1. Biochemistry and Chemical Biology

Repression of ferritin light chain translation by human eIF3

  1. Mia C Pulos-Holmes
  2. Daniel N Srole
  3. Maria G Juarez
  4. Amy S-Y Lee
  5. David T McSwiggen
  6. Nicholas T Ingolia
  7. Jamie H Cate  Is a corresponding author
  1. University of California, Berkeley, United States
  2. Brandeis University, United States
  3. Lawrence Berkeley National Laboratory, United States
https://doi.org/10.7554/eLife.48193
Share this article
Cite this article
  1. Mia C Pulos-Holmes
  2. Daniel N Srole
  3. Maria G Juarez
  4. Amy S-Y Lee
  5. David T McSwiggen
  6. Nicholas T Ingolia
  7. Jamie H Cate
(2019)
Repression of ferritin light chain translation by human eIF3
eLife 8:e48193.
https://doi.org/10.7554/eLife.48193
  2. Download BibTeX
  3. Download .RIS

Abstract

A central problem in human biology remains the discovery of causal molecular links between mutations identified in genome-wide association studies (GWAS) and their corresponding disease traits. This challenge is magnified for variants residing in non-coding regions of the genome. Single-nucleotide polymorphisms (SNPs) in the 5ʹ untranslated region (5ʹ-UTR) of the ferritin light chain (FTL) gene that cause hyperferritinemia are reported to disrupt translation repression by altering iron regulatory protein (IRP) interactions with the FTL mRNA 5ʹ-UTR. Here, we show that human eukaryotic translation initiation factor 3 (eIF3) acts as a distinct repressor of FTL mRNA translation, and eIF3-mediated FTL repression is disrupted by a subset of SNPs in FTL that cause hyperferritinemia. These results identify a direct role for eIF3-mediated translational control in a specific human disease.

https://doi.org/10.7554/eLife.48193.001

Introduction

Iron is essential for a spectrum of metabolic pathways and cellular growth. However, if not properly managed, excess iron catalyzes the production of reactive oxygen species (ROS). To safeguard against these toxic effects, cells sequester iron in ferritin, a cage-like protein complex composed of a variable mixture of two structurally similar but functionally distinct subunits, the ferritin heavy chain (FTH) and the ferritin light chain (FTL) (Harrison and Arosio, 1996), (Knovich et al., 2009). To maintain iron homeostasis, the expression of both ferritin subunits in mammals is regulated post-transcriptionally by iron regulatory proteins that bind a highly conserved RNA hairpin called the iron responsive element (IRE), located in the 5ʹ-UTRs of FTL and FTH1 mRNAs (Figure 1A and B) (Theil, 1994), (Wilkinson and Pantopoulos, 2014). SNPs or deletions in the 5ʹ-UTR that disrupt IRP-IRE interactions are thought to be the primary cause of hereditary hyperferritinemia cataract syndrome, a condition involving an abnormal buildup of serum ferritin in the absence of iron overload (Cazzola et al., 1997).

Figure 1 with 1 supplement see all
Download asset Open asset
Post-transcriptional regulation of FTL mRNA.

(A, B) Iron-responsive regulation mediated by binding of Iron Response Proteins (IRPs) to Iron Response Element (IRE) RNA structures in the 5ʹ-UTR in (A) low-iron conditions and (B) high-iron conditions. In high iron, IRP2 is degraded by the proteasome, whereas IRP1 binds an iron-sulfur cluster to form the enzyme Aconitase (ACO1). (C) General schematic of the luciferase reporter mRNAs. The eIF3 PAR-CLIP site in FTL mRNA spans nucleotides 53–76 (Lee et al., 2015) and the 3RE region spans nucleotides 58–90. (D) Schematic of the IRP and eIF3 interaction sites on the experimentally-determined secondary structure of FTL mRNA (Martin et al., 2012). (E) Luciferase activity in HepG2 cells transfected with luciferase reporter mRNAs 6 hr post transfection, normalized to luciferase luminescence from mRNA with wild-type FTL 5ʹ-UTR. The results are for three biological replicates with error bars representing the standard deviation of the mean.

https://doi.org/10.7554/eLife.48193.002
Figure 1—source data 1

Luciferase reporter readouts.

https://doi.org/10.7554/eLife.48193.005

Although the IRP-IRE interactions have been considered to be the sole post-transcriptional means of regulating ferritin expression, recent studies have provided strong evidence that other presently unknown factors may provide another layer of regulation during FTL translation. For example, the FTL subunit composition of ferritin is altered in response to environmental factors such as hypoxia (Sammarco et al., 2008). We recently found that eIF3 can function beyond its scaffolding role in general translation initiation by acting as either an activator or repressor of translation in a transcript-specific manner (Lee et al., 2015),(Lee et al., 2016). This regulation occurs through interactions with primarily 5ʹ-UTR RNA structural elements (Lee et al., 2015). Notably, we found that FTL mRNA cross-links to eIF3 (Lee et al., 2015), but the role eIF3 plays in regulating FTL translation has not been established.

Here, we report a previously unknown mode of FTL translation regulation with a direct link to disease-related genetic mutations. We show that eIF3 binds to human FTL mRNA by means of sequences in the 5ʹ-UTR immediately adjacent to the IRE, and provides additional regulation of FTL translation independent of IRP-IRE. After using CRISPR-Cas9 genome editing to delete the endogenous eIF3 interaction site in FTL, we monitored direct phenotypic responses of cells under normal and iron modulated conditions. Lastly, we used competitive IRP binding assays to explore the potential role of eIF3 in hyperferritinemia. These experiments reveal that eIF3 acts as a repressor of FTL translation, and disruption of eIF3 interactions with FTL mRNA due to specific SNPs in the FTL 5ʹ-UTR likely contributes to a subset of hyperferritinemia cases.

Results

Identification of the eIF3-FTL mRNA interaction site

In order to understand the functional effect of the interaction between eIF3 and FTL mRNA, we utilized Renilla luciferase (rLuc) reporter mRNAs in which the 5ʹ-UTR from FTL was placed upstream of the Renilla coding sequence (Figure 1C). To measure the importance of the FTL mRNA region identified by PAR-CLIP (Lee et al., 2015), various mutations were introduced into the FTL 5ʹ-UTR to disrupt eIF3 binding. The eIF3 binding site on the 5ʹ-UTR of FTL, as determined by PAR-CLIP, spans a 24 nucleotide sequence that overlaps with the last five nucleotides of the annotated sequence of the FTL IRE (Figure 1—figure supplement 1). Notably, no eIF3 cross-linking site was observed in the 5ʹ-UTR of the mRNA encoding FTH1, which shares the structurally conserved IRE, but not adjacent sequence features (Figure 1—figure supplement 1B) (7). The removal of the eIF3 interaction site dramatically increased translation of rLuc when compared to the full length wild type FTL 5ʹ-UTR, 38-fold when the PAR-CLIP defined sequence was deleted (∆PAR, nucleotides 53–76) and six fold in a deletion that maintained the full IRE sequence (eIF3 repressive element, ∆3RE, nucleotides 58–90) (Figure 1D and E, Figure 1—figure supplement 1D) (Theil, 2015). These results suggest that eIF3 binding to the FTL 5ʹ-UTR represses FTL translation.

Decoupling the repressive role of eIF3 on FTL mRNA from that of IRP

Due to the close proximity between the eIF3 interaction site and the FTL IRE, accompanied by the fact that the 5ʹ-UTR of FTL is prone to large-scale structural rearrangements (Martin et al., 2012), we tested whether the derepression observed in the ∆PAR and ∆3RE mRNAs is a direct result of altering eIF3 binding and not due to disrupting the IRE-IRP interaction. To evaluate the effect of the deletions (∆PAR, ∆3RE) on IRP binding, we carried out RNA-electrophoretic mobility shift assays with near-IR-labeled FTL 5ʹ-UTR RNA and recombinant IRP1. As expected, IRP1 bound to the wild type 5ʹ-UTR of FTL (Figure 2A). IRP1 also bound the ∆3RE RNA, but failed to bind efficiently to the ∆PAR RNA (Figure 2A). The loss of IRP1 binding to ∆PAR RNA could be attributed to disrupted RNA folding, or to the importance of the region of overlap between the IRE and the PAR-CLIP defined eIF3-binding site (Figure 1—figure supplement 1C). We further quantified IRP binding to the ∆3RE 5ʹ-UTR using RNA binding competition assays (Figure 2B and C, Figure 2—figure supplement 1). IRP1 had only slightly attenuated binding to the ∆3RE 5ʹ-UTR RNA when compared to the wild-type FTL 5ʹ-UTR, suggesting that the alleviation of repression observed in the luciferase translation experiments for the ∆3RE mRNA (Figure 1E, Figure 1—figure supplement 1D) could be due to disruption of eIF3 binding.

Figure 2 with 3 supplements see all
Download asset Open asset
Maintenance of 5ʹ-cap and IRP-dependent regulation of 3RE deletions in the FTL 5ʹ-UTR.

(A) Representative native RNA gel shift showing recombinant IRP1 binding activity. Near IR (NIR) labeled RNAs corresponding to the full-length WT FTL 5ʹ-UTR and the FTL 5ʹ-UTR with deletions of the predicted eIF3 interaction site were incubated with recombinant IRP1 and resolved on a native polyacrylamide gel. (B) Dose-response curve of two RNA competition assays, based on gel shifts of NIR-labeled WT IRE-containing RNA, with WT or ∆3RE RNAs serving as cold competitors. Fold excess of competitors extended up to 75,000x. Error bars represent standard deviations for each concentration of competitor. (C) Calculated IC50 values using Prism 7 of the various competitor RNAs, based on the data in (B), with error bars representing the standard deviation from the mean IC50 value. N.A., the IC50 value could not be determined for the Loop mutant due to lack of any detectable competition. (D) Schematics showing the effect of increasing the distance of the IRE from the 5ʹ-cap on IRP regulation of translation initiation (Goossen and Hentze, 1992), (Muckenthaler et al., 1998). The characteristic C (C18 in the wild-type context) is denoted by an asterisk. (E) The luciferase activity of HepG2 cells transfected with mRNAs containing the native and extended spacing between the 5ʹ-cap and IRE, with or without the 3RE site, normalized to the luciferase luminescence of cells transfected with WT FTL mRNA with native spacing from the 5ʹ-cap. The values are from cells that have been harvested 6 hr post-transfection. The results are from three biological replicates, with error bars representing the standard deviation of the mean. (F) The luciferase activity of HepG2 cells transfected with mRNAs containing the native and various combinations of eIF3 (∆3RE) and IRP (Loop, A15G/G16C) disrupting mutations in HepG2 cells, normalized to the luciferase luminescence of cells transfected with WT FTL mRNA with native spacing from the 5ʹ-cap. Double represents an mRNA construct that contains both the Loop and ∆3RE mutations. The results are from six independent transfections, with error bars representing the standard deviation of the mean.

https://doi.org/10.7554/eLife.48193.006
Figure 2—source data 1

EMSA analysis and luciferase reporter readouts.

https://doi.org/10.7554/eLife.48193.011

To further ensure that the alleviation of repression of FTL translation by the ∆3RE mutation results primarily from disruption of eIF3-FTL binding and not IRE-IRP interactions, we modulated the location of the IRE in the 5ʹ-UTR of FTL. It had been shown previously that moving the IRE in the FTH1 5ʹ-UTR further than 60 nucleotides from the 5ʹ m7G-cap partially relieves IRP-dependent inhibition of FTH1 translation (Goossen and Hentze, 1992). Inhibition is lost because bound IRP can no longer fully sterically block the assembly of the 43S preinitiation complex on the mRNA (Muckenthaler et al., 1998),(Paraskeva et al., 1999). To further investigate if this holds for FTL mRNA, we used the FTL and ∆3RE luciferase reporter constructs and placed the characteristic C bulge of the IRE either 32 nucleotides away (native) or 70 nucleotides away (extended) from the 5ʹ-cap (Figure 2D). As with FTH1 mRNA (Goossen and Hentze, 1992), moving the IRE further from the 5ʹ-cap partially derepressed translation of the FTL-based luciferase reporter (Figure 2E). Notably, overall translation was much higher from both ∆3RE mRNAs, even with partial removal of IRP-dependent repression due to the distance from the 5ʹ-cap (Figure 2E). This cap position-independent derepression is unique to FTL as the 3RE region is not conserved in the FTH1 5’-UTR. Furthermore, combining the ∆3RE mutation with a mutation that disrupts IRP binding to the IRE entirely (Loop) (Cazzola et al., 1997) synergistically derepressed luciferase reporter translation (Figure 2F, compare Double to Loop and ∆3RE). We verified the mRNAs were equally stable during the 6 hr time courses (Figure 2—figure supplement 2), ensuring the observed phenotypic changes are caused by changes in translational regulation. The slight inconsistency in the amount of derepression observed with the double mutation (Figure 2F) and the extended ∆3RE mRNA (Figure 2E) may be due to the fact that the extension of the 5’-UTR from its native state does not completely abolish the repression of FTL mRNA translation by IRP, in contrast to the loop mutation which abolishes IRP binding (Cazzola et al., 1997). Furthermore, it is not clear how the distance from the 5’-cap affects repression mediated by eIF3. Taken together, these results indicate that IRP-mediated inhibition of translation is maintained in the ∆3RE mRNA, and that eIF3 confers an additional level of repression beyond that which can be provided by IRP.

We used the luciferase reporter results in Figure 2F to formulate a mathematical model to determine whether eIF3 and IRP can bind and regulate FTL mRNA simultaneously (See Materials and methods for details). Such a model would be useful in conditions with different extents of iron-response regulation. We defined a system in which IRP and eIF3 do not bind the same mRNA. This leaves three possible states in which the FTL mRNA could exist: the fraction bound solely by IRP (x1), the fraction bound solely by eIF3 (x2), and the remainder of the mRNA which is unbound by either factor (x3) (Figure 2—figure supplement 3). We further elaborated this model to include translation efficiency (y). Here we assume the mutations do not affect the translation efficiency of the unbound species (x3), while the other two populations (x1 and x2) have translational efficiencies (y1 and y2) scaled between full repression (yn = 0) and no repression (yn = 1). Lastly, we assumed the mutations that disrupt binding shift the equilibrium of the total mRNA population between bound and unbound fractions of the alternate factor by some amount, either α for those affecting eIF3 binding or β for IRP binding. Using the translational output determined in Figure 2F, we find that the data are inconsistent with the model that eIF3 and IRP cannot bind the same mRNA. Rather, the data indicate that IRP and eIF3 likely act in cis on at least a fraction of the FTL mRNA.

Physiological response to loss of eIF3-dependent repression

To investigate the physiological response to the loss of eIF3-based repression, we genetically engineered either HEK293T or HepG2 cells to generate the ∆3RE mutation in the 5ʹ-UTR in the FTL gene (Figure 3—figure supplement 1A–1C). Notably, we found that the ∆3RE mutation abolished the preferential interaction of eIF3 with FTL mRNA (Figure 3E,F). Furthermore, FTL protein production increased dramatically in the ∆3RE cell lines, as expected from removing the predicted eIF3 repressive element (Figure 3A, Figure 3—figure supplements 1D, 4A and 5A). Importantly, the increase in FTL protein levels was not due to increases in mRNA levels (Figure 3—figure supplement 2A), indicating that the de-repression occurs post-transcriptionally. Interestingly, the increase in FTL levels occurs with a concurrent reduction in FTH levels (Figure 3B, Figure 3—figure supplements 4C and 5A). The decrease in FTH protein levels is also not due to changes in FTH1 mRNA levels (Figure 3—figure supplement 2B).

Figure 3 with 6 supplements see all
Download asset Open asset
Physiological effects of the endogenous removal of the 3RE repressive element.

(A,B) Representative western blots of (A) FTL and (B) FTH levels in the edited (∆3RE) and WT HepG2 cells under normal iron conditions. Serial dilutions were used in order to better visualize the significance of the changes in FTL and FTH protein abundance. (C, D) Representative western blots of FTL levels in the edited (∆3RE) and WT HepG2 cells under high- or low-iron conditions. Iron donor treatment with FAC at 50 μg/mL for 24 hr, and Iron chelation treatment with DFO at 50 μM for 48 hr. The asterisk (*) indicates that lysate from ∆3RE cells were diluted two-fold, due to the higher overall levels of FTL in these cells. All FTL blots are representative of three or more replicates. (E, F) Determination of the preferential binding of eIF3 to FTL mRNA via EIF3B immunoprecipitation (IP) followed by RNA extraction and RT-qPCR in both (E) HEK293T and (F) HepG2 cell lines. Control mRNAs used to normalize IPs were PSMB6 and ACTB. Error bars represent the standard deviation of duplicate qPCR measurements from representative IP reactions. (G) Determination of IRP1 binding to FTL mRNA in WT (HEK + IRP) and ∆3RE (∆3RE + IRP) HEK293T cells via FLAG immunoprecipitation (IP) followed by RNA extraction and RT-qPCR. The ACTB mRNA was used to control for nonspecific binding to FLAG-tagged IRP. HEK – IRP reflects cells that were not transiently transfected, but were carried through the IP and following experiments. Error bars represent the standard deviation for triplicate measurements from representative IP reactions.

https://doi.org/10.7554/eLife.48193.012
Figure 3—source data 1

Data anlysis of eIF3B and FLAG-tagged IRP1 immunoprecipitations.

https://doi.org/10.7554/eLife.48193.022

To test whether IRP maintains its ability to dynamically regulate FTL translation, we treated the cell lines with either ferric ammonium citrate (FAC), an iron donor, or desferoxamine (DFO), an iron chelator, to increase or decrease iron levels, respectively. (Figure 3—figure supplement 3, Figure 3—figure supplement 4B and D, Figure 3—figure supplement 5A and B) (Schneider and Leibold, 2003). FTL levels in the ∆3RE cell lines responded to FAC and DFO treatment in a comparable manner to the unedited (WT) cell lines (Figure 3C–D, Figure 3—figure supplement 4B and D, Figure 3—figure supplement 5A and B), showing that the ∆3RE transcript retains IRP-dependent translational regulation in cells. This iron-responsive regulation is maintained even though the basal levels of FTL protein are much higher in the ∆3RE compared to WT cells (Figure 3D). FTH levels also respond to iron levels in both the WT and ∆3RE cells (Figure 3—figure supplement 3, Figure 3—figure supplement 4D, Figure 3—figure supplement 5B), with basal FTH levels reduced in the ∆3RE compared to WT cells.

To further ensure that IRP-mediated repression was maintained in the ∆3RE cell lines, we transiently transfected the WT and ∆3RE cell lines with plasmids encoding C-terminally FLAG-tagged IRP1. We then used FLAG immunoprecipitation followed by qPCR to determine whether IRP is bound to the edited FTL and other IRE containing mRNAs in vivo. We found that FLAG-tagged IRP bound FTL mRNA similarly in the wild type and ∆3RE cell line (Figure 3G). Intriguingly, FTH1 mRNA was recovered at a considerably higher level in the ∆3RE cell line compared to wild type cells (Figure 3G), suggesting increased IRP binding. This increased binding of IRP to FTH1 mRNA may explain the concurrent decrease in FTH abundance seen in Figure 3B. Notably, this increase in IRP binding to FTH1 mRNA does not appear to be a simple mass action effect, as ferroportin levels–also regulated by an IRP-IRE interaction (Anderson and McLaren, 2012)–are not altered in the ∆3RE cell line (Figure 3—figure supplement 5C). Taken together, these data further support the hypothesis that the observed de-repression in the ∆3RE cells is due to the modulation of eIF3-based regulation of FTL translation, and not due to disruption of IRP-mediated regulation.

We proceeded to investigate whether the lack of eIF3-based repression of FTL translation and concomitant decrease in FTH protein levels had any effects on the assembled ferritin complexes. We purified the ferritin complexes from either wild-type HepG2 cells or the ∆3RE cell line using a fractional methanol precipitation protocol (Cham et al., 1985). We observed that the ferritin complex in the ∆3RE cell line was far more stable than that isolated from wild-type cells. Without ferric ammonium citrate (FAC) treatment to stabilize the complex (Linder, 2013), ferritin purified from WT cells consistently degraded, unlike the stable complexes from the ∆3RE cell line (Figure 3—figure supplement 6). This implicates eIF3 in regulating the dynamics and stability of the ferritin complex, as FTL-enriched ferritin complexes have been shown to be more stable under a wide array of denaturing condition (Santambrogio et al., 1992). Taken together, these results support the hypothesis that removal of the eIF3 interaction site in the 5ʹ-UTR of the FTL mRNA derepresses FTL translation and can have a dramatic effect on ferritin subunit composition in the cell.

SNPs in FTL that cause hyperferritinemia

Although the ∆3RE mutation in FTL revealed eIF3-dependent repression of FTL translation, it is not clear what role eIF3 may play in ferritin homeostasis in humans. The human genetic disease hereditary hyperferritinemia cataract syndrome (HHCS) is an autosomal dominant condition that primarily results in early onset of cataracts due to ferritin amassing in the lens (Millonig et al., 2010). HHCS arises from SNPs or deletions in the 5ʹ-UTR of FTL, which are thought to disrupt the IRP-IRE interaction leading to increased FTL translation. For example, mutations observed in the apical loop of the IRE directly disrupt critical contacts essential for IRP binding to the IRE (Figure 4A) (5). Interestingly, two SNPs identified in certain patients with hyperferritinemia, G51C and G52C, disrupt the nucleotides one and two bases upstream of the annotated eIF3 PAR-CLIP site (Figure 4A) (Camaschella et al., 2000),(Luscieti et al., 2013). Although the PAR-CLIP methodology maps the region of interaction between an RNA and protein of interest, it does not always capture the full interaction site due to the requirement for 4-thiouridine cross-linking and subsequent RNase digestion to generate fragments for deep sequencing (Ascano et al., 2012), (Hafner et al., 2010). Thus, we wondered whether the G51C and G52C SNPs could potentially impact eIF3 repression of FTL mRNA translation, due to the proximity of the G51C and G52C mutations to the eIF3 PAR-CLIP site. We generated luciferase reporter constructs with either of the G51C and G52C SNPs, as well as a control with the previously described mutations in the IRE apical loop, and used mRNA transfections to test their effects on luciferase translation levels. All mutations led to an increase in luciferase levels, indicating that they alleviated translational repression (Figure 4B). We also observed that the G51C and G52C mRNAs do not interact with eIF3, based on eIF3 immunoprecipitations from transfected HEK293T cells (Figure 4C). Furthermore, the G51C and G52C SNPs maintained near wild-type IRP binding (Figure 4D and E), in stark contrast to the mutations in the IRE apical loop (Figure 4A), which completely abolished the interaction between IRP1 and the 5ʹ-UTR element (Figure 4—figure supplement 1). Furthermore, using stable cell lines, we observe iron-responsive regulation of luciferase reporter mRNAs with G51C or G52C SNP-containing FTL 5ʹ -UTRs, as well as IRP binding to these mRNAs (Figure 4—figure supplement 2). These results identify SNPs in FTL that cause hyperferritinemia likely due to disruption of eIF3-dependent repression of FTL translation.

Figure 4 with 2 supplements see all
Download asset Open asset
Role of eIF3 in select cases of hyperferritinemia.

(A) Diagram of IRP binding to the IRE in FTL mRNA (Anderson and McLaren, 2012). Hyperferritinemia mutations are highlighted in orange (Cazzola et al., 1997), green (Camaschella et al., 2000), and blue (Luscieti et al., 2013) with their corresponding serum ferritin levels listed. Normal serum ferritin levels are under 300 µg/L. The 3RE is indicated in purple. Nucleotides that directly interact with the IRP are also identified (i.e. A15, G16, U17). (B) Luciferase activity of HepG2 cells transfected with mRNAs encoding the WT FTL 5ʹ-UTR or various hyperferritinemia mutations (G51C, G52C, or Loop mutant (A15G/G16C)), normalized to WT FTL reporter luciferase luminescence. The results are from three biological replicates, with error bars representing the standard deviation of the mean. (C) Binding of eIF3 to luciferase reporter mRNAs with WT or mutant forms of the FTL 5ʹ-UTR, using EIF3B immunoprecipitation (IP), followed by RNA extraction and RT-qPCR. Cells were harvested 8 hr post-transfection. Data are shown as the percent in the IP, compared to input levels. Error bars are the standard deviation of the mean of duplicate qPCR measurements from a representative IP. (D) Dose-response curve of RNA competition assays, based on gel shifts of NIR-labeled WT IRE-containing RNA, with WT, G51C, G52C, or Loop mutant (A15G/G16C) RNAs serving as cold competitors. Fold excess of competitor WT extended up to 100,000x. Recombinant IRP1 was used. Error bars represent standard deviations for each concentration of competitor. (E) The calculated IC50 values of the various competitor RNAs, based on the data in (D), with error bars representing the standard deviation from the mean IC50 value. N.A., the IC50 value could not be determined for the Loop mutant due to lack of any detectable competition. Note that the data for ∆3RE in panels (D) and (E) are from Figure 2B and C, measured in duplicate. The remaining experiments in (D) and (E) were carried out in triplicate.

https://doi.org/10.7554/eLife.48193.023
Figure 4—source data 1

Luciferase reporter readouts, eIF3B Immunoprecipitation quantification, and EMSA analysis for hyperferritinemia mutants.

https://doi.org/10.7554/eLife.48193.027

Discussion

We have shown that eIF3 represses the translation of FTL mRNA by binding a region of the FTL 5ʹ-UTR immediately adjacent to the IRE. Upon disruption of eIF3 binding, the basal level of FTL production increases dramatically without affecting the iron-responsiveness of FTL translation (Figure 3), or binding of IRP to the remainder of the 5ʹ-UTR containing the IRE (Figure 2, Figure 3, Figure 4). Taken together, these results expand the classical model of FTL mRNA post-transcriptional regulation by the iron-responsive IRP/IRE interaction to include a functionally distinct eIF3-dependent repressive mechanism (Figure 5). The physiological need for a dual repressive system involving IRPs and eIF3 in normal human health remains to be determined. Experiments combining the ∆3RE and apical loop mutations (Figure 2F, Figure 2—figure supplement 3) suggest that eIF3 and IRP can act simultaneously to repress FTL translation. Due to the close proximity of the IRE and eIF3 binding sites, and as suggested by our mathematical modeling (Figure 2—figure supplement 3), it may be possible that eIF3 physically interacts with IRP when they bind to the 5ʹ-UTR in FTL mRNA in certain tissue or cellular contexts. Although we have identified a role for eIF3 in FTL mRNA translation regulation, it is still unclear what role eIF3 may play in response to iron level modulation, a key question to answer in the future. It is possible that both eIF3 and IRP are required for the proper iron responsiveness of FTL translation.

Figure 5
Download asset Open asset
Model of post-transcriptional regulation of FTL mRNA.

IRPs repress FTL mRNA translation in an iron-dependent manner, whereas eIF3 represses FTL translation in an iron-independent manner. Coordination between IRP repression and eIF3 repression may differ by cell and tissue context. Various hyperferritinemia mutations (bolded) listed in the literature are mapped on the experimentall -determined secondary structure of the FTL mRNA 5ʹ-UTR (Martin et al., 2012), (Luscieti et al., 2013). The minimal annotation of the IRE is denoted by with a blue outline and the eIF3 PAR-CLIP defined interaction site is denoted with a purple outline. Mutations that disrupt IRP binding used in this study and determined here to disrupt eIF3 binding are bolded in blue and purple, respectively. (*) indicates nucleotides identified as ribosnitches (Martin et al., 2012).

https://doi.org/10.7554/eLife.48193.028

Our present findings also provide the first molecular evidence for the direct involvement of eIF3 in a human disease caused by SNPs in the human population. We found that disruption of eIF3-mediated regulation of FTL translation could serve as the dominant cause of certain cases of hyperferritinemia. The specific genetic mutations we analyzed (G51C, G52C) map to the less-conserved lower stem of the IRE, and are predicted to have no direct physical interaction with IRP1 (PDB: 3SNP) (Walden et al., 2006). Here, we establish that these mutations minimally interfere with IRP’s interaction with the IRE, whereas they greatly impact the eIF3-based interaction and FTL translational repression (Figure 4C–E). This work highlights how even highly clustered SNPs can contribute to disease through divergent molecular mechanisms. In the case of FTL, these clustered SNPs can disrupt three different aspects of translation regulation: IRP binding, eIF3 binding, or RNA folding (Figure 5).

While our results connect eIF3 translational control to specific examples of hyperferritinemia, they also suggest a broader role for eIF3 in other hyperferritinemias and ferritin-related diseases. For example, we observed that derepression of FTL translation by disruption of the eIF3 interaction site leads to a concomitant decrease in FTH levels, driving an overall increase in ferritin complex stability (Figure 3—figure supplement 6). Ferritin is a known mediator of inflammatory responses, raising the question of whether eIF3 may contribute to ferritin’s role in inflammation (Morikawa et al., 1995),(Recalcati et al., 2008). Our results provide new insights that should help connect molecular mechanisms of translational control to disease-associated SNPs identified in ever expanding genomic databases.

Materials and methods

Plasmids

The FTL 5ʹ-UTR was amplified from human cDNA, and cloned into the pcDNA4 vector with a modified Kozak sequence (Kranzusch et al., 2014), by using In-Fusion HD Cloning Kit (Takara, Cat.# 638911) to generate the starting luciferase reporter plasmids. The FTL transcription start site is derived from the FANTOM5 database (Lizio et al., 2015). Additional mutations were generated through around-the-horn cloning using either blunt primers for deleting regions or primers with overhangs to introduce single or double nucleotide mutations. The IRP1 protein expression plasmid was generated by amplifying the human IRP1 sequence from human cDNA and inserting it into the 2B-T vector (Addgene, plasmid # 29666) following a His6 tag and TEV protease cleavage site. The IRP-FLAG construct was generated by inserting a 1X FLAG tag at the C-terminal end of IRP in the pCMV6-XL4 backbone (OriGENE, SC126974).

In vitro transcription

Request a detailed protocol

RNAs were transcribed using T7 RNA polymerase prepared in-house. For luciferase reporter mRNAs, 5ʹ-capping and 3ʹ-polyadenylation were performed co-transcriptionally by including 3´-O-Me-m 7G(5´)ppp(5´)G RNA Cap Structure Analog (NEB, Cat.# S1411L) and using linearized plasmid template that had a sequence encoding a poly-A tail. Non-labeled RNAs for the IRP1 electrophoresis mobility shift assays were generated in the same manner, except the templates were not polyadenylated. Additionally, RNAs with the 38-nucleotide extension between the 5ʹ -cap and IRE were constructed using a random nucleotide sequence. The exact nucleotide composition 5ʹ of the IRE was previously reported to not significantly impact IRP binding (Goossen and Hentze, 1992). RNAs were purified after DNA template digestion by phenol-chloroform extraction and ethanol precipitation.

For genome editing, we used tandem CRISPR-Cas9 enzymes programmed with single-guide RNAs (sgRNAs) targeting the FTL gene, along with a single-stranded DNA (ssDNA) oligonucleotide homologous to the regions spanning the deleted 3RE sequence (Figure 3—figure supplement 1). sgRNAs were designed using the CRISPR.MIT.EDU program from the Feng Zhang Lab, MIT. CRISPR-Cas9-sgRNA was assembled as RNA-protein complexes (RNPs) (Kim et al., 2014). The DNA for transcription was synthesized by appending the sgRNA sequence downstream of a T7 RNA polymerase promoter. The DNA was then purified using phenol-chloroform extraction followed by isopropanol precipitation. After transcription, the RNA products were treated with DNase I (Promega, Cat.# M6101), run on a 10% denaturing polyacrylamide gel (6 M urea), and extracted from the gel using the crush and soak method and ethanol precipitation.

Luciferase reporter transfections

Request a detailed protocol

Human HepG2 cells were maintained in DMEM (Invitrogen 11995–073) with 10% FBS (Seradigm) and 1% Pen/Strep (Gibco, Cat.# 15140122). Transfections of the luciferase reporter mRNAs were done using the Mirus TransIT-mRNA Transfection Kit (Cat.# MIR 2250), with the following protocol modifications. The day prior to transfection, HepG2 cells were seeded into opaque 96-well plates so that they would reach 80% confluence at the time of transfection. At this point, 200 ng of 5ʹ-capped and 3ʹ-polyadenylated mRNA was added at room temperature to OptiMEM media (Invitrogen, Cat.# 31985–088) to a final volume of 10 µL. All mRNA concentrations were determined using a Nanodrop, and were validated by running samples on an agarose gel. Bands on the gel were quantified to properly ensure all constructs were transfected at equal amounts. Then, 0.6 µL of both Boost reagent and TransIT mRNA reagent were added to the reaction and left to incubate at room temperature for 3 min. The TransIT-mRNA Reagent:mRNA Boost:RNA complex was distributed to the cells in a drop wise manner. Luciferase activity was assayed either 6 hr, or 12 hr post-transfection using the Renilla Luciferase assay kit (Promega, Cat.# E2820) and a Microplate Luminometer (Veritas). We used 6 hr incubation times for these experiments in order to keep readouts for various mutant constructs in a linear range (Bert et al., 2006). The only transfection not shown at 6 hr (Figure 4B) was also carried out for 6 hr in Figure 2—figure supplement 3. In all experiments, we define biological replicates as cells cultured in separate wells, and technical replicates as multiple measurements from cells from the same well.

In order to monitor the stability of transfected mRNAs during the timecourse of the luciferase reporter experiments, these 750 ng of each mRNA was transfected into HEK293T cells at 80% confluency well in a 24-well plate. After 6 hr, the cells were collected, pelleted, and then the RNA was extracted using the Direct-zol RNA Mini prep kit (R2051). cDNA was the prepared using 250 ng of RNA and Superscript IV (Thermo Fisher scientific, Cat. # 18091050). qPCR was carried out with the primers to the luciferase coding sequence with run conditions as described below in the methods for determination of transcript level abundance. We note that a completely accurate quantification of accessible mRNAs in the cell is limited due to technical complications inherent with mRNA transfection protocols and mechanisms (Kirschman et al., 2017).

Cell line generation

Request a detailed protocol

Cell lines (HEK293T and HepG2) were obtained from the University of California Berkeley Cell Culture Facility, validated by STR analysis, and confirmed to be mycoplasma-free by the facility. To generate FTL-eIF3 interaction site null cell lines (∆3RE), we used tandem CRISPR-Cas9 enzymes programmed with single-guide RNAs (sgRNAs) targeting the FTL gene, along with a single-stranded DNA (ssDNA) donor with homology to the regions spanning the deleted ∆3RE sequence. The sgRNAs were generated as described above, and targeted regions on both sides of the eIF3 interaction site (Figure 2A). The RNP complex was generated by incubating 100 pmol Cas9 with the two sgRNAs at a 1:1.2 Cas9 to total sgRNA ratio. This mixture was heated to 37°C for 10 min and then kept at room temperature until use. The ssDNA donor was 90 nucleotides long, with 45-nucleotide homology on either side of the predicted double-strand cut sites allowing it to have perfect homology to the predicted edited sequence.

The Cas9-sgRNA RNP complexes, along with 500 pmol of the ssDNA donor, were transfected into either 5 × 105 HEK293T or HepG2 cells using the Lonza 96-well shuttle system SF kit (Cat. # V4SC-2096). The nucleofection programs used were as follows: CM-130 for HEK293T and EH-100 for HepG2. The transfected cells were left to incubate for either 48 hr for HEK293T cells or 72 hr for HepG2 cells before harvesting and extracting gDNA using QuickExtract (Epicentre: QE09060). The editing efficiency of individual Cas9-sgRNA RNPs was determined using a T7 endonuclease one assay (Reyon et al., 2012). The efficiency of the dual-sgRNA editing approach was determined by PCR-amplifying a 180-base pair region around the eIF3 interaction site, and analyzing the resulting products on a 6% non-denaturing 29:1 polyacrylamide gel. This method achieved an editing efficiency of nearly 100% in HEK293T cells and roughly 85% in HepG2 cells (Figure 3—figure supplement 1B). Monoclonal populations of edited cells were sorted using FACS, screened, and the final edited sequence was determined using TOPO TA cloning (Ramlee et al., 2015).

Transcript level abundance

Request a detailed protocol

Total transcript abundance was determined by lysing 1.25 × 106 cells with Qiazol lysis buffer followed by using the Directzol RNA extraction kit (Zymo Research, Cat. # R2061), according to the manufacturer's instructions. The cDNA was generated by reverse transcription using 350 ng of RNA, random hexamers, and Superscript IV (Thermo Fisher scientific, Cat. # 18091050). Primers for the qPCR were as follows: FTL forward: 5ʹ -ACCTCTCTCTGGGCTTCTAT-3ʹ, FTL reverse: 5ʹ -AGCTGGCTTCTTGATGTCCT-3ʹ (Cozzi et al., 2004), ACTB forward: 5ʹ-CTCTTCCAGCCTTCCTTCCT-3ʹ, ACTB reverse: 5ʹ-AGCACTGTGTTGGCGTACAG-3ʹ (Chen et al., 2008), PSMB6 forward: 5ʹ-GGACTCCAGAACAACCACTG-3ʹ, PSMB6 reverse: 5ʹ-CAGCTGAGCCTGAGCGACA-3ʹ (Mokany et al., 2013), FTH1 forward: 5ʹ-CGCCAGAACTACCACCA-3ʹ, FTH1 reverse: 5ʹ-TTCAAAGCCACATCATCG-3ʹ (Liu et al., 2013), 18S forward: 5ʹ-GGCCCTGTAATTGGAATGAGTC-3ʹ, 18S reverse: 5ʹ-CCAAGATCCAACTACGAGCTT-3ʹ (Lee et al., 2016), RLUC forward: 5’-GGAATTATAATGCTTATCTACGTGC-3ʹ, RLUC reverse: 5ʹ-CTTGCGAAAAATGAAGACCTTTTAC-3ʹ (Kong et al., 2008). Run conditions were: 95 °C for 15 s, followed by 40 cycles of 95 °C for 15 s, 60 °C for 60 s, 95 °C for 1 s.

RNA immunoprecipitation and qPCR

Request a detailed protocol

The EIF3B-RNA immunoprecipitations were adapted from Ramlee et al. (2015) with the following modifications. One 15 cm plate of either HEK293 or HepG2 cells was used to prepare cell lysate using a NP40 lysis buffer (50 mM HEPES-KOH pH 7.5, 150 mM KCl, 2 mM EDTA, 0.5% Nonidet P-40 alternative, 0.5 mM DTT, 1 Complete EDTA-free Proteinase Inhibitor Cocktail tablet per 50 mL of buffer). The lysate was precleared with 15 µL of Dynabeads preloaded with rabbit IgG (Cell Signaling, Cat. # 2729) for one hour at 4 °C. The lysate was collected and then incubated with a fresh 15 µL aliquot of Dynabeads and 7.5 µL of anti-EIF3B antibody (Bethyl A301-761A) for two hours at 4 °C. Preparation of cDNA and qPCR primers are described and listed above.

For EIF3B immunoprecipitations of transfected mRNAs, 2.15 µg of mRNA was transfected into 1 well of either HEK293T or HepG2 cells in a 12-well plate using the protocol above. Cells were then left to incubate for 8 hr before harvesting. The cells were lysed using the NP40 lysis buffer listed above (Lee et al., 2015), and precleared with 2 µL of rabbit IgG-coated Dynabeads for one hour at 4 °C. The lysate was collected and then incubated with a fresh 2 µL aliquot of Dynabeads and 4 µL of anti-EIF3B antibody (Bethyl A301-761A) for 2 hr at 4 °C. RNA was collected, cDNA prepared and qPCR carried out with the primers and run conditions as described in the methods for transcript level abundance.

For FLAG-tagged IRP1 immunoprecipitations, 2.2 µg of plasmid DNA was transfected into a 10 cm dish of 80% confluent HEK239T WT cells or HEK293T ∆3RE mutant cells. Cells were then left to incubate for 24 hr before harvesting. The cells were lysed using the NP40 lysis buffer as described in Ramlee et al. (2015) and then further diluted 3x with the lysis buffer lacking DTT. The lysate was collected and then incubated with pre-equilibrated anti-FLAG antibody conjugated agarose beads (Sigma A2220) for two hours at 4 °C. The beads were then washed with a high salt wash buffer (50 mM HEPES-KOH pH 7.5, 300 mM KCl, 2 mM EDTA, 1% Nonidet P-40 alternative) three times. The protein was eluted from the beads using two washes with 1X FLAG peptide for 30 min each at 4 °C. The RNA was collected using phenol/chloroform extraction followed by ethanol precipitation. cDNA was prepared using Superscript IV (Thermo Fisher scientific, Cat. # 18091050) and qPCR carried out with the primers and run conditions as described above in the methods for determination of transcript level abundance.

Iron level modulation

Request a detailed protocol

In order to modulate the iron levels in cells, HepG2 and HEK293T cells were treated with a final concentration of either 50 µg/mL of ferric ammonium citrate (FAC) (an iron donor) for 24 hr or either 200 µM for 24 hr or 50 µM for 48 hr of desferoxamin (DFO) (an iron chelator) before harvesting.

For the iron treatment of cells with the stably integrated luciferase reporters harboring the FTL 5’UTR SNPs, either H2O, 200 µM of DFO, or 50 ug/mL of FAC was added to an individual 96 well of 80% confluent cells. The cells were allowed to incubate for 24 hr before taking the luminescence reading. BioRender was used for the cell schematic in Figure 4—figure supplement 2A.

IRP1 purification

Request a detailed protocol

IRP1 was purified based on the protocols in Carvalho and Meneghini (2008), Basilion et al. (1994) with modifications. The IRP1-encoding 2B-T plasmid was transformed into chemically-competent BL21 Rosetta pLysS E. coli, using heat shock at 42 °C, and grown on Ampicillin plates. A single colony was used to inoculate a 5 mL LB culture containing Ampicillin, which was then used to inoculate a 50 mL starter culture that was allowed to reach saturation overnight. Approximately 4 × 10 mL of the overnight culture was used to inoculate 4 × 1L cultures using ZY5052 media lacking the 1000x trace metal mix (30 mM HEPES, 5% glycerol, 43 mM KCl, 0.5 mM EDTA, 0.5 mM DTT) (Studier, 2005) plus Carbomicillin. The 1 L cultures were grown at 37 °C to OD600 = 0.36, at which point the temperature was lowered to 18 °C, and allowed to grow at 18 °C for 36 hr prior to harvest.

Pelleted E. coli cells were lysed using sonication in lysis buffer (30 mM HEPES pH = 7.5, 400 mM KCl, 5% Glycerol and 1 mM DTT) along with Protease inhibitor (Roche, Cat. # 5056489001) tablets. Lysate was loaded on a 5 mL Ni-NTA pre-packed HiTrap column (GE, Cat. # 17-5248-02), allowed to incubate at 4 °C for 1 hr, before eluting using 600 mM imidazole in the same buffer as above. Pooled fractions from the elution were then dialyzed overnight into ion-exchange (IEX) buffer (30 mM HEPES pH 7.5, 1 mM DTT, 5% Glycerol, and 1 mM EDTA), for subsequent purification using a 5 mL HiTrap Q-column (GE, Cat. # 17-1154-01). Samples were then loaded on a Q column using IEX buffer, and the column was washed with eight column volumes of IEX buffer without KCl. IRP1 was eluted using 800 mM KCl in IEX buffer.

IRP1 electrophoresis mobility shift assays

Request a detailed protocol

To detect IRP1 binding by native gel shifts, RNA samples were transcribed using the Atto-680 RNA Labeling Kit (Jena Bioscience, FP-220–680) and subsequently purified using RNA Clean and Concentrator-25 columns (Zymo, R1018). This form of labeling has been shown not to disrupt protein-RNA interactions (Köhn et al., 2010). Unlabeled RNA was transcribed and purified as described above.

Binding experiments were carried out with a final concentration of 300 pM of labeled RNA and 225 nM of recombinant human IRP1, which facilitated a 1:1 ratio of RNA binding to IRP1. We first ensured the RNA competition experiments reached equilibrium – which required at least 11 hr of incubation – by measuring the approximate dissociation rate constant (koff) of WT FTL 5ʹ-UTR from IRP1. Heparin was included at the beginning of the reaction at a final concentration of 4.5 µg/mL. The initial binding reaction was carried out in a 1x RXN buffer (40 mM KCl, 20 mM Tris-HCL, pH 7.4, 2 mM MgCl2, 2 mM DTT, and 5% Glycerol) for 30 min at room temperature (Goforth et al., 2010),(Fillebeen et al., 2014). For competition experiments, unlabeled RNA was then added in concentrations 1000x-100,000x that of the labeled RNA. In preliminary experiments, we found the koff to be roughly 0.006 min−1 using an 8 hr incubation time course with competitor. We then tested the fraction of IRP bound after 11 hr and 18 hr incubations with competitor FTL RNA and observed no changes in the residual fraction of IRP bound to RNA (~15%), indicating the reactions had reached equilibrium. Thus, subsequent experiments with competitor RNAs were carried out for 18 hr, after the first 30 min pre-incubation in the absence of competitor. The reactions were resolved on Tris-glycine gels (Thermo Fisher Scientific, Cat.# XP04122BOX) and gels was visualized using an Odyssey Licor set to 700 nM wavelength and an intensity of 6.5. Band intensity quantification was carried out using Image Studio (Licor). The IC50 values for each competitor RNA was determined using Graph Pad Prism 7 (Graph Pad Software) from a set of triplicate experiments, except for the ∆3RE competitor RNA, which was tested in duplicate.

Ferritin complex purification

Request a detailed protocol

The ferritin complex purification procedure was adapted from Cham et al. (1985), with slight modifications. Either one 15 cm dish of ∆3RE cells grown in normal media or eight 15 cm dishes of wild-type HepG2 cells that had been treated with 50 ng/mL FAC for 24 hr were harvested, weighed, and lysed in 4x weight/volume NP40 lysis buffer (50 mM HEPES-KOH pH = 7.5, 150 mM KCl, 2 mM EDTA, 0.5% Nonidet P-40 alternative, 0.5 mM DTT, 1 Complete EDTA-free Protease Inhibitor Cocktail tablet per 10 mL of buffer). Samples were incubated on ice for ten minutes and then centrifuged for 10 min at 21,000xg at 4 °C. Samples were diluted in 1:2 phosphate buffered saline (PBS), and methanol was added to the diluted lysate to a final concentration of 40% (v/v). The sample was then heated to 75 °C for 10 min. After cooling on ice for 5 min, the samples were centrifuged in a microfuge 20R at 1251xg RPM for 15 min at 4 °C. The resulting supernatant was collected and concentrated using a 100 k MW cutoff Amicon filter (Cat.# UFC510024) by centrifugation for 10 min at 14000xg. The sample was washed once with PBS and spun for an additional 4 min at 14000xg. The sample was then collected by inverting the column and centrifuging the sample for 2 min at 1000xg at 4 °C. All samples collected were brought to a final volume of 80 µL with PBS. The purity of the sample was determined by running the sample on a native gel (Thermo Fisher Scientific, Cat.# XP04200BOX) followed by Coomassie staining.

Western blots

Request a detailed protocol

The following antibodies were used for Western blot analysis: anti-EIF3B (Bethyl A301-761A) at 1:1000; anti-FTL (Abcam, ab69090) at 1:800; anti-FTH (Santa Cruz, sc-25617) at 1:400; anti-IRP1 (Abbexa, abx004618) at 1:400; anti-IRP2 (Abcam, ab129069) at 1:800; and anti-b-Actin (Abcam, ab8227) at 1:1000, anti-Ferroportin (Novus, NBP 1–2150255) at 1:300.

Lentiviral transduction

Request a detailed protocol

The G51C and G52C Renilla luciferase reporters were cloned into the NLV103 plasmid. Virus was generated using LentiX cells in a 10 cm dish format and TransIT-LT1 transfection reagent. Virus was harvested and filtered after 48 and 72 hr. Total virus was pooled and 500 µL of fresh virus was added to 106 HEK293T cells along with 10 µg/µL of polybrene (Millipore). Cells were left to incubate for 48 hr before a 4 day selection process with 4 µg/mL of puromycin. Cells were split in non-selective media once before use.

Mathematical modeling of IRP and eIF3 co-occupancy on FTL mRNA

Request a detailed protocol

We tested a mathematical model in which IRP and eIF3 do not bind simultaneously to the same FTL mRNA. This model thus includes three possible states of the FTL mRNA: the fraction bound solely by IRP (x1), the fraction bound solely by eIF3 (x2), and the remainder of the mRNA, which is unbound by either factor (x3) (Figure 2 – figure supplement 2). The model also assumes that the mutations introduced into FTL mRNA do not affect the translational efficiency of unbound mRNA species, that is translation of x3 is identical for all 4 mRNAs. The translational efficiency of IRP-bound mRNA (y1) and eIF3-bound mRNA (y2) ranges from fully repressed (y = 0) to completely unbound and derepressed (y = 1). Thus, the translational efficiency for the bound populations (x1 and x2) is less than 1, while the translational efficiency of x3 is equal to 1. We also include the parameters α and β to account for changes in the distribution, or shifts, of previously bound mRNA (x2 and x1) to new populations (0 ≤ α + β ≤ 1). Taken together, four equations present the luciferase output of each FTL mRNA:

(1) FTL=y1x1+y2x2+x3
(2) Δ3RE=y1(x1+α)+x2−α+x3
(3) Loop=x1−β+y2(x2+β)+x3
(4) Double=x1+x2+x3

Here, FTL represents the luciferase readout of wild-type FTL mRNA; ∆3RE the luciferase readout of FTL mRNA with the ∆3RE mutation; Loop, the luciferase readout of FTL mRNA with the IRE loop mutation; and Double the luciferase readout of FTL mRNA with both the ∆3RE and Loop mutations. In order to solve this system of equations, we proceeded to use experimentally determined values as seen in Figure 2F. We assume the ∆3RE and Loop mutations disrupt regulation by the respective factor (IRP or eIF3) completely, consistent with the biochemical results in Figure 4. In these cases, both y1 and y2 revert to a value of 1, that is the same translational efficiency of x3.

To reduce the number of variables, we rearranged Equation (1) with normalized luciferase values (x3 = 1 - y1x1 - y2x2) and substituted it into Equations (2) through (4)

(5) Δ3RE=y1(x1+α)+x2−α+1−y1x1−y2x2→1+y1α−α+(1−y2)x2
(6) Loop=x1−β+y2(x1+β)+1−y1x1−y2x2→1+y2β−β+(1−y1)x1
(7) Double=x1+x2+1−y1x1−y2x2→1+(1−y1)x1+(1−y2)x2

Further rearrangement and substitution of ∆3RE and Loop into Equation (7) yields:

(8) (1−y2)x2=Δ3RE−1−y1α+α
(9) (1−y1)x1=Loop−1−y2β+β
(10) Double=Loop+Δ3RE+(α+β−y1α−y2β−1)

Given the measured luciferase values (Figure 2F), the above model, which assumes IRP and eIF3 do not bind simultaneously to the same FTL mRNA, is inconsistent with the data, even accounting for measurement error.

Referring to Equation (10):

41.8(±6.5)>12.5(±2.9)+4.4(±0.4)+(α+β−y1α−y2β−1)
41.8(±6.5)>16.9(±2.9)+(α+β−y1α−y2β−1)

Thus, IRP and eIF3 likely can act in cis on FTL mRNAs.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files.

References

  1. Book
    1. Anderson GJ
    2. McLaren GD
    (2012)
    Iron Physiology and Pathophysiology in Humans
    Springer.
    1. Cazzola M
    2. Bergamaschi G
    3. Tonon L
    4. Arbustini E
    5. Grasso M
    6. Vercesi E
    7. Barosi G
    8. Bianchi PE
    9. Cairo G
    10. Arosio P
    (1997)
    Hereditary hyperferritinemia-cataract syndrome: relationship between phenotypes and specific mutations in the iron-responsive element of ferritin light-chain mRNA
    Blood 90:814–821.
    1. Millonig G
    2. Muckenthaler MU
    3. Mueller S
    (2010)
    Hyperferritinaemia-cataract syndrome: worldwide mutations and phenotype of an increasingly diagnosed genetic disorder
    Human Genomics 4:250–262.
    1. Santambrogio P
    2. Levi S
    3. Arosio P
    4. Palagi L
    5. Vecchio G
    6. Lawson DM
    7. Yewdall SJ
    8. Artymiuk PJ
    9. Harrison PM
    10. Jappelli R
    (1992)
    Evidence that a salt bridge in the light chain contributes to the physical stability difference between heavy and light human ferritins
    The Journal of Biological Chemistry 267:14077–14083.

Decision letter

  1. Alan G Hinnebusch
    Reviewing Editor; Eunice Kennedy Shriver National Institute of Child Health and Human Development, United States
  2. James L Manley
    Senior Editor; Columbia University, United States
  3. Leos Shivaya Valasek
    Reviewer; Institute of Microbiology ASCR, Czech Republic

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Repression of ferritin light chain translation by human eIF3" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by James Manley as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Leos Shivaya Valasek (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This report provides evidence that eIF3 binds to ferritin light chain mRNA (FTL) adjacent to and overlapping with the IRE element where IRP1 binds to exert the well-known repression of FTL mRNA translation in low iron conditions. They show that a deletion of the eIF3 binding site that does not include the region of overlap with the IRE derepresses expression of a FTL-5'UTR-LUC reporter by ~10-fold (although the fold-increase seems to vary between 4x and 20X in different experiments). This Δ3RE deletion does not strongly reduce binding of recombinant IRP1 to the FTL 5'UTR in vitro, reducing affinity ~4-fold; and it also still confers derepression of reporter mRNA in the presence of mutations that impair IRP1 binding or function (although the extent of derepression by Δ3RE is diminished when IRP1 function is reduced by moving the IRE further from the cap.) Gene editing was used to introduce Δ3RE into FLT mRNA in cultured cells and it is shown that the mutation derepresses native FTL protein expression apparently without reducing native IRP1 binding to the mRNA (although there are several issues complicating this conclusion), and evidence is presented that the increased FTL expression and concomitant decreased FTH chain expression confers increased stability on the ferritin complex (although there are issues with these results too.) In addition, the expected responses of FTL expression to iron excess or depletion achieved pharmacologically remain intact for the Δ3RE mutated FTL gene. These findings lead the authors to conclude that eIF3 represses translation of FTL mRNA independently of IRP1. They go on to show that two SNPs in FTL associated with hyperferritinemia that are expected to affect IRP1 binding but are close to the eIF3 binding site that they mapped previously by PAR-CliP derepress expression of the FTL-5'UTR-LUC reporter comparably to the Δ3RE mutation and reduce binding of eIF3 to the reporter mRNA in cells, but only weakly reduce binding of recombinant IRP1 to the FTL mRNA in vitro. It was not shown however that the SNP mutations allow for normal IRP1 binding to reporter mRNA in cells; nor was it shown that the mutated reporters respond normally to iron excess or depletion, which appear to be a significant omissions. The results are significant in linking eIF3-mediated repression of a specific mRNA to a human disease and also add an additional layer of regulation to the well-established mechanism of iron-mediated translational control by IRP1.

Essential revisions:

– It should be verified that equal amounts of WT vs delta3RE and WT vs. G51C and G52C reporter mRNAs are present in the transfected cells from which the luciferase expression levels were obtained

– Results shown in panels 2E and 2F appear to be inconsistent. In 2E, delta3RE causes a 32-fold luminescence change; in 2F, the same mutation/construct causes only a 4.4-fold change. Moreover, in 2F, IRP and eIF3 binding have an “additive” (rather than “synergistic”) effect, whereas in 2E, there is no such additive effect (the expectation would be more than 32-fold). The authors need to convincingly address this issue, ideally by performing both sets of experiments in parallel under identical conditions (according to legend 2F, e.g. the incubation times were different).

– Explain the need for the mathematical modelling in Figure 2—figure supplement 2 and the modelling itself in much greater detail in the Results section. The authors may wish to consider if at all, the modelling adds significantly to the scientific quality of the paper.

– Figure 3—figure supplement 1D-E should be presented in the main figures

– Regarding Figure 3, for panels A and D, the results of Western blots from multiple biological replicates must be presented with appropriate loading controls, and statistical analysis of the differences in mean values determined. Also, for panels A and C, a complete set of WB and RNA-Ip experiments from the same WT and mutant cell lines must be presented.

– Discuss a possible mechanism for how delta3RE leads to increased IRP binding to FTH1 mRNA

– The results in Figure 3—figure supplement 4 do not provide convincing evidence that the WT and mutant ferritin complexes differ in stability. The authors should either show increased ferritin stability from delta3RE cells by thermal melting of purified complexes using urea or guanidinium chloride or cite convincing literature where increasing the FTL/FTH ratio was shown to increase ferritin stability.

– It should be shown by analyzing the existing reporters containing the G51C and G52C mutations that these mutations leave intact IRP1 mediated repression of reporter expression in response to changing iron levels achieved using FAC and DFO treatments.

– Make the appropriate revisions of text to address all other major comments of the reviewers.

Reviewer #1:

This report is of interest in providing evidence that eIF3 binds to ferritin light chain mRNA (FTL) adjacent to and overlapping with the IRE element where IRP1 binds to exert the well-known repression of FTL mRNA translation in low iron conditions. They show that a deletion of the eIF3 binding site that does not include the region of overlap with the IRE derepresses expression of a FTL-5'UTR-LUC reporter by ~10-fold (although the fold-increase seems to vary between 4x and 20X in different experiments). This Δ3RE deletion does not strongly reduce binding of recombinant IRP1 to the FTL 5'UTR in vitro, reducing affinity ~4-fold; and it also still confers derepression of reporter mRNA in the presence of mutations that impair IRP1 binding or function (although the extent of derepression by Δ3RE is diminished when IRP1 function is reduced by moving the IRE further from the cap.) Gene editing was used to introduce Δ3RE into FLT mRNA in cultured cells and it is shown that the mutation derepresses native FTL protein expression apparently without reducing native IRP1 binding to the mRNA (although there are several issues complicating this conclusion), and evidence is presented that the increased FTL expression and concomitant decreased FTH chain expression confers increased stability on the ferritin complex (although there are issues with these results too.) In addition, the expected responses of FTL expression to iron excess or depletion achieved pharmacologically remain intact for the Δ3RE mutated FTL gene. These findings lead the authors to conclude that eIF3 represses translation of FTL mRNA independently of IRP1. They go on to show that two SNPs in FTL associated with hyperferritinemia that are expected to affect IRP1 binding but are close to the eIF3 binding site that they mapped previously by PAR-CliP derepress expression of the FTL-5'UTR-LUC reporter comparably to the Δ3RE mutation and reduce binding of eIF3 to the reporter mRNA in cells, but only weakly reduce binding of recombinant IRP1 to the FTL mRNA in vitro. It was not shown however that the SNP mutations allow for normal IRP1 binding and translational repression by IRP1 in cells, and it seems important to show at least that the mutated reporters respond normally to iron excess or depletion. The results are significant in linking eIF3-mediated repression of a specific mRNA to a human disease and also add an additional layer of regulation to the well-established mechanism of iron-mediated translational control by IRP1. However, there are a number of significant shortcomings with the study, as noted above, and described below.

Major comments:

– Figure 1E: Determine the effects of the Δ3RE mutations on reporter mRNA levels, and show that Δ3RE increases the polysome size of the reporter mRNA in the event that it is found to reduce mRNA levels, in order to rigorously establish derepression at the level of translation.

– Figure 3A: the results of Western blots from multiple biological replicates must be presented and statistical analysis of the differences in mean values determined to rigorously establish that the Δ3RE mutation derepresses FTL protein expression.

– Subsection “Physiological response to loss of eIF3-dependent repression” and Figure 3A and C: Why weren't WB results from HEK cells presented, whereas the text would lead one to believe that the effect of the mutation was the same in both cell lines? Also, why were the RNA-Ips shown in panel C done only in HEK cells and not HepG2 cells, as these data are not directly relevant to the WB expression data in panel A? This selective presentation of results is troubling, and a complete set of WB and RNA-Ip experiments from the same WT and mutant cell lines should be presented.

– Figure 3D: It's unclear from the extensive cropping of the data that the results derive from matched WT and Δ3RE cells examined in parallel. In addition, the results from multiple biological replicates must be presented and statistical analysis of the differences in mean fold- derepression or fold-repression values determined between WT and Δ3RE cells.

– Figure 3—figure supplement 4: It's not immediately clear that the smearing attributed to degradation of the WT ferritin complex would not be seen for the mutant complex if more of the latter was loaded on the gel for equal comparison with WT. The amount of intact complex versus degraded species from multiple biological replicates must be presented and statistical analysis of the differences in mean values of the ratios determined.

– Figure 4E: It is necessary to determine the effects of the G51C and G52C mutations on reporter mRNA levels, and show that they increase the polysome size of the reporter mRNA if they are found to reduce mRNA levels, in order to rigorously establish derepression at the level of translation.

– Subsection “SNPs in FTL that cause hyperferritinemia”: they have not shown that the G51C and G52C mutations leave intact IRP1 mediated repression in response to changing iron levels. This should be done at the very least for the reporters using FAC and DFO treatment.

Reviewer #2:

In the presented paper, Pulos-Holmes et al. explore the role of eIF3 in the translational control of the ferritin light chain (FTL) gene, the deregulation of which causes hyperferritinemia. They propose that eIF3 acts as a specific repressor of the FTL mRNA translation that binds to its structured 5' UTR simultaneously with another repressor – the iron regulatory protein (IRP) – and thus adds another, IRP-IRE-independent layer of the negative control. Since a few specific SNPs located in the eIF3-binding site in the FTL 5' UTR that have been associated with the aforementioned disease disturb this eIF3-specific control, authors conclude that this is the first study identifying a direct role for eIF3-mediated translational control in a specific human disease.

It is an interesting and well-executed work that, in my opinion, deserves to be eventually published in the eLife journal.

Major criticism:

The IRE is a well-defined structural element that has been specifically mutated and I think that this work would benefit from the analysis of secondary structures of the deltaPAR and delta3RE mutant mRNAs; it may greatly help with the interpretation of the proposed synergy between IRP and eIF3. Also, I did not understand why the delta3RE deletion goes beyond the PAR-CLIP 3' boundary (nt 76) further downstream to nt 90, when compared to deltaPAR. This was not rationalized.

Were the Luc experiments (like the one shown in Figure 1E) normalized to the mRNA levels? Can the authors rule out that the presence of IRE partially destabilizes FTL mRNA and both tested mutations, highly likely eliminating its structure, at least partially elevate its expression by stabilizing the mRNA?

Figure 2E and F. Taking into account the synergy (additivity) observed in Figure 2F, wouldn't one expect the same in Figure 2E; i.e. a higher bar for delta3RE with the extended form? Also, the magnitude of the repression relieve by this construct differs dramatically between native forms in Figure 1E (10-fold) and 2E (30-fold), why would that be?

Figure 3—figure supplement 1D-E should be presented in the main figure, it is important in my opinion.

Figure 3C. How could delta3RE lead to an increase of IRP binding to the eIF3-independent mRNA of FTH1?

Subsection “Physiological response to loss of eIF3-dependent repression” final paragraph: This implicates eIF3 in regulating not only the abundance of the two ferritin subunits (How could it regulate FTH? Please discuss.), but also the dynamics and stability of the ferritin complex." How could this happen? IRP should be doing the same, shouldn't it? Please discuss.

Figure 4C. I guess it would be informative to examine the binding of the loop mutant to eIF3 as well.

The authors should at least try to discuss what could be the physiological importance of the IRP-IRE-independent eIF3-mediated control of the FTL expression to lift the perception that this study is more descriptive than innovative.

Reviewer #3:

This is a timely and important study by the Cate lab and collaborators. It examines at the molecular level whether and how the general eukaryotic translation initiation factor 3 (eIF3) affects the translation of a specific target mRNAs (human ferritin light chain mRNA, FTL mRNA) to which it makes sequence or structure-specific contacts. This work is a follow-up of a previous genome-wide PAR-CLIP analysis by the Cate lab that identified numerous eIF3 target mRNAs. Pulos-Holmes now demonstrate as a proof-of-principle how particular disease-causing human SNPs in the 5'UTR of FTL mRNA cause translational mis-regulation as a consequence of failed eIF3 interaction and not, as previously assumed, as a consequence of a failed interaction with the iron-response-protein (IRP). IRP is a highly specific RNA-binding protein with a close-by but distinct RNA target site and long known to regulate FTL mRNA in an iron-dependent manner. Conceptually, the present work advances our understanding of how also general translation factors such as eIF3 can cause specific RNA regulation in addition to and in combination with specific RNA-binding proteins. Furthermore, the authors identify the molecular pathway of how particular SNPs cause human disease and they show that even closely-spaced SNPs can act via distinct molecular interactions.

The manuscript is concise, generally well-written and easy to follow also for a general audience. Nevertheless, I have some comments and suggestions, also regarding experimental details, that should be clarified before publication.

1) It should be useful to add nucleotide numbers to Figure 1D and Figure 1—figure supplement 1C, also marking already here the delta3RE and loop dinucleotide mutations (“loop”) that are used throughout the work. Also the legends should be updated accordingly. Otherwise it is for example not quite clear that “loop” means point mutations and not a larger deletion.

2) Also please explain the meaning of “Double” in the legend of Figure 2. What does the line “N.A., the IC50 value.…” in the Figure 2C legend refer to?

3) The Figure referring to the “mathematical modelling” (Figure 2—figure supplement 2) is not very well explained. What is y1 and y2, what are the brown balls, what is meant by “previously”? One can understand the idea of the modelling only after reading through the Materials and methods in detail. Modeling is claimed to “demonstrate that IRP and eIF3 seem to work synergistically”. This sentence sounds odd. In my view, the described modeling approach seems like an “overkill”, unnecessarily complicated and rather indirect, because the data in the end excludes the described model. Isn't it sufficient in Figure 2F to see that, starting from “Double”, the repression in “FTL” (~40-fold) is roughly the product of the repression by IRP in “delta3RE” (~10 fold) and the repression by eIF3 in “Loop” (~4-fold)? If only one of the two binders could act at a time, the maximum repression would be (~10 fold) by IRP. Maybe I am missing something, but then it needs to be better explained.

4) The description of the Western blots and sample dilutions in Figure 3A, B, D and Figure 3—figure supplement 3 is not very clear, and/or it may be advisable to show more convincing blots. It is not always clear which lane shows diluted samples and whether the β Actin controls were co-diluted or not (It seems to me that the b Actin controls were run on separate lanes of the gel and were undiluted). Moreover, in Figure 3D, the lanes in the absence of FAC would be expected to look exactly like the lanes in the absence of DFO (in both cases the cells are still untreated), but this is not really the case. Although all of the cells are still reactive to FAC or DFO, it would be more convincing if the starting material looked the same. Similar in Figure 3—figure supplement 3.

5) In Figure 3B, it might be good to clarify that “delta3RE” refers to the respective mutation in the FTL RNA, e.g. by labelling “delta3RE” in Figure 3A, B as “FTL-delta3RE”. Moreover, the authors do not give any explanation for the puzzling result that FTH protein expression is apparently decreased when FTL RNA is mutated. Is it possible that “delta3RE” cells try to compensate the lack of FTL RNA translational repression by producing generally more IRP protein and that this excessive IRP protein then downregulates FTH RNA? Did the authors try and look at endogenous IRP protein levels in WT and “delta3RE” cells with e.g. a suitable antibody?

6) Is there any explanation for why DFO does not downregulate FTH RNA in Figure 3—figure supplement 2B?

7) It should be clarified in Figure 3—figure supplement 4 that this is a Coomassie-stained native gel (maybe show in blue?) of the purified ferritin complex, presumably containing different ratios of FTL and FTH chains (predominantly FTL in the case of delta3RE cells?). The text claims that the delta3RE cells contain 8x as much ferritin as the WT cells, but this is not obvious from the figure. Moreover, the text also claims the complexes from delta3RE cells to be more “stable”. Presumably this does not refer to the physiological half-life of the complex in the cell but to the “stability” of the complex in the harsh purification procedure using 40% methanol and 75 deg C. Does the ferritin complex remain intact and folded in this process or does it undergo unfolding/refolding and/or disassembly/reassembly? Does protein concentration and purity matter in this process and possibly lead to wrong conclusions regarding “stability”?

8) The text lacks general information for a general reader on what is the physiological difference between FTL and FTH proteins and whether the ratio in which they occur in ferritin has physiological importance. Are FTL-exclusive ferritins more stable than FTH-containing ones? This is important, because eIF3 binding appears to affect not only the total amount of FTL, but even more so the ratio of FTL to FTH, possibly generating ferritin complexes with different properties. This is somewhat mentioned at the end of the Discussion, but the authors could elaborate on this point of physiological significance.

9) Some statements in the text could be revisited:

Abstract: SNPs.… “are thought” to disrupt.…. Do you mean “were described/reported” to disrupt? Otherwise that statement sounds like a hypothesis and creates the impression that the study might continue with a proof of IRP-dependence. See also the first paragraph of subsection “SNPs in FTL that cause hyperferritinemia” for a very similar statement.

Subsection “Identification of the eIF3-FTL mRNA interaction site”: “a secondary structure element that overlaps” or do you mean “a 24 nucleotide sequence.… that overlaps”?

In the same section: “.… in a deletion the/that maintained”.…

Subsection “Decoupling the repressive role of eIF3 on FTL mRNA from that of IRP”: What is “the characteristic C bulge”?

In the same section: List/ explain the loop mutation as a A15G/G16C dinucleotide mutation (clarify numbers in Figures; 40/41 in Figure 5?)

Subsection “Physiological response to loss of eIF3-dependent repression”: “significantly” suggests statistically evaluated significance, maybe use “considerably”.

In the same section check the following statement: “This implicates eIF3 in regulating not only the abundance of the two ferritin subunits, but also the dynamics and stability of the ferritin complex” This statement needs further explanation on how this is thought to occur – for example that different ratios of FTL and FTH within a given ferritin complex could cause such differences (maybe add references if data exists on this).

Figure 3—figure supplement 2 legend: Explain FAC and DFO here instead of later in the Figure 3—figure supplement 3.

Figure 4 legend: small caps for “hyperferritinemia”. “EIF3B” or “eIF3B”? “in the IP compared to/with input levels”. “Loop” mutant with reference to Cazzola et al.? In 4D, clarify that this is competition for IRP-binding. Serum ferritin levels are given for loop mutants, but what is “normal”/physiological? Explain details of the mutants already in 4B, rather than in 4D. If available, add a “Loop”' column in 4C.

https://doi.org/10.7554/eLife.48193.031

Author response

Essential revisions:

– It should be verified that equal amounts of WT vs delta3RE and WT vs. G51C and G52C reporter mRNAs are present in the transfected cells from which the luciferase expression levels were obtained

All mRNA concentrations were determined using a Nanodrop and were validated by running samples on an agarose gel. Bands on the gel were quantified to properly ensure all constructs were transfected at equal amounts. We have clarified this approach in the Materials and methods.

In order to address the concern of RNA degradation, we performed a time course experiment in which we followed both the luminescence and mRNA amount (via qPCR) post-transfection. As shown in Figure 2—figure supplement 2, after 6 hrs posttransfection there is no significant difference in mRNA stability when comparing the different mRNA reporters. With this information we can now definitively conclude that the derepression we see occurs at the level of translation.

– Results shown in panels 2E and 2F appear to be inconsistent. In 2E, delta3RE causes a 32-fold luminescence change; in 2F, the same mutation/construct causes only a 4.4-fold change. Moreover, in 2F, IRP and eIF3 binding have an “additive” (rather than “synergistic”) effect, whereas in 2E, there is no such additive effect (the expectation would be more than 32-fold). The authors need to convincingly address this issue, ideally by performing both sets of experiments in parallel under identical conditions (according to legend 2F, e.g. the incubation times were different).

There are two potential factors that could explain why the derepression values observed in Figure 2F are seemingly not consistent with the extended ∆3RE construct in Figure 2E. First, the extension of the 5′-UTR from its native state does not completely abolish the repression mediated by IRP. It only significantly dampens it (See Goossen and Hentze, 1992). This is in contrast to the loop mutations which abolish IRP binding entirely (Figure 4D and Cazzola et al., 1997). We note that in the original submission, the experiment in Figure 2E was carried out for 12 hrs. We have seen that with HepG2 cells there is still an increase in luminescence beyond shorter (i.e. 6 hr) incubation times. For example, see Figure 1—figure supplement 1D. In a separate experiment, which we now include as the new Figure 2E, we used a 6-hr incubation. In this experiment, derepression of the WT FTL with extended 5′-UTR is about 5-fold. By contrast, the loop mutation derepressed FTL translation 12-fold in the 6-hr incubation in Figure 2F.

Second, it is not clear how the distance from the 5′-cap affects repression mediated eIF3. This would be an interesting mechanistic question to address in a follow-up publication.

– Explain the need for the mathematical modelling in Figure 2—figure supplement 2 and the modelling itself in much greater detail in the Results section. The authors may wish to consider if at all, the modelling adds significantly to the scientific quality of the paper.

Reviewer 3 states that from the data presented in Figure 2F, “If only one of the two binders could act at a time, the maximum repression would be (~10 fold) by IRP,” not the ~40x that we see. However, this assumes that the WT FTL mRNA is fully repressed in these culture conditions. Based on the DFO experiments (i.e. in Figure 3C and Figure 3—figure supplement 4), this is most certainly not the case. This is one of the reasons we developed the model. We think the quantitative modeling provides a framework for understanding FTL translation regulation more generally. For example, in this manuscript we cannot rule out that eIF3 may have a role in iron responsiveness of FTL translation. By having this model present, we can provide possible explanations for why some of the iron responsiveness data is not an exact match between WT and edited cell lines.

– Figure 3—figure supplement 1D-E should be presented in the main figures

These data are now included in Figure 3E-F.

– Regarding Figure 3, for panels A and D, the results of Western blots from multiple biological replicates must be presented with appropriate loading controls, and statistical analysis of the differences in mean values determined. Also, for panels A and C, a complete set of WB and RNA-Ip experiments from the same WT and mutant cell lines must be presented.

The purpose of these experiments was to show that the edited cells retain the ability to respond to iron, which they do. Though quantification of these westerns is not necessary to convey this result, we have included quantification of these data in Figure 3—figure supplements 4 and 5. Figure 3C-D and Figure 3 —figure supplement 4 show a representative set of western blots and quantification of westerns carried out in multiple biological replicates. We have also included representative western blots and quantification of FTL and FTH levels in WT and ∆3RE HEK293T cells, in Figure 3 —figure supplement 5, which show similar results. We note that the increase in FTL abundance in the edited cell lines might alter the cellular response to iron treatment with respect to FTL translation. We therefore cannot expect a fully normal response in these edited cells.

With respect to the IRP-IP results, although we were also interested in preforming the corresponding IRP-FLAG IPs in HepG2 cells, we were only able to achieve efficient plasmid transfections in HEK293T cells. Multiple attempts and various forms of DNA transfection methods were attempted in HepG2 cells, with no success in our hands.

– Discuss a possible mechanism for how delta3RE leads to increased IRP binding to FTH1 mRNA.

This has been included in the main text. Due to the significant increase in FTL protein levels, we initially hypothesized that there is a feedback mechanism causing IRP to bind to FTH1 mRNA to control the overall amount of ferritin there is in the cell. We do show that in the ∆3RE cell lines that IRP seems to bind more FTH1 mRNA when compared to native cells using an IRP-FLAG IP (See Figure 3G and Figure 3—figure supplements 4 and 5). Although we have not checked if the abundance of IRP differs in these cell lines, if it did, other IRP-regulated transcripts such as that for Ferroportin would likely be affected in the same way as FTH. Thus, we have also looked at another protein, Ferroportin, which also has an IRE in the 5′-UTR of its mRNA. Notably, we see no difference in Ferroportin abundance comparing the wild type and ∆3RE cell line, suggesting that the decrease in FTH levels is not due to a mass action process. The mechanism by which IRP would be recruited preferentially to FTH1 mRNA is beyond the scope of this paper.

– The results in Figure 3—figure supplement 4 do not provide convincing evidence that the WT and mutant ferritin complexes differ in stability. The authors should either show increased ferritin stability from delta3RE cells by thermal melting of purified complexes using urea or guanidinium chloride or cite convincing literature where increasing the FTL/FTH ratio was shown to increase ferritin stability.

The precedence for this stems from the fact FTL has a salt bridge in its central hydrophilic region that contributes to its stability (See Santambrogio et al., 1991, which we now cite). We have modified the text to clarify this point.

– It should be shown by analyzing the existing reporters containing the G51C and G52C mutations that these mutations leave intact IRP1 mediated repression of reporter expression in response to changing iron levels achieved using FAC and DFO treatments.

To address whether the G51C and G52C mRNA constructs retain the ability to interact with IRP in vivo,we generated stable HEK293T lines expressing the G51C and G52C luciferase reporter mRNAs using lentiviral constructs. This was done in HEK293T cells because we also wanted to repeat the IRP-FLAG IPs in these cell lines. As shown in Figure 4—figure supplement 2, IRP can still interact with the mutant FTL 5′-UTR elements with the G51C or G52C mutations, as previously determined using the in vitro EMSAs (Figure 4 and Figure 4—figure supplement 1). Although FAC treatment does not seem to release IRP as efficiently from the G51C and G52C reporters compared to endogenous FTL, we suggest that these mutant constructs are more efficiently bound by IRP in cells than the WT, as seen in the IRP-FLAG IP results in Figure 4—figure supplement 2D, or due to possible post-translational regulation of FTL, which is not recapitulated using luciferase as a reporter.

– Make the appropriate revisions of text to address all other major comments of the reviewers.

Additional modifications to the text and figures are described below.

Reviewer #1:

Please see our overall responses above.

Reviewer #2:

[…] It is an interesting and well-executed work that, in my opinion, deserves to be eventually published in the eLife journal.

Major criticism:

The IRE is a well-defined structural element that has been specifically mutated and I think that this work would benefit from the analysis of secondary structures of the deltaPAR and delta3RE mutant mRNAs; it may greatly help with the interpretation of the proposed synergy between IRP and eIF3. Also, I did not understand why the delta3RE deletion goes beyond the PAR-CLIP 3' boundary (nt 76) further downstream to nt 90, when compared to deltaPAR. This was not rationalized.

While we agree that obtaining the secondary structure for the mutant constructs would be quite informative, we think these experiments would be better suited to a follow-up study. We think the functional data in Figure 2 and Figure 2—figure supplement 1 provide convincing data that at least the IRE component of the 5’-UTR retains is structure in the ∆3RE, but not ∆PAR, context.

The ∆3RE extends past the PAR-CLIP site mainly due to CRISPR-Cas9 editing limitations. Since HDR has extremely low efficiency in HepG2 cells, we decided this was the best way to approach the editing. It also achieves both the goals of disrupting eIF3 and maintaining the IRP interaction.

Were the Luc experiments (like the one shown in Figure 1E) normalized to the mRNA levels? Can the authors rule out that the presence of IRE partially destabilizes FTL mRNA and both tested mutations, highly likely eliminating its structure, at least partially elevate its expression by stabilizing the mRNA?

Figure 2E and F. Taking into account the synergy (additivity) observed in Figure 2F, wouldn't one expect the same in Figure 2E; i.e. a higher bar for delta3RE with the extended form? Also, the magnitude of the repression relieve by this construct differs dramatically between native forms in Figure 1E (10-fold) and 2E (30-fold), why would that be?

Figure 3—figure supplement 1D-E should be presented in the main figure, it is important in my opinion.

Figure 3C. How could delta3RE lead to an increase of IRP binding to the eIF3-independent mRNA of FTH1?

Subsection “Physiological response to loss of eIF3-dependent repression” final paragraph: This implicates eIF3 in regulating not only the abundance of the two ferritin subunits (How could it regulate FTH? Please discuss.), but also the dynamics and stability of the ferritin complex." How could this happen? IRP should be doing the same, shouldn't it? Please discuss.

Please see our overall responses above.

Figure 4C. I guess it would be informative to examine the binding of the loop mutant to eIF3 as well.

We think these experiments would be better suited to a follow-up study, and is now covered by our analysis of the model presented in Figure 2—figure supplement 3.

The authors should at least try to discuss what could be the physiological importance of the IRP-IRE-independent eIF3-mediated control of the FTL expression to lift the perception that this study is more descriptive than innovative.

Please see our overall response above. We noted possible broader implications for eIF3mediated repression of FTL translation in the Discussion, and think the molecular insights we have obtained stand on their own as an innovative contribution to understanding human translation regulation.

Reviewer #3:

The manuscript is concise, generally well-written and easy to follow also for a general audience. Nevertheless, I have some comments and suggestions, also regarding experimental details, that should be clarified before publication.

1) It should be useful to add nucleotide numbers to Figure 1D and Figure 1—figure supplement 1C, also marking already here the delta3RE and loop dinucleotide mutations (“loop”) that are used throughout the work. Also the legends should be updated accordingly. Otherwise it is for example not quite clear that “loop” means point mutations and not a larger deletion.

Since this is a cartoon schematic, we did not number the nucleotides in this panel. We added the loop mutation information to Figure 2F, the first time they appear. We have also included numbering in Figure 4A, Figure 5, as well as in the figure legend to Figure 1—figure supplement 1.

2) Also please explain the meaning of “Double” in the legend of Figure 2. What does the line “N.A., the IC50 value…” in the Figure 2C legend refer to?

We have updated the figure legend to Figure 2, to make this clear. Double means a construct that contains the loop mutation that disrupts IRP binding combined with the ∆3RE mutation which disrupts eIF3 binding.

3) The Figure referring to the “mathematical modelling” (Figure 2—figure supplement 2) is not very well explained. What is y1 and y2, what are the brown balls, what is meant by “previously”? One can understand the idea of the modelling only after reading through the Materials and methods in detail. Modeling is claimed to “demonstrate that IRP and eIF3 seem to work synergistically”. This sentence sounds odd. In my view, the described modeling approach seems like an “overkill”, unnecessarily complicated and rather indirect, because the data in the end excludes the described model. Isn't it sufficient in Figure 2F to see that, starting from “Double”, the repression in “FTL” (~40-fold) is roughly the product of the repression by IRP in “delta3RE” (~10 fold) and the repression by eIF3 in “Loop” (~4-fold)? If only one of the two binders could act at a time, the maximum repression would be (~10 fold) by IRP. Maybe I am missing something, but then it needs to be better explained.

4) The description of the Western blots and sample dilutions in Figure 3A, B, D and Figure 3—figure supplement 3 is not very clear, and/or it may be advisable to show more convincing blots. It is not always clear which lane shows diluted samples and whether the β Actin controls were co-diluted or not (It seems to me that the b Actin controls were run on separate lanes of the gel and were undiluted). Moreover, in Figure 3D, the lanes in the absence of FAC would be expected to look exactly like the lanes in the absence of DFO (in both cases the cells are still untreated), but this is not really the case. Although all of the cells are still reactive to FAC or DFO, it would be more convincing if the starting material looked the same. Similar in Figure 3—figure supplement 3.

5) In Figure 3B, it might be good to clarify that “delta3RE” refers to the respective mutation in the FTL RNA, e.g. by labelling “delta3RE” in Figure 3A, B as “FTL-delta3RE”. Moreover, the authors do not give any explanation for the puzzling result that FTH protein expression is apparently decreased when FTL RNA is mutated. Is it possible that “delta3RE” cells try to compensate the lack of FTL RNA translational repression by producing generally more IRP protein and that this excessive IRP protein then downregulates FTH RNA? Did the authors try and look at endogenous IRP protein levels in WT and “delta3RE” cells with e.g. a suitable antibody?

Please see our overall responses above.

6) Is there any explanation for why DFO does not downregulate FTH RNA in Figure 3—figure supplement 2B?

Unfortunately, we do not have an explanation for this at this time.

7) It should be clarified in Figure 3—figure supplement 4 that this is a Coomassie-stained native gel (maybe show in blue?) of the purified ferritin complex, presumably containing different ratios of FTL and FTH chains (predominantly FTL in the case of delta3RE cells?). The text claims that the delta3RE cells contain 8x as much ferritin as the WT cells, but this is not obvious from the figure. Moreover, the text also claims the complexes from delta3RE cells to be more “stable”. Presumably this does not refer to the physiological half-life of the complex in the cell but to the “stability” of the complex in the harsh purification procedure using 40% methanol and 75 deg C. Does the ferritin complex remain intact and folded in this process or does it undergo unfolding/refolding and/or disassembly/reassembly? Does protein concentration and purity matter in this process and possibly lead to wrong conclusions regarding “stability”?

We have clarified the figure legend to address that this is a coomassie stained gel. You are correct that we are referring to the stability of the complex during the purification process and not its half life. Based on literature precedence (cited in the manuscript) we believe that this is an intact and folded complex throughout the purification process. This method of purification is very common for ferritin and because of how harsh it is, we do not feel extremely worried about contaminants in the protein.

8) The text lacks general information for a general reader on what is the physiological difference between FTL and FTH proteins and whether the ratio in which they occur in ferritin has physiological importance. Are FTL-exclusive ferritins more stable than FTH-containing ones? This is important, because eIF3 binding appears to affect not only the total amount of FTL, but even more so the ratio of FTL to FTH, possibly generating ferritin complexes with different properties. This is somewhat mentioned at the end of the Discussion, but the authors could elaborate on this point of physiological significance.

There are physiological differences between FTL and FTH. For example, FTH is the catalytic subunit of the two whereas FTL promotes complex stability and aids in the iron mineralization process. We expanded our description of FTH in various places, i.e. in terms of its effect on Ferritin stability, and the observation of post-transcriptional downregulation of FTH levels, but not Ferroportin levels, based on our new experiments. However, we have not strongly elaborated on this process because we feel that it goes beyond the scope of the paper.

9) Some statements in the text could be revisited:

Abstract: SNPs.… “are thought” to disrupt.…. Do you mean “were described/reported” to disrupt? Otherwise that statement sounds like a hypothesis and creates the impression that the study might continue with a proof of IRP-dependence. See also the first paragraph of subsection “SNPs in FTL that cause hyperferritinemia” for a very similar statement.

This has been corrected.

Subsection “Identification of the eIF3-FTL mRNA interaction site”: “a secondary structure element that overlaps” or do you mean “a 24 nucleotide sequence.… that overlaps”?

This has been corrected.

In the same section: “… in a deletion the/that maintained”.…

This has been corrected. Thank you for finding the typo.

Subsection “Decoupling the repressive role of eIF3 on FTL mRNA from that of IRP”: What is “the characteristic C bulge”?

We have modified Figure 2D by adding an asterisk, and the associated figure legend to clarify where the characteristic C bulge is in the FTL 5’-UTR.

In the same section: List/ explain the loop mutation as a A15G/G16C dinucleotide mutation (clarify numbers in Figures; 40/41 in Figure 5?)

The numbers for the loop mutants A15G/G16C have been noted in the Figure 2F legend, as well as the legend for Figure 4. The numbering represents their space in the IRE sequence, not the full RNA. Unfortunately, there is a difference in the naming system of the hyperferritinemia mutants.

Subsection “Physiological response to loss of eIF3-dependent repression”: “significantly” suggests statistically evaluated significance, maybe use “considerably”.

Changed as suggested.

In the same section check the following statement: “This implicates eIF3 in regulating not only the abundance of the two ferritin subunits, but also the dynamics and stability of the ferritin complex” This statement needs further explanation on how this is thought to occur – for example that different ratios of FTL and FTH within a given ferritin complex could cause such differences (maybe add references if data exists on this).

Please see our overall response above.

Figure 3—figure supplement 2 legend: Explain FAC and DFO here instead of later in the Figure 3—figure supplement 3.

Done.

Figure 4 legend: small caps for “hyperferritinemia”. “EIF3B” or “eIF3B”? “in the IP compared to/with input levels”. “Loop” mutant with reference to Cazzola et al.? In 4D, clarify that this is competition for IRP-binding. Serum ferritin levels are given for loop mutants, but what is “normal”/physiological? Explain details of the mutants already in 4B, rather than in 4D. If available, add a “Loop”' column in 4C.

Hyperferritinemia – fixed.

EIF3B is the human naming convention.

‘IP, compared to input levels’.

Normal levels of ferritin now included in figure legend.

Loop mutations clarified.

Unfortunately, we do not have these data.

https://doi.org/10.7554/eLife.48193.032

Article and author information

Author details

  1. Mia C Pulos-Holmes

    Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8234-4020

  2. Daniel N Srole

    Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    Present address
    Department of Molecular & Medical Pharmacology, David Geffen School of Medicine at UCLA, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation, Writing—review and editing, assisted in carrying out experiments related to screening and obtaining clonal cell lines, as well as plasmid construct generation
    Competing interests
    No competing interests declared

  3. Maria G Juarez

    Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Investigation, Writing—review and editing, aided in plasmid construct generation and accessory biochemical experiments
    Competing interests
    No competing interests declared

  4. Amy S-Y Lee

    Biology Department, Rosenstiel Basic Medical Science Research Center, Brandeis University, Waltham, United States
    Contribution
    Conceptualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared

  5. David T McSwiggen

    Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Conceptualization, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3844-7433

  6. Nicholas T Ingolia

    1. Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    2. California Institute for Quantitative Biosciences, University of California, Berkeley, Berkeley, United States
    Contribution
    Formal analysis, Supervision, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3395-1545

  7. Jamie H Cate

    1. Department of Molecular & Cell Biology, University of California, Berkeley, Berkeley, United States
    2. California Institute for Quantitative Biosciences, University of California, Berkeley, Berkeley, United States
    3. Department of Chemistry, University of California, Berkeley, Berkeley, United States
    4. Molecular Biophysics & Integrated Bioimaging Division, Lawrence Berkeley National Laboratory, Berkeley, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    jcate@lbl.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5965-7902

Funding

National Institute of General Medical Sciences (P50 GM102706)

  • Mia C Pulos-Holmes
  • Daniel N Srole
  • Maria G Juarez
  • Amy S-Y Lee
  • David Trombley McSwiggen
  • Nicholas T Ingolia
  • Jamie H Cate

National Institute of General Medical Sciences (R01 GM065050)

  • Mia C Pulos-Holmes
  • Daniel N Srole
  • Maria G Juarez
  • Jamie H Cate

American Heart Association (16PRE30140013)

  • Mia C Pulos-Holmes

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are immensely grateful to Bruno Martinez for providing us with purified IRP1, Luisa Arake De Tacca for help optimizing the eIF3 immunoprecipitations and for many constructive discussions, Hector Nolla for help with flow cytometry and single cell sorting, and Alison Killilea and the Berkeley tissue culture facility for cells and advice.

Senior Editor

  1. James L Manley, Columbia University, United States

Reviewing Editor

  1. Alan G Hinnebusch, Eunice Kennedy Shriver National Institute of Child Health and Human Development, United States

Reviewer

  1. Leos Shivaya Valasek, Institute of Microbiology ASCR, Czech Republic

Publication history

  1. Received: May 4, 2019
  2. Accepted: August 14, 2019
  3. Accepted Manuscript published: August 15, 2019 (version 1)
  4. Version of Record published: September 3, 2019 (version 2)

Copyright

© 2019, Pulos-Holmes et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,683
    Page views
  • 386
    Downloads
  • 17
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Mia C Pulos-Holmes
  2. Daniel N Srole
  3. Maria G Juarez
  4. Amy S-Y Lee
  5. David T McSwiggen
  6. Nicholas T Ingolia
  7. Jamie H Cate
(2019)
Repression of ferritin light chain translation by human eIF3
eLife 8:e48193.
https://doi.org/10.7554/eLife.48193

Categories and tags

Research organism

Further reading

    1. Biochemistry and Chemical Biology
    2. Developmental Biology

    O-GlcNAc glycosylation orchestrates fate decision and niche function of bone marrow stromal progenitors

    Zengdi Zhang, Zan Huang ... Hai-Bin Ruan
    Research Article Updated

    In mammals, interactions between the bone marrow (BM) stroma and hematopoietic progenitors contribute to bone-BM homeostasis. Perinatal bone growth and ossification provide a microenvironment for the transition to definitive hematopoiesis; however, mechanisms and interactions orchestrating the development of skeletal and hematopoietic systems remain largely unknown. Here, we establish intracellular O-linked β-N-acetylglucosamine (O-GlcNAc) modification as a posttranslational switch that dictates the differentiation fate and niche function of early BM stromal cells (BMSCs). By modifying and activating RUNX2, O-GlcNAcylation promotes osteogenic differentiation of BMSCs and stromal IL-7 expression to support lymphopoiesis. In contrast, C/EBPβ-dependent marrow adipogenesis and expression of myelopoietic stem cell factor (SCF) is inhibited by O-GlcNAcylation. Ablating O-GlcNAc transferase (OGT) in BMSCs leads to impaired bone formation, increased marrow adiposity, as well as defective B-cell lymphopoiesis and myeloid overproduction in mice. Thus, the balance of osteogenic and adipogenic differentiation of BMSCs is determined by reciprocal O-GlcNAc regulation of transcription factors, which simultaneously shapes the hematopoietic niche.

    1. Biochemistry and Chemical Biology
    2. Neuroscience

    Glia-neuron coupling via a bipartite sialylation pathway promotes neural transmission and stress tolerance in Drosophila

    Hilary Scott, Boris Novikov ... Vladislav Panin
    Research Article

    Modification by sialylated glycans can affect protein functions, underlying mechanisms that control animal development and physiology. Sialylation relies on a dedicated pathway involving evolutionarily conserved enzymes, including CMP-sialic acid synthetase (CSAS) and sialyltransferase (SiaT) that mediate the activation of sialic acid and its transfer onto glycan termini, respectively. In Drosophila, CSAS and DSiaT genes function in the nervous system, affecting neural transmission and excitability. We found that these genes function in different cells: the function of CSAS is restricted to glia, while DSiaT functions in neurons. This partition of the sialylation pathway allows for regulation of neural functions via a glia-mediated control of neural sialylation. The sialylation genes were shown to be required for tolerance to heat and oxidative stress and for maintenance of the normal level of voltage-gated sodium channels. Our results uncovered a unique bipartite sialylation pathway that mediates glia-neuron coupling and regulates neural excitability and stress tolerance.

    1. Biochemistry and Chemical Biology
    2. Medicine

    Iron status influences mitochondrial disease progression in Complex I-deficient mice

    CJ Kelly, Reid K Couch ... Anthony S Grillo
    Research Article Updated

    Mitochondrial dysfunction caused by aberrant Complex I assembly and reduced activity of the electron transport chain is pathogenic in many genetic and age-related diseases. Mice missing the Complex I subunit NADH dehydrogenase [ubiquinone] iron-sulfur protein 4 (NDUFS4) are a leading mammalian model of severe mitochondrial disease that exhibit many characteristic symptoms of Leigh Syndrome including oxidative stress, neuroinflammation, brain lesions, and premature death. NDUFS4 knockout mice have decreased expression of nearly every Complex I subunit. As Complex I normally contains at least 8 iron-sulfur clusters and more than 25 iron atoms, we asked whether a deficiency of Complex I may lead to iron perturbations, thereby accelerating disease progression. Consistent with this, iron supplementation accelerates symptoms of brain degeneration in these mice, while iron restriction delays the onset of these symptoms, reduces neuroinflammation, and increases survival. NDUFS4 knockout mice display signs of iron overload in the liver including increased expression of hepcidin and show changes in iron-responsive element-regulated proteins consistent with increased cellular iron that were prevented by iron restriction. These results suggest that perturbed iron homeostasis may contribute to pathology in Leigh Syndrome and possibly other mitochondrial disorders.