Stealth Fluorescence Labeling for Live Microscopy Imaging of mRNA Delivery

Methods for tracking RNA inside living cells without perturbing their natural interactions and functions are critical within biology and, in particular, to facilitate studies of therapeutic RNA delivery. We present a stealth labeling approach that can efficiently, and with high fidelity, generate RNA transcripts, through enzymatic incorporation of the triphosphate of tCO, a fluorescent tricyclic cytosine analogue. We demonstrate this by incorporation of tCO in up to 100% of the natural cytosine positions of a 1.2 kb mRNA encoding for the histone H2B fused to GFP (H2B:GFP). Spectroscopic characterization of this mRNA shows that the incorporation rate of tCO is similar to cytosine, which allows for efficient labeling and controlled tuning of labeling ratios for different applications. Using live cell confocal microscopy and flow cytometry, we show that the tCO-labeled mRNA is efficiently translated into H2B:GFP inside human cells. Hence, we not only develop the use of fluorescent base analogue labeling of nucleic acids in live-cell microscopy but also, importantly, show that the resulting transcript is translated into the correct protein. Moreover, the spectral properties of our transcripts and their translation product allow for their straightforward, simultaneous visualization in live cells. Finally, we find that chemically transfected tCO-labeled RNA, unlike a state-of-the-art fluorescently labeled RNA, gives rise to expression of a similar amount of protein as its natural counterpart, hence representing a methodology for studying natural, unperturbed processing of mRNA used in RNA therapeutics and in vaccines, like the ones developed against SARS-CoV-2.

Subsequently, in the same syringe, DMAP (0.028 g, 0.23 mmol), DIC (719 µl, 4.64 mmol), triethylamine (49 µl, 0.35 mmol) and 2,3,4,5,6-pentachlorophenol (0.309 g, 1.16 mmol) were added to the support and suspended in pyridine (4 mL). The mixture was gently shaken for 4 h at RT before a solution of piperidine (2 mL, 20% in DMF -for capping of the unreacted carboxylic acids on the support) was added for 1 min (longer exposure time will reduce loading as piperidine cleaves the ester bonds with the nucleoside), then quickly washed away with DMF (3x5 mL), dichloromethane (5 mL) and diethyl ether (5 mL). Finally, the resin was shaken in a CAP A + CAP B mix (50/50 v/v) for 2 hours under argon atmosphere, then washed with DMF (5 mL), dichloromethane (5 mL), diethyl ether (5 mL) and argon-dried (final loading: 13 μmol/g -determined by reading optical density of a DMT solution cleaved from a weighed amount of support -=70000 M -1 .cm -1 at 498 nm). Final loading can be increased by performing a second coupling with 1 in the same conditions before capping (typical loading after second coupling 15-20 μmol/g).
Concentrating the reaction mixture and washing the residue multiple times with water and diethyl ether allows recovery of nearly 85 % of unreacted nucleoside 1.
The least-square re-convolution fitting procedure was carried out using the DecayFit software (http://www.fluortools.com/software/decayfit). All decays were fitted to a tri-exponential (n = 3) model.
The presented lifetimes are amplitude-weighted average lifetimes ( ), calculated using the preexponential factors ( ) and lifetimes ( ) according to equation S4. The fitting parameters for the decays are shown in Supplementary Using the Beer-Lambert law, absorption is related to nucleobase concentration according to equations S8 and S9. The following molar absorptivities (unit: M -1 cm -1 ) were adopted: = 12200, = 7400, Assuming that the product RNA is uniform in size (1247 nucleotides), its base composition (A: 408, U: 272, G: 307, C: 260) allows for equations S10 -S13. (S10) Solving the equation system composed of S7 through S13 allows for quantification of / , , (S14) = Using the volume of the cell-free reaction (50 µL) and resulting product solution (100 µL), equation S15 was applied to calculate the incorporation yield ( ). The sequencing (Illumina) of the purified plasmids was carried out using the Eurofins sequencing service with the sequencing primers M13 uni (-43) and M13 rev (-29) for the PCR amplification of the plasmids.
Sequence alignment was performed with Benchling [Biology Software] (2018) using MAFFT Algorithm (https://benchling.com). Motif logos highlighted in the manuscript were made from these alignments using Weblogo 2.0. 11 The sequence alignments obtained (only shown for the products obtained with primer M13 uni (-43)) are shown in Supplementary Fig. 4 and 5.

Generation of H2B:GFP DNA template
The original coding sequence for H2B:GFP was taken from pCS2-H2B:GFP plasmid (Addgene, Plasmid #53744, Supplementary Fig. 6), manually codon-optimized to minimize the occurrence of poly-C n stretches (n<3), in silico-assembled with an additional T7 promoter and other desired features (Shine-S19 Dalgarno/Kozak consensus sequences for enhancement of translation and a 3xStop, respectively at the 5′ and 3′ of the coding sequence itself, plus the needed HindIII/SnaBI restriction sites, to generate the ligation-prone sticky ends) and ordered from Twist Bioscience as a synthetic gene block ( Supplementary   Fig. 6). The obtained sequence was then PCR-amplified, using a Phusion Hot Start High-Fidelity Taq hour) and a Cap 0 analog (using a Vaccinia capping system, NEB protocol #M2080), following the recommended procedures. This was necessary to increase, respectively, affinity for the cellular translation machinery and resistance to endogenous nucleases.