AJCP / Original Article

Loss of CDC14B Expression in Clear Cell Renal Cell Carcinoma Meta-Analysis of Microarray Data Sets Younghye Kim, MD, Jung-Woo Choi, MD, PhD, Ju-Han Lee, MD, PhD, and Young-Sik Kim, MD, PhD From the Department of Pathology, Korea University Ansan Hospital, Ansan, Republic of Korea.

CME/SAM

Key Words: Renal cell carcinoma; Meta-analysis; Microarray; CDC14B DOI: 10.1309/AJCP4PE4JPSRGBQS

ABSTRACT Objectives: To discover significant differentially expressed genes (DEGs) in clear cell renal cell carcinoma (ccRCC) that might be unidentified by single microarray analysis. Methods: The effect sizes of five ccRCC microarray data sets were combined using a random-effects model. The most downregulated gene was validated in paired 80 ccRCC tissues by immunohistochemistry. Results: CDC14B was the most downregulated gene among 1,761 DEGs. CDC14B was strongly expressed in the apical proximal tubules in the nonneoplastic tissues, while it was completely absent in 10 (12.5%) of 80 or downregulated in 70 (87.5%) of 80 ccRCC cases. The complete loss of CDC14B correlated with high T stage (P = .038), advanced TNM stage (P = .027), tumor recurrence (P = .038), and shorter recurrence-free survival (P = .046) compared with the partial loss of CDC14B. Conclusions: Microarray meta-analysis is a useful tool for pathologists. CDC14B expression is downregulated in ccRCC, suggesting its role in renal carcinogenesis.

© American Society for Clinical Pathology

Upon completion of this activity you will be able to: • describe the advantages of meta-analysis of microarray data sets to discover differentially expressed genes in pathologic research. • list the most common genetic alteration in clear cell renal cell carcinoma. • discuss the role of the marker CDC14B in clear cell renal cell carcinogenesis. The ASCP is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The ASCP designates this journal-based CME activity for a maximum of 1 AMA PRA Category 1 Credit ™ per article. Physicians should claim only the credit commensurate with the extent of their participation in the activity. This activity qualifies as an American Board of Pathology Maintenance of Certification Part II Self-Assessment Module. The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose. Questions appear on p 606. Exam is located at www.ascp.org/ajcpcme.

Clear cell renal cell carcinoma (ccRCC) accounts for about 75% of all renal cell carcinomas (RCCs). Histologically, ccRCC is composed of tumor cells with a clear cytoplasm within vascular beds and is believed to originate from the proximal tubular cells.1 Although genes encoding chromatin and histone-regulating proteins are somatically mutated, the most frequent genetic lesion of ccRCC is the von Hippel-Lindau (VHL) mutation with stabilization of hypoxiainducible factor (HIF)–1a and HIF-2a.2,3 Recent studies show that the VHL protein also regulates the biogenesis of primary cilia by orienting microtubules and centrosomes.4-7 The primary cilium is a microtubule-based, nonmotile organelle of almost all interphase mammalian cells that projects from the apical surface of the proximal tubular cells and acts as a phosphoregulated signaling center.8 Loss of the primary cilia in the human kidney is closely linked to

551

Am J Clin Pathol 2014;141:551-558 DOI: 10.1309/AJCP4PE4JPSRGBQS

551 551

Kim et al / CDC14B in Clear Cell Renal Cell Carcinoma

the development of renal cysts and ccRCC with inactivated VHL.8-11 In zebrafish, cell division cycle 14 homolog B (CDC14B) controls the formation of primary cilia, like the VHL. CDC14B, a dual-specific phosphatase, regulates the cell cycle by modulating microtubules. CDC14B-deficient zebrafish leads to the formation of kidney cysts with defective primary cilia.12 However, the mechanisms by which primary cilia are lost in human ccRCC remain to be fully determined. In addition, the expression of CDC14B in human ccRCC has not been investigated. Previous high-throughput microarray studies have reported on differentially expressed genes (DEGs) between ccRCC and the corresponding normal tissues,13-17 between ccRCC with good or poor survival,14,18-20 and for the molecular subclassification of ccRCC.21 However, discordances among the results of individual microarray studies may be caused by the differences in tumor sample sets, microarray platforms and data analysis, and small sample size.22 To overcome the limitations of individual microarray reports, meta-analysis of microarray data sets has been proposed.23-27 The meta-analysis yields more reliable results because of its increased sample size, which is much larger than the number of the sample in each study.28-32 To discover significant genes that might be unidentified by single microarray analysis, we generated DEG lists by combining publicly available ccRCC microarray data sets. Among the DEGs, we validated the most significantly downregulated CDC14B in formalin-fixed, paraffin-embedded tissues using immunohistochemistry and compared its expression with clinicopathologic parameters of ccRCC.

Materials and Methods Microarray Data Sets This meta-analysis combined five microarray data sets of ccRCC retrieved from Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo/) and ArrayExpress (www.ebi. ac.uk/arrayexpress/) ❚Table 1❚.13-17 The five studies presented differential gene expression between ccRCC and

the corresponding renal cortex. The microarray data sets consisted of raw cell intensity files from a total of 77 ccRCC and 64 nonneoplastic renal tissues with a range of 22,283 to 54,675 Affymetrix (Santa Clara, CA) probes. Data Preprocessing and Integration The microarray data sets were prepared and processed according to a previous protocol.28 All analyses were carried out by R version 2.7.2 (R Foundation for Statistical Computing, Vienna, Austria) and Bioconductor release 2.2 (Fred Hutchinson Cancer Research Center, Seattle, WA).33 The effect sizes of data sets were combined using the inverse variance random-effects model. The chosen microarray cell intensity files were preprocessed into a gene expression data matrix using a robust multichip average algorithm. Thus, the raw intensity values were background corrected, log2 transformed, and then quantile normalized.34 Next, the probe sets were mapped to UniGene using the Bioconductor annotation package. Nonidentifiable probes were discarded. For multiple clones matched with the same UniGene accession, the one with the least missing values was selected. The effect size was calculated as Hedges’ adjusted g for every gene in each study, and the effect sizes were combined using the random-effects model. The z statistic, which is the ratio of the pooled effect size to its standard error, was compared with a standard normal distribution to obtain the nominal P value for every UniGene ID. The false discovery rate (FDR)–adjusted P values were used for selecting statistically significant upregulated and downregulated genes. Patients Eighty consecutive cases of ccRCC and the matched nonneoplastic renal tissues were selected at the archive of the Department of Pathology, Korea University Ansan Hospital, from 2001 through 2011. This study was approved by the institutional review board of Korea University Ansan Hospital (reference No. AS12070). All cases were fixed in 10% buffered formalin and embedded in paraffin. Clinicopathologic parameters of the patients with ccRCC, including age, sex, Fuhrman grade, TNM stage, presence of tumor

❚Table 1❚ Microarray Data Sets Used for Meta-Analysis Study

No. of Samples Repository ID

Normal (n = 64)

Tumor (n = 77)

Array Platform

No. of Probes

Lenburg et al13 GSE781 8 9 HGU133 set 44,928 Jones et al14 GSE15641 23 32 HGU133A 22,283 15 Gumz et al GSE6344 10 10 HGU133 set 44,928 Cifola et al16 E-TABM-282 11 16 HGU133plus 2.0 54,675 Wang et al17 GSE14762 12  10 HGU133plus 2.0 54,675

552 552

Am J Clin Pathol 2014;141:551-558 DOI: 10.1309/AJCP4PE4JPSRGBQS

© American Society for Clinical Pathology

AJCP / Original Article

recurrence, and cause of death, were reviewed. Tumor recurrence was defined based on the evidence of imaging modalities (4/7) or biopsy results (3/7). Fuhrman nuclear grades were grouped as low (G1-G2) and high (G3-G4).35 TNM stage, determined according to the guidelines of the American Joint Committee on Cancer,36 was classified as low (I-II) and high (III-IV) for statistical analysis. Primary T stage was divided into low (T1-T2) and high (T3-T4). Immunohistochemistry for CDC14B A tissue microarray was constructed from formalinfixed, paraffin-embedded tissue blocks. Considering tumor heterogeneity, three cores 1 mm in diameter from each tissue block of tumor and the corresponding normal cortical areas were inserted in a grid pattern into recipient paraffin blocks using a tissue arrayer (Beecher Instruments, Sun Prairie, WI). The immunostaining pattern of CDC14B was confirmed in 10 whole sections of normal kidney and ccRCC. Among the top 10 candidate biomarkers, CDC14B was selected for tissue validation. Immunohistochemistry was performed in the tissue microarray sections using a ChemMate Envision Kit (Dako, Carpinteria, CA). Tissue sections were deparaffinized and dehydrated. Sections in 1 mmol/L EDTA buffer (pH 8.0) were heated in a microwave for 10 minutes. After incubation with 0.3% hydrogen peroxide/ methanol for 20 minutes, slides were stained with a 1:20 dilution of anti-CDC14B (Sigma-Aldrich, St Louis, MO). After exposure to secondary antibody at room temperature for 30 minutes, the sections were developed with 3,3′-diaminobenzidine (DAB) and counterstained with hematoxylin. We used positivity in Purkinje cells of the cerebellar cortex as a positive control for CDC14B. As a negative control, immunostaining was carried out with the omission of the primary antibody. CDC14B immunostaining in the nuclei and/ or cytoplasm of normal renal tubular cells and tumor cells was considered positive. The nuclear membranous staining was also interpreted as positive. Image Analysis Digital images were acquired using a BX51 light microscope with a DP70 digital color camera (Olympus, Tokyo, Japan). One digital image for one core was taken under a ×20 objective field and saved in JPEG format at 1,360 × 1,024–pixel resolution. The percentage of DAB-stained area of the total area was calculated using ImageJ (National Institutes of Health, Bethesda, MD). First, the original image was preprocessed by background correction. Each image was separated into a hematoxylin image and a DAB image by the color deconvolution plug-in. The DAB image was segmented by automatic thresholding based on pixel intensities. Finally, the image was converted to black and white images and the area fractions were measured. © American Society for Clinical Pathology

We calculated the representative area percentage of each case by averaging the area percentages of three tissue cores. The cutoff value between a complete loss and a partial loss of CDC14B expression was determined using receiver operator characteristic curves of immunostained area percentages, considering the sensitivity and specificity for the tumor recurrence.37 We regarded a complete loss of CDC14B expression when the immunostained area percentage was 4.5% or less. Statistical Analysis The differential expression of CDC14B between ccRCC and nonneoplastic cortical tissues was evaluated by a paired sample t test. The relationships between CDC14B expression and clinicopathologic parameters were tested by Fisher exact tests. Univariate and multivariate survival analyses were conducted with the Kaplan-Meier method with log-rank tests and the Cox proportional hazards regression model. P values less than .05 were considered significant. Statistical analysis was carried out with SPSS version 10 (SPSS, Chicago, IL).

Results Differentially Expressed Genes by Meta-Analysis We identified novel upregulated or downregulated genes in ccRCC compared with nonneoplastic cortical tissue using a microarray meta-analysis. We found 1,761 significant DEGs at the FDR of 0.1%: 857 upregulated and 904 downregulated genes. Among the selected top 10 genes with the lowest FDR, six upregulated genes included CELF2, HLA-DRB1, ZNF395, SLC1A4, ISG20, and MCM5, while four downregulated genes were CDC14B, AP1M2, HIP1R, and SYNJ2BP ❚Table 2❚. We selected CDC14B, the most downregulated gene, for tissue validation using immunohistochemistry. The forest plot for the CDC14B gene (Hs.40582) is shown in ❚Figure 1❚.13-17 Patient Characteristics Patients with ccRCC included 52 (65.0%) men and 28 (35.0%) women with an age range at diagnosis of 27 to 79 years (median, 58 years). Four cases showed focal sarcomatoid areas. With regard to Fuhrman grade, 11 (13.8%) cases were classified as low (grade I in one and grade II in 10) and 69 (86.3%) cases as high (grade III in 51 and grade IV in 18). Sixty patients were stage I, nine were stage II, nine were stage III, and two were stage IV. Considering only primary tumor extent, 73 patients were T1 or T2 (T1a in 38, T1b in 27, T2a in seven, and T2b in one) and seven were T3a. Seven patients had tumor recurrence during clinical followup. The median duration of follow-up was 33.5 months.

553

Am J Clin Pathol 2014;141:551-558 DOI: 10.1309/AJCP4PE4JPSRGBQS

553 553

Kim et al / CDC14B in Clear Cell Renal Cell Carcinoma

❚Table 2❚ Top 10 Differentially Expressed Genes in Meta-Analysis of Microarrary Data Sets Gene Symbol

Gene Name

UniGene ID

No.a

Effect Size (SE)b

Upregulated CELF2 CUGBP, Elav-like family member 2 Hs.309288 5 1.369 (0.067) HLA-DRB1 Major histocompatibility complex, class II, DR b1 Hs.534322 5 1.102 (0.056) ZNF395 Zinc finger protein 395 Hs.695998 5 2.552 (0.147) SLC1A4 Solute carrier family 1, member 4 Hs.654352 5 1.395 (0.094) ISG20 Interferon-stimulated exonuclease gene 20 kDa Hs.459265 5 2.836 (0.191) MCM5 Minichromosome maintenance complex component 5 Hs.517582 5 2.408 (0.165) Downregulated CDC14B CDC14 cell division cycle 14 homolog B  Hs.40582 5 –1.202 (0.075) AP1M2 Adaptor-related protein complex 1, m2 subunit Hs.18894 5 –2.791 (0.176) HIP1R Huntingtin interacting protein 1 related Hs.524815 5 –1.879 (0.121) SYNJ2BP Synaptojanin 2 binding protein Hs.704691 5 –1.629 (0.115)

t2c pFDR 0 0 0.04 0 0.02 0

9.3E–90 1.1E–83 7.8E–64 7.4E–55 1.8E–53 4.9E–51

0 0 0 0

7.4E–47 8.1E–47 2.0E–45 1.2E–42

pFDR, the false discovery rate–adjusted P value. a The number of studies in which the gene was present. b Effect size is the log -transformed pooled effect size, with the SE of the pooled effect size. 2 c Between-study heterogeneity measure.

CDC14B Expression Cifola et al

16

Gumz et al15 Jones et al14 Lenburg et al13 Wang et al17 Summary –2.0 –1.8 –1.6 –1.4 –1.2 –1.0 –0.8 Standardized Mean Difference (Log2 Scale)

❚Figure 1❚ Forest plot for the most significantly downregulated gene, CDC14B (Hs.40582), discovered from the microarray meta-analysis (Hedges’ g, –1.2; false discovery rate–adjusted P = 7.42E–47). 25

% Area

20

15

10

5

N

T

❚Figure 2❚ Comparison of CDC14B expression between nonneoplastic and clear cell renal cell carcinoma tissues. Percentage areas of CDC14B immunostaining are more than two times lower in tumor tissues (T) than in normal tissues (N) (mean ± SD, 7.6% ± 2.8% vs 15.4% ± 3.6%, paired sample t test, P < .001). 554 554

Am J Clin Pathol 2014;141:551-558 DOI: 10.1309/AJCP4PE4JPSRGBQS

CDC14B expression was completely (12.5%, or 10/80) or partially lost (87.5%, or 70/80) in all ccRCC cases compared with the nonneoplastic renal cortex. The percentage areas of CDC14B immunostaining were significantly lower in ccRCC vs normal cortex (mean ± SD, 7.6% ± 2.8% vs 15.4% ± 3.6%, paired sample t test, P < .001) ❚Figure 2❚. In normal renal cortex, CDC14B was strongly expressed on the apical surface of the proximal tubular cytoplasm and moderately expressed in the nuclei of all proximal and distal tubules ❚Image 1A❚ and ❚Image 1B❚. In contrast, CDC14B expression was absent or markedly reduced in the cytoplasm of tumor cells. However, CDC14B expression was retained in the apical portion of some tumor cells ❚Image 1C❚ and ❚Image 1D❚. Nuclear staining intensities of CDC14B in tumor cells were decreased compared with those in normal tubular cells. The nuclear staining of CDC14B was more intense around the nuclear membranes of tumor cells as well as normal tubular cells (Image 1B). CDC14B was also mildly expressed in the cytoplasm of some endothelial cells in both normal renal parenchyma (data not shown) and tumor stroma (Image 1D). Ten (12.5%) of 80 ccRCC cases showed complete loss of CDC14B expression. The complete loss of CDC14B expression was significantly associated with high T stage (P = .038), advanced TNM stage (P = .027), and tumor recurrence (P = .038) compared with the partial loss of CDC14B expression ❚Table 3❚. In Kaplan-Meier survival analysis, the complete loss of CDC14B expression was highly related to shorter recurrence-free survival compared with the partial loss of CDC14B (log-rank test, P = .046) ❚Figure 3❚. In addition, female sex (P = .0014) and advanced TNM stage (P < .001) were related to a short recurrence-free survival. In multivariate Cox proportional hazards regression, only female sex (hazard ratio, 9.6; 95% CI, 1.1-81.7; P = .039) and TNM stage (hazard ratio, 12.3; 95% CI, 2.3-66.9; P = .004) were associated with recurrence-free survival. © American Society for Clinical Pathology

AJCP / Original Article

A

B

C

D

❚Image 1❚ CDC14B immunohistochemistry. A, The proximal tubular cells express CDC14B in the cytoplasm with strong apical staining. The nuclei of proximal and distal tubular cells also express CDC14B (×400). B, The eight-bit image of CDC14B immunohistochemistry clearly demonstrates the immunostaining pattern of CDC14B, which was made after color deconvolution and segmentation by ImageJ (National Institutes of Health, Bethesda, MD) software (×400). C, Apical staining pattern of CDC14B is focally present in clear cell renal cell carcinoma (ccRCC) (×400). D, CDC14B is also expressed in the nuclei of some tumor cells and the cytoplasm of endothelial cells in ccRCC (×400).

Discussion We discovered 857 upregulated and 904 downregulated DEGs by combining five ccRCC microarray data sets. Among the downregulated DEGs, we first identified that CDC14B expression was completely or partially lost in human ccRCC compared with normal renal cortex. The complete loss of CDC14B expression is highly associated with advanced TNM stage and a short recurrence-free survival of patients with ccRCC in univariate analysis. Among the top 10 upregulated DEGs, ZNF395 is a target of HIF-1a, which is overexpressed in an RCC cell line © American Society for Clinical Pathology

with inactivated VHL38 and in a glioblastoma cell line under hypoxic conditions.39 ISG20 is overexpressed in response to interferon treatment and related to the innate immune response.40 MCM5 is a cell cycle initiator that promotes cellular proliferation.13 This finding strongly supports the view that ccRCC is characterized by the upregulation of HIF-target gene expression, activation of immune pathways, and cell cycle progression. The most significantly downregulated gene was a dualspecific phosphatase, CDC14B. In normal cortical tissues, CDC14B was expressed in all nuclei of the proximal and distal tubules with the accentuation of nuclear membrane staining.

555

Am J Clin Pathol 2014;141:551-558 DOI: 10.1309/AJCP4PE4JPSRGBQS

555 555

Kim et al / CDC14B in Clear Cell Renal Cell Carcinoma

❚Table 3❚ Loss of CDC14B Expression in Clear Cell Renal Cell Carcinoma

CDC14B Expression, No. (%)

Clinicopathologic Parameters

Complete Loss (n = 10)

Incomplete Loss (n = 70)

P Value

Age, y .204

Loss of CDC14B expression in clear cell renal cell carcinoma: meta-analysis of microarray data sets.

To discover significant differentially expressed genes (DEGs) in clear cell renal cell carcinoma (ccRCC) that might be unidentified by single microarr...
628KB Sizes 2 Downloads 3 Views