Xylonolactonase from Caulobacter crescentus Is a Mononuclear Nonheme Iron Hydrolase

Caulobacter crescentus xylonolactonase (Cc XylC, EC 3.1.1.68) catalyzes an intramolecular ester bond hydrolysis over a nonenzymatic acid/base catalysis. Cc XylC is a member of the SMP30 protein family, whose members have previously been reported to be active in the presence of bivalent metal ions, such as Ca2+, Zn2+, and Mg2+. By native mass spectrometry, we studied the binding of several bivalent metal ions to Cc XylC and observed that it binds only one of them, namely, the Fe2+ cation, specifically and with a high affinity (Kd = 0.5 μM), pointing out that Cc XylC is a mononuclear iron protein. We propose that bivalent metal cations also promote the reaction nonenzymatically by stabilizing a short-lived bicyclic intermediate on the lactone isomerization reaction. An analysis of the reaction kinetics showed that Cc XylC complexed with Fe2+ can speed up the hydrolysis of d-xylono-1,4-lactone by 100-fold and that of d-glucono-1,5-lactone by 10-fold as compared to the nonenzymatic reaction. To our knowledge, this is the first discovery of a nonheme mononuclear iron-binding enzyme that catalyzes an ester bond hydrolysis reaction.


ESI FT-ICR MS experimental procedures
The instrument used for measuring protein mass spectra was a Bruker Solarix XR Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany), equipped with an electrospray ionization (ESI) source. The ion source pressure was ∼5 · 10 −6 mbar, and the cell pressure was ∼3 · 10 −10 mbar. Important measurement parameters are shown in Table S1. Calibration was done using ES Tuning Mix (Agilent Technologies, Santa Clara, CA, USA).
The Cc XylC was expressed in E. coli and purified essentially as described by Boer et al. 1 For mass spectrometry measurements, the Cc XylC sample was buffer-exchanged to 10 mM ammonium acetate (ultra-pure in HPLC quality water, pH 6.9) using a PD-10 desalting column (GE Healthcare, Freiburg, Germany). The concentration of this solution was determined using UV absorbance at 280 nm, and it was diluted to approximately 1 μM.
Denaturing conditions were created by mixing enzyme solution, acetonitrile and acetic acid (49.5/49.5/1.0, v/v). A denatured mass spectrum ( Figure S1) was obtained by measuring this solution. Charge state deconvolution ( Figure S2) was done in GNU Octave 2 with a custom function which summed all the observed charge states after transforming them to neutral mass space and aligning the peaks. The neutral mass of the enzyme was calculated by averaging the observed masses of the isotopic peaks in each observed charge state (from 18+ to 33+) corresponding to the most abundant isotopic peak in the deconvoluted spectrum; the mean mass was 31,524.89 Da, and the standard deviation was 0.10 Da.
The native mass spectra were measured first with the enzyme solution in 10 mM ammonium acetate. This resulted in a spectrum ( Figure S3) containing the previously observed apoenzyme as well as another significantly strong signal, the mass of which corresponded to the binding of a single Fe 3+ ion to the enzyme. The pure apo-enzyme was made by adding approximately 20 molar equivalents of EDTA to the enzyme sample before the buffer exchange, which yielded a spectrum ( Figure S4) with no detectable signal of the holo-enzyme. Metal ions were introduced by adding the corresponding metal chloride solutions to the bufferexchanged enzyme solution so that the concentration of the metal was 10 μM. With Fe 2+ ions present, only the iron-complexed holo-enzyme was observed ( Figure S5), whereas with Mg 2+ , Ca 2+ , Fe 3+ , Co 2+ , Ni 2+ or Zn 2+ , no significant change to the previous spectrum of the apo-enzyme was observed ( Figure S6). With Cu 2+ ions present, weak signals corresponding to one to four bound Cu 2+ ions appeared ( Figure S7); this was most likely due to the copper binding to free cysteines on the surface of the enzyme.
To determine the iron affinity to the Cc XylC, a series of solutions with differing Fe 2+ concentrations was prepared. The enzyme concentration was kept constant, at approximately 1 μM, and Fe 2+ concentration was varied from 7.8 nM to 16 μM, increasing in geometric progression. Native mass spectra were measured as before. At below 125 nM Fe 2+ , the holo-enzyme could not be observed, and so those spectra were discarded. At 16 μM Fe 2+ , there was a small amount of apo-enzyme left even though previously it disappeared at 10 μM Fe 2+ ; this discrepancy is due to the rough concentrations used in preliminary tests and subsequent overestimation of enzyme concentration or underestimation of metal concentration. Interpretation of the data is described in Section 3.
To prevent the Fe 2+ from being spontaneously oxidized to Fe 3+ in significant amounts, only fresh solutions were used. Solid FeCl 2 was dissolved and mixed immediately with the apoenzyme. Measurements were done within 30-60 minutes after the mixing to ensure that the equilibrium state would be obtained and that no significant oxidation of Fe 2+ would have time to occur. It was known that the oxidation is slow at pH less than 7 3 and that the presence of various anions at millimolar concentrations slows it down further by up to an order of magnitude 4 , so the oxidation would be negligible within an hour.

ESI QIT MS experimental procedures
The instrument used for reaction progress monitoring was a Bruker Esquire 3000plus quadrupole ion trap (QIT) mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany), also equipped with an ESI source. The ion source pressure was ∼3 mbar, and the cell pressure was ∼1 · 10 −5 mbar. Measurements were done in chromatogram mode: the instrument took scans continuously (about once per second). Important measurement parameters are shown in Table S2.
Samples were prepared in 10 mM ammonium acetate solution. Each sample contained 0.25 mM lactone -either d-glucono-1,5-lactone (99.5 %, Merck) or d-xylono-1,4-lactone (95 %, Sigma-Aldrich) -and 1.0 mM xylitol which was added as an internal standard. The concentrations were chosen so that the intensities of the signals of the product (gluconic or xylonic acid) and xylitol were roughly equal. Due to rapid nonenzymatic hydrolysis, the sample measurements were started immediately after dissolving the lactone to the sample solution. For the reactions with iron present, 10 μM FeCl 2 was added, and for the enzymatic reactions, 0.5 μM EDTA-treated apo-enzyme was also added.
The mass spectra of the protein samples were analyzed using DataAnalysis 5.0 software by Bruker. In order to determine the iron affinity to the Cc XylC, the fractional protein saturation B C (the proportional amount of the holo-enzyme out of all enzyme) for each sample was calculated using the signal intensities of the apo-and holo-enzymes obtained from the mass spectra. The water-bound forms of the holo-enzyme were taken into account as well. Summing the intensities up over the observed charge states, this resulted in the total intensities of the apo-and holo-enzymes (ΣI apo and ΣI holo ) for each sample. The fractional saturation is then The intensities of the signals are assumed to be proportional to the concentrations of the corresponding species. The intensities were obtained using the SNAP peak finding algorithm which evens out noise in the isotopic distributions and gives good estimates of the monoisotopic masses and the intensities of protein signals. This is especially useful when signals are weak and signal-to-noise ratios are low.
For each sample, total enzyme concentration c enzyme , total iron concentration c M and fractional saturation B C are known. Free iron concentration [M] is calculated using the formula The variables [M] and B C are correlated by the 1:1 binding model where K d is the dissociation constant and B max is the level of maximum saturation. This function was fitted to the data points using unweighted orthogonal distance regression in Origin Pro 2018 software 5 . The algorithm minimises the sum of squared orthogonal distances between the curve and the data points and optionally allows giving weights to both coordinates of each data point. Weighting the data points using estimated uncertainties was attempted but did not have a significant effect on the result. The program gave the confidence intervals of the fitted parameters with a 95 % level of confidence: The data points and the fitted curve are shown in Figure 2 in the article.

Monitoring the nonenzymatic and enzymatic lactone hydrolysis
The ion chromatograms of the lactone hydrolysis reactions were analyzed in DataAnalysis 3.3 software by Bruker and exported as line spectra. The signal intensities of the product acid and xylitol were extracted using a custom script and opened in GNU Octave for calculations. These discrete intensity curves were trimmed so that unreliable values in the beginning and in the end were eliminated, and the time of starting the respective reaction was subtracted from each time point, resulting in intensities I acid and I xylitol over time t. To eliminate the majority of noise in the signal intensities, the numbers of data points were reduced by averaging the raw data points over time intervals of constant length, and this length for each respective reaction was chosen to be roughly one fifth of the estimated reaction half-life.
An acid "concentration" value c ′ is defined as Assuming that signal intensities are proportional to concentrations and that xylitol concentration is constant, c ′ is proportional to the concentration of the acid.
The rate equation of the pseudo-first-order reaction where k is the reaction rate constant, and c L and c A are true concentrations of lactone and acid respectively. Solving the differential equation of the second and the fourth expression leads to the integrated rate equation During the reaction, concentration c A will start from zero and rise up to the maximum level c L (0) which is the initial concentration of the lactone. This equation can also be written in terms of c ′ : This function can be fitted to data points (t, c ′ ), and the fit will yield the rate constant k as well as the maximum level c ′ (∞).
The fit was done using GNU Octave with loaded Octave Forge packages io, optim, statistics and struct which are required for using the nlinfit function. Given values of t and c ′ in column vectors at and ac respectively, the following commands fit the function curve S10 from Equation (9) to the data points and return the values and uncertainties of k (k and dk) and c ′ (∞) (cinf and dcinf). curve = @(b, t) b(2) * (1 -exp(-b(1) * t)); [beta R J covb] = nlinfit(at, ac, curve, [0. 1 1]); delta = sqrt(diag(covb)) * tinv(0.5 + 0.95 / 2, length(ac) -2); k = beta(1), dk = delta(1), cinf = beta(2), dcinf = delta (2) The function nlinfit returns, as well as the parameters in vector beta, the variancecovariance matrix covb which is used to estimate the standard deviations of the parameters. They are multiplied by the critical value, given by the function tinv, where the two-tailed integral of Student's t-distribution with the appropriate degrees of freedom reaches 0.95, to get uncertainties with a 95 % level of confidence. The ion chromatograms and the fitted curves are shown in Figure S8.
Rate constants are not very illustrative, and so half-lives of lactone were also calculated for an alternative comparison of reaction progresses. Assigning c A (t) = 0.5 · c L (0) to Equation (8) and solving for t results in the half-life t 1/2 : In multiplication and division, uncertainty ∆ propagates so that relative uncertainty remains: which results in The Cc XylC catalyzed lactone hydrolysis is assumed to follow the Michaelis-Menten model 6,7 where k cat is the turnover number, K M is the Michaelis constant and c E is the enzyme concentration. When c L ≪ K M , the denominator approximately equals K M and the equation becomes similar to the pseudo-first-order rate equation (7). This can be assumed to be true in this case since the integrated first-order rate equation appears to fit well to the data. Therefore, an estimate of the ratio k cat /K M (specificity constant) can be calculated when the rate constant k and c E are known: S11 Figure S8. Measured reaction progresses (concentrations c ′ , proportional to true concentrations, as defined in Equation (5)) and fitted progress curves (Equation (9)

S12
The uncertainties of k and c E cannot be assumed to be uncorrelated, and their covariance cannot be determined. Therefore, the uncertainty of the ratio is estimated using the formula for maximum total uncertainty: which results in The uncertainty ∆c E is estimated to be 0.1 · c E (10 % relative uncertainty) with a 95 % level of confidence, so this becomes 5 References