Quantitative Proteomics Using ¹⁵N SILAC Mouse

InvitrosolTM was purchased from Invitrogen (Carlsbad, CA). Trypsin (modified, sequencing grade) was obtained from Promega, WI. Other laboratory reagents were purchased from Sigma-Aldrich (St. Louis, MO) and Thermo Fisher Scientific (Waltham, MA) unless noted otherwise. ¹⁵N-enriched Spirulina was purchased from Cambridge Isotope Laboratories (Andover, MA). ¹⁵N-enriched Ralstonia was purchased from Silantes GmbH (Germany). All C57/BL mice were obtained from in-house breeding colony. The protein-free mouse diet was purchased from Harlan Laboratories Inc. (Indianapolis, IN). Human brain tissue was purchased from BioChain Institute Inc. (Newark, CA) and homogenized using the same protocol as the ¹⁵N SILAC labeled tissues.

10 grams of Spirulina biomass or Ralstonia were mixed with 30g protein-free diet powder (Harlan TD08152) and H2O to make food pellets as described previously²⁰.

Protocol 1 without adaptation. 6-8 weeks old sexually mature C57/BL female mice were immediately placed on 14N-unlabeled diet (Spirulina or Ralstonia) for 1 week (acclimation) and then mated with C57/BL male mice. Pregnant female mice were continued in the study and fed strictly ¹⁵N-enriched diet throughout the pregnancy and nursing. The weight and size of pups were recorded from day 7 (p7) to day 16 (p16).

Protocol 2 with adaptation. 6-8 weeks old sexually mature C57/BL female mice were fed with a mixed diet starting from 70:30 mix of normal chow and ¹⁵N-enriched Ralstonia, then 40:60 mix of normal chow and ¹⁵N-enriched Ralstonia, then 10:90 mix of normal chow and ¹⁵N-enriched Ralstonia, and finally 100% ¹⁵N-enriched Ralstonia. The weights of female mice on the adaptation diet were monitored closely to ensure no significant weight loss during each stage of diet adaptation before moving on to the next stage of adaptation. Adapted females were then maintained on the 100% ¹⁵N-enriched Ralstonia diet for mating, pregnancy, and nursing pups.

All methods involving animals were approved by the Institutional Animal Research Committee and were accredited by the American Association for Accreditation of Laboratory Animal Care.

Animal tissues were harvested, snap frozen in liquid nitrogen, and stored at -80^oC. 0.5mg of tissue was homogenized in mass spectrometry-compatible lysis buffer (50mM Ammonium Bicarbonate, 4M Urea, 1X invitrosol, protease inhibitors, and phosphatase inhibitors) Protease and phosphatase inhibitors were obtained from Roche Applied Science (Indianapolis, IN). Protein concentration was determined using an EZQ protein assay (Invitrogen, CA). 30µg of protein lysates from the young and old mouse brains were diluted 1:1 in 50mM Ammonium Bicarbonate for trypsin digestion. Trypsin was added to each sample at a ratio of 1:30 enzyme/protein along with 2 mM CaCl₂ and incubated for 16 hours at 37^oC. Following digestion, all reactions were acidified with 90% formic acid (2% final) to stop proteolysis. Then, samples were centrifuged for 30 minutes at 14,000 rpm to remove insoluble material. The soluble peptide mixtures were collected for LC-MS/MS analysis.

Peptide mixtures were pressure-loadedonto a 250 µm inner diameter (i.d.) fused-silica capillary packed first with 3 cm of 3 µm strong cation exchange material (Partisphere SCX, Whatman), followed by 5 µm of 10 mm C18 reverse phase (RP) particles (Aqua, Phenomenex, CA). Loaded and washed microcapillaries were connected via a 2 µm filtered union (UpChurch Scientific) to a 100 µm i.d. column, which had been pulled to a 5 mm i.d. tip using a P-2000 CO₂ laser puller (Sutter Instruments), then packed with 13 cm of 5 µm C18 reverse phase (RP) particles (Aqua, Phenomenex, CA) and equilibrated in 2% acetonitrile, 0.1 % formic acid (Buffer A). This split-column was then installed in-line with a NanoLC Eskigent HPLC pump. The flow rate of channel 2 was set at 300 nl/min for the organic gradient. The flow rate of channel 1 was set to 0.5µl/min for the salt pulse. Fully automated 13-step chromatography runs were carried out. Three different elution buffers were used: 2% acetonitrile, 0.1 % formic acid (Buffer A); 98% acetonitrile, 0.1% formic acid (Buffer B); and 0.5 M ammonium acetate, 2% acetonitrile, 0.1% formic acid (Buffer C). In such sequences of chromatographic events, peptides are sequentially eluted from the SCX resin to the RP resin by increasing salt steps (increase in Buffer C concentration), followed by organic gradients (increase in Buffer B concentration). The last chromatography step consists in a high salt wash with 100% Buffer C followed by an acetonitrile gradient. The application of a 2.0 kV distal voltage electrosprayed the eluting peptides directly into a LTQ-Orbitrap XL mass spectrometer equipped with a nano-LC electrospray ionization source (ThermoFinnigan). Full MS spectra were recorded on the peptides over a 400 to 2000 m/z range by the Orbitrap, followed by five tandem mass (MS/MS) events sequentially generated by LTQ in a data-dependent manner on the first, second, third, fourth, and fifth most intense ions selected from the full MS spectrum (at 35% collision energy). Mass spectrometer scan functions and HPLC solvent gradients were controlled by the Xcalibur data system (ThermoFinnigan, San Jose, CA).

Standard statistical methods were performed to analyze the MS data sets as described and referenced in the manuscript. Pearson Correlation Analysis and Coefficient of Variation analysis were used to survey correlations of overall protein profiles between and within groups. LIMMA was employed to assess statistical significant changes of protein expression in different samples based on linear models ²¹. To match human and mouse peptides, the following corrections were applied to the dataset. First, we allowed the match of identical sequences with interchangeable amino acids such as I/L and K/Q. Then, we allowed the match of peptide sequences wherein two amino acids are switched (e.g. AD and DA) whereas the rest is identical.

The feasibility of labeling rats and mice metabolically with a diet containing ¹³C⁶ lysine or ¹⁵N labeled amino acids ^{20, 22, 23} has been reported. The labeled protein source provides free ¹⁵N amino acids after the animal breaks down the food, which can be used to supply de novo synthesis of proteins. Although stable isotope labeled amino acids have not been reported to cause severe toxicity, we found low fertility of female breeders and lower incorporation of ¹⁵N labeled amino acids using the commonly used acclimation procedure (acclimated to the ¹⁴N versions of both diets for 1 week) using both algae-based ¹⁵N Spirulina and bacteria-based ¹⁵N Ralstonia ²⁴. Furthermore, pups born from the labeled females remained underweight compared to pups born from females on the regular mouse chow (Figure 1A). Consequently, lower incorporation efficiency of ¹⁵N amino acids was found in most tissues derived from these underweight pups (Figure 1B, left panel).

Figure 1A.Metabolic labeling of C57/BL mice using the adaptation scheme. (A). Growth curve of pups from day 7 after birth (p7) to day 16 (p16). Weights of pups born from females acclimated with 14N Spirulina, 14N Ralstonia, or normal chow (ave. n = 15 pups from 3 litters per group) were measured. Weights of pups born from females adapted with 15N Ralstonia (ave. n = 15 pups from 3 litters) were measured.

Figure 1B.Incorporation of 15N amino acids in ATP synthase beta subunit in the heart, lung, and ovary from 15N labeled mice with or without adaptation as shown by the mass spectrum of one of its peptides.

To increase the labeling efficiency of C57BL mice, we optimized the feeding protocol by gradually adapting the female breeders to the ¹⁵N labeled Ralstonia. Instead of one week acclimation with the ¹⁴N diet, however, we gradually mixed ¹⁵N-Ralstonia-based diet with the normal mouse chow and monitored the weights of these mice to determine there was no weight loss before each change in mix (30% every 1-2 weeks) as illustrated in Figure 1C. Using the new feeding protocol, we were able to normalize the weight gain of pups born from ¹⁵N-Ralstonia-adapted females (Figure 1A). Also, a significant improvement of ¹⁵N amino acid incorporation was detected in tissues derived from the newly labeled pups (Figure 1B). The global incorporation efficiency was determined by LC-MS/MS method and analyzed by SEQUEST by identifying both the light and heavy versions of peptides. Lastly, we observed no significant differences in litter sizes between ¹⁵N-Ralstonia-adapted females and unlabeled females (Figure 1D).

Figure 1C.The adaptation scheme of the new metabolic labeling.

Figure 1D.Averaged litter size of 1st and 2nd generations from 15N labeled female mice using the adaptation scheme or female mice fed with the normal chow. Three litters were measured per group.

To evaluate the variability of ¹⁵N SILAC And Label-Free Quantificationgenerated from biological replicates, we acquired quantitative MS data with or without tissue-matched internal standards from ¹⁵N SILAC mice from the same biological samples. Frontal cortex tissues were obtained from three young C57BL male mice (4 weeks old) and three old C57BL male mice (12 months old). Tissue lysates were generated using MS-compatible lysis buffer (see Methods) and analyzed by LC-MS/MS with or without brain lysates generated from ¹⁵N SILAC mice. The overall experimental strategy is summarized in Figure 2A.

Overall, the spectral counting (SC) method yielded 1491 proteins with at least 2 unique peptides at a 1% false discovery rate (FDR). The ¹⁵N SILAC internal tissue standard quantification method yielded 660 quantifiable proteins (at least two quantified peptides and < 20% CV among quantified peptides) at a 1% FDR. 517 proteins were identified in both dataset and used for subsequent evaluations of these two methodologies. For the SC method, normalization was performed based on the sum of unweighted spectrum counts from each MS run, and scaling factors were applied to equalize unweighted spectrum counts of all MS runs. For the ¹⁵N SILAC tissue quantification method, normalization was assessed based on ¹⁴N/¹⁵N ratios between the unlabeled to the ¹⁵N labeled standards to ensure medians are approximately 1 since ¹⁴N and ¹⁵N tissue samples are theoretically mixed at a 1:1 ratio. After normalization, distribution of log₂ transformed expression values are shown in Figure 2B for both ¹⁵N SILAC tissue quantification (log2(ratio)) and SC methods (log2(SC)). The box plot in Figure 2B reveals that log₂ transformed ratios are centered on zero (1:1) and log₂ transformed SC have similar distributions among samples, indicating sufficient normalizations of both datasets.

Next, we performed a series of statistical analyses to study inter and intra-group variations of protein lists from these two methodologies. Using the Pearson Correlation Analysis, ¹⁵N SILAC tissue quantifications showed higher correlations within biological replicates of the same age group and lower correlations among the biological replicates between two age groups, suggesting differential protein expression patterns in the brains of old and young mice (Figure 2C). In contrast, quantitative comparisons based on the spectral counting method showed similar correlations within biological replicates and between the two age groups (uniformly > 0.9). Collectively, our data reveals that SC quantification is less sensitive in detecting subtle changes, such those that occur during aging, than ¹⁵N SILAC mouse quantification.

Figure 2.Statistical analyses of 15N SILAC labeled and label-free quantifications of mouse brain tissues at two different ages. (A). Schematic illustration of the experimental design for the comparison. (B). Quantile normalization of ratios (15N SILAC labeled) or spectra counts (label-free) quantifications of young and old brain tissue lysates. Three biological replicates were analyzed by each quantification method. (C). Pearson correlation between young and old brain tissue lysates by 15N SILAC labeled or label-free quantifications. (D). Comparisons of triplicate values (log2) from proteins commonly identified in 15N SILAC labeled and label-free quantifications. (E). Volcano plot of ratios and p-value of 15N SILAC labeled and label-free quantifications. Fold changes were derived from ratios or spectra counts of proteins from the young brain tissue lysate divided by the old brain tissue lysate. Weighed average of ratios or spectra counts were calculated to derive the log2(fold change) as shown in the x-axis. The y-axis represents the –log(adjusted p-value). Proteins fall above p< 0.05 (above 1.3 on the volcano plot) and above/below 2 fold changes (between 1 to -1 in the volcano plot) are identified as proteins with significant changes (red).

We also examined the within-group variations of these two methodologies. Log2(ratio) values (¹⁵N SILAC tissue quantification) or log2(SC) values (spectra count quantification) from three biological replicates were plotted for the final list of 517 proteins. Again, the SC approach had noticeably higher variations (average coefficients of variation = 23.1% young, 37.6% old) among biological replicates than did ¹⁵N SILAC tissue quantifications (average coefficients of variation = 9.4% young, 13.6% old) (Figure 2D). We speculated that the greater run-to-run variations of the SC approach could mask biologically relevant but subtle changes in protein expression, and therefore further interrogated the significance of both datasets using LIMMA statistics. LIMMA statistics are widely used to analyze complex experiments such as microarray and large-scale proteomic analysis, which involve comparisons among many targets simultaneously ²¹. The empirical Bayes moderated t test was applied to obtain estimated error rates of individual proteins in relation to other proteins in brain tissues derived from the young or old mice. Also, the Benjamini-Hochberg method was used to correct the raw p-values for multiple testing (adjusted p-values) ²¹. Finally, volcano plots based on adjusted p-values and fold change of protein quantification from both methods were used to demonstrate statistically significant changes associated with aging (Figure 2E). Using the criteria of greater than 2 fold changes and FDR < 0.1, 118 (17.9%) proteins were identified as significantly different between the young and old mouse brains (57 upregulated and 61 downregulated) based on the ¹⁵N SILAC mouse quantification (Table 1). Using the same criteria, only 11 (0.7%) proteins were identified as significantly different between the young and old mouse brains (9 upregulated and 2 downregulated) based on the SC quantification. Collectively, we report that the ¹⁵N SILAC mouse quantification method is superior in detecting smaller changes in expression levels given high reproducibility among replicates. However, the SC based quantification method remains useful in detecting large fold changes but with high false positive rate due to large variations among replicates.

Finally, to extend the use of ¹⁵N SILAC mice for global proteome quantification, we studied the possibility of quantifying human tissues using tissue-matched standards from ¹⁵N SILAC mice as the availability of cell lines and concerns of recapitulating tissues with two dimensional tissue culture conditions pose significant limitations on quantitative analysis of human tissues using SILAC labeled cell lysates. As a proof of concept experiment, we spiked an equal amount of protein lysate from ¹⁵N SILAC mouse brains into tissue lysates from human brain and determined the percentage of quantifiable human peptides based on mouse peptides with identical amino acid sequences (see method). Two biological replicates were analyzed. In replicate 1, we found 2993 (60%) quantifiable human peptides from a total of 5027 human peptides identified. In replicate 2, we found 3025 (60%) quantifiable human peptides from a total of 5040 human peptides identified. Data from two technical replications were merged and analyzed for overall ratio distribution. The ratio distribution of quantifiable peptides was unimodal and >75% of peptides were within a 3-fold ratio between human and mouse proteins (Figure 3A). There was also a good agreement with ratios derived from human and mouse peptides between two technical replicates since more than 70% of proteins were within 25% CV (Figure 3B & Suppl. Table 1).

Figure 3.Overall ratios and coefficient variances of human-mouse 15N SILAC quantification. (A). 15N labeled mouse brain lysates were spiked into unlabeled human brain tissue lysates. Protein ratios derived from the human mouse paired peptides from two replications are plotted against the number of proteins (frequency) within the binned ratio. (B). Coefficient variation (CV) analysis of number of peptides within the binned CV.