Ultralight ultrafast enzymes

Inorganic materials depleted of heavy stable isotopes are known to deviate strongly in some physico-chemical properties from their isotopically natural (native) counterparts; however, in biotechnology such effects have not been investigated yet. Here we explored for the rst time the effect of simultaneous depletion of the heavy carbon, hydrogen, oxygen and nitrogen isotopes on the bacterium E. coli and the enzymes expressed in it. Bacteria showed faster growth, with proteins exhibiting higher thermal stability, while for recombinant enzymes expressed in ultralight media, faster kinetics was discovered. At room temperature, luciferase, thioredoxin and dihydrofolate reductase showed a 40-250% increase in activity compared to the native counterparts. The eciency of ultralight Pfu DNA polymerase in polymerase chain reaction was also signicantly higher than that of the normal enzyme. At 10 °C, the advantage factor of ultralight enzymes typically increased by 50%, which points towards the reduction in structural entropy as the main factor explaining the kinetic effect of heavy isotope depletion. Ultralight enzymes may nd an application where extreme reaction rates are required.


Introduction
Isotopes are atoms of the same element that share the same atomic number, but possess different number of neutrons, which is translated in a mass difference. Most elements have several stable isotopes (e.g., carbon has two and oxygen -three stable isotopes), and thus almost all natural compounds have a polyisotopic composition. The question to what extent the isotopic composition plays a role in physical, chemical and biological processes has interested scientists since the discovery of stable isotopes by Aston in 1919. To answer this question best, pure isotopes of different elements have been isolated and studied. The materials and compounds in which the heavier, less abundant isotopes have been signi cantly depleted are called here ultralight or (near-)monoisotopic. In solid state physics, it has been found that near-monoisotopic materials possess strongly deviating properties: e.g., silicon 28 Si (99.87%) has a 60% higher thermal conductivity at 80K than silicon with natural (≈92% 28 Si) isotopic composition 1 . Qualitatively similar results have been obtained for monoisotopic diamond (99.8% 12 C) 2 and gallium arsenide. Strikingly, the thermal conductivity of germanium enriched to 99.99% 70 Ge increased up to 8 times compared to the natural isotopic composition, in which the share of 70 Ge among the ve stable gallium isotopes is 21.2% 3 . Today, monoisotopic 28 Si that is free of nuclear spins causing decoherence in a quantum system emerges as a critical material in the development of quantum information devices, such as quantum computers 3 .
On the molecular level, monoisotopic species can exhibit different kinetics of formation than their polyisotopic counterparts. In the reaction of ozone O 3 synthesis from molecular oxygen O 2 by electrical discharge, the isotope effects alter the reaction rates by typically 10% to 20% depending on the pressure and temperature of the gas in which ozone is formed 4 . The strictly monoisotopic isotopomers 16  In biological systems, each of the four most abundant elements (C, H, O and N) has more than one stable isotope, the lighter one being the most abundant (e.g., the share of 12 C is 98.9%). Microbes, plants and animals, including mammals, can grow in an environment with signi cantly altered ratios of stable isotopes compared to natural, but the phenotype of the organisms can be signi cantly affected [7][8][9][10][11][12][13][14] . The researchers concluded that the "organisms of different isotopic compositions are actually different organisms, to the degree that their isotopic compositions are removed from naturally occurring compositions" 10 .
Among the studies in which the heavy isotopes have been depleted, most reported effects relate to deuterium depletion in water. A major deuterium-depleted water (DDW) phenomenon is the depressed growth of cancer cells 15,16 , which is currently being exploited in a clinical trial 17,18 . Recently, while studying the antiproliferation effect of DDW in human lung adenocarcinoma cells, we found that DDW induces mitochondrial redox imbalance which leads to oxidative stress 19 . In general, deuterium concentration around the natural value of 150 ppm is found to be a cell growth regulator 20 .
It is the easiest to manipulate the isotopic composition of bacteria and yeast, as some strains of these can grow on minimal media composed of water, inorganic salts and a simple organic compound (e.g., an amino acid, glucose or carboxylic acid) as a source of carbon. For instance, the effect of depletion of 13 C on E. coli bacteria growth has been reported 21 .
Simultaneous depletion of several heavy stable isotopes in growth media leading to nearly-monoisotopic organisms have been used for decades to produce near-monoisotopic bacterial proteins for structural studies, e.g. with high-resolution mass spectrometry 22,23 . However, the effect of the ultralight composition on organism's phenotype has not been reported yet to the best of our knowledge. Having acquired considerable experience in measuring the parameters of bacterial growth in various isotopic environments 21, 24-26 , we decided to investigate the effect of the depleted media on the growth and phenotype of E. coli. For this, we formulated a M9 minimum media based on 13 C-depleted glucose and 15 N-depleted salt dissolved in D, 18 O-depleted water (Depleted media). As a control, the growth of E. coli in isotopically natural media (Normal media) was monitored. Four different enzymes were recombinantly produced in Depleted as well as Normal media, and their activity was carefully measured. We found much bigger effects than previously reported for isotope phenomena in molecules, including the kinetics of ozone formation.

Results
The growth of E. coli BL21. The values of lag time, maximum growth rate and maximum density were used 21 for comparison of the E. coli growth in Normal and Depleted media (Fig. 1a). All three parameters indicated that the Depleted media was more bene cial for bacterial growth, providing a signi cantly (all p-values <<10 −6 , n = 21) higher maximum density, faster growth rate as well as a shorter lag time (Fig. 1b).
LC-MS analysis of bacterial lysate showed that in most proteins with MW <20 kDa the monoisotopic mass dominates in the isotopic distribution of molecular ions (Fig. 1c). Fourier Transform Isotopic Ratio Mass spectrometry (FT IsoR MS) of the peptide digest obtained from E. coli lysate that is based on the analysis of the " ne structure" of fragment immonium ions of amino acid residues proved deep depletion of 13 C, signi cant depletion of 15 N as well as a moderate reduction in deuterium in bacterial proteins (Fig. 1d).
Dose dependence in Depleted media dilution. When the Depleted media was diluted by the Normal media to below 80%, the maximum density of E. coli bacteria started to deviate signi cantly from that in pure Depleted media (Fig. 1e), while maximum growth rate was affected only at a dilution below 40% ( Supplementary Fig. 1).
Thermal stability of bacterial proteins. Thermal proteome pro ling is rapidly becoming one of the standard methods of proteome analysis 27 . Brie y, the cellular or bacterial lysate undergoes a 3 min incubation at a set of xed temperature points ranging from 37°C to 83°C, after which the denatured proteins that lost their solubility are spun down and the concentration of each remaining protein in the supernatant is measured by proteomics. The resultant data are tted with a sigmoidal curve (Fig. 2a), and the middle point provides the melting temperature T m . When two growth conditions are compared, the melting temperature difference ∆T m is considered.
The distribution of E. coli ∆Tm values (Fig. 2b, Supplementary Table 1) reveals that more proteins increase their thermal stability upon becoming ultralight than decrease this stability. The median melting temperature of such proteins was ∆T m = 1.0±0.1°C higher than that of the proteins obtained in Normal media ( Supplementary Fig. 2).
We also noticed that the melting curve of many ultralight proteins shows steeper decline with temperature than their isotopically natural counterparts. To test this observation, the melting curves were tted with the equation: (1) where the term B is responsible for the slope steepness (lower values mean steeper curve) and Pl is plateau. Of the 316 proteins found with signi cantly (p<0.05) altered slope, 239 proteins (76%) had a steeper slope in the ultralight form (Fig. 2c, Supplementary Table 2). The median slope of ultralight proteins was 27% higher than that of the proteins in Normal media (Fig. 2d).
Luciferase YY5. We expressed recombinantly luciferase YY5 28 in E. coli grown in both Normal and Depleted media. In order to ensure validity of the comparison, a single peak in size-exclusion chromatography (SEC) was isolated ( Supplementary Fig. 3a) possessing activity. However, mass spectrometry (MS) analysis showed that the isolated peak contains at least two molecules with different molecular weights. Therefore, an additional puri cation step using strong cation exchange (SCX) chromatography was implemented, which gave two well-separated fractions, fracI and fracII ( Supplementary Fig. 2b). Both fractions, especially fracII, showed strong luciferase activity, and MS analysis con rmed that fracII contains just one molecular specie. Tandem MS analysis of that fraction provided isotopic distribution of the y 87 12+ backbone fragment, with the monoisotopic mass dominant in the Ultralight sample, consistent with strong depletion of heavy isotopes in that sample (Fig. 3a). Charge deconvolution provided the average isotopic molecular mass of 61,728.9±0.3 Da of the Normal luciferase in fracII, which was by 240.8±0.3 Da higher than expected ( Supplementary Fig. 3e). The difference was attributed to a combination of biotinylation and methylation, which are common post-translational modi cations (PTMs) in recombinant proteins 29 (Fig. 3d), which was a similar effect to that observed for E. coli growth parameters (Fig. 1e).
The fraction fracI turned out to be somewhat heterogeneous, with the main component corresponding to a combination of double biotinylation and methylation (Supplementary Fig. 3c and 3d). This fraction showed an order of magnitude lower activity than fracII (Supplementary Fig. 3f and 3g). It also demonstrated a signi cant Ultralight/Normal ratio in maximum reaction rate (1.7±0.2, P=2×10 −5 ), and a small reduction in K m (0.87±0.05, P=0.02) ( Supplementary Fig. 3h).
To test the refolding ability of the enzyme after thermal unfolding, the fracII luciferase samples were incubated at 50°C for up to 100 min and then cooled down to RT. As expected, the enzyme activity has dropped, but at all tested time points, the Ultralight enzyme demonstrated a 1.6±0.1 times higher activity than the Normal enzyme (Fig. 3e). Importantly, at lower temperature than RT the ratio of Ultralight/Normal activity for untreated fracII enzyme was increasing, reaching 2.3±0.1 times at 10°C (Fig. 3f).
Thioredoxin. Another tested enzyme was thioredoxin (Trx), an oxidation-reduction active enzyme containing a cystine disul de that is reduced to the dithiol cysteine form by thioredoxin reductase (TrxR) with NADPH as a cofactor 31 . The reduced Trx can oxidize again by reducing its substrate, such as insulin that is composed of peptide chains A and B linked together by two disul de bonds. Reduction of insulin in solution results in white precipitation originating from the free B chain, which is used for measuring Trx activity. Trx ( Supplementary Fig. 4a-c) was expressed recombinantly in E. coli grown in both Normal and Depleted media, puri ed and quanti ed by SEC ( Supplementary Fig. 4d), with the purity veri ed by mass spectrometry (Fig. 4a). Monoisotopic mass dominates in the isotopic distribution of molecular ions in the Ultralight sample, which proved strong depletion of the heavy isotopes in that sample ( Supplementary  Fig. 4e). The monoisotopic mass of the Ultralight protein was 11,784.9±0.1 Da, which was by 1.9±0.1 Da lower than expected. The difference was attributed to a disul de linkage between two cysteine residues, which is a frequent post-translational modi cation in recombinant proteins 32 . The depletion of Ultralight Trx was 65%.
The activity of thioredoxin was measured for different concentrations of the substrate insulin ( Fig. 4c) by tting the Michaelis-Menten equation (Fig. 4d). The maximum reaction rate of the Ultralight thioredoxin was found to be (17±1)% higher than the Normal one (P=0.002, n=3), while the K m value was not signi cantly altered.
The activity of enzyme was also measured at different temperatures in the range 15-40°C. As expected, the enzyme activity dropped at lower temperature, but the Ultralight/Normal activity ratio increased to 2.0±0.3 at 15°C (Fig. 4e).
Dihydrofolate reductase. Dihydrofolate Reductase (DHFR) is a ubiquitous enzyme that is present in all eukaryotic and prokaryotic cells. DHFR catalyzes the transfer of a hydride from NADPH to dihydrofolate with an accompanying protonation to produce tetrahydrofolate, during which NADPH is oxidized to NADP + . By monitoring the decrease in absorbance at 340 nm, at which NADPH has speci c absorbance, the ability of DHFR to catalyze the oxidation of NADPH can be measured. DHFR from Geobacillus stearothermophilus (BsDHFR, Supplementary Fig. 5a and 5b) was expressed recombinantly in E. coli grown in both Normal and Depleted media, puri ed and quanti ed by SEC ( Supplementary Fig. 5c), with the purity veri ed by mass spectrometry (Fig. 5a). Monoisotopic mass dominates in the isotopic distribution of molecular ions in the Ultralight sample, while charge and isotope deconvolution of the Normal BsDHFR mass spectrum provided the average molecular mass of 33820.4±0.2 Da, which was only 0.2 Da lower than expected and within the experimental error. The average mass of the Ultralight protein revealed the isotopic shift of 5.0±0.2 Da ( Supplementary Fig. 5d), consistent with the depletion degree of 76%.
The activity of BsDHFR was measured in the presence of 200 µM DHF and 100 µM NADPH. The decrease in 340 nm UV absorbance for the reaction catalyzed by Ultralight BsDHFR was signi cantly faster compared with the Normal enzyme ( Fig. 5b), which was re ected in a steeper decreasing curve for Ultralight BsDHFR. The tted Michaelis-Menten equation (Fig. 5c) revealed a (70±2)% higher maximum reaction rate for Depleted vs Normal BsDHFR (P=10 −5 , n=4), while the K m value was again not signi cantly changed.
Similar to other tested enzymes, the activity of BsDHFR measured in the range 10-30°C dropped at lower temperature, but the Ultralight/Normal activity ratio increased, reaching 2.1±0.2 at 10°C (Fig. 5e).
Pfu DNA polymerase. Polymerase chain reaction (PCR) proved to be one of the biggest revolutions in molecular biology. Due to its extremely high sensitivity and dynamic range, the potential for high throughout as well as semi-quantitative nature, PCR is now widely used in research, clinical diagnostics, etc. 33 For example, viral nucleic acid detection by real time (RT)-PCR remains the gold standard of current diagnostic tests for the SARS-CoV-2 infection 34 . Although PCR is a mature technology, research is still ongoing to obtain more e cient and DNA polymerase less prone to sequence errors 35 . Pfu DNA polymerase 36 is a high-delity, thermostable enzyme isolated from Pyrococcus furiosus. In this study, Pfu was expressed recombinantly in E. coli grown in both Normal and Depleted media, puri ed and quanti ed by SEC ( Supplementary Fig. 6a), with the purity veri ed by mass spectrometry (Fig. 6a). In the Ultralight sample, monoisotopic mass dominates in the isotopic distribution of b 99 20+ ions, which con rms strong depletion of heavy isotopes (Fig. 6a). Charge deconvolution of the Normal Pfu yielded the average isotopic molecular mass of 100,194.3±0.6 Da, which was by 43.4±0.6 Da higher than expected ( Supplementary Fig. 6b). The difference was attributed to acetylation, possibly combination with deamidation, which both are common PTMs in recombinant proteins 29 . Taking this into account, the Ultralight sample ( Supplementary Fig. 6b) gave the isotopic shift of 12.1±0.6 Da, which corresponds to the depletion degree of 81%.
Human glyceraldehyde-3-phosphate dehydrogenase (GAPDH), one the most used reference genes in RT-PCR assays 37 , was selected as the targeted gene for activity assay. RNAs extracted from human nonsmall lung cancer cell line A549 cells was reverse transcribed to cDNAs which were used as templates for RT-PCR. With the concentration of input cDNAs 2 ng per 20 µL reaction it was found that RT-PCR catalyzed by Ultralight Pfu was signi cantly faster compared to Normal enzyme ( (92.8±2)% (Fig. 6f). Moreover, the amount of nal product in the reaction with the Ultralight enzyme was higher than that in the reaction with the Normal enzyme, which was possibly caused by the lower e ciency of the Normal Pfu.
The activity of Pfu enzyme at different extension temperatures in the range 60-72 o C was studied by detection of the uorescence intensity of SYBR green I which presented the amount of double-stranded DNA (dsDNA) produced in PCR. As expected, the enzyme activity dropped at lower temperature ( Supplementary Fig. 6c), but the Ultralight/Normal activity ratio increased, with a (54 ±5)% higher activity for Ultralight enzyme at 60°C (Fig. 6g).

Discussion
Page 8/25 The most important nding of this work is that at least some enzymes expressed in E. coli grown in the Depleted media show faster kinetics than the identical enzymes produced in the Normal media. This nding is further supported by the faster E. coli growth in the Depleted media, which necessitates faster (on average) enzymatic activity in the growing organism. The size of the heavy isotope depletion effect is large (≈2.5-fold for luciferase and at least 50% for other enzymes). It should be noted that the isotope depletion, however large in relative terms (>5-fold), is miniscule on the absolute scale, as the natural abundance of the most abundant depleted isotope, 13 C, is only 1.1%. Thus the kinetics enhancement due to the normal kinetic isotope effect in strictly monoisotopic enzymes should be <1%. This reasoning is supported by only a 13% reduction in the reaction speed of the DHFR enzyme with fully substituted heavy isotopes of 13 C, 15 N and 2 H at 40°C 38 and a 40% reduction at 5°C 39 . The kinetics acceleration by up to 250% in Ultralight proteins cannot be accounted by the conventional kinetic isotope effect, and the low Pvalues obtained (P=3⋅10 −7 for luciferase and P<10 −11 for the combined results of four enzymes) greatly exceeds the ve-sigma requirement for new discovery accepted in Physics, not to mention the proposed P<0.005 threshold for discoveries in Life Sciences 40 . Importantly, the majority of the Ultralight enzymes are more thermally stable than Normal enzymes, and the superior kinetics of the Ultralight enzymes is preserved even at higher temperatures.
A simple rationalization of the effect of ultralight enzymes is the following. Due to the dominance in nature of light isotopes, every carbon atom is most likely 12 C, every oxygen atom is most likely 16 O, etc. As the amino acid sequences of most enzymes are tightly optimized for function by natural selection, the most likely con guration of a native enzyme should have close to maximum activity. Any deviation from that con guration, such as those arising due to heavy isotopic substitution, is likely reducing this optimal activity, however slightly. This qualitative reasoning explains why monoisotopic enzymes should have close to maximum activity for a given amino acid sequence.
For a more involved and potentially quantitative explanation of this phenomenon, we considered the role of entropy in the magic of enzymatic action, which is the enormous rate enhancement achieved by enzymes compared to the equivalent enzyme-free reactions. This role has been debated for over a century, and the debates are still ongoing 41,42 . There are several well-documented cases where the activation free energy decreases due to enzyme's entropic effects by around 10 kcal/mol, leading to a rate enhancement of ∼10 7 compared to the uncatalyzed reaction. However, despite substantial experimental and theoretical efforts, there is no real consensus on the origin of such large entropic contributions. Recent efforts have shown that the entropy considerations should not be limited to the substrate entropies alone, but must also include all the relevant activation entropy parts, encompassing the whole protein and its surrounding solvent 41 . The presence of heavy isotopes contributes signi cantly to the informational protein entropy, with the monoisotopic molecules having the lowest entropy 43 . Indeed, a single 13 C atom in a protein molecule containing 10,000 carbon atoms (as in a ≈70 kDa enzyme) will have 10,000 different places to reside in, which immediately results in 10,000 distinct isotopomers. At a 1.1% abundance, there will on average be ca. 110 13 C atoms in such a molecule, giving a very large number of isotopologues 44 differing from each other by the number of 13 C atoms. With 2 H, 15 N, 17 O and protein isotopomer will have its own spectrum of vibrational frequencies distinctly different from other isotopomers 46 , vibrational entropy of a poly-isotopic protein could greatly exceed that of the corresponding monoisotopic molecule or of a ultralight protein with much fewer heavy isotopes than natural molecule. Vibrational entropy is known to account for the stability of short alpha-helices 47 and contributes signi cantly to the stability of globular proteins 48 . In general, vibrations of proteins, especially those local to the active site, play a signi cant role in the enzymatic activity 49 . Thus it stands to reason that the protein molecules with much lower vibrational entropy, such as monoisotopic or ultralight enzymes, will exhibit signi cantly altered kinetic parameters than the same enzymes with natural isotopes.
The lower intrinsic entropy due to the isotope depletion can also affect the rates of the protein thermal unfolding and refolding 50 . The thermal pro ling experiments performed here on E. coli lysate suggested that the majority of proteins increased their measured melting temperature, which could be due to a number of scenarios. One possibility is a combination of an unchanged or slightly increased unfolding rate with faster refolding. In favor of the latter scenario, for the same number of amino acid residues the proteins with lower conformational entropy tend to fold faster 50 .
Among other entropy-related explanations of the observed effects is the Isotopic Resonance (IsoRes) phenomenon 21,24,25 . Being a generalization of the ideas of Gao and Marcus, the IsoRes paradigm predicts every type of symmetry (de ned as a condition leading to a reduction in the number of quantummechanical states compared to an asymmetric case) to affect reaction kinetics. At certain isotopic compositions, often corresponding to signi cant heavy isotope enrichments compared to natural isotopic abundances, IsoRes predicted faster kinetics, which has indeed been observed experimentally 21 . The strongest kinetic effect is however predicted by IsoRes for monoisotopic systems, in agreement with the above observations.

Media Preparation
M9 minimum media were prepared as follows. Na 2 HPO 4 .2H 2 O, KH 2 PO 4 , NaCl, NH 4 Cl, MgSO 4 , CaCl 2 and glucose were obtained from Sigma-Aldrich. 13 15.0 g of KH 2 PO 4 and 2.5 g of NaCl in 1000 mL of either normal milli-Q water (for Normal M9 media) or DDW (for Depleted M9 media). Both solutions were autoclaved before proceeding to M9 media preparation. M9 media were prepared by mixing 800 mL of either normal milli-Q water (for Normal M9 media) or DDW (for Depleted M9 media), 200 mL of the corresponding M9 5xSS, 2.0 mL of 1M MgSO 4 , 0.1 mL of 1M CaCl 2 , 5 g of glucose and 1 g of (NH 4 ) 2 SO 4 (isotopically normal for Normal M9 media and 13 C-depleted glucose and 15 N-depleted (NH 4 ) 2 SO 4 for Depleted M9 media). These media were ltered using 0.2 µm polyether sulfone (PES) lters (VWR) before use.

E. coli growth
For each set of experiments, E. coli BL 21 strain was grown on plates containing LB agar (Sigma). A single colony was transferred to a tube containing 6 mL of Normal M9 media and grown for 36 h. These bacteria were further diluted (1:500) into tubes containing either Normal or Depleted M9 media and put into honeycomb well plates (BioScreen, Finland) in 21 replicates. Bacterial growth was monitored by measuring light diffraction on a BioScreen C instrument (BioScreen, Finland) at 37°C. TPP sample preparation E. coli BL21 grown in Normal and Depleted M9 media were diluted (1:500) into tubes containing either Normal or Depleted M9 media in asks. After 36 h of growth, the bacteria in each ask were collected, washed, resuspended in PBS with protease inhibitor (5892791001, Sigma) and lysed by probe sonication. The protein solution was collected after centrifuge and divided into 8 aliquots. These aliquots were incubated for 3 min at either 37, 43, 49, 55, 61, 67, 73 or 79°C. After that the samples were kept at RT for 5 min to cool down and ultra-centrifuged at 35,000 rpm/min at 4°C for 30 min. Afterwards, the supernatant was collected, reduced with 10 mM DTT (10708984001, Sigma) and alkylated with 25 mM IAA (I1149, Sigma). The samples were precipitated using cold acetone at -20°C overnight, digested by Lys C (125-05061, Wako Chemicals GmbH) at a 1:75 enzyme to protein ratio for 6 h at 30°C and then by trypsin (V5111, Promega) (1:50 enzyme to protein ratio) overnight. After labeling using 16 TMTpro reagents (A44520, Thermo Fisher Scienti c) according to manufacturer's instructions, multiplexing and desalting with C18 Sep-pak columns (WAT054960, Waters), the peptides samples were fractionated using a Dionex Ultimate 3000 UPLC system (Thermo Fisher Scienti c) as described before 20 . Every 8 fractions were combined to a single pool and 12 such pools for each TMT-multiplexed sample set were analyzed by nanoLC-MS/MS as customary in shotgun proteomics.

NanoLC-MS/MS analyses were performed on an Orbitrap Fusion Lumos mass spectrometer (Thermo
Fisher Scientific). The instrument was equipped with an EASY ElectroSpray source and connected online to an Ultimate 3000 nano ow UPLC system. The samples were pre-concentrated and desalted online using a PepMap C18 nano-trap column (length -2 cm; inner diameter -75 µm; particle size -3 µm; pore size -100 Å; Thermo Fisher Scientific) with a flow rate of 3 µL/min for 5 min. Peptide separation was performed on an EASY-Spray C18 reversed-phase nano-LC column (Acclaim PepMap RSLC; length -50 cm; inner diameter -2 µm; particle size -2 µm; pore size -100 Å; Thermo Scientific) at 55 °C and a flow rate of 300 nL/min. Peptides were separated using a binary solvent system consisting of 0.1% (v/v) FA, 2% (v/v) ACN (solvent A) and 98% ACN (v/v), 0.1% (v/v) FA (solvent B). They were eluted with a gradient of 3-26% B in 97 min, and 26-95% B in 9 min. Subsequently, the analytical column was washed with 95% B for 5 min before re-equilibration with 3% B. The mass spectrometer was operated in a datadependent acquisition mode. A survey mass spectrum (from m/z 375 to 1500) was acquired in the Orbitrap analyzer at a nominal resolution of 120,000. The automatic gain control (AGC) target for was set as 100% standard, with the maximum injection time of 50 ms. The most abundant ions in charge states 2 + to 7 + were isolated in a 3 s cycle, fragmented using HCD MS/MS with 33% normalized collision energy, and detected in the Orbitrap analyzer at a nominal mass resolution of 50,000. The AGC target for MS/MS was set as 250% standard with a maximum injection time of 100 ms, whereas dynamic exclusion was set to 45 s with a 10-ppm mass window.
Expression and puri cation of luciferase YY5 Part of this work was performed by the Protein Science Facility at Karolinska Institutet, Stockholm. The construct psfLucYY5 was transformed into E. coli BL21 (DE3) T1R pRARE2 cells in Normal and Depleted M9 media containing 50 µg/mL ampicillin and 34 µg/mL chloramphenicol at 37°C. The bacterial growth was monitored by regular OD measurements. Protein expression of the fusion proteins was induced by the addition of 0.5 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG, Merck). The cells were harvested after 16 h by centrifugation, and the pellet was resuspended in lysis buffer including 100 mM HEPES, 500 mM NaCl, 10% glycerol, 10 mM imidazole, 0.5 mM TCEP, pH 8.0, 1 mg/mL lysozyme, 0.1% DDM, 1 mM MgSO 4 , protease inhibitors (complete EDTA-free 0.5 tab/mL), and benzonase (0.125 U/mL). After centrifugation, the samples were loaded onto a resin column (GE Healthcare) equilibrated with binding buffer (20 mM HEPES, 500 mM NaCl, 10% glycerol, 10 mM imidazole and 0.5 mM TCEP, at pH 7.5). The column was washed with binding buffer followed by washing buffer (same as binding buffer but with 30 mM imidazole), and then recombinant luciferase YY5 was eluted with a similar buffer containing 500 mM imidazole. Fractions containing luciferase YY5 were combined and applied to a PD 10 column (GE Healthcare) equilibrated with 50 mM Tris-HCl at pH 7.8. The size and purity of recombinant luciferase YY5 was con rmed by SDS-PAGE.

Expression and puri cation of human Trx
The Trx expression plasmid ( Supplementary Fig. 3a-c) was transformed into E. coli BL21 (DE) in Normal and Depleted M9 media containing 50 µg/mL kanamycin (Merck, K1377) and the expression and puri cation of Trx was performed as described above. Fractions containing Trx according to SDS-PAGE analysis were pooled together and TEV protease (Sigma) was added to the protein sample at a 1:30 molar ratio. The mixture was incubated at 4°C overnight. The His-tag was removed by loading the protein sample onto a 1 mL HisTrap column (GE Healthcare) pre-equilibrated with TEV cleavage reaction buffer supplemented with 20 mM imidazole. The ow-through containing cleaved Trx protein was collected and the puri cation fractions were analyzed by SDS-PAGE.

Expression and puri cation of BsDHFR
The BsDHFR expression plasmid ( Supplementary Fig. 4a and 4b) was transformed into E. coli BL21 (DE) in Normal and Depleted M9 media containing 50 µg/mL kanamycin and the expression and puri cation of BsDHFR was performed as above.

Expression and puri cation of Pfu
The construct Pfu DNA polymerase with Sso7d processivity-enhancing domain was transformed into E. coli BL21 (DE3) T1R cells in Normal and Depleted M9 media containing 50 µg/mL ampicillin and 34 µg/mL chloramphenicol. The expression and puri cation of BsDHFR was performed as above.
Size exclusion chromatography (SEC) SEC puri cation of the recombinant proteins was performed on a Dionex Ultimate 3000 UPLC system (Thermo Fisher Scienti c) with a 25 cm Waters column (2.1 mm ID; particle size 3.5 µm; pore size 300 Å) at a ow rate of 300 µL/min. Fractionation of luciferase YY5, BsDHFR and Pfu were performed using 2× phosphate-buffered saline diluted from 10× PBS (Lonza, 17517Q). Trx was puri ed using 50mM Tris-HCl at pH 7.5 containing 2 mM EDTA. The elution was monitored by UV absorbance at 280 and 214 nm.

Strong cation exchange (SCX) chromatography
Luciferase YY5 isolated with SEC was additionally puri ed by SCX using a Dionex Ultimate 3000 UPLC system (Thermo Fisher Scienti c) with ProPac Strong Cation-Exchange Column (4×250mm, Thermo Fisher Scienti c) at a ow rate of 500 µL/min. Fractionation was performed using a binary solvent system consisting of 12.5 mM sodium phosphate, pH 6.0 (solvent A) and 12.5 mM sodium phosphate, 1 M NaCl, pH 6.0 (solvent B). Elution was done with a 30 min gradient from 1-58% B monitored by UV absorbance at 280 nm. Two fractions showing luciferase activity were collected and used for further analysis.

Measurements of luciferase YY5 activity
Buffered co-factor solution (8 mL) was prepared containing 1 mM DTT, 0.2 mg/mL BSA, 5 mM of MgCl 2 , 4 mM DTT, 100 µM D -luciferin and 100 mM Tris-HCl at pH 7.8. 50 µL of each SCX fraction was mixed with 50 µL buffered co-factor solution in a 96-well microplate well and the luminescence was measured using the Tecan In nite M2000 PRO microplate reader. The luminescence of each fraction was normalized by the total 280 nm UV absorbance of the SCX fraction.
Thermal stability assay of luciferase YY5 The protein concentration of Normal and Ultralight SEC-SCX puri ed luciferase YY5 was determined by microBCA and 50 µL of 0.2 µg/mL enzyme was heated at 50°C for 5-120 min. After cooling to RT, the enzyme solution was mixed with 50 µL of the reaction buffer containing 1 mM DTT, 0.2 mg/mL BSA, 5 mM of MgSO 4 , 4 mM DTT, 100 µM D -luciferin and 100 mM Tris−HCl at pH 7.8. The luciferase activity was measured at 25°C using the Tecan In nite M2000 PRO microplate reader.

Measurement of Trx activity
The assay was performed at RT using 50 mM Tris-HCl containing 2 mM EDTA at pH 7.5 and 200 µM NADPH (Roche, 10107824001). A 60 µL aliquot of each Trx SEC fraction was mixed in a 96-well microplate well with 30 µL buffered enzyme solution containing 15 nM thioredoxin reductase and 10 µL of insulin (Sigma, I0516) at a concentration ranging from 0 to 160 µM (the latter corresponds to 1 mg/mL). NADPH consumption was measured immediately every minute by the decrease of absorbance at 340 nm during the initial linear phase of reaction in the Tecan In nite M2000 PRO microplate reader.
The magnitude of the absorbance decrease at 340 nm of each fraction was normalized by the total UV absorbance at 280 nm of the Trx SEC fraction.

Measurement of BsDHFR activity
The assay was performed at RT using 50 mM Tris-HCl at pH 7. Afterwards, the DNA fragment was visualized under UV light at 280 nm.

Bottom-up proteomics data analysis
The LC-MS/MS raw les were processed by an in-house modi ed version of MaxQuant software (version 1.6.2.3) recognizing TMTpro as an isobaric mass tag using the "Speci c Trypsin/P, Lyc/P" digestion mode with maximum two missed cleavages as described before 20 . The MS/MS spectra were searched against the Uniprot Escherichia coli BL21 database (UP000290528, containing 4,339 entries, last modi ed on January 6, 2020).

Top-down mass spectrometry analysis analysis
The SEC or SCX fractions were concentrated using a 3 kDa centrifugal lter (Merck, UFC500324). The proteins were desalted using ZipTip with 0.6 µL C 4 resin (Merck, ZTC04S096). On-line LC-MS/MS analysis was performed using a Dionex Ultimate NCS-3500RS Nano System coupled to an Orbitrap Fusion Lumos mass spectrometer equipped with a HESI Ion Source (all -Thermo Fisher Scienti c).
Protein separation online with MS was performed using a C 4 reversed-phase PepMap™ 300 micro-ow HPLC column (particle size 5 µm, 1 mm ID × 15 cm, Thermo Fisher Scienti c) enclosed in a column oven compartment maintained at 45°C, at a ow rate of 30 µL/min. 1 µg of protein sample was loaded and separated using a binary solvent system consisting of Solvent A (2% acetonitrile, 98% water and 0.1% formic acid) and Solvent B (2% water, 98% acetonitrile and 0.1% formic acid). The mobile phase composition was 5% B from 0 to 2 min, followed by a 5-30% B gradient from 2 to 5 min, 30-95% B from 5 to 25 min, 95-5% B from 34 to 35 min and 5% B from 35 to 40 min. The electrospray voltage was 3.7 kV with sheath gas at 15 units, auxiliary gas at 5 units and the vaporizer temperature at 75°C. The capillary temperature was set at 300°C. The RF lens amplitude was set at 30% and an additional 15 V voltage difference was applied within the source region to facilitate ion desolvation. All mass spectra were acquired in the positive ion mode with the pro le data type. Application mode was set at "intact protein" FT IsoR MS analysis Bacterial lysate was digested as above, and after desalting the peptides were analyzed by nanoLC-MS/MS. Mass spectra were acquired on an Orbitrap Fusion Lumos Mass in the data-independent mode with z≥2 peptides selected for MS/MS in an isolation window 1000 m/z units wide centered at m/z 800. MS/MS was performed with HCD energy set at 50 NCE. The detection range was from m/z 50 to 200, with the nominal mass resolving power of 60,000. Data processing was done using a home-written R script analyzing the ne isotopic structure of the immonium ions of amino acids Pro, Val and Leu/Ile. Figure 1 <p><strong><em>E. coli</em> growth in Normal and Depleted media. a,</strong> The growth curves of <em>E. coli</em> BL21 grown in Normal and Depleted M9 media (n=21). <strong>b</strong>, The differences in maximum density (left), maximum growth rate (middle) and lag time (right) of <em>E. coli</em> grown in Normal and Depleted media (n=21). <strong>c</strong>, Mass spectrum of the <em>E. coli </em>protein Stringent Starvation Protein A (24.1 kDa) grown in the Normal (left) and Depleted (right) media. The monoisotopic mass position is marked by square. <strong>d</strong>, FT isoR MS analysis of <sup>13</sup>C/<sup>12</sup>C<sup> </sup>(left), <sup>15</sup>N/<sup>14</sup>N (middle), <sup>2</sup>H/<sup>1</sup>H (right) in <em>E. coli</em> grown in Normal and Depleted M9 media (n=3). <strong>e</strong>, Dose-response effect of the Depleted media dilution by the Normal media on maximum density of <em>E. coli </em>bacteria (n=8). </p><p><br></p>  <p><strong>Analysis of luciferase YY5 fracII by mass spectrometry and activity comparison of Normal and Ultralight enzyme. a</strong>, Electrospray ionization mass spectra of fracII of Normal and Ultralight luciferase YY5 (left) and a backbone fragment obtained by MS/MS of molecular ions; square denotes monoisotopic mass. <strong>b</strong>, Activity curve of luciferase YY5 fracII. The luminescence readings (n=4) were normalized to the total UV absorbance in SCX chromatography before being tted with the Michaelis-Menten equation. <strong>c</strong>, Kinetic parameters Vmax (left) and Km (right) of the Normal and Ultralight luciferase YY5, fracII (n=4). <strong>d</strong>, Dose-response effect of the Depleted media dilution by the Normal media on activity of Luciferase YY5 (n=6). <strong>e</strong>, The relative activity of the Ultralight luciferase YY5 fracII compared to Normal enzyme after heating at 50°C for different time intervals and cooling to RT (n=4). <strong>f</strong>, The relative activity of the Ultralight luciferase YY5 fracII compared to Normal enzyme at different temperatures (n=4).</p> The kinetic parameters Vmax (left) and Km (right) of the Normal and Ultralight thioredoxin (n=3).

Declarations Figures
<strong>e</strong>, The relative activity of the Ultralight thioredoxin compared to Normal enzyme at different temperatures (n=4).</p> Figure 5 <p><strong>Analysis of BsDHFR by mass spectrometry and activity comparison of Normal and Ultralight enzyme. a</strong>, Electrospray ionization mass spectra of Normal and Ultralight BsDHFR (left) and a