SON DNA-Binding Protein mediates macrophage autophagy and responses to intracellular infection

Abstract

Intracellular pathogens affect diverse host cellular defence and metabolic pathways. Here, we used infection with Francisella tularensis to identify SON DNA binding protein as a central determinant of macrophage activities. RNAi knockdown of SON increases survival of human macrophages following F. tularensis infection or inflammasome stimulation. SON is required for macrophage autophagy, interferon response factor 3 (IRF3) expression, type I interferon response, and inflammasome-associated readouts. SON knockdown has gene- and stimulus- specific effects on inflammatory gene expression. SON is required for accurate splicing and expression of GBF1, a key mediator of cis-Golgi structure and function. Chemical GBF1 inhibition has similar effects to SON knockdown, suggesting that SON controls macrophage functions at least in part by controlling Golgi-associated processes.

Introduction

Pathogens that invade and replicate within host cells alter diverse cellular processes, including metabolism, gene expression, and cell-type-specific functions. They also interact with host defences – principally cell autonomous immunity but also aspects of the wider innate and adaptive immune systems. As such, they provide powerful tools for studying the regulation and significance of many cellular processes. One such pathogen is Francisella tularensis, a Gram-negative coccobacillus that is highly adept at invading and replicating within macrophages and other cells. Clinically, infection with F. tularensis leads to tularemia, with symptoms and outcomes that vary depending on the route of inoculation ¹. Untreated mortality following pulmonary infection has been estimated over 50%, and this, together with its rapidity of onset, difficulty of diagnosis, and history of development as a biological weapon during the Cold War, makes the deliberate release of F. tularensis a topic of significant concern. At the molecular and cellular level, F. tularensis provides a valuable model because of its interaction with key metabolic and cell-autonomous defence pathways, including autophagy, type I interferon, and inflammasomes.

Autophagy is an evolutionarily ancient process that culminates in lysosomal degradation of cytoplasmic targets. Defective autophagy is linked to a wide range of diseases including neurodegeneration, cancers, and metabolic disorders, as well as infectious disease ². F. tularensis infection has been shown to activate autophagy, but some authors have concluded that this is a host defence mechanism that enables the host cells to eliminate the bacteria ^3,4, while others have shown that autophagy promotes F. tularensis replication, apparently by improving access to host metabolites ^5,6. These contrasting reports may indicate that the effects of inducing autophagy depend on specific contexts such as host cell type and activation. Better understanding of the regulatory mechanisms triggered by, and perhaps subverted by, Francisella may therefore shed light on different physiological roles of autophagy.

Type I interferon secretion and inflammasome activation are key players in the host response to viral and intracellular pathogens. An extensive series of studies has demonstrated the interdependence of these two processes in mouse bone marrow-derived macrophages (BMDMs) pretreated with LPS and infected with the closely related bacterium, F. novicida. In this model, detection of the cytosolic bacteria by the cGAS-STING system leads to IRF3-dependent induction and release of IFNβ. This then enables activation of the AIM2 inflammasome, resulting in release of IL-1β and IL18, and pyroptotic cell death. ^7–16 Studies using human macrophages or different Francisella strains confirm inflammasome activation, but show that pyrin ¹⁷ or NLRP3 ¹⁸ are more important than AIM2 in different situations. Francisella infection therefore provides a valuable model for studying the mechanisms by which both the type I interferon response and inflammasomes are regulated, as well as their functional importance, and how the two systems interact.

SON DNA-Binding Protein (SON) is a large, multifunctional protein that has been studied primarily for its roles in cancer and development. Mutations have been associated with susceptibility to medulloblastoma and a number of developmental syndromes ^19–22. In vitro studies have shown SON to be important for cell cycle progression and maintenance of embryonic stem cell pluripotency, while knockdown ultimately leads to apoptosis ^20,23–28. At the molecular level, SON contains both DNA and RNA binding domains and can control both mRNA transcription and splicing ^23,25,29,30. Furthermore, binding of SON to promoters of leukemia-associated genes is associated with histone methylation and reduced gene expression ²⁰. Knockdown of SON expression in HeLa and embryonic stem cells has been shown to affect both splicing and abundance of a large number of mRNAs ^23,24; it is not yet possible to say to what extent SON affects mRNA abundance directly by regulating transcription, or indirectly, via nonsense mediated mRNA decay following intron retention or exon exclusion. Two studies have proposed that SON may contribute to defence against viral infection: SON protein can bind to and repress transcription from the hepatitis B virus promoter in vitro ³⁰, and knockdown of SON gene expression in macrophages restricts influenza virus replication, apparently by interfering with virus trafficking ³¹. However, a functional relationship, if any, between these observations and wider roles of SON in host defence have not, to our knowledge, been investigated. Nor has SON been studied in the context of bacterial infection.

In this study, an RNAi screen led as to address the role of SON in macrophages during F. tularensis infection. We found that SON regulates inflammasome expression, IFNβ induction, and autophagy, and that the Golgi body may link these different processes.

Results

SON promotes macrophage susceptibility to killing by F. tularensis.

We identified SON in an RNAi screen to identify host genes that enabled infection and killing of macrophages by F. tularensis. As summarised in Figure 1A, a custom-designed, lentivirus encoded shRNA library was prepared against 7,137 human genes of particular relevance to macrophage-pathogen interaction, which were identified using a bioinformatic approach (see methods). Each gene was targeted by 7–8 different shRNAs. The library was expressed in PMA-differentiated THP1 macrophage-like cells which were then infected with virulent F. tularensis SchuS4 under conditions that resulted in ~90% death. The distribution of shRNAs in surviving cells was compared to that in mock-infected cells, with the expectation that shRNAs which protect macrophages from killing would be enriched in survivors. As shown in Figure 1B and Supplemental Dataset 1, survival selection significantly (q<0.1, Fisher’s exact score, Benjamini-Hochberg FDR correction) altered the distributions of 780 shRNAs from a total of 52,689. Of these, 344 were enriched >2 fold in surviving cells, targeting 317 different genes. Independent targeting of a gene by multiple shRNAs increases confidence that the gene is important during infection. As shown in Figure 1B, three genes were each targeted independently by three significantly enriched shRNAs – a stringent criterion.

Figure 1:

RNAi screen identified SON as a susceptibility factor for death of macrophages infected with F. tularensis.

A) Schematic of screen. A custom-designed shRNA library was transduced into THP1 cells. These were then differentiated into a macrophage-like phenotype with PMA and shRNAs that altered host resistance selected by infection with highly virulent F. tularensis SchuS4. Genomic DNA was isolated from the survivors and sequenced to identify encoded shRNAs. shRNAs that were significantly enriched in survivors of F. tularensis infection were identified by comparison with abundance in a mock-infected library then matched to the genes they target.

B) shRNAs enriched in cells that survived infection. Each cross refers to an individual shRNA, mean enrichment across three macrophage donors . X-axis: mean abundance of each shRNA in library-expressing THP-1 macrophages that have survived infection with F. tularensis SchuS4, relative to abundance in mock infected library cells. Y-axis: BH-adjusted p values. Coloured circles indicate shRNAs that target SON (red), EIF3CL (blue), and ADAR (yellow).

C) AM-MDMs were transfected with a pool of 4 siRNAs to knock down expression of the indicated genes (SON, EIF3CL, ADAR) or non-targeting siRNA control (NT), then infected for 48h with F. tularensis SchuS4 at10 bacteria per macrophage (MOI 10). Macrophage mortality was measured by lactate dehydrogenase (LDH) release, and expressed relative to NT transfected, uninfected cells. Mean ± SD, macrophages from 3 different donors. **: p<0.01 relative to NT, 2-way ANOVA with Tukey HSD.

D) AM-MDMs were transfected with individual siRNAs against SON (A-D), the pool of four siRNAs, or non-targeting control (NT) then infected with F. tularensis LVS overnight. Macrophage mortality was measured by LDH release. Mean ±SEM, n=3. *: p<0.05, **: p<0.01, ***: p<0.001, relative to NT, 2-way ANOVA with Tukey HSD.

E) AM-MDMs were transfected with a pool of 4 siRNAs against SON or non-targeting siRNA control (NT), then infected for 3 days spores of B. anthracis ANR-1 at an average MOI of 3.5. Macrophage mortality was measured by LDH release. Mean ±SEM, n=3.

To ensure that our conclusions are not restricted to a specific cell line, we used an siRNA-transfected model of human primary alveolar macrophage-like cells (AM-MDMs) ³² instead of shRNA-transduced THP-1 cells in all validation experiments. As shown in Figure 1C, knockdown of SON also protected AM-MDMs from killing by F. tularensis SchuS4, but knockdown of ADAR was toxic and EIF3CL knockdown conferred no survival benefit in this system. We therefore focused on SON. Figure 1D shows that transfection of siRNAs against SON also protects AM-MDMs from killing by a more experimentally tractable strain of F. tularensis, Live Vaccine Strain (LVS). Protection was observed with AM-MDMs transfected with a mixture of 4 siRNAs against SON (“pool”, as used in all other experiments), or with each siRNA individually. Death in control-transfected cells occurred more rapidly with LVS than SchuS4, indicating differences in the infection kinetics or host response to the different strains. Furthermore, SON knockdown also protected AM-MDMs from death following infection with Bacillus anthracis spores (Figure 1E), although to a substantially smaller extent than F. tularensis. Interestingly, death in this system only occurs 72 hours after infection, by which time no spores or bacilli are visible (data not shown), suggesting that SON-dependent death is secondary to B. anthracis infection. Nonetheless, we conclude that SON expression contributes to infection and/or death of macrophages in response to F. tularensis SchuS4, F. tularensis LVS, and B. anthracis ANR-1. We therefore investigated the roles played by SON in regulating different aspects of macrophage response to infection and stimulation.

Effect of SON knockdown on macrophage gene and protein expression

Because of SON’s documented role in regulation of gene expression, we studied the effect of SON knockdown on expression of a panel of host defence genes. IFIT2, CXCL9 and CXCL10 are markers of a response to type I interferons such as IFNβ, (encoded by IFNB1). Microbial challenge, including F. tularensis infection, also leads to TLR dependent induction of IL1B and TNF and release of interleukin 1β (IL-1β) and Tumour Necrosis Factor α ^33,34. To address the role of SON in the broader innate immune response, we studied the response of these genes to overnight treatment with type II interferon (IFNγ) or the microbial TLR4 ligand LPS, as well as to F. tularensis LVS infection, which stimulates TLR2 among other receptors ^35,36.

As shown in Figure 2A–D, SON knockdown prevented induction of IFIT2 and TNF by all stimuli. This did not reflect a general transcriptional defect since induction of CXCL9 or CXCL10 by IFNγ was not significantly affected. However, inhibition of IL1B and CXCL9, although statistically significant, was partial: these genes were still induced approximately 130- and 15-fold, respectively, on LPS stimulation. We conclude that SON regulates macrophage gene expression in a gene- and stimulus-specific manner. In particular, SON is required for the induction of TNF by all three stimuli, is required for induction of IFNB1 and type I interferon responsive CXCL10 and IFIT2 by F. tularensis and LPS stimulation, but is otherwise largely dispensable for the response to IFNγ.

Figure 2:

Knockdown of SON specifically affects gene and protein expression.

AM-MDMs were transfected with non-targeting control (NT) or SON siRNAs (SONkd) then (A) mock infected, (B) treated with 50ng/ml E. coli LPS, or (C) 100 U/ml IFNγ, or (D) infected with F. tularensis LVS overnight prior to quantitation of the indicated mRNAs (CXCL9, CXCL10, IFIT10, IFNB1, IL1B, TNF, SON) by qPCR. Mean ± SEM of 4 independent experiments. p<0.05 (*), <0.01(**), <0.001(***), SONkd vs NT, 2-way ANOVA with Tukey HSD. (E) Transfected AM-MDMs were mock infected or infected overnight with F. tularensis LVS then lysed and subjected to TMT proteomic analysis. Each point corresponds to an individual protein. Axes indicate the relative abundance of each protein in lysates of SON knockdown or non-targeting control transfected cells. Circles indicate selected interferon regulated proteins, triangles STAT family members, and squares proteins implicated in vesicle shuttling.

We next used tandem mass tag proteomics for an unbiased assessment of the role of SON in macrophage protein expression. AM-MDMs were transfected with SON siRNA or a non-targeting control (NT), and then mock-infected or infected overnight with F. tularensis LVS. A total of 3,545 proteins were detected, of which 198 (5.5%) showed at least a 1.5-fold change in expression following SON knockdown, either with or without subsequent infection (Supplemental dataset 2). Among these, three groups are evident: proteins associated with vesicle trafficking, type I IFN-responsive proteins, and STAT transcription factor family members (Figure 2E). We investigated the contribution of each these groups to modulation of macrophage intracellular defence mechanisms by SON.

SON is necessary for autophagy

The protein showing the strongest fold change on SON knockdown was phosphofuran acid cluster sorting 1, which is associated with retrograde vesicle transport from the trans-Golgi network to the endoplasmic reticulum ³⁷. Other proteins associated with vesicle trafficking included NDRG1 (N-myc downstream regulated gene 1), SASH3 (SAM and SH3 domain containing 3), GBF1 (Golgi Resistance Factor 1), and the autophagy marker LC3. Given prior studies indicating roles for autophagy during Francisella infection, we focused on the increase in LC3 abundance, and investigated whether SON knockdown induced autophagy. Consistent with the proteomic results, immunofluorescence staining showed accumulation of LC3-positive cytoplasmic bodies (Figure 3A). However, this may indicate either induction of autophagy or disruption of flux through the pathway ³⁸. LC3 staining following SON knockdown was noticeably stronger than was evident in AM-MDMs treated with the classical inducer of autophagy, rapamycin, and its morphology was more reminiscent of cells treated with vinblastine, which prevents autophagosome maturation by disrupting microtubules. Moreover, Western blotting showed that, unlike rapamycin treatment, SON knockdown caused reduced expression of the upstream autophagy activators raptor and ULK1, with a consequential reduction in both activating (p-ser555) and inhibitory (p-ser757) ULK1 phosphorylation (Figure 3B). At the RNA level, infection with Francisella or treatment with LPS but not IFNγ induced LC3 gene expression, which is consistent with primed or induced autophagic flux, but SON knockdown strongly inhibited LC3 expression regardless of stimulus (Figure 3C). Both of these lines of evidence suggest that LC3 protein accumulation causes negative feedback on LC3 gene and upstream protein expression, implying that the accumulation results from improper completion of autophagy rather than reflecting increased flux. Finally, successful completion of autophagy results in the degradation of the carrier protein p62/SQSTM1 ³⁸; SON knockdown instead increased p62/SQSTM1 abundance (Figure 3D). While we cannot exclude effects on non-canonical LC3 activities such as exocytosis ³⁸, these experiments, taken together, indicate that SON knockdown does not induce increased autophagy but instead that SON expression is necessary for successful completion of autophagic flux, and knockdown results in accumulation of autophagosomes or partially formed autophagophores.

Figure 3:

Effect of SON knockdown on autophagy.

A) AM-MDMs were transfected with siRNAs against SON (SONkd) or non-targeting control siRNAs (NT, rapamycin, vinblastine) then treated with vehicle (NT, SONkd), 50μM rapamycin or 50 μM vinblastine for 1 hr, prior to immunofluorescence for SON (red) and LC3 (yellow). Grey: Hoechst 33342. Representative of 3 independent experiments.

B) & D) AM-MDMs were transfected as above and extracts were Western blotted for the indicated proteins. Each panel is representative of 3 independent experiments.

C) Transfected cells were mock infected followed by no further treatment (Nil), 50 ng/ml E. coli LPS, or 100 U/ml IFNγ, or infected with F. tularensis LVS (Ft), all overnight, as indicated on the x-axis. LC3 mRNA quantified by qPCR. Mean ±SEM, n=3. (***) p<0.001 2-way ANOVA with Tukey’s HSD, NT vs. SONkd.

Golgi inhibition mimics SON knockdown phenotype

A number of cellular membranes, notably Golgi body, endoplasmic reticulum, and plasma membrane, can contribute the lipid bilayers that make up the autophagic isolation membranes ³⁹. It was therefore striking that the proteomic data showed that GBF1 expression is reduced following SON knockdown. GBF1 is both a marker of the cis-Golgi and a critical enzyme necessary for maintaining Golgi function and integrity. Inhibition of GBF1’s guanine nucleotide transferase activity with either golgicide A (GCA) is known to cause Golgi collapse ⁴⁰. We confirmed reduced expression following SON knockdown by Western blot (Figure 4A). Immunofluorescence staining showed three things (Figure 4B): i) dispersal of GBF1 signalling following staining GCA treatment, consistent with the expected Golgi collapse; ii) reduction in overall GBF1 staining following SON knockdown, consistent with reduced expression; and iii) accumulation of LC3 positive bodies in the cytoplasm following either GCA treatment or SON knockdown, suggesting similar effects on autophagy.

Figure 4:

Relationship between SON knockdown and Golgi body.

A) AM-MDMs were transfected with non-targeting control (NT) or SON siRNAs (SONkd) and Western blot used to determine GBF1 expression without (−) or after infection with F. tularensis LVS (Ft), and treatment with 5μM Golgicide A (GCA) or vehicle. Representative of 3 independent experiments.

B) AM-MDMs were transfected with non-targeting siRNA (NT, GCA) or siRNA against SON and treated overnight with 5μM GCA or vehicle. LC3 (yellow) and GBF1 (green) were detected by immunofluorescence. Grey: Hoechst 33342. Representative of 4 independent experiments.

C) Mis-splicing of GBF1 detected by qPCR. AM-MDMs were transfected and infected as above, and transcripts detected using primers located on exons 29–30 (Total), exon 36, or intron 36 as indicated. Amount of each transcript is expressed relative to control, uninfected cells. Mean ± SEM, n=3. (***) p<0.01, 2-way ANOVA with Tukey’s HSD, NT- vs. SONkd-.

D) AM-MDMs were treated as in (B) and cis-Golgi marker GM130 (cyan) and SON (red) distribution visualized by immunofluorescence. Grey: Hoechst 33342. Representative of 4 independent experiments.

E) Median GM130-postive Golgi area per cell was determined by automated analysis of the same experiments from which images in (D) are taken. Mean ±S.E.M., paired t-test, n=4, ≥300 cells quantified per well.

A global study of the role of SON in mRNA splicing in embryonic stem cells ²⁴ has identified defects towards the 3’-end of GBF1 mRNA, namely skipping of exon 36 and inclusion of introns 35 and 36. By using exon specific qPCR primers, we found that SON knockdown caused a ~70% reduction in GBF1 mRNA (Figure 4C). However, exon 36 expression was inhibited over 90%, and retention of intron 36 went from being essentially undetectable to accounting for about 10% of the total (Figure 4C). This is consistent with the splicing defects reported, and suggests that ~50% of the GBF1 mRNA present after SON knockdown may contain at least one splicing error. Translation of these mis-spliced mRNAs would result in either an in-frame deletion of a highly conserved 34 amino acid motif, or a frameshift and premature termination, respectively. Either of these mutations would occur immediately on the cytoplasmic side of the predicted transmembrane domain, which is encoded by exon 37. They could therefore have substantial effects on GBF1 function, distinct from any effect of reduced protein. To address the effect of SON knockdown on the Golgi, we used a different cis-Golgi marker, GM130. As shown in Figure 4D, SON knockdown caused the Golgi to adopt a looser structure, although the effect was less marked than the complete dissociation caused by chemical inhibition of GBF1’s enzymatic activity using GCA. Automated image analysis confirmed a statistically significant increase in Golgi area (Figure 4E). A similar pattern was found using GBF1 itself as a cis-Golgi marker (data not shown). Taken together, these observations are consistent with a dysregulation of GBF1 and alteration in Golgi structure and function in the absence of SON leading to disruption of autophagic flux and accumulation of LC3 positive bodies, although further studies are necessary to evaluate causality.

Role of STAT family in type I interferon response

The proteomic data showed that SON knockdown strongly inhibited a number of type I interferon inducible proteins, including RSAD2, ISG20, and GBP5 (Figure 2D). In control transfected AM-MDMs, these proteins are expressed at low levels in the absence of bacteria, but are strongly induced by F. tularensis infection (Supplemental dataset 2). However, they are not induced in SON knockdown AM-MDMs, accounting for the relative inhibition shown in Figure 2D. This is consistent with the qPCR data showing that induction of IFNB1 and IFN responsive CXCL9, CXCL10, and IFIT2 by in F. tularensis LVS is largely dependent on SON. Other studies have shown that infection of BMDMs with F. novicida strongly induces type I interferon, which then causes autocrine induction of interferon responsive genes, including Gbp5 ^7,11. Type I interferon receptors signal through the JAK/STAT pathway, in particular STAT1/2 heterodimers. It was therefore interesting to observe that expression of a number of STAT family proteins was dependent on SON in our proteomic study (Figure 2D). qPCR showed that SON knockdown not only repressed STAT family gene expression in uninfected macrophages, but prevented their induction by F. tularensis infection (Figure 5A). Interestingly, STAT1 mRNA expression was dependent on SON expression, but protein abundance was not. We therefore used immunofluorescence to investigate the expression of STAT1 and STAT2, and their translocation to the nucleus in response to interferon stimulation (Figure 5B&C). As expected, type I interferon (IFNβ) caused translocation of STAT2, while type II interferon (IFNγ) caused STAT1 translocation in control-transfected cells. Both expression of STAT2 and its translocation in response to IFNβ were inhibited following SON knockdown, consistent both with STAT2 mRNA and proteomic observations, including the failure to induce interferon responsive proteins in response to infection after SON knockdown. STAT1 expression did not seem to be reduced by SON knockdown, consistent with proteomic data, but its cellular distribution was altered and was no longer responsive to IFNγ. This implies that SON may influence STAT1 dynamics and turnover, but further study is necessary to define this relationship.

Figure 5:

Dependence of STAT transcription factors on SON

A) Expression of the indicated STAT family member mRNAs or SON mRNA was measured by qPCR in AM-MDMs transfected with control (NT) or SON-targeted siRNAs (SONkd), then mock (-) or F. tularensis LVS (Ft) infected. Data are normalised to NT, uninfected. Mean ± SEM, n=3. *** p < 0.001, 2-way ANOVA with Tukey HSD, NT- vs. SONkd-.

B) STAT1&2 expression and nuclear localisation in response to 20 minute stimulation with 100U/ml IFNβ or IFNγ was determined by immunofluorescence, staining for STAT1 or STAT (white), or SON (red). Representative fields from 3 independent experiments shown.

C) Quantification of fluorescence intensities from (B). Mean ± SEM, n=3.

D) mRNA expression was measured by qPCR in AM-MDMs transfected with control or SON siRNAs then treated with vehicle or 1μM JAK inhibitors ruxolitinib or tofacitinib prior to and during infection with F. tularensis LVS. Genes measured (CXCL9, CXCL10, IFIT10, IFNB1, IL1B, TNF, SON) are grouped on the x-axis. Data are normalised to control, untreated cells. Mean ± SEM, n=3.

E) AM-MDMs were transfected and treated as in (D). Survival following infection was measured by LDH release. Mean ± SEM of 3 experiments, performed in triplicate. (***) p<0.001, (n.s.) p>0.05, 2-way ANOVA with Tukey HSD, NT- vs. SONkd-.

To assess the impact of SON knockdown on STAT transcriptional activity, we blocked STAT activation in mock-transfected cells using JAK inhibitors ruxolitinib and tofacitinib. As shown in Figure 5D, JAK inhibition, like SON knockdown, blocked induction of interferon-responsive genes CXCL9, CXCL10, and IFIT1B by F. tularensis infection. This is consistent with STAT dysfunction mediating the failure of SON knockdown cells to induce interferon response proteins following F. tularensis infection. However, expression of the other SON-dependent genes, TNF, IL1B, and, perhaps tellingly, IFNB1, was not affected (Figure 5D). Thus, while STAT dysfunction may contribute to the phenotype of SON knockdown macrophages, other mechanisms account for the SON-dependency of different genes.

Death of F. novicida-infected BMMs has been shown to depend on autocrine stimulation by type I interferon ⁷. Irf3^−/− BMMs, which are unable to produce interferon, do not die when infected but adding exogenous IFNβ restores mortality. Since STAT2 is required for signal transduction from IFN α/β receptors, this raises the possibility that the failure of AM-MDMs to express STAT2 and respond to IFNβ following SON knockdown, described above, accounts for their greater survival following F. tularensis infection. We therefore blocked STAT activity using JAK inhibitors in mock transfected cells (Figure 5E). Neither inhibitor reduced mortality, indicating that the dependence of STAT signalling on SON expression is insufficient to explain the survival advantage following SON knockdown.

Golgi inhibition mimics the effect of SON knockdown on gene expression

Given the previous observation that Golgi inhibition and SON knockdown had similar effects on autophagy, we tested the effect of GCA on gene induction in response to F. tularensis. Figure 6A shows that GBF1 inhibition and SON knockdown have remarkably similar, although not identical, effects on the induction of our panel of genes, perhaps suggesting similar mechanisms.

Figure 6:

IRF3 expression depends on SON.

AM-MDMs were transfected with non-targeting control (NT, GCA) or SON siRNAs (SONkd) then mock (-) or F. tularensis LVS (Ft) infected in the presence of 5μM Golgicide A (GCA) or vehicle (NT, SONkd) as indicated.

A) Gene expression was measured by qPCR. mRNAs quantified (CXCL9, CXCL10, IFIT10, IFNB1, IL1B, TNF, SON) are indicated on the x-axis. Data are normalised to control transfected, mock infected cells. Mean ± SEM of 4 independent experiments.

B) Indicated proteins were detected by Western blot. Representative of 3 independent experiments.

C) IRF3 and SON mRNA measured by qPCR. Mean ± SEM of 3 independent experiments.

The Golgi body plays a key role in the induction of type I interferon in response to various stimuli, including Francisella infection, TLR stimulation, and viral infection. Detection of cytosolic bacteria by cGAS or IFI16 triggers relocation of the signalling intermediate Stimulator of Interferon Genes (STING, TMEM173) from the endoplasmic reticulum to the cis-Golgi ^9,11,12,41. At the Golgi, STING causes autophosphorylation of TBK1 (TANK Binding Protein 1) at Ser-172. This activated TBK1 then phosphorylates IRF3, enabling its dimerization and migration to the nucleus, where it induces expression of genes including type I interferons ⁴¹. Figure 6B shows that, as expected, disruption of the Golgi by GCA prevented TBK1 phosphorylation in response to either F. tularensis infection or LPS. However, SON knockdown did not affect TBK1 phosphorylation. This indicates that SON is not essential for the initial responses to F. tularensis, including bacterial invasion, phagosome escape, and cytoplasmic detection, but that any effect most likely lies distal to the interaction between STING and TBK1.

Figure 6B also shows that IRF3 protein expression is strongly dependent on SON. This is supported by qPCR data, showing that SON knockdown causes a ~75% reduction in IRF3 mRNA (Figure 6C). This reduction may contribute substantially to the effect of SON knockdown on gene induction. However, induction of TNF and IL1B is predominantly driven by NF-κB rather than IRF3, so repression of IRF3 expression may not account for the whole effect observed.

Inflammasome function requires SON

Francisella infection has been shown to activate macrophage inflammasomes, resulting in IL-1β secretion and macrophage death, and this, at least in the case of F. novicida, depends on IRF3 ^7,8,14. We reasoned that, even though JAK inhibition did not protect AM-MDMs from killing by F. tularensis, impaired inflammasome activation may nonetheless contribute to the survival advantage originally observed following SON knockdown. As shown in Figure 7A, F. tularensis infected AM-MDMs and AM-MDMs treated overnight with LPS and the NLRP3 agonists nigericin or gramicidin both released IL-1β, consistent with previous reports. While the amount of IL-1β released differed between AM-MDMs from different donors, in each case secretion was dependent on SON expression. Similarly, SON knockdown inhibited dose-dependent macrophage killing by either nigericin or gramicidin (Figure 7B&C). We therefore measured expression of various inflammasome related genes in control and SON knockdown AM-MDMs (Figure 7D). While expression of the sensor gene NLRP3 was not significantly affected by SON knockdown, mRNAs encoding the effector caspase-1, the adaptor protein PYCARD (ASC), the pyrin and HIN domain (PYHIN) family members AIM2, PYHIN1, and IFI16, and the lytic protein gasdermin D (GSDMD) were all significantly reduced, regardless of whether the cells were subsequently infected. This provides a potential mechanism for the inhibited pyroptosis observed in SON knockdown macrophages.

Figure 7:

SON is required for inflammasome gene expression and function.

AM-MDMs were transfected with non-targeting control (NT) or SON siRNA (SONkd) then infected with F. tularensis LVS (Ft) or mock infected without additional stimulation (-) or with the indicated inflammasome activators.

A) IL-1β release was detected by ELISA. Cells were infected or treated with 100 ng/ml E. coli LPS together with 2 μM gramicidin or 5 μM nigericin. Symbols indicate independent experiments with different donors, each performed in triplicate ±S.D.

B&C) Mortality was measured by LDH release. Mean ±S.E.M. of 3 independent experiments, each performed in triplicate. * p<0.05, ** p<0.01, NT vs. SONkd, paired t test with BH correction.

D) Expression of the indicated inflammasome genes was measured by qPCR. Mean ±SEM, n=3. p<0.05 (*), <0.01(**), <0.001(***), SONkd uninfected vs NT uninfected, 2-way ANOVA with Tukey HSD.

E) Bacterial burden (GFP) was measured by flow cytometry. Percentages are means of 4 experiments ± SD, p<0.01 by 2-tailed t-test; representative images shown.

Pyroptosis is postulated to protect the host from pathogens, including Francisella, in part by eliminating infected cells before bacteria can replicate and by eliciting an inflammatory response in bystander cells. We therefore tested the effect of SON knockdown on bacterial burden. As shown in Figure 7E, there was less fluorescence from SON knockdown AM-MDMs infected overnight with GFP expressing F. tularensis LVS than from infected, control transfected cells, indicating a lower bacterial load. This suggests that another SON-dependent process has a greater effect on bacterial survival and replication within cells than inflammasome activity. Further studies are needed to establish the relative importance of different SON-dependent pathways following F. tularensis infection and other stimuli.

Discussion

Previous studies of SON have focused on its role in development and cancer. Here, we have addressed the roles of SON in innate immunity by studying the effects of its knockdown on an ex vivo differentiated model of human alveolar macrophages. Interfering with SON gene expression improves macrophage survival following infection with F. tularensis. SON modulates the key host defence processes of autophagy, the type I interferon response, and inflammasome gene expression and IL-1β secretion. These effects, in turn, may be related to altered Golgi conformation, and defective splicing and expression of GBF1, which maintains the cis-Golgi. The multiple effects of SON in key aspects of the macrophage response to F. tularensis infection are novel findings with an important limitation: because of SON’s involvement in multiple pathways and targets, it is often difficult to establish clear causality of effects. Thus, it is not clear which of the effects we observe are direct results of SON-dependent transcriptional events and which are secondary, and which have the greatest impact the survival of infected cells.

It is striking that all three aspects of the response to F. tularensis infection closely involve the Golgi body. During autophagy, the Golgi serves as a membrane source during autophagosome development ³⁹. We found that either SON knockdown or GBF1 inhibition caused an accumulation of LC3 positive vesicles. In the case of SON knockdown, our data indicates a disruption in autophagic flux. A previous study proposed that GCA treatment induced a non-canonical autophagy in fibroblasts but other explanations were not excluded ⁴². In contrast, Niso-Santano et al. ⁴³ have reported that the less specific GBF1 inhibitor, brefeldin A, can prevent co-localisation of autophagosomes and the p62 in the context of non-canonical autophagy in an osteosarcoma cell line, which is consistent with our observations. It will therefore be of interest to compare the autophagic lesions, in particular the compositions of the LC3 positive vesicles that accumulate, following SON knockdown and GBF1 inhibition.

Recent data have also shown that activation of some inflammasome types requires an intact Golgi body. The NLRP3 inflammasome forms on the Golgi surface, and that dissipating the Golgi with brefeldin A prevents NLRP3 aggregation, caspase 1 cleavage, and IL1β processing and release following LPS + nigericin treatment in BMDMs ⁴⁴. This is consistent with our observation that GCA treatment prevents IL-1β release from AM-MDMs following treatment with NLRP3 activators or F. tularensis infection. A number of reports show that NLRP3 inflammasome formation depends on trafficking of Golgi-associated vesicles, although different studies have placed these vesicles at the ER-Golgi interface ⁴⁵, between mitochondria-associated membranes and the Golgi ⁴⁴, or emerging from a dispersed trans-Golgi network ⁴⁶. Interestingly, differences are apparent between macrophage models: brefeldin A treatment blocked NLRP3-mediated caspase-1 cleavage and IL-1β release in BMDMs but not in human monocytes or THP-1 cells, nor did it affect activation of the AIM2 or NLRC4 inflammasomes ⁴⁵.

In contrast, the similarity between the effects of SON knockdown and GCA treatment on the induction of innate response genes following F. tularensis infection is less likely to reflect a causal linkage. While the dependence of the type I interferon response on trafficking between ER and Golgi is well documented, this trafficking is required for activation of IRF3 ⁴⁷. Furthermore, our observation that TBK1 phosphorylation is not inhibited by SON knockdown, unlike GCA treatment, indicates that ER to Golgi trafficking, as well as earlier phagocytosis and phagosome escape events, are not dependent on SON. Instead, we observed substantial repression of IRF3 and STAT family member genes following SON knockdown. This repression of key mediators may account for much of the effect on IFNB1 and interferon-responsive genes we observed. We also found that both SON knockdown and GCA treatment inhibited the induction of IL1B and TNF. While these are predominantly driven by NF-κB rather than IRF3, and our data confirm that the induction is not affected by JAK/STAT inhibition, reports from other cell types indicate that IL1B and TNF expression can also be dependent on IRF3 ^48,49. Thus, it is very possible that the dependence of innate response gene induction on SON expression is meditated by IRF3 expression rather than GBF1. Further experiments will be required to clarify this point.

A limitation of the associations between GBF1 splicing and activity on the one hand, and autophagy and inflammasome activity on the other, is that they rely on correlative data. SON knockdown in HeLa or embryonic stem cells has been shown to affect the expression of a large number of genes ^23,24. It is therefore possible that our observations all arise from independent repression of different effectors following SON knockdown, rather than any common mechanism.

Traffic along microtubules provides another alternative explanation for the defects we observe following SON knockdown and GCA treated macrophages. Previous reports have linked SON to microtubule function and organisation, especially at the mitotic spindle ^23,25, so it is possible that the effects we observe are due to defective shuttling of vesicles to or from the Golgi rather than GBF1 misfunction per se. Furthermore, SON has been shown to be necessary for trafficking of influenza virus to late endosomes in A549 lung epithelial cells ³¹. Our observation that SON knockdown does not affect TBK1 phosphorylation following F. tularensis infection indicates that early phagocytic events, up to bacterial escape from the phagosome, are not dependent on SON. In preliminary experiments (not shown), we have observed neither inhibition of gramicidin cytotoxicity by the dynein inhibitor EHNA, nor any gross abnormality in tubulin organisation by immunofluorescence. However, further study is required before a model wherein defective transport along microtubules mediates the cellular effects of SON knockdown can be excluded.

Some support for the importance of SON dependent GBF1 activity comes from the studies of viral infection. An RNAi screen in A549 lung epithelial cells identified SON as necessary for influenza virus infection ³¹. This was attributed to defective trafficking of viral particles to late endosomes in the absence of SON. Others have shown that GBF1 knockdown interferes with influenza trafficking ⁵⁰, but a causal linkage between the two observations has not been made to our knowledge. Indeed GBF1 inhibition has been found to inhibit a large number of viral infections, either by inhibiting the ARF1 and COP-I dependent vesicle trafficking ⁵¹or ARF4 dependent lipid droplet formation ^52,53. Interestingly, given our data showing defective splicing following SON knockdown, replication of poliovirus in HeLa cells was shown to be dependent on GBF1 but not its canonical enzyme activity ⁵⁴.

In summary, we have identified SON as a major control point in the response to an intracellular pathogen. SON governs autophagic flux, the IRF3/IFNβ response, and inflammasome expression and activity. Some of these effects may be mediated through GBF1 and Golgi function. In addition to infection with intracellular bacteria, these pathways are of great importance to a wide range of stimuli, including viral infection, environmental insult, and autoinflammatory disease. Thus, further study of SON may provide insight into mechanisms for a wide range of pathologies, and may offer a novel target for intervention.

Materials and Methods

THP1 library screen

Macrophage genes associated with the response to intracellular pathogens were identified from three sources. We reanalysed publicly available microarray datasets, essentially as described elsewhere ⁵⁵ to identify genes that were repeatedly regulated in different studies using a variety of macrophage types and intracellular pathogens. Other genes were obtained from previous RNAi or yeast two hybrid studies ^56–60. A third group was obtained by network analysis, to give a total of 7137 genes. Each gene was targeted by 7–8 independent shRNAs, designed and packaged in a barcoded, lentiviral library by Cellecta, Inc (Mountain View, CA). The genes targeted and shRNA sequences used are given in Supplemental Dataset 1. The library was transduced into THP1 promyelocytic cells to give approximately 20% transduction with minimal double-transduction, and selected for 3 days with 2 μg/ml puromycin to remove untransduced cells. Six separate transductions were performed to allow triplicate selection. Transduced cells were treated for 48 hr with 20 nM PMA to induce a macrophage-like phenotype, and allowed to recover for 24 hr prior to infection. Transduced macrophages were infected with serum-opsonised F. tularensis SchuS4 at MOI 50 for 72 hr – conditions that killed 90–95% of THP1 cells – or mock infected, in triplicate. Surviving cells were purified by layering on Ficoll-Paque Plus (GE Life Sciences) and centrifuging at 1200xg for 20 min. Elimination of dead cells was confirmed by Trypan blue counting. gDNA was extracted from surviving cells and the representation of each shRNA assessed by PCR amplification and Illumina sequencing of the barcode region. Statistical significance of enrichment was estimated for each shRNA using Fisher’s score with Benjamini-Hochberg FDR correction.

AM-MDM transfection and infection

Alveolar macrophage-like monocyte derived macrophages were matured using 20 ng/ml GM-CSF as described ³². We and others have previously shown that this technique provides an appropriate model of human primary alveolar macrophages ^32,61,62. Leukocytes were harvested from discarded platelet apheresis collars and donor identifying information was not recorded; the procedure was therefore exempted from review by Harvard Longwood Campus Institutional Review Board under Department of Health and Human Services rule 45 CFR 46.101(b)(4). Terminally differentiated AM-MDMs were plated to around 10⁵/cm³ 1–3 days before use. On-Target Plus SmartPools (Dharmacon) of 4 siRNAs, or Non-Targeting control, were transfected using Genmute for Primary Macrophages (Signagen) at a 72.5 nM final concentration. Briefly, media was changed 30–60min before transfection, and AM-MDMs left in presence of transfection complexes for 48 hr. Cells were washed 3× then maintained in RPMI (including 10% serum and GM-CSF) for 48 hr until infection or assay. mRNA knockdown by this method is routinely in excess of 90%.

F. tularensis LVS containing pKK214kanGFP ⁶³ was grown in brain-heart infusion with 1% isovitalex (BD biosciences) and 15 μg/ml kanamycin at 37 °C. F. tularensis ShuS4 was grown overnight in Muller-Hinton broth supplemented with isovitalex, as described elsewhere ⁶⁴. Bacteria were washed and resuspended in PBS, then diluted in RPMI+10% FBS + GM-CSF and added to transfected AM-MDMs to a final MOI 10. Cultures were centrifuged 15’, 3000 ×g, then the infection allowed to proceed 3–4 hr at 37°C + 5% CO₂. Wells were washed 2–3× then pulsed with 50 μg/ml gentamycin for 45–60 min to kill extracellular bacteria, then washed again and cultured overnight in the presence of 2.5 μg/ml gentamycin. Where inhibitors were used, AM-MDMs were pretreated for 1 hr, and the inhibitor maintained throughout the infection. Vehicle controls were usually 0.5% DMSO.

Infections with endospores of B. anthracis ANR-1 were performed essentially as described elsewhere ⁵⁵.

Survival assays

AM-MDM viability was measured using Cytox 96 Non-radioactive Cytotoxicity Assay (Promega) by comparing LDH activity in cell lysates with total LDH (lysate + supernatant). Triplicate wells were used for each condition, and experiments repeated using AM-MDMs from at least three independent donors.

qPCR

Amplicons were detected using Sybr Green, and quantified using the ΔΔCt method relative to housekeeping genes TBP, cyclophilin B (PPIB) and SDHA. See Table 1 for primer sequences. RNA was obtained from triplicate wells, and experiments repeated using AM-MDMs from at least three independent donors.

Table 1:

qPCR primer sequences

Target	Forward	Reverse	Source
AIM2	ACGTCTTCAGGAGGAGAAGGA	GTTCAGGCTTAACATGAGGAGAGAC
CASP1	TGGGACTCTCAGCAGATCAA	CTGCCGACTTTTGTTTCCAT
CXCL9	GGAGTGCAAGGAACCCCAGTAGT	GCAGGAAGGGCTTGGGGCAAAT
CXCL10	TTGTCCACGTGTTGAGATCATTGCT	AGGCAGCCTCTGTGTGGTCCA
GBF1 (total)	GGCACAAGATTTCTGCTTCCT	CAGTAGCAAAATGCGCAGG	UCSC
GBF1 (exon 35)	ACTTACCAAGCTCTTGGAGAACA	GGGATCGCCTCTGACAGTAA
GBF1 (intron 36)	AGCCTGAAGAGCCCCTAATG	TGGTAAGTAGAGGAAACAGCAC
GSDMD	CGGAACCAGACGTGCAGC	TTCGGGCCCACCTCTCAT
IFI16	GTCCGAGGAACAGACTCAGC	CCACTGTTTTCGGGTTCTCA
IFIT2	CAGCATTTATTGGTGGCAGA	TAGTTGCCGTAGGCTGCTCT
IFNB1	AAACTCATGAGCAGTCTGCA	AGGAGATCTTCAGTTTCGGAGG	65
IL1B	GGAGATTCGTAGCTGGATGC	GAGCTCGCCAGTGAAATGAT	Primer Depot
IRF3	ACAAGGAAGGAGGCGTGTTT	CACACAGAACCAGAGGGCAT	UCSC
LC3	CATGAGCGAGTTGGTCAAGA	GTTCACCAGCAGGAAGAAGG
NLRP3	CTGCCTCCTGCAGAACCTG	AAAGACGACGGTCAGCTCAG	UCSC
PPIB	ACCAGGGGAGATGGCACAGGA	TTGCCTGCGTTGGCCATGCT
PYCARD	TCTACCTGGAGACCTACGGC	TCCAGAGCCCTGGTGC
PYHIN	TCAAGATCTGGACTACTGTTGAAGA	TTCCTCCATCAAGTCAGCAA
SDHA	GCACTGTGCATAGAGGACGG	GACGTGCAGCTGAAGTAGGTG	UCSC
STAT1	CAAGTGTTATGGGACCGCAC	ATGCAGGGCTGTCTTTCCAC	UCSC
STAT2	AGCAAAAAGCCTGCATCAGA	AGCTGCCTCAGGTGAAACAA	UCSC
STAT5A	CCAGTACCAGGAGAGCCTGA	AGACACCTGCTTCTGCTGGA	UCSC
STAT6	GATATGGTGCCCCAGGTGTA	GTCAAAGGGCAGGCTCATC
TBP	CACGAACCACGGCACTGATT	TTTTCTTGCTGCCAGTCTGGAC	Pers. comm.
TNF	AGATGATCTGACTGCCTGGG	CAGCCTCTTCTCCTTCCTGA	Primer Depot

UCSC: UCSC genome browser ⁶⁶.

Primer Depot: ⁶⁷.

SON primers were obtained from IDT, catalogue Hs.PT.49a.20721343.

If not otherwise stated, primers were designed using NCBI BLAST and/or Primer3 software.

Immunofluorescence

AM-MDMs were transfected in optical 96 wells plates (Flacon). Cells were fixed in cold methanol for 5’, washed 3× HBSS, and blocked 1–2h 5% with goat serum + 0.3% triton in HBSS. Antibodies are listed in Table 2, and were diluted in HBSS with 2% BSA + 0.3% triton. Images were acquired on a BD Pathway HT microscope a using 40x objective and spinning disk confocal. BD Attovision 1.6 software was used to control acquisition as well as for object quantification and channel false-colouring and assembly. Multiple fields per well were used for quantification, providing a total of 300–1200 cells per well.

Table 2:

Antibodies used for Western Blot and Immunofluorescence

Target	Company	Catalogue
Actin	Cell Signalling	4970
GBF1	BD	612116
IRF3	Santa Cruz	Sc-9082
LC3	Cell Signalling	2775
Raptor	Cell Signalling	2280
SON	Abcam	Ab109472
SON	Santa Cruz	Sc-398508
TBK1	Cell Signalling	3013
TBK1 (pSer172)	Cell Signalling	5483
Tubulin	eBioscience	10-4502-80
ULK1	Santa Cruz	Sc-10902
ULK1 (pSer757)	Cell Signalling	14202
ULK1 (pSer595)	Cell Signalling	5869
α-Rabbit Alexa 594	Molecular Probes	A11012
α-Mouse Alexa 488	Molecular Probes	A11017

Western blotting

AM-MDMs were lysed in RIPA buffer + 2% SDS (which improves recovery of membrane-associated proteins but not LC3-I ³⁸) + 5× protease inhibitor cocktail (Roche) + phosphatase inhibitor cocktail 3 (Sigma) by scraping then passing through a 31G needle. Proteins were denatured in LDS buffer and separated on a Bis-tris gel (Invitrogen) with Biotinylated Molecular Weight Marker (Cell Signalling Technologies catalogue 7727). See Table 2 for antibodies. Images were acquired on an HD2 gel imager (Alpha Innotech) with ALTA U or U4000 camera using FluorChem software at an appropriate zoom. In some cases, contrast was adjusted and ladder bands marked using FluorChem. Images were cropped to show the relevant bands, outlines added to make cropping clear, greyscale inverted, scaled, and contrast adjusted using Gimp 2 and Microsoft PowerPoint. Adjustments were performed uniformly across the entire image. Wherever possible, membranes were reprobed with subsequent antibodies without stripping, to avoid artefacts any loss of protein during stripping.

Proteomics

Tandem mass tag mass spectrometry and peptide assignment and quantification was performed at Harvard FAS Proteomics core. Pooled AM-MDMs from three different donors were transfected then infected/mock infected overnight. Cells were lysed by scraping in RIPA buffer and insoluble debris removed by microcentrifugation.

IL1β ELISA

Capture antibody (mouse monoclonal clone JK1B-1, Biolegend) was diluted 1:200 in PBS. Detection antibody was biotin-conjugated mouse monoclonal clone JK1B-2 (Biolegend), diluted 1:500 in PBS. Streptavidin-HRP (R&D Systems) and hydrogen peroxide/tetramethyl benzidine (colour reagents A & B, R&D Systems) were used to develop.

Inhibitors and miscellaneous reagents

Brefeldin A and rapamycin was purchased from LC Labs. Golgicide A was from Sigma Aldrich. Gramicidin was from Molecular Probes. Nigericin, ruxolitinib, and tofacitinib were from Invivogen. Vinblastine was from Tocris. Human recombinant IFNβ and IFNγ were obtained from Millipore.