Clinical Cancer Research Bridging the Lab and the Clinic in Cancer Medicine Translational Cancer Medicine 2008: Cancer Clinical Trials and Personalized Medicine
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Cell Growth & Differentiation
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chen, Y.-T.
Right arrow Articles by Mazumdar, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chen, Y.-T.
Right arrow Articles by Mazumdar, M.
Clinical Cancer Research Vol. 11, 6558-6566, September 15, 2005
© 2005 American Association for Cancer Research

Imaging, Diagnosis, Prognosis

Messenger RNA Expression Ratios among Four Genes Predict Subtypes of Renal Cell Carcinoma and Distinguish Oncocytoma from Carcinoma

Yao-Tseng Chen1, Jiangling J. Tu1, Jean Kao1, Xi K. Zhou2 and Madhu Mazumdar2

Authors' Affiliations: Departments of 1 Pathology and Laboratory Medicine and 2 Public Health, Division of Biostatistics and Epidemiology, Weill Medical College, Cornell University, New York, New York

Requests for reprints: Yao-Tseng Chen, Department of Pathology, Weill Medical College, Cornell University, Room C-458A, 1300 York Avenue, New York, NY 10021. Phone: 212-746-6472; Fax: 1-212-746-4483; E-mail: ytchen{at}med.cornell.edu.


   Abstract
Top
 Abstract
Materials and Methods
 Results
 Discussion
 References
 
Purpose: Morphologic distinction among clear cell, papillary, and chromophobe types of renal cell carcinoma (RCC) can be difficult, as is the differential diagnosis between oncocytoma and RCC. Whether these renal tumors can be distinguished by their mRNA expression profile of a few selected genes was examined.

Experimental Design: The expression of four genes in renal tumor was evaluated by quantitative reverse transcription-PCR: carbonic anhydrase IX (CA9), methylacyl-CoA racemase (AMACR), parvalbumin (PVALB), and chloride channel kb (CLCNKB). Thirty-one fresh-frozen and 63 formalin-fixed, paraffin-embedded tumor specimens were analyzed.

Results: CA9 expression was highest in clear cell carcinoma and lowest in chromophobe RCC and in oncocytoma. AMACR expression was highest in papillary RCC, and CLCNKB was highest in chromophobe RCC/oncocytoma. PVALB was highest in chromophobe RCC, variable in oncocytoma, and low in clear cell and papillary types. Similar findings were observed in fresh-frozen and formalin-fixed specimens. The mRNA expression ratios among these genes (i.e., CA9/AMACR and AMACR/CLCNKB ratios) further accentuate the gene expression differences among these tumors, and a molecular diagnostic algorithm was established. This algorithm accurately classified the 31 fresh-frozen tumors into 14 clear cell, 5 papillary, 6 chromophobe, and 6 oncocytomas. In the formalin-fixed group, the molecular criteria accurately classified the cases into 15 clear cell, 16 papillary, and 32 in the chromophobe/oncocytoma group but could only separate some, but not all, oncocytomas from chromophobe RCC.

Conclusions: RNA expression ratios based on the four-gene panel can accurately classify subtypes of RCC as well as help distinguish some oncocytomas from chromophobe RCC.

Renal cell carcinomas (RCC) are classified histologically into conventional clear cell, papillary, and chromophobe types (1). Clear cell type is by far the most common, accounting for 70% to 80% of the RCC. In comparison, 10% to 15% of the RCC are of the papillary type and ~5% are of the chromophobe type (2). The distinction of these different types is clinically important, as papillary and chromophobe types have a better prognosis than the clear cell type (3). For instance, Thoenes et al. (4) reported a 5-year overall survival rate of 92% for chromophobe in contrast to 62% for clear cell type. Similarly, a 5-year survival rate of 82% to 90% has also been reported for papillary renal carcinoma (5, 6).

Although the histologic diagnosis of these different subtypes of RCC can be made unequivocally in most cases, selected cases can be difficult, particularly the chromophobe type. Moreover, the differential diagnosis between RCC and renal oncocytoma can pose a diagnostic dilemma. Oncocytoma, comprising 3% to 5% of the renal tumors, is generally considered a benign neoplasm (7). However, some cases of oncocytoma can show atypical histologic features, leading to their being interpreted as RCC, either as chromophobe type or as RCC, unclassified, for fear of underdiagnosing a malignant neoplasm. To help resolve these morphologic dilemmas in surgical pathology, attempts have been made to segregate RCC types by immunohistochemical staining, with variable results (e.g., refs. 8–10).

An alternative to immunohistochemical staining would be to distinguish these tumors at the DNA or RNA level. Despite their overlapping histologic features, different RCC subtypes are biologically distinct. This is evident from their characteristic cytogenetic abnormalities, the most common ones being the chromosome 3p aberrations in clear cell type, polysomes in papillary RCC, most often trisomy 17 and trisomy or tetrasomy 7, and multiple chromosomal losses in chromophobe type (6). These biological differences imply that these tumors should bear distinguishable molecular signatures that could be useful diagnostic tools. This notion is supported by recent gene profiling studies (11–15). In compilation, these studies indicated that clear cell, papillary, and chromophobe RCC could be reliably separated by cDNA microarray technique and hierarchical clustering analysis. However, chromophobe RCC and oncocytoma were found to closely clustered, reflecting their common origin from the collecting duct (6).

Although current evidence indicates that accurate classification of RCC can be achieved by DNA microarray analysis, the complexity and the cost of the assay exclude its use as a routine diagnostic test at present. In this study, we show that by quantitative reverse transcription-PCR (qRT-PCR) analysis of four genes, carbonic anhydrase IX (CA9), methylacyl-CoA racemase (AMACR), chloride channel kb (CLCNKB), and parvalbumin (PVALB) the histologic subtypes of the RCCs could be accurately predicted. In addition, oncocytomas can be separated from clear cell and papillary types in all cases and distinguished from chromophobe RCC in some cases.


   Materials and Methods
Top
 Abstract
 Materials and Methods
Results
 Discussion
 References
 
Tumor tissues. Tumor specimens were obtained from the Department of Pathology at the Weill Medical College of Cornell University. The H&E slides on all cases were reviewed by one of us (J.J.T.), and only histologically unequivocal cases were included in the study analysis. Fresh-frozen tumor specimens were collected following an institutional review board–approved protocol. For formalin-fixed, paraffin-embedded tissue blocks, four 8-µm unstained sections were obtained from one representative block from each case. The nontumor areas on the slides were manually removed with surgical blades, and the remaining tissue on the slide was scraped into an Eppendorf tube for RNA extraction.

In silico analysis to identify candidate genes for quantitative reverse transcription-PCR and primer design. To identify the most promising candidates from the selected gene list, the expression profile of each gene in normal tissues was evaluated using a combination of the Serial Analysis of Gene Expression Anatomic Reviewer and its Virtual Northern tool (http://cgap.nci.nih.gov/SAGE/AnatomicViewer) and database searches using BLASTN (http://www.ncbi.nlm.nih.gov/BLAST). The objective of the analysis was to identify genes with limited expression in normal tissues, focusing on genes with exclusive or predominant expression in kidney. Once such a gene is identified, the exon-intron structure of the gene was defined, and trans-intronic primers for conventional as well as real-time RT-PCR were designed using Primer3 software (http://frodo.wi.mit.edu/primer3/primer3_code.html).

Conventional PCR primers were custom synthesized by Qiagen (Valencia, CA), whereas the qRT-PCR primers and probes were synthesized by Applied Biosystems (Foster City, CA) with 5'-FAM and 3'-TAMRA labeling. The primer sequences are listed in Table 1. Primer-probe premix for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) endogenous control was purchased from Applied Biosystems.


View this table:
[in this window]
[in a new window]
  		Table 1. Primers and probes for qualitative and qRT-PCR

 
RNA extraction from fresh-frozen and paraffin-embedded tissues. Total RNA was extracted from fresh tissues using RNeasy Mini kit (Qiagen) following the manufacturer's protocol using rotor-stator homogenization. Fresh tissues (~30 mg) were used for each sample, with the final RNA yield in the range of 10 to 60 µg based on A260 nm by spectrophotometer. For extraction of RNA from paraffin-embedded tissue, the Optimum FFPE RNA isolation kit (Ambion, Austin, TX) was used using materials derived from four 8-µm sections. The quantity of RNA isolated ranged from 0.14 to 9.70 µg.

Conventional reverse transcription-PCR. Total RNA (2 µg) isolated from fresh-frozen tissues was used for a 20 µL reverse transcription reaction. Synthesized cDNA (2 µL) was then used per 25 µL PCR assay. The PCR was set up using AmpliTaq Gold and its corresponding 10x PCR mix (Applied Biosystems) in an ABI 9600, with 35 cycles of amplification, each cycle consisting of 15 seconds at 94°C, 1 minute at 60°C, and 1 minute at 72°C. The PCR products were visualized by 1.2% agarose gel electrophoresis and ethidium bromide staining.

Quantitative reverse transcription-PCR. qRT-PCR was done using an ABI PRISM 7000 Sequence Detection System. Total RNA (up to 2 µg) was used per 20 µL reverse transcription reaction, and cDNA (2 µL) was then used for each 25 µL PCR, corresponding to 100 ng total RNA. Forty-five two-step cycles of amplification were done, each cycle consisting of 15 seconds at 95°C and 1 minute at 60°C. Control amplification for housekeeping gene GAPDH was done in all samples. All fresh-frozen specimens included in the final analysis have Ct values differing by less than two cycles, indicating similar cDNA quality and quantity. One case of papillary carcinoma had much poorer GAPDH amplification possibly due to inclusion of necrotic areas in the tissue used for RNA extraction. This case was excluded from further study. The GAPDH amplification results were more variable for paraffin-embedded samples, but all samples successfully amplified and were included in the final analysis.

Analysis of quantitative reverse transcription-PCR data. Following the amplification, the same threshold was set for analyzing all experiments to compare Ct values derived from different experiments. The mean Ct values from each sample were normalized against the corresponding GAPDH Ct values calculated as (Ctexperimental gene – CtGAPDH). All data in Results represented normalized Ct values.

For comparing the expression levels of individual gene in various types of tumor, the normalized Ct values were plotted using box-whiskers plots according to each tumor type. These plots showed the median and the range of the Ct values as well as Ct values corresponding to 25% and 75% of the specimens. As such, these plots allowed more quantitative comparisons between groups than other graphic tools, such as scatter plots.

In addition to evaluating individual gene expression difference among tumor types, the relative expression between paired genes was found to be useful predictors of the tumor type (see Results). For this comparison, the relative expression of the paired genes was calculated as (raw CtGene A – raw CtGene B) and expressed as {Delta}Ct (gene A – gene B) or (gene A – gene B) for short. This {Delta}Ct value reflected directly the ratio between mRNA levels of the two genes, which equaled 2–{Delta}Ct if these two genes were amplified with similar efficiencies. For example, a tumor with (CA9 – AMACR) value of –4.0 would have CA9 mRNA level 16-fold (24) of the AMACR mRNA level at the same specimen, as CA9 and AMACR did amplify with similar efficiency.

Statistical analysis and modeling. Regression tree method (16, 17) was used to create the decision rules for classification of renal tumors based on the expression ratios of genes chosen as predictors (see Results), with the goal being to provide highest classification accuracy using the known histologic diagnosis as the "gold standard."

Once the model was established, grouped cross-validation is used to estimate the prediction accuracy and model fitting accuracy. In this analysis, the data are divided into 12 similar size groups. Sequentially, each of the 12 groups is treated as the validation sample and the other 11 groups are treated as training sample. The model is fitted based on data from the 11 groups and the prediction was made for the 12th group. This process was repeated 20 times to compute the average misclassification rate in the training sample and the validation sample.


   Results
Top
 Abstract
 Materials and Methods
 Results
Discussion
 References
 
Identification of target genes. Microarray data from previously published studies on renal cancer (11–13) were examined and a small panel of potentially useful genes was selected (Table 2). The mRNA expression of these genes in normal and neoplastic tissues was evaluated by in silico analysis using bioinformatic tools (see Materials and Methods). Broadly expressed genes, such as vimentin, although showing preferential expression in one type of renal carcinoma (clear cell) versus other subtypes (11), were excluded from further analysis, as RNA from contaminating benign stromal tissue could potentially affect the interpretation of qRT-PCR data derived from tumor tissues. Based on these criteria, CA9, AMACR, PVALB, CLCNKB, and defensin ß1 (DEFB1) were selected, as they all showed evidence of tissue-restricted mRNA expression, minimizing the concern of normal tissue contamination. The preferential expression of these genes in renal tumor was validated by preliminary conventional qualitative RT-PCR analysis on a selective panel of normal and renal cancer tissues (data not shown). Estimated from the yield of the RT-PCR products, normal kidney showed strong expression of PVALB, DEFB1, and CLCNKB, low expression of AMACR, and minimal expression of CA9 (Fig. 1).


View this table:
[in this window]
[in a new window]
  		Table 2. Genes preferentially expressed in different types of renal tumors by DNA microarray studies

 


View larger version (47K):
[in this window]
[in a new window]
  		Fig. 1. RT-PCR analysis for CA9, DEFB1, PVALB, AMACR, and CLCNKB expression. A case of clear cell carcinoma was analyzed, with RNA prepared from two different areas of tumor (T) and uninvolved kidney (N). A 100-bp ladder was used as the molecular weight size standard (MW). All RT-PCR amplicons have the predicted size of 100 to 200 bp. CA9 expression was much stronger in tumor than in normal kidney. The DNA species of ~700 bp in the normal (N) lanes of CA9 represented amplification of contaminating genomic DNA in the sample. In contrast, DEFB1, PVALB, and CLCNKB were all stronger in normal kidney than in tumor tissues. AMACR showed similarly low level of expression in both normal and tumor tissue.

 
Quantitative reverse transcription-PCR on fresh-frozen tissues. qRT-PCR for these five genes was then done on 33 fresh-frozen samples of renal tumor, including 14 clear cell, 5 papillary, 6 chromophobe RCC, and 6 oncocytomas. Two cases were diagnosed as unclassified RCC, and data from these two cases were not used for the initial analysis and statistical modeling.

CA9 was strongly expressed in clear cell RCC, with moderate expression in papillary carcinoma, and low expression in chromophobe RCC and in oncocytoma. A representative qRT-PCR amplification plot is shown in Fig. 2. Shown in normalized Ct (see Materials and Methods), for which a lower Ct value indicates a higher expression, the median Ct for clear cell RCC was 5.612 (range, 4.627-8.022) in contrast to 19.410 for chromophobe RCC and oncocytoma combined (range, 17.109-22.334) and 12.042 for papillary renal carcinoma (Table 3; Fig. 3). As the fold difference at levels of mRNA expression equal 2{Delta}Ct in this experiment, the Ct range difference indicated that the lowest CA9 mRNA expression in clear cell type is >512 (29)–fold that of the highest level observed in chromophobe/oncocytoma, conclusively distinguishing clear cell carcinoma from chromophobe RCC/oncocytoma. In contrast, chromophobe RCC and oncocytoma showed very similar levels of CA9 expression and were essentially indistinguishable.



View larger version (59K):
[in this window]
[in a new window]
  		Fig. 2. An example of qRT-PCR tracing, showing CA9 expression in clear cell carcinoma and in chromophobe RCC/oncocytoma. Two clearly separately clusters were seen, the cluster on the left being from clear cell RCC samples, with Ct at the set threshold ranging between 19.5 and 22.5 (before normalization against GAPDH control). The larger cluster on the right was from chromophobe RCC/oncocytoma samples, with Ct ranging from 32 to 38.

 

View this table:
[in this window]
[in a new window]
  		Table 3. qRT-PCR on fresh-frozen renal tumor specimens

 


View larger version (21K):
[in this window]
[in a new window]
  		Fig. 3. Distribution of normalized Ct (Y axis) depicting data derived from fresh-frozen tissue (see Table 3) in a box-whisker plot. The box depicted the borders of 25% and 75% quartiles, with the horizontal bar showing the median value. The range in each group is shown in vertical lines.

 
CLCNKB gene showed expression profiles opposite to that of CA9. Chromophobe RCC consistently showed high level of CLCNKB transcripts (median Ct, 3.183), followed by papillary RCC (median Ct, 8.965), with clear cell carcinoma showing lowest expression of CLCNKB (median Ct, 13.048). Oncocytomas showed a range of expression overlapping with that of chromophobe type, with the average expression level (median Ct, 4.793) about one-third that of chromophobe but 16-fold higher than papillary type and 256-fold higher than clear cell type.

PVALB expression was similar to CLCNKB in that the expression was highest in chromophobe (median Ct, 2.978) and lowest in clear cell type (median Ct, 17.787). Papillary RCC also showed low PVALB mRNA expression (median Ct, 15.596). Intriguingly, oncocytomas showed a broad range of PVALB expression ranging from 4.777 to 14.024, indicating that expression of this gene in oncocytoma is variable, in some cases much lower than in chromophobe RCC (range, 1.463-3.478). Because chromophobe RCC is always associated with strong PVALB expression, the finding of low PVALB expression is thus supportive of the diagnosis of oncocytoma, if chromophobe RCC is the only alternative diagnosis being considered.

AMACR expression was highest in papillary RCC (median Ct, 3.312). All other categories showed only minor differences in their AMACR levels, with the median for clear cell RCC, chromophobe RCC, and oncocytoma being 6.625, 6.697, and 8.706, respectively.

DEFB1 expression was evaluated in 25 of the 33 cases. It showed highest expression in chromophobe (median Ct, 4.907; range, –0.831 to 5.869) and oncocytoma (median Ct, 3.164; range, 1.251-10.868), with lower expression in papillary type (median Ct, 5.777; range, 3.565-11.549) and clear cell type (median Ct, 9.348; range, 5.293-15.465). The overlapping ranges in these four groups indicate that there is substantial variation in DEFB1 mRNA expression within individual types. Therefore, this gene was judged as having an expression profile similar to CLCNKB and PVALB but less consistent, and DEFB1 was not analyzed in subsequent experiments.

Gene expression ratios highlight differences among tumor types. From the distribution of the Ct values in Fig. 3, it is clear that a very high-level mRNA expression of CA9, AMACR, or CLCNKB/PVALB is essentially diagnostic of clear cell, papillary, and chromophobe RCC/oncocytoma, respectively. The inverse expression profile seen in some of these genes in different tumor types further suggested that evaluation of the expression ratios between paired genes could be more powerful in distinguishing these tumor types. For instance, because clear cell carcinomas express higher CA9 than papillary RCC and because papillary RCC express higher AMACR than clear cell RCC, the expression difference of these two genes, expressed as {Delta}Ct (CA9 – AMACR; see Materials and Methods), could be anticipated to be more reliable than either gene alone in separating these two tumor types. Similarly, {Delta}Ct (CLCNKB – AMACR) should help distinguish papillary RCC from chromophobe RCC/oncocytoma. This notion was proven correct, and Fig. 4 showed that {Delta}Ct (CA9 – AMACR) could distinguish clear cell RC from all three other categories, whereas {Delta}Ct (CLCNKB – AMACR) could distinguish papillary RCC from chromophobe RCC and oncocytoma.



View larger version (12K):
[in this window]
[in a new window]
  		Fig. 4. Distribution of gene expression ratios in the fresh-frozen tissue group as reflected in {Delta}Ct (Y axis; see text). By calculating the {Delta}Ct values between paired genes, clear cell carcinoma (CC) can be easily separated from all other tumor types by {Delta}Ct (CA9 – AMACR), and papillary RCC (Pap) can be separated from chromophobe RCC/oncocytoma by {Delta}Ct (CLCNKB – AMACR). Chromophobe RCC (Chr) can then be separated from oncocytoma (Onc) by PVALB expression.

 
Quantitative reverse transcription-PCR using formalin-fixed, paraffin-embedded tissues. To evaluate whether these findings can be used in routine surgical pathology, RNA extracted from 66 cases of renal tumors were used for qRT-PCR, consisting of 15 clear cell, 19 chromophobe, 16 papillary, 13 oncocytoma, and 3 RCC, unclassified.

In comparison to PCR data using fresh tissues, RNA from formalin-fixed, paraffin-embedded amplified poorer, with Ct values obtained ~4 to 8 cycles behind what was observed in fresh tissues. However, the effects of this poor RNA quality were eliminated by the process of normalization with GAPDH endogenous control, and the values of the normalized Ct were in general similar to those described above for fresh tissues. However, for low abundant mRNA species, such as CA9 in chromophobe/oncocytoma group, no amplification was seen at the end of 45 cycles and an accurate numerical Ct could not be assigned. For such cases, a Ct value of 45 was used for the calculation of gene ratios ({Delta}Ct).

The qRT-PCR data on all 63 cases (excluding the 3 unclassified RCC) were summarized in Table 4, and the critical gene ratios were plotted in Fig. 5. Comparing Tables 3 and 4, it is clear that the gene expression profiles were highly similar between fresh-frozen and formalin-fixed tissues. However, two notable differences were observed. One was that the ranges of mRNA level distribution tend to be wider in the formalin-fixed, paraffin-embedded group for the same tumor category, likely attributed from the greater variation in the RNA quality of formalin-fixed tissues. The second difference observed was the PVALB expression in the chromophobe/oncocytoma group. In the fresh sample group, the chromophobe RCC showed higher level of PVALB expression than oncocytomas, and a Ct cutoff at 4.20 (Figs. 3 and 4), for instance, would clearly separate cases in these two categories. In the formalin-fixed group, all cases of the chromophobe RCC showed strong PVALB expression at levels similar to those seen in the fresh samples. However, the oncocytomas showed a broader range of PVALB expression in this group of 13 cases (median 3.302, range –2.814 to 11.018), significantly overlapping with the range seen in chromophobe RCC (–2.815 to 3.673).


View this table:
[in this window]
[in a new window]
  		Table 4. qRT-PCR on formalin-fixed, paraffin-embedded renal tumor specimens

 


View larger version (12K):
[in this window]
[in a new window]
  		Fig. 5. Distribution of gene expression ratios in the formalin-fixed, paraffin embedded tissue group. Findings were similar to those shown in Fig. 4, with the main difference being the overlapping ranges seen in the expression of PVALB in chromophobe RCC and oncocytoma.

 
Diagnostic algorithm for molecular classification of renal tumors. Based on these findings, {Delta}Ct (CA9 – AMACR), {Delta}Ct (CLCNKB – AMACR), and Ct (PVALB) were selected as predictors, and a diagnostic algorithm was generated using regression tree method and data from all 94 cases (31 fresh and 63 formalin-fixed). The algorithm was first built to classify the cases into clear cell, papillary, and chromophobe/oncocytoma groups based on {Delta}Ct (CA9 – AMACR) and {Delta}Ct (CLCNKB – AMACR). An attempt was then made to separate chromophobe RCC and oncocytoma based on the PVALB expression. The resulting regression tree and the decision rules were shown in Fig. 6.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 6. Diagnostic algorithm for renal tumor based on gene expression characteristics and regression tree analysis.

 
For classification into clear cell, papillary, and chromophobe RCC/oncocytoma groups, the decision rules had 100% classification accuracy in these 94 cases. Using grouped cross-validation (see Materials and Methods), we estimated that this model has 0.7% misclassification rate in the training samples and 3.4% misclassification rate in the validation samples, respectively. The estimated prediction accuracy of this model is thus 96%.

For the 31 fresh-frozen cases, the chromophobe RCC and oncocytoma could be accurately classified using a Ct (PVALB) cutoff of 4.20. Among the 13 oncocytomas in the formalin-fixed group, however, only 6 cases would have been correctly classified. The other 7 cases showed higher level of PVALB expression, indistinguishable from chromophobe RCC. Therefore, we conclude that only oncocytomas with low PVALB expression (Ct > 4.20) can be reliably classified as oncocytoma, whereas cases with high PVALB expression can be either chromophobe RCC or oncocytoma.

This algorithm was then used to evaluate the five unclassifiable RCC in our series, two in the fresh-frozen group and three in the formalin-fixed group. All five cases were found to fall into the chromophobe/oncocytoma category. Their PVALB Ct values were 6.757, 12.915, 1.604, 3.673, and 9.604, respectively. These values predicted that three of the five cases were oncocytoma, whereas two cases would be classified as chromophobe RCC versus oncocytoma, not further classifiable. These predictions were fully compatible with the histologic features of these cases.


   Discussion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
References
 
Several gene expression profiling techniques have been developed in the past decade, including Serial Analysis of Gene Expression (18), massively parallel signature sequencing (19), and DNA microarray (20). The DNA microarray has been most widely used, and one area that will clearly benefit from this type of analysis is the identification of new diagnostic and prognostic tumor markers. The value of this approach, first shown by studies of lymphoma and other hematopoietic malignancies (21, 22), is increasingly recognized in solid tumors. In RCC, several microarray studies have emerged, comparing gene expression profiles between different subtypes of renal cancer (11–15, 23) or between different stages of conventional clear cell carcinoma (24).

From these studies, clear cell carcinoma, papillary carcinoma, and chromophobe carcinoma were found to have discrete microarray patterns. On the other hand, chromophobe RCC has clustered with oncocytoma and cannot be distinguished (11, 13, 14). Moreover, the technical complexity and the cost of DNA microarray prevent its use as a routine diagnostic technique at present, and alternative methodologies to translate the microarray-derived knowledge to daily practice are needed. One strategy is to identify or produce antibodies for the gene product(s) of interest, thus converting the DNA/RNA–based analysis to an immunohistochemistry-based assay that can be routinely done. This was indeed attempted in RCC, and antibodies against DEFB1, PVALB, and vimentin were tested, showing differential staining patterns in different types of RCC (10). A second approach, shown in the present study, is to evaluate the mRNA expression of a few selected genes in tumors by qRT-PCR. This qRT-PCR–based approach has several advantages over the antibody approach. One major advantage is that no antibody is needed. As many of the target genes identified through microarray encoded proteins that had no commercially available antibodies, production of new antibodies would be required. These new antibodies will need to have high sensitivity and specificity on paraffin sections to be useful diagnostic reagents, and this goal is not easily achievable for most new antibodies. Second, unlike qRT-PCR, antibody staining is at best semiquantitative, likely unreliable to reflect the quantitative differences detected by the microarray. The expression level of AMACR, for example, is consistently higher in papillary RCC than in others, but only by several-fold by qRT-PCR. Not unexpectedly, a proportion of clear cell RCC showed positive staining with anti-racemase antibody P504s, although papillary RCC did show preferential staining (24).3 The last but probably the most important advantage of the qRT-PCR strategy is its capability to quantitatively evaluate the expression of more than one gene in a cumulative fashion as illustrated by the gene ratio approach in our study. As expression of multiple genes are often coordinated, the ability of qRT-PCR to evaluate the cumulative effects of multiple gene expression and compare them between specimens is a distinct advantage. This concept was the basis of the recently developed assay for predicting prognosis in breast cancer (25, 26) and in lymphoma (27) in which a panel of genes were analyzed by qRT-PCR, resulting in a calculated index that was used to predict the prognosis and/or likelihood of recurrence.

This qRT-PCR–based approach certainly has its disadvantages in comparison with immunohistochemical assays. It is still technically more demanding and cannot be readily done in current pathology laboratories. However, this will likely change in the near future. For example, user-friendly machines using qRT-PCR technique are being developed for the detection of cytokeratin-positive cells in the sentinel lymph node biopsy specimens of breast cancer (28) and these machines can be easily modified to accommodate the analysis described in our study. Another major disadvantage is the contamination of normal tissues in the RNA preparations. Because tissue microdissection is unlikely a feasible approach for a routine diagnostic test, we had taken two measures to circumvent this drawback: one was to obtain sections on which adjacent nonneoplastic tissue is either absent or could be easily removed manually and the other was to carefully choose the gene targets, limiting the analysis to genes that are expressed exclusively or predominantly in tumor tissues but not in stromal and/or adjacent normal tissues. The choice of CA9, AMACR, CLCNKB, and PVALB reflected such considerations, and the nonneoplastic cells did not seem to adversely affect the data obtained.

CA9 is only expressed in pancreaticobiliary epithelium normally, and no expression was found in normal kidney (29). In contrast, in malignant neoplasm, CA9 is a marker of tumor hypoxia (30), found to be expressed in carcinomas of the kidney, lung, colon, and head and neck squamous cell carcinoma (31–33). Among renal cancer, significant expression was found in the clear cell type (34, 35), and immunotherapeutic monoclonal antibody trials targeting CA9 are currently ongoing (36). Two antibodies against CA9 have been described [i.e., G250 (36) and M75 (37)]. Only M75 works on paraffin sections, but this antibody is not commercially available. Our current study confirmed high level CA9 expression in clear cell carcinoma, and a good commercial anti-CA9 antibody would be highly useful.

AMACR is a racemase that is normally expressed in several normal tissues, including liver, kidney, and salivary gland (38). Using anti-AMACR antibody P504s, renal tubular epithelial cells were found to express AMACR (38). For malignancies, AMACR is best recognized for its expression in prostate cancer (39). RCC and hepatocellular carcinoma have also been found to express AMACR (38). In kidney tumors, AMACR protein expression is most consistently found in papillary carcinoma but has also been observed in some cases of clear cell carcinoma (24), correlating with our observation of AMACR mRNA expression in these tumors.

PVALB, a high-affinity calcium binding protein, is expressed in several tissues, notably brain and kidney (40, 41). In kidney, the expression is limited to the distal nephron, particularly in the early distal convoluted tubules (42). Reflecting this tissue origin, PVALB mRNA expression has been found in chromophobe RCC and oncocytomas (10, 34, 43). It is of interest that we found chromophobe RCC to be more consistent in this expression, whereas some oncocytomas showed lower levels of PVALB. This quantitative difference may not be detectable by antibody staining. Rather, as suggested by Young et al. (10), positive PVALB immunostaining maybe a marker for chromophobe RCC and oncocytoma collectively, distinguishing them from clear cell and papillary RCC.

CLCNKB encodes one of two kidney-specific chloride channels (CLCNKA and CLCNKB), with expression localized to glomeruli, proximal tubules, and collecting ducts (44). Microarray data have suggested higher expression of this gene in chromophobe RCC/oncocytoma than in clear cell type, and this was confirmed by the present study. Given its preferential expression in selective kidney tumors, production of antibodies against this protein would be worthy of pursuing.

Based on the expression of these four genes, we were able to establish a molecular diagnostic model that is highly accurate in classifying renal cancer. As can be expected, chromophobe RCC and oncocytoma remain more difficult to separate. The ability of PVALB expression to predict at least some of the oncocytomas suggests the potential of this approach, and microarray studies to identify additional distinguishing markers between these two tumor types are ongoing.


   Footnotes
 
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.

3 Y.T. Chen, J.J. Tu, S.K. Tickoo, unpublished data. Back

Received 3/22/05; revised 6/ 9/05; accepted 6/21/05.


   References
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 References
 

  1. Mostfi FK. WHO international histological classification of tumours. Berlin: Springer; 1998.
  2. Montironi R, Mikuz G, Algaba F, Lopez-Beltran A, Hamilton PW, Parkinson C. Epithelial tumours of the adult kidney. Virchows Arch 1999;434:281–90.[Medline]
  3. Amin MB, Tamboli P, Javidan J, et al. Prognostic impact of histologic subtyping of adult renal epithelial neoplasms: an experience of 405 cases. Am J Surg Pathol 2002;26:281–91.[CrossRef][Medline]
  4. Thoenes W, Storkel S, Rumpelt HJ, Moll R, Baum HP, Werner S. Chromophobe cell renal carcinoma and its variants—a report on 32 cases. J Pathol 1988;155:277–87.[CrossRef][Medline]
  5. Amin MB, Corless CL, Renshaw AA, Tickoo SK, Kubus J, Schultz DS. Papillary (chromophil) renal cell carcinoma: histomorphologic characteristics and evaluation of conventional pathologic prognostic parameters in 62 cases. Am J Surg Pathol 1997;21:621–35.[CrossRef][Medline]
  6. Zambrano NR, Lubensky IA, Merino MJ, Linehan WM, Walther MM. Histopathology and molecular genetics of renal tumors toward unification of a classification system. J Urol 1999;162:1246–58.[CrossRef][Medline]
  7. Amin MB, Crotty TB, Tickoo SK, Farrow GM. Renal oncocytoma: a reappraisal of morphologic features with clinicopathologic findings in 80 cases. Am J Surg Pathol 1997;21:1–12.[CrossRef][Medline]
  8. Abrahams NA, MacLennan GT, Khoury JD, et al. Chromophobe renal cell carcinoma: a comparative study of histological, immunohistochemical and ultrastructural features using high throughput tissue microarray. Histopathology 2004;45:593–602.[CrossRef][Medline]
  9. Tickoo SK, Amin MB, Linden MD, Zarbo RJ. The MIB-1 tumor proliferation index in adult renal epithelial tumors with granular cytoplasm: biologic implications and differential diagnostic potential. Mod Pathol 1998;11:1115–21.[Medline]
  10. Young AN, de Oliveira Salles PG, Lim SD, et al. ß Defensin-1, parvalbumin, and vimentin: a panel of diagnostic immunohistochemical markers for renal tumors derived from gene expression profiling studies using cDNA microarrays. Am J Surg Pathol 2003;27:199–205.[CrossRef][Medline]
  11. Young AN, Amin MB, Moreno CS, et al. Expression profiling of renal epithelial neoplasms: a method for tumor classification and discovery of diagnostic molecular markers. Am J Pathol 2001;158:1639–51.[Abstract/Free Full Text]
  12. Skubitz KM, Skubitz AP. Differential gene expression in renal-cell cancer. J Lab Clin Med 2002;140:52–64.[CrossRef][Medline]
  13. Takahashi M, Yang XJ, Sugimura J, et al. Molecular subclassification of kidney tumors and the discovery of new diagnostic markers. Oncogene 2003;22:6810–8.[CrossRef][Medline]
  14. Higgins JP, Shinghal R, Gill H, et al. Gene expression patterns in renal cell carcinoma assessed by complementary DNA microarray. Am J Pathol 2003;162:925–32.[Abstract/Free Full Text]
  15. Liou LS, Shi T, Duan ZH, et al. Microarray gene expression profiling and analysis in renal cell carcinoma. BMC Urol 2004;4:9.[Medline]
  16. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and regression trees. Monterey: Wadsworth and Brooks; 1984.
  17. Venables WN, Ripley BD. Modern applied statistics with Splus. New York: Springer-Verlag; 1996.
  18. Velculescu VE, Vogelstein B, Kinzler KW. Analysing uncharted transcriptomes with SAGE. Trends Genet 2000;16:423–5.[CrossRef][Medline]
  19. Brenner S, Johnson M, Bridgham J, et al. Gene expression analysis by massively parallel signature sequencing (MPSS) on microbead arrays. Nat Biotechnol 2000;18:630–4.[CrossRef][Medline]
  20. Eisen MB, Brown PO. DNA arrays for analysis of gene expression. Methods Enzymol 1999;303:179–205.[Medline]
  21. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403:503–11.[CrossRef][Medline]
  22. Staudt LM. Molecular diagnosis of the hematologic cancers. N Engl J Med 2003;348:1777–85.[Free Full Text]
  23. Takahashi M, Sugimura J, Yang X, et al. Gene expression profiling of renal cell carcinoma and its implications in diagnosis, prognosis, and therapeutics. Adv Cancer Res 2003;89:157–81.[Medline]
  24. Boer JM, Huber WK, Sultmann H, et al. Identification and classification of differentially expressed genes in renal cell carcinoma by expression profiling on a global human 31,500-element cDNA array. Genome Res 2001;11:1861–70.[Abstract/Free Full Text]
  25. van de Vijver MJ, He YD, van't Veer LJ, et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347:1999–2009.[Abstract/Free Full Text]
  26. Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 2004;351:2817–26.[Abstract/Free Full Text]
  27. Rosenwald A, Wright G, Chan WC, et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 2002;346:1937–47.[Abstract/Free Full Text]
  28. Green G, Xu M, Major C, et al. Preclinical evaluation of an intraoperative molecular assay for the detection of metastasis in breast sentinel lymph nodes. J Mol Diag 2004;6:429.
  29. Kivela AJ, Parkkila S, Saarnio J, et al. Expression of transmembrane carbonic anhydrase isoenzymes IX and XII in normal human pancreas and pancreatic tumours. Histochem Cell Biol 2000;114:197–204.[Medline]
  30. Potter C, Harris AL. Hypoxia inducible carbonic anhydrase IX, marker of tumour hypoxia, survival pathway and therapy target. Cell Cycle 2004;3:164–7.[Medline]
  31. Miyaji E, Nishimori I, Taniuchi K, Takeuchi T, Ohtsuki Y, Onishi S. Overexpression of carbonic anhydrase-related protein VIII in human colorectal cancer. J Pathol 2003;201:37–45.[Medline]
  32. Kim SJ, Rabbani ZN, Vollmer RT, et al. Carbonic anhydrase IX in early-stage non-small cell lung cancer. Clin Cancer Res 2004;10:7925–33.[Abstract/Free Full Text]
  33. Beasley NJ, Wykoff CC, Watson PH, et al. Carbonic anhydrase IX, an endogenous hypoxia marker, expression in head and neck squamous cell carcinoma and its relationship to hypoxia, necrosis, and microvessel density. Cancer Res 2001;61:5262–7.[Abstract/Free Full Text]
  34. Li G, Cuilleron M, Gentil-Perret A, et al. Rapid and sensitive detection of messenger RNA expression for molecular differential diagnosis of renal cell carcinoma. Clin Cancer Res 2003;9:6441–6.[Abstract/Free Full Text]
  35. Bismar TA, Bianco FJ, Zhang H, et al. Quantification of G250 mRNA expression in renal epithelial neoplasms by real-time reverse transcription-PCR of dissected tissue from paraffin sections. Pathology 2003;35:513–7.[Medline]
  36. Oosterwijk E, Divgi CR, Brouwers A, et al. Monoclonal antibody-based therapy for renal cell carcinoma. Urol Clin North Am 2003;30:623–31.[Medline]
  37. Pastorekova S, Zavadova Z, Kostal M, Babusikova O, Zavada J. A novel quasi-viral agent, MaTu, is a two-component system. Virology 1992;187:620–6.[CrossRef][Medline]
  38. Jiang Z, Fanger GR, Woda BA, et al. Expression of {alpha}-methylacyl-CoA racemase (P504s) in various malignant neoplasms and normal tissues: a study of 761 cases. Hum Pathol 2003;34:792–6.[Medline]
  39. Jiang Z, Woda BA, Wu CL, Yang XJ. Discovery and clinical application of a novel prostate cancer marker: {alpha}-methylacyl CoA racemase (P504S). Am J Clin Pathol 2004;122:275–89.[CrossRef][Medline]
  40. Campean V, Kricke J, Ellison D, Luft FC, Bachmann S. Localization of thiazide-sensitive Na(+)-Cl(–) cotransport and associated gene products in mouse DCT. Am J Physiol Renal Physiol 2001;281:F1028–35.[Abstract/Free Full Text]
  41. Cooper EC, Harrington E, Jan YN, Jan LY. M channel KCNQ2 subunits are localized to key sites for control of neuronal network oscillations and synchronization in mouse brain. J Neurosci 2001;21:9529–40.[Abstract/Free Full Text]
  42. Loffing J, Loffing-Cueni D, Valderrabano V, et al. Distribution of transcellular calcium and sodium transport pathways along mouse distal nephron. Am J Physiol Renal Physiol 2001;281:F1021–7.[Abstract/Free Full Text]
  43. Martignoni G, Pea M, Chilosi M, et al. Parvalbumin is constantly expressed in chromophobe renal carcinoma. Mod Pathol 2001;14:760–7.[CrossRef][Medline]
  44. Takeuchi Y, Uchida S, Marumo F, Sasaki S. Cloning, tissue distribution, and intrarenal localization of ClC chloride channels in human kidney. Kidney Int 1995;48:1497–503.[Medline]



This article has been cited by other articles:


Home page
 Clin. Cancer Res.Home page
S. Rohan, J. J. Tu, J. Kao, P. Mukherjee, F. Campagne, X. K. Zhou, E. Hyjek, M. A. Alonso, and Y.-T. Chen
Gene Expression Profiling Separates Chromophobe Renal Cell Carcinoma from Oncocytoma and Identifies Vesicular Transport and Cell Junction Proteins as Differentially Expressed Genes
Clin. Cancer Res., December 1, 2006; 12(23): 6937 - 6945.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
N. J. Camp, J. M. Farnham, and L. A. Cannon-Albright
Localization of a Prostate Cancer Predisposition Gene to an 880-kb Region on Chromosome 22q12.3 in Utah High-Risk Pedigrees.
Cancer Res., October 15, 2006; 66(20): 10205 - 10212.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chen, Y.-T.
Right arrow Articles by Mazumdar, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chen, Y.-T.
Right arrow Articles by Mazumdar, M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Cell Growth & Differentiation