ALDH isozymes downregulation affects cell growth, cell motility and gene expression in lung cancer cells

A549 and H522 cell lines were obtained from ATCC and maintained in -80°C freezer or cultured in RPMI 1640 medium containing 10% FBS before experiments. Cells were maintained in 5% CO2 incubator at 37°C and used for experiments in the exponential phase of their growth. These cell lines were specifically chosen for the described experiments because they are known to have high levels of ALDH1A1 and ALDH3A1 [32]. These cells were transduced with lentiviral vectors (described below) containing specific siRNA sequences against ALDH1A1 (Lenti 1 cells), ALDH3A1 (Lenti 3 cells), both vectors (Lenti 1+3 cells), and against the green fluorescent protein (GFP) gene (GFP cells, used as a control). Overall total of 5 cell lines from A549 and H522 were used throughout the experiments described here including the parent wild type cell line (WT). Microarray gene profiling was performed only on A549 cells. In order to further validate that the effects seen in the ALDH knock down cells are directly related to ALDH activity, we used known ALDH inhibitors as described below in A549 and H522 cells as well as one other lung cancer cell line, H1299, that lacks any significant expression of ALDH1A1 or ALDH3A1 [32].

The assays described below were used to study the effects of ALDH knock down by siRNA on cell growth and migration as well as colony formation. Triplicates of 2 × 10⁵ cells/ml/well were plated in 6-well tissue culture plates and grown for 72 hours. After this time, the cells were trypsinized, collected and manually counted using hematocytometer and microscope. Viability was determined by trypan blue exclusion. For these experiments, we used the five cell lines from H522 or A549. Means ± SD of growth rates were calculated and differences in growth rates were compared using the Student t-test.

A colorimetric assay kit (MTT based) for the non-radioactive quantification of cell proliferation and viability was purchased from Roche Diagnostics Corporation (Indianapolis, IN) and used for the cell proliferation experiments. In a 96-well plate, 1 × 10³ cells/100 μL/well of each cell line were cultured for 96 hours. After the 96 hour incubation period, the labeling reagent was added to each well, followed by the solubilization solution four hours later. The cells were left overnight in the incubator, and absorbance was measured the following day using a Molecular Devices Kinetic Microplate reader at a wavelength of 570 nm as per the manufacturer's protocol. In this experiment, absorbance directly correlates with number of viable cells per well. Mean ± SD of the absorbance values were calculated for each experimental group and compared using the Student t-test.

For each of the cell lines, triplicates of 200 cells/ml/well in a six-well plate were cultured in RPMI-1640 + 10% FBS. The cells were allowed to grow for 5 days. The number of colonies adhered to the bottom of the plate was then counted using inverted microscope. The mean number of colonies was calculated and compared among all five cell lines originating from the A549 or H522 cell lines.

We measured the effect of down regulation of ALDH on cell migration and motility by using the in vitro scratch wound migration assay [33]. A549 monolayer confluent cells from the five different cell lines were scratched using sterile 200 μl pipette tip in 6-well plastic dishes, and after 20 hr of culture in RPMI-1640 +10% FBS, the migration ability of the cells was evaluated by the width of the wound under inverted microscope fitted with digital camera. Images of the scratch wound were taken immediately after the scratch and 20 hr later. The migration distances (in cm) of the cells were measured on printed images taken at the same magnification and the difference of the width of wounds at 0 and 20 hr was determined and the % wound healing was calculated.

SiRNA constructs were made using pSilencer 2.0-U6 as vector backbone from Ambion. To generate the lentiviral siRNA constructs, the pSilencer 2.0-U6 siRNA plasmid was digested with Pvu II and the U6-siRNA cassette was cloned into the self-inactivating (SIN) pTYF-EFnlacZ lentiviral vector in the Not I site which was blunted with T4 DNA polymerase. For the constructions of ALDH LV siRNA vectors, the target site for ALDH1A1 (Genbank NM_000689) is: 5'-gtagccttcacaggatcaa-3' (nt 777–795) was chosen, and two primers were used to generate the siRNA construct; primer 1: 5'ATA CGC GGA TCC CGT AGC CTT CAC AGG ATC AAT TCA AGA G AT TGA TC-3', primer 2: 5'CGC TAG ACT AGT TAA AAA AGT AGC CTT CAC AGG ATC AAT CTC TTG AA TTG ATC -3'; the target site for ALDH3A1 (Genbank NM_000691) is: 5'-gaagatgattgcagagaca-3' (nt1167-1185), and two primers were used to generate the siRNA construct: primer 1: 5' ATA CGC GGA TCC CGA AGA TGA TTG CAG AGA CAT TCA AGA GAT GTC T-3', primer 2: 5' CGC TAG ACT AGT TAA AAA AGA AGA TGA TTG CAG AGA CAT CTC TTG AA TGT CT -3'. As a control, we similarly generated U6 promoter-driven LV-siRNA targeting GFP, as described before [34]. The lentiviral constructs with the U6-siRNA inserted in the reverse orientation were chosen for vector preparation (see Figure 1).

To produce lentiviral vectors, 293T cells were co-transfected with the three vector plasmids: pNHP, pHEFVSVG, and pTYF siRNA vector, and the virus supernatants were concentrated and titered as previously described [35]. For lentiviral transduction, the cells were plated at a concentration of 10⁵ cells per well in 12-well plate to give 60–80% confluence after overnight culture and then infected at MOI of 50–100 in the presence of polybrene (8 μg/ml). Transduction efficiency was determined by the LacZ reporter gene assay as previously described [35].

After successful transduction with lentiviral constructs, we plated cells for liquid colony culture assay at 200 cells in 1 ml per 35 mm plate. On approximately day 5 of culture, we picked single colonies using inverted microscope and sterile pipette for replating and establishing clones for each of Lenti 1, Lenti 3, Lenti 1+3, and GFP cell lines. The 100% purity of the clones was validated by LacZ staining done on similarly cultured colonies from these established clones.

In order to measure the effects of the siRNA expression, we used the spectrophotometric methods for measurement of ALDH enzymatic activity as well Western blot analysis to detect changes in the protein levels of ALDH1A1 and ALDH3A1. Western blot analysis was performed as described before [17, 18]. In brief, lysates from each of the 5 cell lines mentioned above were size separated in parallel on two 12% denaturing SDS-polyacrylamide gels (BioRad, Hercules, CA), electrotransferred onto nitrocellulose membranes, blocked with 5% milk in TBS, and probed using chicken anti human ALDH1A1 and ALDH3A1 polyclonal antibodies provided generously by Dr. L. Sreerama (St Cloud University, Minneapolis, MN) and Dr. NE Sladek (University of Minnesota, Minneapolis, MN). The specificity of these antibodies has been documented by Dr. Sladek's group [36, 37]. Blots were incubated with chicken anti-human ALDH1A1 and ALDH3A1 primary antibodies at 1:200 or 1:300 dilution, respectively, for 1 hour at room temperature. After washing, the secondary antibody (horseradish peroxidase-labeled rabbit anti-chicken antibody; Sigma Chemical Co., St Louis, MO) was used at 1:4000 dilution for 1 hour. Chemiluminscence method (SuperSignal, Pierce, Rockford, IL) was used for the final visualization of the protein bands on X-ray film (Super Rx, Fuji Photo Film, Tokyo, Japan). After washing and blocking, the same blots were labeled again for visualization of actin as a loading control using anti-actin antibody (Oncogene Research Products, Cambridge, MA).

The cell lysates used for Western blot analysis, were also freshly used to measure ALDH enzyme activity using the spectrophotometric assay as described before [17, 18]. Briefly, the aliquots of 600 μl lysing buffer were incubated at 37°C in Beckman DLC 64 spectrophotometer cuvettes with the addition of cell lysate, 5 mM NAD⁺ and 5 mM propionaldehyde as a substrate. The rate of change in absorbance at 340 nm was measured in 3 replicates over 5 min. A control reaction in which the substrate was not added monitored the endogenous rate of NAD⁺ reduction. The ALDH activity was expressed in nmoles/10⁷ cells.min.

A5 49 cell lines (WT, GFP, and Lenti 1+3) were plated in 25 cm² Corning cell culture flasks at a density of 7.5 × 10⁵ cells/ml with RPMI 1640 medium containing 10% FBS at 37°C. Four culture sets were prepared for each experimental group. Four different clones were used for GFP and Lenti 1+3. After 24 hr incubation, media was removed from the dish and RLT buffer was added directly to the cells. Cells were scraped into this buffer and subsequently homogenized using the QIAshredder kit (Qiagen, Valencia, CA). Total RNA was then harvested using the RNeasy RNA isolation kit (Qiagen). Integrity of total RNA was validated using a Bioanalyzer (Agilent Technologies) as well as OD 260/280 ratios.

Two micrograms of total RNA were utilized for the cDNA synthesis using the cDNA synthesis kit (Affymetrix 900431) and according to the Affymetrix technical protocols. First-strand synthesis was initiated using Superscript II Reverse Transcriptase and the T7-(dt) primer (Affymetrix 900431), which contains the T7 promoter sequence followed by an oligo (dt) tract. Second strand synthesis was carried out using Escherichia coli DNA polymerase I, E. coli DNA ligase, and RNase H (Affymetrix 900431). Five microliters of purified ds cDNA was used to generate biotin-labeled cRNA using the IVT labeling kit (Affymetrix 900449). Biotin labeled cRNA was then purified from unincorporated nucleotides using the RNA cleanup kit (Affymetrix 900371). Quantity and quality of cRNA was determined by OD260/280 ratios and by using a Bioanalyzer (Agilent Technologies).

Before using the array, biotin-labeled cRNA was fragmented for 35 minutes at 94°C. The following reagents were mixed to give the following final concentrations: fragmented cRNA (0.05 μg/ul), control oligonucleotide B2 (50 pM, Affymetrix), 20X eukaryotic hybridization controls (BioB at 1.5 pM, BioC at 5 pM, BioDn at 25 pM, and CreX at 100 pM, Affymetrix 900454), herring sperm DNA (0.1 mg/ml, Promega), acetylated BSA (50 mg/ml, GIBCO-BRL), and 2X hybridization buffer. Probe was heated to 99°C for 5 min, then transferred to 45°C for 5 min, and finally spun at maximum in a microcentrifuge for 5 min. Arrays used for this study were the Human Genome U133 Plus 2.0 Array (Affymetrix) representing over 47,000 transcripts and variants including 38,500 well-characterized human genes. Chips were interrogated with material harvested from A549 WT, Lenti 1+3 and GFP cells and each cell treatment was replicated four times using 4 chips for each experimental group, making the data from each array truly independent biological replicas. Each array was injected with 200 ul of target and rotated in a 45°C oven for 16 hours. Probe was removed from the array and the array was washed and stained using the Euk-WS2 fluidics protocol and the streptavidin-phycoerythrin (SAPE) reagent (Affymetrix).

Scanned images (*.dat files) were analyzed with GCOS software to generate a detection call for each probe set. Hybridization single intensities between GeneChips were normalized and an expression matrix was derived with dChip [38] using the PM only model model-based expression index. Probe sets whose hybridization single intensities were never measured above background levels were removed from the data set to reduce the noise level in the data set. Probe sets whose hybridization single intensities were measured at or below background levels were identified using the Affymetrix detection call algorithm [39].

The modeled-based expression matrix of the probe sets passing the initial expression filter were analyzed in log space with BRB Array Tools 3.2.2 (developed by Richard Simon and Amy Peng Lam and is available online at the Biometric Research Branch of the National Cancer Institute website http://​linus.​nci.​nih.​gov/​BRB-ArrayTools.​html to identify genes that differentiated among the treatment classes at the P < 0.001 significance threshold. The ability of gene identified at the P < 0.001 significance threshold to be differentiated among treatment classes was assessed by using leave-one-out-cross-validation (LOOCV) using a nearest-neighbor prediction model. For visualization purposes the modeled-based expression matrix of probes sets identified as significant among the treatment groups were mean centered and variance normalized using algorithms implemented in dChip.

Probe sets that were significantly different at P < 0.001 and demonstrated ≥ 2 fold change in the level of gene expression were subject to comprehensive, detailed, six-level functional analysis, whereby identity was verified using GenBank accession number, aliases were compiled, and biochemical and biological functions were determined. After the identity of the probe set was confirmed, multiple internet bioinformatics and National Center for Biotechnology Information (NCBI) search engines, including PubMed searches, were used to assign the biological function. Probe sets without confirmed identities were placed in the "Unknown" category. Probe sets that could be identified but that did not have ascertainable function were classified as "Other"

In order to confirm microarray data, we performed real-time RT-PCR [40] for 7 selected genes measuring the change in their level of expression. Primer/probe sets for HMOX1, MAP3K8, MGST1, CXXC5, LHX8, RAB20, STRA6 and 18S rRNA (reference control) were designed by Assays-on-Demand (Applied Biosystems, Foster City, CA, USA) and are listed in Table 1. Total RNA was obtained as described above from all five A549 cells lines described under "Cell Lines".

Two micrograms of total RNA were reverse transcribed to cDNA in a 50 μl reaction containing 4 μl of 100 mM MgCl, 1.0 μl of RNase inhibitor, 4 μl of 10× PCR buffer, 1 μl of 10 mM dNTP mix, 1.0 μl of random hexamers, 1.0 μl of reverse transcriptase (Invitrogen). Mixture was incubated for 50 min at 42°C and the reaction was stopped by heating to 70°C for 15 min. For each of the Genes studied and 18S, 3 μl of 400 ng/mL diluted cDNA was added to 12.5 μl of Taqman Universal Master Mix (Applied Biosystems), 1.25 μl of 20× primer/Probe combination, and 8.25 μl of DEPC-treated water per each reaction. RT-PCR quantification and determination of expression levels were performed on ABI Prism 7900 and Sequence Detection Software 1.6 (Applied Biosystems, Foster City, CA, USA). All PCR reactions were performed in triplicates in our Gene Expression Core Facility and the results were analyzed using comparative method following normalization of expression values to 18S rRNA expression.

For comparison to the siRNA induced knock-down of ALDH activity and as a way of further validation of the microarray results, we also used known inhibitors of ALDH, Tetraethylthiuram disulfide (TT, disulfiram) or diethylaminobenzaldehyde (DEAB), to study the effects of ALDH inhibition on cell growth and expression of selected genes. These inhibitors were purchased from Sigma Biochemicals (Milwaukee, WI) dissolved in DMSO and used in a final DMSO concentration of ≤ 0.1%.

In these experiments, we used the A549 and H522 parent cell lines as well as H1299 cells, a cell line known not to have any measurable ALDH activity [32]. Dose response curves (0, 0.1, 1, 10, and 100 μM) were performed for each inhibitor measuring ALDH activity, cell growth and viability. We used either manual cell counts or the calorimetric (MTT) assay, as described above. In subsequent experiments only 1 and 100 μM concentrations were used. Cells were plated at 2 × 10⁵/well in 6-well plates for 24 hr before adding the inhibitor. After 48 hr incubation with and without the inhibitors (untreated controls), the cells were harvested, counted and then used for ALDH activity assay as described above. The results were expressed as % decrease in cell proliferation or ALDH activity in comparison to untreated control.

In order to confirm that changes in the expression of selected genes are indeed specific to the siRNA induced knock-down of ALDH activity, we performed real-time RT-PCR on 3 of the 7 genes mentioned above (CXXC5, LHX8, RAB20) using RNA obtained from WT A549 cells after incubation with 5 μM of either TT or DEAB for 48 hr. Obtaining similar changes in the expression of these 3 genes after treatment with ALDH inhibitors would strongly suggest direct relationship between ALDH1A1 and ALDH3A1 activity and the changes in the gene profile demonstrated by the microarray analysis.

The expression of the specific siRNAs in A549 cells resulted in a significant decrease in ALDH activity as demonstrated by spectrophotometric enzyme activity assay (Figure 2A). The effect of siRNA against each ALDH isozyme was also demonstrated by Western blot analysis (Figure 2A). With such significant decrease in ALDH activity and proteins in the Lenti 1+3 cell line, we proceeded with the microarray gene profiling experiment comparing gene expression in this cell line to that in GFP and WT control cell lines. Similar decrease in ALDH activity was obtained in H522 lung cancer cell line using the same lentiviral constructs (Figure 2B). However, we have observed the loss of siRNA effect after continuous culture of either cell line for more than 4 weeks. Because of that, fresh cells were thawed every two weeks, and ALDH activity verified before any experiments were performed.

A549 and H522 cells (2 × 10⁵/group) were plated for 72 hours and then counted and viability determined. Figure 3A shows the results which reflect mean ± SD of cell number from at least 4 experiments. The growth rate was significantly lower (P < 0.024) in Lenti-1, Lenti-3, and Lenti-1+3 in comparison to GFP, except for A549 Lenti-1 with P value of 0.052. No significant difference was found between the GFP and the WT. Using the MTT cell proliferation assay, we were able to show similar effects (data not shown).

A549 cells (200/petri dish in triplicates per each group) were plated for 5 days and colonies adhered to the bottom of the plates were counted after this time period. Results show that the mean number of colonies decreased significantly (range, 27–71%) for the Lenti-1, Lenti-3 and Lenti-1+3 cells in comparison to GFP and WT cells of both A549 and H522 (P < 0.016 and 0.003, respectively).

An example of the scratch wound migration test results for A549 cells are shown in Figure 3C which demonstrates delayed wound healing of the Lenti 1+3 cells in comparison to control cells. Summary of at least 4 readings are given for all the five A549 cell lines (Figure 3B). These results show significant inhibition (P < 0.005) of cell migration by knock-down of each enzyme separately as well as both enzymes (Lenti 1+3).

To construct the transcriptional profiles associated with the "knock-down" of ALDH1A1 and ALDH3A1 in A549 cells (Lenti 1+3), two other cell lines, WT and GFP, were used to interrogate the Affymetrix U133 Plus human genome GeneChips containing 38500 well-known genes. The statistical analyses are detailed above. We performed a modified F test to identify probe sets whose hybridization signal intensities varied significantly as a function of the ALDH "Knock-down". Overall about 3735 probe sets out of total of 54675 sets were detected above background and statistically different (P < 0.001) in the Lenti 1+3 experimental group when compared to the WT group (Figure 4) (supplemental material is available on http://​www.​ncbi.​nlm.​nih.​gov/​geo using accession number GSE8045). Many more changes in gene expression (17269 probe sets) were seen when 3 way comparisons were done among the 3 experimental groups mainly due to non-specific changes seen in the GFP group. In order to arrange the significant genes into functional framework, we chose only those genes that were significant (P < 0.001) and differed by ≥ 2 fold in the Lenti 1+3 group when compared to the WT and GFP groups. There were 26 up-regulated and 77 down-regulated genes including ALDH1A1 and ALDH3A1. We summarized the chromosomal location for these genes in Figure 5. We also researched their biological and biochemical properties using multiple internet bioinformatics search engines, web sites, and NCBI resources including BLAST, PubMed, etc. The collective results of those searches were synthesized into color-coded, mechanism-based pies of gene expression (Figure 6). These illustrations provide the relative dominance of each of the biological-biochemical processes most affected by the "Knock-down" of the ALDH isozymes. The names of these genes are tabulated in Tables 2 and 3. In summary, the affected genes are found on multiple chromosomes with the largest cluster (8 genes) observed on chromosome 12. The biological pathway most affected was signal transduction, organ development and Transcription among both the induced and repressed genes, although more pathways were found to be involved with the repressed genes. On the other hand, the biochemical pathways most affected in induced genes were protein binding, cell adhesion and lipid metabolism, while with the suppressed genes protein binding, intracellular signaling and DNA related processes were most affected (see Figure 6). As shown in Table 2, eight of the repressed genes (ID4, DDX3Y, RPS4Y1, EIF1AY, CCL20, GPC6, GPR37, and HMGA2) had the highest decrease (> 6 folds) in expression. At least of these genes, RPS4Y1 and CCL20 are highly expressed in WT A549 cells.

Since the human genome contains 19 ALDH functional genes and three pseudogenes [5], we sought to analyze the effect of the siRNA sequences used here on the other members of the ALDH family. We interrogated the data set of probes that showed changes in gene expression and pulled out the ALDH genes and compared changes among the 3 different experimental groups. The data is summarized in Table 4, and it shows that the siRNA sequences used are quite specific which validates our data further. ALDH1A1 and ALDH3A1 transcripts were decreased by about 7 and 4 folds, respectively. Total of another 12 ALDH genes showed some expression variability, but it was either non-specific (similar in GFP and Lenti 1+3) or minimal.

In order to assess the validity of the microarray results, we used real-time RT-PCR to compare the expression of 8 transcripts with significant expression changes (P < 0.001 and ≥ 2 folds in Lenti 1+3) per the microarray in all 5 cell lines mentioned in Figure 2A include WT, GFP, Lenti 1, Lenti 3 and Lenti 1+3. Figure 7 shows results for 3 genes CXXC5, RAB20, and LHX8 that confirm the results seen in the 3 experimental groups of the microarray. The increase in CXXC5 and RAB20 expression was synergistic with the knock-down of both ALDH1A1 and ALDH3A1, while the effect seen on LHX8 seems to be related to the lentiviral vector integration/effect because it is seen equally in all cell lines except WT. HMOX1 and MGST1 were increased in Lenti 1+3 and to a lesser degree with either Lenti 1 or Lenti 3. On the other hand, MAP3K8 expression was decreased in Lenti 1+3 only, while STRA6 was decreased in Lenti 1+3 and Lenti 1. Thus, the real-time RT PCR measurements of all 7 transcripts showed agreement of 100% with the microarray results. LHX8 was the only gene out of this group exhibiting non specific change in expression, both by microarray as well real-time RT PCR. On the other hand, MGST1 and STRA6 change in expression did not meet the cutoff of ≥ 2 folds.

To further validate our results, cells (1000/well) were incubated with ALDH inhibitors, DEAB or TT, for 72 hrs and cell proliferation measured by the calorimetric MTT based assay (Figure 8A). Absorbance was significantly lower (P < 0.0001) after adding any concentration of TT to A549 cells as well as to H522 cells. Absorbance was significantly lower (P < 0.023) after adding 0.1 μM DEAB to A549 cells, while for H522 cells, the absorbance was significantly lower (P < 0.007) only after adding ≥ 1 μM of DEAB. The slight recovery of cell proliferation at 10 μM concentration of TT has been described before [19]. These proliferation inhibitory effects seem to be specific to the inhibition of ALDH activity since the use of these inhibitors in H1299 cells lacking ALDH1A1 and ALDH3A1 expression or any significant ALDH activity [32] resulted in significantly less effects (P < 0.029) of either TT or DEAB on H1299 cell growth in comparison to H522 and A549 cell lines (Figure 8B). The dose-response effect of both inhibitors (1 and 100 μM) on ALDH activity in A549 and H522 cells is shown in Figure 8C.

Furthermore, 3 of the genes (CXXC5, RAB20 and LHX8, shown in Figure 7) were examined using real-time RT-PCR after 48 hr incubation of A549 cells with 5 μM DEAB or TT. The results show similar changes in the expression of CXXC5 and RAB20 to what was seen in the microarray results and real-time RT-PCR (Figure 8D). Once again change in LHX8 expression was affected differently by the two different inhibitors, which is suggestive of non-specific effect.