moxMaple3: a Photoswitchable Fluorescent Protein for PALM and Protein Highlighting in Oxidizing Cellular Environments

The ability of fluorescent proteins (FPs) to fold robustly is fundamental to the autocatalytic formation of the chromophore. While the importance of the tertiary protein structure is well appreciated, the impact of individual amino acid mutations for FPs is often not intuitive and requires direct testing. In this study, we describe the engineering of a monomeric photoswitchable FP, moxMaple3, for use in oxidizing cellular environments, especially the eukaryotic secretory pathway. Surprisingly, a point mutation to replace a cysteine substantially improved the yield of correctly folded FP capable of chromophore formation, regardless of cellular environment. The improved folding of moxMaple3 increases the fraction of visibly tagged fusion proteins, as well as FP performance in PALM super-resolution microscopy, and thus makes moxMaple3 a robust monomeric FP choice for PALM and optical highlighting applications.

protein and no cysteines. However, the engineering and evolution of mMaple3 incorporated several additional undesirable mutations including two cysteines and a N-linked glycosylation consensus sequence (Fig. 1). We sought to modify mMaple3 for use in oxidizing cellular environments.

Results
Evolution of mMaple3. Our engineering efforts began by modifying the coding sequence of mMaple3 using codons selected for high expression in mammalian cells -a strategy that was first used successfully in the development of EGFP 17 . Following synthesis of codon optimized mMaple3, we began to mutate selected amino acids.
The limiting step in mutagenesis of FPs for oxidizing environments is the replacement of cysteines with alternative amino acids. It was not obvious a priori that the cysteines of mMaple3 could be mutated without disrupting the desirable properties of mMaple3. Similar efforts for EGFP family members often resulted in dark proteins and our previous efforts with the development of moxEos3.2 and moxDendra2 9 had revealed that the cysteine positions were highly sensitive even to seemingly conservative small amino acid substitutions. Serine was never tolerated and an alanine or valine could produce dramatic differences in apparent brightness 3,5,18 . As other related photoswitchable FPs mEos2, Dendra2, and mKikGR have cysteines at the identical positions 19 , we leveraged our previous successful efforts to engineer moxDendra2 whose cysteines align with those in mMaple3 20 .
Initial screening of mutants was carried out by expressing the variants in bacteria and scoring bacterial streaks for whether they remained fluorescent. Using this standard approach, we were surprised by the results of the first cysteine mutants. Based on results with moxDendra2, we anticipated that Cys110 would maintain fluorescence if switched to an Ala or Val. While this hypothesis proved correct, we found that a Val mutation greatly increased the brightness of the bacteria ( Fig. 2 and Table 1). A similar mutation of Cys180 did not increase fluorescence and a Val mutation at that position even decreased fluorescence, whereas an Ala mutation was tolerated with no impact on fluorescence. The cysteine mutations were combined and then glycosylation disrupting N to Q or D mutations were also made without any obvious impact on fluorescence ( Fig. 2 and Table 1). Together, the mutations produced a sugarless, cysteine-free FP we term moxMaple3 (Fig. 2). Bacteria expressing the resulting protein were fluorescent and exhibited the increased brightness observed with the single C110V mutation.
Upon successfully adapting mMaple3 with mutations, we characterized the resulting FP and sought to determine how the mutations impacted bacterial brightness. We considered four possibilities: i) decreased toxicity of the FP resulting in improved cell growth, ii) improved translation and consequently higher protein levels, potentially through improved mRNA structure, iii) increased FP brightness, and/or iv) improved folding resulting in higher levels of correctly folded and thus chromophore-forming protein. We began by asking whether mutations at the Cys110 position improved bacterial growth, potentially as a consequence of decreased FP toxicity. In Fig. 3A, we observed no difference in growth rates for bacteria expressing the parent or mox variants. Therefore, we discarded the hypothesis that the FP might impact cell growth or viability.
Next, we asked whether individual bacteria expressing moxMaple3 were inherently brighter than bacteria expressing mMaple3. moxMaple3 expressing cells were clearly brighter (Fig. 3B,C). This result suggested that more FP might be expressed or the properties of the new FP increased bacterial brightness. To compare absolute protein levels, Coomassie-stained polyacrylamide gels of bacterial lysates were generated for bacterial strains induced and expressing the parental and mox variants. A small, but significant increase of ~25% in band intensity for the moxMaple3 variant argued that a part of the increased bacterial brightness was, indeed, due to enhanced protein expression (Fig. 3D). A similar observation was made by Duwe' et al. when they evolved EGFP to the reversibly photoswitchable rsEGFP and rsEGFP2 and observed increased FP expression in bacteria 21 .
As we could map the enhanced expression to a single amino acid, we considered two potential explanations for increased protein levels-either the changes in the mRNA sequence might enhance translation or the cysteine mutation, itself, improved protein stability. While mRNA structures are notoriously difficult to predict using algorithms, nonetheless, we found that nucleotide changes for moxMaple3 could dramatically alter the predicted secondary structure relative to mMaple3 mRNA (Fig. S1). To further test if these structural differences could alter expression, we generated a series of mutations to alter codons and amino acids at Cys110 and then compared bacterial brightness (Table 1). Substituting the other cysteine codon (TGC to TGT) had no impact on brightness. This argued that the codon wobble position was unlikely to be critical to protein expression or brightness. A series of codons for different amino acids at position 110 had a range of effects on protein brightness, but no obvious correlation between a particular nucleotide at a position and bacterial brightness was observed. While our mutagenesis at this position was not exhaustive, a pattern emerged that suggested that hydrophobic amino acids (Ala, Ile, and especially Val) in position 110 generally improved bacterial brightness, while a hydrophilic Thr did not ( Table 1). The benefit of the Val mutation was unique to Cys110. Mutation of Cys180 to Ala or Val decreased bacterial brightness (Table 1). Interestingly, the original Clavularia-derived FP, mTFP1, has a Val at the equivalent 110 codon position. This observation suggested that the process of evolving new FPs from mTFP1 had inadvertently adversely impacted one or more properties of the resulting FPs due to loss of an optimal amino acid.  Table 1. Brightness of bacteria expressing mMaple3 mutants.
If bacterial brightness could only be partially explained by improved protein expression, we reasoned that FP properties must account for the improved bacterial brightness. Here, we received a surprise. Purification and characterization of mMaple3 and moxMaple3 revealed that the proteins had nearly identical spectral characteristics for both the green and photoactivated red versions of the FPs ( Fig. 4 and Supplementary Table S1). moxMaple3 was not brighter. Therefore, coupled with expression level data, we concluded that the Cys110Val mutant must significantly (accounting for >50% of the increased bacterial brightness for moxMaple3) improve the fraction of protein that successfully folds and produces a stable chromophore. Based on the crystal structure of mTFP1, we hypothesized that V105 in mTFP1 forms a functionally significant contact with the highly conserved Ile65, found in all Clavularia FPs ( Supplementary Fig. S2) 19 . The original mutations to monomerize mMaple3 may have decreased the robustness of mMaple3 folding and our C110V moxMaple3 mutation may have inadvertently restored a hydrophobic interaction with Ile65. Other hydrophobic substitutions at C110 support this interpretation (Table 1). Thus, the moxMaple3 C110V mutation could potentially improve the chromophore folding environment and the overall stability of the protein, accounting for our observations of improved expression and chromophore production. Our fortuitous finding raises the possibility that other Clavularia FPs (e.g. mClavGRs, Kaede, mKikGR) may benefit from the C110V mutation, as well. The outcome of our characterizations was striking in how much a single amino acid change could improve FP folding. Impact of Cys110 on mMaple3 maturation. We hypothesized that the C110V mutation improved either the fraction of proteins capable of chromophore formation or the rate of chromophore maturation. Wang et al. observed a two-fold higher fraction of mMaple vs. mMaple3 that could be imaged 13 . This was in part due to the ability of roughly 60% of mMaple molecules to undergo maturation and form chromophores, while mMaple3 matured less efficiently 13 . The authors also suggested that other PSFPs suffer more severely from issues of incomplete maturation and efficiency of photoconversion.
We investigated whether the apparent difference in mMaple3 and moxMaple3 maturation was due to differences in the rate of maturation or in the completeness of maturation. HeLa cells expressing cytoplasmic mMaple3 or moxMaple3 were treated with cycloheximide for 0 or 6 h to inhibit new protein synthesis. We predicted that if mMaple3 matures more slowly than moxMaple3 that there could be a significant increase in cellular fluorescence  Supplementary Fig. S5 for the original gel). The intense bands in the induced lanes were quantified using ImageJ and then normalized relative to the intensity of the mox band. moxMaple3 is expressed at levels ~25% higher than mMaple3.
over time for mMaple3 expressing cells as pre-existing protein matures and forms chromophores. However, we did not observe any increase in mMaple3 fluorescence intensity in cells after 6 h of translation inhibition, nor did moxMaple3 cell intensity increase, either (Fig. 5). We note that FP chromophore maturation half-times range typically from a few minutes to 2-3 h 22 . Therefore, 6 h should have been sufficient to observe any additional mMaple3 maturation and concomitant increase in cell brightness, if maturation was simply slower. The absence of evidence of delayed maturation of a substantial population of mMaple3 molecules suggests that during folding, nascent mMaple3 molecules could become trapped in a terminal nonfluorescent apoprotein conformation. Moreover, similar to bacterial cells, we observed an almost 2-fold higher cellular fluorescence for mox-Maple3 expressing HeLa cells. In contrast, the overall level of protein expression for both variants, as monitored by the fluorescence signal from mCherry, was the same. There is precedent for incomplete maturation of FPs. Chromophores of some FPs readily form more than one chemical structure 23 and up to 30% of the common GFP fails to undergo the final dehydration reaction to form a fluorescent chromophore and remains in a terminal apoprotein form 24 . Bacteria expressing Venus FP are dramatically brighter than those expressing the parental EYFP despite similar brightness of the purified FPs 7 . Our data are consistent with a model in which the moxMaple3 mutations increase the probability of complete conversion of the apoprotein to a mature chromophore-containing protein.
moxMaple3 performance in mammalian cells. Importantly, we wanted to establish that moxMaple3 performed at least as well as mMaple3 in mammalian cells as a fusion protein and for optical highlighting applications. First, we tested whether our mutations had impacted the monomericity of the FP. We fused the parent mMaple3 and moxMaple3 variant separately to the CytERM construct 21 to assess whether either version induced formation of stacked ER membrane whorls, termed OSER. The monomericity of mMaple3 had been previously assessed in a bacterial protein clumping assay 13 . We observed 1 cell with OSER (n = 100 cells) for mMaple3 and 3 cells with OSER (n = 90) for moxMaple3 ( Supplementary Fig. S3). Overall, these represent negligible levels of FP oligomerization and we classify both variants as monomeric. It is also worth noting that cells expressing the CytERM constructs exhibited qualitatively substantial differences in brightness, similar to what we observed for bacteria and for cytoplasmic expression in HeLa cells. Cells expressing CytERM-moxMaple3 were systematically brighter A similar difference in intensities was observed for cells expressing constructs, GalT-mMaple3 and GalT-moxMaple3, in which the FP was localized in the lumen of the Golgi complex, an oxidizing environment ( Supplementary Fig. S4). Taken together, moxMaple3 label cells and cell structures more intensely than the parental mMaple3. We also asked whether moxMaple3 impaired localization of a Golgi complex reporter. Costantini et al. 5 and Kaberniuk et al. 9 previously reported mislocalization of Golgi-targeted coral-derived red FPs in puncta within and distinct from the Golgi complex. We found that a Golgi complex targeted moxMaple3 reporter robustly co-localized with an endogenous Golgi complex marker, GM130 (Supplementary Fig. S5). Thus, moxMaple3 appears to be an inert reporter in cells and is suitable for use in fusion proteins.
Finally, we confirmed that we could photoswitch the new moxMaple3 variant inside the oxidizing environment of the ER (Fig. 6) and thus moxMaple3's suitability for PALM super-resolution imaging. As the photochemical properties of mMaple3 and moxMaple3 were nearly identical, we did not anticipate any gross differences in performance of the FPs. mMaple3 or moxMaple3 were separately targeted to the ER lumen in HeLa cells. Both constructs were readily photoswitched and usable in PALM assays, though moxMaple3 performed slightly better (Fig. 7). Specifically, for PALM, we observed a 5.5% increase in performance for localization precision means for ER-moxMaple3 relative to ER-mMaple3 (43.1 nm vs. 45.6 nm) (Fig. 7C) and a 29.8% higher mean photon count for ER-moxMaple3 relative to ER-mMaple3 (202.8 vs. 156.3) (Fig. 7D).

Discussion
Attempts to study protein behaviors in living cells depend heavily on the inertness of reporter FPs fused to the endogenous proteins. Furthermore, the properties of FPs including brightness, rate of chromophore maturation, and photostability are top considerations. However, completeness of chromophore maturation is not typically measured in FP development studies. That is, the presence of a fluorescent signal is of primary importance while the number of fluorescent molecules is less critical. Yet, the completeness of FP maturation is especially important for quantitative studies in which investigators assume the number of fluorescent molecules equals the number of molecules tagged with the FP.
Previous studies revealed that a fraction of the FPs can misfold in some cellular environments and not produce chromophores while the rest of the population folds and fluoresces 3,5 . Our finding that C110V significantly increases the fraction of moxMaple3 that forms a chromophore highlights a second mechanism by which the number of fluorescent molecules may differ from the number of FP-tagged molecules. Thus, efforts to quantitatively study endogenous cellular proteins tagged with FPs using CRISPR approaches will benefit from validating or using validated FPs capable of correctly and maximally maturing into fluorescent molecules in the cellular environments to be studied. Alternatively, the incompleteness of FP maturation at least should be characterized and factored into any quantitative analyses, e.g. the population of visible fusion proteins is a fraction of the total population of tagged fusion proteins.
We sought to update mMaple3 for use in a variety of cellular environments through a set of targeted mutations. As such, our goal was not to create a brighter FP or one with new properties. Rather, we simply sought to improve the correct folding and utility of mMaple3 in oxidizing and secretory environments. Our improved mox-Maple3 met these goals and unexpectedly increased successful chromophore maturation. The improved protein should be suitable for use in a wide variety of studies both for PALM microscopy and fusion protein fate studies. moxMaple3 now appears to be the best performing monomeric photoswitchable FP for use in PALM approaches.

Plasmid constructs. See Supplemental Data and Figures.
Spectral and photochemical characterization of mMaple3 and moxMaple3. For bacterial expression of the mMaple3 and moxMaple3 nucleotide sequences encoding these proteins were subcloned into a pBAD/ His-D vector using BglII and EcoRI restriction sites. Proteins with 6x polyhistidine tags on the N-terminus were For absorbance measurements, a Hitachi U-2000 spectrophotometer was used. Fluorescence was measured by a FluoroMax-3 spectrofluorometer (Horiba-Jobin Yvon, Edison, NJ, USA). Photoconversion of purified proteins was performed with a 405/20 nm custom assembled LED array. Proteins were diluted with PBS pH 7.5 to 5 µM and placed on ice in transparent micro tubes with following illumination at 15 mW cm −2 for 20 min.
To determine molar extinction coefficients of the green chromophores, we measured mature chromophore concentrations. Purified proteins were alkali-denatured and extinction coefficients was calculated according to the known extinction coefficient (44,000 M −1 cm −1 ) of the synthetic compound of the tyrosine-containing GFP-like chromophores at 447 nm in 1 M NaOH 25 . Based on the absorbance of the native and denatured proteins, molar extinction coefficients for the native states were calculated. Molar extinction coefficients of the red chromophores were estimated by comparing the absorption bands corresponding to the pre-photoconverted green species and the absorption band of the red species from photoconverted samples.
Fluorescence quantum yield of the green and red chromophores of mMaple3 and moxMaple3 were measured relative to EGFP and TagRFP 26 respectively, according to Williams's method 27 .
To evaluate photostability, photoconversion kinetics and reversible photoswitching of the red form, the BL21-AI bacterial cells containing target proteins expressed from a pBAD/His-D plasmid were grown on LB/ampicillin Petri dishes supplemented with 0.1% arabinose, overnight at 37 °C. Bacteria were photobleached using an Olympus IX81 inverted epifluorescence microscope equipped with a 300 W Xenon arc lamp (LB-LS/30, Sutter), 60 × 1.35 NA oil-immersion objective lens (UPlanSApo, Olympus) and 480/40 nm exciter (Olympus Corporation, Center Valley, PA USA). Since 570/30 nm light caused reversible photoconversion of the chromophore from the red to green state, a 480/40 nm exciter was used for measurement of photostability of proteins before and after photoconversion. The intensity of the photobleaching light was measured as 14.7 mW cm −2 at the back focal plane of objective lens. Fluorescence of the green and red forms was detected using a 480/40 nm exciter and 535/40 nm emission filter and a 570/30 nm exciter and 615/30 nm emission filter sets respectively (Chroma Technology Corporation, Bellows Falls, VT). For measurement of the photoconversion halftime, a 390/40 nm exciter was used. The intensity of the photoconversion light was 1.3 mW cm −2 when measured at the back focal plane of a 60 × 1.35 NA oil-immersion objective lens. After photoconversion with 390/40 nm light, as described above, reversible photoswitching of proteins in the red state was performed using 570/30 nm and 480/40 nm light with intensities 32 mW cm −2 and 27 mW cm −2 respectively measured at the back focal plane of objective lens with exposure time 0.5 s. To test whether mMaple3 matured more slowly than moxMaple3, mMaple or moxMaple3 was cloned into a mCherry-T2A-[maple variant] expression plasmid (derived from the Clontech N1-EGFP plasmid). HeLa cells grown in 12-well plates were transfected using Effectene (Qiagen) and then treated 36 h post-transfection with 50 mg/ml cycloheximide (Acros Organics, cat #357420010) for 0 or 6 h. Cells were trypsinized, collected by centrifugation at 1000 × g 5 min, washed with 1x PBS, and then lysed in M-PER solution (Mammalian Protein Extraction Reagent, Thermo Scientific cat #78501) and fluorescence intensities were measured (for mCherry using excitation 530 nm, emission 545-700 nm and for Maple3 variants excitation 455 nm and emission 475-600 nm) with a FluoroMax-3 spectrofluorometer (Horiba-Jobin Yvon Inc. Edison, NJ, USA). The green fluorescence intensities were divided by the respective mCherry fluorescence intensities to normalize for expression and then all values were normalized to the mean intensity of mMaple3 at 0 h cycloheximide.