| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Regular Articles |
From the Departments of Pathology and Laboratory
Medicine,*
Urology,
and
Biochemistry,
Emory University School
of Medicine, Atlanta; and the Atlanta VA Medical
Center,
Decatur, Georgia
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Oncocytoma and chromophobe RCC are renal epithelial neoplasms that exhibit antigenic characteristics of distal nephron intercalated cells.11-13 Oncocytomas exhibit circumscribed growth of characteristic neoplastic cells (oncocytes), which contain small round nuclei and eosinophilic cytoplasm packed with mitochondria.14 Chromophobe RCCs span a spectrum of morphology.15,16 Some tumors exhibit nested architecture and cells with abundant, eosinophilic granular cytoplasm (eosinophilic variety), whereas others are composed of cells containing clear cytoplasm filled with vesicles that stain with Hale’s colloidal iron (typical variety). There is evidence that the Hale’s colloidal iron-positive vesicles actually represent abortive mitochondria.17 Histologically, the eosinophilic variety of chromophobe RCC can be difficult to distinguish from renal oncocytoma, whereas the typical variety can resemble conventional RCC. The different clinical behaviors of oncocytomas and chromophobe RCCs (the former are invariably benign; the latter are indolent yet malignant) warrant their diagnostic separation. However, overlap in the morphological, immunohistochemical, ultrastructural, and cytogenetic features of these two tumor types suggest that they are closely related.17-19 Cytogenetically, for example, the loss of chromosome 1 is a common finding in both tumor types,2,20,21 with oncocytomas often exhibiting loss of chromosomes 1 and Y, and chromophobe RCCs often exhibiting concurrent loss of multiple chromosomes, including chromosome 1.22,23 Translocations involving chromosome 11q13 may define a distinct subset of oncocytomas.24 Mutations of mitochondrial DNA have also been described in both chromophobe RCCs and oncocytomas.25 However, it is still unclear if these mutations are consistent findings in either lesion, or if they are directly related to the pathogenesis or mitochondrion-rich morphology of these tumors.
The underlying pathogenetic mechanisms of renal neoplasms, and of neoplasms in general, remain a mystery, but new insights into the pathobiology of neoplasia are emerging as technical advances permit large-scale, parallel analysis of eukaryotic gene expression.26 Using cDNA microarrays to analyze total cellular mRNA, one can compare the relative expression levels of several thousand genes in different cell types simultaneously. Several groups have used such expression profiling methods to identify gene expression patterns associated with various tumors and other disease states.27-30 Although this field is still relatively new, it has already succeeded in associating specific gene expression patterns with either neoplastic or non-neoplastic cell types,27-30 allowing researchers to postulate novel gene regulatory circuits and correlate the expression of previously unsuspected genes with specific cell phenotypes. In this sense, gene expression profiling has proved to be an extremely powerful, high-throughput method for identifying specific molecular markers of disease.31,32 In this report, we have applied gene expression profiling to a series of renal epithelial tumors including conventional RCC, chromophobe RCC, and oncocytoma. Based solely on patterns of gene expression, the renal neoplasms were clustered into subtypes consistent with the clinical, histological, and molecular understanding of these tumors. For several individual genes—vimentin, class II major histocompatibility complex (MHC)-associated invariant chain (CD74), parvalbumin, and galectin-3—differential expression patterns identified by cDNA microarrays were validated in a larger series of tumors by immunohistochemistry, thus confirming these gene products as promising pathological markers for the differential diagnosis of renal epithelial neoplasia.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
For microarray experiments, matched specimens of renal tumor and grossly non-neoplastic kidney from the same patient (100–200 mg per specimen) were obtained from the tissue bank maintained by Dr. Fray Marshall. All specimens were promptly frozen and stored at -80°C. Histopathological diagnoses were rendered by the Johns Hopkins University Department of Pathology (Baltimore, MD). The tumors consisted of four cases of conventional (clear cell) RCC (two Fuhrman grade II, one Fuhrman grade III, and one Fuhrman grade IV), one case of chromophobe RCC, and two cases of renal oncocytoma. The conventional RCC patients were a 62-year-old female, a 47-year-old male, a 67-year-old female, and a 57-year-old male; the chromophobe RCC patient was a 73-year-old male; the oncocytoma patients were a 33-year-old female and 72-year-old male. The patients were not followed clinically in this study. For immunohistochemistry, representative tissue blocks were obtained from the Emory University Department of Pathology. The tissues were derived from radical nephrectomies performed at Emory University, and consisted of 20 conventional RCCs (10 Fuhrman grades I-II and 10 Fuhrman grades III-IV), 6 chromophobe RCCs, and 8 oncocytomas. All tissues were fixed in 10% neutral buffered formalin and embedded in paraffin using standard surgical pathology protocols.
Microarray Analysis
Total cellular RNA was prepared from frozen specimens by
mechanical disruption in TRIzol reagent (Gibco BRL, Gaithersburg MD),
followed by chloroform extraction and alcohol precipitation according
to the manufacturer’s instructions. PolyA+ RNA was isolated with
Oligotex oligo-dT beads (Qiagen, Valencia, CA) according to
manufacturer’s instructions. PolyA+ RNA (500 ng) was shipped frozen to
Incyte Genomics (Palo Alto, CA) for labeling and hybridization using
proprietary methods described in detail on the company’s Internet
site, http://www.incyte.com/gem/technology/index.shtml. For each case,
matched polyA+ RNA samples from tumor and non-neoplastic kidney from
the same patient were reverse-transcribed into cDNA, incorporating
deoxynucleotides coupled to distinct fluorescent dyes; cDNAs derived
from non-neoplastic kidneys were labeled with the green dye Cy3, and
cDNAs derived from tumors were labeled with the red dye Cy5.
Differentially labeled cDNAs from matched tumors and controls were
pooled and hybridized simultaneously to Incyte UniGEM v.1 microarrays
containing single-stranded cDNA molecules covalently bound to modified
glass substrates. The UniGEM v.1 microarrays featured targets for 7075
unique human genes, spotted at known positions on the array grids, as
well several proprietary non-human gene targets that served as controls
for reverse transcription and hybridization efficiency. The arrays were
washed after hybridization and scanned by a specialized fluorescent
confocal microscope to detect bound, Cy3-labeled, and Cy5-labeled
cDNAs. Fluorescence intensities at each target position on the array
were balanced to the intensities of internal control targets, for which
known amounts of cognate mRNAs were added to the reverse transcription
reactions. The ratios of balanced Cy5/Cy3 fluorescence intensities at
each target represented the ratios of specific gene expression in the
tumor versus the uninvolved kidney. Reproducibility data
generated by Incyte, available on the company’s Internet site,
indicated that the sensitivity of mRNA detection with UniGEM
microarrays is 2 pg, the dynamic range for mRNA detection is 2–2000
pg, the level of detectable differential expression is 
1.8-fold, and
the average coefficient of variation for Cy5/Cy3 ratios is 15%.
Informatics
Differential expression data were analyzed using Incyte’s
proprietary GemTools software. We used a minimum absolute fluorescence
intensity cutoff of 700 units in either the Cy3 or Cy5 channel, in at
least two hybridization experiments, to select 4906 expressed genes
representing 69% of the array targets. The use of this absolute
fluorescence cutoff was determined through personal communications with
Incyte. Subsequent data analysis was restricted to genes overexpressed
or underexpressed 1.8- to 2.0-fold in tumors relative to matched
non-neoplastic kidneys. Differential expression profiles were analyzed
using the hierarchical average linkage clustering algorithm supplied
with Cluster33
software (Michael Eisen, Stanford
University, Stanford, CA). This algorithm used an iterated,
agglomerative process of similarity measurements based on the Pearson
correlation. In each iteration of the algorithm, the two most similar
data elements (ie, expression profiles) were joined by a node of
a dendrogram, after which the joined elements were averaged and
replaced by a pseudo-element to be used in all subsequent iterations.
Hierarchically clustered gene expression and tumor data were analyzed
graphically using the TreeView33
program bundled with
Cluster software.
Differential expression profiles were also clustered non-hierarchically using the Quality Threshold (QT) clustering algorithm as originally described.34 The advantage of this non-agglomerative approach over hierarchical clustering was that all of the data were compared without the generation of pseudo-elements. In addition, QT clustering is not designed to separate the data into a predetermined, user-defined number of clusters, as would occur with other commonly used clustering algorithms such as self-organizing maps or K-means clustering. Instead, the user input for QT clustering is limited to the QT, which represents how highly correlated each of the members of a given cluster must be. User input QT can range from -1 (completely inversely correlated) to 1 (perfectly correlated). In general, the QT value needed for significance is inversely related to the number of elements to be clustered. For example, a QT of 0.1 to 0.2 may be adequate to cluster tumors reliably using 1000 genes, whereas a higher value of 0.3 to 0.4 may be required to cluster tumors using 100 genes. On the other hand, a QT of 0.6 to 0.7 is likely to be necessary to cluster genes using only 10 or fewer tumors. At QT values higher than 0.7, elements tend not to be clustered even though they are significantly correlated.
For all analyses using the hierarchical and QT algorithms, differential expression values were transformed to log2 before clustering, so that overexpressed and underexpressed genes would have values of opposite sign. In addition, for analyses using the QT algorithm, log2-transformed values were normalized so that every gene had a mean differential expression of 0 and a variance of 1 across the seven experiments.34 Thus, the expression patterns of individual genes were relatively independent of absolute differential expression levels.
Immunohistochemistry
Representative formalin-fixed, paraffin-embedded tissue sections
were dewaxed and subjected to antigen retrieval in citrate buffer, pH
6, using an electric pressure cooker set at 120°C for 5
minutes.35
Sections were incubated for 5 minutes in 3%
hydrogen peroxide to quench endogenous tissue peroxidase.
Immunohistochemistry was performed using primary antibodies directed
against vimentin (mouse monoclonal M0725, 1:80 dilution: DAKO Corp.,
Carpinteria, CA), CD74 (mouse monoclonal LN2, 1:8 dilution; ICN
Biomedicals, Costa Mesa, CA), parvalbumin (goat polyclonal Sc7447, 1:40
dilution; Santa Cruz Biotechnology, Santa Cruz, CA), and galectin-3
(mouse monoclonal GALECT3abm, 1:100 dilution; Research Diagnostics,
Inc., Flanders, NJ). After 25-minute incubations with appropriate
primary antibody, sections were washed and treated with commercial
biotinylated secondary anti-immunoglobulin, followed by avidin coupled
to biotinylated horseradish peroxidase, according to manufacturer’s
instructions (LSAB2 kit for mouse primary antibodies and LSAB+ kit for
goat primary antibody, DAKO). The immunohistochemical reactions were
visualized using diaminobenzidine as the chromogenic peroxidase
substrate. Sections were counterstained with hematoxylin after
immunohistochemistry. Strong positive immunohistochemical staining was
defined as 3+ intensity in at least 30% of tumor cells. Specificity
of the procedure was verified by negative control reactions without
primary antibody and by appropriate staining of positive control
tissues.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
1.8-fold overexpression or underexpression,
based on Incyte’s internal quality control data (see Materials and
Methods). By this definition, 8% of the 4906 detected genes (385
genes) were overexpressed or underexpressed in at least two tumors.
Subsequently, we increased the differential expression cutoff to
2.0-fold to reduce the chance of false positive signals. At this
higher stringency, 189 (4%) of the detected genes were differentially
expressed in at least two tumors. Most of these 189 genes could be
grouped into functional categories such as cell growth and
differentiation, cell adhesion, immune regulation, energy metabolism,
cytoskeleton, vascular biology, and extracellular matrix, whereas 28
sequences corresponded to uncharacterized expressed sequence tags. A
complete listing of the 189 differentially expressed genes, including
microarray fluorescent expression data and differential expression
values in each tumor, is presented as Supplemental Data on our
departmental Internet site www.amjpathol.org.Inspection of this
microarray data revealed that several genes were expressed
preferentially in specific renal tumor subtypes, as listed in Tables 1 and 2
and
discussed further below.
|
|
; note the dendrogram on the
x axis). The same clustering of tumors was obtained when we
used a cutoff of 1.8-fold to define differential expression or when we
eliminated the differential expression cutoff and included all 4906
detected genes in the analysis (data not shown). The chromophobe RCC
and oncocytomas expressed similar patterns of genes, as shown by their
being grouped together in these clustering experiments (Figure 1)
. Of
particular note, the chromophobe RCC/oncocytomas overexpressed genes
coding for ß-defensin-1 (an antibacterial peptide) and parvalbumin (a
calcium-binding protein), the products of which have been localized in
previous studies37,38
to mammalian distal nephron
epithelium (Table 2
and Supplemental Data). In addition, the
chromophobe RCC/oncocytomas overexpressed many sequences related to
mitochondrial biogenesis and oxidative phosphorylation, a finding
consistent with the large numbers of mitochondria per cell in these
particular tumors. Included among these sequences were genes encoding
mitochondrial adenine nucleotide translocator isoforms,
proton-transporting ATP synthase subunits, carbonic anhydrase
isoenzymes, and cytochrome c oxidase subunits (Table 2
and Supplemental Data).
|
. Overall, the two largest QT clusters produced a set of 65
genes that were highly discriminatory between conventional RCCs and
chromophobe RCC/oncocytomas. To demonstrate this point, the 65 genes
were reanalyzed with the hierarchical average linkage clustering
algorithm (shown graphically in Figure 2C
). Comparison of the tumor
dendrograms in Figures 1 and 2C
, in which the relative height of each
node is proportional to 1 minus the Pearson correlation coefficient
R,33
shows that analysis of the 65 genes
identified by the QT algorithm produced a greater distinction between
the two renal tumor categories than analysis of the total group of 189
differentially expressed genes. Thus, the subset of 65 differentially
expressed genes may be particularly useful for classification of renal
epithelial neoplasms, eg, by development of customized lower-density
cDNA microarrays.
|
, y axis). For example, several class
II MHC-related genes, including HLA-DR
, HLA-DRß1, and class
II MHC-associated invariant chain (CD74), were placed into adjacent
nodes by the hierarchical clustering algorithm (Figure 2C
, highlighted
with a blue bar), based on strikingly similar expression patterns with
relative overexpression in the conventional RCCs. Similarly, the genes
encoding two interferon-induced transmembrane proteins were coclustered
on the basis of similar differential expression profiles across the
seven tumors, again with relative overexpression in the conventional
RCCs (Figure 2C
, highlighted with a black bar). In certain cases, the
clustering of genes was predictive of function. An expressed sequence
tag corresponding to GenBank accession number AI311009 was clustered
into a highly correlated node with chloride channel Kb (Figure 2C
,
highlighted with an orange bar). Sequence comparison using the BLAST
algorithm39
showed that these two sequences are highly
homologous.
The protein products of genes overexpressed in conventional RCCs or
chromophobe RCC/oncocytomas, but not in both tumor categories,
represented a set of potential immunohistochemical markers for renal
tumor diagnosis. As described above, the QT algorithm revealed two
large clusters of genes that were overexpressed in only one of the two
tumor categories. The products of several of these genes, including
vimentin, class II MHC-associated invariant chain (CD74), and
galectin-3, have been identified previously in neoplastic and
non-neoplastic renal tissue by immunohistochemistry, making these
antigens candidate pathological markers.31,40,41
Because
of the small number of tumor samples analyzed by microarray, the
variability introduced by a single outlier value for parvalbumin
prevented this gene from being included in the largest QT cluster, even
though it was overexpressed in chromophobe RCC/oncocytomas and
underexpressed in conventional RCCs, similar to most members of that
gene cluster (Table 2
and Supplemental Data). Parvalbumin has been
localized to distal nephron epithelium by
immunohistochemistry,38
making it an interesting candidate
marker for chromophobe RCC/oncocytomas, which appear to be
antigenically related to distal nephron intercalated
cells.11-13
To evaluate the diagnostic utility of
vimentin, CD74, parvalbumin, and galectin-3, as well as to test the
validity of the microarray data, we measured the expression of these
proteins in an independent series of 34 renal tumors by
immunohistochemistry. In agreement with the microarray data, which
showed vimentin mRNA overexpression to be specific to conventional
tumors, vimentin protein was detected in the tumor cells of 17/20
(85%) conventional RCCs versus 0/6 chromophobe RCCs and 0/8
oncocytomas (Figure 3, A–C
, and Table 3
; P 
0.001). Also
consistent with the microarray data, CD74 was detected
immunohistochemically in the tumor cells of 13/20 (65%) conventional
carcinomas versus 0/8 oncocytomas. Interestingly, 4/6 (67%)
chromophobe RCCs stained positively for CD74, a finding not seen in the
single chromophobe tumor analyzed by microarray. These data suggest
that CD74 might be a useful marker in certain cases to discriminate
chromophobe tumors from oncocytomas (Figure 3, D–F
, and Table 3
;
P 0.01). The immunohistochemical detection of
parvalbumin correlated very closely with the microarray results, with
strong tumor cell expression in 6/6 chromophobe RCCs and 8/8
oncocytomas versus only 5/20 (20%) conventional RCCs
(Figure 4, A–C
, and Table 3
;
P 0.001). In the tissue sections, non-neoplastic
kidney showed high background staining, probably related to high
endogenous peroxidase content, which prevented us from confirming if
parvalbumin was indeed expressed selectively by distal nephron
epithelium. Also in agreement with the microarray data, galectin-3 was
consistently detected by immunohistochemistry in chromophobe RCCs (5/6
tumors strongly positive) and oncocytomas (8/8 tumors strongly
positive). Data were discordant, however, for galectin-3 expression in
conventional RCCs. Whereas 0/4 conventional tumors overexpressed
galectin-3 mRNA by microarray, 13/20 (65%) expressed the protein at
high levels by immunohistochemistry (Figure 4, D–F
, and Table 3
).
Interestingly, immunohistochemical detection of galectin-3 in
conventional tumors was correlated with low histological grade (9/10
low-grade tumors versus 4/10 high-grade tumors;
P 0.025).
|
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
. Thus, certain aspects of epidermal growth
factor signaling may be important for the biology of distal nephron
epithelium and/or distal nephron-related epithelial tumors. The
conventional RCCs overexpressed several class II MHC-related genes, as
discussed in greater detail below. In addition, these tumors were
characterized by specific overexpression of several genes related to
vascular biology, including vascular cell adhesion molecule 1, vascular
endothelial growth factor-related protein, von Willebrand factor, and
angiopoietin 2 (Table 1
and Supplemental Data). This finding may be
related to the anastomosing vascular network that distinguishes
conventional RCC from other renal tumors2,42
and may have
implications for anti-angiogenic therapies directed against this tumor
subtype.10
Finally, the conventional RCCs overexpressed at
least three genes on chromosome 5q (CD74, transforming growth
factor-ß, and lysyl oxidase; Table 1
and Supplemental Data),
consistent with previous comparative genomic hybridization data showing
that DNA gains in this tumor type most often involve chromosome
5q.43
Examination of Figure 1
(y axis) shows that several
genes were either overexpressed or underexpressed in the all of the
tumor subtypes examined in this study. This gene group included
ribosomal proteins, HLA-B and insulin-like growth factor binding
protein 3 (overexpressed in conventional RCCs, chromophobe RCC, and
oncocytomas), and c-fos, cathepsin H, and uromodulin
(underexpressed in conventional RCCs, chromophobe RCC, and
oncocytomas). Specific changes in the expression of these genes may be
common in the neoplastic transformation of renal epithelial cells.
Indeed, each of these gene products has been associated with processes
relevant to tumor development and spread.40,44-48
Because our microarray study was based on a limited number of grossly dissected renal tumor samples consisting of heterogeneous cell populations, two obvious concerns were (i) whether the microarray data reflected differential gene expression specific to tumor cells and (ii) if so, whether the data could be generalized to conventional RCCs and chromophobe RCC/oncocytomas as a whole. To try to address these concerns, we validated microarray data for several genes with the cell-specific technique of immunohistochemistry in a larger tumor series. For each of the antigens tested—vimentin, CD74, parvalbumin, and galectin-3—the immunohistochemical signals were detected mainly in neoplastic epithelium, suggesting that the microarray data primarily reflected differential gene expression in tumor cells. Vimentin and CD74 antigen expression was also detected in tumor-associated stroma and vasculature, although these non-neoplastic cells typically comprised a small portion of the total renal tumor cellularity.
Overall, we obtained strong correlations between microarrays and immunohistochemistry for differential gene/protein expression in renal tumor subtypes, although the correlations were not perfect. Most published expression-profiling studies have described good but imperfect correlations between the data from microarrays and gene-specific validation assays.49 This inability to confirm every aspect of microarray data is not surprising, given the number of genes interrogated by typical high-density microarrays and the fact that most array experiments (including ours) have been unable to perform extensive data replication due to limitations in sample size. A recent, comprehensive statistical analysis of microarray experiments has shown that non-replicated expression data are prone to numerous misclassifications, particularly false positive results.50 Thus, appropriate data validation, either by microarray replication or an independent technique such as immunohistochemistry, is clearly essential for accurate interpretation of gene expression profiles.
The cDNA microarray and immunohistochemical experiments in our study both showed vimentin to be a sensitive and specific marker for conventional RCC. In agreement with our results, an independent microarray analysis that compared gene expression between a renal cancer cell line and benign kidney tissue also singled out vimentin as a useful diagnostic and prognostic marker for RCC.31 We also obtained excellent concordance between microarray and immunohistochemical data for the calcium-binding protein parvalbumin, which emerged as a promising marker for chromophobe RCC/oncocytomas. Thus, our study supports the idea that expression profiling of tumors is a potentially powerful method for identifying new pathological markers for tumor diagnosis.31,32
Our data validations were less exact for galectin-3 and CD74. For
example, although the microarrays suggested that galectin-3 mRNA
overexpression was specific to chromophobe RCC/oncocytomas, the
corresponding protein was detected by immunohistochemistry in several
conventional tumors as well. It is possible that galectin-3 mRNA and
protein levels do not correlate precisely in certain tissues. Although
our findings may exclude galectin-3 as a useful marker for
differentiating renal tumor subtypes, the expression of this gene
product appeared to correlate with tumor indolence, being expressed
predominantly in low-grade conventional RCCs, indolent chromophobe
RCCs, and benign oncocytomas. This finding is consistent with several
studies showing reduced galectin-3 expression in clinically aggressive
tumors and may be relevant to the function of galectin-3 as an adhesion
molecule that inhibits metastasis.41,51-53
Interestingly,
our microarray data suggested that the related adhesion molecule
galectin-1 was also expressed differentially in the renal tumor
subtypes, albeit with relative overexpression in conventional tumors
(Table 1
and Supplemental Data). Though we did not confirm galectin-1
data immunohistochemically, other studies have shown that expression of
this lectin may be prognostically or diagnostically relevant to tumor
biology, either alone or in combination with galectin-3.41
Taken together, the microarray and immunohistochemical experiments
suggested that class II MHC-associated invariant chain (CD74) was
expressed in conventional and chromophobe RCCs but not in oncocytomas.
Differential CD74 expression in chromophobe RCCs and oncocytomas is a
finding of potential importance for surgical pathology, since reliable
immunomarkers are not available to distinguish these histologically
related lesions and the benign nature of oncocytomas, compared with the
potential of chromophobe RCCs for metastasis and sarcomatoid
transformation,54,55
makes this an important differential
diagnosis. Intriguing therapeutic implications are also raised by the
expression of class II MHC-related genes in RCCs, given the
responsiveness of many cases to interleukin-2 and/or
interferon-.56
Our results are in general agreement
with a recent report showing class II MHC-associated invariant chain
expression by immunohistochemistry in 53/60 renal
carcinomas.40
In that study, CD74 expression was
correlated with lymphocytic infiltration of tumor, and the authors
speculated that class II MHC-related gene expression might be relevant
to the overall responsiveness of RCC to immunotherapy. Future
microarray studies of primary and metastatic RCC should help determine
whether overall expression patterns of class II MHC-related molecules,
including CD74, are predictive of immunotherapeutic response.
At the time this project began, Incyte UniGEM v.1 microarrays were the most complete cDNA arrays available for gene expression profiling. Their cDNA targets were chosen from expression libraries of all major organ groups and represented genes involved in many biological activities such as cell growth and development, cytoskeletal structure, cell motility, molecular recognition, membrane transport, protein and nucleic acid biosynthesis, and energy metabolism. Despite this, the UniGEM microarrays interrogated a small fraction of the total expected genome.36 Denser arrays are becoming available, both commercially and from individual laboratories, as genome projects and microarray fabrication technologies continue to progress. Thus, expression profiling of the entire genome is likely to be possible in the near future. However, until this becomes a reality, large-scale expression profiling studies will suffer from the somewhat ironic problem of enormous, yet incomplete, data sets. Given these limitations, we expect that we have identified only a fraction of the genes that are relevant to the pathobiology of renal epithelial neoplasms, which might explain why relatively few growth regulatory genes were overexpressed or underexpressed consistently in the tumor subtypes we studied. Even though the UniGEM arrays contained more than 600 genes related to cell growth and development, this still did not represent a complete survey. Notably, for example, the von Hippel-Lindau tumor suppressor gene was not included on the microarrays.
In conclusion, we studied the gene expression profiles of conventional RCCs, chromophobe RCC, and oncocytomas and separated the tumors into reproducible gene expression classes that correlated with histopathological diagnoses. Several functionally related gene clusters were informative for distinguishing conventional RCCs from chromophobe RCC/oncocytomas (eg, class II MHC-related and vascular genes in conventional RCCs; distal nephron and oxidative phosphorylation genes in chromophobe RCC/oncocytomas). These gene clusters offered insights into tumor pathobiology and histogenesis, and highlighted several possible gene regulatory networks that could be important in renal epithelial neoplasia. Several individual genes showed potential utility as immunohistochemical markers for the pathological diagnosis of renal tumor subtypes. Based on these results, we are encouraged to expand our expression profiling studies, using a larger number of specimens that include not only conventional RCCs and chromophobe RCC/oncocytomas, but also other renal tumor subtypes, such as papillary RCCs. We predict that these gene expression profiling experiments will lead to improvements in the basic understanding of renal tumor pathogenesis and will promote the discovery of novel molecular markers for renal tumor diagnosis.
|
Footnotes |
|---|
Supported by the Molecular Based Testing Initiative, Emory University Department of Pathology and Laboratory Medicine.
Accepted for publication February 9, 2001.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  Z. S. Morris and A. I. McClatchey Aberrant epithelial morphology and persistent epidermal growth factor receptor signaling in a mouse model of renal carcinoma PNAS, June 16, 2009; 106(24): 9767 - 9772. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. M. Hegedus, C. F. Skibola, M. Warner, D. R. Skibola, D. Alexander, S. Lim, N. L. Dangleben, L. Zhang, M. Clark, R. M. Pfeiffer, et al. Decreased Urinary Beta-Defensin-1 Expression as a Biomarker of Response to Arsenic Toxicol. Sci., November 1, 2008; 106(1): 74 - 82. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y.-H. Liu, C.-Y. Lin, W.-C. Lin, S.-W. Tang, M.-K. Lai, and J.-Y. Lin Up-Regulation of Vascular Endothelial Growth Factor-D Expression in Clear Cell Renal Cell Carcinoma by CD74: A Critical Role in Cancer Cell Tumorigenesis J. Immunol., November 1, 2008; 181(9): 6584 - 6594. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Gore, D. Starlets, N. Maharshak, S. Becker-Herman, U. Kaneyuki, L. Leng, R. Bucala, and I. Shachar Macrophage Migration Inhibitory Factor Induces B Cell Survival by Activation of a CD74-CD44 Receptor Complex J. Biol. Chem., February 1, 2008; 283(5): 2784 - 2792. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 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]
|
|
|
|
 |
|
 K. Hotakainen, S. Lintula, B. Ljungberg, P. Finne, A. Paju, U.-H. Stenman, and J. Stenman Expression Of Human Chorionic Gonadotropin {beta}-Subunit Type I Genes Predicts Adverse Outcome In Renal Cell Carcinoma J. Mol. Diagn., November 1, 2006; 8(5): 598 - 603. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Starlets, Y. Gore, I. Binsky, M. Haran, N. Harpaz, L. Shvidel, S. Becker-Herman, A. Berrebi, and I. Shachar Cell-surface CD74 initiates a signaling cascade leading to cell proliferation and survival Blood, June 15, 2006; 107(12): 4807 - 4816. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. D. Raman, N. P. Mongan, L. Liu, S. K. Tickoo, D. M. Nanus, D. S. Scherr, and L. J. Gudas Decreased expression of the human stem cell marker, Rex-1 (zfp-42), in renal cell carcinoma Carcinogenesis, March 1, 2006; 27(3): 499 - 507. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y.-T. Chen, J. J. Tu, J. Kao, X. K. Zhou, and M. Mazumdar Messenger RNA Expression Ratios among Four Genes Predict Subtypes of Renal Cell Carcinoma and Distinguish Oncocytoma from Carcinoma Clin. Cancer Res., September 15, 2005; 11(18): 6558 - 6566. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 X. J. Yang, M.-H. Tan, H. L. Kim, J. A. Ditlev, M. W. Betten, C. E. Png, E. J. Kort, K. Futami, K. A. Furge, M. Takahashi, et al. A Molecular Classification of Papillary Renal Cell Carcinoma Cancer Res., July 1, 2005; 65(13): 5628 - 5637. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Togashi, T. Katagiri, S. Ashida, T. Fujioka, O. Maruyama, Y. Wakumoto, Y. Sakamoto, M. Fujime, Y. Kawachi, T. Shuin, et al. Hypoxia-Inducible Protein 2 (HIG2), a Novel Diagnostic Marker for Renal Cell Carcinoma and Potential Target for Molecular Therapy Cancer Res., June 1, 2005; 65(11): 4817 - 4826. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. N. Schuetz, Q. Yin-Goen, M. B. Amin, C. S. Moreno, C. Cohen, C. D. Hornsby, W. L. Yang, J. A. Petros, M. M. Issa, J. G. Pattaras, et al. Molecular Classification of Renal Tumors by Gene Expression Profiling J. Mol. Diagn., May 1, 2005; 7(2): 206 - 218. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Wang, R. Lichtenfels, J. Bukur, Y. Ngalame, M. C. Panelli, B. Seliger, and F. M. Marincola Ontogeny and Oncogenesis Balance the Transcriptional Profile of Renal Cell Cancer Cancer Res., October 15, 2004; 64(20): 7279 - 7287. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M.-H. Tan, C. G. Rogers, J. T. Cooper, J. A. Ditlev, T. J. Maatman, X. Yang, K. A. Furge, and B. T. Teh Gene Expression Profiling of Renal Cell Carcinoma Clin. Cancer Res., September 15, 2004; 10(18): 6315S - 6321S. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. M. Stadler Therapeutic Options for Variant Renal Cancer: A True Orphan Disease Clin. Cancer Res., September 15, 2004; 10(18): 6393S - 6396S. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. R. Scherzer, K. Offe, M. Gearing, H. D. Rees, G. Fang, C. J. Heilman, C. Schaller, H. Bujo, A. I. Levey, and J. J. Lah Loss of Apolipoprotein E Receptor LR11 in Alzheimer Disease Arch Neurol, August 1, 2004; 61(8): 1200 - 1205. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Struckmann, P. Schraml, R. Simon, K. Elmenhorst, M. Mirlacher, J. Kononen, and H. Moch Impaired Expression of the Cell Cycle Regulator BTG2 Is Common in Clear Cell Renal Cell Carcinoma Cancer Res., March 1, 2004; 64(5): 1632 - 1638. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Li, M. Cuilleron, A. Gentil-Perret, M. Cottier, K. Passebosc-Faure, C. Lambert, C. Genin, and J. Tostain Rapid and Sensitive Detection of Messenger RNA Expression for Molecular Differential Diagnosis of Renal Cell Carcinoma Clin. Cancer Res., December 15, 2003; 9(17): 6441 - 6446. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S D Jenson, R S Robetorye, S D Bohling, J A Schumacher, J W Morgan, M S Lim, and K S J Elenitoba-Johnson Validation of cDNA microarray gene expression data obtained from linearly amplified RNA Mol. Pathol., December 1, 2003; 56(6): 307 - 312. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. C. Twine, J. A. Stover, B. Marshall, G. Dukart, M. Hidalgo, W. Stadler, T. Logan, J. Dutcher, G. Hudes, A. J. Dorner, et al. Disease-associated Expression Profiles in Peripheral Blood Mononuclear Cells from Patients with Advanced Renal Cell Carcinoma Cancer Res., September 15, 2003; 63(18): 6069 - 6075. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 X. Zhang, I. H. Diab, and Z. E. Zehner ZBP-89 represses vimentin gene transcription by interacting with the transcriptional activator, Sp1 Nucleic Acids Res., June 1, 2003; 31(11): 2900 - 2914. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P.T. Higgins, R. Shinghal, H. Gill, J. H. Reese, M. Terris, R. J. Cohen, M. Fero, J. R. Pollack, M. van de Rijn, and J. D. Brooks Gene Expression Patterns in Renal Cell Carcinoma Assessed by Complementary DNA Microarray Am. J. Pathol., March 1, 2003; 162(3): 925 - 932. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. C. Ibrahim, Y. Allory, F. Commo, B. Gattegno, P. Callard, and P. Paul Altered Pattern of Major Histocompatibility Complex Expression in Renal Carcinoma: Tumor-Specific Expression of the Nonclassical Human Leukocyte Antigen-G Molecule Is Restricted to Clear Cell Carcinoma While Up-Regulation of Other Major Histocompatibility Complex Antigens Is Primarily Distributed in All Subtypes of Renal Carcinoma Am. J. Pathol., February 1, 2003; 162(2): 501 - 508. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. J. Giordano, D. G. Thomas, R. Kuick, M. Lizyness, D. E. Misek, A. L. Smith, D. Sanders, R. T. Aljundi, P. G. Gauger, N. W. Thompson, et al. Distinct Transcriptional Profiles of Adrenocortical Tumors Uncovered by DNA Microarray Analysis Am. J. Pathol., February 1, 2003; 162(2): 521 - 531. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Weinschenk, C. Gouttefangeas, M. Schirle, F. Obermayr, S. Walter, O. Schoor, R. Kurek, W. Loeser, K.-H. Bichler, D. Wernet, et al. Integrated Functional Genomics Approach for the Design of Patient-individual Antitumor Vaccines Cancer Res., October 15, 2002; 62(20): 5818 - 5827. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. S. Robetorye, S. D. Bohling, J. W. Morgan, G. C. Fillmore, M. S. Lim, and K. S. J. Elenitoba-Johnson Microarray Analysis of B-Cell Lymphoma Cell Lines with the t(14;18) J. Mol. Diagn., August 1, 2002; 4(3): 123 - 136. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. A. Kirschmann, E. A. Seftor, S. F. T. Fong, D. R. C. Nieva, C. M. Sullivan, E. M. Edwards, P. Sommer, K. Csiszar, and M. J. C. Hendrix A Molecular Role for Lysyl Oxidase in Breast Cancer Invasion Cancer Res., August 1, 2002; 62(15): 4478 - 4483. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. T. Gewirtz, L. S. Collier-Hyams, A. N. Young, T. Kucharzik, W. J. Guilford, J. F. Parkinson, I. R. Williams, A. S. Neish, and J. L. Madara Lipoxin A4 Analogs Attenuate Induction of Intestinal Epithelial Proinflammatory Gene Expression and Reduce the Severity of Dextran Sodium Sulfate-Induced Colitis J. Immunol., May 15, 2002; 168(10): 5260 - 5267. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Simonnet, N. Alazard, K. Pfeiffer, C. Gallou, C. Beroud, J. Demont, R. Bouvier, H. Schagger, and C. Godinot Low mitochondrial respiratory chain content correlates with tumor aggressiveness in renal cell carcinoma Carcinogenesis, May 1, 2002; 23(5): 759 - 768. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
