Eukaryotic initiation factor 5A2 regulates expression of antiviral genes

Summary

Translation factors are essential for regulation of protein synthesis. The eukaryotic translation initiation factor 5A (eIF5A) family is made up of two paralogues – eIF5A1 and eIF5A2 – which display high sequence homology but distinct tissue tropism. While eIF5A1 directly binds to the ribosome and regulates translation initiation, elongation, and termination, the molecular function of eIF5A2 remains poorly understood. Here, we engineer an eIF5A2 knockout allele in the SW480 colon cancer cell line. Using ribosome profiling and RNA-Sequencing, we reveal that eIF5A2 is functionally distinct from eIF5A1 and does not regulate transcript-specific or global protein synthesis. Instead, eIF5A2 knockout leads to decreased intrinsic antiviral gene expression, including members of the IFITM and APOBEC3 family. Furthermore, cells lacking eIF5A2 display increased permissiveness to virus infection. Our results uncover eIF5A2 as a factor involved regulating the antiviral transcriptome, and reveal an example of how gene duplications of translation factors can result in proteins with distinct functions.

Introduction

mRNA translation is regulated through the efforts of over a dozen translation factors [1],[2], which coordinate to regulate the loading, elongation, and termination of the ribosome on transcripts. In agreement to their importance, dysregulation of many translation factors is linked to development disorders and carcinogenesis [3]. Intriguingly, several translation factors and ribosomal proteins exist as gene duplications [4],[5]. However, it remains poorly understood if these duplications allow for gain of additional functions, or instead the paralogues exist as functionally redundant factors.

The eukaryotic translation initiation factor 5A (eIF5A) family contains two paralogue genes, eIF5A1 and eIF5A2, which share 84 % identity at the amino acid level [6],[7] and are found duplicated throughout Eukarya (Fig. 1A) [6]. The eIF5A proteins are notably the only proteins to be post-translationally modified with hypusine [6]. eIF5A1 has multiple functions in translation elongation and termination [8]. eIF5A1 binds directly to the ribosome and increases peptidyl transferase activity, along with stimulating the release of newly synthesized peptides during termination [8]. eIF5A1 is also essential for relieving ribosome pausing at motifs such as poly-proline stretches [8]. In agreement with these functions in translation regulation, eIF5A1 depletion leads to alterations to critical cellular pathways, including autophagy, macrophage activation, neurodevelopment [9],[10],[11]. Additionally, eIF5AL1, an eIF5A1 isoform with 98% sequence conservation, is linked to carcinogenesis [12].

Figure 1. Knockout of eIF5A2 from SW480 cells reduces cellular migration.

(A) Sequence alignment of eIF5A1 and eIF5A2. The SH3-fold domain is highlighted in blue, and the OB-fold domain is highlighted in purple. (B) Crystal structure of human eIF5A1 (PDB 3CPF), with amino acid divergence between eIF5A1 and eIF5A2 colored in red [13]. (C) Representative western blot of wild type (WT) and eIF5A2^KO SW480 cells, showing loss of eIF5A2 expression but no change to eIF5A1 levels. (D) Representative image of transwell cell migration by WT and eIF5A2^KO SW480, using 30 % FBS as chemoattractant. (E) Quantification of cell migration. The results are given as the mean ± standard deviation (s.d.) of at least three independent experiments. (F) MTT assay measuring cell proliferation in WT and eIF5A2^KO SW480 cells. The P value is calculated by unpaired t-test. ***P<0.001, **P<0.005.

In contrast to eIF5A1, the molecular function of eIF5A2 remains poorly characterized. Structurally, eIF5A1 is composed of two domains, an N-terminal SRC Homology 3 (SH3)-fold domain, and a C-terminal domain composed of an oligonucleotide-Binding (OB) fold [13]. The majority of the residues that are divergent between eIF5A1 and eIF5A2 are clustered within the C-terminal domain, suggesting these mutations could alter nucleic-acid binding specificity (Fig. 1B). Notably, unlike eIF5A1, which is ubiquitously expressed, eIF5A2 expression is limited to the brain and testes. However, the genomic locus of eIF5A2, chromosome 3q26, is commonly amplified in many cancers and found upregulated in many cancers, including colorectal, bladder, gastric, hepatocellular, pancreatic and lung cancers [14]. In addition, eIF5A1 and eIF5A2 also have differing functions in supporting cancer growth. For example, injection of nude mice with p53^−/− hepatocytes overexpressing eIF5A2 leads to increased tumor volume, when compared with cells overexpressing eIF5A1 [15]. Furthermore, overexpression of eIF5A2, but not eIF5A1, modulates the epithelial mesenchymal transition (EMT) and increases expression levels of EMT-related proteins such as the metastasis-associated protein 1 (MTA1) [16].

Despite these genetic studies highlighting a distinct role of eIF5A2 during cancer development and cell fate decisions, it remains unclear if eIF5A1 and eIF5A2 differ in molecular function. Here, we generated eIF5A2 knockout cell lines and use ribosome profiling to demonstrate that, unlike eIF5A1, eIF5A2 does not regulate mRNA translation. Using RNA-Sequencing, we reveal that knockout of eIF5A2 induces widespread changes to the transcriptome, including reducing levels of factors involved in cellular migration, development, and the antiviral response. We further demonstrate that eIF5A2 knockout leads to increased cellular permissiveness to infection by RNA viruses.

Results

eIF5A2 is not required for global protein synthesis

To identify the gene program controlled by eIF5A2, we used CRISPR-Cas9 gene editing to construct a knockout allele in the colorectal cancer SW480 cell line, which overexpress eIF5A2 (Fig. 1C). In agreement with the function of eIF5A2 in the epithelial to mesenchymal transition [16],[17], eIF5A2 knockout (SW480 eIF5A2^KO) cells expressed lower levels of the mesenchymal marker protein vimentin, and exhibited a 48-fold decrease in cellular migration in a transwell assay, using FBS as the chemoattractant (Fig. 1D, E). Furthermore, knockout of eIF5A2 significantly decreased cellular proliferation (Fig. 1F). Importantly, the expression of eIF5A1 is unaltered in eIF5A2 knockout cells, indicating that the two paralogs are not coregulated in SW480 cells (Fig. 1C).

eIF5A1 was previously shown to be involved in numerous steps of translation, including elongation and termination [8]. Given the high sequence conservation between the two proteins (Fig. 1A), we first examined if eIF5A2 knockout altered global protein synthesis. By [³⁵S]-methionine/cysteine metabolic labeling, we did not observe a significant change in protein synthesis between SW480 and eIF5A2 knockout cell lines (Fig. 2A, B). Furthermore, polysome profile analysis showed similar monosome to polysome ratios between wild type and eIF5A2 knockout cell lines, indicating general translation is unaffected by loss of eIF5A2 (Fig. 2C,D). Altogether, our results reveal that, unlike eIF5A1, eIF5A2 does not regulate general protein synthesis.

Figure 2. eIF5A2 is not required for global translation.

(A) [³⁵S]-Met/Cys labeling of global protein synthesis in WT and eIF5A2^KO SW480 cells. Coomassie staining is showed as a loading control. (B) Quantification of [³⁵S]-Met/Cys metabolic labeling. The intensity of [³⁵S] labeling is normalized to Coomassie staining, and expressed as fold change to WT SW480 cells. (C) Polysome profile of WT and eIF5A2^KO SW480 cells. The results in (A), (C) are representative of three independent experiments. (D) Quantification of the monosome to polysome ratio, as calculated by area under the curve. The results in (B), (D) are given as the mean ± s.d. of three independent experiments. The P value is calculated by unpaired t-test. ns, non-significant.

Knockout of eIF5A2 leads to widespread changes to the transcriptome

Given that eIF5A2 is not essential for cell viability, one possibility is that eIF5A2 regulates transcript-specific translation rather than protein synthesis on a global level. To obtain high resolution information of translational changes that occur upon knockout of eIF5A2, we performed ribosome profiling, an approach that uses deep-sequencing of ribosome-protected footprints to identify changes in the association of ribosomes on mRNAs on a genome-wide level [18]. By mapping ribosome-protected footprints to the transcriptome and normalizing these numbers to total mRNA level, as determined by RNA-Sequencing, we obtained translation efficiency measurements for each mRNA. Less than thirty mRNAs were translationally downregulated upon eIF5A2 knockout, and manual inspection of these genes revealed that the footprints were mapped to intronic regions and thus likely attributed to mapping artefacts (Fig. 3A).

Figure 3. Knockout of eIF5A2 alters expression of genes involved in migration, development, and anti-viral response.

(A) Volcano plot comparing log-fold change in translation efficiency with adjusted P values for mRNAs in WT versus eIF5A2^KO SW480 cells.

(B) Volcano plot comparing log-fold change in transcript expression levels with adjusted P values for genes in WT versus eIF5A2^KO SW480 cells. (C) Gene ontology analysis of transcripts downregulated in eIF5A2^KO SW480 cells. Functional categories below P value = 0.025 are shown. (D) Read mapping to transcripts downregulated in eIF5A2^KO SW480. GAPDH is a negative control mRNA that is unaffected by eIF5A2 knockout.

Given that we do not observe functions for eIF5A2 in mRNA translation, we next asked if eIF5A2 regulates the transcriptome. Analysis of the RNA-Sequencing data from wild type and eIF5A2^KO SW480 cells revealed widespread transcriptional changes, with 377 genes exhibiting upregulated expression and 477 genes being downregulated more than 2-fold and with a P value cutoff of 0.05 (Fig. 3B). We performed gene ontology analysis of the downregulated genes, and found three major functional categories regulated by eIF5A2 (Fig. 3C). In agreement to the decreased cellular migration (Fig. 1D, E), transcript levels of mRNAs that encode proteins involved in migration and adhesion was blocked in eIF5A2 KO cells, including the intermediate filament protein vimentin. Furthermore, eIF5A2 knockout decreases expression of genes involved in development, such as homeobox family proteins [19],[20]. Finally, eIF5A2 regulates genes involved in defense against viruses, a function that has not previously been reported for eIF5A2. Intriguingly, eIF5A2 appears to co-regulate multiple proteins within a gene family. For example, eIF5A2 regulates the level of the interferon-induced transmembrane (IFITM)-Family transcripts (IFITM1, IFITIM2, IFITM3) (Fig. 3D), which are restriction factors that block the cellular entry of various viruses including influenza A virus [21], flaviviruses [21], rhabdoviruses [22], and filoviruses [23]. In addition, eIF5A2 regulates expression of five genes in the Apolipoprotein B Editing Complex (APOBEC3) family, which are cytidine deaminases that restrict retrovirus replication by leading to genomic instability [24].

eIF5A2 regulates basal expression of antiviral genes

We next used quantitative RT-PCR and western blot to validate candidate transcripts identified by RNA-Sequencing as being dependent on eIF5A2 for efficient expression (Fig. 4A, B, S1A). We validated that vimentin was downregulated by 42-fold, and that the migration-related gene Transforming Growth Factor Beta Induced (TGFBi) was downregulated by 11-fold. Notably, the TGF-β pathway is a key regulator of EMT and migration, and modulation of eIF5A2 levels by knockdown [17],[25] or knockout [26] inhibits both cellular processes (Fig. 1C-E). For the development cluster, we validated that homeobox B6 (HoxB6), homeobox B9 (HoxB9), Wnt family member 6 (Wnt6), and nemo like kinase (NLK) are all downregulated by 2.3 to 5 fold upon eIF5A2 knockout (Fig. 4A, B). Intriguingly, the HoxB family members are important for proper skeletal development, and overexpression of eIF5A2 in transgenic mice impairs ossification during development leading to premature skeletal degeneration [27]. In addition, Wnt6 is a critical regulator of organ development and embryogenesis [28] and NLK is a kinase that is involved in diverse developmental cellular processes [29]. We also confirmed that the IFITM and APOBEC3 family proteins, along with interferon alpha inducible protein 27 (IFI27), are all substantially downregulated upon knockout of eIF5A2 (Fig. 4A, B). Finally, as a control for potential off-target effects of gene-editing, we also examined two additional eIF5A2 knockout clonal cell lines, and found that all three lines exhibit the same downregulation of eIF5A2 target genes (Fig. S1B, C). As eIF5A1 and eIF5A2 are divergent in their role on general protein synthesis, we next asked if they exhibit conserved roles in regulating the transcriptome. Depletion of eIF5A1 in SW480 cells does not alter the levels of the eIF5A2 targets HoxB9 or APOBEC3G (Fig. S2). Altogether, our results indicate that eIF5A2 is functionally distinct from eIF5A1 through its regulation of the transcriptome.

Figure 4. SW480 cells display increased permissivity to viral infection upon knockout of eIF5A2.

(A) Fold change in transcript expression levels in eIF5A2^KO versus WT SW480 cells as measured by quantitative RT-PCR. (B) Representative western blot of WT and eIF5A2^KO SW480 cells, showing decreased expression of target genes identified by RNA-Sequencing. Hsp90 is a loading control. (C) Fold change in IFN-β-mediated induction of transcripts in eIF5A2^KO and WT SW480 cells, as measured by quantitative RT-PCR. Fold change induction is compared to corresponding vehicle-treated cells. (D) Representative images of eIF5A2^KO and WT SW480 cells infected with rVSV-EGFP. (E) Fold change in viral titer output from WT and eIF5A2^KO SW480 cells infected with rVSV-EGFP. PFU, Particle forming units. Cells in (D) and (E) were infected at MOI 3 and images or cell supernatant were taken at 4.5 h post infection. (F) Fold change in luciferase activity in WT and eIF5A2^KO SW480 cells infected with murine leukemia virus pseudotyped with the Marburg virus glycoprotein and expressing Firefly luciferase as a reporter for infection. The results in (A), (C), (E–G) are given as the mean ± s.d. of three independent experiments. The P value is calculated by unpaired t-test. **P<0.005, *P<0.05.

Antiviral interferon-stimulated genes (ISG) can be transcriptionally controlled in response to cytokine stimulation or at a constitutive basal level. As ISG levels are already altered in unstimulated wild type versus eIF5A2^KO SW480 cells, we next asked if eIF5A2 is additionally important for transcription upregulation upon treatment with interferon beta (IFN-β). As expected, stimulation with IFN-β led to massive upregulation of the IFITM and IFI27 transcripts [30],[31]. This upregulation is observed regardless of the expression of eIF5A2, and there are not statistical differences in the magnitude of transcriptional activation (Fig. 4C). Thus, these results indicate that eIF5A2 is important for the basal expression of antiviral genes but not for transcriptional upregulation upon IFN-β stimulation.

Loss of eIF5A2 makes cells more permissive for viral infection

Given our findings that eIF5A2^KO SW480 cells exhibit downregulated expression of antiviral genes, we next tested if eIF5A2 controls cellular susceptibility to viral infection and levels of viral replication. Vesicular stomatitis virus (VSV) is a rhabdovirus which has previously been shown to be exquisitely sensitive to interferon [32]. We infected wild type and eIF5A2^KO SW480 with a recombinant VSV strain that is engineered to express eGFP as a reporter for viral replication [33],[34]. eIF5A2^KO SW480 cells display increased numbers of infected cells, as assessed by eGFP expression, when compared to wild type SW480 cells (Fig. 4D). Furthermore. In addition, infected eIF5A2^KO SW480 cells produce ~4-fold higher virus titers compared to wild type cells (Fig. 4E). Thus, knockout of eIF5A2 causes cells to become more permissive for VSV replication and viral output.

As the IFITM family is known to restrict viral entry mediated by filovirus glycoproteins, we next tested if eIF5A2^KO SW480 cells are more susceptible to infection mediated by the Marburg virus (MARV) glycoprotein [23]. We rescued recombinant murine leukemia virus which express Firefly luciferase as a reporter for infection, and were pseudotyped with the MARV glycoprotein. Knockout of eIF5A2 lead to a 3.5-fold increase luciferase activity when compared to WT cells, indicating there was an increase in MARV glycoprotein-mediated entry, in agreement with a decrease in IFITM levels (Fig. 4F). Altogether, our results reveal that eIF5A2 expression regulates expression of an antiviral gene signature important for restriction of viral infection.

Discussion

Here, we reveal that eIF5A2 has a primary function in regulation of the transcriptome but not protein synthesis. In agreement with the genetic link between eIF5A2 upregulation and cancer, we find that eIF5A2 knockout modulates the expression of genes important for cellular migration and differentiation (Fig. 3C, 4A, S1B, C). Unexpectedly, we also discover that eIF5A2 expression is inhibitory towards viral replication, through promoting expression of antiviral genes like the IFITM and APOBEC3 family member. Notably, eIF5A2 knockdown leads to loss of constitutive ISG expression while still permitting IFN-induced signaling, and thus eIF5A2 has a molecular role in intrinsic, but not inducible, antiviral gene regulation. Given that eIF5A2 expression in non-cancer cells is limited to the testes and brain, it is of interest to examine if eIF5A2 is important for the antiviral response in these tissues. Indeed, certain viruses infect testes, including Zika Virus and mumps virus [35],[36]; or are neurotropic, including poliovirus, and rhabdoviruses closely related to VSV such as rabies [37].

Our results show that eIF5A2 functionally differs from eIF5A1 in human cells on a molecular and cellular level (Fig. 2, S2B). Intriguingly, eIF5A1 and eIF5A2 are particularly divergent in their cellular function during viral replication, with eIF5A1 supporting replication of viruses that can be inhibited by eIF5A2 expression. For example, while eIF5A2 promotes the expression of APOBEC3 proteins that block HIV-1 replication (Fig. 4A), eIF5A1 binds to the HIV-1 Rev protein and enhances retroviral genome nuclear export and replication [38],[39]. Furthermore, while eIF5A2 expression is required for basal transcription of IFITM proteins that restrict filovirus entry (Fig. 4C, 4F), eIF5A1 is necessary for the synthesis of the ebolavirus VP30 polymerase component protein and subsequently acts in a pro-viral replication capacity [39]. Finally, modulation of eIF5A1 activity through treatment of cells with chemical inhibitors reduces replication of diverse RNA viruses, including rhabdoviruses, flaviviruses, picornaviruses, and bunyaviruses [41].

How can the two paralogs have different roles despite high sequence conservation? The clustering of divergent residues is in the C-terminal OB-fold, suggesting this is the major domain involved in dictating the unique functions of the eIF5A family proteins. Given that OB-fold domains can be involved in direct nucleic acid binding, this domain could explain why eIF5A2 knockout leads to specific changes in the transcriptome. Additionally, eIF5A2 could have differential functions from eIF5A1 due to cellular localization. Indeed, both eIF5A1 and eIF5A2 show dynamic cellular localization that is regulated through post-translational modifications. Hypusination of the eIF5A proteins at position K50 leads to cytoplasmic localization, while acetylation at position K47 promotes nuclear localization [42]. These post-translational modifications are not equally distributed for both family members, as eIF5A1 has a ~6-fold higher affinity for deoxyhypusine synthase and thus is a better substrate for hypusination than eIF5A2 [6]. Finally, knockdown of the nuclear export protein Xpo4 lead to nuclear accumulation of eIF5A2, and subsequently promotion of tumor growth, suggesting eIF5A2 cellular localization is also important for its role in carcinogenesis [15]. It will be of interest to examine the biochemical mechanisms that allow eIF5A2 to function differentially from eIF5A1 in gene expression.

Indeed, the role of eIF5A2 in controlling the transcriptome could be through direct or indirect regulation of transcription, mRNA stability, or mRNA localization. Notably, prior work showed that eIF5A2 can bind directly to the HIF-1ɑ promoter upon hypoxic stress in esophageal squamous cell carcinomas [43]. Furthermore, eIF5A2 shuttles to the nucleus upon cellular perturbations such as hypoxia, and eIF5A2 overexpression leads to increased nuclear localization of the transcription factor STAT3[17],[43]. Finally, single amino acid mutations in the C-terminus of the yeast eIF5A1 homolog block mRNA decay and lead to accumulation of uncapped mRNAs, suggesting the non-conserved residues in the C-terminal domain of eIF5A2 could also be involved in regulating mRNA stability [44],[45]. Overall, our work reveals that eIF5A2 is molecularly distinct from eIF5A1, and demonstrates how two translation factor paralogs can mediate divergent cellular functions.

Methods

Reagent/Resource	Source	Identifier
Antibodies
Rabbit anti eIF5A	Proteintech	17069-1-AP
Mouse anti GAPDH	Proteintech	60004-1-Ig
Rabbit anti Vimentin	Proteintech	10366-1-AP
Mouse anti E-Cadherin	Proteintech	60335-1-1g
Mouse anti IFITM1	Proteintech	60074-1-Ig
Rabbit anti IFITM3	Proteintech	11714-1-AP
Mouse anti Hsp90	BD bioscience	BD 610418
Rabbit anti APOBEC3G	Cell signaling	43584S
Mouse anti HoxB9	Santa Cruz Biotechnology	sc-398500
Goat anti Mouse IgG HRP	Fisher	G-21040
Goat anti Rabbit HRP	Fisher	G21234
Experimental models organism/strains
SW480	ATCC	CCL-228
Hek293T	ATCC	CRL-11268
BHK-21	ATCC	CCL-10
Vero	ATCC	CCL-81
rVSV-EGFP	Kind gift from Whelan lab
Plasmids
pLBH269 SpCas9 N-Term 5A2 sgRNA	Plasmid derivative of pX330	N/A
pBA392 shRNA 5A1
pBA392 shRNA non targeting
psPAX2
pMD2.G
Chemicals, Peptides and recombinant proteins
DMEM	Invitrogen	11995073
OptiMEM	Fisher	31985070
FBS	Biowest	S1620
Trypsin	Genesee	25-510
100X Penicillin-Streptomycin	Genesee	25-512
Polybrene	Sigma	H9268-5G
Puromycin Dihydrochloride	Fisher	AAJ61278MC
Polyethylenimine, Linear, MW 25,000	Polysciences, Inc	23966-1
Lipofectamine 2000	Fisher	LMRNA001
Emetine Dihydrochloride Hydrate	Med Chem Express	HY-B1479B
Cycloheximide	VWR	97064-724
EXPRE35S35S Protein Labeling Mix	Perkin Elmer	NEG072002MC
Crystal Violet	Fisher	AC405830250
Turbo Dnase	Fisher	AM2238
RNase I	Lucigen	N6901K
SUPERase-In Rnase Inhibitor	Fisher	AM2696
TRIzol Reagent	Fisher	15596026
Glycogen	VWR	97063-256
Mth RNA Ligase	NEB	M2611A
T4 PNK	NEB	M0201L
T4 RNA Ligase 2, truncated K227Q	NEB	M0351L
Yeast 5´ Deadenylase	NEB	M0331S
Rec J Exonuclease	Lucigen	RJ411250
Murine RNase inhibitor	NEB	M0314L
CircLigaseTM ssDNA Ligase	Lucigen	CL4111K
Dynabeads™ Biotin Binder	Life Technologies	11047
Q5 Polymerase	NEB	M0491L
Firefly luciferase kit	Genecopoeia LF012	LF012
DMSO	VWR	97061-250
Thiazolyl blue tetrazolium bromide, 98%	Fisher	AC158990010
qPCR primers	Sequence	Origin
eIF5A2 Forward	TTTGCCAGCTGAAAGTTCCCA	This paper
eIF5A2 Reverse	TGCATGGTCGTCCTTTGAGC	This paper
IFITM1 Forward	CCCCAAAGCCAGAAGATGCACAAGGAG	This paper
IFITM1 Reverse	CGTCGCCAACCATCTTCCTGTCCCTAG	This paper
IFITM2 Forward	AGAAAACGGAACTACTGGGGAA	This paper
IFITM2 Reverse	GAGCATCTCGTAGTTGGGAGG	This paper
IFITM3 Forward	GTCCACCGTGATCCACATCC	This paper
IFITM3 Reverse	CATAGGCCTGGAAGATCAGCAC	This paper
APOBEC3B Forward	AGCGCTTCAGAAAAGAGTGG	This paper
APOBEC3B Reverse	AAGTTTCGTTCCGATCGTTG	This paper
APOBEC3F Forward	TCCGTGGAGATCATGGGCTACAAA	This paper
APOBEC3F Reverse	TGCAGCTTGCTGTCCAGGAATAGA	This paper
APOBEC3G Forward	GCTGTGCCCAGGAAATGGCTAAAT	This paper
APOBEC3G Reverse	ACAAAGGTGTCCCAGCAGTGCTTA	This paper
IFI27 Forward	TCTCCTTCTTTGGGTCTGGC	This paper
IFI27 Reverse	TGGCCACAACTCCTCCAATC	This paper
Vimentin Forward	TGTCCAAATCGATGTGGATGTTTC	This paper
Vimentin Reverse	TTGTACCATTCTTCTGCCTCCTG	This paper
HOXB6 Forward	GGAGCACTGTCGTCCTTCAG	This paper
HOXB6 Reverse	ATTGTCCCAACGTGAGAGCC	This paper
HOXB9 Forward	TACCTCACCAGGGACCGTAG	This paper
HOXB9 Reverse	GGGAGGACTGGGGGTAATCT	This paper
WNT-6 Forward	GGTTATGGACCCTACCAGCA	This paper
WNT-6 Reverse	AACTGGAACTGGCACTCTCG	This paper
NLK Forward	TTGGCCAGAGTGGAAGAATTAG	This paper
NLK Reverse	CCCACAGACCAGATGTCAATAG	This paper
TGFBi Forward	GTCCTGACTGATGAGCTGAAA	This paper
TGFBi Reverse	TGTTGGTGATGGTGGAGATG	This paper
GAPDH Forward	GAAGGTGAAGGTCGGAGTCAAC	This paper
GAPDH Reverse	CAGAGTTAAAAGCAGCCCTGGT	This paper
PGK1 Forward	TTGACCGAATCACCGACCTC	This paper
PGK1 Reverse	CATAACGACCCGCTTCCCTT	This paper
ROBO1 Forward	CCAGAGAGAGCTGGGAAATG	This paper
ROBO1 Reverse	CCAGTCTGATTCTCCGTGGT	This paper
c-myc Forward	TGAGGAGACACCGCCCAC	This paper
c-myc Reverse	CAACATCGATTTCTTCCTCATCTTC	This paper
GPX2 Forward	GACTTCACCCAGCTCAACGA	This paper
GPX2 Reverse	ATGCTCGTTCTGCCCATTCA	This paper
rRNA Depletion oligos	Sequence	Origin
LSU_rRNA_1	ggatctttcccgccccccgttcctcccgacccctccacccgccctcccttcccc	This paper
LSU_rRNA_2	aaggtggctcggggggccccgtccgtccgtccgtccgtcctcctcctcccccgtctccgc	This paper
LSU_rRNA_3	ccgcgaggggggtctcccccgcgggggcgcgccggcgtctcctcgtgggg	This paper
LSU_rRNA_4	ggcgacggggggggtgccgcgcgcgggtcggggggcggggcggactgtccccagtgcgcc	This paper
LSU_rRNA_5	actctggtggaggtccgtagcggtcctgacgtgcaaatcggtcgtccgacctgggtatag	This paper
LSU_rRNA_6	ccgacgcacccccgccacgcagttttatccggtaaagcgaatgattagag	This paper
LSU_rRNA_7	cgcgaagcggggctgggcgcgcgccgcggctggacgaggcgccgccgccccccccacgcc	This paper
LSU_rRNA_8	cggcgggggcggcgcgcgcgcgcgcgcgtgtggtgtgcgtcggagggcggcggcggcggc	This paper
LSU_rRNA_9	ccgcgaggggggcccgggcacccggggggccggcggcggcggcgactctg	This paper
LSU_rRNA_10	ccgggcccttcccgtggatcgccccagctgcggcgggcgtcgcggccg	This paper
LSU_rRNA_11	cgcgtgccccgccgcgcgccgggaccggggtccggtgcggagtgcccttcgtcctgggaa	This paper
LSU_rRNA_12	gagaggcggccgccccctcgcccgtcacgcaccgcacgttcgtggggaacctggcgctaa	This paper
rRNA_depl_1	GGGGGGATGCGTGCATTTATCAGATCA	[49]
rRNA_depl_2	TTGGTGACTCTAGATAACCTCGGGCCGATCGCACG	[49]
rRNA_depl_3	GAGCCGCCTGGATACCGCAGCTAGGAATAATGGAAT	[49]
rRNA_depl_4	TCGTGGGGGGCCCAAGTCCTTCTGATCGAGGCCC	[49]
rRNA_depl_5	GCACTCGCCGAATCCCGGGGCCGAGGGAGCGA	[49]
rRNA_depl_6	GGGGCCGGGCCGCCCCTCCCACGGCGCG	[49]
rRNA_depl_7	GGGGCCGGGCCACCCCTCCCACGGCGCG	[49]
rRNA_depl_8	CCCAGTGCGCCCCGGGCGTCGTCGCGCCGTCGGGTCCCGGG	[49]
rRNA_depl_9	TCCGCCGAGGGCGCACCACCGGCCCGTCTCGCC	[49]
rRNA_depl_10	AGGGGCTCTCGCTTCTGGCGCCAAGCGT	[49]
rRNA_depl_11	GAGCCTCGGTTGGCCCCGGATAGCCGGGTCCCCGT	[49]
rRNA_depl_12	GAGCCTCGGTTGGCCTCGGATAGCCGGTCCCCCGC	[49]
rRNA_depl_13	TCGCTGCGATCTATTGAAAGTCAGCCCTCGACACA	[49]
rRNA_depl_14	TCCTCCCGGGGCTACGCCTGTCTGAGCGTCGCT	[49]
Softwares and informatic resources
Rstudio	1.4.1717
CFX Maestro	Biorad
ImageQuant TL	GE Healthcare lifescience
IGV	Robinson, 2011
Windaq	DataQ intruments
DAVID Bioinformatics Resources 6.8	Huang, 20091; Huang, 20092

Contact for Reagent and Resource Sharing

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact, Amy S.Y. Lee (amysy_lee@dfci.harvard.edu)

Cell culture.

All cells were cultured in complete media (DMEM, high glucose and L-glutamine (Gibco) with 10 % (v/v) FBS (Biowest) and 500 U mL⁻¹ penicillin/streptomycin). For generation of the eIF5A2^KO cell line, SW480 were transfected with pLBH269 SpCas9 N-Term 5A2 sgRNA plasmid, which expresses Cas9 and an eIF5A2-targeting sgRNA, using Lipofectamine 2000 (Thermo Fisher Scientific) following manufacturer’s protocol. After 1 day post transfection, media was changed to complete media supplemented with 1μg mL⁻¹ puromycin to select for transfected cells. After 48 h of selection, cells were trypsinized and seeded in 96-well plates for single cell cloning. Single colony clones were harvested and genotyped by Sanger sequencing to identify eIF5A2 knockout. pLBH269 was a kind gift from L. Harrington and J. Doudna (UC Berkeley). For interferon-β treatment, cells were seeded the day before to be 70 % confluent on the day of treatment. Media was changed to fresh complete media supplemented with 500 U mL⁻¹ interferon-β, and cells were harvested after 24 h of incubation for isolation of total RNA. shRNA-expressing lentiviruses were produced as previously described [46] using lentivirus transfer plasmids encoding the shRNA sequence (non-targeting: 5′-TACAACAGCCACAACGTCTAT-3′, eIF5A1 targeting: 5′-GCTGGACTCCTCCTACACA-3′). To make stable knockdown SW480 cell lines, lentivirus infections were performed in the presence of 10 μg ml−1 hexadimethrine bromide (Sigma), and at 24 h post infection, cells selected with puromycin (Fisher) at 1 μg ml−1 for one week before subsequent experiments.

Quantitative RT-PCR.

Total RNA was isolated by phenol-chloroform extraction and ethanol precipitation. cDNA was reverse transcribed using random hexamers and Moloney Murine Leukemia virus M5 reverse transcriptase (MMLV M5 RT) [47] using standard conditions [46]. Quantitative PCR was performed with Luna Universal qPCR Master Mix (NEB) according to manufacturer’s protocol using a CFX Connect Real-Time PCR Detection System (Biorad).

Western blots.

Cell extracts were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 % (v/v) NP-40, 0.5 % (v/v) sodium deoxycholate, 0.1 % (w/v) SDS) for 5 minutes on ice. Western blot analysis was performed using the following antibodies: anti-eIF5A1/eIF5A2 (Abclonal A4864), anti-Hsp90 (BD bioscience BD 610418), anti-GAPDH (Proteintech 60004-1-Ig), anti-Vimentin (Proteintech 10366-1-AP), anti-E-cadherin (Proteintech 60335-1-1g), anti-APOBEC3G (cell signaling 43584S), anti-IFITM1 (Proteintech 60074-1-Ig), anti-IFITM3 (Proteintech 11714-1-AP), anti-HoxB9 (Santa Cruz Biotechnology sc-398500).

Migration assay.

SW480 cells were trypsinized and resuspended to seed 5 × 10⁵ cells per well in 500 μL of DMEM supplemented with 1 % (v/v) FBS in the upper chamber of 8 μm pore size Corning Transwell (Fisher) inserts. The inserts were placed in a Corning Companion plate (VWR) and 600 μL of DMEM supplemented with 30 % (v/v) FBS was added to the lower chamber as the chemoattractant factor. After 48 h at 37 °C with 5 % CO₂, the cells on the upper surface were removed with a cotton swab, and migrated cells on the lower surface of the insert were fixed with 4 % (v/v) paraformaldehyde in PBS and stained with 0.1 % (w/v) crystal violet. Images of the migrated cells were taken with an EVOS imaging system (Thermo Fisher) and the cells were counted four randomly selected fields.

MTT assay.

Cells were seeded at a density of 10,000 cells per well in a 96-well plate in 100 μL of complete DMEM media and incubated for the indicated timepoints. At each timepoint, media was replaced with 30 μL of complete DMEM supplemented with methyl thiazolyl tetrazolium (MTT) at a concentration of 0.5 mg per well. After 1 h of incubation, 150 μL of DMSO was added to each well and cells were incubated for 10 min with gentle shaking protected from light. The plate was read at a wavelength of 570 nm using a microplate reader (Biotek Synergy H1).

Virus Production and Infection.

Pseudotyped murine leukemia virus (MLV) was rescued as previously described [23]. Plasmids for virus production were a kind gift from M. Farzan (Scripps Research). Briefly, HEK293T cells in a 10-cm plate were transfected with 12 μg of pCMV plasmid DNA encoding MLV gag and pol, 12 μg of pQC plasmid DNA encoding firefly luciferase, and 2.4 μg of plasmid DNA encoding Marburg virus glycoprotein lacking the mucin domain using PEI (1 mg ml⁻¹ linear polyethylenimine (Polysciences)) [47]. Cells were incubated at 37 °C for 8 h and existing media was replaced with fresh complete media. Cells were incubated 32 °C for 72 h, supernatant was harvested, and debris was removed by centrifugation at 3000 × g for 15 min at 4 °C in an Eppendorf 5810R centrifuge. Virus was pelleted by centrifugation in a JA-17 rotor at 28,000 × g for 160 min at 4 °C, and virus was resuspended in NTE buffer (100 mM NaCl, 1 mM EDTA-NaOH pH 8, 10 mM Tris-HCl pH 7.4). For infection of SW480 cells with pseudotyped MLV, cells were seeded the day before in a 96-well plate, and subsequently infected with 1 μL of concentrated virus in 30 μL of complete media supplemented with 10 μg mL⁻¹ polybrene. After 1 h of virus absorption at 37 °C, virus was removed and 100 μL complete media was added. Cells were harvested at 24 h post infection (hpi) and Firefly luciferase activity was measured following manufacturer’s protocol (Genecopoeia) using a Glomax Multi+ plate reader (Promega).

Recombinant VSV virus expressing GFP (rVSV-EGFP) was a kind gift from S. Whelan (Washington University) [33]. Briefly, for propagation of rVSV-EGFP, BHK-21 cells were infected at MOI of 0.01, and supernatant was harvested 16-20 h post infection upon visualization of cytopathic effect. Cellular debris was clarified and virus was pelleted as described above. For infection of SW480 cells with rVSV-EGFP, cells were infected at multiplicity of infection of 3. After 1 h of virus absorption at 37 °C in DMEM, virus was removed and complete media was added. At 4.5 hpi, images were taken on an EVOS imaging system (Thermo Fisher) and supernatant was harvested. Viral titers were determined by plaque assay on Vero cells as previously described [34].

Polysome profiling.

Polysome profiling was performed as previously described with the following alterations [48]. Cells were harvested on ice by scraping into ice cold PBS with 5 μM emetine, and pelleted by centrifugation for 3 min at 300 × g at 4 °C. Cell pellets were lysed in 180 μl ice cold lysis buffer (20 mM HEPES-KOH pH 7.4, 150 mM KOAc, 5 mM MgOAc₂, 0.2 mM spermidine, 1 % (v/v) Triton-X 100, 1 mM DTT, 5 μM emetine) for 10 min on ice, passed five times through a 21-g needle, and nuclei were pelleted by spinning at 15,000 × g for 5 min at 4 °C. Lysates were loaded onto a 10-50 % (w/v) sucrose gradient made in lysis buffer lacking detergent or emetine, and centrifuged for 2 h at 4 °C at 38,000 rpm in a SW41 Ti rotor. The gradient was fractioned using a Brandel gradient fractionator and A254 profiles were acquired using a Dataq data acquisition system. Ribosomes were pelleted from fraction by centrifugation for 2 h at 4 °C at 100,000 rpm in a TLA110 rotor. Quantification of the monosome to polysome ratio was performed through analysis of individual peaks using the trapezoidal rule.

Metabolic labeling.

Cells were seeded the day before to be 80 % confluent the day of the experiment. Media was changed to DMEM without methionine or cysteine (Invitrogen) supplemented with 10 μg mL⁻¹ actinomycin D. After 30 min, 10 μL of EXPRE³⁵S³⁵S Protein Labeling Mix (Perkin Elmer) was added to each plate to start the metabolic labeling. After 2 h, cells were harvested by scraping into ice cold PBS and pelleted by centrifugation for 3 min at 300 × g at 4 °C. Cells were lysed in 3× the pellet volume of NP40-NaCl lysis buffer (50 mM HEPES-KOH pH 7.5, 150 mM NaCl, 2 mM EDTA-NaOH pH 8, 0.5 % (v/v) NP40, 0.5 mM DTT) for 5 min on ice. Nuclei were pelleted by centrifugation for 5 min at 15,000 × g at 4 °C, and equal amounts of clarified lysate were separated by 12 % SDS-PAGE. The gel was Coomassie blue stained prior to imaging using a phosphorimager, and lane intensity was quantified using ImageQuant TL software.

Ribosome profiling and RNA-Sequencing.

Ribosome profiling libraries were constructed as previously described [49]. Briefly, cells were harvested in lysis buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, 100 ug mL⁻¹ cycloheximide, 1 % (v/v) Triton X-100, 25 U mL⁻¹ Turbo DNase1) and ribosome footprints were isolated by RNase 1 digestion and pelleting by centrifugation through a 1 M sucrose cushion at 100,000 rpm for 1 h at 4 °C in TLA-110. 17-34 nucleotide long footprints were size-selected using a 15 % polyacrylamide TBE–urea gel and footprints were dephosphorylated and ligated to a barcoded adenylated RNA linker. Samples were pooled, and cDNA was made by reverse transcription. cDNA products were circularized using CircLigase^™ ssDNA Ligase (Lucigen), contaminating rRNA sequences were removed using biotinylated rRNA depletion oligos [50], and sequencing libraries were constructed by PCR using primers to add Illumina adapters. For RNA-Sequencing libraries, total RNA was isolated using phenol-chloroform extraction and ethanol precipitation. Libraries were prepared and sequenced by Novogene using standard polyA-RNA Illumina sequencing protocol. Experiments were performed in duplicate.

Bioinformatics analysis.

For RNA-Sequencing pre-processing, adapters were trimmed using Cutadapt [51], and contaminating rRNA reads were removed using bowtie2 [52],[53]. For ribosome profiling pre-processing, adapters were trimmed and reads were demultiplexed using Cutadapt. Reads were collapsed with FASTX-Toolkit [54] and UMI were removed with Cutadapt. Contaminating rRNA reads were removed using bowtie2. Reads were mapped to the human transcriptome (GRCh38) using STAR [55] and differential gene analysis was performed using the DESeq2 package [56] and Babel package [57] in R Studio. Gene ontology analysis of RNA-Sequencing data was performed using DAVID [58],[59]. Mapped reads were sorted and indexed using samtools [60], bigwig files were created using DeepTools [61], and reads were visualized using the Integrated Genomics Viewer (Broad Institute) [62].

Supplementary Material

SI Table 1

Table 1. Table of changes in translation efficiency upon knockout of eIF5A2 in SW480 cells. The baseMean indicates the average of the normalized count values, dividing by size factors, taken over all samples. The log2FoldChange is the Fold change between the comparison and control group, reported on a logarithmic base 2 scale. The lfcSE is the standard error estimate for the fold change estimate. The change type column indicates if both the ribosome footprint and RNA-sequencing levels are changed (“both”), or only the ribosome footprint levels are altered (“translational_only”).

SI Table 2

Table 2. Table of changes in transcript levels upon knockout of eIF5A2 in SW480 cells.

3

Figure S1. Analysis of gene expression changes in eIF5A2 knockout clonal cell lines. (A) Fold change in transcript expression levels of eIF5A2^KO versus WT SW480 cells as measured by quantitative RT-PCR of upregulated genes (GPX2 and ROBO1) or non-target genes (PGK1 and c-Myc) as identified by RNA-Sequencing. (B) Representative western blot of WT and two different clonal eIF5A2^KO SW480 cell lines, showing decreased expression of target genes identified by RNA-Sequencing. Hsp90 is a loading control. (C) Fold change in transcript expression levels in two different clonal eIF5A2^KO cell lines versus WT SW480 cells as measured by quantitative RT-PCR.

Figure S2. Analysis of cell proliferation and gene expression changes in SW480 cells upon eIF5A1 knockdown. (A) MTT assay measuring cell proliferation in SW480 cells transduced with a non-targeting or eIF5A1-targeting shRNA. The results are given as the mean ± standard deviation (s.d.) of at least three independent experiments. The P value is calculated by unpaired t-test. **P<0.005. (B) Representative western blot of SW480 cells transduced with a non-targeting or eIF5A1-targeting shRNA, showing that decreased expression of eIF5A1 does not affect eIF5A2 target genes. Hsp90 is a loading control.