Elevated Expression of Stromal Palladin Predicts Poor
Clinical Outcome in Renal Cell Carcinoma
Vivekanand Gupta1, Daniel E. Bassi1,2, Jeffrey D. Simons1, Karthik Devarajan4, Tahseen Al-Saleem2,
Robert G. Uzzo3, Edna Cukierman1*
1Cancer Biology Program, Fox Chase Cancer Center, Philadelphia, Pennsylvania, United States of America, 2Department of Pathology, Fox Chase Cancer Center,
Philadelphia, Pennsylvania, United States of America, 3Department of Surgical Oncology, Fox Chase Cancer Center, Philadelphia, Pennsylvania, United States of America,
4Department of Biostatistics, Fox Chase Cancer Center, Philadelphia, Pennsylvania, United States of America
The role that stromal renal cell carcinoma (RCC) plays in support of tumor progression is unclear. Here we sought to
determine the predictive value on patient survival of several markers of stromal activation and the feasibility of a fibroblast-
derived extracellular matrix (ECM) based three-dimensional (3D) culture stemming from clinical specimens to recapitulate
stromal behavior in vitro. The clinical relevance of selected stromal markers was assessed using a well annotated tumor
microarray where stromal-marker levels of expression were evaluated and compared to patient outcomes. Also, an in vitro
3D system derived from fibroblasts harvested from patient matched normal kidney, primary RCC and metastatic tumors was
employed to evaluate levels and localizations of known stromal markers such as the actin binding proteins palladin, alpha-
smooth muscle actin (a-SMA), fibronectin and its spliced form EDA. Results suggested that RCCs exhibiting high levels of
stromal palladin correlate with a poor prognosis, as demonstrated by overall survival time. Conversely, cases of RCCs where
stroma presents low levels of palladin expression indicate increased survival times and, hence, better outcomes. Fibroblast-
derived 3D cultures, which facilitate the categorization of stromal RCCs into discrete progressive stromal stages, also show
increased levels of expression and stress fiber localization of a-SMA and palladin, as well as topographical organization of
fibronectin and its splice variant EDA. These observations are concordant with expression levels of these markers in vivo. The
study proposes that palladin constitutes a useful marker of poor prognosis in non-metastatic RCCs, while in vitro 3D cultures
accurately represent the specific patient’s tumor-associated stromal compartment. Our observations support the belief that
stromal palladin assessments have clinical relevance thus validating the use of these 3D cultures to study both progressive
RCC-associated stroma and stroma-dependent mechanisms affecting tumorigenesis. The clinical value of assessing RCC
stromal activation merits further study.
Citation: Gupta V, Bassi DE, Simons JD, Devarajan K, Al-Saleem T, et al. (2011) Elevated Expression of Stromal Palladin Predicts Poor Clinical Outcome in Renal Cell
Carcinoma. PLoS ONE 6(6): e21494. doi:10.1371/journal.pone.0021494
Editor: Jo ¨rg D. Hoheisel, Deutsches Krebsforschungszentrum, Germany
Received April 7, 2011; Accepted May 29, 2011; Published June 28, 2011
Copyright: ? 2011 Gupta et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The Keystone Program in Personalized Kidney Cancer Therapy at Fox Chase Cancer Center (Institutional Funds) to EC, VG, TAS and RGU, an American
Urological Association Foundation Research Scholar Program Award to VG (http://www.urologyhealth.org/research/grantprograms.cfm), as well as NCI/NIH
CA113451 to EC and CA06927 to all authors. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: Edna.Cukierman@fccc.edu
In the United States, about 58,240 new renal cancers and 13,040
related deaths took place during2010 . Metastasis to distant sites,
especiallylung,bone,brainand liver,account forthemajorityofthe
morbidity associated with Renal Cell Carcinoma (RCC) .
Because surgery and available targeted therapies have limited
impact on survival in patients with advanced RCC, it is believed
that alternative approaches, such as targeting the primary and/or
secondary tumor microenvironment, could improve clinical out-
comes. A characteristic of renal cancers is that it contains a fibrous-
like stromal reaction that directly intercalates with the cancerous
epithelia . It is well accepted that tumor-associated stroma,
including activated or desmoplastic stromal fibroblasts known as
tumor- or cancer-associated fibroblasts and their self derived
extracellular matrix (ECM), play a major role in cancer develop-
ment, progression, invasion and metastasis [4,5,6]. In order to test if
kidney stroma and its tumor-activated (or tumor-associated) ECM
play pivotal roles in renal tumorigenesis, we screened a cohort of
RCC and normal histological human kidney samples and
determined the clinical significance of scoring stromal activation -
where increased collagen as well as myofibroblastic features were
positively identified- for prognostic purposes. Prompted by results
stemming from this screen, a small sample of human fibroblasts
were harvested from fresh surgical tissues obtained from primary
renal tumors, patient-paired normal adjacent kidney tissue, and
from metastatic sites. We used a fibroblast-derived ECM based
three-dimensional (3D) system of high physiological relevance,
which effectively mimics the in vivo stroma onset of various epithelial
tumors [7,8,9], to stage the stroma into discrete and well char-
to classify RCC and compared it with the established RCC staging
(e.g., TMN classification) systems. The Fuhrman Classification
system used to evaluated RCC progression using TMN classi-
fication has been regarded as an imperfect system that does not
always predict the clinical outcome for individuals at risk .
PLoS ONE | www.plosone.org1June 2011 | Volume 6 | Issue 6 | e21494
Furthermore, despite the successful identification of selected
epithelial marker proteins that characterize the stages of epithelial
transformation , none of these has been shown to serve as useful
prospective markers. We have previously demonstrated that in vitro
stromal sorting can be accomplished by assessing several aspects of
cell-derived 3D cultures such as ECM organization and expression
of known tumor-stroma markers such as alpha-smooth muscle actin
(a-SMA) [7,8,9]. To better assess RCC-associated stromal progres-
sion, here we investigate additional stromal markers which were
simultaneously assessed in vivo using the same original samples:
Palladin, an early myofibroblast differentiation marker localized to
fibroblastic stress fibers ; EDA, a spliced form of fibronectin,
upregulated during renal tumor stromal activation ; and
uPARAP/Endo180, also upregulated in the activated tumor
microenvironment  and tentatively associated to ECM (e.g.,
collagen I) degradation , facilitating tumor invasion and
Stromal palladin is a predictor of poor prognosis in renal
In order to determine the clinical relevance of a-SMA, palladin,
uPARAP/Endo 180 and EDA in RCCs, and hence their possible
usefulness as prospective markers, we analyzed their levels of
expression in the stroma of 53 normal and RCC surgical samples
that constituted a tumor microarray (TMA) described in Table 1.
First, we tested whether our TMA was representative of known
human RCC occurrences by analyzing links between node positive
or metastatic RCC instances and survival rates. Univariate
analyses using CART revealed that survival time was significantly
decreased in node positive compared to node negative (p-value
=0.004, Figure 1A) and metastatic vs. non-metastatic cases
(p-value =0.001, Figure 1B). Next, immunohistological scored
evaluations consisting of 0, 0.5, 1, or 2 values were assigned to
each sample by a blinded pathologist who was instructed to only
score stromal levels of expression for each of the assorted markers
(see examples for palladin in Figure 1e). The correlation between
immunohistological staining and clinical parameters such as
presence of local or distant metastases and survival was analyzed.
Among all the markers tested, the expression of palladin in stromal
fibroblasts showed the strongest correlation with overall survival.
Patients whose stromal palladin staining scores were at or below
0.75 showed a significantly higher survival than those with a
higher level (p-value =0.014, Figure 1c). Importantly, the same
correlation was observed in the subpopulation of patients that did
not present metastatic onsets at the time of their surgical
procedure. Figure 1d shows how using multivariable CART
analysis non-metastatic patients whose palladin levels were 0.75
or below were found to have significantly improved survi-
val compared to those with higher palladin levels (.0.75)
(p-value=0.022). Only 8% of bona fide metastatic patients (M1)
were found to be alive at 60 months while 40% of non-metastatic
patients with a palladin level exceeding 0.75 and 80% of non-
metastatic patients with a palladin level at or below 0.75 were
found to survive 60 months. These results indicate that non-
metastatic patients with lower stromal palladin levels (#0.75)
survive the longest compared to other groups. Correlation between
a-SMA and palladin levels was observed to be 0.58 (95% CI (0.36,
0.73) suggesting a similar stromal distribution for these two
markers. Although a-SMA stromal expression levels presented a
similar distribution to the one observed for palladin, the
background expression of this marker observed in fibroblasts
associated with normal kidneys was rather elevated and thus only
resulted in a marginal statistical relationship between a-SMA
expression and survival. Interestingly, comparison of the stromal
expression levels of these markers in the stroma component of
normal and tumor samples suggested an increase of expression in
the tumor-associated stroma of all markers tested. The results for
the expression of these markers in the stroma from tumor (T) vs.
normal (N) based on the Mann-Whitney test were: significant for
palladin (p-value=0.016; median=0.75 in T and 0.375 in N),
marginally significant for a-SMA (p-value =0.078; median=0.75
in T and 0.5 in N), significant for EDA (p-value=0.010;
median=0.625 in T and 0.125 in N) and relatively unchanged
for stroma control pan-collagen (p-value=0.364; median=0.50 in
T and 0.50 in N). No additional correlations were observed.
Table 1. Types of samples represented in the tumor
Age span Years
Average (range)61.86 (39-80)
Pathology N stage
lymph Node N (%)
0 30 (64)
1 0 (0)
2 8 (17)
Pathologic T stage
1 14 (33)
2 8 (19)
3 15 (35)
Pathology M Stage
M stageN (%)
Met 13 (30)
Non Met28 (65)
Pathology Stage group
Stage N (%)
1 13 (32)
2 6 (15)
3 5 (12)
4 14 (34)
Characteristics of the RCC cohort used to create the tumor microarray used in
this study are listed. N represents the number of cases while percentages are
shown inside parentheses. Note that 62% of tumors were clear cell carcinomas.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org2 June 2011 | Volume 6 | Issue 6 | e21494
Figure 1. RCC distributions and stromal palladin expression levels as they relate to patient outcomes. The depicted graphs (A-D)
correspond to Kaplan-Meier curves where the x-axis denotes the time-scale (in months) and the y-axis denotes the corresponding survival probability
(e.g., survival fraction). A separates the group of cases identified as node positive (N2, Nx) from the node negative (N0) ones. B sorts cases by
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org3 June 2011 | Volume 6 | Issue 6 | e21494
Stromal marker expression is 3D matrix-dependent
Fresh surgical samples from six selected RCC cases (Table 2 and
Table S1) which rendered a total of 59 primary and immortalized
fibroblastic cell lines were used for the in vitro portion of the study
(see Materials and Methods for details regarding these cells). These
fibroblasts were used to obtain 3D cultures in vitro using a previously
published system shown to accurately mimic in vivo physiological
behaviors of normal and tumor-associated fibroblastic cells [7,9,16].
In order to assess the levels of expression of the various stromal
markers and to uncover the conditions needed to achieve in vivo-like
phenotypes, Western Blot analyses were conducted using classic 2D
compared to 3D cultures on the above-mentioned 59 cell lines.
Quantitative analyses were conducted by measuring the marker’s
optical densities obtained from all cells. Figure 2A depicts an
example of a typical experiment, while the graphs in Figure 2 show
the compiled results where each plotted dot corresponds to a single
cell line. When comparing the 3D levels of expression of all four
markers to the corresponding 2D samples, we observed statistically
significant fold differences in a-SMA and palladin (ranging from
2 to 20 folds) in fibroblasts associated with normal, tumor or
metastatic tissues (Table S2). In contrast, uPARAP and EDA-
fibronectin ratios were only significant in samples corresponding to
the primary tumors but not to the normal (non-tumorigenic) or
metastatic tissues (Table S2). Therefore, we concluded that
fibroblastic a-SMA and palladin levels of expression in vitro are
3D matrix-dependent, suggesting that an in vivo-like relevance could
be implied to the self-derived 3D microenvironmental substrates.
Assessment of levels of stromal marker expressions
considering the original RCC stages
Sincethe3D expression levelsofallfourmarkers were distributed
with significantly scattered variations (Figures 2B-2E), we tested
whether these could have stemmed from the original tumor-stages
of the selected RCC cases. For this, the expression levels of the four
markers were plotted for all in vitro 3D samples sorted by the
corresponding original RCC stages as listed in Table 2. Figure 3
shows a cleartendency indicating that levels ofexpressionofa-SMA
and palladin (but not always uPARAP and EDA) stemming from
the lower stage I case were modest when compared to the levels
observed in 3D cultures from fibroblasts of higher tumor stages
(Table S3). Nevertheless, the levels of a-SMA, palladin and
uPARAP were distributed with larger variations of expression
compared with the more even distribution of EDA expression.
Cases showing higher a-SMA and/or palladin did not always
correspond to the ones showing higher uPARAP or EDA
(Figures 3E-3H). Moreover, within each RCC case, uPARAP and
EDA expressions were not significantly changed when cultures of
fibroblasts from normal kidney tissue were compared to tumor-
derived ones. Interestingly, it seemed that each marker presented
different patterns and that none of them truly correlated with the
tumor stages (I, III or IV) or types (histological sarcomatoid (circles)
or clear cell) of the original samples. Results suggest that stromal
progression, although influenced by tumorigenesis, could progress
at rates that are independent from their associated epithelial
components as observed above in stromal palladin using TMAs.
Sorting of stromal 3D stages using indirect
immunofluorescence in vitro
In order to test whether measured expression levels of stromal
markers in 3D cultures correlate to their subcellular localization
and architecture, we conducted indirect immunofluorescence as in
a previous publication . We assessed the topographical features
of cellular fibronectin fibers or of its splice variant EDA looking at
activated or tumor associated parallel-organized signature fiber
patterns, as opposed to the mesh-like and disorganized patterns
formed by these fibers in the cultures from fibrobalsts derived from
normal (i.e., non-activated) stage [7,8,9,17]. Hence, localized stress
fiber and homogenous patterns of a-SMA and palladin charac-
terize the activated (or tumor-associated) stromal stage, while
heterogeneous and cytosolic patterns are consistent with the non-
activated stromal stages [12,18,19,20]. Figure 4 shows represen-
tative images of four identified stromal stages regarded as i)
‘normal’ or non-activated, ii) ‘primed,’ iii) ‘primed/activated’ or
intermediate and iv) ‘activated’ or tumor-associated. Analyses
revealed that the pattern ascribed to activated stromal stages did
not always correspond to the advanced tumor stages, suggesting
that the progression of the stroma may be a somewhat
independent feature from the well established designated tumor
stage similar to the case of stromal palladin in vivo using the TMAs.
Nevertheless, the clear correlation between the immunofluorescent
analyses of 3D stromal cultures to their corresponding Western
Blots with regard to a-SMA and palladin levels supported the
conclusion that stages assessed by indirect immunofluorescence
may be predicted by the corresponding Western Blot results.
In vitro stroma stages, assessed by 3D cultures, correlate
to in vivo equivalents
In order to validate the in vitro sorting of 3D cultures into
progressive stromal stages, paraffin embedded tissue sections
metastatic positive (M1) vs. negative (M0, Mx). C separates the tumor cohort by low (palladin #0.75) and high (palladin .0.75) stromal palladin
expressions and D multivariate CART-based sorting of the cohort where non metastatic patients (M0, Mx) were further sorted by their stromal
palladin expression levels (palladin #0.75 or .0.75). The corresponding p values are provided. E are representative immunohistochemistry images of
palladin staining showing low (left panel) and high (right panel) examples or expression levels. Arrows point to the types of scored fibroblastic cells
that rendered the stromal palladin scores. Barr represents 100 mM. Note how high levels of stromal palladin are indicative of smaller survival
Table 2. RCC cases and paired tissues used to harvest
fibroblastic cell lines.
S.N. Tumor-type Nomenclature Grade StageGender/Age
1 Clear cell carcinomaT1bN0M02I F/74
2 Papillary +
T3bN2M0 HighIII a M/49
3 Clear cell carcinomaT3bN2M04 III b M/49
4 Clear cell carcinoma T3bN2M14IV a M/60
5 Clear cell carcinomaT3NxM14 IV b F/67
6 Papillary +
T3aN2M1HighIV c F/57
Six out of 22 RCCs cases were selected to harvest fibroblasts, which were used
in the in vitro portion of the study. Sample numbers (S.N.), tumor types,
nomenclatures, stages and grades are listed. Also listed are the respective
gender; female (F) and male (M), along with the age of the patients at the time
of surgery. Lower case indexes were assigned to the stages in order to
differentiate among samples obtained from cases that were characterized at
equal pathological and clinical stages.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org4 June 2011 | Volume 6 | Issue 6 | e21494
corresponding to the cases used in the in vitro analyses were
immunostained for a-SMA, palladin, uPARAP, and EDA-
fibronectin. The expression levels of these markers in the tumor-
associated stroma were assessed using a blinded scoring approach
(as done for the TMAs). Representative results are shown in
Figure 5A, while observed expression level values are plotted in
Figure 5B and listed in Table S4. Stroma of normal kidney tissue
presented only low expression or no expression levels of these
Figure 2. Expression levels of in vitro stromal markers are 3D matrix dependent. Lysates from fibroblasts cultured in 2D and 3D conditions
were subjected to Western Blot analyses. A is a representative gel showing a fibroblastic cell line control (Wi38) and samples corresponding to
fibroblasts harvested from normal kidney tissue (N), primary tumor (P) and secondary metastatic lymph node positive tumor (S) corresponding to
case number #6 in Table 2. B-E panels show optical densities (O.D.) calculated for the assorted stromal markers relative to GAPDH. Each dot
represents a single cell line. Median values as well as 25 and 75 percentiles are shown. Note the variation of the medians between 2D and 3D
conditions. See Table S2 for details.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org5 June 2011 | Volume 6 | Issue 6 | e21494
in 3D conditions were subjected to Western Blot analyses. Samples were sorted by tissue type (normal kidney and primary or secondary tumors) and by
tumor stage. Panels A-D show the optical densities (O.D.) of the corresponding stromal markers relative to GAPDH. Each dot represents a single cell line.
andlate(III and IV) stages. Sampledistributiondemonstrates a high level of variation. See Table S3for details. Panels E-H correspondtomedian values and
errors calculated for lysates obtained from 3D cultures of fibroblasts harvested from both normal and primary tumors sorted by the individual cases.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org6 June 2011 | Volume 6 | Issue 6 | e21494
markers. Immunohistochemistry analyses utilizing uPARAP and
EDA were not specific enough and therefore were inconclusive for
these analyses. Notably, results observed for a-SMA and palladin
in vitro showed expression levels and localization patterns that
corresponded to their in vivo counterparts (compare results in
Figure 5 to results in Figures 3 and 4).
Figure 4. In vivo-like 3D kidney stroma system shows at least four progressive stromal stages. Representative reconstituted confocal
images obtained from indirect immunofluorescence of selected un-extracted 3D cultures. A depicts samples labeled with antibodies against a-SMA
and fibronectin while B depicts palladin and EDA-fibronectin. Note that the homogeneity and expression along stress fibers of both a-SMA and
palladin is incremental for normal (i.e., cells harvested from the non-tumorigenic tissue of case #5 in Table 2), primed (cells isolated from the primary
tumor sample of case case #1 in Table 2), primed/activated (cells used were harvested from the primary tumor sample of case case #2 in Table 2)
and activated (cells were harvested from the primary tumor sample of case #5 in Table 2) stages. In addition, parallel patterned organization of
matrices labeled with fibronectin (A) and EDA (B) are also incremental while EDA-fibronectin also shows incremental increases in expression levels.
Overlaid images, in A and B respectively, contain a-SMA or palladin in red, fibronectin or EDA in green and nuclei in blue. Bar represents 25 mm.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org7 June 2011 | Volume 6 | Issue 6 | e21494
Several studies, including some looking at renal cancers, point to
stromal components as key players in inducing tumorigenesis and in
the acquisition of invasive behaviors [5,21,22,23,24,25,26,27].
Hence, investigators have suggested that analysis of the tumor
stroma may render meaningful prognostic data [28,29,30,31,32,33].
In this context, targeting tumor-associated stroma may hamper its
invasion and, consequently, increase the efficacy of chemo- and
radiotherapies . It is well established that tumor activated
fibroblast express myofibroblastic markers, such as a-SMA ,
palladin  uPARAP/Endo180  and EDA-fibronectin . As
the role of these stromal markers in RCC progression is currently
unknown, we investigated the association between the expression of
these markers and RCC prognosis.
In a recent collaborative work, we demonstrated that increased
expression of stromal palladin can be found in fibroblasts
associated with pancreatic ductal adenocarcinoma and other
neoplasias such as lung, skin, breast and kidney . Therefore,
and as previously suggested , one could speculate that palladin
expression may alter some key stromal properties that induce or
Figure 5. In vivo kidney stroma is sorted into progressive stages correlating the ones observed in vitro. Representative a-SMA and
palladin immunohistochemistry (A) labeled images matching samples shown in Figure 4 corresponding to Table 2 cases: #5 for both normal or non-
tumorigenic (normal stroma) and tumor (activated stroma), #1 (primed stroma) and #2 (primed/activated or intermediate stroma). Arrows in A are
pointing to stromal cells. Barr represents 50 mm. Note that, although palladin is localized to the epithelial components of the tissue samples, the
stromal marker expression and patterning of both a-SMA and palladin are incremented according to the stage of the stroma. Graphs in B represent
the in vivo stromal levels of expression of a-SMA and palladin sorted by tumor stage and by tissue type (medians and distributions are shown). See
Table S4 for details.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org8 June 2011 | Volume 6 | Issue 6 | e21494
increase tumorigenesis. To this end after analyzing our TMA data,
we concluded that stromal expression of palladin correlated with
the survival of RCC patients; higher levels of stromal predicted
lower overall survival times. Hence, palladin may constitute an
independent marker for prognosis. Strikingly, the predictive value
of this stromal marker was highly significant in patients who had
not developed metastasis suggesting that it could constitute a high
risk marker in this patient population. Another stromal marker
whose expression correlated with stromal activation, a-SMA,
showed a rather high background staining in the stroma associated
to normal kidney tissue resulting in less impressive differences with
the stroma associated with tumors. In addition, the reduced
number of samples for which survival data are available decreased
the statistical significance between the mean staining score
between these two populations. In this context, palladin seemed
to be a stronger predictor of survival than a-SMA since the
correlation showed by comparing palladin staining in stroma
associated with normal or tumoral tissue remained statistically
significant in spite of the few samples whose survival data were not
The expression of markers suggestive of stromal activation could
serve as indicators of tumor-stromal interactions , a largely
understudied aspect of RCC. Palladin plays important roles in
actin organization, sarcomere integrity, and cytoskeleton archi-
tecture. Palladin has also been associated with stimulation of cell
motility and increased matrix stiffness, both characteristics
attributed to tumor-associated activated stroma . To address
changes associated with the RCC stroma, we developed a novel
renal fibroblast derived 3D culture system that, similar to other
systems [7,9,16,22], effectively mimics many aspects of the in vivo
stromal microenvironments. We used fibroblasts harvested from
five late RCC stage cases with matching non-activated (i.e.,
normal), primary and metastatic tumors and one localized early
stage case as a control. We observed that self-derived 3D cultures
recapitulate the original (i.e., in vivo) expression levels of the four
selected markers. Moreover, 3D matrix-induced a-SMA and
palladin in vitro levels greatly correlated with the stromal activation
of fibroblasts (myofibroblasts). On the other hand, uPARAP and
EDA-fibronectin expressions appeared to be regulated indepen-
dently of 3D ECMs suggesting that perhaps their expression
depends on additional parameters such as tissue type or cellular
location. Previous reports from our laboratory indicated that, in
order to achieve in vivo-like stromal expression patterns, culturing
cells onto fibroblast-derived matrices or as un-extracted 3D
cultures, is often necessary [7,9]. By the use of these 3D cultures,
we demonstrated that levels of stromal markers a-SMA, palladin
and, to some extent, uPARAP are consistently and significantly
increased compared to those of all harvested from normal kidney
tissues, reflecting their usefulness as surrogates of in vivo specimens.
On the other hand, the fibronectin spliced form EDA did not show
In spite of the clear differences in the levels of expression of
a-SMA and palladin observed when comparing 3D cultures
produced by tumor-derived vs. normal kidney-derived fibroblasts,
relatively high distributions among samples were also observed. To
this end, it has been suggested that myofibroblastic activation
occurs in progressive phases  and that perhaps tumor-
associated stroma can also be sorted into various progressive
stages [21,35]. In an effort to determine whether incorporating
stromal features into an enhanced staging scheme could be used
for RCCs, we investigated whether stroma progression accompa-
nies or rather develops at an independent pace from tumor
staging. We questioned if our 3D culturing system effectively
mimics the original in vivo stromal features. Individual analysis of
RCC cases indicated that levels of stromal markers do not
necessarily always correlate with the original stages of the tumors.
Nevertheless, RCC cases showing the greatest expression levels of
both a-SMA and palladin indeed correlated with the most
activated 3D cultures sorted using our published methods
[7,8,9]. Interestingly, the same cases sorted as ‘activated’ presented
in vivo characteristics of activated stroma such as increased levels of
stromal markers palladin and a-SMA indicating that our 3D
system indeed mimics in vivo stromal characteristics.
Our results suggest that the rate of stroma transformation does
not correlate with the tumor stage assigned by the evaluation of
the tumor cells proposing the possibility that the assessment of
stroma progression may complement tumor stage as a clinical
prognostic variable in RCC. Moreover, we believe that we have
identified stromal palladin as an early risk factor predictor for
Materials and Methods
All human tissues used in this study were acquired after
approval of the Fox Chase Cancer Center’s Institutional Review
Board. All samples were decoded to avoid the possibility of patient
identification. It is important to note that a written informed
consent form, where patients agree to have samples taken for
research purposes, was obtained in all the cases.
Dulbecco’s modified Eagle’s medium (DMEM) was obtained
from Mediatech Inc. (Manasas, VA) and fetal bovine serum (FBS)
from Hyclone (South Logan, UT). Mouse anti-a-SMA and rabbit
anti-fibronectin were from Sigma-Aldrich (St. Louis, MO). Mouse
anti-EDA fibronectin was obtained from Abcam (Cambridge, MA)
and rabbit anti-palladin (for immunofluorescence and immuno-
histochemistry) from Proteintech group (Chicago, IL). Monoclonal
mouse anti-palladin (used in Western Blots) was a gift from Dr. C.
Otey from the University of North Carolina [12,38], and mouse
anti-uPARAP was a gift from Drs. T. Bugge (NIDCR/NIH,
Bethesda, Maryland) and L. Engelholm (Copenhagen Biocenter,
Denmark) . Rabbit anti-Vimentin was from Biovision
(Mountain View, CA), mouse anti-pan-keratin from Abcam
(Cambridge, MA) and mouse anti-GAPDH from Millipore. Goat
anti-mouse IRDye 700 and anti-rabbit IRDye 800 were from LI-
COR Biosciences (Lincoln, NE). Anti-mouse Rhodamine red and
anti-rabbit Cy5, donkey F(ab’)2 fragments were from Jackson
Laboratories (West Grove, PA). Plasmids 12245 (pLOX-TERT-
iresTK) and 12240 (pLOX-CWBmi1) were from Addgene (Cam-
bridge, MA, USA). FuGENE 6 Transfection Reagents were from
Roche Diagnostic (Indianapolis, IN, USA).
RCC patient information
53 RCCs samples were collected by the Tumor
Bank and Bio-specimen Repository Facilities at Fox Chase Cancer
Center following IRB approval. Among the 53 RCC cases, 47
constituted malignant tumor samples while 7 represented normal
tissues used as controls. Amid the 47 tumor cases, 62% of the cases
were clear cell carcinomas. A detailed description of the cohort is
presented in Table 1.
Immunohistochemistry of paraffin-embedded samples was
performed using rabbit anti-palladin (1:100) or mouse anti-a-
SMA (1:100) as primary antibodies. An avidin biotin-peroxide kit
(Vectastain Elite; Vector Laboratories, Burlingame, CA), together
with chromogen 39,39-diaminobenzidine, was used following
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org9 June 2011 | Volume 6 | Issue 6 | e21494
manufacturer’s instructions. Negative controls consisted of treated
samples incubated using iso-matched non-specific primary anti-
bodies and normal rabbit or mouse pre-immune sera. All sections
were counterstained with hematoxylin and mounted for inspection.
The score for the intensity of staining in the immunohisto-
chemistry was performed by a ‘blinded’ pathologist using
semiquantitative measurements. The assigned intensities consisted
of four categories; 0, 0.5, 1, and 2. Absence of staining was
assigned with the lowest value (0) and the strongest set of intensities
was assigned the highest (2). The correlation between immuno-
histological staining and clinical parameters such as presence of
local or distant metastases, and survival was analyzed by a
biostatistician (see below).
Isolation of primary fibroblasts from surgical samples
Fresh surgical tissue samples from partial nephrectomies
conducted at the Fox Chase Cancer Center were delivered with
the assistance of the Protocols Laboratory and Biosample
Repository Facility following protocols approved by the Institu-
tional Review Board. Twenty-two primary RCCs, 18 normal
(non-tumorigenic) kidney samples and 5 secondary tumors
(positive lymph nodes) were obtained (Table S1). Tissue samples
were rinsed in cold PBS containing 100 U/ml penicillin and
100 mg/ml streptomycin. Samples were minced and subjected to
overnight digestive dissociation using 0.2% collagenase at 37uC.
Note that samples that were big enough were divided in two
separate harvesting samples at this point. The resultant mixture
was centrifuged at 200 g for 4 min. Supernatants were passed
through 100 mm cell strainer (BD Bioscience) before plating. The
fibroblasts-enriched fraction was cultured for up to 6 passages in
DMEM containing 15% FBS, 100 U/ml penicillin, 100 mg/ml
streptomycin and 2 mM L-glutamine at 37uC using a humidified
atmosphere and 5% CO2. The resultant lines were regarded as
primary cell lines which served as parental for the immortalization
ones (see below). Homogeneity was confirmed by direct micro-
scopic observations, and cells were designated as fibroblastic after
confirmation of mesenchymal marker vimentin expression, as well
as absence of epithelial marker keratin expression. WI38 and
MCF7 cells were used as fibroblastic positive and negative
controls, respectively (Table S1).
Selection of stromal fibroblasts harvested from normal
and matched RCC tissues
Seventy-two primary fibroblastic human kidney cell lines were
obtained from the 22 RCC cases listed in Table S1. The
fibroblastic nature and the level of homogeneity of all cells were
assessed as stated above. Vimentin-positive keratin-negative lines
were kept as primary lines and also served as parental lines for
immortalization by transfection using lipofectamine (Roche,
Indianapolis, IN, USA) with plasmids 12245 (pLOX-TERT-
iresTK) and 12240 (pLOX-CWBmi1) and incubated in OPTI-
MEM (GIBCO) reduced serum according to the manufacturers’
instructions . On average, two immortalized cell lines were
obtained for each selected parental primary one. Successful
transfections were confirmed by Western Blotting assessing high
Bmi1  and low p16 expressions. Six representative cases were
selected (Table 2 and Table S1). The selected cells spanned a total
of 59 lines from which 17 were primary (parental) cells and 42
corresponded to immortalized sets. Five of the 6 selected included
fibroblasts derived from paired samples of normal (non-tumori-
genic) kidney tissue, as well as matched adjacent primary and
secondary (e.g., positive lymph nodes or adrenal gland) tumors.
The sixth case, which included cells harvested only from normal
kidney and adjacent primary tumor, corresponded to a low grade,
low stage (T1b, grade 2) clear cell RCC used as a control. Four of
the six selected cases were histologically assessed as clear cell
carcinomas, while two were identified as papillary with sarcoma-
toid in RCC features. In all, the selected cases contained two stage
III and three stage IV representatives while one was deemed low
nuclear grade and (grade 2) at stage I (Table 2). Experiments were
performed twice using both immortalized and primary lines.
Fibroblast-derived ECM based 3D cultures
Protocol was as previously described [8,41]. Briefly, 250,000
cells/ml were plated on gelatin-coated dishes or cover-slips and
maintained in a state of confluence for 6-8 days. The gelatin served
to stabilize the fibronectin matrices derived by cells cultured onto
this pre-coated surface. This way the cell-derived matrices build in
a multilayer fashion resulting in stable structures which could
withstand the intrinsic forces applied by the cells during 3D matrix
production in the layers. Cells were supplemented every 48 hours
with 50 mg/ml L-ascorbic acid (this also helps to stabilize matrix
production so cells can continue building up). The resultant
‘‘unextracted ECM based 3D cultures’’ were either lysed for
Western Blot analyses or processed for indirect immunofluores-
cence. Controls consisted of matched sample cultures grown
overnight under classic 2D conditions.
Cells were lysed and samples were resolved by SDS-PAGE and
transferred as previously described [7,8]. Blots were incubated
with a combination of primary antibodies against a-SMA (1:1000),
palladin (1:1000), EDA-fibronectin (1:1000), uPARAP (1:1000)
and GAPDH (1:2000) or Vimentin (1:1000) and pan-Keratin
(1:1000). IRDye 700 goat anti-mouse and IRDye 800 goat anti-
rabbit (1:10,000, LI-COR) were used as secondary antibodies and
blots were scanned using the Odyssey Infrared Imaging System
(LI-COR) following the Odyssey User Guide’s (version 2.1)
instructions for membranes. Optical densities were obtained using
the Odyssey 2.1.12 software with median background subtraction
and 3 points selected border width. Ratios showing relative levels
of expression with regards to GAPDH were calculated and plotted
using Prism 5 software (GraphPad Software, San Diego, CA).
Samples were labeled according to standard para-formalde-
hyde/triton procedures [7,8,16]. Briefly, samples were incubated
at room temperature with either rabbit anti-human fibronectin
(1:100) and mouse anti-a-SMA (1:100) or mouse anti-EDA-
fibronectin (1:100) and rabbit anti-palladin (1:100). Samples were
rinsed and incubated with secondary anti-rabbit Cy5 and anti-
mouse Rhodamine Red conjugated donkey F(ab’)2 fragments
(1:100) together with SYBR Green nuclear stain (1:50,000,
Invitrogen) for 30 minutes at room temperature. Samples were
rinsed and then mounted using Invitrogen’s Prolong gold anti-
fading reagent (Carlsbad, CA, USA).
Confocal Image Acquisition and Reconstitution
Samples were scanned using an Ultraview spinning-disc
confocal head (Perkin-Elmer Life Sciences, Boston, MA) mounted
on a Nikon TE-2000U microscope (Optical Apparatus Co.,
Ardmore, PA). Images, corresponding to simultaneous 488 nm,
568 nm, and 647 nm wavelengths, were captured using sequential
Z-slices corresponding to 0.5 mm. Reconstituted projections were
obtained by applying a maximum reconstruction function onto the
acquired Z-planes using MetaMorph Offline 7.0r1 imaging
analysis software (Molecular Devices, Downingtown, PA).
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org10June 2011 | Volume 6 | Issue 6 | e21494
Statistical analyses for Western blot and for normal vs. tumor
immunohistochemistry were performed using the non-parametric
Mann-Whitney test calculated by the Instat statistical software
(GraphPad Software, San Diego, CA). Regarding the tumor micro
array, in order to identify clinical variables related to patient
survival, we performed univariate and multivariable analyses by
constructing decision trees using the Classification and Regression
Trees (CART) methodology. The following clinical variables were
considered as predictors of survival time – levels of the markers a-
sma, palladin, fibronectin EDA and collagen I, as well as T stage,
N stage, M stage and pathological stage. We utilized the unified
CART framework that embeds recursive binary partitioning into
the theory of permutation tests . Significance testing
procedures were applied to determine whether no significant
association between any of the clinical variables and the response
could be stated or whether the recursion would need to stop. We
utilized the open-source R package PARTY (www.r-project.org) in
our computations . Due to the nature of our analyses, no
correction for multiple testing was employed, and a Type I error of
5% was used to test each hypothesis.
lines. List spanning the 22 RCCs cases, tumor types, nomencla-
tures and collaborative (pathological and clinical) stages and
grades used in this study. Also listed are the types of tissues from
where fibroblasts were isolated (e.g., normal kidney as well as
primary and secondary (lymph node or adrenal gland) tumors). K,
stands for keratin while V stands for vimentin. Harvested cells
were sorted by their morphological features as spread, interme-
diate or spindled. ND stands for not determined. * indicate
samples selected for the rest of the study while a, b and c in the
stage column are used to distinguish between the two stage III and
three stage IV samples among the six selected.
RCC cases and their paired/harvested cell
Optical Densities. Median calculated optical densities normal-
ized to GAPDH values are shown in A, while B-E correspond to
the P values obtained using Mann-Whitney test. Relative P value
significances were designated as extremely***, very**, or signifi-
cant*. The tissue sources from where fibroblasts were harvested
are marked as N for normal kidney, P for primary RCC and S for
secondary (metastatic) RCC. 2D and 3D correspond to two-
dimensional and three-dimensional cultures, respectively.
Medians and statistical P values for measured
statistical P values for 3D cultures sorted using the
original RCC’s stages. Median calculated optical densities
Calculated medians, fold differences and
normalized to GAPDH values are shown in A, while B-E
correspond to the indicated median fold differences and
corresponding P values obtained using the Mann-Whitney test.
Relative P value significances were designated as extremely***,
very**, or significant*. Roman numbers (I, III and IV) correspond
to the original collaborative tumor stages from where fibroblasts
were harvested. The tissue sources rendering the fibroblasts used
in the study are marked as N for normal kidney, P for primary
RCC and S for secondary (metastatic) RCC.
SMA and palladin analyzed by immunohistochemistry.
Table listing collaborative stages where * corresponds to samples
used for in vitro analyses while a, b and c serve to differentiate
among the cases as in Table 1. Types of tissues used are depicted
as normal, as well as primary or secondary for tumors. Blinded
assessment of immunohistochemistry expression levels using -, -/+,
+, ++ and +++ as scoring method. The blinded individual
explained the observed stroma (as opposed to epithelial) positive
staining of all samples under ‘‘Description of stromal expression.’’
Expression levels of in vivo stroma markers a-
We thank Dr. A. Klein-Szanto for assistance with histopathology, Mrs. D.
Kister and M. Collins for assertive data management, Mrs. E. Ragan for
proofreading, S. Goldston and Dr. J. Franco-Barraza for discussions, and
Drs. A. Klein-Szanto and G. Hudes for critical comments. We also thank
Drs. C. Otey (UNC, Chapel Hill North Carolina), T. Bugge (NIDCR/
NIH, Bethesda, Maryland) and L. Engelholm (Copenhagen Biocenter,
Denmark) for providing the anti-palladin and anti uPARAP antibodies.
This study used the following FCCC facilities: Biosample Repository,
Histopathology, Cell Culture, Glass-Washing and Talbot Research
Conceived and designed the experiments: VG DEB EC. Performed the
experiments: VG DEB JDS. Analyzed the data: VG DEB KD TA-S EC.
Contributed reagents/materials/analysis tools: VG DEB KD TA-S RGU
EC. Wrote the paper: VG DEB KD EC. Performed the majority of
experiments and analyses and assisted in drafting the manuscript: VG.
Performed the immunohistochemistry experiments, analyzed them and
assisted in drafting the manuscript: DEB. Assisted VG in performing many
of the experiments and participated in drafting the manuscript: JDS. Was
responsible for all statistical analyses and their interpretations and also
assisted in drafting the manuscript: KD. Provided all the pathological
information and reviewed the final draft for accuracy: TA-S. Assisted in
providing the fresh surgical material as well as designing many aspects of
the study and the manuscript draft: RGU. Was responsible for the study
coordination, as well as the design and interpretation of all results and
revision of the drafted manuscript: EC. Read and approved the final
manuscript: VG DEB JDS KD TA-S RGU EC.
1. Jemal A, Siegel R, Xu J, Ward E (2010) Cancer statistics, 2010. CA Cancer J Clin
2. Gupta K, Miller JD, Li JZ, Russell MW, Charbonneau C (2008) Epidemiologic
and socioeconomic burden of metastatic renal cell carcinoma (mRCC): a
literature review. Cancer Treat Rev 34: 193–205.
3. Lohi J, Leivo I, Oivula J, Lehto VP, Virtanen I (1998) Extracellular matrix in
renal cell carcinomas. Histol Histopathol 13: 785–796.
4. De Wever O, Demetter P, Mareel M, Bracke M (2008) Stromal myofibroblasts
are drivers of invasive cancer growth. Int J Cancer 123: 2229–2238.
5. Kalluri R, Zeisberg M (2006) Fibroblasts in cancer. Nat Rev Cancer 6: 392–401.
6. Nelson CM, Bissell MJ (2006) Of extracellular matrix, scaffolds, and signaling:
tissue architecture regulates development, homeostasis, and cancer. Annu Rev
Cell Dev Biol 22: 287–309.
7. Amatangelo MD, Bassi DE, Klein-Szanto AJ, Cukierman E (2005) Stroma-
derived three-dimensional matrices are necessary and sufficient to promote
desmoplastic differentiation of normal fibroblasts. Am J Pathol 167: 475–488.
8. Castello-Cros R, Cukierman E (2009) Stromagenesis during tumorigenesis:
characterization of tumor-associated fibroblasts and stroma-derived 3D
matrices. Methods Mol Biol 522: 275–305.
9. Quiros RM, Valianou M, Kwon Y, Brown KM, Godwin AK, et al. (2008)
Ovarian normal and tumor-associated fibroblasts retain in vivo stromal
characteristics in a 3-D matrix-dependent manner. Gynecol Oncol 110: 99–109.
10. Delahunt B (2009) Advances and controversies in grading and staging of renal
cell carcinoma. Mod Pathol 22(Suppl 2): S24–S36.
11. Pan CC, Chen PC, Ho DM (2004) The diagnostic utility of MOC31, BerEP4,
RCC marker and CD10 in the classification of renal cell carcinoma and renal
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org11 June 2011 | Volume 6 | Issue 6 | e21494
oncocytoma: an immunohistochemical analysis of 328 cases. Histopathology 45: Download full-text
12. Ronty MJ, Leivonen SK, Hinz B, Rachlin A, Otey CA, et al. (2006) Isoform-
Specific Regulation of the Actin-Organizing Protein Palladin during TGF-beta1-
Induced Myofibroblast Differentiation. J Invest Dermatol 126: 2387–2396.
13. Engelholm LH, Ingvarsen S, Jurgensen HJ, Hillig T, Madsen DH, et al. (2009)
The collagen receptor uPARAP/Endo180. Front Biosci 14: 2103–2114.
14. Messaritou G, East L, Roghi C, Isacke CM, Yarwood H (2009) Membrane type-
1 matrix metalloproteinase activity is regulated by the endocytic collagen
receptor Endo180. J Cell Sci 122: 4042–4048.
15. Curino AC, Engelholm LH, Yamada SS, Holmbeck K, Lund LR, et al. (2005)
Intracellular collagen degradation mediated by uPARAP/Endo180 is a major
pathway of extracellular matrix turnover during malignancy. J Cell Biol 169:
16. Cukierman E, Pankov R, Stevens DR, Yamada KM (2001) Taking cell-matrix
adhesions to the third dimension. Science 294: 1708–1712.
17. Provenzano PP, Eliceiri KW, Campbell JM, Inman DR, White JG, et al. (2006)
Collagen reorganization at the tumor-stromal interface facilitates local invasion.
BMC Med 4: 38.
18. Desmouliere A, Chaponnier C, Gabbiani G (2005) Tissue repair, contraction,
and the myofibroblast. Wound Repair Regen 13: 7–12.
19. Goicoechea SM, Bednarski B, Garcia-Mata R, Prentice-Dunn H, Kim HJ, et al.
(2009) Palladin contributes to invasive motility in human breast cancer cells.
Oncogene 28: 587–598.
20. Goicoechea SM, Bednarski B, Stack C, Cowan DW, Volmar K, et al. (2010)
Isoform-specific upregulation of palladin in human and murine pancreas tumors.
PLoS ONE 5: e10347.
21. Beacham DA, Cukierman E (2005) Stromagenesis: the changing face of
fibroblastic microenvironments during tumor progression. Semin Cancer Biol
22. Goetz JG, Minguet S, Navarro-Lerida I, R JJL, Samaniego R, et al. (2011)
Biomechanical remodeling of the microenvironment by stromal caveolin-1
favors tumor invasion and metastasis. Cell (in press); (July 8, 2011).
23. Kaplan RN, Riba RD, Zacharoulis S, Bramley AH, Vincent L, et al. (2005)
VEGFR1-positive haematopoietic bone marrow progenitors initiate the pre-
metastatic niche. Nature 438: 820–827.
24. Pan J, Mestas J, Burdick MD, Phillips RJ, Thomas GV, et al. (2006) Stromal
derived factor-1 (SDF-1/CXCL12) and CXCR4 in renal cell carcinoma
metastasis. Mol Cancer 5: 56.
25. Trimboli AJ, Cantemir-Stone CZ, Li F, Wallace JA, Merchant A, et al. (2009)
Pten in stromal fibroblasts suppresses mammary epithelial tumours. Nature 461:
26. Kwon Y, Cukierman E, Godwin AK (2011) Differential expression of adhesive
molecules and proteases define mechanisms of ovarian tumor cell matrix
penetration/Invasion. PLoS ONE 6: e18872.
27. Castello ´-Cros R, Khan DR, Simons J, Valianou M, Cukierman E (2009) Staged
stromal extracellular 3D matrices differentially regulate breast cancer cell
responses through PI3K and beta1-integrins. BMC Cancer 9: 94.
28. Council L, Hameed O (2009) Differential expression of immunohistochemical
markers in bladder smooth muscle and myofibroblasts, and the potential utility
of desmin, smoothelin, and vimentin in staging of bladder carcinoma. Mod
Pathol 22: 639–650.
29. Sund M, Kalluri R (2009) Tumor stroma derived biomarkers in cancer. Cancer
Metastasis Rev 28: 177–183.
30. Erkan M, Michalski CW, Rieder S, Reiser-Erkan C, Abiatari I, et al. (2008) The
Activated Stroma Index Is a Novel and Independent Prognostic Marker in
Pancreatic Ductal Adenocarcinoma. Clin Gastroenterol Hepatol 6: 1155–1161.
31. Bergamaschi A, Tagliabue E, Sorlie T, Naume B, Triulzi T, et al. (2008)
Extracellular matrix signature identifies breast cancer subgroups with different
clinical outcome. J Pathol 214: 357–367.
32. Kovacs G (1989) Papillary renal cell carcinoma. A morphologic and cytogenetic
study of 11 cases. Am J Pathol 134: 27–34.
33. Witkiewicz AK, Nguyen KH, Dasgupta A, Kennedy EP, Yeo CJ, et al. (2008)
Co-expression of fatty acid synthase and caveolin-1 in pancreatic ductal
adenocarcinoma: implications for tumor progression and clinical outcome. Cell
Cycle 7: 3021–3025.
34. Chometon G, Jendrossek V (2009) Targeting the tumour stroma to increase
efficacy of chemo- and radiotherapy. Clin Transl Oncol 11: 75–81.
35. Desmouliere A, Guyot C, Gabbiani G (2004) The stroma reaction myofibro-
blast: a key player in the control of tumor cell behavior. Int J Dev Biol 48:
36. Salaria SN, Illei P, Sharma R, Walter KM, Klein AP, et al. (2007) Palladin is
overexpressed in the non-neoplastic stroma of infiltrating ductal adenocarcino-
mas of the pancreas, but is only rarely overexpressed in neoplastic cells. Cancer
Biol Ther 6: 324–328.
37. Goicoechea SM, Arneman D, Otey CA (2008) The role of palladin in actin
organization and cell motility. Eur J Cell Biol 87: 517–525.
38. Parast MM, Otey CA (2000) Characterization of palladin, a novel protein
localized to stress fibers and cell adhesions. J Cell Biol 150: 643–656.
39. Salmon P, Oberholzer J, Occhiodoro T, Morel P, Lou J, et al. (2000) Reversible
immortalization of human primary cells by lentivector-mediated transfer of
specific genes. Mol Ther 2: 404–414.
40. Liu J, Cao L, Chen J, Song S, Lee IH, et al. (2009) Bmi1 regulates mitochondrial
function and the DNA damage response pathway. Nature 459: 387–392.
41. Beacham DA, Amatangelo MD, Cukierman E (2006) Preparation of
Extracellular Matrices Produced by Cultured and Primary Fibroblasts. In:
Bonifacino JS, Dasso M, Lippincott-Schwartz J, Harford JB, Yamada KM, eds.
Curr Protocols Cell Biol: John K. Wiley & Sons. pp 10.09.01–10.09.21.
42. Hothorn T, Hornik K, Zeileis A (2006) Unbiased recursive partitioning: A
conditional inference framework. Journal of Computational and Graphical
Statistics 15: 651–674.
43. Team RDC (2005) R: A language and environment for statistical computing. In:
Computing RFfS, ed. Vienna, Austria: R Foundation for Statistical Computing.
pp. The R Project for Statistical Computing.
Stromal Palladin, a High Risk Marker in RCC
PLoS ONE | www.plosone.org12 June 2011 | Volume 6 | Issue 6 | e21494