Abstract
Activation of the serine/threonine kinase Akt contributes to the formation, maintenance, and therapeutic resistance of cancer, which is driving development of compounds that inhibit Akt. Phosphatidylinositol ether lipid analogues (PIAs) are analogues of the products of PI3K that inhibit Akt activation, translocation, and the proliferation of a broad spectrum of cancer cell types. To gain insight into the mechanism of PIAs, time-dependent transcriptional profiling of 5 active PIAs and the PI3K inhibitor LY294002 (LY) was performed in NSCLC cells using high-density oligonucleotide arrays. Gene ontology analysis revealed genes involved in apoptosis, wounding response, and angiogenesis were upregulated by PIAs, while genes involved in DNA replication, repair and mitosis were suppressed. Genes that exhibited early differential expression were partitioned into 3 groups; those induced by PIAs only (DUSP1, KLF6, CENTD2, BHLHB2, PREX1), those commonly induced by PIAs and LY (TRIB1, KLF2, RHOB and CDKN1A), and those commonly suppressed by PIAs and LY (IGFBP3, PCNA, PRIM1, MCM3 and HSPA1B). Increased expression of the tumor suppressors RHOB (RhoB), KLF6 (COPEB) and CDKN1A (p21Cip1/Waf1) was validated as an Akt-independent effect that contributed to PIA-induced cytotoxicity. Despite some overlap with LY, active PIAs have a distinct expression signature that contributes to their enhanced cytotoxicity.
Keywords: Akt, microarray, PIA, RhoB, NSCLC
Introduction
The PI3K/Akt/mTOR pathway is a promising target in cancer since its activation promotes cellular growth, survival and contributes to tumorigenesis in vivo, while inhibition of the pathway promotes apoptosis in cancer cells and increases responsiveness to chemotherapy or radiation (1–5). Akt has an important role in lung cancer since it is activated in response to tobacco components in vitro (6, 7), and the phenotypic progression of tobacco carcinogen-induced lung lesions is dependent upon activation of Akt and mTOR (8, 9). In NSCLC, Akt activation is specific for tumor tissues vs. surrounding normal lung tissues and confers a poor prognosis (10). Despite the strong rationale to target Akt, however, few Akt inhibitors exist.
To address this need, we used molecular modeling to synthesize structurally modified phosphatidylinositol ether lipid analogues (PIAs) designed to interfere with the pleckstrin homology (PH) domain of Akt (11). Five PIAs were identified that rapidly inhibited Akt activation, as well as the phosphorylation of multiple downstream substrates without affecting kinases upstream of Akt (12). PIAs selectively induced apoptosis in NSCLC and breast cancer cell lines with high endogenous levels of Akt activation. Although the PIAs appeared interchangeable in their abilities to inhibit Akt and cause cell death, they induced more cell death than an established PI3K inhibitor, LY294002, despite similar inhibition of the Akt pathway, which suggests PIAs might have additional targets. Support for this hypothesis came from studies of PIAs in the NCI60 cell line panel where activity of PIAs correlated with levels of phosphorylated but not total Akt, but other targets with higher correlation coefficients were identified (13).
Microarrays have been used to query transcriptional programs that underlie processes relevant to cancer such as proliferation (14), transformation (15), senescence (16), metastasis (17), epithelial to mesenchymal transition (18) and activation of oncogenic pathways (19). Elucidation of these programs is important to the development of new therapies. For example, transcriptional profiling of normal vs. tumor tissues has led to the identification of new targets and pharmacodynamic biomarkers to predict efficacy and clinical outcome (20, 21). Profiling of drug-induced gene transcription has been used to uncover the mechanism of action of novel agents, elucidate structure activity relationships, and to determine on-target vs. off-target effects (22, 23).
To elucidate transcriptional changes that could mediate the cytotoxic activity of PIAs, expression profiling in NSCLC cells was performed. We segregated changes in gene expression that were shared by both PIAs and LY294002 and therefore likely due to effects on the PI3K/Akt pathway, from those that were unique to PIAs and therefore might be considered “off-target” effects. Although greatly overlapping with LY294002 in suppression of cell cycle genes, active PIAs uniquely or potently induced a number of tumor suppressor genes that might contribute to biological properties of PIAs that extend beyond inhibition of Akt. This expression profile could underlie their enhanced toxicity and could be utilized in pharmacodynamic studies of PIAs.
Materials and Methods
Cell Lines and Materials
NSCLC cell lines (H157, A549, H1703 and H1155) were obtained from NCI/Navy Medical Oncology (Bethesda, MD). They were maintained in RPMI medium 1640 with 10% (v/v) fetal bovine serum (FBS), and incubated at 37°C in a 5.0% CO2 atmosphere. All lines were recently tested and authenticated by the Core Fragment Analysis Facility (Johns Hopkins University, Baltimore, MD) using a short tandem repeat (STR) profiling in accordance with AACR best practices. The synthesis of the PIAs has previously been described (11). LY294002 was purchased from Calbiochem (San Diego, CA). Antibodies to phospho-Akt (S473), Akt1, Akt2, Akt3, HSP70 and anti-mouse or anti-rabbit secondary antibodies were purchased from Cell Signaling Technologies (Beverly, MA). The DNA primase (p49) antibody was from Lab Vision Corporation (Fremont, CA). Antibodies to KLF6, MCM3, PCNA and IGFBP3 as well as anti-goat secondary antibody were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). RhoB antibody was purchased from Proteintech Group, Inc. (Chicago, IL). Protease inhibitor cocktail tablets were obtained from Roche Diagnostics GmbH (Mannheim, Germany) and the Micro BCA Protein Assay Kit was from PIERCE (Rockford, IL). The pcDNA3-HA-RhoB was a kind gift from Dr. George Prendergast. The pCMV6-KLF6 and -CDKN1A were from OriGene (Rockville, MD). The pcDNA3-Myr-HA-Akt1 (24) was provided by Dr. William Sellers via Addgene (Cambridge, MA). RhoB, KLF6 and CDKN1A On-Target plus human siRNAs were from Dharmacon/Thermo (Lafayette, CO). Protran pure nitrocellulose membranes were purchased from Schleicher & Schuell (Dassel, Germany). All cell culture reagents were purchased from Life Technologies, Inc. (Rockville, MD).
Pharmacological Treatment and oligonucleotide microarray analysis
NSCLC cells were plated 2×105 cells per well in 6-well plates or 2×106 in T-75 flasks in RPMI medium 1640 containing 10% FBS and incubated for 24h. The medium was then changed to RPMI medium 1640 with 0.1% FBS and the cells were incubated overnight. The following morning, cells were treated with 10 μM PIA6 dissolved in DMSO for 0h, 2h, 6h or 12h, and an equal volume of DMSO was added to control samples. For the PIA comparison, 10 μM PIAs (5, 6, 7, 23, 24, 25) or 10 μM LY294002 (LY) were incubated with the cells for 6h. PIA7 (an inert analog consisting of only the lipid side chain) was used as a control. The structures of the PIAs and LY294002 are shown in Figure 1A. Cell viability was not affected in 0.1% FBS for the duration of these experiments. In cells cultured with 5% FBS, PIAs are highly bound to serum proteins and higher concentrations are needed to observe the same effects. Following incubation, the alterations in cellular morphology were photographed, and cells from 6-well plates were harvested for immunoblot analysis. Total RNA was extracted from cells treated in T-75 flasks using TRIzol reagent (Invitrogen) and chloroform and purified according to the RNeasy midiprep spin kit protocol (Qiagen). Oligonucleotide microarray was performed with dye-swap. Microarray chips were generated from the 34,580 longmer probe set Human Genome Oligo Set Version 3.0 (Qiagen). Protocols for cDNA labeling, hybridization, and scanning are available through the National Human Genome Research Institute microarray core. The raw data were deposited in a public functional genomics data repository Gene Expression Omnibus (GSE27911; ref. 25). Immunoblotting analysis was performed as described previously (26).
RT-PCR
Semi-quantitative RT-PCR was performed using the SuperScript™ One-Step RT-PCR System with Platinum® Taq DNA polymerase (Invitrogen). The following forward and reverse primers were used: (1) DUSP1, 5′-ctgccttgatcaacgtctca-3′ and 5′-acccttcctccagcattctt-3′; (2) KLF6, 5′-ggcaacagacctgcctagag-3′ and 5′-aggattcgctgctgacatct-3′; (3) CEND2, 5′-gctttgaggtcaacgagagg-3′ and 5′-gaagtagcgatcgtggaagc-3′; (4) BHLHB2, 5′-ccttgaagcatgtgaaagca-3′ and 5′-gcttggccagatactgaagc-3; (5) PREX1, 5′-ccctggtcagtgaagagagc-3′ and 5′-tcatctccagaccccatctc-3′; (6) TRIB1, 5′-tctggctttgaggcttgttt-3′ and 5′-cagcccagagtccttagtcg-3′; (7) KLF2, 5′-agagggtctccctcgatgac-3′ and 5′-tctcacaaggcatcacaagc-3′; (8) RHOB, 5′-cgacgtcattctcatgtgct-3′ and 5′-cgaggtagtcgtaggcttgg-3′; (9) CDKN1A, 5′-atgaaattcaccccctttcc-3′ and 5′-ccctaggctgtgctcacttc-3′; (10) C21orf58, 5′-cctcttccatcacggaggta-3′ and 5′-ggcacacaggtgtccctagt; (11) IGFBP3, 5′-cagagactcgagcacagcac-3′ and 5′-gatgaccggggtttaaaggt-3′; (12) PCNA, 5′-ggcgtgaacctcaccagtat-3′ and 5′-tctcggcatatacgtgcaaa-3′; (13) PRIM1, 5′-gccatacgcatcattgacag-3′ and 5′-ccaccctttacaaggctcaa-3′; (14) MCM3, 5′-cgcaggaaaaacgagaagag-3′ and 5′-cagaccacacagctgaggaa-3′; (15) HSPA1B, 5′-ccgagaaggacgagtttgag-3′ and 5′-gcagcaaagtccttgagtcc-3′; (16) GAPDH, 5′-gagtcaacggatttggtcgt-3′ and 5′-ttgattttggagggatctcg-3′.
Bioinformatics Tools for Gene Clustering, Visualization and Ontology
The microarray outputs were clustered and visualized by Cluster 3.0 (27) and Java TreeView (28). Gene expression dynamics was analyzed by CAGED program (Cluster Analysis of Gene Expression Dynamics) (29). For gene ontology analysis, the High-Throughput GoMiner web interface (30) was used as described (31).
Cell Transfection and Infection
Transfection of plasmid or siRNA was performed with a Nucleofector device using program T-16 and transfection kit V (Lonza). Cells stably expressing Myr-Akt1 were created following plasmid transfection by G418 (800 μg/ml) selection for 2 weeks. Cell lines expressing Akt isoform specific shRNAs were created by lentiviral infection and shRNA vectors used were from Sigma-Aldrich unless otherwise noted: Akt1, NM_005163.1-628s1c1; Akt2, NM_001626.2-1509s1c1; Akt3, NM_005465.3-671s1c1; non-targeting, pLKO-scr (Addgene). Gene overexpression or knockdown was verified by immunoblotting.
MTS Assay and FACS Analysis
The MTS assay was performed with CellTiter 96 Aqueous One Solution Reagent (Promega) according to the manufacturer’s instructions, and the cell viability was determined by measuring the absorbance at 490 nm using a BioTek ELx800 Microplate Reader. Apoptosis (sub-G1 DNA) was quantified by propidium iodide staining and analysis using a Becton Dickinson FACSort flow cytometer and CELLQuest software.
Results
Optimization of PIA Treatments and Microarray Analysis
Preliminary experiments were performed to optimize conditions for microarray analysis. Previously, we observed that PIAs cause profound morphologic changes in NSCLC cells, including rounding and detachment. To assess the time dependence of these changes, H157 cells were treated with PIA6 and observed over time (Figure 1B). At 2h, there was little morphologic change, but by 6h, the cells had become highly refractile and rounded. Between 6 and 12h, cellular detachment occurred. Similar time dependent changes were observed with other active PIAs, but not an inactive PIA (PIA7) or LY (data not shown). In addition, PIA exposure caused similar morphologic changes in other NSCLC cell lines, but with different kinetics. For example, these changes were delayed in A549 and H1703 cells, but accelerated in H1155 cells (data not shown). In H157 cells treated with PIA6, the surviving fractions measured by MTS assay at 2, 6 and 12h were 95%, 79% and 48%, respectively. These experiments suggest that at treatment times up to 6h, cellular detachment would not confound the measurement of gene expression changes induced by PIAs.
To assess Akt inhibition, immunoblotting was performed with parallel samples prepared from H157 cells (Figure 1C). PIA6 inhibited Akt phosphorylation at S473 at 2, 6 and 12h (left panels). Treatment with any of 5 active PIAs or LY also decreased S473 phosphorylation in H157 cells at 6h (right panels). PIA7, an analog that lacks the inositol ring, did not inhibit Akt phosphorylation. To assure that RNA quality and integrity were maintained with increasing times of exposure to PIAs, analysis using a Bioanalyzer Nanochip was performed. The 28S and 18S rRNA bands were sharp up to 12h and the 28S bands were more intense than 18S bands, indicating the RNA quality was adequate (Fig. 1D, left panels). RNA integrity was also preserved in samples treated with all PIAs or LY for 6h (Fig. 1D, right panels). Based on the assessment of cellular morphology, Akt inhibition and RNA quality, 6h was chosen as the time point at which to compare changes in gene expression with PIAs and LY.
Following microarray analysis, 911 genes were identified that exhibited differential expression by treatment with one or more of the 5 active PIAs in H157 cells (using a cutoff of a 2-fold change in expression). A hierarchical clustered heat map and cluster tree of these data is presented in Figure 1E. The expression of genes in response to LY is shown in the far right column. Clustering of expression signatures revealed that PIA23 and PIA25 showed the highest degree of similarity to each other. Although PIA5 showed similarity to PIA24, it altered fewer genes and many less potently than PIA24. PIA6 shared the most genes in common with the other 4 according to the clustering and was relatively close to the PIA23 and PIA25 subgroup.
To quantify genes that changed in common with PIAs and LY, Venn Mapper (32) was used to calculate the number of differentially expressed genes that overlapped between each PIA treatment and LY, as well as the corresponding Z-scores (Supplemental Table 1). Of genes that increased, PIA23 and PIA25 each shared 33 genes with LY, with the highest positive Z-scores of 5.8 and 7.2, respectively. However, PIA5, 6, and 24 had little overlap with LY, resulting in negative Z-scores. Of genes that decreased, PIA5, 6, 23, 24 and 25 had 13, 11, 43, 25 and 35 genes in common with LY, respectively, all with positive Z-scores (1.0, 4.5, 9.3, 3.3 and 8.2, respectively). Given that the cutoff Z-score for statistical significance (a p-value of <0.05) was 1.96 or greater (http://www.gatcplatform.nl/vennmapper/index.php), these data indicate that PIA6, 23, 24 and 25 shared genes that decreased in common with LY in a statistically significant manner, but only PIA23 and 25 shared genes that increased in common with LY in a statistically significant manner (99.9% confidence level). Interestingly, the Z-scores between any two PIAs were positive for increased or decreased comparisons but negative at increased-decreased or decreased-increased comparisons, suggesting these compounds affected gene expression in a similar direction but to different extents. The complete list of differentially expressed genes and enlarged heat map are shown in Supplemental Table 2 and Supplemental Figure 1.
Early Changes in Gene Expression Caused in Common by PIAs
From the clustered heat map, 83 transcripts were identified that were similarly regulated (either positively or negatively) by all 5 active PIAs. These were partitioned into 4 gene categories (C1-4) using k-Means clustering (Fig. 2A). C1 and C2 were likely due to Akt pathway inhibition, since C1 includes genes that were induced by PIAs and LY, and C2 includes genes that were repressed by both PIAs and LY. C3 and C4 include PIA specific genes that are either repressed or induced by PIAs, respectively.
To identify the earliest changes in gene expression that occurred following PIA treatment, a time course experiment was performed. RNA was isolated after 0, 2, 6 or 12h exposure to PIA6. Sixty genes were extracted from 83 transcripts similarly regulated by PIAs at the 4 time points, and their expression levels with PIA6 exposure are depicted in Fig. 2B. Because temporal patters of gene expression can be useful to identify common regulatory mechanisms, a temporal cluster analysis was performed using the CAGED program (Cluster Analysis of Gene Expression Dynamics). This program identified 15 clusters that exhibited distinct dynamic patterns over time, with 4 genes for each cluster (Fig. 2C, and Supplemental Figure 2). Because we wanted to identify the early and constitutively up-regulated or down-regulated genes that changed with time, clusters 2, 3, 6, 11, 14 and 15 were chosen for further analysis. The 24 genes from these 6 clusters were further partitioned based on data from the PIA and LY comparisons, as well as the time course experiments (10 different conditions using CAGED), which generated 3 groups that exhibited similar patterns (Fig. 2D). The genes in groups I and II were upregulated by PIAs. Group I is characterized by genes not induced by LY. Group II is characterized by genes more strongly induced by PIA treatment, but commonly induced by LY. In contrast, group III genes were downregulated by PIAs and in most cases, LY. The degree of downregulation caused by PIAs was similar to that of LY. The identity of genes comprising these groups, along with attributions of their function, process and subcellular location, is listed in Fig. 2E.
Validation of Microarray Data
Changes in gene expression depicted in Groups I-III were validated using semi-quantitative RT-PCR (Fig. 3A). Time dependence (left panels) and specificity for individual PIAs vs. LY (right panels) were assessed. From group I, DUSP1, KLF6, CENTD2, BHLHB2 and PREX1 were selected for validation. KLF6 and DUSP1 were strongly induced by PIA6 from 2h to 12h, CENTD2 and BHLHB2 were slightly induced at 2h and strongly induced from 6h to 12h, and PREX1 was induced beginning at 6h. The mRNA levels of DUSP1, CENTD2, BHLHB2 and PREX1 were increased in the DMSO-treated sample at 12h, indicating that these genes may be sensitive to culture conditions such as nutrient consumption and/or increases in cell density. DUSP1, KLF6, CENTD2, BHLHB2 and PREX1 were all induced by the 5 active PIAs, and not by LY or PIA7. These data confirm that these genes are rapidly induced by PIAs and are selective for active PIAs.
The 4 genes chosen from group II, TRIB1, KLF2, RHOB and CDKN1A, were induced by PIA6 from 2h to 12h. In particular, RHOB and KLF2 were strongly induced at early time points. The basal level of KLF2 expression was hardly detectable, and its induction appeared transient with peak expression at 2h. Induction of TRIB1 and KLF2 was not selective for PIAs, because LY had similar effects (Figure 3A, middle-right panel). RHOB and CDKN1A were also induced by the 5 active PIAs, but were less potently by LY. Although individual variation was observed, expression of these genes from group II was similarly regulated by PIAs and LY, suggesting that these changes were likely due to inhibition of the PI3K/Akt pathway.
Six genes were chosen from group III for validation. Expression of IGFBP3, PCNA and PRIM1 decreased from 2h to 12h. Decreased expression of C21orf58, MCM3 and HSPA1B was evident from 6h to 12h. When assessed for specificity between PIAs and LY, group III genes were repressed by both PIAs and LY, except for C21orf58, which was not inhibited by LY. Collectively, the expression of these genes as assessed by RT-PCR is in agreement with the microarray data.
To confirm that changes in mRNA expression would lead to protein level alterations, immunoblotting was performed (Fig. 3B). Due to limited availability of reliable antibodies, only a subset of the genes could be assessed. From group I, KLF6 protein expression increased after PIA treatment in 4 NSCLC cell lines. From group II, PIA treatment increased CDKN1A expression and markedly induced RhoB expression. When genes from group III were analyzed, the protein levels of PIA-repressed genes such as IGFBP3, PCNA, MCM3 and HSPA1B did not appreciably decrease at 6h. This could be related to slow protein turnover. To test this, we performed a longer time course and found these proteins were decreased at 12h and 24h (Fig. 3C). These immunoblotting experiments validate protein expression as a readout for genes that are regulated by PIAs.
Identification of biologically-relevant gene categories altered by PIAs
To explore the biological significance of the microarray data, gene ontology analysis was performed by uploading the total gene list represented on the chip, along with the induced or repressed genes, into the High-Throughput GoMiner web interface. A summary of this analysis from the 6h PIA comparison experiment is presented in Table 1. The “total genes” column indicates how many genes on the chip were assigned to the respective categories. The “changed genes” column indicates how many of the genes that changed significantly (>2 fold) during treatment were assigned to the targeted GO categories. The enrichment, p value and false discovery rate (FDR) of the GO categories were calculated and listed with an FDR cutoff of <0.05. Some general categories such as regulation of cellular, physiological or biological processes had the lowest p-values, yet these communicate little meaning. The categories with the next most significant p-values were more relevant, namely apoptosis and cell death, followed by response to wounding, cell cycle, and angiogenesis. Apoptosis is a known cellular outcome of PIA treatment (12). The “under-expressed” GO category result indicates genes associated with DNA-dependent DNA replication were inhibited by PIAs. The individual genes assigned to these GO categories are listed in Supplementary Table 3.
Table 1.
OVEREXPRESSION | |||||
---|---|---|---|---|---|
GO CATEGORY | TOTAL GENES | CHANGED GENES | ENRICHMENT | LOG10(p) | FDR |
GO:0050789_regulation_of_biological_process | 2170 | 24 | 2.1105 | −4.5343 | 0.0000 |
GO:0050791_regulation_of_physiological_process | 1998 | 22 | 2.1012 | −4.0073 | 0.0000 |
GO:0051244_regulation_of_cellular_physiological_process | 1873 | 21 | 2.1395 | −3.8992 | 0.0033 |
GO:0050794_regulation_of_cellular_process | 1963 | 21 | 2.0414 | −3.5890 | 0.0050 |
GO:0006915_apoptosis | 356 | 8 | 4.2882 | −3.3683 | 0.0080 |
GO:0012501_programmed_cell_death | 358 | 8 | 4.2642 | −3.3519 | 0.0067 |
GO:0008219_cell_death | 377 | 8 | 4.0493 | −3.2021 | 0.0086 |
GO:0016265_death | 379 | 8 | 4.0280 | −3.1868 | 0.0075 |
GO:0009611_response_to_wounding | 324 | 7 | 4.1228 | −2.8881 | 0.0300 |
GO:0048519_negative_regulation_of_biological_process | 464 | 8 | 3.2901 | −2.6194 | 0.0256 |
GO:0051243_negative_regulation_of_cellular_physiological_process | 396 | 7 | 3.3732 | −2.3934 | 0.0310 |
GO:0007050_cell_cycle_arrest | 42 | 3 | 13.6304 | −2.8766 | 0.0270 |
GO:0000074_regulation_of_cell_cycle | 327 | 7 | 4.0849 | −2.8648 | 0.0245 |
GO:0007049_cell_cycle | 504 | 8 | 3.0290 | −2.3968 | 0.0326 |
GO:0042330_taxis | 109 | 4 | 7.0028 | −2.6160 | 0.0228 |
GO:0006935_chemotaxis | 109 | 4 | 7.0028 | −2.6160 | 0.0228 |
GO:0048514_blood_vessel_morphogenesis | 46 | 3 | 12.4451 | −2.7616 | 0.0233 |
GO:0001525_angiogenesis | 46 | 3 | 12.4451 | −2.7616 | 0.0233 |
GO:0001568_blood_vessel_development | 46 | 3 | 12.4451 | −2.7616 | 0.0233 |
GO:0001944_vasculature_development | 46 | 3 | 12.4451 | −2.7616 | 0.0233 |
UNDEREXPRESSION | |||||
GO:0006261_DNA-dependent_DNA_replication | 58 | 3 | 26.3207 | −3.7506 | 0.0400 |
Time dependent changes in gene ontology were also assessed (Table 2). The categories of genes induced earliest were death-related. This process continued and was further defined as apoptosis from 6h to 12h. By 12h, several categories were added, including protein and macromolecule biosynthesis, wound healing, and angiogenesis. The first categories of repressed genes included DNA-dependent DNA replication and cell cycle regulation, which were only evident beginning at 6h. By 12h, several other categories were added, similar to the observations with induced genes. Collectively, the gene ontology analysis suggests that although PIAs induce changes in expression of a fairly small number of genes, these changes become manifest by altering many cellular processes in ways that would likely be detrimental to the growth and survival of cancer cells.
Table 2.
GO CATEGORY | Changed/Total genes (FDR) | ||
---|---|---|---|
OVEREXPRESSION | 2h | 6h | 12h |
GO:0001944_vasculature_development | 10/46 (0.0227) | ||
GO:0001568_blood_vessel_development | 10/46 (0.0227) | ||
GO:0048514_blood_vessel_morphogenesis | 10/46 (0.0227) | ||
GO:0001525_angiogenesis | 10/46 (0.0227) | ||
GO:0050878_regulation_of_body_fluids | 16/98 (0.0238) | ||
GO:0007599_hemostasis | 15/87 (0.0209) | ||
GO:0042060_wound_healing | 15/87 (0.0209) | ||
GO:0009611_response_to_wounding | 41/324 (0.0200) | ||
GO:0050817_coagulation | 15/81 (0.0150) | ||
GO:0007596_blood_coagulation | 15/81 (0.0150) | ||
GO:0006412_protein_biosynthesis | 47/371 (0.0200) | ||
GO:0009059_macromolecule_biosynthesis | 49/409 (0.0133) | ||
GO:0006915_apoptosis | 8/356 (0.030) | 43/356 (0.0129) | |
GO:0012501_programmed_cell_death | 4/358 (0.05) | 8/358 (0.015) | 43/358 (0.0111) |
GO:0008219_cell_death | 4/377 (0.04) | 8/377 (0.020) | 45/377 (0.0200) |
GO:0016265_death | 4/379 (0.03) | 8/379 (0.015) | 45/379 (0.0125) |
UNDEREXPRESSION | |||
GO:0006261_DNA-dependent_DNA_replication | 3/58 (0.0100) | 8/58 (0.0000) | |
GO:0006260_DNA_replication | 3/113 (0.0118) | 14/113 (0.0000) | |
GO:0006259_DNA_metabolism | 5/365 (0.0075) | 23/365 (0.0000) | |
GO:0006270_DNA_replication_initiation | 5/14 (0.0000) | ||
GO:0006310_DNA_recombination | 5/48 (0.0076) | ||
GO:0006139_nucleobase__nucleoside__nucleotide_and_nucleic_acid_metabolism | 49/1892 (0.0000) | ||
GO:0043283_biopolymer_metabolism | 29/998 (0.0042) | ||
GO:0006281_DNA_repair | 13/142 (0.0000) | ||
GO:0006974_response_to_DNA_damage_stimulus | 14/157 (0.0000) | ||
GO:0009719_response_to_endogenous_stimulus | 14/166 (0.0000) | ||
GO:0006950_response_to_stress | 24/755 (0.0045) | ||
GO:0007049_cell_cycle | 6/504 (0.0050) | 25/504 (0.0000) | |
GO:0000074_regulation_of_cell_cycle | 5/327 (0.0100) | 14/327 (0.0031) | |
GO:0051244_regulation_of_cellular_physiological_process | 10/1873 (0.0071) | 42/1873 (0.0368) | |
GO:0000279_M_phase | 9/149 (0.0038) | ||
GO:0051301_cell_division | 8/119 (0.0033) | ||
GO:0000910_cytokinesis | 8/119 (0.0033) | ||
GO:0000278_mitotic_cell_cycle | 10/177 (0.0040) | ||
GO:0000086_G2_M_transition_of_mitotic_cell_cycle | 4/37 (0.0245) | ||
GO:0000087_M_phase_of_mitotic_cell_cycle | 7/121 (0.0248) | ||
GO:0007067_mitosis | 7/119 (0.0258) |
Akt dependence and responses to PIAs
Although a COMPARE analysis indicated the cytotoxicity of PIAs correlated with active Akt levels (13), a functional analysis between PIA cytotoxicity and Akt activation in an isogenic system has not been done. To address this, we created an H157 cell line stably transfected with constitutively active MyrAkt1. Although MyrAkt1-expressing cells showed lower basal levels of apoptosis as indicated by cleaved PARP and sub-G1 DNA content, apoptosis was further induced with PIA23 treatment (Fig. 4A). Similar results were observed when other apoptotic assays such as Annexin-V/PI co-staining were employed (data not shown). These findings were confirmed in an A549 isogenic system, in which the 3 Akt isoforms were individually stably knocked down by lentiviral infection with shRNAs. Immunoblotting confirmed Akt isoform specific knockdown, and also demonstrated that Akt1 was the major isoform in A549 cells, because only Akt1 knockdown decreased levels of total and phospho-Akt. Accordingly, only Akt1 knockdown resulted in significantly less apoptotic cell death with PIA treatment (Fig. 4B). These studies demonstrated levels of active Akt, specifically Akt1, correlated with PIA cytotoxicity.
To address the Akt-dependence of PIA-induced genes, we used genetic or pharmacologic approaches to modulate Akt, and measured levels of RhoB, KLF6, and p21 after PIA treatment. In H157 cells transfected with MyrAkt1 or vector, induction of RhoB, KLF6 or p21 by PIA23 was observed (Fig. 4C). Although the induction of KLF6 and p21 by PIA23 in MyrAkt1 transfected cells appears slightly diminished compared to vector transfected cells, this is likely an artifact related to lower expression of p42/44 MAPK under these experimental conditions, which was observed in replicate experiments. When Akt1 was knocked down in A549 cells, the induction of RhoB, KLF6 and p21 by PIA23 was not affected (Fig. 4D). To confirm these results, we pretreated H157 cells with LY for 30 min followed by 6h treatment with PIA23. LY alone slightly induced RhoB, KLF6 and p21 protein levels, but the combination of LY with PIA23 enhanced the expression of the PIA-induced genes over either compound alone (Fig. 4E). These results indicate that induction of these tumor suppressors is only minimally dependent upon the Akt pathway.
An important question is whether any of PIA-induced genes identified contribute to the cytotoxicity of the compounds. To examine this, H157 cells were transiently transfected with RHOB, KLF6 or CDKN1A siRNAs and treated with PIA 48h later. Cell lysates were harvested after 6h to assess knockdown, and sub-G1 DNA analysis was performed after 12h PIA treatment. The results show that although the siRNAs did not entirely block the induction of their target genes, these greatly rescued H157 cells from apoptosis caused by PIA (Fig. 4F). In contrast, overexpression of these genes either individually or in combination significantly decreased the viability of H157 cells (Fig. 4G). Similar results were observed in other NSCLC cell lines such as H1155 and H2882, and in other cancer cell lines with high levels of endogenous Akt activation (data not shown). These data confirm RhoB, KLF6 and p21 induction contribute to the cytotoxicity of PIAs.
Discussion
Using microarray analysis, we identified gene expression profiles that contribute to the biologic effects of PIAs. Validation of individual gene changes using RT-PCR and immunoblotting showed that microarray results could be validated at both the mRNA and protein levels, albeit protein level decreases were delayed as compared to mRNA decreases for the PIA-suppressed genes. The fact that induction of genes by PIAs could be measured by PCR or immunoblotting suggests that these genes could serve as biomarkers for PIA administration.
To place these individual changes in gene expression in a biologic framework, gene ontology analysis was performed and revealed that many cellular processes are altered due to PIA-induced changes in gene expression in a time-dependent manner. Early induction of apoptosis or cell death and repression of DNA replication and cell cycle were observed after PIA administration, which is consistent with development of PIAs as anti-cancer agents. Of the early-induced genes, KLF6, RHOB/RhoB and CDKN1A/p21 were of particular interest because they are known tumor suppressors and their expression reduced overall cell viability and contributed to PIA-induced cytotoxicity.
RhoB is a small GTPase tumor suppressor that regulates actin organization and vesicle transport. It is required for signalling apoptosis in transformed cells that are exposed to chemotherapeutic agents, has a negative modifier function in carcinogenesis (33), and its expression is repressed during NSCLC progression (34–36). Although RhoB has a reciprocal relationship with levels of Akt activation in cells (37, 38), our studies showed that inhibition of Akt through genetic or pharmacologic means did not significantly affect induction of RhoB by PIAs. This suggests a novel mechanism for RhoB induction by PIAs.
CDKN1A/p21 inhibits cell cycle by binding to cyclin/CDK complexes and PCNA in nucleus, which can be prevented by Akt via phosphorylation (39). Because induction of p21 was observed in cell lines that have either WT (A549) or mutant p53 (H157, H1703 and H1155), induction of p21 by PIAs is p53-independent. This observation is consistent with p53-independent induction of p21 by perifosine, an alkylphospholipid whose activity correlated highly with PIAs (13, 40). Interestingly, the induction of RhoB and p21 by PIAs might be related, because PPARγ-mediated induction of p21 in anaplastic thyroid carcinoma is dependent upon up-regulation of RhoB (41).
KLF6 is a member of the Krüppel-like factor family of C2H2-type zinc finger-containing transcription factors implicated in cellular differentiation and tissue development (42). The KLF6 gene encodes a family of proteins generated through alternative splicing, which results in at least four isoforms. Full length KLF6 is a tumor suppressor that is frequently inactivated by loss of heterozygosity (LOH), somatic mutation, and/or decreased expression in human cancer. Its tumor suppressor roles are not completely known, but a number of highly relevant activities have been described such as transactivation of p21 in a p53-independent manner (43, 44), reduction of cyclin D1/cdk4 complexes via interaction with cyclin D1 (45), and induction of apoptosis through up-regulation of ATF3 (46).
Genetic manipulation demonstrated that active Akt1 is a predictor of PIA efficacy in NSCLC cells since expression of constitutively active Akt increased cytotoxicity and knockdown of Akt1 protected against PIA-induced cytotoxicity. PIA treatment induced expression of KLF6, RHOB/RhoB and CDKN1A/p21 in a mostly Akt-independent manner. Nonetheless, these tumor suppressors inhibited cellular viability when overexpressed, and contributed to the cytotoxicity of PIAs. These studies suggest that the broad activity of PIAs against cancer cells is based upon multiple independent mechanisms that include inhibition of Akt (12) and activation of p38α (47) and AMPKα (26), as well as induction of tumor suppressor genes such as KLF6, RhoB and p21. As a result, the development of biomarkers for PIAs will be complex and will need to account for PIA-induced changes in enzymatic activity as well as gene transcription.
Supplementary Material
Acknowledgments
This research was supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research, and in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract NCI Contract NO1-CO-12400. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
Abbreviations
- PI3K
phosphatidylinositol 3′ kinase
- NSCLC
non-small cell lung cancer
- PIA
phosphatidylinositol ether lipid analogue
- LY
LY294002
- MAPK
mitogen-activated protein kinase
Footnotes
The authors have no conflict of interest.
References
- 1.LoPiccolo J, Granville CA, Gills JJ, Dennis PA. Targeting Akt in cancer therapy. Anticancer Drugs. 2007;18:861–74. doi: 10.1097/CAD.0b013e3280cc2c6f. [DOI] [PubMed] [Google Scholar]
- 2.Brognard J, Clark AS, Ni Y, Dennis PA. Akt/protein kinase B is constitutively active in non-small cell lung cancer cells and promotes cellular survival and resistance to chemotherapy and radiation. Cancer Res. 2001;61:3986–97. [PubMed] [Google Scholar]
- 3.Clark AS, West K, Streicher S, Dennis PA. Constitutive and inducible Akt activity promotes resistance to chemotherapy, trastuzumab, or tamoxifen in breast cancer cells. Mol Cancer Ther. 2002;1:707–17. [PubMed] [Google Scholar]
- 4.Granville CA, Memmott RM, Gills JJ, Dennis PA. Handicapping the race to develop inhibitors of the phosphoinositide 3-kinase/Akt/mammalian target of rapamycin pathway. Clin Cancer Res. 2006;12:679–89. doi: 10.1158/1078-0432.CCR-05-1654. [DOI] [PubMed] [Google Scholar]
- 5.LoPiccolo J, Blumenthal GM, Bernstein WB, Dennis PA. Targeting the PI3K/Akt/mTOR pathway: effective combinations and clinical considerations. Drug Resist Updat. 2008;11:32–50. doi: 10.1016/j.drup.2007.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.West KA, Brognard J, Clark AS, Linnoila IR, Yang X, Swain SM, et al. Rapid Akt activation by nicotine and a tobacco carcinogen modulates the phenotype of normal human airway epithelial cells. J Clin Invest. 2003;111:81–90. doi: 10.1172/JCI16147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Tsurutani J, Castillo SS, Brognard J, Granville CA, Zhang C, Gills JJ, et al. Tobacco components stimulate Akt-dependent proliferation and NFkappaB-dependent survival in lung cancer cells. Carcinogenesis. 2005;26:1182–95. doi: 10.1093/carcin/bgi072. [DOI] [PubMed] [Google Scholar]
- 8.West KA, Linnoila IR, Brognard J, Belinsky S, Harris C, Dennis PA. Tobacco carcinogen-induced cellular transformation increases Akt activation in vitro and in vivo. Chest. 2004;125 (Suppl 5):101–2. doi: 10.1378/chest.125.5_suppl.101s. [DOI] [PubMed] [Google Scholar]
- 9.Granville CA, Warfel N, Tsurutani J, Hollander MC, Robertson M, Fox SD, et al. Identification of a highly effective rapamycin schedule that markedly reduces the size, multiplicity, and phenotypic progression of tobacco carcinogen-induced murine lung tumors. Clin Cancer Res. 2007;13:2281–9. doi: 10.1158/1078-0432.CCR-06-2570. [DOI] [PubMed] [Google Scholar]
- 10.Tsurutani J, Fukuoka J, Tsurutani H, Shih JH, Hewitt SM, Travis WD, et al. Evaluation of two phosphorylation sites improves the prognostic significance of Akt activation in non-small-cell lung cancer tumors. J Clin Oncol. 2006;24:306–14. doi: 10.1200/JCO.2005.02.4133. [DOI] [PubMed] [Google Scholar]
- 11.Kozikowski AP, Sun H, Brognard J, Dennis PA. Novel PI analogues selectively block activation of the pro-survival serine/threonine kinase Akt. J Am Chem Soc. 2003;125:1144–5. doi: 10.1021/ja0285159. [DOI] [PubMed] [Google Scholar]
- 12.Castillo SS, Brognard J, Petukhov PA, Zhang C, Tsurutani J, Granville CA, et al. Preferential inhibition of Akt and killing of Akt-dependent cancer cells by rationally designed phosphatidylinositol ether lipid analogues. Cancer Res. 2004;64:2782–92. doi: 10.1158/0008-5472.can-03-1530. [DOI] [PubMed] [Google Scholar]
- 13.Gills JJ, Holbeck S, Hollingshead M, Hewitt SM, Kozikowski AP, Dennis PA. Spectrum of activity and molecular correlates of response to phosphatidylinositol ether lipid analogues, novel lipid-based inhibitors of Akt. Mol Cancer Ther. 2006;5:713–22. doi: 10.1158/1535-7163.MCT-05-0484. [DOI] [PubMed] [Google Scholar]
- 14.Cho RJ, Huang M, Campbell MJ, Dong H, Steinmetz L, Sapinoso L, et al. Transcriptional regulation and function during the human cell cycle. Nat Genet. 2001;27:48–54. doi: 10.1038/83751. [DOI] [PubMed] [Google Scholar]
- 15.Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, et al. Large-scale meta-analysis of cancer microarray data identifies common transcriptional profiles of neoplastic transformation and progression. Proc Natl Acad Sci U S A. 2004;101:9309–14. doi: 10.1073/pnas.0401994101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hardy K, Mansfield L, Mackay A, Benvenuti S, Ismail S, Arora P, et al. Transcriptional networks and cellular senescence in human mammary fibroblasts. Mol Biol Cell. 2005;16:943–53. doi: 10.1091/mbc.E04-05-0392. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasis in primary solid tumors. Nat Genet. 2003;33:49–54. doi: 10.1038/ng1060. [DOI] [PubMed] [Google Scholar]
- 18.LaGamba D, Nawshad A, Hay ED. Microarray analysis of gene expression during epithelial-mesenchymal transformation. Dev Dyn. 2005;234:132–42. doi: 10.1002/dvdy.20489. [DOI] [PubMed] [Google Scholar]
- 19.Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D, et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature. 2006;439:353–7. doi: 10.1038/nature04296. [DOI] [PubMed] [Google Scholar]
- 20.Strand KJ, Khalak H, Strovel JW, Ebner R, Augustus M. Expression biomarkers for clinical efficacy and outcome prediction in cancer. Pharmacogenomics. 2006;7:105–15. doi: 10.2217/14622416.7.1.105. [DOI] [PubMed] [Google Scholar]
- 21.van‘t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature. 2002;415:530–6. doi: 10.1038/415530a. [DOI] [PubMed] [Google Scholar]
- 22.Butcher RA, Schreiber SL. Using genome-wide transcriptional profiling to elucidate small-molecule mechanism. Curr Opin Chem Biol. 2005;9:25–30. doi: 10.1016/j.cbpa.2004.10.009. [DOI] [PubMed] [Google Scholar]
- 23.Marton MJ, DeRisi JL, Bennett HA, Iyer VR, Meyer MR, Roberts CJ, et al. Drug target validation and identification of secondary drug target effects using DNA microarrays. Nat Med. 1998;4:1293–301. doi: 10.1038/3282. [DOI] [PubMed] [Google Scholar]
- 24.Ramaswamy S, Nakamura N, Vazquez F, Batt DB, Perera S, Roberts TM, et al. Regulation of G1 progression by the PTEN tumor suppressor protein is linked to inhibition of the phosphatidylinositol 3-kinase/Akt pathway. Proc Natl Acad Sci U S A. 1999;96:2110–5. doi: 10.1073/pnas.96.5.2110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gene Expression Omnibus. The National Center for Biotechnology Information (NCBI) NIH; Available from: http://www.ncbi.nlm.nih.gov/geo. [Google Scholar]
- 26.Memmott RM, Gills JJ, Hollingshead M, Powers MC, Chen Z, Kemp B, et al. Phosphatidylinositol ether lipid analogues induce AMP-activated protein kinase-dependent death in LKB1-mutant non small cell lung cancer cells. Cancer Res. 2008;68:580–8. doi: 10.1158/0008-5472.CAN-07-3091. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Available from: http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm#ctv.
- 28.Available from: http://sourceforge.net/projects/jtreeview.
- 29.Available from: http://www.genomethods.org/caged.
- 30.Available: http://discover.nci.nih.gov/gominer/htgm.jsp.
- 31.Zeeberg BR, Qin H, Narasimhan S, Sunshine M, Cao H, Kane DW, et al. High-Throughput GoMiner, an ‘industrial-strength’ integrative gene ontology tool for interpretation of multiple-microarray experiments, with application to studies of Common Variable Immune Deficiency (CVID) BMC Bioinformatics. 2005;6:168. doi: 10.1186/1471-2105-6-168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Smid M, Dorssers LC, Jenster G. Venn Mapping: clustering of heterologous microarray data based on the number of co-occurring differentially expressed genes. Bioinformatics. 2003;19:2065–71. doi: 10.1093/bioinformatics/btg282. [DOI] [PubMed] [Google Scholar]
- 33.Prendergast GC. Actin’ up: RhoB in cancer and apoptosis. Nat Rev Cancer. 2001;1:162–8. doi: 10.1038/35101096. [DOI] [PubMed] [Google Scholar]
- 34.Wang S, Yan-Neale Y, Fischer D, Zeremski M, Cai R, Zhu J, et al. Histone deacetylase 1 represses the small GTPase RhoB expression in human nonsmall lung carcinoma cell line. Oncogene. 2003;22:6204–13. doi: 10.1038/sj.onc.1206653. [DOI] [PubMed] [Google Scholar]
- 35.Sato N, Fukui T, Taniguchi T, Yokoyama T, Kondo M, Nagasaka T, et al. RhoB is frequently downregulated in non-small-cell lung cancer and resides in the 2p24 homozygous deletion region of a lung cancer cell line. Int J Cancer. 2007;120:543–51. doi: 10.1002/ijc.22328. [DOI] [PubMed] [Google Scholar]
- 36.Mazieres J, Tovar D, He B, Nieto-Acosta J, Marty-Detraves C, Clanet C, et al. Epigenetic regulation of RhoB loss of expression in lung cancer. BMC Cancer. 2007;7:220. doi: 10.1186/1471-2407-7-220. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Jiang K, Delarue FL, Sebti SM. EGFR, ErbB2 and Ras but not Src suppress RhoB expression while ectopic expression of RhoB antagonizes oncogene-mediated transformation. Oncogene. 2004;23:1136–45. doi: 10.1038/sj.onc.1207236. [DOI] [PubMed] [Google Scholar]
- 38.Jiang K, Sun J, Cheng J, Djeu JY, Wei S, Sebti S. Akt mediates Ras downregulation of RhoB, a suppressor of transformation, invasion, and metastasis. Mol Cell Biol. 2004;24:5565–76. doi: 10.1128/MCB.24.12.5565-5576.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Zhou BP, Liao Y, Xia W, Spohn B, Lee MH, Hung MC. Cytoplasmic localization of p21Cip1/WAF1 by Akt-induced phosphorylation in HER-2/neu-overexpressing cells. Nat Cell Biol. 2001;3:245–52. doi: 10.1038/35060032. [DOI] [PubMed] [Google Scholar]
- 40.Patel V, Lahusen T, Sy T, Sausville EA, Gutkind JS, Senderowicz AM. Perifosine, a novel alkylphospholipid, induces p21(WAF1) expression in squamous carcinoma cells through a p53-independent pathway, leading to loss in cyclin-dependent kinase activity and cell cycle arrest. Cancer Res. 2002;62:1401–9. [PubMed] [Google Scholar]
- 41.Marlow LA, Reynolds LA, Cleland AS, Cooper SJ, Gumz ML, Kurakata S, et al. Reactivation of suppressed RhoB is a critical step for the inhibition of anaplastic thyroid cancer growth. Cancer Res. 2009;69:1536–44. doi: 10.1158/0008-5472.CAN-08-3718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Atkins GB, Jain MK. Role of Kruppel-like transcription factors in endothelial biology. Circ Res. 2007;100:1686–95. doi: 10.1161/01.RES.0000267856.00713.0a. [DOI] [PubMed] [Google Scholar]
- 43.Li D, Yea S, Dolios G, Martignetti JA, Narla G, Wang R, et al. Regulation of Kruppel-like factor 6 tumor suppressor activity by acetylation. Cancer Res. 2005;65:9216–25. doi: 10.1158/0008-5472.CAN-05-1040. [DOI] [PubMed] [Google Scholar]
- 44.Narla G, Heath KE, Reeves HL, Li D, Giono LE, Kimmelman AC, et al. KLF6, a candidate tumor suppressor gene mutated in prostate cancer. Science. 2001;294:2563–6. doi: 10.1126/science.1066326. [DOI] [PubMed] [Google Scholar]
- 45.Benzeno S, Narla G, Allina J, Cheng GZ, Reeves HL, Banck MS, et al. Cyclin-dependent kinase inhibition by the KLF6 tumor suppressor protein through interaction with cyclin D1. Cancer Res. 2004;64:3885–91. doi: 10.1158/0008-5472.CAN-03-2818. [DOI] [PubMed] [Google Scholar]
- 46.Huang X, Li X, Guo B. KLF6 induces apoptosis in prostate cancer cells through up-regulation of ATF3. J Biol Chem. 2008;283:29795–801. doi: 10.1074/jbc.M802515200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Gills JJ, Castillo SS, Zhang C, Petukhov PA, Memmott RM, Hollingshead M, et al. Phosphatidylinositol ether lipid analogues that inhibit AKT also independently activate the stress kinase, p38alpha, through MKK3/6-independent and -dependent mechanisms. J Biol Chem. 2007;282:27020–9. doi: 10.1074/jbc.M701108200. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.