Combining systems and synthetic biology for in vivo enzymology

Enzymatic parameters are classically determined in vitro, under conditions that are far from those encountered in cells, casting doubt on their physiological relevance. We developed a generic approach combining tools from synthetic and systems biology to measure enzymatic parameters in vivo. In the context of a synthetic carotenoid pathway in Saccharomyces cerevisiae, we focused on a phytoene synthase and three phytoene desaturases, which are difficult to study in vitro. We designed, built, and analyzed a collection of yeast strains mimicking substantial variations in substrate concentration by strategically manipulating the expression of geranyl-geranyl pyrophosphate (GGPP) synthase. We successfully determined in vivo Michaelis-Menten parameters (KM, Vmax and kcat) for GGPP-converting phytoene synthase from absolute metabolomics, fluxomics and proteomics data, highlighting differences between in vivo and in vitro parameters. Leveraging the versatility of the same set of strains, we then extracted enzymatic parameters for two of the three phytoene desaturases. Our approach demonstrates the feasibility of assessing enzymatic parameters directly in vivo, providing a novel perspective on the kinetic characteristics of enzymes in real cellular conditions.


Introduction
Enzymes catalyze most of the chemical reactions in living systems.A comprehensive interpretation of enzymatic parameters is therefore of paramount importance to grasp the complexity of cellular metabolism.The vast amount of data collected for thousands of enzymes has contributed to significant progress in our understanding of the remarkable chemical capabilities of biocatalysts and of their roles in cellular reactions.
Enzymatic reactions are traditionally analyzed in vitro, under dilute conditions, using pure or semipure protein samples in buffer solution.In contrast, the cellular medium is most accurately viewed as a heterogeneous, dense, crowded gel, containing various types of macromolecules and cellular lipidic organelles, with potential partitioning effects and variations in substrate and/or product diffusion coefficients (1).The parameters determined in classical enzymology experiments may therefore not be representative of in vivo reaction rates and equilibrium constants (2,3).While some progress has been made in implementing and understanding viscosity and crowding effects in in vitro enzymatic assays, these conditions do not mimic the intrinsic complexity of the cellular environment (2).
In vivo enzymology seems to be the obvious approach to measure enzymatic parameters inside cells.
Early attempts were made using the enzyme photolyase, for which both in vitro and in vivo parameters were determined (4,5).Meanwhile, the in vivo V max values of ten central carbon metabolism enzymes determined in the early 90s revealed significant differences between in vivo and in vitro assays for heteromeric protein complexes (6).The past ten years have seen growing interest from systems biology in determining the in vivo k cat of native enzymes in model organisms.
The in vivo apparent k cat of E. coli enzymes have been determined independently by two research groups, leveraging advances in absolute protein quantification and high throughput metabolomics (7,8).These values were obtained by dividing the flux of enzymatic reactions by the absolute abundance of the corresponding enzymes.Both studies found correlations between in vivo and in vitro data (with correlation coefficients around 0.6), suggesting that this method could serve as an alternative to in vitro assays.While this systems biology approach has provided valuable information, it cannot be employed to determine apparent K M s in vivo.Zotter et al. measured the activity and affinity of TEM1-β lactamase in mammalian cells in vivo with confocal microscopy, using a fluorescently tagged enzyme and a fluorescent substrate and product (9).They observed that the catalytic efficiency (k cat /K M ) of this enzyme differs in vitro and in vivo due to substrate attenuation, indicating that in vitro data are not always indicative of in vivo function.Zotter et al. is probably the most comprehensive in vivo enzymology study to date, but this approach cannot be generalized because of a lack of universally appropriate fluorescent substrates/products for all enzymes.In a recent investigation of the in vivo kinetic parameters of thymidylate kinase (TmK), an interesting finding was the difference in TmK's activity pattern when the substrate (thymidine monophosphate, TMP) wa s supplied in the media versus provided by internal metabolism, with Michaelis-Menten kinetics in the former and Hill-like kinetics in the latter.The authors hypothesized that the limited diffusion of TMP might be due to its confinement in a putative metabolon in E. coli (10).
In vivo enzymology is a particularly attractive prospect for membrane and multimeric proteins (6), which are tedious to purify and for which activity assays are difficult to optimize, mostly because artificial membranes are required (11)(12)(13)(14)(15).In the industrially important carotenoid pathway for example, despite the expression of numerous carotene biosynthesis enzymes, our understanding of their enzymatic behavior remains limited because many are membrane-associated proteins.While kinetic parameters for some phytoene synthases, sourced from plants or bacteria, have been established in vitro, many enzymatic assays require the co-expression of geranylgeranyl pyrophosphate (GGPP) synthase to attain activity, possibly because of the amphiphilic nature of the substrate or the requirement of membranes (12,15).Another example of the difficulty of in vitro assays is phytoene desaturase, for which discernible enzyme activity has only been achieved in engineered environments (e.g.liposomes) with cell-purified substrate (14,(16)(17)(18).
This study combines synthetic and systems biology tools to develop an original in vivo enzymology approach.Synthetic biology offers a remarkable set of genetic engineering tools to precisely modulate the activity of enzymes in synthetic pathways, enabling in vivo control of substrate concentrations for the studied enzymatic reactions.In turn, systems biology can provide quantitative data on these reactions (fluxes and enzyme, substrate and product concentrations) and computational tools to model their behavior.We applied this approach to investigate a synthetic carotenoid production pathway in Saccharomyces cerevisiae, with industrial applications ranging from foods to pharmaceuticals.

General principle of the proposed in vivo enzymology method
Enzymes are usually characterized in terms of their affinity (K M ) and activity (maximal reaction rate V max and turnover number k cat ) (Fig 1).These parameters are typically determined in vitro by varying the substrate concentration across a relatively broad range (about two orders of magnitude) and measuring the reaction rate for each substrate concentration (19).Different mathematical formulas, such as the Michaelis-Menten equation, have been derived to estimate enzymatic parameters from these data.We propose a novel approach wherein the substrate concentration is varied directly within cells (Fig 1 ), by modulating the concentration of the enzyme producing the substrate of the studied reaction.Levels of the substrate-producing enzyme are varied using different combinations of promoter strengths and gene copy numbers, and translate into a wide range of substrate concentrations (2-3 orders of magnitude, as detailed in the following sections).In our experimental setup, each substrate concentration is achieved using a specifically engineered yeast strain.In each of these strains, the gene encoding the enzyme of interest is expressed at a given level, mirroring the conditions of in vitro experiments.Engineered strains are then grown under steady-state conditions (i.e., exponential growth), where reaction fluxes and metabolite concentrations remain constant (20).This stability allows for the accurate measurement of product formation fluxes (equivalent to reaction rates), intracellular substrate concentrations, and enzyme concentrations.The data obtained can then be combined to calculate apparent in vivo kinetics parameters, denoted ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ , V ୫ୟ୶ ୡୣ୪୪ and ݇ ୡୟ୲ ୡୣ୪୪ in analogy with the classical K M , V max and k cat parameters (Fig 1).

Construction of yeast strains to investigate a synthetic carotenoid production pathway
We evaluated our strategy by investigating a synthetic carotenoid pathway in yeast (S. cerevisiae), an industrially important chassis in biotechnology, which includes two membrane-interacting enzymes that are challenging to investigate in vitro (12,16,17,21): phytoene synthase (CrtB) and phytoene desaturase (CrtI).Phytoene synthase, the first enzyme in the carotenoid pathway and considered the bottleneck of carotenoid biosynthesis in plants (21), condenses two molecules of geranylgeranyl pyrophosphate (GGPP) head-to-head to form phytoene.Phytoene synthases have proved difficult to express in soluble and active forms in microorganisms (12,15,21).In vitro rates and affinity constants have been determined for both plant and bacterial phytoene synthases but these parameters may have been altered by the conditions of the in vitro assays-adding detergents for the bacterial enzyme and co-expressing a GGPP synthase or using semi-crude extracts for the plant enzyme (12,15,(21)(22)(23)(24).Following this enzymatic step, three insaturations are introduced in phytoene to produce lycopene.In non-photosynthetic bacteria and fungi, three reaction steps are catalyzed by a single enzyme (CrtI in bacteria and CarB in fungi) (25), whereas in plants and cyanobacteria, four different enzymes are involved in the conversion of phytoene to lycopene (26).
Given that these desaturation steps are a major bottleneck in microbial carotenoid biosynthesis, determining in vivo enzymatic constants is an industrially relevant challenge (10,22,23).
To analyze these enzymes, we assembled a set of strains covering a broad range of substrate concentrations (Fig 2 and Table EV1).To create the collection of strains with different intracellular levels of GGPP, the substrate of phytoene synthase, we first increased GGPP production by modifying the native terpene pathway: strong constitutional expression of ERG20 and of the truncated form of HMG1 and deletion of the main phosphatase gene involved in GGPP and FPP pyrophosphate group removal (DPP1), yielding the strain yENZ15 (29,30).Three different integration cassettes containing different microbial GGPP synthases (CrtE) expressed under different constitutive promoters were integrated into the yENZ15 genome, yielding a series of strains designed to produce a wide range of

GGPP concentrations (Fig 2).
We then verified that the intracellular GGPP concentration was indeed modulated in vivo, as done previously (24,(32)(33)(34).GGPP concentration varied by a factor of 38 in these strains (Fig 3B ), and the strains all had similar growth rates (0.40±0.03 h -1 In vivo enzymatic parameters of phytoene synthase from Pantoea ananas We used the set of strains with different intracellular levels of GGPP (  EV2), indicating that PaCrtB activity (V ୫ୟ୶ ୡୣ୪୪ ) was also similar in all strains.Precise estimates of V ୫ୟ୶ ୡୣ୪୪ (849±71 µM/h, rsd = 13 %) and ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ (19±3 µM, rsd = 14 %) were obtained by fitting these data using an irreversible Michaelis-Menten model (consistent with previous reports (21) and with the very negative ΔG 0 value of this reaction) (Table EV3).As expected from the simulations, parameters could not be determined precisely from the two other datasets (Table EV3).The affinity of PaCrtB for GGPP estimated from our in vivo data is higher than measured in vitro (K M = 41 µM) , Table EV4) (15).
The theoretical maximum phytoene production flux in these strains is therefore 849±71 µM/h.The bottleneck for carotenoid biosynthesis in plants is phytoene synthase, whose activity is regulated by a combination of transcriptional, post-transcriptional, and post-translational mechanisms to adjust carotenoid production (35).In our synthetic system however, our results suggest that GGPP biosynthesis is the limiting step since five copies of the GGPP synthase gene are required to achieve just 51 % of the maximum phytoene production flux, due to the lack of phytoene synthase saturation.

Straightforward extension to analyze downstream enzymes of the carotenoid pathway:
phytoene desaturases As for the investigation of phytoene synthase, the expression of the phytoene desaturases was controlled by different promoters: TDH3p for PaCrtI, PGK1p for XdCrtI, and PGI1p for BtCrtI.These promoters were chosen based on the relative lycopene producing abilities of the different CrtIs in S.
These results highlight the versatility of our approach to compare the in vivo behavior of enzymes from different organisms, using the same set of strains, and provide valuable information to understand and optimize natural and synthetic pathways in vivo.

Discussion
Inspired by conventional in vitro enzyme assays, we developed an innovative approach to measure enzyme kinetics in vivo.We successfully used the proposed method to estimate the in vivo equivalents of Michaelis-Menten parameters for a phytoene synthase and two phytoene desaturases in S. cerevisiae.
These results highlight the reliability and versatility of our in vivo enzymology approach, which hinges on solving two methodological challenges: 1) controlling the substrate pool over a broad concentration range, and 2) measuring the total concentrations of substrate, product and enzyme.
The first challenge was addressed using synthetic biology tools to modulate the concentration of a substrate-producing enzyme (here GGPP synthase, CrtE).The second challenge was overcome thank s to sensitive, quantitative metabolomics and proteomics techniques that can be generalized to various other enzymatic models.Indeed, the coverage of the metabolome and fluxome by current omics approaches is now high and is continuously increasing, which enables the application of the proposed approach to a broad range of enzymes.In this study, we quantified three basic enzymatic parameters ‫ܭ(‬ ଵ/ଶ ୡୣ୪୪ , ݇ ୡୟ୲ ୡୣ୪୪ and ܸ ௫ ).In the absence of absolute quantitative proteomics data, the proposed method still allows for the determination of ܸ ௫ and ‫ܭ‬ ଵ/ଶ values, providing sufficient information for most enzymology and metabolic engineering studies.Similarly, estimating the absolute ‫ܭ‬ ଵ/ଶ values can be achieved using only relative flux values, without requiring absolute concentration of enzymes.
We demonstrate that substrate concentrations can be varied in vivo over a wide range.Here, the GGPP concentration was varied by a factor 167 and the phytoene concentration by a factor 430, allowing clear enzyme saturation curves to be observed for the two fungal phytoene desaturases (XdCrtI and BtCrtI).However, while the range of substrate concentrations explored was significantly wider than those used in in vitro studies of the same enzymes (  (36).In the case of the two P.
ananas enzymes, the mechanisms that cause the absence of saturation may be different.For PaCrtI, the observed concentrations of phytoene (substrate) and lycopene (product) are either equivalent to or lower than those obtained with the two fungal CrtI enzymes.This suggests a true difference in the affinity for phytoene between the bacterial and fungal enzymes.For PaCrtB, given that the GGPP concentration in the cell is high (around 20 µM), the lack of complete saturation is more likely due to a heterogeneous distribution of GGPP within the cell, with only a fraction of GGPP being in the vicinity of the enzyme.Dedicated studies will be required to clarify the effects of cell compartments on enzyme efficiency.
In addition, other explanations for the difficulty in reaching complete saturation merit attention as they also convey general hypotheses on metabolic systems: i) substrate or product toxicity detrimental to cell growth (e.g. the specific growth rate is nearly 50 % lower for lycopene concentrations above 400 µM, (Fig EV2B ), ii) overflow mechanisms that relocate some of the substrate to a different cell compartment or even outside the cell (37,38), iii) the presence of alternative pathways that divert additional substrate produced above a certain concentration threshold (39,40), or iv) intrinsic properties of metabolic systems whereby production fluxes tend to decrease when product concentrations increase (41)(42)(43).While these mechanisms contribute to global metabolite homeostasis (44)(45)(46) and are essential for cell viability, they limit substrate accumulation and thus could prevent complete saturation of the enzyme.Moreover, it is crucial to bear in mind that, similarly to in vitro data, enzymatic parameters are only valid under the conditions used to measure them.We therefore argue that enzymatic parameters should not be considered constants since they depend on the microenvironment (pH, ion concentration, temperature, membrane composition, etc.), be that in vitro or in vivo.Still, as mentioned below, these parameters are useful for pathway engineering and an advantage of our in vivo enzymology method is that kinetics parameters are measured under the exact same conditions as the enzyme is expressed.
Our approach enables the determination of whether a given enzyme is operating at saturation in vivo under different cellular conditions.The tested enzymes displayed various saturation profiles, indicating that they may function under diverse conditions within the cell.Operating at full saturation (far above the ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ , the enzyme works at its maximum rate, being unaffected by changes in substrate concentration or minor environmental variations.This makes the enzymatic reaction highly robust in terms of product formation.Conversely, if an enzyme operates at substrate concentrations much lower than the ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ , any variation in substrate concentration will directly affect the production rate.This could maintain substrate homeostasis, potentially prevent toxic effects due to concentration changes.From a metabolic engineering perspective, achieving the optimal balance between high product formation and metabolic homeostasis is crucial.This balance can be attained by adjusting the levels of both the substrate-forming enzyme and the target enzyme.This process needs to be repeated for each enzyme, potentially under different saturation regimes.In biotechnology, our in vivo enzymology approach may thus guide metabolic engineering strategies to ensure that the overall pathway maintains the desired balance between production and cellular homeostasis, thereby ensuring greater stability and robustness of the engineered microbial strains at maximal production flux. The final step in our in vivo enzymology method involves fitting experimental data with a mathematical model of enzyme kinetics to obtain the corresponding parameters, the same approach as used in vitro.Traditional enzymatic models were derived with in vitro measurements and assumptions in mind.For instance, the classical Michaelis-Menten relationship applies only at steadystate, a condition clearly met in our experimental setup where all data were collected during exponential growth (i.e.metabolite concentrations and fluxes remain constant over time).However, other assumptions do not hold in vivo, in particular regarding the product concentration, which cannot be zero.Nevertheless, for the enzymes investigated here, where the catalyzed reactions are essentially irreversible, the reverse reaction can be neglected, and the Michaelis-Menten formalism still applies.Another important assumption is that the substrate concentration must be higher than the enzyme concentration.In this study, the in vivo substrate/enzyme ratio was between 8 and 1400 for PaCrtB and between 2 and 2500 for XdCrtI (Table EV5).Surprisingly, this criterion was not met for CrtB and CrtI in previous in vitro assays (19,37), with substrate/enzyme ratios between 0.05 and 0.18 for PaCrtB and between 1 and 2.57 for PaCrtI (Table EV4).In our setup, despite some formal assumptions not being fully satisfied, the data were still satisfactorily fit with a Michaelis-Menten equation.This underscores the applicability of our method, from which informative parameters such as the degree of saturation can be inferred to understand in vivo enzyme function and to engineer natural and synthetic pathways for biotechnology.The availability of in vivo data about enzyme kinetics may also lead to the derivation of specific laws that account for the specificities of in vivo studies (e.g., non-negligible product concentrations).
As we advocate for the simplicity of our method, we would like to share some insights on its implementation in future studies.First, to maximize the output of genetic constructs, experimental design should be employed to compare various enzymes that catalyze the same reaction (whether mutants or from different species) or to target multiple enzymes in the same metabolic pathway, in this case phytoene synthase and phytoene desaturase.Second, the number of genetic constructs necessary to reach saturation (ideally 4 to 5) can be minimized by verifying the substrate production range early on.Third, in most cases, testing three enzyme concentrations has been sufficient to obtain a satisfactory saturation curve or at least to precisely estimate the ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ range.Lastly, improvements can be made using high-throughput methods for the construction of plasmids and strains, for growth experiments and for samples preparation steps.The development of single cell metabolomics and proteomics approaches will also significantly increase the throughput of the strain characterization step in future studies.
Sixteen years ago, in their review on enzyme function, Dagmar Ringe and Gregory Petsko wrote: "How do enzymes function in a crowded medium of low water activity, where there may be no such thing as a freely diffusing, isolated protein molecule?In vivo enzymology is the logical next step along the road that Phillips, Koshland, and their predecessors and successors have traveled so brilliantly so far" (3).The work presented here, albeit performed in the context of a synthetic metabolic pathway, touches on the difference between kinetic parameters measured in vitro and in vivo and their interpretation.We notably show how fine-tuning and balancing the expression of the substrateproducing enzyme and the enzyme under study yields datasets from which meaningful and reliable enzymatic parameters ‫ܭ(‬

ଵ/ଶ ୡୣ୪୪
, ݇ ୡୟ୲ ୡୣ୪୪ and ܸ ௫ ) can be obtained.By including additional controlled steps, this method could be applied to a wider range of variables, such as inhibition and activation parameters (by modulating the pool of regulatory metabolites).This method could also benefit from dynamic data (time-course monitoring in response to metabolic or genetic perturbations).New formalisms will be required to account for in vivo conditions, notably the presence of products, similar enzyme and substrate concentrations, local concentration variations, and molecular fluxes within the cell.Our method is particularly valuable for studying membrane-bound and multimeric enzymes, for which the purification and assay optimization steps of classical in vitro enzymology can be extremely challenging.For membrane-bound enzymes, in vivo enzymology offers a realistic environment devoid of detergents or other interferences, with natural membranes rather than liposomes.We sincerely hope that our work will stimulate further studies delving deeper into how enzymes function in their natural environment.

Plasmid construction
Plasmids and primers are listed in Tables EV6-7.Plasmid sequences and annotations are provided in Dataset EV1.The primers were synthesized by IDT (Leuven, Belgium) and the sequences of PaCrtB, PaCrtI and BtCrtI were codon optimized for yeast and synthetized by Twist Bioscience (San Francisco, California).XdCrtE and XdCrtI from Enterobacter agglomerans were amplified from pMRI34-CrtE-Gal1-10-HMG1t, YEplac195 YB/I, and pAC-BETA respectively (47)(48)(49).Sequences of the mutated versions of the TEF1 promoter TEF1mut2p, TEF1mut5p and TEF1mut7p were obtained from Nevoigt et al. (50).Polymerase chain reaction (PCR) was performed using Phusion high fidelity polymerase and Phire Hot start II DNA polymerase (ThermoFisher Scientific, Lithuania).DNA fragments were purified using Monarch DNA Gel Extraction Kit from New England Biolabs.DNA fragments were annealed by isothermal assembly using NEBuilder HiFi assembly kit from New Englands Biolabs.
Clones and plasmids were propagated in homemade calcium-and TOP10-competent Escherichia coli cells.

Construction of yeast strains
All yeast strains used in this study are derived from CEN.PK2-1C and are listed in Table EV1.Yeasts were transformed using Gietz et al.'s high-efficiency transformation protocol (51).Integrative cassettes were obtained by enzyme digestion or PCR and were used without any further purification.
Strains were selected using auxotrophy markers or antibiotic resistance at a concentration of 400 µg/mL.Antibiotic resistance recycling was performed using vector pSH63 as described in the literature (52).Genome integration was verified by colony PCR using the primers listed in Table EV7.
Genomic DNA was extracted using DNA release from ThermoFisher Scientific.

Media and culture conditions
All strains were grown in modified synthetic Verduyn media containing glucose (111 mM), NH 4 Cl (75 mM), KH 2 PO 4 (22 mM), MgSO 4 (0.4 mM) and CSM (ForMedium LTD, Hunstaton, England) at pH 5.0 (53).Sterilization was performed by filtration.Fresh colonies from selective plates were precultured in 350 µL complete synthetic medium at 28 °C for 8 hours and these cells were used to inoculate cultures with a 1:5 medium:flask proportion to an initial OD 600nm of 0.002, grown at 200 rpm at 28 °C.

Carotene quantification
Samples (10 or 20 mL) of yeast culture were harvested with an OD 600nm of approximately 5,

GGPP quantification
GGPP was quantified as detailed previously (55).Briefly, 10 mL of broth was filtered through 0.45 μm Sartolon polyamide (Sartorius, Goettingen, Germany) and washed with 5 mL of fresh culture medium (without glucose).The filters were rapidly plunged into liquid nitrogen and then stored at −80 °C until extraction.Intracellular GGPP was extracted by incubating filters in closed glass tubes containing 5 mL of an isopropanol/H 2 0 NH 4 HCO 3 100 mM (50/50) mixture at 70 °C for 10 min.For absolute GGPP quantification, 50 μL of 13 C internal standard were added to each extract.Cellular extracts were cooled on ice and sonicated for 1 min.Cell debris was removed by centrifugation (5000 g, 4 °C, 5 min).Supernatants were evaporated overnight (SC110A SpeedVac Plus, ThermoFisher, Waltham, MA, USA), resuspended in 200 μL of methanol:NH 4 OH 10 mM (7:3) at pH 9.5 and stored at −80 °C until analysis.
Analyses were carried out on a LC-MS platform composed of a Thermo Scientific Vanquish Focused UHPLC Plus system with DAD HL, coupled to a Thermo Scientific Q Exactive Plus hybrid quadrupole-Orbitrap mass spectrometer (ThermoFisher), as detailed previously (55).Calibration mixture s (prepared at concentrations from 0.08 nM to 10 μM) were used to construct calibration curves from which the absolute concentration of GGPP in the samples was determined.

Western blot
Protein extracts were prepared as described by Zhang et al. (46).Briefly, 1.5 OD 600nm of pelleted cells were pre-treated with 100YµL of a 2YM lithium acetate cold solution, and left to stand for 5Ymin, followed by 5Ymin centrifugation at 5000Yg, 4Y°C.The supernatant was discarded and 100YµL of a 0.4YM solution of NaOH was added.After gentle resuspension, and 5Ymin standing on ice, the samples were centrifuged for 5Ymin at 4Y°C.        Figure EV4.Correlation between cell/mL and OD 600nm (A) and between OD 600nm and mg DCW (B).

Expanded View
Expanded View Table legends Table EV1.Yeast strains used in this study.
Table EV3.Enzymatic parameters measured in vivo.
Table EV4.In vitro parameters of bacterial phytoene synthase and phytoene desaturase.
Table EV5.Molecule number of each component for the in vivo enzymology reaction.
Table EV6.Plasmids used in this study.
Table EV7.Primers used for colony PCRs.
Table EV8.Peptides and PRM transitions used for absolute quantification of the enzymes.

Fig 3 )
Fig 4B.Simulation results suggest that excessively high enzyme concentrations would hinder substrate saturation, rendering parameter estimation impossible.In contrast, an extremely low enzyme concentration would lead to a low production flux, making precise measurements difficult and reducing the chance of obtaining accurate enzyme parameters.Therefore, we decided to perform our saturation experiments using three enzyme concentrations by expressing P. ananas crtB under the control of three different constitutive promoters (low expression: TEF1mut2p, medium expression: PGI1p and high expression: PDC1p).

( 21 ) 1 ,
, highlighting the necessity of measuring enzyme parameters directly within the intracellular environment.Meanwhile, the ݇ ୡୟ୲ ୡୣ୪୪ of 6±1 s obtained from absolute quantitative proteomic s measurements of the enzyme concentration (42 nM) in strains expressing PaCrtB under the control of PGI1p, is comparable to the one obtained in vitro for CrtB in Pantoea agglomerans (14 s -1 caused cell death due to the formation of lycopene crystals (Fig EV2A).PaCrtI and XdCrtl levels did not vary with the phytoene concentration of the strains, but BtCrtI levels were slightly higher in the strains with the lowest phytoene concentrations (Fig EV3).The higher levels of PaCrtI compared with those of the fungal phytoene desaturases is explained in large part by the use of a TDH3 promoter.However, lycopene fluxes were 10 times lower with PaCrtl than with XdCrtl or BtCrtl (Fig 5).Phytoene levels were high enough to saturate both fungal CrtIs (Fig 5) but not PaCrtl, for which lycopene fluxe s remained low for all tested phytoene concentrations.The experimental data from the three enzymes were fitted with an irreversible Michaelis-Menten equation, from which V ୫ୟ୶ ୡୣ୪୪ and ‫ܭ‬ ଵ/ଶ ୡୣ୪୪ values were successfully obtained for XdCrtl and BtCrtl but not for PaCrtl, for which only lower limits for V ୫ୟ୶ ୡୣ୪୪ (13 µM•h -1

centrifuged, and washed with 1
mL of MilliQ water.Cell pellets were freeze-dried and stored at −80 °C until extracted.β-apocarotenal solution (40 µL, 50 µM) was added to the dried cells, and carotenes were extracted with glass beads and 500 μL of acetone in three 20 s rounds of agitation at 0.05 m/s with a FastPrep FP120 cell disruptor (ThermoFisher).The acetone phase was transferred to a new tube and the extraction was repeated twice.Acetone extracts were pooled, centrifuged, dried under nitrogen flux, and resuspended in acetone for HPLC analysis.Analyses were carried out on a Thermo Scientific Vanquish Focused UHPLC Plus system with DAD HL.Extract samples (5 μL) were injected into a YMC carotenoid column (100 × 2.0 mm and 3 μm particle size) equipped with a precolumn (100 × 2.0 mm and 3 μm particle size).The mobile phases used to separate and quantify phytoene, lycopene and β-apocarotenal from ergosterol and derivatives were mixtures of (A) methanol/water (95:5) and (B) dichloromethane.The flow was 0.25 mL/min with the following gradient: 0-0.1 min 5 % B, 0.1-0.5 min 20 % B, 0.5-2 min 60 % B, 2-5 min 80 % B, 5-8 min 80 % B and 8-11 min 5 % B. The absorbance from 210 to 600 nm was measured throughout the run with a data collection rate of 2 Hz and a response time of 2 s.The phytoene concentration was deduced from its absorbance at 286 nm and lycopene and β-apocarotenal concentrations from the absorbance at 478 nm.The reference wavelength (600 nm) was subtracted from each of the wavelengths used for metabolite quantification.Flux calculationPhytoene and lycopene are produced and accumulate in the cells, and their pools are continuously diluted by cell growth.Assuming an absence of degradation or reutilization of these end-products by the cell, phytoene and lycopene production fluxes are balanced solely by their dilution fluxes in the exponential growth phase, where cells are at metabolic steady-state.Thus, phytoene and lycopene production fluxes were determined by multiplying their concentrations by the cell growth rate.This flux calculation method provides results consistent with those obtained by targeted 13 C-fluxomics(54).

Figure 1 .
Figure 1.General principle of the proposed strategy for in vivo enzymology and comparison between in vitro and in vivo enzymology approaches.

Figure 2 .
Figure 2. Scheme of the construction of yeast strains with different GGPP concentration.Red filled dots represent a copy of GGPP synthase, and empty dots represent an empty integration cassette.

Figure 3 .
Figure 3. Construction of the set of yeast strains with different intracellular GGPP concentrations.

Figure EV1 .
Figure EV1.PaCrtB-FLAG expression determined by western blot in yeast strains with different GGPP concentrations (panels A, B and C correspond to three different biological replicates).
(22,23,25,26)-consuming tasks in the proposed in vivo enzymology protocol are the molecular framework to investigate the in vivo kinetics of the next metabolic step with various phytoene desaturases.To demonstrate this point, we chose three widely used microbial lycopene-forming phytoene desaturases (1.3.99.31)(22,23,25,26):P. ananas Crtl (P21685, PaCrtI), which, along with Rubrivivax gelatinosus Crtl (16,21)4)(16,21), complete saturation was not achieved for PaCrtB or PaCrtI.It is conceivable that, when these enzymes are bound to natural membranes, their apparent affinity for the substrate could differ from the in vitro environment.A second explanation concerns substrate availability.While classical in vitro enzyme assays involve homogeneous, highly diluted buffers, the intracellular environment is dense, heterogeneous and compartmentalized.In this first in vivo approach, variations in substrate concentrations between compartments were not considered explicitly, though it may have affected the values obtained for the parameters.For example, a higher concentration of phytoene inside lipid droplets compared with the rest of the membrane would reduce phytoene concentrations around CrtI enzymes and would thus limit substrate saturation.Enzyme localization could also affect measurements as conditions (pH, concentration of ions, etc.) vary between compartments.Iwata-Reuyl et al. found for instance that the measured activity of PaCrtB is 2,000 times lower in the absence of detergents (15), and Fournié et Truan observed that different heterologous CrtI expression systems produced different phytoene saturation patterns