Tristetraprolin Is a Prognostic Biomarker for Poor Outcomes among Patients with Low-Grade Prostate Cancer

While proving beneficial to many men with prostate cancer, testing for serum PSA levels has resulted in an epidemic of prostate cancer overtreatment, causing hundreds of thousands of men with a low risk of disease-specific death to have unnecessary harmful side effects (1–4). Indeed, potentially harmful side effects from overtreatment were an important factor in the United States Preventive Services Task Force's (USPSTF) 2012 recommendation that healthy men should not be screened with PSA testing (5). However, a 2017 update issued by the USPSTF states that clinicians should discuss the potential benefits and harms of PSA testing with men between 55 and 69 years old (6). As an alternative to aggressively treating patients with low-risk prostate cancer, there is a growing emphasis toward utilizing active surveillance. Although this is an appropriate alternative for many low-risk patients, approximately 30% of patients in long-term active surveillance studies were ultimately reclassified as harboring aggressive disease and required therapeutic intervention (7–9). Thus, there is a dire unmet need for improved prognostic prostate cancer biomarkers.

Tristetraprolin (TTP, ZFP36), an RNA-binding protein involved in controlling mRNA stability, impairs tumor growth and development in a variety of malignancies, including prostate cancer (10–19). In a previous study that established TTP's functions as a tumor suppressor in prostate cancer, analysis of TTP RNA expression in a single prostate cancer patient dataset suggested that low TTP-expressing primary tumors may have increased rates of biochemical recurrence (BCR) compared with high TTP-expressing prostate cancer (19). To more fully assess whether TTP is a prognostic biomarker that identifies prostate cancer patients with an increased risk of poor clinical outcomes, we evaluated two prostate cancer patient cohorts from Moffitt Cancer Center (MCC) for TTP RNA or TTP protein expression, as well as six independent prostate cancer genomic datasets having postsurgical clinical follow-up from multiple medical institutions. Our findings establish that TTP alone is a clinically relevant prostate cancer biomarker that provides an improvement over clinical and pathologic variables to predict which patients may harbor aggressive disease.

This study analyzed TTP RNA or protein expression and associated clinical endpoint data (BCR and/or metastatic development) from eight studies (four cohort, two case–cohort, and two case–control) of radical prostatectomy (RP) patients with or without postoperative treatment at multiple U.S. medical centers: Memorial Sloan Kettering Cancer Center (Taylor and colleagues, n = 131; ref. 20), MCC [Das and colleagues, n = 306, and Mahajan and colleagues, tissue microarray (TMA), n = 153; refs. 21–23], Johns Hopkins Medical Institute (Ross and colleagues, n = 260; ref. 24), Mayo Clinic (Karnes and colleagues, n = 235, and Erho and colleagues, n = 545; refs. 25, 26), Cleveland Clinic (Klein and colleagues, n = 182; ref. 27), and The Cancer Genome Atlas prostate adenocarcinoma (TCGA PRAD, n = 333; multiple U.S. medical centers; ref. 28). (An overview of the prostate cancer patient study cohorts is available as Supplementary Table S1.) For the MCC patient cohorts, Das and colleagues, and Mahajan and colleagues, BCR was determined by clinical documentation of a single postoperative PSA ≥ 0.2 ng/mL or two consecutive postoperative PSA values of 0.2 ng/mL.

From the Das and colleagues' cohort, 306 patients were selected for analyses if they had postsurgical clinical follow-up data (Table 1) and quantitative real-time PCR (qRT-PCR) successfully measured TTP and β-actin RNA expression (21, 22). The Das and colleagues' cohort also includes 115 patient-matched normal adjacent tissue samples. From the Mahajan and colleagues' TMA cohort, 153 primary prostate cancer patients whose primary treatment was RP and had postsurgical clinical follow-up data were selected for analyses (Table 1; ref. 23). From the TCGA PRAD cohort, 280 patients had postsurgical clinical follow-up data and were selected for analyses (28). All primary prostate cancer patients in the Taylor and colleagues', Ross and colleagues', Karnes and colleagues', Erho and colleagues', and Klein and colleagues' cohorts were included in these studies. Cancer of the Prostate Risk Assessment postsurgical (CAPRA-S) scores were calculated for the cohort and case–cohort studies, Das and colleagues, Mahajan and colleagues, Taylor and colleagues, Ross and colleagues, and Karnes and colleagues, but not for TCGA PRAD data, which lacks necessary histopathologic data for CAPRA-S. There were no other inclusion or exclusion criteria.

For the Das and colleagues' cohort, total RNA isolation from formalin-fixed, paraffin-embedded (FFPE) prostate tissues was previously described (21, 22). cDNA was synthesized, and qRT-PCR was performed as previously described (29). Data analyses used the ΔΔCt method, where levels of B-actin mRNA served as the internal control, and then the calculated results were converted to log₂ expression. Oligonucleotides for qRT-PCR were as follows: β-actin (β-actin FOR3 5′-AAGGTGACAGCAGTCGGTTG-3′ and β-actin REV3 5′-CGGCCACATTGTGAACTTTG-3′; ref. 30) and TTP (TTP FOR100 5′-CCACTCCTATCAGCGTCT-3′ and TTP REV100 5′-CGCTGCTGGCATATTCAT-3′).

For samples retrieved from the Decipher GRID database (Ross and colleagues, Karnes and colleagues, Erho and colleagues, and Klein and colleagues), specimen selection, RNA extraction, and microarray hybridization were performed in a Clinical Laboratory Improvement Amendments–certified laboratory (GenomeDx Biosciences) and were previously described (24–26). Quality control was performed using Affymetrix Power Tools. Normalization was done using the single-channel array normalization algorithm, and expression was summarized using the median expression of probesets that map to exonic regions of the gene.

Immunohistochemistry (IHC) was performed on a Ventana Medical Systems (VMS) Discovery XT using VMS's UltraMap-AP Kit (VMS #760-153). Antigen retrieval consisted of a timed application of Standard CC1 (VMS # 950-500) at pH of 8.0 and Ventana Antibody Block (VMS #760-4204). LifeSpan Biosciences TTP antibody (LS-B5606) was diluted 1:200, or Ventana Rabbit IgG antibody was diluted 1:1 for a final concentration of 5 μg/mL in Dako antibody diluent (S0809). Primary antibodies were incubated for 32 minutes at room temperature. This was followed by hematoxylin (VMS #760-2021) and Bluing Reagent (VMS #760-2037), 4 minutes for each cellular counterstain. TTP protein levels in the Mahajan and colleagues' TMA samples were semiquantitatively measured by a Modified H Scoring method in a blinded fashion by a research pathologist in the MCC Tissue Core. For this method, staining intensity (0 = negative, 1 = weak, 2 = moderate, and 3 = strong) and percentage of positive cells (0 = 0%, 1 = 1%–33%, 2 = 34%–66%, 3 ≥ 67%) were determined. These two values were multiplied to generate a composite TTP IHC score (range, 0–9), and the scores were stratified as TTP-low = 0 to 3 and TTP-high = 4 to 9. Images were taken using an Aperio AT2 Scanner (Leica Biosystems).

Protein from 293T cells was disrupted in lysis buffer [50 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 2.5 mmol/L EGTA, and 0.1% Tween-20 with 1 mmol/L PMSF, 10 mmol/L β-glycerophosphate, 1 mmol/L NaF, 1 mmol/L NaVO₄, and complete mini tablet protease inhibitor (Roche)] by sonication at 4°C 2 times until all cells were lysed. PANC-1 whole-cell lysate was acquired (Santa Cruz Biotechnology, Inc.; #sc-364380). Protein (30 μg per lane) was separated a 10% SDS-polyacrylamide gel, transferred to a PVDF membrane, and blotted for antibodies specific for TTP (LifeSpan Biosciences; #LS-B5606) and β-actin (Sigma; #AC-15).

Normalized gene expression obtained from the patient datasets was converted to log₂ expression. Differences in TTP RNA expression in normal adjacent versus tumor tissue were tested by the paired t test. Differences in TTP RNA expression in patients that progressed to BCR versus those that remained indolent and patients that developed metastatic disease versus those that remained indolent were tested by the Mann–Whitney U. Both tests were performed using MATLAB version 9.3 (The MathWorks, Inc.). Statistical analyses of time to BCR and time to metastasis were performed using a log-rank test, calculated using R software version 3.4.0. (http://www.R-project.org). In case–cohort studies, survival analysis was weighted using the Lin–Ying method for case–cohort design to estimate cohort parameters (31–33).

HRs and 95% confidence intervals for univariable analysis (UVA) and multivariable analysis (MVA) were estimated using Cox proportional hazard models in R software version 3.4.0. To account for cohort heterogeneity in pooled HR estimates, stratified Cox models were fit.

All statistical tests were two-sided with P values < 0.05 considered statistically significant.

To assess the relationship between TTP mRNA expression and BCR, the Taylor and colleagues', Das and colleagues', Ross and colleagues', Karnes and colleagues', and Klein and colleagues' datasets were evaluated. In all five datasets, patients that progressed to BCR had tumors with lower TTP RNA expression levels at the time of RP than patients whose tumors remained indolent (Fig. 1A). In addition, analysis of the Das and colleagues' cohort found that TTP mRNA levels were lower in prostate tumor tissues versus patient-matched normal adjacent tissues (Supplementary Fig. S1A).

To test if TTP mRNA expression levels at RP correlate with time to BCR, patients in the Taylor and colleagues' and Das and colleagues' cohorts were separated into two subtypes, TTP-high and TTP-low, based on the median TTP expression level in each dataset. Our previous analysis revealed that TTP-low patients in the Taylor and colleagues' study had a decreased time to BCR compared with men with TTP-high prostate cancer (19). Here, using the Taylor and colleagues' cohort, we found that in pathologically low-grade prostate cancer [defined herein as pathologic Gleason score (pGS) ≤ 7(3+4) tumors], patients with low expression of TTP also had a decreased time to BCR compared with TTP-high patients (Fig. 1B). To validate our results, we repeated these analyses and found that Das and colleagues' patients with low TTP-expressing prostate cancer also had a decreased time to BCR versus TTP-high patients (Fig. 1C). This difference was also present in the subset of low-grade patients within the Das and colleagues' cohort, as men in this subset with TTP-low tumors had a decreased time to BCR (Fig. 1D; Supplementary Fig. S1B). The Taylor and colleagues' and Das and colleagues' datasets have too few pathologically high-grade prostate cancer patients (defined here as pGS ≥ 7(4+3) tumors), 36 and 27, respectively, to properly analyze the relationship between TTP expression and BCR in this population; however, from the high-grade tumors that are present in these cohorts, the time to BCR was similar for both TTP-low and TTP-high patients (Supplementary Fig. S1C and S1D). The TCGA PRAD dataset also showed decreased expression of TTP in patients that developed BCR, and a decreased time to BCR in men with TTP-low tumors compared with patients with TTP-high tumors (Supplementary Fig. S2A and S2B). The clinical follow-up time for the TCGA PRAD study is limited (median, 24 months), and there are very few BCR events in the low-grade subset; therefore, it was not possible to analyze this dataset for BCR in pGS subcategories. The Ross and colleagues' and Karnes and colleagues' studies were designed to have metastasis as the study endpoint, and Klein and colleagues is a case–control study, so the time to BCR also cannot be accurately assessed from these cohorts. Overall, these results provide strong evidence that TTP mRNA levels may be a promising prognostic biomarker of prostate cancer that distinguishes which patients with low-grade disease progress to BCR.

An IHC staining assay was developed to measure TTP protein expression in prostate tissues to test if TTP protein levels could be utilized as a clinical biomarker for identifying patients with an increased chance of BCR (Fig. 2A). To validate that the TTP antibody is specific for TTP, an immunoblot assay was performed on cells known to express TTP, PANC-1, and cells lacking TTP protein, 293T (Supplementary Fig. S3A). Specific antigen binding by the TTP antibody was confirmed by IHC staining of sections from the same FFPE prostate cancer sample using the TTP antibody and a nonspecific rabbit IgG antibody (Supplementary Fig. S3B). Using this protocol, TTP protein levels were measured in 153 primary prostate cancer samples from the Mahajan and colleagues' study (Table 1; ref. 23). Similar to TTP RNA analyses, patients with low TTP protein expression had a decreased time to BCR versus patients with TTP-high tumors (Fig. 2B). Once again, in the low-grade population, tumors with low TTP protein levels had a decreased time to BCR compared with those with high TTP expression (Fig. 2C). In addition, TTP protein levels did not predict for BCR in high-grade patients (Supplementary Fig. S3C).

To determine if TTP RNA can be utilized as a prognostic biomarker for identifying RP patients that will have poor outcomes, Cox proportional hazard models were estimated for the Das and colleagues', Taylor and colleagues', Ross and colleagues', and Karnes and colleagues' datasets. UVAs of TTP expression in these cohorts indicate that TTP levels measured at prostatectomy can identify which prostate cancer patients develop BCR (Fig. 3A). Further, stratified Cox regression analysis of the pooled UVA from all of these datasets further supports the conclusion that TTP RNA is a clinically relevant biomarker of BCR (Fig. 3A). In addition, UVA of TTP protein expression in the Mahajan and colleagues' TMA suggests that TTP protein is also a clinically relevant biomarker for BCR (Fig. 3A).

The CAPRA-S score is a clinical risk calculation commonly used by physicians to predict the development of aggressive prostate cancer following RP based on preoperative PSA, pGS, surgical margins, extracapsular extension, seminal vesicle invasion, and lymph node invasion (34). Assessment of CAPRA-S and TTP RNA expression revealed that TTP levels are reduced in patients with high-risk CAPRA-S scores compared with patients with low-risk CAPRA-S in the Taylor and colleagues' cohort. Importantly, no other differences were observed between the Taylor and colleagues' and Das and colleagues' datasets, showing that there is no interaction between these two predictive indicators (Supplementary Fig. S4). Further, to evaluate if TTP RNA can improve clinical risk assessment for RP patients, HRs were calculated for CAPRA-S–adjusted models including TTP expression using MVAs for the same patient cohorts as the UVA. Similar to UVA, stratified Cox regression analysis of the pooled MVA HRs shows that the addition of TTP RNA expression provides independent prognostic information beyond CAPRA-S, suggesting that TTP RNA may improve the ability to predict the patients who are likely to develop BCR after RP (Fig. 3B). In addition, the MVA of CAPRA-S adjusted to include TTP protein in the Mahajan and colleagues' TMA suggests that TTP protein might improve risk prediction (Fig. 3B), but this analysis did not reach the level of significance, so further evaluation is needed in a larger patient cohort to confirm this result.

In addition to BCR, we assessed whether TTP RNA is a potential biomarker that discriminates which patients might progress to metastatic disease by testing four prostate cancer expression datasets: Ross and colleagues, Erho and colleagues, Karnes and colleagues, and Klein and colleagues. Patients in each of these studies that developed metastatic disease had lower TTP RNA levels at RP than men whose tumors remained indolent (Fig. 4A). Erho and colleagues and Klein and colleagues are case–control studies, so the time to metastatic disease cannot be accurately assessed. However, Ross and colleagues and Karnes and colleagues are case–cohort studies with metastasis as the study design endpoint, so patients in these cohorts were analyzed for time to metastasis. These analyses revealed that men with low TTP expression after RP progressed to metastatic disease more rapidly than patients with TTP-high prostate cancer (Fig. 4B and C). Both UVA and MVA of TTP expression and metastatic progression in prostate tumors in these two cohorts demonstrated that TTP RNA can function as a prognostic biomarker for identifying RP patients who are at an increased risk of metastasis (Table 2), both when taken as an independent biomarker and when added alongside CAPRA-S.

The epidemic in overtreatment of low-risk prostate cancer underscores the need to develop superior biomarkers that identify patients with an increased risk of developing aggressive disease. Here, we validate the tumor suppressor TTP as a biomarker for predicting which prostate cancer patients may have poor outcomes following RP. Specifically, by evaluating TTP expression levels and its association with clinical outcomes in seven independent cohorts of prostate cancer patients (n = 1,939) from multiple medical centers, prostate tumors that developed BCR and/or metastatic disease were found to express lower levels of TTP RNA versus tumors that remained indolent. Further, TTP-low prostate cancer progressed to BCR and metastasis at increased rates compared with TTP-high prostate cancer, especially in men with low-grade tumors. In addition, UVAs establish TTP as a clinically relevant prognostic biomarker for identifying which RP patients may have poor outcomes. Finally, stratified proportional hazard modeling found that TTP RNA expression provides prognostic information distinct from what CAPRA-S already provides. Thus, the addition of TTP RNA expression as a variable to CAPRA-S might improve predictions of which RP patients harbor aggressive disease.

The National Comprehensive Cancer Network guidelines classify Stage T1–T2a prostate tumors with a GS ≤ 6 and PSA ≤ 10 ng/mL as very low-risk or low-risk disease, and men in these risk categories that have a life expectancy ≥ 10 years are candidates for active surveillance (35). However, many men in low-risk categories immediately undergo aggressive treatment despite the common adverse long-term side effects. Therefore, new clinical tests are needed to identify men that have low chances of failure on active surveillance. Our data repeatedly demonstrate that patients with low-grade tumors and low TTP expression have a significantly increased rate of BCR following RP versus men with low-grade tumors and high TTP levels. Thus, TTP may be a particularly beneficial biomarker for prognosticating which clinically low-risk prostate cancer patients will have a more aggressive clinical course.

A major strength of our study design is the validation of TTP's ability to function as a biomarker in multiple genomic platforms, as well as in a multi-institutional cohort of prostate cancer patients across the United States. This is one of the larger studies to date of expression and outcomes in prostate cancer (BCR total n = 1,394 and metastatic outcome total n = 1,222). Thus, this study provides strong evidence for the generalizability of TTP as a biomarker to identify inherently aggressive prostate tumors among patients thought to have clinically low-grade disease. However, there is the caveat that valid inference regarding generalizability in minority populations may be limited at this time, and that further study is needed for African-American men in particular due to few known prognostic biomarkers for this population which is unduly affected by poor prostate cancer outcomes. Further, given that Gleason scores are changed approximately 30% of time from the initial biopsy score to the pathologic score following RP (36), an additional strength of our study is that it utilizes pGS, thus providing stringent analysis of prostate cancer outcomes in patients with true low-grade disease. Finally, we developed an IHC method to stain for TTP in prostate tissues and will perform future studies to determine if TTP protein expression from biopsy samples has the ability to predict prostate cancer outcomes. This could be of high clinical relevance as a tool that enables clinicians to more accurately risk-stratify their patients for appropriate treatment recommendations.

We recognize that this study has limitations. Specifically, our studies only assessed tissue from RP, which is more easily accessible than biopsy material. Accordingly, there is as yet no formal proof that TTP is a useful biomarker that will predict success or failure for active surveillance. However, given our findings, we submit that a long-term prospective study examining TTP expression in biopsies in newly diagnosed prostate cancer patients is warranted.

In conclusion, the tumor suppressor TTP is a clinically relevant prognostic prostate cancer biomarker that could be used in determining which patients will have poor clinical outcomes at the time of RP, and may improve the effectiveness of CAPRA-S. Notably, TTP performs particularly well as a biomarker in low-grade prostate cancer patients where much controversy exists on appropriate management options. The ability to use TTP to more accurately identify clinically low-risk patients who may harbor aggressive disease will enable physicians to more accurately risk-stratify their patients for appropriate therapy recommendations and will prevent overtreatment of potentially indolent disease.

E. Davicioni is president and CSO at, and has an ownership interest in, GenomeDx. A.E. Ross has an ownership interest in GenomeDX Biosciences. E.M. Schaeffer is a consultant/advisory board member for GenomeDX. R.J. Karnes reports receiving commercial research funding from and has an ownership interest in GenomeDX. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R.J. Rounbehler, A.E. Berglund, E.M. Schaeffer, J.L. Cleveland, J.Y. Park, K. Yamoah

Development of methodology: R.J. Rounbehler, E.M. Schaeffer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Awasthi, W. Li, E. Davicioni, A.E. Ross, E.M. Schaeffer, E.A. Klein, R.J. Karnes, R.B. Jenkins, J.Y. Park, K. Yamoah

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.J. Rounbehler, A.E. Berglund, T. Gerke, M.M. Takhar, S. Awasthi, N.G. Erho, R.B. Jenkins, J.Y. Park, K. Yamoah

Writing, review, and/or revision of the manuscript: R.J. Rounbehler, A.E. Berglund, T. Gerke, M.M. Takhar, S. Awasthi, W. Li, E. Davicioni, N.G. Erho, A.E. Ross, E.M. Schaeffer, E.A. Klein, R.J. Karnes, R.B. Jenkins, J.L. Cleveland, J.Y. Park, K. Yamoah

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Gerke, E. Davicioni, R.J. Karnes, R.B. Jenkins, J.L. Cleveland, J.Y. Park

Study supervision: R.J. Rounbehler

This work was supported in part by a Moffitt Cancer Center Team Science grant (to J.Y. Park, A.E. Berglund, and R.J. Rounbehler), NIH/NCI R01 grants CA128813 (to J.Y. Park) and CA167093 (to J.L. Cleveland), Prostate Cancer Foundation Young Investigator Award (to K. Yamoah), American Cancer Society MRSG-17-108-01-TBG (to K. Yamoah), the Cortner Couch Endowed Chair, College of Medicine at the University of South Florida (to J.L. Cleveland), and NCI Comprehensive Cancer Center Grant P30-CA076292 awarded to the H. Lee Moffitt Cancer Center & Research Institute.

The authors thank Dr. Nupam Mahajan for generously providing the prostate cancer TMA utilized in these studies. We also thank Dr. Alex Lopez and Jeanette Rheinhardt of the Moffitt Cancer Center Tissue Core for their expert technical assistance. The authors thank Hyun Y. Park for the technical assistance that she provided for this study. Dr. John L. Cleveland also acknowledges the generous support of Lesa France Kennedy.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.