Claudin 4 Is Differentially Expressed between Ovarian Cancer Subtypes and Plays a Role in Spheroid Formation.

Abstract

Claudin 4 is a cellular adhesion molecule that is frequently overexpressed in ovarian cancer and other epithelial cancers. In this study, we sought to determine whether the expression of claudin 4 is associated with outcome in ovarian cancer patients and may be involved in tumor progression. We examined claudin 4 expression in ovarian cancer tissues and cell lines, as well as by immunohistochemical staining of tissue microarrays (TMAs; n = 500), spheroids present in patients' ascites, and spheroids formed in vitro. Claudin 4 was expressed in nearly 70% of the ovarian cancer tissues examined and was differentially expressed across ovarian cancer subtypes, with the lowest expression in clear cell subtype. No association was found between claudin 4 expression and disease-specific survival in any subtype. Claudin 4 expression was also observed in multicellular spheroids obtained from patients' ascites. Using an in vitro spheroid formation assay, we found that NIH:OVCAR5 cells treated with shRNA against claudin 4 required a longer time to form compact spheroids compared to control NIH:OVCAR5 cells that expressed high levels of claudin 4. The inability of the NIH:OVCAR5 cells treated with claudin 4 shRNA to form compact spheroids was verified by FITC-dextran exclusion. These results demonstrate a role for claudin 4 and tight junctions in spheroid formation and integrity.

Full text

Page 1

Int. J. Mol. Sci. 2011, 12, 1334-1358; doi:10.3390/ijms12021334

International Journal of
Molecular Sciences
ISSN 1422-0067
www.mdpi.com/journal/ijms
Article
Claudin 4 Is Differentially Expressed between Ovarian Cancer
Subtypes and Plays a Role in Spheroid Formation
Kristin L. M. Boylan 1, Benjamin Misemer 1, Melissa S. DeRycke 1, John D. Andersen 1,
Katherine M. Harrington 1, Steve E. Kalloger 2, C. Blake Gilks 2, Stefan E. Pambuccian 1 and
Amy P. N. Skubitz 1,*
1 Department of Laboratory Medicine and Pathology, University of Minnesota, Minneapolis, MN
55455, USA; E-Mails: boyla002@umn.edu (K.L.M.B.); benjamin.misemer@gmail.com (B.M.);
deryc004@umn.edu (M.S.D.); ande0555@umn.edu (J.D.A.); harr0750@umn.edu (K.M.H.);
pambu001@umn.edu (S.E.P.)
2 Cheryl Brown Ovarian Cancer Outcomes Unit, British Columbia Cancer Agency, Vancouver,
Canada; E-Mails: skalloger@mac.com (S.E.K.); Blake.Gilks@vch.ca (C.B.G.)
* Author to whom correspondence should be addressed; E-Mail: skubi002@umn.edu;
Tel.: +1-612-625-5920; Fax: +1-612-625-5622.
Received: 17 January 2011; in revised form: 12 February 2011 / Accepted: 12 February 2011 /
Published: 22 February 2011

Abstract: Claudin 4 is a cellular adhesion molecule that is frequently overexpressed in
ovarian cancer and other epithelial cancers. In this study, we sought to determine whether
the expression of claudin 4 is associated with outcome in ovarian cancer patients and may
be involved in tumor progression. We examined claudin 4 expression in ovarian cancer
tissues and cell lines, as well as by immunohistochemical staining of tissue microarrays
(TMAs; n = 500), spheroids present in patients’ ascites, and spheroids formed in vitro.
Claudin 4 was expressed in nearly 70% of the ovarian cancer tissues examined and was
differentially expressed across ovarian cancer subtypes, with the lowest expression in clear
cell subtype. No association was found between claudin 4 expression and disease-specific
survival in any subtype. Claudin 4 expression was also observed in multicellular spheroids
obtained from patients’ ascites. Using an in vitro spheroid formation assay, we found that
NIH:OVCAR5 cells treated with shRNA against claudin 4 required a longer time to form
compact spheroids compared to control NIH:OVCAR5 cells that expressed high levels of
claudin 4. The inability of the NIH:OVCAR5 cells treated with claudin 4 shRNA to form
OPEN ACCESS

Page 2

Int. J. Mol. Sci. 2011, 12


1335
compact spheroids was verified by FITC-dextran exclusion. These results demonstrate a
role for claudin 4 and tight junctions in spheroid formation and integrity.
Keywords: ovarian cancer; claudin 4; biomarker; spheroid; ascites

1. Introduction
Ovarian cancer is the most lethal gynecological malignancy, resulting in approximately 125,000
deaths yearly, worldwide [1]. Due to the paucity of specific symptoms and the lack of an effective
screening method, the majority of ovarian cancers are diagnosed at late stages of malignancy, after the
tumor has spread beyond the ovary [2]. Although initial response to treatment (surgery and
chemotherapy) is favorable, most patients will relapse with tumors that are chemoresistant and
ultimately die of their disease.
In contrast to other solid tumors, the most common method for ovarian cancer metastasis is direct
peritoneal spread. Tumor cells slough off the ovary into the peritoneal fluid where they are
disseminated throughout the abdominal cavity, and subsequently attach to the mesothelial cell lining
and invade, forming metastatic outgrowths [3]. Late stage cancers are frequently associated with
ascites, and tumor cells can be shed into the ascites fluid either as single cells or multicellular
aggregates called spheroids. Spheroids suspended in the ascites fluid were previously thought to be
quiescent. However, we and others have shown that ovarian cancer spheroids are a source of tumor
invasion and metastasis [4–8]. Spheroids are also chemoresistant [9,10], implicating spheroids as a
factor in disease persistence or recurrence.
In an effort to find novel biomarkers for ovarian cancer using gene expression profiling, we and
others [11–15] have identified claudin 4 as a gene that is highly overexpressed in ovarian cancer, and
thus may contribute to tumor formation and metastasis. Claudins are a family of cellular adhesion
molecules that are components of tight junctions, which play important roles in cell polarity,
paracellular transport, and the formation of epithelial cell sheets which serve as a barrier in tissues.
Claudin expression in normal cells is tissue specific, and altered claudin expression has been identified
in multiple cancer types [16,17]. Consistent with the idea that tumor formation is associated with tight
junction disruption, downregulation of claudin family members has been reported in some cancers, and
is associated with a poor prognosis or metastatic disease [18–27]. In contrast, claudin expression may
also be elevated in different types of cancers and associated with a metastatic phenotype [28–33].
In this study, we sought to validate the overexpression of claudin 4 that we previously observed in
ovarian cancer tissues in our gene microarray experiments [12]. We have evaluated the levels of
claudin 4 RNA and protein in ovarian cancer tissues and cell lines using RT-PCR, qRT-PCR, Western
immunoblotting, and immunohistochemistry, in order to assess the potential of claudin 4 as an ovarian
cancer biomarker. We also sought to determine whether claudin 4 overexpression could be correlated
with relevant clinical outcomes using tissue microarrays comprised of 500 cases of clinically annotated
ovarian cancer. Finally, because of its role in intercellular adhesion, and the importance of ovarian
cancer spheroids in ovarian cancer metastasis and chemoresistance, we examined the expression of
claudin 4 in ovarian cancer spheroids and its potential role in spheroid formation.

Page 3

Int. J. Mol. Sci. 2011, 12


1336
2. Results and Discussion
2.1. Claudin 4 RNA Is Overexpressed in Ovarian Cancer
Our previous analysis of gene expression in human tissues identified claudin 4 as one of 66 genes
that were upregulated in serous ovarian cancer relative to normal ovaries and over 300 other normal
and diseased tissues [12]. Our initial validation identified claudin 4 as one of three genes that best
distinguished between ovarian carcinoma and normal ovary tissues [12]. To further validate our gene
expression studies, we examined the expression of claudin 4 RNA in ovary tissues and cell lines.
Consistent with our microarray data, RT-PCR showed 5/5 serous ovarian cancer tumor tissue samples
tested were positive for claudin 4 compared to only 2/6 normal ovary samples (data not shown). Other
groups have also identified claudin 4 RNA upregulation in ovarian cancer tissues [11,13–15] as well as
pancreatic, prostate, and squamous cell carcinomas [29,31,34–36].
Because surface epithelial cells comprise only a minor fraction of the normal ovary, we examined
the expression of claudin 4 in ovarian cancer and immortalized normal ovarian surface epithelial
(NOSE) cell lines. We also used ovarian cancer cell lines as a more pure population of tumor cells,
without contaminating stroma and other cell types. Similar to what others have reported [32,37], we
found the level of claudin 4 RNA in ovarian cancer cell lines was varied. By qRT-PCR, we found the
cell lines OVCA433, C-13, OVCAR5, OV2008, CAOV3, and SKOV3 express high levels of claudin 4
mRNA, while the cell lines OVCA 429, ES-2, MA148, HEY, and A2780-CP (as well as all of the
NOSE cell lines) were found to express low levels of claudin 4 mRNA (Figure 1A). The variable
expression of claudin 4 in ovarian cancer cell lines suggests that claudin 4 expression may be
associated with functional characteristics of the cell lines, such as proliferation rate or aggressive
behavior. In other studies, the expression of claudin 4 in ovarian cancer cell lines has been shown to
increase cell migration and invasion [28].
2.2. Claudin 4 Protein Is Overexpressed in Ovarian Cancer Cell Lines and Tissues
We next examined claudin 4 protein expression by Western immunoblot analysis in both cell lines
and tissues. Claudin 4 protein was detected in the six ovarian cancer cell lines that expressed high
levels of claudin 4 RNA (OVCA433, C-13, OVCAR5, OV2008, CAOV3, and SKOV3), whereas
ovarian cancer cell lines with low levels of claudin 4 RNA expression and all of the NOSE cell lines
were negative for claudin 4 protein by Western blot (Figure 1B). These results, coupled with the
qRT-PCR data, led us to select the NIH:OVCAR5 and MA148 cell lines for subsequent experiments in
this study. We observed that all seven primary tumors from women with stage III/IV serous ovarian
cancer tested were positive for claudin 4 protein expression (Figure 1C). Although some claudin 4
transcripts were detected in normal ovaries by RT-PCR, by Western blotting, none of the five normal
ovary tissues tested were positive for expression of claudin 4 protein (Figure 1C). These results are
supported by other recent studies in which claudin 4 protein expression was demonstrated in lysates
from ovarian cancer cell lines, but not in cultures of NOSE cells [32,37,38]. One caveat to this is the
emerging concept that some ovarian cancers, in particular serous subtype tumors, arise from the
fimbria of the fallopian tubes [39], which could limit the validity of normal ovaries for comparison of
gene expression in these and other experiments [11,13–15,32,37,38].

Page 4

Int. J. Mol. Sci. 2011, 12


1337
Figure 1. Claudin 4 RNA and protein expression in ovarian cancer tissues and cell
lines. (A) Claudin 4 RNA expression in 13 ovarian cancer cell lines and 7 NOSE cell lines
as determined by qRT-PCR. Expression values shown as fold change over the lowest
expressing cell line (1816-575), and are the average of two experiments (see Experimental
section). (B) Claudin 4 protein expression was determined by Western immunoblot
analysis of ovarian cancer and NOSE cell lines (50 µg protein/lane). -actin serves as a
loading control. (C) Claudin 4 protein expression was determined by Western immunoblot
analysis of 7 primary stage III/IV serous ovarian cancer tissues (C1–C7) and 5 normal
ovaries (N1-N5) (50 µg protein/lane). -actin, loading control.

2.3. Claudin 4 Protein Expression in Ovarian Cancer Tissues
In previous immunohistochemical (IHC) studies, we observed that claudin 4 staining was localized
to the cell membrane of frozen sections from 15 serous ovarian cancer primary tumors, and 15 serous
ovarian cancer tumors metastatic to the omentum; no claudin 4 staining was observed in the surface
epithelial cells of the 15 normal ovaries that were examined [12]. These earlier studies used fresh,
snap-frozen tissues that were embedded in OCT, so that the antigens would not be destroyed by
fixatives, increasing the likelihood that the antibodies would recognize the antigens in the tissues. In
the current study, we initially performed IHC staining using a test set of 58 formalin-fixed
paraffin-embedded (FFPE) tissue blocks including 21 FFPE cases of normal ovaries with intact surface
epithelial cells. The FFPE tissue sections were used in order to determine if formalin fixation would
interfere with the detection of claudin 4, and to optimize the staining methodology for subsequent
FFPE tissue microarrays. In the test set of FFPE tissues, we observed that staining for claudin 4 in

Page 5

Int. J. Mol. Sci. 2011, 12


1338
ovarian cancer cells was localized to the cancer cell membranes with some cytoplasmic blush
(Figure 2A). Overall, the ovarian cancer tumor cells, but not stroma, had a high percentage of claudin
4 staining in the individual sections. Normal ovarian surface epithelium was either negative or had a
slight blush of staining present. Claudin 4 staining was observed in 64% of the serous ovarian cancer
tissues (21/33), 75% of the clear cell ovarian cancer tissues (3/4), and only 19% of the normal ovary
tissues (4/21). The successful optimization of IHC staining for FFPE tissues and the results showing
increased levels of claudin 4 expression in serous ovarian cancer tissues compared to normal ovaries,
not only validated our previous results [12], but led us to examine claudin 4 expression in a much
larger cohort of patients.
Figure 2. Claudin 4 immunohistochemical staining of FFPE tissues. (A) Representative
claudin 4 staining of whole mount sections of serous ovarian cancer and normal ovary.
Top, 200× magnification; bottom, enlargement to show detail. (B) Examples of claudin 4
staining and scoring for TMA samples: 0, no cancer cells staining; +1, <10% of cancer
cells staining; +2, 10–50% of cancer cells staining; +3, >50% of cancer cells staining. In
some cases, positive scores (1, 2, and 3) were binarized for analysis.

Page 6

Int. J. Mol. Sci. 2011, 12


1339
2.4. Claudin 4 Protein Is Differentially Expressed between Subtypes of Ovarian Cancer
To determine whether other subtypes of ovarian cancer would also show increased levels of
expression and if claudin 4 expression in ovarian cancer was associated with outcome or other clinical
parameters, we performed immunohistochemical staining of claudin 4 on tissue microarrays (TMA).
The TMAs encompassed 500 cases of epithelial ovarian cancer of different subtypes (serous,
mucinous, endometrioid and clear cell; Table 1; Figure 2B); each tissue was associated with patient
clinical data, including up to 20 years of follow-up [40]. Overall, claudin 4 expression was observed in
69.9% of ovarian cancer patients, with differential expression observed between the different ovarian
cancer subtypes (p = 0.0026; Figure 3A), due to a lower percentage of cells stained in clear cell
tumors. The highest percentage of expression was observed in the endometrioid and mucinous
subtypes (both 77.4% positive), compared to serous (72.17% positive), and clear cell (57.58%
positive) subtypes. These results extend the previous analysis of claudin 4 expression in ovarian cancer
subtypes [32,37,38,41], which are generally in agreement with our data. In these prior studies, claudin
4 expression was elevated in the majority of cases of epithelial ovarian cancer, with approximately
70% of serous ovarian cancers staining positively for claudin 4 [32,37,38,41]. However, in contrast to
our results showing that the endometrioid and mucinous subtypes of ovarian cancer had the highest
levels of claudin 4 expression, Litkouhi et al. found the highest percentage of claudin 4 staining in
endometrioid and clear cell subtypes, although with a much smaller sample size [37].
Table 1. Subtype, stage, Silverberg grade, and Claudin 4 score of ovarian cancer tissue
microarrays.
Subtype Median Age
(range)
Stage N Claudin 4
Positive (%)
Silverberg
Grade
N Claudin 4
Positive (%)
Claudin 4 Positive
Overall (%)
Serous
(n = 212)
59.6
(33.5–86.0)
I 50 33 (66.0%) 1 12 7 (58.3 %) 153 (72.2 %)
II 93 69 (74.2%) 2 56 37 (66.1%)
III 69 51 (73.9%) 3 144 109 (75.7%)
Endometrioid
(n = 125)
54.1
(29.4–88.1)
I 69 54 (78.3%) 1 82 66 (81.5%) 96 (77.4 %)
II 50 37 (75.5%) 2 35 27 (77.1%)
III 6 5 (83.3%) 3 8 3 (37.5%)
Clear Cell
(n = 132)
55.0
(28.1–89.0)
I 68 45 (66.3%) 1 0* N/A 76 (57.6 %)
II 56 26 (46.4%) 2 0* N/A
III 8 5 (62.5%) 3 132 76 (57.6 %)
Mucinous
(n = 31)
56.4
(25.4–76.7)
I 18 15 (83.3%) 1 11 8 (72.7%) 24
(77.4%)
II 12 8 (66.7%) 2 18 14 (77.8%)
III 1 1 (100%) 3 2 2 (100.0%)
Total
(n = 500)
56.6
(25.4–89.0)
I 205 147 (71.7%) 1 105 81 (77.1%) 349
(69.9%)
II 211 140 (66.4%) 2 109 78 (71.6%)
III 84 62 (73.8%) 3 286 190 (66.4%)
* All clear cell carcinomas are considered high grade.

Page 7

Int. J. Mol. Sci. 2011, 12


1340
Figure 3. Claudin 4 expression in ovarian cancer subtypes in tissue microarrays.
(A) Claudin 4 is differentially expressed in ovarian cancer subtypes (p = 0.0026). Percent
of TMA cases positive for claudin 4 staining in ovarian cancer subtypes. Staining score:
black, score +1 (<10% of cancer cells staining); white, score +2 (10–50% of cancer cells
staining); gray, score +3 (>50% of cancer cells staining). (B) Claudin 4 expression by
Silverberg Grade. Percent of TMA cases positive for claudin 4 by Silverberg Grade (grade
1 is black; grade 2 is diagonal stripe; grade 3 is gray). Positive scores (1, 2, and 3) were
binarized. Claudin 4 expression is significantly lower in high grade endometrioid ovarian
cancer (p = 0.0178). (C) Percent of TMA cases positive for claudin 4 by stage (stage I is
black; stage II is diagonal stripe; stage III is gray). Positive scores (1, 2, and 3) were
binarized. Claudin 4 expression was not significantly different between the stages.


Claudin 4 expression was also analyzed in each subtype by Silverberg grade and stage (Figures 3B
and 3C). In the endometrioid subtype, claudin 4 was differentially expressed by Silverberg grade
(p = 0.0178), and was inversely associated with grade; 81.48% of Grade 1 tumors were claudin 4
positive, 77.14% of Grade 2 tumors were claudin 4 positive, and 37.5% of Grade 3 tumors were
claudin 4 positive. In contrast, previous studies that examined primarily serous tumors suggested that
claudin 4 expression was increased in undifferentiated tumors [37,38]; however, those studies included
only a small number of non-serous subtype tumors. No other associations between claudin 4
expression and stage or Silverberg grade were observed.

Page 8

Int. J. Mol. Sci. 2011, 12


1341
Figure 4. (A) Kaplan-Meyer survival curve of TMA data for 500 ovarian cancer patients
showing disease-specific survival measured in days. (B) Kaplan-Meyer survival curve of
TMA data for 500 ovarian cancer patients showing relapse-free survival measured in days.
Red, staining score 0; green, staining score +1; blue, staining score +2; orange, staining
score +3.

Figure 5. Disease-specific survival by ovarian cancer subtype. Kaplan-Meyer survival
curve of TMA data for ovarian cancer patients shown by subtype. Disease-specific survival
measured in days. Positive scores were binarized and are shown in gray. Black, claudin 4
negative.

Page 9

Int. J. Mol. Sci. 2011, 12


1342
As shown by Kaplan-Meier curves (Figure 4), there was no association between claudin 4
expression and disease-specific survival or relapse-free survival in ovarian cancer overall. When the
survival data was examined by histological subtype (Figure 5), there was no significant association
between claudin 4 expression and survival in any of the ovarian cancer subtypes. In a smaller cohort of
42 high grade serous tumors, Litkouhi et al. also found no association between claudin 4 expression
and survival [37]. In contrast, Lanigan et al. recently reported that overexpression of claudin 4 was
associated with an adverse outcome in breast cancer [30]. Similarly, claudin 4 overexpression is
associated with poor outcome in clear cell renal cell carcinoma [42]. Although the results from the
TMA studies did not provide an association between claudin 4 protein expression and survival, an
important finding was made. Namely, the data shows that claudin 4 protein is expressed by the
majority of ovarian cancer tissues, but not by normal surface epithelial cells of the ovary, and thus may
serve as a target for therapy. The implications for the role of claudin 4 in other functional aspects of
ovarian cancer dissemination were therefore pursued.
2.5. Claudin 4 Plays a Role in Spheroid Formation/Integrity
Our finding that claudin 4 is localized primarily to the membrane in immunohisotchemically stained
slides corresponds to previous reports by ourselves and others [13,15,37]. However, other studies have
shown claudin 4 staining of both the cytoplasm and membrane in some serous ovarian cancer tumors,
suggesting that in addition to its role in tight junction formation, claudin 4 may have additional
functions regulating proliferation or differentiation [32,38]. In an effort to explore the potential
functional role of claudin 4 in ovarian cancer dissemination and determine whether functional tight
junctions are formed, we examined claudin 4 protein expression and localization by
immunocytochemistry in cells from the ascites of ovarian cancer patients. Late stage ovarian cancers
are frequently associated with the accumulation of peritoneal ascites fluid, which may contain ovarian
cancer cells present either singly or as multicellular spheroids. Eight of 10 ascites samples showed
positive claudin 4 membrane staining, either in single cells or multicellular aggregates (spheroids) or
both, with strong staining visible at the points of cell-cell contact (Figure 6, arrows). Our findings are
consistent with those of Kleinberg et al., who observed claudin 4 staining in over 90% of 218 ovarian
cancer effusions examined [43].
To determine whether the expression of claudin 4 may affect the formation of the spheroids which
are frequently found in the ascites of ovarian cancer patients, we engineered two ovarian cancer cell
lines to express different levels of claudin 4. The Western blot in Figure 7A shows no endogenous
expression of claudin 4 in the MA148 cell line, while the NIH:OVCAR5 cell line expresses high levels
of endogenous claudin 4. A claudin 4 transgene was then ectopically expressed in MA148 cells, and
claudin 4 protein was shown to be expressed at high levels (Figure 7A). Conversely, NIH:OVCAR5
cells were transfected with an shRNA directed against claudin 4, resulting in a decrease in claudin 4
expression levels (Figure 7A).
Immunocytochemistry of spheroids formed in vitro from cultured cells expressing different levels
of claudin 4 show claudin 4 localized to the membrane in cultured spheroids, similar to spheroids from
patient ascites (Figure 7B). Although the NIH:OVCAR5 cells expressing an shRNA against claudin 4
have substantially reduced levels of claudin 4 protein, they were still able to form spheroids. Similarly,

Page 10

Int. J. Mol. Sci. 2011, 12


1343
the ovarian cancer cell line MA148 had undetectable levels of claudin 4 expression, yet these cells
formed compact multicellular spheroids in vitro; demonstrating that claudin 4 expression is not
essential for spheroid formation. Previous analysis of breast and ovarian cancer cell lines suggests that
a diverse array of adhesion molecules, including cadherins and beta 1 integrin, are involved in
spheroid formation in vitro, depending on the cell line [8,44].
Figure 6. Claudin 4 expression in patient spheroids. Immunocytochemical staining of
spheroids isolated from the ascites of two representative stage III/IV serous ovarian cancer
patients. Spheroids were stained with either an antibody against claudin 4 or normal mouse
IgG, followed by a FITC-conjugated secondary antibody (green). Nuclei were stained with
DAPI (blue). Arrows indicate claudin 4 staining at sites of cell-cell contact. Bar = 20 µm.

2.6. Claudin 4 Increases the Rate of Ovarian Cancer Spheroid Formation
Although claudin 4 is not absolutely required for spheroid formation, we examined in vitro spheroid
formation over time in ovarian cancer cells expressing different levels of claudin 4. For both ovarian
cancer cell lines tested (NIH:OVCAR5 and MA148), tight, round multicellular aggregates or spheroids
formed from single cells after approximately 24 hr. The size of the spheroids formed in vitro was at
least partially dependent upon the number of cells plated (data not shown). As shown in Figure 8, the
three-dimensional structures formed by NIH:OVCAR5 cells differed according to their expression of
claudin 4 shortly after seeding. The parental NIH:OVCAR5 ovarian cancer cell line expressed high
levels of claudin 4 and was able to form compact spheroids in vitro after 24 hr in culture, while the
spheroids formed from NIH:OVCAR5 cells treated with shRNA targeting claudin 4 remained as

Page 11

Int. J. Mol. Sci. 2011, 12


1344
loosely associated aggregates. The time required to form “true spheroids” (defined as tight round,
regular, large, nonpermeable structures) increased in the absence of claudin 4 from about 24 hr for
NIH:OVCAR5 cells to over 60 hr for NIH:OVCAR5 cells treated with shRNA targeting claudin 4.
Although the differences in spheroid structure between NIH:OVCAR5 cells and NIH:OVCAR5 cells
treated with shRNA diminished over time, the size of the spheroids after 72 hr suggests that increased
levels of claudin 4 expression contribute to compact spheroid formation. In addition to the parental
NIH:OVCAR5 cells, an empty vector control could also have served as a negative control; however,
the Western blot in Figure 7A shows similar levels of beta actin in the parental and shRNA cells,
suggesting that the shRNA does not have any generalized “off-target” effects in NIH:OVCAR5 cells.
Figure 7. (A) Western blot showing claudin 4 expression levels in ovarian cancer cell lines
engineered to express different levels of claudin 4 (10 µg protein/lane): MA148 cells
transfected with claudin 4; MA148 cells transfected with an empty vector; NIH:OVCAR5
cells; and NIH:OVCAR5 cells treated with shRNA targeted to claudin 4. (B) In vitro
cultured spheroids from cell lines engineered to express different levels of claudin 4.
Spheroids were cultured for 48 hours, then stained with an antibody against claudin 4
followed by a FITC-conjugated secondary antibody (green). Nuclei were stained with
DAPI (blue). Top left, MA148 cells transfected with an empty vector; bottom left, MA148
cells transfected with claudin 4; top right, NIH:OVCAR5 cells; and bottom right,
NIH:OVCAR5 cells treated with shRNA targeted to claudin 4. Bar = 50 µm.

Our analysis of the time to spheroid formation shows that claudin 4 expression contributes to
compact spheroid structure. Sodek et al. previously showed that the formation of compact spheroids
by ovarian cancer cells was associated with contractile behavior and an invasive phenotype [8]. Cancer
cells grown as spheroids are also known to be chemoresistant, which is due, in part, to their structure
[9]. Perhaps claudin 4 and tight junctions contribute to this by functioning as a barrier to
chemotherapy. Alternatively, cell-cell adhesion could activate prosurvival signaling in spheroids

Page 12

Int. J. Mol. Sci. 2011, 12


1345
[10,45–47]. Interestingly, claudin 4 was identified in a proteomic analysis of chemoresistance in
ovarian cancer as one of 58 proteins that were overexpressed in cisplatin resistant cells [48].
Figure 8. Time course of spheroid formation in NIH:OVCAR5 cells. Spheroids were
formed in vitro using the liquid overlay method. Cells were grown in 96-well plates, with
2400 cells plated per well. Spheroid formation was observed at intervals over 72 hr.
Experiments were repeated at least twice. Bar = 500 µm.

The time required for MA148 cells to form spheroids was not dependent upon the presence of
transfected claudin 4 (data not shown). The ability of MA148 cells to form spheroids in the absence of
claudin 4 suggests that a redundant system of cell-cell adhesion may be used for spheroid formation.

Page 13

Int. J. Mol. Sci. 2011, 12


1346
Other claudins, especially claudin 3, are also overexpressed in ovarian cancer and could be part of this
redundancy [11,13–15,32,38,44,49].
2.7. Claudin 4 Decreases Paracellular Permeability
We also examined the paracellular permeability of spheroids expressing different levels of claudin 4
by testing the ability of spheroids formed in vitro to exclude FITC-dextran. At early time points
(1–3 days), more compact spheroids capable of dye exclusion were formed in NIH:OVCAR5 cells
compared to NIH:OVCAR5 cells treated with shRNA targeting claudin 4 (Figure 9); suggesting that
claudin 4 levels are related to paracellular permeability and tight junction barrier function in spheroids.
No difference in dye exclusion was observed at 8 days (data not shown). Differences in paracellular
permeability between the empty vector and the claudin 4 transfected MA148 cells were more subtle.
Slight differences in dye infiltration were observed, but overall the 3D structure and tightness of the
spheroids did not appear to be affected (data not shown).
Figure 9. Paracellular permeability of cultured spheroids expressing different levels
of claudin 4. (A) Photograph taken on fluorescent microscope en face of NIH:OVCAR5
spheroids incubated in FITC-dextran. (D) Photograph taken on fluorescent microscope en
face of NIH:OVCAR5 spheroids treated with shRNA targeting claudin 4 incubated in
FITC-dextran. Spheroids in (A) that express high levels of claudin 4 are tighter than those
in (D) that express low levels of claudin 4. The fluorescence level of the perpendicular (Z)
planes indicated by the horizontal yellow lines in (A) and (D) are shown in panels (B) and
(E); providing a cross-sectional view of the tight aggregates (A, B) vs. the loose aggregates
(D, E). Bar = 20 µm. Panels (C) and (F) show fluorescence profiles of spheroids in panels
(B) and (E) as described in the Experimental section.

Page 14

Int. J. Mol. Sci. 2011, 12


1347
Paracellular resistance in ovarian cancer cell monolayers has been shown to be directly related to
levels of claudin 4 expression [37]. Further studies have shown that phosphorylation of claudin 4
decreases the assembly of claudin 4 in tight junctions, thereby enhancing paracellular permeability
[50,51]. In colon cancer cells, overexpression of claudin 4 decreased paracellular permeability and
increased invasiveness [33]. Together with the observation that compact ovarian cancer spheroids are
more invasive than diffuse spheroids [8], our results suggest that increased claudin 4 expression could
be associated with invasiveness in ovarian cancer as well. Again, our TMA findings that claudin 4
protein was expressed by the majority of ovarian cancer tissues suggest that claudin 4 may serve as a
target for therapy.In the past, spheroids have been shown to play an important role in ovarian cancer
dissemination and invasion [5,6,8,52,53] and may also contribute to chemoresistance [9,54]. In this
work, we showed that cells expressing high levels of claudin 4 were able to form compact spheroids
more rapidly than cells with lower levels of claudin 4 expression, and paracellular permeability was
increased in spheroids expressing reduced levels of claudin 4. These results suggest that claudin 4 may
mediate chemoresistance in spheroids by increasing tight junction barrier function, and implicate
claudin 4 as a target for therapy. The ability of the C-terminal fragment of the Clostridium perfringens
enterotoxin (CPE), a polypeptide that causes food poisoning and binds to claudin 4 as a cellular
receptor [55,56], to disrupt tight junction formation and increase paracellular permeability has been
shown in embryogenesis [57] and ovarian cancer cell lines [37], and could potentially be used to
increase the sensitivity of ovarian cancer cells to standard chemotherapy [58].
3. Experimental Section
3.1. Reagents
Cell culture media and supplements were purchased from Invitrogen Corporation (Carlsbad, CA)
unless otherwise stated. Chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless
otherwise stated.
Antibodies used were mouse anti-human claudin 4 (clone 3E2C1; Invitrogen), mouse anti-human
-actin (clone AC-74; Sigma-Aldrich), and normal mouse IgG (clone 3-5D1-C9; AbCam). Secondary
antibodies used were FITC-conjugated goat anti-mouse IgG + IgM (Roche Diagnostics, Indianapolis,
IN), stabilized horseradish peroxidase-conjugated goat anti-mouse IgG (Thermo Fisher Scientific,
Rockford, IL), and biotinylated horse anti-mouse IgG (Vector Laboratories, Burlingame, CA).
3.2. Cell Lines
Ovarian cancer cell lines SKOV3, ES-2, NIH:OVCAR3, HEY, C-13, OV2008, OVCA429,
OVCA433, A2780-S, and A2780-CP (provided by Dr. Barbara Vanderhyden, University of Ottawa,
Canada), NIH:OVCAR5 (provided by Dr. Judah Folkman, Harvard Medical School, Boston, MA),
CAOV3 (provided by Dr. Robert Bast Jr., University of Texas, Houston, TX), and MA148 (provided
by Dr. Sundaram Ramakrishnan, University of Minnesota, Minneapolis, MN) were maintained as
previously described [6,59–61]. SKOV3, ES-2, and OVCA429 cell lines were derived from clear cell
carcinomas; OV2008 and C-13 cell lines were derived from endometrioid tumors; NIH:OVCAR3,

Page 15

Int. J. Mol. Sci. 2011, 12


1348
NIH:OVCAR5, OVCA433, CAOV3, HEY, MA148, A2780-S, and A2780-CP cell lines were derived
from serous adenocarcinomas [59,60].
Immortalized normal ovarian surface epithelial (NOSE) cell lines 1816-575, 1816-686, IMCC3,
IMCC5, and HIO117 (provided by Dr. Patricia Kruk, University of South Florida, Tampa, FL), and
IOSE-29 and IOSE-80 (provided by Dr. Nelly Auersperg, University of British Columbia, Vancouver,
BC, Canada) were also maintained as described [62,63]. Cells were maintained in a humidified
chamber at 37 °C with 5% CO2 and were routinely subcultured with trypsin/EDTA.
3.3. shRNA Knockdown of Claudin 4
NIH:OVCAR5 cells were stably transfected with shRNA clone TRCN0000116631 (Open
Biosystems, Huntsville, AL) plasmid DNA using Lipofectamine 2000 (Invitrogen) according to the
manufacturer’s instructions.
3.4. Transfection of MA148 Cells with Claudin 4
The claudin 4 coding sequence was amplified from CAOV3 total RNA using the Access one step
RT-PCR kit (Promega, Madison, WI) with primers; Forward, AGATCTATGGCCTCCATGGGG;
Reverse, TCTAGATTACACGTAGTTGCTGGCAGC, and cloned into the TA cloning vector pCR2.1
(Invitrogen) according to the manufacturer’s instructions. The claudin 4 coding fragment was excised
from pCR2.1 by digestion with BglII and XbaI and ligated into the pcDNA3.1 expression vector
(Invitrogen), and the sequence and orientation were verified by sequencing with vector primers. The
pcDNA3.1-claudin 4 plasmid was transfected into MA148 cells using Lipofectamine 2000 (Invitrogen)
according to the manufacturer’s instructions. Stable clones were selected with neomycin.
3.5. Tissue Samples
Snap-frozen tissue samples and formalin-fixed, paraffin-embedded (FFPE) tissue blocks were
obtained from the University of Minnesota Tissue Procurement Facility (TPF) after IRB approval.
Snap-frozen tissues were used for isolation of RNA and protein; FFPE tissues blocks were used to
optimize immunohistochemical staining. The seven snap-frozen ovarian cancer tissues used for RNA
and protein analysis were derived from the primary ovarian tumors of women with stage III/IV ovarian
cancer of the serous subtype. The five snap-frozen normal ovarian tissues were obtained from patients
with benign leiomyomas, endometriosis, benign peritubal cysts, or other non-ovarian diseases. For
immunohistochemistry, 33 serous tumors, 4 clear cell tumors, and 21 normal ovaries were examined.
All tissue samples underwent strict quality control measures prior to use in these studies. Namely,
tumors were diagnosed by a pathologist at the time of surgery using OCT embedded tissue. The
following day, the FFPE H&E slides were reviewed by a pathologist to confirm the accuracy of the
diagnosis. A third pathologist reviewed the quality control H&E slides of all TPF cases to confirm the
diagnosis of the samples prior to distribution to researchers. Additionally, a pathologist (S.E.P.)
reviewed the slides while scoring the IHC staining.