Eukaryotic initiation factor 6 modulates myofibroblast differentiation at transforming growth factor-β1 transcription level via H2A.Z occupancy and Sp1 recruitment

Eukaryotic initiation factor 6 (eIF6) is an evolutionary conserved protein, which functions in ribosomal biogenesis and translation control by regulating the binding of 40S and 60S ribosomal subunits (Ceci et al., 2003; Miluzio et al., 2009). Previous evidence has shown that eIF6 has a dual function: (1) nucleolar eIF6 is necessary for the biogenesis of 60S subunits in both yeast and mammals, and (2) cytoplasmic eIF6 is required for insulin and growth factor-stimulated translation in mammals (Brina et al., 2015). eIF6 is constitutively expressed in vitro, but its expression is variable in vivo. A high level of eIF6 expression has been found in rapidly proliferating cells (Donadini et al., 2001). Dysregulation of eIF6 expression has proved to be relevant to some pathological states. Ectopic expression of Drosophila eIF6 (DeIF6) in wing discs results in a Wg phenotype due to eIF6 selectively regulating Wnt signaling and β-catenin protein synthesis (Ji et al., 2008). The injection of eIF6 mRNA into one blastomere of two-cell-stage embryos resulted in a bent phenotype and fewer apoptotic cells during Xenopus laevis development, because eIF6 seemed to regulate the translation of specific mRNAs as an anti-apoptotic factor (De Marco et al., 2010). Moreover, eIF6 overexpression has been reported to cause an aberrant eye development in X. laevis (Miluzio et al., 2011). Furthermore, eIF6 overexpression has been found in selected cancer types, such as colorectal carcinomas (Sanvito et al., 2000), head and neck cancer (Rosso et al., 2004), and malignant mesothelioma (Biffo et al., 1997). It has been proposed that eIF6 is a rate-limiting controller of the initiation of translation, able to affect tumorigenesis and tumor growth (Gandin et al., 2008) and that the impairment of eIF6 activity in cytoplasm restricts lymphomagenesis and tumor progression (Miluzio et al., 2011). In addition, the physiological significance of eIF6 also impacts on inherited Shwachman Diamond syndrome (SDS). The deletion of chromosome 20 (q11.21q13.32) in the bone marrow of SDS patients exhibiting the benign prognosis is due to a gene/dosage effect of the eIF6 protein (Finch et al., 2011).

Recently, interest in the study of eIF6 has increased because this protein has been identified to interact with a neuronal protein 3.1 (P311; also known as NREP), promoting hypertrophic scar formation and contraction by enhancing the expression of transforming growth factor-β1 (TGF-β1) (Tan et al., 2010; Peng et al., 2012). The protein P311 was identified as the most intensely up-regulated gene in fibrosis-associated myofibroblasts (MFBs), as revealed by gene expression profiling (Wu et al., 2004) and comparative proteomic analysis (Ma et al., 2014). Myofibroblasts are key mediators of pathogenesis of all fibrotic diseases that significantly contribute to connective tissue remodeling by exerting traction forces and synthesizing extracellular matrix (ECM) components (Wynn and Ramalingam, 2012). Myofibroblasts are characterized by the expression of α-smooth muscle actin (α-SMA), which is a common key element for detection of myofibroblast differentiation and a major player in contractile force production. Profibrotic cytokine TGF-β1, along with the mechanical resistance of the extracellular matrix, is an amply documented primary stimulus for myofibroblast differentiation and persistence (Hinz et al., 2012). Thus, the data suggest that eIF6 might be relevant in TGF-β1 expression and myofibroblast differentiation, and which has yet to be examined.

In this study, our results show that eIF6 inversely modulates the expression of TGF-β1, mainly at transcription level. Our findings demonstrate that this regulatory effect is associated with eIF6 altering the occupancy of the TGF-β1 promoter by H2A.Z (Swiss-Prot P0C0S6) and Sp1, leading to altered transcription activity of TGF-β1. We also provide evidence that the TGF-β/Smad signaling pathway was involved in eIF6-regulated myofibroblast differentiation and activation. We also demonstrate that eIF6 is down-regulated in hypertrophic scar-derived fibroblasts, while ectopically overexpressed eIF6 in these cells could suppress their differentiation to myofibroblasts. These findings further support the hypothesis that eIF6 might play a relevant role during fibrogenesis. Thus, these data suggest a novel role of eIF6 as a regulator of TGF-β1 expression at transcription level in myofibroblast differentiation.

To investigate whether eIF6 plays a role in the expression TGF-β1, we measured the transcription and protein levels of TGF-β1 using eIF6 WT and eIF6^+/− fibroblasts. Using western blotting and quantitative polymerase chain reaction (qPCR) analysis, we confirmed that isolated fibroblasts from the skin of eIF6^+/− mice exhibited an approximately 50% reduction of eIF6 expression at both the protein and mRNA levels (Fig. 1A,B). Both TGF-β1 (Tgfb1) mRNA and TGF-β1 protein levels were found to be significantly increased in eIF6^+/− fibroblasts versus WT fibroblasts (mRNA: 2.7-fold increase, P<0.001, n=5; protein: 1.8-fold increase, P<0.001, n=5; Fig. 1C,D), as revealed by qPCR and enzyme-linked immunosorbent assay (ELISA) analyses. To verify these observations, we further expressed eIF6 ectopically in scar-derived fibroblasts (sFB) with a recombinant adenoviral eIF6-expressing vector [eIF6-enhanced green fluorescent protein (EGFP), and using EGFP as the control vector]. The ectopic overexpression of eIF6 was confirmed in sFB as shown in Fig. 1E and F. In contrast, eIF6 overexpression significantly decreased mRNA and protein levels of TGF-β1 in eIF6^+/− fibroblasts in comparison with WT cells (mRNA: 60% decrease, P=0.008, n=5; protein: 65% decrease, P=0.017, n=5; Fig. 1G,H). As expected, the level of TGF-β1 was inversely correlated with the level of eIF6, indicating that eIF6 modulates the expression of TGF-β1.

One objective of this study was to determine how TGF-β1 levels are increased when there is a deficiency of eIF6. To this end a polysomal profile was determined in order to evaluate the eIF6 effect on TGF-β1 translation as eIF6 is known as a crucial translational regulator. Contrary to what was initially assumed, TGF-β1 mRNA associated with heavier polysomes did not show significant differences between eIF6 knockdown cells and control cells, indicating that this effect might not be relevant to translation control (Fig. 2A and Fig. S1). To determine whether eIF6 regulated TGF-β1 mRNA at post-transcription level, we measured the level of TGF-β1 mRNA by qPCR followed by treatment with actinomycin D for various times to inhibit de novo RNA synthesis. The half-life of TGF-β1 mRNA was not altered by eIF6 knockdown in eIF6^+/− fibroblasts (Fig. 2B). Since the 5′ end of the TGF-β1 promoter region is very rich in G+C, we evaluated the DNA methylation patterns of the core region of TGF-β1 promoter in WT and eIF6^+/− fibroblasts via bisulfite sequencing. However, the methylation level of the TGF-β1 promoter region was similar for WT fibroblasts (61.9%) and eIF6^+/− fibroblasts (73.8%; Fig. 2C). To investigate whether eIF6 could regulate TGF-β1 promoter activity, HEK293 cells were co-transfected with expression vectors for eIF6 (pCMV-eIF6 plasmid; Fig. S2) and TGF-β1 promoter constructs in pGL3basic and analyzed by luciferase reporter gene assays. As shown in Fig. 2D, eIF6 overexpression slightly inhibited the TGF-β1 promoter-mediated luciferase activity (Fig. 2D; 28% reduction for phTGF1; 21% reduction for phTGF2; 19% reduction for phTGF3; 36% reduction for phTGF4). However, this partial regulatory effect of eIF6 on TGF-β1 promoter activity could not explain the profound effect of eIF6 on TGF-β1 mRNA expression as shown in Fig. 1G and H, leading to the supposition that there could be other underlying mechanisms.

As the general mechanisms such as promoter activity, translation activity and mRNA stability were not proven to be significantly involved in TGF-β1 expression in the above studies, we then moved to use an iTRAQ (isobaric tags for relative and absolute quantitation)-based quantitative proteomic approach to screen other potential mechanisms. We thus identified a total of 302 differently expressed proteins induced by eIF6 deficiency. Notably, proteins down-regulated by eIF6 deficiency were involved in chromosome organization, chromatin organization, chromatin assembly or disassembly, DNA packaging, protein–DNA complex assembly and nucleosome assembly. Moreover, we found that there was an association between the expression of eIF6 and histone proteins confirmed by western blot analysis (Fig. S3a), especially histone variant H2A.Z (Fig. 3A, Table 1 and Fig. S3b,c). Recently, it has been reported that H2A.Z co-operates with transcription factors to mediate gene expression in response to diverse cellular cues (Subramanian et al., 2015). It is reasoned that H2A.Z depletion from gene bodies led to a more open chromatin state, facilitating the accessibility of the coding sequence to binding factors. These observations support the idea that an epigenetic mechanism of eIF6 is involved in TGF-β1 regulation. While the transcription factor Sp1 is the primary determinant for the control of constitutive expression of the TGF-β1 gene (Li et al., 1998), we hypothesized that eIF6 might influence the presence of H2A.Z with the TGF-β1 promoter facilitating Sp1 accessibility to the TGF-β1 promoter. To test this hypothesis, we performed a TGF-β1 promoter assay and chromatin immunoprecipitation (ChIP)-qPCR analysis. Previous analysis of the mouse TGF-β1 promoter showed two major transcription start sites (TSS, +1 and +290) and five putative Sp1 binding sites (Sp1-1: −122/−117; −66/−61; Sp1-3: −35/−30; Sp1-4: −15/−10; Sp1-5: +12/+17 around the TSS; Fig. 3B) (Geiser et al., 1991). After the stimulation with exogenous TGF-β1, H2A.Z binding to the different regions of TGF-β1 promoter was greatly reduced in eIF6^+/− fibroblasts but enriched at the corresponding regions of TGF-β1 promoter in WT fibroblasts (region 1: −120/+73; region 2: −37/+175; region 3: −118/+246) (Fig. 3C). It was demonstrated using ChIP-qPCR that Sp1 recruitment to TGF-β1 promoter was remarkably increased in eIF6^+/− fibroblasts compared with that in WT fibroblasts after TGF-β1 stimulation. Moreover, this enhanced recruitment induced by eIF6 deficiency occurred at all the Sp1 binding sites (Fig. 3D). Concomitantly with these chromatin modifications, TGF-β1-auto-induction TGF-β1 mRNA expression was dramatically enhanced by eIF6 deficiency (eIF6^+/− fibroblasts versus WT fibroblasts upon the exogenous TGF-β1 stimulation, 3.4-fold increase, P=0.001, n=5; Fig. 3E). These data were consistent with the above-described parallel ELISA and qPCR data, which showed that increased expression of TGF-β1 was found in eIF6^+/− fibroblasts.

The discovery that eIF6 could regulate TGF-β1 expression at transcription level prompted the question: did this regulation affect cell function? The in vitro myofibroblast differentiation model was chosen to test this possibility because TGF-β1 plays a critical role in myofibroblast differentiation. We first examined the expression of myofibroblast marker proteins (α-SMA, the most used marker of myofibroblastic phenotype) and collagen I in eIF6^+/− and WT fibroblasts. The expression of α-SMA was significantly increased in eIF6^+/− fibroblasts compared with WT fibroblasts at both the mRNA (2.9-fold difference, P=0.020, n=5; Fig. 4A) and protein (1.8-fold difference, P=0.029, n=5; Fig. 4B) levels. Expression of collagen I (Col1a1 and Col1a2) mRNA was also increased in eIF6^+/− fibroblasts (Col1a1: 4.9-fold difference, P=0.001, n=3; Col1a2: 4.6-fold difference, P=0.007, n=3; Fig. 4C). In line with these observations, increased stress fiber formation was found in eIF6^+/− fibroblasts compared with WT fibroblasts (Fig. 4D). Using an established in vitro collagen contraction assay in a fibroblast-populated collagen lattice (FPCL) model (Carlson and Longaker, 2004), eIF6^+/− fibroblasts were shown to contract collagen lattice more extensively than WT fibroblasts (P=0.001, n=3; Fig. 4E). When the isolated fibroblasts were examined by flow cytometry, there was a reduction in the percentage of eIF6^+/− fibroblasts in S-phase (6.9 vs 11.1% in WT; Fig. 4F) but the percentage of eIF6^+/− fibroblasts in G2/M-phase was increased (42.13 vs 27.59% in WT; Fig. 4F), suggesting a higher percentage of mature fibroblasts in the G2/M-phase. There was no significant difference in apoptosis observed between eIF6^+/− and WT fibroblasts (Fig. 4G), suggesting that the mature fibroblasts have the same cellular lifespan as WT fibroblasts. Collectively, these observations indicate that eIF6 plays a role in myofibroblast differentiation in vitro.

Having discovered that eIF6 regulated TGF-β1 mediated myofibroblast differentiation, we proposed that TGF-β1/Smad signal pathway might also be affected and consequently involved in the generation of MFB in eIF6^+/− fibroblasts. Firstly, the levels of type II transforming growth factor-β1 receptor (TGF-βRII) (Tgfbr2) mRNA expression did not differ in eIF6^+/− fibroblasts versus wild-type fibroblasts (P=NS, n=3; Fig. 5A). The activity of TGF-β1 is mediated downstream through activation and nuclear translocation of effectors, including Smad2, Smad3 and Smad4 (Massagué, 2000) and suppression of inhibitory Smad7 (Zhang et al., 2007). Secondly, we examined the Smad mediator expression upon exogenous stimulation with TGF-β1 (at 5 ng ml⁻¹). The level of phosphorylated Smad2 (p-Smad2) was increased 1.5-fold (P=0.008, n=3; Fig. 5B) and coupled with a decreased Smad7 level of 50% (P=0.002, n=3; Fig. 5D) in eIF6^+/− fibroblasts compared with WT fibroblasts. The level of phosphorylated Smad3 was slightly but not significantly elevated in eIF6^+/− fibroblasts (P=NS, n=3; Fig. 5C). A TGF-βR-I-specific pharmacological inhibitor SB431542 (30 μM), which interferes with Smad2 phosphorylation activation (Mori et al., 2004), significantly inhibited p-Smad2 in eIF6^+/− fibroblasts to the WT level (Fig. 5E), leading to the suppression of the up-regulation of α-SMA expression in eIF6^+/− fibroblasts, in which the level was similar to that in WT fibroblasts (Fig. 5F). These data suggest that the axis downstream of TGF-β1 is intact and can be modulated under eIF6 deficiency.

To determine whether in vitro findings on the role of eIF6 in the induction of TGF-β1 and regulation of MFB differentiation are reproducible in vivo, we performed a circular full thickness excision wound on eIF6^+/− and their littermate WT mice. As it has previously been proven that MFB differentiation contributed to wound contraction for wound closure (Kramann et al., 2013), we evaluated the rate of wound closure after injury. The wound closure rate of eIF6^+/− mice was much more rapid than that of WT mice at day 1 after injury (P=0.045, n=6; Fig. 6A), indicating improved wound contractibility in eIF6^+/− mice. Consistent with this, immunohistochemistry results showed that expression of α-SMA was increased throughout the wound and, in particular, beneath the hyperproliferative epidermis of wound edges in eIF6^+/− mice in 6-day-old wounds (Fig. 6E). However, only a faint staining for α-SMA was detected at the edges of WT wounds (Fig. 6E). These data were confirmed by western blot analysis; α-SMA protein level was increased in the wound of eIF6^+/− mice at day 6 after injury (1.4-fold increase, P=0.040, n=3; Fig. 6C). Similarly, distinctly strong TGF-β1-expressed staining was found in the newly formed granulation tissue at day 6 after wounding in eIF6^+/− mice compared with wounds of the WT mice (Fig. 6D). These data were confirmed by western blot analyses. The level of TGF-β1 in the wound of eIF6^+/− mice was shown to have increased at day 0 (3.2-fold increase, P=0.029, n=3), and at day 1 (4.5-fold increase, P=0.015, n=3) and day 6 (2.4-fold increase, P=0.017, n=3) post-wounding compared with that in WT mice (Fig. 6B). Our results thus suggest that increased amounts of TGF-β1 may account for the increased numbers of α-SMA-positive myofibroblasts and the enhanced wound contraction in eIF6^+/− mice, indicating that eIF6 deficiency could regulate the expression of TGF-β1 and MFB differentiation in vivo.

Whilst the activation of ECM-producing myofibroblast is a feature common to all fibrotic diseases regardless of the initiating events (Wynn and Ramalingam, 2012), the results outlined above implicated a role for eIF6 in this process. To investigate this further, we first examined whether the eIF6 expression differed in the fibrotic skin tissue and non-fibrotic skin tissue. When hypertrophic scar-derived fibroblasts (labeled as sFB in the figures) from surgical samples and normal skin-derived fibroblasts (labeled as FB in the figures) were examined by qPCR and western blotting analyses, eIF6 expression was found to be significantly lower in sFB compared with fibroblasts at both the mRNA level (53% reduction, n=10, P=0.041; Fig. 7A) and the protein level (50% reduction, n=10, P=0.028; Fig. 7B). The over-expression of eIF6 by adenoviral vector in sFB resulted in a significant decrease in α-SMA expression, both at the mRNA level (67% reduction, P<0.001, n=5; Fig. 7C) and protein level (67% reduction, P=0.029, n=5; Fig. 7D). In addition, eIF6-gene transduction significantly decreased COL1A1 mRNA expression in sFB (52% reduction of COL1A1, P<0.001, n=3; Fig. 7E) and significantly lowered the capacity for the sFB to contract the floating collagen lattice (P<0.001, n=3; Fig. 7F) compared with the control vector transduced cells.

The factor eIF6 is a highly conserved protein necessary for cell life, and can mediate the ribosome biogenesis and assembly (Ceci et al., 2003). It has been previously demonstrated that eIF6 controls translation at the rate-limiting step of initiation in response to extracellular signals (Gandin et al., 2008). TGF-β1 is a multifunctional cytokine that participates in numerous biological processes, including cell proliferation, differentiation, immune modulation and extracellular matrix production (Sporn et al., 1986). However, the molecular mechanisms underlying TGF-β1 regulation are not fully understood. Similarly, the potential role for eIF6 in TGF-β1 regulation and the associated mechanism has not been examined. In the present study, we found an inverse correlation between the levels of eIF6 and TGF-β1 (Fig. 1). Unexpectedly, there was no significant change in TGF-β1 mRNA associated with heavier polysomes in eIF6^+/− fibroblasts, as revealed by polysomal profile analysis, which is inconsistent with previous evidence for eIF6 control translation initiation (Fig. 2A). These data led to the hypothesis that eIF6 might participate in the transcriptional regulation of TGF-β1. However, neither the stability of TGF-β1 mRNA nor the methylation levels of the TGF-β1 promoter have revealed significant differences between eIF6^+/− fibroblasts and WT fibroblasts (Fig. 2B,C). The promoter assay showed only partial inhibition of TGF-β1 promoter-mediated luciferase activity under the condition of eIF6 over-expression (Fig. 2D), which could not explain the profound influence of eIF6 on TGF-β1 expression. The promoter-reporter assay may not accurately reflect regulation of the endogenous TGF-β1 gene, where chromatin effects may represent another level of control. Interestingly, our Sp1 ChIP data showed that Sp1 recruitment is highly increased within the 5′-UTR promoter region of TGF-β1, and supports an indirect role of eIF6 in the regulation of TGF-β1 transcription.

How could the Sp1 bind to the TGF-β1 promoter region more when eIF6 was deficient? We know that eukaryotic genes are organized into assemblies of core histone-containing nucleosomes. Transcription factors co-operate with RNA polymerases searching packed chromatin efficiently, then locating and interacting with their target DNA-binding sites to regulate gene transcription. Recently, histone variant H2A.Z has garnered particular interest because it integrates information from histone post-translational modifications, other histone variants, and transcription factors to mediate gene induction across various tissue types or environmental conditions (Subramanian et al., 2015). It has been shown that H2A.Z incorporation can alter nucleosome structure and dynamics, the activity of chromatin remodeling enzymes, and histone modification patterns (Weber and Henikoff, 2014). Intriguingly, our iTRAQ-based proteomic data revealed an association between eIF6 and H2A.Z, thus indicating a role for eIF6 in chromosome organization, chromatin organization, chromatin assembly or disassembly, DNA packaging, protein–DNA complex assembly and nucleosome assembly. Prompted by these observations, we carried out H2A.Z ChIP assays at the TGF-β1 promoter region and confirmed that H2A.Z occupancy was remarkably decreased in eIF6^+/− fibroblasts, while TGF-β1 mRNA expression was dramatically enhanced in these cells, as confirmed by qPCR (Fig. 3C,E). As such, it is proposed that H2A.Z eviction from TGF-β1 promoter facilitates the Sp1 recruitment because H2A.Z converts chromatin from a transcriptionally repressed state to an active state under eIF6 deficiency. This idea is yet to be formally tested; however, here we provide an important clue for epigenetic modulation of TGF-β1 transcription by eIF6. This type of mechanism has recently been described as the expression of TGF-β1 is well correlated with chromatin conformation at the TGF-β1 promoter, and the open chromatin conformation might facilitate the binding of Sp1 and Sp3 to the TGF-β1 promoter region (Lee et al., 2011). Future studies, of particular interest, will be directed at determining the exact role by which eIF6 regulates H2A.Z incorporation and chromatin structure altering. In addition, H2A.Z is reported to be essential for survival in many organisms and is important for lineage commitment during differentiation (Creyghton et al., 2008). It has been demonstrated that H2A.Z-deficient mice embryos develop normally until implantation (Faast et al., 2001; Wu et al., 2014). Gandin et al. (2008) have shown that eIF6 null mice embryos are lethal at pre-implantation, indicating that eIF6 is essential for proper development at the stage of pre-implantation. Taken together, this indicates that there might be a connection between eIF6 and H2A.Z in early development and pre-implantation that remains to be clarified.

Another interesting implication of the proteomics data is that eIF6 deficiency leads to 180 proteins being significantly up-regulated and 122 proteins down-regulated in the fibroblasts (data not shown). Although a positive role for eIF6 in global translation has been proved in knockdown mice, there is evidence for a selective role of eIF6 in gene regulation. In line with these data, eIF6 has been demonstrated to selectively regulate Wnt signaling and β-catenin protein synthesis (Ji et al., 2008). However, verification of this concept requires further study.

Myofibroblasts, characterized by the presence of stress fibers and α-SMA, are the common feature of fibrotic disease and are responsible for recruitment of inflammatory cells, remodeling of the ECM, and a favorable environment for tumor cell growth and invasion (Eyden et al., 2009). TGF-β1 is a key fibrogenic regulator of the myofibroblast phenotype that favors their differentiation and resistance to apoptosis. In the light of our findings, it is possible to speculate that eIF6 exerts control over the myofibroblast differentiation. Supporting this concept, the data confirm that the expression of α-SMA and collagen I are markedly increased coupled with the enhanced formation of stress fiber and cell contractibility in eIF6^+/− fibroblasts (Fig. 4A–E). It is generally accepted that the proliferative status of a cell is closely related to its differentiation, i.e. the organism develops but the proliferation becomes restricted (Andreeff et al., 2000). In agreement with this concept, our results demonstrated that a smaller fraction of eIF6^+/− fibroblasts were in S-phase versus that of WT cells, whereas a greater fraction of eIF6^+/− cells than WT cells were arrested at the G2/M-phase. Moreover, eIF6^+/− fibroblasts showed no significant difference in apoptosis from WT cells. These data indicate that a more differentiated and functional cell cycle takes part in the fibrogenic phenotype of eIF6^+/− fibroblasts. In partial keeping with this observation, eIF6 down-regulation in pre-B-cells reduced the S-phase fraction and increased the G0/G1 fraction, thereby impairing mitotic cycle progression (Miluzio et al., 2011). In accordance with our results, the high eIF6 expression observed in dysplastic carcinomas suggests that eIF6 is involved in the transition to malignancy by impairing the cell's differentiated capacity (Rosso et al., 2004). Thus the partially hindered mitotic cell cycle progression in eIF6^+/− fibroblasts may favor their differentiation into myofibroblasts. Furthermore, we examined the role of eIF6 in vivo using a skin wound model in eIF6^+/− mice, and found that eIF6 deficiency induced the expression of TGF-β1 and myofibroblast differentiation. Taken together, we therefore propose that eIF6 deficiency dramatically enhances TGF-β1 mRNA expression in the manner of TGF-β1 auto-induction. The increased level of TGF-β1 then might contribute to facilitate eIF6^+/− fibroblast differentiation to myofibroblast. To the best of our knowledge, this is the first study on the role of eIF6 in myofibroblast differentiation and the underlying mechanisms. Thus, our study might provide insights into the novel mediators (eIF6) linking myofibroblast differentiation, which is closely related to fibrosis development, tumor progression, cardiovascular pathological alteration, etc.

Furthermore, the use of pharmacological inhibitors of the Smads pathway confirmed a key function of p-Smad 2 and Smad 7 in the transduction of fibrogenic signals by eIF6 deficiency (Fig. 5B–E). The TGF-β/Smad signal pathway was identified as a main fibrogenic signaling signature induced in eIF6^+/− fibroblasts in response to exogenous TGF-β1 stimulation, as its inhibition largely blocked the increased expression of α-SMA induced by eIF6 deficiency (Fig. 5F). Collectively, these results characterized the TGF-β/Smad pathways as the important signaling components activated under eIF6 deficiency. Moreover, eIF6-gene transduction in hypertrophic scar-derived fibroblasts significantly decreased the expression of myofibroblast marker proteins and suppressed their contractibility, raising the possibility of eIF6 as a repressor of controlling the myofibroblast phenotype during wound healing and fibrogenesis. Furthermore, the subcellular localization of eIF6 was switched from perinucleus in the fibroblasts in normal skin sample to cytoplasm of fibroblasts in hypertrophic scar sample from the same patient (Fig. S4c,d). Thus, the reduced expression (Fig. S4a) and the altered subcellular distribution of eIF6 we detected in the hypertrophic scar tissue seem to further support the above hypothesis.

In summary, we provide novel insights of the functional role of eIF6 in modulating the transcription of TGF-β1 by regulating Sp1 recruitment to TGF-β1 promoter via H2A.Z occupancy. Our data mechanically explain the eIF6 deficiency-induced myofibroblast differentiation through the TGF-β/Smad pathway. The role of eIF6 in the regulation of TGF-β1 and myofibroblast differentiation has been proposed in Fig. 7G, which extends the roles for the multifunctional eIF6 in different and as yet undescribed cellular processes.

All studies were approved by the Ethics Committee of Southwestern Hospital and were performed according to the Declaration of Helsinki (with the Edinburgh 2000 revisions). The animals were treated strictly in accordance with the National Institutes of Health Guide concerning the Care and Use of Laboratory Animals. The surgery on animals was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize suffering.

The following antibodies were used in this study: rabbit antibody to eIF6 (Cell Signaling Technology, Beverly, MA); rabbit antibody to α-SMA (Abcam, Cambridge, UK); rabbit antibody to TGF-β1 (NOVUS Biologicals, Littleton, CO); rabbit antibody to Smad7 (Epitomics, Burlingame, CA); antibody to p-Smad2 Ser465/467, antibody to Smad2, antibody to p-Smad3 Ser423/425 and antibody to Smad3 (Cell Signaling Technology); mouse antibody to β-actin, mouse antibody to GAPDH (Santa Cruz Biotechnology, Santa Cruz, CA); and rabbit antibody to Sp1, rabbit antibody to H2A.Z and mouse antibody to H2B (Abcam).

Hypertrophic scar-derived fibroblasts (labeled as sFBs in the figures) from scar-excision surgical samples and normal skin-derived fibroblasts (labeled as FBs in the figures) from autologous skin grafting were isolated and cultured as previously described (Tan et al., 2010). Primary dermal fibroblasts from mouse skin were obtained using neonatal (3-day-old) eIF6^+/− and wild-type mice and were called mouse neonatal fibroblasts (MNFs) in this paper. All analyses were performed at least three times to ensure reproducibility with various genetic backgrounds. After washing in 75% ethanol and phosphate-buffered saline (PBS), we excised dorsal and abdominal skin from the region posterior to the occipital bone to anterior to the tail base. The remaining subcutaneous fat and muscle were trimmed, and the skin was washed twice in sterile, ice-cold PBS. The skin was cut into 1 mm×1 mm squares and washed in sterile, ice-cold 1× PBS. The skin was digested with 0.25% trypsin at 37°C for 30 min in a constant-temperature shaker at 250 r.p.m. The fibroblasts were filtered using a 200-mesh cell strainer and cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco, Gaithersburg, MD) supplemented with 10% fetal bovine serum, 100 mg ml⁻¹ streptomycin, 100 U ml⁻¹ penicillin and 2 mM glutamine, in a humidified cell culture incubator at 37°C supplemented with 5% CO₂ and 2% O₂. The culture medium was replaced every 2 days, and experiments were performed between the first and third passages (P1–P3). TGF/Smad pathway analysis was performed as follows: MNFs were synchronized for 24 h with serum-free medium (SFM, Sciencell, Carlsbad, CA) and were then left untreated or treated with 5 ng ml⁻¹ TGF-β1 (R&D Systems, Minneapolis, MN) for 24 h or with 30 µM SB431542 (TGF-β1 kinase inhibitor VI; Calbiochem, La Jolla, CA) for 1 h and followed by 5 ng ml⁻¹ TGF-β1 treatment for 24 h.

The hypertrophic scar-derived fibroblasts (labeled as sFBs) were transfected with recombinant eIF6 adenoviral vector (eIF6-EGFP) or with the control empty cassette vector (EGFP) as previously described (Tan et al., 2010). The full-length eIF6 gene was cloned into the pEGFP-N2 vector in-frame with a C-terminal EGFP tag; subsequently, eIF6-EGFP and EGFP were cloned into the multiple cloning site of the pShuttle-CMV vector, and homologous recombination was performed using pAdEasy-1 in Escherichia coli BJ5183 to obtain the recombinant adenovirus pAd-eIF6-EGFP and the control empty cassette adenovirus pAd-EGFP. All plasmids were sequenced to confirm sequence fidelity. As previously described (Setoguchi et al., 1994), the recombinant adenoviral vectors were used to transduce HEK293 cells using Lipofectamine™ 2000 transfection reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The primary adenoviruses were then propagated in HEK293 cells. All adenoviruses used for subsequent experiments were obtained from passage 4–5, and the viral titer was determined in plaque-forming units (PFU). The adenoviruses were used to infect hypertrophic scar fibroblasts directly by co-culture.

A TGF-β1 promoter fragment spanning nucleotides −1328 to +812 was synthesized from human genomic DNA (Promega) by PCR using the primers 5′-GAT TCG ACG CGT AGA TCA CTT TGG CTG CTG T-3′ (forward primer) and 5′-TAG ACC AGA TCT GAG CGC GAA CAG GGC-3′ (reverse primer). Mlu I and Bgl II restriction sites were respectively added to forward primer and reverse primer by PCR. The amplified PCR products were ligated into the Mlu I and Bgl II sites of the pGL3-basic vector (Promega), yielding pGL3-TGF (−1328/+812). A series of deletion constructs of human TGF-β1 promoter fragments were synthesized by PCR using the pGL3-TGF (−1328/+812) plasmid as a template. The primer sequences are listed in Table S1. The PCR products were digested with Mlu I and Bgl II, and ligated into the Mlu I and Bgl II sites of the pGL3-basic vector. All the resultant constructs were verified by DNA sequencing and by restriction enzyme digests.

This was a retrospective study. The inclusion criteria for patients with hypertrophic scar and preparation of scar specimens have been described previously (Tan et al., 2010). Sections of human hypertrophic scar samples were obtained from scar-excision surgery and from autologous skin grafting preserved in tissue bank of the Institute of Burn Research, Southwest Hospital (Chongqing, China). The study was conducted according to the Helsinki Declaration and approved by the Ethics Committee of Southwest Hospital of the Third Military Medical University, PLA [Approval number: TMMU Scientific Research 2014 (57)]. Samples used in this study were anonymous, and it is impossible for anyone to link the samples to the sources.

The eIF6^+/− mice (CL57Bl6 background) and corresponding control wild-type mice (eIF6^+/+) were purchased from Fondazione Centro San Raffaele del Monte Tabor (Italy). The eIF6 mutation was repeatedly backcrossed. The mice were maintained in a pathogen-free environment and under light-, temperature- and humidity-controlled conditions. Food and water were made available ad libitum. Male eIF6^+/− mice and WT mice aged 8–10 weeks weighing 20–22 g (CL57BL/6 background) were used for wound model and constantly maintained under specific pathogen-free conditions (n=6 male mice per group, and the experiment was repeated at least 3 times). Mice were anesthetized with an intraperitoneal injection of 1% (w/v) sodium pentobarbital (0.05 ml g⁻¹ body weight). After shaving the dorsal hair and cleaning the exposed skin with 70% ethanol, two sets of two circular (4.5 mm) full-thickness skin wounds were punched on each side of the spine in the middle of the dorsum, as described elsewhere (Peters et al., 2005). The animals were killed by administering an overdose of anesthetic 6 days after the surgery. Wound healing and contraction were evaluated post-operatively at six time points: 1 and 6 days from images taken together with a mm measuring scale in the field of view. The wound areas were calculated using Photoshop software (Adobe Systems) and wound sizes at any given time point after wounding were expressed as a percentage of initial (day 0) wound area. The mean of the measurements of the wounds was used in the statistical analyses. For expression analyses, one wound sample from each animal was frozen in liquid nitrogen immediately after excision.

Total cellular RNA was isolated from primary cells using RNeasy Mini kits (Qiagen, Valencia, CA). Quantitative RNA analysis was performed using SYBR-Green I (Toyobo) and a 7500 Real-Time PCR System (Applied Biosystems) as previously described (Tan et al., 2010). The primer pairs used are shown in Table S1. Real-time PCR analysis was performed as follows: initial denaturation for 5 min at 95°C, 40 cycles of 15 s at 95°C, 15 s at 60°C and 20 s at 72°C. Cycle threshold values of mRNAs were normalized to the GAPDH (or β-actin) internal control, and the values were plotted as the relative transcript abundance. Three biological replicates were used for all experiments, and representative figures are shown. The primer pairs are listed in Table S2.

The whole-cell lysates from fibroblasts and wound tissue were extracted using the KGP250 Kit (KeyGen Biotech, Nanjing, People's Republic of China), and western blotting was performed as previously described (Tan et al., 2010) with antibodies against eIF6, α-SMA, TGF-β1, Smad7, p-Smad2 Ser465/467, Smad2, p-Smad3 Ser423/425, Smad3, β-catenin, β-actin and GAPDH.

Fibroblasts (passage 2) from skin derived from the eIF6^+/− neonatal mice and their wild-type littermate controls were synchronized for 24 h in serum-free medium (SFM). The medium was then replaced with fresh SFM and supernatant samples collected from cells for another 24 h. The total TGF-β1 in the cell culture supernatant was measured by an ELISA kit (R&D Systems, Minneapolis, MN). The TGF-β1 concentration was normalized to the number of cells, which was measured by cell counting using a hemocytometer. The data are presented as TGF-β1 pg ml⁻¹ per 10⁵ cells.

Tissue samples were fixed in 4% paraformaldehyde in 0.1 M PBS at pH 7.4 and embedded with paraffin. Serial 3-mm sections were then made with the paraffin-embedded tissues. For immunohistochemistry, the sections were then incubated with Proteinase K (Millipore, Bedford, MA) for 20 min (37°C), and endogenous peroxidase was quenched with 3% H₂O₂ in methanol for 10 min at room temperature. The sections were then blocked with 1% BSA (Sigma-Aldrich, St Louis, MO) in PBS and incubated overnight with the primary antibodies (anti-eIF6 1:2500; anti-α-SMA 1:2000; anti-TGF-β1 1:1000) diluted in blocking solution (4°C), followed by incubation with a peroxidase-coupled secondary antibody plus streptavidin–peroxidase complex and diaminobenzidine staining (ZSGB-BIO, Beijing, People's Republic of China). The sections were then counterstained with hematoxylin.

Fibroblasts (passage 3) from eIF6^+/− and wild-type mice were cultured on coverslips and fixed using cold acetone and methanol with subsequent rinsing in PBS. The coverslips were then stained using the In Situ Cell Death Detection Kit (Roche, Mannheim, Germany) according to the manufacturer's instructions. Counterstaining was performed using 4′,6-diamidino-2-phenylindole (DAPI). The TUNEL-stained coverslips were imaged, and positive cells were counted in five random high-power fields for at least two sides for each group.

Cell cycle analysis was performed as previously described (Darzynkiewicz et al., 2001). Briefly, cells were harvested and washed twice with PBS followed by fixation in 80% ethanol for 30 min at room temperature. The cells were then collected by centrifugation and stained with 50 μg μl⁻¹ propidium iodide. The cells were then treated with 100 µg µl⁻¹ RNase for 15 min at 37°C and analyzed using a FACScan flow cytometer. The cell cycle distribution was analyzed using the ModFit LT program.

Fibroblast contraction studies were performed as previously described (Tan et al., 2010). The fibroblast-populated collagen lattices (FPCL) contraction index was calculated as follows (Akhmetshina et al., 2012):

Contraction index=[1 – (D/D₀)²]×100%,

where D is the diameter of the collagen protein gel block, and D₀ is the initial diameter of the gel block.

The eIF6^+/− and WT fibroblasts were stained by phalloidin-tetramethylrhodamine and evaluated for stress fiber confocal microscopy, as previously described (Laplante et al., 2010). In brief, to observe the characterization of stress fiber with confocal microscopy, cells were grown on 14-mm glass slides in 24-well plates, rinsed with PBS and fixed with 4% formaldehyde. They were washed three times in PBS before permeabilization and after each subsequent step. Permeabilization was performed with 0.2% Triton X-100 in PBS for 15 min. Microwell dishes were blocked with PBS/BSA 3% for 30 min and incubated with phalloidin-tetramethylrhodamine (Sigma) for 60 min at room temperature. The cells were then visualized at room temperature under a Leica SP5 confocal microscope (emission detector 570–620 nm) and analyzed using Zeiss LSM Image Examiner software.

HEK293 cells were seeded onto 6-well plates and co-transfected with the pCMV-Tag3B-eIF6/pCMV-Tag3B and the constructs of TGF-β1 promoter using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. At 24 h post-transfection, the levels of firefly and Renilla luciferase activity were measured sequentially from a single sample using the Dual-Glo Luciferase Assay System (Promega). The firefly luciferase activity was normalized to Renilla activity, and the relative amount of luciferase activity in the untreated cells was designated as 1. The luminescence was measured with a luminometer (Centro LB960; Berthold Technologies, Bad Wildbad, Germany).

Polysome profiles were performed as previously described (Carrieri et al., 2012). HEK293T cells were infected with packaging plasmid PMDG, ΔR8.74, pILLV and shRNA-eIF6 in this analysis. The mature antisense sequences of the constitutive shRNA for eIF6 (pGIPZ plasmid; Open Biosystems) is: 5′-AGCTTCCTACTAGCACCTG-3′. The cytoplasmic lysates were fractionated on a sucrose gradient. RNA was extracted from each of the fractions and subjected to qPCR. Data represent the percentage of the total amount of corresponding mRNA on each fraction.

TGF-β1 promoter methylation analysis was performed with the bisulfite sequencing PCR (BSP) method. Bisulfite modification was performed with a EZ DNA Methylation-Gold Kit (Zymo Research). A DNA fragment in the TGF-β1 promoter region was amplified by PCR with the primers listed in Table S3. The obtained DNA fragments were ligated with pMD19-T (Takara) and transfected into E. coli. Plasmid DNA purified from each colony was submitted for sequencing. The sequencing results were analyzed for DNA methylation using the website http://quma.cdb.riken.jp/.

For quantitation of the rate of decay of TGF-β1 mRNA, eIF6^+/− and WT fibroblasts were treated with the transcription inhibitor actinomycin D (Act-D) at a final concentration of 2 μg ml⁻¹. At various times (0, 6, 12 and 24 h) after the addition of Act-D, medium was removed and total RNA was isolated as described above. The rate of mRNA degradation was subsequently determined using real-time RT-PCR, as described above.

eIF6^+/− and WT fibroblasts were synchronized for 24 h with serum-free medium and were then treated with 5 ng ml⁻¹ TGF-β1. After stimulation, protein–DNA complexes were cross-linked by formaldehyde (1% final concentration). Cells were resuspended in 200 μl sodium dodecyl sulfate (SDS) lysis buffer [50 mM Tris (pH 8.1), 1% SDS, 5 mM EDTA, and complete proteinase inhibitor mixture] and subjected to five cycles of sonication on ice with 10-s pulses. Sonicated samples were centrifuged to spin down cell debris, and the soluble chromatin solution was immunoprecipitated using antibodies specific for H2A.Z, Sp1 (Abcam, Cambridge, UK). Protein-bound immunoprecipitated DNA was washed with LiCl wash buffer and 10 mM Tris and 1 mM EDTA (pH 8.0), and immune complexes were eluted by adding elution buffer (1% SDS and 0.1 M NaHCO₃), followed by incubation for 4 h at 65°C in 200 mM NaCl and 1% SDS to reverse cross-links, and incubation for 1 h at 45°C with 70 μg ml⁻¹ proteinase K (Sigma-Aldrich). DNA was extracted with phenol/chloroform, precipitated with ethanol/0.3 M NaHCOOH/20 μg glycogen, and resuspended in 100 μl of nuclease-free water. qPCR was performed with 8 μl of DNA sample for quantification. The primers for quantification are listed in Table S4.

iTRAQ Protein Profiling was performed as previously described (Lu et al., 2008), and the complete description of sample preparation, peptide labeling, cation exchange fractionation, Nano-LC and MALDI-TOF/TOF mass spectrometry analysis is given in a previous report (Ruppen et al., 2010). To decrease eventual biases caused by biological variations, fibroblasts isolated from three to five mice from each group (eIF6^+/− mice and WT mice) were pooled, yielding one sample. The complete set of data files (*.RAW) from the iTRAQ experiments were searched against the mouse ipi.MOUSE.v3.87.fasta database (version 28/9/2011, containing 59,534 peptides) using Mascot2.2 software. The false discovery rate (FDR) was used to filter the results (protein FDR≤0.01, peptide FDR≤0.01) (Sheng et al., 2012). The relative quantification and statistical analysis were provided by ProteomicsTools software (version 3.0.5), and proteins with a P-value <0.05 were considered differentially expressed proteins. An additional 1.3-fold change cut-off for all iTRAQ ratios (ratio <0.79 or >1.33) was selected to classify proteins as up- or down-regulated. Proteins with iTRAQ ratios below the low range (0.79) were considered to be under-expressed, whereas those above the high range (1.33) were considered to be over-expressed. All the significant down-regulated proteins were subjected to gene ontology classification by publicly available DAVID 2.0 software (http://david.abcc.ncifcrf.gov/). All gene ontology groups demonstrated enhanced statistical representation (P<0.01). In the figures, bars represent the proportion of genes involved in each category, and the number of genes is indicated in the figure.

Statistical analyses were performed using two-tailed Student's t-test or one-way analysis of variance (ANOVA) for multiple comparisons as appropriate. Data are expressed as means±s.e.m. Values of P<0.05 were considered significant. Prism 5.0 software was used to perform statistical analysis.

We thank Professor Stefano Biffo (Istituto Nazionale di Genetica Molecolare, ‘Romeo ed Enrica Invernizzi’, Milano, Italy) and other members of his laboratory for helpful discussion and suggestions on the experiments.