Patent Application
20130059755
Prior to undergoing radical prostatectomy or other aggressive treatments for prostate cancer, it is useful to know with as much accuracy as possible whether the procedure is likely to be curative. Typically, a physician may provide responses to a patient's request for prognostic information by declining to make specific predictions, or providing overall averages, or providing a subjective assessment or assigning the patient to a defined risk group (such as a high or low risk group) based on a model. Providing the most accurate assessment is almost always the proper response. To provide the best possible assessment, since the pathologic stage of cancer correlates with the probability of recurrence after surgery or treatment, many efforts have been made to predict the final pathologic stage or outcome in prostate cancer patients. To this end, many nomograms, algorithms and markers have been employed with the Kattan Nomogram being the most successful. Preferably, the concordance index is used to select a particular strategy from the considered models/nomograms/algorithms and the like. The Kattan nomogram is described in U.S. Pat. No. 5,993,388 and U.S. Pat. No. 6,409,664, both of which are incorporated herein by reference.
An aggressive therapy for the treatment of clinically localized prostate cancer is radical prostatectomy. Unfortunately, many men treated with radical prostatectomy later experience progression of their disease. Starting with an increase in serum PSA, which indicates recurrence, the cancer returns in many men months or years following surgery. Early identification, prior to detectable PSA, of men likely to ultimately experience recurrence would be useful in considering additional treatments while preserving quality of life. Further, accurate estimation of a likelihood of recurrence will also be useful in clinical trials to identify candidates for control groups or for an investigational treatment of interest.
Increased accuracy in the classification of newly diagnosed clinically localized prostate cancers is needed if treatment is to be better tailored to this subgroup of patients. As in other cancers, a number of molecular markers and gene signatures of phenotype and prognosis have been developed recently for prostate cancer. These have provided some significant insights into the existence of distinct classes of aggressive prostate cancer and a number of potential candidate gene markers. A clinically viable test that incorporates the expression values of a small number of gene markers is useful either standing alone or in conjunction with other tools such as the Kattan post-operative nomogram to assess risk of PSA recurrence is desirable.
In one aspect, a gene expression based assay is used to provide a prognosis of the recurrence of prostate cancer. The gene expression based assay is preferably based on a small number of genes. A three gene profile is preferred. In a preferred embodiment, the expression level of the three genes MYH11, SSBP1, and DPT provides prognostic information about the likelihood of prostate cancer recurrence. In alternative embodiments, the expression level of genes such as Filamin C, gamma, RAS like family 12 and Filamin A may be used. Other possible marker combinations include (i) Growth Arrest Specific 1, Smoothelin, Leiomodin 1 and Histone 1 H3d, and (ii) Sorbin and SH3 domain containing 1, PDZ and LIM domain 7, LIM and senescent cell-antigen like domains 2. Additional useful combinations of expression level based markers may be readily selected from Table 2.
In another aspect, a gene expression level based assay is used in conjunction with a clinical tool such as a nomogram that is based on multiple clinical indicators of prostate cancer prognosis.
In a preferred embodiment, a measurement of the expression of three genes and three control genes is determined. Preferably, the gene measurements are measured relative to the average of three control genes. These gene measurements along with the predictive probability from the Kattan nomogram are incorporated into a statistical model to generate a score reflecting the probability of recurrence or the risk of recurrence. This information can also be used to determine the likelihood of PSA levels rising following radical prostatectomy. This probability of recurrence can help patients make personalized decisions about their choice in therapy.
More specifically disclosed is a classifier for distinguishing between subjects in a high recurrence risk category of prostate cancer and subjects in a low recurrence risk category of prostate cancer. The preferred classifier comprises a classification procedure including measuring an expression level of at least two diagnostic markers selected from TABLE 2 with each expression level measured relative to an average of the expression levels of two or more control markers. The expression level of the at least two diagnostic markers is then combined with a predictive probability based on a Kattan nomogram to generate the classifier, which may be also be in the form of a nomogram. Herein the Kattan nomogram is a composite measure, described, for instance in U.S. Pat. Nos. 6,409,664 and 5,993,388.
The Kattan nomogram is typically based on two or more members of a group consisting of PSA value prior to surgery (and prior to hormone therapy, if received), PSA at the time of prostatectomy, primary Gleason at surgery, secondary Gleason at surgery, prostatectomy Gleason Sum, the year of prostatectomy, the months disease free, whether or not surgical margins were positive, whether or not cancer was found in seminal vesicles, whether or not there was extra-capsular extension, whether or not cancer was found in lymph nodes (if any were removed), pre-radiotherapy PSA, radiation dose (if applicable), whether surgical margins were positive or negative (if applicable), whether there was seminal vesicle involvement (if applicable), whether there was lymph node involvement (if applicable), whether there was extra capsular involvement (if applicable), whether or not neo-adjuvant hormones were prescribed, whether or not neo-adjuvant radiation was prescribed.
The Kattan nomogram may also be a composite measure based on two or more members of a group consisting of pretreatment PSA level, combined Gleason grade, specimen Gleason sum, clinical stage, surgical margin status, prostatic capsular invasion maximum cancer length in a core, total length of cancer in the biopsy cores, percent of cores positive level, extraprostatic extension, level of extraprostatic extension, apoptotic index, percent of cancer in one or more cores, percent of high grade cancer in one or more cores, total tumor volume, zone of location of the cancer, presence of seminal vesicle invasion, type of seminal vesicle invasion, p53, Ki-67, p27, DNA ploidy status, lymph node status, and lymphovascular invasion.
A preferred classifier for distinguishing between subjects in a high recurrence risk category of prostate cancer and subjects in a low recurrence risk category of prostate cancer requires measuring the expression levels of diagnostic markers MYH11, SSBP1 and DPT.
Preferred control markers are TUBA, ALAS1 and ACTG1. Alternative preferred control markers are selected for relatively steady expression levels as detected by RT-PCR.
Another preferred classifier for distinguishing between subjects in a high recurrence risk category of prostate cancer and subjects in a low recurrence risk category of prostate cancer requires measuring the expression levels of at least one diagnostic marker selected from TABLE 2 in addition to MYH11, SSBP1 and DPT.
In another preferred embodiment two or more diagnostic markers are selected from the set of markers presented in TABLE 2. More specific sets of diagnostic markers include
Also disclosed is a method of predicting the recurrence of prostate cancer. The method includes the steps of (i) determining an expression level of DPT relative to a standard; (ii) determining expression level of at least one additional markers from TABLE 2; and (iii) transforming the expression level of DPT and the expression level of at least one additional markers from TABLE 2 into a score corresponding to a probability of recurrence of prostate cancer. The method may alternatively, to the step of transforming the expression level of DPT and the expression level of at least one additional markers from TABLE 2, have a step of combining the expression level of DPT and the expression level of at least one additional markers from TABLE 2 with at least one additional indicator of prostate cancer recurrence to determine a composite score. Such a composite score reflects a likelihood of recurrence of prostate cancer.
Preferably, the expression level of DPT and the expression level of at least one additional markers from TABLE 2 is determined prior to at least one treatment selected from the group consisting of prostectomy, hormone therapy, single agent chemotherapy, two agent chemotherapy and treatment with a farnesyl transferase inhibitor.
This disclosure also covers, without limitation, a kit for detecting an expression level of at least DPT, the kit comprising reagents and a first probe set comprising a first probe specifically recognizing DPT and a second probe specifically recognizing a second Marker selected from TABLE 2. The kit may include a device for converting the expression level of DPT and the expression level of at least one additional markers from TABLE 2 into an indicator of a likelihood of recurrence of prostate cancer. The device may be in the form of a nomogram. More specifically, a preferred kit is implemented as one or more members of the group consisting of a mechanical device, a graphical representation, and software instructions to implement a user interface for providing a representation of the indicator of the likelihood of recurrence of prostate cancer. In yet a further aspect of the invention a kit is provided containing reagents for conducting a measurement of three genes and three control genes from a prostate tumor. Instructions (optionally as computer code) are provided to enable the gene measurements to be normalized to the average of three control genes. These measures along with the predictive probability from a nomogram are incorporated into a statistical model that generates a probability of recurrence.
Nomograms are widely used to predict prostate cancer recurrence. The most widely used nomogram is the Kattan nomogram described in U.S. Pat. Nos. 6,409,664 and 5,993,388 each of which is incorporated in its entirety in this specification. For those who have not received surgical treatment for their prostate cancer, these nomograms incorporate the following information: most recent PSA (prostate specific antigen) value, primary and secondary Gleason grade, physician's assessment of clinical stage (using the 1992 or 1997 UICC system), radiation therapy dose that is recommended (if applicable), the number of positive cores found during biopsy, the number of negative cores found during biopsy, whether or not neo-adjuvant hormones had been prescribed, and whether or not neo-adjuvant radiation had been prescribed. For those for whom surgery was performed, the factors include: PSA value prior to surgery (and prior to hormone therapy, if received), PSA at the time of prostatectomy, primary Gleason at surgery, secondary Gleason at surgery, prostatectomy Gleason Sum, the year of prostatectomy, the months disease free, whether or not surgical margins were positive, whether or not cancer was found in seminal vesicles, whether or not there was extra-capsular extension, whether or not cancer was found in lymph nodes (if any were removed), pre-radiotherapy PSA, radiation dose (if applicable), whether surgical margins were positive or negative (if applicable), whether there was seminal vesicle involvement (if applicable), whether there was lymph node involvement (if applicable), whether there was extra capsular involvement (if applicable), whether or not neo-adjuvant hormones were prescribed, whether or not neo-adjuvant radiation was prescribed. This information is combined in a spreadsheet, for example, and a simple statistical treatment provides an analysis used to determine prognosis.
The methods and kits of this invention are most preferably used in conjunction with these nomograms and the information they provide. The method and kits of the invention also involve the detection of a group of genes. Preferably, the expression level one of the genes in this group is measured relative to at least one of the control markers. More preferably the expression levels of three genes and three controls are measured. The most preferred genes are MYH11, SSBP1, and DPT. The preferred controls are ALAS1, TUBA, and ACTG1. Higher expression of the genes MYH11 or DPT indicates a stronger likelihood of remaining free of prostate cancer recurrence than if such over-expression is not seen as described in the examples. A higher expression level of SSBP1 indicates a stronger likelihood of recurrence. The nucleic acid sequences that correspond to the genes whose expression level is measured as well as the sequences used to measure such expression levels are referred to in this specification as Markers. Other sequences of interest include genes useful as assay controls such as ALAS1, TUBA, and ACTG1. Markers are detected with any of the methods used to detect gene expression; preferably these are amplification based methods such as PCR, its variants, and alternative methods. Most preferably, RTPCR is used.
Nucleic acid probes or reporters specific for certain Markers are preferably used to detect the expression of the Marker gene in tumor tissue. Other biological fluids or tissues can be used including prostate tissue, urine, urethral washings, blood and blood components such as serum, ejaculate, and other samples from which prostate proteins could be expected. Any specimen containing a detectable amount of the relevant polynucleotide can be used.
One disclosed method includes contacting a target cell containing a Marker with a reagent that binds to the nucleic acid. The target cell component is a nucleic acid such as RNA. The reagents preferably include probes and primers such to amplify and detect the target sequence. For example, the reagents can include priming sequences combined with or bonded to their own reporter segments such as those referred to as Scorpion reagents or Scorpion reporters and described in U.S. Pat. Nos. 6,326,145 and 6,270,967 to Whitcombe et. al. (incorporated herein by reference in their entirety). Though they are not the same, the terms “primers” and “priming sequences” may be used in this specification to refer to molecules or portions of molecules that prime the amplification of nucleic acid sequences.
Preferred primers are capable of initiating synthesis of a primer extension product, which is substantially complementary to a polymorphic locus strand. The primers and/or probes may be prepared using any suitable method including automated methods. Environmental conditions conducive to synthesis include the presence of nucleoside triphosphates and an agent for polymerization, such as DNA polymerase, and a suitable temperature and pH. The priming segment of the primer or priming sequence is preferably single stranded for maximum efficiency in amplification, but may be double stranded. If double stranded, the primer is first treated to separate its strands before being used to prepare extension products. The primer must be sufficiently long to prime the synthesis of extension products in the presence of the inducing agent for polymerization.
Preferred primers are most preferably eight or more deoxyribonucleotides or ribonucleotides. The exact length of primer will depend on factors such as temperature, buffer, and nucleotide composition. The oligonucleotide primers most preferably contain about 12-20 nucleotides although they may contain more or fewer nucleotides.
When complementary strands of nucleic acid or acids are separated, regardless of whether the nucleic acid was originally double or single stranded, the separated strands are ready to be used as a template for the synthesis of additional nucleic acid strands. This synthesis is performed under conditions allowing hybridization of primers to templates to occur. Generally synthesis occurs in a buffered aqueous solution, preferably at a pH of 7-9, most preferably about 8. A molar excess (for genomic nucleic acid, usually about 108:1, primer: template) of the two oligonucleotide primers is preferably added to the buffer containing the separated template strands. The amount of complementary strand may not be known if the process of the invention is used for diagnostic applications, so the amount of primer relative to the amount of complementary strand cannot always be determined with certainty. As a practical matter, however, the amount of primer added will generally be in molar excess over the amount of complementary strand (template) when the sequence to be amplified is contained in a mixture of complicated long-chain nucleic acid strands. A large molar excess is preferred to improve the efficiency of the process.
The agent for polymerization may be any compound or system that will function to accomplish the synthesis of primer extension products, preferably enzymes. Suitable enzymes for this purpose include, for example, E. coli DNA polymerase 1, Klenow fragment of E. coli DNA polymerase I, T4 DNA polymerase, other available DNA polymerases, polymerase mutants, reverse transcriptase, and other enzymes, including heat-stable enzymes (e.g., those enzymes which perform primer extension after being subjected to temperatures sufficiently elevated to cause denaturating). A preferred agent is Taq polymerase. Suitable enzymes will facilitate combination of the nucleotides in the proper manner to form the primer extension products complementary to each locus nucleic acid strand. Generally, the synthesis will be initiated at the 3′ end of each primer and proceed in the 5′ direction along the template strand, until synthesis terminates, producing molecules of different lengths. There may be agents for polymerization, however, which initiate synthesis at the 5′ end and proceed in the other direction, using the same process as described above.
In another aspect of the invention an expression ratio is used. Establishing a ratio between the amount of amplified Marker attained and the amount of amplified reference Marker or control Marker region amplified can do this. This can be done using quantitative real-time PCR. Ratios can be plugged into a statistical model to determine a likelihood of prostate cancer recurrence.
The kits of the invention can be configured with a variety of components, preferably such that they all contain at least one primer or probe or a detection molecule (e.g., Scorpion reporter). In one embodiment, the kit includes reagents for amplifying and detecting Marker segments. Optionally, the kit includes sample preparation reagents and/or articles (e.g., tubes) to extract nucleic acids from samples.
In a preferred kit, reagents necessary for RTPCR are included such as, a corresponding PCR primer set, a thermostable polymerase, such as Taq polymerase, and a suitable detection reagent(s) such as hydrolysis probe or molecular beacon. In optionally preferred kits, detection reagents are Scorpion reporters or reagents. A single dye primer or a fluorescent dye specific to double-stranded DNA such as ethidium bromide can also be used. Additional materials in the kit may include: suitable reaction tubes or vials, a barrier composition, typically a wax bead, optionally including magnesium; necessary buffers and reagents such as dNTPs; control nucleic acid (s) and/or any additional buffers, compounds, co-factors, ionic constituents, proteins and enzymes, polymers, and the like. Optionally, the kits include nucleic acid extraction reagents and materials.
In preferred kit of the invention, instructions to conduct the assay on patients with prostate samples are provided. It is most preferred that an article encoded with computer instructions for preparing a prediction from a Cox Proportional Hazard analysis or other statistical comparator is provided. The instructions are loaded into a computer such as a general purpose computer such that when the values of the gene analysis and, optionally, the Kattan parameters are input into the program, the computer provides as output the likelihood of recurrence (high versus low or a numerical indicator).
In a preferred kit of the invention, mechanical implementations of a nomogram may be also provided. Such implementations may use paper based components or moving mechanical parts to allow consideration of the individual variables used by the nomogram.
Purpose: Accurate estimates of the risk of recurrence are needed for the optimal management of patients with clinically localized prostate cancer. A nomogram and novel molecular predictors were combined into a new prognostic model of prostate specific antigen (PSA) recurrence.
This study was designed to identify genes that correlate with PSA recurrence in patients with clinically localized prostate cancer with the goal of developing an accurate predictive classifier that can be readily applied in current routine clinical practice for management of patients with organ-confined disease. Here we report the development of a clinically viable test that incorporates the expression values of three novel gene markers, measurable by RTPCR, and the Kattan post-operative nomogram (11), a widely used tool in the clinical management of prostate cancer, to assess risk of PSA recurrence. Finally, we show that this new classifier provides improved accuracy compared with the Kattan nomogram for predicting biochemical recurrence in this lower risk patient population.
Materials and Methods: Gene expression profiles from formalin-fixed, paraffin-embedded (FFPE) localized prostate cancer tissues were analyzed to identify genes associated with PSA recurrence. The profiles of the identified markers were reproduced by reverse-transcriptase-polymerase-chain-reaction (RTPCR). The RTPCR profiles from three of these genes, along with the output from the Kattan post-operative nomogram, were used to produce a predictive model of PSA recurrence.
Results: After variable selection, a model of PSA recurrence was built that combined expression values of three genes and the post-operative nomogram. The 3-gene plus nomogram model predicted 5-year PSA recurrence with a concordance index (c-index) of 0.77 in a validation set compared to a c-index of 0.67 for the nomogram. This model identified a subgroup of patients that were at high risk for recurrence which were not identified by the nomogram.
Conclusions: This new gene-based classifier has superior predictive power when compared against the 5-year nomogram to assess risk of PSA recurrence in patients with organ-confined prostate cancer. This classifier should provide a more accurate stratification of patients into high and low risk groups for treatment decisions and adjuvant clinical trials.
Materials and Methods
Patients and Tumor Samples
Patient information was obtained from the St. Vincent's Campus Prostate Cancer Group (SVCPCG) database (Human Research Ethics Committee Approval H00/088). From January 1990 to December 2001, 960 patients were treated for prostate cancer with radical prostatectomy (RP) with no preoperative therapy at St Vincent's Hospital, Sydney. The subgroup of 316 consecutive patients with clinically localized disease assessed in the current study are those patients of the 960 for which the pathological stage ranged from pT2A to pT3A; the minimum follow-up for censored patients was five years; RP was the primary treatment; and tissue blocks could be accessed from the RP specimens for use in gene expression profiling experiments. The date of PSA recurrence was defined as the date of the first increase in serum PSA≧0.2 ng/mL after RP. These patients were randomly split into training and test sets. The test set was used solely for validation purposes. Differences in the distribution of the clinical variables between the training and test sets were evaluated by either a t-test, log-rank, or Chi-square test depending on whether the variable was continuous, time-to-event, or categorical. All statistical tests were two-sided and significance was defined as p<0.05.
Gene Expression Profiling
Six μm sections from each of the FFPE tissue blocks were submitted to pathology review (JGK) and macrodissected to ensure >30% malignant epithelium was used for total RNA extraction using the High Pure RNA Paraffin kit (Roche Diagnostics, Indianapolis, Ind.).
Gene expression profiling experiments were performed on all total RNA specimens in the training series with a 1200 gene custom designed DNA-mediated annealing, selection, ligation and extension microarray (DASL) (Illumina, San Diego, Calif.)(22). Both the gene expression profiles and the design of the DASL array can be accessed in the GEO database (accession pending). Three control genes, ALAS1, TUBA and ACTG1, were selected for the microarray and RTPCR analysis based on earlier studies in prostate cancer (12, 23).
Significance Analysis of Microarrays (SAM) analysis using the survival mode was used to measure prognostic significance for each probe on the array (24-26). The probes were ranked by the absolute value of the test statistic. False discovery rates were calculated by data permutation.
RTPCR assays were designed for the top-ranked prognostic marker candidates (n=30), including all genes under the lowest reported false discovery rates for both increasing and decreasing probes. Genes with a poor Pearson correlation (<0.4) between the array and RTPCR data among the training samples were excluded from further analysis.
Construction of the Prognostic Model
Using the Memorial Sloan Kettering Cancer Centre on-line calculator (http://www.mskcc.org/mskcc/html/10088.cfm), the five-year nomogram recurrence score was calculated for each individual patient using the post-operative historical model. To select variables for a multivariate model, the delta CT values of the candidate genes from RTPCR as well as the predicted probabilities from the nomogram were processed by the L1 regularization path algorithm using the training samples (25). By cross-validating the training series using the path algorithm to set different limits on the potential for over-fitting the Cox model, the signature with the least error was selected. The final predictive model for deployment was built by fitting these selected variables to the training set using a Cox proportional-hazards model.
A cut-off for high and low risk stratification of the model, and a cut-off for the nomogram were both chosen under the assumption that the costs for false positives and false negatives are equivalent. Under this assumption, a cut-off giving the highest training set accuracy (defined by the total number of correctly classified patients) was chosen.
Validation of the Prognostic Model
To evaluate the accuracy of predictive prognostic models with respect to the actual freedom from recurrence in the test set, a calibration curve was generated from the predicted 5-year recurrence-free probability estimated by Cox proportional-hazards regression and the Kaplan-Meier estimates of the actual recurrence-free probability at 5-years (24, 27). The performance of the final prognostic model in the test set was assessed by Kaplan-Meier curves and hazard ratios by stratifying the test set patients into a low risk and high risk group based on the pre-selected cut-off from the training set. All statistical analyses were performed in R, version 2.5.0 (www.r-project.org).
Results
Patient Characteristics
Total RNA was isolated from 316 prostatectomy FFPE tissues and 20 samples were excluded due to RNA degradation. The clinical and pathological characteristics of patients included in the training and test sets are summarized in Table 1. Median follow-up was 72 months, and median time from radical prostatectomy to biochemical recurrence was 34 months among those who recurred. Ninety-eight of 296 patients developed recurrence, including 74 patients who developed recurrence within 5 years of surgery. The training series consisted of 138 patients with the remaining 158 set aside for the test set. There was no statistically significant difference in the clinicopathological characteristics between these two sets of patients (Table 1).
Gene Expression and Univariate Analysis
The RNA samples from the training set of 138 patients were analysed by DASL array. The permutation of the SAM algorithm revealed false discovery rates of 0% for 20 genes on the DASL array. The top 30 genes as ranked by the score value from the SAM analysis had a false discovery rate of 6.8%. These 30 genes were then assessed by RTPCR analysis using the same training set. Six of the 30 selected genes displayed a correlation between DASL and RTPCR of less than 0.4 and thus were removed from further analysis. The effect of each gene on recurrence free probability was measured by Cox regression. The hazard ratio quantified the relative risk of PSA recurrence for each increase of 1 normalized CT. The hazard ratio and P value are recorded for both the training and test sets (Table 2). Twenty-three of 24 markers (except for marker HIST1H3D) continued to have a significant association to recurrence in the test set. In the same analysis, a 5-year postoperative nomogram was also a significant predictor of PSA recurrence in both the training and test sets (P value of 0.001 and 0.005, respectively).
Further variable selection was performed on the RTPCR training set of Example 1 to build a multivariate prognostic classifier. Four variables were selected by the L1 Regularization algorithm: 3 genes (DPT, SSBP1 and MYH11) and the 5-year nomogram. These 4 variables were then modeled on the training set using Cox regression analysis.
Classifier Validation and Survival Analysis
When testing the prognostic model on an independent test series of 157 patients, the c-index for the classifier was apparently higher than the c-index for the nomogram (0.77 vs. 0.67) (
The cut-off from the training set was used to place test set patients into either a high- or low-risk group. Kaplan-Meier curves were generated for the test set samples by stratifying the patients into a low risk and high risk group based on a cutoff from the training set that produced the highest accuracy on the training samples. A calibration curve was generated from the predicted 5-year disease-free survival of the test set and the Kaplan-Meier estimates at 5years based on cuts in the predicted probability at 0.3,0.5,0.7 and 0.9. The Kaplan-Meier analysis for PSA recurrence-free probability showed a highly significant difference in time to PSA recurrence for the predicted low- and high-risk groups (HR 6.85, 95% CI=3.77 to 12.43, P<0.001,
The clinical and pathological characteristics of patients included in the training and test sets are summarized in Table 1. Median follow-up was 72 months. Mean and median times from radical prostatectomy to biochemical recurrence were 40 and 34 months, respectively. Ninety-eight of 296 (33%) patients developed recurrence, including 74 (25%) patients who developed recurrence within 5 years of surgery.
Application of an Improved Prognostic Model
In order to evaluate the potential impact of the model on patient management, we compared the accuracy of prognostic stratification using the classifier compared to the 5-year post-operative nomogram on the test cohort. We used the cut-off based on each model's highest accuracy and then applied these to the test set (
RTPCR assays were designed for the candidate markers and the three (3) control genes. Multiple PCR primers and probes were designed for the markers. CT values were normalized to the average of 3 control genes.
Each normalized gene was plotted against the DASL signal values. Total RNA was reverse 15 transcribed, and pre-amplified with the gene-specific primers. Pre-amplified cDNA was then quantified using ABI PRISM® 7900 sequence detection system (APPLIED BIOSYSTEMS).
Also identified were additional panels, which while suboptimal compared to the preferred panel, are also useful in their own right. Each of the alternative panels can be used as a stand alone panel or in combination with the Kattan nomogram, the performance of which is improved by each of the alternative panels. The arrow symbol indicates the c-index corresponding to the model.
Also identified were probes and primer pairs for the remaining markers. They are presented in Table 3.
An alternative set of primers and probes were also developed for MYH11, DPT and SSBP1 as well as ALAS, ACTG and TUBA. The results were at least consistent or superior compared to the primer and probe sets of Example 3. They alternative primer and probes are:
Discussion:
The relatively low level of complexity of the disclosed classifiers is also important with the ability to measure expression of a small set of genes in FFPE tissue with a diagnostic-approved platform. These significant advances address some key factors affecting the likelihood of successfully implementing this predictive tool in a clinical diagnostic setting. The need for very small concentrations of RNA derived from FFPE tissue will also facilitate its potential long-term applicability to routine pathology specimens including preoperative transrectal biopsies.
This systematic assessment of prostate cancer-related gene expression correlates of PSA recurrence in order to develop a gene-based classifier of recurrence in clinically localized prostate cancer of potential broad clinical utility. The novel gene predictors were identified using a custom DASL array, a microarray platform that allows high-throughput gene expression profiling of RNA derived from FFPE tissues (22). A key component of this study was the design of the custom DASL microarray gene set that is based on gene markers that were identified by re-analysis of published datasets (12, 13), in-house gene expression data (unpublished), and markers previously implicated in prostate cancer progression (14,16,17). This affords a degree of independent validation for those 30 genes that were most significant in this study. The 24-gene markers that correlated with PSA recurrence in the gene expression array analysis were further validated by RTPCR to produce a preferred 3-gene signature (DPT, MYH11 and SSBP1) that when combined with an established nomogram, resulted in a new classifier of PSA recurrence. This classifier was subsequently validated in an independent group of patients. DPT and MYH11 are novel prostate cancer prognostic markers while SSBP1 has previously been associated with aggressive prostate cancer (17).
An assessment of the value of this new classifier over a widely used nomogram for prostate cancer recurrence showed that the new classifier identified patients with both low- and high-risk of recurrence with much greater accuracy than the postoperative nomogram alone (9). The classifier presented here was also able to stratify patients within clinically relevant subgroups based on conventional clinicopathological parameters into high- and low risk-recurrence groups. Of note, the ability to stratify patients with Gleason 6 and 7 cancers represents a significant advance in predictive accuracy over current approaches. The use of PSA recurrence as a significant endpoint for prostate cancer has been disputed since only a proportion of patients who experience recurrence progress to clinically significant disease. These relationships will be more clearly defined as this cohort matures with data on metastases and death from PrCa. However, it is clear that the detection of a rising PSA post-prostatectomy is an important decision point when most physicians and patients consider further treatment options (5). In this context, the use of the 5-year nomogram to evaluate the potential impact of this classifier is valid with the majority of biochemical recurrences post-prostatectomy occurring within 5 years.
Of significance is the impact of this new classifier as a decision tool when considered against other published signatures and gene markers of molecular phenotype and prognosis in prostate cancer (12-20). This study is unique in being developed specifically to aid prediction of risk of recurrence in prostate cancer patients with clinically localized disease, since these patients represent >80% of newly diagnosed cases of prostate cancer in the United States. While published signatures and gene markers have been identified from cohorts of patients representing the spectrum of pathological stages, both the training and test cohorts in this study were restricted to organ-confined prostate cancer. The importance of having concordance between the patient group used in the development of a predictive tool, with the anticipated target group is reinforced by the relatively low performance of the 5-year nomogram in this cohort of clinically localized patients compared with previous reports. On further analysis, this is likely due to the limitation of assessing only organ-confined cases in this study. When tested on our consecutive prostate cancer patient cohort from the same institution that was not limited to organ-confined disease, the nomogram performance was consistent with previous studies (9, 11).
The relatively low level of complexity of the classifier is also important. With the ability to measure expression of a small set of genes in FFPE tissue and the use of a platform that is approved for diagnostic testing in archival specimens, it is a significant advance as it addresses some key factors affecting the likelihood of successfully implementing this predictive tool to a clinical diagnostic setting. The requirement for low concentrations of RNA derived from FFPE tissue will also facilitate its potential long-term applicability to routine pathology specimens including preoperative transrectal biopsies. Further validation in external cohorts of both surgical and preoperative biopsies, including replicating the gene selection in biopsies, is now required to confirm the wider applicability of this classifier in the preoperative setting.
The implementation of an accurate predictive classifier for localized prostate cancer has important implications for patient management of early prostate cancer. Patients with localized disease and high-risk features are likely to benefit from adjuvant therapies including hormone, radiation and systemic treatments and the benefits should be evaluated in treatment trials (7, 28, 29). Early phase clinical trials employing such agents are underway in the hormone-refractory setting but may ultimately be tested in localized prostate cancer as adjuvant therapies (30). Intrinsic to these studies is the accurate identification of high risk patients to ensure homogeneous patient groups8. Conversely, the improved identification of patients of low risk of recurrence will reduce the number of patients who are exposed to the morbidity of therapy as a result of the identification of increasing numbers of indolent cancers through PSA screening.
Finally, the development of an improved prognostic model for localized prostate cancer has the potential to facilitate better treatment decisions, either to forego treatment of indolent disease or offer adjuvant chemotherapy for men with high risk of recurrence.
All of the references in this disclosure are hereby incorporated by reference herein for all of their relevant teachings.
| Number | Date | Country | Kind |
|---|---|---|---|
| PCT/US2011/030686 | Mar 2011 | US | national |
