Longitudinal Claudin Gene Expression Analyses in Canine Mammary Tissues and Thereof Derived Primary Cultures and Cell Lines

Human and canine mammary tumours show partial claudin expression deregulations. Further, claudins have been used for directed therapeutic approaches. However, the development of claudin targeting approaches requires stable claudin expressing cell lines. This study reports the establishment and characterisation of canine mammary tissue derived cell lines, analysing longitudinally the claudin-1, -3, -4 and -7 expressions in original tissue samples, primary cultures and developed cell lines. Primary cultures were derived from 17 canine mammary tissues: healthy, lobular hyperplasia, simple adenoma, complex adenoma, simple tubular carcinoma, complex carcinoma, carcinoma arising in a benign mixed tumour and benign mixed tissue. Cultivation was performed, if possible, until passage 30. Claudin mRNA and protein expressions were analysed by PCR, QuantiGene Plex Assay, immunocytochemistry and immunofluorescence. Further, cytokeratin expression was analysed immunocytochemically. Cultivation resulted in 11 established cell lines, eight showing epithelial character. In five of the early passages the claudin expressions decreased compared to the original tissues. In general, claudin expressions were diminished during cultivation. Three cell lines kept longitudinally claudin, as well as epithelial marker expressions, representing valuable tools for the development of claudin targeted anti-tumour therapies.


Introduction
The claudin (CLDN) protein family is a major structural and functional component of the tight junction complex in epithelial and endothelial cells [1][2][3]. Changes in the CLDN expressions were found in human and canine epithelial neoplasms of the mammary gland [4][5][6], human prostate [7,8], canine pancreas [9,10], colorectum of both species [11][12][13][14][15] and evaluated regarding the diagnostic potential of those proteins. Furthermore, CLDN proteins act as a receptor for the enterotoxin of Clostridium perfringens (Clostridium perfringens enterotoxin, CPE), a feature that enables a directed CLDN targeting for therapeutic approaches [16,17]. This feature was already used to successfully kill human CLDN-3 and -4 expressing cells in vivo and in vitro [7,[16][17][18][19]. Directed recombinant mutation of the CPE sparing the cytotoxic domain leads to a protein that allows a directed reversible binding of CLDN proteins without killing the targeted cell [19][20][21]. This enables drug delivery to CLDN expressing cells [22]. Thus, CLDN targeting has been discussed to represent a new therapeutic option for CLDN expressing tumours, although side effects appearing in CLDN expressing non-neoplastic cells have to be considered. Consequently, a potential application should be carefully chosen, e.g., intratumoural injections for local application [5,7,18,19,[23][24][25]. Tumour derived cell lines are a key tool for in vitro evaluation of CLDN targeting as a therapeutic option. Furthermore, understanding the regulation of CLDN expressions in tumour cells increases general insights in tumour biology. In general, cell lines can also serve in comparative cancer genetics and translational medicine, as reported for canine lung tumour derived cell lines, showing genetic and biochemical similarities to human lung tumours and promising effects of a drug effective in human tumours on these cell lines [26].
Several studies indicated that cells derived from human and canine tumours reduce their CLDN expressions during cultivation [4,7,18,27]. Consequently, the CLDN expression profiles of cell lines do not always reflect the expression profiles of the original tissues. Thus CLDN-expressing tumour derived cell lines, keeping the CLDN expressions, are of major value for tumour research.
This study reports comparative CLDN-1, -3, -4 and -7 expression analyses of non-neoplastic and neoplastic canine mammary gland tissue samples and thereof derived primary cultures and cell lines. The aim of the study was to identify stable CLDN expressing canine mammary gland derived cell lines, keeping the initial epithelial character.

Gene Expression Analyses: QuantiGene Plex Assay
The QuantiGene Plex Assay allowed the simultaneous multiplexing of all genes of interest (CLDN-1, -3, -4, -7, β-Actin (ACTB) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH)) in RNAs of all examined passages of the primary cultures, cell lines and available original tissue samples via amplified fluorescence detection. Furthermore, all passages of the primary cultures, cell lines and the original tissue samples were analysed in one run. CLDN expressions were normalized to the reference genes ACTB and GAPDH revealing similar results. ACTB and GAPDH were chosen as they were already used for CLDN gene expression analyses in canine prostate tumour and mammary tissue and tumour derived cell lines [27]. Furthermore, they were used for CLDN gene expression analyses of human tissues and intestinal carcinomas [13,14,28,29]. If a CLDN expression was normalized only to ACTB or GAPDH it is clearly stated within the text. Table 3. This table shows the cytokeratin (CK) distribution of the cell lines. "Histological Classif." refers to the histological classification of the original tissue samples from which the cell lines were derived, "p.n" refers to a certain passage, "AE1/AE3" refers to Pan-CK (CK1-8, CK10, CK13-16, CK19),  "MNF116" refers to Pan-CK (CK5, CK6, CK8, CK17, CK19), "HMW" refers to Pan-CK (CK1, CK5, CK10,  CK14). "+" refers to positivity, "(+)" to a weak positivity and "−" to negativity for protein distribution. "NA" refers to "not available".
CLDN Gene Expression Alterations from Tissue Samples to Passage 00 Alterations of the CLDN expressions presumably result from dissociating cells from the tissue matrix in the process of exposing them to cultivation. These alterations were analysed in three primary cultures and three cell lines by comparing the gene expressions of the original tissue samples and the corresponding first passage (p.00).

CLDN Gene Expressions in the Latest Analysed Passage
High CLDN gene expression levels (cut off was set at level 0.1, referring to level 1 as gene expression of the reference genes ACTB and GAPDH) in the latest analysed passage were displayed for CLDN-1 by four cell lines (T124, DT14/04T, T120A and DT14/10), for CLDN-3 by two cell lines (T124 and T120A) and for the CLDNs-4 and -7 by three cell lines (T124, DT14/04T and T120A). For details see Figure 2.

CLDN Gene Expressions in the Latest Analysed Passage
High CLDN gene expression levels (cut off was set at level 0.1, referring to level 1 as gene expression of the reference genes ACTB and GAPDH) in the latest analysed passage were displayed for CLDN-1 by four cell lines (T124, DT14/04T, T120A and DT14/10), for CLDN-3 by two cell lines (T124 and T120A) and for the CLDNs-4 and -7 by three cell lines (T124, DT14/04T and T120A). For details see Figure 2.

Discussion
Mammary tumours commonly occur in female dogs [30]. Treatment options are restricted to invasive options such as surgery and, additionally, chemotherapy in the case of metastasis [31]. Retrospective studies exploring cytotoxic agents are mostly restricted to a limited number of dogs, included in the respective studies. Complex prospectivly conducted studies evaluating experimental therapeutic protocols are practically missing [32]. However, the establishment of non-invasive therapeutic approaches for mammary carcinomas is currently addressed in various ways. A recent study reported a non-invasive treatment option for mammary tumours in dogs and cats using plasmonic photothermal therapy [33]. Due to disease similarities between humans and dogs, similar environmental conditions, but easier performable trials (guidelines are less strict than for humans), dogs are considered to represent an excellent model for human diseases [34]. Accordingly, data obtained by canine tumour research may also lead to progress in human tumor research. Further experimental therapeutic approaches targeting human cancer cells include the use of the enterotoxin of Clostridium perfringens (CPE) in order to kill cancer cells by directed CLDN binding. In a xenografical mouse model, the intratumourally injected CPE reduced the size of CLDN expressing tumours initiated by a human breast cancer cell line [5]. A recombinant CPE-tumour necrosis factor fusion protein inhibited growth of CLDN expressing human ovarian tumour cells [22]. Human ovarian tumour cells overexpressing CLDNs-3 and -4, injected into mice, were selectively killed by the administration of CPE [18]. Further, CLDN-3 and -4 expressing cells from a human prostate carcinoma bone marrow metastase were killed in vitro by adding the CPE to the culture medium [7]. Additionally, mutated recombinant CPE fragments are in experimental evaluation as for example the non-toxic but CLDN binding C-terminal fragment (C-CPE), which is able to inhibit the growth of CLDN expressing tumour cells [5,16,22,24]. A combination of the C-CPE and taxol enhanced chemosensitivity of human ovarian cancer cells and inhibited tumour growth of such cells in a xenografic mouse model [24].
In order to develop specific CLDN targeting therapeutic approaches for canine mammary gland tumours, the availability of CLDN expressing cell lines is a precondition. In the present study, comparative CLDN gene and protein expression analyses on eleven cell lines and six primary cultures derived from canine mammary gland non-neoplastic and neoplastic tissues were performed. In the majority of the herein analysed cell lines and primary cultures (fourteen of seventeen) the CLDN-1, -3, -4 and -7 gene expressions decreased after transfer to culture conditions when compared to the original tissue samples. Furthermore, CLDN expressions remained low during longitudinal cultivation over 30 passages. These results match our previously published study, showing low to missing CLDN expression in long term established canine mammary non-neoplastic and neoplastic tissue derived cell lines ZMTH3, MTH53A and MTH52C [27]. Further studies evaluating CLDN expression in different tumour types reported a decrease or lack of CLDN-3, -4 and -7 expressions of in vitro cultured cells compared to uncultured cells [4,7,18]. Human prostate carcinoma metastases derived cell lines PC3 and LNCaP lack CLDN-3 and -4 expressions, whereas human non-neoplastic prostate epithelium expresses the CLDNs-3 and -4 [7]. A 1000-fold lower CLDN-7 expression was found in cultivated human mammary endothelial cells (HMECs) compared to uncultured HMECs [4]. A comparable decrease regarding CLDN-3 and -4 gene expression patterns was reported for cultured human ovarian cancer cell lines that show a lower CLDN expression than primary ovarian tumours [18]. The reason for CLDN expression or repression for the tumour cells is not fully understood. It has been proposed that the loss of CLDNs allows the loss of cell polarity, thereby promoting the tumorigenesis [35]. Metastasis is a process linked to advanced tumour stage, in which cells dissociate from a tumour, enter blood or lymphatic vessels, attach in distant sites and form new tumours [36]. Epithelial-to-mesenchymal transition (EMT) is linked to metastasis [37] and it is suggestive that loss of CLDNs contributes to this process [12,29]. However, the role of CLDNs in EMT is not yet elucidated. As in cultured cells, the complex three dimensional structure of tissues is non-existent, the observed down regulation can be caused by the loss of required CLDN functions.
However, in three of the herein evaluated new established canine mammary cell lines (originating from respectively a lobular hyperplasia, a carcinoma and an adenoma) expression of the CLDN genes increased comparing an early passage and p.30, but did not reach constant expression levels before p.30. Further, these cell lines kept the epithelial character throughout the observed cultivation period. To the best of our knowledge, this is the first study to longitudinally analyse the CLDN expressions up to passage 30. Comparisons to other studies are difficult as information about the analysed passages are missing.
Correlation coefficient analyses of the three CLDN expressing cell lines identified in the present study indicated that culturing conditions do not affect all CLDNs in the same manner. These findings might be associated to the functional ability of different CLDNs to specifically heterophilically interact with each other. CLDN-3 is able to interact heterophilically with CLDN-1 or CLDN-2, whereas the latter are not able to interact with each other [38]. Immunofluorescence revealed that CLDN-1 and -4 proteins in cell line T124 were located in the cytoplasm, whereas the CLDN-3 and -7 proteins were located at the cell membranes, where the CLDN proteins are supposed to be located [39]. Such apparent mislocalizations were also described for the CLDN-3 protein in human breast cancer derived cell lines and may be related to local invasiveness [40]. These findings should receive attention in studies evaluating the CLDNs as a target for anti-tumour therapies.
In summary, the data reported here show that cultivation of canine mammary cells has negative effects on the expression of different CLDNs. However, three epithelial canine mammary lobular hyperplasia, adenoma and carcinoma derived cell lines kept the expression of the analysed CLDN genes and proteins during cultivation until p.30. These three canine mammary gland tissue derived CLDN expressing epithelial cell lines represent valuable tools for further studies regarding CLDN targeted tumour therapeutic approaches.

Canine Mammary Gland Tissue Samples
Five normal canine mammary gland tissue samples, two lobular hyperplasias, one simple adenoma, one complex adenoma, two simple carcinomas, three complex carcinomas, one carcinoma arising in a benign mixed tumour and two benign mixed mammary tumours were provided by the Small Animal Clinic, University of Veterinary Medicine Hannover, Germany, the Institute of Veterinary Medicine, Georg-August-University Göttingen, Germany, and the veterinary practice of Dr. M. Schilling, Bielefeld, Germany; in total 17 canine mammary gland tissue samples.
Samples for diagnostic procedures were stored in formalin. Histological classification was performed in the Department of Pathology of the University of Veterinary Medicine Hannover, Germany, according to the proposed classification of canine mammary tumours [41]. Depending on the size of the tissue samples, portions for gene expression analyses were shock frozen in liquid nitrogen and stored at −80 • C until RNA was isolated using the RNeasy ® Mini Kit (Qiagen, Hilden, Germany). For details see Table 1.

Establishment of Primary Cultures
Non-neoplastic and neoplastic tissue samples were stored in Hanks Medium (Hank's salts L201-00, Biochrom GmbH, Berlin, Germany), 2% penicillin and streptomycin (Biochrom GmbH, Berlin, Germany) until being processed. A small portion of the initial tissue sample was chopped into small pieces followed by collagenase treatment (0.35% Collagenase NB 8 Broad Range, SERVA Electrophoresis GmbH, Heidelberg, Germany) at 37 • C until cell connections dissolved, for a maximum of 4 h. The dissociated cells were washed with medium 199 (Gibco by Life technologies TM , Darmstadt, Germany) containing 20% fetal calf serum (FBS Superior, Biochrom GmbH, Berlin, Germany) and 200 IU/mL penicillin and 200 mg/mL streptomycin (Biochrom GmbH, Berlin, Germany). Afterwards, cells were transferred into sterile 25 cm 2 flasks containing 5 mL medium 199 (20%). The cultures were incubated in 5% CO 2 /air at 37 • C. During the first days, the cells were monitored on a daily basis observing cell growth, depending on respective characteristics. The culture medium was changed at least twice per week. Reaching confluency, the cells were passaged by adding 1 mL TrypLE TM Express ([-] Phenol Red, Gibco by Life technologies TM , Darmstadt, Germany) and then the cell suspension was split and transferred to sterile cell culture flasks, containing 5 mL medium 199 (20%) per flask.
The cells dissolved from the initial tissue sample were herein indicated as passage (p.) 00 (p.00). After first splitting, the cells were termed p.01, higher numbers indicate the number of the certain passage. The term "primary culture" referred to cell cultures between p.00 and p.19. Reaching p.20 and above, the certain "primary culture" was furthermore designated as "cell line".
Primary cultures were analysed in p.00 and additionally in another early passage (p.01, p.02). In case cell growth discontinued before the tenth passage, the latest viable passage was analysed. In case cells grew beyond p.10 but discontinued growth before p.19, then p.10, and additionally the latest viable passage, were analysed.
Cell lines were analysed in p.00, in an early passage (p.01-p.03) and then in p.10, p.20 and p. 30. Depending on the cell number of a passage, pellets were generated according to Table 1 and stored adequately.

Generation of Cell Pellets
The cells were detached by adding 1 mL TrypLE TM Express ([-] Phenol Red, Gibco by Life technologies TM , Darmstadt, Germany) to the flasks. The cells were counted using Cellometer SD100 cell counting chambers (Nexcelom Bioscience, Lawrence, MA, USA) and Cellometer TM Auto T4 (Nexcelom Bioscience), pelleted at 1000 g for 10 min and stored adequately until further usage.
For histological analyses, pellets were stored in 4% paraformaldehyde. For gene expression analyses, using either QuantiGene Plex Assay (Affymetrix, Santa Clara, CA, USA) or conventional PCR, cell pellets were stored at −80 • C. 5-30 mg of frozen tissue samples were homogenized using the Tissue Lyser II (Qiagen, Hilden, Germany) and 5 mm Stainless Steel Beads (Qiagen) according to the manufacturer's instructions (Qiagen). RNA was isolated using the RNeasy ® Mini Kit (Qiagen) including a digestion step for genomic DNA using the RNase-Free DNase Set (Qiagen) according to the manufacturer´s instructions. An additional step for digestion of genomic DNA was performed using RQ1-DNAse (Promega, Mannheim, Germany) according to the manufacturer's instructions. The total amount of RNA was quantified using Synergy 2 (Biotek, Bad Friedrichshall, Germany) controlled by Gen5™ Reader Control and Data Analysis Software. RNAs were stored at −80 • C until usage.

RNA-Isolation from Cultured Cells for Gene Expression Analyses Using Conventional PCR
Cells were homogenised using QIAshredder TM columns (Qiagen). RNA was isolated using the RNeasy ® Mini Kit (Qiagen), RNase-Free DNase Set (Qiagen) and RQ1-DNAse (Promega) as described in Section 4.1.5. Measurement of the amount of RNA and storage was performed as described in Section 4.1.5. For each run, background controls and process controls were performed as duplicates. As process controls, samples that were to be analysed were chosen. Samples were run as unicates. Normalization was performed towards two reference genes, ACTB [13] and GAPDH [42]. Available tissue samples and passages of the corresponding cell cultures were analysed according to Table 1.

Gene Expression Analyses
Probes for the QuantiGene Plex Assay (Affymetrix) were ordered based on the mRNA sequences provided by the National Center for Biotechnology Information (NCBI) and custom designed by Affymetrix (Santa Clara, CA, USA). The accession numbers for the mRNA sequences were as follows: CLDN Expression in the Latest Analysed Passage CLDN expression was declared as "high", if the expression level in the latest analysed passage was higher than cut off 0.1, as "low", if the expression level was lower than cut off 0.1. Normalization was performed towards the reference genes ACTB and GAPDH, all values refer to the reference gene value 1.

Linear Correlation of CLDN Expression during Cultivation
The correlation coefficient according to Pearson is an empirical correlation coefficient and can be applied to calculate linear correlations between two time-dependent gene expression profiles, which were obtained from the cultivation of cell lines T120A, T124 and DT14/04T. The correlation coefficient is defined by the following equation: in which x and y are passage series for two different CLDNs of the same cell line. "xm" and "ym" are the mean values of the CLDN gene expression data of a passage series in a certain cell line. For a better visibility of correlations, each value of each CLDN gene expression profile was scaled with a factor. The value of the factor is equal to the mean value of the expression time series such that all scaled expression profiles have the same mean value.

CLDN Gene Expression Analyses Using Conventional PCR
cDNA-Synthesis cDNA was synthesised using M-MLV-Reverse Transcriptase (Promega), 500 ng of total RNA and AP2-Primer according to the manufacturer´s instructions. Two negative process controls were included. To check reagents for contamination, a negative control was performed once for each run that included the reverse transcriptase but no RNA, so that no cDNA was synthesisable from sample RNA. To check the RNA for contamination, an additional negative control was performed for each sample and included RNA but no reverse transcriptase, so that cDNA could not be synthesised from the RNA. Genomic DNA contamination was excluded in cDNA and negative controls using a PCR assay (CXCR4 upVIIIa/CXCR4 lo VIII) for an intron sequence of the reference gene C-X-C chemokine receptor type 4 (cxcr4) which can only be found on genomic DNA, but not on cDNA. Integrity of the RNA and thereby cDNA quality of all samples was tested using a PCR assay for the reference gene ACTB (ACTB up/ACTB lo). cDNA was further analysed if contaminations were excluded in the cDNA and negative controls and if integrity of RNA was verified. Primer sequences have been published previously [27].

Immunocytochemistry
For immunocytochemical characterisation of the cytokeratin distribution in the cell cultures, cell pellets from early and late passages of the cultured cells were detached with TrypLE TM Express Enzyme (Thermo Fischer Scientific, Darmstadt, Germany) followed by mechanical scraping (Cell Scraper M, TPP, Trasadingen, Switzerland) if enzymatic detachment was incomplete. Detached cells were centrifuged at 1000× g for 10 min, fixed in 4% paraformaldehyde and embedded in paraffin wax. Pellets of all primary cultures and cell lines were immunocytochemically stained with a panel of antibodies detecting various cytokeratins (CK), for details see Tables 2 and 3. In addition, pellets of the primary culture DT14/04R and cell lines T120A, T121, T124, DT14/04T, DT14/06T and DT14/10 were further investigated using antibodies specific for the CLDN-1, -3, -4 and -7 proteins according to Table 5. Immunocytochemistry was performed on serial sections according to standard procedures as described previously [43]. Briefly, antigen retrieval was achieved by incubation in a citrate buffer (pH 6.0, 20 min, 95 • C) followed by inhibition of endogenous peroxidase activity with 0.5% H 2 O 2 in 70% ethanol for 30 min, and blocking of non-specific binding with inactivated goat serum (diluted 1:5 in phosphate-buffered saline, PBS, pH 7.2). Primary antibodies were applied for 1 h at ambient temperature (for anti-CLDN antibodies) or overnight at 4 • C (for anti-CK antibodies) followed by incubation with biotinylated goat anti-mouse or goat anti-rabbit secondary antibodies (Vector Laboratories, Burlingame, CA, USA), respectively. Subsequently, avidin-biotin-peroxidase reagent (Vector Laboratories) was used according to the manufacturer's instructions for 30 min and for anti-CLDN antibodies amplification was achieved using biotinylated tyramine as described elsewhere [44]. Supplemented with H 2 O 2 , 3,3 -diaminobenzidine (Sigma Aldrich, Munich, Germany) was applied to induce a brown colour reaction, and sections were counterstained with Mayer's haematoxylin. Table 5. List of antibodies used for immunocytochemistry (ICC) and immunofluorescence (IF). Staining pattern of all antibodies is identical to that described in the literature. "*" cross-reactivity with canine according to manufacturer's information, "**" the canine KRT14 gene is a homolog of the human KRT14 gene (National Center for Biotechnology Information HomoloGene). Cross reactivity of the antibody is highly presumable as amino acid sequences of the human (NP_00517.2) and canine (NP_001240670.1) cytokeratin 14 reveal an identity of 97% (according to National Center for Biotechnology Protein Blast).

Immunofluorescence
Immunofluorescence was performed for cell cultures T120A, p.37-42; T124, p.15; DT14/04T, p.4 and p. 16-19; and DT14/06T, p.44-49. The cells were cultivated on collagen coated cover slips. After achieving a confluency of about 75%, the cells were fixed with 4% formaldehyde for 10 min at room temperature and permeabilized with 0.3% Triton X-100 in PBS for 15 min at 37 • C. BSA (bovine serum albumin, 1%) in PBS was used for 30 min at 37 • C to block and saturate the non-specific binding positions. Then the cells were stained over night at 4 • C for CLDN-1, -3, -4 and -7 proteins with the respective primary antibodies according to Table 5. The secondary fluorescein conjugated anti-rabbit (Merck Millipore, Darmstadt, Germany) and fluorescein isothiocyanate (FITC) conjugated anti-mouse (Merck Millipore) were diluted 1:100 in PBS containing 1% BSA and added to the cells with 2 µM DAPI (Sigma-Aldrich, Munich, Germany) for 1 h at 37 • C. The cells were stored in PBS at 4 • C for further analysis. As a control for unspecific binding sites, cells were also incubated only with the secondary antibodies. The fluorescent images of the cells were collected with a Nikon Eclipse TE2000-E confocal laser scanning microscope (400 nm for Dapi and 488 nm for the CLDN proteins), with a 60× water immersion objective and EZ-C1 3.80 software program (Nikon, Düsseldorf, Germany). Image processing was performed using the ImageJ plugin named FigureJ [51].