Irp2 regulates insulin production through iron-mediated Cdkal1-catalyzed tRNA modification

Regulation of cellular iron homeostasis is crucial as both iron excess and deficiency cause hematological and neurodegenerative diseases. Here we show that mice lacking iron-regulatory protein 2 (Irp2), a regulator of cellular iron homeostasis, develop diabetes. Irp2 post-transcriptionally regulates the iron-uptake protein transferrin receptor 1 (TfR1) and the iron-storage protein ferritin, and dysregulation of these proteins due to Irp2 loss causes functional iron deficiency in β cells. This impairs Fe–S cluster biosynthesis, reducing the function of Cdkal1, an Fe–S cluster enzyme that catalyzes methylthiolation of t6A37 in tRNALysUUU to ms2t6A37. As a consequence, lysine codons in proinsulin are misread and proinsulin processing is impaired, reducing insulin content and secretion. Iron normalizes ms2t6A37 and proinsulin lysine incorporation, restoring insulin content and secretion in Irp2−/− β cells. These studies reveal a previously unidentified link between insulin processing and cellular iron deficiency that may have relevance to type 2 diabetes in humans.

Mechanistically, Irp2-difficiency suppressed the mRNA and protein levels of transferrin receptor (TfR1) and upregulated the protein level of ferritin, leading to the decrease in the total iron content in the pancreatic beta-cells. The decrease of intracellular iron subsequently impaired the activity of Cdkal1, an iron-sulfur cluster-dependent enzyme required for the methylthio-modification of tRNALys(UUU). Dysregulation of Cdkal1 then impaired proinsulin translation at Lys codon. Importantly, supplement of iron in Irp2-deficient cells enhanced Cdkal1 activity, which improved the proinsulin translation fidelity and increased insulin contents. These results demonstrate that Irp2 is essential for the iron homeostasis and controls translational fidelity of proinsulin via tRNA modification enzyme Cdkal1. Overall, these findings are interesting and novel.
Major concerns: 1. The 5-month old Irp2-knockout mice showed glucose intolerance after intraperitoneal (i.p.) injection of glucose (Figure 1a), but the plasma insulin levels of knockout mice did not differ from wild-type mice at the basal level and after the i.p. injection of glucose ( Figure 2a). The plasma insulin data was not consistent with the hyperglycemic clamp data (Figure 2d), which showed that insulin secretion was markedly decreased in Irp2-knockout mice. Authors should explain this discrepancy.
2. The finding that Irp2-deficiency resulted in the accumulation of proinsulin is particularly important in this study. Authors should examine the plasma proinsulin level or plasma proinsulin/insulin ratio in Irp2-deficient mice with i.p. administration of glucose. Authors also need to examine the proinsulin secretion in isolated islets after glucose stimulation.
Minor concerns: 1. Figure 3e and 3f. Authors showed that the average islet area and the beta-cell mass are comparable between Irp2-knockout mice and wild-type mice. HE staining of pancreas and islets need to be presented to support these data.
2. Figure 3. There are two "b" panels. The upper right panel should be panel "c".
3. Figure 4d. Authors claimed that TfR1 staining was reduced in Irp2-knockout beta-cells (page 7, line 164). This is not convincing because the quality of the TfR1 staining is very poor. A majority of betacells was not stained by the antibody even in the wild-type islets.
4. Figure 5a. Authors claimed that TfR1 was modestly reduced in Irp2-knockdown INS cells (page 8, line 175). This claim is not correct because the intensity of TfR1 band in Irp2 knockdown cells was comparable to the TfR1 band in control cells (lane 1 versus lane 2). Please do not overstate the result, or author should provide the quantitative data. 5. Figure 5b. Please provide the method for this experiment.
6. Figure 5c. In addition to Calcein-AM, authors need to use ICP-MS to measure the iron content in Irp2-knockdown INS cells (figure 5) and Irp-knockout INS cells (figure 6). 7. Page 7, line 145: It is not appropriate to state that there is no general defect in ER protein maturation just because Glut2 showed no change between Irp2-knockout islets and wild-type islets. Please rephrase. 11. Page 13, line 362. "min 120" should be "120 min".
12. The units for glucose and insulin are not consistent. For example, insulin is shown as ng/ml in figure 2b, but is shown as pmol/l in figure 2d. Please be consistent.
13. Please provide sequencing data to show how Irp2 gene was edited in the two lines of Irp2knockout cells The main point of this paper is that beta cell deficiency of labile iron pertubs the fidelity of proinsulin translation leading to diabetes. Specifically, in this manuscript, the authors utilize mice with whole body knockout of the RNA-binding protein Irp2, and show that these mice develop glucose intolerance with an increase in intracellular proinsulin and a decrease in intracellular insulin. They indicate that Irp2 loss of function leads to increased sequestered iron but decreased labile iron that they link with destabilization of the Fe-S enzyme Cdkal1 (itself a T2D susceptibility gene), thus resulting in improper translation of lysine codons in proinsulin that could account for impaired proinsulin processing and possible beta cell ER stress. Using Irp2 deficient INS1 cells as a model, some of these adverse effects are improved by preincubation for 18 hours with ferric ammonium citrate (FAC) to increase labile iron. The connection of Irp2 levels to regulation of Cdkal1 to ultimately drive regulation of proinsulin processing through methylthiolation and proper lysine codon translation is novel, and the story is well written. However, a major concern is insufficient validation and the lack of more physiological evidence to support the mechanism that they are hypothesizing. At this point, additional studies are necessary to support the conclusions and enhance the significance of the manuscript.
Specific Comments:

The authors show that 18 hours of FAC supplementation rescues insulin-related phenotypes in
Irp2-deficient beta cells. To demonstrate physiological significance, the authors should try to determine if iron supplementation (even a relatively brief course to prevent toxicity) can restore insulin production or secretion in vivo. 2. Independent of #1, the authors should determine if FAC is capable to restore insulin processing and Cdkal1 levels in isolated Irp2 null islets instead of relying solely on results from cell lines. 3. Immunofluorescence microscopy with co-stained organelle markers, and transmission electron microscopy of the islets of Irp2 null pancreata are needed to understand where in beta cells the increased proinsulin may be accumulating. 4. Many of the mechanistic connections drawn between Irp2 and impaired insulin processing may be linked by effects on Cdkal1, but some attempts should be made to exclude other possibilities. a. Fe-S clusters regulate key mitochondrial proteins and proinsulin intracellular transport and processing is an energy intensive process. Low Complex I activity in shIrp2 INS-1 cells could lead to the generation of ROS, which could then lead to untoward downstream effects. Thus, pharmacologic approaches to enhance mitochondrial fuel utilization (such as mitochondrial targeted substrates including methylpyruvate or amino acids leucine/glutamine/alanine) and/or reduce ROS (antioxidants) independent of using an iron supplement are needed to determine if it is truly Cdkal1 deficiency that impairs insulin processing in Irp2 null islets or simply bioenergetic failure. b. Can overexpression of Cdkal1 more efficiently rescue proinsulin processing deficits in Irp2 null cells plus and minus FAC? 5. As best I can tell, the Irp2 null mice have been around for 15 years, but implicating deficiency of Irp2 directly in human disease has not really happened yet. As far as I know, Irp2 is not even on the long list of diabetes GWAS candidates. With this in mind, it would be best if the authors would attempt to demonstrate a proinsulin processing defect in primary human islets with deficiency of labile iron, perhaps by utilizing iron chelation on the isolated islets. 6. The activation of eIF2a may reflect ER stress; additionally iron status might independently affect eIF2a phosphorylation -these points are poorly developed. At minimum, the authors should recheck phospho-eIF2a levels in IRP2 deficient cells treated ± FAC, as well as a similar experiment after Cdka1 overexpression.

Minor:
7. The Figure 6G legend was omitted. 8. The FAC abbreviation on line 473 is misplaced in the sentence.

Response to reviewers' comments:
We appreciate the thorough critique of our manuscript that we found insightful and constructive. Based on the reviewers' comments and suggestions, we have revised the manuscript. New experiments and responses to reviewers' comments are described below.
New experiments: 2. When impaired Cdkal1 function causes defective modification of the tRNA and consequent misreading of lysine codons is another amino acid inserted in place of lysine, or is the protein frameshifted/truncated? If the latter, wouldn't the amount of pro-insulin be decreased, or at least the amount of normal proinsulin?
Response: Wei et al. (JCI, 121:3598, 2011) reported that loss of Cdkal1 function causes the insertion of another amino acid rather than producing a frameshift/truncated protein. They used a dual luciferase assay to detect frameshifts occurring during decoding of AAA and AAG in Bacillus subtilis wildtype and ΔygeV (Cdkal1 bacterial ortholog) and found no significant frameshift activity in the ΔygeV strain. They found, however, that when constructs were induced with IPTG, translation of AAA and AAG was reduced. They concluded that the ms 2 t 6 A modification is critical to prevent misreading of lysine codons when protein synthesis is high. We discussed this study in the Discussion. We are currently working with a colleague at the university who has experience in mass spec analysis of RNA modifications to identify the amino acid inserted in PI in place of lysine.
3. The description of Fig. 4a doesn't make sense to me. First, there doesn't seem to be any Fth1 in WT cells -is that correct? The amount of Fth1 appears to be increased in Irp2-/-islets but the text in line 159 describes them as decreased. Similarly, for Fig. 4b the text in line 162 describes the level of Fth1 mRNA as increased in Irp2-/-islets but the figure shows it to be decreased. There also appear to be errors in the sentence "Ftl1 mRNA levels were similar …, but, unexpectedly, Fth1 mRNA levels increased, which may be a transcriptional response to compensate for increased Fth1 protein (Fig. 4b)." Response: Sentences were corrected. "Ftl1 was similarly expressed in WT and Irp2 -/islets, while Fth1 was not detected in WT islets, but was notably increased in Irp2 -/islets (Fig. 4a)." "Ftl1 mRNA levels were similar in WT and Irp2 -/islets, but, unexpectedly, Fth1 mRNA levels decreased, which may…" 4. Figure 5a is reported to show a decrease in Tfr1 in cells treated with shIrp2 but that is not apparent in the second lane of the western blot. Could the results be quantitated, or is there a better blot?
Response: A new western blot (Fig. 5a) is included that better demonstrates reduced TfR1 levels in shIrp2 cells.

Nancy Andrews
Reviewer #2 (Remarks to the Author): The present study investigated the role of Irp2 in the regulation of insulin biosynthesis and glucose metabolism. Irp2-knockout mice exhibited glucose intolerance due to an impairment in insulin secretion from pancreatic betacells. The decrease of insulin secretion was resulted from an increase in the proinsulin content and the decrease in the total insulin content in the beta-cells. Mechanistically, Irp2-difficiency suppressed the mRNA and protein levels of transferrin receptor (TfR1) and upregulated the protein level of ferritin, leading to the decrease in the total iron content in the pancreatic beta-cells. The decrease of intracellular iron subsequently impaired the activity of Cdkal1, an iron-sulfur cluster-dependent enzyme required for the methylthio-modification of tRNALys(UUU). Dysregulation of Cdkal1 then impaired proinsulin translation at Lys codon. Importantly, supplement of iron in Irp2-deficient cells enhanced Cdkal1 activity, which improved the proinsulin translation fidelity and increased insulin contents. These results demonstrate that Irp2 is essential for the iron homeostasis and controls translational fidelity of proinsulin via tRNA modification enzyme Cdkal1. Overall, these findings are interesting and novel.
Major concerns: 1. The 5-month old Irp2-knockout mice showed glucose intolerance after intraperitoneal (i.p.) injection of glucose (Figure 1a), but the plasma insulin levels of knockout mice did not differ from wild-type mice at the basal level and after the i.p. injection of glucose (Figure 2a). The plasma insulin data was not consistent with the hyperglycemic clamp data (Figure 2d), which showed that insulin secretion was markedly decreased in Irp2-knockout mice. Authors should explain this discrepancy.

Response:
We repeated plasma insulin measurements in 7-month old WT and Irp2 -/mice after an imp. injection of glucose that showed blunted insulin secretion in Irp2 -/mice (Fig. 2a). In this experiments, plasma insulin levels were measured by a Mouse Insulin ELISA. The discrepancy in the original data is likely due to the use of an RIA KIT to measure plasma insulin, which we no longer use.
2. The finding that Irp2-deficiency resulted in the accumulation of proinsulin is particularly important in this study. Authors should examine the plasma proinsulin level or plasma proinsulin/insulin ratio in Irp2-deficient mice with imp. administration of glucose. Authors also need to examine the proinsulin secretion in isolated islets after glucose stimulation.

Response:
The plasma proinsulin/insulin ratio was quantified in 7.5-month old WT and Irp2 -/mice after an i.p. glucose injection using a proinsulin ELISA (Fig. 3d) and showed increased P/I ratio in Irp2 -/mice vs WT. We also measured proinsulin secretion was in WT and Irp2 -/islets from 7.5 months old mice that shows increased proinsulin secretion in Irp2 -/islets under basal glucose and after stimulation with high glucose (Fig. 3l and n).
Minor concerns: 1. Figure 3e and 3f. Authors showed that the average islet area and the beta-cell mass are comparable between Irp2-knockout mice and wild-type mice. HE staining of pancreas and islets need to be presented to support these data.
Response: H&E staining of representative pancreas sections from 7.5-month WT and Irp2 -/mice is shown in Supplemental Fig. 3a-b. We also included representative insulin-stained pancreatic sections isolated from 7.5month WT and Irp2 -/mice that were used to measure islet area and mass (Fig. 3d-e) 2. Figure 3. There are two "b" panels. The upper right panel should be panel "c". Response: Corrected 3. Figure 4d. Authors claimed that TfR1 staining was reduced in Irp2-knockout beta-cells (page 7, line 164). This is not convincing because the quality of the TfR1 staining is very poor. A majority of beta-cells was not stained by the antibody even in the wild-type islets.

Response:
We repeated immunostaining of our paraffin-embedded pancreatic sections using two other commercial TfR1 antibodies, but the images were not any better than the one in our manuscript. In addition to the antibodies that are not ideal, TfR1 is expressed at much lower levels than insulin, which may affect the image quality. Our graphic designer tried to obtain a better image, which is in the manuscript.
4. Figure 5a. Authors claimed that TfR1 was modestly reduced in Irp2-knockdown INS cells (page 8, line 175). This claim is not correct because the intensity of TfR1 band in Irp2 knockdown cells was comparable to the TfR1 band in control cells (lane 1 versus lane 2). Please do not overstate the result, or author should provide the quantitative data.
Response: A new western blot (Fig. 5a) is included that better demonstrates reduced TfR1 levels in shIrp2 cells.
5. Figure 5b. Please provide the method for this experiment.
Response: A sentence to explain this method is included in the legend of Fig. 5b. "Whole cell lysates from EV and shIrp2 cells were incubated with a 32 P-labeled ferritin IRE RNA probe followed by the resolution of the Irp1-and Irp2-IRE complexes by non-denaturing polyacrylamide gels". A more complete description of the method is included in Materials and Methods (RNA-electrophoretic mobility shift assays (RNA-EMSAs)).