EGF-Induced Acetylation of Heterogeneous Nuclear Ribonucleoproteins Is Dependent on KRAS Mutational Status in Colorectal Cancer Cells

KRAS mutational status is considered a negative predictive marker of the response to anti-EGFR therapies in colorectal cancer (CRC) patients. However, conflicting data exist regarding the variable response to EGFR-targeted therapy. The effects of oncogenic KRAS on downstream targets were studied in cell lines with different KRAS mutations. Cells harboring a single KRASG13D allele showed the most tumorigenic profile, with constitutive activation of the downstream pathway, rendering them EGF-unresponsive. Conversely, KRASA146T cells showed a full EGF-response in terms of signal transduction pathways, cell proliferation, migration or adhesion. Moreover, the global acetylome of CRC cells was also dependent on KRAS mutational status. Several hnRNP family members were identified within the 36 acetylated-proteins. Acetylation status is known to be involved in the modulation of EGF-response. In agreement with results presented herein, hnRNPA1 and L acetylation was induced in response to EGF in KRASA146T cells, whereas acetyl-hnRNPA1 and L levels remained unchanged after growth factor treatment in KRASG13D unresponsive cells. Our results showed that hnRNPs induced-acetylation is dependent on KRAS mutational status. Nevertheless hnRNPs acetylation might also be the point where different oncogenic pathways converge.


Introduction
Colorectal cancer (CRC) is one of the most prevalent tumors worldwide [1] and despite many advances in therapy, long-term survival for patients with metastatic disease is still poor [2]. Antibodies against the Epidermal Growth Factor Receptor (EGFR) have been successfully used

Cell culture
Colon cancer cell lines HCT116 and their derivatives HAE6 and HAF1 were commercially acquired from the GRCF Biorepository and Cell Center, at Johns Hopkins University. All cell lines were cultured in McCoy's 5A Modified Medium (Gibco) supplemented with 10% FBS, penicillin/streptomycin and L-glutamine and maintained under standard conditions.

Protein extraction and Immunoblot analysis
Total proteins were extracted in RIPA buffer supplemented with inhibitors and sodium butyrate. Equal amounts of proteins were electrophoresed in SDS-PAGE gels and transferred onto nitrocellulose membranes and immunoblotted as described elsewhere [18]. Blots were developed by enhanced chemoluminiscence (ECL Detection Kit, GE Healthcare). Protein levels were quantified by densitometry and normalized by β-actin or GAPDH expression.

2D-electrophoresis and acetylome analysis
Protein extracts (100μg) were separated by two-dimensional (2D) electrophoresis, as previously described [18]. Proteomic images were acquired with a Typhoon Trio Imager. In parallel, 2D-gels were electroblotted on nitrocellulose membranes and analyzed by western blot with anti-acetylated lysine antibody. Acetylated proteins overlapping with the SYPRO Ruby-stained gels were excised for MS analysis. Gel specimens were processed with a MassPrep station (Waters). NanoLC-ESI-MS/MS and data analysis were performed as described [19] by the Proteomics and Bioinformatics Unit, (CIMA-University of Navarra, Spain).

Immunoprecipitation
500μg of proteins were immunoprecipitated overnight with specific antibodies or normal serum IgG (Dako) at 4°C on a rotating device. Protein-antibody complexes were pulled-down with 50% (v/w) Dynabeads (Invitrogen, Life Technologies). After several washes in PBS, immunocomplexes were recovered by boiling in LB and then analyzed by Western blot as described. 1% of protein extracts used for immunoprecipitation was loaded as Input.

RNA isolation and expression analysis by qPCR
RNA extraction, complementary DNA synthesis and qRT-PCR were carried out as preciously described. [18]. Primer sequences for hnRNPA1, hnRNPA3, hnRNPA2/B1, hnRNPL and GAPDH are shown in supplementary data. Pre-developed Taqman primers for KRAS, ADAM19, CATHEPSIN C, CATHEPSIN L and 18S were from Applied Biosystems. Relative gene expression was expressed as 2 -Δ(ΔCt) Data were normalized by 18S quantification in the same sample reaction. All reactions were carried out in triplicate.

Microarray gene expression. Hybridization and data analysis
Total RNA (5μg) was used to get biotinylated cRNA according to the manufacturer's protocol (Affymetrix). Quality of the labelled cRNA was confirmed by hybridization to an Affymetrix test chip (Test3-chip). Labelled cRNA was hybridized to an Affymetrix Genechip Human Gene 1.0ST. Microarray data were obtained by AGCC Affymetrix software. Raw data of three replicates for each cell line were analyzed by R/Bioconductor using the Limma package [20]. F-statistics, t-statistics and log odds of differential expression were calculated by empirical Bayes moderation of the standard errors towards a common value. Differentially expressed genes were identified by pairwise comparison, (p-value cutoff = 0.05). Only genes with a fold change above 1.5 and below -1, were considered differentially expressed for further analyses and classified according to the biological process category following the criteria of the Gene Ontology Consortium [21].

BrdU Cell proliferation assay
Cell proliferation was examined by a bromodeoxyuridine (BrdU) incorporation assay kit (Calbiochem). Briefly, the cells were seeded at a density of 5x103 cells/ well in a 96-well culture plate. After 24h starvation, cells were treated with EGF (20 ng/ml) in serum-free medium. The BrdU reagent was added to the wells for 4h in the presence or absence of EGF. At the indicated time points, cells were fixed during 30min in a fixative/denaturing solution. Cells were incubated with anti-BrdU antibody (1:100) for 1h and after several washes, incubated with HRPconjugated IgG for 30 min. Tetra-methyl benzidine substrate was then added for 15 min and the reaction was stopped by the addition of 1.25M sulfuric acid stop solution. Absorbance was measured in each well at dual wavelengths 450-540nm.

Wound-healing assay
The three cell lines were cultured under standard conditions. Confluent monolayers were scratched with a 10μl pipette tip to induce a wound. The wounded edges were imaged using an inverted microscope Nikon Eclipse Ti (10X Magnification). Images were collected at 0, 24 and 48 h after scratch.

Statistical analysis
Data are the mean ± SEM of at least three independent experiments. Significant differences were determined by one-sample Student's t test. Differences were considered significant at p<0.05. Representative blots are shown.

3.1.
Differential effect of KRAS A146T and KRAS G13D in EGF-treated CRC cells HCT116 CRC cell line, harboring an endogenous activating KRAS G13D/wt mutation, and two isogenic HAF1 (KRAS wt/-) and HAE6 (KRAS G13D/-) clones were used for this study. All those genes recommended to predict a clinical outcome for cancer therapy were sequenced in the three cell lines [22] (S1 Table). Interestingly, while all mutations found in HCT116 were also observed in HAE6 cells, we found that in HAF1 cells a KRAS A146T hotspot mutation had spontaneously arisen in what so far was intended to be a wild-type allele. Given this result, we characterized KRAS A146T cells and compared them with HCT116 or HAE6 cell lines (S1 Fig and S1 File). HAE6 cells showed the highest proliferation rate and the lowest levels of cell adhesion. A wound healing analysis of cell migration revealed differences among the three cell lines (, with HCT116 showing a higher rate than HAF1 but lower than HAE6 cells. In contrast to the previous results, KRAS mutational status had no apparent effect on ERK and AKT phosphorylation, under basal conditions. This is not surprising, since the three cell lines harbor a PIK3CA activating mutation. KRAS A146T seemed to be the less tumorigenic mutation. Nevertheless, unsynchronized cell culture under basal conditions could mask the effect of KRAS mutation in response to a specific stimulus. Therefore, to further explore the downstream effect of KRAS A146T versus the well-characterized KRAS G13D mutation, HAF1 and HAE6 cells were grown in serum-deprived media and then stimulated with EGF. To rule out the possibility that the mutation could affect KRAS expression, qPCR was performed and KRAS mRNA levels found in both cell lines were similar (data not shown). Although EGF-treatment induced the rapid phosphorylation of ERK1/2 and AKT in HAF1, this response was poor in HAE6 cells ( Fig 1A). Noticeable, pERK levels were already high in untreated HAE6 controls. In addition, as already described in other cell lines with an hyperactivated ERK pathway [23], EGF decreased pAKT levels in this HAE6 cell line.
Cell adhesion experiments could not be performed in serum-free media because HAF1 cells were sensitive to trypsinization after 24h-serum deprivation. A higher talin-cleavage has been associated with lower cell adhesion [24]. Therefore, we analyzed cleaved-talin as an indirect method to measure cell adhesion. In unstimulated cells, talin-cleavage was already higher in HAE6 than in HAF1 cells. Moreover, while talin-cleavage was induced in response to EGF in HAF1 cells, this proteolytic cleavage was EGF-independent in HAE6 cells ( Fig 1B). We next analyzed cell migration and found that HAE6 cells were also unresponsive to EGF, in contrast to the cell migration induced by EGF in HAF1 cell line ( Fig 1C). Finally, cell proliferation, although induced by EGF in both cell lines, was increased to a lesser extent in HAE6 than in HAF1 cells (Fig 1D). We reasoned that hyperactivation of ERK1/2 induced by KRAS G13D could reach a plateau beyond which cells cannot be further stimulated by EGF.

Differential mRNA levels in KRAS G13D vs KRAS A146T CRC cell lines
We examined those possible KRAS-dependent transcriptional effects in both cell lines using Affymetrix DNA microarrays (S2 and S3 Files). Under the presence of KRAS G13D mutation, most differentially expressed genes were down-regulated (Fig 2A and 2B; S2 Fig) when compared to KRAS A146T mutation (HAF1). However, the expression of the majority of genes did not substantially change, and only few of them were up-regulated. In order to validate these data, the expression of a selection of up-regulated genes was analyzed by qPCR. These genes were selected for their prominent role in tumor metastasis, growth factor shedding or CRC drug-resistance [25,26]. The expression of these genes was dramatically up-regulated in cells with a mutated KRAS G13D allele, thus confirming the microarray data ( Fig 2C and S2 Fig).
Among the genes down-regulated in HAE6 vs. HAF1 cells, the most representative pathway affected was EGFR1 signaling; 43 genes were down-regulated out of the 156 described for the EGFR1 canonical pathway. However, other pathways triggered by cytokines and growth factors, such as VEGF, PDGF, HGF, IGF1,TGFβ, TNFα and several interleukins, were also downregulated. This is not surprising since most of them share several steps within the RAS signaling pathway.

Effect of KRAS mutations on the global acetylome of CRC
One of the potential pathways affected by KRAS mutational status which may regulate several biological processes is protein acetylation. We sought to study whether the global acetylation profile in CRC was differentially affected by activated KRAS G13D and activated KRAS A146T . Isolated proteins from both cell lines grown in 10%FBS, were immunoprecipitated and analyzed by western blot with anti-acetyl-Lys antibodies (Fig 3A). A 2D western blot proteomic approach was used to deeply analyze acetylated-proteins from the two cell lines. Global protein Lys-acetylation was increased in HAE6 compared to HAF1 cells, according to increased number of detected spots (Fig 3B).
Positive spots from the HAE6 acetylome were excised and identified by mass spectrometry. 67 spots were detected but only those with a confident Mascot ion score and a high peptide matching were selected (S2 Table). The (GO) biological function analysis of our 36 acetylated proteins revealed that target proteins were enriched in four main categories related to cell growth, energy metabolism, RNA metabolic process and protein metabolism (Fig 3C). Many   of these proteins were already described to be targets of putative acetyltransferases or deacetylases and to become acetylated under different experimental conditions (reviewed in S2 Table). Our microarray results confirmed that differences in the acetylation pattern were not due to increased mRNA expression. Indeed, those genes whose products were identified in our acetylome analysis remained unchanged in HAE6 when compared to HAF1 cells (S2 and S3 Files).

Acetylation of hnRNP family members
Among the acetylated-targets there was an interesting group of RNA binding proteins, all of them members of the family of heterogeneous nuclear ribonucleoproteins (hnRNPs), involved in a variety of molecular functions going from mRNA processing, transport, stability or even translation [27]. We analyzed in the two cell lines whether the basal acetylation of hnRNPs was dependent on KRAS mutational status. Acetylation was measured by immunoprecipitation experiments with acetyl(Lys) antibodies followed by western blot with antibodies recognizing the specific hnRNP member. As shown in Fig 4A, hnRNPA1 and A2/B1 were found to be hyperacetylated under growing conditions (10% FBS) in HAE6 when compared to HAF1; however, this pattern was not so clear in other family members tested such as hnRNPA3 or L. The mRNA and protein levels of hnRNPs were analyzed by qPCR and western blot respectively (Fig 4B and 4C); the expression of none of the hnRNPs showed a significant change when compared among the different cell lines. Therefore, the different levels of acetylation observed are not the result of a higher rate of gene expression or protein stabilization.

Acetylation of hnRNPA1 and L in response to EGF
Although acetylation of hnRNPs has been already described in other cell systems, at present it is not known whether this acetylation is part of the EGF response in CRC cells. Should it be the case, it will be important to establish whether the KRAS mutational status could condition a differentially induced-acetylation of hnRNPs upon EGF stimuli. Both cell lines were grown in serum-free medium and then stimulated with EGF for a short (20min.) and a long (2h) period of time (Fig 5A). Samples from each condition were immunoprecipitated with either anti-hnRNPA1 or anti-hnRNPL antibodies. These hnRNPs were selected since they represent those isoform showing the highest and the lowest acetylation differences when comparing HAF1 and HAE6 cells. To establish the acetylation ratio (acetyl-hnRNP/total hnRNP), pull-down proteins were further analyzed by western blot with specific antibodies against acetyl-Lys or the corresponding hnRNP.
Under serum deprivation, basal acetylation levels of both hnRNPs were higher in HAE6 than in HAF1 cells (Fig 5A, lane 0). However, the acetylation of both, hnRNPA1 and hnRNPL at the end point of EGF-treatment was the same in the two cell lines. Interestingly, when the acetylation of EGF-treated samples was compared to untreated controls, acetylated-hnRNP levels did not significantly change throughout the time course of EGF-treatment in HAE6 cells (Fig 5A). Conversely, EGF induced a dramatic acetylation of both hnRNPs in HAF1 cells. Interestingly the same response was observed after EGF treatment in HCT116 parental cell line (S3 Fig). Finally, mRNA and protein levels from hnRNPA1 or-L were not induced in any EGFtreated cell line at the tested time points (Fig 5B and 5C; S3 Fig). Consequently, EGF-induced acetylation observed in HAF1 cannot be attributed to the transcriptional or translational modulation of hnRNPs.

Role of KRAS A146T mutation on hnRNP acetylation
To study whether KRAS A146T mutation may account for EGF-induced acetylation of hnRNPs, we specifically inhibited the KRAS signaling pathway in the HAF1 cell line. Independently of KRAS A146T mutation, HAF1 cells also encompass a PI3KCA mutation. Therefore we analyzed the acetylation of hnRNPs induced by EGF in cells pretreated with inhibitors of both, MEK1/2 (U0126) and PI3K (LY294002). Inhibition of one single pathway either MEK1/2 or PI3K, by using the specific inhibitor alone, had no effect on protein acetylation (Fig 6A). The ability of RAS to influence PI3K activity and vice versa has been well documented [28,29]. Indeed, blocking MEK1/2 activity has been shown to induce AKT phosphorylation in CRC cell lines    [30]. On the other hand, blocking PI3K activity has been shown to increase pERK1/2 levels in response to a variety of stimulus [28]. In agreement with this, when both inhibitors were used simultaneously, the inhibition of RAS/PI3K signaling pathways completely blocked the EGFinduced acetylation of both hnRNPA1 and hnRNPL (Fig 6B). Interestingly, basal levels of acetylated hnRNPs in serum-starved controls were not affected by the kinases inhibitors.
Basal levels of acetyl-hnRNPs could be dependent on the half-life of acetylations. Actually, acetylation of proteins is not only the result of acetyltransferase-induced activity, but also of deacetylases. To examine the role of deacetylases on basal acetylation, we investigated whether the deacetylase inhibitor trichostatin A (TSA) could mimic the effect of EGF in HAF1 cells. Indeed, TSA increased the acetylation of both hnRNPs to the same levels as those reached by the EGF-treatment, although no synergistic effect was observed (Fig 6C).

Discussion
Herein we uncover several aspects of the molecular events underlying colon carcinogenesis that will be further discussed. We provide the first report indicating that the acetylation of 36 specific proteins is dependent on KRAS mutational status; among them, several members of the hnRNP family. Moreover, our data indicate that hnRNP acetylation is part of the EGFinducible response. Secondly, we show the molecular consequences of allele imbalance by deletion of either, a wild-type or a mutant allele in KRAS G13D/WT CRC cells. Deletion of the wild type allele (HAE6) leads, not only to hnRNP acetylation, but also to the more tumorigenic and EGF-unresponsive profile of these cells. Conversely, deletion of the mutant allele (HAF1) induces a spontaneous KRAS A146T mutation in the wild-type allele.
Collectively our observations seem to parallel the clinical findings indicating that disruption of the KRAS wild-type allele is an adverse prognostic factor for CRC [31]. Surprisingly, in HAF1 cells, a spontaneous KRAS A146T point mutation was found. In agreement with this, the most recent reports show dynamic KRAS genotypic changes throughout the progression of metastatic CRC [11]. This point mutation also renders a constitutive activation of KRAS [32], In the near future, it would be important to analyze possible spontaneous mutations in any other CRC cell line used as experimental model to study the molecular mechanisms of therapy-resistance or new drug development.
Mutation screening tests based on the hotspot KRAS codons 12 and 13 have resulted in the mis-classification of up to one-third of patients that were erroneously considered as KRAS wild-type and were resistant to anti-EGFR treatments [33,34]. KRAS mutations in codons 61 and 146 are significantly associated with shorter progression-free survival compared with wildtype KRAS in CRC patients treated with a combination of cetuximab and chemotherapy [10]. Although many questions remain to be answered to uncover the impact of KRAS A146T mutation in CRC drug resistance, to our knowledge this is the first description of the effect of this recently discovered mutation on downstream signaling pathways, proliferation, migration and cell adhesion in EGF-stimulated CRC cells. Thus, our results further reinforce the importance of extending the mutation screening tests to the less frequent hotspot mutations such as codon 146 or 61 [5,22], which could determine the EGF-response going from signal transduction to the global acetylome of tumor cells.
Indeed, the acetylation/deacetylation process is involved in the modulation of the EGFresponse in experimental and clinical CRC studies [15]. However, how the mutational status of KRAS can affect the acetylome of CRC is so far not known. In agreement with our observations about hnRNPs acetylation, one of the largest acetylome studies showed that the main functional networks modulated by acetylation were related to RNA maturation [35]. All in all, although the prominent role of hnRNPs in tumor progression has been already described [36][37][38], this is the first report showing a relation between KRAS mutational status and hnRNP acetylation. The highest level of acetyl-hnRNPs, and also the largest number of acetylated proteins, was found in cells harboring one single allele of KRAS G13D . Accordingly, our molecular and biological characterization of KRAS G13D cell line confirmed the more oncogenic behavior of these cells (HAE6).
Indeed, under basal conditions, while proliferation and migration rates were dramatically higher, cell adhesion was lower in HAE6 (KRAS G13D/-) than in HAF1 cells (KRAS A146T/-). It has been proposed that oncogenic KRAS might promote tumor progression by limiting the efficacy of RAS/RAF/MEK/ERK1/2 signaling, whereas KRAS-responsive tissues exhibit a full activated signaling pathway, and trigger potent antitumor responses [31]. In agreement with this, HAE6 cells showed an EGF-unresponsive MAPK signaling pathway. Consistently, the acetylation of hnRNPL and A1 remained unchanged after EGF-treatment. Furthermore, EGF could not further induce in these cells, any of the biological responses described for this growth factor, i.e. cell proliferation, adhesion or migration.
The functional consequences of hnRNP acetylation are still unknown; however, in agreement with a KRAS-dependent modulation of hnRNP acetylation, several hnRNP family members are modulated by the MAPK pathway [39,40]. This phosphorylation may condition hnRNP subcellular distribution and function [28]. The possible interplay between acetylation and phosphorylation of specific hnRNPs remains elusive. In the near future, it would be important to study whether this putative interplay could determine mRNA splicing, stability, transport or translation of specific genes involved in tumorigenesis or resistance to antitumor therapies. On the other hand, since CRC cells also harbor KRAS-independent mutations, we cannot rule out the suggestive idea of hnRNPs acetylation as the point where different signaling pathways converge. The mutation-induced activation of one or more of these pathways might result in hnRNPs hyperacetylation. Should this hypothesis be confirmed, the relevance and extension of our data will be increased from both, the molecular and clinical point of view.
Supporting Information S1 Fig. Proliferation, adhesion and migration of CRC isogenic cells in basal conditions. A. HAE6, harboring a single KRAS G13D mutated allele, showed increased proliferation compared to HAF1 and HCT116 isogenic cell lines, as determined by MTT cell proliferation assays. A viability assay was conducted daily from 1 to 4 days. The results are expressed as fold versus day 1 for each cell line. B. Effect of KRAS G13D mutation on cell adhesion in fibronectin-coated plates (2,5 μg/cm 2 ). The results are expressed as fold compared to HAF1 cell line. C. Woundhealing assay was performed to analyze cell migration. The micrographs show the wound (left panels), cell migration at 24h (middle panels) and at 48h (right panels) for the three cell lines. Tumor cells, which migrated to the wound area, were counted at 48h. The results are expressed as fold compared to HAF1 cell line. D. Phosphorylated and total ERK1/2 protein levels, as well as AKT and pAKT levels were analyzed by western blot in the three cell lines cultured in basal conditions with 10%FBS. Image shown is representative of three independent experiments. HCT116 cells were grown in serum-free medium and then stimulated with EGF (20ng/mL) for 20 and 120min. A. Protein extracts from control or EGF-treated cells were isolated and immunoprecipitated with α-hnRNPA1 or hnRNPL antibodies. The immunoprecipitated samples were then analyzed by western blot with antibodies recognizing acetyl-lysine residues and, either hnRNPA1 or hnRNPL. Inputs of each specific hnRNP in the different cell lines are shown. B. mRNA levels of hnRNPA1 and hnRNPL were analyzed by qPCR in control and 2h EGFtreated cells. No statistical significance was found (n = 3). C. Western blot analysis showing protein levels of both hnRNPs in control and 2h EGF-treatment conditions. The intensity of hnRNPs bands was measured and normalized by β-actin; graph shows the quantification for both hnRNPL and A1 in HCT116 cell line. No statistical significance was found (n = 3).