Semiquantitative Proteomics Enables Mapping of Murine Neutrophil Dynamics following Lethal Influenza Virus Infection

This article is featured in In This Issue, p.773

A causal relationship exists clearly between lung neutrophils and influenza A virus (IAV)-induced lung injury (1–5). Neutrophil contributions shift from being protective to deleterious with increased viral load and lung inflammation levels, and lung neutrophilia is a well-established hallmark of fatal IAV infections (2, 6). In vivo depletion studies in mice, using the neutrophil-selective anti-Ly6g mAb, have demonstrated a nonredundant role for neutrophils in limiting IAV proliferation during the early days following infection (1, 5). However, partial neutrophil depletion using low doses of anti-Ly6g mAb conversely decreases the prevalence of IAV-induced mortality, implying that a hazard threshold exists where excessive neutrophil recruitment exacerbates IAV-induced lung tissue injury (2). Using transcriptomics, an IAV lethality–associated and neutrophil-derived module was previously identified, which featured genes involved in neutrophil chemotaxis and IL-1, IL-6, and TNF-α production (2). Based on this, a neutrophil-CXCL2 feedforward recruitment axis was proposed to drive IAV-induced disease mortality (2). However, whether IAV-induced neutrophil-driven injury is merely a consequence of increased neutrophil numbers or additional changes in neutrophil phenotype remains unclear. Understanding the dynamics of the IAV lethality–associated lung neutrophil proteome may also allow us to identify novel IAV lethality–associated proteins and refine neutrophil-centric strategies for attenuating IAV-induced lung injury.

Proteomics provides advantages over transcriptomics when characterizing neutrophil functions. Only a modest correlation (R² = 0.4–0.6) exists between global mRNA and protein levels (7–9), with correlations varying further between different protein classes (7, 9). This is significant, as proteins, rather than mRNA, act as cellular effectors and determinants of environmental adaptation (10). Traditionally regarded as terminally differentiated cells, mature neutrophils also contain significantly less mRNA content and mRNA heterogeneity compared with other leukocytes (11–13), with >10% contamination by other leukocytes disproportionately confounding a previous study of tissue neutrophil transcriptomes (14). Transcriptomics also cannot recapture the degranulation process, despite its central importance to neutrophil function (15). Neutrophil proteomes have already been characterized in the peripheral blood of healthy humans (16, 17) or after acute inflammation (18) in human patients with rare monogenic diseases (19) or traumatic injuries (20) and in rodent and murine models of trauma (20) and bacterial infection (21). These studies have highlighted the diversity of neutrophil effector proteins and that changes in protein biosynthesis, immune cell signaling, cellular stress, and apoptosis pathways appear in neutrophils between steady-state and disease. Our study focuses on the IAV lethality–associated neutrophil proteome, which has not been characterized before despite abundant evidence linking lung neutrophilia to increased lung injury (1–5). Using semiquantitative proteomics, we can map the degree to which the bronchoalveolar lavage (BAL) neutrophil phenotype is predetermined by bone marrow (BM) maturation or additionally modulated across different tissue compartments during lethal IAV infection.

In this study, we used multiplex peptide stable isotope dimethyl labeling (22) to characterize BM, blood, and BAL neutrophil proteomes during homeostasis or lethal IAV infection (1000 PFU H1N1 A/PR/8/34). A total of 1372 proteins were identified across all neutrophil samples, with 155 proteins differentially expressed between at least two tissue compartments. Following lethal IAV infection, three compartment-specific changes were highlighted; 1) translation of IFN-stimulated genes (ISGs) was significantly increased in BM neutrophils, 2) blood neutrophils then underwent a context-independent downregulation of ribosomal proteins, and 3) BAL neutrophils were characterized by protein degranulation and a limited increase in type I IFN–stimulated and metal-sequestering proteins. A comparison of BAL neutrophil proteomes following sublethal (100 PFU PR8) or lethal (1000 PFU PR8) IAV infection or Gram-negative Pseudomonas aeruginosa infection also revealed that neutrophil-derived ISGs and lipid metabolism-associated proteins were selectively increased following IAV infection in a dose-dependent manner. Altogether, these data provide comprehensive insight into how the IAV lethality–associated BAL neutrophil response is deployed from BM neutrophil maturation to alveolar airspace infiltration.

All experimental procedures were approved and performed in accordance with guidelines from the La Trobe University Animal Ethics Committee. C57BL/6 (B6) mice were purchased from the Walter and Eliza Hall Institute Bioservices (Kew, Melbourne, Australia) or bred in house at the La Trobe Animal Research and Teaching Facility (LARTF) (Bundoora, Melbourne, Australia) from B6 breeders purchased from the Walter and Eliza Hall Institute Bioservices. B6.Ifnar1^−/− mice were bred inhouse at LARTF and backcrossed onto the B6 background for at least 12 generations (23). All mice were maintained under specific pathogen-free conditions. Female mice aged 10–12 wk old were used for all experiments.

The IAV strain A/PR/8/34 (PR8, H1N1) was propagated in 10-d-old embryonated hens’ eggs, and the PR8 virus was titrated using Madin-Darby canine kidney cells as described (24). P. aeruginosa was a kind gift from Associate Professor Hamsa Puthalakath (25). For each experiment, an aliquot of P. aeruginosa was grown in 4 ml of Luria-Bertani media for 16 h at 37°C with shaking, washed once in PBS, and resuspended in 4 ml of PBS. CFU was determined by measuring the OD600 using a spectrophotometer.

For intranasal infections, mice were anesthetized with methoxyflurane and infected with 100 or 1000 PFU PR8 or 3 × 10⁶ P. aeruginosa diluted to a final volume of 30 μl using PBS. Control mice received 30 μl of PBS intranasally. Mice were weighed daily and euthanized at the indicated timepoints postinfection or at a humane end point of <75% of original body weight for survival studies.

BAL samples were obtained by inserting a cannula through the trachea of euthanized mice and flushing the lungs with 4 × 0.5 ml of PBS supplemented with 2% FCS and 2 mM EDTA (FACS buffer). Whole blood was collected via the vena cava (150 μl unless otherwise specified), and BM cells were flushed with RPMI media from the left femur. All samples were incubated in the RBC Lysing Buffer Hybri-Max (Sigma-Aldrich) for 5 min room temperature (RT) and resuspended into single-cell suspensions with FACS buffer for flow cytometry analysis or cell sorting. Cell numbers and cell viability were assessed via trypan blue exclusion using a hemocytometer.

Single-cell suspensions were incubated with BD Fc Block (anti-CD16/32) for 10 min at 4°C, before incubation with fluorochrome-conjugated primary Abs or nonbinding Ab controls (listed in Supplemental Table I) for 20 min at 4°C. Cells were washed in FACS buffer and resuspended in 100 μl of FACS buffer, with 5 μl of DAPI solution added 30 s prior to data acquisition unless LIVE/DEAD Fixable Yellow Dead Cell Stain was used. Data were acquired on either a BD Canto II or a Cytoflex S flow cytometer. Flow cytometry data analysis was conducted using FlowJo v10 (Tree Star). Nonbinding Ab controls were used as controls for nonspecific IgG binding and/or nonspecific fluorophore binding.

Single-cell suspensions were incubated with BD Fc Block (anti-CD16/32) for 10 min at 4°C before incubation with fluorochrome-conjugated primary Abs (described in Supplemental Table I) for 20 min at 4°C. Cells were washed in 1 ml of PBS and fixed in 200 μl of 1% PFA dissolved in PBS for 20 min at RT, washed, and stored in 1 ml of PBS until further use. Cell membrane permeabilization was achieved using 100 μl of ice-cold 100% methanol and kept on ice for 30 min. Cells were washed in 2 ml of PBS and blocked in 10% goat serum diluted in PBS for 60 min at RT before incubation with primary intracellular Abs diluted in 10% goat serum at 4°C overnight. Cells were washed in PBS and incubated with secondary Ab (goat anti-rabbit IgG Alexa Fluor 647) for 60 min at RT, washed, and resuspended in 100 μl of PBS. Data were acquired on a Cytoflex S flow cytometer. Flow cytometry data were analyzed using FlowJo v10 (Tree Star).

Neutrophils were obtained from age-matched, 1) PBS control–treated, 2) 1000 PFU PR8-infected, 3) 100 PFU PR8-infected, or 4) 3 × 10⁶ P. aeruginosa–infected B6 mice at day 3 postinfection. To obtain sufficient cell numbers for proteomics, a pool of 11 mice per group was used for PBS control versus 1000 PFU PR8 infection experiments (comparing BM, blood and, where applicable, BAL neutrophils), and a pool of six mice per group was used for 100 versus 1000 PFU PR8 infection versus 3 × 10⁶ P. aeruginosa infection experiments (comparing only BAL neutrophils). Single-cell suspensions were incubated with BD Fc Block (anti-CD16/32) for 10 min at 4°C and incubated with anti-mouse CD45.2 FITC, CD11b PerCP5.5, Ly6g PE, and Siglec-F allophycocyanin primary Abs for 20 min at 4°C (Supplemental Table I). FACS sorting was conducted using a BD FACSAria III, with gating of neutrophils as all live (DAPI^neg)/singlet/Siglec-F^neg/Ly6g^pos and CD11b^pos cells. Cell purity was checked postsort. Sorted cells were then washed in PBS, snap-frozen in liquid nitrogen, and stored at −80°C before use.

Cell pellets were dried using a SpeedVac Concentrator and Savant Refrigerated Vapor Trap (Thermo Fisher Scientific) for 45 min RT. Samples were dissolved in digestion buffer (8 M urea, 50 mM ammonium bicarbonate, 2 mM tris(2-chloroethyl) phosphate before incubation for 5 h at 25°C. A total of 10 mM iodoacetamide was then added to alkylate thiol groups at 20°C for 35 min in the dark. The alkylated preparation was diluted to 1 M urea with 25 mM ammonium bicarbonate (pH 8.5) before sequencing grade trypsin (Promega) was added to 5 μM final concentration. Trypsin digestion was performed overnight at 37°C. Digests were acidified with 1% (v/v) trifluoroacetic acid and the peptides desalted on SDB-XC (Empore) StageTips as previously described (26). Peptides were modified by stable isotope dimethyl labeling for quantitative proteomics according to (22). A label switch was performed for each replicate. Labeled samples were mixed in a 1:1:1 ratio, desalted as described above, and dried before being fractionated off-line by high-pH reversed-phase fractionation (27) into 12 fractions. Peptides from each fraction were reconstituted in 0.1% trifluoroacetic acid and 2% acetonitrile (ACN) and loaded onto C18 PepMap 100-μm inner diameter × 2 cm trapping column (Thermo Fisher Scientific) at 5 μl/min for 6 min and washed for 6 min before switching the precolumn in line with the analytical column (Acquity Ethylene Bridged Hybrid [BEH] C18, 1.7 μm, 130 Å and 75-μm inner diameter × 25 cm; Waters). The separation of peptides was performed at 250 nl/min using a nonlinear ACN gradient of buffer A (0.1% formic acid, 2% ACN) and buffer B (0.1% formic acid, 80% ACN), starting at 5% buffer B to 55% over 55 min, then 100% B for 5 min followed by an equilibration step of 15 min (0.1% formic acid, 2% ACN). Data were collected on a Q Exactive HF (Thermo Fisher Scientific) in Data-Dependent Acquisition Mode using m/z 350–1500 as mass spectrometry (MS) scan range. Higher-energy collision dissociation tandem MS (MS/MS) spectra were collected for the 10 most intense ions per MS scan at 15,000 resolution. Dynamic exclusion parameters were set as follows: exclude isotope on, duration 30 s, and peptide match preferred. Other instrument parameters for the Orbitrap were MS scan at 120,000 resolution, maximum injection time 30 ms, automatic gain control target 3 × 10⁶, collision at 28% energy for a maximum injection time of 110 ms with automatic gain control target of 1 × 10⁵.

Identification and isotopic quantification of proteins was performed on raw output files from liquid chromatography (LC) positive ion electrospray ionization MS/MS using MaxQuant (version 1.5.8.3) (28) together with its built-in search engine Andromeda. The Mouse Uniprot FASTA database (downloaded on March 29, 2017; 50,943 protein entries) together with common contaminants were used for analysis. Carbamidomethylation of cysteines was set as a fixed modification, acetylation of protein N-termini, methionine oxidation was included as variable modifications. For quantitative analysis, dimethyl labeled DimethlyLys0, DimethylNter0, DimethlyLys4, DimethylNter4, DimethlyLys8, and DimethylNter8 were used as labels together with the optional iBAQ (Intensity-Based Absolute Quantification) calculation. Parent mass tolerance was set to 4.5 ppm (after refinement by MaxQuant) and fragment mass tolerance to 20 ppm. Trypsin was set as the digestion enzyme with up to two missed cleavages allowed. The match between runs feature of MaxQuant was used to transfer peptide identifications from one run to another based on retention time and mass-to-charge ratio, and the “requant option” of MaxQuant was enabled. Both peptide and protein identifications were reported at a false discovery rate of 1%. To identify proteins that were differentially expressed in at least one comparison type, we filtered for protein expression ratios ≥2-fold and a raw p value <0.05. Multiple testing corrections bluntly reduce the detection of true positives in proteomics, especially for semiquantitative methodologies subject to ratio compression (29). The alternative incorporation of an effect size cut-off can be useful, as it still yields small false positives for low-power medium scale experiments [i.e., for samples with triplicates and 1000 protein entries (28)]. Mass spectrometry proteomics data were deposited to the ProteomeXchange Consortium via the PRIDE (30) partner repository with the dataset identifier PXD012681: http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD012681, and the entire proteomics R data analysis workflow is available at https://github.com/codetrainee/Neutrophil-systems-biology.

For proteomic studies, experiment 1 comprised comparisons between 1) 1000 PFU PR8 BAL versus BM neutrophils, 2) 1000 PFU PR8 BAL versus blood neutrophils, and 3) 1000 PFU PR8 blood versus BM neutrophils. Experiment 2 comprised comparisons between 1) 1000 PFU PR8 BM versus PBS control BM neutrophils and 2) PBS control blood versus BM neutrophils. Because a strong correlation was observed for 1000 PFU BM neutrophil samples between experiments 1 and experiment 2 (R² = 0.887), the comparison 1000 PFU PR8 blood versus PBS control blood neutrophils could be additionally modeled in silico using the following equation:

where

is a sample comparison obtained from experiment 1 and

is a sample comparison obtained from experiment 2. Experiment 3 comprised comparisons between BAL neutrophils obtained from B6 mice infected with 1) 3 × 10⁶ CFU P. aeruginosa versus 100 PFU PR8, 2) 3 × 10⁶ CFU P. aeruginosa versus 1000 PFU PR8, and 3) 1000 PFU PR8 versus 100 PFU PR8.

For comparisons between two groups, the Welch t test (two-tailed, Welch degrees of freedom modification) was used. When comparing three or more groups, a one-way ANOVA test was applied followed by the Tukey posttest. A p value <0.05 was considered as statistically significant for all tests used. Statistical analyses were performed using R 3.5.1, and data depicted as mean or median ± SEM as indicated.

B6 mice were characterized at day 3 and 6 post-PR8 infection using a sublethal (100 PFU) or lethal (1000 PFU) dose of IAV infection, respectively (Fig. 1A, 1B). In agreement with previous observations (1), BAL immune cell numbers greatly increased with IAV infection dose (Fig. 1C). BAL and blood neutrophil numbers and proportions were highly increased following IAV infection compared with PBS control (Fig. 1C), with negligible BAL neutrophils observed in PBS control (Fig. 1C, Supplemental Fig. 1A). However, no statistically significant differences in neutrophil proportions and numbers were observed between 100 PFU and 1000 PFU PR8 infection, except for changes in blood neutrophil numbers at day 6 postinfection (Fig. 1D, Supplemental Fig. 1B). BM neutrophil numbers and proportions were not altered following IAV infection compared with PBS control (Supplemental Fig. 1B, 1C). In concordance with an activated tissue-infiltrating phenotype (31), BAL neutrophils upregulated surface levels of CD11b, CXCR2, and MHC class II (MHC II), and downregulated surface levels of CD62L compared with blood and BM neutrophils at day 3 and 6 post-PR8 infection (Fig. 1E, 1F and statistics in Supplemental Fig. 2). In contrast, Ly6c expression remained unchanged between BM, blood, and BAL neutrophils. Discrete neutrophil subsets were not observed based on a lack of bimodally distributed neutrophil activation markers. Neutrophil surface marker expression levels were visualized along a biexponential scale in all flow cytometry panels (Fig. 1E, 1F). Changes in neutrophil surface marker expression levels were also calculated as median fluorescence intensities (MFIs) and compared along a linear scale (Supplemental Fig. 2). No statistically significant differences in MFIs were observed between sublethal and lethal IAV infection, except for CXCR2 (p = 0.0258), CD62L (p = 0.0198), and Ly6C (p = 0.0256) in BAL neutrophils at day 6 (Supplemental Fig. 2). As changes in cell surface neutrophil activation markers present very limited insight into overall neutrophil function, we used high-resolution MS to comprehensively profile comparative changes in BM, blood, and BAL neutrophil proteomes during homeostasis or lethal IAV infection.

Multiplex peptide stable isotope dimethyl labeling uses isotopomeric dimethyl labeling of trypsin-digested peptides to quantitate relative changes in the proteomes of three different samples per experimental run (22) (Supplemental Fig. 3A). Pragmatic advantages include its scalability for use with lower protein amounts (i.e., from 1 × 10⁶ rather than 1 × 10⁷⁻⁸ cells), and reduced experimental variability as multiple samples are simultaneously quantitated per experiment. We characterized the proteome of BM, blood, and BAL neutrophils following lethal IAV infection. We chose day 3 post-IAV infection, as this strides the peak of lung neutrophilia and provides a potential therapeutic window where key effectors can still be targeted before severe lung damage occurs (32). To study the lung neutrophil proteome, BAL instead of digested lung tissue neutrophils were obtained to avoid contamination by pulmonary capillary-sequestered blood neutrophils and the confounding influence of lung tissue digestion on neutrophil activity. As BAL neutrophil recruitment is negligible in PBS control–treated mice (Supplemental Fig. 1A), the PBS BAL neutrophil proteome was not characterized as a comparison point.

Using FACS, 1 × 10⁶ BM, blood, and BAL neutrophils were isolated from pooled B6 mice with a purity >98% (Supplemental Fig. 3B–D). As each experimental run was limited to three sample comparisons (via peptide labeling with “light,” “intermediate,” or “heavy” stable isotopes), we designed two experimental studies with 1000 PFU PR8 BM neutrophils acting as an internal constant (Fig. 2A). Experiment 1 comprised comparisons between 1) 1000 PFU PR8 BAL versus BM neutrophils, 2) 1000 PFU PR8 BAL versus blood neutrophils, and 3) 1000 PFU PR8 blood versus BM neutrophils. Experiment 2 comprised comparisons between 1) 1000 PFU PR8 BM versus PBS control BM neutrophils and 2) PBS control blood versus BM neutrophils, allowing us to compare neutrophil changes between homeostasis and lethal IAV infection. A robust correlation was observed for 1000 PFU BM neutrophil samples between experiments 1 and 2 (R² = 0.887) (Supplemental Fig. 4A). A total of 1797 and 1591 proteins were identified from experiments 1 and 2, respectively, with 1372 proteins shared between both experiments and used for downstream analyses (Fig. 2B). In terms of cellular location, most identified proteins were associated with organelle, membrane, or extracellular regions (Supplemental Fig. 4B). Canonical cell surface activation markers CD11b and CXCR2 were detected in the LC-MS/MS data but did not pass the filtering process, whereas CD62L and MHC II were not identified in LC-MS/MS data, suggesting a potential detection bias against cell surface membrane proteins (33).

Principal component analysis (PCA) revealed that replicates of each sample comparison clustered together, indicating low variability between individual replicates, whereas different sample comparison types predominantly clustered apart, indicating distinct changes observed between different neutrophil compartments. A notable exception occurred for two different sample comparison types that clustered together (1000 PFU PR8 blood versus BM neutrophils and PBS control blood versus BM neutrophils), indicating that neutrophil proteome changes following BM egress into peripheral blood are indistinguishable between homeostasis and IAV infection (Fig. 2C). This was also evident following unsupervised hierarchical clustering of all identified proteins (Supplemental Fig. 4C) or all proteins differentially expressed in at least one comparison type (Fig. 2D). An additional comparison of 1000 PFU PR8 blood versus PBS control blood neutrophils could also be separately modeled in silico and visualized within the context of all experimentally acquired sample comparisons (Supplemental Fig. 5A, 5B). This in silico comparison clustered very similarly with the experimental comparison 1000 PFU PR8 BM versus PBS control BM neutrophils, highlighting that neutrophil proteome differences between PR8 infection and homeostasis are first acquired in BM neutrophils and then retained in blood neutrophils (Supplemental Fig. 5A, 5B).

To identify proteins that were differentially expressed in at least one comparison type (i.e., between at least two different neutrophil compartments), we filtered for protein expression ratios ≥2-fold with a raw p value <0.05, identifying 155 proteins that could be organized into nine protein clusters using k-means clustering (Fig. 2D, Supplemental Table II). Calculation of the optimal number of clusters was determined using the elbow method (i.e., the total within-cluster sum of square [WSS]). A notable finding was cluster 9, which contained proteins with increased expression in BM neutrophils following IAV infection compared with PBS control. These proteins predominantly comprised ISGs that directly sense and/or inhibit virus proliferation and propagation, such as ubiquitin-like protein ISG15, IFN-induced protein with tetratricopeptide repeats 1 (IFIT1), IFN-activated gene 204 (IFI204; also known as IFI16), DExD/H-Box helicase 58 (DDX58), 2′-5′-oligoadenylate synthetase 3 (OAS3), and deltex E3 ubiquitin ligase 3L (DTX3L). ISGs associated with antimicrobial activity, such as guanylate binding protein 7 (GBP7) and galectin 9 (LGALS9), or with terminal differentiation, such as Schlafen (SLFN1) and cytidine/uridine monophosphate kinase 2 (CMPK2), were also found in this cluster.

Other notable clusters included cluster 6, which contains proteins that are further increased in BAL neutrophils compared with blood or BM neutrophils following IAV infection, and cluster 3, which contains proteins that were conversely decreased in BAL neutrophils compared with blood or BM neutrophils following IAV infection (Fig. 2D). Cluster 6 comprised proteins associated with antiviral (guanylate binding protein 2 [GBP2]) or immunomodulatory activity (IL-1R antagonist [IL1RN]). Interestingly, proteins involved in iron (ferritin H chain 1 [FTH1] and ferritin L chain 1/2 [FTL1/2]) and selenium sequestration (selenium binding protein 1/2 [SELENBP1/2]) were also selectively upregulated in BAL neutrophils following peripheral blood recruitment and are novel IAV lethality–associated protein effectors. Cluster 3 predominantly comprised of secretory proteins with known extracellular distributions, including MMP8, MMP9, leucine-rich α-2-gp1 (LRG1), and the putative myeloid cell chemotaxin resistin-like γ (RETNLG) (34). In contrast to clusters 3 and 6, cluster 8 comprised proteins that were increased in BM compared with blood neutrophils in a context-independent manner but also modestly increased in 1000 PFU PR8 BM versus PBS control BM neutrophils. Unsurprisingly, these proteins were associated with cell proliferation (proliferating cell nuclear Ag), mitotic regulation (cytoskeleton-associated protein 4) (35), or neutrophil differentiation (C/EBPε) (36)]. Altogether, these data provide a comprehensive profile of how the BAL neutrophil proteome is shaped from BM neutrophil maturation to alveolar airspace infiltration following lethal IAV infection.

We next focused on all proteins upregulated between 1000 PFU PR8 versus PBS control BM neutrophils (Fig. 3A; no significantly downregulated proteins detectable. Analysis of protein–protein interactions using the STRING database (37) confirmed that a network of mutually interacting ISGs were upregulated in BM and consequently also blood neutrophils following lethal IAV infection (Fig. 3A, Supplemental Fig. 6), agreeing with previous findings that antiviral signaling was elevated in BM cells following influenza or Sendai virus infection (38). By comparison, a context-independent decrease in ribosomal protein expression was observed in blood compared with BM neutrophils during both homeostasis or lethal IAV infection, agreeing with previous transcriptomic observations that the blood neutrophil phenotype is predominantly characterized by a comparative decrease in the transcription of translation-associated genes (11, 12) (Supplemental Figs. 6, 7, Supplemental Table III). To specifically investigate whether type I IFN signaling contributed to the upregulation of ISGs in BM, blood, and BAL neutrophils following lethal IAV infection, we measured BM, blood, and BAL neutrophil GBP2, ISG15, and OAS3 expression levels in 1000 PFU PR8-infected type I IFN–α receptor 1–deficient (B6.Ifnar1^−/−) mice compared with wild-type B6 mice (Fig. 3B). Interestingly, neutrophil expression of OAS3, but not GBP2 or ISG15, was decreased across all neutrophil compartments in B6.Ifnar1^−/− compared with wild-type mice (Fig. 4). This may reflect previous observations that ISG15 and GBP2 induction are type I IFN independent (39), with the latter also caused by type II IFNs (40, 41).

We next focused on proteins upregulated or downregulated between 1000 PFU PR8 BAL versus blood neutrophils. In contrast with IAV-induced changes at the BM level, proteins upregulated in BAL compared with blood neutrophils following IAV infection were less densely interconnected and more likely represented disparate effector proteins (Fig. 3C). Interestingly, proteins downregulated in BAL compared with blood neutrophils interacted centrally through vascular endothelial growth factor A (VEGFA), TGF β 1, and TNF. This observation complements the well-established proangiogenetic activities of MMP8, MMP9, and LRG1 and the antimicrobial activities of lactotransferrin (LTF) and chitinase 3–like 1 (Fig. 3C). Proteins downregulated in BAL compared with blood neutrophils were also known to be extracellularly distributed and included those previously identified in neutrophil granules (LTF, neutrophilic granule protein [NGP], serglycin [SRGN]).

An alluvial diagram was also used to visualize all differentially expressed proteins ordered by the direction of expression between sequential neutrophil compartments (Fig. 4A). This enabled a global map of all compartmental protein changes occurring from BM neutrophil maturation to alveolar airspace infiltration during lethal IAV infection. To identify whether specific signaling pathways were overrepresented in differentially expressed neutrophil proteins between different neutrophil compartments, we searched all candidates, except for those downregulated in PBS control blood versus BM neutrophils (separately shown in Supplemental Fig. 8) against the Reactome database (42). As expected, IFN, antiviral, and cytokine signaling pathways were predominantly overrepresented among proteins that were upregulated in PR8 1000 PFU BM compared with PBS control BM neutrophils (Fig. 4B, Supplemental Fig. 8). Degranulation was overrepresented among proteins that were downregulated in PR8 1000 PFU lung versus blood neutrophils (Fig. 4B).

Neutrophils are rapidly recruited in response to both bacterial and viral lung infections and are considered as a nonspecific first line of defense. Whether BAL neutrophils selectively modify parts of their defense repertoire in response to bacterial versus viral infections has not been studied before using proteomics. We therefore characterized the proteome of BAL neutrophils following sublethal (100 PFU) or lethal (1000 PFU) PR8 infection or sublethal P. aeruginosa bacterial infection (Fig. 5A–C, Supplemental Table IV). A decreased total number of 582 proteins were identified but overlapped with those previously increased in PR8 1000 PFU BAL neutrophils. ICAM1, cluster of differentiation 14 (CD14) (the LPS coreceptor), and FTH1 expression was highly increased in BAL neutrophils following P. aeruginosa compared with PR8 (100 PFU or 1000 PFU) infection, in concordance with the LPS-induced upregulation of ICAM1 observed in activated neutrophils (43). Expression of SELENBP1/2 and the ISGs: IFIT1, ISG15, ISG20, and radical S-adenosyl methionine domain containing 2 (RSAD2) were increased in BAL neutrophils following PR8 compared with P. aeruginosa infection and in an IAV dose–dependent manner, indicating a modest capacity for pathogen response tuning in BAL neutrophils (Fig. 5B, 5C). Other BAL neutrophil proteins that unexpectedly increased in PR8 compared with P. aeruginosa infection included lipid metabolism-associated proteins (carboxylesterase 1D [CES1D], carbonyl reductase 2 [CBR2], aldehyde dehydrogenase 1 family member A1 [ALDH1A1]), and the lung protective proteins surfactant protein B (SFTPB), and secretoglobin family 1A member 1 (SCGB1A1).

BAL neutrophil protein expression differences were also separately characterized using intracellular protein staining following 100 or 1000 PFU PR8 infection or 1 × 10⁴ PFU X-31 infection or LPS administration (Fig. 5D). The MFI of ISG15, GBP2 was significantly increased following 1000 PFU PR8 infection compared with LPS administration. The MFI of SELENBP1/2 was significantly decreased following LPS administration compared with IAV infection, with the highest level of expression observed following 1000 PFU PR8 infection, in agreement with the proteomics data. The MFI of MMP9 was significantly decreased in BAL neutrophils following LPS administration compared with 100 PFU PR8 infection, also agreeing with the proteomics data. Altogether, our study provides new and comprehensive insight into both the dynamics and selectivity of the IAV lethality–associated BAL neutrophil proteome. This knowledge may help us to identify new IAV lethality–associated neutrophil effectors and refine neutrophil-centric strategies for attenuating IAV-induced lung injury.

To our knowledge, our study provides the first proteomic characterization of neutrophil dynamics following lethal IAV infection, from BM maturation and egress to alveolar airspace infiltration. In the era of rapid advances in transcriptomics, proteomics can be underused despite only a modest correlation (R² = 0.4–0.6) existing between global mRNA and protein levels (7, 9, 18). A key reason stems from the absence of amplification steps prior to sequencing in proteomics, which prevents characterization of less abundant primary immune cell populations. Semiquantitative proteomics using multiplex peptide stable isotope dimethyl labeling can help reduce this difficulty threshold by enabling proteomic studies on samples with less starting material compared with label-free proteomics (i.e., from 1 × 10⁶ rather than 2 × 10⁶ to 1 × 10⁷⁻⁸ cells). Neutrophils possess significantly reduced mRNA and heterogeneity content compared with other leukocytes, with only ∼1500 genes, explaining 80% of all mRNA content (12). Consequently, it is impossible to predict how many neutrophil mRNA changes, even disease lethality-associated ones, translate into significant protein-level changes using transcriptomics alone. Interestingly, whereas ∼4000 genes explain 80% of all mRNA content in T cells (12), only 250 proteins account for 75% of the cytotoxic T cell protein mass (7), highlighting that protein heterogeneity may be comparatively restricted in effector immune cells. Considering this, our identification of 1372 proteins from triplicate BM, blood, and BAL neutrophil samples may be viewed as a good representation of the neutrophil proteome (44, 45).

Neutrophil proteomes have been previously characterized in studies of neutrophil development (46), response to pathogens (45, 47), chronic disease (18, 48, 49), or physical stress (20, 50). These studies highlight the dynamism of the neutrophil response, with a degree of nonoverlapping changes captured between different disease conditions. In a heterogeneous condition like chronic obstructive pulmonary disease, the blood neutrophil proteome has been used to segregate patients with increased or decreased bacterial responsiveness independent of changes in clinical progression (49). Proteomics has also enabled the reclassification of distinct human blood neutrophil subsets, with CD16^highCD62L^low neutrophils identified as a separate and inflammation-induced subset distinguishable from banded or segmented neutrophils (45). A comparison with other proteomic studies (45, 51) indicates that IAV infection and LPS administration induce overlapping changes in neutrophil protein expression, albeit at different levels. In humans, LPS administration greatly decreased neutrophil MMP8/9 protein levels concurrent to increased anti-inflammatory IL1RN and iron-sequestering FTH1 levels. These changes were also observed in our study of BAL neutrophil proteomes following IAV infection, albeit at different fold changes. The nonspecific nature of neutrophil defense is also reinforced by observations that ISG15 is upregulated in neutrophils following both LPS administration or IAV infection (45, 52), necessitating caution when using neutrophil-derived ISGs as patient virus response predictors (53). These results also provide two important implications for the identification of IAV disease severity biomarkers in human patients. First, increased ISG expression is a neutrophil-derived component of the anti-IAV immune response and gene modules separating IAV disease severity based on increased ISG versus neutrophil activity may actually be capturing nonconventional or dysregulated antiviral neutrophil activity in the latter module (53). Second, the use of blood-derived neutrophil biomarkers requires careful consideration, as a cluster of IAV lethality–associated proteins are only increased following pulmonary microvasculature transmigration and/or alveolar airspace infiltration.

Our studies highlight that two infection-dependent and one context-independent compartmental transitions constitute the final IAV lethality–associated BAL neutrophil response. These two infection-dependent transitions consist of 1) an increased translation of ISGs in BM neutrophils and 2) a further increase in ISGs and metal-sequestering protein levels concurrent with the degranulation of secretory proteins in BAL neutrophils. The first observation confirms previous findings that BM leukocyte ISG expression is increased following either IAV or Sendai virus lung infection in a process dependent on peripheral blood type I IFN signaling (38). Following BM egress, ribosomal protein expression is decreased in peripheral blood neutrophils in agreement with transcriptomic findings that neutrophils express notably decreased translation-associated genes compared with other leukocytes (11, 13). This process, however, occurs independent of homeostatic or inflammatory signaling. Instead, a second and final IAV-induced proteome transition is observed in BAL neutrophils, potentially triggered by pulmonary microvasculature transmigration and/or selective pathogen- or damage-associated molecular patterns or G protein–coupled receptor ligands present in the lung microenvironment (54, 55). Of note, although our study identifies clear differences between the IAV lethality–associated BAL versus blood neutrophil proteome, we cannot speculate on the extent to which these changes are coordinated by pulmonary microvasculature transmigration, pulmonary epithelial cell transmigration, or the release of IAV infection–induced signals. This area merits further characterization, potentially through spatial transcriptomics and proteomics, as neutrophil transmigration is organ-selective and uses different selectins and integrins depending on the lung pathogen stimulus (56).

Following lethal IAV infection, proteins with decreased expression in lung compared with blood neutrophils include proangiogenic MMP-8, proangiogenic LRG1 (57), proangiogenic and neutrophil migration promoting MMP-9 (58), the putative chemotactic hormone RETNLG (34), antimicrobial iron-sequestering proteins LTF and LCN2, coagulation cofactor F5, and the hemoglobin subunit HBA2. In addition to neutrophil CXCL2 chemokine production, TNF-α–induced MMP-9 release may also contribute to the pathological IAV-neutrophil feedforward recruitment axis (58). In BAL neutrophils, however, concurrently increased anti-inflammatory IL1RN expression and the release of proangiogenetic factors may serve as counterbalancing attempts to promote resolution and tissue repair. This mechanism may be selectively subverted in chronic inflammatory situations to promote tumorigenesis (59); for instance, if BAL neutrophil release of MMP8, MMP9, and LRG1 exceeds that of the antiangiogenic tissue inhibitor of metalloproteinases 2 (TIMP2). From our proteomics study, an important distinction emerges that it is not solely the production or release of proinflammatory or angiogenetic proteins but rather the balance of inflammatory versus anti-inflammatory or angiogenic versus antiangiogenic proteins that characterize neutrophil function.

It is unsurprising for ISGs such as IFIT1, ISG15, and ISG20 to be further increased in BAL neutrophils following IAV infection, as these proteins function to directly inhibit virus propagation (60). SELENBP1/2, however, are potentially interesting IAV lethality–associated candidate proteins. Following lethal IAV infection, neutrophil SELENBP1/2 expression is only increased after pulmonary microvasculature transmigration and/or alveolar airspace infiltration and occurs in an IAV infection selective and dose-dependent manner. Apart from selenium sequestration, SELENBP1 can convert methanethiol into the reactive oxygen species H₂O₂, formaldehyde, and H₂S (61) and function to suppress tumor growth (62). Although its role during IAV infection is currently unclear, reactive oxygen species generation is a hallmark of activated neutrophils, and it will be interesting to examine whether SELENBP1 contributes to this pathological process. Conflicting reports currently exist regarding whether selenium deficiency beneficially, adversely, or indifferently impacts the course of IAV infection and merit further investigation (63–65). Similarly, it will be interesting to investigate whether altered iron levels, through increased BAL neutrophil LTF release and/or FTH1 and FTL1/2 expression, acts as a nonspecific pathogen response or modulates IAV-induced lung injury.

Single-cell studies have highlighted that blood neutrophil heterogeneity exists in the form of phenotypically distinct cell subsets during inflammation in humans (45) and cancer in mice (31, 66). Such heterogeneity is further compounded by the possibility of reverse neutrophil migration from regions of tissue injury to the vasculature (67). As our bulk cell sorting approach cannot distinguish between a scenario where only a subset of cells is phenotypically altered and one where all cells are altered along a continuous spectrum, it is not known whether the compartment-specific changes observed manifest in all or a major or minor subset of Ly6g^posCD11b^pos mature murine neutrophils. Clarifying this distinction is an important area of future research (31). As single-cell transcriptomics is also limited by the modest correlation between mRNA and protein expression (9, 47), candidate proteins identified in our study may additionally aid Ab selection for the characterization of infection-associated neutrophil heterogeneity using MS-based single-cell profiling.

Altogether, our study provides new insight into the dynamic changes that occur in BM, blood, and BAL neutrophil proteomes following lethal IAV infection. Our findings highlight that the IAV lethality–associated BAL neutrophil proteome is shaped by increased BM neutrophil ISG expression, context-independent blood neutrophil ribosomal protein downregulation, and a limited increase in ISG and metal-sequestering protein expression concurrent to cell degranulation following lung infiltration. A comprehensive understanding of the compartment-specific dynamics of the neutrophil proteome may allow us to develop more targeted neutrophil-centric strategies for attenuating IAV-induced lung injury.

We thank Megan Maher (La Trobe University) and Hugh Harris (University of Adelaide) for discussion of selenium studies. We especially thank Sam Kelly, Nicole Caldera, Sheree Brown, Adam Azzopardi, and LARTF (La Trobe Animal Research and Teaching Facility) for invaluable contributions to animal care, monitoring, and management.

The authors have no financial conflicts of interest.