Supplementary Materials for:

An untargeted metabolomics method for archived newborn dried blood spots in epidemiologic studies

Lauren Petrick1, William Edmands1, Courtney Schiffman2, Hasmik Grigoryan1, Kelsi Perttula1, Yukiko Yano1, Sandrine Dudoit2,3, Todd Whitehead4,5, Catherine Metayer4,5, Stephen Rappaport1,5*.

1.  Department of Environmental Health Sciences, School of Public Health, University of California, Berkeley, CA, USA

2.  Division of Biostatistics, School of Public Health, University of California, Berkeley, CA, USA

3.  Department of Statistics, University of California, Berkeley, CA, USA

4.  Division of Epidemiology, School of Public Health, University of California, Berkeley, CA, USA

5.  Center for Integrative Research on Childhood Leukemia and the Environment, University of California, Berkeley, CA, USA.

*Corresponding Author: S. Rappaport,

1 Summary statistics for newborn DBS Summary statistics were calculated from the metadata to describe the relationships among the population covariates of the 103 subjects used in the statistical analysis (Table S1). These statistics were obtained using the ‘t.test’ function in R, specifying a two-sided alternative. P- values for the associated t-statistics should be taken with caution, as the set of assumptions behind the two sample t-tests was not evaluated. The males in this sample were typically 282 g heavier than females (p-value = 0.010) and their DBS were typically archived two years longer (p-value = 0.023). No significant differences were detected across groups regarding potassium levels or ethnicity.

Table S1. Summary statistics of characteristics from the 103 newborn DBS used in the statistical analysis (3 of the total 106 subjects were not included due to missing covariate data).

Sex / Ethnicity
Characteristic / Male / Female / p-value / Hispanic / Non-Hispanic / p-value
Number of samples / 61 / 42 / 53 / 50
Potassium (mM) / Min / 3.33 / 4.05 / 0.496 / 4.09 / 3.33 / 0.293
Mean / 5.83 / 6.00 / 6.02 / 5.72
Max / 9.94 / 8.69 / 8.69 / 9.94
Child birth year (y) / Min / 1988 / 1985 / 0.023 / 1985 / 1985 / 0.803
Mean / 1996 / 1994 / 1995 / 1995
Max / 2005 / 2006 / 2006 / 2005
Birth weight (g) / Min / 2041 / 2438 / 0.010 / 2438 / 2041 / 0.518
Mean / 3594 / 3312 / 3445 / 3515
Max / 4961 / 4451 / 4961 / 4933

2 Optimization of DBS extraction

Triplicate 4-mm punches obtained from experimental DBS and the adjacent filter paper were placed in vials and extracted with 400 mL of each of the following solvent mixtures: acetonitrile/water (2/1), acetonitrile/water (4/1), methanol/water (2/1), methanol/water (4/1), methanol/ethanol/water (1/1/1), and methanol/ethanol/water (2/2/1) (Bruce et al. 2009; Raju et al. 2016; Want et al. 2006). The tubes were agitated at 1,400 rpm at 37°C for 1 h and stored at -20°C overnight to precipitate proteins that were removed by filtration prior to analysis. Each extract was analyzed in duplicate by nanoLC-HRMS using an Agilent 1260 Infinity nano/capillary-flow HPLC system and Chip Cube™ interface connected to an Agilent 6550 QTOF HRMS (Santa Clara, CA, USA) operated in nESI (+) mode. An Agilent SmMol-Chip-43 Chip was utilized for chromatographic separation, consisting of a 43-mm 80 Å C18 analytical column combined with a 40 nL trapping column. Mobile phases were A: 0.1% formic acid in water and B: 0.1% formic acid in methanol. After loading 0.5 mL of an extract on the trapping column at 3 mL/min, the analytical separation was performed at 0.4 mL/min with the following gradient: 1 to 50% B from 0 - 3 min, 50 to 99% B from 3 - 12 min, 99% B for 12-16 min, followed by return to initial conditions from 16-21 min. Mass spectra were acquired at 2 spectra/s between 50-1000 m/z, with source gas temp = 290°C, gas flow = 12 L/min, Vcap = 1850 V, and fragmentor = 175 V.

Optimal extraction conditions were determined by comparing numbers of features in DBS extracts that were present with mean abundances of at least twice those of blank-filter extracts and with CVs < 30%, as shown in Table S2. The largest numbers of features were observed in extracts obtained with the two mixtures of acetonitrile/water. While the 2/1 acetonitrile/water had about 20% more features than the 4/1 mixture (4857 vs. 4076), broad chromatographic peaks characteristic of proteins were observed in the 2/1 extracts but not in the 4/1 extracts. Thus, extraction with 4/1 acetonitrile/water (at 37°C) was chosen for analysis of archived DBS.

Table S2. Comparison of features detected by extracting DBS and filter blanks with different solvents.

Solvent / No. featuresa
MeOH/H2O (2/1) / 2,655
MeOH/H2O (4/1) / 3,609
ACN/H2O (2/1) / 4,857
ACN/H2O (4/1) / 4,076
MeOH/EtOH/H2O (1/1/1) / 3,639
MeOH/EtOH/H2O (2/2/1) / 3,366

aNumbers of features shown are those that had at least a two-fold higher abundance in blood extracts compared to blank-filter extracts and a coefficient of variation less than 30%.

3 UHPLC-MS details Reference masses for real time mass calibration were m/z 119.0363 and 805.9859 and were delivered by an isocratic pump (Agilent, Santa Clara, USA). Dual ESI source settings included gas temp = 290°C, drying gas = 14L/min, nebulizer pressure = 45 psig, Vcap = 3500V, and fragmentor = 175 V. High purity (>95%) nitrogen for drying gas and nebulizer pressure was supplied by a N2 gas generator (Peak Scientific, Billerica, MA). Tandem MS analysis was performed in auto MSMS mode as follows: MS and MSMS acquisition rates of 4 and 2 spectra/s with a medium isolation width (4 m/z), fixed collision energies of 10, 20, 40 and 50V. Dynamic exclusion mode was enabled using the following scheme: 2 spectra per precursor with a release after 0.5 min. Precursor threshold was set at 2000 counts or 0.01%, charge state preference was 1 using an algorithm for recognition of small molecules, precursor purity stringency was set at 75% with a 25% cutoff. The precursor TIC target was 25000 counts/s.

Fig S1. Sample calibration curve used for converting mV measured with an ion selective electrode in the DBS extracts to potassium concentration in M. Note the log scale on the x-axis. The slope was similar to that suggested by the manufacturer (~55 mV).

Fig S2. Plot of potassium concentration versus total usable signal for each subject after adjustment for run order and log TUS.

References

Bruce, S. J., Tavazzi, I., Parisod, V., Rezzi, S., Kochhar, S., & Guy, P. a. (2009). Investigation of human blood plasma sample preparation for performing metabolomics using ultrahigh performance liquid chromatography/mass spectrometry. Analytical Chemistry, 81(9), 3285–3296. doi:10.1021/ac8024569.

Raju, K. S. R., Taneja, I., Rashid, M., Sonkar, A. K., Wahajuddin, M., & Singh, S. P. (2016). DBS-platform for biomonitoring and toxicokinetics of toxicants: proof of concept using LC-MS/MS analysis of fipronil and its metabolites in blood. Scientific Reports, 6, 22447. doi:10.1038/srep22447.

Want, E. J., O’Maille, G., Smith, C. A., Brandon, T. R., Uritboonthai, W., Qin, C., et al. (2006). Solvent-dependent metabolite distribution, clustering, and protein extraction for serum profiling with mass spectrometry. Analytical Chemistry, 78(3), 743–752. doi:10.1021/ac051312t.