Phosphorylation-dependent regulation of ALDH1A1 by Aurora kinase A: insights on their synergistic relationship in pancreatic cancer
© The Author(s). 2016
Received: 22 July 2016
Accepted: 24 November 2016
Published: 13 February 2017
Epithelial-to-mesenchymal transition (EMT) and cancer stem cell (CSC) formation are key underlying causes that promote extensive metastasis, drug resistance, and tumor recurrence in highly lethal pancreatic cancer. The mechanisms leading to EMT and CSC phenotypes are not fully understood, which has hindered the development of effective targeted therapies capable of improving treatment outcomes in patients with pancreatic cancer.
We show a central role of Aurora kinase A (AURKA) in promoting EMT and CSC phenotypes via ALDH1A1, which was discovered as its direct substrate using an innovative chemical genetic screen. AURKA phosphorylates ALDH1A1 at three critical residues which exert a multifaceted regulation over its level, enzymatic activity, and quaternary structure. While all three phosphorylation sites contribute to its increased stability, T267 phosphorylation primarily regulates ALDH1A1 activity. AURKA-mediated phosphorylation rapidly dissociates tetrameric ALDH1A1 into a highly active monomeric species. ALDH1A1 also reciprocates and prevents AURKA degradation, thereby triggering a positive feedback activation loop which drives highly aggressive phenotypes in cancer. Phospho-resistant ALDH1A1 fully reverses EMT and CSC phenotypes, thus serving as dominant negative, which underscores the clinical significance of the AURKA-ALDH1A1 signaling axis in pancreatic cancer.
While increased levels and activity of ALDH1A1 are hallmarks of CSCs, the underlying molecular mechanism remains unclear. We show the first phosphorylation-dependent regulation of ALDH1A1, which increases its levels and activity via AURKA. Recent global phospho-proteomic screens have revealed increased phosphorylation of ALDH1A1 at the T267 site in human cancers and healthy liver tissues where ALDH1A1 is highly expressed and active, indicating that this regulation is likely crucial both in normal and diseased states. This is also the first study to demonstrate oligomer-dependent activity of ALDH1A1, signifying that targeting its oligomerization state may be an effective therapeutic approach for counteracting its protective functions in cancer. Finally, while AURKA inhibition provides a potent tool to reduce ALDH1A1 levels and activity, the reciprocal loop between them ensures that their concurrent inhibition will be highly synergistic when inhibiting tumorigenesis, chemoresistance, and metastasis in highly aggressive pancreatic cancer and beyond.
KeywordsAurora kinase A ALDH1A1 Chemical genetic Epithelial-to-mesenchymal transition Therapy
Pancreatic cancer is an exceptionally aggressive disease with the majority of its victims dying within 1 year of diagnosis . Effective early detection and screening are not available, and tumors are typically diagnosed after metastasis. Systemic chemotherapy, radiation therapy, or a combination of both therapies is used; however, patients rapidly develop resistance, leading to a poor 5-year survival rate of only 6%. Accumulating evidence suggests that the acquisition of epithelial-to-mesenchymal transition (EMT) and cancer stem cell (CSC) phenotypes in pancreatic tumors is an important underlying cause for extensive local tumor invasion, early metastasis, drug resistance, and tumor recurrence . However, the mechanisms leading to EMT and CSC phenotypes are not fully understood, which has hindered the development of effective targeted therapies and sensitive biomarkers to improve treatment outcomes in patients with pancreatic cancer.
We focused on Aurora kinase A (AURKA), as it is overexpressed in a vast majority of pancreatic tumors [3–5]. AURKA depletion inhibits in vivo tumorigenicity, enhances taxane chemosensitivity, and induces apoptosis in pancreatic cancer cells . Data such as these have resulted in ongoing clinical trials of more than a dozen AURKA inhibitors. Despite these encouraging findings, AURKA remains an essential kinase required for mitosis in all dividing cells. Consequently, its inhibition in Phase II clinical trials has been associated with several adverse side effects, suggesting that collateral inhibition of AURKA in rapidly proliferating normal tissues is responsible for the undesirable side effects . These findings indicate that selective therapies against oncogenic targets of AURKA exploited in cancer may be a superior option for developing effective drugs and combating collateral toxicity.
To this end, we identified aldehyde dehydrogenase 1 (ALDH1A1) as a direct substrate of AURKA using an innovative chemical genetic screen that we developed. ALDH1A1 is a pancreatic CSC marker and is highly enriched in a subpopulation of cells which are extremely resistant to chemotherapy. Furthermore, ALDH1 is highly enriched in surgical specimens from patients with pancreatic cancer who had undergone preoperative chemo-radiation therapy compared to untreated patients . However, ALDH1A1 levels have not been analyzed in pancreatic cancer. As AURKA overexpression is also associated with enhanced chemoresistance, we hypothesized that AURKA may upregulate ALDH1A1 leading to EMT, CSC phenotypes, and chemoresistance in pancreatic cancer.
Antibodies against AURKA (H-130), actin (C-2), tubulin (TU-02), and ALDH1A1 (B-5) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA) (for RRIDs and/or lot numbers see Additional file 1). Cell culture BxPC3, Panc1, and HEK293T cells were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). Validated antibodies against Snail (40084), Slug (40088), N-cadherin (39429), and CD44 (39435) were purchased from One World Lab (San Diego, CA, USA) and were used at 1:5000 dilution. Antibodies against vimentin (bs-0756R), E-cadherin (bs-10009R), and matrix metalloproteinase-2 (MMP-2) (bs-4599R) were purchased from Bioss Antibodies and used at 1:1000 dilution (Additional file 1). BxPC3 and Panc1 cells were cultured in RPMI with 10% fetal bovine serum (FBS) supplemented with 2 mM glutamine and antibiotics. HEK293T and Phoenix cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS supplemented with 2 mM glutamine and antibiotics.
Expression plasmids and constructs
Human influenza hemagglutinin (HA)-tagged wild-type and mutant ALDH1A1 were cloned into VIP3 mammalian vector and bacterial petDuet vector at EcoRI and XhoI sites. HA-tagged ALDH1A1 mutants were generated using site-directed mutagenesis.
Expression and purification of AURKA, TPX2, and ALDH1A1
AURKA was prepared using the Baculovirus Bac-to-Bac Expression System according to the manufacturer’s instructions (Invitrogen). 6x-His-TPX2 and 6x-His-tagged wild-type and mutant ALDH1A1 were expressed in Escherichia coli and purified using the procedures previously described [9, 10].
Transfection and retroviral infection
For generating stable cell lines, AURKA and ALDH1A1 plasmids were transiently transfected using calcium phosphate into Phoenix cells. The retroviruses were harvested and used to infect BxPC3 cells as reported previously .
In vitro kinase assays
For in vitro labeling, AURKA-TPX2 complex (on Ni-NTA beads) was pre-incubated with 100 μM of ATP for 1 h in a 1× kinase buffer (50 mM Tris, 10 mM MgCl2) to activate AURKA. The beads were washed extensively with 1× kinase buffer to remove excess ATP, and then subjected to an in vitro kinase assay with 2 μg of 6x-His-tagged recombinant protein (wild-type or mutant ALDH1A1) in the presence of 0.5 μCi of [γ-32P]ATP for 15 min. Reactions were terminated upon the addition of sodium dodecyl sulfate (SDS) loading buffer and subsequently separated by SDS-PAGE gel, transferred to a polyvinylidene difluoride (PVDF) membrane, and exposed for autoradiography.
AURKA and ALDH1A1 shRNA
AURKA short hairpin RNAs (shRNAs) were generated in our previous study . Both AURKA and ALDH1A1 shRNAs were cloned into the pLKO.1 TRC vector, which was a gift from David Root . The sequences are as follows:
AURKA shRNA1 (forward): 5′-CCGG GGC TTT GGA AGA CTT TGA AAT CTCGAG ATT TCA AAG TCT TCC AAA GCC TTTTTG-3′. AURKA shRNA1 (reverse): 5′- AATTCAAAAA GGC TTT GGA AGA CTT TGA AAT CTCGAG ATT TCA AAG TCT TCC AAA GCC-3′. AURKA shRNA2 (forward): 5′- CCGG GCA CCA CTT GGA ACA GTT TAT CTCGAG ATA AAC TGT TCC AAG TGG TGC TTTTTG-3′. AURKA shRNA2 (reverse): 5′-AATTCAAAAA GCA CCA CTT GGA ACA GTT TAT CTCGAG ATA AAC TGT TCC AAG TGG TGC-3′. AURKA shRNA3 (forward): 5′-CCGG GCC AAT GCT CAG AGA AGT ACT CTCGAG AGT ACT TCT CTG AGC ATT GGC TTTTTG-3′. AURKA shRNA3 (reverse): 5′-AATTCAAAAA GCC AAT GCT CAG AGA AGT ACT CTCGAG AGT ACT TCT CTG AGC ATT GGC-3′. ALDH1A1 shRNA1 (forward): 5′ C CGG AGC CTT CAC AGG ATC AAC AGA CTC GAG TCT GTT GAT CCT GTG AAG GCT TTT TTG 3′. ALDH1A1 shRNA1 (reverse): 5′ A ATT CAA AAA AGC CTT CAC AGG ATC AAC AGA CTC GAG TCT GTT GAT CCT GTG AAG GCT 3′. ALDH1A1 shRNA2 (forward): 5′ C CGG ACC TCA TTG AGA GTG GGA AGA CTC GAG TCT GTT GAT CCT GTG AAG GCT TTT TTG 3′. ALDH1A1 shRNA2 (reverse): 5′ A ATT CAA AAA ACC TCA TTG AGA GTG GGA AGA CTC GAG TCT GTT GAT CCT GTG AAG GCT 3′. Control shRNA (scrambled shRNA), AURKA, and ALDH1A1 shRNA lentiviruses were generated and used for infecting BxPC3 cells. Stable cells were generated following puromycin selection.
Soft agar colony formation
BxPC3, Panc1, and different stable cell lines were plated in RPMI (103, 104, and 105 cells per dish in triplicate), 0.3% agar, and 10% FBS six-well plates as reported previously . Transformed colonies were counted after 3 weeks using crystal violet staining.
Cells were lysed in modified radioimmunoprecipitation assay (RIPA) buffer, supplemented with protease inhibitors. Equal amounts of cell extracts were used for western blotting.
BxPC3 cells were co-transfected with ALDH1A1 or AURKA shRNA along with 6x-His-ubiquitin. After 36 h, MG132 (Sigma) was added at 10 μM final concentration for an additional 12 h. Cells were then harvested, and ubiquitylated proteins were isolated using Ni-NTA beads. The proteins were separated by SDS-PAGE and analyzed using antibodies against AURKA and ALDH1A1.
Cell migration was determined using Boyden chambers as reported previously . The assays were performed in replicates of three, in four independent experiments. To allow for comparison between multiple assays, the data were normalized and expressed as a percentage of the number of cells present on the membrane.
Cells were seeded in 96-well plates at 5 × 103 cells per 100 μL per well and cultured for 24, 48, and 72 h. The MTT assay was conducted as reported previously . Experiments were repeated three times in triplicate wells to ensure reproducibility.
BxPC3 and Panc1 cells were grown on poly-l-lysine-coated coverslips for 24 h, fixed with 4% formaldehyde in PBS for 15 min at room temperature, and then washed three times with PBS. The cells were blocked in 1% FBS, 2% bovine serum albumin (BSA), 0.1% triton X-100 in PBS for 1 h at 25 °C. Cells were labeled with antibodies (Actin, AURKA, or ALDH1A1) for 3 h in PBS, followed by incubation with fluorescein isothiocyanate- or Texas Red-conjugated secondary antibody. Cells were visualized using a Nikon Eclipse E600 microscope (Nikon Instruments, Melville, NY, USA).
ALDH1A1 activity assay
The ALDH1A1 activity assay was conducted in 100 mM Hepes buffer (pH 8.0) containing 2 mM DTT, 5 mM NAD+, 1.5 mM MgCl2, and 37.5 mM imidazole. The relative concentrations of wild-type and each mutant were determined by SDS-PAGE densitometry to ensure equal concentration in all samples. To measure the consequences of AURKA-mediated phosphorylation on ALDH1A1 activity, it was phosphorylated using AURKA/TPX2 in kinase buffer (37.5 mM Hepes pH 8.00, 300 μM ATP, 4 mM MgCl2). Each reaction was then diluted with Hepes-BSA buffer so that each sample contained 100 mM Hepes pH 8.00, 77.4 μM ATP, 1.03 mM MgCl2, 0.65% BSA, and 5 mM NAD+. Reactions were initiated upon the addition of propanal to a final concentration of 4 mM. The change in absorbance at 360 nm due to NADH formation was measured in a 96-well plate format (SpectraFluor PLUS, TECAN). The data are shown as the fractional conversion of NAD+ to NADH to highlight the relative intrinsic activity of wild-type and mutants and effect of AURKA-dependent phosphorylation. For the dephosphorylation experiment, ALDH1A1 activity was measured for 45 min, at which time 1 μL of calf-alkaline phosphatase was added to the sample and the measurements were performed for another 1.5 h. For untreated samples, 50% glycerol in 1x CutSmart Buffer (New England Biolabs) was added as a control.
For experiments coupling the percentage of ALDH1A1 phosphorylation with a change in its dehydrogenase activity (Fig. 3g–i), recombinant AURKA-TPX2 and ALDH1A1 (0.74 mg/mL) were reacted on a 10 μL scale in the indicated combinations in 1x kinase assay buffer containing 0.5 mM ATP at 30 °C. An identical reaction mixture for samples containing ATP was treated with a tracer amount (~0.01 μCi) of [γ-32P] ATP. After incubation for varying periods, the kinase reactions were terminated by adding EDTA. For activity measurements, each sample was simultaneously treated with freshly prepared 5 mM NAD+ and the initial reaction absorbance was recorded. Reactions were initiated upon the addition of propanal to a final concentration of 5 mM. The reaction progress curves were monitored at 360 nm every 10 min for 7 h. The reaction progress curves were plotted as (At-Ao/Ao) to reflect the relative amounts of NADH produced over time. For measuring percent phosphate incorporated into ALDH1A1 (Fig. 3g), the radioactive reactions were terminated via trichloroacetic acid (TCA) precipitation, washed two times with 200 μL of 90% acetone, and counted. To determine the percent phosphate incorporated into ALDH1A1, the background signal from AURKA in the absence of ALDH1A1 was subtracted from that containing both ALDH1A1 and AURKA. As AURKA phosphorylates ALDH1A1 at three sites, we calculated the percentage mol phosphorylation of ALDH1A1 per mole of the protein accordingly.
Native gel analysis of AURKA-mediated ALDH1A1 phospho-oligomeric regulation
AURKA was immunoprecipitated from BxPC3 cells, and a kinase reaction was carried out in 50 mM Tris pH 8.00, 10 mM MgCl2, 50 mM NaCl, and 50 mM imidazole containing ~0.3–0.6 mg/mL ALDH1A1. Notably, when high concentrations of ALDH1A1 were used (>1 mg/mL), the change from tetramer to monomer was less pronounced. The kinase reaction was initiated upon the addition of ATP to a final concentration of 1 mM. The reaction was spun down, and 10 μL of the supernatant was mixed with 5 μL of 3x native running buffer containing 10 mM EDTA for the first time point. The reaction tube was then placed on a thermomixer (800 rpm) at room temperature. 10 μL aliquots of the supernatant were collected at each time point indicated. The samples were separated using clear-native PAGE (CN-PAGE) with a 7.5% resolving gel (pH 8.9) and a 4.3% stacking gel (pH 6.7). Following electrophoresis, the gel was soaked in 1x SDS-PAGE running buffer for 5 min to denature complexes, followed by 1x transfer buffer containing 20% methanol to remove excess SDS. Proteins were transferred to a PVDF membrane, and ALDH1A1 was visualized using a monoclonal anti-6x-His antibody.
In-gel ALDH1A1 activity assay
AURKA kinase was immunoprecipitated from BxPC3 cells and reacted for 30 min with ALDH1A1 as previously described. ALDH1A1 in the absence of AURKA was used as a control. Samples were separated using CN-PAGE 1.5 mm gels. Following electrophoresis, the gel was soaked 2x for 5 min in 100 mM Tris pH 8.8, and for an additional 5 min in 50 mL of 100 mM Tris pH 8.8 containing 15 mg nitrotetrazolium blue chloride (NBT) and 25 mg NAD+. Colorimetric detection of ALDH1A1 activity was carried out by incubating the gel in the same buffer as above with the addition of 1 mg of 1-methoxy phenazine methosulfate (PMS) and 5 mM benzaldehyde in the dark at 37oC with agitation. Bands were clearly visible after 5 min and were allowed to intensify over 45 min. The reaction was terminated by soaking the gel in a 10% acetic acid solution. The enzymatic activity was visualized as a purple band due to formazan precipitation upon NAD+ consumption. The gel was then fixed in 10% acetic acid and 40% methanol and stained with Coomassie G-250.
Oligomer-specific phosphorylation of ALDH1A1
AURKA kinase was reacted with ALDH1A1 as previously described for 0, 10, and 30 min. Following CN-PAGE, the gels were fixed in 10% acetic acid and 30% methanol for 20 min and washed 3 times for 5 min with water. Phos-tag staining was carried out according to the manufacturer’s protocol. Gels were imaged using a Typhoon© scanner equipped with a Phos-tag filter. Following fluorescent detection, total protein levels were visualized with Coomassie G-250 staining. Protein quantification was determined by SDS-PAGE densitometry. The Phos-tag gel was colorized using ImageJ to denote it as a phosphorylation-specific signal.
BxPC3 cells were plated in a 96-well plate at a density of 3000 cells/well. After 12 h, the cells were transfected with 50 ng/well of Snail, Slug, CD44, or E-cadherin promoter-driven luciferase plasmids using lipofectamine (Invitrogen). Cells were simultaneously co-transfected with 50 ng/well of the pRL-SV40 Renilla luciferase plasmid (Promega) as an internal control. After 48 h, firefly and Renilla luciferase activities were measured with the Dual Luciferase kit (Thermo Fisher Scientific). The firefly luciferase signal was normalized to the Renilla luciferase signal to account for variations in transfection efficiency. All experiments were performed in triplicate, three independent times.
Bar graph results are plotted as the average ± standard error of the mean (SEM). Significance was evaluated using a two-tailed Student’s t test and two-way analysis of variance and is displayed as follows: *p < 0.05, **p < 0.01, ***p < 0.001.
ALDH1A1 is a direct target of AURKA
AURKA regulates the subcellular localization of ALDH1A1 in pancreatic cancer cells
We next examined the subcellular localization of AURKA and ALDH1A1 in BxPC3 and Panc1 cells. ALDH1A1 displayed cytoplasmic localization, which is consistent with previous findings showing it to be a cytoplasmic enzyme (Fig. 1b and c). AURKA too displayed cytoplasmic localization in both BxPC3 and Panc1 cells (Fig. 1d and e). More importantly, when AURKA levels were knocked down using AURKA-shRNA or inhibited using MLN8237 (aka alisertib, an AURKA-specific inhibitor), ALDH1A1 adopted somewhat perinuclear localization (compare actin staining versus ALDH1A1 staining), suggesting that AURKA regulates the subcellular localization of ALDHI1A1 to some extent in BxPC3 cells (Fig. 1f and h). As shown in Fig. 1g and i, ~35% and 22% of BxPC3 cells displayed perinuclear localization of ALDH1A1 in AURKA shRNA-treated and MLN8237-treated cells, respectively. To examine whether AURKA-mediated regulation of ALDH1A1 was common in other pancreatic cancer cells, we investigated ALDH1A1 subcellular localization in Panc1 cells in the absence or presence of either AURKA shRNA or MLN8237. Similar to the results obtained in BxPC3 cells, AURKA depletion or inhibition resulted in moderate perinuclear localization of ALDH1A1 (Fig. 1j–m). Data used to generate the summary statistics shown in Fig. 1g, i, k, and m are reported in Additional file 2.
AURKA and ALDH1A1 associate in pancreatic cancer cells
AURKA positively regulates ALDH1A1 levels
Our previous studies revealed that AURKA-mediated phosphorylation of PHLDA1 at S98 degrades it, whereas AURKA-mediated phosphorylation of LIMK2 at S283, T494, and T505 increases its protein stability [12, 20]. Thus, we examined whether AURKA affects the protein levels of ALDH1A1. Ectopic overexpression of AURKA increased the levels of ALDH1A1 (Fig. 2e), and its depletion using corresponding shRNAs decreased it (Fig. 2f), suggesting that AURKA positively regulates ALDH1A1 levels. Figure 2g shows quantification of alterations in ALDH1A1 levels upon AURKA knock-down from three independent experiments. We also inhibited AURKA using MLN8237, which too resulted in substantial decrease in ALDH1A1 levels, suggesting that AURKA regulates ALDH1A1 using its kinase activity (Fig. 2h). We observed similar AURKA-mediated positive regulation of ALDH1A1 in Panc1 cells, suggesting that it is a common mechanism in pancreatic cancer cells (Fig. 2i–n). Data used to generate the summary statistics shown in Fig. 2g, j, l, and n are reported in Additional file 3.
AURKA inhibits ALDH1A1 degradation
AURKA increases the enzymatic activity of ALDH1A1
Increased ALDH1A1 levels and activity are hallmarks of cancer stem cells. Thus, we investigated whether AURKA modulates the enzymatic activity of ALDH1A1. We designed a two-step approach to phosphorylate ALDH1A1 and subsequently measure its dehydrogenase activity. Due to the nature of the assay, we selected five controls to account for any phosphorylation-independent change in dehydrogenase activity. Upon phosphorylation by AURKA, we observed a robust increase in ALDH1A1 activity relative to the phosphorylated controls (Fig. 3e). During our initial experiments, we observed an increase in activity when ALDH1A1 was incubated with ATP alone; however, this change was less than that observed when in the presence of AURKA and ATP (Fig. 3e). We hypothesized that this increase was likely due to phosphorylation by bacterial kinases co-purified with ALDH1A1 and/or a direct interaction with ATP itself. We thus carried out a high-stringency purification of ALDH1A1 to eliminate potential impurities, which resulted in a robust increase in ALDH1A1 activity only when AURKA was able to phosphorylate ALDH1A1 (Fig. 3f).
We observed that the increase in ALDH1A1 activity in these experiments is highly influenced by the activity of AURKA. Accordingly, if the ratio of phosphorylated to unphosphorylated ALDH1A1 was too low, the change in activity could not be observed. Thus, we needed to address the stoichiometry of phosphorylation and correlate it with the observed increase in activity.
To answer this question, we measured the activity and the percent mol phosphate incorporated per mole of ALDH1A1 (Fig. 3g). As expected, there was virtually no observable change in ALDH1A1 activity after 1.5 h (Fig. 3h). This was not surprising, as the stoichiometry of phosphorylation was a mere 3.1% (Fig. 3g). However, after a 6-h kinase reaction, the stoichiometry of phosphorylation reached 38%, which was accompanied by an observable increase in ALDH1A1 activity (Fig. 3i). This increase was prevalent in the first several hours of the ALDH1A1 activity, as seen in the projections of each plot after 2 h (Fig. 3h, i, right panel).
As an additional approach to confirm that the increase in ALDH1A1 activity was indeed due to direct phosphorylation by AURKA, we treated phosphorylated ALDH1A1 with calf-intestinal alkaline phosphatase (CIP) and monitored the impact on dehydrogenase activity (Fig. 3j). As predicted, the addition of CIP caused a decrease in the overall activity, supporting that the change in activity observed following the kinase assay was the result of direct phosphorylation by AURKA. Combined, these findings serve to validate that ALDH1A1 enzymatic activity is regulated via phosphorylation.
ALDH1A1 also positively regulates AURKA levels, triggering a feedback activation loop
AURKA phosphorylates ALDH1A1 at T267, T442, and T493
AURKA regulates the subcellular localization of ALDH1A1 via phosphorylation
To investigate the significance of AURKA-mediated phosphorylation of ALDH1A1, we initially investigated the subcellular localization of HA-tagged wild-type and mutant 3A-ALDH1A1 (T267A, T442A, T493A) in BxPC3 cells. While wild-type ALDH1A1 displayed cytoplasmic localization similar to the endogenous enzyme, the 3A mutant revealed perinuclear localization, showing that AURKA-mediated phosphorylation contributes to the cytoplasmic residence of ALDH1A1 (Fig. 5c). We further analyzed total ALDH1A1 (including endogenous levels) and ectopically expressed wild-type and mutant ALDH1A1 in the corresponding cell lines using ALDH1A1 and HA antibodies, respectively. While wild-type ALDH1A1-expressing cells showed diffused cytoplasmic staining of both endogenous and ectopically expressed ALDH1A1, 3A mutant-expressing cells showed distinct perinuclear staining of the 3A mutant but cytoplasmic staining for the endogenous enzyme (Fig. 5d). Together, these results underscore a role of AURKA in maintaining the cytoplasmic localization of ALDH1A1.
To investigate the contribution of each of the ALDH1A1 phosphorylation sites in regulating its subcellular localization, we generated HA-tagged wild-type and phospho-resistant single mutants of ALDH1A1-expressing BxPC3 cells, and examined their subcellular localization. Surprisingly, all single phospho-resistant mutants of ALDH1A1 showed predominantly cytoplasmic localization, suggesting that inhibition of phosphorylation at all sites is required for its perinuclear localization (Fig. 5e).
AURKA increases ALDH1A1 levels via phosphorylation at all three sites
We next examined the consequences of AURKA-mediated phosphorylation of ALDH1A1 on its expression levels in BxPC3 cells. These cells were transfected with wild-type ALDH1A1 and the 3A-ALDH1A1 allele for 30 h, and their protein levels analyzed. As shown in Fig. 5f, while wild-type ALDH1A1 was highly expressed, 3A-ALDH1A1 was expressed at a much lower level, suggesting that AURKA-mediated phosphorylation of ALDH1A1 is required to stabilize its protein levels. Furthermore, as 3A-ALDH1A1 was expressed at a lower level, it led to a concomitant decrease in the AURKA level, presumably due to the feedback activation loop.
Our data revealed that AURKA increases ALDH1A1 levels by inhibiting its ubiquitylation; thus, we examined whether 3A-ALDH1A1 was impervious to AURKA-mediated protein stability. We transiently depleted AURKA from wild-type ALDH1A1-BxPC3 and 3A-ALDH1A1-BxPC3 cells and examined the relative ubiquitylation levels of ALDH1A1. While wild-type ALDH1A1 was significantly degraded via ubiquitylation upon AURKA depletion, 3A-ALDH1A1 showed slight ubiquitylation, confirming that AURKA-mediated phosphorylation is responsible for increased ALDH1A1 stability (Fig. 5g).
AURKA-mediated stabilization of ALDH1A1 via phosphorylation prompted us to analyze potential degron sites in the ALDH1A1 sequence, which revealed T267 within a D box motif (RVT*L). Although APC-Cdh1-mediated degradation of D-box-containing proteins can occur independently of any modification, we hypothesized that AURKA-mediated stabilization of ALDH1A1 is likely due to the phosphorylation at T267. Thus, we generated stably expressing T267A-ALDH1A1-BxPC3 cells. As controls, we also generated the other two single-mutant expressing cells (T442A-ALDH1A1-BxPC3 and T493A-ALDH1A1-BxPC3 cells). ALDH1A1 levels were analyzed in wild-type, the three single mutant expressing BxPC3 cells, and 3A-BxPC3 cells. As hypothesized, the ALDH1A1 level was reduced in T267A-ALDH1A1-BxPC3 cells compared to wild-type ALDH1A1-expressing cells; however, interestingly, it was also reduced in T442A-ALDH1A1-BxPC3 and T493A-ALDH1A1-BxPC3 cells, with the triple mutant showing minimal protein levels, suggesting that phosphorylation of each of the three sites contributes to increased ALDH1A1 levels (Fig. 5h and i).
As our data showed that AURKA increases ALDH1A1 levels by inhibiting its ubiquitylation, we analyzed the contribution of each of these three sites in affecting protein stability. The mutant ALDH1A1-expressing cell lines displayed a similar steady state decrease in ALDH1A1 levels compared to wild-type ALDH1A1-expressing cells, as shown in Fig. 5h and i. We then transiently knocked down AURKA in these cells using AURKA shRNA, which resulted in robust ubiquitylation of wild-type ALDH1A1, followed by each of the three single mutants with relatively less ubiquitylation (Fig. 5j). The 3A-ALDH1A1 mutant displayed minimal ubiquitylation upon AURKA depletion, confirming its independence from AURKA-mediated phosphorylation. These results demonstrate that AURKA-mediated phosphorylation of ALDH1A1 at each of the three sites (T267, T442, and T493) contributes to increased protein stability.
AURKA increases ALDH1A1 activity predominantly via phosphorylation at the T267 site
AURKA-mediated phosphorylation of ALDH1A1 regulates its oligomeric states
ALDH1A1 can exist in monomeric, dimeric or tetrameric forms; however, their relative activities remain unknown. As phosphorylation is known to regulate the oligomeric distribution of several proteins, we hypothesized that AURKA-mediated phosphorylation of ALDH1A1 alters its oligomeric state, and that this change was coupled with the observed change in activity. Similar to most multimeric proteins, ALDH1A1 exists predominantly as a tetramer at high concentrations; however, as the dilution increases, the equilibrium becomes more dynamic and begins to populate the monomeric form. Surprisingly, very little dimer was observed at all concentrations, suggesting that the tetramer and the monomer are its most abundant forms.
We monitored the abundance of each species upon phosphorylation as a function of time to analyze the kinetics of the phosphorylation-dependent oligomeric redistribution of ALDH1A1. To this end, we chose a concentration which contained both tetrameric and monomeric species. As the phosphorylation progressed, the ALDH1A1 oligomeric distribution drastically shifted to the monomeric form (Fig. 6c). The concomitant decrease in tetramer abundance serves as an internal control to validate that it is indeed dissociating upon phosphorylation. While Fig. 6d shows average relative changes in tetramer and monomer abundance, Fig. 6e displays the average ratio of normalized monomer to tetramer as a function of time derived from three independent experiments.
We then sought to explore which oligomeric species had the highest activity with and without phosphorylation. To this end, we conducted a similar kinetic analysis and measured the dehydrogenase activity using a native in-gel activity assay (Fig. 6f, g). Upon phosphorylation the monomeric species that formed harbored the highest activity (a greater than sevenfold increase), whereas the tetramer showed little to no change. Figure 6h shows quantification of oligomer-specific ALDH1A1 activities from three independent experiments. This is the first example to identify the oligomer-specific catalytic activities of ALDH1A1 complexes following phosphorylation.
We next sought the oligomeric species that was most phosphorylated to examine a potential correlation between phosphorylation and activity. As predicted, the monomeric population that formed was highly phosphorylated (Fig. 6i–k). Collectively, these data suggest that phosphorylation of ALDH1A1 triggers the dissociation of the tetramer into its monomeric form and that this highly phosphorylated monomer harbors the highest enzymatic activity among all of the oligomeric species. Data used to generate the summary statistics shown in Fig. 6d, e, h, and k are reported in Additional file 5.
ALDH1A1 and AURKA feedback activation loop promotes highly aggressive pancreatic cancer phenotypes
The impact of ALDH1A1 was further evaluated in BxPC3, ALDH1A1-BxPC3 and 3A-ALDH1A1-BxPC3 cells under anchorage-independent conditions. ALDH1A1 expression led to a robust increase in colony formation in BxPC3 cells compared to parental BxPC3 cells. By contrast, expression of 3A-ALDH1A1 acted as dominant negative and exhibited minimal number of colonies in the soft agar assay (Fig. 7h). Similar results were also obtained in Panc1 cells (Fig. 7i). These findings show that AURKA-mediated phosphorylation of ALDH1A1 is crucial for promoting cell proliferation both under attached and anchorage-independent conditions in pancreatic cancer cells.
ALDH1A1 and AURKA feedback activation loop enhances cell motility
We next determined the contribution of ALDH1A1 in promoting chemotaxis using serum-starved BxPC3, ALDH1A1-BxPC3, and 3A-ALDH1A1-BxPC3 cells. A robust increase in cell motility was observed upon wild-type ALDH1A1 expression (Fig. 7j). However, expression of 3A-ALDH1A1 significantly impaired cell motility compared to vector-expressing BxPC3 cells, suggesting that 3A-ALDH1A1 may act as a dominant negative and inhibit cell motility. To further explore the impact of AURKA on ALDH1A1-mediated motility, we depleted AURKA in BxPC3, ALDH1A1-BxPC3, and 3A-ALDH1A1-BxPC3 cells, which considerably reduced cell motility in ALDH1A1-BxPC3 cells, but not in 3A-ALDH1A1-BxPC3 cells, suggesting that the ALDH1A1-triggered increase in cell motility is due to AURKA-mediated positive regulation (Fig. 7k). We further tested this hypothesis by overexpressing AURKA in BxPC3, ALDH1A1-BxPC3, and 3A-ALDH1A1-BxPC3 cells. AURKA overexpression increased cell migration in wild-type ALDH1A1-BxPC3 cells, but not in 3A-ALDH1A1-BxPC3 cells (Fig. 7l). This result was further validated in Panc1 cells, where wild-type ALDH1A1 expression showed a drastic increase in cell motility while 3A-ALDH1A1 showed significant inhibition when compared to wild-type Panc1 cells (Fig. 7m, n). Similarly, AURKA depletion significantly reduced cell motility in ALDH1A1-BxPC3 cells, but not in 3A-ALDH1A1-BxPC3 cells (Fig. 7o, p), while AURKA overexpression increased cell migration in wild-type ALDH1A1-Panc1 cells, but not in 3A-ALDH1A1-Panc1 cells (Fig. 7q, r). Data used to generate the summary statistics shown in Fig. 7c, f, g, i, n, p, and r are reported in Additional file 6. Together, these results corroborate that AURKA-mediated phosphorylation of ALDH1A1 plays a key role in increasing cell motility. As chemotaxis plays a key role in cancer metastasis, these results underscore a crucial oncogenic role of the AURKA-ALDH1A1 feedback activation loop in pancreatic malignancy.
ALDH1A1 upregulation is a key mechanism by which AURKA increases EMT and CSC phenotypes
To further confirm these results, we analyzed the protein levels of N-cadherin, CD44, Slug, Snail, and E-cadherin in BxPC3, wild-type ALDH1A1-BxPC3, and 3A-ALDH1A1-BxPC3 cells. E-cadherin is an epithelial cell marker, which is downregulated upon EMT, while Snail, Slug, and CD44p are increased in cells undergoing EMT and CSC. Similar to the promoter activation assays, N-cadherin, CD44, Slug, and Snail levels increased considerably upon ALDH1A1 expression, but were prevented in the presence of 3A-ALDH1A1 in BxPC3 cells, confirming that the phosphorylation-resistant mutant serves as a dominant negative (Fig. 8b). In addition, E-cadherin levels decreased in ALDH1A1-overexpressing cells, but showed a significant increase in 3A-ALDH1A1 cells. We also analyzed the levels of vimentin and matrix metalloproteinase-2 (MMP-2), levels of which are known to increase in EMT-undergoing cells and promote tumor invasion. ALDH1A1 overexpression increased both vimentin and MMP-2 levels in BxPC3 cells, whereas phospho-resistant mutant expression abolished their expression. Collectively, these results validate that AURKA-mediated phosphorylation plays a key role in ALDH1A1-mediated aggressive phenotypes.
In parallel, we also analyzed the levels of E-cadherin, N-cadherin, CD44, vimentin, MMP-2, Slug, and Snail in BxPC3 cells treated either with DMSO control or MLN8237. AURKA inhibition reduced the levels of N-cadherin, CD44, vimentin, MMP-2, Slug, and Snail, but increased E-cadherin levels in pancreatic cancer cell lines (Fig. 8c). These findings further underscored a key role of AURKA in promoting EMT in pancreatic cancer, presumably by modulating ALDH1A1 levels and activity.
We next conducted a sphere-forming assay to gauge the self-renewal capacity of BxPC3, ALDH1A1-BxPC3, and 3A-ALDH1A1-BxPC3 cells using ultralow attachment conditions. Under these conditions, cancer stem cells grow in suspension and form independent spheres or colonies. As shown in Fig. 8d, parental BxPC3 mainly formed aggregates of cells with no pancreatosphere formation. In contrast, overexpression of ALDH1A1 induced large pancreatosphere formation, confirming that ALDH1A1 overexpression causes the CSC phenotype. Importantly, 3A-ALDH1A1-BxPC3 cells also showed no pancreatosphere formation, thereby confirming that AURKA-mediated phosphorylation of ALDH1A1 is crucial for inducing the CSC phenotype.
ALDH1A1 phosphorylation contributes to drug resistance
As both EMT and CSC contribute to drug resistance, we examined doxorubicin sensitivity in BxPC3 cells, which revealed about 50% loss in cell viability in 72 h at 1 μM concentration (Fig. 8e). Expression of wild-type ALDH1A1 in BxPC3 rendered these cells significantly resistant to doxorubicin (~30% loss in cell viability in 72 h), whereas 3A-ALDH1A1 expression conferred high sensitivity to doxorubicin-induced toxicity (~70% loss in cell viability) (Fig. 8e). These findings suggest that AURKA-mediated phosphorylation of ALDH1A1 considerably contributes to the drug resistance observed in patients with pancreatic ductal adenocarcinoma (PDAC). These results led us to investigate whether ablation of AURKA sensitizes BxPC3 cells to ALDH1A1 inhibition. We used N,N-diethylaminobenzaldehyde (DEAB) to inhibit ALDH1A1, although it is not very selective and inhibits other ALDH isozymes as well. As shown in Fig. 8f, AURKA-depleted cells exhibit higher sensitivity to ALDH1A1 inhibition compared to scrambled shRNA-expressing cells, suggesting that targeting the AURKA-ALDH1A1 axis is likely to be more effective than individually targeting either AURKA or ALDH1A1.
To test this hypothesis, we treated BxPC3 cells with 100 μM DEAB for 24 h, which showed minimal toxicity, suggesting inhibition of ALDH1A1 alone does not confer toxicity to cells (Fig. 8g, h, bar graph and line graph, respectively). In contrast, AURKA inhibition showed ~32% and ~35% loss in cell viability at 1 and 1.5 μM MLN8237 concentrations, respectively. More importantly, co-inhibition of ALDH1A1 and AURKA was highly effective, particularly, at 1.5 μM MLN8237 concentration, which showed more than 65% loss in cell viability. These findings underscore the significance of the AURKA-ALDH1A1 feedback activation loop in pancreatic carcinoma and suggest that concurrent inhibition of AURKA and ALDH1A1 is likely to be highly effective in targeting highly chemoresistant PDAC (Fig. 8i).
ALDH1A1 is a cytosolic enzyme which catalyzes the oxidation of various aldehydes including trans- and cis-retinal to their corresponding less reactive acids, thereby aiding the cell in detoxifying these compounds. ALDH1A1 also binds to endobiotics and xenobiotics and possesses antioxidant activity [25, 26]. ALDH1A1 is highly expressed in normal stem cells, but is also widely used as a marker to identify and isolate various types of CSCs. ALDH1A1 level and activity are upregulated in patients treated with neoadjuvant chemotherapy and are associated with extreme chemoresistance and poor prognosis in pancreatic cancer and beyond. Thus, there is a critical need to develop therapies that target ALDH1A1 alone or in combination to eliminate these “stem cell-like” tumor cells in pancreatic and other types of cancer.
While several studies have shown increased mRNA or protein levels of ALDH1A1 in different cancers, the molecular mechanism remains unclear. There is no report that shows phosphorylation-dependent regulation of ALDH1A1 to date. We show that AURKA directly phosphorylates ALDH1A1 at T267, T442, and T493, which stabilizes ALDH1A1. ALDH1A1 also stabilizes AURKA levels by inhibiting its ubiquitylation, thereby triggering a reciprocal feedback activation loop. Interestingly, each of the three sites contributes towards ALDH1A1 stability. Furthermore, ectopic expression of phospho-resistant ALDH1A1 prevents the promoter activation of Slug, Snail, CD44, and N-cadherin while simultaneously failing to repress E-cadherin promoter, suggesting that inhibition of AURKA-mediated phosphorylation impairs its transcriptional activity. Importantly, ectopic expression of phospho-resistant ALDH1A1 not only fully inhibits ALDH1A1-mediated oncogenic pathways, but even reduces the basal levels of Snail, Slug, N-cadherin, vimentin, MMP-2, and CD44 in BxPC3 cells, highlighting a vital role of AURKA in ALDH1A1-mediated signaling in cancer cells.
We also show that AURKA-mediated phosphorylation at the T267 site increases ALDH1A1 activity. Recent global phospho-proteomics analysis studies indeed observed increased phosphorylation at the T267 site on ALDH1A1 in human colorectal cancer and non-small lung cancer tissues [27, 28]; however, the kinase responsible remains unknown. Similarly, Han et al. reported increased phosphorylation at T267 on ALDH1A1 in human liver tissues . ALDH1A1 is highly expressed in the liver and is crucial for the prevention of dietary retinol-induced liver toxicity , suggesting that AURKA-mediated increase in ALDH1A1 levels and activity may also be essential for normal tissues which require high ALDH1A1 activity. Thus, these studies not only underscore the clinical significance of AURKA-mediated regulation of ALDH1A1 in several cancers, but also highlight that the AURKA-ALDH1A1 axis may have an important role in normal tissue homeostasis.
Furthermore, our study showed that AURKA phosphorylates ALDH1A1 at T493, which is located within its oligomerization domain. ALDH1A1 exists as a tetramer, which is a dimer of a dimer (A-B + C-D). However, the relative enzymatic activities of its monomeric, dimeric, and tetrameric forms still remain unclear. One study reported that heating of ALDH1A1 increases the conversion of the dimeric form to the monomeric form, which is associated with decreased activity, although this study did not analyze the activity of the tetrameric form of ALDH1A1 . Class II ALDH enzymes, such as ALDH2, display half-of-the-sites reactivity in tetrameric form, which can assume full reactivity depending on the pH or in the presence of divalent cations. Accordingly, increasing the Mg2+ concentration from 0 to 0.4 mM resulted in a two-fold increase in horse liver ALDH2 activity which was accompanied by the dissociation of tetrameric enzyme into dimers . Similarly, with increasing pH, the tetrameric ALDH2 enzyme dissociated to the more active dimeric state even in the absence of Mg2+ ions . These findings suggest that Class II ALDHs display higher activity in the dimeric state; however, it is not known whether ALDH1A1 (which belongs to Class I) displays similar kinetics in the dimeric and tetrameric states. Furthermore, the activity of a Class I ALDH such as ALDH1A1 is inhibited in the presence of Mg2+, suggesting that Class I and Class II ALDHs are regulated differently. To date, no study has investigated phosphorylation as a potential mechanism which governs these reported changes in ALDH1A1’s oligomeric state and enzymatic activity.
Our data show that AURKA-mediated phosphorylation of ALDH1A1 shifts the total population to its monomeric form, which harbors higher dehydrogenase activity relative to the tetramer. Surprisingly, unlike Class II ALDH, we did not observe substantial dimeric species in either the unphosphorylated or phosphorylated state. It is likely that in cells, the monomeric phosphorylated form may be re-incorporated into high order oligomers to form the active tetramer when ALDH1A1 levels are high. Alternatively, the cells may keep ALDH1A1 activity in check by storing excess enzyme in the less active tetrameric form, which is capable of readily dissociating to form the highly active phospho-monomer for rapid temporal response. While future studies are needed to uncover this mechanism, the phospho-oligomeric regulation of ALDH1A1 could be exploited as a novel therapeutic intervention.
In conclusion, we unraveled a novel mechanism of ALDH1A1 regulation that is triggered by AURKA. While extreme chemoresistance and CSC formation are the defining features of pancreatic cancer, the mechanism by which AURKA may be involved in these processes has not been examined. Our discovery of ALDH1A1 as an AURKA substrate provides a novel mechanism by which AURKA promotes chemoresistance and stem cell formation in pancreatic cancer via ALDH1A1 and vice versa. Thus, targeting AURKA and ALDH1A1 in combination is expected to be highly effective in inhibiting tumorigenesis, chemoresistance, and metastasis in highly lethal pancreatic carcinoma. Furthermore, analysis of ALDH1A1 phosphorylation levels should aid in the development of diagnostic and prognostic biomarkers for identifying patients with suitable molecular phenotypes for maximal therapeutic benefit, avoiding treatment of unresponsive patients and improving overall quality of life.
Aldehyde dehydrogenase 1A1
Cancer stem cell
Nitrotetrazolium blue chloride
The pcDNA3-HA-ADH was a gift from Steven Johnson (Addgene plasmid # 11610).
This work was supported by the National Institute of Health (R03 CA 166912).
Availability of data and materials
All the data on which the conclusions of the paper are based are presented in the paper and the additional files.
JW conducted most of the BxPC3 cell experiments. KN performed all Panc1 and several BxPC3 experiments. LC performed a few BxPC3 cell experiments. KV generated wild-type and various ALDH1A1 mutants, conducted kinase and dehydrogenase assays, and performed and analyzed oligomerization data as depicted in Fig. 5j. JWh participated in the in vitro ALDH1A1 assay. KS analyzed the data and wrote the manuscript with KV. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Alian OM, Philip PA, Sarkar FH, Azmi AS. Systems biology approaches to pancreatic cancer detection, prevention and treatment. Curr Pharm Des. 2004;20(1):73–80.View ArticleGoogle Scholar
- Castellanos JA, Merchant NB, Nagathihalli NS. Emerging targets in pancreatic cancer: epithelial-mesenchymal transition and cancer stem cells. Onco Targets Ther. 2013;6:1261–7.PubMedPubMed CentralGoogle Scholar
- Fukushige S, Waldman FM, Kimura M, Abe T, Furukawa T, Sunamura M, Kobari M, Horii A. Frequent gain of copy number on the long arm of chromosome 20 in human pancreatic adenocarcinoma. Genes Chromosomes Cancer. 1997;19(3):161–9.View ArticlePubMedGoogle Scholar
- Li D, Zhu J, Firozi PF, Abbruzzese JL, Evans DB, Cleary K, Friess H, Sen S. Overexpression of oncogenic STK15/BTAK/AURKA kinase in human pancreatic cancer. Clin Cancer Res. 2003;9(3):991–7.PubMedGoogle Scholar
- Rojanala S, Han H, Muñoz RM, Browne W, Nagle R, Von Hoff DD, Bearss DJ. The mitotic serine threonine kinase, Aurora-2, is a potential target for drug development in human pancreatic cancer. Mol Cancer Ther. 2004;3(4):451–7.PubMedGoogle Scholar
- Hata T, Furukawa T, Sunamura M, Egawa S, Motoi F, et al. RNA interference targeting Aurora kinase A suppresses tumor growth and enhances the taxane chemosensitivity in human pancreatic cancer cells. Cancer Res. 2005;65(7):2899–905.View ArticlePubMedGoogle Scholar
- Durlacher CT, Li ZL, Chen XW, He ZX, Zhou SF. An update on the pharmacokinetics and pharmacodynamics of alisertib, a selective Aurora kinase A inhibitor. Clin Exp Pharmacol Physiol. 2016;43(6):585–601.View ArticlePubMedGoogle Scholar
- Mizukami T, Kamachi H, Mitsuhashi T, Tsuruga Y, Hatanaka Y, Kamiyama T, Matsuno Y, Taketomi A. Immunohistochemical analysis of cancer stem cell markers in pancreatic adenocarcinoma patients after neoadjuvant chemoradiotherapy. BMC Cancer. 2014;14:687.View ArticlePubMedPubMed CentralGoogle Scholar
- Sun KH, de Pablo Y, Vincent F, Johnson EO, Chavers AK, Shah K. Novel genetic tools reveal Cdk5′s major role in Golgi fragmentation in Alzheimer’s disease. Mol Biol Cell. 2008;19(7):3052–69.View ArticlePubMedPubMed CentralGoogle Scholar
- Sun KH, de Pablo Y, Vincent F, Shah K. Deregulated Cdk5 promotes oxidative stress and mitochondrial dysfunction. J Neurochem. 2008;107(1):265–78.View ArticlePubMedGoogle Scholar
- Shah K, Shokat KM. A chemical genetic screen for direct v-Src substrates reveals ordered assembly of a retrograde signaling pathway. Chem Biol. 2002;9(1):35–47.View ArticlePubMedGoogle Scholar
- Johnson EO, Chang KH, de Pablo Y, Ghosh S, Mehta R, Badve S, Shah K. PHLDA1 is a critical negative regulator and effector of AURKA kinase in breast cancer. J Cell Sci. 2011;124(Pt 16):2711–22.View ArticlePubMedGoogle Scholar
- Moffat J, Grueneberg DA, Yang X, Kim SY, Kloepfer AM, Hinkle G, Piqani B, Eisenhaure TM, Luo B, Grenier JK, Carpenter AE, Foo SY, Stewart SA, Stockwell BR, Hacohen N, Hahn WC, Lander ES, Sabatini DM, Root DE. A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen. Cell. 2006;124(6):1283–98.View ArticlePubMedGoogle Scholar
- Shah K, Vincent F. Divergent roles of c-Src in controlling platelet-derived growth factor-dependent signaling in fibroblasts. Mol Biol Cell. 2005;16(11):5418–32.View ArticlePubMedPubMed CentralGoogle Scholar
- Sun KH, Lee HG, Smith MA, Shah K. Direct and indirect roles of cyclin-dependent kinase 5 as an upstream regulator in the c-Jun NH2-terminal kinase cascade: relevance to neurotoxic insults in Alzheimer’s disease. Mol Biol Cell. 2009;20(21):4611–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Shah K, Liu Y, Deirmengian C, Shokat KM. Engineering unnatural nucleotide specificity for Rous sarcoma virus tyrosine kinase to uniquely label its direct substrates. Proc Natl Acad Sci U S A. 1997;94(8):3565–70.View ArticlePubMedPubMed CentralGoogle Scholar
- Shah K, Shokat KM. A chemical genetic approach for the identification of direct substrates of protein kinases. Methods Mol Biol. 2003;233:253–71.PubMedGoogle Scholar
- Kim S, Shah K. Dissecting yeast Hog1 MAP kinase pathway using a chemical genetic approach. FEBS Lett. 2007;581(6):1209–16.View ArticlePubMedGoogle Scholar
- Chang KH, Multani PS, Sun KH, Vincent F, de Pablo Y, Ghosh S, Gupta R, Lee HP, Lee HG, Smith MA, Shah K. Nuclear envelope dispersion triggered by deregulated Cdk5 precedes neuronal death. Mol Biol Cell. 2011;22(9):1452–62.View ArticlePubMedPubMed CentralGoogle Scholar
- Johnson EO, Chang KH, Ghosh S, Venkatesh C, Giger K, Low PS, Shah K. LIMK2 is a crucial regulator and effector of Aurora-A-kinase-mediated malignancy. J Cell Sci. 2012;125(Pt 5):1204–16.View ArticlePubMedGoogle Scholar
- Chang KH, Vincent F, Shah K. Deregulated Cdk5 triggers aberrant activation of cell cycle kinases and phosphatases inducing neuronal death. J Cell Sci. 2012;125(Pt 21):5124–37.View ArticlePubMedGoogle Scholar
- Shi C, Viccaro K, Lee HG, Shah K. Cdk5-FOXO3a axis: initially neuroprotective, eventually neurodegenerative in Alzheimer’s disease models. J Cell Sci. 2016;129:1815–30.View ArticlePubMedGoogle Scholar
- Ferrari S, Marin O, Pagano MA, Meggio F, Hess D, et al. Aurora-A site specificity: a study with synthetic peptide substrates. Biochem J. 2005;390:293–302.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang F, Li H, Yan XG, Zhou ZW, Yi ZG, He ZX, Pan ST, Yang YX, Wang ZZ, Zhang X, Yang T, Qiu JX, Zhou SF. Alisertib induces cell cycle arrest and autophagy and suppresses epithelial-to-mesenchymal transition involving PI3K/Akt/mTOR and sirtuin 1-mediated signaling pathways in human pancreatic cancer cells. Drug Des Devel Ther. 2015;9:575–601.PubMedPubMed CentralGoogle Scholar
- Vasiliou V, Nebert DW. Analysis and update of the human aldehyde dehydrogenase (ALDH) gene family. Hum Genomics. 2005;2:138–43.PubMedPubMed CentralGoogle Scholar
- Marchitti SA, Brocker C, Stagos D, Vasiliou V. Non-P450 aldehyde oxidizing enzymes: the aldehyde dehydrogenase superfamily. Expert Opin Drug Metab Toxicol. 2008;4:697–720.View ArticlePubMedPubMed CentralGoogle Scholar
- Klammer M, Kaminski M, Zedler A, Oppermann F, Blencke S, Marx S, Müller S, Tebbe A, Godl K, Schaab C. Phosphosignature predicts dasatinib response in non-small cell lung cancer. Mol Cell Proteomics. 2012;11(9):651–68.View ArticlePubMedPubMed CentralGoogle Scholar
- Shiromizu T, Adachi J, Watanabe S, Murakami T, Kuga T, Muraoka S, Tomonaga T. Identification of missing proteins in the neXtProt database and unregistered phosphopeptides in the PhosphoSitePlus database as part of the Chromosome-centric Human Proteome Project. J Proteome Res. 2013;12(6):2414–21.View ArticlePubMedGoogle Scholar
- Han G, Ye M, Liu H, Song C, Sun D, Wu Y, Jiang X, Chen R, Wang C, Wang L, Zou H. Phosphoproteome analysis of human liver tissue by long-gradient nanoflow LC coupled with multiple stage MS analysis. Electrophoresis. 2010;31(6):1080–9.PubMedGoogle Scholar
- Molotkov A, Duester G. Genetic evidence that retinaldehyde dehydrogenase Raldh1 (Aldh1a1) functions downstream of alcohol dehydrogenase Adh1 in metabolism of retinol to retinoic acid. J Biol Chem. 2003;278(38):36085–90.View ArticlePubMedGoogle Scholar
- Gondhowiardjo TD, van Haeringen NJ, Kijlstra A. Molecular weight forms of corneal aldehyde dehydrogenase. Curr Eye Res. 1992;11(4):377–81.View ArticlePubMedGoogle Scholar
- Takahashi K, Weiner H. Magnesium stimulation of catalytic activity of horse liver aldehyde dehydrogenase. Changes in molecular weight and catalytic sites. J Biol Chem. 1980;255(17):8206–9.PubMedGoogle Scholar
- Takahashi K, Weiner H, Filmer DL. Effects of pH on horse liver aldehyde dehydrogenase: alterations in metal ion activation, number of functioning active sites, and hydrolysis of the acyl intermediate. Biochemistry. 1981;20(21):6225–30.View ArticlePubMedGoogle Scholar