Background
Colorectal cancer (CRC) is one of the major causes of cancer-related death in the Western world. When localized, CRC is often curable by surgery, but the prognosis of patients with metastatic CRC (mCRC) remains very poor [
1]. Standard chemotherapy regimens (FOLFIRI or FOLFOX) used in combination with targeted therapies have improved response rates and survival [
2‐
4]. However, 30 to 50% of patients are intrinsically resistant to treatments and nearly all patients who are initially responsive will eventually develop resistance. Therefore, other first-line treatment options for patients with mCRC are required.
Claudins (CLDN) are integral membrane proteins that determine the barrier features of tight junctions [
5,
6] and are considered potential therapeutic targets for antibody-based treatments [
7]. They have four transmembrane domains, two extracellular loops and cytoplasmic tails [
8,
9]. Claudin-1 (CLDN1) is one of the 27 members of the claudin family [
10]. Several groups have reported increased expression of CLDN1 in primary CRC and metastases as well as in CRC cell lines but with marked differences in its localization [
11‐
14]. Indeed, immunohistochemistry analyses in CRC and normal mucosa samples showed that cancer cells display a membranous [
11] or membranous/cytoskeletal [
12] staining of CLDN1. Other studies found a strong CLDN1 expression at the cell–cell boundaries and in the cytoplasm of cancer cells [
13]. Moreover, CLDN1 nuclear localization in tumor samples was reported [
14]. Aberrant CLDN1 expression in tumor cells can lead to alteration of the tight junction structure and function, or dysregulation of cell signaling pathways [
15]. CLDN1 could be involved in WNT and NOTCH signaling. Indeed, CLDN1 is a known target of TCF/LEF signaling [
13], but seems also to participate in the regulation of the WNT signaling pathway [
16]. Through NOTCH signaling upregulation, CLDN1 has a role in colon epithelium homeostasis [
17] and in promoting colon tumorigenesis [
18].
Much effort has been focused on understanding CLDN1 function and regulation in cancer; however, little is known about its potential usefulness as a therapeutic target. Therefore, the aim of this study was to accurately assess CLDN1 expression in CRC samples and to determine the potential of CLDN1 as a therapeutic target for antibody-based treatments in CRC.
Methods
CRC samples
For gene expression profiling, we selected 143 tumor samples from 143 patients included in three cohorts: the prospective single-center study REGP (19 patients) (GSE62322) [
19,
20], the retrospective multi-center study COSIVAL (68 patients) and the prospective multi-center study BIOCOLON (56 patients) (GSE62080 and GSE72970) [
21]. For these three studies, Inclusion criteria were: histologically proven colon adenocarcinoma, advanced and bidimensionally measurable tumor (stage IV), age between 18 and 75 years, and World Health Organization (WHO) performance status ≤2. Before any treatment, all patients underwent surgery for primary tumor resection or endoscopic biopsy.
For western blot analysis, 13 additional tumor samples from the prospective single-center study REGP but not included in the 143 samples were used.
For immunohistochemistry analysis, tissue samples from 52 additional patients with CRC were retrospectively selected from the Institute of Cancer Research of Montpellier pathology files only when normal mucosa, adenoma and adenocarcinoma samples were available for the same patient.
All the studies using human tissue samples were approved by the relevant ethics committees and all participants were informed about the study objectives and methods and signed a written informed consent before enrolment.
Gene expression analysis
Colon samples (normal colon, primary tumor and hepatic metastasis samples for the REG/P study, but only primary tumor specimens for the COSIVAL and BIOCOLON trials) were collected at the time of surgery following a standardized procedure to obtain high quality RNA [
22]. Samples were then hybridized to human genome U133 Plus 2.0 arrays (Affymetrix Inc., Santa Clara, CA).
To identify new therapeutic targets for antibody-based therapy in mCRC, we compared the gene expression profiles of normal mucosa (n = 17), primary tumor (n = 20) and hepatic metastases (n = 19) tissue samples.
As CRC heterogeneity must be taken into account when comparing gene expression profiles, primary tumor samples (
n = 143) were classified using the CRC molecular classifications based on gene expression profiles that have been proposed by three independent groups [
23‐
25] and the recent consensus classification [
26]. Briefly, De Sousa E. Melo et al. proposed to group tumors in three classes: CCS1 (CRC with microsatellite instability, MSI), CCS2 (cancer with chromosomal instability, CIN) and CCS3 (new subtype) [
25]. Sadanandam et al. identified five molecular subtypes, based on the cell phenotype: Goblet-like, Transit-Amplifying (TA), Enterocyte, Stem-like, and Inflammatory [
24]. Marisa et al. described six molecular subtypes (C1 to C6) with the following main features: C1 = CIN and downregulation of immune pathways, C2 = MSI, C3 = mutated KRAS, C4 = stem cell phenotype-like, C5 = CIN and upregulation of the WNT pathways, and C6 = CIN and normal-like gene expression profile [
23]. Finally, starting from six previously published signatures, an international consortium presented a classification with four consensus subtypes: MSI (CMS1), canonical (CMS2), metabolic (CMS3), and mesenchymal (CMS4) [
26] (for review [
27]). The CRC sample distribution according to the molecular subtype is shown in Additional file
1: Table S1.
Immunohistochemistry analysis
Samples were assembled in a tissue micro-array (TMA) using three tissue cores (0.6-mm diameter each) as previously described [
28]. Briefly, 3-μm sections of the TMA were de-paraffinized and rehydrated in graded alcohols. Heat-induced antigen retrieval was performed by incubating TMA sections in EDTA buffer (pH 9) at 98 °C in a water bath for 20 min. After neutralization of the endogenous peroxidase activity, TMA sections were incubated with a polyclonal anti-CLDN1 antibody (JAY-8, Zymed laboratories Inc, CA, USA) or antibody diluent (Dako, Glostrup, Denmark) alone for 60 min. Primary antibody binding was visualized using the Envision® system and the Dako Autostainer® (Dako, Glostrup, Denmark). The percentage of CLDN1-positive cells and the staining intensity (0, no staining; 1, yellowish; 2, brown; and 3, dark brown) were evaluated for each individual TMA spot.
Western blot analysis
Tumor tissue samples from patients were directly grinded in lysis buffer (150 mM NaCl, 10 mM Tris pH 7.4, 1 mM EDTA, 1 mM EGTA, 1% SDS, 1% Triton X-100, 0.5% NP-40, 2 mM PMSF, 100 mM NaF, 10 mM sodium orthovanadate, one cocktail protease inhibitor tablet for 10 ml) using a Mixer Mill® MM 300 unit (Qiagen, Valencia, CA). Protein concentration was determined with the Bradford assay (Pierce Coomassie Plus Protein Assay). Then, 50 μg of total proteins were resolved by 12% SDS-PAGE and transferred onto nitrocellulose membranes (Whatman® Protran®, pore size 0.45 μm). Non-specific binding sites were blocked with 5% (wt/vol) nonfat milk in PBS with 0.1% (vol/vol) Tween 20 (PBS-T) at room temperature for 1 h and then membranes were incubated at 4 °C with a polyclonal anti-CLDN1 antibody (JAY-8) overnight. Membranes were then washed and incubated with the appropriate horseradish peroxidase-conjugated secondary antibody for 1 h. Revelation was performed with a chemiluminescence system (Amersham Biosciences); β-tubulin expression was used for normalization.
Protein extraction was carried out as previously described [
29]. For each sample, 20-μm thick sections were cut with a cryotome, mixed in liquid nitrogen and gently ground with a micro-pestle. For subcellular protein extraction, the ProteoExtract Subcellular Proteome Extraction Kit was used according to the manufacturer’s instructions (Calbiochem). Subcellular fractions (10 μg/each) were loaded on 12% SDS-PAGE gels. Immunoblotting was done as described above with the following primary antibodies: anti-CLDN1 (JAY-8), -CD71 (Invitrogen), -Histone H3 (Pierce) and -β-tubulin (Sigma T4026).
Cell lines
The following human CRC cell lines were used: SW480 (ATCC CCL-228), SW620 (ATCC CCL-227), Caco-2 (ATCC HTB-37), Difi [
30] (a gift from C. Montagut, Department of Medical Oncology, Hospital del Mar, Barcelona, Spain), HCT116 (CCL-247), and LS174T (ATCC CL-188). To obtain the CLDN1-positive SW480 cell line (SW480-CLDN1), SW480 cells were stably transfected with the human CLDN1 cDNA clone (Invitrogen MGC collection) or with empty vector (pcDNA) using the jetPRIME™ transfection reagent (Polyplus-transfection Inc., France). CLDN1-positive clones were selected by growing transfected cells in the presence of 500 μg/ml of geneticin. For CLDN1 silencing, SW620 was transduced with the pSIREN vector containing the shRNA against CLDN1 (SW620shCLDN1) or against luciferase (shLUC, negative control). After 24 h, cells were selected with 1 μg/mL puromycin and stable clones were pooled. All transient transfections were done using the jetPRIME™ transfection reagent.
Production of anti-CLDN1 mAb 6 F6
For antibody production, 6–8 weeks-old female BALB/c mice (Harlan, Gannat, France) were challenged by intra-peritoneal (i.p.) injection of 4 million mouse NIH cells transiently transfected with CLDN1 cDNA (NIH-CLDN1 cells) every two weeks (five injections in total). NIH-CLDN1 cells were mixed with complete Freund’s adjuvant (Sigma) for the first injection, and with incomplete Freund’s adjuvant (Sigma) for the other four injections. An intravenous booster injection of NIH-CLDN1 cells was given three months after the fifth immunization. Three days later, spleen cells from immunized mice were fused with the mouse myeloma cell line P3-X63-Ag.8.653 to produce mouse hybridomas. Supernatants from newly generated clones were screened by fluorescence-activated cell sorting (FACS) using SW480-CLDN1 and SW480 cells (negative control). The screening results were confirmed by performing and additional screening using SW620 and SW620-shCLDN1 cells. The anti-CLDN1 hybridoma 6 F6 clone was selected and cloned by limiting dilution. Antibody isotyping showed that 6 F6 was an IgG3k.
All animal experiments were performed in compliance with the French government guidelines for experimental animal studies (agreement CEEA-LR-12052).
Radiolabeling and SPECT-CT imaging
Female athymic nude mice (6–8-week-old) were purchased from Harlan. The 6 F6 mAb was radiolabeled with 125I (Perkin Elmer) at the specific activity of 370 MBq/mg for single-photon emission computed tomography (SPECT) imaging, using the IODO-GEN (Pierce Chemical Co.) method. After tail vein injection of 16 MBq/50 μg 125I-labeled 6 F6, whole-body single photon emission tomography/computed tomography (SPECT/CT) images were acquired with a 4-head multiplexing multi-pinhole NanoSPECT camera (Bioscan Inc., Washington, USA) at various times (48, 72 and 96 h). Concomitantly, whole-body micro-CT images were acquired for anatomic co-registration with SPECT data. Reconstructed SPECT and CT data were visualized and co-registered using Invivoscope®.
Clonogenic assay
Colorectal cancer cells were seeded in a 6-well plate (150, 250 or 400 cells/well) and allowed to adhere at 37 °C overnight. Then, 1 ml of RPMI with or without 6 F6 mAb (final concentration: 100 μg/ml) was added to each well and cells were cultured for six days. After six more days in medium without antibody, plates were read using a Celigo™ imaging cytometer and the “Single colony verification” application. The Celigo™ cytometer is a benchtop in situ cellular analysis system that provides images of wells using bright-field illumination (Nexcelom Bioscience, MA, USA).
Establishment of three-dimensional (3D) spheroid cultures
Ultra-low attachment, round-bottomed 96-well plates (Corning Costar) were used for spheroid formation. SW480, SW480-CLDN1 or SW620 cells were seeded at a density of 5 × 104. Cells aggregated and merged in 3D spheroids within 24–72 h. Images of wells were taken with a phase-contrast microscope using a 5× objective or captured with the Celigo™ imaging cytometer using the “Tumorosphere” application. Cell viability was assessed with the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI, USA). After addition of 100 μl of CellTiter Glo reagent to each well for 10 min, luminescence was measured on a 1450 MicroBeta TriLux Luminescence microplate reader (Perkin Elmer).
Cell cycle and proliferation analysis in spheroids
Spheroids were prepared by plating 1000 DiFi cells per well in ultra-low attachment 96-well plates, and growing them in the presence of 100 μg/ml of the 6 F6 mAb or irrelevant mAb (retuximab) for 5 days. For cell cycle analysis, cells were pelleted, trypsinized, washed with PBS, fixed in 75% ethanol, and stained with 40 μg/ml propidium iodide in the presence of 100 μg ml−1 RNAse (Qiagen). The cell cycle distribution was determined with a FC500 Beckman Coulter Flow Cytometer using the FL-3 channel. Cells were gated on a dot plot that displayed the DNA-pulse-peak vs the DNA-pulse area to exclude doublets. Cell cycle distributions were illustrated using the Flow Jo analysis software (Treestar, FLOWJO, Ashland, OR, USA).
At day 4 of culture, cell proliferation was measured by incubating cells with 5-ethynyl-2′-deoxyuridine (EdU) for 24 h. EdU is incorporated into DNA during active DNA synthesis. Then, after cell trypsinization and fixation/permeabilization in 75% ethanol/PBS, incorporated EdU was labeled and detected with the Click-iT EdU Alexa Fluor 488 Flow Cytometry Assay Kit (Invitrogen). Cells were then incubated with 1 μg/ml of 4′,6-diamidino-2-phenylindole (DAPI) in PBS/0.1% Triton X100 at 37 °C for 30 min. The Celigo™ “Expression Analysis” (Target 1 + Mask) application was used to quantify the fluorescent signal and for data analysis. Cells were identified using the DAPI nuclear stain and DNA synthesis was quantified by measuring EdU incorporation.
Mouse xenograft models
1.5 × 106 SW620 cells or 3 × 106 DiFi cells were suspended in culture medium and injected subcutaneously (s.c.) into the right flank of 6–8-week-old female athymic nude mice from Harlan. When the tumor volume reached approximately 100 mm3, mice were randomized in different groups and treated by i.p. injection of 0.9% NaCl or 6F6 mAb (15 mg/Kg per injection) twice a week for three consecutive weeks for the first experiment and thrice a week for the second experiment. Tumors were measured bi-weekly with a caliper and volumes calculated with the formula: D1 x D2 x D3/2.
Intrasplenic hepatic colonization model
In each experiment, 2 million luciferase-expressing SW620 cells (SW620-LUC cells) were injected in the spleen of 6–8-week-old female athymic nude mice. Spleen was removed 2 min after cell injection. On day 1 after injection, mice were randomly divided in two groups of 10 mice that were treated either with 15 mg/kg of 6 F6 mAb or 0.9% NaCl by i.p. injection, thrice per week. To evaluate metastatic formation and dissemination, luciferase expression was monitored by luminescence imaging after injection of luciferin (Camera Ivis Lumina II, PerkinElmer®) once per week. At week 5 after surgery, mice were sacrificed, livers were removed, photographed and macroscopically visible metastases on the liver surface were counted.
Statistical analysis
Statistical analysis was done using the STATA 13.0 software (StataCorp). For gene expression or immunohistochemistry experiments, differences between groups were analyzed using the Kruskall Wallis/Dunn’s test. Correlations between CLDN1 gene expression and progression-free survival (PFS) and overall survival (OS) were evaluated in the entire group (n = 143 patients) and according to the tumor molecular subtype. To this end, the 143 patients were divided in two groups based on the median CLDN1 gene expression (i.e., 9.75 [arbitrary units]). PFS and OS values were compared using the Kaplan-Meier method and differences between survival distributions assessed using the log-rank test.
The paired t-test was used to compare the effect of incubation with the 6 F6 mAb in in vitro experiments.
In in vivo experiments, a linear mixed regression model was used to determine the relationship between tumor growth and number of days after injection. The fixed part of the model included variables corresponding to the number of post-graft days and different treatment groups. Interaction terms were built into the model; random intercepts and random slopes were included to take into account the time effect. The coefficients of the model were estimated by maximum likelihood. Survival rates were estimated from the date of the injection until the date when the tumor reached a volume of 1500 mm3 using the Kaplan–Meier method. Survival curves were compared using the log-rank test. For the hepatic colonization experiments, differences between groups were evaluated with the Mann-Whitney U test. For all experiments, differences were considered to be significant when P < 0.05.
Discussion
In this study, we assessed CLDN1 as a potential therapeutic target in CRC. We found that CLDN1 is mainly localized at the cell membrane of CRC cells and described, for the first time, CLDN1 expression in the different CRC molecular subtypes. We also found that CLDN1 expression has a prognostic value in the C3 and C5 Marisa’s subtypes. Using a new mAb (6F6) that specifically recognizes the extracellular part of human CLDN1, we demonstrated that CLDN1 targeting reduces CRC xenograft growth and liver metastasis formation. Finally, we showed that 6F6 activity is mediated through inhibition of cell proliferation.
CLDN1 membrane overexpression in all tested primary tumor samples from patients with mCRC supports the hypothesis that CLDN1 could represent a target for antibody-based therapy. Analysis of CLDN1 expression in the new CRC molecular subtypes [
23‐
25] and the recent consensus subtypes [
26] highlighted important variations among the different CRC classes, in agreement with CRC heterogeneity. Specifically,
CLDN1 expression is significantly higher in CRC subtypes associated with marked WNT signaling activation, such as the Marisa’s C5, Sadanandam’s TA and CMS2 consensus subtypes. These findings are in accordance with the already described involvement of CLDN1 in WNT signaling [
13,
14,
16]. Moreover, we demonstrated for the first time that
CLDN1 expression level could be used for outcome prediction in patients with mCRC and that this prognostic value is dependent on the molecular subtype. CLDN1 could become a marker to guide the choice of targeted therapy in mCRC. Indeed, CRCs belonging to the Marisa’s C5 and C3 subtypes could be targeted with an anti-CLDN1 mAb, particularly the C3 subtype because it is enriched in CRCs from patients with mutated
KRAS who are not eligible to anti-EGFR targeted therapy [
33]. Moreover, the in vivo anti-tumor effect (tumor growth delay and survival improvement) of the 6F6 mAb in mice xenografted with CLDN-positive CRC cells seems to be independent of
KRAS mutational status.
In vitro, the 6 F6 mAb negatively influences cell growth when CRC cells are grown in a 3D culture system. Growing evidence indicates that cancer cells cultured in 3D spheroids are closer to in vivo models than cell cultured in 2D systems and that they might better predict the in vivo outcome [
34]. Indeed, CLDN1 is critical for maintaining cell growth in 3D, but does not affect cell growth in monolayer cultures [
35]. Our findings indicate that in 3D culture, the 6F6 mAb has a cytostatic effect on tumor cells. As both treated and untreated spheroids showed similar cell cycle distribution, this cytostatic effect reflects a global slowing down of cell cycle progression and not an arrest in a particular phase.
In the intrasplenic model of liver metastases, which results in the rapid liver colonization by tumor cells with extremely aggressive growth [
36], treatment with the 6F6 mAb decreased the number of liver metastases. These results are supported by the in vitro finding that the 6F6 mAb decreases the number of colonies and reduces cell survival in a large panel of CLDN1-positive cancer cell lines. Moreover, the 6F6 mAb affected the migration capacity of all tested CLDN1-positive CRC cell lines. In agreement, siRNA-mediated
CLDN1 knock-down in metastatic CRC cells inhibits migration [
14]. Conversely, CLDN1 overexpression increases cell motility [
37]. Our anti-CLDN1 mAb, by affecting tumor cell migration, could play an important role in the control of cancer cell invasiveness.
The efficacy of the 6F6 mAb could be improved by increasing its Fc-dependent properties. Indeed, the 6 F6 mAb is an IgG3 and, therefore, only binds with low affinity to FcgammaRI [
38] that mediates antibody-dependent cell-mediated cytotoxicity. Another approach to enhance 6 F6 mAb efficacy would be to associate a cytotoxic effector agent in order to create an antibody–drug conjugate [
39]. Before that, its toxicity in healthy tissues must be evaluated, although a recent study excluded major toxicity or side effects induced by another anti-CLDN1 mAb in mice [
40].
Acknowledgments
The authors thank Imade Ait Arsa, Véronique Garambois, Christelle Larbouret and Charles Vincent for their help with animal experiments. The animal facility staff is greatly acknowledged. We thank Patrick Balaguer for providing the SW620-LUC cell line and Hervé Watier for helpful discussions. Flow cytometry analyses were performed using equipment at the Montpellier RIO Imaging facility.