Our previous experiments with mCherry as a reporter gene (Lale et al., 1101) led us to suspect the presence of a shorter protein isoform of mCherry. We first analyzed the sequence of the codon-optimized version of mCherry and noticed three methionine residues in a relatively close proximity from the annotated start codon (Figure 1). In addition, we recognized an SD-like sequence ranging from −12 to −6 nucleotides upstream of the first internal methionine residue (Figure 1). This led us to hypothesize that a short functional isoform of mCherry is produced from one of the three downstream methionines and not from an alternative start codon.

To determine which of these methionine residues function as an ATIS that still renders a functional mCherry protein, we designed three versions of mCherry (V1, V2, and V3), with increasing N-terminal truncations (Figure 1). Each version of mCherry was expressed in E. coli with a constitutive promoter/5′-UTR. The fluorescence measurements of the different mCherry versions are presented in Figure 2. The smallest truncation (V1) retains 73% of the fluorescence intensity of the original codon-optimized version (mCherry-CO); whereas, the other versions (V2 and V3) do not show any fluorescence. In addition, proteomics analysis of the red fluorescent protein mCherry V1 confirmed that the first 10 amino acids were absent (Supplementary Figure S3), although the fragment between M10 and M17 could not be detected due to the short length of the digested fragments. This is conclusive evidence that the mCherry gene produces a short functional protein isoform starting at the methionine in position 10 of the amino acid sequence.

Because most red fluorescent proteins originate from the modifications of DsRed (Matz et al., 1999), we imagined that the dual-isoform issue of mCherry could affect other members of the red fluorescent protein family. We performed protein sequence alignments to determine the extent of the issue across DsRed-derived fluorescent proteins (Figure 1). We were able to trace back the apparition of the second isoform to the engineering of mRFP1.3 (Figure 3).

DsRed was modified into mRFP1 to overcome obligate tetramerization, improve protein maturation, and modify the excitation/emission wavelength couple (Campbell et al., 2002). Then, mRFP1 was further modified into mRFP1.1 by the notable Q66M substitution in the chromophore to improve parameters such as photostability, quantum yield, and extinction coefficient (Shaner et al., 2004). During the same round of modifications, in an attempt to improve the poor N-terminal fusion properties of mRFP1.1, the initial residues of eGFP (MVSKGEE) followed by the linker (NNMA) were added to mRFP1.1 to yield mRFP1.3¹². This manipulation is responsible for the apparition of the short mCherry isoform. Indeed, the linker (NNMA) provides the alternative methionine start codon, and the eGFP fragment offers an SD-like sequence for ribosome entry (Figure 1).

The phylogenic overview of red fluorescent proteins (Figure 3) shows all the proteins engineered from mRFP1.3 that are affected by the dual-isoform issue. In addition to the mRFP1.3-derived proteins, mBanana, mGrape2, and mGrape3 also received the N-terminal extension described above. The problem becomes even more concerning because it affects all commonly used variants in the red color spectrum. Furthermore, the timeline of engineering is important. As mRFP1.3 and mCherry were created in 2004, this entails that the results from a large number of publications using red fluorescent proteins since 2004 may have been affected by this issue.

In prokaryotes, translation initiation shares quasiuniversal principles (Nakagawa et al., 2010; Rodnina, 2018), with some variations in the SD sequences recruiting ribosomes. Because the short mCherry isoform is encoded in the mCherry gene sequence, it is highly likely that other prokaryotes also produce the short mCherry isoform. In addition, depending on the codon usage of the N-terminal extension, the short mCherry expression may be stronger or adapted to other organisms (as demonstrated by mCherry with SD-deoptimized) (Figure 1). We found that the short mCherry isoform (V1) was functional in Vibrio natriegens and Pseudomonas putida (Supplementary Figures S5 and S6), and another group had found the mCherry ATIS occurring in Mycoplasma (Carroll et al., 2014). This demonstrates that the mCherry defect occurs across a wide range of bacteria.

Translation initiation significantly differs between prokaryotes and eukaryotes. Bacteria mainly rely on the annealing of 16 S ribosomal RNA with the SD sequence and mRNA structures surrounding the start codon, whereas eukaryotes lack this type of interaction with the mRNA transcript. Instead, the small ribosomal subunit (40 S), charged with tRNA_Met,i, recognizes an AUG start codon through a scanning mechanism of codons within the 5′UTR of mRNA strands (Sonenberg and Hinnebusch, 2009; Hinnebusch, 2014). The nucleotide context around the start codon also plays a role in initiating translation, favoring specific nucleotide motifs, sequence named after Kozak who first identified it (Kozak, 1986). The consensus yeast Kozak sequence was determined as (A/U)A (A/CA (A/C)A AUG UC(U/C) (Hamilton et al., 1986); however, this sequence varies a lot both within and across organisms (Nakagawa et al., 2008; Cuperus et al., 2017). To assess if the short mCherry isoform is also produced in eukaryotes, we replaced the AUG start codon with the UAG stop codon from a S. cerevisiae codon-optimized mCherry. In addition, we modified the codon usage surrounding M10 to GAAGAAGACAAC AUG GCC to resemble the optimal yeast Kozak sequence described by de Boer et al. (2020) and Hamilton et al. (1986). We demonstrated that this mCherry sequence provides in average a 19% background fluorescence compared with the original sequence (Figure 4; Supplementary Figure S7). It conclusively proves that the eukaryotic scanning mechanism of translation initiation allows the production of the short mCherry isoform. The nucleotide context surrounding the start codon may influence the level of translation in different eukaryotic cells. However, the ability of yeast to translate the short mCherry isoform raises concerns regarding studies using mCherry as a fluorescent marker across eukaryote species.

To provide a version of mCherry that is usable for gene expression and protein localization without conferring background expression, we tested different solutions in the context of fusion protein. The first solutions consisted in a substitution of the problem-causing M10 to glutamine or leucine, which aims to preserve the structural properties of the protein (Figure 5). As M10 stabilizes mCherry by interacting with the tyrosine residue Y43, the substitutions of M10 with glutamine or leucine allow similar stabilization through the formation of an H-bond with Y43 or with Van der Walls interactions, respectively, while occupying an equivalent steric space as methionine. The second solution was simply to use mCherry V1 as a reporter as it resembles mRFP1.1.

Fusion proteins composed of sfGFP fused with C-terminal engineered versions of mCherry were constructed to assess the performance of each solution. First, sfGFP and mCherry versions were linked with an alanine/glycine linker to estimate the C-terminal fusion properties of the engineered mCherry versions (Figure 6). Second, the background expression due to the short isoform was assessed by placing two stop codons in the linker peptide right upstream the mCherry versions.

In both cases, GFP is used as an internal standard of gene expression to compare the mCherry production as fusion or discrete protein between samples. The fusion proteins composed of the mutated amino acids in position 10 or the V1 isoform display an mCherry/GFP ratio of 1.3, which constitutes a reference under the assumption of an equimolar production of mCherry and GFP because they are expressed as a fusion protein (Figure 6). The inclusion of stop codons in the protein linker confirms this assumption, as none of these constructs displays mCherry fluorescence when the full-length fusion protein is not produced.

In the context of fusion protein, the original mCherry-(CO) shows a nearly twofold increase in mCherry abundance that can be attributed to the short isoform; however, this was not clearly detected for the mCherry-(DeOpt) version that should present the same characteristics (Figure 6). This may suggest a higher proportion of V1 isoform and truncated translation of sfGFP-mCherry fusions because ribosomal entry at the internal SD sequence of mCherry-(DeOpt) may cause upstream-translating ribosomes to fall off at the protein junction with mCherry. The two M10-mutated versions of mCherry present the same expression intensities as the original protein sequences in the fusion context; whereas the fluorescence of the short mCherry V1 isoform is slightly lower but still reasonable (Figure 6). This demonstrates that all solutions provided to resolve the mCherry issue are functional in a fusion protein context.

For the fusion proteins containing stop codons, the original mCherry-(CO) sequence and the “SD-deoptimized” version present a background fluorescence due to the production of the short isoform, which is amplified in the case of “SD-deoptimized” due to higher ribosomal entry on the 5′end of the mCherry sequence as its internal SD sequence favors V1 isoform translation. On the contrary, the M10 substitutions and V1 isoform do not show any fluorescence, which proves that these solutions successfully abolish the production of the short mCherry isoform in the context of C-terminal fusion (Figure 6). Likewise, the short mCherry V1 did not show any background fluorescence but appeared to be less efficient in C-terminal fusions, as reported for mRFP1.1 (Shaner et al., 2004).

Subtle changes in the amino acid sequence of fluorescent proteins can cause changes in their spectral properties as well as their brightness and maturation (Arpino et al., 2014). The removal of the N-terminal sequence before M10, or the substitutions of M10 by leucine and glutamine, seems to affect the brightness and/or maturation of the protein and potentially its ability to be used as soluble fusion partner. However, we detected no changes in the excitation/emission spectra for the different variants (Supplementary figure S8). Further investigation and engineering are necessary to recover the properties of mCherry without generating the short protein isoform that creates background fluorescence.

Experiments were performed in E. coli DH5-α (NEB), grown in LB-Lennox (Oxoid) (10-g/L casein peptone, 5-g/L yeast extract, 5-g/L NaCl with additional 15-g/L agar for plates) supplemented with 100-µL/mL ampicillin (Sigma-Aldrich). PCR amplifications were performed using Q5 polymerase (NEB). All other necessary enzymes were also purchased from NEB. Primers were ordered from Eurofins Genomics – Sigma-Aldrich. Plasmids and PCR products were purified using the QIAprep Plasmid Miniprep Kit and QiaQuick PCR Purification Kit, respectively (Qiagen). Plasmid Sanger sequencing was performed by Eurofins Genomics. The E. coli codon-optimized sequence of mCherry was a gift from Yanina R. Sevastsyanovich (University of Birmingham).

A pUC19 backbone containing a constitutive promoter/5′-UTR expressing sfGFP, generated in a previous study (Lale et al., 1101), was used as template for cloning the different versions of mCherry. The backbone and the different versions of mCherry were amplified by PCR using the respective primers presented in Supplementary Table S1. To test for short isoforms, sfGFP was replaced by the codon-optimized mCherry gene or its shorter versions (V1, V2, and V3) via Golden Gate assembly (Engler et al., 2008). To build fusion constructs, the different mCherry versions were fused downstream sfGFP using the Golden Gate assembly.

The Golden Gate assembly mixture was chemically transformed into competent E. coli by heat shock (45 s at 42°C). Cells were placed on LB plates containing 100-µg/mL ampicillin and grown overnight at 37°C. Positive clones were grown in 5-mL LB supplemented with 100-µg/mL ampicillin, and their respective plasmids were purified. DNA sequences were confirmed by Sanger sequencing.

For S. cerevisiae, we used the yGPRA plasmid kindly provided by Regev et al. (de Boer et al., 2020). In brief, it contains a yeast codon-optimized mCherry gene expressed by the TEF1 promoter. The AUG start codon was replaced by an UAG stop codon, and the codon usage surrounding M10 was modified to GAAGAAGACAAC AUG GCC to resemble yeast Kozak sequence via PCR using primers presented in Supplementary Table S1. The PCR products were subject to overnight DpnI digestion at 37°C, subsequently purified, assembled using the Gibson assembly, and transformed into E. coli Dh5α. Correct assembly was confirmed by restriction digest with the BbsI and speI enzymes of the plasmid as well as sequencing (Eurofins Genomics). The original yGPRA plasmid and the modified yGPRA_mCherry_V1 plasmid were transformed into the uracil-deficient S. cerevisiae strain using the previously described electroporation method (Benatuil et al., 2010). The transformants were selected onto SD–UT media (6.7-g/L yeast nitrogen base without amino acids, 1.6-g/L yeast synthetic dropout medium supplement without uracil, and 20-g/L agar).

Each E. coli strain bearing a given plasmid was inoculated into LB supplemented with 100-µg/mL ampicillin on 96-well plates and incubated overnight at 37°C with 800-rpm agitation in a Multitron Pro plate-shaking incubator (Infors HT). For yeast, positive clones for both yGPRA and yGPRA_mCherry_V1 plasmids were cultivated in triplicates in 5-mL SD–UT media. The uracil-deficient S. cerevisiae strain was cultivated in triplicates in 5-mL YPD media, collected via centrifugation at 5,000 rpm, and then washed with SD–UT media before fluorescence measurements. Each sample (100 µL) was used for fluorescence measurements. Fluorescence was measured using an Infinite M200 Pro TECAN fluorimeter (Noax Lab AS). The excitation/emission wavelengths were 488/525 nm for GFP (gain 67) and 576/610 nm for mCherry (gain 97 for E. coli and 163 for yeast). Fluorescence intensity was normalized by OD₆₀₀ of the corresponding well.

Cell cultures of E. coli and S. cerevisiae carrying the respective plasmids expressing mCherry and the V1 isoform were placed between glass slides for the microscopic analysis. An inverted fluorescence microscope (Zeiss Axio Observer. Z1, 2.3.64.0) with a ×20 air objective (NA 0.8) was used to detect mCherry and was localized within the cells using the bright field filter. Image processing was performed using the Zeiss image analysis software (2.3.64.0).

The amino acid sequence of mCherry-(CO) was used as template to perform preliminary protein BLAST searches (blast.ncbi.nlm.nih.gov/Blast.cgi). Phylogeny of red fluorescent proteins was consulted on FPbase [27] (fpbase.org).

The strain carrying the constitutively expressed mCherry V1 was grown overnight in 50 mL of LB media supplemented with 100-µg/mL ampicillin at 37°C, 225 rpm. Cells were harvested by centrifugation (5,000 rpm, 5 min) and resuspended in lysis buffer (50 mM Tris-HCl, 50 mM NaCl, 0.05% Triton X-100, pH 8.0) supplemented with 1 tablet EDTA-free cOmplete ULTRA protease inhibitor (Roche). Cell debris were eliminated via centrifugation (7,500 rpm, 15 min), and the soluble fraction was collected.

Two samples of 200-µL cell lysate each were separated for LC-MS analysis, one to be digested by trypsin alone and the other by both trypsin and Lys-C. Soluble proteins were precipitated by methanol/chloroform/water precipitation. In brief, 800-µL methanol was added to the sample, followed by the addition of 200-µL chloroform and vortexing. After the addition of 600-µL ultrapure water (18 MΩ), samples were thoroughly mixed by vortexing and centrifuged at 16000 g for 2 min. The upper layer was discarded without disturbing the protein layer, and further 800-µL methanol was added, followed by vortexing and centrifugation as above. After removing the supernatant, the protein pellet was air-dried for 10 min and was then reconstituted in 150 µL of 50-mM ammonium bicarbonate (BioUltra; Sigma-Aldrich, Germany). Next, the proteins were reduced with 1.5 µL of DTT (Sigma-Aldrich, Canada) for 20 min at 70°C, brought back to room temperature, and then alkylated using 6 µm of iodoacetamide (BioUltra; Sigma-Aldrich, United States) in the dark at room temperature for 30 min. Excess iodoacetamide was quenched by adding 3.5 µm of DTT (Sigma-Aldrich, Canada) and incubating in the dark for 20 min at room temperature. At last, the proteins were digested with endoproteinase at 37°C overnight, one sample digested with 1.25-µg trypsin and the other by 1.25-µg trypsin and 1.25-µg Lys-C. After overnight digestion, 5 µL of formic acid was added to each sample, and then the peptides were dried in a vacuum concentrator at 60°C.

The samples were resuspended in 100-µL 0.1% formic acid and desalted in C18 stage tip columns, unless otherwise specified. The chemicals were of Optima grade obtained from Fisher Chemicals, and centrifugations were performed at 2000 g. In brief, stage tip columns consisting of three C18 plugs (Empore C18 47 mm SPE Disks, 3 M, United States) were made and activated using 50-µL methanol via centrifugation for 2 min; then, methanol activation was repeated. Then, the stage tip column was equilibrated with 60 µL of 0.1% formic acid in water via centrifugation for 2 min; then, the equilibration was repeated. The peptide samples were centrifuged (16000 g for 25 min), supernatants were loaded onto stage tip columns and centrifuged for 4 min, and flow-through solutions were reloaded to stage tip columns. The stage tip columns were washed with 60 µL 0.1% formic acid and centrifuged for 2 min; the wash was then repeated. The peptides were then eluted from the stage tip column using 40 µL of 70% acetonitrile, washed with 0.1% formic acid, and then centrifuged for 2 min; the elution step was then repeated. At last, desalted peptides were dried in a vacuum concentrator at 60°C and stored at −20°C until LC-MS analysis.

Dried peptides were reconstituted in 50 µL of 0.1% formic acid in water and shaken at 6°C at 900 rpm for 1.5 h. The samples were centrifuged at 16000 g for 10 min, and 40-µL supernatants were transferred to MS vials for LC-MS analysis. LC-MS analysis was performed on an EASY-nLC 1200 UPLC system (Thermo Fisher Scientific) interfaced with a Q Exactive mass spectrometer (Thermo Fisher Scientific) via a Nanospray Flex ion source (Thermo Fisher Scientific). Peptides were injected onto an Acclaim PepMap100 C18 trap column (75 µm i. d., 2-cm long, 3 µm, 100 Å, Thermo Fisher Scientific) and further separated on an Acclaim PepMap100 C18 analytical column (75 µm i. d., 50-cm long, 2 µm, 100 Å, Thermo Fisher Scientific) using a 180-min multistep gradient (150 min 2%–40% B, 15 min 40%–100% B, 15 min at 100% B, where B is 0.1% formic acid and 80% CH₃CN and A is 0.1% formic acid) at 250-nL/min flow. The peptides were analyzed in the positive ion mode under data-dependent acquisition using the following parameters: electrospray voltage, 1.9 kV; HCD fragmentation with normalized collision energy, 28. Each MS scan (200–2000 m/z, 2-m/z isolation width, profile) was acquired at a resolution of 70,000 FWHM in the Orbitrap analyzer, followed by MS/MS scans at a resolution of 17,500 (2 m/z isolation width, profile) triggered for the 12 most intense ions, with a 30-s dynamic exclusion, and analysis using the Orbitrap analyzer. Charge exclusion was set to unassigned, 1, and >4.

The proteins were identified by processing the LC-MS data using Thermo Fisher Scientific Proteome Discoverer (Thermo Fisher Scientific) version 2.5. The following search parameters were used: enzyme specified as trypsin with maximum two missed cleavages allowed; acetylation of protein N-terminus with methionine loss, oxidation of methionine, and deamidation of asparagine/glutamine were considered as dynamic and carbamidomethylation of cysteine as static posttranslational modifications; precursor mass tolerance of 10 parts per million with a fragment mass tolerance of 0.02 Da. Sequest HT node was used to query the raw files against sequences for mCherry (original and short); E. coli (strain K-12) proteins were downloaded from UniProt (www.uniprot.org/proteomes/UP000000625) in September 2020 and a common LC-MS contaminants database. For downstream analysis of this peptide-spectrum matches (PSMs), for protein and peptide identifications the PSM FDR was set to 1% and as high and 5% as medium confidence; thus, only unique peptides with these confidence thresholds were used for the final protein group identification and for the labeling of the confidence level, respectively.

The 3D structure modeling of mCherry was performed using the software PyMOL and the PDB file 2H5Q. In all mCherry PDB files (2H5Q, 6YLM, 6IR1 and 2, 6MZ3, 4ZIN), the first eight amino acids were unmodeled, residues 9 and 10 were computationally added, but M10 was mismodeled. We corrected residue 10 to model methionine. Then, we changed M10 into glutamine and leucine on the 3D structure with PyMOL. Rotamers with 2–3 Å proximity with Y43 are shown to model their interactions.