|Home | About | Journals | Submit | Contact Us | Français|
The mechanism underlying the sensing of varying degrees of physiological folate deficiency, prior to adaptive optimization of cellular folate uptake through the translational up-regulation of folate receptors (FR) is unclear. Because homocysteine, which accumulates intracellularly during folate deficiency, stimulated interactions between heterogeneous nuclear ribonucleoprotein E1 (hnRNP-E1) and an 18-base FR-α mRNA cis-element that led to increased FR biosynthesis and net up-regulation of FR at cell surfaces, hnRNP-E1 was a plausible candidate sensor of folate deficiency. Accordingly, using purified components, we evaluated the physiological basis whereby l-homocysteine triggered these RNA-protein interactions to stimulate FR biosynthesis. l-Homocysteine induced a concentration-dependent increase in RNA-protein binding affinity throughout the range of physiological folate deficiency, which correlated with a proportionate increase in translation of FR in vitro and in cultured human cells. Targeted reduction of newly synthesized hnRNP-E1 proteins by siRNA to hnRNP-E1 mRNA reduced both constitutive and l-homocysteine-induced rates of FR biosynthesis. Furthermore, l-homocysteine covalently bound hnRNP-E1 via multiple protein-cysteine-S-S-homocysteine mixed disulfide bonds within K-homology domains known to interact with mRNA. These data suggest that a concentration-dependent, sequential disruption of critical cysteine-S-S-cysteine bonds by covalently bound l-homocysteine progressively unmasks an underlying RNA-binding pocket in hnRNP-E1 to optimize interaction with FR-α mRNA cis-element preparatory to FR up-regulation. Collectively, such data incriminate hnRNP-E1 as a physiologically relevant, sensitive, cellular sensor of folate deficiency. Because diverse mammalian and viral mRNAs also interact with this RNA-binding domain with functional consequences to their protein expression, homocysteinylated hnRNP-E1 also appears well positioned to orchestrate a novel, nutrition-sensitive (homocysteine-responsive), posttranscriptional RNA operon in folate-deficient cells.
Folates are essential for the synthesis of DNA, so all proliferating human cells are dependent on the adequate provision of dietary folate (1). High affinity folate receptors (FR),2 which were first introduced to the biomedical literature 30 years ago (2), are critical for the cellular uptake of 5-methyltetrahydrofolate in several human (normal and cancer) cells (3). Among three human FR (FR-α, FR-β, and FR-γ), the FR-α are ubiquitous and physiologically relevant for proper development of the neural tube and neural crest (4, 5), for hematopoiesis (6), for transplacental maternal-to-fetal folate transport (7), for cerebral folate transport (8), and for renal conservation of folates (3, 9, 10).
Because nutritional folate deficiency induced a marked (100-fold) increase in cell surface FR protein without a proportionate increase in FR mRNA in a cervical cancer cell line, we pursued the possibility of posttranscriptional regulation of FR. Initially, we identified an interaction between an 18-base cis-element in the 5′-UTR of FR-α mRNA and a trans-factor protein that was critical for the biosynthesis of FR (11). This trans-factor was later identified as hnRNP-E1 (12), which is also known as poly(rC)-binding protein 1 (PCBP1) and αCP1 (13–23). Subsequent studies led to a model that established a linkage between perturbed folate metabolism and coordinated up-regulation of FR at the translational level (24); in this scheme, the accumulated intracellular homocysteine resulting from folate deficiency triggered the interaction of FR-α mRNA cis-element with hnRNP-E1, which then stimulated the biosynthesis and up-regulation of FR (24). However, the physiological relevance remained unclear because such RNA-protein interactions must be highly responsive to even small increases of homocysteine beyond 14 μm, as would be found during mild folate deficiency (1), but we could only consistently detect RNA-protein interactions at dl-homocysteine concentrations of 25 μm and beyond (24). Whether this stemmed from the use of dl-homocysteine rather than physiological l-homocysteine was not clear. Moreover, we had not yet captured cellular RNA-protein complexes in direct response to the small increases in intracellular concentration of l-homocysteine found in mild folate deficiency. Furthermore, the underlying molecular mechanism(s) whereby physiological concentrations of homocysteine triggered these RNA-protein interactions was also obscure. The latter issue is of particular importance because the binding of FR-α mRNA cis-element by hnRNP-E1 is a necessary prelude to its translation (11). Indeed, the factors and events that trigger the initial RNA-protein interaction can be operationally conceived as comprising the “afferent (sensing) limb” of translation, whereas all subsequent events (leading to recruitment of the ribosome and initiation of polypeptide synthesis) constitute the more complex “efferent (executing) limb” of translation. In this context, any entity that can recognize a fall in intracellular folate with a high degree of sensitivity and respond by facilitating the restoration of folate levels toward normalcy by up-regulating FR would fulfill criteria of a bona fide “folate sensor.”
Therefore, to increase understanding of the molecular basis for an induction of the RNA-protein interaction (reflecting the afferent sensing limb of translation of FR), we tested the hypothesis that hnRNP-E1 incurred one or more dose-dependent posttranslational modifications as a result of the increasing intracellular concentrations of homocysteine accompanying physiological cellular folate deficiency and that this led to a progressive increase in affinity of homocysteinylated hnRNP-E1 for binding FR-α mRNA cis-element, which determined the extent of translation of FR. Collectively, our results suggest that hnRNP-E1 fulfills the characteristics of a physiological sensor of folate deficiency and is an intrinsic component of the afferent limb of translation of FR and shed new light on the molecular basis underlying the physiological activation of a nutritional folate deficiency-sensitive (and homocysteine-responsive) posttranscriptional RNA operon that is orchestrated by homocysteinylated hnRNP-E1.
Folate-replete HeLa-IU1-HF cells and folate-deficient HeLa-IU1-LF cells were obtained (25) and propagated as described (24). Briefly, HeLa-IU1-HF cells were propagated long term at 37 °C in 5% CO2 in high folate minimum essential medium (MEM-HF) (Invitrogen) containing 10% nondialyzed fetal bovine serum and 9 nm 5-methyltetrahydrofolate plus 2.3 μm pteroylglutamate. HeLa-IU1-LF cells, which were derived from HeLa-IU1-HF cells, were stably propagated in low folate MEM-LF medium, which contained 10% nondialyzed fetal bovine serum added to folate-free MEM, that resulted in a final concentration of 9 nm 5-methyltetrahydrofolate.
1 × 105 cpm of each of the 32P-labeled mRNA cis-elements (see supplemental Methods) were reacted with 0.5 μg of purified recombinant glutathione S-transferase (GST)-hnRNP-E1 protein in the absence and presence of increasing concentrations (0–100 μm) of dl- or l-homocysteine followed by analysis of RNA-protein complexes by native PAGE (60:1), autoradiography, and densitometric analysis (24).
To determine the influence of homocysteine on the dissociation constant (KD) of the RNA-protein interaction, binding assays between 0.1 μg of purified recombinant GST-hnRNP-E1 and 35S-labeled 18-base FR-α mRNA cis-element (0–12 nm) were carried out in a final volume of 750 μl of binding buffer in the absence and presence of increasing concentrations (0–100 μm) of dl- or l-homocysteine at 4 °C for 1 h. BSA (0.1 μg) was used as a control in place of hnRNP-E1 for background determination. The mixture was then filtered through a Microcon YM-30 column by centrifugation at 12,000 × g followed by three consecutive washes, each with 500 μl of binding buffer. The retentate containing RNA-protein complexes (in 50 μl) was counted in a liquid scintillation counter. Counts from samples containing hnRNP-E1 reflected total binding, and those from samples containing BSA reflected nonspecific binding. Specific binding for each concentration of radioligand was determined by subtracting the values of nonspecific binding from those of total binding. The KD was calculated from a Scatchard plot (26) using GraphPad Prism 4 from GraphPad Software (San Diego, CA).
The binding affinity of 35S-labeled 18-base FR-α mRNA cis-element and purified recombinant GST-hnRNP-E1 in the presence of various thiol amino acids was also determined to compare the influence of l-cysteine and l-homocysteine on the KD of this RNA-protein interaction (Table 1).
XhoI-linearized pSPTFR (0.5 μg/reaction) was added to the transcription mixture and incubated at 30 °C for 15 min. After adding 1 μl of N-ethylmaleimide (to achieve a final concentration of 2 mm) for 15 min, 1 μl of dl- or l-homocysteine (to achieve final concentrations of 0–100 μm), 8 μl of transcription reaction mixture, and 1.6 μl of [35S]methionine were added to 38.4 μl of the translation mixture followed by incubation for 1 h at 30 °C and analysis of the translation product by 10% SDS-PAGE and autoradiography.
The pXL37 plasmid DNA and pYST19 plasmid DNA were both linearized first by EcoRI. After inclusion of [α-35S]UTP in each transcription reaction, transcripts containing 35S-labeled antisense RNA to 18-base FR-α mRNA cis-element or 35S-labeled antisense RNA to 19-base 15-lipoxygenase mRNA cis-element were purified.
HeLa-IU1-HF cells cultured at ~80% confluence in 100-mm plates were treated for 2 h at 37 °C in MEM-HF (24) containing increasing concentrations of l-homocysteine (0, 12.5, 25, and 50 μm). After removal of medium, the plates (20 plates for each l-homocysteine concentration) were irradiated once at 120 mJ/cm2 in a UV Stratalinker 1800 cross-linker set at auto-cross-link mode (Stratagene) with the cover off to enable RNA-binding proteins to cross-link bound RNA (27). The cells were then harvested, and those from each concentration of l-homocysteine used were pooled and centrifuged at 1000 × g, and pellets were frozen at −80 °C. The cross-linked RNA·hnRNP-E1 protein complexes from S-100 cytosolic proteins derived from these cell pellets were then isolated (11) and specifically immunoprecipitated using anti-hnRNP-E1 antiserum (described below). The pellets from these immunoprecipitates were then treated with 100 μl of RNase A (1:5000 dilution) for 15 min and washed four times with 1 ml of Buffer A (11). (The addition of RNase led to cleavage of RNA that was not protected through direct cross-linking with hnRNP-E1.) The pellet was finally resuspended in 0.2 ml of Proteinase K (4 mg/ml), and the mixture was incubated at 37 °C for 18 h on a shaker. (This step led to the release of small RNA fragments that were originally cross-linked to hnRNP-E1 in response to various concentrations of l-homocysteine (27).) Next, extraction of these small RNA fragments was carried out using the PureLink microRNA isolation kit (Invitrogen). The supernatant (control) sample to which non-immune serum had been added was also treated similarly. The final concentration of total extracted cellular RNA was determined using a NanoDrop spectrophotometer ND-1000 (NanoDrop Technologies, Wilmington, DE), which allowed measurement of RNA concentration in a volume of 1 μl. Whereas a sufficient amount of small RNA fragments (for slot blot hybridization) was recovered from the anti-hnRNP-E1 antiserum-treated sample, the RNA concentration of non-immune serum-treated sample was undetectable.
Equal amounts (maximum of 100 ng) of small RNA fragments that originally reacted with hnRNP-E1 in response to l-homocysteine were isolated, as described above, together with control plasmids containing the cDNAs of either 18-base sense FR-α cis-element, 19-base 15-lipoxygenase cis-element, or HPV-16 L2 cis-element were transferred to Nylon Membranes using a Hybri-SlotTM Manifold (BRL Invitrogen). In addition, the entire sample representing RNA that was eluted from non-immune serum samples was also transferred to nylon membranes. After UV cross-linking of transferred RNA to the membrane (27), these were tested for hybridization with 35S-labeled antisense probes to either the 18-base FR-α mRNA cis-element or the 19-base 15-lipoxygenase mRNA cis-element (28). Briefly, membranes for RNA cross-linking were wet in 1× TBE buffer before insertion into Stratagene hybridization tubes. Then 20 ml of hybridization buffer (6% SDS, 200 mm Na2HPO4, pH 7.2) was added, and prehybridization was carried out at 60 °C for 2 h in a hybridization oven, followed by the addition of relevant probes and incubation at 60 °C overnight. The blots were washed sequentially three times with 25 ml of prewarmed 2× SSC buffer at 60 °C for 20 min, followed by 25 ml of prewarmed 1× SSC buffer at 60 °C for 20 min before autoradiography and densitometric scanning analysis.
Prior studies have employed RNAi to reduce hnRNP-E1/PCBP1 in Huh7 cells (29). Because HeLa-IU1-HF cells have a basal expression rate of FR that is further increased by l-homocysteine at the translational level (24), we transfected these cells with siRNA directed against hnRNP-E1/PCBP1 mRNA (29) and a non-targeted scrambled RNA sequence (from Invitrogen). Two days after transfection, when the hnRNP-E1 mRNA level was reduced by 72%, we assessed the biosynthesis of newly synthesized hnRNP-E1 and FR (24). In a second set of experiments, the same procedure was followed, except that 2 h before assessment of the biosynthesis of FR, cells were treated with 50 μm l-homocysteine. Briefly, HeLa-IU1-HF cells were trypsinized, counted, and plated in 6-well dishes with 1.5 × 105 cells in serum-free MEM-HF. Transfection using Lipofectamine RNAiMAX was carried out on the second day, when cells achieved a confluence of ~30%, by following the manufacturer's protocol (Invitrogen). This transfection reaction mixture, containing either 180 nm siRNA duplex or scrambled control and 5 μl of Lipofectamine RNAiMAX in a final volume of 500 μl of Opti-MEM, was first incubated at 22 °C for 30 min and then added to cells containing 2.5 ml of MEM-HF at 37 °C. After 6 h, the medium was replaced with MEM-HF, and incubation continued until cells were harvested on each day until the fifth day. RNA isolated from these cells was tested by quantitative real-time RT-PCR (qRT-PCR) to determine the extent of hnRNP-E1 mRNA inhibition. The hnRNP-E1 mRNA level was determined by qRT-PCR using the primers 5′-GGATATGCTGCCCAACTCCA-3′ and 5-Carboxyfluorescein isothiocyanate-labeled 5′-GACGCCGGAGACTGGGAGAGCG5C-3′ (Invitrogen). Primers for a housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (GADPH), which was labeled with 6-Carboxy-4′,5′-dichloro-2′,7′-dimethoxyfluorescein (Invitrogen) and used as the internal control, were run in parallel to standardize the input. In other experiments, hnRNP-E2 mRNA levels were also determined on various days by qRT-PCR following transfection of HeLa-IU1-HF cells with siRNA against hnRNP-E1 mRNA (supplemental Fig. S4). qRT-PCR was performed on a TaqMan ABI 7900HT sequence detection system (PE Biosystems, Forest City, CA) using the SuperScript III Platinum one-step qRT-PCR kit (Invitrogen). Cycle conditions were as follows. After the first step of 15 min at 50 °C and 10 min at 95 °C, samples were cycled 40 times at 95 °C for 15 s and at 60 °C for 60 s. For all quantitative analyses, we used the comparative CT method using instructions from PE Biosystems. All PCRs were performed in triplicate for at least three independent studies. Controls consisting of RNase-free water were negative in all runs.
The biosynthesis of FR was assessed on the second day after transfection of siRNA directed against hnRNP-E1 mRNA or non-targeted scrambled RNA sequence (24). Briefly, cells were starved overnight in cysteine-free MEM-HF and exposed to 2 ml of cysteine-free medium either without or with 50 μm l-homocysteine for 2 h the next day. Cells were then pulsed with 100 μCi of l-[35S]cysteine in 1 ml of cysteine-free MEM-HF/well and harvested at various times (0.5, 1, 2, and 4 h), and the biosynthetic rate of FR was assessed after immunoprecipitation using anti-FR antiserum (24). Similar methods were also employed to evaluate the biosynthesis of hnRNP-E1 using anti-hnRNP-E1 antiserum (12) substituted for anti-FR antiserum (24).
We determined if less l-[35S]homocysteine was bound to hnRNP-E1 isolated from cells stably propagated in low folate medium (HeLa-IU1-LF cells) (24) compared with cells in high folate medium (HeLa-IU1-HF cells), where endogenous cellular homocysteine was lower (i.e. 126 μm versus 47 μm, respectively) (24). Our anti-hnRNP-E1 peptide antiserum (12) developed against the hnRNP-E1 peptide sequence SLAQYLINARLSSEKGMGC contained shared sequences (underlined) with hnRNP-E2 (a closely related protein with 83% homology in its total amino acid sequence), predicting that this antiserum would recognize shared epitopes in both proteins, which are expressed in these HeLa cells (30). Indeed, Western blots of cytosolic S-100 fractions of HeLa-IU1 cells using this antiserum (see Fig. 5B) confirmed a major band corresponding to hnRNP-E1 and a minor band corresponding to hnRNP-E2 (and no other contaminant proteins); this pattern was not dissimilar to the doublet found on SDS-PAGE of hnRNP-E1 isolated from human placenta (12) or Western blots of rat brain probed with anti-hnRNP-E1 and -E2 antiserum (31). Because there was no change in the expression of hnRNP-E1 per mg of cytosolic (S-100 fraction) proteins from HeLa-IU1-HF and HeLa-IU1-LF cells on Western blots (Fig. 5B) and Northwestern blots (24), this allowed for direct comparison of the amount of l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) between these cells. Moreover, because all cytosolic S-100 fractions were routinely sequentially pre-exposed to IgGsorb (a 10% suspension of formalin fixed Staphylococcus aureus cells bearing protein A) to remove nonspecific bound proteins that interacted with this matrix and then to non-immune serum followed by IgGsorb, the fraction of l-[35S]homocysteine that was specifically bound to hnRNP-E1 (and hnRNP-E2) was the difference in values obtained when aliquots of the samples were finally incubated with either anti-hnRNP-E1 antiserum followed by IgGsorb or non-immune serum followed by IgGsorb. Briefly, cytosolic S-100 fraction (25 μg) from ~80% confluent HeLa-IU1-HF or HeLa-IU1-LF cells was incubated with 100 μm l-[35S]homocysteine in 100 μl of binding buffer. These S-100 fractions are identified as “in vitro” (Fig. 5A) to distinguish them from other “in vivo” studies that involved the addition of 100 μm l-[35S]homocysteine in 4 ml of medium to 3 × 106 intact HeLa-IU1-HF or HeLa-IU1-LF cells cultured to ~80% confluence. Following incubation for 2 h at 37 °C and extensive rinsing with Dulbecco's phosphate-buffered saline (D-PBS), cells were harvested into 10 ml of D-PBS containing 20 mm EDTA and protease inhibitors, and cell pellets were further washed in 5 volumes of D-PBS before preparation of cytosolic S-100 fractions (12). These in vitro and in vivo S-100 fractions were first exposed to 200 μl of IgGsorb that was sedimented by centrifugation (13,600 × g for 5 min). Then 100 μl of non-immune rabbit serum was added to the supernatant, and after the addition of 200 μl of IgGsorb and centrifugation, the resulting supernatants from the in vitro and in vivo samples were divided equally into six different tubes from each sample. Three aliquots each were incubated with either (undiluted) anti-hnRNP-E1 antiserum (10 μl) or non-immune rabbit serum (10 μl) for 16 h at 4 °C, followed by the addition of 200 μl of IgGsorb, incubation for 2 h, and centrifugation at 13,600 × g for 5 min at 4 °C. After the pellets (20 μl each) were washed four times with 1 ml of D-PBS containing 20 mm EDTA and protease inhibitors and resuspended in 0.48 ml of D-PBS, an aliquot (200 μl) was mixed with 10 ml of counting mixture and analyzed for radioactivity in a β-scintillation counter. The results for l-[35S]homocysteine that was specifically bound to hnRNP-E1 (and hnRNP-E2) were calculated based on the amount of 35S incorporated into each sample by subtracting nonspecific binding using non-immune serum from results of total binding using anti-hnRNP-E1 antiserum and conversion of data to nmol of l-homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of protein (based on protein concentration of the S-100 fraction).
Because some fraction of l-[35S]homocysteine could have been converted to l-[35S]methionine in HeLa-IU1-HF cells by methionine synthase (EC 22.214.171.124), we determined the amount of l-[35S]methionine that could have become incorporated into newly synthesized hnRNP-E1 (and hnRNP-E2) during a 2-h period of incubation with l-[35S]homocysteine. Therefore, HeLa-IU1-HF cells in MEM-HF medium (which contains 96.7 μm l-methionine) were incubated with 100 μCi of l-[35S]methionine (specific activity, 1175 Ci/mmol) for 2 h and processed similarly to l-[35S]homocysteine-incubated cells, as described above. The final amount of l-[35S]methionine incorporated into newly synthesized hnRNP-E1 was corrected for the amount of unlabeled l-methionine present in the medium (i.e. the final specific activity of l-[35S]methionine was 259 mCi/mmol).
One μg of dialyzed, purified recombinant GST-hnRNP-E1 and 50 μm l-homocysteine were mixed in 10 mm Hepes, pH 7.6, containing 3 mm MgCl2, 40 mm KCl, 5% glycerol in a final volume of 15 μl. After incubation at room temperature for 30 min, the mixture was assayed by SDS-PAGE without the addition of dithiothreitol (DTT) in the loading buffer followed by Coomassie Blue staining. The Coomassie Blue-stained gel pieces were then cut out, incubated in 100 μl of 50% 0.1 m ammonium bicarbonate and 50% acetonitrile (by volume) at 37 °C for 30 min, and dehydrated with 100 μl of pure acetonitrile for 5 min. After excess acetonitrile was decanted, the gel pieces were dried under vacuum for 15 min and then rehydrated with a trypsin solution (0.5 μg in 100 mm ammonium bicarbonate) and incubated overnight at 37 °C. Following digestion, peptides were extracted with 0.1% trifluoroacetic acid for 30 min at 37 °C. Tryptic digests (10 μl) were injected onto a 75-μm (internal diameter) × 5-cm C-18 reverse-phase column (Waters, Picofrit column), and peptides were eluted with a gradient from 5 to 45% acetonitrile developed over 30 min at a flow rate of 250 nl/min using an Agilent 1100 series nanopump. The column was interfaced with a LTQ ion trap mass spectrometer (Thermo Electron Corp., Waltham, MA), and data were collected in the triple play mode. MS/MS spectra were searched against the IPI human protein data base with SEQUEST to confirm protein identity. Data analysis for disulfide bond detection was carried out manually.
See supplemental Methods for details on preparation of unlabeled and radiolabeled l-homocysteine; isolation of recombinant GST-hnRNP-E1 and GST-hnRNP A1; the homocysteine stability assay; preparation of radiolabeled 18-base FR-α mRNA cis-element, 19-base 15-lipoxygenase mRNA cis-element, WA1 RNA cis-element, HPV-16 L2 minor viral capsid protein mRNA, 27-base tyrosine hydroxylase mRNA cis-element, and 19-base neurofilament-M mRNA cis-element; a comparison of the effects of dl- and l-homocysteine on various chloramphenicol acetyltransferase (CAT) reporter constructs transfected into HeLa-IU1-HF cells; analysis of the rate of degradation of hnRNP-E1 in HeLa-IU1-HF cells; animals and animal care; and immunohistochemical analysis of paraffin-embedded sections of the fetal adrenal medulla and cerebellum.
Because all previous studies employed crude S-100 preparations of hnRNP-E1 as well as dl-homocysteine in concentrations that were higher than values found in mild folate deficiency (24) and because dl-homocysteine contains an unknown fraction of physiologically relevant l-homocysteine, we focused on a likely direct role of physiological concentrations of l-homocysteine on purified recombinant GST-hnRNP-E1 fusion protein (supplemental Fig. S2) in triggering RNA-protein interactions. Gel shift assays involving interaction of 18-base FR-α mRNA cis-element with purified recombinant GST-hnRNP-E1 demonstrated a dose-dependent increase in generation of RNA-protein complexes with increasing concentrations of dl- and l-homocysteine (0–100 μm) that were within the physiologic range of mild, moderate, or severe folate deficiency (Fig. 1A). Improved sensitivity of detection of the effects of lower concentrations of l-homocysteine on the RNA-protein interaction was achieved by extensive dialysis of DTT that was used in purification of recombinant GST-hnRNP-E1 protein. Nevertheless, prolonged exposure of the gels could still detect faint RNA-protein signals in the absence of l-homocysteine. Direct comparison of the densitometry signals of the RNA-protein complexes at each homocysteine concentration revealed a consistently greater signal using l-homocysteine (between 1.5- and 2-fold greater) when compared with dl-homocysteine within the range studied.
Because purified preparations of FR-α mRNA cis-element and hnRNP-E1 were employed for these RNA-protein interactions and because homocysteine can covalently bind to a variety of proteins by forming protein-cysteine-S-S-homocysteine mixed disulfide bonds (32–35) and alter the function of some proteins (33, 34, 36, 37), we determined if homocysteine altered the affinity of purified recombinant GST-hnRNP-E1 for interaction with an 18-base FR-α mRNA cis-element. As shown in Fig. 1B, the increase in RNA-protein signal in the presence of l-homocysteine and dl-homocysteine was also associated with a dose-dependent increase in binding affinity, so from a base line dissociation constant value (KD = 4.3 nm) there was a progressively lower KD value at each increasing concentration of l-homocysteine when compared with dl-homocysteine. Thus, l-homocysteine enhanced the RNA-protein binding affinity 7-fold and at a lower and physiologically relevant concentration through direct effects on hnRNP-E1.
Earlier, we demonstrated the specificity of hnRNP-E1 for translation of FR by showing that the addition of anti-hnRNP-E1 antiserum during in vitro translation inhibited biosynthesis of [35S]FR (12). Because rabbit reticulocyte lysate is rich in hnRNP-E1 (11, 12, 38), we compared the relative effectiveness of dl- and l-homocysteine in stimulating the in vitro translation of FR protein that was biosynthetically radiolabeled with [35S]cysteine. After quenching excess (4.1 mm) β-mercaptoethanol extant in the commercial translation kit with 2 mm N-ethylmaleimide (Fig. 1C, lane 2), the densitometric analysis of the autoradiogram of these [35S]FR protein signals revealed a dose-dependent enhancement of FR protein synthesis with both forms, but l-homocysteine had a consistently greater impact than dl-homocysteine. (See supplemental material on the stability of dl- and l-homocysteine and analysis of the likely basis for requiring higher concentrations of dl- than l-homocysteine to demonstrate RNA-protein interactions).
As shown in supplemental Fig. S1A, although purified recombinant GST-hnRNP-E1 bound to the radiolabeled 18-base FR-α mRNA cis-element in the presence of 100 μm DTT as well as in a dose-dependent manner in the presence of l-homocysteine (lanes 2–5), purified recombinant GST-hnRNP A1 (a control mRNA-binding protein) did not react with this FR-α mRNA cis-element probe (lanes 6–8). Conversely, the radiolabeled WA1 RNA cis-element did interact with purified recombinant GST-hnRNP A1 in the presence of either 100 μm l-homocysteine or DTT (lanes 12 and 13), but failed to react with purified recombinant GST-hnRNP-E1 (lanes 10 and 11). Whereas these studies confirmed the specificity of interaction of FR-α mRNA cis-element and homocysteinylated hnRNP-E1, they also eliminated the possibility of interaction of the FR-α mRNA cis-element with another control mRNA-binding protein (hnRNP A1) in the presence of l-homocysteine. As shown in supplemental Fig. S1B, the addition of anti-hnRNP-E1 antiserum to a translation mixture in vitro progressively quenched the translation of FR-α in a dose-dependent manner (lanes 2–4). The addition of non-immune serum had no effect (not shown). Moreover, this inhibition of FR-α biosynthesis by anti-hnRNP-E1 antiserum was reversed by the addition of purified recombinant GST-hnRNP-E1 in a dose-dependent manner (lanes 7 and 8). However, the addition of purified recombinant hnRNP A1 had no such effect in reversing the inhibitory effect of anti-hnRNP-E1 antiserum (lanes 5 and 6). These experiments indicated that there was no functional influence of hnRNP-A1 on the translation of FR-α, which only involved the interaction of the 18-base FR-α RNA cis-element with hnRNP-E1 present in the reticulocyte lysate.
Homocysteine enters cells by an active cysteine transporter system (39). Therefore, we determined if the acute entry of physiologically relevant concentrations of homocysteine could trigger biological effects using an FR-α mRNA cis-element-driven CAT reporter and via direct effects on the translation of FR in cells (24). (Supplemental Fig. S3 depicts a schematic representation of a CAT reporter construct driven by cDNA of the 18-base FR-α mRNA cis-element, pFR18-CAT.) As shown in Fig. 2A, transfection of cells by a pCAT (control) vector and exposure of HeLa-IU1-HF cells to increasing concentrations of l- versus dl-homocysteine led to comparable CAT activity across the entire range studied. By contrast, when pFR18-CAT DNA-transfected HeLa-IU1-HF cells were exposed to increasing concentrations of l- or dl-homocysteine for 2 h at 37 °C, there was a progressively greater CAT activity in the presence of l-homocysteine when compared with dl-homocysteine (Fig. 2B). Moreover, as shown in Fig. 2, C and D, mRNA levels of pCAT or pFR18-CAT were not significantly changed after the addition of 0, 20, and 50 μm l-homocysteine. These data indicated that the effect of l-homocysteine on CAT protein expression (Fig. 2B) was not caused by any changes in CAT mRNA levels. Because we had identified a direct effect of l-homocysteine in increasing the affinity of purified recombinant GST-hnRNP-E1 for FR-α mRNA cis-element (Fig. 1B), these CAT data also supported the possibility that l-homocysteine directly modified hnRNP-E1 within cells and that this, in turn, led to an increased interaction with pFR18-CAT mRNA that resulted in increased CAT protein expression.
The results from Figs. 1 and and22 suggested that increasing concentrations of l-homocysteine modified hnRNP-E1 into a moiety with increased affinity for FR-α mRNA cis-element. Therefore, it was necessary to demonstrate the dose-dependent formation of such RNA-protein complexes within cells in direct response to increasing concentrations of l-homocysteine that mimicked various grades of physiological folate deficiency. However, there were several additional mRNA cis-elements with common poly(rC)- or poly(U)-rich sequences that could potentially bind to hnRNP-E1 (20, 24, 31, 40–48) in cells containing higher than basal concentrations of l-homocysteine. Therefore, it was imperative to capture only FR-α mRNA cis-element bound to hnRNP-E1 (as RNA-protein complexes) within cells in direct response to increasing intracellular l-homocysteine concentrations. This could be shown if the RNA-protein complexes that formed in response to increasing concentrations of l-homocysteine (which were subsequently UV-cross-linked and immunoprecipitated by anti-hnRNP-E1 antiserum-bound S. aureus protein A) contained a progressive dose-dependent increase of FR-α mRNA cis-element when evaluated using slot blot hybridization analysis. However, this large (RNA-protein-antibody-protein A) complex was in a suspension, which precluded accurate quantification of immunoprecipitated hnRNP-E1 protein. Therefore, an alternative approach was to first release the small mRNA cis-element fragments (that were originally UV-cross-linked to hnRNP-E1 as a function of l-homocysteine dose) by proteolysis of the immunoprecipitate. This step could then be followed by the addition of equal aliquots of soluble RNA fragments per sample and finally probing for evidence of enrichment of FR-α mRNA cis-element using a specific 35S-labeled antisense FR-α mRNA cis-element under high stringency conditions (to exclude interaction with other cellular mRNA that may have bound hnRNP-E1).
As shown in Fig. 3, lanes 2–4, there was an l-homocysteine-induced dose-dependent increase in signal when compared with control samples to which l-homocysteine was not added (lane 1). In addition, there was no signal in the sample treated with non-immune serum, confirming the specificity of immunoprecipitation of hnRNP-E1 by antiserum (data not shown). Furthermore, there was no change in the expression of hnRNP-E1 in cells acutely exposed to l-homocysteine over 2 h (data not shown). In concordance with these data, there was also a greater amount of FR-α mRNA cis-element that was in complex with hnRNP-E1 within folate-deficient cells, which contain 2.7-fold more homocysteine than folate-replete cells (24); this was demonstrated by the greater hybridization signal with radiolabeled antisense FR-α mRNA cis-element probe when tested against FR-α mRNA cis-element that had complexed with hnRNP-E1 isolated from HeLa-IU1-LF cells compared with HeLa-IU1-HF cells (Fig. 3A, lane 6 versus lane 5, respectively). The specificity of hybridization with the radiolabeled antisense FR-α mRNA cis-element probe was shown by the lack of hybridization with various plasmid DNAs containing cDNAs that encode mRNAs known to interact with hnRNP-E1. Thus, there was only a signal using the plasmid DNA containing FR-α cDNA (lane 7) and no signals with the plasmid DNAs containing 15-lipoxygenase cDNA (12) (lane 8) or human papillomavirus type-16 (HPV-16) L2 cDNA (44) (lane 9). Yet another corollary control to test the level of stringency of hybridization and the veracity of data in Fig. 3A was demonstrated by the data in Fig. 3B, where 35S-labeled antisense probe to 15-lipoxygenase mRNA cis-element failed to hybridize with the plasmid DNA containing FR-α cDNA (Fig. 3B, lane 1) or HPV-16 L2 cDNA (Fig. 3B, lane 3); however, there was a signal with the plasmid DNA containing 15-lipoxygenase cDNA (Fig. 3B, lane 2). Collectively, these results of increased specific hybridization signals with radiolabeled antisense FR-α mRNA cis-element probe reflected the capture of FR-α mRNA cis-element-bound hnRNP-E1 protein complexes within cells as a function of increasing intracellular l-homocysteine concentrations. Fig. 3C demonstrates that when HeLa-IU1-HF cells that were stably propagated and maintained in high folate medium were acutely exposed to 50 μm l-homocysteine, there was an increase in the rate of FR biosynthesis from basal values of 39.0 ± 5.6 to 70.3 ± 5.8 fmol of l-[35S]cysteine incorporated into FR/mg of protein/h, respectively (p = 0.003). These results, which were concordant with the influence of l-homocysteine on RNA-protein binding, dissociation constant, and in vitro translation data (Fig. 1), CAT reporter data (Fig. 2B), and intracellular RNA-protein complex formation (Fig. 3A), all suggested that the putative homocysteinylation of hnRNP-E1 within cells allowed for interaction with the FR-α mRNA cis-element to trigger the biosynthesis of FR. Earlier, we determined that although cysteine can stimulate the interaction of FR-α mRNA cis-element and hnRNP-E1, the concentration of cysteine did not significantly increase in folate-deficient cells, and homocysteine was the only intracellular thiol that progressively increased during the development of folate deficiency (24). Accordingly, we directly compared the affinity of radiolabeled FR-α mRNA cis-element binding by purified recombinant GST-hnRNP-E1 in the presence of equimolar concentrations (25 μm) of either l-cysteine or l-homocysteine (Table 1); the results revealed KD values of 12.42 ± 0.26 and 8.11 ± 0.34 nm, respectively (p < 0.001). (The differences in absolute KD values from Fig. 1 probably arise from use of a different batch of components (RNA, protein, and l-homocysteine) in these later experiments.) Moreover, despite maintenance of a concentration of 25 μm l-cysteine in the FR-α mRNA cis-element and GST-hnRNP-E1 reaction mixture, the addition of either 25 or 100 μm l-homocysteine (Table 1) led to a progressive lowering of the KD value from 7.52 ± 0.29 to 3.21 ± 0.23 nm, respectively (p < 0.001). Thus, a progressive increase in l-homocysteine concentration (under conditions where the concentration of l-cysteine was unchanged) led to a further increase in affinity between FR-α mRNA cis-element and GST-hnRNP-E1 and was consistent with the possibility of a further unmasking of the RNA-binding site in hnRNP-E1. Taken together, these findings of a greater affinity of the RNA-protein complex under experimental conditions that simulate the intracellular milieu of folate-deficient cells (24) can therefore explain the more prominent signal on slot blot hybridization, signifying more RNA-protein complexes in folate-deficient cells compared with folate-replete cells (Fig. 3A, lane 6 versus lane 5, respectively).
Experiments were then designed to define the effect of depletion of hnRNP-E1 mRNA by RNAi on the biosynthetic rate of FR. Using an earlier protocol (29), which involved a high ratio of siRNA to cell number, led to significant cell death of HeLa-IU1-HF cells that began within 2 days and eventually involved nearly all cells by the fifth day. When a smaller ratio of siRNA to cell number was employed, the hnRNP-E1 mRNA was depleted by 66, 72, 63, 64, and 67% on days 1, 2, 3, 4, and 5, respectively (Fig. 4A).
Despite the marked homology between hnRNP-E1 and hnRNP-E2 (83% amino acid identity), it was unlikely that hnRNP-E2 mRNA was reduced following RNAi of hnRNP-E1 mRNA in HeLa-IU1-HF cells because there were only 12 of 25 homologous RNA sequences in the siRNA used. Indeed, when specifically measured, hnRNP-E2 mRNA was not reduced following RNAi of hnRNP-E1 mRNA (supplemental Fig. S4), predicting that hnRNP-E2 protein level was unaffected in these cells. The experimentally determined rate of degradation of [35S]cysteine incorporated into hnRNP-E1 (supplemental Methods) was 0.6 fmol/mg of protein/h, and the calculated half-life (t½) of the hnRNP-E1 protein was 52 h. This predicted that one-half of preformed hnRNP-E1 protein would be present in cells at 48 h after exposure to siRNA against hnRNP-E1 mRNA. Accordingly, we determined if the biosynthesis of newly synthesized hnRNP-E1 protein was adversely affected within 2 days after HeLa-IU1-HF cells were transfected with either a non-targeted scrambled sequence or siRNA directed against hnRNP-E1 mRNA. As shown in Fig. 4B, when compared with basal control values using scrambled RNA, there was a highly significant (p = 0.0002) reduction in the rate of biosynthesis of hnRNP-E1 after 72% reduction of hnRNP-E1 mRNA by siRNA at 48 h. Specifically, there was a reduction in the rate of newly synthesized hnRNP-E1 from 51.5 ± 2.9 to 27.2 ± 1.9 fmol of l-[35S]cysteine incorporated into hnRNP-E1/mg of protein/h in scrambled and siRNA-treated cells, respectively. In addition, Fig. 4C confirmed that depletion of 72% of hnRNP-E1 mRNA by siRNA, leading to a reduction in hnRNP-E1 protein, also led to a significant reduction of the basal FR biosynthetic rate from 49.4 ± 3.2 to 21.2 ± 3.0 fmol of l-[35S]cysteine incorporated into FR/mg of protein/h (p = 0.0004). These results involving the targeted reduction of hnRNP-E1 mRNA unambiguously confirmed our earlier studies (12, 24) which suggested that hnRNP-E1 was directly involved in the biosynthesis of FR within cells. Moreover, because there was a reduction of the basal rate of FR biosynthesis even in folate-replete HeLa-IU1-HF cells (Fig. 4C), these data suggest that the constitutive expression of FR in these cells is also mediated by hnRNP-E1 interaction with FR-α mRNA cis-element. The results of a second set of experiments involved a comparison of the biosynthesis of FR after the transfection of HeLa-IU1-HF cells with either scrambled RNA or siRNA directed against hnRNP-E1 mRNA for 2 days followed by exposure of cells to 50 μm l-homocysteine for 2 h. As shown in Fig. 4D, the l-homocysteine-stimulated values of FR biosynthesis in cells transfected with either scrambled RNA or siRNA directed against hnRNP-E1 mRNA were 75.0 ± 3.6 and 47.3 ± 4.6 fmol of l-[35S]cysteine incorporated into FR/mg of protein/h, respectively (p = 0.0012). These results also confirmed that targeted reduction of hnRNP-E1 mRNA by siRNA, which led to a reduction in hnRNP-E1 protein, also prevented l-homocysteine from exerting a maximal effect on FR biosynthesis; this was consistent with the critical role of hnRNP-E1 plus l-homocysteine in mediating the biosynthesis of FR under conditions of up-regulation of FR during folate deficiency. (See supplemental Discussion on the likelihood that measurement of the extent of reduction of the rate of newly synthesized hnRNP-E1 protein following the siRNA-induced reduction of hnRNP-E1 mRNA was underestimated.) Taken together, the data in Fig. 4 confirmed that hnRNP-E1 was essential for both the constitutive biosynthesis of FR in HeLa-IU1-HF cells and for the homocysteine-induced up-regulation of FR during folate deficiency.
The finding that l-homocysteine enhanced the FR-α mRNA cis-element interaction with purified recombinant GST-hnRNP-E1 (Fig. 1B) raised the possibility that endogenous cellular homocysteine (24) could also bind hnRNP-E1 and transform this (homocysteinylated hnRNP-E1) moiety into a high affinity specific mRNA-binding protein; indeed, this possibility was also supported by our CAT data (Fig. 2) and RNAi studies (Fig. 4). If this was the case, then hnRNP-E1 from folate-deficient HeLa-IU1-LF cells, which contain a higher intracellular concentration of endogenous homocysteine, should have less capacity to bind to exogenously added radiolabeled homocysteine when compared with folate-replete HeLa-IU1-HF cells, which have lower intracellular concentrations of endogenous homocysteine. Accordingly, we compared the direct binding of l-[35S]homocysteine to hnRNP-E1 that was isolated from HeLa-IU1-HF and HeLa-IU1-LF cells (24). Although these two cell lines have a similar concentration of hnRNP-E1 (and hnRNP-E2) per mg of cytosolic (S-100 fraction) protein (Fig. 5B), HeLa-IU1-HF cells contain 47 μm total extractable homocysteine, whereas HeLa-IU1-LF cells contain 126 μm total extractable homocysteine (24). Therefore, when we immunoprecipitated hnRNP-E1 (and hnRNP-E2)-bound l-[35S]homocysteine from both of these cell lines by specific anti-hnRNP-E1 antiserum (12, 24), because the amount of immunoprecipitated hnRNP-E1 (and hnRNP-E2) proteins would be similar, this allowed us to directly compare the amount of l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of total S-100 cytosolic fraction between these cell lines. We first determined that by directly exposing hnRNP-E1 (and hnRNP-E2) from equivalent amounts of cytosolic S-100 fractions from HeLa-IU1-HF or HeLa-IU1-LF cells to l-[35S]homocysteine, the nmol of l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of protein was 72.6 and 39.0, respectively; this is marked “in vitro” in Fig. 5A. Then, using a physiologic in vivo approach, we detected more l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) within intact HeLa-IU1-HF than HeLa-IU1-LF cells (38.7 versus 4.7 nmol of l-homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of protein, respectively); these data are denoted “in vivo” in Fig. 5A. The data in Fig. 5B confirmed no changes in hnRNP-E1 (and hnRNP-E2) levels between HeLa-IU1-HF and HeLa-IU1-LF cells. Thus, under both conditions, binding of l-[35S]homocysteine to cellular hnRNP-E1 (and hnRNP-E2) appeared to be inversely proportional to prevailing endogenous cellular homocysteine concentrations, suggesting competition for available homocysteine binding sites within hnRNP-E1 (and hnRNP-E2).
Because in vitro incubation of hnRNP-E1 (and hnRNP-E2) from the S-100 fraction of HeLa-IU1-HF or HeLa-IU1-LF cells with l-[35S]homocysteine for 2 h was under conditions where no new protein synthesis took place, this provided an accurate estimate of the ~2-fold greater amount of l-[35S]homocysteine bound by hnRNP-E1 (and hnRNP-E2) derived from HeLa-IU1-HF cells when compared with HeLa-IU1-LF cells. By contrast, the in vivo studies, which identified an 8-fold greater difference in binding of l-[35S]homocysteine by hnRNP-E1 (and hnRNP-E2) from HeLa-IU1-HF cells, could be explained (in part) if a large fraction of l-[35S]homocysteine was converted by cellular methionine synthase to [35S]methionine and subsequently incorporated into newly synthesized hnRNP-E1 (and hnRNP-E2) protein. However, formal control studies that evaluated the extent of [35S]methionine incorporated into hnRNP-E1 (and hnRNP-E2) under these conditions determined that only 1.02 nmol of [35S]methionine could have become incorporated into newly synthesized hnRNP-E1 (and hnRNP-E2) protein within 2 h of incubation with cells. Therefore, correcting for this value, there was 37.7 (38.7 minus 1.02) nmol of l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of protein from HeLa-IU1-HF cells. This value is still 8-fold greater than the 4.7 nmol of l-[35S]homocysteine bound to hnRNP-E1 (and hnRNP-E2) per mg of protein from HeLa-IU1-LF cells. Therefore, these in vitro and in vivo studies were internally consistent with one another and strongly suggested that endogenous homocysteine competed with l-[35S]homocysteine for binding cellular hnRNP-E1 (and hnRNP-E2).
All preceding studies suggested that the posttranslational homocysteinylation of hnRNP-E1 resulted in the unmasking and optimization of its FR-α mRNA cis-element-binding site. This could conceivably occur by replacement of critical cysteine-S-S-cysteine disulfide bonds (that normally preclude high affinity RNA binding to hnRNP-E1 in the folate-replete state) by protein-cysteine-S-S-homocysteine mixed disulfide bonds (that keep open this RNA-binding pocket in hnRNP-E1 for optimum mRNA binding in the folate-deficient state). Accordingly, we searched for such protein-cysteine-S-S-homocysteine mixed disulfide bond signatures after incubating hnRNP-E1 with l-homocysteine. When purified recombinant GST-hnRNP-E1 was incubated with 50 μm l-homocysteine and the mixture was digested with trypsin (pH 7.0), subsequent analysis by nano-LC-MS/MS identified the presence of cysteine-S-S-homocysteine mixed disulfide bonds at Cys-54, Cys-109, Cys-158, Cys-163, and Cys-194 of hnRNP-E1 (Fig. 6A). As one example of the identification of a cysteine-S-S-homocysteine mixed disulfide bond at Cys-158, trypsinization of hnRNP-E1 alone yielded a peptide fragment from the KH2 domain with the sequence AITIAGVPQSVTECVK, which was captured by nano-LC-MS/MS analysis (Fig. 6B); the sequence is noted in the left-hand column labeled Seq in charts A and B. However, upon incubation of hnRNP-E1 with l-homocysteine, this particular peptide fragment was detected by nano-LC-MS/MS analysis with an increased m/z (Δm/z = 134), indicating that it had covalently bound another moiety that had a molecular weight of 135. These data are highlighted in charts A and B by arrows where all three types of fragment ions (a-14, a-15, and a-16; b-14, b-15, and b-16; and y-3 to y-16) showed consistent results (third from the bottom). The Δm/z of 134 is identical to the modification of homocysteine on Cys-158 and therefore indicates the presence of a cysteine-S-S-homocysteine mixed disulfide bond at Cys-158 of hnRNP-E1.
Because l-homocysteine progressively increases binding affinity of hnRNP-E1 for FR-α mRNA cis-element, the data suggest that covalent binding of homocysteine to several additional cysteine residues within hnRNP-E1 contributes to fully unmasking the high affinity RNA-binding site. Collectively, these data support a direct effect of homocysteine on the hnRNP-E1 protein.
Because hnRNP-E1 binds several mRNAs in the presence of conventional reducing agents (24, 38, 40–44, 49, 50), we determined if some of these mRNAs could also bind to the RNA-binding site in hnRNP-E1 that was apparently unmasked by homocysteine. For example, a domain in the 3′-coding region of the L2 mRNA of HPV-16 minor viral capsid protein (44) has significant homology to the 18-base cis-element in the 5′-UTR of FR-α mRNA (11), and both of these mRNAs have poly(rC)-rich domains that are bound by hnRNP-E1 (11, 44). As shown in Fig. 7, A and B, with increasing concentrations of dl- and l-homocysteine, there was a proportionate increase in RNA-protein complexes using rabbit 15-lipoxygenase mRNA and the HPV-16 L2 minor viral capsid protein mRNA, with a consistently stronger effect of l-homocysteine at every concentration used, similar to that noted for the FR-α mRNA cis-element. (The additional smaller RNA-protein signal in Fig. 7A probably represents a breakdown product of GST-hnRNP-E1). Two independent laboratories have recently shown that hnRNP-E1/PCBP1/αCP1 can bind to cis-elements in the 3′-untranslated region of tyrosine hydroxylase (TH) mRNA (51) and the neuronal intermediate neurofilament-middle molecular mass (neurofilament-M, NF-M) mRNA (31) to induce a posttranscriptional up-regulation of these proteins; of significance, these investigators, as most others, employed artificial thiols (such as β-mercaptoethanol and/or DTT) to trigger these RNA-protein interactions. Accordingly, we determined if similar interactions could be triggered using physiologically relevant concentrations of l-homocysteine found in murine folate deficiency (52) and whether the functional consequences of RNA-protein interactions were detected by altered protein expression in the organs of gestation day 17 murine fetuses whose mothers had experienced gestational folate deficiency, having been fed a 400-nmol folate/kg diet for 2 months prior to and during pregnancy, when compared with control fetuses whose mothers were consistently fed a folate-replete diet composed of 1200 nmol folate/kg throughout.
Fig. 7C documented a dose-dependent formation of RNA-protein complexes following the interaction of purified recombinant GST-hnRNP-E1 with either the 27-base TH mRNA cis-element or the 19-base NF-M mRNA cis-element in the presence of physiologically relevant concentrations of l-homocysteine found in murine dietary folate deficiency (52). Although the folate-replete murine fetal adrenal medulla (Fig. 7D, left) exhibited significant base-line TH staining, the folate-deficient murine fetal adrenal medulla (Fig. 7D, right) had further increased staining reflecting TH protein up-regulation. Similarly, although the folate-replete murine fetal cerebellum (Fig. 7E, left) exhibited significant base-line NF-M staining, the folate-deficient murine fetal cerebellum (Fig. 7E, right) had further increased staining reflecting up-regulation of NF-M protein expression.
These data confirmed that the (patho)physiologically relevant thiol, l-homocysteine, was as capable of triggering RNA-protein interactions as artificial thiols, strongly suggesting that l-homocysteine is a physiologically significant thiol in activating hnRNP-E1 in folate-deficient tissues in vivo. Furthermore, these studies supported the likelihood that TH and NF-M mRNAs were also prototypical members of this novel nutrition-sensitive, posttranscriptional RNA operon that is orchestrated in vivo by homocysteinylated hnRNP-E1.
This paper affirms the existence of a physiologically relevant mechanism for the sensing and subsequent restoration of cellular folate concentration during mild, moderate, or severe folate deficiency through a proportionate up-regulation of FR. This homeostatic process is initiated when the metabolite homocysteine, which accumulates intracellularly during folate deficiency (1, 24), covalently binds to and eventually transforms hnRNP-E1 into a moiety with even higher affinity for the FR-α mRNA cis-element (Fig. 8). The resulting formation of these intracellular RNA-protein complexes triggers a proportionate increase in the biosynthesis of FR and its up-regulation on cell surfaces that serves to maximize folate uptake from the folate-depleted external milieu.
Our experimental model involved the primary induction of folate deficiency that led to an increased intracellular rise in homocysteine concentration (24), which, in turn, triggered the homocysteinylation of hnRNP-E1 and an increase in biosynthesis of FR. This suggested that hnRNP-E1 probably functions as a sensitive sensor of cellular folate deficiency (Fig. 8). However, in a similar fashion, a rise in intracellular homocysteine from other causes (1, 53), including nutritional deficiencies of vitamin B12 (cobalamin) and vitamin B6 (pyridoxine), which together with folate deficiency comprise the major clinically relevant nutritional deficiencies involving one-carbon metabolism worldwide (1), as well as other rare congenital alterations in one-carbon metabolism (1, 53), could also trigger homocysteinylation of hnRNP-E1 and result in up-regulation of FR, but in these conditions, the effect on FR would be a secondary (collateral) effect.
Although we had earlier determined that the major metabolite that increased in cells in folate deficiency was homocysteine (24), it was important to investigate the relative effectiveness of homocysteine in triggering RNA-protein complex formation independently of and in the presence of cysteine. Increasing concentrations of l-homocysteine induced a progressive increase in affinity of the derivatized hnRNP-E1 for the FR-α mRNA cis-element (Fig. 1B). Moreover, physiologically relevant equimolar concentrations of l-homocysteine were more effective in reducing the KD of the RNA-protein interaction when compared with l-cysteine, and the addition of increasing concentrations of l-homocysteine in the presence of fixed concentrations of l-cysteine led to an even greater affinity (Table 1). These data are consistent with the work by Jacobsen et al. (32), who have argued in favor of the greater stability of a protein-cysteine-S-S-homocysteine mixed disulfide bond compared with a protein-cysteine-S-S-cysteine disulfide bond based on higher sulfhydryl pKa values; because the pKa of the sulfhydryl of homocysteine is 10.0, whereas that of cysteine is 8.3, this results in greater disulfide bond stability for homocysteine in comparison with cysteine (32). (In this context, see supplemental Discussion, “Limitations in measurement of protein-bound versus free homocysteine within cells”).
The finding that the binding of l-[35S]homocysteine to cellular hnRNP-E1 was inversely proportional to the prevailing endogenous homocysteine concentration (Fig. 5) suggested the existence of finite homocysteine-binding sites in hnRNP-E1. Indeed, the identification of peptide-cysteine-S-S-homocysteine mixed disulfide bonds at Cys-54, Cys-109, Cys-158, Cys-163, and Cys-194 of hnRNP-E1 following incubation of homocysteine with hnRNP-E1 (Fig. 6) suggested that homocysteine disrupted more than one cysteine-S-S-cysteine disulfide bond; the concomitant dose-dependent increase in affinity for the FR-α mRNA cis-element by homocysteinylated-hnRNP-E1 suggests that the progressive homocysteinylation of native cysteine-S-S-cysteine disulfide bonds in hnRNP-E1 unmasked an underlying RNA-binding domain. Because hnRNP proteins generally mediate DNA/RNA binding by K-homology (KH) domains (54, 55), there are 6 cysteine residues in the three KH domains of hnRNP-E1 that are candidates for physiologic interaction with homocysteine and can potentially contribute to fully unmasking the RNA-binding pocket of hnRNP-E1 during folate deficiency. Previous studies have demonstrated that KH domains 1 and 3 in hnRNP-E1 bind to poly(rC) (55), whereas the cis-element of neurofilament-M mRNA is bound by KH1 and KH2 (31). Hence, it is possible that further studies will uncover the involvement of more cysteine residues in additional KH domains in hnRNP-E1 that participate in binding the 18-base FR-α mRNA cis-element. This would be consistent with the earlier suggestion that most, possibly all, KH domains have the ability to bind nucleic acids and that some may do so as single isolated entities, whereas others may work cooperatively with other KH domains (55). Indeed, the formation of more than one protein-cysteine-S-S-homocysteine mixed disulfide bond can explain the progressive reduction in KD as more l-homocysteine was added to the RNA-protein mixture (Fig. 1B). Collectively, these results suggest that under folate-replete conditions, the RNA-binding pocket of hnRNP-E1 is partially masked by critical cysteine-S-S-cysteine disulfide bonds. There is, however, a level of basal FR-α mRNA cis-element binding by endogenous hnRNP-E1 in HeLa-IU1-HF cells that was captured as RNA-protein complexes (Fig. 3A, lane 1), which accounts for the base-line biosynthesis rate of FR (Fig. 3C) and abundant cell surface FR expression (24, 25). Indeed, Fig. 4C suggests that it was this fraction that was perturbed by RNAi of hnRNP-E1 mRNA in HeLa-IU1-HF cells under basal conditions. However, during folate deficiency, the accumulated cellular homocysteine probably binds progressively to critical cysteine-S-S-cysteine disulfide residues by forming protein-cysteine-S-S-homocysteine mixed disulfide bonds, which keep the RNA-binding pocket in an open conformation (Fig. 8B). Accordingly, we have embarked on a systematic study to determine the sequence of disruption of these cysteine-S-S-cysteine disulfide bonds in hnRNP-E1 by l-homocysteine during the progressive development of folate deficiency using a combination of site-directed mutagenesis and dynamic studies in vivo.
Because recent studies have identified that hnRNP-E1/PCBP1 can function as a chaperone of iron to ferritin (29), it remains to be determined if there is an interaction between a potential iron-sulfur cluster involving cysteine residues and l-homocysteine. Moreover, the potential effect of phosphorylation of homocysteinylated hnRNP-E1 and the modulation of its affinity for target mRNAs is another important area for study (56).
hnRNP-E1 is a multifunctional mRNA-binding protein that has the potential to bind a large number of diverse mRNAs (11–24, 38, 40–44, 49, 50). Although most investigators have used artificial thiols (β-mercaptoethanol and DTT) to trigger RNA-protein interactions, we employed the more (patho)physiologically relevant and effective thiol, l-homocysteine, to activate hnRNP-E1 in this paper. Thus, it was possible that the unmasked mRNA-binding domain in the homocysteinylated hnRNP-E1 could bind the FR-α mRNA cis-element as well as many of these mRNA cis-elements during conditions of cellular folate deficiency as a collateral effect. Our data suggest that like the FR-α mRNA cis-element, these other mRNAs (Fig. 7) are also likely to be prototypical members of the same nutrition-sensitive posttranscriptional RNA operon that is modulated by homocysteinylated hnRNP-E1. However, the consequences for protein expression will probably differ depending on the individual mRNA. For example, whereas binding of FR-α mRNA cis-element by hnRNP-E1 stimulates biosynthesis of FR protein at the translational level (24), binding of HPV-16 L2 mRNA by hnRNP-E1 apparently leads to inhibition of the synthesis of the L2 protein (44), and binding of rabbit 15-lipoxygenase mRNA by hnRNP-E1 can prevent differentiation (49). Moreover, the interaction of murine TH and NF-M mRNA cis-elements with hnRNP-E1 led to increased protein expression in the murine fetal adrenal medulla and cerebellum, respectively, via established posttranscriptional regulatory mechanisms (45, 57). Therefore, although the possibility that hnRNP-E1 modulates a posttranscriptional RNA operon was raised recently (58, 59), our physiologically relevant data highlight the critically important variable of accumulation of cellular homocysteine that conditionally activates the principal orchestrator (hnRNP-E1) of this posttranscriptional RNA operon. Our studies also suggest that this posttranslational modification of hnRNP-E1 by homocysteine activates a novel, nutrition-sensitive (homocysteine-responsive), posttranscriptional RNA operon that is dynamically functional and sensitive to varying degrees of physiological folate deficiency. In this model, the RNA-binding domain of homocysteinylated hnRNP-E1 has the capacity to accommodate a plethora of mRNA that are quite unrelated except for common poly(rC)- or poly(U)-rich cis-elements (12, 17, 24, 40) that have varying degrees of affinity for interaction with this protein. (Indeed, the possibility of different degrees of binding affinity between hnRNP-E1 and various mRNAs (in Fig. 7, A–C) is hinted at by the varying strengths of RNA-protein signals at 50 μm l-homocysteine.) Any true differences in KD would serve to modulate the expression of their respective proteins at the posttranscriptional level at various levels of folate deficiency. In turn, this could contribute to changes in expression of several critical proteins involved in cellular nutrition, cell proliferation, differentiation, apoptosis, and dysmorphogenesis at different degrees of folate deficiency.
There are many parallels between human and murine FR and between hnRNP-E1 of both species. For example, the mouse homologue of FR-α is folate-binding protein-1 (Folbp1) (4). FR also functions in murine cells (4, 5, 60, 61), and mice have a similar regulatory unit involving the interaction between a 17-base cis-element in the 5′-UTR of FR mRNA and a trans-factor called mouse poly(rC)-binding protein (αCP1), which is homologous with human hnRNP-E1 and found in several murine tissues, including the brain (12, 52, 62, 63). Based on our data on the interaction of homocysteinylated hnRNP-E1 with mRNA cis-elements from TH (45) and NF-M (31, 64–67), it is plausible that homocysteinylation of hnRNP-E1 within murine tissues can also trigger the binding of several diverse mRNAs with common signature poly(rC)-rich or poly(U)-rich sequences (18, 20, 31, 38, 40, 43, 47, 48, 58, 68, 69) that are part of a nutrition-sensitive (homocysteine-responsive), posttranscriptional RNA operon in mice.
Because the KH domains between hnRNP-E1 and hnRNP-E2 are 93% identical (58) and there is 100% identity in the location of cysteine residues of all three KH domains between hnRNP-E1 and hnRNP-E2 (also called PCBP2/αCP2), it appears very likely that a similar formation of cysteine-S-S-homocysteine mixed disulfide bonds in hnRNP-E2 during folate deficiency could also modulate the interaction of this internal ribosome entry site-trans-acting factor with many of its mRNA targets (28, 38). In a broader context, the presence or absence of accessible cysteine residues within KH or RNA recognition domains among other hnRNP family members and RNA-binding proteins, respectively, would probably be the primary determinant as to whether there is covalent binding by homocysteine during folate deficiency. Moreover, because there are so many other mRNA-binding proteins that are closely related to hnRNP-E1 (38, 40), these too are potential targets for homocysteinylation. Our data (supplemental Fig. S1) also suggest that hnRNP A1 is one such candidate because purified recombinant GST-hnRNP A1 clearly bound to the WA1 RNA cis-element in the presence of either DTT or l-homocysteine. There was a distinct difference in the strength of the signal with each of these thiols although the concentrations of RNA, protein, and thiols were similar, suggesting different affinities. Therefore, we can reasonably predict that some among the entire spectrum of RNA cis-element sequences that bind to hnRNP A1 with a >100-fold range of affinities (70) may be modulated during folate deficiency. Parenthetically, however, it should be noted that, apart from hnRNP-E1, there is no evidence to indicate that homocysteinylation of any of these other hnRNPs would trigger an increase in FR synthesis leading to augmented cellular folate uptake in response to folate deficiency; indeed, this was the basis for considering hnRNP-E1 as a candidate sensor of physiological folate deficiency. See supplemental Discussion, “Additional evidence for the specificity of interaction between the FR-α mRNA cis-element and hnRNP-E1.” Nevertheless, it is yet possible that the function of additional proteins that directly contribute to the cellular uptake of folates by increasing other FR isoforms or other folate transporters (10) could also be influenced by homocysteinylation during folate deficiency. In this event, these too could be considered additional candidate sensors of folate deficiency.
Based on these considerations, the posttranslational homocysteinylation of additional RNA-binding proteins would further amplify the number of RNAs that are regulated through the activation of multiple nutrition-sensitive (homocysteine-responsive), posttranscriptional RNA operons; and taken together, these would comprise a novel, higher order, nutrition-sensitive (homocysteine-responsive), posttranscriptional RNA regulon (71) during clinical folate or vitamin B12 deficiency. Confirmation of such networks would further open the field for a deeper voyage of discovery that could help to explain the expression of diverse genes in the fetal brain during gestational folate deficiency, especially those genes (such as TH and NF-M) that can be predicted to lead to the imprinting of neural circuits in utero or to induce neurohormonal-neurotransmitter imbalance that predisposes to altered behavior in postnatal life (52, 72, 73). See supplemental Discussion, “Homocysteinylation of hnRNP-E1/PCBP1 as an alternative determinant in enabling IRES-dependent translation”.
Finally, our studies also raise the possibility that, in addition to other proposed mechanisms (74), subtle genetic polymorphisms within either the FR-α mRNA cis-element or the RNA-binding site of hnRNP-E1 (75) could interfere with physiologic sensing of reduced folate availability and subsequent triggering of up-regulation of FR in fetal tissues during maternal folate deficiency; such polymorphisms in the regulatory components that dictate FR expression would remain undetected using conventional searches for mutations in the coding region of FR (42, 76) and could yet be the basis for folate-responsive neural tube defects.
GST-hnRNP-A1 plasmid DNA was a kind gift from Professor Benoit Chabot (Université de Sherbrooke, Québec, Canada). The generosity and encouragement of Professor Donald W. Jacobsen (Lerner Research Institute, Cleveland Clinic) in providing l-[35S]homocysteine during the early phase of this work is gratefully acknowledged.
*This work was supported, in whole or in part, by National Institutes of Health Grant CA120843 and HD39295 (to A. C. A.). This work was also supported by a Veterans Affairs Merit Review Award (to A. C. A. and H. N. J.).
The on-line version of this article (available at http://www.jbc.org) contains supplemental Methods, Results, Discussion, and Figs. S1–S4.
2The abbreviations used are: