|Home | About | Journals | Submit | Contact Us | Français|
Translation initiation of eukaryotic mRNAs typically occurs by cap-dependent ribosome scanning mechanism. However, certain mRNAs are translated by ribosome assembly at internal ribosome entry sites (IRES). Whether IRES-mediated translation occurs in stressed primary human endothelial cells (EC) is unknown.
We performed microarray analysis of polyribosomal mRNA from EC to identify IRES-containing mRNAs. Cap-dependent translation was disabled by poliovirus (PV) infection and confirmed by loss of polysome peaks, detection of eIF4G cleavage, and decreased protein synthesis. 87.4% of mRNAs were dissociated from polysomes in virus-infected EC. 12% of mRNAs remained associated with polysomes and 0.6% were enriched ≥2-fold in polysome fractions from infected EC. Quantitative RT-PCR confirmed the microarray findings for 31 selected mRNAs. We found that enriched polysome associations of PDCD8 and JunB mRNA resulted in increased protein expression in PV-infected EC. The presence of IRES in the 5’UTR of PDCD8 mRNA, but not of JunB mRNA, was confirmed by dicistronic analysis.
We show that microarray profiling of polyribosomal mRNA transcripts from PV-infected EC successfully identifies mRNAs whose translation is preserved in the face of stress-induced, near complete cessation of cap-dependent initiation. Nevertheless, internal ribosome entry is not the only mechanism responsible for this privileged translation.
Cap-dependent initiation, also known as the scanning model, is the mechanism by which most eukaryotic mRNAs are translated. All nuclear encoded eukaryotic mRNA is processed by capping with 7-methyl guanosine (m7G) at the 5’-terminus prior to transport out of the nucleus and into the cytoplasm for translation. In the cytoplasm, the m7G cap structure is bound by eukaryotic initiation factor 4F (eIF4F), a complex of three polypeptides including eIF4E, A, and G. The m7G cap structure is recognized by eIF4E which serves to anchor the eIF4F complex while the secondary structure of the mRNA is unwound by the helicase activity of eIF4A. eIF4G functions as a scaffolding protein that bridges the mRNA-bound eIF4F to the 40S ribosomal subunit through eIF3. This pre-initiation complex then scans the 5’ untranslated region (UTR) until an AUG start codon is recognized 3.
An alternative, and less common, form of translation initiation is a cap-independent mechanism that requires an internal ribosome entry site (IRES), a specialized internal RNA structure in the 5’UTR. During cap-independent translation, 40S ribosomal subunits assemble directly on the IRES near the start codon, thereby bypassing the requirement for the mRNA 5’cap structure and eIF4F to initiate translation 4.
First identified in the 5’UTRs of mRNAs of encephalomyocarditis virus and poliovirus (PV) 5, 6, IRES allow the efficient translation of un-capped viral messages into proteins while cap-dependent translation initiation is inhibited in the host cell. Since then, a number of cellular mRNAs have been recognized to contain an IRES 4. These IRES-containing mRNAs encode proteins involved in multiple biological processes such as mitosis, differentiation, apoptosis, hypoxia, heat shock and oxidant injury 7. IRES maintain or even induce the synthesis of specific proteins while cap-dependent translation is severely impaired in cells under stress conditions. Dysregulation of IRES-mediated translation has been linked to the development of several human diseases, including diabetes 8, cardiovascular diseases 9, multiple myeloma 10, Charcot-Marie-Tooth disease 11, and congenital X-linked dyskeratosis 12. Therefore, delineation of conditions in which IRES influence patterns of protein expression has substantial biologic and pathologic significance.
Endothelial cells (EC), which line the extensive human vasculature, respond to a multitude of signals from their environment and directly influence complex homeostatic mechanisms that regulate vessel injury, vascular tone, inflammation and coagulation. EC are capable of both rapid and prolonged responses to extracellular stimuli, resulting in alterations in cellular function and phenotype. Pathologic conditions such as sepsis, ischemia-reperfusion injury and vascular thrombosis may arise in part because of dysregulated endothelial responses to these stimuli. Most studies in EC have focused on transcriptional mechanisms that regulate gene expression, which may require hours for a response to a stimulus because new mRNA production is a requisite event. We and others have shown that EC can also regulate gene expression by translational control as a means of responding to stimuli in relatively short time periods and with diversity and precision that is not afforded by transcriptional regulation alone 2, 13–18. Whether EC utilize IRES as a translational control mechanism when cap-dependent translation is inhibited by cellular stress is unknown. We speculated that human EC are capable of IRES-mediated protein synthesis and that this mechanism might be used when EC are activated under pathophysiologic conditions or subjected to pathologic stress.
To approach this question, we established PV infection of cultured human EC as a model for internal ribosome entry when cap-dependent translation is inhibited in stressed vascular endothelium. We then performed polyribosomal (polysome) profiling and high-throughput microarray analysis 19 and identified several IRES-containing mRNAs in EC infected with PV. PDCD8, JunB, and angiomotin-like 2 (AMOTL2) emerged as intriguing candidates for IRES-mediated translation as the presence of an IRES has not been described in the mRNAs encoded by these genes in any cell type. We show that EC utilize cap-independent translation as an important alternative mechanism to selectively synthesize gene products that may be necessary for the cell’s response to stress when cap-dependent translation is inhibited.
Additional assays including immunofluorescence, immunoblot analysis, polysome profiling, [35S]-methionine incorporation experiment, quantitative real-time RT-PCR (qRT-PCR), construction of dicistronic vectors, transient transfections and luciferase reporter assays, and northern analysis were performed according to standard techniques. An expanded Methods section is available in the Online Data Supplement.
Primary human umbilical vein EC (HUVEC) were isolated and cultured as described 20. EaHy 926 cells have been propagated in our laboratory since the original gift from Edgell 21. PV 1 (live-attenuated vaccine strain) was purchased from ATCC. The preparation of PV1 and its infection of EC are described in the online supplement material.
RNA from polysome fractions was isolated from mock or PV-infected EC using Trizol LS (Invitrogen, Carlsbad, CA), labeled with Cy3 (mock-infected) or Cy5 (PV-infected) and then subjected to whole human genome-wide microarray analysis (43,203 transcripts, Agilent, Santa Clara, CA) to obtain translational profiles. The procedures were performed in the microarray core facility in University of Utah as described in the online data supplement.
Each Cy3 or Cy5 signal was normalized by total Cy3 or Cy5 signal on the slide respectively. The ratios of normalized Cy5 to Cy3 were then calculated. We arbitrarily used PV-infected/mock-infected ratio thresholds of ≥ 2.0 as indicating translational upregulation, 1.0 – 2.0 as preserved translation and ≤ 1.0 as disrupted translation in response to PV infection.
PV disrupts host cap-dependent translation while employing IRES to translate its own viral message 6. Host messages that continue to be translated despite PV infection are also likely to do so via an IRES-dependent mechanism 19. In order to determine whether PV infection can be used to identify IRES candidates in human EC, we first investigated the potential for PV to infect primary EC in culture. Using immunocytochemistry, we demonstrated the presence of PV antigen in the cytoplasm of EC as early as two hours post infection (Fig1). Between four to six hours after inoculation, PV protein dramatically increased in the cytoplasm of most cells indicating active and efficient translation of PV mRNA into protein during this time period. By 8 hours post-infection, EC lysis was apparent suggesting that the PV had completed its first life cycle 22.
PV inhibits translation of host cellular mRNAs by proteolytic cleavage of eIF4G, a key component of cap-dependent translation initiation machinery 19. There are two isoforms of eIF4G in eukaryotic cells, I and II, both of which need to be depleted to completely inhibit host cap-dependent translation 23, 24. In order to determine whether eIF4G cleavage results from PV infection in human EC, Western analysis for eIF4G isoforms was performed. We found cleavage of both isoforms beginning two hours after infection. After four hours of PV infection, eIF4GI was completely cleaved along with the majority of eIF4GII (Fig 2A). Thus, a four hour time point was chosen to use in later microarray experiments in which we examined candidate IRES-containing messages in virally-infected EC. Interestingly, complete cleavage of eIF4GII was never observed in EC infected with PV even six hours post-inoculation.
Further evidence for inhibition of host translation by PV was provided by polyribosomal profiling of PV-infected EC. In mock-infected EC, multiple polysome peaks were observed, consistent with efficient translation of a multitude of host EC proteins 1, 2 (Fig 2B, left). Four hours after PV infection, the polysome peaks were dramatically reduced in parallel with a corresponding increase in the abundance of the 40S, 60S and 80S fractions (Fig 2B, right). These changes in the polyribosomal profiling pattern of PV-infected EC are consistent with major inhibition of cap-dependent translation at the initiation stage in the host cell 19, 25, 26.
To further assess whether the loss of polysome peaks is an indicator of global inhibition of protein production in PV-infected EC, labeling of new protein products using [35S]-methionine incorporation was performed. As expected, there was a dramatic inhibition of protein synthesis in EC four hours post infection (Fig 2C). Nevertheless, several protein bands were readily detected, providing evidence for synthesis of viral proteins and, potentially, a subset of host proteins via alternative mechanisms when cap-dependent translation is disabled by the infecting virus (Fig 2C).
The most likely mechanism of preserved synthesis of host proteins in PV-infected cells is IRES-mediated translation 19. To examine this issue, candidate genes were first identified in EC by comparative genome-wide microarray profiling of mRNAs that remained associated with polysomes (sucrose gradient fractions 6–10, with three or more associated ribosomes per mRNA) following PV infection (Fig 2B and Fig 3A). For each array element, the amount of polysome-associated mRNA derived from PV-infected EC was compared with the amount from mock-infected EC to obtain polysome ratios (PR). PR values greater than 2.0 were taken to indicate increased association with polysomes in PV-infected cells and the mRNA was considered a strong candidate for the presence of an IRES (Fig 3A).
As expected in the polysome arrays from PV-infected EC, the majority (37,756 of 43,203, 87.4%) of host mRNAs exhibited a PR less than 1 (Fig 3B), indicating reduced or absent association with polysomes when cap-dependant translation was inhibited. An example is the house-keeping gene, β-actin (ACTB), with a PR of 0.69 +/− 0.08. In contrast to the large number of genes that had reduced association with polysomes, 12% of mRNAs remained associated with polysomes (1<PR<2) from EC infected by PV. Interestingly, 277 out of 43,203 (0.64%) mRNAs showed 2-fold or more enrichment in the polysome fractions from PV-infected EC compared with mock-infected cells (Fig 3B). These 277 messages represented 241 unique genes and encoded proteins involved in multiple biological functions, including: angiogenesis, cell signaling, growth and apoptosis, oncogenesis and inflammation (Supplement Table 1). Some of these messages have previously been reported to contain IRES, including Cyr61 and Pim1 19. (The microarray data have been deposited in NCBI’s Gene Expression Omnibus (GEO. http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE15356.)
To verify the microarray data, 31 genes with a wide range of PR values derived from microarray analysis were chosen and PCR primers were designed accordingly (Supplement Table 2). A PR value for each sequence was generated by utilizing qRT-PCR to compare the relative abundance of mRNA isolated from the polysome fractions of PV-infected EC to mock-infected EC. Supplement Table 3 shows PR values generated from both the array and qRT-PCR for these 31 genes. Overall, PR generated from qRT-PCR were highly similar to those obtained from microarray experiments with a correlation coefficient of 0.92 (Fig 4). Specifically, the house-keeping gene β-actin had a PR of 0.69 +/− 0.08 from array analysis, which was confirmed by qRT-PCR with a calculated PR 0.46 +/− 0.06 (Supplement Table 3). These data indicate that β-actin mRNA is dissociated from polysomes following PV infection. Conversely, some messages increased their association with polysomes following inhibition of cap-dependent translation by PV. For example, PDCD8 and AMOTL2 had a PR of 2.4 and 3.0 from array analysis and 2.1+/−1.0 and 1.8+/−0.6 from qRT-PCR respectively (Supplement Table 3). Although PR of JunB from qRT-PCR (0.9) was less than that from array analysis (2+/−0.8), the data still indicated its retained association with polysomes in PV-infected EC. Thus, JunB was still considered a candidate to contain an IRES in its 5’UTR.
To determine if preserved or increased association of mRNAs with polyribosomes in PV-infected EC is also associated with translation of those transcripts and synthesis of the corresponding proteins, immunodetection assays were performed. Immunocytochemistry was used to examine accumulation of PDCD8 in PV-infected EC. Mock-infected EC were found to have basal levels of PDCD8 protein diffusely distributed in the cytoplasm (Fig 5A). Following six hours of PV infection, increased PDCD8 protein was observed throughout the cytoplasm of infected EC, consistent with new translation of PDCD8 mRNA into protein. Similarly, JunB protein increased in PV-infected EC compared to mock-infected EC as demonstrated by immunocytochemistry (Fig 5B) and Western blot analysis (Fig 5C). AMOTL2 protein expression was not examined in these experiments because no appropriate anti-AMOTL2 antibody was available.
No unequivocal consensus sequences for IRES have been defined 27. One approach to identify mRNAs with a functional IRES sequence in the 5’UTR is the use of a dicistronic assay in which expression of a reporter driven by IRES activity is compared to expression of a second reporter upstream of the candidate IRES sequence 28. To examine PDCD8, JunB, and AMOTL2 mRNAs for functional IRES elements, the 5’-UTR sequences from these genes were positioned in the intercistronic space between rLUC and fLUC cistrons. These constructs were then transfected into EaHy 926 cells, a human endothelial cell line 21, and expression levels were determined. In this assay, the translation of upstream rLUC is cap-dependent, while the translation of downstream fLUC depends on the intercistronic 5’UTR containing an IRES (Fig 6A) 28. An empty dicistronic vector and a dicistronic plasmid containing the β-actin 5’UTR were used as controls. The dicistronic data are displayed as the ratio of fLUC to rLUC activity with the ratio obtained from the empty dicistronic vector set to 1. Constructs containing the PDCD8 and AMOTL2 5’UTRs yielded fLUC/rLUC ratios of 10.5 and 125.7 respectively, while the constructs containing the actin 5’UTR only increased fLUC/rLUC ratio 3.8- fold compared to the empty vector (Fig 6B). These data suggested that the PDCD8 and AMOTL2 5’UTRs contained IRES activities. Interestingly, inserting the JunB 5’UTR into the intercistronic region only increased the fLUC/rLUC activity ratio by 5.8-fold (approximately the same level as the actin 5’UTR, Fig 6B). This result suggests that preserved translation of JunB mRNA in PV-infected cells did not require an IRES.
Apart from the presence of an IRES, alternative explanations for increased synthesis of the downstream fLUC reporter product in the dicistronic assay are cryptic transcriptional promoter activity in the cloned 5’UTR or alternative splicing introduced by the candidate 5’UTR 27. To exclude these possibilities, Northern analysis was performed using polyA mRNA isolated from the transfected EaHy cells to determine if monocistronic fLUC mRNA was produced from these constructs, thus suggesting unexpected promoter activity. The size of the Northern product would also indicate whether alternative splicing had occurred. Northern blots revealed that intact dicistronic transcripts were present in the RNA samples isolated from cells transfected with empty vector, actin, PDCD8, and AMOTL2 dicistronic constructs (Fig 6C). However, in cells transfected with the AMOTL2 dicistronic construct, an equal amount of monocistronic intact fLUC mRNA was found. Thus, monocistronic transcripts likely contributed to the fLUC activity observed in the AMOTL2 dicistronic transfection probably explaining the particularly dramatic increase in downstream fLUC synthesis (126-fold) seen with this construct. Because of this artifact, we are unable to conclusively determine whether AMOTL2 contains an IRES.
EC respond to inflammation, injury and stress signals by undergoing key changes in function and phenotype, many of which require new or altered gene expression 29. Survival and function of mammalian cells exposed to environmental and toxic stress requires reprogramming of mRNA translation in order to sustain expression of key gene products 30, but these intricate processes are largely unexplored in human endothelium. Our studies provide new insights regarding human EC and their translational responses under conditions of experimental stress imposed by viral infection.
The combination of polysome profiling and microarray analysis has been used to identify candidate IRES-containing messages in several other cell systems using different mechanisms of cellular stress 19, 25, 26. Approximately 3–5% of cellular mRNAs were found to be translated utilizing cap-independent initiation under each condition tested. In these three studies, there was no significant overlap among the genes identified, suggesting that up to 10–15% of all mRNAs are capable of using cap-independent translation initiation mechanisms 31. In our EC PV infection model, we found that approximately 12% of EC mRNAs remain associated with the polysomes. This relatively higher percentage of candidate mRNA molecules in our system could be due to the experimental procedures used to pool polysome fractions, different cell types, virus strain, stress stimulators, microarray procedures, and materials used in the studies.
Because complete cleavage of eIF4GII was never observed in PV infected EC, we cannot exclude the possibility that a very small amount of cap-dependent initiation persisted. Nonetheless, these studies indicate that translation of up to 15% of mRNA transcripts may be preserved under severe, ultimately lethal, cell stress. Adaptive mechanisms to allow ongoing translation initiation of privileged transcripts represent an under-appreciated pathway to gene expression in stressed human EC.
PV is the causative agent of poliomyelitis, in which motor neuron death leads to paralysis. PV-induced motor neuron death was recently found to be mediated through an apoptotic process 32–34. PV was also reported to induce apoptosis in vitro in other cell types, including CaCo-2 colon cancer cell line 35, U937 promonocytic cell line 36, dendritic cells and macrophages 37, and HeLa cells 38, 39. However, the mechanisms by which PV induces apoptosis are not totally clear. Induction of PV 2A protease has been reported to result in caspase-independent apoptotic cell death 40. The proposed mechanism was PV-induced preferential cap-independent translation of cellular mRNAs that encode apoptotic factors. This hypothesis is supported by the finding that several IRES-containing mRNAs encoded proteins regulate apoptosis 41–44. A recent paper reported that approximately 3% of mRNAs remain associated with the polysomes in apoptotic cells 25. In this report, we identified another apoptotic factor, PDCD8, that contains an IRES element in its 5’UTR. PDCD8, also called apoptosis-inducing factor (AIF), has been reported to mediate caspase-independent human coronary EC apoptosis induced by ox-LDL 45. In our PV-infected EC model, we found that PDCD8 mRNAs were preferentially associated with heavy polysomes and efficiently translated during PV infection. Ongoing PDCD8 synthesis in infected cells is possibly a host defense mechanism to induce apoptosis in virally infected cells.
Once infected by virus, host cells initiate an antiviral defense response. This includes the increased expression of immediate-early genes, including the AP-1 transcription factor family, which affect cell survival and the outcome of the viral infection 46. JunB, a member of the AP-1 family, is translationally upregulated in thrombin-stimulated ECs (Schmid, et al, 2011, in submission, 47) and is induced upon virus infection 46.
Using microarray and qRT-PCR, we demonstrated that JunB message remains associated with polysomes in PV-infected EC. Moreover, the persistent polysome association of JunB mRNA correlates with increased JunB protein expression in PV-infected cells. However, the dicistronic assay did not confirm the presence of an IRES in the JunB 5’UTR since the fLUC/rLUC activity ratio approximated that of the actin 5’UTR, a transcript that dissociates from polysomes in virus-infected EC. This result indicates that persistent translation of JunB mRNA in PV-infected cells may not require an IRES and is consistent with previous reports that JunB 5’UTR does not contain an IRES 42, 48.
Despite the lack of evidence for an IRES in the 5’UTR of JunB, evidence for its preserved translation in the face of PV infection is compelling. Both microarray and qRT-PCR analysis demonstrated that JunB message remains associated with polysomes in PV-infected EC. Moreover, the persistent polysome association of JunB mRNA correlates with increased JunB protein expression in PV-infected cells as shown by both Western blotting and immunocytochemistry.
We speculate that increased JunB protein expression in PV-infected EC occurs as a result of particularly avid affinity of the JunB 5’UTR to surviving components of the cap complex, possibly eIF4GII since this key protein was incompletely proteolyzed (Fig 1B). We also found that total JunB mRNA levels increased 1.8-fold upon PV infection in EC (data not shown) which would increase the number of JunB transcripts available to interact with eIF4GII. Interestingly, our microarray data showed that other AP-1 members: FosB, c-Fos, and JunD exhibit a PR of 3.7, 2.1, and 1.1 respectively, consistent with this gene family’s biological role in cell stress 47. Thus, our studies with JunB in PV-infected EC suggest that a privileged population of cellular mRNA exist that are able to complete translation initiation under stressful conditions possibly by maintaining particularly avid affinity to the cap complex.
AMOTL2 belongs to the motin family which is comprised of 3 polypeptides: angiomotin, angiomotin-like 1 (AMOTL1) and AMOTL2 49. Angiomotin binds to angiostatin and regulates angiogenesis 50 and AMOTL 2 is essential for cell movement in vertebrate embryos 51. Our microarray data imply that the motin family mRNAs are subjected to differential translational regulation in EC during PV infection. Angiomotin and AMOTL1 exhibit a PR of 0.9+/−0.3 and 1+/−0.3 respectively, while AMOTL2 has a PR of 3. qRT-PCR confirmed the increased polysome association of AMOTL2 mRNA in EC during PV infection, again suggesting that AMOTL2 might contain an IRES.
The dicistronic assay has been considered “a gold standard for detecting IRES activity” 27, but it has limitations 52. Following performance of the assay, Northern blot analysis is necessary to exclude several possible artifacts. Insertion of the 5’UTR of interest into the intercistronic region might unintentionally introduce a cryptic transcriptional promoter for the downstream cistron. Alternative splicing might excise the intercistronic segment and allow the two cistrons to be translated as one. Our dicistronic assay did show that AMOTL2 5’UTR dramatically increased fLUC/rLUC ratio when compared to the empty dicistronic plasmid and when compared to the activity of the actin 5’UTR. However, Northern analysis revealed abundant monocistronic fLUC message along with intact dicistronic transcripts in cells transfected with the rLUC-AMOTL2-fLUC reporter construct. It is likely that the monocistronic transcripts contributed to the fLUC activity observed in the AMOTL2 dicistronic assay clouding our ability to attribute the increased fLUC expression to the presence of an IRES in the AMOTL2 5’UTR. Other experimental approaches, such as intact dicistronic mRNA transfection, will be needed to definitively determine whether AMOTL2 5’UTR contains an IRES. Regardless, the notion that AMOTL2 might be preferentially translated when EC are stressed is consistent with the role of angiogenesis as a reparative process.
We have conducted preliminary experiments (data not shown) to assess whether similar mechanisms of translational control are active in EC subjected to non-viral stress such as hydrogen peroxide or arsenite. Oxidative stress produces an overall decrease in protein synthesis and a shift of mRNA to the monosomal fraction similar to the pattern depicted in Fig 2B, C. Additionally, oxidative stress induces expression of PDCD8 and JunB protein. These results suggest that mechanisms to preserve translation of key stress-related gene products are not unique to viral infection.
In summary, a model of PV infection of human EC was successfully established and produced evidence that up to 12% of cellular transcripts remain associated with the translational machinery, either through internal ribosome entry or other mechanisms. These translational control mechanisms further expand the diversity of regulated gene expression displayed by EC under severe stress.
Nahum Sonenberg and Elliot Spencer contributed important reagents, for which we are grateful. The University of Utah School of Medicine Cell Imaging Facility was used to obtain confocal and fluorescent images, and we greatly appreciate the aid of the core’s director Christopher K. Rodesch. We appreciate the aid of Donnell Benson and Jessica Phibbs for cell culture, and the significant contributions of Diana Lim in preparing the figures for this manuscript.
SOURCES OF FUNDING
This work was supported by NIH grants HL075507 (Larry W. Kraiss), HL66277 (Andrew S. Weyrich), and R37HL44525 (Guy A. Zimmerman). Hansjörg Schwertz was supported by a Beginning-Grant-in-Aid (09BG1A 2250381) from the American Heart Association Western States Affiliate.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.