Ajuba is a LIM-domain protein which contains a unique N-terminal region, a pre-LIM region and three tandem C-terminal LIM domains [1
]. LIM domains are tandem zinc-finger structures that function as a protein-binding interface and are associated with cytoskeletal organization and signal transduction from the plasma membrane to the nucleus [4
]. LIM-domain proteins are highly conserved between species and Ajuba is most closely related to its family members LIMD1 and WTIP [5
Ajuba has been found to be significantly upregulated in several cancers such as esophageal squamous cell carcinoma, cervical cancer and colorectal cancer [6
]. In the literature, there are discrepancies as to whether Ajuba is a driver or suppressor of tumor cell proliferation. In hepatocellular carcinoma and malignant mesothelioma, Ajuba was shown to be a negative regulator of the proto-oncogene YAP and therefore was classified as a tumor suppressor [10
]. Also, Sato et al. supported its tumor suppressor function in small cell lung cancer by demonstrating that loss of Ajuba expression resulted in enhanced tumor growth [12
]. Whereas Ajuba has also been identified as a tumor promotor in cervical and colorectal cancer through positive regulation of YAP and TAZ and therefore negatively regulating the Hippo pathway [6
Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths, with metastasis leading to a 5-year patient survival rate of only 14% [14
]. The aim of this study was to investigate the role Ajuba expression in colon cancer growth and metastasis. We found Ajuba to be highly expressed in human colon cancer and its expression is negatively correlated with patient survival. We modified Ajuba expression in the cancer cell line SW480 established from a primary Dukes’ type B colon adenocarcinoma [15
] and performed high-throughput transcriptomics. Transcriptomic results were supported with mass cytometry by time-of-flight (CyTOF) and validated with biological assays. Our data demonstrates that Ajuba promotes colon cancer cell proliferation, migration and tumor metastasis in vivo.
Using high-throughput sequencing, we provide a genome wide description of how Ajuba may be involved in pathways essential for colon cancer growth and potential for metastasis. Throughout our study, we observed stronger changes with the shRNA lentiviral construct targeting the coding region of Ajuba (shAjuba1) compared to the one targeting the non-coding region (shAjuba2).
The differences between the two constructs was evidenced with data describing Ajuba mRNA and protein expression, outcomes in biological assays, and transcriptional changes. With RNA-seq we observed that with the more efficient KD there were a greater number of DEGs and both KD lines had significant overlap between the DEG. However, the same DEGs were not always affected in the two conditions, but many were involved in the same pathways, capturing different parts of the same biological process, thereby demonstrating robustness in the pathways identified.
Comparing the OE cell line with its control showed only few DEG. This may be explained by the fact that the parental SW480 cell line already expresses high amounts of Ajuba. Interestingly, the pathway enrichment analysis of both KD and OE lines identified similar pathways affected by altering Ajuba expression, thus strongly supporting its role in these processes. The main branches of pathways identified were proliferation, cell differentiation and epithelial-to-mesenchymal transition (EMT) which are all pathways known to be important in cancer development and metastasis. Proliferation pathways includes pathways such as epithelial cell proliferation, Wnt signaling and MAPK cascade. This coincides with a previous report that Ajuba is an important regulator of the Wnt signaling pathway [17
]. In addition, there are numerous reports that Ajuba is involved in the regulation of the Hippo pathway, a signaling cascade that inhibits proliferation and promotes apoptosis and is often found to be dysregulated in cancer [5
]. Finally, the cell differentiation pathways such as negative regulation of cell differentiation and epithelial cell differentiation are also well known to be dysregulated in cancer. EMT pathways include activities such as actin filament-based processes, cell projection morphogenesis, regulation of cell adhesion and wounding. Ajuba has been found to be involved in EMT and cell adhesion, a process crucial for migration and metastasis formation of tumors [8
]. During EMT, cancer cells acquire more stem cell features, for example, they lose markers of differentiation and they tend to become more mobile and invasive [25
]. And finally, to assess the role of Ajuba in differentiation we measured the stem cell marker ALDH [26
]. We found that cells with lower Ajuba expression have fewer ALDH positive cells. This is in agreement with Lang et al. demonstrating that high NFATC2 expression enhances YAP activity and promotes stemness via upregulation of Ajuba [13
We also investigated the effect of Ajuba KD using a single cell proteomics approach with CyTOF [16
]. The t
-SNE plots clustered according to antigen expression show a slight shift between the different samples. Many of the targets were not remarkably regulated however, the proteins that were differentially regulated were p53, pS6, EPHA2 and EGFR and correlated with the RNA sequencing results further supporting the involvement of Ajuba in these pathways. In agreement with Kalan et al., we observed that pp.53 was increased in KD cell lines [27
]. Ribosomal protein S6 (S6) was decreased in KD cell lines. S6 is a major substrate of different protein kinases in the ribosome such as ribosomal protein S6 kinase (S6K) [29
]. S6K acts downstream in the PI3 kinase pathway and its phosphorylation induces protein synthesis at the ribosome and thereby controls cell growth, proliferation and survival [30
]. This supports the pathway analysis of the RNA-seq data as well as the cell proliferation biological assay that decreased Ajuba expression decreases cell proliferation. This finding however could only be shown in the more efficient KD shAjuba1 and no significant decrease of proliferation rate could be observed in shAjuba2. We hypothesized, that the remaining Ajuba expression was sufficient to maintain cell proliferation. The two receptors EPHA2 and EGFR are reported to be increased in CRC patients and have a critical role in oncogenic signaling [31
]. The canonical function of EPHA2 is to inhibit cancer proliferation as well as motility whereas the non-canonical pathway EPHA2 promotes tumor survival and metastasis and drives the cells to be more dedifferentiated [34
]. Another role of EPHA2 was investigated by Dohn et al. who found EPHA2 protein to be regulated by TP53 and to induce apoptosis [35
]. EPHA2 was elevated in shAjuba1 cells supporting its potential role in the canonical function of EPHA2.
EGFR protects cancer cells from apoptosis, facilitate invasion and promote angiogenesis [36
]. Surprisingly, we observed more EGFR in shAjuba2 cells then its control which is not congruent with our biological assays in which cells with loss of Ajuba expression have decreased migration capacities. The increase of EGFR might be due to a compensatory reaction of the cells.
Our findings agree with the current state of literature describing Ajuba as pro-proliferative in colorectal cancer [8
]. We can state that only Ajuba and not its closely related LIM domain family members WTIP and LIMD1 are significantly increased in colon cancer compared to adjacent non-tumor tissues [8
]. We demonstrate that tumor samples of CRC metastasis from the liver were highly proliferative, as shown by Ki-67 staining, and Ajuba was highly expressed in the actively proliferating cells. Nevertheless, others report Ajuba to be anti-proliferative in other cancers such as malignant mesothelioma and HCC [10
]. The diverse functions of Ajuba are defined by its cellular localization. In the cytoplasm, the role of Ajuba is to stabilize cell junctions [41
], centrosome formation [42
] and to repress the Hippo Signaling pathway [5
]. We also observed focal points of Ajuba in the nucleus of colon cancer cells suggesting a nuclear role of Ajuba, supporting it reported function as a transcription factor [22
]. Ajuba does contain a nuclear export sequence and therefore can be shuttled between the nucleus and the cytoplasm [1
]. In the nucleus, Ajuba can interact with the transcription factor SNAIL to repress E-cadherin gene expression and epithelial-to-mesenchymal transition [22
In summary, Ajuba was found to be highly expressed in colorectal tumors. Its knockdown led to decreased cell proliferation, migration and colony formation and to a decreased tumor burden in a model colon cancer metastasis to the liver. Taken together, our data demonstrates the crucial role of Ajuba in driving colon cancer proliferation and its dissemination.
4. Materials and Methods
4.1. Cell Lines
Human colorectal cancer cell lines (SW480™, SW620™ and HCT-116™) were purchased from ATCC. SW480 and HCT116 are primary colon adenocarcinoma cell lines. SW480 has been classified as classified as Dukes’ B meaning that the cancer has grown through the muscle layer of the bowel. SW620 is a colon adenocarcinoma cell line isolated form a lymph node to which the primary tumor metastasized. It has been isolated from the same patient as SW480 after one year and was classified as Dukes’ C meaning that the cancer has spread to at least 1 lymph node. The cells were cultured in Dulbecco’s Modified Eagle’s Medium GlutaMAX (with 10% FBS, 100 μg/mL penicillin/streptomycin (Life Technology, Carlsbad, CA, USA) at 37 °C in a humidified incubator with 5% CO2. All lines have been tested and are negative for mycoplasma contamination using PCR Mycoplasma Test Kit).
4.2. Clinical Samples
Primary human colon tissues and CRC tumor metastases from liver were obtained from patients of the University Hospital Bern (Inselspital, Bern, Switzerland). Informed consent was obtained prior to surgery in compliance with the local ethics regulations and under approval of local ethics commission (Project-ID Nr. 2019-00157).
4.3. Public Data Acquisition
On the 24th of September 2019, colon cancer RNA-seq expression (counts) and survival data was downloaded from The Cancer Genome Atlas [47
] specifying Primary site = Colon and Projects = TCGA-COAD.
4.4. Differential Expression
Differentially expressed genes from RNA-seq samples were computed with the R package DESeq2 [48
]. Genes with adjusted p
-value < 0.05 were considered statistically significant. TCGA: Primary Tumors (n
= 471) were compared with Healthy Tissue Sample (Adjacent non-tumor tissue n = 41). For colon cancer cell lines we compared shScrambled (n
= 2) with shAjuba1 (n
= 3), shScrambled with shAjuba2 (n
= 3) and Control (n
= 3) with Ajuba OE (n
= 3). 2 samples were discarded due to metastatic and recurrent tumors (2).
4.5. Survival Analysis
The survival curves were calculated with the R function survfit from the R package survival [49
] with the formula Surv (time, vitalstatus)~categorie and plotted with the R function ggkm from the R package ggkm [50
] with options pval = T). For patients with more than one tumor, the gene expression of the multiple tumors were averaged. The data was separated in high expression (top 20%) and low expression (bottom 80%). Using interactive tools on publicly available data visualization tools such as the Human protein Atlas the effect of different cutoff threshold can also be tested. Three samples were discarded due to missing clinical information (2) and missing days of follow up information (1).
4.6. Western Blot
Total protein extraction was performed using RIPA cell lysis buffer (10 mM Tris with pH 8, 1 mM EDTA pH 8, 150 mM NaCl, 0.5% NP40) with addition of protease inhibitors (1 mM NaF, 10 mM NaVO3, 1 mM PMSD, 1X protease inhibitor cocktail—P1860, Sigma, (St. Louis, MO, USA). The cell lysates were sonicated (Sonopuls, Bandelin, Berlin, Germany) then centrifuged. Snap frozen tissue pieces were dissociated using a TissueLyser (Qiagen, Hilden, Germany) for 2 min at 20 Hz in RIPA buffer. The protein lysate concentrations were determined with the Bio-Rad Protein Assay System (Bio-Rad, Hercules, CA, USA) as described by the manufacturer. Equal amounts of proteins were separated by SDS–PAGE and transferred onto a nitrocellulose membrane using the iBlot2 Gel transfer device. The membrane then was blocked in 5% non-fat dry milk dissolved in PBS for 1 h followed by incubation with the primary antibody overnight at 4 °C. After incubation with the HRP conjugated secondary antibody, chemiluminescent reaction was performed with Western Lightning Plus-ECL from Perkin Elmer ((Waltham, MA, USA). Membranes were developed using the x-ray film processor Curix 60 (AGFA, Mortsel, Belgium). The band size was estimated using Page Ruler™ Prestained Protein Ladder (Fermentas, Waltham, MA, USA) and Precision Plus Protein™ DUAL Color Standards (Bio-Rad, Hercules, CA, USA #161-0374). Primary antibodies used were rabbit monoclonal anti-Ajuba (1:1000 dilution, Cell Signalling Danvers, MA, USA) and HRP-conjugated secondary antibodies used were goat anti-rabbit (Dako, Glostrup, Denmark). β-actin-HRP (1:100,000 dilution, Sigma, St. Louis, MO, USA) was used as a loading control.
4.7. Quantitative Real-Time Reverse Transcription PCR
Total RNA was isolated from human samples and cell lines using NucleoZOL (Macherey-Nagel, Macherey-Nagel, Dürren, Germany) according to manufacturer’s protocol. The quality and concentration of RNA were measured using Nanodrop 2000 Spectrophotometer Thermo Scientific, Waltham, MA, USA). Five hundred ng of total RNA was used for cDNA synthesis using Omniscript RT Kit 200 (Qiagen, Hilden, Germany). mRNA was analysed by quantitative RT–PCR with TaqMan gene expression assays and reagents according to the standard protocols (Applied Biosystems, Foster City, CA, USA). using specific primers and housekeeping genes 18S FAM as control. We used the TaqMAN ViiA TM 7 Real-time PCR system from Applied BioSystems for the amplification steps and data collection. Log 2-fold changes were computed using the ΔΔCt method. Ct values of target genes (TG) were calculated relative to a reference gene (RG, 18S) using the following formula: ΔCtTG = CtTG − CtRG. Experimental groups (TG) are normalized to control group (CG): ΔΔCt = ΔCtTG − ΔCtCG, and fold increase = 2−ΔΔCt.
4.8. Lentiviral Transduction
Due to the fact that the Ajuba KO cells were not viable we decided to use shRNA to knockdown and overexpress Ajuba. The CRC cell line SW480 was transduced with two independent shRNAs targeting Ajuba and one lentiviral Ajuba OE construct. All experiments were carried out on cells at 25–50% confluence. Cells were transfected with lentiviral supernatant, containing shRNA targeting Ajuba or Ajuba OE construct in DMEM with 10% FBS. The viral supernatant was added onto the cultured cells in a total volume of 900 μL and incubated for 3 h then centrifuged at 1400 rpm for 5 min, incubated for an additional 3 h, then complete medium was added to a final volume of 2 mL and incubated for 48 h. shRNA lentiviral constructs were purchased from MISSION© ((Sigma, St. Louis, MO, USA).
Clone shAjuba1: targeting the coding region NM-032876.4-1385s1c1
Clone shAjuba2: targeting the non-coding region 2URT NM_032876.4-2786s1c1
Clone shScrambled: pLKO.1-puro Non-Mammalian shRNA control plasmid
To overexpress Ajuba, SW480 cells were transduced with lentivirus containing an Ajuba overexpressing construct
Clone Ajuba OE Harvard PlasmID:Phage_CMV_C_FLAG_HA_IRES_PURO
As a control for the Ajuba overexpressing cells, SW480 cells were transduced with the Ajuba overexpressing construct, in which we clonally excised the exogenous Ajuba sequence. Stable cell lines were positively selected using 1.5 μg/mL puromycin (Life Technologies, Carlsbad, CA, USA). The efficiency of the transduction was assessed by real-time qPCR and immunoblot.
Human colon and colon cancer metastases from the liver were OCT preserved and 4μm cryosections were cut. The cryocut sections or cultured CRC cell lines were fixed using 4% paraformaldehyde for 10 min at room temperature. Cells were then permeabilized and blocked with 0.2% Triton-X and 5% goat serum in PBS for 20 min. After that monoclonal Ajuba ab (1:200 dilution, Cell Signalling, Danvers, MA, USA) was used and incubated over night at 4 °C. After three washing steps with PBS, 0.25%BSA and 0.1% Triton-X, secondary antibody (anti-rabbit conjugated to cy5 (1:1000, Life Technologies, Carlsbad, CA, USA), was incubated for 1 h at room temperature. Finally, nuclei were stained with DAPI (1:2000, Cell Signalling, Danvers, MA, USA) for 30 min at room temperature. Immunohistochemistry was imaged using fluorescence microscopy (LCI DMI4000 B, Leica, Wetzlar, Germany).
RNA was isolated from SW480 cells using the ReliaPrep RNA cell Miniprep System kit (Promega, Madison, WI, USA). The quality and concentration were measured with a 2100 BioAnalyzer (Agilent, Santa Clara, CA, USA). The isolated RNA was then sent for sequencing according to the following parameters: paired-end with reads of 50 bp, TruSeq Stranded mRNA. The RNA was sequenced with a NovaSeq6000 (Illumina, San Diego, CA, USA).
FASTQ files were aligned to the human reference genome hg38 with HISAT2 [51
]. Resulting sam files were transformed into bam with SAMtools [52
]. The reads were counted with the R function featureCounts from the R package Rsubread [53
] with options isPairedEnd = TRUE, GTF.featureType = “exon” and GTF.attrType = “gene_id”.
4.12. Data Visualization
Data was transformed to reads per million (RPM) for visualization. Principal component analysis was done with the R function prcomp on the log(1+x) transformed data. Heatmaps were done with R function heatmap.2 from the R package gplots [54
] with average linkage and Pearson distance; for gene expression, data was scaled from 0 to 1.
4.13. Venn Diagram
Venn diagrams were drawn using custom Venn diagrams from Bioinformatics & Evolutionary Genomics web tools [55
4.14. Pathway Enrichment Analysis
Pathway enrichment analysis was performed using Metascape Gene Annotation & Analysis Resource [56
]. We used the significantly differentially expressed genes looking at the two Ajuba KD compared with the control, shScrambled and Ajuba OE compared with its control. For shAjuba1 we restricted the enrichment analysis to the top 2500 DEG sorted by p
-value. For multiple gene lists, we used the Metascape multiple gene list function. The thresholds applied to select for the DEG genes was adjusted p
-value below 10−10
for shAjuba1 and adjusted p
-value below 0.003 for shAjuba2. The top 20 statistically significant family of pathways, clustered by Metascape, were displayed. The DEG were also displayed as a circos plot in order to show genes that are common or are part of the same pathways. The gene list of the selected pathways were obtained with the R function gconvert from the R package gProfileR2 [57
]. And displayed as stackplots showing the fraction of upregulated and downregulated genes, the total number of DEG and the significance of the pathway.
4.15. Mass Cytometry by Time-of-Flight (CyTOF)
A total of two million cells per cell line were used for CyTOF analysis. The cells were stained with cisplatin to identify live cells and incubated for 10 min at RT. The samples were then fixed, permeabilized and barcoded using the Pd 20-plex barcoding kit (Fluidigm, San Francisco, CA, USA) according to manufacturer’s protocol. The samples were pooled in one tube and stained with metal-conjugated cell surface antibodies according to previously established titrations, cells were then fixed and permeabilized with the FoxP3 intranuclear staining kit as directed in the kit and intracellularly and intranuclearly stained with the targets in these compartments. Lastly, cells are placed in DNA intercalation solution (iridium in Fix and Perm buffer) overnight at 4 °C. The following morning cells were washed 3 times to remove salts and proteins and acquired on the Helios mass cytometer in Maxpar water containing 4-element beads. A minimum of 150,000 cells per cell line was recorded.
4.16. Mass Cytometry Analysis
The raw fcs file was normalized with the R function normalizer_GUI from the R package premessa [58
] and debarcoded with the R function debarcoder_GUI from premessa. The resulting files were gated in Cytobank to discard beads, dead cells, pressure spikes over time and doublets. The cleaned files were analyzed with the R as follows: files were first read as a flowFrame object with the package flowCore [59
]. To visualize the data, 5000 cells from each sample were randomly selected, their expression was arcsinh transformed, and we performed a t
-SNE dimensionality reduction the with R function Rtsne from the R package Rtsne [60
4.17. Comparing RNA-Seq with Mass Cytometry Results
In order to compare the mRNA and the protein expression, first we averaged the RPM values of the RNA-seq among the replicantes. Second, we normalized the RPMs values and the arcsinh transformed CyTOF data to range in [0,1] by dividing by the maximum observed value in Scrambled, sh1 and sh2, or equivalently, by applying function f(x) = x/max(x), independently for RNA-seq and CyTOF. The rescaled values where represented in barplots.
4.18. Cell Proliferation Assay
Cells were seeded in a 96-well plate at a density of 2000 cells per well in 200 μL of medium. Cells were incubated for at least 4 h to attached to the plate. Every day at the same time, MTT (5 mg of thiazolyl blue dissolved in 5 mL DMEM) was added in one tenth of the original culture volume (20 μL for 200 um plated) in each well (4 wells per time point and condition) and incubated for 1 h. The medium then was discarded and replaced by 200 μL of DMSO. Using an Infinite 2000 (Tecan, Männedorf, Switzerland) the plate was shaken and read at an absorbance of 570 nm.
4.19. Colony Formation Assay
One thousand cells per well were platted in 6-well plates and incubated for 7 days. To stop the colony formation, the medium was removed and cells were washed twice with DPBS before being dried. Crystal violet was used to stain the cells (3 g crystal violet, 99.9 mL methanol, 49.9 mL acetic acid) by incubating each well for 30 min at room temperature. The number of colonies were counted using the Colcount (Oxford Optromix, Abingdon, UK).
4.20. Migration Assay
Migration distance was assessed using silicon stoppers in a 96-well plate, plated with a density of 50,000 cells in 200 μL medium per well. Cells were plated in quadruplicates and were left to adhere for 6 h with the silicon stoppers. Afterwards, the stoppers were carefully removed and pictures were taken at 0 h and 24 h under 4× magnification using a light microscope. The radius of the silicon stopper area was calculated by measuring the cell free area using ImageJ package. The migration distance was finally computed by calculating the difference in radius at the different time points in micrometres (μm).
In 6-well plates, cells were platted with a density of 1000 cells in 200 uL medium per well. After letting the cells adhere for 6 h, the plates underwent irradiation in a Gammcell 40 (Best Theratronics, Ottawa, Canada) at the following doses: 0, 2, 3, 4, 5, 6 and 8 Grays. Non-irradiated cells were used as a control. The sensitivity after irradiation was assessed by the cells capacity to form colonies as described under Colony formation assay.
4.22. ALDH Assay
Two Million cells were seeded into 10 cm diameter dishes and let to adhere overnight. The samples were processed according to protocol using the AldefluorTM kit from Stemcell Technologies (Vancouver, BC, Canada). The samples were analysed using FACS (LSR II, BD Biosciences, Franklin Lakes, NJ, USA) recording a total number of 10,000 events. Data was analysed and gated using FlowJoTM 10 in order to remove debris, and duplet cells. Finally, gates were set according to the negative control of the cell lines where DEAB was added. The gates are set to contain exactly 3% positive cells in the negative control and gates were kept the same for the same cell line in order to assess the percentage of cells that have shifted from the negative population.
4.23. In Vivo Metastases
RAG 2−/−y chain KO mice were injected with 1 million of SW480 cells line either shScrambled, shAjuba1 or shAjuba2 in order to assess metastatic behavior of the different cell lines. All surgical procedures were performed under laminar flow and under sterile conditions using a general anaesthesia with intraperitoneal injection of fentanyl, midazolam and medetomidin. Anesthetized mice were immobilized in a supine position and the abdomen was entered through a midline incision. After exposure of the spleen, the cells were injected in a total volume of 100 μL in PBS directly into the spleen, similarly as previously described by Soares et al. [61
]. After injection, a cotton swap was applied of the place of injection for 30 s in order to avoid reflux of the cells. The abdomen was closed with a two-layer running suture. Tumor formation in the spleen and metastatic development the liver, were visually detected at 7 weeks post-operation. Mice were sacrificed using intraperitoneal injection of terminal anaesthesia and organs harvested weighed and photographed. Samples of liver and spleen were snap frozen and paraformaldehyde conserved for further histological, mRNA and protein analysis.
4.24. Graphs and Statistical Analysis
R version 3.5.1 was used for displaying and computation of publicly available data, RNA-seq and CyTOF graphs using the R-packages ggplot2 [62
]. The graphs and the statistics for qPCR, and in Figure 6
were done by using GraphPad Prism software (San Diego, CA, USA). p
-values were calculated using an unpaired, two-tailed Student’s t
-test or two-way ANOVA with no repeated measures and Tukey adjusted for multiple comparison. For all analyses NS denotes p
> 0.05, * p
< 0.05, ** p
< 0.01, *** p
< 0.001, **** p
4.25. Data Availability
All data are available on Genome Expression Omnibus repository with the GEO accession number GSE147111.