The human embryonic kidney cell line HEK 293 T was cultured in Dulbecco’s modified Eagle’s medium (DMEM). After the cells grew to 80% density in a 10-cm petri dish, the medium was washed out and the cells were harvested by centrifugation at 1,000 g for 5 min under room temperature. Then the cell pellets were washed with cold PBS buffer twice.

HEK 293 T cell pellets were suspended in lysis buffer [6 M guanidine hydrochloride, 100 mM HEPES (pH 8.0), and 1% (v/v) protease inhibitor cocktail (Roche, Swiss)] on ice for 30 min, followed by sonication for 2 min with 2 s on and 5 s off at 30% power. Then the lysates were centrifuged at 20,000 g at 4°C for 10 min and the supernatant was collected. Protein concentration was measured by using the BCA assay (Beyotime, China). For PBC technical evaluation, the 900 µg extracted proteins were used and divided into three equal parts (300 µg protein/each part). The proteins were reduced by using 5 mM dithiothreitol (DTT) at 56°C for 30 min and alkylated by using 15 mM iodoacetamide (IAA) in darkness at 25°C for 30 min; 20 mM DTT was added to the protein solution to quench the excess IAA. Then the proteome sample was digested with LysC (Hualishi, China) with an enzyme/protein ratio of 1:50 (w/w) at 37°C overnight.

For 2-PCA labeling, 10 mM 2-PCA (J&K Scientific, China) dissolved in 100 mM HEPES buffer (pH 8.0) was added to the 300 µl peptide solution (300 µg peptide) at 37°C for 16 h (MacDonald et al., 2015) For sulfo NHS-biotin labeling, 2 mM sulfo NHS-biotin (APExBIO, United States ) dissolved in 100 mM HEPES (pH 8.5) was added to the 2-PCA–labeled peptide sample at 37°C for 30 min. The labeled-peptide samples were dried in SpeedVac and then desalted by using Sep-Pak C₁₈ cartridges (50 mg sorbent per cartridge, Waters, United States ).

A volume of 1 ml streptavidin beads (GE Healthcare, United States) were washed twice with 800 µl PBS. The 2-PCA- and biotin-labeled peptide was re-suspended to 800 µl PBS, pH was adjusted to 7.5, and then it was incubated with streptavidin beads at room temperature for 1 h. The supernatant was collected by centrifugation at 300 g for 2 min, and the beads were washed with 800 µl PBS twice. The supernatant and washing solution was combined and dried in SpeedVac.

The C-terminal peptides were fractionated by using a home-made StageTip C₁₈ column. The StageTip C₁₈ column was made as follows. C₁₈ disks (3M, United States ) were cut by a hypodermic needle and pushed into P200 pipet tips. Then 2mg C₁₈ resin (Durashell C₁₈, Agela, China) was re-suspended in 200 µl acetonitrile (ACN), loaded into prepared pipet tips, and then centrifuged at 400 g for 10 min. The StageTip column was equilibrated with 150 µl water of 0.1% ammonium hydroxide (NH₃·H₂O) for three times, respectively; the centrifuge time was controlled in 10 min. Then the peptide sample was loaded onto the column, the column was washed by 150 µl water (0.1% NH₃·H₂O), and the peptides were eluted with 2%, 9%, 15%, 20%, 24%, 30%, and 80% ACN in water (0.1% NH₃·H₂O). The fractionations were dried in SpeedVac and desalted with ZipTip C₁₈ (Millipore, United States).

The sample was analyzed by using an EASY-nLC 1200 HPLC tandem with the Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific, United States). The peptide was resolved in buffer A (2% ACN in water and 0.1% formic acid) and separated by using a home-made C₁₈ capillary column (25 cm × 75 μm, 1.9 µm particle size, and 100 Å pore size) (Li et al., 2020b), A column oven was used and the heating temperature was set at 60°C (Kyte and Doolittle, 1982).

For PCA- and biotin-labeled evaluation, the peptide sample before and after labeling were detected by using 1 h gradient LC-MS/MS. The LC gradient was set as follows: 8 %–13% buffer B (90% ACN in water and 0.1% formic acid) for 20 min; 26% buffer B for 31 min, with a raise to 45% in 5 min; and finally 80% buffer B for 60 min. The flow rate was set to 300 nL/min. Then ions were scanned over 350–1,300 m/Z at a resolution of 12,000 (200 m/Z) with the automatic gain control (AGC) target of 5.0e5 and maximum injection time of 50 ms. The charge state included was 2–6 and dynamic exclusion was 60 s. The data-dependent mode was set up with a cycle time of 3 s, and MS2 data were acquired by higher-energy collisional dissociation (HCD) fragmentation and normalized collision energy (NCE) of 32%. The AGC target was set to 7.0e3 and maximum injection time was set to 35 ms.

The enriched C-terminal peptide was detected by 2 h gradient LC-MS/MS. The LC gradient was set as follows: 2%–5% buffer B for 3 min; to 16% buffer B for 52 min; to 35% buffer B for 50 min, with a raise to 47% for 10 min; and finally 80% buffer B for 120 min. The flow rate was set as 300 nL/min. Then peptides were scanned over 350–1,800 m/z at a resolution of 60,000 with the automatic gain control (AGC) target of 3e6 and maximum injection time of 45 ms. The charge state included was 1–5 and dynamic exclusion was 20 s. The data-dependent mode was set up with the top 10 most abundant precursors and subjected to MS/MS fragmentation, and MS2 data were acquired by higher-energy collisional dissociation (HCD) fragmentation and normalized collision energy (NCE) of 28%. The AGC target was set to 1e5 and the maximum injection time was set to 80 ms.

The raw data were searched against the UniProt homo proteome database (version 201,812) through Proteome Discoverer (version 2.2, Thermo Fisher Scientific) and loaded into the Mascot search engine (version 2.3, Matrix Science). The enzyme type was Lys-C/P. Up to two maximum missed cleavage was used. The precursor mass tolerance was set as 10 ppm, and the fragment mass tolerance was set as 0.02 Da. For samples before enrichment, carbamidomethyl (C) was set as static modification, and acetyl (protein N-term) and oxidation (M) were set as dynamic modifications. In addition, PCA (N-term), biotinylation (K), and biotinylation (N-term) were set as dynamic modifications for analyzing sulfo NHS-biotin–labeled samples. For sample of enrichment, carbamidomethyl (C) and biotinylation (K) were set as static modifications, and PCA (N-term) and oxidation (M) were set as dynamic modifications. The minimal peptide length was filtered with six amino acids. The results were filtered by a 1% false discovery rate (FDR) at PSM, peptide, and protein levels (Zhang et al., 2018; Wang et al., 2021).

For the open-search, all raw data files were processed using pFind software (version 3.1.0) with an open-search mode (Kahl et al., 2018). The enzyme type was set as Lys-C/P, and the maximum missed cleavage was 2. The precursor mass tolerance was 10 ppm and fragment mass tolerance was 10 ppm. The results were filtered by a 1% false discovery rate (FDR) at both PSM and protein levels (Chi et al., 2018; Guangcan et al., 2021).

The determination of 2-PCA and biotin labeling efficiency analyses and other character results were carried out with GraphPad (version 8.0). All statistical tests were analyzed using student t-tests. The GRAVY scores were calculated using the online tool (https://web.expasy.org/protparam/) (Kyte and Doolittle, 1982). The analysis of peptide sequences was conducted by iceLogo (Colaert et al., 2009), and a p-value < 0.05 was used. The bioinformatic analysis was performed using DAVID (version 6.8) with an adjusted p-value < 0.05 (Huang et al., 2007).

All the original proteomics raw data and proteome database result files in this study have been deposited to the iProX Consortium with the subproject ID IPX0003710000.

Chemical derivatization on protein or peptide is an enabling technology for a proteomics study. 2-PCA reported could selectively label the α-amino group on protein/peptide N-terminus over the ε-amino group of lysine residue through an N-terminal amine-specific cyclization reaction (MacDonald et al., 2015). By taking advantage of such reaction selectivity of 2-PCA, we designed a PBC strategy for high-throughput profiling of protein C-termini by bottom-up shotgun proteomics (Figure 1). The extracted whole cell proteome is digested into peptides by LysC protease, which cleaves peptide bond C-terminal to Lys residue. As a result, each non-C-terminal peptide ends with lysine and thus contains two amino groups on the peptide N-terminus and the lysine side chain. In contrast, C-terminal peptides lack lysine and only contain one α-amino group on peptide N-termini. Next, the peptides are tandemly reacted with 2-PCA and NHS-biotin. Ideally, all the α-amino groups on the peptide N-terminal are blocked with 2-PCA and ε-amino groups on the lysine side chain are labeled with biotin. The chemically derivatized peptides are then incubated with streptavidin beads. The original protein N-terminal peptides and internal peptides containing biotin-labeled lysine are all captured by streptavidin beads. Finally, the C-terminal peptides are enriched by filter-aided separation from streptavidin beads and detected by LC-MS/MS. Thus, our PBC strategy not only took advantage of LysC, which has higher digestion efficiency and lower cost than other commonly used proteases (LysargiNase and ArgC) (Giansanti et al., 2016) but also employed only two chemical derivatization steps at the peptide level, in contrast to other reported negative C-terminal peptide enrichment strategies requiring at least three chemical derivation steps at both protein and peptide levels (Schilling et al., 2010; Zhang et al., 2015; Hu et al., 2019). Therefore, the PBC strategy is less labor- and time-consuming for C-terminal peptide enrichment.

In a previous study (MacDonald et al., 2015), only limited number of peptides (20 peptides) with similar sequence composition (only varying at N-terminal amino acid) were used to evaluate the characteristics of the reaction between 2-PCA and the peptide N-terminal α-amino group. Whether other factors could affect the selectivity and efficiency of this reaction needs to be deeply studied by using a large-scale and highly complex peptide sample. In this study, we first used the mass spectrometry–based shotgun proteomics approach to systemically study the characteristics of the selective reaction with the α-amino group on the peptide N-terminus. The highly complex peptides were generated by using LysC digested whole HEK 293 T cell proteome. The evaluation for 2-PCA labeling efficiency was performed in technical triplicates from the same proteome sample. The peptides were reacted with 2-PCA and then detected by LC-MS/MS. Compared to the unlabeled sample, the number of identified protein and peptide decreased in the 2-PCA–labeled sample under same LC-MS/MS conditions. We reasoned that the LC gradient for 2-PCA–labeled peptide sample was unsuitable and needed to be further optimized because the peptide retention time was changed after 2-PCA labeling (Supplementary Figures S1A,B; Supplementary Tables S1, S2). In order to evaluate whether 2-PCA could react with the ε-amino group on lysine, we both set the 2-PCA labeling at the peptide N-terminal and lysine as variable modification for database searching. The bioinformatic analysis results showed 2-PCA dominantly labeled at the peptide N-terminal, while only 2.3% lysines were labeled (Supplementary Figure S1C; Supplementary Table S3). Such results agreed with previously reported result that 2-PCA labeling could selectively react with the α-amino group on the peptide N-terminal. (MacDonald et al., 2015)

Then we evaluated the N-terminal labeling efficiency (the percentage of 2-PCA–labeled peptides in all peptides). Among 4,704 identified non-redundant peptides (peptides of the same sequence with different 2-PCA modification status were considered as two different peptides), 75% peptides (3,529 peptides) were fully labeled with 2-PCA, 8.1% peptides (380 peptides) were partially labeled, and 16.9% peptides (795 peptides) were completely unlabeled (Figure 2A, Supplementary Table S2). For the 795 unlabeled peptides, 25.1% peptides (199 peptides) were N-terminally acetylated and 48.0% (382 peptides) were identified to possess proline at the second amino acid position. We used spectral counting–based quantification to calculate the N-terminal labeling efficiency for the 380 partially 2-PCA–labeled peptides and found 122 peptides (occupied 32.1%) with labeling efficiency higher than 90% and 187 peptides (occupied 49.4%) with labeling efficiency between 50% and 90%. As mentioned before, by excluding the N-acetylated and proline-containing peptides which could not react with 2-PCA in theory, we found more than 90.7% peptides were fully labeled or with labeling efficiency higher than 80%. Such results showed that 2-PCA could react with peptide N-terminal with high efficiency and high selectivity.

We next analyzed factors that influenced the 2-PCA-labeling efficiency, such as peptide length, hydrophilic/hydrophobic properties, and amino acids composition. As shown in Figure 2B, the peptide length significantly affected the 2-PCA–labeling efficiency, with short peptides more preferred. The length of unlabeled peptides was significantly longer than the fully labeled or partially labeled peptides. We also explored the relationship between peptide hydrophilicity (evaluated by GRAVY score) and labeling efficiency (Figure 2C). The data showed the peptides with a higher hydrophilic value achieved higher labeling efficiency. We reasoned the peptides with shorter and higher hydrophilic values could be more easily dissolved in aqueous buffer, which resulted in higher reaction efficiency.

The 2-PCA labeling efficiency was reported to be impaired when the second position of peptide was proline. However, whether different amino acids on other positions could influence the labeling efficiency is still unknown. In this study, we systemically analyzed the influence of first three amino acids in the peptide on reaction efficiency. We found the reaction efficiency was higher when the first amino acid as L/A/E than others (Figures 2D,E, Supplementary Figures S1D,E). Perhaps, less steric hindrance (A/L) and more hydrophilicity (E) of the first amino acid (A) benefited the chemical conversion of 2-PCA coupling. Interestingly, our results showed glycine as the first amino acid supplied less steric hindrance but dramatically inhibited the “reaction efficiency”. Such lower reaction efficiency observed for peptides containing N-terminal glycine may be mainly due to the Thorpe–Ingold effect that led to kinetically slower cyclization (Kaneti et al., 2004). Our results showed proline as the second amino acid could inhibit the reaction, which was agreed with the previous study reported (MacDonald et al., 2015). In addition, we found lysine as the first amino acid could also affect the reaction efficiency. The iceLogo analysis indicated the peptides with first amino acid as A/E were overrepresented in fully labeled peptides (Figure 2F). These results indicate the 2-PCA labeling with high α-amino group labeling efficiency is applicable for the proteomic analysis.

According to the PBC strategy design, the blocking efficiency of the amino group on the peptide N-terminal and the lysine side chain is the key to highly efficient C-terminal peptide enrichment. In order to evaluate the amino group blocking efficiency of the peptide tandem reaction with 2-PCA and biotin, we performed the C-terminal peptide blocking experiments in technical triplicates from same started proteome. We found more than 99.7% α-amino group on the peptide N-terminal and more than 97.7% ε-amino group on K were blocked (Figures 3A,B, Supplementary Table S4). Interestingly, the peptides with second amino acid as proline that could not react with PCA were labeled with biotin, which would help to remove such peptide by streptavidin beads and increase the C-terminal peptide enrichment efficiency. In addition, we analyzed the whole human proteome digested with LysC in silico and found only 695 C-terminal peptides with second amino acid as proline, which account for 6.3% in all 11,106 C-terminal peptides (comprised of 6–50 amino acid residues) (Figure 3C). So, we reasoned the C-terminal peptide with proline as the second position amino acid was minimal, which largely would not influence C-terminome peptide profiling using the PBC strategy. Since the chemical derivatization of peptides will influence the chromatography separation and peptide scoring (reflecting the peptide identification confidence), the retention time and the ion score of the same peptide with and without 2-PCA labeling were systematically compared. We found that the retention time was significantly increased after labeling. The retention time distribution clearly showed that the labeled peptides were dominantly eluted in the later stage of the LC gradient (Figure 3D, Supplementary Figure S1F). We reasoned the 2-PCA molecule contains a pyridine ring structure, which would increase the peptide hydrophobic value after labeling. The peptide matching scores showed no significance difference between peptide with and without 2-PCA labeling (Figure 3E), which indicates the 2-PCA labeling would not influence the C-terminal peptides identification by our PBC strategy.

We used the PBC strategy to enrich C-terminal peptides from cell sample and systematically evaluated its performance. In total, 300 µg proteins were used as the starting materiel, and three technical replicates were performed. Finally, we identified an average of 1,433 C-terminal peptides belonging to 919 proteins in each experiment (Figures 4A,B, Supplementary Table S5). It should be noted that our results showed a lot of C-terminal peptides were not blocked by 2-PCA at N-termini. We systematically evaluated the hydrophilic/hydrophobic values and peptide lengths of the enriched C-terminal peptides. We found the unlabeled C-terminal peptides displaying significantly higher hydrophobic values and peptide lengths than the 2-PCA–labeled or partially labeled C-terminal peptides (Supplementary Figure S2). These results were all consistent with our conclusions that the efficiency of 2-PCA labeling was related to the peptide hydrophilic/hydrophobic value in Figure 2C. A total of 1,975 C-terminal peptides belonging to 1,190 proteins were identified, which achieved 180% more C-terminal peptide identification than the highest reported C-terminal peptide number based on the chemical derivatization–based C-terminomics study. (Hu et al., 2019) We found that 46.9% peptides (926 peptides) were identified in all the triplicates, and 70.8% peptides (1,398 peptides) were identified in at least two replicates (Figure 4C). We also evaluated the quantification linear correlation among the replicates and found the Pearson correlation coefficients of each two pair among the three replicates were all remarkably high (>0.85) (Figure 4D). These results indicated that the PBC strategy provided a high-throughput and reproducibility for C-terminal peptide enrichment. The abundance dynamic range of the enriched C-terminal peptides from the PBC strategy was larger than seven orders of magnitude (Figure 4E). We compared our results with published C-terminome datasets of human cells (Table 1, Supplementary Figure S3A). Our PBC strategy achieved the highest identification number of C-terminal peptides among all reported chemical derivatization–based strategies. In addition, the C-terminal peptide enrichment efficiency of our PBC method is 68% (the percent of C-terminal peptides among all identified peptides), which is highest in the C-terminomics study up to date. Thus, although the enriched C-terminal peptides were not completely labeled with 2-PCA, the PBC provides a novel and simple strategy for high efficiency and reproducibility C-terminal peptides enrichment.

The C-terminomic study could construct the landscape of C-termini in cell with high throughput and efficiency, which could be used to study the C-end rules in cells with high accuracy. In our study, the iceLogo analysis of the C-terminal dataset showed that the P/Q/S is highly enriched in the protein C-terminal sequence (Figure 5A). This result was consistent with our previous study using the LAACter method. (Hu et al., 2019) In addition, the lysine was observed under-represented at −10 to −1 position, which could be explained by that the peptide sample used in the PBC strategy was lysC digestion generated. In order to study whether the C-terminal peptide enrichment bias existed in our PBC strategy, gene ontology enrichment was used to systematically evaluate our C-terminal dataset. We found the enriched C-terminal peptides distributed among different cellular compositions (Figure 5B, Supplementary Table S6), such as cytosol, membrane, ribosome, and nuclear. “C-end rules” was reported as the key factor influencing protein functions in cell (Lin et al., 2018). In addition, we also systemically compared the distribution of first three amino acids in enriched C-terminal peptide and theoretical C-terminal peptide. The theoretical C-terminal peptides were generated from in silico digestion of whole human proteome by using LysC (Figure 3C). The results showed the similar distribution of first three N-terminal amino acids between enriched and theoretical C-terminal peptide (Supplementary Figures S3B–D), which also indicated the PBC strategy could enrich C-terminal peptides without bias.

Our C-terminomic study enabled us to examine C-terminal truncated proteins (Supplementary Table S7). Totally, 217 C-terminal truncation peptides corresponding to 50 proteins were identified. For example, six truncation peptides were identified from HSP90AB1 (Table 2). The annotated MS/MS spectra of these six truncation peptides are showed in Supplementary Figure S3E. In addition, we also found the truncated histone protein H4 in our data (Table 2). The annotated MS/MS spectra of the truncation peptides of H4 are showed in Supplementary Figures S4A,B. In order to further analyze the cellular composition and molecular functions of these 50 truncated proteins, the gene ontology (GO) and KEGG pathway enrichment were performed (Figure 5C). We found these proteins significantly enriched in extracellular exosome, vesicle, and ribosome, suggesting these proteins mainly partook in the biological processing of translation and translational initiation. The ribosome pathway was also significantly enriched through the KEGG pathway analysis. We inferred the truncated proteins were mainly generated from two origins. One was from immature proteins due to the incomplete translation step in ribosome; another might come from the proteolytic fragments by exopeptidases and metallocarboxypeptidase enzyme in exosome or vesicle.

PTMs on protein C-terminal region play important roles in protein functions (Marino et al., 2015). However, the reported PTMs studies on C-terminal were limited due to the lack of in-depth and large-scale C-terminome data. In this study, we used the open-search method to analyze the PTMs distribution in our C-terminomic dataset. All the raw data were reanalyzed by using pFind 3.1 search engine through open-search searching mode. Totally 44,178 PSMs were identified and more than 30% PSM were identified with containing PTMs (Figure 5D, Supplementary Table S8). The PTM types and their frequency distributions is showed in Figure 5E. The dominant PTM was 2-PCA modification on peptide N-terminus, which was introduced by chemical divinization in vitro at the peptide level. Deamidation on asparagine was frequently occurred in our data. Deamidation was reported could occur spontaneously on proteins both in vivo and in vitro (Hao et al., 2011; Brown et al., 2017). In vitro, the deamidation rate of protein or peptides closely related with pH and temperature of sample preparation buffer. In our PBC method, the peptides were reacted with 2-PCA and biotin all performed at pH 8.5 under 37°C. We reasoned that such high frequency of deamidation were mainly artefacts introduced in sample preparation. Amino acid substitutions were reported to alter physiological properties of protein (Huang and Gromiha, 2010; Gromiha et al., 2019), such as the enzymatic stability and protein folding rate. We found the acetylation on lysine and amino acid substitution highly occurred in the protein C-terminal region (Figures 5E,F). Interestingly, we also found many lysines at protein C-termini with acetylation, which was also reported benefit for protein stability (Wang et al., 2017). In addition, we also found 176 PSMs were identified as the PTM-labeled histone peptides (Supplementary Figure S4C). The main PTM types were deamination, amino acid substitution, and dehydration. Thus, the PBC strategy provides a high-throughput analysis of the PTMs on the protein C-terminal region.