En masse analysis of nascent translation using microarrays

Abstract

We report a robust method for studying en masse changes in translation using cDNA arrays. The relative distribution of messenger RNAs (mRNAs) along polysome gradients was monitored by performing cDNA array analysis of each gradient fraction and quantifying the mRNA translational status by regression analysis. Using this strategy to study human carcinoma cells exposed to short-wavelength ultraviolet light (UVC), we identified a subset of 17 translationally induced mRNAs and a subset of 69 translationally repressed mRNAs following UVC irradiation. We describe an effective approach for globally investigating changes in protein biosynthesis.

INTRODUCTION

In mammalian cells, environmental stimuli trigger a series of tightly orchestrated transcriptional and posttranscriptional processes. Among the latter events, the contribution of RNA splicing, messenger RNA (mRNA) transport and stability, as well as protein translation and stability are increasingly recognized. While several high-throughput methodologies allow en masse analysis of steady-state mRNA levels in various physiologic and pathologic paradigms, there is growing appreciation for the discordance between mRNA levels and protein abundance (the steady-state mRNA level of a particular gene may increase while its translation decreases and vice versa) and for the role of translational control in the implementation of gene expression patterns (1,2).

Given the fundamental advantage of elucidating protein expression profiles over mRNA expression patterns, efforts have been made in recent years to study the translational status of collections of transcripts. To this end, a variety of methods have been employed for the separation of actively translated mRNAs, which are recruited into polysomes that have high sedimentation rates, from translationally inactive mRNAs, which exist as free cytoplasmic ribonucleoproteins with significantly lower sedimentation rates. Global analyses of the corresponding mRNA populations using array-based approaches have had critical limitations. According to some methodologies, polysomes of all sizes were separated from the soluble fraction (cytosol) and pooled together through centrifugation (pellet) and were then assessed as two separate fractions in which critical changes in polysome sizes could not be discerned (3). In other cases, the preparation of continuous polysome gradients that were collected as multiple fractions of increasing molecular weight has allowed the analysis of subtle alterations in the abundance of given transcripts, but the comparisons have been challenging to perform and have typically been limited to a small number of fractions (4,5). Here we present a comprehensive approach for assessing the translational status of large collections of transcripts by taking into consideration their abundance throughout the entire gradient. The robust predictive value of this strategy permitted the identification and successful verification of subsets of transcripts whose translation was jointly regulated in response to ultraviolet irradiation (UVC) stress.

MATERIALS AND METHODS

Cell Culture, Treatment, and Small Interfering RNA Transfection

Human RKO colorectal carcinoma cells were cultured in minimum essential medium (MEM; Invitrogen, Carlsbad, CA, USA). For UVC treatment, the medium was removed and saved, and the cells were rinsed with phosphate-buffered saline (PBS). After irradiation (20 J/m²), the medium was added back, and the cells were returned to the incubator for the times indicated in the figures.

Polysome Preparation and Northern and Western Blot Analyses

Samples were collected and fractionated from RKO cells (5 × 10⁶ per sample), and RNA from polysomal fractions was extracted as previously described (6).

For Northern blot analysis, RNA was isolated either from whole cells or from each fraction in the sucrose gradients using standard methodologies. For the detection of c-myc mRNA, a cDNA fragment was labeled using random primers in the presence of [α-³² P]dATP and Klenow enzyme. Complementary oligonucleotides (see the supplementary materials available online at www.BioTechniques.com) were end-labeled using [α-³² P]dATP and used for the detection of EEF1B2, SFN (14-3-3σ), ID1, PLK, STK12, ENO1, CALM2, and CDKN1A (p21).

For Western blot analysis, whole-cell lysates (20-50 μg) were resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene difluoride (PVDF) membranes. Hybridizations were carried out using monoclonal antibodies recognizing Plk (Zymed Laboratories, San Francisco, CA, USA), 14-3-3σ (Research Diagnostics, Flanders, NJ, USA), p21, calmodulin (Upstate Biotechnology, Waltham, MA, USA), β-Actin (Abcam, Cambridge, MA, USA), or using polyclonal antibodies recognizing c-Myc (Cell Signaling Technology, Beverly, MA, USA), Eef1b2 (ProteinTech Group, Chicago, IL, USA), Id1 (C-20), CCN1 (H-279), enolase (C-19), or STK12 (E-15) (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Following incubation with the appropriate secondary antibodies, signals were detected with the ECL™ reagent (Amersham Biosciences, Piscataway, NJ, USA).

cDNA Array Analysis

Whole-cell RNA, RNA from each sucrose fraction, and immunoprecipitated RNA were reverse transcribed in the presence of [α-³³ P]dCTP. Radiolabeled cDNA was used to hybridize cDNA array membranes (MGC arrays, 9600 genes; www.grc.nia.nih.gov/branches/rrb/dna/array.htm). All of the data were analyzed using the ArrayPro® software (Media Cybernetics, San Diego, CA, USA), and normalized by Z score transformation (7). Z scores were used to calculate Z ratios in order to ascertain changes in gene expression. A Z ratio is the average Z score in UVC-treated cells minus the average Z score in control cells, divided by the standard deviation of all of the differences in signal intensity in the array. Z ratios of steady-state RNA signals and RNA in each sucrose gradient fraction were obtained by comparing values in UVC-treated cells (3, 6, 12 h after irradiation) with untreated samples, and were visualized using Cluster and TreeView software, respectively (available online at rana.lbl.gov/EisenSoftware.htm; Eisen Lab, University of California, Berkeley, Berkeley, CA, USA). The criteria used to deem individual mRNAs subject to regulated protein synthesis were: (i) significant Z ratios (≥ 0.5 for induced translation; ≤-0.5 for repressed translation) for at least two polysomal fractions of high molecular weight [fractions (8 and 9) or (9 and 10)] that adequately identified changes in mRNA abundance in heavy gradient fractions (P < 0.01) and (ii) regression slopes resulting from Z ratios calculated from all the fractions > 0.2 (induced translation) or <-0.2 (repressed translation). Best-fit lines calculated from the Z ratios of all of the fractions for a given mRNA (regression lines) provided a measure of the translational trend of a given mRNA: positive for translationally induced, negative for translationally repressed. The equation for the lines is: y = mx + b, where m is the slope and b is the intercept.

Analysis of Newly Translated Protein

Newly translated p21, 14-3-3σ, Eef1b2, c-Myc, and Calm2 proteins were investigated by incubating 10⁶ cells with 1 mCi L-[³⁵S]methionine and L-[³⁵S]cysteine (Easy Tag™ EXPRESS, NEN/PerkinElmer, Boston, MA, USA) per 60-mm plate for 20 min, whereupon the cells were lysed using TSD lysis buffer [50 mM Tris, pH 7.5, 1% sodium dodecyl sulfate (SDS), and 5 mM dithiothreitol (DTT)]. Immunoprecipitation reactions were carried out as previously described (8) for 1 h at 4°C using appropriate antibodies and immunoglobulin G (IgG) as a control. Following extensive washes in TNN buffer [50 mM Tris, pH 7.5 250 mM NaCl, 5 mM EDTA, 0.5% Nonidet™, (NP40)], immunoprecipitation material was resolved by 12% SDS-PAGE, transferred onto PVDF filters, and visualized using a PhosphorImager.

RESULTS AND DISCUSSION

Experimental Assessment of Translational Status

A scheme was devised to monitor shifts in the relative abundance of cytoplasmic transcripts along polysomes of increasing molecular weight (Figure 1A). A size-dependent fractionation of components of the translational machinery on sucrose gradients was followed by the isolation and reverse transcription of RNA in each of 11 fractions and the hybridization of 11 cDNA arrays using the resulting radiolabeled material. Evidence that these fractions comprised mRNAs with different degrees of polysomal association was obtained through the use of translational inhibitors and agents that disrupt polysomes (Supplementary Figure S1). In this study, comparisons were performed between human colorectal carcinoma RKO cells that were irradiated with UVC (20 J/m², and collected 3, 6, or 12 h later) and cells that were left unirradiated (Untr.). Three pair-wise comparisons (3, 6, and 12 h after UVC, each compared with Untr.) were performed by Z score transformation and Z ratio analysis (7). Two criteria for inclusion of a given mRNA in the “Induced” translation group were chosen: first, Z ratios for fractions (8 and 9) or (9 and 10) must be significantly higher when comparing UVC-treated with untreated cells, and second, the slope of the linear regression (best-fit) line resulting from each of 11 Z ratios should have a value of ≥0.2. Conversely, “Repressed” translation was considered for transcripts displaying significantly lower Z ratios for fractions (8 and 9) or (9 and 10) (lower abundance in UVC-treated than in untreated cells) and ≤-0.2 slopes of the linear regression lines. Illustrative examples are shown in Figure 1B.

Figure 1.

Strategy to study changes in translation using cDNA arrays. (A) RNA was isolated from each of 11 cytoplasmic fractions: fractions 1 and 2 lacked any ribosome components (Unb.), fractions 3-5 contained ribosome subunits or single ribosomes (Mono., monosomes), and fractions 6-11 spanned low and high molecular weight (LMW and HMW, respectively) polysomes. The quality of the fractionation was monitored by measuring the absorbance of each fraction (A₂₅₄), and the purity, composition, and integrity of the RNA extracted from each fraction were visualized by ethidium bromide staining. RNA was reverse transcribed to prepare [α-³³ P]dATP-radiolabeled cDNA for hybridization of cDNA arrays. Experiments were performed three independent times, and data were analyzed as previously described (7). (B) The cells were either left untreated (Untr.) or were treated with 20 J/m² of UVC and collected 3, 6, or 12 h later, whereupon fractions were collected. Transcripts were deemed to be subject to translational induction or repression if they met the two criteria detailed in the text. Boxes, fractions (8 and 9) or (9 and 10), which must be significantly higher when comparing UVC-treated with untreated cells. SFN (14-3-3σ) and MYC (c-Myc) mRNAs (12 h after UVC) exemplify translationally induced and repressed genes, respectively. mRNA, messenger RNA; UVC, ultraviolet irradiation.

Regression lines for all of the genes exhibiting either induced (top) or repressed (bottom) translational status are shown (Figure 2A). The same set of transcripts are represented at all time points in each category (Induced, Repressed). For each mRNA, Z ratios at 12 h (Total RNA) are listed next to the regression slopes (R). Steady-state level changes for numerous mRNAs follow the same trend as the R values (Figure 2B). However, in many instances, reduced steady-state levels are seen for mRNAs exhibiting upward R values (FTL, GOLGA2, etc., Induced group) while elevated steady-state levels are seen for some mRNAs exhibiting downward R values (NRAP, RPL7, UCC1, ID1, etc., Repressed group). The findings that steady-state mRNA levels often increase or remain unchanged while the translational status (and consequently the translation) decreases, and vice versa, strengthen the notion that the assessment of total mRNA levels provides an incomplete account of gene expression changes. Instead, ascertaining the degree of translational engagement affords critical information regarding which genes are ultimately expressed into protein. It is important to note that the experimental system described here identifies changes in translational status regardless of whether total mRNA levels increase, decrease, or remain unaltered.

Figure 2.

Changes in relative polysomal distribution of mRNAs after ultraviolet irradiation (UVC). (A) Regression lines corresponding to mRNAs in the induced translation (top) or repressed translation (bottom) categories. (B) List of genes encoding transcripts that were translationally induced (left) and translationally repressed (right) after UVC treatment. R, linear regression slopes; Total RNA, Z ratios (UVC vs. control) for total RNA. Values correspond to 12 h after UVC. Complete array information is available (see Materials and Methods and the supplementary material available online at www.BioTechniques.com). mRNAs, messenger RNAs.

Verification of Translational Regulation

To test the value of this approach in predicting changes in protein biosynthesis, Northern blot analysis was initially used to monitor mRNA abundance in each of the 11 gradient fractions following UVC (Figure 3A); steady-state mRNA abundance in whole-cell preparations (Total) was also assessed. For mRNAs encoding p21 (CDKN1A) and 14-3-3σ (SFN), predicted to undergo enhanced protein synthesis (Induced group), transcript abundance was elevated in the high molecular weight polysome fractions after UVC treatment. For mRNAs deemed to be subject to repressed translation, transcript abundance was lower in high molecular weight fractions and higher in fractions spanning low molecular weight polysomes, monosomes, and free ribosome subunits following UVC, as shown for mRNAs encoding c-Myc (MYC), Eef1b2 (EEF1B2), Calm2 (CALM2), and Eno1 (ENO1).

Figure 3.

Validation of translational regulation by UVC. (A) Northern blot analysis of the changes in abundance of selected mRNAs in whole cells (Total) and on gradient fractions after 20 J/m² ultraviolet irradiation (UVC) of RKO cells. (B) Representative Western blot analysis to assess the abundance of selected proteins following UVC. Whole-cell protein extracts were size-fractionated by SDS-PAGE and signals detected using specific antibodies; β-Actin signals on the same filters served to monitor the even loading of samples. (C) Nascent protein analysis. Six hours after either no treatment (Untr.) or 20 J/m² UVC treatment, RKO cells were incubated in the presence of L-[³⁵S]methionine and L-[³⁵S]cysteine for 20 min and nascent proteins subjected to immunoprecipitation assay. The samples were resolved by 10%-15% SDS-PAGE for visualization of radiolabeled signals. mRNAs, messenger RNAs; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis.

Importantly, as detected by Western blot analysis, for those mRNAs predicted to be engaged in increased protein biosynthesis, protein levels were indeed found to be elevated, while for those mRNAs predicted to be subject to reduced protein synthesis, protein levels were found to be lower (Figure 3B). The preferred method for monitoring protein translation involves the assessment of nascent protein synthesis, whereby a brief (20-min) incubation in the presence of [³⁵S]-labeled amino acids is followed by an immunoprecipitation assay using specific antibodies (6,8). This approach allows for the direct measurement of newly synthesized protein without a significant contribution from additional confounding processes (protein degradation, cleavage, etc.). Given the inefficient signal incorporation due to the short labeling period, the usefulness of this method relies on the availability of highly sensitive and specific antibodies. For the panel of proteins tested (for which several such antibodies were available), the incorporation of radiolabeled amino acids followed the pattern predicted from the analysis of polysome profiles: it was significantly higher for p21 and 14-3-3σ, revealing that their translation was induced by UVC treatment, and was considerably lower for c-Myc, Eef1b2, and Calm2, indicating that their translation was suppressed by UVC (Figure 3A). While other array-based methods have been sought to study alterations in translation, the approaches described here are the first to predict changes in protein biosynthesis and, consequently, in protein steady-state levels (Figure 3, B and C) in a robust, accurate, and demonstrable manner.

The consequences of translationally controlling the levels of these gene products following UVC treatment remain to be studied on a protein-by-protein basis. However, it can be postulated that the increased translation of antiproliferative and anti-apoptotic proteins p21 and 14-3-3σ, together with the decreased translation of proliferative proteins c-Myc and cyclin I (ClnI), and the pro-mitotic proteins Plk and Stk12, etc., likely acting in concert with changes in their degradation rates, facilitate the immediate onset of a growth-inhibited state while the damage incurred by UVC is assessed. In this paradigm, subsequent changes in mRNA levels, brought upon by regulated transcription and/or mRNA turnover, probably contribute to the mid- to long-term regulation of gene expression during the stress response as cells commit to a path of death or repair macromolecules and attempt to survive the injury.

We propose that the array-based methodologies described in this investigation will provide powerful tools for studying global alterations in protein biosynthesis as we seek a more complete understanding of the influence of posttranscriptional events in the implementation of gene expression programs.