Introduction

Influenza virus is a major cause of respiratory disease, causing > 300,000 deaths in a typical year, millions to tens of millions of deaths in pandemic years, and extensive loss of economic activity [1]. An obligate step in viral infection is fusion of the viral membrane with the endosomal membrane, which allows for release of the viral RNA-polymerase complex (RNPs) into the host cytosol. This fusion is driven by an acid-induced conformational change in the viral coat protein hemagglutinin (HA). Because fusion is essential for infection, fusion inhibition has been an intensively pursued approach to block influenza infection using small molecules. Although recent studies have elucidated the biochemical mechanism of HA-mediated fusion [2, 3], the sequence dependence and particularly intracellular factors that govern HA fusion competence are still unclear. Largely, this is due to a lack of tools to readily characterize HA conformational changes inside the cell.

Mature HA (cleaved to two domains: HA1 and HA2) mediates viral endocytosis into the cell through engagement of the sialic acid receptor by the HA1 domain. Under the acidic conditions of the early endosome, HA undergoes a global conformational change [4], including extensive remodeling of both HA1 and HA2 domains. As a result, an N-terminal region of HA2, termed the fusion peptide, is exposed. The fusion peptide embeds into the endosomal membrane, promoting fusion with the viral membrane and subsequent release of the viral genome into the cytosol. The most dramatic structural change takes place over the HA2 region in the stalk domain near the C-terminus; acidification and conversion to the post-fusion state lead to a substantial increase in solvent exposure in this region. Given the large change in solvent engagement over the entirety of HA, we hypothesized that profiling reactions would be sensitive to the fusion event.

Hence, we set out to develop a label-free, targeted approach to interrogate HA fusion competence. The gold standard for profiling protein conformation is hydrogen-deuterium exchange (HDX) [5], which has been successfully applied to fully characterize solvent exposure throughout HA in both pre- and post-fusion states [2]. HDX has several limitations, however, including incompatibility with biologically complex matrices, the requirement for rapid, high-resolution separation, and usually high nanogram to microgram levels of sample. To overcome these challenges, oxidative and electrophilic labeling approaches have been taken to footprint water accessibility on protein surfaces, both in buffer and in complex media [6,7,8,9,10]. Among oxidative approaches, SPROX employs peroxide to oxidize methionine residues [9, 11], while FPOP generates oxidizing radicals through photolysis [12,13,14,15]. Electrophilic approaches include targeting lysine with NHS esters [16, 17], oxyimidates, and thioimidates [18,19,20], and targeting glutamate and aspartate with glycine ethyl ester [21,22,23]. Thioimidate reagents are notable in that they are cell-penetrable, selectively label an abundant amino acid (lysine), provide a homogenous reaction product allowing straightforward spectral assignment, promote ionization during electrospray, and do not modify residue charge, which can itself alter protein conformation [24].

Previously, thioimidates have been applied to discover changes in intracellular protein conformation using data-dependent analysis, yielding similar performance to SPROX. However, when targeting specific modifications, greater sensitivity and robustness can be achieved by using targeted, data-independent methods. Although being required to assay for a specific peptide, as opposed to the best performing peptides in an entire protein, sharply constrains target peptide choice, multiple reaction monitoring (MRM) and parallel reaction monitoring (PRM) have proven effective for quantifying PTMs [25, 26] and adduct sites [27]. Hence, we considered that electrophilic profiling by S-methylthioacetimidate (SMTA), followed by reaction monitoring, might be an effective strategy to identify pre- and post-fusion HA conformational states. Herein, we implement a targeted MS/MS technique of pseudo-parallel reaction monitoring to quantify labeling of the pre- and post-fusion HAs. We find that SMTA efficiently labels HA, distinguishing the pre- and post-fusion states, and that we can measure this difference using reaction monitoring. This will enable the further development of assays to assess HA viral fusion competence.

Materials and methods

Materials

Sodium phosphate, sodium chloride, HEPES, Tris, MS-grade methanol, MS-grade acetonitrile, and chloroform were obtained from Fisher (Pittsburgh, PA, USA). Tissue-culture grade dimethyl sulfoxide was obtained from Corning (Corning, NY, USA). Urea, formic acid, TCEP, and iodoacetamide were obtained from Acros (a division of Fisher). Sodium citrate was obtained from VWR (Radnor, PA, USA). Imidazole was obtained from Sigma (Burlington, VA, USA). Phosphate-buffered saline (PBS; 10 mM phosphate, 150 mM sodium, pH 7.4) was obtained from GE Healthcare (Chicago, IL, USA).

Preparation and purification of recombinant hemagglutinins

HEK293T was obtained from ATCC (Manassas, VA, USA). The cells were cultured in complete media containing DMEM (Corning) supplemented with 10% FBS (Seradigm, a division of VWR), 1% L-glutamine (Corning), and 1% penicillin streptomycin (Corning). HA was prepared as previously described [28]. Briefly, PR8HA-6xHistag (strain A/Puerto Rico/8/1934 H1N1) expressing plasmid was transfected to HEK293T cells by the PEI transfection method. After 72 h, the cultured media were cleared by low speed centrifugation (3500 rpm, 20 min, 4 °C) and the HA was purified by His6 Ni Superflow resin (Clontech) overnight at 4 °C. The resin media were passed through 10 mL polypropylene columns (Qiagen, Germantown, MD, USA). The retained resin was washed 4 times with wash buffer (50 mM NaH2PO4, 30 mM NaCl, 20 mM imidazole, pH 8.0). The protein was eluted with elution buffer (50 mM NaH2PO4, 30 mM NaCl, 250 mM imidazole, pH 8.0). The eluate was concentrated using an Amicon 10 kDa MWCO spin column (Merck Millipore, Burlington, MA, USA) and dialyzed 4 times with PBS and quantified by Thermo Nanodrop 3.0 (Thermo, Waltham, MA, USA). Purity was evaluated by Silver Stain [29].

Acid-induced activation of recombinant HA

PR8 was chosen as the HA wild-type model for the study. Purified recombinant PR8 HA was diluted in PBS buffer. To pre-cleave the precursor HA0 to HA1 and HA2, TPCK-treated trypsin was added to the protein at a ratio of 10:1 (protein-to-protease) and incubated with 600 rpm shaking at 37 °C for 5 min. Pre-cleavage was quenched with 1 mM PMSF (Sigma), which remains in the sample solution through later steps. The pH of PR8 HA solution was brought to pH 4.9 by adding citrate buffer at pH 3.0 to a final citrate concentration of 7.0 mM. The acidified HA was incubated at ambient temperature for 3 h, then brought to pH 8.0 by 20 mM phosphate pH 11.00. The neutral control sample was treated in the same manner with citrate pH 7.4 to a final citrate concentration of 7.0 mM. Both acidified and neutral samples were brought to equal volumes by PBS. The fusion inhibitor [30] arbidol-HCl (Fisher) was added as indicated in DMSO following pre-cleavage at a concentration of 400 μM and incubated for 30 min at ambient temperature prior to acidification. Control samples received 4% DMSO vehicle alone.

Trypsin susceptibility of acid-induced HA

The treated HA above was incubated at 37 °C for 3 h to deactivate PMSF, then was digested with TPCK-trypsin (Sigma) at 5:1 (protein:protease ratio) for 30 min at 37 °C, shaken at 600 rpm [31]. Immediately after being digested, the samples were denatured and reduced in 1X Laemmli buffer (12% SDS (w/v), 0.06% bromophenol blue (w/v), 47% glycerol (v/v), and 60 mM Tris pH 6.8) and 20 mM DTT. The denatured protein was boiled at 100 °C for 5 min. The digest was separated on 12% reducing SDS-PAGE gel for silver staining.

SMTA labeling of recombinant HA

SMTA was prepared as the hydrogen iodide salt from thioacetamide (Sigma) and methyl iodide (Sigma), as previously described [18], and aliquots stored under Ar at − 80 °C. Aliquots were freshly dissolved prior to labeling in 25 mM HEPES, 50 mM NaCl pH 7.4 at 200 mM SMTA, which was then neutralized with sodium hydroxide. SMTA solution was added to PR8 HA samples, mixed well and let sit for 30 min at ambient temperature. The reaction was quenched by methanol-chloroform precipitation, by adding 4 equivalents methanol, 1 equivalent chloroform, and 3 equivalents water with vortexing after each addition, followed by sedimentation of protein to the interphase at 10,000×g for 5 min. The aqueous layer was removed and the protein pellet washed well with methanol and air-dried.

Sample preparation for mass spectrometry

Air-dried protein pellets were solubilized in freshly prepared 9 M urea in 50 mM Tris pH 8.0. Ten millimolar aqueous TCEP was added to reduce the disulfide bonds for 30 min in the dark at room temperature. Five millimolar aqueous iodoacetamide was added to alkylate the sulfhydryl groups for 15 min in the dark at room temperature. The solutions were diluted with 50 mM Tris pH 8.0 to 2 M urea and digested with sequencing grade porcine trypsin (Thermo) or Glu-C (Thermo) at a ratio of 20:1 protein-to-protease for 18 to 24 h at 37 °C, shaken at 600 rpm. After digestion, the samples were acidified with 5% formic acid and debris eliminated at 21,100×g for 30 min.

Preparation of complex mixtures containing digested HA

HEK293T cells were harvested by scrapping in PBS and lysed in RIPA buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate, 0.1% SDS) or by freeze-thaw in PBS. The lysate concentration was quantified by Bradford assay (Bio-rad, Hercules, CA, USA) with the measurement by Varian Cary 60 UV/Vis spectrophotometer (Agilent, Santa Clara, CA, USA). The protein was precipitated by methanol-chloroform precipitation and prepared as digested peptides similar to the above description. Digested peptides from either acidified or neutral as described above was spiked in lysate at 10% (w/w). One microgram lysate was analyzed by pseudo-PRM in the LTQ Orbitrap Velos Pro (Thermo).

Data-dependent analysis of SMTA-labeled PR8 HA

Digest was injected from the autosampler of an Easy-nLC 1000 to the trapping column containing a Kasil frit and 2-cm 5 μm C18 resin (Phenomenex). High salt concentration was removed by washing with buffer A through the waste line. The analytical column is a pulled 5 um tip packed with 15- to 20-cm 3 μm C18 resin (Phenomenex, Torrance, CA, USA). The peptides were injected directly onto the trapping column at a flow rate of 5.0 μL/min and separated on the analytical column at a flow rate of 500 nL/min. Buffer A was 5% acetonitrile, 0.1% formic acid, and buffer B was 80% acetonitrile, 0.1% formic acid. The gradient conditions were as follows: 0–5 min: 1–7%B; 5–95 min: 7–55%B; 95–115 min: 55–100%B; 115–120 min: 100%B; 120–125 min: 100–1%B; 125–145 min: 1%B. The eluate was sprayed at 3.0 kV and the ion transfer tube maintained at 275 °C. The mass spectrometer (LTQ Orbitrap Velos Pro) precursor scan was acquired with the AGC set to 105 and 50 ms maximum injection time. In all data acquisition strategies, MS2 scans were acquired by CID at a normalized collision energy set to 35 kV.

Targeted library generation

Data-dependent acquisition of MS/MS spectra with the linear ion trap in the LTQ Orbitrap Velos Pro were performed with the following settings: MS/MS on the 10 most intense ions per precursor scan, dynamic exclusion repeat count, 1, repeat duration, 30 s; exclusion list size, 500; and exclusion duration, 5 s. The raw data was extracted using Raw Converter to MS1 and MS2 spectral files. These files were searched against a database of the PR8 HA sequence and its reversed sequence by the ProLucid algorithm in the IP2 software suite (Integrated Proteomics, San Diego, CA) [32]. The database was supplemented with forward and reverse (decoy) sequences for 400 common contaminants (e.g., keratins, immunoglobulins, albumen). The precursor mass tolerance was set to 3000.0 m-amu, and the fragment mass tolerance was set to 600.0 ppm. The search space included up to 3 missed cleaved tryptic peptides. Carbamidomethylation (+ 57.02146) of cysteine was considered as a static modification. Amidination of lysine (+ 41.02655) was permitted as a differential modification. Peptide candidates were filtered using DTASelect for less than 1% false positive peptides [33].

Pseudo-PRM analysis

From data-dependent analysis, the precursors were chosen such that their spectral counts are > 3 and they are fully tryptic peptides (Table 1). From each precursor, the 3 most intense fragments were chosen for monitoring. The peptides were injected directly to the trapping column at a flow rate of 5.0 μL/min and separated on the analytical column at a flow rate of 350 nL/min. Buffer A was 5% acetonitrile, 0.1% formic acid, and buffer B was 80% acetonitrile, 0.1% formic acid. The gradient conditions were as follows: 0–1 min: 1–7%B; 1–59 min: 7–35%B; 59–64 min: 35–100%B; 64–69 min: 100%B; 69–74 min: 100–1%B; 74–94 min: 1%B. The monitoring time for each peptide was 10 min. In the linear ion trap, the precursor scan was acquired at normal resolution, with the AGC set to 104 and 50 ms maximum injection time and isolation width of 1.0 m/z. In the FT Orbitrap, the precursor scan was acquired at a resolution of 15,000, with the AGC set to 5 × 104, isolation width of 1.0 m/z, and 25 ms maximum injection time. The three most intense product ions were chosen for quantification. The integration of each transition was measured from the extracted ion chromatogram from Xcalibur and summed to obtain the total integration of each peptide. The m/z windows used to extract the transitions for quantification in the linear ion trap and Orbitrap are 4 m/z and 0.2 m/z, respectively. For each peptide, the evaluation of differential amidination between acidified and neutral conditions of HA was calculated as integration of modified peptide/integration of unmodified peptide.

Table 1 Peptide sequences for reaction monitoring
Full size table

Multiple reaction monitoring

MRM was performed on a Thermo TSQ Vantage with Easy-nLC II. The digest was injected (Easy-nLC II) onto a C18 column and washed using a similar trapping set-up as described above. Buffer A was 0.1% formic acid, and buffer B was 100% acetonitrile, 0.1% formic acid. The peptides were injected onto the trapping column at a flow rate of 2.50 μL/min and separated on the analytical column at a flow rate of 230 nL/min. The gradient conditions were as follows: 0–1 min: 0–7%B, 1–46 min: 7–33%B, 46–47 min: 33–80%B, 47–77 min: 80–90%B. Trapping and analytical columns were equilibrated before every analysis. The elution time window set for each peptide was 6 min. The spray voltage was set at 1.8 kV and the discharge current was 4.0 μA for all 3 quadrupoles. Collision energies were calculated with the following formula: CE = 0.03 × (precursor m/z) + 2.905 for 2+ charge ions and CE = 0.038 × (precursor m/z) + 2.281 for 3+ charge ions. The Q1 and Q3 peak width (FWHM) was set to 0.70 and 0.80, respectively. The cycle time was set to 800 ms. Each peptide was integrated directly from its extracted ion chromatogram on Xcalibur.

Statistical analysis

Statistical analysis was performed in Excel. To compare the amidination profile between acidified and neutral conditions, we employed Student’s t test with the p value set to 0.05. For the comparison of multiple treatments, one-way ANOVA was employed with the alpha factor set to 0.05, followed by Tukey’s test. The sample size is indicated in the figure captions for each analysis. All data are expressed as means ± standard deviations.

Results and discussion

Generation of a library of amidinated HA peptides

We chose PR8 HA (strain A/Puerto Rico/8/1934 H1N1) as our nominal “wild-type” HA, as this subtype’s structure and fusion has been widely studied [34]. To increase recombinant PR8 yield by allowing secretion, the C-terminal trimerization and membrane pass regions were replaced with a bacterial foldon trimerization domain [35, 36]. This form has been validated to form PR8 HA trimers that still undergo acid-dependent fusion, as demonstrated by protease susceptibility following acidic incubation [31]. We optimized cleavage of full-length HA (HA0) to mature HA1/HA2 by brief TPCK trypsinolysis (see Electronic Supplementary Material (ESM) Fig. S1A). The two bands for HA2 represent glycoforms, as confirmed by coalescence upon PNGase treatment (data not shown). Furthermore, citrate acidification of mature HA1/HA2 renders the protein more sensitive to further tryptic degradation (Fig. 1a; lanes 3, 4) as opposed to the mature protein in PBS (lanes 1, 2) or treated with an isotonic neutral citrate buffer (Fig. 1a; lanes 5, 6). The tryptic sensitivity validates that acidification promotes the conformational change to the post-fusion state.

Fig. 1
figure 1

SMTA labeling of pre- and post-fusion rHA. Silver stain of SDS-PAGE separated recombinant PR8 HA (rHA), demonstrating trypsin susceptibility following acid-induced fusion. rHA subjected to brief trypsin precleavage (1:10 w/w TPCK-treated trypsin in PBS for 5 min., 37 °C, 600 rpm) was incubated in either acidic or neutral citrate buffer. Afterwards, all samples were brought to pH 8.0, followed by trypsin digestion at 37 °C for 30 min. as indicated. Dagger symbol indicates small degradation products. Asterisk indicates residual protein from the media purification of HA, presumably bovine serum albumen. Flow chart for HA pre-cleavage, fusion, neutralization, and SMTA labeling. Structures are of H3N2 from PDB 1HA0 [37], 3HMG [38], and 1HTM [39] and visualized in NGL viewer [40]. Reaction mechanism for SMTA modification of the lysine amine. At neutral or greater pH, deamination to the imide is disfavored, driving amidination as shown [24]

Full size image

PR8 HA contains 37 lysine residues. To determine their amidination profile, we prepared PR8 HA in both the pre- and post-fusion states and treated each with 20 mM SMTA for 30 min at pH 8.0 (Fig. 1b). SMTA selectively amidinates lysine residues as this pH [24] (Fig. 1c). The reaction was quenched by chloroform-methanol precipitation, and the protein reduced, alkylated, and digested by either Glu-C or trypsin. Peptide sequences were determined by data-dependent LC-MS/MS. Qualitatively similar results in peptide identifications were observed for both pre- and post-fusion HA; hence, the following summarizes what was observed for the post-fusion HA. With trypsin modification, we observed 75–85% coverage of the protein (theoretical digest yields 86% coverage by peptides with between 5 and 30 residues long), with 0 lysines observed only with the modification, 9 lysines observed solely without the modification, and 25 lysines observed both with and without amidination (ESM Fig. S1B). By contrast, Glu-C digestion yielded 60–65% coverage of the protein (theoretical digest yields 68% coverage by peptides between 5 and 30 residues long), with 0 lysines observed solely with the modification, 14 lysines observed solely without the modification, and 12 lysines observed both with and without amidination (ESM Fig. S1B). These results are consistent with previous findings that Glu-C yields peptides with poorer chromatographic profile, greater variability in digestion, and up to a threefold drop in unique peptide identifications [41]. Furthermore, the high density of Glu sites in the HA2 region inherently limits the coverage that can be obtained with this protease. Hence, we moved forward with trypsin, despite its inability to cleave amidinated lysine [19]. Interestingly, we found lysine amidination to be distributed throughout the protein sequence, indicating that the PR8 HA is accessible to SMTA across all domains (ESM Fig. S1C). Labeling does not appear to be saturating, as although some lysines were never observed to be modified, and other lysines are identified from peptide-spectral matches in either the modified or unmodified state, there are no lysines that we only observed to be amidinated.

Pseudo-PRM quantification of fusion-dependent HA amidination

To identify peptides that could serve as markers of lysine amidination, we evaluated the tryptic library (3 runs) for peptides that met the following criteria [42]: (1) robustness of identification, with ≥ 3 spectral counts in each run; (2) fully tryptic; (3) no missed cleavage except for at the amidination site; (4) containing lysines that were identified as both modified and unmodified, indicating that there is a dynamic range of modification; (5) lacking methionine; (6) no K/R within 2 amino acids of the cleavage site (except for − 1 of course); and (7) charge state of either 2 or 3. We identified 4 amidinated peptides that were worth further consideration based on these criteria. The three most intense transitions from the library were used for each (Table 1, Fig. 2a, b and ESM Fig. S2A, B). For normalization, we considered three internal reference peptides [25]: the equivalent unlabeled peptide with missed cleavage at the central lysine, the predicted cleavage product for the unlabeled peptide with the lysine now available for tryptic cleavage, and a peptide elsewhere in the protein generated through arginine cleavage (and hence presumably unaffected by SMTA labeling). Surprisingly, despite the inherently increased variance at missed cleavages, normalizing amidinated peptides by the fully cleaved, non-amidinated peptides yielded higher CV values (ESM Fig. S2C). Therefore, we quantified the modification ratio at each lysine by taking the ratio of the intensity of modified peptide over the intensity of unmodified peptide. SMTA concentration was set to 20 mM, the minimum value that maximized this modification ratio for any of the analyzed peptides (ESM Fig. S2D). Because we are using an ion trap to characterize three transitions in each peptide, as opposed to the Orbitrap traditionally used for parallel reaction monitoring, we describe this approach as pseudo-PRM, analogous to pseudo-SRM, when single transitions are monitored using an ion trap [25, 43]. Using this approach, we found that amidination at Peptide 1 (K62) on PR8 HA is significantly higher in the post-fusion (acid-induced) conformational state (Fig. 2c). Peptide 1 is located in the fusion (F′) subdomain of HA1, far from the receptor binding site but in a region that undergoes substantial variation with fusion and that stabilizes HA2 in the pre-fusion state [2, 44]. More modest effects were observed at Peptide 2 (K170; increased amidination) and Peptide 4 (K459; decreased amidination), with no change observed at Peptide 3 (K394) (Fig. 2c). The change in amidination in part contrasts with a previous HDX study on viral H3 hemagglutinin [2], wherein Peptide 4 was found to become far more solvent accessible upon fusion. The same study found a modest increase in solvent accessibility post-fusion for the Peptide 1 region, with little change in the vicinity of the lysines in Peptides 3 and 4. These differences could reflect differences between serotypes, or the differences between D2O accessibility and local lysine reactivity. Also, it is worthwhile to note that these peptides are chosen as reporters of the conformational change that are best suited to reaction monitoring, rather than being regions that themselves drive the fusion-associated conformational change.

Fig. 2
figure 2

Differential amidination of Peptides 1–4 by SMTA. CID fragmentation spectra collected on the LTQ for unmodified and amidinated Peptide 1 during DDA analysis of acidified, SMTA-labeled rHA digests prepared as in Fig. 1b. Identified b/y ions are indicated above and below the peptide sequence. Scan ranges were determined by the instrument. The transitions chosen for MRM, pseudo PRM and PRM analysis are indicated by boxes. Extracted ion chromatograms for unmodified and amidinated Peptide 1 analyzed during pseudo PRM in a LTQ Velos ion trap, summed from transitions indicated in Fig. 2a. Pseudo-PRM quantification of the amidination ratio for Peptides 1–4 in digests of pre- or post-fusion rHA as indicated. The ratios of amidination are determined from the AUC, as measured by pseudo-PRM, for each target modified and unmodified peptide as described in Table 1. “Pre-fusion” and “post-fusion” preparations are as in Fig. 1b. SMTA amidination is performed with 20 mM SMTA, pH 7.0, 30 min at ambient temperature. Error bars represent standard deviation. p values were determined by two-tailed unpaired Student’s t test, n = 6 replicates prepared independently

Full size image

To validate that the conformational switch itself drove the change in SMTA labeling, we considered two orthogonal methods of blocking viral fusion. HA requires pre-cleavage to free the fusion peptide and enable the conformational change [45]. This pre-cleavage occurs physiologically by endogenous host proteases, but can be accomplished in vitro through brief tryptic digestion [46]. In the absence of tryptic pre-cleavage, no difference on Peptide 1 amidination of PR8 HA was observed between acidic and neutral incubation (Fig. 3a). Because of the central role of fusion in viral entry and infection, fusion inhibitors have been intensively investigated as antiviral candidates. Arbidol (umifenavir) has been reported to inhibit fusion of H1 serotype HA, including PR8, at 400 μM [47]. We compared the amidination of target peptides in the presence or absence of arbidol, as well as under neutral or acid incubation conditions. Consistent with the amidination fingerprint serving as a reporter of fusion competence, acidification in the presence of arbidol yielded an amidination profile similar to that of HA that was never exposed to low pH in Peptide 1 (Fig. 3b). Similar, though not statistically significant (ANOVA p = 0.07), effects were observed for Peptide 2 (ESM Fig. S3A). Furthermore, in support of this effect being mediated by a direct binding event, Peptides 3 and 4, which are present near the arbidol binding pocket (Fig. 3c), demonstrated decreased amidination across conditions (Fig. 3d and ESM Fig. S3B), consistent with arbidol’s presence in the pocket lowering the accessibility of this residue. In summary, both biochemical and chemical inhibition of fusion yield an electrophilic footprint resembling the pre-fusion state.

Fig. 3
figure 3

The amidination assay is sensitive to conditions that interfere with viral fusion. Amidination ratios for rHA that has not been tryptically pre-cleaved, subject to acidification and labeling as in Fig. 1. Amidination ratios at Peptide 1 in rHA prepared in either the pre- and post-fusion states following pre-incubation with Arbidol (400 μM, 30 min), a fusion inhibitor, as indicated. “Pre-fusion” and “post-fusion” preparations are as in Fig. 1. Error bars represent standard deviation. Significance was first assessed by single factor ANOVA, followed by two-tailed Tukey’s post hoc test. ANOVA F = 12, p < 10−4, **Tukey p < 0.002. n = 6 replicates prepared independently. Amidination ratio at Peptide 4, as described in Fig. 3b. ANOVA F = 8.3, p < 10−3, *Tukey p < 0.02. Arbidol structure superimposed in the binding pocket on a crystal structure for unbound PR8/HA, and with the amidination-prone lysines of Peptides 3 and 4 noted in green. Note that the PR8/HA structure would undoubtedly be substantially distorted, as the binding pocket on this isotype is too narrow to accommodate the inhibitor [47]

Full size image

Sensitivity and robustness

To determine the sensitivity of the approach, we measured the LOD for modified and amidinated Peptide 1 with respect to rHA, finding 190 and 180 fmol respectively (Table 2, Fig. 4a), with linear response up to 3 pmol. Similar LOD were identified for the other profiled peptides (Table 2, ESM Fig. S4). Although these LODs are high with respect to typical MRM and PRM thresholds, it should be stressed that these are with respect to the total HA protein concentration, rather than the levels of peptide. The amidination ratio is robust to concentration over this range (Fig. 4b). We also considered that the injection time could bias the observed ratio [25]; we maintained an AGC of 104 while varying injection time (Fig. 4c). Longer injection times allow variance in cycle time when the AGC target is reached, while weaker signals with lower injection times can lead to measurement artifacts due to the Xcalibur noise filter [48]. Interestingly, the ratio does not change with injection time, indicating that the AGC target is not getting achieved; the cycle time is consistent across each peak. However, at lower injection times, the variance does increase. To compare to a more traditional MRM approach, we further evaluated amidination of Peptide 1 on a TSQ Quantiva triple quadrupole mass spectrometer. Similar results to those obtained using pseudo-PRM (Fig. 3b) were observed, but with greater variation (Fig. 4d).

Table 2 LODs (from standard deviation of the response) for HA peptides
Full size table
Fig. 4
figure 4

Sensitivity of pseudo-PRM measurement of Peptide 1. Linear relationships between AUCs for modified and amidinated Peptide 1 and the amount of PR8 injected. Minimal dependence of the modification ratio of Peptide 1 on the amount of PR8 HA loaded between 13 and 200 ng. Minimal dependence of the modification ratio of Peptide 1 on the injection time into the ion trap. Error bars represent standard deviation. MRM quantification of Peptide 1 in rHA digests on a triple-quadrupole mass spectrometer. Significance was first assessed by single factor ANOVA, followed by two-tailed Tukey’s post hoc test. ANOVA F = 6.0, p < 0.005. n = 6 replicates prepared independently

Full size image

Evaluation of a mutant form of rHA

A chimeric rHA (cHA) is a mutant form of rHA, which is composed of the head from one HA subtype and the stalk domain from the other [28]. These play an important role in emerging strategies for universal influenza vaccine development [49]. The lack of a fusion assay has limited our understanding of the function of the cHA. To this end, we evaluated amidination of one such chimera, cH6/1 (Fig. 5a, b; head from A/mallard/Sweden/81/02 and stalk from A/Puerto Rico/08/34), which contains Peptide 4 in its PR8 HA stalk domain and whose virus is challenging to generate de novo, implying impaired HA function. Unlike in native PR8 HA, amidination of Peptide 4 in cH6/1 does not decrease following acidification (Fig. 5b), indicating that its conformation is less influenced by acidification, and perhaps less fusion-competent, than that of native PR8 HA.

Fig. 5
figure 5

Schematic illustrating the design of chimeric c6/1 HA, and comparison to the PR8 HA sequence. Locations of Peptides 1–4 are indicated. Peptides 1 and 2 are located in the head region of PR8 HA and hence are absent in the c6/1 HA chimera. Differential amidination of Peptide 4 between pre- and post-fusion states of c6/1 HA

Full size image

PRM for detection in complex mixtures

Ultimately, the benefit of PRM profiling of HA fusion, as compared to HDX, would be to identify the fusion-associated conformational switch in complex biological milieu. To evaluate our methods’ performance in complex mixtures, we spiked rHA digest into a HEK293T lysate, and compared the measurement of amidination to rHA digest in buffer, as well as between the use of the LTQ ion trap and the Orbitrap (Fig. 6a). Both traps perform well, but the high-resolution Orbitrap does decrease variance as opposed to the ion trap. As with the ion trap, no dependence of the modification ratio on injection time was observed (ESM Fig. S5A). Lysate added to labeled PR8 HA prior to tryptic digestion does not affect the Peptide 1 modification ratio. We also considered adding PR8 HA to the lysate prior to labeling. Interestingly, lysate prepared with high detergent RIPA buffer does affect the observed SMTA labeling, but lysate prepared by freeze-thaw of cells in PBS does not affect the PR8 HA modification ratios measured by PRM (Fig. 6b).

Fig. 6
figure 6

Pseudo-PRM (ion trap) and PRM (orbitrap) in lysate. Differential amidination at peptide 1 on rHA in the presence or absence of tryptically digested HEK293T lysate. Error bars represent standard deviation. Significance determined from 2-tailed unpaired Student’s t test (n = 3 independently prepared samples). Differential amidination of 1 μg HA spiked into 5 μg HEK293T lysate prepared by the freeze-thaw method, then SMTA-labeled, digested and analyzed by PRM in the Orbitrap. Error bars represent standard deviation. Significance determined from two-tailed unpaired Student’s t test (n = 6 independently prepared samples)

Full size image

Conclusions

This study represents the first reported effort to combine electrophilic profiling and reaction monitoring. We profiled two different conformations of rHA: the pre-fusion state and the acid-induced post-fusion state. These two states demonstrate different susceptibility to electrophilic modification at lysine by the labeling reagent SMTA. The reliance of differential amidination on fusion is validated by its dependence on HA pre-cleavage and the loss of differential amidination upon pre-incubation with the fusion inhibitor Arbidol. Furthermore, the homogeneity of the labeling product allowed straightforward mass spectrometric quantification through MRM on a triple quadrupole instrument, pseudo-PRM in a linear ion trap, and PRM in an Orbitrap mass analyzer. Surprisingly, we find that normalization to non-amidinated peptides harboring a missed cleavage as the internal reference peptides provides robust quantification. The ability to discriminate between pre- and post-fusion states, and the robustness of the assay in the presence of complex matrix interferences, suggests that this approach will allow quantification of intracellular hemagglutinin fusion competence.

One potential application of this approach is to characterize how sequence influences HA fusion competence. Another long-term goal of this methodology is to identify HA fusion yield inside the cell. Under this scheme, we find limits of detection in the 100 to 300 fmol range with respect to HA. At 1000 molecules of HA per virion [50], profiling cells at a multiplicity of infection of 10 would yield about 2 fmol HA in a well of a 96-well plate, above the 0.1 to 1 fmol range often achieved for peptide detection using isotope dilution MRM [43, 51]. This method could similarly benefit from isotopically labeled standards. One inherent limitation is the constriction of peptide choice as compared to total protein quantification, wherein the peptides with the best signal, robustness, and freedom from interferences can be chosen. It is possible that in this respect, the product heterogeneity associated with more promiscuous labeling methods such as FPOP might be superior, as the diversity of products increases the number of potential precursor ions that report on individual conformational changes. On the other hand, greater diversity of products might lower the sensitivity to any individual product. Further investigations will compare amidination to other protein labeling strategies as a method to identify HA conformational changes by reaction monitoring of covalent labeling products.