Human Ribosomal G-Quadruplexes Regulate Heme Bioavailability

The in vitro formation of stable G-quadruplexes (G4s) in human ribosomal RNA (rRNA) was recently reported. However, their formation in cells and their cellular roles have not been resolved. Here, by taking a chemical biology approach that integrates results from immunofluorescence, G4 ligands, heme affinity reagents, and a genetically encoded fluorescent heme sensor, we report that human ribosomes can form G4s in vivo that regulate heme bioavailability. Immunofluorescence experiments indicate that the vast majority of extra-nuclear G4s are associated with rRNA. Moreover, titrating human cells with a G4 ligand alters the ability of ribosomes to bind heme and disrupts cellular heme bioavailability as measured by a genetically encoded fluorescent heme sensor. Overall, these results suggest ribosomes are central hubs of heme metabolism.


INTRODUCTION
Cells tightly control heme concentration and bioavailability (1)(2)(3) because it is essential but potentially cytotoxic. Proteins that regulate heme concentration are relatively well understood; structures and mechanisms of all eight heme biosynthetic enzymes and the heme degrading heme oxygenases are known (1)(2)(3). However, regulation of heme bioavailability, including intracellular trafficking from sites of synthesis in the mitochondrial matrix or uptake at the plasma membrane, is poorly understood. Current paradigms for heme trafficking and mobilization involves heme transfer by unknown proteinaceous factors and largely ignore contributions from nucleic acids. Given that the first opportunity for protein hemylation occurs during or just after translation, ribosomal RNA (rRNA) or proteins (rProteins) may be critical for shepherding labile heme to newly synthesized proteins.
We hypothesized that intracellular heme bioavailability is regulated in part by rRNA quadruplexes (G4s). G4s are nucleic acid secondary structures that are composed of four guanine columns surrounding a central cavity that sequesters monovalent cations. Our hypothesis is based on the high affinity of heme for G4s (KD ~10 nM) (4-6), our work demonstrating that rRNA forms extensive G-tracts in vitro (7,8), the extreme stabilities of rRNA G4s in vitro (7,8) and the extraordinary abundance of rRNA in vivo (9). DNA G4s are proposed to help regulate replication (10), transcription (11), and genomic stability (12). In RNA, G4s are associated with untranslated regions of mRNA and have been proposed to regulate translation (13)(14)(15). However, the in vivo folding state and functional roles of G4s are under debate. Eukaryotic cells contain helicases that appear to unfold RNA G4s (16) although counter arguments have been put forth (17,18).
The density of G4 sequences on surfaces of the human ribosome, which is extremely abundant, is high, with 17 G4 sequences in the 28S rRNA and 3 in 18S rRNA ( Figure   1A). Previous to this report, it was not known if human ribosomes form G4s in vivo or what their functions might be.
Indeed, herein we present evidence that rRNA forms G4s in vivo that regulate cellular heme homeostasis. Results of immunofluorescence experiments with a G4 antibody, RNA pulldowns and competition experiments with G4 ligands provide strong support for in vivo formation of G4s by rRNA tentacles. We find that G4s on ribosomes bind heme in vitro ( Figure 1B) and that perturbation of G4s in vivo with G4 ligands affects in vivo heme interactions and heme bioavailability, as measured by heme affinity reagents and genetically encoded heme sensors. Taken together, the results here indicate that surface-exposed rRNA G4s interact with heme in cells and suggest that ribosomes are hubs for cellular heme metabolism.

RESULTS
Ribosomal RNA forms G4s in vivo. Confocal microscopy and G4-pulldowns were used to determine if human ribosomes form G4s in vivo. For confocal microscopy, we used the BG4 antibody, which selectively targets G4s (19,20) and has been broadly used for visualizing DNA G4s and non-ribosomal RNA G4s in cells. (20)(21)(22)(23) Our method of permeabilizing cells for antibody treatment does not permeabilize the nuclei (24). Therefore, DNA G4s were not anticipated or observed. To identify ribosome associated G4s, we determined the extent to which antibodies to rProtein L19 (eL19) and to G4s colocalize and how this is altered when cells are subjected to RNase or G4 ligand PhenDC3, which are expected to modulate G4-L19 colocalization. Prior to antibody addition, cells were crosslinked with paraformaldehyde to lock G4s in situ. This procedure is intended to prevent induction of G4s by the antibody and has been shown to reduce levels of detection of G4s (18). The extent of L19 and G4 antibody colocalization suggests that a fraction of ribosomes form G4s (Figure 2A,C) and that most G4s are associated with ribosomes. Specifically, we find that ~83% of BG4 pixels colocalize with L19, indicating that the vast majority of G4s in vivo are associated with ribosomes ( Figure 2C, green bar) and are therefore rRNA G4s. Conversely, only 5% of L19 pixels colocalize with BG4 ( Figure 2C, WT red bar), indicating that only a specialized fraction of ribosomes contains G4s. Similar results were obtained using an antibody against rProtein uL4 instead of L19 (not shown).
PhenDC3, which is known to induce and stabilize G4s, (25,26) appears to increase ribosomal G4 formation in vivo; treating cells with PhenDC3 increases L19-BG4 colocalization from 5 to ~24% ( Figure 2C). The increase in colocalization upon PhenDC3 treatment supports formation of G4s by ribosomes. By contrast, treating cells with RNase A abolishes the L19-BG4 colocalization signal ( Figure 2C). Together, these results indicate the colocalized BG4 signal is coming from a G4 forming RNA in close proximity to L19. mRNA in the cytosol, in the unlikely event that they form G4s at high frequency (16), may confound our ability to selectively detect rRNA G4s. The high density of ribosomes on the surface of the endoplasmic reticulum (ER) and the lower abundance of mRNA in this location as compared to the cytosol (27) motivated us to investigate if G4s colocalize with the ER. Toward this end, we determined the extent to which BG4 colocalizes with an antibody against an ER membrane protein (calnexin) ( Figure 2B).
Indeed, we find that ~45% of the BG4 signal colocalizes with the ER marker ( Figure 2D, green bar), indicating a significant presence of RNA G4s at the ER membrane. As with L19, the fraction of the ER signal that colocalizes with G4s (~2%) is completely abolished by RNase (undetectable) and enhanced by PhenDC3 (12%) ( Figure 2D). Altogether, the data are consistent with formation of RNA G4s by ER-bound ribosomes. Figure 2. Ribosomal G4s in HEK293 cells. Colocalization of (A) ribosomal protein L19 or (B) endoplasmic reticulum (red) with RNA G4s (green). Nuclei were stained with DAPI (blue). (C) Extent of colocalization is quantitated as the ratio of colocalized pixels over total L19 pixels (red bars) or as the ratio of colocalized pixels over total BG4 pixels (green bar). Same analysis was done for ER-BG4 colocalization (D). The statistical significance relative to WT is indicated by asterisks using an ordinary one-way ANOVA with Dunnett's post-hoc test. Each dot represents a biological replicate. (E) G4 ligand BioTASQ binds to 28S and 18S rRNAs. In the presence of BioTASQ and streptavidin beads, human rRNAs do not enter the agarose gel. (F) Schematic representation of the BioTASQ pulldown protocol. (G) RT-qPCR analysis of rRNAs pulled down by BioTASQ. The statistical significance relative to a fold enrichment value of 1 is indicated by asterisks using a one sample t and Wilcoxon test. Each dot represents a biological replicate. Data in (G) are represented as RNA enrichment under "BioTASQ + streptavidin beads" conditions relative to control streptavidin beads. * P < 0.05. n.s. = not significant.
In an orthogonal approach, we pulled down RNA with BioTASQ (18,28), which is a G4 ligand linked to biotin. BioTASQ captures G4s. We previously used BioTASQ to demonstrate that human rRNA forms G4s in vitro ( Figure 2E) (8). Here, we captured rRNA G4s from crosslinked HEK293 cells by methods summarized in Figure 2F.
BioTASQ captures 28S rRNA from cell lysates (Figure 2G), in agreement with our previous in vitro BioTASQ data and with observations of G4-L19 colocalization above.
BioTASQ also captures 18S rRNA although the signal is significantly weaker. Taken together, our immunofluorescence and BioTASQ pulldown experiments provide strong evidence that human ribosomes form G4s in vivo.
Human ribosomes bind hemin in vitro. It has been suggested that G4s might associate with heme in vivo (29). In vitro, heme binds with high affinity to G4s by endstacking (30-32) ( Figure 1B). We used UV-visible spectroscopy to assay the binding of hemin to human rRNA. rRNA oligomers GQES7-a ( Figure 3A  PhenDC3 was used to confirm binding of hemin to ribosomal G4s. PhenDC3, like hemin, end-stacks on G4s (29) and therefore competes with heme for binding to G4s.
With fixed GQES7-a and hemin, addition of PhenDC3 causes a decrease in the intensity of the hemin Soret peak ( Figure 3D) due to dissociation of heme. The same phenomenon is observed with assembled ribosomal particles (LSU: Figure 3E Human ribosomes bind heme in vivo. We developed an assay that exploits differential interactions with hemin-agarose, an agarose resin covalently linked to heme, to report in vivo heme binding to ribosomes and rRNA. The degree to which any biomolecule interacts with heme in cells is inversely correlated with the extent to which it interacts with hemin-agarose upon lysis due to competition between endogenous heme and hemin-agarose for the heme-binding site. Therefore, the effects of heme binding factors in vivo can be monitored by determining if their interaction with hemin-agarose changes upon depletion of intracellular heme. Accordingly, HEK293 cells were conditioned with and without succinylacetone (SA (33)), an inhibitor of heme biosynthesis. Lysates of these cells were incubated with heminagarose, and hemin-agarose interacting rRNA was quantified by RT-qPCR. Consistent with previous work (34), treatment with 0.5 mM SA for 24 hours caused a 7-fold decrease in total cellular heme in HEK293 cells (results not shown). The results reveal that rRNA binding to hemin-agarose relative to control agarose lacking heme increases by ~4-fold in cells depleted of heme ( Figure 4A). This result suggests that, under heme-depleted conditions, a greater fraction of rRNA heme binding sites are free and available to bind hemin agarose. In short, the data are consistent with a model in which ribosomal RNAs associate with endogenous heme. RT-qPCR analysis from untreated (WT) and SA-treated human cells. Statistical significance relative to WT is represented by asterisks using Student's t-test. Each dot represents a biological replicate. (B) RT-qPCR analysis from PhenDC3-treated HEK293 cells. Statistical significance relative to no treatment conditions is represented by asterisks using ordinary one-way ANOVA with Dunnett's post-hoc test. Each dot represents a technical replicate coming from individual biological replicates. The experiment was performed a total of 2 times with similar dose-dependent trends (Fig. S.3A). Data in (A) and (B) are represented as RNA enrichment in hemin agarose beads relative to control sepharose beads. (C) Single cell analysis of HS1-transfected HEK293 cells grown in heme deficient media containing succinylacetone (HD+SA), regular media containing 5-aminolevulinic acid (R +ALA), or regular media (regular) with the indicated concentrations of PhenDC3. Statistical significance relative to regular conditions is represented by asterisks using the Kruskal-Wallis ANOVA with Dunn's post-hoc test. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001; n.s. = not significant; (n ≅ 1500 cells). (D) Median HS1 sensor ratios obtained in (C) as a function to PhenDC3 concentration.

In vivo PhenDC3 increases binding of ribosomes to hemin agarose. To determine
if rRNA G4s bind heme in vivo, we treated HEK293 cells with the G4 ligand PhenDC3 (48 hrs at 37 °C). PhenDC3 and heme compete for binding to G4 rRNA in vitro (Figure 3).
Thus, if rRNA G4s bind heme in vivo, PhenDC3 is expected to displace any rRNA bound heme. After cell lysis, excess hemin agarose is expected to out compete rRNA bound PhenDC3. Indeed, RT-qPCR reveals that PhenDC3 treatment of HEK293 cells causes a dose-dependent increase in binding of the LSU to hemin agarose ( Figure 4B). A corresponding, but weaker signal is seen for the SSU, in agreement with the higher abundance of G4 regions in the LSU than in the SSU ( Figure 1A). rRNA G4s regulate heme bioavailability in vivo. To determine if rRNA G4s regulate heme homeostasis, we deployed a previously described genetically encoded ratiometric fluorescent heme sensor, HS1. HS1 is a tri-domain fusion protein consisting of a heme binding domain, cytochrome b562, fused to fluorescent proteins, eGFP and mKATE2, whose fluorescence is quenched or unaffected by heme, respectively. Thus, the eGFP:mKATE2 fluorescence ratio is inversely correlated with bioavailable heme as measured by HS1. HS1 was previously used to characterize heme homeostasis in yeast, bacteria, and mammalian cells, and was instrumental in identifying new heme trafficking factors and signals that alter heme biodistribution and dynamics (33,35,36). We asked if cytosolic heme bioavailablity is altered in response to PhenDC3 (33). As indicated in Figure 4C, single cell analysis of a population of ~1500 HEK293 cells per condition indicate the median HS1 eGFP/mKATE2 ratio increases upon heme depletion in heme deficient media containing SA (HD+SA) and decreases upon increasing intracellular heme when cells are conditioned with the heme biosynthetic precursor 5-aminolevulinic acid (ALA) to drive heme synthesis. Titration of PhenDC3 results in a dose dependent increase in HS1 sensor ratio, indicating heme is less bioavailable when it is displaced from G4s in rRNA. The fractional heme saturation of HS1 decreases by ~15% ( Figure  4D). Together, our data indicate that rRNA G4s bind heme and regulate intracellular heme bioavailability.

DISCUSSION
The results here provide strong evidence that ribosomal tentacles form G4s in human cells, and that these G4s are involved in appropriating heme.
Immunofluorescence experiments with BG4 and L19 antibodies suggest a specialized fraction of cytosolic ribosomes (~5%) form G4s and that most extra-nuclear G4s (~83%) are ribosomal. The small fraction of ribosomes observed to form G4s in vivo contrasts with the high stability of ribosomal G4s in vitro (7,8). This difference is consistent with Guo and Bartel, who propose that eukaryotic cells have a robust machinery that unfolds G4s (16). However, the high concentration of rRNA acts in opposition to the low frequency per ribosome, so the RNA G4s are reasonably abundant. The RNA G-quadruplexome appears to be ribosome-centric.
We previously reported that surfaces of both the SSU and the LSU contain G4 sequences (7,8). A broad variety of data are consistent with more extensive formation of We propose that heme-rRNA G4 interactions may be important for protein hemylation reactions or buffering cytosolic heme. Indeed, PhenDC3, which competes for heme binding in G4s, causes a decrease in heme bioavailability as measured by the heme fluorescent sensor, HS1. This could be due to displacement of heme from rRNA G4s to a site that is less exchange labile, resulting in the observed decrease in HS1 heme binding. Alternatively, the upregulation in heme oxygenase due to PhenDC3 treatment Hemin Agarose Binding. HEK293 cells were seeded onto a 6-well plate at an initial confluency of 20% in Dulbecco's modified Eagle's medium (DMEM) with 10% Fetal Bovine Serum (FBS) and allowed to seed for 48 hrs at 37 °C. Media was then replaced for DMEM with 10% heme-depleted FBS supplemented with 0.5 mM succinyl acetone (for SA-treated cells). For untreated cells, media was changed for DMEM in 10% regular FBS. Both treated and untreated samples were allowed to incubate at 37 °C for 24 hrs. Cells were then collected by scrapping and lysed using 1.5 mm zirconium Beads (Benchmark). Lysates were quantified by Bradford assay. In the meantime, hemin agarose beads and sepharose beads were equilibrated 3 times by centrifugation with Lysis buffer (0.1% Triton X-100, 10 mM Sodium Phosphate, 50 mM KCl, 5 mM EDTA, pH 7.5, 1X protease arrest, RNasin RNase Inhibitor (Promega)). 100 µL of beads (50 µL bed volume) were used per biological replicate. After bead equilibration, each lysate was divided into two and 10 µg were loaded to hemin agarose and 10 µg to sepharose beads.
Mixtures were allowed to bind for 60 min, rotating at 20 rpm at room temperature. Then, three washes were performed using Lysis buffer and supernatants were discarded. Each wash consisted of 10 min incubation at room temperature with 20 rpm rotation followed by centrifugation at 700g for 5 min. Bead bound fractions were eluted by a 15 min incubation at room temperature with 20 rpm rotation in 50 µL of 1M imidazole in Lysis buffer followed by centrifugation at max. speed for 2 min and supernatants were collected.
RNA was then extracted from eluted fractions with TRIZOL using the manufacturer's protocol. For the PhenDC3 titration in HEK293 cells experiment, the same protocol was followed with the difference that cells were allowed to seed for 24 hrs (20% initial confluency) and then PhenDC3 was added in increasing concentrations (5 µM, 10 µM, 20 µM). DMSO carrier treatment was performed the same way but with equivalent DMSO volumes. Cells were left at 37 °C for 48 hrs and collected and lysed as described above.
RT-qPCR. The sets of primers used can be found in Table S and yellow-green laser (ex 561 nm). EGFP was excited using the argon laser and was measured using a 530/30 nm bandpass filter, mKATE2 was excited using the yellowgreen laser and was measured using a 610/20 nm bandpass filter. Data evaluation was conducted using FlowJo v10.4.2 software. Single cells used in the analysis were selected for by first gating for forward scatter (FSC) and side scatter (SSC), consistent with intact cells, and then by gating for cells with mKATE2 fluorescence intensities above background were selected. The fraction of sensor bound to heme may be quantified according to the following equation (33): where R is the median eGFP/mKATE2 fluorescence ratio in regular media and Rmin and Rmax are the median sensor ratios when the sensor is depleted of heme or saturated with heme. Rmin and Rmax values are derived from cells cultured in heme deficient media conditioned with succinylacetone (HD+SA) or in media conditioned with ALA (33). The plot in Figure 4D was obtained by fitting the median sensor ratios in Figure 4C to the following 1-site binding model (33,54): where x is the independent variable, [PhenDC3] BG4 purification. pSANG10-3F-BG4 was a gift from Shankar Balasubramanian and with the ER (yellow pixels) and the one that does not colocalize (green pixels). "L19", "ER", and "BG4" images only present their respective fluorescence signals.       Gene Sequence (5' to 3') GQES7-a GAATTCTAATACGACTCACTATAGGGCGGAGGGGGCGGGCTCCGGC GGGTGCGGGGGTGGGCGGGCGGGGCCGGGGGTGGGGTCGGCGGG GGACCGAAGCTT