FIELD OF THE INVENTION

The invention relates to a method of pre-surgical risk stratification of a prostate cancer subject. Moreover, the invention relates to a diagnostic kit, to a use of the kit in a method of pre-surgical risk stratification of a prostate cancer subject, to a use of a gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in pre-surgical risk stratification of a prostate cancer subject, and to a corresponding computer program product.

BACKGROUND OF THE INVENTION

Cancer is a class of diseases in which a group of cells displays uncontrolled growth, invasion and sometimes metastasis. These three malignant properties of cancers differentiate them from benign tumors, which are self-limited and do not invade or metastasize. Prostate Cancer (PCa) is the most commonly-occurring non-skin malignancy in men. It displays as a heterogeneous disease with varying potential to develop progressively to deadly forms of the disease. Of the estimated 417,000 annual new cases in Europe, around 92,000 will die from their disease (see Ferlay J. et ah, GLOBOCAN 2012 vl .0, Cancer

Incidence and Mortality Worldwide: I ARC CancerBase No. 11 [Internet], Lyon, France, International Agency for Research on Cancer, 2013).

Clinically, various schemes for pre-surgical risk classification have been developed based upon longitudinal biological patient outcomes (see Rodrigues G. et al,“Pre- treatment risk stratification of prostate cancer patients: A critical review”, Canadian

Urological Association Journal, Vol. 6, No. 2, pages 121-127, 2012). While active surveillance (AS) is recommended by the various national and international guidelines for men with very low and low risk prostate cancer (see Mohler J. et al.,“NCCN clinical practice guidelines in oncology: Prostate cancer, Version 1.2016”, Journal of the National

Comprehensive Cancer Network, Vol. 14, No. 1, pages 19-30, 2016), there is a significant sub-group in this patient population with a risk of 10 to 25% cancer recurrence after primary treatment (see, for example, Hernandez D.J. et al.,“Contemporary evaluation of the D’amico risk classification of prostate cancer”, Journal of Urology, Vol. 70, No. 5, pages 931-935, 2007). These patients suffer from the burden of follow-up treatments that are typically triggered by biochemical relapse. Likewise, in the intermediate risk group there is a sub- population with low risk of biochemical progression (see, for example, Jung J.W. et al., “Stratification of patients with intermediate-risk prostate cancer”, BJU International, Vol.

115, No. 6, pages 907-912, 2015). Nevertheless, this group is heterogeneous, comprising patients with varied outcomes, including those with aggressive pathological characteristics (see Abem M.R. et al.,“Delayed radical prostatectomy for intermediate-risk prostate cancer is associated with biochemical recurrence: Possible implications for active surveillance from the SEARCH database”, The Prostate, Vol. 73, No. 4, pages 409-417, 2013).

Clinical risk descriptors do not delineate effectively either the extent of the disease or its aggressiveness for all patients. Thus, there is a need for better patient stratification in order to optimize primary treatment decisions.

SUMMARY OF THE INVENTION

It is an object of the invention to provide a method of pre-surgical risk stratification of a prostate cancer subject, which may allow making better primary treatment decisions. It is a further object of the invention to provide a diagnostic kit, a use of the kit in a method of pre-surgical risk stratification of a prostate cancer subject, a use of a gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in pre-surgical risk stratification of a prostate cancer subject, and a corresponding computer program product.

In a first aspect of the present invention, a method of pre-surgical risk stratification of a prostate cancer subject is presented, comprising:

determining a gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in a biological sample obtained from the subject,

determining an expression based risk score for the subject based on the gene expression profile, and

determining a pre-surgical prognostic risk score for the subject based on the expression based risk score and pre-surgical clinical variables of the subject.

The cAMP signaling pathway is known to play an important role in both the development and progression of prostate cancer (see Merkle D. and Hoffmann R.,“Roles of cAMP and cAMP-dependent protein kinase in the progression of prostate cancer: Cross-talk with the androgen receptor”, Cellular Signalling, Vol. 23, No. 3, pages 507-515, 2011).

While a family of adenylate cyclases is responsible for the synthesis of cAMP, cyclic nucleotide phosphodiesterases (PDEs) appear to represent the only cellular mechanism for its destruction. PDEs provide both signal termination and, importantly, the

compartmentalization of cAMP signaling within the 3D matrix of cells. This is achieved through the spatially discrete destruction of cAMP via sub-populations of distinct PDE iso forms sequestered by localized anchor proteins/signalosomes (see, for example, Conti M. and Beavo J.,“Biochemistry and physiology of cyclic nucleotide phosphodiesterases:

essential components in cyclic nucleotide signaling”, Annual Review of Biochemistry, Vol. 76, pages 481-511, 2007). Thus changes in the expression and/or activity of distinct PDE iso forms can alter downstream signaling pathways during disease development and progression, providing potential targets for novel biomarkers and for targeted therapeutic intervention. Indeed, alterations in the expression of members of the cAMP-degrading PDE4 family appear to be associated with a number of different diseases, including stroke, acrodysostosis, schizophrenia, and COPD. Recently, it was shown that down-regulation of a particular PDE4 iso form (PDE4D7) may have an impact on prostate cancer (see, for example, Bottcher R. et ah,“Human phosphodiesterase 4D7 (PDE4D7) expression is increased in TMPRSS2-ERG positive primary prostate cancer and independently adds to a reduced risk of post-surgical disease progression”, Britisch Journal of Cancer, Vol. 113, No. 10, pages 1502-1511, 2015). PDE4D7 iso form is a so-called long iso form as it contains both the UCR1 and UCR2 regulatory domains. UCR1 is found in long, but not short, PDE4 iso forms and allows for regulation by various protein kinases, including PKA and MK2 and also determines the functional outcome of catalytic unit phosphorylation by ERK.

Functionally, it provides part of the cellular desensitization system to cAMP and enables cross-talk between signaling pathways that lead to the activation of ERK and AMPK, for example.

By determining an expression based risk score for a prostate cancer subject based on the gene expression profile of PDE4D7, additional molecular information representing the biology of the disease is obtained. The prognostic power of PDE4D7 is utilized in pre-surgical patient risk assessment by determining a pre-surgical prognostic risk score that is not only based on pre-surgical clinical variables of the subject but that is further based on the expression based risk score. This may allow for an improved stratification of the subject in a pre-surgical setting that may result in better primary treatment decisions. For instance, the pre-surgical prognostic risk score may allow to make better recommendations on whether to select active surveillance vs. active intervention, e.g., radical prostatectomy, for certain sub-populations of prostate cancer patients. The term“phosphodiesterase 4D7” or“PDE4D7” refers to the splice variant 7 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D7 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001165899.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO: 19, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D7 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:20, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001159371.1 encoding the PDE4D7 polypeptide. The term “phosphodiesterase 4D7” or“PDE4D7” also relates to the amplicon that can be generated by the primer pair PDE4D7_forward (SEQ ID NO:2l) and the PDE4D7_reverse (SEQ ID NO:22) and can be detected by probe SEQ ID NO:23.

The PDE4D7 polypeptide can also be detected with primer pair PDE4D7- 2_forward (SEQ ID NO:24) and the PDE4D7_reverse (SEQ ID NO:25) and can be detected by probe SEQ ID NO:26.

The term“PDE4D7” also comprises nucleotide sequences showing a high degree of homology to PDE4D7, e.g., nucleic acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO: 19 or amino acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO:20 or nucleic acid sequences encoding amino acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO:20 or amino acid sequences being encoded by nucleic acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO: 19.

The term“biological sample” or“sample obtained from a subject” refers to any biological material obtained via suitable methods known to the person skilled in the art from a subject, e.g., a prostate cancer patient. The biological sample used may be collected in a clinically acceptable manner, e.g., in a way that nucleic acids (in particular RNA) or proteins are preserved.

The biological sample(s) may include body tissue and/or a fluid, such as, but not limited to, blood, sweat, and urine. Furthermore, the biological sample may contain a cell extract derived from or a cell population including an epithelial cell, such as a cancerous epithelial cell or an epithelial cell derived from tissue suspected to be cancerous. The biological sample may contain a cell population derived from a glandular tissue, e.g., the sample may be derived from the prostate of a male subject. Additionally, cells may be purified from obtained body tissues and fluids if necessary, and then used as the biological sample. In some realizations, the sample may be a tissue sample, a urine sample, a urine sediment sample, a blood sample, a saliva sample, a semen sample, a sample including circulating tumor cells, extracellular vesicles, a sample containing prostate secreted exosomes, or cell lines or cancer cell line.

In one particular realization, biopsy or resections samples may be obtained and/or used. Such samples may include cells or cell lysates.

It is also conceivable that the content of a biological sample is submitted to an enrichment step. For instance, a sample may be contacted with ligands specific for the cell membrane or organelles of certain cell types, e.g., prostate cells, functionalized for example with magnetic particles. The material concentrated by the magnetic particles may

subsequently be used for detection and analysis steps as described herein above or below.

Furthermore, cells, e.g., tumor cells, may be enriched via filtration processes of fluid or liquid samples, e.g., blood, urine, etc. Such filtration processes may also be combined with enrichment steps based on ligand specific interactions as described herein above.

The term“prostate cancer” refers to a cancer of the prostate gland in the male reproductive system, which occurs when cells of the prostate mutate and begin to multiply out of control. Typically, prostate cancer is linked to an elevated level of prostate-specific antigen (PSA). In one embodiment of the present invention the term“prostate cancer” relates to a cancer showing PSA levels above 4.0. In another embodiment the term relates to cancer showing PSA levels above 2.0. The term“PSA level” refers to the concentration of PSA in the blood in ng/ml.

The term“non-progressive prostate cancer state” means that a sample of an individual does not show parameter values indicating“biochemical recurrence” and/or “clinical recurrence”.

The term“progressive prostate cancer state” means that a sample of an individual shows parameter values indicating“biochemical recurrence” and/or“clinical recurrence”.

The term“biochemical recurrence” generally refers to recurrent biological values of increased PSA indicating the presence of prostate cancer cells in a sample.

However, it is also possible to use other markers that can be used in the detection of the presence or that rise suspicion of such presence. The term“clinical recurrence” refers to the presence of clinical signs indicating the presence of tumor cells as measured, for example using in vivo imaging.

The term“prognosticating prostate cancer” as used herein refers to the prediction of the course or outcome of a diagnosed or detected prostate cancer, e.g., during a certain period of time, during a treatment or after a treatment. The term also refers to a determination of chance of survival or recovery from the disease, as well as to a prediction of the expected survival time of a subject. A prognosis may, specifically, involve establishing the likelihood for survival of a subject during a period of time into the future, such as 6 months, 1 year, 2 years, 3 years, 5 years, 10 years or any other period of time.

It is preferred that the pre-surgical clinical variables comprise one or more of: (i) an age of the subject; (ii) a prostate-specific antigen (PSA) level; (iii) a primary and secondary biopsy Gleason score; (iv) a clinical stage; and (v) a percentage of tumor positive biopsies.

Treatment decisions in primary, localized prostate cancer are largely subject to a combination of the risk of future disease progression and life expectancy. The National Comprehensive Cancer Network (NCCN) has defined five risk categories based on pre- surgical clinical variables (see Mohler J. et al.). For each risk group, ranging from very low, low, intermediate (dichotomized into favorable vs. unfavorable intermediate), high and very high risk, several options of interventions are presented in current practice prostate cancer guidelines. More advanced tools of clinical risk prediction have been presented in the past in the form of mathematical models, which combine the value of clinical variables into a single score (see Lughezani G. et al.,“Predictive and prognostic models in radical prostatectomy candidates: A critical analysis of the literature”. European Urology, Vol. 58, No. 5, pages 687-700, 2010). One of the most extensively validated clinical risk algorithms for pre- surgical decision support is the pre-surgical CAPRA score (see Cooperberg M.R.,“The UCSF Cancer of the Prostate Risk Assessment (CAPRA) Score: A straightforward and reliable pre-operative predictor of disease recurrence after radical prostatectomy”, Journal of Urology, Vol. 173, No. 6, pages 1938-1942, 2005). The score is a combination of clinically available information, i.e., patient age, pre-operative PSA, biopsy Gleason, percentage of tumor positive biopsies, and clinical stage. Initially published in 2005, this score has been validated in several studies since then (see Brajtbord J.S. et al.,“The CAPRA score at 10 years: Contemporary perspectives and analysis of supporting studies”, European Urology, Vol. 71, No. 5, pages 705-709, 2017). By combining the molecular information provided by the expression based risk score with the information from such extensively validated pre- surgical clinical variables, a pre-surgical prognostic risk score with an improved prognostic power may be obtained.

It is also preferred that the method further comprises:

determining a pre-surgical Cancer of the Prostate Risk Assessment (CAPRA) score for the subject,

wherein the pre-surgical prognostic risk score is determined by combining the expression based risk score and the pre-surgical CAPRA score.

As mentioned above, the pre-surgical CAPRA score is one of the most extensively validated clinical risk algorithm for pre-surgical decision support in prostate cancer. It provides a categorical score between 1 and 10 with three categories of low risk (pre-surgical CAPRA scores 0 to 2), intermediate risk (pre-surgical CAPRA scores 3 to 5), and high risk (pre-surgical CAPRA scores 6 to 10). In view of its current level of validation as a prognostic algorithm in prostate cancer as well as its easy-to-interpret single score output, the pre-surgical CAPRA score can advantageously be combined with the expression based risk score into a pre-surgical prognostic risk score that may easily be determined in clinical practice and that may allow for a further improvement in pre-surgical prognosis compared to the use of the pre-surgical CAPRA algorithm alone.

It is further preferred that the expression based risk score and the pre-surgical CAPRA score are combined with a regression function derived from a population of prostate cancer subjects.

Regression analysis helps one understand how the typical value of the dependent variable (or“criterion variable”) changes when any one of the independent variables is varied, while the other independent variables are held fixed. This relationship between the dependent variable and the independent variables is captured in the regression function, which can be used to predict the dependent variable given the values of the independent variables. The dependent variable can be, for example, a binary variable, such as biochemical relapse within 5 years after surgery. In this case, the regression is a logistic regression that is based on a logit function of the independent variables, which, here, comprise or consist of the expression based risk score and the pre-surgical CAPRA score. By means of the regression function, an improved prediction of e.g. the 5-year risk of biochemical recurrence after surgery may be possible.

In an alternative, it is preferred that the pre-surgical prognostic risk score is determined as a modified pre-surgical Cancer of the Prostate Risk Assessment (CAPRA) score for the subject, in which a primary and secondary biopsy Gleason score is replaced by the expression based risk score.

The biopsy Gleason score has, on one hand, a significant impact to the pre- surgical CAPRA score. At the same time, however, it was found by the present inventor that Gleason scoring is subject to substantial variability amongst pathologists. By replacing the information of the biopsy Gleason score within the pre-surgical CAPRA score with the molecular information provided by the expression based risk score, a modified pre-surgical CAPRA score may be obtained that can be more reliable and less susceptible to variations in the assessment provided by different pathologists.

It is further preferred that the expression based risk score is a value in a predefined range, wherein depending on the value a number of points in the range from 0 to 3 are added in the modified pre-surgical CAPRA score.

The primary and secondary biopsy Gleason score is considered in the pre- surgical CAPRA score as follows: If both the primary and secondary biopsy Gleason score are in the range of 1 to 3, no point is added in the modified pre-surgical CAPRA score.

Alternatively, if the primary biopsy Gleason score is in the range of 1 to 3 and the secondary biopsy Gleason score is in the range of 4 to 5, one point is added in the modified pre-surgical CAPRA score. Finally, if the primary biopsy Gleason score is in the range of 4 to 5 and the secondary biopsy Gleason score is in the range of 1 to 5, three points are added in the modified pre-surgical CAPRA score. By adding, depending on the value of the expression based risk score, a number of points in the range from 0 to 3 in the modified pre-surgical CAPRA score, the overall structure of the resulting modified pre-surgical CAPRA score can be kept the same with a minimum total score of 0 and a maximum total score of 10.

In one preferred example, the expression based risk score is a value in the range of 1 to 5 and three points are added in the modified pre-surgical CAPRA score if the value is in the range of 1 to <2, whereas two points are added if the value is in the range of 2 to <3, one point is added if the value is in the range of 3 to <4, and no point is added if the value is in the range of 4 to <5.

It is preferred that the method further comprises:

proposing a primary treatment for the subject based on the pre-surgical prognostic risk score,

wherein the primary treatment is selected from the group consisting of: (i) at least a partial prostatectomy; (ii) an active therapy selected from radiation treatment, hormone therapy, chemotherapy, and a combination thereof; and (iii) active surveillance. As mentioned above, the various national and international guidelines recommend different treatments for a prostate cancer subject depending on the risk of future disease progression and life expectancy. For example, for men with very low and low risk prostate cancer active surveillance (AS) is generally recommended, whereas for high risk cancer a radical prostatectomy could be indicated. However, as also mentioned before, the known clinical risk descriptors do not delineate effectively either the extent of the disease or its aggressiveness for all patients. For instance, it has been found that in the NCCN very low and low risk groups, there is a significant sub-group of patients population with a risk of 10 to 25% cancer recurrence after primary treatment. Likewise, it is known that in the intermediate risk group there is a sub-population with low risk of biochemical progression. By basing the proposing of a primary treatment for the subject on the pre-surgical prognostic risk score, better recommendations on e.g. whether to select active surveillance vs. active intervention, e.g., radical prostatectomy, may be made for certain sub-populations of prostate cancer patients.

It is further preferred that the method comprises:

normalizing the gene expression profile with respect to one or more reference genes selected from the group consisting of: Homo sapiens hypoxanthine

phosphoribosyltransferase 1 (HPRT1), Tubulin- Alpha- lb (TUBA1B), Homo sapiens pumilio RNA-Binding Family Member (PUM1), and Homo sapiens TATA box binding protein (TBP),

wherein the expression based risk score is determined based on the normalized gene expression profile.

By normalizing the gene expression profile with respect to one or more reference genes and by determining the expression based risk score is determined based on the normalized gene expression profile, variability in the determination of the expression based risk score can be reduced. This enables differentiation between real variations in gene expression profiles and variations due to the measurement processes. In this respect, it has been found that HPRT1, TUBA1B, PUM1, and TBP are particularly well suited as reference genes for normalizing the PDE4D7 gene expression profile.

The gene expression profile may be determined by detecting mRNA expression using one or more primers and/or probes and/or one or more sets thereof.

Moreover, the gene expression profile may be determined by an amplification based method and/or microarray analysis and/or RNA sequencing. The determining of the gene expression profile may include performing Real-Time Quantitative Polymerase Chain Reaction (RT- qPCR) on RNA extracted from the biological sample. In other embodiments, the gene expression profile is determined by RNA sequencing, conventional PCR (using, e.g., end point analysis by gel electrophoresis), or multiplex-PCR. In the case of RT-qPCR, the determining of the gene expression profile may include determining a threshold cycle (Ct) value for PDE4D7 and each of the one or more reference genes. The PCR may be performed with at least one primer and/or probe for measuring a reference gene selected from HPRT1, TUBA1B, PUM1, and TBP.

It is preferred that the one or more reference genes comprise at least two, or at least three of HPRT1, TUBA1B, PUM1, and TBP. In a particularly preferred realization, the one or more reference genes comprise all of HPRT1, TUBA1B, PUM1, and TBP.

Other reference genes which may be additionally or alternatively used for normalizing the PDE4D7 gene expression profile include: Homo sapiens actin, beta, mRNA (ACTB); Homo sapiens 60S acidic ribosomal phosphoprotein P0 mRNA (RPLP0);

Polymerase (RNA) II (DNA Directed) Polypeptide A, 220kDa (POLR2A); Beta-2- Microglobulin (B2M); and Aminolevulinate-Delta-Synthase (ALAS-l).

It is further preferred that the expression based risk score is determined with a scoring function, based on the gene expression profile, the scoring function having been derived from gene expression profiles for biological samples of prostate cancer subjects.

Herein, it is particularly preferred that the scoring function is based on the normalized gene expression profile, e.g., the gene expression profile normalized with respect to all of HPRT1, TUBA1B, PUM1, and TBP, and that the scoring function is derived from correspondingly normalized gene expression profiles for biological samples of prostate cancer subjects. In one preferred realization, the scoring function is a linear transformation that transforms the normalized gene expression profile into a predefined range of values, such as the above-mentioned range of 1 to 5. Such a transformation can be determined by considering the frequency distribution of the normalized gene expression profile values for PDE4D7 for biological samples of a population of prostate cancer subjects and by determining the transformation that transforms the frequency distribution into the desired range. By making use of such a scoring function, the expression based risk score can be expressed in a way that is intuitive to a user, such as in a small positive value range. This is similar to other categories used in the clinical routine, e.g., in histo-pathology grading (Gleason) or multi-parametric MRI radiology scoring (PIRADS).

In one particular realization, the expression based risk score is determined as follows: EBRS = (((PDE4D7_norm + A)*B)+l), (1) where“EBRS” is the expression based risk score,“PDE4D7_norm” is the normalized PDE4D7 gene expression profile value, and A and B are variables.

In one example, A may be about 6-8, such as 6.7167499999999, B may be 0.4-0.45, such as 0.420780231744713, and the expression based risk score may be a value in the range of 1 to 5 (as mentioned above). The expression based risk score can may also be classified or categorized into one of at least two risk groups, based on the value of the expression based risk score. For example, there may be two risk groups, or three risk groups, or four risk groups, or more than four predefined risk groups. Each risk group covers a respective range of (non-overlapping) expression based risk score. For example, a risk group may include all expression based risk scores from 1 to <2, another risk group from 2 to <3, another risk group from 3 to <4, and another risk group from 4 to <5.

It is particularly preferred that the determining of the gene expression profile comprises performing RT-qPCR on RNA extracted from the biological sample, wherein a Cq value is determined for PDE4D7 and for each of the one or more reference genes, and wherein the determining of the expression based risk score includes normalizing the Cq value for PDE4D7 using the Cq value for each of the one or more reference genes and computing the expression based risk score as a linear function of the normalized Cq value.

For example, the normalized Cq value for PDE4D7 may be generated by applying the following:

N(CqpDE4D7) - Mean(Cq _ref _genes) - (CqpDE4D7), (2) where N(Cqpo _E 4 _D 7) is the normalized genes expression profile value

(quantification cycle, Cq) of PDE4D7, Mean(Cq _{ref genes} ) is the arithmetic mean of the PCR Cq values of the one or more reference gene, and Cqpo _E 4 _D 7 is the PCR Cq value of PDE4D7.

It is preferred that the gene expression profile further includes expression information from phosphodiesterase 4D variant 5 (PDE4D5) and/or from phosphodiesterase 4D variant 9 (PDE4D9), wherein an expression based risk score is determined for the subject for each of the phosphodiesterase 4D variants based on the gene expression profile, and wherein the pre-surgical prognostic risk score for the subject is determined based on the expression based risk scores and the pre-surgical clinical variables of the subject. In a further aspect of the present invention, a diagnostic kit is presented, comprising:

at least one primer and/or probe for determining the gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in a biological sample obtained from a prostate cancer subject; and

optionally, at least one primer and/or probe for determining the gene expression profile for one or more reference genes selected from the group consisting of: Homo sapiens hypoxanthine phosphoribosyltransferase 1 (HPRT1), Tubulin- Alpha- lb (TUBA1B) Homo sapiens pumilio R A-Binding Family Member (PUM1), and Homo sapiens TATA box binding protein (TBP); and

optionally, at least one agent for determining a prostate-specific antigen (PSA) level in a biological sample obtained from the subject; and

optionally, instructions for computing a pre-surgical prognostic risk score based on the gene expression profile for PDE4D7 and pre-surgical clinical variables of the subject, the instructions optionally being stored on a computer program product which, when executed by a computer, perform a method comprising:

determining an expression based risk score for the subject based on the gene expression profile for PDE4D7, and

determining the pre-surgical prognostic risk score for the subject based on the expression based risk score and the pre-surgical clinical variables of the subject,

optionally, wherein the method comprises:

normalizing the gene expression profile for PDE4D7 with respect to the one or more reference genes,

wherein the expression based risk score is determined based on the normalized gene expression profile for PDE4D7,

optionally, wherein the pre-surgical clinical variables comprise the prostate- specific antigen (PSA) level.

The at least one agent for determining the prostate-specific antigen (PSA) level can be, e.g., a PSA specific antibody or the like.

In a further aspect of the present invention, a use of the kit as defined in claim 13 in a method of pre-surgical risk stratification of a prostate cancer subject is presented.

In a further aspect of the present invention, a use of a gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in pre-surgical risk stratification of a prostate cancer subject is presented, comprising: determining the gene expression profile in a biological sample obtained from the subject,

determining an expression based risk score for the subject based on the gene expression profile, and

determining a pre-surgical prognostic risk score for the subject based on the expression based risk score and pre-surgical clinical variables of the subject.

In a further aspect of the present invention, a computer program product is presented comprising instructions which, when the program is executed by a computer, cause the computer to carry out a method comprising:

determining a gene expression profile for phosphodiesterase 4D variant 7 (PDE4D7) in a biological sample obtained from a prostate cancer subject,

determining an expression based risk score for the subject based on the gene expression profile, and

determining a pre-surgical prognostic risk score for the subject based on the expression based risk score and pre-surgical clinical variables of the subject.

It shall be understood that the method of claim 1 , the diagnostic kit of claim 13, the use of the diagnostic kit of claim 14, the use of a gene expression profile of claim 15, and the computer program of claim 16 have similar and/or identical preferred embodiments, in particular, as defined in the dependent claims.

It shall be understood that a preferred embodiment of the present invention can also be any combination of the dependent claims or above embodiments with the respective independent claim.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following drawings:

Fig. 1 shows schematically and exemplarily a flowchart of an embodiment of a method of pre-surgical risk stratification of a prostate cancer subject,

Figs. 2 to 7 show results of Kaplan-Meier survival analysis of the time to PSA relapse or start of any salvage therapy after radical prostatectomy in the DB patient cohort

(#151),

Figs. 8 to 10 show results of ROC curve analysis of 5-year biochemical recurrence (BCR) after surgery in the DB patient cohort, Fig. 11 shows results of decision curve analysis of the net benefit of five different treatment strategies for men who experience disease recurrence within 5 years after surgery (in total 45 of the 151 investigated patients failed the initial primary treatment of surgery by PSA recurrence (29.8%)),

Figs. 12 to 17 show results of Kaplan Meier survival analysis of the time to PSA relapse (endpoint: BCR) in the RP patient cohort (n=536) for the PDE4D5, PDE4D7, and PDE4D9 scores,

Fig. 18 shows a heatmap of TMPRSS2-ERG negative tumor samples of the RP patient cohort (n=256),

Figs. 19 and 20 show results of Kapaln Meier survival analysis of biochemical recurrence (BCR) progression free survival amd secondary treatment (salvage radiation and or androgen deprivation) free survival (STFS) in the DB patient cohort for the categorized CAPRA & PDE4D5/7/9 combination score,

Figs. 21 and 22 show results of ROC curve analysis of 5-year biochemical recurrence (BCR) after surgery and 5 -year secondary treatment free survival (STFS) after surgery in the DB patient cohort for the CAPRA model vs. the CAPRA & PDE4D7 model vs. the CAPRA & PDE4D5/7/9 combination model,

Fig. 23 shows results of decision curve analysis in the DB patient cohort of the net benefit of four different treatment decision strategies for men at risk to experience disease recurrence within 5 years after surgery, and

Fig. 24 shows results of decision curve analysis in the DB patient cohort of the net reduction of two different treatment decision strategies for men at risk to experience disease recurrence within 5 years after surgery.

DETAIFED DESCRIPTION OF EMBODIMENTS

Overview of pre-surgical risk stratification

Fig. 1 shows schematically and exemplarily a flowchart of an embodiment of a method of pre-surgical risk stratification of a prostate cancer subject.

The method begins at step S100.

At step S102, a biological sample is obtained from each of a first set of patients (subjects) diagnosed with prostate cancer. Preferably, monitoring prostate cancer has been performed for these prostate cancer patients over a period of time, such as at least one year, or at least two years, or about five years, after obtaining the biological sample. At step S104, a gene expression profile for PDE4D7 is obtained for each of the biological samples obtained from the first set of patients, e.g., by performing RT-qPCR (real time quantitative PCR) on RNA extracted from each biological sample. The exemplary gene expression profile includes an expression level (e.g., value) for PDE4D7 which can be normalized using value(s) for each of a set of reference genes, such as HPRT1, TUBA1B, PUM1, and/or TBP. In one realization, the gene expression profile value of PDE4D7 is normalized to with respect to one or more reference genes selected from the group consisting of HPRT1, TUBA1B, PUM1, and TBP, e.g., at least one, or at least two, or at least three, or, preferably, all of these reference genes.

At step S106, a scoring function for assigning an expression based risk score is determined, based on the gene expression profile for PDE4D7 obtained for at least some of the biological samples obtained for the first set of patients and respective results obtained from the monitoring. In one preferred realization, the scoring function is a linear

transformation that transforms the normalized gene expression profile into a predefined range of values, such as the above-mentioned range of 1 to 5. As mentioned above, such a transformation can be determined by considering the frequency distribution of the normalized gene expression profile values for PDE4D7 for biological samples of a population of prostate cancer subjects (here, the first set of patients) and by determining the transformation that transforms the frequency distribution into the desired range. In one particular realization, the expression based risk score is determined as specified in Eq. (1) above.

At step S108, a biological sample is obtained from a patient (subject or individual). The patient can be a new patient or one of the first set.

At step Sl 10, a gene expression profile is obtained for PDE4D7, e.g., by performing PCR on the biological sample. In one realization, the gene expression profile value of PDE4D7 is normalized to with respect to one or more reference genes selected from the group consisting of HPRT1, TUBA1B, PUM1, and TBP, e.g., at least one, or at least two, or at least three, or, preferably, all of these reference genes. This is substantially the same as in step S104.

Other reference genes which may be additionally or alternatively used in steps S104 and Sl 10 include: Homo sapiens actin, beta, mRNA (ACTB); Homo sapiens 60S acidic ribosomal phosphoprotein P0 mRNA (RPLP0); Polymerase (RNA) II (DNA Directed) Polypeptide A, 220kDa (POLR2A); Beta-2-Microglobulin (B2M); and Amino levulinate- Delta-Synthase (ALAS-l). At step Sl 12, an expression based risk score is determined for the patient, based on the gene expression profile, using the derived scoring function.

At step Sl 14, a pre-surgical prognostic risk score is determined for the patient based on the expression based risk score and pre-surgical clinical variables of the patient.

This will be described in more detail later in the description.

At Sl 16, a therapy recommendation may be provided, e.g., to the patient or his or her guardian, to a doctor, or to another healthcare worker, based on the pre-surgical prognostic risk score. To this end, the pre-surgical prognostic risk score may be categorized into one of a predefined set of risk groups, based on the value of the pre-surgical prognostic risk score. Providing a therapy recommendation may include one or more of: a) proposing a therapy for the patient based on the assigned risk group, with at least two of the risk groups being associated with different therapies, b) computing a disease progression risk prediction of the patient before or after prostate surgery; and c) computing a therapy response prediction for the patient before or after prostate surgery. Example therapies include at least a partial prostatectomy, an active therapy selected from radiation treatment, chemotherapy, and a combination thereof, and observation alone, i.e., without performing prostatectomy or active therapy (i.e., active surveillance).

The method ends at Sl 18.

Each of the risk groups may be associated with a respective proposed therapy, which differs in its aggressiveness. Each proposed therapy may be based on the results of the patients from the first set that were assigned to that risk group and is one which is predicted to provide the least aggressive therapy which does not exceed a threshold clinical risk for development of prostate cancer. In some cases, this enables a new patient to be assigned to a risk group associated with a less aggressive proposed therapy than would be the case for other risk profiling methods, such as that using the Gleason score, the NCCN risk categories, or the pre-surgical CAPRA score.

In one embodiment, the gene expression profiles at steps S104 and Sl 10 are determined by detecting mR A expression using one or more primers and/or probes and/or one or more sets thereof.

A detailed description of PDE4D7 and the one or more reference genes including their Transcript ID (NCBI RefSeq) and the corresponding amino acid sequences for the primer pair and probe are shown in TABLE 3. This table also shows, for each gene, a sense primer, and antisense primer, and a probe sequence that specifically binds to the amplicon. TABLE 1 : Exemplary primer and probe nucleic acid sequences

To explore the prognostic power of PDE4D7 in pre-surgical patient risk assessment, the correlation to disease recurrence in the context of pre-surgical risk variables and algorithms like the pre-surgical CAPRA score were investigated. Combination models of the expression based risk score together with pre-surgical variables were developed in a surgery cohort and the model was validated in independent patients on diagnostic biopsy tissue. The results show that PDE4D7 may add additional information to pre-surgical variables or prognostic scores that are based on such variables, such as the pre-surgical CAPRA score, that may allow for better patient stratification in order to optimize primary treatment decisions.

EXAMPLES

Patient cohorts and samples

Two patient cohorts, a radical prostatectomy (RP) patient cohort and a diagnostic biopsy (DB) patient cohort, with the demographics shown in TABLE 2, were employed. For the RP patient cohort, a small biopsy punch (approximately 1 millimeter by 2 millimeters) of tissue was collected of a representative tumor area from the resected prostate from 550 patients who had been consecutively operated on between 2000 and 2004 at a single high- volume clinical center in Germany. After quality control of the study data based on pre-defmed criteria and removal of patients who underwent adjuvant hormone therapy 503 patient samples were found eligible for statistical analysis. For the DB patient cohort, a single biopsy punch (approximately 1 millimeter by 2 millimeters) was collected from the tumor positive diagnostic biopsy with the highest Gleason grade per patient. The 168 patients in this case were diagnosed with prostate cancer and operated on between 1995 and 2011 at the University Klinik Muenster, Germany. In total, diagnostic needle biopsy tissues of 151 patients were found eligible for statistical analysis.

TABLE 2: Demographics of the radical prostatectomy (RP) patient cohort and the diagnostic biopsy (DB) patient cohort

For patient age, preoperative PSA, percentage of tumor in biopsy, prostate volume, and PSA density, the minimum and maximum values in the each cohort are shown, while the median and IQR values are depicted in parentheses. For the CAPRA risk categories, the number of patients and percentage per risk group are shown. In case of pre- surgical pathology, the biopsy Gleason scores and the Gleason grade groups as well as clinical stages are indicated (by number and percentage of patients). Post-surgical pathology is represented by the pathology Gleason scores and Gleason grade groups, the pathology stages, the surgical margin status after prostatectomy, the tumor invasion status of the seminal vesicles and pelvic lymph nodes (by number and percentage of patients). In this respect, it is noted that the extracapsular extension was not provided as a primary parameter but was derived from pathology stage pT3a. The follow-up demonstrates the mean and median follow-up periods in months after surgery for all patients. The outcome category illustrates the cumulative 5- and lO-year biochemical recurrence (BCR) and clinical recurrence to metastases (CR) post-surgical primary treatment. The treatment category lists the cumulative 5- and lO-year start to salvage radiation therapy (SRT) or salvage androgen deprivation therapy (SADT) after surgery. Mortality is shown as prostate cancer specific survival (PCSS) as well as overall survival (OS). For all outcomes, the number of men experiencing the outcome per total number of men with the respective 5- or lO-year follow are shown, wherein the percentage of events is given in parentheses. (B) Demographics of the diagnostic biopsy patient cohort. (N/A=not available).

Laboratory methods All used laboratory methods including oligonucleotide primers and probes for RT-qPCR (quantitative real-time PCR), RNA extraction, and quality control and procedures to include/discard samples from the statistical analysis were as described previously in Bottcher R. et ah. The primers and probes used for the RT-qPCR to measure the genes of interest as well as the reference genes are also given in TABLE 1.

RESULTS

Correlation of the expression based risk score to longitudinal clinical outcomes

The continuous expression based risk score (i.e., the reference gene normalized and transformed expression of the PDE4D7 transcript) was correlated to pre- surgical clinical variables in the two patient cohorts: the RP patient cohort and the DB patient cohort.

TABLES 3 to 8 show the uni- and multivariate Cox regression analysis of the biochemical recurrence (BCR) free survival and overall survival (OS) of the continuous expression based risk scores. In the multivariate analysis, the expression based risk scores were adjusted by pre-surgical clinical variables, namely, by the age at surgery, the pre operative PSA, the PSA density, the biopsy Gleason grade group, the percentage of tumor positive biopsy cores, the percentage of tumor in biopsy, and the clinical stage (see TABLES 3, 5, and 7) or by the pre-surgical Cancer of the Prostate Risk Assessment (CAPRA) score (see TABLES 4, 6, and 8). The biopsy Gleason grade group and the clinical stage were modeled as categories with the lowest category used as a reference. All other demographic and clinical variables as well as the expression based risk scores were modeled as continuous variables. All variables were entered into the multivariate model and the respective statistical measures are given in the tables.

Univariate Cox regression analysis demonstrated a very significant correlation of the continuous expression based risk score to time to BCR in the two investigated patient cohorts with HR=0.53; 95% CU0.41-0.67; pO.OOOl, and HR=0.43; 95% CU0.33-0.65, p<0.000l, respectively (see TABLES 3 and 7). Adjusting the multivariate Cox regression analysis for the pre-surgical variables or the pre-surgical CAPRA score resulted in a significant independent contribution to the prediction of post-surgical BCR for the continuous expression based risk score (see TABLES 3, 4, 7, and 8). Moreover, when testing the endpoint overall survival (OS) in the RP patient cohort, we observed a similar

independent predictive power of the continuous expression based risk score when adjusting to the relevant pre-surgical clinical variables or the pre-surgical CAPRA score (see TABLES 5 and 6). Interestingly though, only age (HR=l.l; 95% 0=1.03-1.2; p=0.007) and clinical stage cT2 and cT3 remained significant predictors in the multivariate model apart from the expression based risk score (HR=0.43; 95% 0=0.29-0.62; p<0.000l). The pre-surgical CAPRA score was a significant predictor of overall survival in the univariate analysis (HR=l .2; 95% 0=1.3-1.6; p=O.Ol), while it fell below statistical significance in the multivariate analysis (HR=l.2; 95% Cl 0.99-1.4; p=0.06). The continuous expression based risk score remained very significant also in the multivariate modeling (HR=0.4; 95% 0=0.28-0.58; pO.OOOl; TABLES 5 and 6).

TABLE 3: Uni- and multivariate Cox regression analysis of the biochemical recurrence (BCR) free survival of the expression based risk score in the RP patient cohort (#503), wherein the expression based risk score was adjusted for pre-surgical variables in the multivariate analysis.

TABLE 4: Uni- and multivariate Cox regression analysis of the biochemical recurrence (BCR) free survival of the expression based risk score in the RP patient cohort (#503), wherein the expression based risk score was adjusted for the pre-surgical CAPRA score in the multivariate analysis.

TABLE 5: Uni- and multivariate Cox regression analysis of the overall survival (OS) of the expression based risk score in the RP patient cohort (#503), wherein the expression based risk score was adjusted for pre-surgical variables in the multivariate analysis.

TABLE 6: Uni- and multivariate Cox regression analysis of the overall survival (OS) of the expression based risk score in the RP patient cohort (#503), wherein the expression based risk score was adjusted for the pre-surgical CAPRA score in the multivariate analysis.

TABLE 7: Uni- and multivariate Cox regression analysis of the biochemical recurrence (BCR) free survival of the expression based risk score in the DB patient cohort (#151), wherein the expression based risk score was adjusted for pre-surgical variables in the multivariate analysis.

TABLE 8: Uni- and multivariate Cox regression analysis of the biochemical recurrence (BCR) free survival of the expression based risk score in the DB patient cohort (#151), wherein the expression based risk score was adjusted for the pre-surgical CAPRA score in the multivariate analysis.

Pre-surgical prognostic risk score based on the expression based risk score and pre-surgical clinical variables

To further explore the prognostic power of PDE4D7, the benefit of a combination of the expression based risk score with pre-surgical clinical variables used to prognosticate prostate cancer patients for various treatment regimens was tested. Based on the multivariate Cox regression data, it was hypothesized that a combination of the pre- surgical CAPRA score together with the expression based risk score will provide a significant improvement in prognostic power over pre-surgical clinical variables alone. To evaluate this hypothesis, a sub-cohort of 449 patients (92 events; 20.5%) of the RP patient cohort with complete 5-year outcome histories was selected and a logistic regression model to combine the expression based risk score with the pre-surgical CAPRA score to predict the 5 -year risk of biochemical recurrence after surgery was generated. The logit(p) regression function was transformed to p=l/(l+e ^A (-logit(p)) in order to calculate the probability p for an individual patient to experience a biochemical relapse within 5 years after surgery.

The modeling proved the independent predictive value of the expression based risk score to the pre-surgical CAPRA metric (Odds ratio 0.46; 95% Cl 0.3-0.69; p=0.0002; data not shown). Next, this pre-surgical CAPRA score and expression based risk score logistic regression model was tested independently on 151 patients who were eligible for statistical data analysis. All patients had a minimum of 60 months of follow-up after operation. The methods used to investigate the power of the pre-surgical CAPRA score and expression based risk score logistic regression model were (i) Kaplan-Meier survival analysis, (i) ROC curve analysis, as well as (iii) re-classification and decision curve analysis. Figs. 2 to 7 show results of Kaplan-Meier survival analysis of the time to PSA relapse or start of any salvage therapy after radical prostatectomy in the DB patient cohort (#151). Figs. 2 to 4: Kaplan-Meier analysis of the biochemical recurrence (BCR) free survival in the DB patient cohort of the categorized expression based risk score (indicated as “PDE4D7 Score” in Fig. 2), the pre-surgical CAPRA score categories (indicated as“CAPRA Score” in Fig. 3), and the pre-surgical CAPRA score and expression based risk score combination model (indicated as“CAPRA & PDE4D7 Score” in Fig. 4). Figs. 5 to 7:

Kaplan-Meier analysis of post-surgical time to start of any salvage therapy free survival in the DP patient cohort of the categorized expression based risk score (indicated as“PDE4D7 Score” in Fig. 5), the pre-surgical CAPRA score categories (indicated as“CAPRA Score” in Fig. 6), and the pre-surgical CAPRA score expression based risk score combination model (indicated as“CAPRA & PDE4D7 Score” in Fig, 7). In all figures, censored patients are indicated by vertical bars. The Hazard Ratio’s (HR) compared to the reference group and the logrank p-values are provided. The number of men at risk are given in the tables below the Kaplan-Meier survival graphs. The expression based risk score categories were defined as: “PDE4D7 Score (1-2)”: expression based risk score in the range of 1 to <2;“PDE4D7 Score (2-3)”: expression based risk score in the range of 2 to <3;“PDE4D7 Score (3-4)”:

expression based risk score in the range of 3 to <4; and“PDE4D7 Score (4-5)”: expression based risk score in the range of 4 to <5. The highest expression based risk score category “PDE4D7 Score (4-5)” was used as the reference category. The pre-surgical CAPRA score categories were defined as:“CAPRA scores (0-2)”: pre-surgical CAPRA scores in the range of 0 to 2;“CAPRA scores (3-5)”: pre-surgical CAPRA scores in the range of 3 to 5; and “CAPRA scores (>5)”: pre-surgical CAPRA scores in the range of 6 to 10. The lowest pre- surgical CAPRA score category“CAPRA scores (0-2)” was used as the reference category. The categories of the pre-surgical CAPRA score and expression based risk score combination model were defined according to the probability to experience PSA failure after surgery based on the logit(p) function of the logistic regression model as:“Probability (<0.l5)”; “Probability (0.15 to <0.25)”;“Probability (0.25 to <0.5)”; and“Probability (0.5 to 1.0)”.

The Kaplan-Meier survival studies on the DB patient cohort showed a significant separation of patients into different risks to experience post-operative PSA relapse or start of any salvage therapy after radical prostatectomy based on the pre-surgical measurement of the expression based risk score in a tissue punch of a diagnostic biopsy (see Fig. 2). Concerning the logistic regression model, which combines the pre-surgical CAPRA score and the expression based risk score in a regression function, the model-calculated probability to experience the endpoint was classified or categorized into four groups (as described above) and the hazard ratios between the lowest vs. highest recurrence risk groups were compared with the patient groups stratified by the pre-surgical CAPRA score categories alone. The addition of the expression based risk score to the pre-surgical CAPRA categories proved to significantly increase in the hazard ratio from the lowest to the highest risk group (9.0 vs. 39.6, respectively) while the patient numbers in these risk categories remained virtually the same (see Figs. 3 and 4).

When testing the clinically relevant endpoint of start of any salvage therapy (radiation, hormone ablation) after PSA failure, it was found that, as before, the results show an improved risk separation between patients with lowest vs. highest risk to receive post- surgical secondary treatment when combining the pre-surgical CAPRA score with the expression based risk score (see Figs. 5 to 7).

Furthermore, the prediction of 5-years PSA relapse after primary treatment of the pre-surgical CAPRA score vs. the logistic regression model, which combines the pre- surgical CAPRA score and the expression based risk score in a regression function and which was previously developed using the RP cohort with complete 5-year follow-up (#449), was tested. Using a ROC analysis, the 5-year AUCs (area under the curve) were calculated as 0.77 for the pre-surgical CAPRA score alone and as 0.82 for the logistic regression model (see Fig. 8). The AUCs of the two models were tested to be significantly different (p=0.004).

In another test, the benefit of adding the expression based risk score to the pre- surgical CAPRA score was explored. The pre-surgical CAPRA metric categorizes its individual components and gives weighted points to these categories. The pre-surgical CAPRA score is the sum of these weighted points (see Cooperberg M.R.). As mentioned above, the biopsy Gleason score has, on one hand, a significant impact to the CAPRA score. At the same time, however, it was found by the present inventor that Gleason scoring is subject to substantial variability amongst pathologists. Therefore, the information of the biopsy Gleason score within the pre-surgical CAPRA score was replaced with the molecular information provided by the expression based risk score. In particular, depending on the value of the expression based risk score, a number of points in the range from 0 to 3 were in a modified pre-surgical CAPRA score. By doing so, the overall structure of the resulting modified pre-surgical CAPRA score could be kept the same with a minimum total score of 0 and a maximum total score of 10 (see TABLE 9). TABLE 9: Replacement of the biopsy Gleason score by the expression based risk score in the pre-surgical CAPRA model.

When testing this model for 5-year BCR outcome prediction in ROC analysis, the resulting AUC of the“CAPRA (-BxGl/+PDE4D7) Score” remained the same as compared to the pre-surgical CAPRA score which includes the information of the biopsy Gleason score (AUC=0.77; see Fig. 9). While replacing the biopsy Gleason score with the expression based risk score does not impact the model performance, the 5 -year prediction of PSA failure of the pre-surgical CAPRA model drops significantly in case the biopsy Gleason information is removed from the clinical model (AUC=0.77 vs. AUC=0.69 with and without the biopsy Gleason score; p=0.008; see Fig. 9, where the pre-surgical CAPRA score without the biopsy Gleason information is indicated as“CAPRA Score (-BxGl)”). In the patient cohort with biopsy Gleason <6, only a limited number of men experience PSA relapse after surgery (21.2% in the RP cohort and in the DB cohort, respectively) or, even more relevant, progress to metastases (1.9% in the RP cohort), or suffer from disease specific death (1.3% in the RP cohort). Therefore, a sub-cohort analysis (#74) of the pre-surgical CAPRA score vs. the pre-surgical CAPRA score and expression based risk score combination model was performed in the DP patient cohort including only patients with a biopsy Gleason score >7. The combination model showed equivalent performance compared to testing the complete cohort (AUC=0.8 vs. 0.82, respectively), while there was more performance drop in the pre-surgical CAPRA model alone in the sub-cohort analysis (AUC=0.73 vs. 0.77, respectively; see Fig. 10). This may indicate an improved

discrimination performance of the combination model in patients with higher risk

characteristics.

Recently, the concept of decision curve analysis (DCA) was introduced into the field of testing the value of a biomarker or prediction model in clinical practice (see Vickers A.J. and Elkin E.B.,“Decision curve analysis: A novel method for evaluating prediction models”, Medical Decision Making, Vol. 26, No. 6, pages 565 to 574, 2006). DCA is a net benefit analysis, which compares the true-positive to the weighted false-positive rates across different risk thresholds to decide on the start of secondary treatment after surgery due to risk of future PSA failure. Here, the net benefit of secondary treatment triggered by PSA relapse after surgery was explored for the three models (i.e., the expression based risk score, the pre-surgical CAPRA score, and the pre-surgical CAPRA score and expression based risk score regression model) in decision curve analysis (DCA). Five different treatment strategies were compared: (1, 2) treat all or none patients; (3 to 5) treat according to the expression based risk score, the pre-surgical CAPRA score, or the combined regression model of the pre-surgical CAPRA score and the expression based risk score. For this, the true (TPR) and the false positive rate (FPR) was determined for all strategies between varying decision thresholds ranging from 0% to 50% in 5% steps (see Fig. 11). The analysis demonstrated that all used models show better net benefit compared to the“treat all” strategy while the pre- surgical CAPRA score and expression based risk score regression model revealed the best net benefit across all modeled decision thresholds.

The provided results show that a predictive model of the pre-surgical clinical risk algorithm CAPRA with quantitative measurements of the prostate cancer biomarker PDE4D7 in a biological sample of a prostate cancer subject may provide an improved risk stratification. It was demonstrated in multiple analyses that this risk prediction model performs better in stratifying prostate cancer patients to treatment relevant risk categories compared to using risk schemas based solely on pre-surgical clinical parameters, as recommended by the various currently employed national prostate cancer guidelines.

Discussion

Treatment decisions in primary, localized prostate cancer are largely subject to a combination of the risk of future disease progression and life expectancy. The provided data illustrate that the expression based risk score adds independent information to the pre- surgical CAPRA metric to predict disease recurrence, while in prediction of overall survival (of which -80% of the events is due to non-disease specific death) the expression based risk score remains the only variable that significantly contributes to survival prediction. Thus, PDE4D7 may be adding prognostic value to clinical prediction models based on pre-surgical variables, like the pre-surgical CAPRA score, for disease specific outcomes as well as to the prediction of survival to support treatment decision making.

Recently, the long-term results of the active surveillance cohort within the Goteborg randomized prostate cancer screening trial were published. This indicated that men with clinically low risk disease may have a considerable risk to experience progressive disease under a deferred treatment regime (see Godtman R.A. et ah,“Long-term Results of Active Surveillance in the Goteborg Randomized, Population-based Prostate Cancer Screening Trial”, European Urology, Vol. 70, No. 5, pages 760-766, 2016). Therefore, the present inventor questioned whether men other than those with very low risk disease would be eligible for expectant management strategies. The recent publication of the lO-year outcomes of the ProtecT study indicates similar conclusions in the active monitoring arm of the trial (see Hamdy F.C. et al,“lO-Year Outcomes after Monitoring, Surgery, or

Radiotherapy for Localized Prostate Cancer”, New England Journal of Medicine Vol. 375, pages 1415-1424, 2016). Although there is some debate about the validity of these results to contemporary practice (see Bergh R.C.N. van den, Murphy D.G., Poel H.G. van der, “Expectant Management for Prostate Cancer: Lessions from the past, Challenges for the Future”, European Urology, Vol. 70, pages 767-770, 2016), they may suggest that only patients with the very lowest risk are safe of any progression during deferred treatment management. While the use of clinical criteria like the pre-surgical CAPRA model allow the selection of such a low risk patient cohort of (45% of the RP cohort with a 9.3% risk of 5- year post-surgical BCR; 27.7% of the DP cohort with a 4.8% risk of 5-year post-surgical BCR), the addition of molecular markers may allow to enlarge this very low risk patient group. In fact, the developed pre-surgical CAPRA score and expression based risk score regression model defines a very low risk cohort of 38 out of 151 patients (25.2%) in the DB patient cohort with a NPV (negative predictive value) of 100% for a 5 -year risk of post treatment PSA recurrence.

Active surveillance (AS) has been established as a suitable and safe treatment alternative for men with low risk prostate cancer over the last years (see Garisto J.D. and Klotz L.,“Active surveillance for prostate cancer: How to do it right”, Oncology (Williston Park), Vol. 31, No. 5, pages 333 to 340 2017). A big challenge though associated with AS are the strict monitoring schedules that men are advised to follow in order to not miss signs of progressive disease like raise in PSA or up-grading in biopsy Gleason which are typically protocol triggers to switch from AS to active treatment. Longitudinal AS studies have published decreasing patient compliance to the monitoring protocols in AS over time in particular when it comes to additional biopsy procedures (see Bokhorst L.P. et ah, “Compliance Rates with the Prostate Cancer Research International Active Surveillance (PRIAS) Protocol and Disease Reclassification in Noncompliers”, European Urology, Vol.

68, No. 5, 2015). This issue can be addressed with a selection algorithm as proposed here, i.e., the combination of a clinical model like the pre-surgical CAPRA score with the prognostic genomic biomarker PDE4D7 to define a patient cohort with virtually no risk to progress over a period of 5 years. This provides a way forward to include men into active surveillance on the basis of very limited (or no) follow-up for a given time period after AS start.

Other variations to the disclosed realizations can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

In the claims, the word“comprising” does not exclude other elements or steps, and the indefinite article“a” or“an” does not exclude a plurality.

One or more steps of the method illustrated in Fig. 1 may be implemented in a computer program product that may be executed on a computer. The computer program product may comprise a non-transitory computer-readable recording medium on which a control program is recorded (stored), such as a disk, hard drive, or the like. Common forms of non-transitory computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tape, or any other magnetic storage medium, CD-ROM, DVD, or any other optical medium, a RAM, a PROM, an EPROM, a FLASH-EPROM, or other memory chip or cartridge, or any other non-transitory medium from which a computer can read and use.

Alternatively, the one or more steps of the method may be implemented in transitory media, such as a transmittable carrier wave in which the control program is embodied as a data signal using transmission media, such as acoustic or light waves, such as those generated during radio wave and infrared data communications, and the like.

The exemplary method may be implemented on one or more general purpose computers, special purpose computer(s), a programmed microprocessor or microcontroller and peripheral integrated circuit elements, an ASIC or other integrated circuit, a digital signal processor, a hardwired electronic or logic circuit such as a discrete element circuit, a programmable logic device such as a PLD, PLA, FPGA, Graphical card CPU (GPU), or PAL, or the like. In general, any device, capable of implementing a finite state machine that is in turn capable of implementing the flowchart shown in Fig. 1, can be used to implement one or more steps of the method of risk stratification for therapy selection in a patient with prostate cancer is illustrated. As will be appreciated, while the steps of the method may all be computer implemented, in some embodiments one or more of the steps may be at least partially performed manually.

The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified herein.

While the invention has described so far based on the gene expression profile for PDE4D7, which can include an expression level (e.g., value) for PDE4D7 which can be normalized using value(s) for each of a set of reference genes, the gene expression profile may further include expression information from other PDE4D variants. For example, the other PDE4D variant(s) may include one or more of PDE4D1, PDE4D2, PDE4D3, PDE4D4, PDE4D5, PDE4D6, PDE4D8 and PDE4D9. The diagnostic kit may then additionally comprise at least one primer and/or probe for determining the gene expression profile for each of the other PDE4D variant(s) in the biological sample obtained from the prostate cancer subject. Preferably, however, only the gene expression profile for PDE4D7, in particular, an expression level (e.g., value) for PDE4D7 which can be normalized using value(s) for each of a set of reference genes, is employed. The term“phosphodiesterase 4D1” or“PDE4D1” relates to the splice variant 1 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D1 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197222.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO:l, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D1 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:2, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184151.1 encoding the PDE4D1 polypeptide. The term “phosphodiesterase 4D1” or“PDE4D1” also relates to the amplicon that can be generated by the primer pair PDElDlD2_forward (SEQ ID NO:3) and the PDElDlD2_reverse (SEQ ID NO:4) and can be detected by probe SEQ ID NO:5.

The term“phosphodiesterase 4D2” or“PDE4D2” refers to the splice variant 2 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D2 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197221.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO:6, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D2 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:7, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184150.1 encoding the PDE4D2 polypeptide.

The term“phosphodiesterase 4D3” or“PDE4D3” refers to the splice variant 3 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D3 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 006203.4, specifically, to the nucleotide sequence as set forth in SEQ ID NO:8, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D3 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:9, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 006194.2 encoding the PDE4D3 polypeptide.

The term“phosphodiesterase 4D4” or“PDE4D4” refers to the splice variant 4 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D4 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001104631.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO: 10, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D4 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO: 11 , which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001098101.1 encoding the PDE4D4 polypeptide.

The term“phosphodiesterase 4D5” or“PDE4D5” refers to the splice variant 5 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D5 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197218.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO: 12, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D5 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO: 13, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184147.1 encoding the PDE4D5 polypeptide. The term “phosphodiesterase 4D5” or“PDE4D5” also relates to the amplicon that can be generated by the primer pair PDE4D5_forward (SEQ ID NO: 14) and the PDE4D5_reverse (SEQ ID NO: 15) and can be detected by probe SEQ ID NO: 16.

The term“phosphodiesterase 4D6” or“PDE4D6” refers to the splice variant 6 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D6 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197223.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO: 17, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D6 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO: 18, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184152.1 encoding the PDE4D6 polypeptide.

The term“phosphodiesterase 4D8” or“PDE4D8” relates to the splice variant 8 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D8 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197219.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO:27, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D8 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:28, which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184148.1 encoding the PDE4D8 polypeptide.

The term“phosphodiesterase 4D9” or“PDE4D9” relates to the splice variant 9 of the human phosphodiesterase PDE4D, i.e., the human phosphodiesterase PDE4D9 gene, for example, to the sequence as defined in NCBI Reference Sequence: NM 001197220.1, specifically, to the nucleotide sequence as set forth in SEQ ID NO:29, which corresponds to the sequence of the above indicated NCBI Reference Sequence of the PDE4D9 transcript, and also relates to the corresponding amino acid sequence for example as set forth in SEQ ID NO:30 which corresponds to the protein sequence defined in NCBI Protein Accession Reference Sequence NP 001184149.1 encoding the PDE4D9 polypeptide. The term “phosphodiesterase 4D9” or“PDE4D9” also relates to the amplicon that can be generated by the primer pair PDE4D9_forward (SEQ ID NO:3l) and the PDE4D9_reverse (SEQ ID NO:32) and can be detected by probe SEQ ID NO:33.

The terms“PDE4D1,”“PDE4D2,”“PDE4D3,”“PDE4D4,”“ PDE4D5,” “PDE4D6,”“PDE4D8,” and“PDE4D9” also comprises nucleotide sequences showing a high degree of homology to PDE4D1, PDE4D2, PDE4D3, PDE4D4, PDE4D5, PDE4D6,

PDE4D8 and PDE4D9 respectively, e.g., nucleic acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NOs: 1, 6, 8, 10, 12, 17, 27 or 29 respectively or amino acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO:2, 7, 9, 11, 13, 18, 28 or 30 respectively or nucleic acid sequences encoding amino acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO:2, 1, 9, 11, 13, 18, 28 or 30 or amino acid sequences being encoded by nucleic acid sequences being at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98% or 99% identical to the sequence as set forth in SEQ ID NO:l, 6, 8, 10, 12, 17, 27 or 29.

Adding information from PDE4D5 and PDE4D9 to the pre-surgical prognostic risk score

Like the PDE4D7 iso form, the PDE4D5 and PDE4D7 iso forms are long iso forms that each contain both the UCR1 and UCR2 regulatory domains. Here, we investigated the added value of including information from the gene expression profiles for PDE4D5 and PDE4D9 in the CAPRA & PDE4D7 score combination model to predict longitudinal clinical outcomes.

EXAMPLES

Patient cohorts and samples

Two patient cohorts, a radical prostatectomy (RP) patient cohort and a diagnostic biopsy (DB) patient cohort, with the demographics shown in TABLE 10, were employed. For the RP patient cohort, two small biopsy punches (approximately 1 millimeter by 2 millimeters) of tissue were collected of a representative tumor area (index lesion of the tumor) from the resected prostate from 606 patients who had been consecutively operated on between 2000 and 2004 at a single high-volume clinical center in Germany. For the DB patient cohort, a single biopsy punch (approximately 1 millimeter by 2 millimeters) was collected from the tumor positive diagnostic biopsy with the highest Gleason grade per patient. The 168 patients in this case were diagnosed with prostate cancer and operated on between 1995 and 2011 at the University Klinik Muenster, Germany. After quality control of the RNAseq and the RT-qPCR data 536 and 151 patient samples were found eligible for statistical analysis in the RP cohort and the DB cohort, respectively.

TABLE 10: Demographics of the radical prostatectomy (RP) patient cohort and the diagnostic biopsy (DB) patient cohort.

For patient age, preoperative PSA, percentage of tumor in biopsy, prostate volume, and PSA density, the minimum and maximum values in the each cohort are shown, while the median and IQR values are depicted in parentheses. For the CAPRA risk categories, the number of patients and percentage per risk group are shown. In case of pre- surgical pathology, the biopsy Gleason scores and the Gleason grade groups as well as clinical stages are indicated (by number and percentage of patients). Post-surgical pathology is represented by the pathology Gleason scores and Gleason grade groups, the pathology stages, the surgical margin status after prostatectomy, the tumor invasion status of the seminal vesicles and pelvic lymph nodes (by number and percentage of patients). In this respect, it is noted that the extracapsular extension was not provided as a primary parameter but was derived from pathology stage pT3a. The follow-up demonstrates the mean and median follow-up periods in months after surgery for all patients. The outcome category illustrates the cumulative 5- and lO-year biochemical recurrence (BCR) and clinical recurrence to metastases (CR) post-surgical primary treatment. The treatment category lists the cumulative 5- and lO-year start to salvage radiation therapy (SRT) or salvage androgen deprivation therapy (SADT) after surgery. Mortality is shown as prostate cancer specific survival (PCSS) as well as overall survival (OS). For all outcomes, the number of men experiencing the outcome per total number of men with the respective 5- or lO-year follow are shown, wherein the percentage of events is given in parentheses. (B) Demographics of the diagnostic biopsy patient cohort. (N/A=not available).

Laboratory methods

To account for potential tumor heterogeneity the two tissue punches of the RP cohort were combined before nucleic acid extraction. A potential difference in tumor cellularity of the tissue punches was addressed by normalization of the RT-qPCR results of the PDE4D transcripts to the four reference genes HPRT1, TUBA1B, PUM1, and TBP. All used molecular laboratory methods including oligonucleotide primers and probes for RT- qPCR (quantitative real-time PCR), RNA extraction, and quality control and procedures to include/discard samples from the statistical analysis were described before (see Alves de Inda M. et al.,“Validation of Cyclic Adenosine Monophosphate Phosphodiesterase-4D7 for its Independent Contribution to Risk Stratification in a Prostate Cancer Patient Cohort with Longitudinal Biological Outcomes”, European Urology Focus, 2017).

RNA sequencing

RNA Sample Processing: 100 ng of total RNA was used as input to remove ribosomal RNA using Ribo-Zero Gold (Human/Mouse/Rat) rRNA Removal Kit (Illumina Inc.) according to the instructions of the manufacturer. For library construction, we used the total of the depleted RNA as input into the Scriptseq V2 RNA-Seq Library Preparation Kit (Epicentre/Illumina Inc.). Prepared RNAseq libraries were sequenced using a NextSeq 500 sequencing system (paired-end; 2 x 75 bp read length; approx. 80 million total reads per sample).

RNAseq Data Processing: The RNAseq raw data were pre-processed using Illumina’s bcl2fastq software incorporating a filtering by phred scores, thereby reducing low quality reads. Since FFPE may degenerate the bases the sequencing results have been filtered using a scoring algorithm to select reads representing the high-quality fraction. The score was calculated for a set of reads in a sample as follows: The set of reads was aligned against a reference genome. The alignment result for each read (i.e. the number of bases mapping correctly to the reference genome) was counted per read. The total number of successfully mapped bases was summed over all reads of the set. This sum was divided by the total number of bases of the set. The resulting relative number was called the score: Score = sum over all reads in sample (N BASES ALIGNED per

(2) read) / N BASES SAMPLE;

A score filter was used to select the subset of reads which contribute to the EQ Score by their good alignment result (all or most of the bases map correctly to the genome). The derived subset of high-quality reads was selected for further processing. If reads were mapped by fragmenting them (which may be required when aligning RNA) the measure was calculated based on the fragments alignment quality and the fragments were selected accordingly.

Read quality filtering: To retain only high quality reads the following filtering steps were applied. Reads were discarded when >50% of the bases had a phred score <11; bases at the read ends cut read if phred score <11; reads <63 bases discarded; reads with unknown (N) base calls were deleted; only read were kept pairs where each read passed the quality filter.

Gene expression calculation: To ensure comparability of expression values between samples all read counts were normalized by the transcripts per million method (TPM) as implemented in the RSEM algorithm (see Li B. et al,“RNA-Seq gene expression estimation with read mapping uncertainty”, Bioinformatics, Vol. 26, No. 4, pages 493-500, 2010).

Data analysis and statistics

Generation of normalized PDE4D transcript expression was performed by subtracting the RT-qPCR Cq of the respective PDE4D transcript from the averaged RT- qPCR Cq of the reference genes. Normalized PDE4D5, PDE4D7, and PDE4D9 expression profiles were transformed to PDE4D5, PDE4D7, and PDE4D9 scores, as outlined for PDE4D7 in detail above. In correlation analysis for various available biological and treatment related outcomes (see TABLE 10), the PDE4D transcript scores were either used as a continuous or as a categorical variable defined as: a) PDE4D5/7/9 score (l<2); b)

PDE4D5/7/9 score (>2 and <3); c) PDE4D5/7/9 score (>3 and <4); and d) PDE4D5/7/9 score (>4 and <5). The CAPRA risk score and corresponding low (1), intermediate (2), high-risk (3) categories were calculated as described in Cooperberg M.R.. Uni- and multivariate Cox regression and Kaplan Meier analyses were applied to correlate biochemical recurrence (BCR) progression free survival, or secondary treatment (salvage radiation and or androgen deprivation) free survival (STFS) to the PDE4D7 score in the RP cohort (n=536) and the RP* cohort from Taylor B.S. et al,“Integrative genomic profiling of human prostate cancer”, Cancer Cell, Vol. 18, No. 1, pages 11-22, 2010 (178 patients of which 130 were included in our study) and the DB cohort (h=151). To determine the TMPRSS2-ERG status of patient samples in Exon Array cohorts, we used relative ERG expression values and applied

Partitioning Around Medoids (PAM, R-package‘cluster’, k = 2) to assign the patient samples to the ERG positive or negative group based on expression. Decision curve analyses was performed as described in Vickers A.J. et al,“Extension to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers”, BMC Medical Informatics and Decision Making, Vol. 8, pages 1-17, 2008. For statistical analysis the software package MedCalc (MedCalc Software BVBA, Ostend, Belgium) was used.

RESULTS

Association of PDE4D transcript scores to longitudinal clinical outcomes depends on the TMPRSS2-ERG fusion status

Firstly, we set out to do Kaplan-Meier survival analysis of the PDE4D7 score categories in TMPRSS2-ERG rearrangement positive vs. gene fusion negative patient samples. In total 536 patient samples with data on TMPRSS2-ERG status were included of which 280 (52.2%) were defined as fusion positive while 256 samples (47.8%) were defined to be absent of this prostate specific fusion event. Biochemical recurrence (BCR) was selected as a surrogate endpoint for post-surgical disease progression due to the significant number of events for this outcome in the studied patient cohorts (see TABLE 10).

A clear difference in BCR progression free survival analysis was observed between the fusion positive vs. negative tumors with a highly significant logrank p-value (<0.000l) for the PDE4D7 categories in the presence of the rearranged TMPRSS2-ERG gene fusion (see Fig. 12). The patient group with the highest level of PDE4D7 expression (i.e., PDE4D7 scores 4-5) showed the lowest risk of disease progression after surgery in the TMPRSS2-ERG fusion positive cancers. In contrast, in prostate tumors without an ERG gene fusion event the discrimination in Kaplan-Meier survival between the defined four different PDE4D7 categories was non-significant (logrank p-value = 0.08; see Fig. 13). Interestingly, when looking at BCR progression free survival analysis of the PDE4D5 score we found the opposite situation compared to what was observed for the PDE4D7 scores. Only in gene fusion free tumors the PDE4D5 score categories significantly (logrank p-value<0.000l) predicted biochemical relapse (see Figs. 14 and 15). We observed a similar result as for the PDE4D5 scores also for the PDE4D9 score categories with a logrank p-value<0.000l in survival analysis in TMPRSS2-ERG negative tumors. However, in contrast to the analysis of PDE4D5 scores, the survival analysis of PDE4D9 score categories in gene fusion positive cancers also resulted in a significant association to biochemical recurrence although with a somewhat weaker p-value compared to the TMPRSS2-ERG negative tumors (logrank p- value=0.005 vs. logrank p<0.000l, respectively; see Figs. 16 and 17).

Next, it was investigated to what extent the score categories for the three different prostate cancer expressed PDE4D transcripts were determined to be mutually exclusive in individual patient samples or whether the same score category (e.g., [1-2] or [4- 5]) was seen across the same samples for the three studied PDE4D splice variants. For this, we plotted a heatmap (not shown in the figures) including all 536 patient samples with an initial split of between TMPRSS2-ERG gene fusion positive vs. fusion negative samples. While the samples within the TMPRSS2-ERG negative samples were ordered according to their PDE4D5 or PDE4D9 score category the samples with positive for the gene fusion were ordered according to their PDE4D7 score category from low to high. The heatmaps replicated the results of the Kaplan-Meier survival analysis with more events in the lower PDE4D iso form score categories. However, the PDE4D transcript score categories were to some extent non-overlapping within a patient sample. When focusing on the lowest score category (i.e., all scores for PDE4D5 / PDE4D7 / PDE4D9 between [1-2]) we identified 31 samples with at least one of the three PDE4D transcripts with a score category [1-2] (see Fig. 18). For three samples (marked in bold) we measured the lowest score category for all three PDE4D transcripts while for two samples (marked in bold italic) at least two PDE4D transcripts belonged to the lowest score category. For the other 26 samples only one PDE4D splice variant was measured as expressed very low while the two respectively other isoforms showed higher expression levels in these samples. The risk to develop metastases or to die from prostate cancer (6 and 5 out of the 31 patients, respectively) increases strongly with reduced expression levels of multiple long prostate expressed PDE4D iso forms (see Fig. 18). Also, the time scale after surgery to an event like BCR was generally shorter (< 2 years) in those patients with at least two PDE4D transcript scores [1-2] and/or [2-3] Vice versa, the higher the expression level of at least one of the three PDE4D splice variants the less likely was the chance to experience BCR after surgery. Or, in case the event happened, it was typically on a longer time scale (2-5 years, in some cases >5 years after primary treatment). Taken together, these data indicate that, next to PDE4D7, other long transcripts of PDE4D, namely PDE4D5 and PDE4D9, may also have significant prognostic value in prostate cancer. Therefore, we hypothesized that the addition of the PDE4D5 and PDE4D9 score, to that of PDE4D7, might increase the power to predict post-surgical risk of disease progression either over the clinical variables or over the above-described PDE4D7 based model (see also Alves de Inda M. et al. and van Strijp D. et al.,“The Prognostic PDE4D7 Score in a Diagnostic Biopsy Prostate Cancer Patient Cohort with Longitudinal Biological Outcomes”, Prostate Cancer, 2018).

Logistic regression model of clinical variables and prostate cancer expressed long PDE4D transcripts

To test this concept, we developed a prognostic model to include the clinical CAPRA score (see Cooperberg M.R.) together with the gene expression profile scores of PDE4D5, PDE4D7 and PDE4D9. For model development the RP (n=536) and RP* cohorts (n=l30) were used. Logistic regression analysis was performed to predict post- surgical biochemical relapse in the RP and RP* cohorts to estimate the weights for the CAPRA score as well as for the PDE4D transcripts. The coefficients were calculated by logistic regression. Next, we adjusted the initial coefficients after logistic regression analysis of the four model inputs on the RP* cohort by calculating an average of the coefficients for the RP and RP* cohorts, thus taking the heterogeneity of different patient groups into account. The final CAPRA & PDE4D5/7/9 model (coi · PDE4D5 score + co ₂ · PDE4D7 score + co ₃ · PDE4D9 score + co ₄ · CAPRA score, where coi, co ₂ , co ₃ and co ₄ are the regression coefficients) was tested for its prognostic power to predict BCR as well as start of secondary treatment of after surgery (i.e., radiation, or hormone deprivation) in the independent DB patient cohort. For any other outcome, like metastases or death, we used the RP and RP* cohorts (note: these clinical endpoints were not used during model development). An overview of the logistic regression modelling is provided in the following TABLE 11.

TABLE 11 : The CAPRA & PDE4D5/7/9 score combination model was developed by logistic regression of the CAPRA score and the normalized PD4D5, PDE4D7, and PDE4D9 expression values in the RP patient cohort (n=480 with completed 5-years follow-up after surgery) and in the RP* cohort (n=l30; Taylor et al, 2010). The logistic regression coefficients for the two cohorts were averaged to generate a mean coefficient for PD4D5, PDE4D7, and PDE4D9 and the CAPRA score based on the data of two independent patients cohorts in order to take the variation of different patient groups into account. These coefficients were used as weights in the final CAPRA & PDE4D5/7/9 regression model. Note: The CAPRA score is calculated based on Cooperberg M.R.; however, as the information on the number of positive biopsy cores was missing for the RP* cohort, the CAPRA score for this cohort was calculated using only patient age, pre-operative PSA, biopsy Gleason score, and clinical stage. The influence of the missing information on the biopsy cores was very limited as tested on the RP cohort as well as the DB cohort (data not shown).

Logistic Regression RP Cohort (PDE4D5, PDE4D7, PDE4D9, CAPRA)

Coefficients and Standard Errors

Logistic Regression RP* Cohort (PDE4D5, PDE4D7, PDE4D9, CAPRA)

Coefficients and Standard Errors

Average Model

Coefficients

a 459 samples for CAPRA analysis

In a particularly preferred realization,

the coefficient coi for PDE4D5 is in the range from -1.26 to -0.26, preferably, in the range from -1.16 to -0.36, more preferably, in the range from -1.06 to -0.46, more preferably, in the range from -0.96 to -0.56, more preferably, in the range from -0.86 to -0.66, most preferably, -0.76, and/or

the coefficient co ₂ for PDE4D7 is in the range from -1.21 to -0.21, preferably, in the range from - 1.11 to -0.31 , more preferably, in the range from - 1.01 to -0.41 , more preferably, in the range from -0.91 to -0.51, more preferably, in the range from -0.81 to -0.61, most preferably, -0.71, and/or

the coefficient co ₃ for PDE4D9 is in the range from -1.23 to -0.23, preferably, in the range from -1.13 to -0.33, more preferably, in the range from -1.03 to -0.43, more preferably, in the range from -0.93 to -0.53, more preferably, in the range from -0.83 to -0.63, most preferably, -0.73, and/or

the coefficient co ₄ for the pre-surgical CAPRA score is in the range from 0.15 to 1.15, preferably, in the range from 0.25 to 1.05, more preferably, in the range from 0.35 to 0.95, more preferably, in the range from 0.45 to 0.85, more preferably, in the range from 0.55 to 0.75, most preferably, 0.65.

Kaplan-Meier survival analysis of the CAPRA & PDE4D5/7/9 model

In Kaplan-Meier survival analysis the CAPRA & PDE4D5/7/9 model stratified 29 men (19.2%) of the total DB cohort (h=151) within the lowest score class [1-2] into a patient group with no risk over the follow-up period of 60 to 200 months of PSA relapse nor any risk of starting secondary treatments (see Figs. 19 and 20). By slightly increasing the cut-off of this model score category from [1-2] to [1-2.1] the number of men in this group with no risk of post- surgical disease progression will increase from 29 to 36 subjects (23.8%; data not shown). In contrast, the patient with the highest categories of CAPRA & PDE4D5/7/9 scores of [3-4] and [4-5] experience a risk of biochemical progression within 5 years after surgery of 63.9% and 83.3%, respectively. Similarly, the risk to undergo secondary treatment estimated from the survival analysis within a period of 5 years post-surgery was 44.1% and 75% for these patient groups, respectively (see Figs. 19 and 20).

ROC curve analysis of the CAPRA & PDE4D5/7/9 model

For the DB cohort (as above) BCR and start of secondary therapy were tested as outcomes. We compared the CAPRA & PDE4D5/7/9 model with the above-described CAPRA & PDE4D7 model (see also van Strijp D. et ah). For both tested clinical endpoints we identified an increase in the AUC (Area Under the Curve) of 10% and 6%, respectively, compared to the CAPRA score alone and 5% and 4%, respectively compared the CAPRA & PDE4D7 model (see Figs. 21 and 22).

To further explore this effect we tested outcomes other than biochemical relapse. As we developed the combination model of the CAPRA and the PDE4D transcript scores using BCR as an endpoint in the two radical prostatectomy cohorts (RP and RP*) we did not test the model on that endpoints in these cohorts. Instead we used other outcomes for testing like the progression to metastases after surgery or death from prostate cancer after primary (i.e., RP) or secondary treatments (i.e., SRT - salvage radiation therapy; SADT - salvage androgen deprivation therapy) to investigate the potentially added value of PDE4D5 and PDE4D9 to the earlier described CAPRA & PDE4D7 model. TABLE 12 provides an overview of the increase in AUCs (areas under the curves) of up to 5% and up to 12% compared to the CAPRA score or the CAPRA & PDE4D7 score model, respectively when using the additional prognostic value of PDE4D5 and PDE4D9 in addition to the CAPRA & PDE4D7 score model. This data indicate that the use of additional prostate relevant PDE4D transcripts may increase the prognostic power of our previously published combination model of the CAPRA and PDE4D7 score. TABLE 12: Overview of the AUCs for the CAPRA score, the CAPRA & PDE4D7, and the

CAPRA & PDE4D5/7/9 regression models to predict multiple endpoints in various patient cohorts.

The patient cohort that was used for the respective endpoint is indicated including the number of patients with respective follow-up periods. The tested clinical endpoints are given including the number and percentage of the respectively tested events. Note: The CAPRA score is calculated based on Cooperberg M.R.; however, as the information on the number of positive biopsy cores was missing for the RP* cohort the CAPRA score for this cohort was calculated using patient age, pre-operative PSA, biopsy Gleason score, and clinical stage only. The influence of the missing information on the biopsy cores was very limited as tested on the RP as well as the DB cohort.

Decision curve analysis of the CAPRA & PDE4D5/7/9 model Decision curve analysis is a net benefit analysis, which compares the true- positive to the weighted false-positive rates across different risk thresholds which a clinician / patient might want to accept (see Vickers A.J. et ah,“Net benefit approaches to the evaluation of prediction models, molecular markers, and diagnostic tests”, BMJ, Vol. 25, 2016). We explored the net benefit of avoiding primary treatment based on the predicted risk of a PSA relapse after surgery for the CAPRA score and the CAPRA & PDE4D7 score vs. the here presented CAPRA & PDE4D5/7/9 combination model. The analysis demonstrated that all models showed better net benefit compared to the“treat all” strategy while the combination model of CAPRA with the three PDE4D transcripts revealed the best net benefit across all modeled decision thresholds (see Fig. 23). Similarly, the net reduction analysis in primary treatment revealed a substantial difference in treatment reduction between the CARPA score and the CARPA & PDE4D5/7/9 combination model across all decision thresholds (see Fig. 24). Furthermore, the addition of PDE4D5 and PDE4D9 to the CAPRA & PDE4D7 model clearly improves the net benefit in decision curve analysis as well as would more effectively reduce the number of interventions per 100 patients compared to the CAPRA combination model with PDE4D7 alone.

We illustrated that the earlier presented CAPRA & PDE4D7 risk model can be further improved by adding other long PDE4D transcripts into the model. The rational for this added prognostic benefit of PDE4D5 and PDE4D9 is supported by the differences in prediction power between TMPRSS2-ERG positive vs. gene fusion negative patient tumors. By complementing PDE4D7 with the two other prostate cancer relevant PDE4D transcripts PDE4D5 and PDE4D9 we built a more comprehensive prognostic model to assess the risk of disease progression before primary intervention.

Discussion

It has been shown that a predictive model of the clinical risk algorithm

CAPRA in combination with the prostate cancer biomarker PDE4D7 provides value to prostate cancer risk stratification (see also Alves de Inda M. et al. and van Strijp D. et ah). The provided experimental results demonstrate that PDE4D7 adds independent value to the clinical CAPRA model and significantly improves the prognostic power to predict post- surgical disease progression. In order to further increase the value of this CAPRA &

PDE4D7 combination mode we identified expression differences of various long PDE4D iso forms in primary tumor material which were different for the prostate cancer specific TMRPSS2-ERG gene rearrangement (see Bottcher R. et al.,“Human PDE4D isoform composition is deregulated in primary prostate cancer and indicative for disease progression and development of distant metastases”, Oncotarget, Vol. 7, No. 43, pages 70669-70684, 2016).

The study data presented here enabled us to dissect the impact of three different PDE4D transcripts PDE4D5, PDE4D7 and PDE4D9 on the risk of post- surgical disease progression depending on the genomic background of the patient’s tumor.

Interestingly, PDE4D7 was found to be associated significantly with post-treatment disease recurrence in a TMRPSS2-ERG fusion positive background while less prognostic in patients without this particular gene fusion event. In contrast, PDE4D5 and PDE4D9 were highly prognostic in a non- fusion genomic background while PDE4D5 was not and PDE4D9 was less significantly associated with disease progression when the genomic fusion was present.

Active surveillance (AS) has become an accepted treatment alternative and is recommended by the national guidelines for men with low- and very-low risk prostate cancer (see Briganti A. et al,“Active Surveillance for Low-risk Prostate Cancer: The European Association of Urology Position in 2018”, European Urology, Vol. 74, No. 3, pages 357-368, 2018). The guiding principle of AS is to delay, not to avoid the primary treatment. The switch from AS to active intervention should be taken while the treatment intent is still curative. Consequently, men in AS have to follow strict monitoring schedules as

discontinuation and switch to active treatment takes place at the earliest sign of disease progression like a raise in PSA, a biopsy Gleason score or clinical stage migration. However, taken the low mortality risk of men in the active monitoring arm of the ProtecT trial into consideration it is questionable to what extent these observed changes in clinical presentation of the disease correlate with true biological disease progression (see Hamdy C.F. et al.,“10- Year Outcomes after Monitoring, Surgery, or Radiotherapy for Localized Prostate Cancer”, The New England Journal of Medicine, Vol. 375, pages 1415-1424, 2016).

Currently, new technology like multi-parametric MRI, or genomics is considered for stratification of men to active surveillance or for monitoring of men in AS (see Eineluoto J.T.,“Repeat multiparametric MRI in prostate cancer patients on active

surveillance”, PLoS ONE, Vol. 12, No. 12, 2017, and Canfield S. et al.,“Active Surveillance Use Among a Low-risk Prostate Cancer Population in a Large US Payer System: 17-Gene Genomic Prostate Score Versus Other Risk Stratification Methods”, Reviews in Urology,

Vol. 19, No. 4, pages 203-212, 2017). While the longitudinal cost of AS has been estimated to reach the same order of magnitude as primary interventions (see Keegan K.A. et al., “Active surveillance for prostate cancer compared with immediate treatment: an economic analysis”, Cancer, Vol. 118, No. 14, pages 3512-3518, 2012) driven by the cost of repeated biopsies in particular (see Dall’Era M.A.,“The economics of active surveillance for prostate cancer”, Current Opinion in Urology, Vol. 23, No. 3, pages 278-282, 2013), any newly implemented technical tool might only be cost-effective if its use will lead to less men discontinuing AS with a switch to definitive treatment and/or if surveillance schedules will be minimized (or even avoided in some patients). We believe that the combination of a clinical metric like the CAPRA score with genomic biomarkers like those presented here, namely evaluation of long form

PDE4D(5/7/9) transcripts to predict the future risk of a patient to experience disease progression, may provide future support for selecting patients to be included into active surveillance that require very limited (little or no) follow-up for a defined time period after start of (active) surveillance.

Any reference signs in the claims should not be construed as limiting the scope.

The invention relates to a method of pre-surgical risk stratification of a prostate cancer subject, comprising determining a gene expression profile for

phosphodiesterase 4D variant 7 (PDE4D7) in a biological sample obtained from the subject, determining an expression based risk score for the subject based on the gene expression profile, and determining a pre-surgical prognostic risk score for the subject based on the expression based risk score and pre-surgical clinical variables of the subject. This may allow for an improved stratification of the subject in a pre-surgical setting that may result in better primary treatment decisions. For instance, the pre-surgical prognostic risk score may allow to make better recommendations on whether to select active surveillance vs. active intervention, e.g., radical prostatectomy, for certain sub-populations of prostate cancer patients.

The attached Sequence Listing, entitled 2017PF02734_Sequence

Listing_ST25 is incorporated herein by reference, in its entirety.