A palette of bright and photostable monomeric fluorescent proteins for bacterial time-lapse imaging

Fluorescent proteins (FPs) are pivotal for examining protein production, localization, and dynamics in live bacterial cells. However, the use of FPs in time-lapse imaging is frequently constrained by issues such as oligomerization or limited photostability. Here, we report the engineering of four new fluorescent proteins: mChartreuse (green), mJuniper (cyan), mLemon (yellow), and mLychee (red), using site-directed mutagenesis. These fluorophores outperform current standards for photostability and aggregation properties while retaining high levels of brightness in living E. coli bacteria. Starting with superfolder GFP (sfGFP), we developed mChartreuse, which shows significantly enhanced brightness, stability, and monomericity compared to other GFP variants. From mChartreuse, we derived mJuniper, a cyan fluorescent protein with rapid maturation and high photostability, and mLemon, a yellow fluorescent protein with improved photostability and brightness. We also identified a mutation that eliminates residual oligomerization in red fluorescent proteins derived from Discosoma species, such as mCherry and mApple. Incorporating this mutation into mApple, along with other substitutions, resulted in mLychee, a bright and photostable monomeric red fluorescent protein. These novel fluorescent proteins advance fluorescence time-lapse analysis in bacteria and their spectral properties match current imaging standards, ensuring seamless integration into existing research workflows.


Introduc.on
Fluorescent proteins (FPs) are essen4al tools in life science, enabling the tracking of protein localiza4on and gene expression at the single-cell level (Chudakov et al., 2010).Green fluorescent protein (GFP) was ini4ally cloned from the jellyfish Aequorea victoria and ini4ated a revolu4on in the study of protein localiza4on (Chalfie et al., 1994).However, wild-type A. victoria GFP (avGFP) showed several shortcomings since it matured poorly at 37°C and required the use of phototoxic ultraviolet light (400 nm) to be excited (Ogawa et al., 1995).Mutagenesis of avGFP enabled the discovery of wellfolded variants that were excited by blue light (488 nm), such as EGFP, GFPmut2, or superfolder GFP (sfGFP) (Cormack et al., 1996;Pédelacq et al., 2006;Zhang et al., 1996).GFP derivates also displayed a tendency to dimerize and cause mis-localiza4on of fusion partners, which could be eliminated by disrup4ng a small hydrophobic dimeriza4on interface with the A206K subs4tu4on (Zacharias et al., 2002).Subs4tu4on of residues cri4cal for fluorescence also enabled the engineering of blue, cyan, and yellow variants of GFP, expanding the paleke of usable FPs (Heim et al., 1994;Ormö et al., 1996).The discovery of a red fluorescent protein (RFP) in Discosoma sp.paved the way for mul4color analysis.Indeed, this Discosoma sp.red fluorescent protein (DsRed) was homologous to avGFP but excited by green light (558 nm), which therefore enabled simultaneous imaging of GFP and DsRed with minimal spectral overlap (Matz et al., 1999).However, engineering usable deriva4ves of DsRed was extremely challenging since this red FP matures slowly, oligomerizes as a 4ght tetramer, and can stochas4cally mature as a green side-product that overlaps with the spectral proper4es of GFP (Bevis and Glick, 2002).Bevis & Glick overcame the slow matura4on of DsRed by random mutagenesis, yielding DsRed-Express, which matured 15 4mes faster than DsRed (Bevis and Glick, 2002).A tour-de-force study by the Tsien lab reported the monomeriza4on of DsRed-Express into a fast-maturing monomeric RFP (mRFP1) through structure-guided disrup4on of its AB and AC tetrameriza4on interfaces (Campbell et al., 2002).While mRFP1 was dim and bleached quickly when excited, subsequent studies evolved a paleke of DsRed deriva4ves with a wide range of proper4es, which were called mFruits (Shaner et al., 2008(Shaner et al., , 2004)).Of all mFruits, mCherry became the standard RFP due to its fast matura4on, high photostability, and lack of GFP-like side product (Shaner et al., 2004).
Over the years, poten4ally useful fluorescent proteins were iden4fied in other organisms.For example, an extremely bright yellow FP was iden4fied in the cephalochordate Branchiostoma lanceolatum and engineered in a monomeric green FP called mNeonGreen (mNG) (Shaner et al., 2013).mNeonGreen was twice as bright as EGFP, with similar photostability and matura4on kine4cs as the laker (Shaner et al., 2013), thereby highligh4ng the poten4al for bright and exploitable FPs to be discovered in organisms beyond Discosoma sp. or A. victoria.Other recently released FPs such as AausFP1 from Aequorea australis, mStayGold from Cyteais uchidae or AzaleaB5 from Mon=pora monasteriata show great promise as bright fluorescent probes (Ando et al., 2023b(Ando et al., , 2023a;;Lambert et al., 2020).Novel FPs could also be engineered de novo from a synthe4c template as was the case for the mScarlet family (Bindels et al., 2016).Further evolu4on of mScarlet yielded mScarlet3 and mScarlet-I3, the brightest red FPs to date (Gadella et al., 2023).
All FPs have different sets of parameters that dictate setups for which they should be used.These include fluorescence spectrum, brightness, matura4on rate, photostability and dispersibility (Cranfill et al., 2016).The brightness of FPs is measured as the product of product of its ex4nc4on coefficient and quantum yield and represents the amount of signal that can be obtained from a given FP.
However, another cri4cal parameter that influences in vivo brightness is matura4on rate, which defines the kine4cs by which a folded dark FP is converted into a fluorescent species by cycliza4on of three amino acids that become its chromophore.Because bacteria tend to be fast-growing organisms, the use of fast-maturing FPs is of utmost importance in these organisms to achieve high brightness, since dark species of slow-maturing FPs would quickly get diluted by fast doublings (Balleza et al., 2018).

Site-directed mutagenesis improves all proper=es of sfGFP
We first aimed to obtain an improved GFP derived from the well-characterized A. victoria GFP (avGFP) scaffold.We chose sfGFP as a star4ng template due to its improved folding, fast matura4on, high brightness and broad use in bacterial models (Balleza et al., 2018;Pédelacq et al., 2006).We first introduced the V206K monomerizing muta4on and the F145Y photostability-improving muta4on as previously described for the construc4on of msGFP2 (Valbuena et al., 2020).To further improve the brightness of this star4ng template, we used high-throughput flow cytometry to screen several muta4ons found in bright avGFP deriva4ves.Aser four rounds of site-directed mutagenesis, we iden4fied our brightest variant containing muta4ons N39I, I128S, D129G from the Achilles YFP (Yoshioka-Kobayashi et al., 2020) and N149K from the Emerald GFP (Teerawanichpan et al., 2007) (Figure 2A, Supplementary Figure 1).We called this variant mChartreuse based on the similarly colored liquor from southeast France.(Bindels et al., 2016), where the tested FP is fused to a spectrally dis:nct FP.The ra:ometric measurement of the tested FP normalized to that of the reference FP enables to offset any varia:on in FP signal due to differences in expression levels.The rigid spa:al linker that separates both FPs prevents fluorescence resonance energy transfer (FRET).C: In vivo brightness quan:fica:on of green FPs. Green FP fluorescence was measured in exponen:ally growing cultures and normalized to that of mCherry.
Data shows mean and standard devia:on of six independent replicates.D: Descrip:on of the matura:on assay.Exponen:ally growing cells were treated with 100 µg/mL erythromycin, 10 µg/mL tetracycline hydrochloride & 10 µg/mL rifampicin to inhibit protein synthesis and enable full matura:on of dark unmatured fluorophores.Matura:on kine:cs were measured from this point and fiaed to a pseudo-first order kine:cs, from which matura:on half-:mes were computed.E: Matura:on half-:mes of green FPs.Purified mChartreuse displayed an ex4nc4on coefficient of 71 mM -1 cm -1 .and a quantum yield of 0.75 (Table 1).However, evalua4ng in vivo brightness is a beker indicator of FP performance in living systems.By fusing mChartreuse and other GFPs with mCherry, we can evaluate the in vivo brightness of these FPs in growing cells by normalizing the green fluorescence signal to that of mCherry, therefore allowing us to offset any varia4on in the expression level of these constructs (Bindels et al., 2016) (Figure 2B).Our results show that mChartreuse is 30% brighter than sfGFP in E. coli (Figure 2C).mChartreuse is also brighter than mNeonGreen and mGreenLantern, two bright and top-performing GFPs that will be used hereaser as comparison points to sfGFP and mChartreuse (Campbell et al., 2020;Shaner et al., 2013) (Figure 2C).
To assess how well mChartreuse matured in exponen4ally growing E. coli, we quan4fied the matura4on half-4me of mChartreuse and other green FPs by shuung down protein synthesis using a cocktail of erythromycin, tetracycline and rifampicin, followed by measuring fluorescence matura4on kine4cs aser protein synthesis shutoff (Figure 2D).In this setup, protein synthesis inhibi4on arrests fluorophore biosynthesis, allowing remaining dark unmatured fluorophores to fully mature (Supplementary Figure 2A).By fiung a pseudo-first order kine4cs curve to the matura4on data, we determined the matura4on half-4me of mChartreuse to be 4.9 min, faster than its sfGFP parent (10.7 min), or than mNeonGreen (10.3 min) and mGreenLantern (7.4 min) (Table 1, Figure 2E).
The excita4on and emission spectra of untagged mChartreuse were determined aser three phase par44on purifica4on (Jain et al., 2004).mChartreuse displayed a broad absorbance peak with a maximum at 487 nm, while its emission peaked at 510 nm, similar to that of superfolder GFP or EGFP (Pédelacq et al., 2006;Zhang et al., 1996) (Table 1, Figure 2F).The fluorescence of mChartreuse displays a pKa of 4.9, showing that it is resistant to fluctua4ons in cytosolic pH (Table 1,   Supplementary Figure 3) We evaluated the dispersibility and oligomeriza4on of mChartreuse and other green FPs by fusing them to the ClpP protease as previously described (Landgraf et al., 2012).In this system, oligomeric FPs cause coopera4ve clustering of ClpP homo-tetradecamers into fluorescent foci, while monomeric FPs do not perturbate the homogenous cytosolic localiza4on of ClpP (Landgraf et al., 2012).To offer an unbiased quan4fica4on of foci forma4on by these ClpP fusions, we measured the skewness of pixel intensity distribu4ons in single bacteria as a proxy for fluorescence inhomogeneity (Ducret et al., 2016).As previously described, fusion of ClpP with sfGFP (a dimerizing FP) induced foci forma4on, exemplified by the high fluorescence skewness observed in such cells (Figure 2G-H).On the other hand, ClpP fusions with mChartreuse, mNeonGreen, and mGreenLantern all displayed homogeneous fluorescence as shown by the low fluorescence skewness values of these fusions (Figure 2G-H).Our results therefore show that mChartreuse is monomeric in E. coli.
Conclusively, the evolu4on of mChartreuse improved all aspects of sfGFP (i.e brightness, matura4on, dispersibility and photostability), making it a superior choice for all applica4ons.

Monomeriza=on and enhancement of the mApple red FP.
We also sought to engineer a bright and photostable monomeric red fluorescent protein.While mScarlet-I3 was monomeric and offered unrivaled brightness, it showed lower photostability compared to other red FPs (Gadella et al., 2023).On the other hand, mFruits like mCherry and mApple offered higher photostability (Shaner et al., 2008(Shaner et al., , 2004)).However, previous data shows that mFruits aggregate when fused to ClpP in E. coli (Landgraf et al., 2012).Indeed, our data shows that fusions of mCherry and mApple with ClpP had a high propensity to form foci, confirming that these FPs are not monomeric (Figure 4A).On the other hand, ClpP-mScarlet-I3 did not form foci, indica4ng that this FP is truly monomeric (Figure 4A).Since mScarlet-I3 was derived from a mFruit-based synthe4c template, we introduced muta4ons specific to mScarlet-I3 into mApple, then screened for muta4ons conferring a non-aggrega4ve behavior in ClpP fusions.The S131P muta4on (numbering rela4ve to DsRed) abolished foci forma4on in ClpP fusions with mApple or mCherry (Figure 4A).In DsRed tetramers, S131 is part of the AB tetramer interface between two DsRed protomers, with the hydroxyl sidechain and the amide nitrogen of S131 each implicated in a hydrogen bond with the carboxyl sidechain of D154 located in trans (Figure 4B).It is therefore likely that the subs4tu4on of S131 with a proline prevents H-bond forma4on with D154 and eliminates residual dimeriza4on of mFruits in E. coli.we used mApple-S131P as a star4ng template to engineer a brighter RFP.During the previous engineering of mFruits, the original N and C-termini of DsRed were replaced with those of avGFP (Shaner et al., 2004).While this allowed proper localiza4on of some fusions, restora4on of DsRed-like termini was shown to improve brightness and relieve RFP-induced cytotoxicity (Gadella et al., 2023;Valbuena et al., 2020).We therefore synthe4zed a codon-op4mized mApple template, replaced the MVSKGEENNM N-terminus and TGGMDE C-terminus of mApple with the MDSTE N-terminus and GSQGGSGGS C-terminus of mCherry2C (Supplementary Figure 5) (Valbuena et al., 2020).We subsequently performed rounds of site-directed mutagenesis to iden4fy muta4ons conferring increased in vivo brightness.Muta4ons V71A, L85Q, K139R, A145P and I210V were incorporated in our final variant, bap4zed mLychee (Figure 4C, Supplementary Figure 5).The mLychee red FP is 28% brighter than mApple, and six 4mes brighter than mCherry under excita4on at 565 nm, while mScarlet-I3 outperforms other FPs, being more than thrice as bright as mApple (Figure 4D).mLychee showed a matura4on half-4me of 36.4 min, faster than its mApple parent (55.0 min) or than mCherry (44.4min), but slower than mScarlet-I3 (21.5 min) (Table 1, Figure 4E, Supplementary Figure 2D).Similarly to what was published for mScarlets and mApple (Bindels et al., 2016;Gadella et al., 2023;Shaner et al., 2008), mLychee displays an absorbance peak at 568 nm, an emission peak at 596 nm and a pKa of 6.3 (Table 1, Figure 4F, Supplementary Figure 3).mCherry displayed the highest photostability (t1/2 = 3.6 s) while mScarlet-I3 bleached the quickest (t1/2 = 0.4 s) (Figure 4G).mApple and mLychee displayed complex bleaching behaviors, with an abrupt loss of 15% of their signal aser the first 100 ms exposure, followed by slow bleaching kine4cs similar to that of mCherry (t1/2 = 2.7 s & 2.8 s, respec4vely) (Figure 4G).Previous results showed that this fast ini4al bleaching of mApple was a photochromic phenomenon reversible either by 4me or by illumina4ng this fluorophore with blue light (Shaner et al., 2008).We therefore quan4fied the bleaching of mLychee by alterna4ng illumina4ons between 555 nm and 474 nm (Figure 4H).In these condi4ons, the fast ini4al bleaching of mLychee was prac4cally eliminated, yielding ini4al bleaching kine4cs similar to that of mCherry (Figure 4H).Therefore, mLychee is a truly monomeric red FP that offers an excellent balance between brightness and photostability.

Discussion
In this manuscript, we introduce a novel paleke of fluorescent proteins tailored for bacterial applica4ons.This toolkit that includes mChartreuse, mJuniper, mLemon, and mLychee, represents a significant advancement in 4me-lapse imaging with its improved proper4es compared to tradi4onal and commonly used fluorescent proteins.The enhanced photostability and minimal aggrega4on of these novel proteins make them ideal fusion tags, and their spectral characteris4cs enable straighyorward integra4on into exis4ng research workflows without requiring specialized equipment.
Contras4ng with other modern bright FPs like mNeonGreen or StayGold, mChartreuse and its deriva4ves were engineered from sfGFP, a well-folded green FP that is broadly described and used in bacterial imaging studies.Applica4ons that currently use sfGFP can therefore be quickly to use mChartreuse, mJuniper and mLemon.Further tes4ng of these novel FPs as fusion tags in E. coli and other bacterial species will enable us to assess how well such fusions preserve na4ve localiza4on and ac4vi4es of essen4al prokaryo4c proteins.
Our findings confirm that the F145Y muta4on significantly improves the photostability of sfGFP deriva4ves, consistent with previous research (Valbuena et al., 2020).However, the rela4onship between oligomeriza4on and residue subs4tu4on appears more intricate.To examine FP oligomeriza4on in E. coli, we developed a quan4ta4ve method based on the skewness in fluorescence distribu4on of ClpP fusions within single bacteria, enabling unbiased and robust quan4fica4on of FP aggrega4on and foci forma4on.Our results confirm that the V206K muta4on abolishes oligomeriza4on in sfGFP deriva4ves (Valbuena et al., 2020).However, since the A206K muta4on is also present in mVenus and mYpet, which tend to aggregate in the ClpP assay (Landgraf et al., 2012), this suggests that aggrega4on behavior is influenced by factors beyond this muta4on, such as superfolder muta4ons present in our ini4al sfGFP template (S30R, F99S, N105T, I171V) or muta4ons that were introduced during the development of mChartreuse (N39I, I128S, D129G, F145Y, N149K).
Previous research showed that mCherry and mApple are monomers in mammalian cells when assayed using the Organized Smooth Endoplasmic Re4culum (OSER) approach (Bindels et al., 2016;Cranfill et al., 2016).However, fusions of mCherry with ClpP, IbpA or RpoS in E. coli all caused substan4al aggrega4on (Goormagh4gh et al., 2018;Govers et al., 2018;Landgraf et al., 2012), sugges4ng that red fluorescent proteins derived from Discosoma sp, are not monomeric in bacterial cells.The reasons for these discrepancies are not known but might stem from differences in cytosolic composi4on and crowding between mammalian and bacterial cells, or in differences of sensi4vity between the OSER and ClpP assays.Nevertheless, the subs4tu4on of the serine 131 of mApple by a proline abolishes the aggrega4ng behavior of the mLychee variant developed from mApple.While mLychee is dimmer than mScarlet-I3, it compensates for this lack of brightness with robust photostability.mLychee would therefore be more appropriate for 4me-lapse applica4ons while mScarlet-I3 would remain preferable for endpoint methods (e.g., epifluorescence microscopy snapshots, flow cytometry) where photobleaching due to repeated excita4on is not an issue.
Although these novel fluorescent proteins were ini4ally developed for bacterial applica4ons, they have the poten4al to significantly enhance imaging techniques in eukaryo4c cells as well.One major considera4on in transferring these FPs to eukaryo4c systems is their compa4bility with the cellular environment, including cytosolic composi4on, protein expression, and folding dynamics, which differ significantly from bacterial cells.Therefore, op4mizing these FPs to eukaryo4c systems would require adapta4ons such as codon usage op4miza4on for efficient expression, as well as poten4al adjustments to accommodate eukaryo4c post-transla4onal modifica4ons and cellular environments.
All fluorescence measurements were performed on a Spark plate reader (Tecan) using a 5 nm bandpass monochromator.Cyan fluorescence was excited at 450 nm and collected at 480 nm, green fluorescence was excited at 480 nm and collected at 510 nm, yellow fluorescence was excited at 510 nm and collected at 540 nm, and red fluorescence was excited at 565 nm and collected at 595 nm.

Molecular cloning
All plasmid constructs were obtained by Gibson assembly unless specified otherwise and are detailed in Supplementary Table 1.All primers used to construct these vectors are detailed in Supplementary Table 2.All enzymes were purchased from New England Biolabs.Polymerase chain reac4ons were performed using PrimeSTAR Max (Takara).Synthe4c genes for mApple with modified N and C-termini and mGreenLantern were ordered from Integrated DNA Technologies.pFN01 vectors, which encode all tested FPs fused to a reference FP and separated with a rigid linker, were constructed by amplifying a mTurquoise2-linker-mScarlet-I3 fragment from pDress-mTq2-link-mSc-I3 (Gadella et al., 2023) with primers IpFN01 F & R and a mini-F backbone from pNF02 using VpFN01 F & R. Red and yellow FPs were amplified using primer pairs 01miscF & R (mCherry, mApple, SYFP2), 01YPet F & R (mYPet), and 01sfGFP F & R (mLemon) on their respec4ve templates (Supplementary Table 1) (Gadella et al., 2023;Shaner et al., 2008).To clone green and cyan FPs, mTurquoise2 was first replaced by mCherry on pFN01 by amplifica4on of the backbone using VmTq2mCh F & R and by amplifica4on of mCherry from pROD62 using ImTq2mCh F & R. Subsequently, FPs were cloned in this vector using primers 01sfGFP F & R (sfGFP, mChartreuse, mJuniper), 01mNGb F & R (mNeonGreen), mGL F & 01sfGFP R (mGreenLantern), and 01misc F & R (mTurquoise2) amplified from their respec4ve templates (Supplementary Table 1) (Gadella et al., 2023;Rousseau et al., 2023).
The pCLP vector used to fuse FPs to ClpP was constructed by amplifying clpP and its na4ve promoter using primers clpP F & R.This fragment was digested with NheI and BamHI and cloned in pUA66 (Zaslaver et al., 2006) digested with AvrII and BamHI.To construct ClpP-FP fusions, FPs were amplified from corresponding pFN01 vectors with primers insCLP F & R and cloned in a pCLP backbone amplified with pCLP F & R. Muta4on S131P was introduced on pCLP-mCherry and pCLP-mApple using primers mFruitS131P F & R. pET151 vectors used for untagged FP produc4on were constructed by amplifica4on of a pET151 backbone using primers pET151 F & R, and of mChartreuse, mLemon, mJuniper or mLychee from pNF02 vectors using primers 151FP F & R.
Lysates were separated from beads by piercing the bokom of the tube and by collec4ng the liquid phase in a bigger tube by centrifuga4on.Aser clearing by centrifuga4on, lysates are processed by three-phase par44on (TPP) purifica4on (Jain et al., 2004).A first TPP step was performed with 20% ammonium sulfate satura4on & 1 volume t-butanol at room temperature.Aser phase separa4on by centrifuga4on, the aqueous phase of this TPP was subjected to a second step with 60% ammonium sulfate satura4on & 2 volumes of t-butanol at room temperature, from which the interphase was isolated by centrifuga4on.This interphase was resuspended in 10 mM Tris-HCl pH 8.0 and ridden of insoluble impuri4es by centrifuga4on.A final desal4ng step was performed using Sephadex G-25 spin columns (Cy4va).Purified FPs were stored at 4°C in the dark and were les to mature at least 24 hours before experiments.Centrifuga4on of FP mother liquors was performed before all experiments to remove insoluble interfering impuri4es.All absorbance measurements were performed using FPs diluted in 10 mM Tris-HCl pH 8.0 to achieve a peak absorbance of 0.2, while all samples for fluorescence measurements were diluted to a peak absorbance of 0.04.Ex4nc4on coefficients were determined as described (Cranfill et al., 2016), by normalizing the absorbance of FP samples to that of FPs denatured with 1 M NaOH, under the assump4on that the molar ex4nc4on coefficient of the denatured chromophore at 447 nm (mChartreuse, mLemon) or 457 nm (mLychee) is 44 mM -1 cm -1 .Since mJuniper could not be denatured by strong bases, its concentra4on was determined by BCA assay (Pierce BCA Protein Assay Kit, Thermo Scien4fic).
Buffer 4tra4on to determine pKas was performed by mixing FP samples with 1 volume of citratephosphate-borate buffer at a given pH as described (Gadella et al., 2023).These buffers were prepared from a solu4on of 100 mM citric acid, 100 mM boric acid & 100 mM monosodium phosphate 4trated with NaOH.Buffers were les to stabilize 24h at room temperature, yielding pH values of 3. 01, 3.92, 4.92, 5.93, 7.07, 8.03, 8.90 & 9.94.Each replicate of fluorescence measurements was fiked to a Hill func4on (F(pH)=Fmax/(1+10 n*(pKa-pH ) using the least squares method, with F(pH) (fluorescence at a given pH) and pH as variables, and Fmax (fluorescence at plateau), n (Hill coefficient) and pKa as parameters.

Microscopy procedures
Bacteria grown as described in general procedures were sealed on M9GC pads containing 1% agarose using GeneFrames (ThermoFisher).Cells were imaged using an Eclipse Ti2 microscope (Nikon) equipped with a Plan Apo λ 100x/1.45objec4ve, a motorized stage, Z-dris correc4on (Perfect Focus System, Nikon), hea4ng (Okolab), a solid-state light source (Spectra X, Lumencor) and a sCMOS camera (Orca-Fusion BT, Hamamatsu).Cyan fluorescence was excited using a 438/24 excita4on filter and collected using a 482/25 emission filter, green fluorescence was excited using a 474/27 excita4on filter and a 515/30 emission filter, yellow fluorescence was excited using a 509/22 excita4on filter and a 544/25 emission filter.Red fluorescent proteins were excited using a 578/21 excita4on filter and a 641/75 emission filter for ClpP fusions or with a 553/24 excita4on filter for photostability experiments.
All fluorescent proteins were imaged using ad-hoc LEDs at 50% power intensity with 100 ms exposure 4mes, except RFP photostability experiments which used 12% power of the 555 nm LED.Images were subsequently processed using MicrobeJ by automa4c detec4on of cells and quan4fica4on of singlecell fluorescence parameters (i.e.mean for photostability experiments, skewness for ClpP fusions) (Ducret et al., 2016).Photobleaching experiments were performed by bleaching live cells with 100 ms steps using appropriate excita4on seungs as described above.Photobleaching half-4mes were determined as the 4me at which the mean fluorescence of the popula4on dropped below 0.5.

Figure 1 :
Figure 1: Novel fluorescent proteins detailed in this study A: Flow chart of fluorescent proteins developed in this study.Each arrow shows a development step with its associated subs:tu:ons.New N-ter/C-ter refers to the replacement of mApple N and C-termini with those of mCherry2C (Valbuena et al., 2020).B: Images of bacterial colonies carrying pNF02 vectors encoding each indicated fluorescent proteins, transilluminated with UV light.

Figure 2 :
Figure 2: Development and characterizaCon of the mChartreuse green fluorescent protein.A: All six muta:ons introduced in sfGFP to obtain mChartreuse.B: Representa:on of the brightness quan:fica:on system developed by(Bindels et al., 2016), where the tested FP is fused to a spectrally dis:nct FP.The Data shows mean and standard devia:on of three independent replicates.F: Spectral proper:es of mChartreuse.The dashed line shows the absorbance spectrum, the light-colored line shows the excita:on spectrum, and the dark-colored line shows the emission spectrum.Data is shown as the mean of three independent spectrum acquisi:ons.G: Representa:ve images of ClpP fusions.Fusion of ClpP with a dimeric FP (i.e.sfGFP) causes substan:al aggrega:on, which leads to the forma:on of fluorescence foci and high fluorescence skewness in single cells.Fusion with a monomeric FP (i.e.mChartreuse) does not cause aggrega:on, leading to homogenous FP distribu:on in the cytosol and low fluorescence skewness.H: Fluorescence distribu:on skewness of single cells expressing ClpP-FP fusions.Bars show the mean and standard devia:on.Data were acquired from three independent experiments, with the total numbers of analyzed bacteria shown in brackets.I: Photostability of FP signals in live bacteria illuminated with 100 ms steps and normalized to intensity at :me 0. The dashed grey line shows the cutoff at which t1/2 were calculated.Data were acquired from three independent experiments, with the total numbers of analyzed bacteria shown in brackets.

Figure 3 :
Figure 3: Development and characterizaCon of cyan and yellow variants of mChartreuse.A: Muta:ons introduced in mChartreuse to obtain the mJuniper cyan variant and the mLemon yellow variant.B: In vivo brightness quan:fica:on of cyan and yellow FPs.Fluorescence was measured in exponen:ally growing cultures and normalized to that of mCherry (for cyan FPs) or mTurquoise2 (for yellow FPs).Data shows mean and standard devia:on of six independent replicates.C: Matura:on half-:mes of cyan and yellow FPs.Data shows mean and standard devia:on of three independent replicates.D & G: Spectral proper:es of mJuniper (D) and mLemon (G).The dashed line shows the absorbance spectrum, the light-colored line shows the excita:on spectrum and the dark-colored line shows the emission spectrum.Data is shown as the mean of three independent spectrum acquisi:ons.E & H: Fluorescence distribu:on skewness of single

Figure 4 :
Figure 4: Development and characterizaCon of the mLychee red fluorescent protein.A: Fluorescence distribu:on skewness of single cells expressing ClpP-FP fusions.Bars show the mean and standard devia:on.Data were acquired from three independent experiments, with the total numbers of analyzed bacteria shown in brackets.B: Loca:on of serine 131 (S131) in the AB dimeriza:on interface of a DsRed tetramer (PDB 1ZGO) shown in ribbon representa:on.The inlet shows the interac:on between S131 and aspartate 154 (D154) located in trans at the atomic level, with puta:ve hydrogen bonds shown as dashed lines.C: All six muta:ons introduced in mApple to obtain mLychee.D: In vivo brightness quan:fica:on of red FPs.Fluorescence was measured in exponen:ally growing cultures and normalized to that mTurquoise2.Data shows mean and standard devia:on of six independent replicates.E: Matura:on half- pNF02 vectors used for unfused FP expression were constructed by amplifying a mini-F backbone from pNF02-mScarlet-I (Goormagh4gh et al., 2018) using primers bbpNF02 F & R. sfGFP was introduced in pNF02 using primers sfGFP02 F & R while mApple with new N and C-termini (mAppleNC) was directly inserted as a synthe4c gene.Muta4ons to construct mChartreuse were introduced in pNF02-sfGFP using primers N39I F & 39 R, D129G F & I128S, N149K F & Y145F R, and V206K F & 206 R. mJuniper was generated from pNF02-mChartreuse using primers S72A F & Y66W, and H148D F & N146F R while mLemon was constructed using primers S72A F & S65GT63S F, V206K F & T203Y, and V224L F & 221 R. mLychee was obtained from pNF02-mAppleNC using primers V71A F & L85Q R, S131P F & R, A145P F & K139R R, and 219 F & I210V R.