|Home | About | Journals | Submit | Contact Us | Français|
During translation, stop codon read-through occasionally happens when the stop codon is misread, skipped, or mutated, resulting in the production of aberrant proteins with C-terminal extension. These extended proteins are potentially deleterious, but their regulation is poorly understood. Here we show in vitro and in vivo evidence that mouse cFLIP-L with a 46-amino acid extension encoded by a read-through mutant gene is rapidly degraded by the ubiquitin-proteasome system, causing hepatocyte apoptosis during embryogenesis. The extended peptide interacts with an E3 ubiquitin ligase, TRIM21, to induce ubiquitylation of the mutant protein. In humans, 20 read-through mutations are related to hereditary disorders, and extended peptides found in human PNPO and HSD3B2 similarly destabilize these proteins, involving TRIM21 for PNPO degradation. Our findings indicate that degradation of aberrant proteins with C-terminal extension encoded by read-through mutant genes is a mechanism for loss of function resulting in hereditary disorders.
Translation from an mRNA to a protein is normally terminated at stop codons (UAA, UAG, and UGA). However, reading through a stop codon occasionally happens, the frequency of which is affected by the nucleotides around the stop codon (1, 2). When read-through of a stop codon occurs, the translation continues to the next in-frame stop codon or to the poly(A) tail at the 3′ terminus of the mRNA, resulting in the production of aberrant proteins with C-terminal extension. Similar C-terminally extended proteins are constitutively produced when a stop codon is mutated to code for an amino acid; in both cases, this can lead to potentially deleterious proteins.
Translation from mRNAs lacking a stop codon (nonstop mRNA) is suppressed by several mechanisms, including accelerated degradation of nonstop mRNA and translational repression triggered by stalled polysomes in the vicinity of the poly(A) tail (3,–5). However, genome-wide bioinformatics analysis of seven vertebrate animal species indicated that a relatively small number of mRNAs lack a downstream stop codon and that most mRNAs contain additional in-frame stop codons in the 3′-untranslated region (3′-UTR) (Table 1). Therefore, even if a canonical stop codon is mutated to code for an amino acid, translation is terminated at a downstream stop codon in many cases. Currently, more than 400 read-through single nucleotide polymorphisms (SNPs) that result in encoding C-terminally extended proteins terminated at a downstream stop codon have been identified in humans (supplemental Table 1), and some of these SNPs are related to hereditary disorders. However, regulation of these C-terminally extended proteins is poorly understood.
As a model system to study the regulation of the read-through mutant genes, we found a read-through mutation in the long form of mouse cFLIP-L (cellular FLICE-like apoptosis-inhibitory protein) gene that results in a mutant cFLIP-L protein containing a C-terminal 46-amino acid extension (cFLIP-L+46). Structurally, cFLIP-L resembles caspase-8 but has no caspase activity, and therefore, it inhibits apoptosis signaling initiated by death receptor ligation (6). In this study, we demonstrate that homozygous cFlip-l+46 mutant mice die at around 13.5 days postcoitum and undergo extensive apoptosis in the liver. Mechanistic analysis revealed that the extended 46-amino acid peptide functions as a degron to destabilize the cFLIP-L protein and that an E3 ubiquitin ligase, TRIM21 (tripartite motif-containing 21), mediates the ubiquitylation of cFLIP-L+46. We also show that C-terminally extended PNPO (pyridoxamine 5′-phosphate oxidase) and HSD3B2 (3β-hydroxysteroid dehydrogenase type II) proteins, encoded by read-through mutant genes found in patients of congenital diseases, are degraded by a similar mechanism. These findings indicate that degradation of aberrant C-terminally extended proteins that are encoded by genes with read-through mutations is a mechanism for loss of function and can lead to hereditary disorders.
Tissue culture plastics were purchased from Greiner Bio-One (Tokyo, Japan). Cycloheximide (CHX)2 and anti-FLAG-agarose (M2) were from Sigma-Aldrich. MG132 was from Peptide Institute (Osaka, Japan).Anti-Myc-agarose (9E10) was from Santa Cruz Biotechnology, Inc. Cell Counting Kit-8 was from Dojindo Laboratories (Kumamoto, Japan). TNFα was provided by Prof. Tsujimoto (Teikyo Heisei University, Tokyo, Japan). cDNAs encoding mouse cFLIP-L, human PNPO, and HSD3B2 were amplified by PCR and cloned into the p3xFLAG-CMV-10 (Sigma-Aldrich) or pcDNA3 expression vectors (Invitrogen). cDNAs encoding extended peptides were PCR-amplified and inserted into the pEGFP-C2 expression vector (Clontech). pcDNA4/His/LacZ was purchased from Invitrogen. A human TRIM21 expression vector was provided by Prof. Hatakeyama (Hokkaido University, Sapporo, Japan). To generate siRNA-resistant TRIM21 expression vectors, synonymous nucleotide substitutions were introduced using PfuUltra high-fidelity DNA polymerase (Agilent Technologies) according to the manufacturer's protocol.
Animal experiments were approved by the Animal Experiment Committee of the National Institute of Health Sciences, Japan. N-Ethyl-N-nitrosourea (ENU)-mutagenized cFlip-l+46 mutant mouse lines were originally generated by the RIKEN BioResource Center, as described previously (7). cFlip-deficient mice were provided by Dr. Yeh (Amgen Inc.) (8). All mice were housed under specific pathogen-free conditions.
Embryos at E12.5 were fixed in 10% neutral phosphate-buffered formalin, embedded in paraffin, sectioned at 6 μm, and stained with hematoxylin-eosin. For TUNEL staining, sections were stained using an in situ apoptosis detection kit (Takara Bio, Shiga, Japan).
Mouse embryonic fibroblasts (MEFs) and HeLa cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10% fetal bovine serum (Gibco), and 50 μg/ml kanamycin (Sigma-Aldrich). Transient transfections were carried out using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. siRNA transfections were carried out using Lipofectamine RNAi/MAX (Invitrogen) according to the manufacturer's instructions. Typically, cells were transfected, and 48 h later, the cells were treated with 10 μm MG132 for 6 h. Target sequences for mouse cFlip-l, human TRIM21, and mouse Trim21 were as follows (5′ to 3′): mouse cFlip-l, GAATAGACTTGAACACAAA; human TRIM21 1, AAGCAGGAGTTGGCTGAGAAG; human TRIM21 2, GCTCCCTCATCTACTCCTT; mouse Trim21 1, CCTAAACCCTCATCTCTTT; mouse Trim21 2, CCTGGACACGTTAGATATT.
For immunoblot analysis, cells were lysed in a lysis buffer containing 0.1 m Tris/HCl (pH 7.5), 1% SDS, 10% glycerol. For coimmunoprecipitation, cells were lysed in a lysis buffer containing 0.02 m Tris/HCl (pH 7.5), 0.2% Nonidet P-40, 0.15 m NaCl, 10% glycerol, and 1× complete protease inhibitor mixture (Roche Applied Science), and the lysates were immunoprecipitated with the indicated antibodies. Lysates or immunoprecipitates were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. The following antibodies were used to detect proteins: anti-cFLIP-L (Dave-2) (Enzo Life Sciences, Plymouth Meeting, PA); anti-HSP90 (catalog no. 610419) (BD Transduction); anti-cleaved caspase-3 (catalog no. 9664), anti-β-catenin (catalog no. 9562), and anti-Myc (catalog no. 2276) (Cell Signaling); anti-human TRIM21 (12108–1-AP) (Proteintech Group, Chicago, IL); anti-β-tubulin (ab6046) (Abcam); anti-FLAG (M2) (Sigma-Aldrich); anti-HA (3F10) (Roche Applied Science); and anti-GAPDH (FL-335), anti-β-actin (C-2), anti-GFP (B-2), anti-His (H-3), and anti-mouse TRIM21 (M-20) (Santa Cruz Biotechnology).
Apoptosis was analyzed with an annexin V-FITC apoptosis detection kit (BioVision, Milpitas, CA), as described previously (9). Briefly, after treatment, cells were gently trypsinized and washed with serum-containing medium. Cells were collected by centrifugation, additionally washed with PBS, and resuspended in binding buffer. The cells were stained with annexin V-FITC and propidium iodide at room temperature for 5 min in the dark, according to the manufacturer's instructions, and analyzed on a FACScan flow cytometer (BD Biosciences).
HeLa cells transfected with the indicated vectors were treated with 10 μm MG132 and lysed in lysis buffer containing 0.1 m Tris/HCl (pH 7.5), 1% SDS, 10% glycerol. After heat denaturation, the lysates were diluted 10 times with 0.1 m Tris/HCl (pH 7.5). Proteins were immunoprecipitated with anti-FLAG or anti-Myc agarose-conjugated beads, and the immunoprecipitates were analyzed by Western blotting with anti-HA.
Lysates from HeLa cells expressing GFP+46 or GFP+22 were immunoprecipitated with antibody against GFP. Following SDS-PAGE, gels were stained with SYPRO ruby protein gel stain (Lonza, Rockland, ME), and protein bands that specifically bind to the GFP+46 were excised from the gel and subjected to in-gel trypsin digestion. The peptides extracted from the gel pieces were analyzed by LC/MS/MS using an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) connected to a Paradigm MS4 HPLC system (Michrom BioResources, Auburn, CA). The analytical column for LC/MS was a reversed-phase column (L-column 2 ODS; 150 × 0.075 mm, 3 μm; Chemicals Evaluation and Research Institute, Tokyo, Japan). The mobile phase was 0.1% formic acid containing 2% acetonitrile and 0.1% formic acid containing 90% acetonitrile (A buffer). The peptides were eluted at a flow rate of 300 nl/min with a gradient of 2–65% of A buffer over 50 min. MS/MS conditions were as follows: an electrospray voltage of 2.5 kV in positive ion mode, a capillary temperature of 275 °C, and a collision energy of 35% for MS/MS. The spectra data obtained by MS/MS were subjected to database search analysis with the SEQUEST algorithm (Proteome Discoverer version 1.4, Thermo Fisher Scientific) using the UniProt database.
Total RNA was prepared from MEFs with RNeasy (Qiagen). First strand cDNA was synthesized from 1 μg of total RNA with an oligo(dT) primer using the SuperScript first-strand synthesis system (Invitrogen). Quantitative real-time PCR was performed with an ABI Prism 7300 sequence detection system using TaqMan real-time PCR master mix (Applied Biosystems, Foster City, CA) with TaqMan probes/primers for mouse cFlip-l and 18S ribosome RNA (Applied Biosystems). The relative amounts of cFlip-l mRNAs were calculated by using the comparative Ct method. Mouse 18S ribosomal RNA was used as an invariant control.
Data for human, mice, rat, cow, pig, frog, and zebrafish full-length mRNAs were retrieved from the NCBI RefSeq database in January 2015. Read-through SNP data and associated full-length human mRNA data were retrieved from dbSNP and RefSeq in the NCBI database in March 2014. Stop mRNAs that contain an in-frame termination codon in the 3′-UTR were selected. 3′-UTR sequences were translated in silico using a Perl translate module.
In a large scale archive of ENU-mutagenized mice (10), we found a read-through mutation (X482W) in the cFlip-l gene that results in encoding a mutant cFLIP-L protein containing a C-terminal 46-amino acid extension (cFLIP-L+46) (Fig. 1A). To understand how the extended protein is regulated and the phenotypic consequences of the read-through mutation, we analyzed the mutant mice and the cFLIP-L+46 protein.
Heterozygous cFlip-l+46 mutant (+/Mut) mice developed normally and were fertile, but homozygous mutant (Mut/Mut) mice died at around E13.5 (Fig. 1B and Table 2A). To exclude the possibility that abnormalities found in the mutant (Mut/Mut) mice are due to ENU-generated mutation of another, unknown gene that happened to be homozygous in the cFlip-l+46 Mut/Mut mice, we crossed heterozygous cFlip-l+46 mutant (+/Mut) mice onto heterozygous cFlip-deficient (+/−) mice. The resulting transheterozygote (−/Mut) mice died at around E11.5 (Table 2B), which is slightly later than the embryonic lethality observed in cFlip null (−/−) mice (Table 2C). Because any mutation except for cFlip-l is heterozygous in the transheterozygotes and gives no phenotypic changes, the lethality observed in the transheterozygote embryos is due to the complication in the cFlip-l gene. Accordingly, these results indicate that the homozygous read-through mutation of the cFlip-l gene is embryonic lethal. Histological analysis of cFlip-l+46 Mut/Mut embryos at E12.5 showed hemorrhage with many condensed nuclei and TUNEL-positive cells in the liver (Fig. 1C). Thus, homozygous cFlip-l read-through mutant (Mut/Mut) embryos survived beyond E10.5 when cFlip null mice die because of impaired yolk sac vasculature development (8, 11) but eventually died at around E13.5, probably because of extensive apoptosis in the liver. Because the extended cFLIP-L+46 protein is destabilized (see below), these results suggest that a tiny amount of cFLIP-L+46 protein enables the embryos to survive beyond E10.5 but that it is not sufficient to suppress hepatocyte apoptosis at a later embryonic stage.
Next, we analyzed the behavior of the cFLIP-L+46 protein. Levels of cFLIP-L+46 protein were dramatically reduced in the Mut/Mut whole embryo (Fig. 2A) and liver (Fig. 2B) at E12.5. In the liver, the active form of caspase-3 was observed (Fig. 2B), which is in good agreement with the observed hepatocyte apoptosis (Fig. 1C). Reduced cFLIP-L+46 protein levels were similarly observed in Mut/Mut MEFs (Fig. 2C). Consistent with the reduced levels of cFLIP-L+46 protein, the Mut/Mut MEFs were highly sensitive to TNFα-induced apoptosis (Fig. 2D) as were wild-type MEFs depleted of cFLIP-L using siRNA (Fig. 2E).
To examine whether the residual cFLIP-L+46 can inhibit apoptosis signaling in Mut/Mut MEFs, we down-regulated cFLIP-L+46 in Mut/Mut MEFs using siRNA. As shown in Fig. 3, A and B, depletion of the residual cFLIP-L+46 protein significantly sensitized the cells to TNFα-induced apoptosis. Conversely, exogenous expression of cFLIP-L+46 protein in Mut/Mut MEFs suppressed TNFα-induced caspase activation (Fig. 3C) and increase of annexin V-positive apoptotic cells (Fig. 3D). These results indicate that cFLIP-L+46 protein has an activity to inhibit apoptosis but that the amount of cFLIP-L+46 protein is severely reduced, resulting in higher sensitivity to apoptosis in the Mut/Mut MEFs and hepatocytes.
We next investigated why levels of cFLIP-L+46 protein are severely reduced. In MEFs, the turnover of cFLIP-L+46 protein was much faster than that of wild-type cFLIP-L protein. When the MEFs were treated with CHX, cFLIP-L+46 protein disappeared within 1 h, whereas wild-type cFLIP-L protein was retained for over 12 h (Fig. 4A). Similarly, cFLIP-L+46 exogenously expressed in HeLa cells turned over more rapidly than wild-type cFLIP-L (Fig. 4B). Treatment with a proteasome inhibitor, MG132, increased the accumulation of cFLIP-L+46 in Mut/Mut MEFs (Fig. 4C). In addition, ubiquitylation of cFLIP-L+46 exogenously expressed in HeLa cells was highly enhanced compared with that of wild-type cFLIP-L (Fig. 4D). On the other hand, levels of cFlip-l+46 mRNA in Mut/Mut MEFs were not reduced compared with cFlip-l mRNA in wild-type (WT) MEFs (Fig. 4E). These results collectively indicate that the expression of cFLIP-L+46 protein is severely reduced because of the rapid degradation by the UPS.
To examine whether the extended peptide can promote proteasomal degradation when grafted onto a heterologous protein, we genetically fused the extended peptide to the C terminus of GFP and expressed it in HeLa cells. GFP+46 levels were much lower than those of GFP, and this difference was reduced by proteasome inhibitor MG132 (Fig. 5A), indicating that the extended peptide functions as a degradation signal. Trimming 10 C-terminal amino acids from the extended peptide to form GFP+36 did not affect the degradation signal activity, whereas deletion of 24 C-terminal amino acids to form GFP+22 abolished the activity (Fig. 5A).
Because the extended peptide does not contain a lysine residue that can be ubiquitylated, we reasoned that the peptide functions as a degron that interacts with a ubiquitin ligase to destabilize cFLIP-L protein. To identify the ubiquitin ligase involved in the destabilization of cFLIP-L+46 protein, we searched for proteins that interact with GFP+46 but not with GFP+22 by liquid chromatography-mass spectrometry and identified TRIM21, a RING-finger E3 ubiquitin ligase involved in innate and acquired immunity (12,–14). Fig. 5B shows that GFP+46, but not GFP+22, co-precipitates TRIM21 protein in the cells. In addition, cFLIP-L+46 strongly, but cFLIP-L minimally, co-precipitates TRIM21 in the cells (Fig. 5C). These results confirmed the selective interaction of TRIM21 with the extended peptide harboring the degron activity.
We next investigated whether TRIM21 is required for the degradation of GFP+46 and cFLIP-L+46 proteins. siRNA-mediated down-regulation of human TRIM21 in HeLa cells increased the levels of GFP+46, which was suppressed by restoring TRIM21 levels by co-transfecting an siRNA-resistant FLAG-tagged human TRIM21 construct (Fig. 5D). In addition, knockdown of mouse Trim21 in cFlip-l+46 Mut/Mut MEFs significantly increased the level of cFLIP-L+46 (Fig. 5E). The rapid protein turnover of cFLIP-L+46 exogenously expressed in HeLa cells was suppressed by knockdown of human TRIM21 (Fig. 5F). Furthermore, the ubiquitylation of cFLIP-L+46 protein was suppressed in cells depleted of TRIM21 and restored by co-transfection of the siRNA-resistant Trim21 gene (Fig. 5G). These results indicate that TRIM21 is a ubiquitin ligase responsible for the destabilization of cFLIP-L+46 and GFP+46 proteins.
In humans, more than 400 read-through SNPs result in encoding C-terminally extended proteins that terminate at downstream in-frame stop codons (supplemental Table 1). Among them, 20 are related to hereditary disorders (Table 3). We hypothesized that some of the read-through mutant proteins could be regulated by a mechanism similar to that which degrades cFLIP-L+46 protein. To test this hypothesis, we examined in HeLa cells the expression of GFP proteins fused to the extended peptides encoded by the 20 genes involved with hereditary disorders. We found that the extended peptides from PNPO (28 amino acids) and HSD3B2 (95 amino acids) destabilized the GFP protein, which was abrogated by MG132 (Fig. 6A).
Consistent with this, levels of PNPO+28 expressed in HeLa cells were suppressed compared with wild-type PNPO, and this suppression was abrogated by MG132 (Fig. 6B, left panels). PNPO+28 protein turned over more rapidly than wild-type PNPO (Fig. 6C, left panels), and PNPO+28 was more heavily ubiquitylated compared with wild-type PNPO (Fig. 6D, left panels). Similar results were obtained with HSD3B2+95 and the corresponding wild-type HSD3B2 proteins (Fig. 6, B–D, right panels). These results indicate that C-terminal extension of PNPO and HSD3B2 proteins caused by read-through mutations results in protein instability through the UPS.
Finally, we examined whether TRIM21 is involved in the destabilization of the extended PNPO+28 and HSD3B2+95 proteins. siRNA-mediated knockdown of human TRIM21 increased the levels of PNPO+28, but not of HSD3B2+95 protein (Fig. 6E). Furthermore, the rapid turnover of PNPO+28 protein was suppressed by knockdown of human TRIM21 (Fig. 6F), and TRIM21 strongly interacted with PNPO+28 in the cells (Fig. 6G). These results suggest that TRIM21 plays a role in the ubiquitylation of PNPO+28 but that another E3 ligase is responsible for the ubiquitylation of HSD3B2+95 protein.
Currently, 20 stop codon read-through mutations are related to hereditary disorders in humans; however, it is not understood how the C-terminally extended read-through mutant proteins are regulated. In this study, we analyzed a read-through mutant of mouse cFlip-l as a model system and showed that the homozygous read-through mutant embryos undergo extensive apoptosis in the liver. The C-terminally extended peptide interacts with a ubiquitin ligase TRIM21 and promotes ubiquitylation and proteasomal degradation of cFLIP-L+46 protein. We also showed that the C-terminally extended proteins of human PNPO and HSD3B2 encoded by read-through mutant genes found in patients are degraded by a similar mechanism involving TRIM21 for PNPO degradation.
PNPO plays an essential role in brain metabolism, catalyzing the conversion of pyridoxine 5′-phosphate and pyridoxamine 5′-phosphate to pyridoxal 5′-phosphate, a metabolically active form of vitamin B6 (15). Patients with homozygous read-through mutation in the PNPO gene (X262Q) suffer from neonatal epileptic encephalopathy (16) and show no PNPO activity. The null PNPO activity can be explained by degradation of the extended PNPO+28 protein via the UPS. Because TRIM21 plays an important role in the degradation of the extended PNPO+28 protein (Fig. 6, E and F), inhibition and/or suppression of TRIM21 could be a strategy to restore PNPO protein levels in these cases.
HSD3B2 catalyzes the oxidation and isomerization of Δ5-3β-hydroxysteroid precursors into Δ4-ketosteroids and is essential for the formation of progesterone, a precursor for all classes of steroid hormones. Deficiency in the activity of HSD3B2 in the adrenal cortex causes congenital adrenal hyperplasia (17). A read-through mutation in HSD3B2 (X373C) was found in patients with congenital adrenal hyperplasia, resulting in dramatic reduction of HSD3B2 protein levels and activity (18). It is likely that the mutant protein is degraded by the UPS, although the E3 ligase involved is currently unknown.
Among 20 read-through mutant proteins that are related to hereditary disorders in humans, C-terminally extended PNPO+28 and HSD3B2+95 proteins are degraded by the UPS as well as mouse cFLIP-L+46 protein. Interestingly, the extended peptides of PNPO+28, HSD3B2+95, and cFLIP-L+46 share the Leu-Xaa-Xaa-Leu-Leu (LXXLL) sequence, whereas other extended peptides encoded by the mutant genes in Table 3 do not have the sequence. The LXXLL sequence is well known as a protein recognition motif widely used in transcriptional regulation (19). To investigate whether the LXXLL sequence is involved in the degradation of C-terminally extended proteins, we generated point or deletion mutants of the sequence in cFLIP-L+46, but the mutations did not affect the protein degradation of cFLIP-L+46 (data not shown). Therefore, the LXXLL sequence in the C-terminally extended proteins is not likely to play an important role in regulating the protein stability.
The 3′-UTR is known to regulate the stability of mRNAs and the translation (20) but is rarely considered to encode functional protein sequence. However, in some cases, functional peptides are encoded in the 3′-UTR, downstream of the canonical termination codon, and they can influence the function of the mother protein. Examples include an isoform of vascular endothelial growth factor A (VEGF-Ax) and a read-through mutant of CRYM. VEGF-Ax with a C-terminal 22-amino acid extension is generated by a programmed translational read-through mechanism depending on heterogeneous nuclear ribonucleoprotein A2/B1. The extended peptide converts a proangiogenic VEGF-A into an anti-angiogenic VEGF-Ax protein (21). A read-through mutation in the CRYM gene is found in patients with non-syndromic deafness, and subcellular localization of CRYM protein is altered by the C-terminal extension (22). In addition to these observations, our study provides evidence for the destabilization of proteins by C-terminally extended peptides encoded in the 3′-UTRs of human PNPO and HSD3B2, which result in congenital diseases, and in mouse cFlip-l, which causes embryonic lethality. Because there are more read-through mutations known to be involved in human hereditary disorders (Table 3), further study will clarify the function of the extended peptides and how the C-terminally extended proteins are regulated.
N. S. and M. N. designed the experiments and wrote the manuscript. Y. Sugaki performed bioinformatics analysis. N. S., N. O., Y. Sugaki, C. O., M. I., and M. N. performed the experiments. Y. Sakuraba and Y. G. generated cFlip-l+46 mutant mice. D. T., N. H., and N. K. performed LC/MS experiments. M. N. supervised all research. All authors approved the final version of the manuscript.
We thank Drs. T. Hattori and K. Okuhira for helpful discussions. We thank Dr. W.-C. Yeh for kindly providing cFlip-deficient mice, Prof. Hatakeyama for FLAG-TRIM21 plasmid, and Prof. Tsujimoto for TNFα. Histological analysis was carried out by Genostaff Co., Ltd. (Tokyo, Japan).
*This work was supported by Japan Society for the Promotion of Science KAKENHI Grants 25112521 (to M. N.), 26860050 (to N. S.), and 25241016 (to Y. G.) and by Health and Labor Sciences Research Grant 26401201 (to N. K.). The authors declare that they have no conflicts of interest with the contents of this article.
This article contains supplemental Table 1.
2The abbreviations used are: