Hypusinated eIF5A is required for the translation of collagen

The translation elongation factor eIF5A, the eukaryotic homolog of prokaryotic elongation factor P (EF-P), is an essential and highly conserved protein. It is the only known protein modified by hypusination, which is required for eIF5A activity and occurs in two sequential enzymatic steps catalyzed by a deoxyhypusine synthase (DHPS) and a deoxyhypusine hydroxylase (DOHH), two enzymes that are also essential in most eukaryotic cells (Park and Wolff, 2018). In mammals, eIF5A is encoded by the paralogous genes Eif5a1 and Eif5a2, for which amino acid sequences of the corresponding protein isoforms show high identity. Overexpression of each human isoform has been linked to different types of cancer, and high levels of eIF5A2 enhance metastasis (Mathews and Hershey, 2015; Nakanishi and Cleveland, 2016; Ning et al., 2020).

Hypusinated eIF5A binds to the ribosome in the E-tRNA site, where it interacts with the P-tRNA to promote productive positioning for peptide bond formation (Dever et al., 2018). Experiments with bacterial EF-P (Doerfel et al., 2013; Ude et al., 2013) and Saccharomyces cerevisiae (Gutierrez et al., 2013; Li et al., 2014) show that eIF5A facilitates translation through polyproline sequences containing three or more consecutive prolines, and this function seems to be conserved through evolution (Muñoz-Soriano et al., 2017). Ribosome profiling (Schuller et al., 2017) and 5PSeq assays (Pelechano and Alepuz, 2017) have shown that eIF5A also alleviates ribosome pauses at tripeptide combinations of proline, glycine and charged amino acids. A search in the human proteome identified that the extracellular matrix (ECM) compartment is highly enriched in potential eIF5A targets (Pelechano and Alepuz, 2017).

Collagen is the most abundant protein in vertebrates, constituting more than 25% of human body weight. Collagens are essential in the ECM and function in tissue structure, development and remodeling, cell adhesion and migration, cancer and angiogenesis. The collagen family, which comprises around 28 different types in mammals, has characteristic triple-helical folding formed by three collagen α chains. In mammals, the more than 40 distinct α chains contain abundant repetitions on the so-called collagenic motif (X-Y-Gly), with proline or hydroxylated proline frequently in the first (X) and second (Y) positions and glycine in the third, which enables the formation of the collagenous triple helix (Brodsky and Persikov, 2005; Kadler et al., 2007; Myllyharju and Kivirikko, 2001).

Collagen α chains are co-translationally inserted into the endoplasmic reticulum (ER) and are thus targeted to the secretory pathway. In the ER, the correct folding of the α chains to form procollagen requires the action of co- and post-translational modification enzymes, such as hydroxylases and glycosyltransferases, and the assistance of several molecular chaperones (Ito and Nagata, 2019). Secretion of the collagen triple helix involves the packaging and formation of big COPII vesicles (Malhotra and Erlmann, 2015). Incorrect folding or assembling of collagen results in the intracellular retention of immature procollagen and the induction of ER stress and the unfolded protein response (UPR) (Wong and Shoulders, 2019).

Deficient production of mature collagen can lead to severe diseases, collectively called collagenopathies (Arseni et al., 2018; Jobling et al., 2014; Myllyharju and Kivirikko, 2001), and to defects in wound healing when injured skin requires a physiological increase in collagen production (Nyström and Bruckner-Tuderman, 2019). By contrast, excessive and uncontrolled synthesis of ECM proteins, especially collagen, results in fibrosis. Collagen I represents 80–90% of the proteins in the fibrotic matrix and disrupts normal tissue architecture and function. Fibrotic diseases, primarily liver fibrosis, followed by cardiac, renal and pulmonary fibrosis, present a major health problem because of a lack of effective curative treatments, and they are a leading cause of morbimortality (Ricard-Blum et al., 2018; Weiskirchen et al., 2019).

In the present study, we tested the hypothesis that the translation elongation factor eIF5A is necessary for collagen synthesis. First, we used yeast cells carrying temperature-sensitive eIF5A to determine levels of heterologously expressed collagen type I α1 chain (Col1a1) and translation stops at collagenic tripeptide motifs cloned in a dual-luciferase system. Second, we used cultured mouse fibroblasts and depleted active eIF5A, either through treatment with the DHPS inhibitor N¹-guanyl-1,7-diaminoheptane (GC7) or with siRNA against Dhps and Eif5a1, to evaluate the expression of Col1a1 and to localize it. We also explored the induction of ER stress upon depletion of active eIF5A. Third, we used TGF-β1 as a profibrotic stimulus in cultured human hepatic stellate cells (HSCs) to assess the extent to which Col1a1 overproduction depends on eIF5A expression. In yeast, our data suggest that ribosomes stall when encountering collagenic motifs in eIF5A-depleted cells. Our results with mammalian cells show that active eIF5A is required to maintain Col1a1 homeostasis and to increase its production during fibrotic processes. We propose that eIF5A could be a therapeutic target for regulating collagen production in fibrotic diseases.

Our previous study showed that the gene ontology terms ‘ECM organization’ and ‘collagen metabolism’ were significantly enriched in the human proteins with a high content (>25) of eIF5A-dependent motifs (Pelechano and Alepuz, 2017). To determine whether only collagens or also other proteins of the mammalian ECM compartment contain a significant number of eIF5A-dependent tripeptides, and to identify which ones, we analyzed the amino acid sequence of the Mus musculus ECM proteins. ECM proteins are classified as fibrous proteins, proteoglycans, glycoproteins and other proteins (Frantz et al., 2010), and we further subdivided fibrous proteins into collagens and others (Fig. 1A). We examined the abundance of the 43 tripeptide motifs with the highest score for eIF5A-dependent ribosome pausing (Pelechano and Alepuz, 2017) and found that, although the average distribution of eIF5A-dependent motifs in mouse ECM proteins was 22.6 motifs per protein, collagens had a much higher average frequency (86.0 motifs per protein) than the rest of the ECM protein groups (motifs per protein: 21.7 in non-collagen fibrous proteins, 9.4 in proteoglycans, 13.0 in other proteoglycans and 6.8 in others) (Fig. 1B). Collagens also contained different eIF5A-dependent motifs from the rest of the ECM proteins, such as the non-polyproline motifs PGP, PPG, EPG and DPG that constitute 78% of the motifs, whereas the well-known PPP (Dever et al., 2018) motif was almost absent in collagen sequences (Fig. 1B,C). Those four non-polyproline eIF5A motifs appear in stretches in the collagenic regions of the polypeptide sequences that form the triple helix of the collagen proteins (Kadler et al., 2007). Our analysis points to mammalian collagens as important targets of eIF5A during their translation.

To obtain molecular evidence supporting the role of eIF5A in the translation of collagenic sequences carrying putative elF5A-dependent motifs, we followed two different approaches using S. cerevisiae cells. Human and yeast eIF5A proteins share >60% homology and are functionally interchangeable (Schnier et al., 1991; Schwelberger et al., 1993). Yeast cells do not contain collagen, but heterologous expression of human collagen has been successfully achieved (Brodsky and Ramshaw, 2017). First, using a tetracycline-regulated (tetO₇) expression system (Garí et al., 1997) to avoid the deleterious effects of constant collagen expression in yeast cells, we expressed the first 420 amino acids of mouse Col1a1, containing 15 PPG, 15 PGP and four EPG motifs (Fig. 2A; Fig. S1A), fused to the β-galactosidase reporter. As a positive control of eIF5A-translation dependency, we constructed another fusion with a fragment of the yeast Bni1 protein containing three documented eIF5A-dependent polyproline sequences (Li et al., 2014) (Fig. 2A; Fig. S1A). We analyzed the expression of non-fused lacZ, Col1a1-lacZ and Bni1-lacZ after inducing expression by incubating for 6 h in tetracycline-free media in wild-type yeast cells and eIF5A temperature-sensitive mutant cells (tif51A-1) carrying a single point mutation (Pro83 to Ser) (Li et al., 2011). β-Galactosidase activity corresponding to Col1a1-lacZ and Bni1-lacZ was similar in wild-type and mutant cells when incubated at a permissive temperature (25°C) (Fig. 2A). However, when incubated for 6 h at a restrictive temperature (37°C) (Li et al., 2014), Col1a1-lacZ expression was reduced by ∼40% in the mutant with respect to the wild type; a stronger reduction (∼80%) was observed in Bni1-lacZ expression (Fig. 2A). We did not observe any difference in the expression of non-fused lacZ between wild-type and mutant cells at 37°C (Fig. 2A). These results indicate that eIF5A is required for the expression of the cloned Col1a1 fragment, and confirm that it is a strong requirement for the translation of the polyproline motifs of Bni1.

To further define the role of eIF5A in the translation of specific collagenic motifs, we used a second approach based on the use of the dual-luciferase reporter system developed in yeast cells (Letzring et al., 2010). Dual-luciferase plasmids express a single mRNA that contains the Renilla luciferase at the 5′ end and the firefly luciferase open reading frames (ORFs) at the 3′ end, joined in-frame by a short sequence from which the motifs to study can be cloned (Fig. 2B; Fig. S1B). In the dual-luciferase assay, a premature aberrant translation termination caused by the inserted motif, but without a stop codon, will result in a Renilla polypeptide without the C-terminal part containing the firefly luciferase. Although the synthesis of a truncated polypeptide may affect its stability, the detection of reduced firefly:Renilla luciferase ratio is indicative of motif-induced translation termination. For these analyses, we used a plasmid with no insertions; a plasmid with ten consecutive optimal codons for proline inserted (10P) (Letzring et al., 2010); and four plasmids we constructed with insertions of three non-consecutive repetitions of optimal codons for the polyproline motif PPP (3PPP), for the collagenic motifs PGP (3PGP) and EPG (3EPG), and for the control motif TQA (3TQA), for which ribosome pausing is not predicted upon eIF5A depletion (Pelechano and Alepuz, 2017) (Fig. 2B; Fig. S1B). We also constructed a dual-luciferase plasmid in which we inserted a short stretch of mouse collagen IV sequence (Col4a1) containing several PPG and PGP motifs (Fig. 2B; Fig. S1B). None of the resulting combinations of three amino acids in the inserted sequences, except the motif under investigation, were predicted to induce ribosome pausing by eIF5A depletion (Fig. S1C) (Pelechano and Alepuz, 2017). Dual-luciferase reporter constructs were introduced into isogenic strains expressing wild-type eIF5A or temperature-sensitive eIF5A-S149P and grown at a semi-permissive temperature (33°C) as described (Gutierrez et al., 2013). We hypothesized that if eIF5A promoted translation of the inserted motifs in the dual-luciferase plasmid, then the firefly:Renilla luciferase ratio would be lower in the eIF5A-S149P mutant strain grown at 33°C. As shown in Fig. 2B, the firefly:Renilla luciferase ratio for each construct was different depending on the cloned motif. The empty plasmid and the control TQA motif showed no difference in the firefly:Renilla luciferase ratio between the wild type and eIF5A mutant. As previously reported (Gutierrez et al., 2013), a very low ratio was observed for the 10P construct upon eIF5A depletion (Fig. 2B). Low ratios in the eIF5A mutant were also obtained with 3PPP, indicating that three non-consecutive polyproline tripeptide motifs are sufficient to impose a requirement for eIF5A, although without reaching the strong reduction observed with the longer polyproline motif (10P). Similarly, insertion of 3PGP and 3EPG and the Col4a1 fragment resulted in a lower luciferase ratio in the eIF5A mutant than in the wild type (Fig. 2B). These results confirm that eIF5A promotes the translation of the collagenic tripeptide motifs.

To investigate the role of eIF5A in collagen synthesis in mammalian cells, we used a mouse fibroblast line. Fibroblasts are prototypical collagen producer cells, and we focused on collagen 1 (Col1) because it is the most abundant collagen in animal tissue (Myllyharju and Kivirikko, 2001). Col1 is a heterotrimer with two α1 chains and one α2 chain. The Col1a1 subunit contains 50 PGP, 45 PPG, ten EPG and one DPG motif (Fig. 1C). Using these mouse fibroblasts, we investigated the effect of inhibiting eIF5A hypusination with GC7, which acts as a competitive inhibitor of DHPS (Park and Wolff, 2018), and quantified Col1a1 levels. As seen in Fig. 3A and B, treatment of fibroblasts with 30 µM or 65 µM GC7 reduced eIF5A hypusination and concomitantly lowered Col1a1 levels. A reduction in Col1a1 was already evident at 24 h of GC7 treatment and was still visible at 96 h. Importantly, treatments with GC7 did not modify total eIF5A content, except for a slight reduction at 72 h (Fig. 3A; Fig. S2A). Col1a1 synthesis is mainly regulated at the level of mRNA stability and translation (Zhang and Stefanovic, 2016). To investigate whether the effects of eIF5A depletion were attributable to a negative effect on Col1a1 translation rather than transcription, we quantified mRNA levels after treatment with 30 µM GC7. No significant differences were observed in Col1a1 mRNA levels with GC7 (Fig. S2B), suggesting that the effect of eIF5A inhibition on Col1a1 occurs at the level of translation. Because eIF5A is an essential protein in eukaryotic cells, we checked whether GC7 treatment reduced fibroblast viability. However, treatment with 30 µM GC7 up to 96 h did not significantly reduce fibroblast viability when compared to control (untreated) cells (Fig. S2C). Low cytotoxicity at 30 µM GC7 has also been described for human cells in culture (Xu et al., 2014); therefore, we used this GC7 concentration in the following experiments.

Because off-target effects have been described for GC7 (Oliverio et al., 2014), as an alternative to using GC7 to deplete functional eIF5A, we transfected mouse fibroblasts with Dhps siRNA to reduce the first enzymatic step of eIF5A hypusination. After 72 h, we observed a >50% reduction in Dhps mRNA levels and a similar reduction in hypusinated eIF5A (Fig. 3C). This depletion of functional eIF5A correlated with a <50% reduction in Col1a1 protein levels, whereas no reduction in Col1a1 mRNA was observed (Fig. 3B). We then transfected mouse fibroblasts with Eif5a1 siRNA to deplete Eif5a1 mRNA, which is highly expressed in mammalian cells (Park and Wolff, 2018). The reduction in Eif5a1 mRNA (∼50% at 72 h of transfection) was paralleled by a strong reduction in Col1a1 protein (∼70% compared to the control with scrambled siRNA) with, again, no change in Col1a1 mRNA level (Fig. S3).

Together, these results show that hypusinated eIF5A is necessary to maintain high levels of Col1a1 protein in mouse fibroblasts and suggest a role for eIF5A in the translation of Col1a1 mRNA.

Translation of collagen polypeptides occurs in the ER, where the polypeptide chains are translationally and post-translationally modified to form the triple helix (Malhotra and Erlmann, 2015; Sharma et al., 2017; Zhang and Stefanovic, 2016). To examine the effects of hypusinated eIF5A depletion on collagen synthesis, we analyzed Col1a1 distribution in mouse fibroblasts treated with 30 µM GC7. Immunostaining showed a lower Col1a1 signal in the treated fibroblasts than in untreated cells (Fig. 4A,B). Moreover, in the control cells, the Col1a1 signal was visualized as dots concentrated more in the perinuclear region but also distributed all over the cytoplasm. By contrast, although the GC7-treated fibroblasts showed a concentration of Col1a1-stained dots in the perinuclear region, suggesting localization in the ER, the signal was almost absent in the rest of the cytoplasm (Fig. 4C). Similar results of Col1a1 signal intensity reduction and predominant perinuclear localization were observed by depleting functional eIF5A in mouse fibroblasts with Dhps siRNA (Fig. 4D–F).

In order to clearly assign an ER localization to the Col1a1 dots observed in functional eIF5A-depleted fibroblasts, we co-visualized Col1a1 with Hsp47 (also known as SERPINH1), the collagen-specific chaperone that aids during collagen folding and maturation (Ito and Nagata, 2019). In control and GC7-treated fibroblasts, a similar Hsp47 perinuclear signal was observed (Fig. S4). However, all Col1a1 signal co-localized with Hsp47 signal in GC7-treated cells, whereas in control cells Col1a1 signal was distributed throughout the fibroblasts cells and part of the signal did not co-localize with Hsp47 signal. Again, these data suggest that Col1a1 is retained at the fibroblast ER upon functional eIF5A depletion. Moreover, we also investigated whether lack of functional eIF5A provokes a general defect in ER and the Golgi secretory pathway because a correlation between the function of eIF5A in translation and secretion has been described in yeast (Rossi et al., 2014). With this aim, we visualized nidogen-1, which is a major component of basement membranes and ECMs (Mao et al., 2020) and lacks eIF5A-dependent motifs. We observed no changes in nidogen-1 signal intensity and localization in GC7-treated fibroblast cells (Fig. S5). This result supports the specific effect of eIF5A on collagen translation against a more general effect on ER translation and secretion, without excluding the possible specific effect of eIF5A in the ER-coupled translation of other proteins containing eIF5A-dependent motifs.

It is well documented that misfolding and accumulation of mutated collagen in the ER lead to ER stress and the induction of collagenopathies (Gawron, 2016). Thus, if the lack of functional eIF5A is hindering the translation of the Col1a1 collagenic segments, it would stall translating ribosomes at the ER membrane and trigger the ER stress response. To test this, we analyzed the expression of several factors induced during the UPR that are activated in response to ER stress: the chaperone GRP78/BiP (also known as HSPA5), a sensor of unfolded proteins in the ER; the ER stress transducer ATF6, a transcription activator of genes involved in protein folding, secretion and degradation; and CHOP (also known as DDIT3), a proapoptotic transcription factor induced upon severe ER stress (Hetz et al., 2020; Oyadomari and Mori, 2004; Ron and Walter, 2007). A quick and transient induction of BiP, Atf6 and Chop mRNA levels was observed in fibroblasts after 24–48 h of GC7 treatment (Fig. 5A). Correspondingly, CHOP protein levels increased ∼100-fold at the same time in GC7-treated fibroblasts (Fig. 5B). Moreover, there was intense nuclear localization of CHOP protein in fibroblasts 24 h after GC7 treatment (Fig. 5C,D); this correlates with its previously described change from cytoplasmic to nuclear localization during stress (Oyadomari and Mori, 2004; Ron and Walter, 2007).

We confirmed in yeast that depletion of eIF5A was sufficient to induce the ER stress response (Fig. S6), in agreement with previous results (Rossi et al., 2014). Induced mRNA levels of ER stress markers (SSB2, SSE1, HSP31) were observed in the temperature-sensitive eIF5A mutant at restrictive temperature, whether expressing the Col1a1-lacZ construct or not. Moreover, we observed that expression of the Col1a1-lacZ construct in wild-type yeast cells also induced the ER stress response, probably due to the inability to obtain proper folding of the Col1a1 fragment fused to lacZ (Fig. S6).

In sum, these results suggest that partially synthesized Col1a1 is retained at the ER in cells with a diminished amount of hypusinated eIF5A, causing a reduction in procollagen type I export from the ER to the Golgi complex. Either because there is an accumulation of translationally paused ribosomes at the ER or because partially synthesized Col1a1 peptides expose hydrophobic domains that may deplete components of the quality control system (Gawron, 2016), the UPR is induced in functional eIF5A-depleted cells. Additionally, a reduction in eIF5A may cause a defect in the ER-coupled translation of other proteins containing eIF5A-dependent motifs in yeast and mammalian cells.

Hepatic fibrosis is the most common fibrotic process, in which HSCs are the major cellular type responsible for producing high levels of collagen. HSCs in normal liver are quiescent and synthesize trace amounts of Col1, but an increase in the major profibrogenic inducer TGF-β1 (Meng et al., 2016) results in the transdifferentiation of HSCs into myofibroblast-like cells, which acquire a strong contractile phenotype and secrete and remodel large amounts of Col1 (De Minicis et al., 2007).

To assess the effect of eIF5A hypusination inhibition on Col1a1 production during hepatic fibrogenesis in vitro, LX2 cells, an immortalized cell line of human HSCs, were treated with TGF-β1 (2.5 ng/ml) for 48 h, or with its vehicle (DMSO), with or without 30 µM GC7. We observed an ∼90% reduction in hypusinated eIF5A in GC7-treated cells compared to the corresponding controls of HSCs, treated with or without TGF-β1, while total eIF5A protein levels were not significantly changed (Fig. 6A,B). The level of hypusinated eIF5A was higher in TGF-β1-treated than in control cells, but the difference was not statistically significant. Importantly, Col1a1 protein was undetectable in GC7-treated HSCs under basal conditions. As expected, a huge increase in Col1a1 protein levels was observed in TGF-β1-treated cells, which was fully prevented by co-treatment with GC7 (Fig. 6A,C). We also analyzed COL1A1 mRNA levels under these conditions to distinguish between transcriptional and post-transcriptional effects. A small increase in COL1A1 mRNA levels was observed in TGF-β1-treated cells (Fig. 6C), in correlation with the profibrotic activation of HSCs, although most of the huge increase in Col1a1 protein level with TGF-β1 can be attributed to post-transcriptional effects. In GC7-treated cells, we observed a slight reduction in COL1A1 mRNA, with and without TGF-β1 treatment, that does not justify the large drop in Col1a1 protein levels when active eIF5A is lacking (Fig. 6A,C). These results strongly suggest that eIF5A is required for Col1a1 synthesis during TGF-β1-mediated hepatic fibrogenesis.

Additionally, we investigated further the effect of GC7 on the TGF-β transdifferentiation process in HSCs. This process involves the reprogramming of transcription with upregulation of markers such as α-smooth muscle actin (α-SMA, also known as ACTA2) (De Minicis et al., 2007) and also downregulation of peroxisome proliferator-activated receptor-gamma (PPAR-γ) in TGF-β-activated HSCs (De Minicis et al., 2007; Hazra et al., 2004). As expected, we observed an increase in α-SMA mRNA and decrease in PPARG mRNA expression in TGF-β-treated HSCs (Fig. 6D). Interestingly, GC7-treated HSCs showed the opposite pattern, with significant reduction of α-SMA mRNA and increase in PPARG mRNA, in cells incubated with or without TGF-β (Fig. 6D). These results suggest that inhibition of eIF5A hypusination yields antifibrotic effects as it not only reduces collagen synthesis but also inhibits the profibrotic transdifferentiation of HSCs.

Expression of collagen, the most abundant protein in animals, is regulated at the level of transcription, mRNA stability and translation (Lindquist et al., 2000). Our results here, using different approaches, provide evidence that eIF5A is essential for the translation of tripeptide motifs that are heavily present in collagens. In yeast, fusions of mouse Col1a1 fragments with β-galactosidase show low expression levels upon eIF5A depletion. In a dual-luciferase reporter system, PGP, PPG and EPG non-polyproline tripeptide motifs, abundant in collagens, stopped translation between Renilla and firefly luciferases when eIF5A was lacking. In mouse fibroblasts, depleting functional eIF5A by treatment with GC7 or by interfering with the Eif5a1 or Dhps mRNAs results in reduced Col1a1 levels, with no reduction in the Col1a1 mRNA. In human HSCs, inhibiting eIF5A hypusination with GC7 eliminates the Col1a1 overproduction caused by profibrotic treatment with TGF-β1. Together, these results link eIF5A activity with translation of the PGP, PPG and EPG motifs that are highly abundant in the Col1a1 collagenic regions. The need for eIF5A in translating these non-polyproline tripeptide motifs has been suggested in yeast by ribosome profiling (Schuller et al., 2017) and 5PSeq assays (Pelechano and Alepuz, 2017), in which it was noted that eIF5A depletion resulted in ribosomes stalling in these motifs. However, no physiological confirmation of a stop to translation causing the deficient protein synthesis in these motifs has been shown until now. Although our work has focused on the study of Col1a1, the existence of similar collagenic stretches in all collagens (Fig. 1) suggests a similar dependency on eIF5A for the translation of the different collagens. Supporting this, a short polypeptide stretch of mouse Col4a1 containing several PPG and PGP motifs clearly stops translation in a yeast dual-luciferase system (Fig. 2; Fig. S1).

We observed that inhibition of eIF5A hypusination led to an accumulation of Col1a1 around the nuclei of mouse fibroblasts and provoked ER stress. Mutations in collagen, collagen-maturation enzymes or proteins that inhibit proper collagen folding, packing and secretion to induce ER stress and intracellular retention of collagen have been observed in collagenopathies, such as skeletal chondrodysplasia, osteogenesis imperfecta and osteoarthritis (Gawron, 2016). Intracellular retention of procollagen induces the ER stress proteins BiP and CHOP, which may lead to apoptosis (Schulz et al., 2016). eIF5A has also been implicated in ER function. In yeast and HeLa cells, depletion of eIF5A causes ER stress and upregulates stress-induced chaperones (Mandal et al., 2016). We propose a direct role of eIF5A in facilitating ER-coupled collagen translation; similarly, eIF5A may facilitate the co-translational translocation into the ER of other proteins containing eIF5A-dependent motifs. Although most collagenopathies have a genetic etiology, with a mutation in a collagen or collagen-metabolism protein (Forlino and Marini, 2016; Jobling et al., 2014; Marini et al., 2017), it would be of interest to determine whether patients with a clinical, but not a genetic, diagnosis have defects in eIF5A expression. Interestingly, recent reports have linked impaired eIF5A function caused by the presence of human EIF5A or DHSP genetic variants with developmental and neurological rare disorders (Faundes et al., 2021; Ganapathi et al., 2019).

The excessive and uncontrolled synthesis of ECM proteins, mainly collagen type I, is the hallmark of fibrotic diseases; thus, the severity of the fibrotic disease depends on the amount of Col1 produced (Zhang and Stefanovic, 2016). Nearly 45% of all deaths in the developed world are attributed to chronic fibroproliferative diseases. Therefore, the demand for effective antifibrotic drugs will likely continue to increase in the coming years (Wynn, 2008). In our study, using an in vitro model of fibrogenesis with human HSCs, TGF-β1 induced the expected huge increase in Col1a1 levels, which was almost completely abolished upon GC7 treatment. Additionally, we observed that GC7 inhibits the profibrotic transdifferentiation expression program of TGF-β1-activated HSCs. In vivo studies are required to confirm the profibrotic role of eIF5A and test whether downregulation of eIF5A may be beneficial for the treatment of fibrosis of the liver and other organs.

S. cerevisiae strains used are listed in Table S1. Yeast cells were grown in YPD (2% glucose, 2% peptone, 1% yeast extract) or synthetic complete (SC) media lacking the indicated amino acid [2% glucose, 0.7% yeast nitrogen base (YNB) and required Drop-Out percentage]. For experiments, cells were grown exponentially until they reached the required optical density at a wavelength of 600 nm (OD₆₀₀) at 25°C, when they were transferred to a semi-permissive (33°C) or restrictive (37°C) temperature.

Fusions with lacZ and under the control of a tetracycline-repressible operon (tetO₇) were constructed in the plasmid pCM179 (Garí et al., 1997). Dual-luciferase reporter constructs were generated by cloning between the in-frame Renilla and firefly luciferase ORFs in the plasmid pDL202 (Letzring et al., 2010). Inserts were generated by PCR amplification using M. musculus cDNA or S. cerevisiae genomic DNA. pDL202 and 10P plasmids were kindly supplied by Elizabeth Grayhack (University of Rochester, Rochester, NY) (Letzring et al., 2010). Oligonucleotides used are listed in Table S2. The GAP-REPAIR homologous recombination method was used for cloning (Joska et al., 2014).

Wild-type eIF5A (BY4741) or temperature-sensitive eIF5A-P83S (tif51A-1) containing pCM179 and derivative plasmids were cultured in SC-URA medium at 25°C with 2 μg/ml doxycycline to keep the tetO₇ promoter switched off. At an OD₆₀₀ of 0.2, the cell culture was washed with medium lacking doxycycline, resuspended in fresh SC-URA medium to activate tetO₇ transcription, and incubated at 25°C or 37°C for 6 h. Then, β-galactosidase assays were performed as described previously (Garre et al., 2012).

The dual-luciferase assay with yeast strains containing wild-type eIF5A (J697) or temperature-sensitive eIF5A-S149P (J699) were performed as described previously (Gutierrez et al., 2013) with some modifications. Yeast was cultured in SC-URA medium at 33°C to an OD₆₀₀ of 0.8, harvested and frozen. Subsequently, cell pellets were resuspended in 200 µl lysis buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 5% NP-40, Complete Protease Inhibitor Cocktail (Roche)] and mixed with one volume of glass beads for further homogenization in a Precellys tissue homogenizer (Bertin Corp.). Lysates were cleared by centrifugation at 19,515 g for 5 min at 4°C, and protein extracts were assayed for firefly and Renilla luciferase activity sequentially in a 96-well luminometry plate (Thermo Fisher Scientific) using a microplate luminometer (Thermo Fisher Scientific) and the Dual-Glo Luciferase Assay System (Promega). Finally, the firefly:Renilla luciferase ratio for each construct was calculated.

The fibroblast cell line was isolated from mouse kidney and immortalized by retroviral delivery of the SV40 large T (Benito-Jardon et al., 2017). These cells were cultured in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal calf serum (FCS; Thermo Fisher Scientific) and 1% penicillin–streptomycin (Gibco), and maintained at standard conditions of 37°C and 5% CO₂ in a humidified atmosphere. Cells were cultured with or without treatment with the DHPS inhibitor GC7 (Calbiochem) at the indicated concentration to deplete functional eIF5A (Park et al., 1994).

LX2 (immortalized cell line of HSCs) cells were gifted by Dr Scott L. Friedman, Icahn School of Medicine at Mount Sinai, New York, NY. Cells were cultured in DMEM with high glucose (Sigma-Aldrich), supplemented with 10% FCS, 2 mM L-glutamine, 50 U/ml penicillin and 50 μg/ml streptomycin (Gibco, Invitrogen) in a humidified atmosphere with 5% CO₂ at 37°C. For the experiment, cells were seeded in a six-well plate (0.18×10⁶ cells/well) 24 h before treatment and treated for 48 h with TGF-β1 (2.5 ng/ml) or its vehicle DMSO (Sigma-Aldrich). GC7 was used at 30 µM alone or together with TGF-β1.

Non-treated cells and 30 µM GC7-treated cells were grown in DMEM, and samples were taken at 1, 2, 3, 4, 5 and 7 days. Cells from suspension, PBS washes and the fibroblast monolayer detached with trypsin were collected for analysis. A 4% Trypan Blue/PBS dilution was used to stain dead cells. The fibroblast population was analyzed by counting live and dead cells in a Neubauer chamber. The cell death percentage was calculated from the 50–400 cells counted.

A synthetic pool of siRNAs was used to target and knock down DHPS (Sigma-Aldrich, EMU150671) or eIF5A (Santa Cruz Biotechnology, sc-40560) expression according to the manufacturers’ protocols. Briefly, 3×10⁴ cells/well were seeded in a six-well culture plate 1 day prior to transfection until 70–90% confluence. For Dhps silencing, 250 µl Opti-Mem I Reduced Serum Medium (Thermo Fisher Scientific, 31985062) containing 7 µl Lipofectamine 3000 reagent (Thermo Fisher Scientific, L3000015) and 10 µl of either Dhps or control siRNA (Sigma-Aldrich, SIC001), previously incubated together for 15 min, was added to the culture. Cells were transfected and incubated for 72 h. For eIF5A silencing, the culture medium was replaced with transfection medium (Santa Cruz Biotechnology, sc-36868) containing 6 µl transfection reagent (Santa Cruz Biotechnology, sc-29528) and 12 µl of either Eif5a or control siRNA (Santa Cruz Biotechnology, sc-36869), previously incubated together for 30 min. The cells were transfected and incubated with the transfection mixture for 6 h, after which the transfection medium was replaced with 2× fibroblast usual medium. After 24 h, the medium was replaced with 1× usual medium and incubated until 72 h post-transfection. Finally, cells were collected for quantitative reverse transcription PCR (RT-qPCR) and western blot analysis.

For western blot analysis of fibroblast protein content, proteins were extracted in RIPA buffer [150 mM NaCl, 0.05 M Tris-HCl pH 8.0, 0.005 mM EDTA, 0.1% SDS, 0.5% sodium deoxycholate, 1% NP-40, cOmplete Protease Inhibitor Cocktail (Roche)] for 15 min. The lysates were centrifuged at 22,000 g at 4°C for 5 min to remove the cell debris and insoluble proteins. Protein content was quantified in a Nanodrop device (Thermo Fisher Scientific), and 100 µg of each sample was loaded into 4–20% gradient precast SDS-PAGE gels (Bio-Rad).

To analyze protein expression in LX2 cells, whole-cell protein extracts were obtained by lysis of the cell pellets in complete lysis buffer [20 mM HEPES pH 7.4, 400 mM NaCl, 20% (v/v) glycerol, 0.1 mM EDTA, 10 μM Na₂MoO₄, 10 mM NaF]. Immediately prior to their use, 1 mM DTT, 5 mM broad-spectrum serine, cysteine protease inhibitors (Complete Mini™ and Pefabloc^®, Roche) and 0.05% detergent solution (NP-40 Surfact-Amps™, Thermo Fisher Scientific) were added. Samples were then vortexed twice at maximum speed (10 s), incubated on ice (15 min), vortexed again at maximum speed (30 s) and centrifuged (16,000 g, 15 min, 4°C).

Western blot membranes were incubated with primary antibodies against Col1 (1:500, rabbit polyclonal, Merck, AB765P for fibroblasts and 1:1000, Cell Signaling Technology, 84336 for LX2 cells), hypusinated eIF5A (1:600, FabHpu antibody, Genentech), total eIF5A (1:500, rabbit monoclonal anti-eIF5A antibody, Abcam, b32407), Gadd153 (1:200, anti-CHOP antibody, Santa Cruz Biotechnology, sc-7351), α-tubulin (1:5000, Proteintech, 66031-1-Ig) or GAPDH (1:5000, rabbit polyclonal, Invitrogen, PA1-987 for fibroblasts and 1:10,000, mouse monoclonal, Sigma-Aldrich, G8795 for LX2 cells). Appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies were used (1:10,000, Promega). Chemiluminescent signals were detected and digitally analyzed, and normalized against α-tubulin or GAPDH.

To analyze mRNA levels, total RNAs were isolated from fibroblasts using a PureLink RNA Mini Kit (Invitrogen). The reverse transcription and quantitative PCR reactions were performed as detailed previously (Garre et al., 2013). Specific primers are listed in Table S2. Gapdh or ACT1 mRNA levels were used for normalization.

For indirect immunofluorescence of fibroblasts, cells (10³) were seeded onto laminin-coated glass coverslips and cultured for 1 or 4 days at 37°C in medium containing 30 µM GC7 or Dhps siRNA. Cells were fixed with 4% paraformaldehyde for 10 min, washed four times with PBS, permeabilized with 0.1% Triton X-100 in PBS for 10 min, and blocked for nonspecific protein binding with 3% bovine serum albumin (BSA)/PBS for 1 h at room temperature. Cells were incubated with primary antibodies against Col1 (1:250, Merck, AB765P), Gadd 153/CHOP (1:40, Santa Cruz Biotechnology, sc-7351), HSP47 (1:50, Santa Cruz Biotechnology, sc-5293) or nidogen-1 (1:500, anti-serum, a gift from Rupert Timpl, Max-Planck Institute, Munich, Germany) (in 1% BSA/PBS), and secondary antibodies Alexa Fluor 488-conjugated anti-rabbit (1:2000) or Alexa Fluor 546-conjugated anti-mouse (1:800) IgG. Rhodamine-conjugated phalloidin coupled with tetramethylrhodamine (TRITC; 1:500, Sigma-Aldrich) was used for actin filament counterstaining (in 1% BSA/PBS). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 1:10,000, Merck). Immunofluorescence images were acquired using an Olympus FLUOVIEW FV 1000 confocal microscope equipped with ×40 and ×60 objectives on a Nikon ECLIPSE E200. The same exposure times were used to acquire all images. Image analysis was carried out with ImageJ and Photoshop CC 2019.

Statistical evaluation was carried out with Microsoft Office Excel 2016. Data are expressed as the mean±s.d. Significant differences between the treated groups and the control were determined using Student's t-test (two-tailed, paired), at a significance level of P<0.05.

We thank Thomas Dever for his useful comments and suggestions. We are grateful to Irene Gimeno-Lluch for her initial input and to Sheila Otega-Sanchís for technical help with the fibroblast experiments. We are also thankful to all laboratory members for their comments and support.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258643