We have previously described that P-Ser⁸³-HP1γ by PKA mediates extracellular signals during interphase 8 . In the current study, we uncover a new Aurora kinase A-mediated pathway that phosphorylates Ser⁸³-HP1γ during mitosis, which is necessary for the proper execution of this process. For this purpose, we initially analyzed the kinetics of phosphorylation in HeLa cells arrested in different phases of the cell cycle. Treatment with roscovitine, a membrane permeable cyclin-dependent kinase (CDK) inhibitor, that arrests cell cycle progression at the G₁/S and G₂/M checkpoints 16 , resulted in dose-dependent inhibition of P-Ser⁸³-HP1γ (Figure  1A). To better define the temporal pattern of these events, we treated with either aphidicolin to arrest cells in S phase, or nocodazole to obtain mitotic arrest (G₂/M). The mitotic population demonstrated a striking increase in P-Ser⁸³-HP1γ levels in comparison to the normal cycling population and S phase arrested cells (Figure  1B). To define these events in the absence of kinase inhibitors, we synchronized HeLa cells by double thymidine block to obtain cell extracts at subsequent time points of release from cell cycle arrest. These experiments revealed that the levels of P-Ser⁸³-HP1γ peaked twice, the first at 2 hours post-release (G₁/S boundary, Figure  1C, Additional file 1: Figure S1 A). As this peak was likely the phosphorylation event coinciding with the previously described involvement of PKA during interphase 8 , we utilized the PKA-specific inhibitor, KT5720, to treat HeLa cells upon release from double thymidine block. Upon KT5720 treatment, P-Ser⁸³-HP1γ levels at 2 hours post-release were significantly diminished (Additional file 1: Figure S1 B). However, of greater interest, a more prominent second peak across 8 to 10 hours post-release from cell cycle arrest, which coincided with G₂/M, was observed (Figure  1C, Additional file 1: Figure S1A). The lower P-Ser⁸³-HP1γ levels seen in-between these two peaks (4 to 6 hours post-release, Figure  1C) corresponded with S phase (Additional file 1: Figure S1A), similar to aphidicolin treatment. These results demonstrate that levels of P-Ser⁸³-HP1γ peak significantly at G₂/M phase during the cell cycle, suggesting that phosphorylation of this protein may play a role in cell division.

Subsequently, we sought to complement the biochemical assays of phosphorylation described above by mapping the temporal pattern of staining for P-Ser⁸³-HP1γ during cell cycle progression. For this purpose, we performed immunofluorescence using confocal microscopy in cells co-stained with the anti-P-Ser⁸³-HP1γ and different cell cycle markers. We utilized cyclin D as a marker of G₁, 5-ethynyl-2´-deoxyuridine (EdU)-pulse labeling for S phase, and cyclin B to indicate the G₂ and M phases of the cell cycle. Figure  2A,B,C, which represents a low magnification field of cells stained with the anti-P-Ser⁸³-HP1γ, demonstrates that the level and distribution of the signal for this modified form of HP1γ is variable in epithelial cells growing under normal conditions. Thus, we examined more carefully the levels and distribution of P-Ser⁸³-HP1γ signals in relationship to key cell cycle markers. P-Ser⁸³-HP1γ localization in cyclin D-positive cells (G₁) appeared in the euchromatic compartment of the nucleus as a fine punctate pattern (Figure  2D,E,F). Quantification of cyclin D-positive cells demonstrated that 76.6% of this population (160/209) had P-Ser⁸³-HP1γ staining. However, staining was relatively negligible in cells that were positively marked by a short pulse of EdU, indicative of S phase (Figure  2G,H,I) with only 22.7% of EdU-positive cells (34/150) having any P-Ser⁸³-HP1γ signal. The strongest P-Ser⁸³-HP1γ signal was found in 88.3% of cyclin B-positive cells (182/206), which corresponded to G₂ (Figure  2J,K,L), and the signal continued through M in prometaphase, metaphase and anaphase, until returning to similar levels as G₁ during telophase and cytokinesis (Figure  2M,N,O,P,Q,R). Thus, these results were congruent with our biochemical studies and confirmed that P-Ser⁸³-HP1γ occurs as two peaks, beginning at G₁ and ending at S, and the second peak which begins at G₂ and continues during M. Interestingly, a conspicuous feature of P-Ser⁸³-HP1γ localization was its staining in cyclin B-positive cells for which the nuclear membrane has not yet disassembled (late G₂ prophase), in which the P-Ser⁸³-HP1γ punctate pattern was stronger and present not only in euchromatin but also within centrosomes (Figure  2L). Although the cyclin B-positive cells found in M demonstrated reduced P-Ser⁸³-HP1γ signal on chromosomes, a strong signal continued to localize at the centrosome region of the mitotic spindle (Figure  2M,N,O,P). In all these cases, P-Ser⁸³-HP1γ coincided with the presence of cyclin B at the centrosome. As several mitotic kinases are highly enriched at this organelle 17 , these studies prompted us to identify the kinase responsible for the significant P-Ser⁸³-HP1γ event found during mitotic progression.

While PKA was implicated in the first peak of P-Ser⁸³-HP1γ levels that occur at G₁, the kinase that mediates the second peak of P-Ser⁸³-HP1γ at G₂/M, described here, remained unknown. Interestingly, we found that the temporal pattern of P-Ser⁸³-HP1γ coincided with phosphorylation of histone H3 at serine 10 (P-Ser¹⁰-H3, Figure  1C). P-Ser¹⁰-H3 initiates during G₂ in pericentric foci and spreads along the chromosome arms, thus serving as a hallmark of mitosis 18 . Previously derived consensus sequences for Aurora kinases suggested that, similar to P-Ser¹⁰-H3, Ser⁸³-HP1γ might be a target of Aurora kinases 19 . Additional experiments demonstrated that the temporal pattern of P-Ser⁸³-HP1γ was similar to both Aurora A and Aurora B (Figure  1C). These results led us to hypothesize that the newly described P-Ser⁸³-HP1γ at G₂/M was achieved through the activity of an Aurora kinase. Thus, we first performed immunofluorescence experiments to determine whether P-Ser⁸³-HP1γ co-localized with any of these kinases at G₂/M. Indeed, we found that P-Ser⁸³-HP1γ localized to areas rich in Aurora A (Figure  3A,B,C), but not Aurora B (Figure  3D,E,F). P-Ser⁸³-HP1γ was also confirmed to be present at the Aurora A-rich area of the spindle poles through colocalization with γ-tubulin (Figure  3G,H,I) and α-tubulin (Figure  3J,K,L). More importantly, we found that critical regulators of G₂/M progression, which are also targets of Aurora A, namely cyclin B1, cyclin B2 and their partner kinase, CDK1, also colocalized with P-Ser⁸³-HP1γ (Figure  3M,N,O,P,Q,U). Together, these results demonstrated that mitotic phosphorylation confers a distinct localization of this HP1γ subpopulation to the spindle poles that is marked by the G₂/M Aurora A-cyclin B-CDK1 pathway, supporting the idea that this kinase may be the enzyme involved in P-Ser⁸³-HP1γ at G₂/M.

To mechanistically test this hypothesis, we initially incubated glutathione S-transferase (GST) fusion wild type and nonphosphorylatable mutant HP1γ proteins with each Aurora kinase, Aurora A or Aurora B, followed by western blot using the phospho-specific P-Ser⁸³-HP1γ antibody. These in vitro kinase assays demonstrated that the wild type HP1γ, but not the dominant negative, nonphosphorylatable S83A-HP1γ mutant 8 , could be phosphorylated in vitro by both Aurora A and Aurora B (Figure  4A). To determine whether Aurora kinases also phosphorylate HP1γ in vivo, we performed western blots of siRNA-treated HeLa cells against Aurora A and B, separately (Figure  4B). We found that Aurora A siRNA can inhibit the P-Ser⁸³-HP1γ in vivo, whereas Aurora B siRNA demonstrated only a slight reduction in levels of P-Ser⁸³-HP1γ (56% of control levels). Of note, Aurora A kinase depletion by siRNA also leads to arrest of cells at G₂/M 20 , thus eliminating the influence of the G₁ phosphorylation in these experiments. To further investigate the participation of Auroras in this event, Chinese hamster ovary (CHO) cells, which have relatively low basal levels of P-Ser⁸³-HP1γ, were transfected with either wild type Aurora A or Aurora B (Figure  4C). As a result, levels of P-Ser⁸³-HP1γ were higher in the Aurora-transfected cells than control. This occurred with both Aurora A and Aurora B transfection, as expected due to their effects on cell cycle progression. In contrast, transfection of epithelial cells, BxPC3, which have high basal levels of P-Ser⁸³-HP1γ, with the dominant negative form of Aurora A 21 22 resulted in reduced levels of P-Ser⁸³-HP1γ (Figure  4D). Similar to the siRNA experiments, dominant negative Aurora B 21 23 had less effect on P-Ser⁸³-HP1γ levels than Aurora A. Therefore, we utilized the dominant negative Aurora A in HeLa cells to confirm this phenomenon by immunofluorescence. Compared to control cells (Figure  4E), transfections with dominant negative Aurora A (Figure  4F) abolished the localization of the P-Ser⁸³-HP1γ in cell compartments rich in this kinase (arrow). In contrast, transfection with dominant negative Aurora A did not affect P-Ser⁸³-HP1γ levels or localization in interphase cells (Figure  4F). Furthermore, we utilized the Aurora A- and Aurora B-specific pharmacological inhibitors, MLN8237 and hesperidin, respectively, to determine the participation of each kinase in P-Ser⁸³-HP1γ. We found that specific inhibition of Aurora A with 300 nM MLN8237, which was confirmed by loss of activated phosphorylation of Aurora A at threonine 288 (P-Thr²⁸⁸) 24 , diminished P-Ser⁸³-HP1γ levels without affecting total HP1γ protein levels (Figure  4G). However, treatment with hesperidin (200 nM) for specific inhibition of Aurora B did not reduce P-Ser⁸³-HP1γ levels, while still inhibiting the Aurora B target P-Ser¹⁰-H3 (Figure  4H, and personal communication with H Dormann and CD Allis). Combined, these results demonstrate that Aurora A kinase is primarily responsible for the localization and increased level of P-Ser⁸³-HP1γ observed in G₂/M. Together with the biochemical experiments described above, these data implicate, for the first time, Aurora kinase in the cell cycle-regulated P-Ser⁸³-HP1γ. This observation also represents the first evidence describing mammalian HP1 at the spindle poles, a localization that is preferred by a large amount of proteins involved in the regulation of cell cycle transitions.

Functionally, HP1γ has been previously shown to play a role in cell cycle progression 10 11 12 13 , although how this protein is regulated to modulate this function remains unknown. Inhibition of Aurora A leads to mitotic spindle defects and misaligned chromosomes 25 26 . Thus, as phosphorylation of HP1γ is downstream of this pathway during mitosis, we investigated whether disrupting the function of this protein also coincides with this effect. For this purpose, we performed stable lentiviral-mediated shRNA knockdown of HP1γ (shHP1γ) in HeLa cells. HP1γ knockdown was confirmed by western blot with approximately 90% reduction in protein levels (Figure  5A). These cells also displayed a significant decrease in P-Ser⁸³-HP1γ staining by immunofluorescence (Figure  5B), demonstrating that localization of P-Ser⁸³-HP1γ to the mitotic spindle pole was unambiguous. We found that 25.5% of shHP1γ cells in mitosis displayed abnormalities (n = 200, Figure  5C), including multipolar spindles, centrosome disruption or lagging, unorganized chromosomes (Figure  5D). shRNA control cells (shCTRL) displayed abnormalities in only 1% (n = 200). However, in spite of this informative outcome, since HP1γ knockdown depleted all forms of the protein, the contribution of Ser⁸³ phosphorylation to this effect could not be assessed by this experimental manipulation. Thus, to better determine the role that phosphorylation of Ser⁸³ plays in this function, we sought to rescue the knockdown phenotype with wild type and Ser⁸³ mutant HP1γ. Transduction with empty vector (EV) control did not change the number of abnormalities observed with shHP1γ. Reintroduction of wild type HP1γ (+WT-HP1γ) rescued, to a significant extent, the abnormal mitotic effects seen with knockdown of this protein (10% abnormal, n = 200). Notably, an alanine substitution, which rendered HP1γ unable to undergo phosphorylation at Ser⁸³ (+S83A), was unable to rescue the knockdown phenotype (23% abnormal, n = 200). This data indicates that first, normal HP1γ levels are necessary for proper mitotic functions and second, HP1γ must be amenable to Aurora A-mediated Ser⁸³ phosphorylation to achieve these effects.

Normal mitotic cell division is a prerequisite for proliferative homeostasis and proper cell cycle progression 27 . Thus, based on our results demonstrating the role of P-Ser⁸³-HP1γ in mitosis, we examined the resultant effects of P-Ser⁸³-HP1γ on cell proliferation by analyzing cells transfected with wild type, S83A-HP1γ or S83D-HP1γ mutant via EdU incorporation. We found that wild type HP1γ had a slight increase in EdU incorporation compared to EV control (103.9% ± 2.6% of EV control, Figure  6A). However, nonphosphorylatable S83A-HP1γ mutant decreased the levels of EdU (94.2% ± 1.6% of EV control, P <0.05, Figure  6A). Notably, an aspartic acid substitution (S83D), designed to mimic Ser⁸³ phosphorylation, had a significant increase in levels of EdU incorporation over control cells (111.2% ± 2.6% of EV control, P <0.05, Figure  6A). Thus, these results support the idea that phosphorylation of Ser⁸³ is necessary for the regulation of cell cycle progression by HP1γ.

We next investigated whether the changes observed in EdU incorporation by both phosphomimetic and nonphosphorylatable Ser⁸³-HP1γ mutants were accompanied by changes in other biochemical surrogates for cell cycle progression, such as known mitotic gene networks. For this purpose, we performed a genome-wide query using Affymetrix (Santa Clara, CA, USA) profiles as transcriptional readouts of their effects. Hierarchical clustering of targets significantly altered by HP1γ (526 targets), S83A-HP1γ (492 targets) or S83D-HP1γ (1,727 targets) overexpression demonstrated that gene networks modulated by HP1γ experienced deregulation in the presence of the Ser⁸³ mutation, indicating dependence of these processes on regulation of Ser⁸³ phosphorylation (Figure  6B). Based on Euclidean distance calculation and the resulting dendrogram, both control and nonphosphorylatable S83A-HP1γ mutant samples were statistically the most similar (Figure  6B). The fact that the EV and the S83A-HP1γ mutant possessed the closest relationship suggested that the latter worked predominantly as either an inactive or dominant negative mutant. However, the phosphomimetic S83D-HP1γ mutant, for the most part, reversed the effect of the S83A-HP1γ mutant, suggesting that it likely worked in a constitutively active manner thereby mimicking Aurora A-mediated Ser⁸³ phosphorylation. Pathway-specific RT-PCR was used to validate a subset of significant targets (Additional file 2: Table S1). These experiments revealed that HP1γ and its phosphorylated form have the ability to change the levels of transcripts related to mitosis.

Gene Ontology (GO) ANOVA analysis was utilized to probe for differentially expressed functional groupings of genes (Figure  6C). Overall, HP1γ overexpression resulted in significant enrichment of targets related to regulation of cellular proliferation, cell division, and mitosis (P <0.05). S83A-HP1γ mutant overexpression yielded differential expression in targets related to protein localization to the chromosome, regulation of the S phase of the mitotic cell cycle and regulation of the G₂/M transition of mitotic cell cycle. S83D-HP1γ mutant overexpression showed significant alteration in genes related to the regulation of the mitotic cell cycle, regulation of the G₂/M anaphase-promoting complex, maintenance of centrosome location and spindle pole structure, among others. Consequently, from these data, we conclude that disruption of phosphorylation status of HP1γ has diverse effects on multiple aspects of the mitotic cell cycle, which is congruent with its cell cycle-associated phosphorylation pattern (Figures  1 and 2) indicating a pervasive role of the regulation of HP1γ in cell division.

Interestingly, previous studies have shown that depletion of HP1γ in primordial germ cells reduces their number as a result of impaired cell cycle progression 13 . Comparison of our expression dataset with a published dataset in primordial germ cells revealed that the expression of the nonphosphorylatable S83A-HP1γ mutant displayed a highly similar pattern as HP1γ depletion, including targets related to cell cycle, proliferation and growth. This ability of the S83A-HP1γ mutation to mimic conditions of absolute HP1γ depletion at the level of gene expression networks, combined with the inability of the S83A-HP1γ mutant to rescue the mitotic defects observed with HP1γ knockdown, indicates that posttranslational modification of this residue is needed for proper progression through mitosis. Furthermore, it may be concluded from our genome-wide analysis that HP1γ participates in the regulation of processes, which support proper cell division and proliferation through phosphorylation-dependent and phosphorylation-independent mechanisms.

Cell lines were obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA) and maintained according to the manufacturer’s protocol. The human LX2 cell line was obtained as a generous gift from Dr Steve Freeman (Mount Sinai, NY, USA). Roscovitine (Sigma-Aldrich, St Louis, MO, USA) treatment was added at increasing concentrations (0, 5, 10 and 20 μM) for 8 hours, and lysates were harvested. Cells were treated with 3 μg/ml aphidicolin or 2 μg/ml nocodazole (both from EMD Millipore, Billerica, MA, USA) for 16 hours to arrest at G₁/S and G₂/M, respectively. Control cells were treated with vehicle, dimethyl sulfoxide (DMSO). HeLa cells were synchronized by double thymidine block. Thymidine (2 mM, Sigma-Aldrich) was added to asynchronous cells for 18 hours. Cells were subsequently released for 9 hours in regular growth media prior to the second thymidine (2 mM) block. After 17 hours, cells were released for the thymidine block and lysates were collected at the indicated time points. KT5720 was obtained from EMD Millipore. MLN8237 and hesperidin were purchased from Selleckchem (Houston, TX, USA). For hesperidin treatment, HeLa cells were arrested in mitosis by treatment with nocodazole for 16 hours. Arrested cells were treated with 200 nM hesperidin for the indicated times in the presence of 10 μM of the proteasome inhibitor MG132 (Sigma-Aldrich) to prevent mitotic exit 28 .

Standard molecular biology techniques were used to clone HP1γ into the pGEX and Ad5CMV vectors. For HP1γ-specific transient shRNA-mediated knockdown, complementary oligonucleotides were synthesized for the target sequence (GCAAATCAAAGAAGAAAAG), annealed and ligated into the pCMS3 vector (kindly provided by Dr Daniel Billadeau, Mayo Clinic, Rochester, MN, USA). For stable shRNA-mediated HP1γ knockdown, control or HP1γ-specific shRNA lentiviral particles (Santa Cruz Biotechnology, Inc, Santa Cruz, CA, USA) were used to infect cells according to the manufacturer’s protocol, followed by puromycin selection (2 μg/ml). Myc-tagged wild type and dominant negative constructs for Aurora A and Aurora B were a kind gift from Dr Paolp Sassone-Corsi 21 . S83A-HP1γ and S83D-HP1γ mutations were obtained using the QuickChange Site-Directed Mutagenesis Kit, as suggested by the manufacturer (Agilent Technologies, Inc, Santa Clara, CA, USA). All constructs were verified by sequencing at the Molecular Biology Core at Mayo Clinic, Rochester, MN, USA. Aurora A (AURKA) and Aurora B (AURKB) Silencer validated siRNAs were purchased from Ambion-Life Technologies (Carlsbad, CA, USA). Epitope-tagged (6xHis-Xpress) HP1γ, S83A-HP1γ and S83D-HP1γ, as well as EV (Ad5CMV), were generated as recombinant adenovirus in collaboration with the Gene Transfer Vector Core at the University of Iowa, IA, USA.

Samples were run on 4 to 20% gradient SDS-PAGE gels (Lonza, Walkersville, MD, USA) or 12% SDS-PAGE gels and electroblotted onto polyvinylidene difluoride (PVDF) membranes (EMD Millipore). The membranes were blocked in 5% BSA in tris-buffered saline Tween-20 (TBST) for 1 hour at room temperature. The blots were incubated for 2 hours at room temperature or overnight at 4°C with primary antibody (P-Ser⁸³-HP1γ 8 , 1:1,000; HP1γ, 1:1,000; and P-Ser¹⁰-H3, 1:5,000 (all from EMD Millipore); Aurora A, 1:1,000 (BD Biosciences Pharmingen, San Diego, CA, USA); Aurora B, 1:500; cyclin B1, 1:1,000; cyclin B2, 1:1,000; and CDK1, 1:1,000 (Abcam, Cambridge, MA, USA); β-actin, 1:1,000; and α-tubulin, 1:1,000 (Sigma-Aldrich); c-Myc (9E10) for Myc-tagged proteins, 1:1,000 (Thermo Scientific, Rockford, IL, USA); and OMNI D8 for His-tagged proteins, 1:1,000 (Santa Cruz Biotechnology)). After repeated washes in TBST, horseradish peroxidase (HRP)-conjugated anti-rabbit or mouse IgG secondary antibody (1:5,000) was added for 1 hour at room temperature. Blots were developed by Pierce ECL chemiluminescent substrate (Thermo Scientific).

Immunofluorescence and confocal microscopy were performed as previously described 42 . The primary antibodies were used at the following dilutions: P-Ser⁸³-HP1γ, 1:200; and γ-tubulin, 1:500 (Sigma-Aldrich); Aurora A, 1:50; and Aurora B, 1:50 (BD Biosciences Pharmingen); cyclin B1, 1:500; cyclin B2, 1:100; and CDK1, 1:40 (Abcam); and cyclin D3, 1:200 (Cell Signaling Technology, Danvers, MA, USA). For localization of P-Ser⁸³-HP1γ during S-phase, EdU incorporation was combined with immunofluorescence. Prior to fixation, cells were incubated for 30 minutes in media containing 10 uM EdU. Subsequently, cells were processed for immunofluorescence, followed by EdU labeling using the Click-iT EdU Imaging Assay Kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. For mitotic aberrations, spindle poles were labeled by immunofluorescence with γ-tubulin and counterstained with 4',6-diamidino-2-phenylindole (DAPI) containing mounting media (Vector Laboratories, Burlingame, CA, USA). For each condition, at least 200 mitotic cells were analyzed to quantify mitotic aberrations.

GST fusion protein purification was done as previously described 8 . For Aurora A and Aurora B in vitro kinase assays, HP1 fusion proteins (10 μg) were incubated with recombinant kinases (EMD Millipore) and 10 mM ATP (Sigma-Aldrich) for 10 minutes at 30°C, in either the supplied buffer (Aurora A) or buffer containing 50 mM Tris pH 7.5, 0.1 mM ethylene glycol tetraacetic acid (EGTA), and 15 mM dithiothreitol (DTT, Aurora B). Kinase reactions were terminated by the addition of SDS loading dye and then resolved by western blot as described above.

Cell proliferation was measured by EdU incorporation using both fluorescence-activated cell sorting (FACS) and microscopy. Cells were infected with adenovirus carrying control, HP1γ, S83A-HP1γ or S83D-HP1γ vectors. Forty-eight hours post-plating, cells were pulsed with 10 μM EdU (Invitrogen) for 1 hour. Subsequently, cells were processed using the Click-iT EdU Flow Cytometry or Imaging Assay Kits (Invitrogen) according to the manufacturer’s protocols. EdU incorporation was measured by FACS analysis at the Mayo Flow Cytometry Research Core Facility, Rochester, MN, USA, or confocal microscopy. Each experiment was performed at least five different times in triplicate, expressed as means with standard error of mean (SEM) and statistical analyses were performed using unpaired t-test.

Global gene expression profiling was carried out at the Microarrays Facility of the Research Center of Laval University, CRCHUL, QC, Canada, utilizing the Affymetrix Human Gene 1.0 ST arrays (28,869 well-annotated genes and 764,885 distinct probes). Intensity files were generated by Affymetrix GCS 3000 7G and the GeneChip Operating Software (Affymetrix, Santa Clara, CA, USA). Data analysis, background subtraction and intensity normalization was performed using robust multi-array analysis (RMA) 43 . Genes that were differentially expressed along with false discovery rate were estimated from t-test (>0.005) and corrected using Bayesian approach 44 45 . Data analysis, hierarchical clustering and ontology were performed with the oneChannelGUI to extend affylmGUI graphical interface capabilities 46 , and Partek Genomics Suite, version 6.5 (Partek Inc, St Louis, MO, USA) with ANOVA analysis. Final fold changes were calculated as x = 2^log2value. Probes with P value <0.05 and fold change ± 2.2 among HP1γ versus EV, S83A-HP1γ versus EV, and S83D-HP1γ versus EV were selected for further analysis. For GO ANOVA, a minimum threshold of three genes and P <0.05 was used to identify significant functional groups. To validate the Affymetrix microarray, targets with significant alteration (P <0.05) were compared to the real-time data using an arbitrary cutoff of ± 2.2 fold change compared to EV control.