TAILS N-terminomics of human platelets reveals pervasive metalloproteinase-dependent proteolytic processing in storage

Anna Prudova, Katherine Serrano, Ulrich Eckhard, Nikolaus Fortelny, Dana V. Devine and Christopher M. Overall

Key Points

  • TAILS proteomics identified 2938 human platelet proteins, pervasive proteolytic processing, and precise proteolytic cleavage sites in stored platelets.

  • During storage, metalloproteinases were predominantly involved in proteolytic processing, while other proteinases were mainly involved in degradation.


Proteases, and specifically metalloproteinases, have been linked to the loss of platelet function during storage before transfusion, but the underlying mechanisms remain unknown. We used a dedicated N-terminomics technique, iTRAQ terminal amine isotopic labeling of substrates (TAILS), to characterize the human platelet N-terminome, proteome, and posttranslational modifications throughout platelet storage over 9 days under blood-banking conditions. From the identified 2938 proteins and 7503 unique peptides, we characterized N-terminal methionine excision, co- and posttranslational Nα acetylation, protein maturation, and proteolytic processing of proteins in human platelets. We also identified for the first time 10 proteins previously classified by the Human Proteome Organization as “missing” in the human proteome. Most N termini (77%) were internal neo-N termini (105 were novel potential alternative translation start sites, and 2180 represented stable proteolytic products), thus highlighting a prominent yet previously uncharacterized role of proteolytic processing during platelet storage. Protease inhibitor studies revealed metalloproteinases as being primarily responsible for proteolytic processing (as opposed to degradation) during storage. System-wide identification of metalloproteinase and other proteinase substrates and their respective cleavage sites suggests novel mechanisms of the effect of proteases on protein activity and platelet function during storage. All data sets and metadata are available through ProteomeXchange with the data set identifier PXD000906.


Platelets are small, anucleate cells that play an important role in hemostasis and immune function. Platelet concentrates (PCs) are commonly used as supportive therapy for thrombocytopenia, cancer, AIDS, and organ transplant patients. However, the shelf life of platelet products is typically limited to 5 days due to a risk of bacterial contamination and a loss of platelet function, known as the platelet storage lesion (PSL). Recent advances in pathogen inactivation can extend storage up to 7 days, yet PSL remains a significant constraint.1

The onset of PSL is a multifactorial process that appears to be triggered by blood collection and component processing and results in gradual loss of platelet posttransfusion recovery and hemostatic function. The characteristics of PSL include altered metabolic activity and agonist response, microvesiculation, shape change, surface expression of several protein markers, and cytoskeleton rearrangement.2 Due to the anucleate nature of platelets, transcriptional studies of PSL provide limited mechanistic insight, making proteomics a more suitable tool. However, to date, proteomics has focused mostly on freshly isolated platelets, identifying up to 4000 proteins.3 The few storage studies suggested apoptosis and cytoskeleton as affected systems that may be associated with shortened survival of transfused platelets.4-6 Whereas some changes were attributed to active translation from preexisting messenger RNAs,7 others were due to altered protein posttranslational modifications (PTMs), such as phosphorylation8,9 and proteolysis.4,8 However, despite the importance of PTMs to protein and platelet function, the occurrence and regulation of most PTMs during platelet storage remains unknown.

Proteolysis represents one of the important, but often overlooked, PTMs that is ubiquitous, irreversible, and affects every protein in the cell through excision of the initiator methionine, removal of signal peptides or propeptides, and eventually degradation. As opposed to degradation to completion, proteolytic processing by highly controlled, precise cleavage (often at a single site) of proteins yields stable, structurally and functionally distinct protein chains.10 Proteolysis plays key roles in platelet signal transduction.11,12 Although activation of caspase13 and complement14 proteolytic cascades has been linked to PSL, platelet storage with respective cysteine15,16 and serine17 protease inhibitors failed to prevent PSL features, thus questioning the mechanistic importance of these proteases. In contrast, storage with the broad-spectrum metalloproteinase inhibitor GM6001 improved platelet posttransfusion function and recovery in a murine model.18 Metalloproteinases are a family of membrane or extracellular zinc-dependent proteases that include matrix metalloproteinase (MMP) and a disintegrin and metalloproteinase (ADAM) protease subfamilies, implicated in proteolytic shedding and processing of many bioactive molecules.19 Whereas platelets express several MMPs (1, 2, 3, 9, and 14),20 murine ADAM knockout studies supported by in vitro biochemical assays suggest an important role of ADAMs in PSL development.18,21,22 However, a mechanistic understanding of the actual role metalloproteinases play in PSL requires knowledge of their in vivo substrates and the corresponding cleavage sites that often change protein bioactivity or localization.23

Despite its importance, proteolytic processing is inherently difficult to detect using conventional proteomics.10 Recently, several specialized proteomics approaches for protease substrate characterization have been developed,24-26 including an N-terminal enrichment method from our laboratory, terminal amine isotope labeling of substrates (TAILS).27 We used TAILS to characterize proteolytic processing and N-terminal PTMs in human platelets stored under blood-banking conditions for the first time. We identified 3321 N termini by TAILS. The majority (77%) of all identified N termini represented stable proteolytic products (many previously unreported), suggesting a pervasive and potentially important role of global proteolytic processing in platelets during storage. Furthermore, we used broad-spectrum and metalloproteinase inhibitors to define proteome-wide protease-related changes. The identified cleavage sites in structural, metabolic, and other signaling proteins suggest potential mechanisms of metalloproteinase and other proteinase involvement in PSL.

Materials and methods

Blood collection

Blood was collected from healthy consenting volunteers according to a protocol approved by the Canadian Blood Services research ethics board. Leukoreduced buffy coat PC pools (4 donors/pool) were prepared at the NetCAD laboratory according to the Canadian Blood Services standard operating procedures as described in supplemental Methods (available on the Blood Web site).

Platelet storage

After resting, PCs were subdivided into several minibags (maintaining the same surface-to-volume ratio). Where larger volumes were needed, several PCs were pooled prior to their subdivision. For inhibitor studies, after a 1-hour rest, equal volumes of sterile-filtered inhibitor or vehicle alone (dimethylsulfoxide [DMSO]) were added through a sample port under sterile conditions. For broad protease inhibition, cOmplete mini EDTA-free protease inhibitor cocktail tablets (1 tablet/10 mL) (Roche) were used. For metalloproteinase inhibition, 10 μM marimastat in 0.16% DMSO final was used. Platelets were stored in a controlled-temperature cabinet (20°C to 24°C) with agitation. Samples were taken aseptically through bags sample ports at the same time on different days. After 9 days, PCs were tested for sterility using BacT/ALERT bottles. All reported units were culture negative. In total, 19 platelet units were analyzed, 2 of them by proteomics.

Platelet proteome preparation

Platelets were sedimented (500g, 10 min) and washed as described previously,5 with minor modifications. To prevent proteolysis during washing steps, all buffers were supplemented with protease inhibitors (10 mM EDTA+ cOmplete miniprotease inhibitor cocktail by Roche; 1 tablet/10 mL). Following washes, different cell types were counted using hematology analyzer (Advia-120; Siemens, Erlangen, Germany) to ensure consistent preparations. Proteomes were prepared for TAILS analyses as described in supplemental Methods.

TAILS workflow and mass spectrometry analysis

Platelet proteomes were analyzed by iTRAQ-TAILS using 4-plex iTRAQ reagents (Applied Biosystems) as described previously28,29 and in supplemental Methods. TAILS utilizes a commercially available hyperbranched polyaldehyde polymer (HPG-ALD) to bind and remove internal trypsin-generated peptides under reductive conditions. HPG-ALD is available without commercial or company restriction from Flintbox Innovation Network, The Global Intellectual Exchange and Innovation Network ( For 8-plex analyses, 8-plex iTRAQ reagents (Applied Biosystems) were used in fivefold reagent/protein weight excess with 50% DMSO final. The 8-plex TAILS was performed essentially as with 4-plex reagents (30-minute labeling), except that labeling was for 1 hour at room temperature. For 4-plex and 8-plex analyses, 0.5 and 0.3 mg protein/condition per iTRAQ channel were used, respectively.

The resulting TAILS and pre-TAILS samples were fractionated offline using a strong cation exchange and analyzed by tandem mass spectrometry (MS/MS) using a QSTAR XL Hybrid electrospray ionization mass spectrometer (Applied Biosystems/MDS SCIEX) as described previously.29 In total, 5 platelet proteome analyses of 2 platelet units were performed (one 4-plex TAILS, one 4-plex pre-TAILS on one unit, and three 8-plex TAILS experiments on the other). MS/MS data were searched against a human UniProt protein database (v3.42) by Mascot v2.3.0 (Matrix Science) or X! Tandem and combined by iProphet (false discovery rate [FDR] ≤5%) after PeptideProphet analyses as described previously28 (in full in supplemental Methods). Missing proteins in the human proteome were analyzed at FDR ≤1%, and the spectra were manually inspected. All data have been deposited to the ProteomeXchange30 with identifier PXD000906.

Platelet in vitro tests

Platelet concentration and mean platelet volume were obtained by processing buffy coat samples on a hematology analyzer (Advia-120; Siemens, Erlangen, Germany). The extent of shape change (ESC) assay was used to assess platelet quality during storage and reflects the platelet responsiveness to activation with adenosine 5′-diphosphate. The ESC measure was made in triplicate by light-scattering assessment using an aggregometer (SPA 2000; Chronolog, Havertown, PA). pH, pO2, pCO2, glucose, and lactate were measured directly from a syringe immediately after sampling PC on a blood gas analyzer (GEM Premier 3000; Instrumentation Laboratory, Orangeburg, NY). For lactate results exceeding the normal range, a 1:1 dilution of the sample was made with phosphate-buffered saline, and the test was repeated. Platelet swirl was observed and recorded on all samples by the same investigators to avoid subjective differences. Western blotting and flow cytometry analyses were performed as described in supplemental “Methods.”

Results and discussion

Experimental design

We aimed to analyze clinically relevant and transfusion-relevant changes during platelet storage. Therefore, PCs were prepared, handled, and stored using standard blood-banking protocols. In contrast to studies analyzing platelets isolated directly from fresh blood donations, the earliest platelet data we obtained were from PCs sampled on the day of production (day 1), 1 day postcollection, and therefore do not represent native platelets but rather reflect their postproduction state.

To assess both global proteome and protease-specific N-terminome changes during platelet storage, we used a 2-pronged quantitative proteomics approach in combination with protease inhibitor studies (Figure 1A).31 We used 4-plex and 8-plex iTRAQ labeling for relative quantification of time-resolved changes between multiple conditions in a single analysis. In TAILS analyses, enrichment of natural N termini (protein termini as they were translated or after methionine excision or protein maturation) and those generated by proteolysis (internal or neo-N termini) expands the proteome dynamic range, increases coverage of low-abundance proteins, and enables simultaneous identification of protease cleavage sites and assessment of N-terminal PTMs (supplemental Figure 1). This also renders the approach well suited for detecting “missing” proteins32 in the human proteome as part of the Human Proteome Project. In addition, analysis of the samples before the N-terminal enrichment (pre-TAILS) provides traditional shotgun-like data.

Figure 1

Characterization of platelet proteome and degradome during storage. (A) Experimental setup. Blood was collected from healthy volunteers on day 0. Platelet units were prepared using buffy coat method and according to Canadian Blood Services blood-banking protocols on day 1. After resting, platelet units were subdivided into smaller units and vehicle or protease inhibitors were added. The resulting platelet units were stored for 9 days under the blood-banking conditions at 22°C with agitation. Platelet aliquots were sampled throughout storage. Platelets were washed, lysed, and subjected to proteomic analyses using TAILS and pre-TAILS approaches (similar to shotgun analysis) to yield the platelet N-terminome and proteome, respectively. In total, 4 TAILS analyses and 1 pre-TAILS analysis were performed. (B) Overlap in the number of total peptides (left) and proteins (right) identified by pre-TAILS and TAILS analyses. (C) Distribution of internal and N-terminal peptides among 5648 and 1878 peptides identified before (left) and after (right) N-terminal enrichment in 4-plex pre-TAILS and TAILS analyses of the same sample, respectively. N-terminal peptides originally present in the sample include naturally blocked (by Nα-acetylation) and free N termini. Free N termini become iTRAQ-labeled in TAILS protocol unless cyclization renders these unreactive (cyclization occurs with peptides with N-terminal glutamine, asparagine, and carbamidomethylated cysteine residues). The unlabeled peptides represent internal peptides released by trypsin during the proteome digest prior MS analysis. (D) Western blot validation of 2 “missing” proteins in the human proteome identified proteomically: nuclear pore membrane glycoprotein 210-like (Nup210L; left panel) and aquaporin-6 (right panel). Protein expression in 2 individual platelet units is shown. Arrows indicate different positions of the glycosylated and oligomerization forms of aquaporin-6 as previously assigned.31

The platelet proteome and N-terminome

From 5 platelet proteomes, we identified 7503 unique peptides corresponding to 2938 proteins (Figure 1B and supplemental Table 1), comprising the most extensive human platelet data set among storage studies and the second-largest data set after a recent study of fresh human platelets.3 TAILS yielded a unique 35% fraction of all identified peptides and proteins (2627 and 1033, respectively) (Figure 1B and supplemental Tables 1-3). For comparison, an earlier COFRADIC N-terminomics study identified 345 proteins in human platelets.33 The utility of TAILS N-termini enrichment was further demonstrated by comparison of TAILS and pre-TAILS analyses of the same sample (Figure 1C and supplemental Tables 5 and 6). Whereas shotgun-like analyses of the sample before N-terminal enrichment (pre-TAILS) returned only few N-terminal peptides (combined 18%), TAILS yielded 93% of N-terminal peptides composed of naturally acetylated N termini (17%), cyclized N termini (24%), and N termini that were originally free αN termini and now were iTRAQ labeled (52%) in the TAILS workflow.

Combining pre-TAILS and TAILS data, we identified with high confidence (FDR ≤1%) 10 peptides from 10 proteins that were previously classified as missing32 with little or no evidence at the protein level (Table 1, supplemental Table 12). These included 2 “uncertain” and 8 detected-only-as-transcripts proteins, expression of 2 of which was validated by western blotting (Figure 1D). Among the 10 peptides, 7 were N termini, with 5 detected by TAILS. Such proteins, previously missed by shotgun proteomics, could be identified by TAILS due to its 5 orders of magnitude dynamic range27 and enrichment for different semitryptic peptides that have altered m/z, ionization, and fragmentation properties compared with the tryptic peptides analyzed by traditional proteomics.34 Thus, TAILS enables comprehensive proteome characterization toward the Human Proteome Project goal to experimentally detect all of the protein sequences, including PTMs, expressed from the human genome throughout different cell stages.35

Table 1

Mass spectrometric identification of proteins classified by neXtProt as “missing”

N-terminal processing in platelets

Among the 2960 annotated N termini identified in 5 experiments (supplemental Table 4), 676 (23%) represented natural protein N termini, including naturally acetylated (397, or 59%) and free N termini (279, or 41%) (Figure 2A). As expected, acetylated peptides were mostly represented by natural N termini (397, or 59%), whereas internal neo-N termini prevailed among the free N termini (2010, or 88%) (Figure 2B). Consistent with previous reports,28,29,34 the majority of natural N termini started after methionine excision (125, or 56%; 296, or 74% for free and acetylated natural N termini, respectively). Free natural N termini also started after signal or propeptide removal (55, or 24%; 7, or 3%, respectively). Noteworthy, many new instances of N-terminal PTMs that were not previously observed were identified here.

Figure 2

Characterization of N-terminal modifications in platelets. (A) Distribution of platelet N termini as natural (as synthesized or after maturation) and internal (proteolysis-derived neo-N termini) termini among the 2960 high-confidence positionally annotated peptides identified in total by TAILS and pre-TAILS analyses. Fractions of identified Nα-acetylated (Ac) vs free (which become iTRAQ labeled in TAILS protocol) N termini within each category are indicated. Frequency distribution of N-terminal processing after removal of the initiator methionine, propeptides, and signal peptides is shown. (B) Distribution of natural and internal N termini among 2290 free and 670 acetylated peptides among 2960 high-confidence positionally annotated peptides. (C) Amino acid frequency distribution in P2′ position following intact initiator methionine [ie, (Ac)M-X)] (acetylated or not) for natural N termini (open bars) vs internal N termini (black bars). (D) Amino acid frequency distribution in P1′ position (acetylated or not) following initiator methionine excision [ie, M.(Ac)X] in natural N termini (open bars) vs internal N-termini (black bars). (E) Amino acid distribution in P2′ position following intact initiator methionine in acetylated (ie, AcM-X) natural N-termini (open bars) vs internal N termini (black bars). (F) Amino acid frequency distribution in P1′ position following initiator methionine excision for acetylated (ie, M.AcX) natural N termini (open bars) vs internal N termini (black bars). Gray bars show amino acid frequency distribution in P1′ position for acetylated internal N termini lacking putative initiator methionine in P1 position (ie, Z.AcX, where Z is any amino acid other than methionine). Number of peptides in each category is indicated in brackets. *Indicates N termini that are potential products of translation at alternative start sites predicted based on their N-terminal amino acid distribution analysis.

The amino acid distribution of the N-terminal peptides demonstrated that aspartic and glutamic acids after the initiator methionine prevented its excision by methionine aminopeptidases (Figure 2C), whereas alanine and serine promoted this activity (Figure 2D), consistent with previous reports.28,29,34 The similarity of the distribution profiles of natural vs internal N termini starting with or after putative initiator methionine (Figure 2C-D white vs black bars) suggested that the majority (105/155) of the N termini annotated as internal (Figure 2C-D black bars) could be natural N termini resulting from alternative translation start sites.29

Analysis of the in vivo acetyltransferase specificity showed preference for methionine (where it was intact; Figure 2E) and alanine and serine (when initiator methionine was excised; Figure 2F), as seen previously.29 Amino acid profiles of natural N termini (Figure 2E-F white bars) were in good agreement with those of internal peptides starting after or with a putative initiator methionine (Figure 2E-F black bars). This suggests that the latter are products of alternative translation and cotranslational acetylation. In contrast, 26 acetylated internal N termini without putative initiator methionine (Figure 2F gray bars) showed a unique profile, suggesting that a different acetyltransferase was responsible for posttranslational acetylation. Such posttranslational acetylation of internal N termini has been recently proposed to stabilize the respective cleaved proteins, thus allowing functional diversification of the proteome under conditions of limited de novo protein synthesis such as in mature erythrocytes34 or platelets.

Strikingly, internal N termini that mostly are products of proteolysis accounted for 77% (2284) of all N termini (Figure 2A), thus reflecting a previously uncharacterized high proteolytic activity in stored platelets. This prompted us to define changes in proteolysis over time and in response to protease inhibitors.

Changes to the platelet proteome and degradome during storage

We used 4-plex iTRAQ-TAILS to compare relative amounts of proteolytic products after 2, 4, or 9 days of platelet storage. Of 1878 peptides (1014 proteins; supplemental Table 6), 1102 could be quantified (supplemental Table 7); the others were N-acetylated and had no lysine side chains to bear the iTRAQ tag. To determine the significance cutoff ratio for differentially abundant peptides, we analyzed the distribution of iTRAQ ratios for natural N termini on day 4 with or without protease inhibitors, as this specific peptide subset is least likely to be dramatically affected by protease inhibitors. Whereas protease inhibitors cannot directly cause high iTRAQ ratios for original protein N termini, they prevent degradation and therefore their disappearance, thus resulting in lower iTRAQ ratios and a slightly asymmetrical distribution (Figure 3A). We chose 2 standard deviations as our change significance cutoff, which corresponded to iTRAQ ratios of 2 and 0.5.

Figure 3

Extensive proteolytic processing in human platelets during storage. Platelets were stored under blood-banking conditions for 9 days with or without EDTA-free cOmplete protease inhibitor cocktail, and the platelet proteome from different days was characterized by 4-plex iTRAQ TAILS (day 2, day 4, day 4 + inhibitors [Inh], day 9). Of 1878 high-confidence peptides identified by TAILS, 1102 were quantified and had positional annotation in UniProt/Swiss-Prot. (A) Distribution of log2 iTRAQ (day 4/day 4+Inh) ratios for 414 natural N termini identified in 4-plex TAILS analysis. Two times the standard deviation corresponding to the ratios of 2 and 0.5 were chosen as statistical significance ratio cutoffs and are indicated by a dashed line. Low iTRAQ (D4/D4+Inhibitors) ratios (≤0.5) of natural N termini indicate decrease in their abundance due to proteolysis in the absence of inhibitors. (B) Distribution of log2 iTRAQ (D4/D2) ratios for internal N termini identified in 4-plex TAILS analysis. The statistical significance cutoffs of 2 and 0.5 are indicated by dashed lines. High iTRAQ (D4/D2) ratios (≥2) indicate increased production of internal neo-N-terminal peptides due to proteolytic processing in the absence of inhibitors. N termini found with intermediate iTRAQ ratios (0.5 ≥ ratios ≤ 2) represent products of proteolysis that are not changing in time. (C) Quantitative overview of the 481 N-terminal peptides that were significantly up- (ratios ≥2; right side of the graph) or downregulated (ratios ≤0.5; left side of the graph) in at least 1 comparison (Day 4/Day 2, Day 9/Day 2, or Day 4/Day 4+Inh). Percentile fractions of internal N termini (black) and natural N termini (white) among downregulated or upregulated N termini are shown. Total number of peptides in each category is indicated. (D) Fold enrichment of internal (black bars) vs natural (white bars) N termini for significantly up- (ratios ≥2; right side) or downregulated (ratios ≤0.5; left side) N termini as summarized in Figure 3C. For each category, percentile distribution of internal and natural N termini was normalized for the total levels of natural and internal N termini (ie, 38% and 62%, respectively) observed in the subset of 1102 quantifiable peptides and expressed as log2 ratio.

We categorized stable products of processing (ratios ≥2), degradation resulting in low ratios (ratios ≤0.5), and proteolytic products that do not change with the inhibitors or in time (0.5 ≥ ratios ≤ 2) (Figure 3B). Here, isotopic labeling is key to distinguishing between processing vs degradation and between changes in proteolysis (eg, between control and inhibitor or between day 1 and day 3 samples) and products of background proteolysis present in all samples at the same level. Some other N-terminal enrichment methods24-26 do not offer this option. Whereas some gel-based methods differentiate processing from degradation, they suffer from the usual gel-method limitations (eg, limited sensitivity and resolution), and they rarely identify the actual cleavage site (reviewed in Doucet et al10 and Agard and Wells36). Specifically, we identified 340 (31%) and 405 (37%) N termini with significantly altered abundance on days 4 and 9 compared with day 2, respectively (Figure 3C). Interestingly, the majority of the affected N termini were upregulated and therefore represented products of proteolytic processing (Figure 3C, right vs left side). Most of these upregulated peptides were internal neo-N termini (87% and 77% on days 4 and 9, respectively) compared with the total levels of neo-N termini identified in this sample (62%) (Figure 3D right side). In contrast, downregulated peptides were largely represented by natural N termini (66% and 62% on days 4 and 9, respectively) compared with the total levels of natural N termini identified (38%) (Figure 3D left side). These quantitative findings further support high proteolytic activity in general and highlight the predominant role of proteolytic processing as opposed to degradation of proteins in stored platelets.

Effects of broad-spectrum protease inhibitors during platelet storage

To gauge contributions of different protease classes to proteolysis during storage, we incubated platelets for 4 days with or without the EDTA-free cOmplete protease inhibitor cocktail, which targets serine and cysteine proteases but is ineffective toward metalloproteinases or proteasome threonine proteases. In total, 83 peptides (7%) were significantly (≥twofold) affected with 27 (2%) upregulated and 56 (5%) downregulated N termini (Figure 3C and supplemental Table 8). Whereas neo-N termini dominated among the upregulated peptides (70%), they were slightly underrepresented among the downregulated peptides (54%), compared with the level of 62% observed in the whole data set (Figure 3D). Although a significant fraction of peptides downregulated from day 2 to day 4 was rescued by the inhibitors (19/50 or 38%), ratios of only 5% (15/290) of stable proteolytic products (high iTRAQ ratios) were affected (supplemental Figure 2). Thus, serine and cysteine proteases are likely more involved in protein degradation during platelet storage, whereas other classes of proteases are responsible for proteolytic processing resulting in stable neo-N termini.

Examples of inhibitor-responsive proteolytic processing are shown in Figure 4. The isotope ratios for the mature N terminus of tropomyosin-4 starting after removal of the initiator methionine and subsequent Nα-acetylation (Figure 4A) were unaltered throughout storage and in response to protease inhibitors. In contrast, a neo-N-terminal peptide was upregulated from day 2 to 4 (3.1-fold), and this was reduced by the protease inhibitors (2.2-fold). Thus, this cytoskeletal protein showed constant levels of expression that was accompanied by increased proteolytic processing at position Arg105↓106Leu during storage. In another example, a significantly upregulated (≥twofold) mature N terminus and neo-N terminus of l-lactate dehydrogenase A-chain indicated a concerted increase in expression and processing of this metabolic enzyme during storage (Figure 4B).

Figure 4

Examples of proteolytic processing identified by TAILS in stored platelets. Platelets were stored under blood-banking conditions for 9 days with or without EDTA-free cOmplete protease inhibitor cocktail, and the platelet proteome from different days was characterized by 4-plex iTRAQ TAILS (day 2, day 4, day 4 + inhibitors, day 9). Examples of the identified N termini are shown for tropomyosin-4 (A), l-lactate dehydrogenase A (B), and TCP-1β (C). Peptides identified by TAILS are underlined within the respective protein sequences; sequence positions of internal neo- and natural N termini are indicated by arrows. Start and end position of the identified peptides and their N-terminal modifications, sequences, preceding amino acids, and iTRAQ ratios are shown. Ac denotes N-terminal acetylation.

In contrast, low iTRAQ ratios of mature protein N termini could be due to increased degradation/processing or downregulated expression or both. Thus, correct interpretation of diminished ratios requires quantitative data for additional peptides from the same protein or a follow-up using biochemical approaches. For instance, mature N terminus (Nα-acetylated after initiator methionine excision) of T-complex protein 1 subunit β (TCP-1β) decreased on day 4, whereas a second, overlapping, neo-N terminus dramatically increased (Figure 4C), thus consistent with cleavage at the Ser5↓6Leu bond. TCP-1β is a chaperone assisting folding of cytoskeletal proteins (eg, tubulin and actin),37 which mediate platelet shape change upon activation. Thus, TCP-1β processing may also play a role in PSL.

Global effects of a metalloproteinase inhibitor during platelet storage

To test for any potential effects of metalloproteinases during storage, we incubated human platelets for up to 7 days with marimastat, a zinc-chelating broad-spectrum metalloproteinase inhibitor drug. We detected no clinically significant changes in platelet swirl, pH, soluble gases, glucose, and lactate levels or P-selectin expression (data not shown). However, marimastat prevented time-dependent increase in platelet volume and resulted in higher platelet concentration compared with controls (Figure 5A-B), likely by blocking activation-related platelet shape change and aggregation throughout storage, respectively. Consistently, marimastat-incubated platelets showed an improved agonist response using the ESC assay (Figure 5B), further supporting the notion of inhibition of platelet activation by marimastat during storage. Although differences in agonist response were statistically significant (P = .030, n = 18, paired t test), not all platelet units reacted to marimastat equally, where less responsive units also had dampened changes in platelet volume (P = .022; Figure 5C). Together, these results suggest that although metalloproteinases appear to dominate some platelet processes (aggregation), others (agonist-induced shape/volume change) are also controlled by other factors.

Figure 5

Effect of exogenous broad-spectrum metalloproteinase inhibitor on stored platelets. Platelets were stored under blood-banking conditions for 7 days in the absence or presence of 10 μM marimastat (n = 18). (A) Platelet volume changes from day 1 to day 7 and in response to marimastat. Mean platelet volume (n = 18) for platelets stored with 10 µM marimastat (black bars) or DMSO vehicle control (empty bars). *P < .05 (paired t test with Bonferroni correction). Error bars indicate standard deviation of the mean. (B) Marimastat effect on platelet volume (P = 2.2 × 10−5), concentration (P = 1.3 × 10−5), and ESC (P = .03) on day 7 expressed as log2(marimastat/DMSO). Gray line indicates no change (ratio of 1); P values represent results of paired t tests. Arrows point out the parameters of the platelet unit analyzed by proteomics. (C) Linear correlation between platelet volume and ESC on day 7 expressed as log2(marimastat/DMSO). P value and correlation coefficient are shown.

Next, we compared changes in the platelet N-terminome during storage with marimastat or with vehicle alone using 8-plex iTRAQ. Among the 1999 N termini (1000 proteins) identified in 2 technical replicates (supplemental Table 9), 1420 N termini (739 proteins) were annotated and 1111 N termini (558 proteins) were quantifiable (supplemental Table 10). Among the quantifiable N termini, 178 peptides (16%) were significantly changed (≥twofold) on at least 1 sampling day (day 3, 5, or 7) when marimastat-stored platelets were compared with their age-matched vehicle controls (supplemental Table 11). The numbers of affected peptides were approximately equal among sampling days, suggesting steady activity of metalloproteinases throughout storage (Figure 6A). Further, whereas protease inhibitor cocktail (mostly affecting cysteine and serine proteases) showed a bias toward inhibiting degradation (as seen by a twofold prevalence of low-ratio peptides; Figure 3C), metalloproteinase inhibition equally affected proteolytic processing and degradation (as evident from approximately equal numbers of low- and high-ratio peptides; Figure 6A). These observations suggest a more prominent role of metalloproteinases in proteolytic processing during platelet storage compared with other protease classes.

Figure 6

Effect of exogenous broad-spectrum metalloproteinase inhibitor on platelet N-terminome during storage. Platelets were stored under blood-banking conditions for 7 days in the absence or presence of 10 μM marimastat. Two technical replicates were performed. Platelet proteomes from different time points and conditions were analyzed by 8-plex iTRAQ-TAILS analyses. Out of 1999 high-confidence peptides identified by TAILS in 2 technical replicates, 1111 could be quantified and had positional annotation in the UniProt/Swiss-Prot database. (A) Quantitative overview of the 178 N-terminal peptides that were significantly up- (ratios ≥2; right side of the graph) or downregulated (ratios ≤0.5; left side of the graph) in response to marimastat on at least 1 sampling day (ie, as indicated by iTRAQDMSO/Marimastat on days 3, 5, and 7). Percentile fraction of internal or neo-N termini (black bars) and natural N termini (white bars) is shown. Number of peptides in each category is shown. (B) Fold enrichment of neo- (black bars) vs natural (white bars) N termini for significantly up- (ratios ≥2; right side of the graph) or downregulated (ratios ≤0.5; left side of the graph) N termini are shown. For each category, percentile distribution of internal and natural N termini was normalized for the total levels of these N termini observed among a total of 1111 quantifiable and positionally annotated peptides (ie, 31% and 69% of natural and internal N termini, respectively) and expressed as a log2 ratio. (C) Western blot validation of metalloproteinase substrates identified by TAILS: tubulin, ρ-GDI, gelsolin, talin. Protein expression was evaluated in 2 individual platelet units in at least 2 technical replicates, and representative blots are shown. Arrows and arrowheads indicate position of the full-length protein and cleavage products, respectively.

Interestingly, neo-N termini were overrepresented among both high- and low-ratio peptides (Figure 6B right vs left sides of the graph, respectively). Affected proteins, iTRAQ ratios for the corresponding peptides and types of marimastat responses are further summarized in Figure 7. Gene ontology analysis of the proteins affected by marimastat did not show enrichment of any specific category (data not shown), suggesting that direct and downstream effects of metalloproteinase activity influence a wide range of platelet pathways.

Figure 7

Effect of marimastat on the platelet N-terminome during storage. Heatmap overview of time-dependent changes in relative abundance of peptides from platelets stored in the absence (3 columns on the left) or presence (3 columns in the middle) of marimastat. iTRAQ values for 1111 quantifiable and positionally annotated high-confidence peptides were normalized for the starting value on day 1. Only the peptides that were significantly regulated up (ratios ≥2, yellow) or down (ratios ≤0.5, blue) in at least 1 condition (when compared with day 1) are shown. Black indicates intermediate iTRAQ ratios (0.5 ≥ ratios ≤ 2) that did not reach our statistical significance cutoff. Last three columns on the right show marimastat response of the same peptides on specific days (ie, DMSO/marimastat ratios on days 3, 5, and 7). Original and internal N-terminal peptides were sorted based on their ratios and presented separately. Generalized responses of the main peptide clusters are described on the left-hand side of the heatmap. *Denotes natural N termini of proteins where expression is affected by metalloproteinases/marimastat via an indirect mechanism (ie, through control of protein synthesis or uptake and not by direct cleavage of these proteins). Gene symbols of the corresponding proteins are shown.

Marimastat-sensitive proteolytic processing of tubulin, ρ-GDI, gelsolin, and talin, which were proteomically identified as substrates and were mechanistically relevant to the observed changes in platelet agonist response, was confirmed by western blotting (Figure 6C). Some of the proteolytic products decreased in the presence of marimastat (ie, fragments at ∼51, 70 kDa for talin; ∼80 kDa for gelsolin; ∼20 and 35 kDa on day 5 for tubulin; ∼15 and 20 kDa for ρ-GDI), suggesting that these were products of metalloproteinase activity. In contrast, other fragments were unresponsive to marimastat and therefore represented activity of other protease classes (ie, fragments at ∼62 and 95 kDa for talin, ∼80 kDa for gelsolin, and ∼30 kDa for tubulin). Some other fragments (ie, ∼110-kDa band for talin on day 5) increased in intensity in the presence of marimastat, potentially due to an initial cleavage by a protease of another class and then further degradation by metalloproteinases, blockade of which by marimastat would lead to fragment accumulation. Western blot analyses show time-dependent changes in protein levels that reflect protein synthesis and proteolysis. Increased full-length protein levels of talin and gelsolin from day 5 to day 7 are consistent with active translation of these proteins during platelet storage, which has been shown for platelet proteins by incorporation of isotopically labeled amino acids during storage.7,38 Neoepitope antibodies to specific cleavage sites in these proteins do not exist, rendering it difficult to unambiguously assign each band to a specific neo-N termini identified by MS. Nonetheless, the western blots are consistent with processing by metalloproteinases, either MMPs or ADAMs, during platelet storage.

Examples of metalloproteinase-dependent processing during platelet storage

We found evidence of ectodomain shedding of several transmembrane glycoproteins, including 3 peptides of GPIbα, a ligand-binding component of the von Willebrand factor receptor complex, GPIb-V-IX39 (Figure 8A). The PLHP(16)↓(17)HPICEVSKVASHLEVNCDKR peptide mapped to the mature protein N terminus after removal of the signal peptide and was equally abundant at all times and marimastat unresponsive. Thus, expression and maturation of this protein does not change significantly throughout platelet storage. In contrast, the second peptide, KLRG(480)↓(481)VLQGHLESSR, mapped to the internal GPIbα sequence and showed time-dependent accumulation that was completely inhibited by marimastat. This indicated time- and metalloproteinase-dependent processing during platelet storage. Notably, this cleavage event has been previously attributed to ADAM17 by in vitro studies utilizing recombinant ADAM17 and GPIbα-based peptides22; we now detect this processing in vivo. The third identified GPIbα peptide, KLR(479)↓(480)GVLQGHLESSR, was N-terminally extended by 1 amino acid compared with ADAM17-generated cleavage and showed no changes in time or in response to inhibition. The unaltered abundance of this neo-N terminus confirmed steady expression of GPIbα (as was indicated by levels of the mature N terminus), whereas the lack of marimastat response indicates a metalloproteinase-independent processing at this site. Proteolytic processing of GPIbα and GPV, another component of the same complex, was confirmed by flow cytometry (Figure 8D-E).

Figure 8

Examples of metalloproteinase-dependent proteolytic processing identified by TAILS and flow cytometry in stored platelets. Platelets were stored under blood-banking conditions for 7 days in the absence or presence of 10 μM marimastat. Two technical replicates from the same platelet unit were performed. Platelet proteomes from different time points and conditions were analyzed by 8-plex iTRAQ-TAILS analyses. Examples of the identified N termini are shown for glycoproteins 1bα (A), 1bβ (B), and VI (C). Peptides identified by TAILS are underlined; sequence positions of internal neo- and natural N termini are indicated by arrows. Peptide sequences show start and end position of the identified peptides and their N-terminal modifications, sequences, preceding amino acids, and iTRAQDMSO/Marimastat ratios on specified days. Graph inserts show time-resolved changes in relative peptide levels expressed as % of peptide levels on day 1 in the presence (dashed line) or absence (solid line) of marimastat. Processing of GPIbα (D) and GPV (E) was confirmed by flow cytometry. Receptor density for platelets stored with 10 µM marimastat (black bars) or DMSO vehicle control (empty bars). Values for individual platelets units and mean values (GPIbα, n = 7; GPV, n = 5) are shown. *P < .05 (paired t test with Bonferroni correction). Error bars indicate standard deviation of the mean.

TAILS also yielded multiple novel cleavage sites. The second component of the GPIb-V-IX complex,39 GPIbβ, was identified by 2 neo-N termini, LRGR(127)↓(128)LLPYLAEDELR and PALR(82)↓(83)TAHLGANPWR (Figure 8B). The first peptide did not show time- or marimastat-dependent changes, but the second was upregulated during storage, which marimastat blocked. This indicated a novel metalloproteinase-dependent proteolytic processing event during platelet storage. Whereas complete loss of the GPIbα ligand-binding ectodomain results in significantly impaired thrombosis in transgenic mice,40 the consequences of trimming of GPIbβ ectodomain are unknown. Notably, cleavage occurs between the leucine-rich-repeat domains involved in protein-protein interactions and a flanking stabilizing leucine-rich-repeat C-terminal domain. We hypothesize that loss of this domain may induce conformational changes and result in altered protein-protein interactions within the remaining leucine-rich-repeat domain of the trimmed GPIbβ molecule.

Another novel example is a metalloproteinase-dependent processing of GPVI, a collagen receptor on the platelet surface39 (Figure 8C). We detected a neo-N-terminal peptide, KEGD(150)↓(151)PAPYKNPER, that decreased during the first 5 days of storage, which marimastat blocked. On day 7, the response was reversed, suggesting that the relative contribution of GPVI degradation was negligible from day 5 and the observed changes rather reflect a metalloproteinase-dependent processing at the Asp150↓151Pro bond. This site is within an immunoglobulin-like C2-type 2 domain that is important for ligand binding and cell adhesion, and so processing here is likely to affect platelet hemostatic function posttransfusion. The soluble shed ectodomains might also have additional roles in signal transduction, independent of the function of the remaining membrane-bound glycoprotein. For instance, soluble fragments of GPIbα (also called glycocalicin),41 GPV,42,43 and GPVI44 found in human plasma are postulated as markers and mediators of thrombosis.

We detected family-wide activity of MMPs in PCs using a nonselective general MMP peptide substrate, which decreased upon addition of marimastat (supplemental Figure 4A). Gelatin zymography analyses failed to detect active forms of MMP-2 and MMP-9 (supplemental Figure 4B); however, 1 or more of the other 21 members of the MMP family or ADAMs could be active in the concentrates.

Overall, we have characterized and quantified changes undergoing in the platelet proteome and N-terminome during storage. Among the 7503 identified peptides corresponding to 2938 platelet proteins, we have defined co- and posttranslational N-terminal processing and identified 105 putative products of alternative translation and 10 proteins previously undetected anywhere in the human proteome. Our study provided an unbiased system-wide surveillance of global proteolytic processing during storage. We observed high levels (77%) of stable proteolytic products among N termini that exceeded previous observations in normal erythrocytes (64%)33 and inflamed skin (61%),29 where multiple proteolytic processes are activated. We propose that such extensive proteolytic processing represents a general regulatory mechanism to structurally modify and functionally diversify the proteome when de novo protein synthesis is limiting. Protease inhibitor experiments highlighted a predominant role of metalloproteinases in proteolytic processing as opposed to degradation (compared with other proteinases) and identified novel system-wide targets. The precise cleavage sites identified in this study provide a baseline for further studies by suggesting mechanistic connections between protease activity and functional deficits observed in PSL, although other regulatory factors at the crossroads with metalloproteinases remain unknown. A broad-spectrum inhibitor of metalloproteinases, marimastat was originally developed as an antineoplastic drug. Although marimastat was clinically safe in thousands of patients for up to 2 years, it failed to improve survival in cancer clinical trials and resulted in musculoskeletal pain in some patients and so was discontinued.45,46 Although the concentration used here is expected to be safe for transfusion, it remains to be tested whether storage of human platelets with metalloproteinase inhibitors offers posttransfusion advantage and could be implemented as a new regimen to improve the quality of platelets for life-saving transfusions.


Contribution: A.P. and K.S. participated in the project design, performed platelet experiments, and interpreted the data; A.P. performed TAILS, data analyses, activity assays, and western blotting and drafted the manuscript; K.S. performed flow cytometry analyses and revised the manuscript; U.E. and N.F. performed data and bioinformatics analyses and revised the manuscript; and C.M.O. and D.V.D. participated in the project design and were responsible for the project supervision, data interpretation, and manuscript revision and provided grant support.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christopher M. Overall, Centre for Blood Research, University of British Columbia, 4.401 Life Sciences Centre, 2350 Health Sciences Mall, Vancouver, BC V6T 1Z3, Canada; e-mail: chris.overall{at}


From the University of British Columbia Centre for Blood Research, the authors thank Dr W. Chen for mass spectrometer operation, Dr J. Kizhakkedathu for kindly providing the HPG-ALD polymer, and Dr Nestor Solis and Dr Philipp Lange for valuable discussions. They are grateful to the anonymous volunteers who donated blood for this study as well as the dedicated personnel at the Vancouver-based NetCAD laboratory of Canadian Blood Services.

A University of British Columbia Centre for Blood Research Strategic Training Program in Transfusion Science and the Centre for Blood Research collaborative award supported A.P. U.E. was supported by the postdoctoral fellowship from the Michael Smith Foundation for Health Research. C.M.O. holds a Canada Research Chair in Metalloproteinase Proteomics and Systems Biology. K.S. and D.D. acknowledge support from the Canadian Blood Services. This work was supported by grants from the Canadian Institutes of Health Research as well an Infrastructure Grant from MSHFR and the University of British Columbia Centre for Blood Research.


  • This article contains a data supplement.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

  • Submitted April 10, 2014.
  • Accepted October 1, 2014.


View Abstract