Comprehensive analysis of protein digestion using six trypsins reveals the origin of trypsin as a significant source of variability in proteomics

Abstract

Trypsin is an endoprotease commonly used for sample preparation in proteomics experiments. Importantly, protein digestion is dependent on multiple factors, including the trypsin origin and digestion conditions. In-depth characterization of trypsin activity could lead to improved reliability of peptide detection and quantitation in both targeted and discovery proteomics studies. To this end, we assembled a data analysis pipeline and suite of visualization tools for quality control and comprehensive characterization of pre-analytical variability in proteomics experiments. Using these tools, we evaluated six available proteomics-grade trypsins and their digestion of a single purified protein, human serum albumin (HSA). HSA was aliquoted and then digested for 2 or 18 hours for each trypsin, and the resulting digests were desalted and analyzed in triplicate by reversed phase liquid chromatography - tandem mass spectrometry. Peptides were identified and quantified using the NIST MSQC pipeline and a comprehensive HSA mass spectral library. We performed a statistical analysis of peptide abundances from different digests, and further visualized the data using the principal component analysis and quantitative protein “sequence maps”. While the performance of individual trypsins across repeat digests was reproducible, significant differences were observed depending on the origin of the trypsin (i.e., bovine vs. porcine). Bovine trypsins produced a higher number of peptides containing missed cleavages, whereas porcine trypsins produced more semi-tryptic peptides. In addition, many cleavage sites showed variable digestion kinetics patterns, evident from the comparison of peptide abundances in 2 hour vs. 18 hour digests. Overall, this work illustrates effects of an often neglected source of variability in proteomics experiments: the origin of the trypsin.

Introduction

Mass spectrometry (MS) based proteomics is a key technology used in biomedical research. Both targeted and discovery proteomics strategies are being increasingly used for characterization of peptides and proteins using samples derived from cells, tissues, or biological fluids such as plasma or urine ^{1, 2}. MS technology is also being increasingly applied in the analysis of biological pharmaceuticals, food allergen detection, and related applications ^3-5. All these applications rely on the ability to identify and quantify biological molecules such as peptides and proteins in the analyzed samples with a high degree of accuracy, sensitivity, and reproducibility. At the same time, a typical quantitative proteomics workflow is a complex process consisting of multiple pre-analytical (sample processing) and analytical steps⁶. It includes proteolytic digestion of proteins into peptides, separation of the resulting peptide mixtures using liquid chromatography (LC), identification and quantification of peptides by MS, and computational analysis of MS data. As a result, technical variability in proteomics experiments can be high. Understanding and measuring all major sources of variability is essential for the success of MS-based proteomics as a reliable measurement platform ^7-10.

In a typical analysis, proteins are assayed by the detection of their corresponding fragments - peptides - produced by proteolytic digestion of proteins using an endoprotease ¹¹. The abundance of a peptide is estimated using the intensity of the peptide ion extracted from MS data ¹² (or using fragment ion intensities in the case of targeted strategies such as selected reaction monitoring (SRM) ¹³ or SWATH-MS ¹⁴), with an additional (optional) normalization to the intensity of a spiked-in labeled reference peptide. Non-detection of peptide ions corresponding to high charge states of the selected peptide, unanticipated chemical modifications, incomplete digestion resulting in peptides containing missed cleavage sites, and run-to-run variation in the ionization processes such as in-source fragmentation of a parent peptide sequence, all contribute to errors in the measurements of the peptide abundance. Total measurement variance has been studied indicating that inter-laboratory CVs for quantified peptides can be as high as >25%, albeit these are both laboratory and target sequence dependent ^{7, 9, 10}. More recent work developed statistical models to elucidate primary contributions toward variance while also showing the advantage of using heavy isotope labeled reference material^{15, 16}. These studies have shown that the proteolytic digestion is a primary source of error in the abundance measurement ^{7, 17, 18}. Gaining a better understanding of proteolytic digestion and its contribution toward measurement variability in proteomics is the main motivation behind our work.

Trypsin is the most commonly used endoprotease for proteomic analysis, and several commercially available enzymes have been produced. Most commonly used are ‘proteomics grade’ trypsins where lysines are methylated to protect against autolytic degradation. Functionally, trypsin is a serine protease that is specific (cleaves C-terminal of K and R amino acid residues), active in semi-denaturing conditions, and maintains its activity across a selected pH range ^{19, 20}. The mechanism of trypsin digestion is well understood and the activity maintains specificity in part due to the highly conserved catalytic triad and the binding pocket arginine. Recent studies have attempted to further refine the rules for trypsin specificity ^19,21. Regardless, trypsin specificity remains highly conserved across species due to its conserved secondary structure, catalytic motif, and substrate binding pocket ²². This conservation is present in two trypsin sequences of different origin, porcine and bovine, which represent the majority of commercially available MS grade trypsin enzymes. The combination of biochemical stability and specificity, suitability for a broad range of sample preparation and analysis methods, and the relatively low cost of the enzyme explain its widespread use in proteomics, although the utility of other enzymes including LysN, LysC, GluC, chymotrypsin, and pepsin have also been explored ^{23, 24}. There has been an increased interest in developing alternative methods for trypsin digestion ^{20, 25, 26}, however in-solution and in-gel digestion using a commercially available trypsin remain the primary methods for sample processing in proteomics.

Importantly, protein digestion depends on multiple factors, including the origin of the trypsin, digestion conditions, denaturing conditions, and the presence of post-translational modifications on the protein that may interfere with trypsin digestion ^{27, 28}. Thus, evaluation of the reproducibility of proteolytic digestion and the degree of digestion completeness is of high importance. In recent reports, variability that is attributed to pre-analytical steps such as reduction, alkylation, and trypsin digestion of the target proteins have produced upwards of 30% error for inter sample variation²⁶. The referenced study, however, was not designed to determine the specific contribution of protein digestion efficiency toward the measurement accuracy and reproducibility. So far, most recent studies have focused on the development of improved digestion strategies and protocols ^28-37 for complex protein sample analysis. Additionally, recent work has profiled the performance of trypsins from a variety of commercially available sources and for multiple proteins in a mixture ^{38, 39}. However, there is a need for a more in-depth analysis of trypsin digestion performed under more strictly controlled conditions and using well-defined protein samples. As MS-based proteomics is poised to become more influential in discovery and diagnostic research for clinical proteomics in the near future, and with protocols for biofluid analysis approaching standardization ^{7, 40}, it is becoming imperative to develop metrological assays to ascertain performance of pre-analytical steps such as protein digestion.

In this work, we comprehensively assayed the performance of six commercially available trypsins of bovine and porcine origin using digests of a highly pure native human serum albumin (HSA). A comprehensive spectral library of tandem mass (MS/MS) spectra of HSA peptides was assembled from thousands of independent analyses of HSA digests, and used for robust and comprehensive peptide identification of the HSA digestion products. First, using two of the six trypsins, we investigated the reproducibility of replicate digests and the performance of the LC-MS/MS measurement platform to assess the variability due to sample handling, the instrumentation, and vial to vial differences. Second, and as the primary analysis performed in this work, the abundances (or, more precisely, MS signal intensities) of HSA peptides, as well as the overall performance metrics, were monitored across HSA digests obtained using all six trypsins and under two different digestion conditions (2 and 18 hours digestion time). This analysis revealed significant differences in the propensity of trypsins for missed and irregular cleavages dependent on their origin (bovine or porcine) and digestion time.

Experimental Methods

Trypsins

Six proteomics grade trypsins were used in the study (M1: G Biosciences #786-245, mass spectrometry grade; M2: Princeton Separations #EN-151, sequencing grade; M3: Promega #V5111, sequencing grade; M4: G Biosciences #786-245B, mass spectrometry grade; M5: Roche Applied Science #11418025001, sequencing grade; M6: Worthington Biochemical Corp. #LS02120). These were all TPCK treated to reduce the chymotryptic activity and methylated for resistance to autolysis. Each of the trypsins was of either porcine (M1-M3) (UniprotKB: P00761) or bovine (M4-M6) (UniprotKB: P00760) origin. As the main focus of this work was evaluation of the performance of different trypsins for gaining a better understanding of the sources of measurement variability, we do not disclose a preference for any manufacturer (see Disclaimer).

Human Serum Albumin

Human serum albumin purchased from Sigma (catalog number: A3782, purity: > 99%) was used for the initial assessment of reproducibility of replicate digests. The HSA purchased from Lee Biosolutions (catalog number: 101-12, purity: >98%) was used for the main experiment.

Sample digests

Digests for both the initial reproducibility analysis using two trypsins and the main experiment using all six trypsins were performed under the same conditions, except the digestion time (18 hours for the initial reproducibility analysis, 2 hours and 18 hours for the main experiment). In these experiments, 6 mg of human serum albumin (from Sigma in the case of the initial reproducibility analysis, and from Lee Biosolutions for the rest of the experiments) was dissolved in 600 μl of 6M Urea in 100mM Tris buffer. Then, 30 μl of a 200 mM DTT solution was added at room temperature for 1h to reduce the protein mixture, followed by addition of 120 μl of a 200 mM iodoacetamide solution with incubation at room temperature in the dark for 1 hour. Then 120 μl of a 200 mM DTT solution was added and samples were incubated for 1 hour to eliminate excess iodoacetamine. The resulting protein solution was separated into 145 μl aliquots. The urea concentration was reduced by diluting the reaction mixture with 755 μl of water in each vial. Then, 100μl aliquot containing 20 ug of trypsin from a particular source was added to the vial. Samples were mixed by gentle vortexing and the digestion was carried out at 37 °C. For each vial, 500 μl samples were extracted after 2 or 18 hours and then quenched with 10 μl of formic acid (50%) to give a pH < 3.

LC-MS/MS

For each run, a 1 μl aliquot of each 1μg/μl digest mixtures was injected into a Dionex Ultimate 3000 HPLC (Acclaim pepmap300 column, 150mm × 300um, C18, 5um, 300Å, Dionex, Sunnyvale, CA) and passed through a nanospray source into a Finnigan LTQ mass spectrometer (Thermo Fisher Scientific, Waltham, MA; mass resolution 0.4 m/z). Mobile phase A consisted of 0.1% formic acid in water and mobile phase B consisted of 0.1% formic acid in acetonitrile. The peptides were eluted by increasing mobile phase B from 1% to 90% over 50 minutes. Data were collected using a data dependent mode with a dynamic exclusion time of 20 seconds. The top 8 most abundant precursor ions were selected for ion-trap fragmentation over the m/z range 250-2000. Each sample was analyzed using LC-MS/MS in triplicate.

HSA spectral library

The reference spectral library (Dong et al., in preparation) was compiled from over three thousand digest runs for a wide range of digestion conditions. A set of peptide identification search engines were used to find all identifiable peptide products. This included multiple missed cleavages and up to six charge states as well as semi-tryptic products (from in-source fragmentation as well as those from irregular cleavages). A wide range of common modifications were included in the search, ranging from sodiated aspartic acid residues to cyclization of N-terminal cysteine. Additional, less common modifications were found using programs capable of finding untargeted modifications (i.e., so-called ‘blind’ mode). High-scoring identifications (passing a 5% local FDR filter) from all analyses were then combined and their MS/MS spectra were clustered to produce consensus spectra [40, 41]. These consensus spectra were then subjected to a battery of quality assessment filters. These, for example, eliminated MS/MS spectra corresponding to MS1 features that were rarely detected, or identifications with having multiple rare modifications, inconsistent retention time, or with large mass errors. Summary statistics for the HSA library are shown in Supplementary Figure 1.

Data analysis

Raw data files were converted to mzXML and MGF using the ReadW4Mascot2 converter ^{9, 41} (an extension of ReadW.exe). MSPepSearch (v.0.9) ⁹ was used to match spectra against the HSA library described above. Search tolerances were 1.8 m/z units for the precursor and 0.8 m/z units for fragment peaks. A minimum MSPepSearch score of 450 was required for inclusion of identified spectra in the subsequent analysis using nistms_metrics.exe. ReadW4Mascot2.exe, ProMS.exe, MSPepSearch.exe, MergePepResults.exe, and NISTMS_metrics.exe were all run and the data were combined using the NISTMSQC v1.2.0 data analysis pipeline⁹. These tools are available for download at http://peptide.nist.gov web site. Quality metrics and their deviations for each raw data file were also calculated. Intraseries and interseries deviations were reported as coefficients of variation (CV). CVs for the LC-MS/MS replicates were computed using the values across three LC-MS/MS replicates. The interseries (the comparison between all the trypsins) and intraseries (the comparison only between trypsins of the same origin, porcine or bovine) CVs were computed using the mean values from each sample. All data were uploaded into a custom MySQL (v.5.0.95) database and additional statistics were calculated using PHP (v.5.3.3) scripts. Any further analysis and plots were completed using R (2.14) or Microsoft Excel 2007.

Peptide intensities (area calculations from extracted ion chromatograms) were calculated using ProMS (v0.9). Peptide intensities were normalized to the total intensity of all ions identified in each LC-MS/MS run. The intensity for each unique peptide sequence in each LC-MS/MS run was calculated as the sum of the normalized intensities of all identified peptide ions detected for that sequence in that run (i.e. summing the normalized intensities of peptide ions containing different modifications and or identified in different charge states). Finally, the intensity of each unique peptide sequence in each trypsin digest (and each digestion time point) was computed as the average intensity across the three LC-MS/MS replicates for each sample. In doing so, missing values were dropped from the computation with the mean computed only from observed values. As such, the reported CVs were only included for ions with at least 2 measurements (out of 3 LC-MS/MS replicates) per sample. Comprehensive lists of the peptides identified in the main experiment, including their abundances, are included in Supplementary Tables 1 and 2 for the 2 and 18 hour digests, respectively. The lists of all the ions detected in the analysis for each of the six trypsins and used in computations of peptide intensities are included in Supplementary Tables 3 and 4.

Results and Discussion

Overview

An overview of the analysis is shown in Figure 1. The main application of the data analysis pipeline and visualization tools assembled in this work was to evaluate the performance of trypsins from different manufacturers (M1-M3: porcine, M4-M6: bovine, see Methods for detail). As such, digests of a single substrate, human serum albumin (HSA), were used to test individual enzyme performance. This ensured the greatest sensitivity to the differences between the trypsins. The initial analysis (Figure 1A) investigated the reproducibility of replicate digests performed on different days. For this part, two of the six trypsins (M2 and M5) were used to digest HSA for 18 hours four times on different days, and analyzed in triplicate using LC-MS/MS. For the primary experiment (Figure 1B), single digests for each of the six trypsins were used, with 2 and 18 hours digestion time, and analyzed using LC-MS/MS in triplicate. The resulting data was used for the following three primary analyses: 1) analysis across all trypsins regardless of origin (interseries M1 through M6), 2) separate analysis for the two trypsin sequences (intraseries, porcine: M1-M3 and bovine: M4-M6) and 3) comparison between the two digestion times (2 versus 18 hours digestion).

Figure 1. Overview of the experiments.

(A) Digest replicates were performed on 3 separate days, 4 times, and analyzed by LC-MS/MS in triplicate. (B) Trypsin digests were performed using six different trypsins of bovine or porcine origin, for 2 or 18 hours digestion time, and then analyzed by LC-MS/MS in triplicate. Interseries analysis describes analysis of trypsin digests across all six different trypsins; intraseries analysis is done separately across three porcine (M1-M3) or three bovine (M4-M6) trypsin digests.

Digestion reproducibility and LC-MS/MS platform stability

To assess the reproducibility of the digests and the stability of the entire measurement platform (Figure 1A), repeated digest using the M2 and M5 trypsins were analyzed by LC-MS/MS in triplicate. The acquired data were processed using the NIST MSQC tools (see Methods). The results were summarized using various metrics as described in Rudnick et al ⁹. The metrics of most relevance to the aims of this study are listed in Tables 1 and 2. These included the total number of MS/MS scans acquired (DS2-B), the number of identified unique ions (P-2B) and peptides (P-2C), the ratios of the number of ions observed in different charge states compared to that in charge state z =2⁺ (IS-3A through IS-3C), and the mean length of peptides (PL-1 to PL-4) detected for each charge state (Table 1).

Table 1. Metrics for analysis of reproducibility: repeated digest experiment.

The metrics from Rudnick et al. 9 used in this work. Mean and % CV of each metric are reported for either the M2 or M5 trypsin, 4 digests for each trypsin. Mean and % CVs were calculated for each replicate digest (labeled Digest 1 through Digest 4). Additionally, the mean and % CVs were calculated for each trypsin using the mean values for each replicate digest (far right two columns). DS: dynamic sampling, P: peptide, IS: ion source, PL: peptide length.

Metric	Description	M2 Digest 1		M2 Digest 2		M2 Digest 3		M2 Digest 4		M5 Digest 1		M5 Digest 2		MS Digest 3		M5 Digest 4		M2		M5
		mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV
DS2-B	MS/MS scans	10259.7	2.1	10786.7	0.2	10796.0	05	10497.3	0.3	10373.3	0.7	10774.0	0.6	10823.0	0.6	10467.7	0.3	10584.9	24	10609.5	2.1
P-2C	Unique peptides	109.7	2.1	121.0	0.8	119.3	1.7	116.7	1.3	127.7	1.8	129.7	1.9	132.3	2.7	121.3	1.9	116.7	4.3	127.8	3.7
P-2B	Unique ions	343.7	3.7	350.7	1.4	352.7	0.9	342.7	2.1	394.3	3.9	391.3	0.5	383.3	0.8	373.0	1.4	347.5	1.4	385.5	2.5
IS-3A	Ratio lz/2z	0.3	5.4	0.3	12.5	0.3	8.4	0.3	2.1	0.3	8.5	0.3	5.0	0.3	3.3	0.3	7.3	0.3	12.8	0.3	9.5
IS-3B	Ratio 3z/2z	0.7	6.5	0.8	5.8	0.9	5.2	0.8	7.9	0.9	8.9	1.0	6.9	1.0	8.8	0.8	14.1	0.8	8.6	0.9	10.5
IS-3C	Ratio 4z/2z	0.4	3.1	0.4	2.7	0.5	5.7	0.4	0.5	0.5	3.8	0.6	5.7	0.6	4.4	0.5	11.2	0.4	9.9	0.5	12.0
PL-1	AVG length z=l	7.8	1.0	7.7	3.1	7.5	2.3	7.5	0.9	7.5	3.1	7.2	0.8	7.1	0.5	7.4	2.3	7.6	1.5	7.3	2.3
PL-2	AVG length z=2	10.1	0.6	9.9	0.7	9.9	1.2	9.9	1.3	10.3	1.0	10.1	0.8	10.0	1.4	10.2	0.2	9.9	1.1	10.1	1.6
PL-3	AVG length z=3	15.6	1.8	14.9	1.1	14.6	0.4	15.0	1.2	15.4	1.9	14.7	0.2	14.6	0.5	15.1	1.7	15.0	2.7	15.0	2.6
PL-4	AVG length z=4	22.0	0.7	21.3	1.0	21.1	2.4	22.1	1.3	22.0	1.4	21.7	1.8	21.6	0.7	22.0	2.7	21.6	2.3	21.8	0.9

Table 2. Metrics for analysis of reproducibility: six trypsins study.

Mean and % CV of each metric are reported for each trypsin (M1-M6). Interseries mean ± % CV are calculated for all trypsins combined. Porcine or Bovine ± % CV are the reported values for the intraseries analysis (between porcine (M1-M3) or bovine (M4-M6) trypsins). A) 2 hour digests. B) 18 hour digests. Bov/Por % change is the percent difference between the mean reported values for the bovine and porcine digests. DS: dynamic sampling, P: peptide, IS: ion source, PL: peptide length.

A. 2h
Metric	Description	Ml		M2		M3		M4		MS		M6		Intcrscries		Porcine		Bovine		Bav/Por
		mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	%diff
DS2-B	MS/MS scans	10081.7	0.7	10085.3	0.5	10073.3	0.6	10187.0	0.8	10099.0	0.5	10121.3	0.5	10107.9	0.4	10080.1	0.1	10135.8	0.5	0.6
P-2C	Unique peptides	108.7	3.0	107.7	4.2	96.0	1.8	139.3	2.5	130.3	1.9	146.3	2.1	121.4	17.2	104.1	6.8	138.6	5.8	33.1
P-2B	Unique ions	312.0	2.0	296.0	3.3	266.3	1.5	333.3	2.9	320.0	1.7	349.7	0.9	312.9	10.4	291.4	8.0	334.3	4.4	14.7
IS-3A	Ratio lz/2z	0.5	5.2	0.5	3.2	0.4	3.4	0.3	2.4	0.4	6.9	0.4	6.2	0.4	11.0	0.4	9.2	0.4	8.3	−14.2
IS-3B	Ratio 3z/2z	1.0	0.5	1.0	10.7	0.8	4.1	1.2	2.1	1.1	6.5	I.I	6.0	1.0	15.1	0.9	13.9	1.1	4.1	21.6
IS-3C	Ratio 4z/2z	0.5	8.3	0.6	3.4	0.4	9.1	0.7	4.0	0.7	6.6	0.7	10.8	0.6	22.1	0.5	15.7	0.7	1.9	42.9
PL-1	AVG length z=l	7.9	0.9	7.7	1.9	7.8	1.7	7.5	1.3	7.6	1.6	7.5	0.8	7.7	1.6	7.8	0.8	7.5	0.6	−3.3
PL-2	AVG length z=2	10.8	1.2	11.0	2.0	10.7	1.5	10.8	1.2	10.9	0.5	10.9	1.4	10.9	1.0	10.8	1.3	10.9	0.5	0.5
PL-3	AVG length z=3	16.1	0.7	16.1	2.0	15.8	3.0	16.1	1.2	16.1	3.4	16.6	1.2	16.2	1.8	16.0	1.0	16.3	1.8	1.7
PL-4	AVG length z=4	22.0	0.6	21.7	0.5	22.1	0.3	22.3	1.7	22.3	2.1	22.1	1.5	22.1	1.1	21.9	0.9	22.2	0.4	1.5

B. 18 h
Metric	Description	Ml		M2		M3		M4		M5		M6		Interseries		Porcine		Bovine		Bov/Por
		mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	mean	CV	%diff
DS2-B	MS/MS scans	10026.0	0.9	10115.0	1.0	10077.7	1.2	10085.7	1.2	10143.0	0.7	10113.3	1.4	10093.5	0.3	10072.9	0.4	10114.0	0.3	0.4
P-2C	Unique peptides	98.3	2.6	90.7	4.2	87.7	1.7	142.7	1.6	132.0	0.8	148.7	0.4	116.7	24.2	92.2	5.9	141.1	6.0	53.0
P-2B	Unique ions	370.3	1.6	357.3	1.3	320.0	1.9	432.7	1.8	406.7	3.1	427.7	3.2	385.8	12.5	349.2	7.5	422.4	3.3	21.0
IS-3A	Ratio lz/2z	0.4	4.9	0.4	0.5	0.4	3.9	0.3	2.9	0.4	1.5	0.3	7.1	0.4	6.5	0.4	5.7	0.3	4.5	−8.3
IS-3B	Ratio 3z/2z	0.8	0.7	0.8	4.9	0.7	3.9	1.2	2.8	1.0	3.7	1.1	3.3	0.9	22.7	0.8	12.5	1.1	6.7	43.7
IS-3C	Ratio 4z/2z	0.3	4.9	0.4	3.4	0.3	4.3	0.7	6.4	0.5	3.2	0.7	6.5	0.5	35.9	0.3	13.0	0.6	11.7	91.7
PL-1	AVG length z=l	7.7	0.8	7.8	1.2	8.0	0.5	7.5	1.4	7.6	1.5	7.6	0.5	7.7	2.6	7.8	2.0	7.6	0.7	−3.4
PL-2	AVG length z=2	10.8	0.8	10.8	1.5	10.6	1.2	11.1	0.3	11.1	0.6	11.2	1.8	10.9	2.2	10.7	0.9	11.1	0.4	3.6
PL-3	AVG length z=3	15.9	1.0	15.7	0.9	15.8	0.9	16.4	0.8	16.5	1.1	16.7	1.0	16.2	2.7	15.8	0.6	16.5	0.9	4.5
PL-4	AVG length z=4	21.1	1.2	20.4	0.9	20.4	0.3	23.0	1.3	23.2	0.9	23.3	0.8	21.9	6.8	20.6	1.8	23.2	0.6	12.2

For the LC-MS/MS replicates, the CVs for the DS2-B metric were consistently less than 2.1% and the CVs for the total detected unique ions and peptides never exceeded 3.9% CV. Higher CVs (up to 14.1%) were reported for the IS-3 metrics. This reflects the more variable nature of the ionization process with regard to higher charge ions (z = 3⁺ or higher). However, it is z = 2⁺ ions that account for the majority (~70%) of the total detected ions. The PL-1 to PL-4 CVs were below 2.7%. The CVs for the repeated digests completed on separate days, for each of the two trypsins, were similar or increased slightly compared to the CVs for the LC-MS/MS replicates of the same digest. For the digest replicates, the DS2-B CVs were 2.4% and 2.1% for the M2 and M5 trypsins, respectively. Excluding the IS-3 metrics, the CVs increased from an average of 1.3% for the LC-MS/MS replicates to an average of 2.2% for digest replicates. Taken together, the data showed the high reproducibility of the LC-MS/MS measurement platform, and that performing replicate digests under the same conditions, was not a major source of variability in this study. Thus, the subsequent analysis across all six trypsins, described below, was performed using a single digest for each trypsin.

Six trypsins study

The main experiment of this study, performed using six different trypsins (Figure 1B), is discussed in the remainder of this manuscript. Reported measures of variability (Table 2) were low for most metrics when considering the LC-MS/MS replicates (labeled M1-M6 CV columns in Table 2), as expected based on the results presented above. The greatest variances were observed in the analysis across all the trypsins (labeled as Interseries in Table 2). The total number of MS/MS scans remained stable throughout the experiment for all the comparisons (2 hours vs. 18 hours, bovine vs. porcine, interseries). The numbers of unique peptide sequences (P-2C) and unique peptide ions (P-2B) that were produced at 2 hours were 121.4 ± 17.2% and 312.9 ± 10.4%, respectively, and decreased slightly at 18 hours. As previously discussed, the highest CVs were observed for the IS-3A-C metrics at both time points.

Based on Principal Component Analysis (PCA), discussed later in this manuscript, the origin of the trypsin (bovine vs. porcine) was found to be a significant factor contributing to the overall variability of peptide abundances. At 2 hours, the classification of the trypsins into two subclasses resulted in significantly decreased CVs within each class (Intraseries analysis in Figure 1B; see Table 2, columns ‘Porcine’ and ‘Bovine’) as compared to the interseries analysis. For example, the CVs were less than 7% for the reported number of unique peptides when analyzed separately for each subclass (compared to 17.2% for the interseries analysis). Interestingly, the mean values of the total unique peptides and total unique ions were 33.1% and 14.7% higher for the bovine trypsin group than for the porcine group (Table 2, column ‘Bov/Por’). The bovine trypsins also produced more peptides that ionized into high charge state ions (3⁺ and 4⁺) and with lower CVs (IS-3B and IS-3C metrics). At 18 hours, significant differences were observed for the number of unique ions. These increased, compared to 2 hour digestion time, by ~20% and 26% for the porcine and bovine trypsins, respectively, and the % difference (bovine vs. porcine) increased by 6%. The average peptide lengths were slightly different between bovine and porcine trypsin digests: bovine trypsins produced longer peptides when considering the identifications resulting from high charge state peptide ions (e.g., longer by 2.5 amino acids, PL-4 metric).

The increase in the average peptide length and in the number of peptide identifications from high charge state ions in bovine vs. porcine digests suggested that bovine trypsins were more likely to produce peptides containing missed cleavages. Protease specificity is in part determined by its substrate binding pocket, the surface fit between the substrate and enzyme, and the primary sequence of the substrate ^{19, 42}. Thus, the two enzyme sequences (porcine vs. bovine) may exhibit differences in their substrate affinities that should be observable in these data. To test this further, different trypsins were compared with respect to the types of cleavage that led to the formation of the peptides (Figure 2). The total numbers of identified unique peptide sequences were counted considering whether they were fully tryptic peptides with no missed cleavages (FT), fully tryptic peptides containing one or more missed cleavage (MC), or semi-tryptic peptides (ST).

Figure 2. Unique peptide identifications.

Unique peptide counts (mean ± standard deviation across three replicates) in three different peptide categories (FT: fully tryptic, MC: missed cleavage, ST: semi-tryptic) are plotted for each trypsin (M1-M6) for both the 2 hour (A) and 18 hour (B) digestion time points. C) Same as above, data grouped by porcine or bovine trypsin category (mean ± standard deviation across all trypsins and replicates for each trypsin category), 2 and 18 hour digests. M1-M3: porcine trypsins, M4-M6: bovine trypsins.

At 2 hours, the interseries analysis (across all trypsins, M1-M6, Figure 2A) demonstrated that total counts of FT peptides were consistent between all trypsins (43.8 ± 0.2% unique peptide identifications). Peptides classified as MC and ST, on the other hand, were significantly more varied: 84.2 ± 21.0% and 59.9 ± 12.0%, respectively. Most notable were a higher number of MC peptides in the M4-M6 digests and more ST peptides in the M2 digest. At 18 hours (Figure 2B), the number of MC peptides decreased slightly (14 fewer peptides) for all the trypsins, but the differences between the M1-M3 and M4-M6 trypsins became more visible. The number of ST peptides increased in 18 hour digests compared to 2 hour digests for all the trypsins, and the M2 trypsin produced a more marked increase in the number of these peptides (about 50% more ST peptides over the other trypsins).

Grouping the trypsins by their origin (porcine or bovine) made these trends even more apparent (Figure 2C). The bovine group collectively produced a significantly higher number of unique MC peptides (66.2 ± 9.9% in porcine vs. 102.1 ± 9.4% in bovine; 54% more in bovine) whereas the porcine trypsins produced slightly more ST peptides (62.1± 27.8% in porcine vs. 57.0 ± 10.5% in bovine; 9% more in porcine). The number of FT peptides stayed the same across trypsins and for the 18 vs. 2 hour digests. The total numbers of MC peptides were higher for the bovine vs. porcine trypsins regardless of the digestion time, but decreased with increased digestion time for both classes of trypsin (down 33% in porcine and 5.6% in bovine). ST peptides, on the other hand, increased 62% in porcine and 29% in the bovine trypsin digests from 2 to 18 hours.

Analysis of peptide abundances

The above analysis of the counts of the identified peptides was extended by considering the quantitative measure of peptide intensity. Here, the intensities of peptides were estimated based on the intensities of peptide ions extracted from the MS1 data using the ProMS tool of the NIST MSQC pipeline which were then summed for each unique peptide sequence. The intensities were normalized to the total intensity of all peptides in the LC-MS/MS run (see Methods for detail). As such, the peptide intensity numbers discussed below are given as the fraction of the total HSA intensity detected in each sample.

At 2 hours, the peptide intensity trends (Figure 3A) were similar to those observed for the unique peptide counts (Figure 2A), except that the differences between FT peptide intensities in bovine and porcine trypsins became discernable (M1-M3: 0.34 ± 1.4%; M4-M6: 0.25 ± 8.0%). Relative to the 2 hour digests, FT peptide intensities in 18 hour digests (Figure 3B) increased by 15.3% to 0.34 ± 20.6%, MC intensities decreased 17.1% to 0.44 ± 30.0% and ST peptide intensities increased by 28.5% to 0.21 ± 33.3%. The MC and ST peptides were more variably produced than FT peptides across all the trypsins. Similar to the count data shown in Figure 2, there were significantly different intensities of MC and ST peptides between the M1-M3 vs. M4-M6 trypsins. These differences become more apparent in Figure 3C. In porcine trypsin digests, the total FT and ST peptide intensity increased with longer digestion time by 19.4% and 41.3%, respectively. This was significantly more than the corresponding 9.7% and 12.1% increases in bovine trypsin digests. At the same time, the intensity of MC peptides decreased greatly in porcine (29.9%) but only slightly (6.7%) in bovine trypsin digests. The differences between the bovine and porcine trypsins generally were more pronounced at 18 hours.

Figure 3. Peptide intensities.

Peptide intensities computed using normalized peptide ion intensities (mean ± standard deviation across three LC-MS/MS analysis) in three different peptide categories (FT: fully tryptic, MC: missed cleavage, ST: semi-tryptic) are plotted for each trypsin (M1-M6) for both the 2 hour (A) and 18 hour (B) digestion time points. Intensities are shown as the fraction of the total intensity of all HSA peptides (for each digest) contributed by peptides from a particular category. C) Same as above, data grouped by porcine or bovine trypsin category (mean ± standard deviation across all trypsins and replicates for each trypsin category), 2 and 18 hour digests. M1-M3: porcine trypsins, M4-M6: bovine trypsins.

Considering all of the data together, FT peptides as a category were found to be the most reproducible type of peptides. The increase in the intensity of FT peptides in 18 vs. 2 hour digests, accompanied by a corresponding decrease in the number and intensities of MC peptides, indicated more complete HSA digestion with longer digestion time. The increase in both the intensity and in the number of unique ST peptides with increased digestion time may indicate an increased probability of trypsin producing an irregular cleavage (non K/R). However, this could also be attributed to other factors such as peptide degradation after trypsin digestion or increased activity of contaminating enzymes such as chymotrypsin. The ST peptides significantly contributed toward the total summed HSA intensity (~21%) and the number of identified unique ST peptides was higher than that of MC or FT peptides. It should also be noted that the high number of identified ST peptides in these data was due to low sample complexity and is not representative of the numbers of ST peptides observed in a typical analysis of complex protein samples ^38,43. A detailed list of the detected peptides and their intensities for both the 2 and 18 hour digests are included in Supplementary Tables 1 and 2.

Analysis of trypsin digestion using PCA

We sought to use a statistical and visual approach to highlight the global peptide intensity trends. To this end, we first performed a Principal Components Analysis (PCA) (Figure 4). The input in the exploratory PCA analysis was a matrix consisting of 267 peptide intensity measurements (peptides identified and quantified in both time points). The PCA analysis identified three statistically significant principal components (PC). The first and most significant component (PC1; explained 89% of the total peptide intensity variance) summarized the differences between the intensities of different peptides in these data (see PC1 vector coefficients listed in Supplementary Figure 2). This component could be explained by the biases in the entire measurement system, including the effect of the physio-chemical properties of a peptide on its ionization efficiency. The second component, PC2, explained 7% of the variance after considering the variance explained by PC1. This component accounted for the differences between the porcine and bovine trypsins (see PC2 vector coefficients listed in Supplementary Figure 2).

Figure 4. Principle Components Analysis.

A) PC1 vs. PC2 plot. PC2 component describes the differences between peptide intensities in digests using porcine (M1-M3) vs. bovine (M4-M6) trypsins. Selected peptides whose intensities contribute most significantly toward differentiating between the bovine and porcine trypsins are indicated. B) Same as above, except PC3 vs. PC2 are plotted.

Figure 4A plots the data using the first two principal components and visually demonstrates the separation between bovine and porcine trypsins for both 2 hour and 18 hour digests. While some of the MC peptides were produced with greater variance and were low in intensity (centered around zero on PC1 and PC2 axes), several MC peptides were very abundant (high PC1 values) and were reliably detected by most or all trypsins (low PC2 values). This was exemplified by the MC peptide R.LVRPEVDVMCTAFHDNEETFLKK.Y, the most abundant peptide in the data. Both FT and MC peptides contributed to the separation between the M1-M3 (porcine) from the M4-M6 (bovine) trypsin digests. FT peptides were more abundant in porcine trypsin digests (positive PC2 values), whereas MC peptides were more prevalent in bovine trypsin digests (negative PC2 values). ST peptides contributed much less toward the total intensity despite their high count, and were not informative for distinguishing the bovine from the porcine trypsin digests. The peptides that contributed the most toward the separation between bovine and porcine trypsins are labeled in Figure 4A. The most striking example is the MC peptide K.RMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK.C (3 missed cleavages) and its shorter MC sibling K.RMPCAEDYLSVVLNQLCVLHEKTPVSDR.V (2 missed cleavages), which were favored by the bovine trypsins, whereas the FT subsequence of this peptide, R.MPCAEDYLSVVLNQLCVLHEK.T was favored in the porcine digests.

The third principal component (PC3; which explained approximately 2% of the remaining variance) accounted for the differences between the 2 and 18 hour digests (Figure 4B). One interesting example is the MC peptide K.RMPCAEDYLSVVLNQLCVLHEK.T which was highly abundant in the 2 hour porcine digests. This peptide became digested into the FT peptide R.MPCAEDYLSVVLNQLCVLHEK.T at 18 hours. Overall, the digestion time effect was notable for the porcine trypsins but largely negligible for bovine trypsins.

To further analyze the peptides, fold changes (FC_bov/por) were calculated to identify which peptides were most different in abundance between the bovine and porcine trypsin digests. For each time point (2 hour and 18 hour digestion time), Table 3 lists the top 30 peptides: the 15 peptides with the highest FC values (more abundant in the bovine digests) and the 15 peptides with the lowest FC values (more abundant in the porcine digests). Included in this list are the peptide K.RMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK.C and its subsequences already noted using the PCA analysis described above. At 2 hours, Table 3 lists 12 FT and 18 MC peptides, of which all FT peptides were more abundant in the porcine digests and 15 out of 18 MC peptides were more abundant in the bovine digests. A similar trend was observed at 18 hours (10 FT peptides - all more abundant in the porcine digests, and 17 MC peptides - 2 of which were more abundant in the porcine digests), except the list also included 3 ST peptides that were more abundant in the porcine digests. The full list of identified peptides and their fold changes are shown in the Supplementary Tables 3 (2 hour digests) and 4 (18 hour digests).

Table 3. Peptides with most altered intensities between bovine and porcine trypsin digests.

Peptide intensities are reported for each trypsin (M1-M3: porcine, M4-M6: bovine), along with mean intensity for the porcine (μ_por) and bovine (μ_bov) group. Position is the start position for each peptide in the sequence of HSA. Flanking residues are the immediate N or C terminal residues of the peptide in the protein sequence. FCbov/por indicates fold change for the mean intensity in the bovine vs. porcine digests. To account for missing values, fold changes were computed after addition of a 0.25 background factor to all M1-M6 intensities. Top 30 peptides are listed for each time point (the 15 peptides with the highest FC_bov/por and no missing values across bovine trypsin digests, and the 15 peptides with the lowest FC_bov/por and no missing values across porcine trypsin digests). A) 2 hour digests. B) 18 hour digests. FT: fully tryptic, MC: missed cleavage, ST semi-tryptic.

2h
Peptide	Flanking Residues	Cat	Pos	Ml	M2	M3	M4	M5	M6	μ por	μ bov	FCbov/por
RMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK	K.C	MC	468	0.10	0.10		31.68	28.33	24.34	0.10	28.12	89.7
LDELRDEGKASSAK	K.Q	MC	205			0.18	11.29	6.99	10.77	0.18	9.68	31.9
RMPCAEDYLSVVLNQLCVLHEKTPVSDR	K.V	MC	468	0.92	0.28	0.68	18.93	23.65	26.49	0.62	23.03	26.6
AEFAEVSKLVTDLTK	K.V	MC	249				5.83	2.68	9.63		6.05	25.2
VFDEFKPLVEEPQNLIKQNCELFEQLGEYKFQNALLVR	K.Y	MC	396	0.23		0.27	11.12	4.10	14.32	0.25	9.85	24.2
AFKAWAVAR	R.L	MC	233				4.48	0.10	5.07		3.22	13.9
MPCAEDYLSVVLNQLCVLHEKTPVSDR	R.V	MC	469	0.50	0.22	0.30	7.03	5.72	8.20	0.34	6.98	12.3
LSQRFPKAEFAEVSK	R.L	MC	242				2.59	0.34	4.86		2.60	11.4
LVRPEVDVMCTAFHDNEETFLKKYLYEIAR	R.R	MC	138	0.14		0.11	3.49	1.46	5.63	0.13	3.53	11.3
MPCAEDYLSVVLNQLCVLHEKTPVSDRVTK	R.C	MC	469	0.06			3.07	1.94	3.35	0.06	2.78	11.2
VFDEFKPLVEEPQNLIKQNCELFEQLGEYK	K.F	MC	396	0.20	0.07	0.21	4.71	3.43	3.81	0.16	3.98	10.3
CCKADDKETCFAEEGKK	K.L	MC	581	1.50	0.05	1.17	9.35	9.81	13.47	0.91	10.87	9.6
AACLLPKLDELRDEGKASSAK	K.Q	MC	198				1.50	0.74	2.82		1.69	7.8
LVTDLTKVHTECCHGDLLECADDRADLAK	K.Y	MC	257				1.02	1.04	2.64		1.57	7.3
YLYEIARR	K.H	MC	161	0.66		0.55	4.99	3.42	4.59	0.61	4.33	7.0
DDNPNLPR	K.L	FT	130	1.74	4.54	1.84	0.64	1.80	0.79	2.71	1.08	0.45
TCVADESAENCDK	K.S	FT	75	3.51	3.48	2.17	1.14	1.41	0.88	3.06	1.14	0.42
FQNALLVR	K.Y	FT	426	19.23	24.30	19.07	8.08	12.26	4.71	20.87	8.35	0.41
LDELRDEGK	K.A	MC	205	12.42	13.47	11.52	4.19	7.61	2.96	12.47	4.92	0.41
AACLLPK	K.L	FT	198	12.24	15.97	11.43	5.10	6.18	4.07	13.21	5.12	0.40
QNCELFEQLGEYK	K.F	FT	413	7.19	8.84	7.26	2.36	4.19	1.35	7.76	2.64	0.36
NECFLQHKDDNPNLPR	R.L	MC	122	1.90	3.13	1.92	0.69	0.62	0.40	2.32	0.57	0.32
RPCFSALEVDETYVPK	RE	FT	508	21.83	23.15	21.98	4.37	12.34	3.89	22.32	6.87	0.32
AAFTECCQAADK	K.A	FT	186	6.44	9.46	5.87	1.85	2.54	1.72	7.26	2.04	0.30
RMPCAEDYLSVVLNQLCVLHEK	K.T	MC	468	39.40	44.60	41.90	12.92	14.25	10.06	41.97	12.41	0.30
LCTVATLR	K.E	FT	97	7.02	11.80	6.64	1.57	3.63	0.89	8.49	2.03	0.26
MPCAEDYLSVVLNQLCVLHEK	R.T	FT	469	19.29	19.70	22.07	5.32	5.46	4.30	20.35	5.03	0.26
EFNAETFTFHADICTLSEK	K.E	FT	524	4.75	4.99	6.10	0.50	0.86	0.33	5.28	0.56	0.15
SLHTLFGDK	K.L	FT	88	6.81	12.45	6.77	0.51	1.70	0.34	8.68	0.85	0.12
TPVSDR	K.V	FT	490	4.02	4.82	3.96	0.71			4.27	0.71	0.11
18h
Peptide	Flanking Residues	Cat	Pos	Ml	M2	M3	M4	M5	M6	μ por	μ bov	FCbov/por
RMPCAEDYLSVVLNQLCVLHEKTPVSDR	K.V	MC	468	0.19			18.90	31.06	34.58	0.19	28.18	91.1
TCVADESAENCDKSLHTLFGDKLCTVATLR	K.E	MC	75				14.40	11.37	16.85		14.21	57.8
RMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK	K.C	MC	468				17.78	13.86	10.08		13.91	56.6
QNCELFEQLGEYKFQNALLVR	K.Y	MC	413	0.21	0.18	0.09	19.96	15.45	23.17	0.16	19.52	48.4
DVCKNYAEAKDVFLGMFLYEYARR	K.H	MC	337	0.11	0.07	0.12	13.11	19.29	17.65	0.10	16.68	48.2
AAFTECCQAADKAACLLPK	K.L	MC	186	0.28	0.18	0.15	18.94	21.11	17.51	0.20	19.19	42.8
LDELRDEOKASSAK	K.Q	MC	205				10.03	6.07	10.46		8.85	36.4
VFDEFKPLVEEPQNLTKQNCELFEQLGEYKFQNALLVR	K.Y	MC	396				7.24	3.71	14.49		8.48	34.9
YKAAFTECCQAADKAACLLPK	R.L	MC	184				6.80	2.74	6.58		5.37	22.5
CCKADDKETCFAEEGKK	K.L	MC	581				7.23	8.11	0.40		8.96	22.0
AEFAEVSKLVTDLTK	K.V	MC	249				3.85	1.90	8.37		4.70	19.8
MPCAEDYLSVVLNQLCVLHEKTPVSDR	R.V	MC	469	0.44		0.12	6.10	7.23	11.03	0.28	8.12	19.1
CCTESLVNRRPCFSALEVDETYVPK	K.E	MC	499				6.49	0.85	6.24		4.53	19.1
YLYETARR	K.H	MC	161				4.22	2.92	4.66		3.94	16.8
TPVSDRVTK	K.C	MC	490	0.08			4.62	4.08	3.55	0.08	4.08	15.7
RPCFSALEVDETYVPK	R.E	FT	508	28.07	29.41	27.78	6.16	13.53	4.71	28.42	8.13	0.29
QNCELFEQLGEYK	K.F	FT	413	10.75	12.36	12.18	2.67	4.35	1.44	11.77	2.82	0.26
ALVLTAF	K.A	ST	44	0.65	1.17	0.57				0.80		0.24
QEPERNECFLQHK	K.D	MC	117	2.51	4.31	2.82	0.25	0.66	0.35	3.21	0.42	0.19
DDNPNLPR	K.L	FT	130	4.64	8.97	6.23	0.70	1.45	0.78	6.61	0.97	0.18
MPCAEDYLSVVLNQLCVLHEK	R.T	FT	469	44.47	41.92	52.23	9.89	7.98	6.25	46.21	8.04	0.18
NECFLQHKDDNPNLPR	R.L	MC	122	4.84	3.45	4.78	0.58	0.63	0.42	4.36	0.54	0.17
VHTECCHGDLLECADDR	K.A	FT	264	4.26	7.76	6.40	0.83	1.03	0.49	6.14	0.78	0.16
NECFLQHK	R.D	FT	122	0.62	2.54	LOO				1.38		0.15
FSALEVDETYVPK	C.E	ST	511	2.51	4.72	3.90	0.41	0.37	0.28	3.71	0.35	0.15
LCTVATLR	K.E	FT	97	15.36	20.28	17.91	1.68	3.28	0.87	17.85	1.94	0.12
TFHADTCTLSEK	F.E	ST	531	0.07	5.32	0.12				1.84		0.12
EFNAETFTFHADICTLSEK	K.E	FT	524	16.57	6.43	20.79	1.13	1.57	0.71	14.59	1.14	0.09
SLHTLFGDK	K.L	FT	88	20.39	25.28	22.26	1.31	2.11	0.25	22.65	1.23	0.06
TPVSDR	K.V	FT	490	4.63	4.54	4.01				4.39		0.05

Visualization of trypsin digestion using quantitative “sequence maps”

The significant differences between the activities of porcine and bovine trypsins, despite their high sequence identity (82.5%), might be due to their slight structural differences (see structures Porcine: PDB: 2A31, Bovine: PDB: 3MI4 online at www.rscb.org). In addition, the denaturing conditions for the digests may produce different changes in the activity or selectivity of the two trypsins for HSA cleavage sites. Here, selectivity of trypsin is defined as efficiency of digestion which is dependent on factors such as the relative position of amino acid residues to the K/R cleavage site and the secondary structure of the substrate. This secondary structure could also include structures reproducibly formed in urea. Therefore, we sought to develop a strategy that would allow us to visualize how well each enzyme digested HSA across the HSA sequence. We also sought to determine whether these trends changed over time, dependent on the position of the cleavage site in the HSA sequence, and whether these trends were different for the two trypsins.

The intensities of the FT, MC and ST peptides dependent on the HSA sequence position were plotted in Figure 5. The intensities for each amino acid (AA) position along the HSA sequence were computed as the sum of the intensity of all peptides from a particular category (FT, MC, or ST) at that amino acid position. These intensities were plotted for the 2 hour (Figure 5A) and 18 hour (Figure 5B) digests (for clarity, only the M3 digests are shown, the others responded similarly). At 2 hours, the most abundant FT peptide (Figure 5A; AA positions 397-413, peptide K.VFDEFKPLVEEPQNLIK.Q) had an intensity value of 0.0028 (i.e., it contributed 0.28% to the total HSA abundance). The maximum MC peptide intensity - the most abundant peptide overall - (position 139-161, missed cleavage site at AA position 160, peptide R.LVRPEVDVMCTAFHDNEETFLKK.Y) was 0.0046 (see also Figure 4A). The maximum ST peptide intensity was 0.0001 (sequence positions 173-183, peptide Y.FYAPELLFFAK.R). For the 18 hour digests, the previously referenced FT peptide increased in intensity to 0.0030, whereas the intensity of the referenced MC peptide remained unchanged. A different peptide, R.MPCAEDYLSVVLNQLCVLHEK.T (AA positions 470-490), was then the most abundant FT peptide (0.0035, an increase from 0.0015 at 2 hours). The longer, overlapping peptide, K.RMPCAEDYLSVVLNQLCVLHEK.T (missed cleavage at AA position 469), proportionally decreased from 0.0029 to 0.0005, similar to the results that were observed in the PCA analysis (Figure 4B).

Figure 5. Peptide intensity as a function of amino acid position along the HSA protein sequence.

Stacked bars (y axis) represent the summed peptide intensities of fully tryptic (FT, black), missed cleavages (MC, red) and semi-tryptic (ST, blue) peptides covering a particular amino acid position (x axis). ST peptides are shown on the separate scale from FT and MC peptides. Data for M3 trypsin only, 2 hour (A) and 18 hour (B) digestion time points. Selected peptide sequences with the highest measured intensity for each of the cleavage rules are indicated and discussed in the text.

To further compare the differences between the two sequences of trypsin, the sequence maps of 18 hour digests were plotted for all the bovine and porcine trypsins (Figure 6). At 18 hours, the total sequence coverage was 99.7% and 99.2% (598aa and 592aa) for the porcine and bovine trypsins. Only two positions in the HSA sequence were not detected in any digest (Q220, R221) whereas 3 additional HSA amino acids were not detected in any bovine digest (546Q, 547I, and 548K). This visual representation of the data provides a complementary way to identify sequence regions showing most significant differences such as for the KAAFTECCQAADKAACLLPKLDELRDEGKASSAKQ and KRMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK sequences.

Figure 6. Percent of total intensity contributed by different categories of peptides for each amino acid position.

Data shown are for the 18 hour time point and are an extension of the analysis in Figure 5. A) The shaded colors in each graph represent each manufacturer listed from the lowest number outward: M1-M3 and M4-M6. Porcine peptides are plotted in the positive y axis and bovine in the negative y axis. The x axis represents the amino acid position of HSA, and the y axis is the relative % intensity of the total intensity detected at that amino acid position. Examples of sequence regions showing significant differences between porcine and bovine trypsins are indicated (see text). FT: fully tryptic, MC: missed cleavage, ST: semi-tryptic.

The sequence KRMPCAEDYLSVVLNQLCVLHEKTPVSDRVTK and its subsequences, are a representative example of peptides containing missed cleavages (Figure 7) which produced complex digestion dynamics, dependent on the length of the digest and the origin of trypsin. For example, in the 2 hour digests, both bovine and porcine trypsins were less likely to produce the fully tryptic peptide R.MPCAEDYLSVVLNQLCVLHEK.T, thus favoring the K.R cleavage site over the R.M site (as would be expected according to previous studies⁴⁴). With a longer digestion time, the porcine trypsins become more effective at producing this specific R.M cleavage, whereas the bovine trypsins’ ability to cleave at that site did not significantly change. Regardless of the content and relative location of amino acid residues near the cleavage site, digests from bovine trypsins produced nested sequence sets containing more abundant MC peptides.

Figure 7. Heat map of peptide abundances.

The longest sequence and its subsequences identified in different digests are shown for one exemplary case (see text for detail). The heatmaps were plotted using the sum of the intensities for each peptide ion that was detected for that sequence. The mean intensity for the 3 LC-MS/MS replicates were then calculated for each trypsin.

The differences in the specificities of several commercially available trypsins were also recently observed in the analysis of more complex samples ^{38, 39}. Due to significantly increased sample complexity, the number and relative intensity of the ST and MC peptides identified in those studies was lower (e.g., ST peptides contributed ~5% of total protein abundance in the analysis of eight-protein mixtures³⁹ vs. ~ 20% in our study). The experimental design used in our work (including lot-to-lot analysis and two digestion time points for each trypsin), and the computational pipeline and visualization tools assembled for the analysis of the data, were developed to best ascertain the differences in the efficiency of digestion between different trypsins. Importantly, by using a single protein as the substrate and performing peptide identification using a comprehensive HSA spectral library we were able to identify more MC and ST peptides from the target protein. Additionally, the sample preparation steps in our study did not include fractionation of the digested samples, thus reducing the sources of measurement variability. As a result, our study allowed a more consistent and robust detection of the differences amongst the ST, MC, and FT peptides when considering the origin of the trypsin. On the other hand, because we used a single protein as the substrate, we did not attempt to ascertain how digestion may be influenced by the presence of other proteins as is the case for complex protein samples ³³.

Conclusion

We characterized the performance of six commercially available trypsins which produced distinct differences depending on the origin (porcine vs. bovine). By limiting the substrate to a single protein (HSA), and digesting it for 2 and 18 hours, we were able to gain a more complete picture of the complex digestion process. An initial experiment suggested high reproducibility amongst digest replicates whereas the main analysis revealed that porcine and bovine trypsins reproducibly produced a different complement of peptides. Of these, the fully tryptic peptides were observed with the lowest variance of peptide counts and abundance. There were significant differences for how the bovine and porcine trypsin digests produced different sequences and sub-sequences arising from missed cleavages. The bovine trypsin digests produced a higher number and intensity of MC peptides, whereas porcine trypsins produced more ST peptides. Overall, our analysis suggested that peptide release during the protein digestion depends on multiple factors, including the digestion conditions, the sequence properties of the substrate, and the activity of the protease. Further work is required, using well defined protein substrates such as HSA, but also using complex protein samples in order to obtain a more complete characterization of proteolytic digestion in a typical proteomic experiment. Further studies of the structural relationship between enzyme and substrate together with other causative effects (e.g., conditions for solubility, denatured structure of HSA) to which these differences arise are also needed. The strategies and tools for visualization and data analysis presented in this work, including quantitative sequence maps and PCA, should be useful in these efforts. Further characterization of the complex patterns of trypsin activity and selectivity, together with the improved computational methods for detection of ‘proteotypic’ peptides based on revised trypsin digestion rules, should improve the reliability of peptide detection and quantification. This, in turn, is expected to lead to improved statistical outcomes for proteomic studies involving targeted (e.g., SRM-based) and untargeted experimental workflows.