Multifunctional Reagents for Quantitative Proteome-Wide Analysis of Protein Modification in Human Cells and Dynamic Profiling of Protein Lipidation During Vertebrate Development

Novel multifunctional reagents were applied in combination with a lipid probe for affinity enrichment of myristoylated proteins and direct detection of lipid-modified tryptic peptides by mass spectrometry. This method enables high-confidence identification of the myristoylated proteome on an unprecedented scale in cell culture, and allowed the first quantitative analysis of dynamic changes in protein lipidation during vertebrate embryonic development.

Abstract: Novel multifunctional reagents were applied in combination with al ipid probe for affinity enrichment of myristoylated proteins and direct detection of lipid-modified tryptic peptides by mass spectrometry.T his method enables high-confidence identification of the myristoylated proteome on an unprecedented scale in cell culture,and allowed the first quantitative analysis of dynamic changes in protein lipidation during vertebrate embryonic development.
Although advancements in protein mass spectrometry (MS) have enabled global profiling of numerous post-translational modifications (PTMs), confident high-throughput identification of lipidated proteins remains problematic.L ipidation adversely affects solubility,c hromatographic properties,a nd ionization of modified peptides,a nd routine proteome-wide detection by MS and assignment of the modification sites remains an unsolved challenge. [1] Furthermore,l ipidated proteins are often present at relatively low abundance in cells,a nd lipid-specific enrichment is required to reduce sample complexity and improve discovery.Enrichment strategies can exploit natural features of the lipid, for example,the use of anti-farnesylpeptidea ntibodies; [2] indirect analysis through acyl-biotin exchange for S-acylation; [3] or more generalizable introduction of af unctional handle through metabolic incorporation of abioorthogonal chemical tag. [4] Co-translational myristoylation is as pecific class of protein lipidation whereby N-myristoyl transferases (NMTs) catalyze the covalent and irreversible addition of atetradecanoyl chain at an N-terminal glycine (Gly), revealed by the action of methionine aminopeptidases during protein synthesis at ribosomes. [5] Metabolic tagging has become the technique of choice for whole-proteome profiling of the cellular myristoylated proteome.A sp reviously demonstrated, [6] metabolic tagging of ac ellular proteome with tetradec-13-ynoic acid (YnMyr) and subsequent elaboration through ligation with secondary reporters (capture reagents) enables visualization, enrichment, and MS-based identification of myristoylated proteins.This chemoproteomic methodology delivers greatly improved processing time and sensitivity over traditional radioisotope or direct MS approaches, but there are also limitations.F irstly,a sw ith any affinity enrichment technique,unspecific protein background is often observed when coupled to high-sensitivity MS detection. Secondly,p rotein fatty acyl transferases that utilize longer acyl-CoAs as substrates may also use YnMyr either natively or following chain extension, which may lead to false positive identifications. [7] Finally,r obust and general methods for confirmation of the lipidation site have yet to be established.
To overcome these limitations,w er eport ap ortfolio of reagents for the enrichment and visualization of myristoylated proteomes and direct MS detection of lipid modification of protein Ntermini. Thedesign of these reagents is based on compound 1 (Figure 1a), arobust tool for enrichment and ingel fluorescence (igFl) analysis of metabolically tagged proteomes. [8] To make 1 amenable to the detection of lipidmodified peptides,weenvisioned the introduction of alinker ( Figure S1 in the Supporting Information) to facilitate cleavage as an integral part of ap roteomic workflow.I n astandard tagging experiment (Figure 1b), cells or organisms are treated with YnMyr or am yristic acid (Myr) control to metabolically tag proteins via activation by cellular acyl-CoA synthase followed by transfer by NMTs.T he cells/organisms are harvested and the metabolically tagged proteins ligated to ac apture reagent through copper-catalyzed alkyne azide cycloaddition (CuAAC) and affinity-enriched. Thee nriched proteomes are visualized by igFl to assess metabolic tagging and enrichment efficiency, or analyzed through on-bead proteolytic digestion, MS analysis,and software-aided protein identification.
We synthesized al ibrary of reagents bearing enzymatically cleavable linkers ( Figure 1a,T able S1 in the Supporting Information). Reagents 2-5 contain arginine (Arg) residues and reagent 6 contains al ysine (Lys) residue to enable proteolysis by trypsin and simultaneous enhancement of the ionization properties of lipidated peptides.R eagent 7 in turn incorporates phenylalanine (Phe) and reagents 8-9 incorporate aspartic acid (Asp) to facilitate cleavage by chymotrypsin and AspN,r espectively.T hese proteases are less commonly used in proteomic workflows,h owever they may offer complementary protein sequence coverage,t hus improving overall modified peptide discovery.S imilarly to 1,c ompounds 2 and 6 were equipped with the TAMRA fluorophore (R 1 ) for fast visualization by igFl. All other reagents lack the fluorophore to reduce bulk and possible steric hindrance,s ince biotin (contained in R 2 )i ss ufficient for visualization (e.g., by using streptavidin conjugated to horseradish peroxidase (streptavidin-HRP)). In addition, 5 was equipped with a-azido-arginine to further minimize the size of the cleaved product. Finally,w ea imed to enhance the ionization properties of the lipidated peptides still further by incorporating trimethyl (4,7,8)o rd imethyl (9)l ysine residues.
Taking advantage of both the TAMRA and biotin moieties,w ef irst investigated the efficiency of 2-9 for the capture and enrichment of myristoylated proteins.H EK293 cells were treated with YnMyr or Myr for 24 hfollowed by lysis,CuAAC with 1-9,a nd affinity enrichment on streptavidin-conjugated beads.C aptured proteins were eluted by thermal denaturation, resolved by SDS-PAGE, and visualized by igFl or Western Blotting with streptavidin-HRP ( Figure S2a and b). We were pleased to observe that the majority of the reagents

Angewandte
Chemie showed comparable enrichment efficiencyto1and that there was negligible background signal, as shown by the Myr control. However,t he performance of 8 and 9 was clearly diminished, which may indicate lower efficiency of CuAAC, and these reagents were excluded from further analysis.W e also observed that despite their bulkier structures, 2 and 6 performed as robustly as 1,w hich was demonstrated by two alternative readouts.
We then evaluated the capacity of these reagents to identify proteins with aconsensus myristoylation sequence (MG-proteins) and, most importantly,N -terminally myristoylated peptides.S amples were prepared as described above,s ubjected to CuAACw ith 2-7,a ffinity-enriched, and digested with either trypsin (2-6)o rc hymotrypsin (7), followed by MS analysis.
Five reagents (2-4 and 6-7)i dentified 38 lipid-modified peptides with high (> 99 %) confidence ( Figure S2c and Table S2 in the Supporting Information). Nearly 90 %ofthese peptides were discovered with 2,t hus demonstrating its high effectiveness despite the presence of the TAMRA dye and lower predicted net charge compared to 3 and 4,r espectively. Reagent 2 was also superior to 6,t hus suggesting that the proteolysis step and/or the ionization properties of the Arg-containing conjugates were improved compared to Lyssince CuAACand enrichment were similar ( Figure S2). These steps also seemed efficient for 5,h owever,n ol ipidated peptides were detected;wespeculate that proximity of the triazole to the cleavage site reduces the affinity for trypsin. Regardless of the reagent type,a pproximately 30 %ofall of the proteins identified carried an N-terminal MG motif (Table S3), which is more than four-fold higher than in analysis of the native proteome,w here the abundance of MG proteins is approximately 7%. [9] The remaining 70 %o ft he detected proteins represent tagging at sites other than the Nterminus as aresult of the promiscuity of lipid transferases and proteins binding to beads and/or tagged proteins,t hus highlighting the utility of methodologies that allow either quantification of NMT inhibition (if reliable inhibitors are available) [6b] or direct detection of lipidated peptides to improve modified protein ID confidence.
Having selected the best performing capture reagent (2), we next aimed to maximize the discovery of myristoylated peptides and the corresponding endogenous protein substrates of human NMT.L arger-scale chemical proteomics experiments were undertaken with three commonly used cell lines:H eLa, MCF7, and HEK293. Twoc omplementary software packages were employed for MS data analysis, with primary identification of lipid-modified peptides per-formed with PEAKS7, and MaxQuant1.5 used as asecondary platform. [10] Thelatter was also used to quantify protein levels between YnMyr-a nd Myr-treated samples,with Myr serving as ac ontrol for both nonspecific protein identification and reliability of the modified peptide identifications.8 1l ipidmodified peptides were detected in PEAKS7 analysis,o f which approximately 40 %were common to all three cell lines (Figure 2a). MaxQuant analysis delivered 62 lipidated peptides,f our of which were not observed in PEAKS7. [11] Importantly,e ach search engine returned only one lipidated peptide common to both YnMyr and Myr control samples, which was assigned as afalse positive and excluded from the list of myristoylated peptides (Table S4).
Based on statistically significant enrichment of MG proteins pulled down from YnMyr-treated samples compared to Myr controls,w ei dentified 170, 163, and 145 candidate myristoylated proteins for HEK293, HeLa, and MCF7 cells, respectively (Table S5). However, ap revious comprehensive study of quantitative dose response to NMT inhibition showed that enrichment is only asimple and indirect measure of N-terminal myristoylation since only 70 out of 169 enriched MG HeLa proteins robustly responded to NMT inhibition. [6b] In agreement with this alternative approach, our dataset . .
shows direct evidence for 87 of the candidate proteins in the three cell lines.H erein, we provide direct proof for myristoylation in Hela cells for 47 previously reported proteins,as well as significantly increased coverage of since direct MS/MS evidence for myristoylated peptides was detected for 69 enriched MG targets in HeLa, 65 in HEK293, and 50 in MCF7 cells (Table S5). This represents the largest directly validated database reported to date for the N-myristoylated human proteome.
We next applied 2 to profile the myristoylated proteome for the first time in ad eveloping organism. Following evaluation of YnMyr tagging specificity ( Figure S3), Danio rerio (zebrafish) embryos were metabolically tagged in specific time windows (0-24, 48-72, and 96-120 hours postfertilization (hpf)) and processed according to the standard workflow.C hemical proteomics experiments revealed the scope of myristoylation in developing zebrafish;7 2p otential myristoylation targets were identified based on statistical analyses of YnMyr/Myr enrichment and 56 lipid-modified peptides were discovered (Tables S6 and S7). Notably,int he absence of validated NMT inhibitors for zebrafish, the methodology presented herein is currently the only approach for confident assignment of myristoylated proteins in this animal model.
We also investigated the dynamic nature of myristoylation through pulsed YnMyr tagging as described above,c oupled with visualization by igFl and quantification through triplex dimethyl labeling [12] (Figure 2b and c, respectively). IgFl analysis indicates development-stage-specific D. rerio protein myristoylation profiles that reflect differential protein expression during embryonic development. Further MS-based analysis allowed the quantification of myristoylation dynamics for 54 zebrafish proteins (Figure 2c), which revealed that myristoylation is most prominent in early development. Several proteins myristoylated within 24 hpf (e.g., prkaca and gnai families) participate in progesterone-mediated maturation, the hedgehog and wnt signaling pathways, melanogenesis,a nd meiosis,a ll of which are critical in early development (Table S8).
In summary,t he multifunctional capture reagents reported herein enable robust identification of metabolically-tagged myristoylated proteomes,w ith unprecedented confidence resulting from the combination of chemicalprobe-based enrichment and release and direct detection of lipid-modified peptides by MS.P reviously reported reagents [13] typically required an extra proteolytic step and their capacity to enable the detection of lipidated peptides has not been demonstrated. Herein, we report the largest database (87 counts) of experimentally validated human proteins that are myristoylated at an endogenous level in living cells. We also present the first profile of myristoylation in al iving multicellular organism and the confident identification of over 50 novel targets.Importantly,this work provides the first example of the analysis of any protein lipidation event during vertebrate development. We envision that our reagents will find application in the quantitative and dynamic analysis of other PTMs,a nd in related workflows such as activity-based protein profiling,both in cells and in multicellular organisms.