Decision letter: Dynamic post-translational modification profiling of Mycobacterium tuberculosis-infected primary macrophages

Article Figures and data Abstract Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Macrophages are highly plastic cells with critical roles in immunity, cancer, and tissue homeostasis, but how these distinct cellular fates are triggered by environmental cues is poorly understood. To uncover how primary murine macrophages respond to bacterial pathogens, we globally assessed changes in post-translational modifications of proteins during infection with Mycobacterium tuberculosis, a notorious intracellular pathogen. We identified hundreds of dynamically regulated phosphorylation and ubiquitylation sites, indicating that dramatic remodeling of multiple host pathways, both expected and unexpected, occurred during infection. Most of these cellular changes were not captured by mRNA profiling, and included activation of ubiquitin-mediated autophagy, an evolutionarily ancient cellular antimicrobial system. This analysis also revealed that a particular autophagy receptor, TAX1BP1, mediates clearance of ubiquitylated Mtb and targets bacteria to LC3-positive phagophores. These studies provide a new resource for understanding how macrophages shape their proteome to meet the challenge of infection. Introduction Mycobacterium tuberculosis (Mtb) is among the most successful human pathogens and forms long-term, chronic infections that can span decades. Macrophages are central to tuberculosis pathogenesis as they are both the major site of Mtb replication yet also trigger antimicrobial functions required for host resistance to infection (Russell, 2011). The mechanisms by which Mtb exploits macrophages and the cell-intrinsic immune effectors that limit Mtb replication, as well as how these two properties are balanced during chronic infection, is only partially understood. Uncovering these intimate interactions, which have coevolved over nearly 70,000 years (Comas et al., 2013), may reveal novel therapeutic intervention strategies to treat the nearly ten million people who fall ill to tuberculosis infection each year (World Health Organization, 2018). Mtb infection of macrophages engages several pattern recognition receptors including toll-like receptor 2 (TLR2) leading to expression of inflammatory mediators (Gopalakrishnan and Salgame, 2016). After phagocytosis, the bacterial-containing phagosome enters into the canonical endosomal/lysosomal pathway, which is under the control of Rab GTPases, including Rab5 and Rab7. However, the typical maturation of this compartment is actively blocked by Mtb and fails to acidify or acquire markers of late endosomes/lysosomes (Sturgill-Koszycki et al., 1994), an activity observed in early electron microscopy studies (Armstrong and Hart, 1971). While this effect of virulent Mtb on host intracellular vesicle trafficking requires the type VII protein secretion system ESX-1 (MacGurn and Cox, 2007), the mechanism of phagosomal maturation arrest remains mysterious. Importantly, the ESX-1 system mediates limited perforation of the phagosomal membrane, which in turn activates two cytosolic pathways in host cells (Watson et al., 2015; Watson et al., 2012; Manzanillo et al., 2012). First, activation of the cGAS/STING/TBK1 signal transduction cascade leads to production of type I IFN and a profound antiviral transcriptional response that inhibits host resistance (Watson et al., 2015). Second, cytosolic access by Mtb also activates ubiquitin-mediated selective autophagy targeting, an evolutionary ancient anti-microbial cellular response that counteracts phagosome maturation arrest by actively targeting microbes to lysosomes (Via et al., 1997; Choy and Roy, 2013). Although some components of the host autophagic machinery are critical for controlling infection (Watson et al., 2012; Zheng et al., 2009), it remains unclear whether this antimicrobial effect is dependent on canonical autophagy (Kimmey et al., 2015). Global, unbiased approaches to probe the Mtb-macrophage interface have primarily relied on measuring changes in mRNA levels that, while facile, naturally limit the analysis of cellular state toward signal transduction cascades that directly lead to transcriptional outputs (Ehrt et al., 2001). While these studies have certainly uncovered insights into the mechanisms of Mtb pathogenesis (Ehrt et al., 2001; Berry et al., 2010), monitoring changes in protein post-translational modifications (PTMs) represents a more comprehensive way to assess changes in cellular state during infection, as essentially all cell biological pathways, including intracellular trafficking, autophagy, nuclear import, and metabolism are regulated by PTMs. Indeed, many intracellular bacterial pathogens hijack or alter normal PTMs of host proteins to manipulate cells and promote their pathogenesis (Roy and Mukherjee, 2009; Patel et al., 2009; Fiskin et al., 2016). However, global PTM analyses require proteomics-based assays that are inherently more difficult than measuring mRNA levels. Selective autophagy is predominantly regulated by PTMs rather than transcription (Lamark et al., 2017), as it is under the control of several types of modification, especially ubiquitylation and phosphorylation (Herhaus and Dikic, 2015). Initial targeting of intracellular structures (intracellular pathogens, damaged organelles, protein aggregates, etc.) to autophagy is mediated by ubiquitylation of the cargo. Autophagy receptors then recognize these ubiquitin signals and recruit phagophore membranes to the cargo via interactions with LC3/GABARAP proteins embedded in these vesicles, ultimately enveloping cargo inside a continuous autophagic vacuole competent for fusion with lysosomes. Importantly, PTMs of receptors also govern their activity. For example, ubiquitylation of p62 and phosphorylation of Optineurin increase their activity upon recruitment to cargo, forming a positive feedback loop promoting autophagosome completion (Peng et al., 2017; Matsumoto et al., 2011; Heo et al., 2015). In the case of Mtb, some ubiquitin ligases required for selective autophagy have been identified (Manzanillo et al., 2013; Franco et al., 2017), but their substrates and roles are still mysterious and other ubiquitin ligases are likely required. Likewise, some autophagy receptors have been implicated in autophagy of the Mtb-containing vacuole (Watson et al., 2012; Manzanillo et al., 2013; Franco et al., 2017), but many have not been tested (Rogov et al., 2014). Thus, global PTM analysis would be likely to reveal components of autophagy involved in Mtb infection (Sarraf et al., 2013). Here, we report our findings using mass spectrometry (MS) to comprehensively identify changes in host protein abundance, phosphorylation, and ubiquitylation during a 24 hr time course of primary murine macrophages infected with virulent Mtb. These datasets represent a significant resource for future studies and expands upon previously published PTM studies at a single time point in the RAW macrophage cell line (Penn et al., 2018) and of tyrosine phosphorylation in primary macrophages (Sogi et al., 2017), and has uncovered thousands of significantly modified proteins in response to infection. Bioinformatic analysis indicated that while some PTM changes were enriched in pathways known to be important in Mtb pathogenesis (e.g. autophagy), other pathways identified were surprising (e.g. nucleosome assembly). In particular, we identified significant enrichment in the phosphorylation of several autophagy receptors, which guided directed studies that revealed a unique role for TAX1BP1 in cell intrinsic control of Mtb infection. Collectively, these findings indicate that our MS/genetics approach is a powerful way to identify the ways by which macrophages attempt to control intracellular bacterial infection. Results Proteome-level evaluation of primary macrophage responses to Mtb infection To identify new innate immune pathways modulated during Mtb infection, we sought to obtain a deep data set of changes in host protein abundance and post-translational modifications during a time course of macrophage infection. To this end, we infected primary murine bone-marrow-derived macrophages with Mtb in biological triplicate, harvested infected cells at 2-, 4-, 6-, 8-, and 24 hr post-infection, and prepared protein lysates (Figure 1A). We also performed time-matched mock infections of the same macrophages harvested at an early (0 hr), middle (6 hr), or late (24 hr) time points. Because we identified extremely few changes between uninfected cells at all three time points, we simply matched the Mtb-infected cell data from the 2- and 4 hr time points with the data from the 0 hr mock sample, and the 6- and 8 hr time points with the data from the 6 hr mock sample (Figure 1—figure supplement 1A–C). Cellular lysate samples were digested with trypsin and a portion of the resulting peptides were used for measurement of protein abundance, with the remaining peptides subjected to separate enrichments using phospho-peptide (Swaney and Villén, 2016) or diGly remnant (Udeshi et al., 2012) affinity technologies. The diGly-MS approach, however, cannot distinguish ubiquitin modification from modification with either of two different ubiquitin-like proteins, ISG15 and Nedd8, as they leave identical diGly remnants after trypsinization. Thus, although we will refer to diGly remnant peptides as evidence of 'ubiquitylation', it is important to note the possibility that these events represent these rarer modifications. We subjected all peptide samples (abundance, phospho- and diGly-modified peptides) to liquid chromatography-mass spectrometry (LC-MS). Based on the measured m/z ratio of the parent ions and fragment ions for each peptide coupled with a database search of the mouse proteome, we determined the amino acid sequence, modification sites, and cognate protein for each peptide. Pair-wise comparison of parent ion MS intensity measurements between technical and biological replicates from the same condition were highly reproducible, as indicated by strong correlation coefficients from global abundance, phospho-enriched, or diGly-enriched samples (Supplementary file 1; Figure 1B). Figure 1 with 4 supplements see all Download asset Open asset Proteomics and RNAseq profiling of Mtb infected bone-marrow-derived macrophages. (A) Schematic for the experimental design indicating the hours post-infection at which point mock-infected or Mtb infected macrophages were harvested and peptides were generated for global protein abundance measurements by LC-MS. Phosphorylated peptides (Ph) or ubiquitylated peptides containing the diGly-remnant (GG) were separately enriched. (B) Replica plots of MS intensity measurements for individual peptides in global protein abundance (AB), phosphopeptide (PH), or ubiquitylated peptide (UB) samples. The correlation coefficient is displayed (ρ). (C) Percent of total changes (log2(fold change) greater than one or less than −1, p value less than 0.05) occurring at each time point 2–24 hr post-infection for global protein abundance, phosphorylation, or ubiquitylation. (D–F) Correlation plot of changes in gene transcription and protein abundance (AB; D), phosphorylation (PH; E), or ubiquitylation (UB; F) at 24 hr post-infection with Mtb. Proteins with no statistically significant changes during Mtb infection are colored black. Proteins with statistically significant changes during Mtb infection are colored blue. The correlation coefficient (ρ) for proteins with statistically significant changes during Mtb infection and mRNA levels are shown. Multiple phosphorylation sites were detected in Gpr84, Fcgr1, Ms4a6d, Vim, Map7d1, Mrc1, Tk1, and Mki67 (Supplementary file 5). To determine the statistically significant changes in response to infection, we analyzed the combined set of biological replicates with the in-house Bioconductor package artMS (Jimenez-Morales et al., 2019) to determine the proteins with the most significant changes in abundance and PTMs upon infection (p<0.05 and log2-fold-change≥1 or ≤ −1). Importantly, in many cases, we identified significant peptide counts in one experimental condition (e.g. infected) but no corresponding peptides in the cognate control (e.g. uninfected or mock), indicating that these changes were likely biologically relevant, but could not be represented by a mathematical ratio between infected and uninfected due to zero values. To address this issue, we used a method described by Waters et al. in which missing values are imputed by the limit of MS1 detection to calculate an 'imputed value' (Webb-Robertson et al., 2015), which allowed us to estimate changes in levels for this category of peptides. The peptides from the global protein abundance or PTM enrichments mapped to thousands of unique proteins (Figure 1—figure supplement 1D–F), consistent with deep coverage of the proteome relative to prior Mtb proteomics studies (Hoffmann et al., 2018). Comparison of peptide levels in Mtb infected samples versus uninfected controls revealed over 1000 statistically significant changes in abundance or PTMs during Mtb infection (Figure 1—figure supplements 2–3; Figure 2C-G), indicating pronounced changes to the macrophage proteome. Figure 2 Download asset Open asset Autophagy receptor phosphorylation during Mtb infection. (A) Table showing the number of unique phosphosites in each autophagy receptor at 2–24 hr post-infection. * p value less than 0.05 and log2(fold change) greater than one or less than −1 for at least one of the non-imputed phosphosites. (B) Domain organization displaying post-translational modifications in autophagy receptors. The LC3-interacting region (LIR), ubiquitin binding domain (UBAN, UBZ), myosin-6 binding domain (MYO6), SKIP carboxyl homology domain (SKICH), and TANK-binding kinase-1 (TBK1) binding domains are labeled. (C–G) Volcano plots highlighting changes in autophagy receptor phosphorylation at each time point 2–24 hr post-Mtb infection. Proteins with a log2(fold change) greater than one are colored red. Proteins with a log2(fold change) less than −1 are colored blue. Proteins with a p value less than 0.05 (or -log10 (p value) greater than 1.3) are above the horizontal black line. Phosphorylated residues and ubiquitylated lysine residues are noted. Statistical analysis of phosphopeptide data was performed with artMS version 1.4.0. The number of statistically significant changes in protein abundance or ubiquitylation progressively increased over the time course (Figure 1C). In contrast, phosphopeptide changes were bimodal with the greatest number of phosphopeptide changes occurring early (2- or 4 hr post-infection) or late (24 hr) after Mtb infection (Figure 1C). Although protein abundance levels are largely determined by mRNA levels under steady state conditions (Edfors et al., 2016), environmental changes can result in a transient lack of correlation between transcript and protein levels (Liu et al., 2016). However, after cells adapt to their new physiological state, the correlation between mRNA and protein levels is restored (Liu et al., 2016). To determine the extent of correlation between the relative changes in RNA and protein abundance after Mtb infection, we compared transcriptomics datasets from previous experiments using the same macrophages and Mtb strain (Braverman et al., 2016) with our proteomic measurements. Overall, our data is consistent with previous findings of dendritic cells treated with LPS (Berg-Larsen et al., 2013), as we observed weak correlation between relative protein and RNA abundance early after infection (6 hr time point, Pearson coefficient ρ = 0.36, Figure 1—figure supplement 4), but at 24 hr post-infection the correlation was much stronger (ρ = 0.74, Figure 1D) indicating that by this time point the changes in the proteome are largely reflective of changes in the transcriptome. Importantly, this correlation was similar for both non-imputed (ρ = 0.81) and imputed (ρ = 0.78) proteins, indicating that the imputation method accurately captured abundance changes. Interestingly, we observed poor correlations between relative changes in mRNA compared to changes in phosphorylated (ρ = 0.37, Figure 1E) or ubiquitylated proteins (ρ = 0.52, Figure 1F) at this same late time point. Thus, these results suggest that while changes in RNA abundance is a moderate predictor for changes in protein abundance in response to infection, other biological process controlled by PTMs are at play during macrophage infection, reinforcing the notion that proteomics is a valuable method to identify biological processes during infection that are independent of gene expression. Mtb infection elicits unique and overlapping changes in protein abundance, phosphorylation, and ubiquitylation Next, we sought to compare the proteins that changed in abundance, phosphorylation, or ubiquitylation during Mtb infection. A complete list of changes in macrophage global protein abundance, phosphopeptide, and diGly-remnant peptides at 2–24 hr post-infection with Mtb are found in Supplementary files 4–6. To visualize the differentially modified proteins (global abundance) or peptide (PTMs), we plotted the log2-fold change in each protein or peptide versus its statistical significance during Mtb infection (-log10(p value); Figures 2A,3G, and Figure 1—figure supplements 2–3). Non-imputed proteins or peptides with p values < 0.05 and a log2-fold change >1 or<1 were deemed statistically significant (Figure 3A). Imputed proteins or peptides were conservatively assigned a p value just above the level of statistical significance and thus these values are plotted immediately above the horizontal line demarcating the cut-off p value of 0.05 (-log10(p value)=1.3; Figure 3A). Many of the proteins that changed in global abundance and ubiquitylation, but to a lesser degree in their phosphorylation levels, are encoded by interferon-regulated genes such as Ifit1, Ifit2, Stat1, or Ifih1 (Figure 3A C). A comparison of individual proteins that changed significantly in abundance (n = 470), phosphorylation (n = 1489), or ubiquitylation (n = 606) during Mtb infection, indicated limited overlap between the three categories (Figure 3B). Figure 3 with 5 supplements see all Download asset Open asset Comparison of macrophage proteins changing in abundance, phosphorylation, or ubiquitylation during Mtb infection. (A) Volcano plots displaying proteins changing in abundance, phosphorylation, or ubiquitylation at 24 hr post-infection. Proteins with a log2(fold change) greater than one are colored red. Proteins with a log2(fold change) less than −1 are colored blue. Proteins with a p value less than 0.05 (or -log10(p value) greater than 1.3) are above the horizontal black line. Phosphorylated residues and ubiquitylated lysine residues are noted. (B) Venn diagram displaying the number of unique and overlapping proteins changing in abundance (AB), phosphorylation (PH), or ubiquitylation (UB) in the aggregate measurements 2–24 hr post-infection. (C) Enriched ontogeny clusters highlighting commonly enriched pathways. Pathways are colored coded into common groups (trafficking, metabolism, signal transduction, immunity). The dendrogram displays clustering of the gene ontogeny pathways. To identify distinct pathways that may be under the control of phosphorylation or ubiquitylation (Ashburner et al., 2000), we performed functional pathway enrichment analysis with the bioinformatics pipeline, Metascape (Tripathi et al., 2015). Functional pathways involving immunity, cellular trafficking, metabolism, and signal transduction were among the top 20 most statistically significant enriched pathways (Figure 3C). As expected based on previous literature, antiviral responses were highly enriched (p=4.53×10−20 in the global abundance dataset; Figure 3C), including response to type I (IFN-β p=4.25×10−16, and IFN-α 3.81 × 10−10 in the global abundance dataset; Figure 3C), likely due to phagosome perforation and activation of the cGAS/STING/TBK1 pathway (Flynn et al., 1993; Cooper et al., 1993; Moreira-Teixeira et al., 2018). Small GTPase mediated signal transduction was also highly enriched in the phosphopeptide datasets, which included Rab GTPases (Rab7, 8b, and 14), guanine nucleotide exchange factors (RabGef1, ArfGef1-2, ArhGef2, Dock1, 2, 5, 8, and 11), and GTPase activators (ArfGap 1, 2, 3, 12, 17, 18, 24, 25, 30, 35) (p=2.12×10−26; Figure 3C). As Rab GTPases are involved in endosomal trafficking, and GAPs and GEFs regulate membrane transport, phagocytosis, and control the actin cytoskeleton, enrichment of these functional categories indicates profound reorganization of intracellular trafficking upon infection with Mtb. We identified an enrichment in modified proteins in the category 'regulation of catabolic processes' (p=2.39×10−25 in the phosphopeptide dataset; Figure 3C), which includes components of the autophagy machinery such as Atg4b, Rubicon, autophagy kinases (Ulk1, Ulk2), and autophagy receptors (Optineurin, Bnip3l). We identified other enriched functional pathways not previously implicated in Mtb pathogenesis such as nucleocytoplasmic and response to Figure and Figure supplement Thus, our unbiased proteomic approach components of pathways known to be important in the host response to Mtb and identified new pathways that may play roles in the of macrophages the functional enrichment analysis the of measuring changes in the proteome using separate enrichment one to unique host responses to infection that measuring mRNA levels not in protein abundance reveal enrichment of antiviral and inflammatory pathways during tuberculosis infection The above analysis was performed by of data over the time course of infection. Next, we sought to from our data the of proteome response by using analysis and the of changes over the individual time points each of the three data (Figure supplements analysis was performed on the imputed and non-imputed proteins (Figure supplements abundance changes into two groups and several clusters (Figure supplement The of proteins that increased in abundance during Mtb infection was enriched for and to they are predominantly targets of Figure supplement × × the antiviral the proteins and increased in protein abundance at and 24 hr post-infection, and proteins and Supplementary file increased at 24 hr post-infection. among all of the non-imputed proteins that changed in abundance, the greatest change (Figure and was also a increase in mRNA at the same time point (Figure While proteins of acid et al., 2015), murine the type I IFN response by binding to and of mRNA et al., 2013). of the type I IFN response may play a critical role in Mtb pathogenesis as mRNA was described to be increased in macrophages infected with a Mtb strain et al., The of proteins in abundance and included and replication pathways (Figure supplement × consistent with studies indicating that Mtb infection may arrest macrophage et al., 2017), a by the intracellular et al., et al., Although the autophagy pathway was not significantly enriched in our abundance data (Figure supplement was a statistically significant increase in the autophagy receptors p62 and TAX1BP1 during Mtb infection (Supplementary file PTM changes during tuberculosis infection are enriched for immune autophagy, and cellular trafficking the 24 hr time ubiquitylated substrates into that in ubiquitylation during infection, that increased ubiquitylation early hr) after infection, that were ubiquitylated between hr after infection, and that were identified as ubiquitylated only at the 24 hr time point (Figure supplement for example, was enriched at the time points of infection 3 and Figure supplement We also identified an enrichment in among the of proteins with ubiquitylation during Mtb infection (Figure supplement 1, × which is likely reflective of the dramatic changes in the transcriptional of macrophages they bacteria and 2014). Phosphorylated substrates also into Proteins that increased early after infection, Proteins that increased after infection, Proteins that early after infection, and Proteins that late after infection (Figure supplement This analysis revealed that actin organization and GTPase mediated signal transduction changed the 24 hr time course of infection as they were significantly enriched in all of the clusters (Figure supplement several functional categories were only enriched in the early clusters 1 and cytosolic transport, of or in the clusters 2 and enrichment of receptor tyrosine pathways is consistent with the role for in Mtb macrophages (Sogi et al., 2017; et al., 2014). 2 and 4 were enriched for responses to and organization (Figure supplement which we could be to the dramatic transcriptional of macrophages and ubiquitylation observed at the same time points, as above 1; Figure supplement this data indicates that phosphorylation mediates in cellular trafficking pathways and during the host response to Mtb infection. of and ubiquitin that may change during tuberculosis infection Although the functional significance of the phosphorylation changes we observed during Mtb infection we sought to our data to kinases that may be during Mtb infection. To this end, we used the to compare the differentially phosphorylated substrates we identified during Mtb infection with datasets that phosphorylation changes determined under a of experimental et al., 2016). identified in this way can be used to from proteomic data and lead to that can be tested Importantly, this approach identified kinases expected to be during Mtb infection, including involved in inflammatory and responses as well as type I IFN we also activity of several dependent consistent with the that primary macrophages upon infection et al., 2017; Figure supplement Interestingly, several kinases were identified at the including and that these may regulate previously pathways that macrophage responses to infection. with kinases and known for their role in cellular responses et al., 2014). with proteins and and a cascade that respond to damaged and at the et al., and et al., Indeed, activation of has been found to promote macrophage a of indicating that this cascade play an important role in of chronic infection et al., 2018). this analysis also revealed of the cascade that actin an effect that has not been described during Mtb infection (Figure supplement et al., et al., 2001). major of actin was during infection of macrophages with a Mycobacterium and that virulent may changes in actin to host To ubiquitin ligases involved in host responses to