Coming together: liver fibrosis, collagen mRNAs and the RNA binding protein

Liver fibrosis is characterized by the excessive and uncontrolled production of type I collagen by activated hepatic stellate cells (HSCs). Major complications of liver fibrosis are caused by the deposition of type I collagen; however, most current research in liver fibrosis is directed towards understanding the activation of HSCs rather than the mechanism of collagen synthesis. Current models of the synthesis of type I collagen postulate that the procollagen α1(I) and α2(I) polypeptides are synthesized separately, and independently post-translationally modified within the lumen of the endoplasmic reticulum. Two α1(I) peptides and one α2(I) peptide then find each other and fold into a triple helix. However, two facts contradict this simplified model. First, more than 99% of naturally synthesized type I collagen is comprised of heterotrimers of two α1(I) chains and one α2(I) chain; homotrimers of α1(I) chains almost never form [1]. However, collagen α1(I) homotrimers readily form triple helices in humans who have a complete absence of the α2(I) chain [2] and in knockout mice where the gene coding for α2(I) has been inactivated [3]. Therefore, α1(I) chains have the propensity for folding into a functional triple helix, and a certain fraction of homotrimers would form if the interactions between the α1(I) and α2(I) polypeptides are not strictly coordinated. Second, the rate of post-translational modifications and the rate of folding into triple helices are coupled, since the mutations that delay folding result in hypermodifications of the chains and severe forms of osteogenesis imperfecta [4,5]. Therefore, some coordination of synthesis of the procollagen α1(I) and α2(I) polypeptides must take place.

It’s a competitive world

In mammalian cells, there is an excess of mRNAs, beyond the capacity of the translational machinery. This means that multiple mRNAs compete for a limited number of ribosomes. Features at the 5´ untranslated region determine the competitiveness of an mRNA; secondary structures are inhibitory and the translation start codon must be in the optimal sequence context for efficient initiation [6]. In the 5´ untranslated regions of mRNAs encoding the collagen α1(I) and α2(I) polypeptides, there is a stem-loop structure encompassing the translation start codon [7,8]. There has been an evolutionary pressure to maintain the 5´ stem loop and the sequence around the start codon was sacrificed to maintain the stem loop. Thus, it seems that mRNAs coding for type I collagen are designed to be inefficiently translated, containing a secondary structure and a bad start codon. How then, can activated HSCs synthesize large amounts of type I collagen protein from the mRNAs that are poor messages?

We are not competitive but there are ways

One method to synthesize large amounts of type I collagen protein is to maintain a high steady-state level of type I collagen mRNAs. The half-life of collagen α1(I) mRNA increases from less than 1.6 h in quiescent HSCs to more than 24 h in activated HSCs. This stabilization is caused by the binding of αCP protein to the C-rich sequence in the 3´ untranslated region [9]. This binding is only seen in activated HSCs. αCP stabilizes some other mRNAs [10] but not collagen α2(I) mRNA and, at present, it is not clear what contributes to the long half-life of collagen α2(I) mRNA.

The other method is to increase the local concentration of collagen mRNAs. This can be achieved by binding collagen α1(I) and α2(I) mRNAs together and localizing them at discrete regions of the cell. By aggregating and translating collagen mRNAs in coordination, a high local concentration of α1(I) and α2(I) polypeptides can be achieved. The high local concentration of α1(I) and α2(I) polypeptides can facilitate folding into the heterotrimer, which is concentration dependent [11]. We believe that this coupling at the mRNA level is the key for the high levels of collagen synthesis observed in activated HSCs. So how can collagen α1(I) and α2(I) mRNAs be bundled together?

Coming together

The 5´ stem loop is the sequence element specific for collagen mRNAs of all vertebrate species. We have cloned LARP6 as the protein that binds the 5´ stem loop; mutation of a single nucleotide in the 5´ stem loop completely abolishes the binding. LARP6 has the unique RNA-binding domain and binds the 5´ stem loop of collagen α1(I) and α2(I) mRNA with high affinity (K_d = 1.7 nM) [Stefanovic et al.; Unpublished Data]. These facts suggest that the primary function of LARP6 is in the metabolism of collagen mRNAs. LARP6 can also interact with itself to form dimers and possibly multimers, suggesting that by LARP6–LARP6 interactions, two collagen mRNAs can be brought together to form a complex [Stefanovic et al.; Unpublished Data]. We were able to pull down collagen α1(I) mRNA using the collagen α2(I) 5´ stem loop as bait; the interaction between the two mRNAs was greatly enhanced when LARP6 was overexpressed. LARP6 can also aggregate collagen mRNAs into discrete granules; these granules become prominent when the translational machinery is disassembled and collagen mRNAs cannot feed into the normal biosynthetic pathway. Both findings indicate that LARP6 can aggregate collagen mRNAs.

At the same time, the binding of LARP6 to the 5´ stem-loop is of high enough affinity to prevent ribosomal scanning towards the start codon [Stefanovic et al.; Unpublished Data]. Therefore, even the inefficient translation of collagen mRNAs is completely abrogated by the binding of LARP6. So, how do HSCs produce collagen protein when the collagen mRNAs are bundled and inaccessible to ribosomes?

An unexpected help

Nonmuscle myosin is a motor protein whose function is to move actin filaments. The protein is comprised of two heavy chains and four light chains, forming hexamers that polymerize into a fiber along the actin filaments. By hydrolyzing ATP, hexamers acquire their motor function to slide actin filaments, resulting in cell locomotion and contractility [12]. Nonmuscle myosin is not expressed in quiescent HSCs; however, upon activation, high levels of the protein are found in HSCs [13]. The temporal expression of nonmuscle myosin in culture activation of HSCs parallels that of collagen. Nonmuscle myosin is also expressed in myofibroblasts of various sources. We have discovered that the carboxy terminal domain of LARP6 interacts with nonmuscle myosin and that collagen mRNAs can be immunoprecipitated with antibodies against nonmuscle myosin in a 5´ stem-loop-dependent manner [Stefanovic et al.; Unpublished Data]. Thus, LARP6 not only bundles collagen mRNAs together, it also associates them with nonmuscle myosin; however, for this task, it uses a different protein domain.

Myosin light chains need to be phosphorylated in order for the hexamers to form. ML-7 is a specific inhibitor of myosin light-chain kinase, and treatment of HSCs with ML-7 results in the disassembly of myosin fibers [14]. When we treated HSCs with ML-7, the network of myosin fibers collapsed and there was a dramatic decrease in collagen mRNA and protein levels, with almost complete cessation of collagen protein secretion into the medium [Stefanovic et al.; Unpublished Data]. This indicates that intact myosin filaments are a prerequisite for the stabilization of collagen mRNAs and for the synthesis and secretion of collagen protein. Since nonmuscle myosin confers motility to myofibroblasts, we postulate that the ability to migrate towards the site of injury and to synthesize type I collagen are integrated processes of the wound-healing mechanism. So how does the association of collagen mRNAs with nonmuscle myosin help the synthesis of collagen protein?

The final act

Polyribosomes (i.e., polysomes) assemble when mRNAs are actively translated into protein. We have discovered that a significant amount of nonmuscle myosin is found in polysomal preparations [13]. When polysomes are dissociated with puromycin, nonmuscle myosin disappears from these preparations. When HSCs are treated with ML-7, the amount of nonmuscle myosin found in polysomal fractions is greatly reduced. This indicates that myosin filaments participate in translation and that activated HSCs develop filaments not only for cell motility but also to support protein synthesis. The majority of proteins synthesized by activated HSCs are extracellular matrix proteins, so it is likely that nonmuscle myosin filaments help the translation of collagen mRNAs by facilitating ribosomal loading.

The LARP6 protein has to be removed in order for collagen ribosomal mRNAs to initiate translation. We have observed that the affinity of LARP6 for 5´ stem loop is higher when the protein is phosphorylated, so one method of removing LARP6 from collagen mRNA is by dephosphorylation. Although the phosphatase responsible for this is unknown, we hypothesize that, upon association of collagen mRNA with nonmuscle myosin, LARP6 is dephosphorylated and leaves the 5´ stem loop. Ribosomes attached to the filaments engage collagen mRNAs and, since these mRNAs are sequestered from the competing cytosolic mRNAs, the translation initiation is under noncompetitive conditions. This results in the coordinated initiation of the synthesis of collagen α1(I) and α2(I) polypeptides, starting with the signal peptides. It is likely that collagen signal peptides are recognized by signal-recognition particles when they emerge from the ribosome and that signal-recognition particles bring the elongating ribosomes, the nascent collagen chains and the collagen mRNAs to the membrane of the endoplasmic reticulum [15]. The motor function of myosin does not to contribute to this relocation, since it can be inhibited by blebbistatin without affecting collagen synthesis. Once on the membrane of the endoplasmic reticulum, the coordinated syntheses, modifications and folding of collagen polypeptides are completed.

Conclusion

A high level of type I collagen synthesis in liver fibrosis is facilitated by the aggregation of collagen α1(I) and α2(I) mRNAs and their association with nonmuscle myosin filaments. These processes are mediated by the binding of LARP6 to the conserved 5´ stem loop of collagen mRNAs. The acquisition of myosin filaments and motility are prerequisites for high levels of type I collagen synthesis. We propose that this mechanism is set in motion in pathological fibrosis when HSCs are activated. A default pathway for constitutive, low-level collagen synthesis is LARP6 independent and operates in quiescent collagen-producing cells. If proven to be specific for fibrotic conditions, the LARP6-dependent mechanism may be an attractive target for the development of antifibrotic drugs.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.