Stable-Isotope Probing Reveals That Hydrogen Isotope Fractionation

May 28, 2013 - The fractionation of hydrogen stable isotopes during lipid biosynthesis is larger ...... 43: Stable Isotope Geochemistry, Chapter 3, pp...
0 downloads 0 Views 1MB Size
Articles pubs.acs.org/acschemicalbiology

Stable-Isotope Probing Reveals That Hydrogen Isotope Fractionation in Proteins and Lipids in a Microbial Community Are Different and Species-Specific Curt R. Fischer,†,∥ Benjamin P. Bowen,§,∥ Chongle Pan,⊥ Trent R. Northen,§ and Jillian F. Banfield*,†,‡ †

Department of Earth and Planetary Science and ‡Department of Environmental Science, Policy and Management, University of California, Berkeley, California 94720, United States § Life Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States ⊥ BioSciences Division and Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States S Supporting Information *

ABSTRACT: The fractionation of hydrogen stable isotopes during lipid biosynthesis is larger in autotrophic than in heterotrophic microorganisms, possibly due to selective incorporation of hydrogen from water into NAD(P)H, resulting in D-depleted lipids. An analogous fractionation should occur during amino acid biosynthesis. Whereas these effects are traditionally measured using gas-phase isotope ratio on 1H-1H and 1H-2H, using an electrospray mass spectrometry-based technique on the original biomolecular structure and fitting of isotopic patterns we measured the hydrogen isotope compositions of proteins from an acidophilic microbial community with organism specificity and compared values with those for lipids. We showed that lipids were isotopically light by −260 ‰ relative to water in the growth solution; alternatively protein isotopic composition averaged −370 ‰. This difference suggests that steps in addition to NAD(P)H formation contribute to D/ H fractionation. Further, autotrophic bacteria sharing 94% 16S rRNA gene sequence identity displayed statistically significant differences in protein hydrogen isotope fractionation, suggesting different metabolic traits consistent with distinct ecological niches or incorrectly annotated gene function. In addition, it was found that heterotrophic, archaeal members of the community had isotopically light protein (−323 ‰) relative to growth water and were significantly different from coexisting bacteria. This could be attributed to metabolite transfer from autotrophs and unknown aspects of fractionation associated with iron reduction. Differential fractionation of hydrogen stable isotopes into metabolites and proteins may reveal trophic levels of members of microbial communities. The approach developed here provided insights into the metabolic characteristics of organisms in natural communities and may be applied to analyze other systems.

V

into nonexchangeable C−H bonds through enzyme-catalyzed reductions with NAD(P)H. However, the isotope fractionation during protein biosynthesis and differences in lipid versus protein fractionation between autotrophs and heterotrophs have not been systematically assessed. An analytical challenge associated with measurement of hydrogen isotope fractionation in proteins is the difficulty of coupling aqueous-phase liquid chromatography methods to traditional isotope ratio mass spectrometry (IR-MS) instrumentation. In addition to technical challenges, natural system studies need to address contributions from many coexisting organisms, some of which may be very closely related yet play different ecosystem roles. For example, different levels of fractionation may be associated with different organisms, and fractionations for closely related organisms may differ if the organisms have

ariations in the abundance of stable isotopes in biomolecules and in their biosynthetic precursors, that is, isotopic fractionation, have been widely used to study the geochemistry and trophic structure of both modern and ancient ecological systems.27,22,17,21,25,10 A recent study showed that D/ H fractionation between lipids and water correlated with trophic function for a range of bacteria and eukaryotes in pure cultures.30 Organisms growing chemolithoautotrophically or phototrophically were found to synthesize lipids strongly depleted in deuterium relative to water in the growth medium, i.e., εlipids/water (see Methods for nomenclature) was as negative as −370 ‰. Organisms growing heterotrophically, even if capable of autotrophic growth under alternate conditions, fractionated H relative to D less strongly. A hypothesis raised to support these findings is that hydrogen fractionation occurs during the synthesis of central metabolic redox cofactors, such as NADH or NADPH, and this fractionation is transmitted to lipids during biosynthesis.30 As with lipid biosynthesis, protein biosynthesis entails the fixation of hydrogen atoms from water © XXXX American Chemical Society

Received: March 27, 2013 Accepted: May 28, 2013

A

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 1. Workflow for measurement of D/H fractionation in the metabolomic and proteomic analysis of a model microbial community. A bioreactor with a constant, known loading of D-enriched water as the sole hydrogen source is seeded with planktonic cells from a mixed-species chemoautotrophic microbial community. Biofilm develops at the air−medium interface in the bioreactor after several weeks. The biofilm is harvested, whereupon proteins and metabolites are harvested through multistep purification procedures. Extraction of both proteins and metabolites includes multistep treatments of samples with large volumes of isotopically nonenriched protic solvents, thus causing labile hydrogen atoms that exchange rapidly with solvent in the proteomics and metabolomics samples to isotopically re-equilibrate to near naturally occurring abundance of deuterium. Nonexchangeable hydrogen atoms would not be expected to re-equilibrate during the time scale of this experiment.

measured deuterium isotope fractionation between both water and lipids and water and proteins in a chemoautotrophic mixed microbial community that has served as a model system for a variety of prior molecular ecology studies2,5,18,26,29,30 (Figure 1). This community is hydrotrophic, meaning that water serves as the dominant source of hydrogen. A laboratory bioreactor system was inoculated with a mixed microbial community and allowed to grow in the presence of a medium with 4.00 % v/v D2O as the sole hydrogen source. After 2 weeks, a high developmental stage biofilm had developed on the air−medium interface. This biomass was harvested, and protein and lipid extracts were analyzed by LC−MS. Use of high-resolution mass spectra collected from aqueousphase electrospray-ionization instruments for measurement of deuterium isotopic abundance, whether in proteins or in lipids, requires accounting for both hydrogen ions that are labile and rapidly exchange with solvents and those that do not. In protic solvents, hydrogen atoms bound to nitrogen, oxygen, and sulfur atoms in small molecules exchange very quickly with hydrogen atoms from the solvent.7 Hydrogen atoms bound to carbon do not. (Exceptions to the latter rule, such as terminal alkynes or 1,3-dicarbonyl compounds, are encountered neither in the proteinogenic amino acids nor in most lipids.) Amide bonds in intact globular protein often exchange slowly in protic solvents, but the rate of exchange is accelerated dramatically by protein denaturation and by high pH. Given the precision of the techniques utilized here, the exchangeable hydrogens in both lipids and proteins will be isotopically equilibrated and nearly equal to the natural isotopic abundance found in the solvents used in the extraction steps (Figure 1).9 The exchangeable and nonexchangeable hydrogen atoms each contribute to observed mass isotopomer distributions. Figure 2A shows simulated data demonstrating how the mass isotopomer distribution for an example peptide, Arg-Ala, and phospholipid (lyso C16:0 PE) change when deuterium abundance at 10 atom % is assumed

different metabolic traits or ecological niches. The chemoautotrophic biofilm communities such as those studied here include bacteria that share 94% 16S rRNA gene sequence identity. These are ecologically distinct,13,5,26,1 but understanding of their metabolic differences is incomplete. Thus, strain-specific measurement of isotopic fractionation of proteins provides the ability to distinguish coexisting organisms, something that would be difficult using lipids (except in rare cases). In the current study, we developed a method to measure isotope fractionation in both proteins and lipids from a microbial community containing both closely related and highly divergent organisms. Our approach used conventional metabolomic and proteomic based mass spectrometry rather than the isotope ratio mass spectrometry normally used for geochemical analysis. This required adaptation of a recently developed liquid chromatography electrospray mass spectrometry (LC−MS/ MS) method for quantification of 15N enrichment in proteins16 to measure D enrichment in proteins. The resulting data, as well as data from lipid analyses, were used as inputs into a computational algorithm to determine isotopic abundances from high-resolution mass spectral data. We tested the hypothesis that D/H fractionation from water into proteins in two coexisting autotrophs and associated heterotrophs correlates with their known trophic levels and the extent to which fractionation is predicted by the hydrogen isotope composition in lipids. Our findings suggest that biosynthetic steps in addition to NAD(P)H formation contribute to D/H fractionation in microbial biomolecules and that patterns of fractionation differ between closely related community members.

1. RESULTS AND DISCUSSION 1.1. Exchangeable and Nonexchangeable Hydrogen Atoms in Measurements of Proteins and Lipids. We B

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 2. Theoretical approach to determining D/H fractionation by fitting isotope abundances to isotopomer distributions for intact molecules measured by high-resolution electrospray mass spectrometry. (A, B) Mass isotopomer distributions of a model peptide (A) and a model phospholipid (B) at low resolution, assuming D abundance of 10 atom % in both exchangeable and nonexchangeable positions (red) or of 10 atom % in only nonexchangeable positions and 0.00156 atom % in exchangeable positions. (C) The intrinsic sensitivity of determination of isotopic fractionation between a product P and a substrate S (αP/S) to measurements of the abundance FP of a stable isotope in product molecules, as a function of isotopic abundance in the substrate, FS. (D) The effect of relative intensity error (stemming from instrumental limitations) on estimates of FP from mass spectral data. As discussed in the text, the combined effects shown in panels C and D lead to measurement errors in αP/S of ±0.076 and δP/S of ±75 ‰, which is well inside the range of α values observed for fractionation of hydrogen isotopes by biosynthetic processes.

abundance (FS = 0.000156 for Vienna Standard Mean Ocean Water [VSMOW]) and for α = 0.6 is over 8600, meaning that any error, σFP, in measured values of FP is magnified 8600 times in the derived error σα of measured values of α. In contrast, at optimal FS (62.5 atom %) for a value of α of 0.6, ∂α/∂FP has a value of 2.4, an improvement of over 3600-fold. Thus, experiments with isotopically enriched precursors are essential when measuring fractionation by measurement of FP by ESIMS methods. Figure 2C also shows that measurement of α (at optimal labeling) will be more precise for processes with strong fractionation than for processes with little fractionation. The precision of measurements of FP from ESI-MS data, σF, depends on instrumental performance in a number of ways: (i) numerical errors arising from our convex optimization algorithm, (ii) background subtraction error caused by finite signal-to-noise ratios,14,15 and (iii) random noise in intensity values.12,4 Given that this study is focused on signals with signal-to-noise ratios often greater than 100, it is most likely that the third factor is the largest contributor to uncertainty. Figure 2D shows the impact of the third factor, random noise in intensity values. The x-axis is the 95% confidence interval (2× the standard deviation) of a normally distributed random intensity error added to calculated mass spectral data. The yaxis is the 95% confidence interval in the best-fit D-abundance. The observed precision depends on isotope abundance, and for 5% D abundance, the uncertainty in best-fit D abundance is less than 0.2 atom %, even at relative intensity error as large as 12%.

either for all hydrogen atoms (red circles) or when deuterium abundance in exchangeable positions is assumed to be native abundance and only nonexchangeable positions are assumed to have 10 atom % deuterium (black circles). The difference in the spectra illustrates the importance of accounting for both exchangeable and nonexchangeable hydrogen atoms. 1.2. D/H Fractionations Are Large Enough to Be Detected by ESI-MS When Using Enriched Substrates. To determine the fractional labeling differences that can be discriminated using this approach, we estimated the precision with which isotopic fractionation can be measured from ESIMS-derived mass isotopomer distributions for biomolecules. As discussed in the Methods section, the measurement error σα in a calculated fractionation factor α between a biosynthetic product P and a substrate S depends on two factors (when the isotopic composition of S, FS, is assumed to be known precisely), as shown in below in the Methods section (eq 4). The first factor, ∂α/∂FP, represents the intrinsic relationship between α and the measured isotopic enrichment or depletion in the product FP. The second factor, σFP represents the precision in assessments of FP. The first factor is a strong function of isotopic abundance in the substrate FS and the fractionation factor α, as shown in Figure 2C, which plots ∂α/ ∂FP. As FS approaches either zero or 100%, α cannot be measured accurately. The value of FS that minimizes σα is 1/(1 + α). At this optimal value of FS, the derivative ∂α/∂FP is equal to 4α. Notably, the value of ∂α/∂FP at natural deuterium C

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Figure 3. Compound-specific D/H fractionation between growth water and lipids from a mixed microbial community as measured by electrospray ionization mass spectrometry. Communities were grown in medium with 4% D2O (dotted orange line). Four abundant phosphatidylethanolamine lipids, differing in the degree of N-methylation, saturation, and fatty acid chain length, were analyzed by LC−ESI-MS, and the resulting data were used to calculate deuterium abundance in each lipid. The black bars represent the range of values calculated from four replicate analyses, and the dotted gray line represents the median D abundance. The lipids were strongly depleted in D relative to the growth water, which is the sole source of hydrogen to the community, a finding consistent with previous analyses of lipid biosynthesis in chemoautotrophs.30

1.4. Organism-Specific D/H Fractionation during Protein Biosynthesis Revealed by Shotgun Proteomics. The fractionation of hydrogen between water and proteins was measured using shotgun proteomics. The Sipros algorithm16 was used to simultaneously determine the sequence and isotopic abundance of detected peptides. We modified Sipros to account for the number of exchangeable hydrogens in each amino acid residue type, as shown in Supplementary Table S1. The algorithm assumes that all nonexchangeable H atoms in a given peptide have the same D atom %. A total of 36,090 spectra for 4,477 unique peptide sequences were detected by Sipros at a false discovery rate of 1.5%, with a mean D abundance of 2.58 atom %, although peptides with D abundances as high as 5 or as low as 0 atom % were identified. The Sipros-estimated deuterium atom % for each peptide, as well as the Sipros-derived peptide sequence, was used to initiate the convex optimization algorithm in the same manner as described above for lipids. Best fits from the optimization are compared to the original Sipros estimate in Figure 4. The distribution of best-fit D abundance is narrower than the distribution of initial Sipros estimates. The convex optimization also provides a sum-squared error parameter, which represents the goodness of fit between best-fit spectra and raw data. We observed that this sum-squared error (sse) parameter was significantly higher for matches to reverse database spectra, i.e., to known false positives. In contrast, the sse for spectra matching to trypsin, an undeuterated protein added as part of the proteomics workflow, was low and in the same range as that observed for sse values of partially deuterated peptides from acid mine drainage organisms. Thus, fits with high sum-squared error can be identified as false positive assignments. Measured D atom % values of proteins were used to examine the dependence of fractionation on organism type, as shown in Figure 5. The analysis considered the four dominant organisms present in the biofilm communities: the bacteria Leptospirillum Group II UBA and 5-way CG genotypes and Leptospirillum Group III and the archaeaon Ferroplasma acidarmanus. Results show significant interorganism differences (Table 1). Peptides were also detected for less-abundant organisms, including the archaeal G-plasma lineage as well as for fungi. However, the number of unique peptides was low, and the mean sse for

Instrument manufacturers report that standard deviations in relative isotopomer intensities are 2−3%. We also assume an enrichment error of 0.1% on the commercial D2O and an experimental error of 0.1% for preparing dillutions for an overall error in FS of 0.2%, Combining the effects of relative intensity error and signalto-noise derived truncation error at 4% D abundance leads to an estimate for σFP of about 0.3 atom %. Assuming 4% D abundance, the resulting value of σα is 0.076, which corresponds to ±76 ‰. Although certainly large by the standards of isotope ratio mass spectrometry, this is well below the observed range of fractionation between water and biomolecules and therefore allows fractionation studies to be performed using conventional metabolomic and proteomic instrumentation. This error is for measurement of D abundance in a single molecular species. 1.3. D/H Fractionation Measured in Lipids from a Microbial Community. The most abundant lipids detected in methanol/chloroform extracts of biofilm community samples are a family of lyso phosphatidylethanolamine and lyso Nmethyl phosphatidylethanolamine lipids. As described previously,9 these lipids may be important in adaptation to high metal concentrations and very low pH conditions characteristic of their growth environment. The structure of these lipids and the best-fit D abundance for each are shown in Figure 3. Raw single-replicate best-fit D abundances for these lipids ranged between 2.63 and 3.18 atom %, with mean values for the four lipids between 2.87 and 3.11 atom %. While these numerical values might be due to a physical effect, their values do not have statistically significant differences. For some lipid peaks, we found evidence for weak chromatographic fractionation of isotopologues (data not shown), i.e., that the deuterium abundance of the early-eluting portion of the peak was different from that of the late-eluting portion of the peak, but the effect was slight and we controlled for the effect by considering only the portion of the peak well above the full-width half-maximum in intensity. Likewise, it is noteworthy that many lipid peaks will contain isomers that are coeluting or nearly coeluting. The mean D abundance across the lipids shown in Figure 3 is 3.00 atom %, which corresponds to αl/w = 0.739, or an εL/W of −261 ‰. The highly negative εL/W value is consistent with a chemoautotrophic origin for these lipids.30,9 D

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

Table 1. Organismal Dependence of D/H Isotopic Fractionation between Growth Water and Proteins in a Mixed Microbial Communitya organism

n

avg FP (atom %)

SEM (‰)

mean εP/S (‰)

SD (‰)

l2cg l2uba lepto3 fer

2000 1620 1748 125

2.49 2.49 2.67* 2.75*

2.4 2.8 2.8 12.1

−388.5 −389.6 −343.6 −323.3

107.7 112.4 118.3 134.9

a

Peptides unique to the listed genomes were culled from a data set of 36,090 peptides and associated D abundances for each. The SEM is the standard error of the mean D abundance; SD is standard deviation. The four organisms have statistically significant differences in mean D abundance as assessed by an analysis of variance with multiple comparisons at the p = 0.05 level.

ssp. or Ferroplasma, suggesting that some may be false positive identifications. 1.5. Discussion. Through theoretical and experimental approaches, we have demonstrated the ability to measure and quantify hydrogen isotope fractionation in proteins as well as lipids. The method relies upon (i) high-mass accuracy mass spectrometry measurements, (ii) the use of deuterated water in microbial community growth experiments, and (iii) the development and refinement of spectral interpretation and fitting tools. 1.5.1. Sources of Fractionation. Bacteria of the Leptospirillum genus are the primary producers in the microbial communities studied here. They are thought to fix carbon by the reverse tricarboxylic acid cycle and to use reverse electron transport as means to generate highly reduced biosynthetic precursors (NAD(P)H and/or ferredoxins) from the lowpotential electron donor iron(II). The D/H fractionations that arise from these particular chemoautotrophic pathways have not been studied previously. The finding that Leptospirillum

Figure 4. D/H fractionation between growth water and proteins in a mixed microbial community as measured by electrospray ionization mass spectrometery. A chemoautotrophic microbial community in which water is the sole hydrogen source was cultivated in 4 atom % D2O, proteins were extracted, and tryptic digests were analyzed by LC−ESI-MS using shotgun proteomics techniques. The plot shows the distribution of D abundance in proteins as assessed by Sipros (yaxis) or by a convex optimization refinement of the Sipros data (xaxis). The Sipros algorithm, which simultaneously identifies tryptic peptide sequences and estimates isotopic abundance in identified peptides, provides a broad range of D abundances in unique peptides from the community. The Sipros estimates are refined by convex optimization to yield a more precise distribution of D abundance in the community proteome.

unique peptides from these lesser-abundant community members was higher than that observed for Leptospirillum

Figure 5. D/H fractionation observed in community metaproteomics data is organism-dependent. Since detected peptides retain sequence information that can be matched to metagenomics data, the source organism for many detected peptides can be uniquely determined. A 2-D histogram of the best-fit D abundance (y-axis) versus the log-scale sum-squared error between the best fit and the observed spectrum (x-axis) can be made for each of the four most dominant organisms in the community. The histograms show that for the most abundant organisms, fitting errors are usually less than 10−2. The bacterium Leptospirillum group III and the archaeon Ferroplasma acidaramnus have significantly different average D abundance levels in their proteome than the initial biofilm colonists, the Leptospirillum group II bacteria. E

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

1.6. Conclusion. Our analyses demonstrate that conventional mass spectrometry based metabolomic and proteomic methods can be employed to measure hydrogen isotope fractionation in microbial communities and reveal organismspecific isotopic fractionation patterns in coexisting members. In the specific case studied here, water (and much lesser amounts of species such as ammonium, bisphosphate, and bisulfate that are equilibrated with water) is the dominant hydrogen source for the community. Water is an important source of nonexchangeable hydrogen in nearly all ecosystems on Earth, and so the method should be generally applicable. A key aspect of our approach is the use of isotopically labeled substrates. The ESI-MS and isotopic pattern fitting techniques described here show that autotrophic bacteria sharing 94% 16S rRNA gene sequence identity had statistically significant differences in protein hydrogen isotope fractionation, suggesting different metabolic traits consistent with distinct ecological niches or incorrectly annotated gene function. In addition, it was found that heterotrophic, archaeal members of the community had isotopically light protein composition (−323 ‰) and were significantly different from coexisting bacteria. This is attributed to metabolite transfer and iron reduction. Differential fractionation of hydrogen stable isotopes into metabolites and proteins may reveal trophic levels of members of microbial communities.

proteomes are strongly depleted in D relative to water in the growth medium is consistent with previous studies involving other chemoautotrophic pathways.30 Abundant phospholipids in the acid mine drainage (AMD) community, which in all likelihood derive predominantly from Leptospirillum, were also strongly depleted in D. Figure 3 shows that lipid biosynthesis from water resulted in a D/H fractionation of about −260 ‰, and Figure 5 shows that protein biosynthesis from water resulted in a D/H fractionation that was organism-dependent but averaged −370 ‰. Proteins were thus significantly depleted in D relative not only to water in the growth medium but also to lipids. Fractionation arising mainly from NAD(P)H biosynthesis would give rise to lipids and proteins with similar D abundances, as NAD(P)H is the primary reductant in both lipid and amino acid biosynthesis. Thus, the difference in lipid and protein D abundance suggests that steps other than the biosynthesis of NAD(P)H from water contribute to the observed fractionation. The finding of isotopically lighter proteins compared to lipids in Leptospirillum spp. contrasts to studies of phototrophic algae in pure culture, which found that D/H values of lipids and proteins, while both D-depleted relative to photosynthetically generated carbohydrates, were isotopically similar.8 One possible explanation for this difference may be that the reverse TCA cycle can lead directly to synthesis of amino acid precursors in Leptospirillum, whereas in algal photosynthesis all lipids and amino acids must be derived from a common carbohydrate precursor. 1.5.2. Organism-Dependent Fractionation. Ferroplasma acidarmanus is an iron-oxidizing archaeon believed to be incapable of autotrophic growth6 and thus to rely on heterotrophy for growth in the environment,2 although it is capable of oxidizing iron(II) and use of this electron donor improves growth. Lipid biosynthesis by heterotrophs was previously found to result in a weak isotope fractionation or in a slight preference for deuterons over protons.30 As such, the strong depletion of D in the Ferroplasma proteome, similar to the D depletion observed in the proteomes of the chemoautotrophic Leptospirillum, was particularly interesting. One possibility may be that during chemoheterotrophic growth by iron oxidation, reducing power (NADPH) is synthesized by the same mechanism as in chemoautotrophs, reverse electron transport. This would result in similar D fractionation in Ferroplasma and Leptospirillum proteomes. An alternative explanation is that most of the amino acids used by Ferroplasma for protein biosynthesis originate from Leptospirillum, thus its proteins have comparable isotope fractionation patterns. This is not unreasonable given that the best-studied relative of Ferroplasma, Thermoplasma acidophilum, requires peptides for growth.24 Pinpointing the exact mechanism of D fractionation in Ferroplasma will require further study. Regardless of the source of D/H fractionation, it is clear that the proteomes of closely related organisms can have distinct D/ H abundance patterns. The D/H ratio in proteomes from UBA-subtype Leptospirillum group II and the 5-way CGsubtype Leptospirillum group II are significantly different from the same ratio for the Leptospirillum group III proteome (Figure 4). The extent of fractionation corresponds to the order in which the organisms have been shown to appear during biofilm establishment. This finding indicates a different balance between fractionating and nonfractionating pathways in these strains.

2. METHODS 2.1. Calculation of High-Resolution MS1 Mass Spectra As a Function of Isotope Abundance. Our starting point for calculating high-resolution isotopic abundance in the presence of hydrogen exchange was an implementation of the fast Fourier transform-based method20,19 provided by Mathworks, Inc. as part of MATLAB R2010b. The isotopic composition was calculated in terms of the natural isotopic composition of all elements, and this function was modified so that user-specified isotopic abundances for any given isotope could be specified and so that a 93rd chemical element, “Hn”, corresponding to nonexchangeable hydrogens and having isotopes with masses equivalent to “H”, which was now taken to indicate exchangeable hydrogens. In this way, populations of exchangeable and nonexchangeable hydrogens were treated as isotopically independent. For fitting isotopic abundances to raw mass spectral data, the monoisotopic mass (MIM) of the model formula was used to filter raw data. Only m/z values between MIM − 1 Da and MIM + 12 Da were considered. Retention time filtering was also applied so that the scans used for fitting came from retention times where the total intensity from the feature of interest was greater than the full width at halfmaximum. Although a very small shift in the retention time could be observed due to deuterium incorporation, it was much less than the integration window. Data in this (m/z, RT) window was unit-vector normalized. The sum-squared error between normalized spectral data and model data generated as described above was the cost function minimized. The sum-squared error was calculated as the sum-squared difference between normalized intensities of the eight most abundant peaks generated by the model and the corresponding peaks from the data. 2.2. Estimating Precision in the ESI-MS-Based Determination of Isotopic Abundance and Fractionation. The relationships used in this work to relate measured fractional isotopic enrichments of deuterium incorporated into various analytes were εi = the isotopic enrichment or depletion relative to a standard of analyte or element i (parts per thousand); αi = the dimensionless isotopic fractionation factor; RPi = the ratio of the amount of heavy to the light isotope in the product analyte; RS = the ratio of the amount of heavy to light isotope in the substrate; FPi = the atomic fraction of heavy isotope in the product analyte; FS = the atomic fraction of the heavy isotope in the substrate; σαi = the uncertainty in the fractionation factor; σFS = the F

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

volume 10 mL of planktonic cells in natural abundance isotopic medium was used to inoculate 1000 mL of 4% v/v D2O in the 9K-BR growth medium (pH 0.9; 400 mM iron(II) sulfate as the electron donor). Growth was conducted in medium recirculation mode until a low development stage biofilm was visible on the surface of the bioreactor medium, after which time operation was switched to a single-pass configuration. When the iron(II) in the medium was depleted, the medium was replaced with freshly prepared 4% v/v D2O medium. A total of 4.0 L of 4% v/v D2O medium was used over a 2week period until the attainment of high developmental stage biofilm. Approximately 6 g of wet biomass was harvested from the bioreactor and stored at −80 °C as a wet pellet. 2.5. Protein Extraction and Proteomic Analysis. One gram of frozen wet biomass was extracted for protein analysis, as schematized in Figure 1 and described previously.18 Sample preparation involves biofilm disruption by sonication, treatment with sodium carbonate at pH 11, TCA precipitation, resuspension in denaturing (6.0 M guandinium chloride) solvents, trypsin digestion, and solvent exchange on solid-phase extraction resins. The denaturing conditions and high pH ensure rapid equilibration of exchangeable hydrogens with nativeabundance solvents used in the extraction. Shotgun proteomics analysis was performed as described previously.16 Briefly, peptides were separated using 2-dimensional liquid chromatography (strong cation exchange LC/reverse-phase LC). Unlabeled HPLC-grade solvents (Burdick & Jackson, Muskegon, MI) were used for LC separation, which further facilitates isotopic exchange and dilution of labile (exchangeable) deuterons for protons. The LC eluent was directly electrosprayed into a Thermo LTQ Orbitrap Velos instrument. Data-dependent MS/MS were acquired using the following parameters: every MS1 scan followed by five MS2 scans; MS1 acquired in Orbitrap at resolution 30,000 with twomicroscan averaging; MS2 scans acquired in Orbitrap at resolution 7,500 with two-microscan averaging; 35% normalized collision energy; ±2.5 Da isolation window; dynamic exclusion enabled with ±3 Da exclusion window. 2.6. Phospholipid Extraction and Analysis. Biomass samples were freeze-dried overnight (Labconco FreeZone2.5, Labconco, Kansas City, MO), and 100 mg of lyophilized biomass was extracted with 800 μL of a 2:1 methanol/chloroform solvent mixture. The resulting extracts were briefly disrupted by agitation with a steel ball (Biospec Products Mini-beadbeater 96, Biospec Products, Bartlesville, OK), vortexing, and low-power sonication (Branson Sonifier 250, Branson Ultrasonics, Danbury, CT). Solvents were removed from centrifugation-clarified extracts by vacuum centrifugation. Dried extracts were then resuspended in 100 μL of 2:1 methanol/chloroform to further facilitate isotopic equilibration of exchangeable hydrogens and then redried by vacuum centrifugation. Final reconstitution prior to LC−MS analysis was with 100 μL resuspended in 2:2:1 isopropanol/methanol/water; 4 μL of the resulting resuspension was analyzed by LC−MS. LC−MS analysis was by reverse-phase (C18) column and an Agilent ESI-qTOF Model 6520 as described previously.9 When MS/MS experiments were conducted, the collision-induced dissociation (CID) cell voltage was set to 10 V. 2.7. Data Processing. The proteomic SIP data were processed using the Sipros algorithm.16 There were two types of hydrogen atoms in every amino acid, exchangeable hydrogen atoms and nonexchangeable hydrogen atoms. In the Sipros configuration file, the number of exchangeable hydrogen atoms was provided for every amino acid type under the column of element P and the number of nonexchangeable hydrogen atoms under the column of element H. The D atom % of the exchangeable hydrogen was fixed at the natural abundance, 0.0115%. The D atom % of the nonexchangeable hydrogen was searched from 0% to 5% at 0.1% intervals. The genomic databases used for Sipros search of proteomic data have been previously described.5 The Sipros search results were filtered at a score cutoff of 40. Peptide identifications were clustered into protein isotopologues at a D atom % distance of 1%. A minimum of two peptides is required for a protein-isotopologue identification. The false discovery rate of protein identification was then calculated as 2× of the percentage of reverse protein IDs out of all protein IDs.

uncertainty in the substrate isotopic enrichment (atom % or fraction); and σFPi = the uncertainty in the product analyte isotopic enrichment (atom % or fraction). The partitioning of hydrogen isotopes between two chemical species, a product P and a substrate S, is usually quantified by a fractionation factor αP/S, defined as αP/S = RP/RS, where R is the ratio of deuterium nuclei to protium nuclei in all molecules of chemical species. Shown in eq 3, αP/S can also be expressed as a function of fractional isotopic abundances Fi instead of isotope ratios, since Ri = Fi/(1 − Fi). Here αi ranges between the lipids and the SMOW reference scale.

εi = 1000(αi − 1)

αi =

αi =

σαi =

(1)

R Pi RS

(2)

FPi /(1 − FPi) FS/(1 − FS)

⎞2 ⎛ ∂αi ⎞2 ⎛ ∂αi ∂αi σFS⎟ + ⎜⎜ σFP ⎟⎟ ≈ σF ⎜ i ∂FPi Pi ⎝ ∂FS ⎠ ⎝ ∂FPi ⎠

−1 ⎛ ∂αi FS ⎞ (F (α − 1) + 1)2 = ⎜(FPi − 1)2 ⎟ =− S i ∂FPi 1 − FS ⎠ (FS − 1)FS ⎝

(3)

(4)

(5)

Equation 3 implies that measuring fractional isotopic abundances of the substrate (FS) and the product (FP) is sufficient to determine αP/S and thus that uncertainties in measured values for FS and FP determine uncertainties in αP/S. For this study, we assume that FS can be controlled and/or measured to high precision, so that error σα in measurements of αP/S depends solely on error σFP in FP, as shown in eq 3. In experiments with labeled substrate, any imprecision in substrate fractional abundance FS stems mainly from imperfect precision of laboratory pipettes or imprecision in the extent of enrichment provided by manufacturers of commercially available isotopically enriched materials. Both sources of uncertainty are negligible in comparison to the uncertainty stemming from measurement of FP. Even if FS is not known a priori for a given experiment, it can be measured to high accuracy using traditional stable isotope mass spectrometry performed on the pure substrate. Equations 4 and 5 show that σα depends on two factors: (i) a partial derivative of αP/S with respect to FP and (ii) σFP, the random error in measurements of F P. The first factor does not depend on instrumentation, fitting procedures, or any other experimental parameter except for FS and FP (or equivalently, FS and αP/S) and can be obtained through simple differentiation and eliminating FP from the result by using eq 3. 2.3. Limits on the Accuracy and Precision of Determining Isotopic Abundance from High-Resolution Mass Spectral Data. To estimate σFP as a function of FP and as a function of fitting and instrumental parameters, we used a two-step procedure. First, the theoretical distribution of isotopologue abundances for a model compound was calculated at a given D abundance. The predicted isotopologue abundances are given as a list of mass values mi and relative abundances αi.The goal of the error model is to transform the list of (mi, ai) values into a list of measured values of the form (Mi′, Ii′) that simulates real mass spectral data. Most instruments suffer from mass inaccuracy, dynamic range limitations, and insufficient resolution. These technical limitations mean that Mi′ and mi may differ (although usually by less than 5−25 ppm) and that ai and Ii′ may differ as well. Extra “noise” peaks may be present, and therefore, more peaks may be present in the mass spectrum than in the model. 2.4. Cultivation of a Mixed-Species, Chemolithoautotrophic, Acidophilic Biofilm in D-Enriched Medium. The bioreactor system used has been described previously.16,3 In order to ensure that the biofilm had reached an isotopic steady state, an inoculum G

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

To fit deuterium abundance to observed data, a convex optimization routine was used that minimized the sum-squared error between the observed data and model-derived calculations of mass isotopomer distribution. MATLAB R2010b was used for all calculations.



(11) Horibe, Y., and Craig, H. (1995) D/H fractionation in the system methane-hydrogen-water. Geochim. Cosmochim. Acta 59, 5209− 5217. (12) Kind, T., and Fiehn, O. (2006) Metabolomic database annotations via query of elemental compositions: Mass accuracy is insufficient even at less than 1 ppm. BMC Bioinf. 7, 1 DOI: 10.1186/ 1471-2105-7-234. (13) Lo, I., Denef, V. J., VerBerkmoes, N. C., Shah, M. B., Goltsman, D., DiBartolo, G., Tyson, G. W., Allen, E. E., Ram, R. J., Detter, J. C., Richardson, P., Thelen, M. P., Hettich, R. L., and Banfield, J. F. (2007) Strain-resolved community proteomics reveals recombining genomes of acidophilic bacteria. Nature 446, 537−541. (14) Pan, C. L., Kora, G., McDonald, W. H., Tabb, D. L., VerBerkmoes, N. C., Hurst, G. B., Pelletier, D. A., Samatova, N. F., and Hettich, R. L. (2006) ProRata: A quantitative proteomics program for accurate protein abundance ratio estimation with confidence interval evaluation. Anal. Chem. 78, 7121−7131. (15) Pan, C. L., Kora, G., Tabb, D. L., Pelletier, D. A., McDonald, W. H., Hurst, G. B., Hettich, R. L., and Samatova, N. F. (2006) Robust estimation of peptide abundance ratios and rigorous scoring of their variability and bias in quantitative shotgun proteomics. Anal. Chem. 78, 7110−7120. (16) Pan, C. L., Fischer, C. R., Hyatt, D., Bowen, B. P., Hettich, R. L., and Banfield, J. F. (2011) Quantitative tracking of isotope flows in proteomes of microbial communities. Mol. Cell. Proteomics 10, No. M110.006049. (17) Post, D. M. (2002) Using stable isotopes to estimate trophic position: Models, methods, and assumptions. Ecology 83, 703−718. (18) Ram, R. J., VerBerkmoes, N. C., Thelen, M. P., Tyson, G. W., Baker, B. J., Blake, R. C., Shah, M., Hettich, R. L., and Banfield, J. F. (2005) Community proteomics of a natural microbial biofilm. Science 308, 1915−1920. (19) Rockwood, A. L., and VanOrden, S. L. (1996) Ultrahigh-speed calculation of isotope distributions. Anal. Chem. 68, 2027−2030. (20) Rockwood, A. L., Vanorden, S. L., and Smith, R. D. (1995) Rapid calculation of isotope distributions. Anal. Chem. 67, 2699−2704. (21) Ruess, L., and Chamberlain, P. M. (2010) The fat that matters: Soil food web analysis using fatty acids and their carbon stable isotope signature. Soil Biol. Biochem. 42, 1898−1910. (22) Schmidt, H. L., Werner, R. A., and Eisenreich, W. (2003) Systematics of 2 H patterns in natural compounds and its importance for the elucidation of biosynthetic pathways. Phytochem.y Rev. 2, 61− 85. (23) Sessions, A. L., Jahnke, L. L., Schimmelmann, A., and Hayes, J. M. (2002) Hydrogen isotope fractionation in lipids of the methaneoxidizing bacterium Methylococcus capsulatus. Geochim. Cosmochim. Acta 66, 3955−3969. (24) Smith, P. F., Langworthy, T. A., and Smith, M. R. (1975) Polypeptide nature of growth requirement in yeast extract for Thermoplasma acidophilum. J. Bacteriol. 124, 884−892. (25) Summons, R. E., Jahnke, L. L., and Roksandic, Z. (1994) Carbon isotopic fractionation in lipids from methanotrophic bacteria Relevance for interpretation of the geochemical record of biomarkers. Geochim. Cosmochim. Acta 58, 2853−2863. (26) Tyson, G. W., Chapman, J., Hugenholtz, P., Allen, E. E., Ram, R. J., Richardson, P. M., Solovyev, V. V., Rubin, E. M., Rokhsar, D. S., and Banfield, J. F. (2004) Community structure and metabolism through reconstruction of microbial genomes from the environment. Nature 428, 37−43. (27) Vander Zanden, M. J., and Rasmussen, J. B. (2001) Variation in delta N-15 and delta C-13 trophic fractionation: Implications for aquatic food web studies. Limnol. Oceanogr. 46, 2061−2066. (28) Washburn, M. P., Wolters, D., and Yates, J. R. (2001) Largescale analysis of the yeast proteome by multidimensional protein identification technology. Nat. Biotechnol. 19, 242−247. (29) Wilmes, P., Remis, J. P., Hwang, M., Auer, M., Thelen, M. P., and Banfield, J. F. (2009) Natural acidophilic biofilm communities reflect distinct organismal and functional organization. ISME J. 3, 266−270.

ASSOCIATED CONTENT

S Supporting Information *

This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: jbanfi[email protected]. Author Contributions ∥

These authors contributed equally to this work.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research was funded by the U.S. Department of Energy, Office of Biological and Environmental Research CarbonCycling Program (DE-SC0004665), the DOE Genomics:GTL Program grant number DE-FG02-05ER64134 and ENIGMA Scientific Focus Area No. DE-AC02-05CH11231. The authors thank Banfield lab members for assistance with biofilm sampling in the field, and JM Hayes for valuable technical discussions.



REFERENCES

(1) Baker, B. J., and Banfield, J. F. (2003) Microbial communities in acid mine drainage. FEMS Microbiol. Ecol. 44, 139−52. (2) Baumler, D. J., Jeong, K. C., Fox, B. G., Banfield, J. F., and Kaspar, C. W. (2005) Sulfate requirement for heterotrophic growth of ″Ferroplasma acidarmanus″ strain fer1. Res. Microbiol. 156, 492−8. (3) Belnap, C. P., Pan, C., VerBerkmoes, N. C., Power, M. E., Samatova, N. F., Carver, R. L., Hettich, R. L., and Banfield, J. F. (2010) Cultivation and quantitative proteomic analyses of acidophilic microbial communities. ISME J. 4, 520−30. (4) Bowen, B. P., Fischer, C. R., Baran, R., Banfield, J. F., and Northen, T. (2011) Improved genome annotation through untargeted detection of pathway-specific metabolites. BMC Genomics 12 (Suppl 1), S6. (5) Denef, V. J., Kalnejais, L. H., Mueller, R. S., Wilmes, P., Baker, B. J., Thomas, B. C., VerBerkmoes, N. C., Hettich, R. L., and Banfield, J. F. (2010) Proteogenomic basis for ecological divergence of closely related bacteria in natural acidophilic microbial communities. Proc. Natl. Acad. Sci. U.S.A. 107, 2383−90. (6) Dopson, M., Baker-Austin, C., Hind, A., Bowman, J. P., and Bond, P. L. (2004) Characterization of Ferroplasma isolates and Ferroplasma acidarmanus sp. nov., extreme acidophiles from acid mine drainage and industrial bioleaching environments. Appl. Environ. Microbiol. 70, 2079−88. (7) Eigen, M. (1964) Proton transfer acid-base catalysis + enzymatic hydrolysis . I. Elementary processes. Angew. Chem., Int. Ed. 3, 1−&. (8) Estep, M. F., and Hoering, T. C. (1981) Stable hydrogen isotope fractionations during autotrophic and mixotrophic growth of microalgae. Plant Physiol. 67, 474−477. (9) Fischer, C. R., Wilmes, P., Bowen, B. P., Northen, T. R., and Banfield, J. F. (2012) Deuterium-exchange metabolomics identifies Nmethyl lyso phosphatidylethanolamines as abundant lipids in acidophilic mixed microbial communities. Metabolomics 8, 566−578. (10) Hayes, J. M. (2001) Fractionation of the isotopes of carbon and hydrogen in biosynthetic processes, in Reviews in Mineralogy and Geochemistry (Valley, J. and Cole, D., Eds.) Vol. 43: Stable Isotope Geochemistry, Chapter 3, pp 225−277, Mineralogical Society of America, Chantilly, VA. H

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX

ACS Chemical Biology

Articles

(30) Zhang, X. N., Gillespie, A. L., and Sessions, A. L. (2009) Large D/H variations in bacterial lipids reflect central metabolic pathways. Proc. Natl. Acad. Sci. U.S.A. 106, 12580−12586. (31) Osburn, M. R., Sessions, A. L., Pepe-Ranney, C., and Spear, J. R. (2011) Hydrogen-isotopic variability in fatty acids from Yellowstone National Park hot spring microbial communities. Geochim. Cosmochim. Acta 75, 4830−4845.

I

dx.doi.org/10.1021/cb400210q | ACS Chem. Biol. XXXX, XXX, XXX−XXX