1.1Metabolite Extraction and LC-ESI-Qtof-MS Analysis

SUPPLEMENTAL INFORMATION FOR

Deuterium-exchange Metabolomics Reveals N-Methyl lysoPhosphatidylethanolamines As Abundant Lipids in Acidophilic Mixed Microbial Communities

Curt R. Fischer1*, Paul Wilmes1,5*, Benjamin P. Bowen3,4, Trent Northen3,4, Jillian F. Banfield1,2

1Supplemental Methods

1.1Metabolite extraction and LC-ESI-qTOF-MS analysis.

Natural biofilm samples.Approximately 500 mg of frozen biomass (-80 oC) was removed from each sample and analyzed by LC-ESI-qTOF-MS as described previously[3]. Cells were pelleted by centrifugation (4,000 rpm) and resuspended in an equivalent volume of ice cold 1 x PBS buffer at pH 1 to avoid inconsistent cell lysis prior to sonication. The pH of the 1 x PBS buffer was adjustedusing a 5 M H2SO4 solution. Micro-tip sonication was performed on ice for four 30 s bursts. This was followed by the addition of 1,000 μL of hot methanol (65 oC), vortexing and incubation at 65 oC for 3 min. The biomass was pelleted by centrifugation (14,000 rpm), 750 μL of supernatant were removed and the pellets re-extracted with 500 μL of hot isopropanol (65 oC) for 3 min. The biomass was again pelleted by centrifugation, the extracts pooled, dried by vacuum centrifugation, redissolved in a 1:2:2 mixture of water:methanol:IPA (100 μL/mg of wet biomass), and filtered (0.45 mm PVDF, National Scientific, Rockwood, TN, USA).

Bioreactor biofilm samples. Frozen 250 mg to 500 mg of -80 °C conserved bioreactor biofilm pellets were removed from bulk frozen samples and weighed. One volume (w/v) of PBS and two volumes of methanol were added to each biofilm sample. The sample was sonicated on ice at high power (~200 W) for 5 min using a microtip-equipped Sonicator 350 (Heat Systems Ultrasonics, Farmingdale, NY). Frequent pauses were inserted in each sonication cycle to avoid overheating. After sonication, samples were vortexed, and centrifuged for 10 min at 10,000 g. Two volumes of methanol/PBS-rich supernatant were removed and transferred to a separate container. Two volumes of isopropanol were added to the sonicatedlysate , and the sample was vortexted and re-sonicated. The isopropanol extract was centrifuged, and two volumes of supernatant were removed from above the cell debris and added to the separate container containing MeOH/PBS supernatant.

These mixed supernatants were dried by vacuum centrifugation, and stored at -20 °C. Prior to LC-MS analysis, stored dried extracts were reconstituted in 100 uL of 2:2:1 (v:v:v) methanol:isopropanol:water and filtered through 0.22 um PVDF microfilters prior to LC-MS analysis.

Reverse phase chromatography was performed using 4 μL of extract injected onto a 150 × 0.5 mm Zorbax C-18 column (Agilent Technologies, Santa Clara, CA, USA) with 5 mm particles at a flow rate of 20 L/min. Buffer A was HPLC water with either 0.1% formic acid or 0.1% acetic acid, and buffer B was acetonitrile with either 0.1% formic acid or 0.1% acetic acid. Chromatography was performed as follows: the column was equilibrated in 3% buffer B, held for 3 min, and then the gradient was 3 - 50% buffer B over 5 min, 50 - 99% buffer B over 25 min, held at 99% for 10 min, and then re-equilibrated in 3% buffer B for 10 min. Sample run orderwas randomized to control for carry-over and instrument drift.

Positive mode data was collected over a range of m/z 50-2000 using an Agilent Technologies (Santa Clara, CA, USA) ESI-qTOFModel 6520 (typical resolution 14k)with a capillary voltage of 4,000 V, a drying gas flow rate of 3.5 L/min, scan rate of 1.03s/spectra, and a nebulizer gas pressure of 10 psi. Each raw data file, was converted to mzXML using Trapper (Agilent Technologies).

1.2Data visualization in Kendrick mass defect diagrams

The Kendrick mass or Kendrick m/z ratio is the mass or m/z value in Daltons multiplied by 14.000000 / 14.015650. On this mass scale, CH2 units have an integer mass. A mass defect diagram plots the integer mass of a feature vs. the mass “defect” or decimal part of the mass [1]. Addition or subtraction of CH2 units (e.g., by replacement of a hydrogen with a methyl group) will result in purely horizontal shifts on a Kendrick mass defect diagram. Addition or subtraction of other moieties will result in diagonal shifts.

1.3Chemical standards.

Authentic standards of ethanolamine, N-methyl ethanolamine, and N,N dimethyl ethanolamine were purchased from Sigma-Aldrich. Authentic standards of 1-palmitoyl-d31-2-oleoyl-sn-glycero-3-phosphoethanolamine, 1,2-diphytanoyl-sn-glycero-3-phosphoethanolamine, and 1-stearoyl-2-Hydroxy-sn-Glycero-3-phosphoethanolamine were purchased from Avanti Polar Lipids, Inc.

1.4Estimation of absolute abundance for lyso phospholipids and comparison to diacyl lipid abundance

Absolute quantification of the lyso lipids in bioreactor-derived AMD biofilm was made by spiking between 7.6 and 253 nmol of 18:0 lysoPE (Avanti Polar Lipids) into lyophilized biomass samples before extraction. To ensure complete lipid extraction, a lipid-specific extraction protocol was used: biomass samples were freeze-dried overnight (Labconco FreeZone2.5, Labconco, Kansas City, MO) and extracted with a 2:1 methanol:chloroform solvent system. 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). Extracts clarified after centrifugation were centrifuged under vacuum and resuspended in 2:2:1 isopropanol:methanol:water and analyzed by LC-MS as described above.

The 18:0 lysoPE lipid is isobaric with the methylated 17:0 lipid naturally found in the sample and elutes at a very similar chromatographic retention time. It fragments by CID in analogous manner to the methylated 17:0 lipid, but fragment ions have distinct m/z values because the C18:0 lipid has a longer fatty acid chain and no N-methylation. For example, the natural methylated 17:0 lipid spectrum has a prominent peak at 327 Da, while the 18:0 lyso PE standard has an internal has a peak at 341 Da. Across the range in internal standard loadings used (between 7.6 and 253 nmol), instrument response was approximately linear. Absolute abundance of the 482 lipid was estimated from the peak intensity ratio of the internal standard and the sample-derived methylated 17:0 lipid and knowledge of the absolute amount of internal standard loaded.

1.5Data processing.

For metabolomics profiling, data for each sample were converted to mzXML format and analyzed using the software xcms[2], which provides peak-finding, integration, and feature grouping functionalities. Features co-eluting with the solvent front or a very intense, highly complex envelope of features at 500 seconds were discarded to avoid the complexities of ion suppression and mass red-shifting; only features with retention times of > 600 seconds were retained for further analysis.

For 15N- or proton-painting experiments, MassHunter software (Agilent Technologies, Santa Clara, CA) was used to extract ion chromatograms (at +/- 5 ppm mass tolerance) at m/z values predicted from the m/z value of the unlabeled monoisotopic mass of the target feature and the number of label-able atoms in the target formula.

Hierarchical clustering (Pearson uncentered) of unit vector normalized PE and MMPE MS1 signal intensities was carried out in the MultiExperiment Viewer package (Linkage method: average linkage; leaf order optimization: none;

Strained normalized spectral abundance factor (S-NSAF) values were used as the metric for protein abundance and unit vector normalized metabolite feature intensities were used as the metric for metabolite feature abundance, as decribed previously [3]. In order to avoid potential biases in the data arising from zero values, we focused our analyses on protein and metabolite features that were detected in all of the sampled communities. The R programming language was used to calculate the Pearson product-moment correlation coefficients (r; cor function) and corresponding p-values (cor.pvals function) for all possible protein-metabolite feature pairs. Correlated features and their associated r- and p-values were extracted using custom Ruby scripts (available from the authors upon request). Only r-values with an associated statistical p-value of ≤ 0.0001 were considered. To achieve a balanced analysis, an r value was chosen that relates an equal number of proteins and metabolite features [36]. Regression analysis revealed an equivalent sampling of proteins and metabolite features at r ≥ 0.9285 (n = 370).

Correlation networks at r ≥ 0.9285 for correlated protein-metabolite features, proteins and metabolite features were constructed in Cytoscape version 2.6.3 using pairs of correlated proteins and/or metabolite features (nodes) with their corresponding r (edges) as input and using the circular layout function in the y-files plugin for Cytoscape.

2Mass spectra of authentic ethanolamine standards and MS2 fragments of detected lipids

3Absolute quantification

4Proton painting data for all lipids and example spectra of mixed isomers for m/z = 466 and m/z = 468 features.

5Blast information for putative pmtA genes

Table S1. Blast summary statistics for AMD genes with high homology to known phosphatidylethanolamineN-methyl transferase genes.

ORF / Best hit vs. K00551 / K00551 best-hit Eval
UBA_LeptoII_Scaf_8524_GENE_228 / ava:Ava_2605 / 9.00E-23
5wayCG_LeptoII_Cont_11276_GENE_62 / ava:Ava_2605 / 7.00E-23
UBA_LeptoIII_Cont_9568_GENE_79 / ava:Ava_2605 / 3.00E-26
ORF / Best hit vs. nr / nr best hit Eval
UBA_LeptoII_Scaf_8524_GENE_228 / A. manganoxydans S185-9A1 methyltransferase, EAS49908 / 1.00E-21
5wayCG_LeptoII_Cont_11276_GENE_62 / A. manganoxydans S185-9A1 methyltransferase, EAS49908 / 4.00E-22
UBA_LeptoIII_Cont_9568_GENE_79 / S. viridis DSM43017 methylase in quinone biosynthesis YP_003132138.1 / 3.00E-25

6Correlation analysis of combined metabolomic and proteomic data

7Supplemental References

[1]C.A. Hughey, C.L. Hendrickson, R.P. Rodgers, A.G. Marshall, and K.Qian. Kendrick mass defect spectrum: a compact visual analysis for ultrahigh-resolution broadband mass spectra. Anal. Chem, 73(19):4676–4681, 2001.

[2]ColinA Smith, ElizabethJ Want, Grace O'Maille, Ruben Abagyan, and Gary Siuzdak. Xcms: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal Chem, 78(3):779–787, Feb 2006.

[3]P.Wilmes, B.P. Bowen, B.C. Thomas, R.S. Mueller, V.J. Denef, N.C. VerBerkmoes, R.L. Hettich, T.R. Northen, and J.F. Banfield. Metabolome-Proteome Differentiation Coupled to Microbial Divergence.mBio, 1(5):e00246, 2010.

1