Metabolic RNA labeling in non-engineered cells following spontaneous uptake of fluorescent nucleoside phosphate analogues

Abstract

RNA and its building blocks play central roles in biology and have become increasingly important as therapeutic agents and targets. Hence, probing and understanding their dynamics in cells is important. Fluorescence microscopy offers live-cell spatiotemporal monitoring but requires labels. We present two fluorescent adenine analogue nucleoside phosphates which show spontaneous uptake and accumulation in cultured human cells, likely via nucleoside transporters, and show their potential utilization as cellular RNA labels. Upon uptake, one nucleotide analogue, 2CNqAXP, localizes to the cytosol and the nucleus. We show that it could then be incorporated into de novo synthesized cellular RNA, i.e. it was possible to achieve metabolic fluorescence RNA labeling without using genetic engineering to enhance incorporation, uptake-promoting strategies, or post-labeling through bio-orthogonal chemistries. By contrast, another nucleotide analogue, pAXP, only accumulated outside of the nucleus and was rapidly excreted. Consequently, this analogue did not incorporate into RNA. This difference in subcellular accumulation and retention results from a minor change in nucleobase chemical structure. This demonstrates the importance of careful design of nucleoside-based drugs, e.g. antivirals to direct their subcellular localization, and shows the potential of fine-tuning fluorescent base analogue structures to enhance the understanding of the function of such drugs.

Graphical Abstract

Graphical Abstract.

Introduction

RNA is a key player in biology, not only for its role as an information transmitter and regulator in cellular processes, but also as the carrier of genetic information in RNA viruses. As such, RNA is emerging as an increasingly important therapeutic agent and target. This is, for example, reflected by the growing number of clinically approved oligonucleotide-based therapeutics (1–3), and especially the recent success of SARS CoV-2 mRNA vaccines (4–6), which will likely lead the way for a broader translation of mRNA-based therapeutics during the coming decade (7). RNA is also an important target for nucleoside analogue antiviral drugs that inhibit replication during RNA viral infections (8) and hence carries significant potential as prevention for future pandemics as well as for treatment of current severe virus infections such as Ebola and Marburg (9). Being able to probe RNA molecules, including their building blocks, inside cells is important to understand the dynamics of RNA-involving processes as well as to track RNA delivery.

The total levels and sub-cellular distributions of coding and non-coding RNAs strongly influence the nature and current states of cells (10,11). Consequently, probing cellular transcriptomes with different levels of complexity is important for studying both healthy and diseased cells. Moreover, better understanding of the uptake, export, and intracellular distribution of both RNA molecules and nucleosides would considerably facilitate the development of new RNA-targeted drugs, including nucleoside-based antivirals. Notably, the subcellular distribution of such antivirals is important for their efficacy since the localization of RNA replication depends on the nature and lineage of the RNA virus (12–14).

RNA analytical methods that are live-cell compatible and thereby can probe RNA dynamics and spatiotemporal distribution are needed to study abovementioned RNA processes and to complement static methods such as RNA sequencing (15,16). Fluorescence microscopy offers spatiotemporal monitoring of cellular and molecular dynamics on biologically relevant time scales. However, whilst there is a wealth of genetically encoded fluorescent proteins and tags to study proteins of interest, as well as live-cell compatible probes to stain lipid structures (17) and DNA, corresponding options to visualize RNA and its building blocks remain scarce and highly reliant on ex situ labeling or cell fixation. Examples of in cellulo labeling of RNA include the monitoring of the dynamics of individual RNA transcripts by use of GFP-fusion proteins tethered to MS2 stem–loop sequences (18), fluorescent RNA aptamers (19,20) or RNA-FISH (21). A more direct approach to advance RNA imaging is to utilize cell-endogenous machineries for metabolic labeling of RNA (22,23) via enzymatic incorporation of fluorescent labels. Notably, since enzymatic incorporation mimics the modes of action of nucleoside drugs, this approach has potential to also facilitate structure-activity relationship (SAR) studies of, for example, nucleoside-based antivirals.

Metabolic labeling relies on the promiscuity of the biological machinery to allow incorporation of synthetic analogues of natural building blocks into proteins, RNA or DNA (22,23). RNA metabolic labeling has typically been performed using radioactive or isotope labels (24). To extend applications to live-cell microscopy intrinsically fluorescent or activated building blocks (22,23) have also been devised. An example of an intrinsically fluorescent building block which was compatible with in vitro RNA labeling using common RNA polymerases was recently reported by us (25) but this is to date a rare exception (26). As reviewed by Spitale et al. (22) and Kleiner (23), fluorescent metabolic labeling of RNA has traditionally been performed using pyrimidine and purine nucleoside analogues that must be activated after cellular transcription to introduce fluorescent labels via bio-orthogonal chemical reactions (27–33). The advantage of this approach is that the activated group (typically an azide, alkyne or vinyl) generally infer minimal structure perturbations to the nucleoside analogue, which facilitates their cellular processing via the nucleotide salvage pathway and enable efficient RNA incorporation. A weak point is that many bio-orthogonal reactions require cell fixation before the readout. Strain-promoted azide-alkyne cycloaddition (SPAAC) (34,35) chemistry overcomes this disadvantage but has proven to proceed with slow kinetics (22,23).

Other impeding problems for efficient live-cell metabolic RNA labeling relate to the poor cellular uptake of many modified nucleotides (36) and the associated risk of cytotoxicity caused by interference with RNA processing, translation, or modification (37,38). Nucleotides are natively negatively charged and therefore typically unable to diffuse directly across the plasma membrane and, as such, require transfection, microinjection, or electroporation to enter cells (23). To circumvent this problem, Luedtke et al. developed the nanoparticle-forming fluorescent TAMRA-dATP, which was reportedly internalized via endocytosis (39).

A more common approach has been to use uncharged nucleoside analogues, but since they are nevertheless polar, they are generally not particularly membrane permeable. If the chemical modifications of the nucleoside analogues are sufficiently small, they can be internalized via cellular nucleoside transporters in the concentrative transporter (CNT) or equilibrative transporter (ENT) families. These are particularly important for adenosine uptake (40). After cell entry, such nucleosides need to be re-phosphorylated three times by cytosolic kinases to function as substrates in transcription. Alternatively, prodrug approaches have been developed to reduce the charge and increase the lipophilicity of nucleosides (41–43) and hence facilitate their passive diffusion across the plasma membrane. This has been successful for some antivirals (e.g. remdesivir (44,45)). Moreover, prodrug technologies such as ProTide uses monophosphate analogue forms of a nucleoside and hence decreases the number of re-phosphorylation steps that are needed inside of the cell (41–43).

To enhance the efficiency of metabolic RNA labeling, genetic engineering tools have been developed based on the overexpression of one or several of the enzymes in the nucleotide salvage pathways to enhance the re-phosphorylation of the modified nucleoside. For example, Wang et al. showed that upon overexpression of uridine-cytidine kinase 2 (UCK2) it was possible to incorporate the intrinsically fluorescent tricyclic cytosine analogue tC (46,47) into the RNA of living cells (48). We have previously developed several fluorescent base analogues (FBAs) as tools for nucleic acid investigations in vitro (47,49–56). More recently, we applied this innovation for spatiotemporal monitoring of the uptake and subcellular distribution of antisense oligonucleotide (ASO) gapmers and furthermore expanded the technology for monitoring biological functions of full-length mRNAs using live-cell confocal fluorescence imaging of the emission from different FBAs (25,57). This development was, in part, enabled by the establishment of a new robust route to synthesize analogues of NTPs, enabling enzymatic production of RNAs of interest using in vitro transcription (25).

The herein presented study builds on our previous observations that the tricyclic cytosine analogue tC^O was very well accepted by RNA polymerases in vitro and further triggered by the serendipitous finding in the lab that two fluorescent adenosine triphosphates, namely pATP and 2CNqATP, but not tC^OTP (Figure 1A), were spontaneously taken up and effectively accumulated intracellularly by various cultured cell lines of human origin. This prompted us to study their uptake pathways, subcellular distributions, and processing. Based on this, we also explored a straightforward approach for fluorescent metabolic labeling of cellular RNA. Furthermore, through live-cell fluorescence imaging, we observe notable differences in the subcellular distributions and accumulation of pA- and 2CNqA-nucleotides, influencing their metabolic labeling applicability. These findings emphasize the value of cell-compatible fluorescent analogues and their potential in optimizing nucleoside analogue drugs and underscore the significance of comprehending the cellular distribution and localization of these compounds in the rational design of novel nucleoside-based antivirals, among other applications.

Figure 1.

Molecular structures and spectra of fluorescent nucleoside triphosphates. (A) Molecular structures with the modifications, compared to the natural compound, in color: quadracyclic adenosine triphosphate 2CNqATP (green), pentacyclic adenosine triphosphate pATP (blue), tricyclic cytosine triphosphate tC^OTP (yellow). (B) UV–vis absorption (ϵ, solid line) and normalized emission (dotted line) spectra of the triphosphates in (A), with the same color code.

Materials and methods

Synthesis

The cytosine analogue tC^OTP and the adenosine analogue 2CNqATP were synthesized following literature procedures (25,58). Synthesis of pA nucleoside and pATP is described in detail in the supplementary information. Briefly, the triphosphate synthesis is a solid-support reaction procedure using the appropriate 5′-protected ribonucleoside as starting material. The monophosphate of 2CNqA was prepared using the same method, but without the pyrophosphate treatment.

Stock preparation and general handling

The nucleosides were dissolved in Dulbecco's phosphate-buffered saline 1× (DPBS, Gibco, 14190144, containing no calcium or magnesium) for cell culture. The stock concentrations were measured using a Cary 50 spectrometer (Agilent Technologies). Absorption was measured at 260 nm the concentration determined by Beer–Lambert law (ϵ_260nm of the modified bases in Supplementary Table S1). Hereafter the aliquots for cell work were used and stored in cold sterile conditions.

Photophysical characterization

UV–vis absorption spectrum of pATP in Dulbecco's phosphate-buffered saline 1× (DPBS, Gibco 14190144, containing no calcium or magnesium) were recorded on a Cary 4000 spectrometer (Agilent Technologies) with a wavelength interval of 1 nm, integration time of 0.1 s, and a spectral band width (SBW) of 1 nm. A cuvette with an optical path length of 3.0 mm and working volume of 60 μL was used. The molar absorptivity spectrum ((λ) in M⁻¹ cm⁻¹, Figure 1B) was calculated using equation 1, with A(λ) being the absorption, A₂₆₀ the absorption at 260 nm, and ₂₆₀ the molar absorptivity at 260 nm (₂₆₀= 22 300 M⁻¹ cm⁻¹ for pATP) (56).

(1)

The fluorescence quantum yield (Φ_F) was determined relative to the quantum yield of quinine sulphate in 0.5 M H₂SO₄ (Φ_F,r = 0.546) (59), and calculated using equation 2, in which subscripts s and r refer to sample and reference, respectively. I is the integrated fluorescence intensity, A the absorption of the fluorophore at the excitation wavelength and n is the refractive index of the solvent (1.33 for H₂O or 1.43 for H₂SO₄) (59).

(2)

The emission spectra were measured on a SPEX Fluorolog 3 spectrofluorometer (Jobin Yvon Horiba) in a quartz cuvette with 3 mm optical pathlength. Three consecutive emission spectra were collected and averaged. The spectra were recorded from 360 nm to 700 nm with excitation at 355 nm and 1.0 nm wavelength interval with an integration time of 0.1 s. To avoid bleaching, the monochromator slits on the excitation side were set to 0.8 nm, and on the emission side to 2.5 nm. The spectrum of PBS was subtracted from all sample emission spectra to account for Raman scattering. The absorption at the excitation wavelength for both sample and reference were below 0.02 to avoid inner filter effects. The quantum yield determination was done in two independent experiments with 0.47 and 0.51, which results in a mean value of 0.49 (S.D. = 0.03).

Fluorescence lifetimes were determined using time-correlated single photon counting (TCSPC). The excitation light was provided by an LDH-P-C-375 pulsed laser diode (PicoQuant), centered at 377 nm (FWHM pulse width was 1 nm and 70 ps with respect to wavelength and time, respectively). System was operated with a PDL 800-B (PicoQuant) laser driver at a repetition frequency of 10 MHz. The sample emission was collected at a right angle with respect to the excitation light beam using a monochromator that was set to 460 nm (SBW = 10 nm) and an emission polarizer set to magic angle detection (54.9° with respect to excitation polarization). A microchannel-plate photomultiplier tube (R3809U-50; Hamamatsu) collected the photons and the signal was fed into a LifeSpec multichannel analyser (Edinburgh Analytical Instruments) using 2048 active channels (24.4 ps/channel), and a stop condition at 10 000 counts in the top channel. The instrument response function, , was determined using a frosted glass (scattering) insert while observing the emission at 377 nm (SBW = 10 nm). The intensity decays, , were fitted with re-convolution according to equation 3, where and are the amplitude and time constant of the i:th exponential, respectively.

(3)

The DecayFit software (DecayFit – Fluorescence Decay Analysis Software 1.3, FluorTools, www.fluortools.com) was used to perform the least-square fitting procedure. All decays were satisfactorily fitted to a monoexponential expression ( = 1).

Cell lines and cell culture conditions

Wild-type Huh-7 cells were cultured at 37°C and 5% CO₂ in Dulbecco's modified Eagle's medium (DMEM GlutaMax, Gibco 21885-025) with an addition of 10% fetal bovine serum (Gibco 10270-106, origin Brazil). Wild-type HEK-293T cells were cultured using the same medium composition. Wild-type SH-SY5Y human neuroblastoma cells were grown at 37°C and 5% CO₂ in a 1:1 mixture of minimal essential medium (Gibco 41090-028) and nutrient mixture F-12 Nut Mix Ham containing l-glutamine (Gibco 21765-029) supplemented with 10% fetal bovine serum (as above), 1% non-essential amino acids (Lonza 13-114E). For sub-cultivation the adherent cells were washed with calcium and magnesium free DPBS (Gibco 14190-144), and detached with 0.25% trypsin–EDTA (Gibco, 25200-056). All cells were cultured in a mycoplasma free lab and verified to be mycoplasma free. For seeding, the cells were counted after trypsin neutralization and diluted to the desired number of cells and thereafter incubated at 37°C with 5% CO₂ for 24 h before experiments.

Confocal laser scanning microscopy

Images were captured using a Nikon eclipse Ti microscope with a Nikon C2plus scanner, two parallel Gallium arsenide phosphide (GaAsP) detectors, and an Apo 60x Oil λS DIC N2 objective. To excite the FBAs the 405 nm laser line was used, and the emission was collected between 407 and 607 nm. To optimize the signal while avoiding possible photobleaching of the compounds, the laser power was kept low.

The cells were seeded in four-compartment dishes with glass bottoms (CELLview Dishes, Greiner Bio-One) and, unless stated differently, at a concentration of 0.18 million cells/ml in a working volume of 250 μl. During imaging the dish was placed in a stage top incubation chamber (OKO lab) maintaining 37°C and 5% CO₂. Exposure time started by adding pre-warmed (37°C) cell culture medium (CCM) containing 2.5 μM of FBA to the cells. Control cells were treated with an equal volume of DPBS in CCM. To image the uptake of FBA-TPs over time, the time-lapse setup of the NIS software was used. The time between exposing the cells to the compounds under sterile conditions to recording the first image of the time lapse was generally 3–5 min, because of the need of mounting the dishes in the stage top incubator.

To test if FBA uptake occurred via an active mechanism, cells were analyzed at 4°C. This was achieved by keeping the cells in the fridge (4°C) for 20 min prior to exposure. Treatment solutions were prepared as described above, but on ice. After 1.5 h of incubation at 4°C the cells were washed twice with ice cold DPBS and fresh cold (4°C) CCM was added. During the subsequent image acquisition, the heating of the stage control was turned off and the chamber stage was pre-cooled with an ice block.

To investigate the cellular clearance of the FBAs upon exchange of the cell medium (to remove extracellular FBA), Huh-7 cells or SH-SY5Y cells were exposed to 2.5 μM of the FBAs over 24 h and imaged as described above. Thereafter, medium was removed, fresh CCM (without FBA) was added, and cells were imaged again at indicated time points.

To investigate the effect of exposure with hydroxyurea, cells were exposed to 15 nM (previously used in our lab for long-term exposures), 50 μM (60), or 10 mM (48) of hydroxyurea for 30 min prior to the exposure with 2.5 μM 2CNqATP. Cell uptake was thereafter imaged as described above.

Pulse-chase

Typically, 2CNqATP was first added to the cells to facilitate imaging the effect of NaAsO₂ or Actinomycin D (Act D) in the fluorescence channel. Cells were seeded in a microscopy dish as described above. At the start of the experiment, the cells were exposed to 2.5 μM 2CNqATP for 3 h. Cells were then washed once with PBS whereafter 4 μM Act D or 50 μM NaAsO₂ was added and incubated for 3 or 2 h, respectively. A test series for these conditions was run and is presented in Supplementary Table S3.

Image analysis

Microscopy images were analyzed using ImageJ. For line-profile analysis the fluorescent channel was extracted, and a line was drawn through a single cell crossing at least one nucleolus. Distance vs. intensity data was extracted and normalized so that the average value of the cytoplasmatic signal was unity. This was done for three images in which five cells were selected (i.e. a total of 15 intensity profiles). The line profiles were thereafter analyzed using the batch peak analyzer in the Origin software with a theme created for this analysis: Baseline mode was set to median, and baseline was automatically subtracted. To find the peaks, only positive direction was allowed with a local maximum of two points, whereafter, the peaks were filtered by height value with the threshold set to 1. For analysis of nucleoli area 30 cells across three different images per condition were selected and analyzed manually. The polygon selection tool in ImageJ was used to outline nucleoli and thereby measure their area. Mean areas and standard deviations were calculated from the resulting values.

Cytotoxicity assays

Huh-7 cells were seeded in a 96-well plate at a concentration of 0.18 million cells/ml and a working volume of 100 μl per well, one day prior to exposure. Treatment solutions of indicated concentrations of FBA in CCM were prepared as a dilution series and cells were treated with these solutions for 24 h. Unexposed cells were treated in the same way, albeit with an equal amount of DPBS added to cell culture medium instead of FBA. Two different cytotoxicity assays (AlamarBlue cell viability and LDH membrane leakage) were performed in parallel on each sample using the cells and the culture medium from each sample, respectively. All treatments were done in technical triplicates and the experiment was repeated twice for triphosphate exposure and in a single experiment for 2CNqAMP.

The AlamarBlue assay was performed in the following way: the conditioned medium was removed from the treated cells (and saved for the LDH leakage assay, see below). Resazurin reagent (AlamarBlue cell viability reagent, Invitrogen 1960012) was freshly diluted in cell culture medium (1:10) and added to the cells. Thereafter, the cells were incubated for 3 h at 37°C and 5% CO₂. The resulting resorufin fluorescence was read using an Optima Fluostar plate reader (BMG Labtech) with an excitation filter of 544 nm (±9 nm) and an emission filter of 590 nm (±7 nm). Dimethyl sulfoxide 5% and CCM alone (i.e. no FBA added) were used as positive and negative controls for cell death, respectively. Relative metabolic activities in the cell samples were calculated by subtracting DBPS () as background from all measured fluorescence intensities of FBA samples (). The measured fluorescence intensity of all replicates of untreated cells was averaged and cell viability of the treated cells were expressed as fraction of the signal from untreated cells ( (equation 4).

(4)

To evaluate the effect of FBAs on cell membrane integrity the CyQUANT LDH Cytotoxicity Assay kit (Thermo Scientific, C20300) was used according to the manufacturer's instructions. The absorption was measured at 490 nm and 630 nm on a Thermo Scientific multiscan plate reader GO, within an hour after stopping the reaction. To determine LDH activity, i.e. LDH present in cell culture medium the background absorption at 630 nm was subtracted from the absorption at 490 nm for each sample. Biological triplicates of all measurements were averaged, signal from untreated cells subtracted and expressed as fraction of the signal from lysed cells (i.e. maximum LDH release) (equation 5).

(5)

Flow cytometry

Huh-7 cells were seeded in 96-well plates at 0.4 million cells/ml and a working volume of 100 μl one day prior to the experiment. Following treatment, the cells were washed twice with DPBS, and thereafter harvested by addition of 0.25% trypsin-EDTA. The trypsin was neutralized by adding DPBS with 2% FBS and the samples were transferred to 96 U-bottom well plates for high throughput readout using flow cytometry.

Measurements were performed on a Luminex CellStream flow cytometer with a high throughput sampler (HTS). The FBAs were excited using a 405 nm laser and the emission collected through a 456/51 nm bandpass filter. To ensure a good cell dispersion and to disrupt eventual clusters, a mixing step was added before measuring. Stop criteria for cell counting was set to 5000 events of single cells. Untreated cell samples were used to adjust instrument settings and set the gate for single living cell based on forward scatter (FSC) and side scatter (SSC) area. The detector gain was adjusted so that the background signal (cell autofluorescence) appeared at around 10² in the fluorescence histogram.

Concentration dependent uptake

Cells were exposed to 0–10 μM 2CNqATP or 2CNqAMP in CCM for 24 h. 2CNqATP uptake was quantified in whole cell lysates, following a wash (1×) with DPBS addition of passive lysis buffer (Promega, diluted in DPBS) at 4°C overnight. The fluorescence in whole cell lysates was determined using an Optima Blue Fluostar plate reader (BMG Labtech). 2CNqATP was excited using 355 nm (±20 nm) filter and emission was collected through a bandpass filter centered at 460 nm (±12 nm). Lysate buffer fluorescence was subtracted as background, and the fluorescence intensities of technical triplicates were averaged. 2CNqAMP uptake was instead quantified using flow cytometry (vide supra). To confirm that the plate-reader and flow cytometer-based methods gave comparable results, a control experiment was performed where Huh-7 cells were exposed to 2.5 μM or 10 μM of 2CNqATP or 2CNqAMP for 24 h the cells thereafter analyzed by both methods (Supplementary Figure S5). Finally, to compare the results, data from all experiments were normalized to the highest added concentration of FBA-TP (10 μM).

Uptake kinetics

For uptake kinetics experiments, Huh-7 WT cells were exposed to solutions of 2.5 μM FBA in CCM for different time points. As control, an equal volume of DPBS was added to the CCM instead of FBA. Flow cytometry experiments were conducted in 96-well plates (vide supra), where all treatments were measured in technical triplicates and in three independent biological experiments. The timing of the experiment was set up so that all incubations ended at the same time, and all cell samples were thus harvested at the same time. The mean cellular fluorescence from three technical replicates was averaged and plotted as a function of time. Three such kinetic curves, recorded on separate days were collected. A mono-exponential function (ExpDec1 model) was globally fitted to the data using Origin software (OriginPro Version 9.7.0.188) and a Levenberg Marquard iteration algorithm (all fitting details in Supplementary Table S2). This yielded time constants for 2CNqATP and 2CNqAMP uptake. Cells were also seeded in four-chamber microscopy dishes and exposed as described above to record time-lapse videos of the uptake.

Competition of uptake with canonical nucleotides

The uptake of 2CNqAMP and 2CNqATP into Huh-7 cells were competed with natural adenosine triphosphate (ATP, Sigma-Aldrich) or adenosine monophosphate (AMP, Sigma-Aldrich). 2.5 μM 2CNqATP was competed with 625 μM ATP and 2.5μM 2CNqAMP with 625 μM AMP, i.e. 250-fold excess of the canonical nucleoside triphosphate. The readout is based on the fluorescence of the FBA compounds, detected in living cells by flow cytometry following co-administration of the analogue and the canonical nucleotide. For imaging, cells were seeded as described above and then exposed as described for the flow cytometry experiment. Images were recorded as described in the microscopy section after 4 h of exposure without changing the medium.

RNA extraction from cells exposed to FBA

Huh-7 cells were seeded in 12-well plates (working volume 1.0 ml/well) so that a cell number of 2 × 10⁶ was achieved for extraction. For exposure, conditioned medium was removed from the cells and replaced by treatment solutions containing 2.5 μM 2CNqATP or 2CNqAMP in CCM. For control cells CCM alone was added. Cells were incubated for indicated times at 37°C and 5% CO₂. For RNA extraction and purification, the QIAGEN RNeasy Mini Kit was used, following the manufacturer's protocol albeit with slight modifications described here. Briefly, the cells were washed three times with DPBS, detached, counted, and pelleted. Lysis buffer was added to the pellets, followed by homogenization by 12 times passing it through a 20-gauge needle (0.9 mm outer diameter). An equal volume of ethanol was then added to the lysed cells and the resulting solution was transferred to a RNeasy spin column and centrifuged. Then, binding buffer was added on top of the column which was centrifuged again. The column-bound RNA was washed 5X with washing buffer, with a centrifugation step between each addition. To elute the extracted RNA, 30 μl of the provided RNase-free water was applied to the column, followed by centrifugation. This elution step was repeated three times using the same collection tube, rendering a total volume of 90 μl. For controls unexposed cells were lysed as described above and FBA was thereafter added directly to the lysate to make up a 2.5 μM final concentration. All control samples were applied to RNeasy spin columns and treated the same way as the compound-exposed cell samples, following the purification protocol described above.

For co-exposure of the cells with hydroxyurea, a solution containing 2.5 μM 2CNqATP or 2CNqAMP and hydroxyurea was prepared and exposed to Huh-7 WT cells for 24 h. Low hydroxyurea concentration (15 nM) was chosen to minimize toxic effects associated with long exposure times. Hereafter RNA extraction was performed as described above.

Analysis of extracted total cellular RNA

Absorption spectra of extracted RNA samples and controls were recorded as described above. Emission spectra were recorded on a SPEX Fluorolog 3 spectrofluorometer (Jobin Yvon Horiba) by exciting the FBA at 350 nm and recording the emission from 358 nm to 690 nm. The excitation slit was set to 2 nm, and the slit on the emission side to 3 nm with an integration time of 0.05 s. The emission spectra were normalized to the corresponding absorption value at 260 nm, to compensate for the RNA concentration variation. Excitation spectra were measured on the same instrument with the emission wavelength set to 443 nm, and the excitation was scanned between 235 nm and 445 nm. The slits on the excitation side were set to 2 nm and on the emission side to 3 nm with an integration time of 0.05 s. For comparison, excitation spectra were overlaid with absorption spectra of artificial 2CNqA-containing 25mers of single-stranded RNA with the following sequence: 5′-CGA CAA AAU CAA [2CNqA]AU GCG UGA UUG G-3′ (55), hereafter referred to as 2CNqA-ssRNA. For comparison to 2CNqA-containing double-stranded RNA, the same strand was hybridized with a fully complementary RNA sequence (U complementary to 2CNqA), hereafter referred to as 2CNqA-dsRNA. Absorption spectra were recorded in water as solvent at room temperature as described above under ‘photophysical characterization’.

To estimate the degree of 2CNqA incorporation in the total RNA extracts, we generated a standard curve by diluting pure 2CNqATP solutions in milliQ water followed by recording of emission spectra using the same instrument settings as those used for the readout of the RNA extracts. The emission spectra were integrated over the full wavelength range and background corrected by subtracting the integrated area of a corresponding spectrum of milliQ water. The resulting intensity values were corrected for the change in fluorescent quantum yield when 2CNqA is incorporated into dsRNA (Φ_F,2CNqATP = 0.47(58) versus Φ_{F 2CNqAdsRNA} = 0.12 (55)) and for the slight change in excitation due to a shift in absorption (a factor of 1.1). The emission data for the extracted RNA were treated in the same way and their 2CNqA concentrations were estimated using the standard curve. The total RNA base concentration in the samples were calculated based on the absorption at 260 nm as described above. Finally, by dividing the 2CNqA concentration with the total nucleotide concentration, the incorporation degree was determined.

It should be noted that accurate quantification and an exact comparison of the incorporation kinetics for 2CNqAMP and 2CNqATP based on spectral readout poses a considerable challenge as many parameters and controls (ranging from strict consistency in cellular treatment and thorough spectroscopic characterization of the compound inside total RNA extracts) need to be considered and, hence, must be performed carefully and interpreted with caution.

Addition of RNA to 2CNqATP or 2CNqAMP

To investigate potential effects of non-covalent binding on the emission of the 2CNqA fluorophore, various concentrations of purified RNA from an in vitro transcription reaction was added to a constant amount of 2.5 μM 2CNqATP or 2CNqAMP (i.e. one solution per condition). Excitation spectra were recorded using a SPEX Fluorolog 3 spectrofluorometer (Jobin Yvon Horiba) with emission at 460 nm and excitation from 230 to 450 nm with an integration time of 0.1 s. The slits on the excitation and emission side were set to 1 and 2 nm, respectively.

Gel electrophoresis

To examine the size distribution of the total RNA extract, gel electrophoresis was used. 2% agarose gel (% w/v) were cast in 1x TBE with 1× SYBR Safe dye (Thermo Scientific). 500 ng of the RNA samples in formamide containing 1× Loading dye (Thermo Scientific) were loaded. A standard RiboRuler High Range RNA ladder (Thermo Scientific) in 1× loading dye was used for size comparison. The RNA samples and ladder were denatured in loading dye (containing formamide) for 15 min at 65°C followed by immediate cooling on ice for 5 min prior to loading onto the gel. The gel was run in 1x TBE for 1.0 h at a constant voltage of 100 V and then imaged using a ChemDoc Touch (BioRad) gel imager with SYBRsafe settings (defined by the instrument).

Results

Synthesis and photophysical characterization of fluorescent nucleoside phosphates

We have previously reported the synthetic routes towards the tC^O- and 2CNqA-triphosphate derivatives (Figure 1A) (25,58), whereas the synthesis of the pATP (Figure 1A) variant was developed here. Briefly, the pA nucleoside was obtained via Silyl-Hilbert-Johnson glycosylation of 4-chloro-5-iodo-7H-pyrrolo[2,3-d]pyrimidine, followed by a Miyaura borylation on the C5 position of the scaffold. Subsequent Suzuki coupling was performed to introduce the Boc protected 2-napthylamine at position C5. Ring closure in basic conditions followed by global deprotection provided the final pentacyclic nucleoside, pA. pATP (Figure 1A) was synthesized from the pA using the same protocol as used for tC^OTP (25) and 2CNqATP (58), highlighting the robustness and generality of the protocol.

The photophysical properties of FBA-TPs in aqueous solution are critical for understanding and interpreting fluorescence-based in vitro and in cellulo data. The UV-vis absorption and emission spectra of pATP were, therefore, recorded and compared to corresponding spectra of tC^OTP (25) and 2CNqATP (58) (Figure 1B). A common feature for the FBA-TPs is that they primarily absorb in the near-UV region, with their lowest energy transition tailing out into the visible region, making them excitable using the 405 nm light source that is available on most conventional fluorescence microscopes. The pATP emission appears as an unstructured band centered at 420 nm and is moderately blue shifted compared to the slightly broader, but equally unstructured, spectra of tC^OTP (453 nm) and 2CNqATP (463 nm).

The fluorescence quantum yields (Φ_F) of tC^OTP and pATP were determined to 0.27 and 0.48, respectively, which are comparable to that of 2CNqATP (0.49) (58). The brightness values, for which molar absorptivity (ϵ) is factored in: B = Φ_F × ϵ, range from 2700 to 6000 M⁻¹cm⁻¹ at the lowest transition band maxima (Supplementary Table S1) and are exceptionally high for FBAs. To characterize and prime the FBA-TPs for possible applications in lifetime-imaging based techniques, we also determined the fluorescence lifetimes (τ_F) of tC^OTP (3.4 ns) and pATP (6.0 ns). The fluorescence lifetime of 2CNqATP is 9.9 ns as published in Nilsson et al. (58).

Fluorescent adenosine analogue triphosphates are spontaneously taken up into human cells

With substantial amounts of the three FBA triphosphates at hand we decided, contrary to what is a rational investigation considering their high negative charge (net charge of –3.3 at physiological pH (61)), to add tC^OTP, pATP, and 2CNqATP to the medium of adherent cultures of three human cell lines (Huh-7, SH-SY5Y and HEK-293T) and explored their interactions with the cells using confocal fluorescence microscopy (Figure 2A). This led to the surprising finding that the two ATP analogues 2CNqATP and pATP were spontaneously taken up by the cells.

Figure 2.

Spontaneous cellular uptake of fluorescent nucleoside triphosphates. (A) Confocal fluorescence microscopy images of the cellular uptake of 2CNqATP, pATP, and tC^OTP. Huh-7, SH-SY5Y or HEK-293T wildtype cells were exposed to 2.5 μM of the compounds in CCM and images were captured after 24 h at 37°C. (B) Representative fluorescence intensity line profile to depict the subcellular distribution of 2CNqATP following uptake. The nucleoli display 2.3 ± 0.1 times higher emission intensity compared to the nucleus and cytosol. (C) Cytotoxicity of 2CNqATP, pATP, and tC^OTP to Huh-7 cells as determined by changes in metabolic activity relative to untreated cells at 24 h, using the AlamarBlue assay. All treatments were done in triplicates and the experiment was performed twice. The data are presented as mean ± standard deviation. (D) The effect of Actinomycin D treatment visualized using 2CNqATP. Cells were treated for 3 h with 2.5 μM 2CNqATP only or with 2.5 μM 2CNqATP and 4 μM Actinomycin D in a pulse-chase experiment. Thirty nucleoli in three different images were measured for each condition using the polygon selection tool in ImageJ. Arrows highlight two representative nucleoli. The scatter plot to the right shows the data points with the mean and standard deviation. P-value was calculated using a two-tailed heteroscedastic t test; **** P< 0.0001. The scale bars in the confocal images represent 20 μm.

On the other hand, tC^OTP was unable to spontaneously traverse the cellular membranes, which is more in line with common notion. tC^OTP is a close derivative of the tC nucleoside which was recently used in metabolic labeling by Wang et al. (48). The apparent lack of spontaneous uptake of the tC^OTP could potentially be related to emission differences between the A- and C-analogues in the intracellular environment such that tC^O emission would be strongly quenched. However, since all three FBAs have been thoroughly characterized in cell-free buffers and studied as incorporated labels in nucleic acids inside live cells without major effects on their emissive properties (25,57), and since these FBAs unlike many others maintain a bright emission irrespectively of their environment (17,26,50,51), this cannot reasonably explain the significant differences in intracellular fluorescence observed here.

Further, we found that the subcellular distributions of 2CNqATP and pATP fluorescence are remarkably different. This observation was highly consistent across the three cell types (Figure 2A). 2CNqATP accumulated in both the cytoplasm and the nucleus of the cells, with no apparent concentration gradient. Additionally, 2CNqATP was found to accumulate in nuclear sub-compartments (nucleoli). Line profile intensity analyses showed that these have a 2.3 (±0.1) times higher intensity compared to the rest of the nucleus and the cytoplasm (Figure 2B).

By contrast, pATP fluorescence was entirely absent in the cell nuclei and unevenly distributed across the cell body with an apparent higher accumulation in perinuclear locations that may correspond to the endoplasmic reticulum or Golgi. These observations emphasize that 2CNqATP and pATP are sensed as different molecules by the cells and show, importantly, how relatively small changes in the chemical structure of their nucleobase part (Figure 1A) can have significant impact on their biological action and potential. As judged from the observation of normal cell morphologies (Figure 2A), the absence of aberrant effects on cell membrane integrity (tested using the LDH leakage assay, data not shown), and the unaffected metabolic activity (tested using the AlamarBlue assay, Figure 2C) 24 h post exposure, none of the three FBA triphosphates appear to be toxic to the cells.

Following the observation of the ease and consistency of visualizing the nucleoli in several cell lines by merely adding 2CNqATP to the cells, we decided to investigate its applicability as a label to probe dynamics of this membrane-less compartment. We exposed cells, which had been preincubated with 2CNqATP, to Actinomycin D (Act D) (Figure 2D), a drug that has previously been shown to affect nucleolar structure by arresting transcription due to DNA intercalation (62). The Act D treated cells had smaller nucleoli in accord with previous reports using immunostaining or electron microscopy to visualize their structure (62,63). By analyzing cross-sectional areas, we found that size of the nucleoli decreased by a factor of 2–3 as a result of Act D exposure (Figure 2D, Supplementary Table S2). The same effect was seen when the addition order is reversed, i.e. when exposing cells to Act D first, followed by 2CNqATP (Supplementary Figure S1), suggesting that the morphological change to nucleoli was independent of the presence 2CNqATP. On the contrary, when instead exposing the cells to NaAsO₂, known to trigger the integrated stress response without altering nucleoli integrity and structure (63), the nucleoli stayed intact until the cells eventually died (Supplementary Figure S1, Supplementary Table S3).

2CNqATP, pATP and their monophosphate derivatives are actively internalized by cells

To better understand the spontaneous and selective uptake of the two adenosine analogues (2CNqATP and pATP), we explored cellular uptake into Huh-7 cells at low temperature (4°C) to disfavor energy-dependent cellular processes, including membrane transport mechanisms and endocytosis (64). No uptake was observed after 1.5 hours of treatment at 4°C (Figure 3A, top panels) whereas cells kept at 37°C for the same period of time showed intense intracellular fluorescence (Figure 3A, bottom panels), in accord with the data presented in Figure 2. This result strongly suggests that both 2CNqATP and pATP are internalized via active cellular processes rather than by passive transmembrane diffusion.

Figure 3.

Investigations on fluorescent adenine analogue uptake pathways in Huh-7 cells. (A) Confocal microscopy images of the uptake of 2CNqATP and pATP after 1.5 h at 4°C compared to the corresponding uptake at 37°C. (B) Confocal microscopy images of pATP and 2CNqATP uptake before and after removal of the treatment solution and adding fresh CCM. (C) Molecular structure of 2CNqAMP and confocal microscopy images of 2CNqAMP uptake at 37°C and 4°C at indicated time points. (D) Uptake of 2CNqATP (yellow) and 2CNqAMP (red) at 37°C as function of extracellular FBA concentration. The data were recorded after 24 h of incubation. Readout of the 2CNqATP uptake was performed in cell lysates using a plate reader (mean ± S.D., n = 3) while readout of 2CNqAMP was done using flow cytometry (mean ± S.D., n = 3). Lines were added to guide the eyes. Comparison of the methods are presented in Supplementary Figure S5. (E) Confocal images show the uptake of 2CNqAMP (top) and 2CNqATP (bottom) at the indicated time points after exposure at 37°C. Analysis using flow cytometry presented in Supplementary Figure S7 and Supplementary Table S4. (F) Uptake of 2CNqATP and 2CNqAMP after 3 h at 37°C in the presence (+AXP) and absence (-AXP) of a 250-fold excess (625 μM) of canonical ATP (for 2CNqATP) or canonical AMP (for 2CNqAMP) measured using flow cytometry (mean ± S.D., n = 3). Representative intensity histograms from the flow cytometry are shown in Supplementary Figure S8. Right panel shows representative confocal microscopy images of the uptake. The FBA concentration was 2.5 μM in all exposures, except in D). All scale bars represent 20 μm.

Furthermore, the broad subcellular distributions in the cytoplasm, and for 2CNqATP in the nucleus, is not consistent with uptake via endocytosis which results in the appearance of bright cytoplasmic foci corresponding to endosomes and/or lysosomes (65). In a second set of experiments, we explored the cellular clearance of 2CNqATP and pATP fluorescence by exposing Huh-7 and SH-SY5Y cells to the compounds for 24 h, followed by cell culture media exchange (Figure 3B and Supplementary Figure S2, respectively). We found a remarkable difference in retention between the two analogues, which was consistent across both cell lines, where 2CNqATP uptake resulted in stable fluorescence intensity inside the cells, whereas pA fluorescence disappeared within 15 min. This indicates that the intracellular pA concentration is dependent on an equilibrium between uptake and export, whereas 2CNqA has a stronger tendency to accumulate.

Cells can export and import adenine derivatives via various transmembrane transporters and, moreover, control their phosphorylation states via the so-called purinergic signaling pathway. The pathway includes a number of extracellularly active dephosphorylating enzymes (ecto-ATPases, ectopyrophosphatases and ectonucleotidases) (66). Focusing on the 2CNqA analogue, which rapidly accumulated in cells (Figure 3B), we compared the uptake of 2CNqATP to 2CNqAMP, the corresponding monophosphate derivative of 2CNqA, which also showed low toxicity to the cells (Supplementary Figure S3). We observed that the intracellular fluorescence resulting from 2CNqAMP cell treatment (Figure 3C) was distributed similarly to 2CNqATP (Figure 2A), with an enhanced accumulation in the nucleoli (2.3 ± 0.1 times intensity enhancement, Supplementary Figure S4). Like 2CNqATP, the uptake of 2CNqAMP was also inhibited at 4°C (Figure 3C).

Furthermore, we found that the concentration dependence of the 2CNqATP and 2CNqAMP uptake was similar and displayed saturation above ca. 2.5 μM (Figure 3D). It is worth noting that this was observed and reproduced using two readout techniques which show consistency of our results and highlight the possibility to analyze our probes with different readout methods (Supplementary Figure S5). Taken together, this strongly indicates that both 2CNqATP and 2CNqAMP are taken up and accumulated by the same energy-dependent, non-endocytic mechanisms, and suggests uptake via purinergic nucleoside importers, following dephosphorylation at the cell surface.

Using flow cytometry (Supplementary Figure S6) as well as confocal imaging (Figure 3E) we explored the kinetics of 2CNqATP and 2CNqAMP uptake, which were found to increase exponentially and thereafter asymptotically approach steady state at longer time points. However, the kinetic curves have distinctly different shapes, with 2CNqAMP being internalized faster and reaching a plateau value after ca. 2 h of incubation, whereas 2CNqATP uptake had a slower onset and continued to increase up to ca. 4 h. The kinetic data were well fitted to a single exponential model (Supplementary Figure S7), yielding time constants (τ) of 35 min and 74 min for 2CNqAMP and 2CNqATP uptake, respectively (Supplementary Table S4).

To further test the hypothesis that the uptake pathway of 2CNqATP and 2CNqAMP is related to adenine transporters, we performed competition experiments with excess addition of canonical ATP and AMP, respectively. Addition of a 625 μM of ATP or AMP to 2.5 μM of the corresponding FBA treatment solutions (250-fold excess), resulted in an approximate 2-fold decrease in uptake of the compounds (Figure 3F, Supplementary Figure S8). This demonstrates that competition occurs but with a rather low efficiency. A similar behavior was observed when pATP was added to cells together with a 250-fold excess of ATP, which indicates that both FBA derivatives use the same pathway to enter cells (Supplementary Figure S9). We also noted that the subcellular localization of the 2CNqA derivatives were altered by the presence of excess ATP or AMP (Figure 3F, images), suggesting competition in processes not only at the uptake stage but also inside the cells as well as a possible higher pressure for export in cells exposed to high ATP concentrations (physiological levels of extracellular ATP are typically in the nM range (67)). Altogether, the competition experiments support the conclusion that 2CNqA and pA derivatives can enter cells via the same pathway as natural adenosine derivatives and hence that their chemical modifications do not hinder them from being dephosphorylated by the enzymatic machinery on the cell surface and recognized and accepted by adenine importers.

2CNqATP is spontaneously incorporated into cellular RNA following cellular uptake

Having observed the uptake and intracellular accumulation of 2CNqATP, 2CNqAMP and pATP, we next investigated if the FBA derivatives could be incorporated into cellular RNA. We exposed Huh-7 cells to 2.5 μM solutions of 2CNqATP, 2CNqAMP, pATP or tC^OTP for up to 48 h and thereafter collected the total cellular RNA using spin column extraction as depicted in Figure 4A and detailed in the Materials and methods section. Direct additions of the FBA nucleotides to the collected cell lysate were included as negative controls. The extracted total cellular RNA was thereafter analyzed using fluorescence spectroscopy.

Figure 4.

Spectroscopic analysis of RNA extracted from cells exposed to 2CNqATP. (A) Workflow of the cell treatment and subsequent RNA extraction. Huh-7 cells were exposed to 2.5 μM FBA at 37°C for 24 h. The cells were then washed and lysed, whereafter total cellular RNA was extracted and purified using a spin column-based method. As a control, 2.5 μM FBA was added directly to the cell lysate and the same purification procedure was performed. (B, C) Representative emission spectra of purified RNA from Huh-7 cells exposed to 2.5 μM 2CNqATP (B) or 2CNqAMP (C). Increase of emission with exposure time is indicated by arrow with extraction after 2, 8, 12, 24 and 48 h (lightest to darkest color) of exposure, compared to control, prepared as shown in (A) (grey). Spectra are normalized to the corresponding absorption at 260 nm to adjust for differences in concentration of the samples. (D) Excitation spectra of the extracted RNA from cells exposed to 2CNqATP (yellow) or 2CNqAMP (red) for 24 h. For comparison, the absorption spectra of the 2CNqATP monomer (black dashed line), 2CNqA-ssRNA (dark grey solid line), and 2CNqA-dsRNA (light grey solid line) were included (see methods for details). The spectra were normalized to unity at the peak maxima, represented by vertical lines at 352 nm and 360 nm for 2CNqATP and 2CNqA-containing RNA, respectively. (E) Titration of unlabeled RNA (0 to 512 μM in 8 steps) to 10 μM of 2CNqATP to investigate if any non-specific binding events would affect the excitation spectrum of 2CNqATP. The excitation band maxima (352 nm) are indicated by the vertical line. (F) Agarose gel (2.0 mass-%) electrophoresis of total RNA extracts of Huh-7 cells exposed to 2.5 μM 2CNqATP over indicated time. In the control, 2CNqATP was added to cell lysate before extraction. 500 ng RNA of each sample was loaded and SYBRsafe was added to the gel for detection. Indicated bands are 28S rRNA (I), 18S rRNA (II), xylene cyanol dye present in the loading dye (grey III), and small RNAs (IV). Corresponding gel of 2CNqAMP samples in Supplementary Figure S11. (G) Representative emission spectra of purified RNA from cells co-exposed to 2.6 μM 2CNqATP and 15 nM hydroxyurea over 24 h (dotted dark yellow line). For comparison RNA from cells exposed to 2.5 μM 2CNqATP only over 24 h was isolated in the same experiment. (solid dark yellow line; compare A). A control, where 2CNqATP was added to lysate, was performed as well (solid light-yellow line; compare A). Corresponding isolation of RNA exposed to 2CNqAMP in Supplementary Figure S12.

The FBA emission (centered around 460 nm) in the total RNA extracts from cells treated with 2CNqATP or 2CNqAMP increased as function of exposure time (Figure 4B, C). The lower emission in the control rules out the possibility that 2CNqA fluorescence in the RNA extracts was a result of non-covalent interactions between 2CNqA and cellular RNA and supports that the FBA has been endogenously incorporated. Notably, the 2CNqA emission was lower in the RNA extracts of 2CNqAMP-treated cells compared to 2CNqATP-treated cells. This observation is surprising given the faster uptake of 2CNqAMP and suggests that both derivates are relevant as reagents for cellular RNA labeling.

The incorporation of 2CNqA into RNA was further supported by red-shifts (ca. 8 nm) in the excitation spectra of the cell-extracted RNA relative to the 2CNqATP monomer (Figure 4D). This shift is related to the slight sensitivity of 2CNqA to its microenvironment and agrees well with the position of the absorption spectra of single- and double-stranded RNAs containing 2CNqA (55). Furthermore, the titration of unlabeled long RNA to solutions of 2CNqATP (Figure 4E) or 2CNqAMP (Supplementary Figure S10) did not result in any excitation shift but the peaks remain centered at 352 nm. These observations strongly suggest that 2CNqA, following cellular uptake, is processed by cellular kinases for phosphorylation and thereafter incorporated into newly synthesized RNA by the cellular transcription machinery.

We tested whether the addition of 2CNqA to cells influenced the RNA size distribution using agarose gels (Figure 4F, Supplementary Figure S11). The resulting bands of samples collected after 2–48 h of exposure are identical to the control (untreated cells) and display the typical pattern of total RNA extracts from human cells: 28S rRNA at 5 kb, 18S rRNA at 1.9 kb with 28S rRNA giving about twice as intense signal, and smaller RNAs (e.g. 5S rRNA, tRNA, snRNA) below the 200 nt band (Figure 4F indications I, II, IV).

To test if the ribose version of the FBAs could be converted into the corresponding deoxyribose, cells were co-exposed to 2CNqATP or 2CNqAMP and hydroxyurea. Hydroxyurea inhibits ribonucleotide reductase, which converts ribonucleotide diphosphates into their 2′ deoxy derivatives (68). In a first experiment three different concentrations (15 nM, 50 μM, 10 mM) of hydroxyurea were tested and the effect of co-exposure was observed under the confocal microscope (Supplementary Figure S12B). In the next step RNA from cells that were exposed to 2CNqATP or 2CNqAMP and hydroxyurea for 24 h was extracted and spectroscopically analyzed. Over long time periods toxic effects were observed for the 50 μM and 10 mM exposures, hence the lowest concentration (15 nM) was chosen for long-term exposure (24 h). There were no differences in 2CNqA emission in RNA extracts from the hydroxyurea co-treated versus untreated cells (Figure 4G, Supplementary Figure S12A), nor did hydroxyurea induce any changes to the sub-cellular distribution of 2CNqA (Supplementary Figure S12B). This indicates that 2CNqA remains as a ribonucleotide inside cells and enters RNA-specific pathways. It should be noted, however, that hydroxyurea can have other effects on cells than just the inhibition of the ribonucleotide reductase (69) and a detailed screening would be needed to draw more conclusions on the effects of co-exposure of nucleotide analogues and hydroxyurea.

Lastly, to estimate the efficiency of the 2CNqA incorporation into cellular RNA we constructed a 2CNqATP emission standard curve (Supplementary Figure S13) as described in the Materials and Methods section to determine the 2CNqA concentration in the RNA extracts. This was then related to the total RNA concentration (which was determined by the absorption of total RNA bases at 260 nm). This estimate, which includes several assumptions and details complicating an exact assessment as described in the Materials and Methods section, suggests that 2CNqA was spontaneously incorporated in approximately one of every 10⁴–10⁵ bases after 24 h of cellular exposure to 2CNqATP. This number reflects the 2CNqA nucleotides incorporated into all kinds of newly synthesized RNA strands. Hence, the number is also affected by the fact that total RNA extracts also contain a considerable amount of long-lived RNA present in the cell prior to the exposure, a portion of which remains intact throughout the experiment.

In contrast to 2CNqATP or 2CNqAMP exposure, RNA extracts from cells treated with pATP or tC^OTP (Supplementary Figure S14) did not display enhanced FBA fluorescence, suggesting that these FBAs could not be incorporated into cellular RNA. The latter is highly expected given that tC^OTP is not internalized by the cells. For pATP, our data suggests that the subcellular localization, which appears to exclude the nuclear sites of cell-endogenous RNA synthesis, effectively prevents incorporation of pA into cellular RNA. This, again, indicates that subtle variations in nucleotide chemistries are decisive for the applicability of RNA analogues in metabolic labeling applications and that the 2CNqATP version presented here has specific applicability as a fluorescent RNA label.

Discussion

RNA processing and synthesis are highly dynamic molecular processes which underlie cellular regulation and function. This study explores new strategies for metabolic fluorescence labeling of cellular RNA using ribonucleotide derivatives of the fluorescent base analogues tC^O, 2CNqA, and pA and investigates the cell uptake kinetics and dynamics of these compounds. Furthermore, we report structure-dependent variations that may be useful for future design of nucleotide-based labels and therapeutics targeting the transcription machinery, such as small molecule RNA probes and nucleoside antivirals.

Adenine analogue structure variations influence their cellular uptake, localization, and retention

This study shows that the ribonucleotides of two fluorescent adenosine analogues, pA and 2CNqA, are spontaneously taken up into cells but that they display substantial and functionally important differences in uptake, subcellular localization, and retention (Figures 2 and 3). The difference in sub-cellular localization of 2CNqA and pA (Figure 2A) provides an example of how relatively minor changes in the design of a functional nucleoside probe or drug can profoundly alter its bioavailability at the target site and hence its efficacy.

The observation of poor cell retention of pA (Figure 3B) furthermore suggests that certain modifications may target nucleoside-based probes or drugs for active excretion. This could occur via the same mechanisms as for modified RNA bases, which are excluded from the salvage pathways and excreted (70), in the form of nucleosides or monophosphates (71) and where hydrophobicity-increasing adenine methylation is common (72). Notably, pA is a more hydrophobic modification than 2CNqA. Another possibility is that the high hydrophobicity of pA leads to increased non-specific interactions with other, lipophilic, cell constituents, diminishing pA’s ability to serve as an A-analogue in, for example, the nucleotide salvage pathway. In addition, it is possible that pA, due to the larger chemical modification compared to 2CNqA, may be sterically restricted from the active site of one or several of the kinases that re-phosphorylate nucleotides after cellular uptake, and that this would further facilitate excretion. We have previously observed that pA can be less well accepted as an adenine analogue compared to 2CNqA (57). All of these suggested effects would make pA less efficiently re-phosphorylated inside cells than 2CNqA and therefore more quickly excreted.

Our observation that even the minor molecular structure changes in the fluorescent nucleotides presented here result in significant differences in biological activity is interesting. The finding suggests that these derivatives, new extensions to their chemical design, and even novel FBAs could be useful model compounds to better understand, through image-based visualization, how nucleoside-based drugs with altered nucleobase moieties (73,74) should be designed to reach their intended intracellular targets.

The number of phosphate groups influences the cellular uptake mechanism

We report that 2CNqATP and 2CNqAMP uptake is concentration-dependent and saturable (Figure 3D) but proceed with significantly different kinetics (Figure 3F, Supplementary Figures S6 and S7). The ca. two times slower uptake of 2CNqATP is reasonable considering that the purinergic uptake pathway consists of a complex and fine-tuned series of dephosphorylations of ATP via membrane-bound ectonucleotidases (75), like CD39, which cleave off the first two ATP phosphate groups and that have been reported to be the rate-limiting enzyme in the dephosphorylation cascade (76).

Furthermore, the observation that 2CNqATP and 2CNqAMP uptake (Figure 3F), as well as pATP uptake (Supplementary Figure S9), can be competed with ATP and AMP, respectively. Also qualitatively, this supports uptake via the purinergic pathway. However, we also note that, with the multiple roles that ATP has in cell biology (77), an increased extracellular concentration of ATP may induce other responses in the cells. For example, under conditions of excess extracellular ATP, we observe a small difference in the intracellular localization of the 2CNqA derivatives, seemingly more like that of pA (Figure 2A). This change in cellular localization could indicate a potential use for 2CNqA and pA nucleotides as fluorescent co-exposure probes for studying the effects of other adenine-mimicking effectors. Altogether, our results point to that the two 2CNqA derivatives and pATP are accepted as analogues of native ATP by a range of enzymes in the uptake pathway.

RNA incorporation and possible applications of the metabolic fluorescence labeling approach

The metabolic fluorescence labeling of RNA (Figure 4) using non-engineered cells and unassisted uptake of nucleotides of 2CNqA (Figures 2 and 3) could have several direct applications, such as our image-based example where the effect of Act D on nucleolar size can be followed in live cells (Figure 2D). The preferential localisation of 2CNqA to nucleoli facilitates imaging and detection of these membrane-less compartments, which play a major role primarily in ribosome biogenesis, and are likely to be involved in cell cycle regulation, DNA damage repair, and pre-mRNA processing (78). Hence, to study the nucleoli is an important contribution for understanding cellular mechanisms of several diseases (79–81).

Nucleoli are conventionally visualized using immunostaining towards nucleoli-abundant proteins (e.g. fibrillarin (82) or nucleophosmin (63)) or by genetically engineering cells to express nucleolar proteins fused to fluorescent proteins. This approach is challenging when monitoring the effect of chemicals or drugs on the nucleolar RNA, rather than proteins, as for example investigated by Szaflarski et al. (63). Approaches to label RNA inside nucleoli in live-cell imaging have been made by designing, for example, fluorescent intercalators that have high affinity to ribosomal RNA (rRNA) (83,84) with one dye being commercially available (SYTO RNASelect, Thermo Fisher Scientific). However, its wider applicability in different cell cultures and various microscopy techniques remains unverified.

In this work, we show that 2CNqA enables straightforward staining of nucleoli and allows direct tracking of morphological changes to these compartments in live cells, shown here in response to Act D (Figure 2D). Act D binds to DNA in the transcription initiation complex and prevents the elongation of RNA (85) leading to reduced nucleolar RNA stability. This prevents liquid-liquid phase separation which is the basis for nucleoli formation and morphology (62). The changes in nucleoli size were visualized at non-toxic concentrations, making 2CNqA a promising tool in the development of drugs targeting rRNA and rRNA synthesis inhibitors. It also suggests possibilities to track RNA involvement in the formation of other membrane-less cell compartments such as stress granules.

It is well known that different RNA species have different turnover rates and lifetimes in cells, depending on their function and microenvironment, as well as cell type and growth conditions (86,87). The potential for monitoring such kinetics using our metabolic labels is exemplified in this work by the time-dependent increase in 2CNqA emission in extracted RNA following exposure to the 2CNqA nucleotides (2–48 h; Figure 4B, C), which we propose reflects the in-cell synthesis of total RNA. Considering that rRNA accounts for as much as 80% of the total cellular RNA (88), and that 2CNqA localizes to the nucleoli – the center of rRNA transcription, formation, and maturation (89), it is likely that a significant amount of the 2CNqA ends up in nascent rRNA. The presence of the 28S and 18S subunits of human rRNA in our samples were confirmed by gel electrophoresis (Figure 4F). Although we could not, at this time, verify the presence of 2CNqA in the agarose gel bands due to low gel scanner sensitivity, the gel electrophoresis analysis using SYBR staining shows similar band intensities for the 28S and 18S rRNA irrespective of exposure time. This indicates that the presence of 2CNqA in the cells does not adversely affect the synthesis of certain RNA types significantly, which is important for the reporting capabilities of the label.

Although the RNA extractions performed in this work make no effort to capture and analyze specific RNA species in the cells, combining our metabolic labelling approach with RNA separation and analysis protocols such as fluorescence-detected capillary electrophoresis (90), liquid chromatography (91), or hybridization capture technologies (e.g. beads or microarrays) (92), could provide access to more specific RNA kinetics. The transcription dynamics of rRNA, for example, which is critical for a variety of cell biology processes, could be one attractive target for this development (93). A distinct advantage of the direct expose–extract–measure method suggested here lies in its simplicity, as the protocol can be executed in standard biochemistry labs, using common spectrofluorometers and plate readers. This is in contrast to existing gene expression profiling methods that rely on RNA sequencing, which indeed can provide detailed information, but depend on laborious workup and analysis protocols (94).

Number of phosphates affects incorporation into cell-endogenous RNA

An interesting observation when comparing the details of the RNA extraction from the cells exposed to 2CNqATP (Figure 4B), to the ones exposed to 2CNqAMP (Figure 4C), is the lower emission that results from the use of the monophosphate derivative, which we interpret as a lower degree of incorporation of 2CNqA. Given the faster uptake kinetics for 2CNqAMP compared to 2CNqATP (τ = 35 min versus τ = 74 min, respectively, vide supra) one could expect the opposite, especially if both compounds are converted to the 2CNqA nucleoside during the internalization and thereafter follow the same pathway of intracellular re-phosphorylation and nuclear transport. This suggests that a more complex process occurs, which we at present cannot fully explain. It is, however, known that purinergic signaling can affect RNA synthesis (95), and considering that our adenine analogues herein are accepted by various cellular enzymes, we speculate that the RNA production and incorporation is not solely governed by the concentration of 2CNqA nucleotide inside the cell, but may also be influenced by the triphosphate version of 2CNqA having an indirect stimulating effect on the cell-endogenous labeling. A thorough understanding of these observations warrants further investigations of both derivatives. It has not, however, escaped our notice that co-exposures with ATP could be an interesting avenue to explore for controlling RNA incorporation of these analogues.

Conclusions

We show that it is possible to achieve unassisted cellular delivery of fluorescent nucleotide analogues to live human cells and, by using these as metabolic labels, both better understand structure-activity relations relevant for the design of nucleoside probes and drugs, and explore the dynamics of cellular process and structures (nucleoli) related to RNA synthesis. Our study shows that efficient uptake can be achieved with nature-mimicking chemical modifications to both mono- and triphosphate variants of adenine, but not with variants of cytosine suggesting that the choice of base type may be a critical factor to facilitate bioavailability.

Furthermore, we show that fluorescence readouts are highly useful to understand the relationship between critical factors for successful metabolic RNA labelling such as efficient uptake, nuclear accumulation, and cellular retention, and the exact structure of a nucleotide. Such knowledge is undoubtedly useful for future design of RNA metabolic drugs to improve not only their target interaction but also bioavailability inside of cells. One such line of improvement for 2CNqA and related analogues could be to explore prodrug-inspired designs to even further increase cellular availability. However, already at the fairly low incorporation degree of 2CNqA that we achieved here (one in every 10⁴–10⁵ bases) it was possible to monitor cellular production of nascent total RNA, as well as to image nucleoli structures inside of live cells.

2CNqA is a spectroscopically versatile and bright probe for this class of nature-mimicking labels and could, in addition, be suitable for fluorescence correlation spectroscopy (FCS) (58) and fluorescence recovery after photobleaching (FRAP) detection of RNA dynamics in, for example, nuclear and cytosolic protein condensates. Furthermore, the significant difference in lifetimes between free and nucleic acid incorporated 2CNqATP (55,58), could be used for live-cell fluorescence lifetime imaging (FLIM) applications which may enable imaging-based monitoring of transcriptional activity at various localizations inside of cells.

Metabolic fluorescence labeling of RNA in the way reported here, i.e. relying on spontaneous uptake, using non-engineered cells and a bio-orthogonal chemistry-free approach, has to the best of our knowledge not previously been achieved. Together with the recent report by Wang et al., in which human cells were engineered to metabolically label RNA with fluorescent cytosine analogues (48), our findings highlight the general potential of derivatives of fluorescent base analogues to be used for in cellulo labeling of nucleic acids. Moreover, the present study also suggests that fluorescent base analogues can further the understanding of how modified nucleotides function in a cellular context, for example by aiding structure-activity relationship type studies that will eventually facilitate the design of novel nucleoside-based drugs (73,74), including antivirals.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Supplementary data

Supplementary Data are available at NAR online.

Funding

Area of Advance Nano at Chalmers University of Technology to P.P. and L.M.W. and conducted as part of the FoRmulaEx research center for nucleotide delivery and with associated financial support to E.K.E. and L.M.W. from the Swedish Foundation for Strategic Research (SSF) [IRC15-0065]; Swedish Foundation for Strategic Research (SSF) [ID14-0036 to M.G.]; Swedish Research Council [VR, grant No. 2021-04409 to L.M.W.]. Funding for open access charge: Swedish Universities Fund (Bibsam).

Conflict of interest statement. L.M.W., J.R.N., E.K.E., T.B., A.G. and P.P. own shares in Stealth Labels Biotech AB (LanteRNA); J.R.N. is also an employee there. A.G., M.B., M.L., A.D. and T.B. may own shares and are employees at AstraZeneca.