Duplexing metabolic deuterated water-labeled samples using dimethyl labeling to estimate protein turnover rates

Abstract

Metabolic labeling with deuterated water (D₂O), combined with liquid chromatography coupled to mass spectrometry (LC-MS), is used to study turnover rates of individual proteins in vivo. Often, protein turnover rates from two (treatment and control) conditions are compared. The turnover rates are obtained from time course LC-MS experiments for each condition. Here, we explore dimethyl duplexing to directly compare label enrichment from two D₂O-labeled samples in a single LC-MS experiment. Light and heavy dimethyl channels carry two metabolically labeled samples. Protein turnover is modeled as a time course of the ratio of the monoisotopic relative abundance of a peptide in the heavy dimethyl channel to that in the light dimethyl channel. The turnover rates computed using the duplexed samples were close to those of non-duplexed samples. We illustrate the advantages of this method through the analysis of liver proteome dynamics in rapamycin-treated mice compared to control mice.

Dimethyl duplexing of metabolically deuterated water-labeled samples addresses challenges in dynamic proteome studies by enabling direct comparison of label enrichment. Here, the authors demonstrate that this approach allows accurate protein turnover modeling and efficient analysis of liver proteome dynamics in rapamycin-treated versus control mice.

Introduction

Protein abundances change in response to internal or external stimuli by adjusting the balance of protein synthesis and degradation. Metabolic labeling with deuterated water, ²H₂O, (also referred to as “heavy” water or D₂O), followed by liquid chromatography coupled with mass spectrometry (LC-MS), is a powerful high-throughput tool to study the turnover of individual proteins in vivo^1–4. The deuterium from the ²H₂O labels amino acids that are subsequently incorporated into peptides and proteins. Deuterium-enriched water is safely provided as drinking water in low enrichments (less than 8% enrichment of body water)⁵ and does not require diet adaptation. The isotope equilibrium in the body water is achieved nearly instantaneously⁶. The ease of use and cost-effectiveness are some of the advantages of deuterated water labeling compared to other in vivo labeling approaches, such as ¹⁵N labeling^7,8 (labeling using ¹⁵N algae diet) or labeling with stable isotopes of an essential amino acid (e.g., ¹³C₆-Lys⁹ or ²H₃-Leu¹⁰). Deuterated water has been used to label various living organisms, including chicken^11,12, rodents^13–16, human^17–21, and dog²², as well as cell cultures^23–27.

In labeling experiments, deuterated water is administered for a certain duration of time that considers the species and expected turnover rates. At the conclusion of labeling, cell or tissue samples are collected, and proteins are extracted and enzymatically digested. The resulting peptides are analyzed using LC-MS²⁸. The peptides are identified from their tandem mass spectra (MS/MS) and mass-to-charge ratios (m/z) using protein sequence databases, while deuterium enrichment is measured using isotope profiles²⁹. As the newly synthesized proteins incorporate deuterium-labeled amino acids, the relative abundance (RA) of the monoisotope decreases. The turnover rate is estimated by modeling the time course of the monoisotopic RA depletion with an exponential decay function³⁰. The labeling results in a composite isotope profile of labeled and unlabeled forms of a peptide in LC-MS^31,32. Combining two samples (e.g., labeled and unlabeled samples, or samples labeled under control and treatment conditions) in a single run (and after normalization for the plateau of the label enrichment in each sample) will allow a direct comparison of fractional syntheses between samples.

In this study, we examined a new experimental approach to study protein turnover using LC-MS. We duplexed the unlabeled and deuterium-labeled samples using dimethyl labeling³³, Fig. 1. In dimethyl labeling, the side-chain amino group of the Lys residue and the N-terminal amino group of a peptide are chemically labeled using formaldehyde and cyanoborohydride. Up to three samples can be multiplexed in dimethyl labeling³⁴ by using stable isotope-labeled formaldehyde and sodium cyanoborohydride. We used the light (to label unlabeled samples, a mass difference of 28 Da) and heavy channels (to label deuterium-labeled samples, a mass difference of 36 Da), which are separated by 8 Da (per each labeling site). By combining the deuterium-labeled and unlabeled samples (two-plex samples) in the same LC-MS experiment, we determined the ratio of a peptide’s monoisotopic RAs in the deuterium-labeled (I₀(t)) and unlabeled (I₀(0)) samples. The time course of the ratio of monoisotopic RAs (I₀(t)/I₀(0)) was modeled as an exponential decay function to determine the turnover rate. In another application, we duplexed proteomes of rapamycin-treated and control samples. Both samples were metabolically labeled with deuterated water for 8 days. The isotope profiles in the heavy and light dimethyl channels were used to compare the fractional syntheses and estimate the effect of the rapamycin treatment.

Fig. 1. Two approaches to estimate protein turnover rates using deuterated water-labeled samples.

Mice are given deuterated water for specific labeling durations, and tissues are collected at preplanned time points to determine the level of deuterium incorporation into newly synthesized proteins. In the traditional approach, the unlabeled (day 0) and deuterated water-labeled samples (day 21) are prepared separately and analyzed in different liquid-chromatography coupled to mass spectrometry (LC–MS) experiments. In the dimethyl duplexing approach, unlabeled and deuterated samples undergo dimethyl labeling with distinct mass differences, are combined into a single mixture, and analyzed in a single LC–MS run. This duplexing strategy reduces instrument time and enables direct comparison of isotope distributions of a peptide in two samples obtained under two different conditions (e.g., unlabeled/metabolically labeled, control/treatment). Representative spectra illustrate the shift in isotopic patterns between day 0 and labeled samples for each workflow. Created with BioRender.com.

Results and discussions

We combined deuterated water metabolic labeling with dimethyl labeling to facilitate the estimation of protein turnover rates. The experimental approach provides several possible routes for analysis of proteome dynamics. Using dimethyl labeling (light and heavy channels), we duplexed each labeled sample with the unlabeled sample. In seven experiments, the samples were duplexed in pairs as follows: (0, 0), (0, 1), (0, 3), (0, 5), (0, 7), (0, 15), and (0, 21). The numbers in each parentheses denote the metabolic deuterated water-labeling duration. The light dimethyl channel always contained unlabeled samples. The heavy dimethyl channel contained a complete time course of the samples metabolically labeled with deuterated water. For example, (0, 5) denotes the experiment in which five-day metabolically labeled and unlabeled samples were duplexed using dimethyl heavy and light channels, respectively.

Dimethyl labeling labels the N-termini of peptides and the side chain of Lys residue. The light dimethyl labeling channel results in a 28.0313 Da mass increase per labeling site. The heavy dimethyl labeling channel results in a 36.0757 Da mass increase (in comparison with the light dimethyl channel, per dimethyl group, six ¹H are substituted with six ²H, and two ¹²C are substituted with two ¹³C) per labeling site³⁵. A representative LC-MS scan of a pair of channels of unlabeled and labeled (with deuterium) forms of a peptide is shown in Fig. 2. The figure depicts the two-plex MS1 spectrum of the AIAEELAPER⁺² peptide of the betaine-homocysteine S-methyltransferase 1 protein. The peptide, AIAEELAPER⁺², has one labeling site at the N terminus. The spectrum was obtained from deuterium-labeled and unlabeled murine liver samples, which were duplexed using stable isotope dimethyl labeling^33,34. The unlabeled samples were labeled with the dimethyl light channel, and the metabolically D₂O-labeled (for 5 days) sample was labeled with the heavy dimethyl channel. The m/z of the monoisotope of the unmodified peptide is 549.79312 Th (Thompson). The labeling with a dimethyl light channel shifts the monoisotopic m/z to 563.8088 Th. For the peptide AIAEELAPER⁺² of the deuterium-labeled sample, the monoisotopic m/z is 567.831 Th. Figure 2 also shows the isotope profiles of the peptide from unlabeled and labeled (with deuterium) samples, which were recorded in a single MS1 scan. This combination allows us to compute the label incorporation from a single experiment. Supplementary Fig. 1 shows the duplexed isotope profiles of this peptide from unlabeled and labeled samples for the rest of the labeling durations used in this study (0, 1, 3, 7, 15, and 21 days), which were duplexed in pairs as (0,0), (0,1), (0,3), (0,7), (0,15), and (0,21). From each pair of isotope profiles in these experiments, we compute the ratio (I₀(t)/I₀(0)) of the RAs of the monoisotopes at that time point of labeling. Supplementary Fig. 2 presents the experimental time course of the I₀(t)/I₀(0) ratio for the peptide AIAEELAPER⁺². Theoretical fits, derived from Eqs. (2) and (3) in the “Methods” section, alongside the experimental time course of the depletion of the monoisotopic RA obtained exclusively from the heavy dimethyl channel for this peptide, are displayed in Supplementary Figs. 1, 3, and Fig. 2 demonstrate a distinct separation between the light and heavy dimethyl labeling channels, which enables us to quantify the RAs of the mass isotopomers from different labeling durations in the same experiment. Furthermore, the spectral plots show the depletion of monoisotopic RA as the labeling duration increases in comparison with the unlabeled samples. The ratio of monoisotopic RAs between deuterium-labeled and unlabeled samples obtained from the same experiment is used to quantify the turnover rates of the peptides.

Fig. 2. MS1 spectrum of AIAEELAPER⁺² peptide of the betaine-homocysteine S-methyltransferase 1 protein from duplexed labeling of dimethyl and deuterated water.

The light dimethyl channel carries unlabeled (with deuterium) peptide, and the heavy dimethyl channel carries the labeled form of the peptide (metabolic labeling with deuterated water for 5 days). The relative abundance of the monoisotopes of the labeled (heavy dimethyl channel) and unlabeled (light dimethyl channel) forms of a peptide can be determined from the mass spectrum.

In Supplementary Fig. 4A–F, we show scatter plots of monoisotopic RAs of pairs of unlabeled versus theoretical and labeled peptides versus experimental unlabeled peptides for each time point of labeling for all peptides. The figures clearly show the gradual labeled enrichment (gradual reduction in the monoisotopic RA), as the labeling duration increases. The data qualitatively demonstrate the validity of our approach to the estimation of protein turnover rates from duplexed samples.

We compared the turnover rates computed by the quantification using mass isotopomers in light (unlabeled sample) and heavy (labeled samples) channels and the rates computed from the mass isotopomers in the heavy (labeled samples) channel of dimethyl experiments (using a single-sample approach). The pairing of samples allows turnover rate calculations (described in the “Methods” section) using both methods: from heavy channel only using Eq. (2), and from duplexed samples using Eq. (3). Thus, as discussed in detail in Supplementary Note, for Eq. (3), we determined the time course of the I₀(t)/I₀(0) ratio for every peptide using the isotope distribution of a peptide in the heavy channel for deuterated form of the peptide (I₀(t)), and the isotope distribution in the light channel for the unlabeled form of the peptide (I₀(0)). On the other hand, the heavy channels in the duplexed samples contained all time points for computing the turnover rate of a peptide using Eq. (2) from the labeling time course of the isotope distributions of a peptide in this channel. As an example, in Supplementary Table 1, we show the monoisotopic RAs from each channel for peptide AIAEELAPER⁺². This selection of experiments for dimethyl labeling channels allows the comparison of the turnover rates from Eqs. (2) and (3). The scatter plot of the rates is shown in Fig. 3. Pearson’s correlation between the natural logarithms of the turnover rates was 0.945, and its 95% confidence interval was [0.938, 0.952]. The two methods produced similar results. The density plot of the distribution of relative differences between the two rates is shown in Supplementary Fig. 5 (the relative difference of turnover rates for every peptide was defined as the difference of the turnover rates of the peptide from two methods divided by the average of the turnover rates). As seen from the figure, the mean and median of the differences were less than 4% of the mean of the rates from the two approaches.

Fig. 3. Turnover rates computed using the duplexed samples agree well with those of non-duplexed samples.

The scatter plot of the natural logarithms of turnover rates obtained from quantifications of labeled and unlabeled pairs of mass isotopomers (light and heavy channels in the dimethyl labeling), the x-axis, with the labeled only mass isotopomers (heavy channel in the dimethyl labeling), the y-axis, in dimethyl experiments. The broken red line is the identity line. The color bar on the right displays a range of colors corresponding to different levels of data density.

For the same tissue samples (unlabeled and labeled with deuterium), we also generated LC-MS data for analysis with a traditional approach of one sample per experiment. Tissues from all seven time points of labeling (0, 1, 3, 5, 7, 15, and 21 days) were used. We analyzed these data for protein/peptide turnover rate determination with a traditional approach, Eq. (2). The traditional approach resulted in more peptide identifications (7232 distinct peptides at four or more time points of labeling) than did the dimethyl approach (3304 distinct peptides). The reduction in the number of peptide identifications (20% for precursor ions) due to the increased proteome complexity, which resulted from the addition of a second channel of dimethyl labeling, has recently been reported³⁵ for Orbitrap mass analyzers. An additional factor contributing to the reduced numbers of distinct (for quantification) peptides is the redundancy in the identification of the same peptide in the light and heavy dimethyl channels. Further discussions are provided in the Supplementary Note.

The number of quantified peptides (the coefficient of determination (R²) larger than 0.8) was also higher in the traditional approach (2517) than in the duplexed samples (1263). The scatter plot of the rates for highly abundant peptides (monoisotopic peak higher than 1.5 × 10⁶ in at least one MS1 scan) is shown in Supplementary Fig. 6. For these peptides, we observed a good agreement between the rates computed from the two methods. The dynamic range of turnover rates for these peptides was approximately 62. Overall, in the traditional approach, 60% of all peptide quantifications had R² larger than 0.8, and for 30% of peptides, R² was larger than 0.95 (high goodness of fit between the experimental time course data and the corresponding theoretical fit). In the analyses of the dimethyl labeled samples, these numbers were 77% (R² ≥ 0.8) and 53% (R² ≥ 0.95), respectively. Supplementary Data 1 contains turnover rates of murine liver peptides and proteins computed using the traditional and dimethyl labeling duplexed samples analyzed in this work.

Turnover rate estimations are dependent on the spectral accuracy. The latter often changes with peptide abundance³⁶. We compared abundances of the same peptides between two-plex and single-sample experiments. For several ranges of absolute peptide abundances, we generated box plots of the absolute value of relative differences between the monoisotopic RAs of the same peptides quantified in two-plex and single-sample experiments (for each duration of labeling with deuterated water). The relative difference of RAs (of a peptide in two-plex and single-sample experiments) was defined as the difference between the RAs divided by their average. The box plots are shown in Supplementary Fig. 7. We grouped the peptides into four different groups based on their abundances: [0, 10⁶), [10⁶, 10⁷), [10⁷, 10⁸), [10⁸, 10¹²). The most variability between the RAs was found for peptides whose abundances were 10⁶ or less. This variability increased with the labeling duration. When we analyzed the turnover rate estimations in different abundance groups, the rates in lower abundance groups differed the most (Supplementary Fig. 8). The figure contains scatter plots of the turnover rates from the two-plex (x-axis) and single-sample (y-axis) approaches, as well as the distributions of their relative differences of the turnover rates in three groups of abundances. For the peptides in the highest abundance group (10^7.5; 10¹²), the mean of relative differences was only 10%. In this group, the turnover rates agreed well. As the peptide abundances decreased, the relative difference became large. The medians of relative differences for each abundance group are shown in Supplementary Fig. 8. For 62% of all common peptides, the turnover rates (actual rates, not log-transformed) estimated from both approaches had less than 30% relative difference. The corresponding number for the 50% relative difference was 80%.

The value in the duplexing approach lies in the ability to compare two groups within the same run. To test this idea, we used rapamycin, which is an inhibitor of mTORC1 and a known modulator of protein turnover rates. We compared the label enrichments of murine liver proteins obtained under rapamycin treatment and control conditions. The control and treated samples were duplexed using light and heavy dimethyl channels, respectively. The experiments were done in triplicate. We compared fractional synthesis (fs) for each peptide/protein quantified in both conditions. The computation of fractional synthesis requires the monoisotopic RA of the unlabeled sample, for which we used the theoretical distribution of natural isotopes of atoms. In the formula for fractional synthesis, label enrichment is normalized for the enrichment at the plateau of labeling of a peptide, I₀^asymp:

$$f\mathrm{s}=\frac{I_0\left(0\right)-I_0\left(t\right)}{I_0\left(0\right)-I_0^{\mathrm{asymp}}}$$ 1

I₀^asymp, too, was calculated using the theoretical monoisotopic RA of the peptide (see the “Methods” section for details). The scatter plot of the fs values (computed using Eq. (1)) of all peptides in control and rapamycin-treated samples is shown in Supplementary Fig. 9. However, the raw data does not account for the experimental errors. To account for the experimental fluctuations in the isotope measurements, we generated a null distribution for fractional synthesis. The null distribution assumes that there is no change (due to the metabolic labeling) in fractional synthesis, and the changes are purely due to experimental fluctuations. To generate the null distribution, we used two unlabeled (with deuterium) samples and duplexed them using dimethyl labeling. For data from each channel, we generated a distribution of differences (errors) between experimental and theoretical monoisotopic RAs of all identified peptides. We then divided the differences by ($I_0\left(0\right)-I_0^{\mathrm{asymp}}$) for each peptide to obtain the null distribution of fs values for each dimethyl channel. We used individual null distributions for heavy and light channels, as the distribution of monoisotopic RA errors was different for each channel. The null distributions of fs for heavy and light channels, and the fs distributions of samples labeled for 8 days are shown in Supplementary Fig. 10A, B. The p-values of the fs values of peptides metabolically labeled for 8 days were computed using the corresponding null distribution, as shown in the figures.

Figure 4 shows the scatter plot of the fs values of proteins whose peptides were filtered at a 0.20 significance level in each channel. Protein fs values were obtained as the median of the fs values of their peptides. The scatter plot of the fs values of these peptides is shown in Supplementary Fig. 11. For 72% of the proteins (Fig. 4), fractional synthesis in the rapamycin-treated sample (the y-axis) was less than that in the control sample (the x-axis). In Supplementary Data 2, we provide the fs values of all proteins quantified in the samples labeled for 8 days. In a previous study of skeletal muscle and heart of a heterogeneous line of adult mice (UM-HET3), we showed that male and female mice treated with rapamycin had slightly suppressed bulk protein synthesis in muscle, but minimal effects in heart, and that these outcomes slightly differed by sex³⁷. In the same mouse strain, we showed that bulk protein synthesis rates hid the heterogeneity of the differences in individual protein synthesis rates between rapamycin-treated and controls³⁸. Using the current dimethyl-labeling approach³⁴ in the liver of male adult mice, 71% of proteins showed a decrease in synthesis rate with rapamycin treatment. This change closely mirrors the 67% of proteins that showed a decrease in synthesis rate when using a non-duplexed sample preparation method on the livers of male mice in our previous publication³⁸. No proteins showed opposing directional changes. In the livers from our previous study³⁸, roughly twice as many mitochondrial proteins had lower synthesis rates with rapamycin treatment (log₂FC < 0.95) as had greater synthesis rates (log₂FC > 1.05). In the current study with duplexing, 63% of mitochondrial proteins had lower synthesis rates with rapamycin treatment, while 37% of mitochondrial proteins had higher synthesis rates. Therefore, the duplexing results resemble the overall results from previous non-duplexing approaches.

Fig. 4. Rapamycin treatment slows protein synthesis in murine liver.

The scatter plot shows fractional synthesis (fs) of proteins in the rapamycin-treated (the y-axis) and control (the x-axis) samples. Treatment and control samples were metabolically labeled with deuterated water for 8 days and duplexed using dimethyl labeling. The fs values of peptides of these proteins were filtered to retain only those that had p-values of 0.2 or less (based on the null distribution of peptide fs values). For 72% of proteins, fs in the control sample was higher than that in the rapamycin-treated sample. The broken black line is the identity line. The bar on the right indicates the range of data density represented by different colors.

Conclusion

In this work, we explored the use of dimethyl labeling-based sample duplexing to determine protein turnover rates from LC-MS data of metabolically deuterated-water labeled samples. Peptide turnover rates computed using two-plex samples and single-sample experiments were in good agreement. Dimethyl duplexing can be used in comparative studies, when two samples obtained under different conditions (e.g., treated and untreated, disease and healthy) are compared for differential protein turnover. In such cases, duplexing of treated and untreated samples obtained from labeling for the same duration of time allows computation of fractional syntheses for both samples and directly compares isotope distributions obtained in the same MS1 scans. Such duplexing of unlabeled and labeled samples can be useful in experiments when samples are in short supply (e.g., in human studies) and/or turnover rates are determined from one labeled sample only³⁹. Overall, the number of peptide identifications in dimethyl duplexed samples were lower than that in the non-duplexed samples. Partially, it is due to the redundancy in the identification of peptides in the light and heavy channels.

Methods

Animal labeling

Sample origin

Time-course experiment

We purchased female mice from The Jackson Laboratory (Bar Harbor, ME, USA) and housed them in the OMRF vivarium under veterinary care. These mice were a subset of mice from a previously published study¹⁵. The mice were group housed (3 animals/cage) in ventilated cages in a temperature-controlled room maintained at 22 ± 3 °C on 14 h:10 h light/dark cycles with ad libitum access to chow and water. When the mice were 16 weeks of age, we began feeding them a low-fat diet (Research Diets, New Brunswick, NJ; D12450Ji; 10 kcal% Fat) for 4 weeks and labeled them with deuterium oxide for 1, 3, 5, 7, 15, or 21 days (n = 1/time point). We labeled mice with deuterium oxide (D₂O) with an initial intraperitoneal bolus injection of 99% D₂O, followed by ad libitum access to drinking water enriched with 8% D₂O. We staggered diet and deuterium labeling initiation such that the duration of diet treatment and age at the time of sacrifice were the same for all animals regardless of labeling period. We collected liver and blood after euthanasia by exsanguination via cardiac puncture under isoflurane anesthesia. To collect plasma, we centrifuged blood at 2000 × g for 10 min and collected the supernatant. We aliquoted the plasma and froze it at −80 °C until further analysis.

Rapamycin experiment

We used male Pax7-Cre mice generated via in-house breeding from Pax7-Cre mice (Strain #010530) obtained from The Jackson Laboratory (Bar Harbor, ME, USA) and housed them in the OMRF vivarium under veterinary care. These mice were a subset of mice from an ongoing experiment to isolate satellite cells in skeletal muscle, thus, the liver was left unaffected by Cre recombinase activity. Animals were group housed (maximum five per cage) with ad libitum access to food and water in a room on a 14:10 h light: dark cycle with constant temperature and humidity control, with ad libitum access to chow and water. When mice were 30 weeks of age, they received a rapamycin-supplemented chow at 42 ppm or remained on a control diet. The rapamycin was supplied by Rapamycin Holdings (San Antonio, TX, USA) and then incorporated into chow by Lab Supply (Fort Worth, TX, USA). At the onset of rapamycin treatment, we initiated labeling with D₂O as described above. In this case, we took a single time point at 8 days (n = 1/time point). Mice were euthanized by carbon dioxide inhalation followed by cervical dislocation and cardiac puncture for blood collection. Liver and plasma were collected as described above.

Body water enrichment

To determine body water enrichment, we placed 50 μL of plasma and 0–12% D₂O standards (Sigma, 151882) into the inner well of an O-ring screw cap and inverted on a heating block overnight at 80°C. We diluted collected condensate 1:300 in ddH₂O and analyzed it on a liquid water isotope analyzer (Los Gatos Research, Los Gatos, CA, USA) against a standard curve prepared with samples having different concentrations of D₂O (0–12%).

Sample homogenization

We powdered the frozen livers with a Bessman Tissue Pulverizer (Repligen, Rancho Dominguez, CA) on liquid nitrogen. We added 1 mL of RIPA buffer supplemented with HALT protease inhibitor (Thermo Fisher, Waltham, MA) to approximately 30 mg of powdered liver. We then sonicated the samples for 15 cycles of 30 s on, 30 s off, on the high-intensity setting (Bioruptor Plus; Diagenode, Denville, NJ). Following sonication, we incubated the samples at 90 °C for 10 min and then sonicated them again using the previously mentioned sonication protocol. We centrifuged the homogenized liver lysates at 500× g for 5 min to remove any debris and then collected the supernatant for our analyses.

Preparation for LC-MS/MS analysis

Using the Pierce 660 nm Protein Assay (Thermo Fisher, Waltham, MA), we determined the protein concentration of each homogenized liver lysate. We took protein aliquots from the homogenized liver lysates and precipitated them in acetone overnight at −20 °C. Aliquots were adjusted so that 10 µg of protein was available. After acetone precipitation, we centrifuged the protein samples at 10,000 × g for 10 min and decanted the acetone. We then dried the samples via vacuum centrifugation (Thermo Scientific SpeedVac) at 45 °C until a dry protein pellet formed.

We gently resuspended the dried protein pellets in 100 µL 8 M urea supplemented with 50 mM DTT and 5 mM Tris HCl and incubated the resuspended protein lysates at 37 °C for 1 h. Following this incubation, we added iodoacetamide to a final concentration of 15 mM and incubated the samples at room temperature in the dark for 30 min. We then diluted each sample with three volumes of 50 mM ammonium bicarbonate, added trypsin to reach a 1:20 trypsin to protein lysate ratio, and incubated overnight at 37 °C. Following digestion, we desalted the samples using Pierce C18 Spin Columns (Thermo Fisher) using the manufacturer’s protocol. After desalting, peptides were gently dried via vacuum centrifugation (Thermo Scientific SpeedVac) at 45 °C until a dry peptide pellet formed. For samples that did not undergo dimethyl labeling, we resuspended the dried peptide pellet in 200 µL of 1% acetic acid and proceeded directly to LC-MS/MS analysis. For samples that did undergo dimethyl labeling, we performed dimethylation using formaldehyde and sodium cyanoborohydride as described in Boersema et al.³⁴ Following dimethylation, we dried the samples again via vacuum centrifugation and resuspended each member of the duplex in 100 µL of 1% acetic acid. We then combined the duplex members to bring the total volume to 200 µL and proceeded to LC-MS/MS analysis.

LC-MS analysis

All samples were analyzed on a ThermoScientific Orbitrap Exploris 480 interfaced to a Vanquish Neo HPLC, an EasySpray Source, and an EasySpray C18 PepMap Neo column. 10 µL sample volumes were injected and eluted from the column at 125 nL/min with a linear gradient of 2% CH₃CN in water with 0.1% formic acid to 40% CH₃CN in 50 min. The instrument was operated in data-dependent acquisition (DDA) mode, acquiring a full scan mass spectra from m/z 350 to 1300 with a resolution of 60,000 and subsequent MS/MS with an m/z resolution of 15,000. Selected ions were isolated at 2 m/z and higher charge dissociation (HCD) at 25% was used. A top 20 strategy was used with dynamic exclusion that excluded precursor ions for 30 seconds after 2 tandem mass spectra had been acquired. All data were acquired in profile mode. Source volts were 2500 with the ion transfer tube at 320 °C. For a full scan, automated gate control (AGC) was 3e6 with max fill equal to 25 msec. For MS/MS, AGC was set to “standard” with max fill set to “auto”.

Protein/peptide identification

Raw mass spectral data were converted to mzML format using the MSConvert tool of Proteowizard⁴⁰ version 3.0.22048. Peptides were identified from tandem mass spectra using the database search engine Mascot⁴¹ (Matrix Science, Boston, MA) version 2.5. The SwissProt protein sequence database (12 February 2022 release) was used. The database search parameters were set to Mus musculus for taxonomy, 15 ppm (parts per million) mass tolerance for peptide precursors, 40 ppm mass tolerance for fragment ions, fixed modification of cysteine residues with the carbamidomethylation, and variable oxidation of methionine (+15.9949). Light (28.0313 Da) and heavy (36.0757 Da) channels of dimethyl labeling of the peptide N-terminus and lysine residues were set as variable modifications. # ¹³C was set to 2, which allows for 0/1/2 isotopes. This setting permits the consideration of precursor selections from higher mass isotopomers. Trypsin was specified as the protease with up to 3 missed cleavage sites. The false discovery rate (FDR) was controlled by a reversed sequence database approach. Peptide FDR was set to 1%. Only the peptides with the highest score (Mascot bold red peptides) of matching to tandem mass spectra were retained.

Data processing for protein turnover rate

In traditional approaches, protein turnover rates are computed using the time-course of monoisotopic RA, I₀(t), obtained from LC-MS data of deuterated water-labeled peptides. At every time point of labeling, the monoisotopic RA is computed as the normalized abundance of the monoisotope from the complete isotope profile of a peptide^6,42. The time series of the depletion of monoisotopic RA is modeled as an exponential decay function:

$$I_0\left(t\right)=I_0^{\mathrm{asymp}}+\left({I_0\left(0\right)-I}_0^{\mathrm{asymp}}\right)e^{-\mathrm{kt}}$$ 2

In Eq. (2), I₀(0) is the monoisotopic RA of the unlabeled peptide, I₀^asymp is the monoisotopic RA achieved at the plateau of labeling, t is the labeling duration, and k is the turnover rate (also referred to as degradation rate constant^43,44). I₀^asymp is calculated from the number of hydrogens accessible to deuterium from deuterated water (N_EH), the deuterium enrichment (in excess of natural abundance of deuterium) of body water (p_W), and the natural abundance of deuterium (p_H)³⁰:

$$I_0^{\mathrm{asymp}}=I_0\left(0\right){\left(1-\frac{p_{\mathrm{W}}}{1-p_{\mathrm{H}}}\right)}^{N_{\mathrm{EH}}}$$

The turnover rate, k, is obtained from the non-linear least squares regression of the time course of I₀(t) as modeled in Eq. (2). Equation (2) was obtained under the assumptions of steady-state for proteins and the first order and zeroth-order kinetics of protein degradation and synthesis, respectively³⁰.

Dimethyl labeled samples

As was discussed above, unlabeled and labeled samples were duplexed in a dimethyl experiment. For each peptide, we obtained the corresponding isotope profiles and determined the monoisotopic RAs. From Eq. (2), one obtains the time course of the ratio, I₀(t)/I₀(0):

$$\frac{I_0\left(t\right)}{I_0\left(0\right)}={\left(1-\frac{p_{\mathrm{W}}}{1-p_{\mathrm{H}}}\right)}^{N_{\mathrm{EH}}}+\left(1-{\left(1-\frac{p_{\mathrm{W}}}{1-p_{\mathrm{H}}}\right)}^{N_{\mathrm{EH}}}\right)e^{-\mathrm{kt}}$$ 3

The ratio of the monoisotopic RAs from labeled and unlabeled samples are measured in the same mass spectrum. To obtain the turnover rate, we fit the time course of the experimental values of I₀(t)/I₀(0) to the model in Eq. (3) using non-linear least-squares regression. For peptide, AIAEELAPER⁺², Supplementary Fig. 3 shows the experimental time course of the monoisotopic RA and theoretical fits using Eq. (2) and Eq. (3). The turnover rates for this peptide are 0.27 day⁻¹ and 0.244 day⁻¹, respectively, for models using Eqs. (2) and (3). The difference between the rates from these models is less than 10% (relative to the result from Eq. (2)). This example demonstrates components of and the applicability of the turnover rate estimation using mass spectral data from dimethyl coupled samples.

Ethical approval

We performed all animal procedures in accordance with protocols approved by the Institutional Animal Care and Use Committee at the Oklahoma Medical Research Foundation (OMRF) and the guidelines provided by the National Research Council’s Guide for the Care and Use of Laboratory Animals.

Author contributions

H.M.D. extensively tested models of turnover rate estimations, implemented them into the software, contributed to writing the manuscript, and prepared figures. R.G.S. conceptualized, designed the study, and wrote the manuscript. M.E.T. carried out data analysis, contributed to the writing of the paper, and confirmed the manuscript. K.A.K. contributed to data collection and confirmed the manuscript. M.T.K. contributed to data collection, contributed to the writing of the paper, and confirmed the manuscript. B.F.M. contributed to the data collection, contributed to the conception of the paper, and confirmed the manuscript.

Peer review

Peer review information

Communications Chemistry thanks Dong-Gi Mun and the other, anonymous, reviewers for their contribution to the peer review of this work.

Data availability

The data used for testing the approach reported in this work is available within the work and as Supplementary Data. The mass spectral data and data analysis results have been deposited in the MassIVE repository (http://massive.ucsd.edu). The results of the peak detection and quantification for each peptide in every protein (Protein_Name.csv) and their corresponding rate constants (Protein_Name.RateConst.csv) are also available in the repository. The dataset identifier is MSV000098797.

Code availability

The Software for Protein turnover determination from two-plex (dimethyl labeled) samples is available on GitHub https://github.com/rgsadygov/Dimethyl_protein_turnover/releases/tag/v1.0.2 and can be found by Zenodo generated 10.5281/zenodo.17316529.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42004-025-01762-1.