Prognostic biomarkers

Kidney cancer. Principles and practice. Second edition. Primo N. Lara, Jr. Eric Jonasch (Editors). Springer International Publishing (2015)

Prognostic biomarkers have been studied in parallel with advances in the tumorigenesis of this cancer. A summary of the potential molecular prognostic biomarkers that have been investigated for RCC is provided (Table 4.1). We will focus the following discussion on the broad spectrum of prognostic biomarkers.

Table 4.1. Potential individual molecular prognostic biomarkers for renal cell carcinoma (RCC)

__Kidney Cancer_ Principles and Practice-Springer International Publishing (2015) T 4.1-1

__Kidney Cancer_ Principles and Practice-Springer International Publishing (2015) T 4.1-2

Clinical biomarkers

Historically, multiple clinical algorithms were used to estimate prognosis, including the UCLA Integrated Staging System (UISS) to predict risk for disease recurrence or disease-associated death [55] and the Memorial Sloan Kettering Cancer Center (MSKCC) risk criteria for estimating survival for patients with metastatic disease [56]. The UISS incorporates the TNM staging systems, performance status, and the Fuhrman grade of the tumor and is heavily weighted based on tumor stage. While valuable, this staging system does little to risk stratify those patients with nonmetastatic but sizeable primary tumors. For patients with metastatic disease, which remains incurable with current therapeutic options, the MSKCC algorithm is a valuable clinical tool to establish prognostic intervals for a disease that can range from indolent to rapidly lethal. This system also takes into account the Karnofsky performance status (which can be highly subjective and variable), time from diagnosis to treatment, and laboratory values of hemoglobin, lactate dehydrogenase, and corrected serum calcium. With the widespread clinical use of targeted therapies in RCC, it is necessary for those criteria, which were validated in the era of cytokine therapies, to recruit new biomarkers to match deregulated pathways with effective inhibitors.

In a recent revision of the model, Motzer et al. developed a nomogram that includes both statistically significant and insignificant factors as biomarkers to create a non-biased prognostic model for patients receiving sunitinib [56]. The additional factors included were the number of metastatic sites (p < 0.01), the presence of hepatic metastases (p < 0.1), thrombocytosis (p < 0.01), prior nephrectomy (p = 0.37), the presence of lung metastases (p = 0.74), and serum alkaline phosphatase levels (p = 0.82) [56].

Histological biomarkers

Tumor stage is widely considered by many clinicians as the most important prognostic factor. Historically, effort has focused on identifying critical features in addition to tumor size, such as extracapsular extension, renal vein invasion, inferior venous cava invasion, lymph node involvement, and presence or absence of adrenal gland metastases. It is only recently that the histologic subtyping of RCC into clear cell, papillary, and chromophobe variants gained its long-deserved attention. Aggregation of data has shown that each tumor subtype is associated with different pathophysiology and clinical behavior. In the largest and most comprehensive retrospective review to date, a group of 3,062 cases was identified between 1970 and 2003, among them 2,466 patients (80.5%) with clear cell, 438 (14.3%) with papillary, and 158 (5.2%) with chromophobe RCC. A significant difference in metastasis-free and cancer-specific survival existed between patients with ccRCC and the two other dominant subtypes. Even after multivariate adjustment, the ccRCC subtype remained a significant predictor of metastasis and cancerspecific death [57].

In an effort to estimate prognosis within the ccRCC group, the Fuhrman grading system has been used to further categorize tumors according to tumor cell morphology and correlates tumor grade to mortality [58]. Other histologic features, including the presence of alveolar features, lymphovascular invasion [59], and sarcomatoid dedifferentiation [60] play pivotal roles in prognosis as well, although the degree to which each of these affect prognosis is uncertain.

Genetic biomarkers

Traditional cytogenetic karyotyping studies have altered the approach used in classifying RCC subtypes. Characteristic karyotypes have been consistently associated with each of the most common subtypes of RCC (clear cell, papillary, and chromophobe) [61–63]. In ccRCC, the most frequently observed cytogenetic abnormalities were loss of 3p (60%), gain of 5q (33%), loss of 14q (28%), trisomy 7 (26%), loss of 8p (20%), loss of 6q (17%), loss of 9p (16%), loss of 4p (13%), and loss of chromosome Y in men (55%) [64]. It is interesting that tumors with loss of 3p typically presented at lower TNM stages. Loss of 4p, 9p, and 14q were all associated with higher TNM stages, higher grade, and greater tumor size. A deletion of 3p was associated with better prognosis, while loss of 4p, 9p, and 14q were each associated with worse prognosis [64]. With regard to the less common RCC variants, in papillary RCC, trisomies of chromosomes 7 and 17 were found to be specific genetic alterations irrespective of their size, grade, and cellular differentiation [65]. Another study indicated trisomy 16 and chromosome Y were specifically involved in papillary RCC [66]. The rarest subtype of the three, chromophobe RCC, predominantly showed loss of whole chromosomes, such as loss of chromosomes 1, 2, 6, 10, 13, 17, and 21 [67]. A recent evaluation of the somatic mutation spectrum of chromophobe RCC showed these tumors have commonly mutated TP53 and PTEN genes, although less than half of all tumors had one of these mutations [68]. Further analysis revealed frequent TERT promoter genomic rearrangements in chromophobe RCC, as well as alterations in mitochondrial DNA including increased mitochondrial genome copy numbers and electron transport gene complex 1 mutations [68].

Karyotyping provides a piece of the genetic puzzle of RCC tumorigenesis by elucidating some chromosomal changes. However, in order to complete the puzzle and identify the stepwise progression of RCC carcinogenesis, we have to rely on genomic or exomic sequencing, array comparative genomic hybridization (a-CGH), or SNP analysis.

Recent advances in sequencing technology have made large-scale genomic sequencing rapid and cost-effective. As above, several genes located on chromosome 3p (PBRM1, SETD2 and BAP1) have recently been identified as commonly mutated in ccRCC, along with the frequently mutated VHL gene. These results indicate that large-scale gene sequencing is no longer limited by cost and can provide substantial genetic information to identify heterogeneity in ccRCC.

The presence of these genetic mutations has been shown to have prognostic and predictive significance. Patients with BAP1-mutated tumors have significantly worse median overall survival with a nearly threefold increased hazard ratio for death than those with PBRM1 mutations [50]. BAP1 is also an independent marker of poor prognosis in patients with low-risk disease and may be able to help risk stratify this group of patients [51]. Presence of BAP1 is also associated with metastatic disease at presentation [45]. The combination of BAP1and PBRM1-mutated tumors is rare and has been associated with an even worse overall survival than either mutation alone in most studies, although not in one small study [45, 50]. The BAP1 mutation was originally described via genetic sequencing [41], but immunohistochemical testing has now been validated and also correlates with poor overall survival and adverse clinicopathological tumor features [52]. SETD2 mutations are associated with worse cancer-specific survival in a cohort of patients from the Cancer Genome Atlas, but not an MSKCC cohort [46]. The presence of PBRM1 mutation does not seem to be associated with a change in cancer-specific survival [47], although it has been associated with advanced tumor stage in some earlier studies [69]. It therefore has been suggested to play a more prominent role in tumor initiation instead of disease progression [46].

Gene expression profiles

Multiple studies have used traditional gene profiling using RT-PCR to quantify RNA expression. In 2001, Takahashi et al. studied the expression profile of 29 ccRCC samples and found 51 genes, which could categorize RCC for prognostic purposes [70]. More recently, an analysis of gene expression profiles using machine learning algorithms refined the notion that more than one type of ccRCC was present and used 49 ccRCC samples to define a panel of 120 genes which can accurately define two groups of ccRCC, designated ccA and ccB [71]. This model was refined for application using a NanoString platform using archival renal tumor tissues, demonstrating the feasibility of the approach and showing an advantage of molecular classification using the ClearCode34 biomarker for ccA and ccB integrated with stage and grade over conventional clinical algorithms [72].

Using an RT-PCR platform adapted for fixed tissue analysis, 931 archival formalin-fixed tumor tissues from patients with localized ccRCC were examined across 732 candidate genes [73]. With a median follow-up of 5.6 years, 448 genes were found to be associated with a longer recurrence-free interval (p < 0.05). Sixteen genes had a strong association after consideration of clinical pathologic covariates and false discovery adjustments (HR 0.68–0.80). Among the 16 genes, increased expression of angiogenesisrelated genes (EMCN and NOS3) was associated with lower risk of recurrence, as was increased expression of immune-related genes (CCL5 and CXCL9). This profile provides a feature set readily adaptable to validation studies and has additional promise as a potential predictive biomarker as well. Several of the recently discovered 3p genes commonly mutated in ccRCC also have unique gene expression profiles, but they have been thus far indistinguishable from nonmutant tumors using unsupervised hierarchical clustering algorithms and are therefore not ready for clinical use at this time [42].

Hybrid strategies

The current trend is to incorporate multiple complementary approaches for better identification and understanding of cancer-related genes. Cifola et al. performed the first integrated analysis of DNA and RNA profiles of 27 RCC samples [74]. Seventy-one differentially expressed genes (DEGs) were found in aberrant chromosomal regions and 27 upregulated genes in amplified regions. Among them, the transcripts encoding LOX and CXCR4 were found to be upregulated. Both are implicated for cancer metastasis. Such combinations of genomic and transcriptomic profiling may potentially provide us a more powerful tool for prognostic estimation.

Another trend is to combine epigenetic data with gene expression profiling for better understanding of these interactions. In a preliminary study, an 18-gene promoter methylation panel using quantitative methylation-specific PCR (QMSP) for 85 primarily resected RCC was evaluated [75]. Significant differences in methylation among the four subtypes of RCC were found for CDH1 (p = 0.0007), PTGS2 (p = 0.002), and RASSF1A (p = 0.0001). CDH1 and PTGS2 hypermethylation levels were significantly higher in ccRCC compared to non-ccRCC. RASSF1A methylation levels were significantly higher in papillary RCC than in normal tissue (p = 0.035).

Further validation of epigenetic data in larger cohorts is needed to explore the true prognostic value.

Copy-number analysis

Array comparative genomic hybridization (a-CGH) has been used to identify the specific copy number changes associated with RCC. A comprehensive analysis incorporated a-CGH and gene expression profiles from 90 tumors in order to identify new therapeutic targets in ccRCC [76]. There were 14 regions of nonrandom copy-number change, including seven regions of amplification (1q, 2q, 5q, 7q, 8q, 12p, and 20q) and seven regions of deletion (1p, 3p, 4q, 6q, 8p, 9p, and 14q). An analysis aimed at identifying the relevant genes revealed VHL as one of three genes in the 3p deletion peak, CDKN2A and CDKN2B as the only genes in the 9p deletion peak, and MYC as the only gene in the 8q amplification peak. An integrated analysis to identify genes in amplification peaks that are consistently overexpressed among amplified samples confirmed MYC as a potential target of 8q amplification and identified candidate oncogenes in the other regions.

a-CGH may also improve the diagnostic accuracy for RCC. A recent study examined a-CGH on ex vivo fine-needle aspiration (FNA) biopsies and tumor fragments of 75 RCC patients. The pattern of genomic changes identified by a-CGH was used blindly to classify the renal tumors and the genetic findings were subsequently compared with the histopathologic diagnosis. a-CGH was successful in 82.7% of FNA biopsies and in 96% of tumor fragments. The genetic pattern correctly recognized 93.5% of ccRCC, 61.5% of chromophobe RCC, 100% of papillary RCC, and 14.3% of oncocytoma, with the negative predictive value being above 90% [77]. As RCC histology is an independent predictor of prognosis, one could postulate that a-CGH will have powerful prognostic value as well.

SNP genotyping

Single nucleotide polymorphism (SNP) genotyping has been used to detect cytokine gene polymorphisms in RCC patients to determine its prognostic significance. A panel of 21 SNPs within the promoter regions of 13 cytokine genes were analyzed in a single-center study of 80 metastatic RCC patients [78]. IL4 genotype -589T-33T/-589C-33C was identified as an independent prognostic risk factor in metastatic RCC patients with a median overall survival decreased 3.5-fold (3.78 months, p < 0.05) compared with patients homozygous for IL4 haplotype -589C-33C (13.44 months). An association was also found between three SNPs (-2578C/A, -1154G/ A, and -634C/G) in the VEGF gene and survival of 213 RCC patients [79]. A more recent study found an SNP in IL-8 was associated with survival in patients treated with pazopanib, and these results were validated using data from the COMPARZ trial in sunitinib-treated patients [80, 81]. Multiple VEGF SNPs have also been associated with response and survival as well [80, 82]. These studies contribute evidence that SNP genotyping could be used to develop prognosis algorithms in patients with metastatic RCC.

VHL and HIF as prognostic biomarkers

Based on the extensive discussion of the derangement of this pathway as a result of VHL mutation, it is not surprising then that VHL loss or HIF stabilization might provide a prognostic resource. Perhaps owing to the high prevalence of VHL mutation among ccRCCs, numerous efforts to demonstrate that VHL mutation is a prognostic indicator have been unfruitful. Klatte and colleagues showed preliminary evidence that HIF-1α expression can provide an independent prognostic factor for patients with ccRCC. Patients with high (>35%) tumor immunostaining of HIF-1α had shorter survival than patients with low (=35%) immunostaining of HIF-1α [83]. However, more recent studies have suggested that higher expression of HIF-1α and HIF-2α are associated with improved prognosis [84, 85]. Whether tumor expression of HIF-1α provides substantial prognostic information with respect to the natural history of ccRCC remains to be determined, as does the role of HIF-2α in this setting.

Circulating cells

Levels of circulating endothelial cells and circulating tumor cells have been recently gaining attention as prognostic biomarkers. Several studies have shown that higher levels of circulating endothelial cells or circulating endothelial progenitor cells during the first cycle of VEGFtargeted therapy were associated with improved PFS [86, 87]. However, this technology remains investigational for assessing disease at this time.

[contact-form-7 id=»5168″ title=»Контактная форма 1″]


Добавить комментарий

Войти с помощью: 

Ваш e-mail не будет опубликован. Обязательные поля помечены *