3) + (-0.915 x X4) + (0.731 x X5); and wherein the independent covariates XrX5were coded as follows:

[0062] Xi: Stage of the tumour

[0063] X2: Histological Grade X3: percentage of Ki67 / mib1 positive cells

[0064] X4: miR3916 expression level

[0065] X5: miR3613-5p expression level; and and wherein the regression coefficients (estimated from the logistic model) bi and b2have the following values: if X1 = “stage 1A” then b1 = 0; or if X1= “stage 1 B” then b1= -14.317; or if X1= “stage 2A” then b1= 0.054; or if X1= “stage 2B” then b1= 0.739; or if X1 = stage “3A” then b1 = 1 .294; or if X1= “stage 3B” then b1= 1.087; or if X1= “stage 3C” then b1 = 2.989, and if X2= “histological grade 1” then b2= 0 or if X2= “histological grade 2” then b2= -0.231 or if X2= “histological grade 3” then b2= -0.474.

[0066] In a further preferred embodiment the expression level of miR-3916 and miR3613-5p in step a) is detected by quantitative RT-PCR, ddPCR or probe based methods such as nCounter (Nanostring)

[0067] Preferably, said miRNA expression level is detected by measuring the amount of the total RNA, comprising mRNA and miRNA, extracted with conventional RNA extraction methods.

[0068] Preferably, the qRT-PCR analysis is performed by using the TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Sc) and specific stem-loop RT primers or with an analogue method of quantitative RT-PCR, by using the specific primers and probes selected for amplifying miR-3916 and miR3613-5p analyzed according to the present invention.

[0069] Preferably, the normalization of qRT-PCR analysis may be performed by using, as endogenous control, small nucleolar RNA (snoRNAs) or small nuclear RNA (snRNA) selected from RNU48, RNU44, U47 and RNU6B.

[0070] More preferably, miRNA quantitation in each sample is performed by using a relative quantification method with standard curve.

[0071] More preferable for both miRNAs and RNU48 endogenous control, standard curves are constructed by plotting the threshold cycle (Ct) values against Iog10 of the copy number and fitting by linear least square regression.

[0072] More preferable miRNA expression in each samples is determined as the ratio of the miRNA copy number to the RNU48 copy number and then multiplied by 1000 for easier tabulation: hsa-miR3613 / RNU48 x 1000 and hsa-miR3916-5p / RNU48 x 1000 (9).

[0073] In a further preferred embodiment the stage of the tumor in step b) is determined by analyzing the tumor size and the presence / absence of lymph node metastasis, preferably according to the AJCC staging system, as reported below:

[0074]

[0075] According to the AJCC pathological classification

[0076] Tumour dimension (T)

[0077] T1= tumour <2 cm in greatest dimension

[0078] T2= tumour > 2.0 cm and <5 cm in greatest dimension

[0079] T3= tumour >5.0 cm in greatest dimension

[0080] T4= tumour of any size with direct extension to skin or chest wall

[0081] Lymp hnode (N)

[0082] N0= no nodal metastasis.

[0083] N1= cancer has spread to 1 to 3 axillary (underarm) lymph node(s), and / or cancer is found in internal mammary lymph nodes (those near the breast bone) on sentinel lymph node biopsy.

[0084] N2= cancer has spread to 4 to 9 lymph nodes under the arm, or cancer has enlarged the internal mammary lymph nodes.

[0085] N3= cancer has spread to 10 or more axillary lymph nodes, with at least one area of cancer spread greater than 2 mm; or cancer has spread to the lymph nodes under the collarbone (infraclavicular nodes), with at least one area of cancer spread greater than 2 mm; or cancer is found in at least one axillary lymph node (with at least one area of cancer spread greater than 2 mm) and has enlarged the internal mammary lymph nodes; or Cancer has spread to 4 or more axillary lymph nodes (with at least one area of cancer spread greater than 2 mm), and to the internal mammary lymph nodes on sentinel lymph node biopsy; or cancer has spread to the lymph nodes above the collarbone (supraclavicular nodes) on the same side of the cancer with at least one area of cancer spread greater than 2 mm.

[0086] Metastases (M)

[0087] MO: No distant (distant organs or diatant lymphnodes) spread is found on x-rays (or other imaging tests) or by physical exam.

[0088] According to a preferred embodiment the percentage of KI67 / Mib1 positive cells in step c) is detected by immunohistochemistry.

[0089] According to a further preferred embodiment the histological grade of point d) is detected with the Nottingham histological grading system (3,4). This system, evaluates three morphological features of the tumour: glandular / tubular differentiation, nuclear pleomorphism and mitotic count. A score is given to each morphological feature and the scores are added to get a number between 3 and 9 that is used to get a grade of 1 , 2, or 3.

[0090] Grade 1 or well differentiated (score 3 to 5). The cells are slower-growing, and look more like normal breast cells.

[0091] Grade 2 or moderately differentiated (score 6, 7). The cells are growing at a speed of and look like cells somewhere between grades 1 and 3.

[0092] Grade 3 or poorly differentiated (score 8, 9). The cancer cells look very different from normal cells and will probably grow and spread faster.

[0093] In a further preferred embodiment the subject is classified at high, intermediate and low risk of developing metastasis, based on the calculated risk score.

[0094] Preferably said metastasis are distant metastasis. According to a preferred embodiment the method according to the present invention calculate the risk of developing metastases in a patient within 15 years of follow up.

[0095] Preferably a subject with a risk score > 38.2% is considered at high risk of developing metastasis, a subject with a risk score between 10.6 and 38.1% is considered at an intermediate risk of developing metastasis, and a subject with a risk score < 10.6% is considered at a low risk of developing metastasis.

[0096] In a further preferred embodiment said breast cancer is of the molecular subtype selected from Luminal A, Luminal B, HER2-amplified and Basal-like or of the surrogate molecular subtype selected from Hormone Receptor positive, HER2 amplified and Triple negative.

[0097] According to a preferred embodiment, the method of the present invention classify breast cancer patients of any molecular or surrogate subtype of breast cancer in the predicted class of subjects at low, intermediate and high risk to develop metastases.

[0098] According to a preferred embodiment the method of the present invention is used for the prognostic risk stratification of breast cancer patients to avoid systemic adjuvant treatment in low risk patients.

[0099] According to a preferred embodiment the method of the present invention is used for the prognostic risk stratification of breast cancer patients to identify alternative therapeutic options after surgery in patients at intermediate / high risk classification; breast cancer is normally treated with systemic adjuvant chemotherapy, using antracyclin combinations, such as for instance doxorubicine or epirubicine, which may be used together with taxanes, such as for instance paclitaxel or docetaxel, and / or cyclophosphamide.

[0100] The method of the present invention allows therefore to select the most appropriate treatment based on the risk score obtained for each patient and, consequently, to reduce the risk of side-effects which is normally associated with systemic adjuvant chemotherapy.

[0101] A further embodiment of the present invention is a computer medium comprising instructions which, when executed by a computer, cause the computer to carry the following steps: a) receiving raw data from the expression level obtained from the method of the present invention; b) integrating said data with clinical date determined at point b), c), and d) of the method of the present invention; c) calculating the LP; d) transforming the LP to individual risk score.

[0102] EXPERIMENTAL SECTION

[0103] A global miRNA expression profiling of a discovery cohort of 34 breast cancer cases was performed by using GeneChip® miRNA 4.0 Array (Thermo Fisher Scientific). Breast cancer cases were selected on the basis of clinical behaviour: 17 cases were breast cancers without metastases after 10 years follow-up (BC-MO), and 17 (BC- M1 ) were cases who had developed distant metastases within 5 years follow-up (n=14) or showed synchronous metastases at diagnosis (n=3) (Table 1 ). As control, 3 normal breast tissues from reductive mammoplasty (Normal Breast Tissue, NBT) and 3 in situ carcinomas (CIS) were also analysed. As shown in Figure 1, miRNA expression levels were able to differentiate NBTs, CIS and non-metastatic BC from patients with metastases. By this analysis, 8 miRNAs were found differentially expressed microRNAs (fold change >2 among MO and M1 cases (Table 2). Table 1. Baseline clinical-pathological characteristics of the discovery cohort (n=34).

[0104] Abbreviations - IQR: Interquartile Range (i.e. first-third quartiles); Table 2 miRNA differentially expressed in non-metastatic (BC-MO) versus metastatic (BC-M1 ) patients by Affymetrix GeneChip analysis.

[0105] An RT-qPCR approach was used to evaluate the accuracy of the 8 miRNAs in discriminating subjects who developed distant metastases in an extended patient cohort including 223 BC (Validation cohort) cases without metastases at diagnosis with a median follow up of 6.6 years (Table 3).

[0106] Table 3. Baseline clinical-pathological characteristics of the Validation cohort

[0107] (n=223)

[0108] Abbreviations - IQR: Interquartile Range (i.e. first-third quartiles).

[0109] A stepwise multivariable logistic regression was used to identify which of all the 8 miRNAs were the most associated with the development of metastases. The procedure starts with a null model (no covariates are included) and, at each step, the model is updated including (forward movement) or removing (backward movement) a specific covariate from the previous model, in order to achieve an optimal set of covariates that minimize the model’s Akaike Information Criteria (AIC), which quantifies the amount of information loss due to this update. This stepwise logistic regression robustly identified miR-3916 and miR-3613-5p as the chosen covariates, providing the minimum model’s AIC along with the highest discriminatory ability within 15 years of follow up (12, 13, 14, 15).

[0110] As shown in Table 4, compared with a higher risk of developing metastases, the expressions of miR-3916 and miR-3613-5p were down-regulated (Odds Ratio (OR)=0.56, 95%CI: 0.37-0.80, p=0.004) and up-regulated (OR=1.50, 95%CI: 1.03- 2.22, p=0.038), respectively. ORs were reported for each unitary increase in one standard deviation of the natural log transformed miRNA values.

[0111] Table 4. Results from stepwise multivariable logistic regression model.

[0112] Abbreviations: OR: Odds Ratio; SD: Standard Deviation; Cl: Confidence Interval. *OR computed for each unitary increase in one SD of miRNA log-expression.

[0113] To assess the robustness and consistency of the set of identified covariates, this model was also performed in bootstrap datasets, showing the number of times, out of 5000 resamples of data, that each covariate: 1 ) was selected, 2) had a positive or negative sign of the estimated regression coefficient, and

[0114] 3) was statistically significant (i.e. p<0.05) out of 5000 bootstrap resamples of data. Out of 5000 bootstrap resamples, the stepwise model which considered all miRNAs as candidate variables chose miR-3916 and miR-3613-5p in 88.6% and 62.5% of times, respectively (Table 5), thus providing high consistency in identifying the final model, which included only two covariates. Moreover, when adjusted for the main tumor characteristics (Stage, histological grade, ki67 / Mib1 ) the expressions of miR- 3916 and miR-3613-5p continued to be down-regulated (OR=0.42, 95%CI: 0.23-0.70, p=0.002) and up-regulated (OR=2.06, 95%CI: 1.27-3.50, p=0.005), respectively.

[0115] Table 5. Statistics on bootstrap data resamples (n=5000 resamplings)

[0116] To determine whether miR3916 and miR3613-5p expression improves the ability to discriminate patients at risk to develop distant metastases, were evaluated the performance in terms of AUC of: 1 ) a multivariable logistic model including the expression level of miRNA3916 and miR3613-5p, and 2) a multivariable logistic model including only the main tumor characteristics stage (AJCC 7th edition), histological grade and Ki67 (“base model”). The AUC computed using the predicted individual probabilities estimated from the multivariable logistic models were: 0.68 (95%CI: 0.59-0.77) for the model including only the two miRNAs (Figure 2A), and 0.76 (95%CI: 0.68-0.84) for the “base model” (Figure 2B). However by adding the expression levels of miR-3916 and miR-3613-5p, the base model significantly improved its discriminatory power to AUC=0.85 (95%CI: 0.79-0.91 ) (delta-AUC p=0.001 , Figure 2B). This model including main tumour characteristics and miRNAs expression was defined as the “extended” model. As shown in Figure 3, the “extended model” is better calibrated (ICI=0.0324) with respect to the “base model” (ICI=0.0398).

[0117] On the base of the extended model, for each subject, a LP (i.e. the Linear Predictor) value, which corresponds to the logit of the probability (i.e. the natural logarithm of the odds) of developing a distant metastasis, can be calculated directly from the multivariable logistic model by simply performing a weighted linear combination of the values of the independent covariates (i.e. X1-X5) using the estimated regression coefficients provided by the Maximum Likelihood Estimation procedure (i.e. -2.498, bi, b2, 0.018, -0.915 and 0.731 for X1-X5 respectively) as weights, as follows:

[0118] LP = -2.498 + (bi x Xi) + (b2x X2) + (0.018 x X3) + (-0.915 x X4) + (0.731 x X5)

[0119] The independent covariates XrX5were coded as follows:

[0120] X1 : Stage of the tumour

[0121] X2: Notthingham Histological Grade

[0122] X3: Ki67 / mib1 (%)

[0123] X4: The natural logarithm of the miR3916 expression

[0124] X5: The natural logarithm of the miR3613-5p expression With regard to the covariates Xi and X2, the corresponding regression coefficients (bi and b2) can take on different values, depending on the categories defined for each covariate under consideration, as detailed below: if X1= “1A” then b1= 0 or if X1= “1 B” then b1= -14.317 or if X1= “2A” then b1= 0.054 or if X1= “2B” then b1= 0.739 or if X1= “3A” then b1= 1.294 or if X1= “3B” then b1= 1.087 or if X1= “3C” then b1= 2.989 if X2= “1” then b2= 0 or if X2= “2” then b2= -0.231 or if X2= “3” then b2= -0.474

[0125] By applying the inverse formula (the 'expit' function) to LP, an estimate of the individual’s risk (%) of developing distant metastases within 15 years after the diagnosis is provided, as shown in the following formula:

[0126] 1

[0127] Risk (%) = 100 x -

[0128] 1 + e ’LP

[0129] As shown in Figure 4, on the basis of predicted probabilities estimated from the “extended” model, patients are classified as follows:

[0130] - low risk of developing distant metastases if they have a risk (%) <10.6%;

[0131] - intermediate risk of developing distant metastases if they have a risk (%) between 10.6% and 38.1%: or - high risk of developing distant metastases if they have a risk (%) >38.2% of developing distant metastases.

[0132] The patients’ classification is also summarized in table 6.

[0133] Table 6. Patients’ classification

[0134] *Risk score estimation could not take place for 24 patients, as they had missing values for some covariates included in the extended model.

[0135] References

[0136] 1. Sung H, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021 71 :209-229 doi.org / 10.3322 / caac.21660.

[0137] 2. Edge SB et al AJCC cancer staging manual(7thedition), New York: Springer 2010.

[0138] 3. Bloom HJ, Richardson WW. Histological grading and prognosis in breast cancer: A study of 1409 cases of which 359 have been followed for 15 years. Br J Cancer 1957 11 :359-77. doi: 10.1038 / bjc.1957.43.

[0139] 4. Elston CW, Ellis IO Pathological prognostic factors in breast cancer I: The value of histological grade in breast cancer: Experience from a large study with long-term follow-up. Histopathology 1991 19:403-10. doi: 10.1111 / j.1365- 2559.1991. tb00229.x.

[0140] 5. Duffy MJ, et al. Clinical use of biomarkers in breast cancer: Updated guidelines from the European Group on Tumor Markers (EGTM). Eur J Cancer. 2017 75:284-298. doi: 10.1016 / j.ejca.2O17.01 .017. Epub 2017 Feb

[0141] 28. Pasculli B, et al. Epigenetics of breast cancer: Biology and clinical implication in the era of precision medicine. Semin Cancer Biol 2018 51 :22-35. doi: 10.1016 / j.semcancer.2O18.01 .007. Esteller M Non-coding RNAs in human disease, Nat Rev Genet 2011 12:861 - 74. doi: 10.1038 / nrg3074. Berindan-Neagoe I et al. MicroRNAome genome: a treasure for cancer diagnosis and therapy. CA Cancer J Clin. 2014 64:311 -36. doi: 10.3322 / caac.21244. Epub 2014 Aug 7. Barbano R, et al. Stepwise analysis of MIR9 loci identifies miR-9-5p to be involved in Oestrogen regulated pathways in breast cancer patients. Sci Rep. 2017 7:45283. doi: 10.1038 / srep45283. Fontana A et al Combined analysis of miR-200 family and its significance for breast cancer Sci Rep. 2021 ;11 (1 ):2980. doi: 10.1038 / s41598-021 -82286-1 . Pasculli et al. Hsa-miR-155-5p Up-Regulation in Breast Cancer and Its Relevance for Treatment With Poly[ADP-Ribose] Polymerase 1 (PARP-1) Inhibitors Front Oncol. 2020 10:1415. doi: 10.3389 / fonc.2020.01415. eCollection 2020. Vergouwe Y, et al. Substantial effective sample sizes were required for external validation studies of predictive logistic regression models. J Clin Epidemiol 2005 58):475-83. doi: 10.1016 / j.jclinepi.2004.06.017. Elisabeth R. DeLong, et al. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988 44(3):837-45. Austin PC, Steyerberg EW. The Integrated Calibration Index (ICI) and related metrics for quantifying the calibration of logistic regression models. Stat Med. 2019 20;38:4051 -4065. doi: 10.1002 / sim.8281 . Epub 2019 Jul 3. Gbnen M. Analyzing Receiver Operating Characteristic Curves with SAS. Cary, NC: SAS Institute Inc, 2007. Hosmer, D. W., Lemeshow, S. (2000). Applied logistic regression. John Wiley and Sons. ISBN: 0471356328, 9780471356325

Claims

Claims1. A method in-vitro or ex-vivo for calculating the risk score of developing metastases, in a patient affected by breast cancer, comprising the following steps: a) detecting the expression level of miR-3916 and miR3613-5p in a biological sample obtained from a patient; b) determining the stage of the tumor; c) determining the percentage of KI67 / Mib1 positive cells; d) determining the histological grade of the tumor; e) calculating the risk score of said patient.

2. A method according to claim 1 wherein the risk score of the patient is calculated in step e) by a multivariate logistic regression model according to the following formula:1 risk score (%) = 100 x -1 + e ’LPwhereinLP = -2.498 + (bi x Xi) + (b2x X2) + (0.018 x X3) + (-0.915 x X4) + (0.731 x X5); and wherein the independent covariates XrX5were coded as follows:Xi: Stage of the tumorX2: Histological GradeX3: percentage of Ki67 / mib1 positive cellsX4: miR3916 expression levelX5: miR3613-5p expression; and and wherein the regression coefficients bi and b2have the following values:if X1 = “stage 1A” then b1 = 0, or if X1= “stage 1 B” then b1= -14.317, or if X1= “stage 2A” then b1= 0.054, or if X1= “stage 2B” then b1= 0.739, or if X1 = stage “3A” then b1 = 1 .294, or if X1= “stage 3B” then b1= 1.087, or if X1= “stage 3C” then b1= 2.989, if X2= “histological grade 1” then b2= 0, or if X2= “histological grade 2” then b2= -0.231 , or if X2= “histological grade 3” then b2= -0.474.

3. A method according to claim 1 , wherein the expression level of miR-3916 and miR3613-5p in step a) is detected by quantitative RT-PCR, ddPCR or probebased methods.

4. A method according to claim 1 , wherein the stage of the tumor in step b) is identified by analyzing the tumor size and the presence / absence of lymph node metastasis, preferably according to AJCC staging system.

5. A method according to claim 1 , wherein the percentage of KI67 / Mib1 positive cells in step c) is detected by immunohistochemistry.

6. A method according to claim 1 , wherein the histological grade of point d) is detected with the Nottingham histological grading system.

7. A method according to any of the previous claims, wherein a subject with a risk score > 38.2% is considered at high risk of developing metastasis, a subject with a risk score between 10.6 and 38.1% is considered at an intermediate risk of developing metastasis, and a subject with a risk score < 10.6% is considered at a low risk of developing metastasis.

8. A method according to the previous claims, wherein the biological sample is a tissue sample, preferably a fresh tissue sample, a frozen tissue sample or a formalin fixed paraffin embedded sample (FFPE), or liquid biopsy (plasma or serum).

9. A method according to any of the previous claims, wherein said breast cancer is of the molecular subtype selected from Luminal A, Luminal B, HER2- amplified and Basal-like or of the surrogate molecular subtype selected from Hormone Receptor positive, HER2 amplified and Triple negative.

10. A computer medium comprising instructions which, when executed by a computer, cause the computer to carry the following steps: a) receiving raw data from the expression level obtained from the method of claim 1 ; b) integrating said data with clinical date determined at point b), c), and d) of claim 1 ; c) calculating the LP; d) transforming the LP to individual risk score; wherein the computer calculates a risk score of developing metastases in a patient affected by breast cancer.

Images