SUMMARY
Despite many connections between mutations in leucine-rich repeat kinase 2 (LRRK2) and susceptibility to mycobacterial infection, we know little about its function outside of the brain, where it is studied in the context of Parkinson’s Disease (PD). Here, we report that Lrrk2 controls peripheral macrophages and brain-resident glial cells’ ability to respond to and express inflammatory molecules. LRRK2 KO macrophages express elevated basal levels of type I interferons, resulting from defective purine metabolism, mitochondrial damage, and engagement of mitochondrial DNA with the cGAS DNA sensing pathway. While LRRK2 KO mice can control Mycobacterium tuberculosis (Mtb) infection, they exhibit exacerbated lung inflammation and altered activation of glial cells in PD-relevant regions of the brain. These results directly implicate Lrrk2 in peripheral immunity and support the “multiple-hit hypothesis” of neurodegenerative disease, whereby infection coupled with genetic defects in LRRK2 create an immune milieu that alters activation of glial cells and may trigger PD.
INTRODUCTION
Mutations in leucine rich repeat kinase 2 (LRRK2) are a major cause of familial and sporadic Parkinson’s Disease (PD), a neurodegenerative disease characterized by selective loss of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNc) region of the midbrain (Cookson, 2017; Kim and Alcalay, 2017; Martin et al., 2014; Schulz et al., 2016). Despite Lrrk2 having been implicated in a variety of cellular processes including cytoskeletal dynamics (Civiero et al., 2018; Kett et al., 2012; Pellegrini et al., 2017), vesicular trafficking (Herbst and Gutierrez, 2019; Sanna et al., 2012; Shi et al., 2017), calcium signaling (Bedford et al., 2016; Calì et al., 2014), and mitochondrial function (Ryan et al., 2015; Singh et al., 2019; Yue et al., 2015), its precise mechanistic contributions to triggering and/or exacerbating PD are not known.
Of all the cellular pathways affected by LRRK2 mutations, dysregulation of mitochondrial homeostasis has emerged as a centrally important mechanism underlying PD pathogenesis and neuronal loss (Cowan et al., 2019; Panchal and Tiwari, 2019). Indeed, other PD-associated genes such as PARK2 (Parkin), PINK1, and DJ1, all play crucial roles in mitochondrial quality control via mitophagy. Lrrk2 has been implicated in mitophagy directly through interactions with the mitochondrial outer membrane protein Miro (Hsieh et al., 2016), and several lines of evidence support roles for Lrrk2 in controlling mitochondrial network dynamics through interactions with the mitochondrial fission protein Drp1 (X. Wang et al., 2012). Accordingly, a number of different cell types, including fibroblasts and iPSC-derived neurons from PD patients harboring mutations in LRRK2 exhibit increased oxidative stress, increased ROS, and defects in mitochondrial network integrity (Sison et al., 2018; Smith et al., 2016). Because DA neurons in the SNc have high bioenergetic needs and a unique highly-branched morphology, they are thought to be particularly sensitive to defects in mitochondrial homeostasis conferred by mutations in LRRK2 (Surmeier et al., 2017). In spite of these well-appreciated links, Lrrk2’s contribution to mitochondrial health in cells outside of the brain remains vastly understudied.
There is mounting evidence that mutations in LRRK2 contribute to immune outcomes both in the brain and in the periphery. For example, mutations in LRRK2 impair NF-κB pathways in iPSC-derived neurons and render rats prone to progressive neuroinflammation following administration of peripheral innate immune triggers (López de Maturana et al., 2016). Likewise, chemical inhibition of LRRK2 attenuates inflammatory responses in microglia ex vivo (Moehle et al., 2012). In addition to these strong connections between LRRK2 and inflammatory responses in the brain, numerous GWAS studies suggest that LRRK2 is an equally important player in the peripheral immune response. Numerous single nucleotide polymorphisms (SNPs) in LRRK2 are associated with susceptibility to mycobacterial infection (Fava et al., 2016; Marcinek et al., 2013; D. Wang et al., 2015; F.-R. Zhang et al., 2009), inflammatory colitis (Umeno et al., 2011), and Crohn’s Disease (Van Limbergen et al., 2009). Consistent with a role for LRRK2 in pathogen defense and autoimmunity, LRRK2 is abundant in many immune cells (e.g. B cells, dendritic cells, monocytes, macrophages) and expression of Lrrk2 is induced in human macrophages treated with IFN-γ (Gardet et al., 2010). Loss of LRRK2 reduces IL-1β secretion in response to Salmonella enterica infection in macrophages ex vivo (Liu et al., 2017) and enhances expression of pro-inflammatory cytokines in response to Mtb infection (Härtlova et al., 2018). However, the precise mechanistic contribution(s) of LRRK2 to controlling immune responses in the periphery remain poorly understood.
Here, we provide the first evidence that LRRK2’s ability to influence inflammatory gene expression in macrophages is directly linked to its roles in maintaining mitochondrial homeostasis. Specifically, we demonstrate that depolarization of the mitochondrial network and hyper-activation of Drp1 in LRRK2 KO macrophages leads to release of mtDNA, engagement of the cGAS-dependent DNA sensing pathway, and abnormally high basal levels of interferon β (type I interferon (IFN)) and interferon stimulated genes (ISGs). These high basal levels of type IFN appears to completely reprogram LRRK2 KO macrophages, rendering them refractory to a number of distinct innate immune stimuli, including infection with the important human lung pathogen, Mtb. While Mtb-infected LRRK2 KO mice did not exhibit significant differences in bacterial burdens, we did observe exacerbated pathology in the lungs. Remarkably, although no bacilli were present in the brains of CT or KO mice, both exhibited dramatic signs of neuroinflammation, evidenced by activation of microglia and astrocytes in the dorsolateral striatum (DLS), a PD-relevant region of the brain. Collectively, these results (1) demonstrate that LRRK2’s role in maintaining mitochondrial homeostasis is critical for proper induction of inflammatory gene expression in both peripheral macrophages and brain-resident glial cells and (2) provide strong support for the “multiple-hit hypothesis” of neurodegeneration, whereby peripheral infection coupled with specific genetic mutations may trigger or exacerbate neuronal loss.
RESULTS
RNA-SEQ analysis reveals that LRRK2 deficiency in macrophages results in dysregulation of the type I interferon response following Mtb infection
To begin to implicate LRRK2 in the peripheral immune response, we took an unbiased approach to determine how loss of LRRK2 impacts innate immune gene expression during Mtb infection of macrophages ex vivo. Briefly, primary murine bone marrow-derived macrophages (BMDMs) derived from littermate heterozygous (control, CT) and knockout (KO) LRRK2 mice were infected with Mtb at a MOI=10 and RNA-seq analysis was performed on total RNA collected from uninfected and infected cells 4h post-infection. Previous studies have identified 4h as a key innate immune time point during Mtb infection, corresponding to the peak of transcriptional activation downstream of sensing molecules including the cytosolic DNA sensor cGAS (Manzanillo et al., 2012; Watson et al., 2015; 2012).
Following analysis with CLC Genomics Workbench, we identified hundreds of genes that were differentially expressed in LRRK2 KO vs. CT BMDMs during Mtb infection, with 192 genes significantly up- or down-regulated (179 up, 13 down) (p<0.001) (Fig. 1A, Table S1). Although a number of genes were significantly induced during infection in both genotypes, the level of induction for a number of transcripts was noticeably lower in LRRK2 KO cells compared to CT (Fig. 1C). Canonical pathway analysis of differentially expressed genes revealed significant enrichment for genes involved in type I IFN and other related pathways (−log(p)=12.98), including Activation of IRF by cytosolic Pattern recognition receptors (−log(p)=11.81), RIG-I signaling (−log(p)=4.69) and autophagy (−log(p)=3.815) (Fig. 1B, S1B). Indeed, a majority of the top differentially expressed genes were well-known interferon-stimulated genes (ISGs), including Ifit1, Mx1, and Isg15 (Fig. 1C). Mtb is a potent activator of type I IFN expression, thought to occur mostly through permeabilization of the Mtb-phagosome and release of bacterial dsDNA into the cytosol, where it is detected by DNA sensors like cGAS, activating the STING/TBK1/IRF3 axis (Collins et al., 2015; Wassermann et al., 2015; Watson et al., 2015; Wiens and Ernst, 2016). Follow-up RT-qPCR analysis confirmed that Ifnb and ISGs like Mx1, Isg15, and Gbp7 were induced to lower levels in LRRK2 KO macrophages compared to controls following Mtb infection (Fig. 1D). This differential response seemed to be specific for type I IFN and ISGs since the transcripts of Mtb-induced cytokines like Tnf and Il1b reached similar levels in both genotypes (Fig. 1E).
Resting LRRK2 KO macrophages express elevated levels of type I IFN
To begin to investigate the nature of this defect in type I IFN induction, we again used CLC Genomics Workbench analysis to identify transcripts affected by loss of LRRK2 in uninfected or “resting” macrophages. We observed higher basal expression of a number of innate immune transcripts in LRRK2 KO BMDMs relative to control, including several type I IFN family genes (e.g. Oas3, Irf7, Oasl2, Isg15, Zbp1) (Fig. 1F, 1H). Indeed, differential gene expression analysis and unbiased canonical pathways analysis revealed that again, “Interferon Signaling” was the most significantly impacted pathway in resting macrophages (−log(p)=12.95) (Fig. 1G). In fact, almost all the pathways and families of genes that were differentially expressed in LRRK2 KO BMDMs in Mtb-infected cells were also impacted in uninfected cells (Fig. 1B and 1G, S1C-D). RT-qPCR analysis confirmed significantly elevated levels of type I IFN transcripts including Ifnb, Irf7, Isg15, and Ifit in LRRK2 KO BMDMs (Fig. 1I and S1E). Interestingly, several transcripts, including Apoe, a gene associated with Alzheimer’s and cardiovascular disease, had decreased expression in LRRK2 KO BMDMs compared to controls (Fig. 1J). Transcripts like Tnf were again expressed at similar levels in the two genotypes (Fig. 1K). Importantly, the resting and induced levels of interferon and ISGs were similar between WT and HET LRRK2 BMDMs, validating our use of HETs as controls in future experiments (Fig. S1F-G).
We also found increased basal expression and decreased induction of interferon and ISGs upon Mtb infection in the human monocyte cell line U937 (Fig. 2A and S2A) and in RAW 264.7 macrophages when LRRK2 was knocked down by shRNA (Fig. S2C). Infection of RAW 264.7 cells in which LRRK2 is knocked out (RAW 264.7 LRRK2 KO) with Mycobacterium leprae (Mlep), which shares the virulence-associated ESX-1 secretion system with Mtb and also induces type I IFN through cytosolic nucleic acid sensing (de Toledo-Pinto et al., 2016), showed a similar defect in type I IFN induction compared to CT RAW cells (Fig. 2B and S2B). Together, these transcriptome-focused analyses reveal that LRRK2 KO macrophages have a higher interferon signature at baseline but are unable to induce the type I IFN response to the same levels as control cells when infected with Mtb or Mlep.
LRRK2 KO macrophages fail to induce type I IFN in response to diverse innate immune stimuli
Because both Mtb and Mlep stimulate type I IFN through the cGAS/STING/TBK1 axis, we hypothesized that loss of LRRK2 may cause defects in this pathway. Briefly, CT and KO BMDMs were treated with a series of agonists that stimulate a variety of innate immune pathways and elicit a type I IFN response. LRRK2 KO BMDMs failed to fully induce Ifnb and Irf7 transcripts following transfection of interferon stimulating DNA (ISD) (Fig. 2C), as did LRRK2 KO peritoneal macrophages (Fig. 2D), LRRK2 KO or KD RAW macrophages (Fig. 2E and S2C), and LRRK2 KO MEFs (Fig 2F and S2E). Likewise, stimulating the DNA sensing adapter STING directly using the agonist DMXAA or transfection of the cGAS product/STING agonist, cGAMP also failed to induce type I IFN in LRRK2 KO BMDMs (Fig. 2G and S2D) and in LRRK2 KO PEMs (Fig. S2F).
We next tested whether loss of LRRK2 impacted the ability of cells to respond to other activators of the type I IFN response outside of the cytosolic DNA sensing cascade. To these ends, we treated LRRK2 KO and control BMDMs with transfected polyI:C (to activate RNA sensing), LPS (to stimulate TRIF/IRF3 downstream of TLR4), and CpG and CL097 (to stimulate nucleic acid sensing via TLR9 and TLR7, respectively). In all cases, we observed a defect in the ability of LRRK2 KO BMDMs to induce Ifnb and ISGs (Fig. 2G). LRRK2 KO BMDMs were also defective in ISG expression (Isg15 and Irf7) following recombinant bioactive IFN-β treatment (which directly engages with IFNAR) (Fig. 2I). Consistent with our RT-qPCR data, western blot analysis of IRF3 (Ser 395) after ISD transfection showed a significant defect in the ability of LRRK2 KO macrophages to respond to cytosolic DNA (Fig. 2H). LRRK2 KO BMDMs were also defective in ISG expression (Isg15 and Irf7) following recombinant bioactive IFN-β treatment (which directly engages with IFNAR) (Fig. 2I). Based on these collective results, we concluded that LRRK2 KO macrophages are reprogrammed in such a way that they cannot properly induce type I IFN expression regardless of the innate immune stimulus received.
We next hypothesized that the increased basal levels of type I IFN transcripts were driving the inability of LRRK2 KO macrophages to properly induce a type I IFN response. To test this, CT and LRRK2 KO BMDMs were treated with an IFN-β neutralizing antibody to prevent engagement of IFNAR and downstream signaling. As predicted, this IFN-β blockade decreased basal transcript levels of Irf7 and Isg15 in LRRK2 KO cells to levels similar to CT cells (Fig. 2J) and remarkably, restored the ability of LRRK2 KO BMDMs to fully induce type I IFN expression following ISD transfection (Fig. 2K). We further validated that loss of IFNAR signaling rescues the LRRK2 KO BMDM phenotype by crossing LRRK2 KO mice to IFNAR KO mice and observed a significant reduction in ISG levels in LRRK2/IFNAR double KO BMDMs (Fig. 2L) and upon stimulation normalized their induction of type I IFNs and ISGs to that of controls (Fig 2M). Together, these results indicate that cytosolic nucleic acid and IFNAR signaling pathways are all intact in LRRK2 KO macrophages, but chronic elevated basal type I IFN expression renders these cells refractory to innate immune stimuli.
Increased basal IFN in LRRK2 KO macrophages is dependent on cytosolic DNA sensing through cGAS
Because IFN-β blockade and loss of IFNAR normalized ISG expression in LRRK2 KO macrophages, we hypothesized that LRRK2 contributes to basal type I IFN expression upstream of the cytosolic DNA sensing pathway. To directly test the involvement of cGAS in generating elevated resting levels of type I IFN in LRRK2 KO macrophages, we crossed LRRK2 KO and cGAS KO mice and compared type I IFN transcript levels in double KO BMDMs with those of littermate controls. As expected, loss of cGAS led to lower resting Ifnb and Isg15 expression (Schoggins et al., 2014) and knocking out cGAS in a LRRK2 KO background rescued the elevated basal IFN and ISG expression (Fig. 3A and S3A). With lowered resting type I IFN levels, cGAS/LRRK2 double KOs were able to respond normally to IFN/ISG-generating innate immune stimuli like LPS or cGAMP and DMXAA (both of which bypass cGAS and stimulate STING directly) (Diner et al., 2013), but not ISD (Fig. 3B and S3A, respectively). Together, these results indicate that the high basal levels of type I IFN in LRRK2 KO macrophages are due to engagement of the cGAS-dependent DNA sensing pathway.
Cytosolic sensing of mtDNA contributes to basal type I IFN expression in LRRK2 KO macrophages
We next sought to identify the source of this cGAS-activating signal. Mitochondrial DNA (mtDNA) has been shown to be a potent activator of type I IFN downstream of cGAS (Yang et al., 2014), and LRRK2 is known to influence mitochondrial homeostasis, albeit through mechanisms that are not entirely clear. To begin to implicate mtDNA in type I IFN dysregulation in LRRK2 KO macrophages, we first observed the status of the mitochondrial network in MEFs from control and LRRK2 KO mice. As has previously been described for cells overexpressing LRRK2 or mutant alleles of LRRK2 (Yang et al., 2014), LRRK2 KO MEFs had a more fragmented mitochondrial network, especially around the cell periphery (Fig. 3C). We hypothesized that this fragmentation was a sign of mitochondrial damage and could allow mitochondrial matrix components, including mtDNA, to leak into the cytosol. Therefore, we isolated the cytosolic fraction of control and LRRK2 KO MEFs and measured cytosolic mtDNA levels. We found that LRRK2 KO MEFs had ∼2-fold higher levels of total mtDNA compared to controls, and importantly, the LRRK2 KO cells had ∼2-fold more cytosolic mtDNA (Fig. 3D-E). To exacerbate the proposed defect, we treated BMDMs with the Bcl2/caspase inhibitors ABT737/OVD-PH, which together induce mitochondrial stress and spillage of mtDNA into the cytosol. As expected, type I IFN and ISG gene expression was induced but was decreased in LRRK2 KO cells while Tnf transcript levels remained unaffected (Fig. 3F and S3B-C). We also sought to exacerbate the LRRK2 defect by crossing LRRK2 KO mice with TFAM HET mice, which are deficient in the mitochondrial transcription factor required for maintaining the mitochondrial network (Kasashima et al., 2011; West et al., 2015). Double LRRK2 KO/TFAM
Het BMDMs had even further elevated type IFN and ISG gene expression and they were similarly defective at inducing ISGs upon innate immune stimulation (ISD or LPS) (Fig. 3G). Together, these data suggest that spillage of mtDNA into the cytosol in LRRK2 KO cells contributes to their defective type I IFN expression.
To further implicate mtDNA in contributing to abnormal type I IFN responses in LRRK2 KO cells, we sought to rescue the defect by depleting mtDNA using ddC, an inhibitor of mtDNA synthesis, or ethidium bromide (EtBr), an intercalating agent shown to deplete mtDNA in dividing cells (Leibowitz, 1971; Meyer and Simpson, 1969). Treating LRRK2 KO RAW 264.7 cells with ddC or EtBr reduced mtDNA copy number (Fig. 3H-I, S3D, and S3E-F respectively), and normalized IFN and ISG expression in resting LRRK2 KO cells. Importantly, LRRK2 KO macrophages depleted of mtDNA showed a restored type I IFN response when stimulated with ISD or treated with DMXAA (Fig. 3I and S3D-F). Collectively, these results demonstrate a critical role for mtDNA in driving both the high basal levels of type I IFN and the inability to properly induce type I IFN expression in LRRK2 KO macrophages.
Previous studies of microglia have shown that Lrrk2 contributes to mitochondrial homeostasis through interaction with the mitochondrial fission protein Drp1 (Ho et al., 2018). Thus, we hypothesized that the loss of LRRK2 may compromise mitochondrial stability via misregulation of Drp1 activity and spillage of mtDNA into the cytosol. Looking grossly at Drp1, we observed Drp1+ puncta via immunofluorescence microscopy at the ends of fragmented mitochondria in LRRK2 KO MEFs, but the total levels and overall distribution did not differ between the CT and KO cells (Fig. S3G). Drp1 fission is positively regulated by its phosphorylation at Ser616 (Taguchi et al., 2007). Therefore, to measure Drp1 activity in control and LRRK2 KO cells, we performed flow cytometry with an antibody specific for phospho-S616 Drp1. We found increased p616 in resting LRRK2 KD RAW and KO MEFs compared to CT cells (Fig. 3J-K). Next, to test whether Drp1 activity influences ISG expression in LRRK2 KO cells, we chemically inhibited Drp1 with Mdivi-1 and measured basal gene expression by RT-qPCR. In LRRK2 KO BMDMs and LRRK2 KD RAW macrophages, Drp1 inhibition returned ISG expression in LRRK2 KO cells to control levels (Fig. 3L and S3H). Moreover, Drp1 inhibition also restored the cytosolic mtDNA levels in LRRK2 KO cells to those of CT cells (Fig. S3I and S3J) This is consistent with an indirect function of Drp1 in mtDNA replication (Parone et al., 2008). Together, these data indicate that aberrant ISG basal expression and induction in LRRK2 KO cells is caused by leakage of mtDNA into the cytosol, downstream of excessive Drp1-induced mitochondrial fission.
LRRK2 KO macrophages are susceptible to mitochondrial stress and have altered cellular metabolism
Given that cytosolic mtDNA contributes to type I IFN defects in LRRK2 KO macrophages, we predicted that mitochondria in LRRK2 KO cells may be more damaged or more prone to damage. To better understand the health of the mitochondrial network in LRRK2 KO vs. CT macrophages, we first used the carbocyanine dye, JC-1. JC-1 dye accumulates in mitochondria dependent on membrane potential resulting in red fluorescent J-aggregates. Upon loss of membrane potential JC-1 diffuses into the cytosol where it emits a green fluorescence as a monomer. Thus, a reduction in red/green fluorescence intensity signifies mitochondrial depolarization allowing for a sensitive assessment of mitochondrial membrane potential. Flow cytometry analysis of resting CT and LRRK2 KO cells revealed lower levels of JC-1 dye aggregation (i.e. lowered mitochondrial membrane potential) in LRRK2 KO BMDMs (Fig. 4A-B), LRRK2 KD RAW 264.7 macrophages (Fig. S4A), and primary LRRK2 KO MEFs (Fig. S4B-C), compared to CT cells. In addition, LRRK2 KO cells were more sensitive to the mitochondrial damaging and depolarizing agents, rotenone and ATP, which is consistent with the LRRK2 KO cells harboring a baseline mitochondrial defect (BMDMS: Fig. 4C-D, RAW264.7 and MEFs: Fig. S4B-C).
Previous reports have indicated that LRRK2 dysfunction alters reactive oxygen species (Pereira et al., 2014; Russo et al., 2019). To test whether ROS could contribute to the defective type I IFN signature in LRRK2 KO cells, we treated control and LRRK2 KO BMDMs with mitoTEMPO (mitoT), a mitochondrially-targeted scavenger of superoxide (Liang et al., 2010). LRRK2 KO cells treated with mitoT had normal basal ISG expression levels (Fig. 4E) and induced IFN and ISG to levels comparable to control cells following ISD transfection (Fig. 4F). We also hypothesized that mitochondrial defects may render LRRK2 KO macrophages incapable of meeting metabolic demands in response to carbon sources. To test this idea, we added sodium pyruvate, an intermediate metabolite of glycolysis and the TCA cycle, to LRRK2 KO and control RAW 264.7 cell media (DMEM + 10% FBS) and indeed, observed that sodium pyruvate exacerbated high basal levels of type I IFN and decreased induction of ISGs in LRRK2 KO macrophages in a dose-dependent fashion (Fig. S4D-E).
We next used the Agilent Seahorse Metabolic Analyzer to further investigate the nature of the mitochondrial defect in LRRK2 KO macrophages. In this assay, oxidative phosphorylation (OXPHOS) and glycolysis are assayed by oxygen consumption rate (OCR) and extracellular acidification rate (ECAR), respectively. We found that OCR in LRRK2 KO BMDMs was defective both in terms of maximal and reserve capacity (Fig. 4G upper graph), indicating reduced mitochondrial metabolism. Defects in non-glycolytic acidification, maximal, and reserve capacity were also detected, indicating reduced glycolysis (Fig. 4G lower graph). This result was surprising as cells typically switch from utilizing OXPHOS to glycolysis when activated (Kelly and O’Neill, 2015). Remarkably, co-treatment of LRRK2 KO BMDMs with mitoT and IFN-β neutralizing antibody completely restored OCAR and ECAR. This rescue was greater than treatment of either IFN-β Ab or mitoT alone (Fig. 4G). We also measured OCR and ECAR in the presence of sodium pyruvate and observed an even greater defect in reserve and maximal capacity in LRRK2 KO vs. CT macrophages (Fig. S4F). Collectively, these data demonstrate that loss of LRRK2 in macrophages has profound impact on the mitochondria, whereby loss of mitochondria are less capable of effectively processing high energy electrons produced by the TCA cycle.
Reduced antioxidants and purine biosynthesis metabolites contribute to mitochondrial damage and type I IFN expression in LRRK2 KO macrophages
To better understand possible mechanisms driving, or resulting from, damaged mitochondria in LRRK2 KO macrophages, we performed an unbiased query of metabolites using LC/MS/MS (Zhou et al., 2012). Looking at metabolites with significant differences, we noted lower levels of IMP and hypoxanthine, intermediates in the purine biosynthesis pathway (Fig. 5C) in LRRK2 KO cells (Table S2), which we confirmed using pure molecular weight standards (Fig. 5A-C and S5A-B). Interestingly, purine metabolism is tightly associated with generation of antioxidant compounds and several metabolites in this pathway are well-characterized biomarkers of PD (Chen et al., 2012). Consistent with these defects in purine metabolism, we observed significantly fewer puncta containing formylglycinamidine ribonucleotide synthase (FGAMS, also known as PFAS), a core purinosome component, per LRRK2 KO cell compared to control cells (Fig. 5D-E). Consistent with lower levels of antioxidants, we detected increased oxidized glutathione and glutamate metabolism compounds in LRRK2 KO macrophages (Table S2).
Because depleted antioxidant pools (and concomitant accumulation of ROS) can lead to mitochondrial damage, we hypothesized they might contribute to the mitochondrial and type I IFN defects we observe in LRRK2 KO macrophages. To test this, we first supplemented cells with antioxidants directly in order to rescue the type I IFN defect in LRRK2 KO macrophages. Addition of urate consistently reduced basal ISG levels in LRRK2 KO BMDMs (Fig 5F), and urate treatment restored the ability of LRRK2 KO BMDMs to induce type I IFN upon stimulation with LPS or ISD (Fig 5G). In addition, treatment with urate or mitoT restored the resting mitochondrial membrane potential of LRRK2 BMDMs (Fig. 5H, histogram in 5I). Collectively, these results suggest that the depletion of antioxidant pools in LRRK2 KO macrophages downstream of defective purine metabolism contributes to their mitochondrial dysfunction and aberrant type I IFN expression signature.
LRRK2 KO mice control Mtb infection similarly but have altered infection-induced neuroinflammation
Previous reports have linked SNPs in LRRK2 with susceptibility to mycobacterial infection in humans, and our studies indicate that LRRK2 plays a key role in homeostasis of macrophages, the first line of defense and replicative niche of Mtb. Therefore, we sought to understand how LRRK2 deficiency influences innate immune responses in vivo during Mtb infection. We infected control and LRRK2 KO mice with ∼150 CFUs via aerosol chamber delivery. At 7, 21, 63, and 126 days post-infection, we observed no significant differences in bacterial burdens in the lungs or spleen of infected mice (Fig. S6A). We also measured serum cytokines and tissue cytokine expression and found no major differences (Fig. S6D-E). However, upon inspection of lung tissues via H&E staining, we observed significantly more neutrophils (polymorphonuclear leukocytes, PMNs) in the lungs of LRRK2 KO mice 21 days post-infection (Fig. 6A-B). Furthermore, the percentage of neutrophils that were necrotic (degenerate PMNs) was higher in the infected LRRK2 KO mice than in the controls (Fig. 6B). The LRRK2 KO mice also had more granulomatous nodules, indicating more macrophage-like cells had infiltrated the infected lungs (Fig. S6B-C). Together, these results indicate that while control and LRRK2 KO mice did not display different bacterial burdens, the KO mice had an overly robust innate immune response early during Mtb infection.
Effect of LRRK2 and TB infection on microglia reactivity in the dorsal lateral striatum (DLS), substantia nigra pars compacta (SNc), and ventral tegmental area (VTA)
Several lines of evidence point to a connection between persistent infections and neurodegenerative disease and several links between LRRK2 and neuroinflammation have been previously reported (De Chiara et al., 2012; Schildt et al., 2019). Therefore, we set out to investigate markers of neuroinflammation in the brains of infected mice and uninfected controls. We first focused on microglia since these are the cells of the central nervous system (CNS) that play important roles in neuroimmune surveillance, similar to macrophages in the periphery (Ousman and Kubes, 2012). To assess the extent to which Mtb infection alters microglia reactivity in LRRK2 CT and KO mice, we focused on three brain structures relevant to PD.
These are the dorsolateral striatum (DLS) which contains dopaminergic (DA) terminals, the substantia nigra pars compacta (SNc) and the ventral tegmental area (VTA), both of which contain DA cell bodies (Fig. 6C). Microglia reactivity in the DLS, SNc and VTA was assessed by measuring endogenous Iba1 fluorescence intensity in microglia and normalizing these values to NeuN fluorescence in the DLS or TH fluorescence in the SNc and VTA.
First, we examined the effect of systemic Mtb infection on LRRK2 CT mice. It is important to note that at all time points examined, there is no evidence of Mtb bacilli in the brains, as measured by acid-fast staining (Fig. S6J). Compared to uninfected mice, infected LRRK2 CT mice showed significant increases in normalized Iba1 fluorescence in the DLS at 21 and 126 days post-infection (3-fold at day 21) (Fig. 6D & 6F). In contrast, Iba1 fluorescence in the two midbrain structures, SNc and VTA, remained unaltered relative to age-matched uninfected LRRK2 CT mice at 21 days post-infection, but significantly increased at 126 days post-infection (5-fold in SNc; 2-fold in VTA) (Fig. 6F and S6G).
We next examined microglia reactivity in the DLS, SNc and VTA of LRRK2 KO mice following systemic Mtb infection. We found that Mtb-infected LRRK2 KO mice showed a pattern of microglia reactivity in the DLS that was similar to LRRK2 CT mice. Relative to uninfected LRRK2 KO mice, we observed a 2-fold increase in Iba1 fluorescence within the DLS at day 21 (p<0.001) and a 2.5-fold increase at day 126 (p<0.001) post Mtb-infection (Fig. 6F). Compared to age-matched uninfected LRRK2 KO control mice, Iba1 fluorescence in the SNc and VTA from Mtb infected LRRK2 KO mice was unchanged at 21 days post-infection (p=0.89, SNc; p=0.33, VTA), but increased by 3-fold in the SNc (p<0.001) and VTA (p=0.03) at 126 days post-infection (Fig. 6F and Fig. S6G). Together these data show for the first time that microglia in the three PD-relevant brain regions, the DLS, SNc, and VTA, become reactive following chronic systemic infection with Mtb. Intriguingly, upon close inspection, we observed a high degree of colocalization between Iba1 and TH staining in the Mtb-infected SNc (Day 126) (indicated by arrows in Fig. 6F) and accumulation of TH-positive cellular debris. We speculate that these events correspond to phagocytosis of damaged neurons or neuronal debris, which selectively occurs in the SNc of Mtb-infected mice.
Furthermore, midbrain microglia become reactive at later time points post-Mtb infection than microglia in the DLS with no additional effect of LRRK2 KO on the pattern or extent of reactivity. In addition, we did not observe any baseline differences in the reactivity profile of microglia across all brain regions or time points between uninfected LRRK2 KO and LRRK2 CT mice. The sole exception was a small but significant 1.5-fold decrease in Iba1 in the DLS of LRRK2 KO compared to CT mice at 126 days-post infection (p=0.01) (Fig. 6E-F).
Since the mice were aged for an additional 126 days in our experiments with or without Mtb infection and because aging can alter the expression of Iba1 in microglia, we examined the effect of age on baseline expression of Iba1 in LRRK2 CT and KO mice. Both groups of mice showed a significant age-dependent reduction in microglia reactivity in the DLS, SNc and VTA when comparing 3.5 month old to 6.5 month old uninfected LRRK2 CT and KO mice (Fig. S6I) (p<0.001 for DLS and SNc; p=0.016 for VTA; ANOVA). Mtb infection failed to reduce microglia reactivity in the SNc and VTA from day 21 to 126, while preserving the natural decline of microglia reactivity in the DLS (Fig. 6K and S6J).
Effect of LRRK2 and Mtb infection on astrocyte reactivity in the dorsal lateral striatum (DLS), substantia nigra pars compacta (SNc), and ventral tegmental area (VTA)
We also assessed the extent to which Mtb infection alters astrocyte reactivity in LRRK2 CT and KO mice in the DLS, SNc and VTA. Although once considered supporting cells in the CNS, emerging evidence suggests that astrocytes play vital roles in modulating neural circuit activity during physiological and pathological states (Khakh and Sofroniew, 2015). Since GFAP expression levels are known to increase in reactive astrocytes, we measured GFAP fluorescence and normalized values to NeuN fluorescence in the DLS and TH fluorescence in SNc and VTA.
We found that at 21 days post-infection, GFAP levels in LRRK2 CT astrocytes in the DLS were similar to uninfected mice, but significantly increased after 126 days post Mtb-infection (p=0.52 at day 21; 4-fold at day 126, p=0.002) (Fig. 6J and I), suggesting that chronic systemic TB infection increases astrocyte reactivity in the DLS. In contrast to the DLS, the SNc showed a more complex profile of astrocyte reactivity. When compared to age-matched uninfected CT mice, SNc astrocytes showed 4-fold lower GFAP fluorescence at day 21 post-infection (p<0.001), which increased by 3-fold 126 days post-infection (p=0.005) (Fig. 6J). The astrocyte reactivity profile within the VTA of LRRK2 CT mice was similar to the SNc, but differences at both time points were not significant (p=0.41 at day 21; p=0.10 at day 126) (Fig. S6H).
Having found large differences in astrocyte reactivity with systemic Mtb infection in LRRK2 CT mice, we assessed the extent to which astrocytes in LRRK2 KO mice with systemic Mtb infection showed reactivity in DLS, SNc and VTA. In contrast to LRRK2 CT mice, Mtb infected LRRK2 KO mice displayed no significant GFAP differences in the DLS at either time point when compared to uninfected LRRK2 KO mice (p=0.10 at day 21; p=0.13 at day 126) (Fig. 6J and I). In addition, we observed that LRRK2 KO mice followed a pattern of astrocyte reactivity in the SNc and VTA that was similar to that seen in LRRK2 CT mice. Astrocytes in the SNc showed significantly lower GFAP at day 21 post-infection, but had significantly higher levels than uninfected LRRK2 KO mice by day 126 post-infection (3-fold at day 21, p=0.03; 2-fold at day 126, p=0.009) (Fig. 6J). GFAP fluorescence in the VTA were similar at both time points when compared to uninfected mice (p=0.31 at day 21; p=0.08 at day 126) (Fig. S6H). Together these data suggest that astrocytes in CT mice become reactive in the DLS and at later stages of infection, in the SNc following Mtb infection. This pathology mirrors the progression of neurodegeneration in idiopathic PD(Stephens et al., 2005; Villalba et al., 2009; Zaja-Milatovic et al., 2005). In addition, the reactivity of DLS astrocytes following Mtb infection is dependent on LRRK2 expression, while SNc astrocytes do not depend on LRRK2 for induction of reactivity following Mtb infection.
We also found an effect of age on astrocyte reactivity in the DLS and SNc. Astrocyte reactivity in uninfected mice showed a significant age-dependent reduction in the DLS and SNc of mice such that old mice showed significantly lower GFAP fluorescence than young mice (p<0.001 for DLS; p=0.023 for SNc; ANOVA) (Fig. S6I). This age-dependent reduction was maintained in the DLS of TB infected mice from day 21 to 126. We observed that Mtb infection caused a loss of age-dependent GFAP reduction in the SNc and VTA (Fig. 6K and S6J). Together these data suggest that systemic TB infection dramatically alters the reactivity profile of astrocytes in the DLS, SNc and VTA and that astrocyte reactivity following Mtb infection is dependent on the expression of LRRK2.
Loss of LRRK2 impacts the capacity of astrocytes to respond to stimuli ex vivo
Because we observed dynamic changes to astrocyte reactivity and microglial activation during the course of Mtb infection in vivo we wanted to better understand the response of these cells ex vivo, especially in terms of the mitochondrial and type I IFN phenotypes we uncovered in LRRK2 KO peripheral macrophages. To this end, we differentiated primary cell cultures enriched in astrocytes and microglia from the brains of neonatal CT and LRRK2 KO mice. As expected, astrocyte cultures were positive for Gfap mRNA whereas microglial cultures expressed Iba1 (Fig. 7A), confirming successful enrichment. LRRK2 KO astrocyte cultures expressed a mild increase in Gfap and Ccl5 mRNA over CT ex vivo (Fig. 7B), indicating an increased reactivity at rest. Remarkably, when stimulated with IFN-β directly or the supernatants of CT macrophages infected with Mtb, LRRK2 KO astrocyte cultures selectively failed to upregulate these same markers to the extent of CT astrocytes (Fig. 7C, S7A). As we observed in LRRK2 KO BMDMs, LRRK2 KO astrocytes also expressed increased ISGs at rest, failed to robustly upregulate these genes upon stimulation (Fig. 7D), and were sensitive to mitochondrial depolarizing agents relative to CT astrocytes (Fig 7E). Taken together, these results strongly suggest that LRRK2 KO astrocytes are defective in many of the same ways we observed for peripheral macrophages and begin to elucidate how peripheral infection, coupled with genetic defects, may precipitate neuroinflammation.
We also examined LRRK2 KO microglia ex vivo. LRRK2 KO microglia exhibited a modest reduction of Iba1 protein levels, compared to CT, upon stimulation with IFN-β (Fig. 7G), consistent with in vivo results at D126 post-Mtb infection (Fig. 6F). Surprisingly, LRRK2 KO microglia had a reduced type I IFN signature at rest and dramatically upregulated ISGs upon stimulation (Fig. 7G). Consistent with this phenotype, LRRK2 KO microglia were not sensitive to depolarizing stressors (Fig. 7H). LRRK2 KO microglia also significantly upregulated proinflammatory chemokines Ccl5 and KC over time (Fig. 7I), but failed to upregulate Tnf to the extent of LRRK2 CT microglia (Fig. 7J), indicting that LRRK2 KO microglia may have an alternative proinflammatory polarization compared to CT cells. Together, these data suggest that LRRK2 KO microglia and astrocytes have altered ability to sense and respond to innate immune cues.
DISCUSSION
Despite being repeatedly associated with susceptibility to mycobacterial infection and other inflammatory disorders in GWAS studies, very little is known about how Lrrk2, a massive, multifunctional protein, functions outside of the central nervous system. Here, we provide evidence that loss of LRRK2 influences the ability of immune cells—both in the periphery and in PD relevant regions of the brain—to respond to and express inflammatory molecules. During Mtb infection, in peripheral organs like the lungs, these defects manifest themselves at the level of local neutrophil/macrophage infiltration and necrosis without significantly impacting bacterial replication. In the brains of Mtb-infected mice, loss of LRRK2 sensitizes glial cells like astrocytes and microglia, inducing a hyper-reactive phenotype even when these cells are exposed to the same circulating cytokine milieu as CT mice. Together, these results argue strongly for the “multiple-hit hypothesis” of neurodegenerative disease, whereby genetic susceptibility (e.g. loss of LRRK2) coupled with environmental stressors (e.g. Mtb infection (Shen et al., 2016), mitochondrial stress (Tanner et al., 2011), exhaustive exercise (Sliter et al., 2018)) can trigger neuroinflammation and potentially cause downstream damage to neurons (Balin and Appelt, 2001; Patrick et al., 2019).
We propose that dysregulation of type I IFN expression in LRRK2 KO macrophages is the consequence of two distinct cellular defects conferred by loss of Lrrk2. First, in the absence of Lrrk2, decreased levels of purine metabolites and urate contribute to oxidative stress, leading to damage of the mitochondrial network. A recent human kinome screen identified Lrrk2 as a kinase involved in dynamics of the purinosome, a cellular body composed of purine biosynthetic enzymes that assembles at or on the mitochondrial network (French et al., 2016). Specifically, shRNA knockdown of Lrrk2 in HeLa cells inhibited purinosome assembly and disassembly. As purinosomes are posited to form in order to protect unstable intermediates and increase metabolic flux through the de novo purine biosynthetic pathway (An et al., 2008; Schendel et al., 1988; Zhao et al., 2013), we propose that Lrrk2-dependent defects in purinosome assembly lead to lower levels of IMP and hypoxanthine. Lower levels of these purine nucleotide intermediates in LRRK2 KO macrophages are especially notable in the context of PD, as the plasma of patients (both Lrrk2 mutant and idiopathic) has been shown to contain significantly less hypoxanthine and uric acid levels (uric acid/urate being the final product of the purine biosynthetic/salvage pathway (Fig. 5C) (Johansen et al., 2009) and Lrrk2 mutation carriers with higher urate plasma levels are less likely to develop PD (Bakshi et al., 2019). Indeed, urate is currently being investigated as a potential therapeutic of PD, highlighting the importance of purine biosynthesis in maintaining healthy neurons.
Second, we propose that loss of LRRK2 contributes to type I IFN dysregulation through phosphorylation of the mitochondria-associated fission protein Drp1. Previous reports have shown that Lrrk2 can physically interact with Drp1 and that Lrrk2 mediates mitochondrial fragmentation through Drp1 (Bakshi et al., 2019). Overexpression of both wild type Lrrk2 and the G2019S allele have been shown to cause mitochondrial fragmentation (X. Wang et al., 2012). Curiously, we observe a similar phenotype in macrophages lacking LRRK2 (Fig. 3C), suggesting that the balance of Lrrk2 protein levels is crucial for maintenance of the mitochondrial network. Treatment of a microglia cell line (BV2) with the TLR4 agonist LPS has previously been shown to enhance mitochondrial fission and neuroinflammation, which Ho et al. propose occurs by increasing Lrrk2 and DRP1 levels (Ho et al., 2019; Perez-Carrion et al., 2018; Su and Qi, 2013; X. Wang et al., 2012). These results, coupled with our own observations that the Mtb-induced peripheral cytokine milieu can activate microglia and astrocytes, begin to paint a complex picture whereby tipping the balance of Lrrk2 and Drp1 levels in any number of ways can trigger a pathogenic feedback loop, leading to fragmentation of mitochondria and activation of type I IFN responses. In the absence of Lrrk2, lower antioxidant levels (via aforementioned purinosome abnormalities) likely exacerbate this defect, leading to higher oxidative stress and additional damage to the mitochondrial network.
Although we observed a striking type I IFN defect (both higher basal levels and an inability to induce ISG expression downstream of cytosolic nucleic acid sensing and/or engagement of IFNAR) in a number of primary cells and cell lines, we did not measure a major difference in IFN-β expression, either in circulating serum or in select tissues, at select key time points during Mtb infection. These results were surprising to us, as Mtb is a potent activator of cytosolic DNA sensing (Manzanillo et al., 2012; Watson et al., 2015) and type I IFN is an important biomarker of Mtb infection associated with poor outcomes (Berry et al., 2010). We previously observed an apparent disconnect between type I IFN expression defects in Mtb-infected macrophages and Mtb-infected mice, as loss of the cytosolic DNA sensor cGAS almost completely abrogates type I IFN expression in macrophages but only has minor effects in the serum and lungs in whole KO mice (Watson et al., 2015), suggesting that mechanisms exist in vivo to overcome even major defects in nucleic acid sensing and downstream type I IFN expression. Importantly, consistent with our macrophage data, another recent publication that investigated the role of LRRK2 in controlling Mtb infection reported a significant decrease in IFN-α in the lungs of LRRK2 KO infected animals at Day 56 post-infection (Härtlova et al., 2018).
Because our wild type and LRRK2 KO mice were able to control Mtb replication to similar levels (no significant differences in CFUs), the changes we observe in the brains of these mice cannot be attributed to bacterial loads. Nor can they be readily attributed to differences in circulating cytokines, although it is possible that cytokine levels are significantly different at time points other than those that we measured (as is the reported in Hartlova et al.). Therefore, it is rather remarkable that we observed an increase in microglial reactivity in Mtb infected CT and LRRK2 KO mice in the DLS, a region that is implicated in the initiation of PD (Villalba et al., 2009). Our data also suggest that reactive microglia are intimately associated with TH+ neuronal debris in the SNc of Mtb-infected mice (Fig. 6F). It is possible that the microglia are phagocytosing pieces of damaged neurons or perhaps damaging the neurons themselves. In either case, the fact that this behavior is only apparent in the brains of Mtb-infected mice strongly suggests that peripheral infection alters interactions between microglia and neurons in potentially pathological ways. It is tempting to speculate that this increase in reactive microglia serves as a mechanism by which persistent infections can precipitate neurodegeneration (Patrick et al., 2019).
In contrast, astrocytes in the DLS become reactive with Mtb infection only in CT mice and not LRRK2 KO mice. This suggests that astrocytes in the DLS depend on LRRK2 to initiate reactivity, while microglia do not require LRRK2 to respond to Mtb infection. Indeed, LRRK2 mRNA expression in astrocytes in mice is ∼6 fold higher than microglia (Y. Zhang et al., 2014). We observed that Mtb infection increases microglial and astrocytic reactivity at later time points (126 days) in the SNc when compared to the DLS. Taken together with our data that KO of LRRK2 in macrophages and MEFs increases fragmentation of peripheral mitochondria, the differential profile of reactivity in astrocytes between DLS and SNc strongly suggest that astrocytic mitochondria in the SNc are more resilient and likely functionally different from astrocytic mitochondria in the DLS.
These data are the first to directly connect LRRK2’s role in maintaining mitochondrial homeostasis to its emerging role in inflammation, both in the brain and in the periphery. Given the striking phenotypes we observed in the brains of Mtb-infected LRRK2 KO mice, it is tempting to speculate that LRRK2s contribution of neuroinflammation and glial cell activation is a major driver of PD, thus opening the door for novel immune-targeted therapeutic interventions designed to halt or slow neurodegenerative disease progression.
METHODS
Mice
LRRK2 KO mice (C57BL/6-Lrrk2tm1.1Mjff/J) stock #016121, and IFNAR KO mice (B6(Cg)-Ifnar1tm1.2Ees/J) stock #028288 were purchased from The Jackson Laboratories (Bar Harbor, ME). Tfam Het (Woo DK 2012), and cGAS KO (B6(C)-Cgastm1d(EUCOMM)Hmgu/J) mice were provided by A. Phillip West, Texas A &M Health Science Center (Bryan, TX). All mice used in experiments were compared to age and sex matched controls. In order to ensure littermate controls were used in all experiments LRRK2 KO crosses were made with (Het) Lrrk2−/− x (KO) Lrrk2+/− mice. Mice used to generate BMDMs and PEMs were between 8-12 weeks old. Mice were infected with Mtb at 10 weeks. Mice used to make glial cultures were P0.5 days old. Embryos used to make primary MEFs were at 14.5 days post coitum. All animals were housed, bred and studied at Texas A&M Health Science Center under approved IACUC guidelines.
M. tuberculosis infections
The Erdman strain was used for all M. tuberculosis infections. Low passage lab stocks were thawed for each experiment to ensure virulence was preserved. M. tuberculosis was cultured in roller bottles at 37°C in Middlebrook 7H9 broth (BD Bioscienes) supplemented with 10% OADC, 0.5% glycerol, and 0.1% Tween-80 or on 7H11 plates (Hardy Diagnostics). All work with M. tuberculosis was performed under Biosafety level 3 containment using procedures approved by the Texas A&M University Institutional Biosafety Committee.
Prior to infection, BMDMs were seeded at 1.2×106 cells/well (6-well dish) or 3×105 cells/well (12-well dish), RAW cells at 5×105 cells/well (12-well dish), and U937s at 1×106 cells/well. U937s were cultured with 10ng/ml phorbol 12-myristate 13-acetate (PMA) for 48 hr to induce differentiation and then recovered in fresh media for an addition 24 hr prior to infection.
To prepare the inoculum, bacteria grown to log phase (OD 0.6-0.8) were spun at low speed (500g) to remove clumps, and then pelleted and washed with PBS twice. Resuspended bacteria were briefly sonicated and spun at low speed once again to further remove clumps. The bacteria were diluted in DMEM + 10% horse serum and added to cells at an MOI of 10. Cells were spun with bacteria for 10 min at 1000g to synchronize infection, washed twice with PBS and then incubated in fresh media. RNA was harvested from infected cells using 0.5-1ml Trizol reagent 4h post-infection unless otherwise indicated.
Mouse infections
All infections were performed using procedures approved by Texas A&M University Institutional Care and Use Committee. The M. tuberculosis inoculum was prepared as described above. Age- and sex-matched (approximately 3-month-old male and female littermates) were infected via inhalation exposure using a Madison chamber (Glas-Col) calibrated to introduce 100-200 CFUs per mouse. For each infection, approximately 5 mice were euthanized immediately, and their lungs were homogenized and plated to verify an accurate inoculum. Infected mice were housed under BSL3 containment and monitored daily by lab members and veterinary staff.
At the indicated time points, mice were euthanized, and tissue samples were collected. Organs were divided in order to maximize infection readouts (CFUs: left lobe lung and ½ spleen; histology: 2 right lung lobes and ¼ spleen; RNA: 1 right lung lobe and ¼ spleen). For histological analysis organs were fixed for 24 hr in neutral buffered formalin and moved to ethanol (lung, spleen) or 4% paraformaldehyde and moved to 30% sucrose (brain). Organs were further processed as described below. For cytokine transcript analysis, organs were homogenized in Trizol, and RNA was isolated as described above. For CFU enumeration, organs were homogenized in 5 ml PBS + 0.1% Tween-80, and serial dilutions were plated on 7H11 plates. Colonies were counted after plates were incubated at 37° for 3 weeks. Blood was collected in serum collection tubes, allowed to clot for 1-2 hr at room temperature, and spun to separate serum. Serum cytokine analysis was performed by Eve Technologies.
Histopathology
Lungs and spleens were fixed with paraformaldehyde, subjected to routine processing, embedded in paraffin, and 5-µm sections were cut and stained with hematoxylin and eosin (H&E) or acid-fast stain (Diagnostic BioSystems). A boarded veterinary pathologist performed a masked evaluation of lung sections for inflammation using a scoring system: score 0, none; score 1, up to 25% of fields; score 2, 26-50% of fields; score 3, 51-75% of fields; score 4, 76-100% of fields. To quantify the percentage of lung fields occupied by inflammatory nodules, scanned images of at least 2 sections of each lung were analyzed using Fiji Image J (Schindelin et al., 2012) to determine the total cross-sectional area of inflammatory nodules per total lung cross sectional area. For acid fast staining, one brain hemisphere was fixed with paraformaldehyde for 48 hours, then transferred to a cryoprotective buffer (30% sucrose in a phosphate buffer), and frozen for coronal slicing into 40-µm sections. At least two sections per mouse were stained with an acid-fast stain (Diagnostic BioSystems) according to the manufacturer’s instructions and visualized by an Olympus BH2 light microscope.
Tissue Immunohistochemistry
Immunohistochemistry was performed on uninfected C57/BL6NJ mice and mice that were infected with mycobacterium tuberculosis (TB). Two genotypes of mice were utilized for these experiments: heterozygous LRRK2+/− (Het) or knockout LRRK2−/− (KO). Brains from Het (CT) and KO mice were collected on days 7, 21, or 126 after initiation of tuberculosis infection. To specifically evaluate the effect of TB on astrocytes and microglia, we collected brains from sham infected Het and KO mice 21 and 126 days after sham infection.
Striatal and midbrain sections from these mice were stained for glial fibrillary acid protein (GFAP) to label astrocytes, Iba-1 to label microglia, and NeuN for neurons in striatal slices. In addition to Iba1 and GFAP, tyrosine hydroxylase (TH) labeling was used for visualizing dopaminergic neurons in the midbrain.
For immunohistochemistry, mice were anesthetized with isoflurane and quickly decapitated. The brain was gently removed from the skull and postfixed in 4% paraformaldehyde overnight at 4°C. The tissue was cryoprotected in 30% sucrose + PBS solution for 48-72 hours. 40 µm thick coronal sections were obtained using a cryostat microtome (Leica) and preserved in 0.01% sodium azide + PBS at 4°C.
Immunohistochemistry was performed using previously published techniques (Srinivasan et al., 2015; 2016). Briefly, sections were washed 3x for 10 min in 1X TBS, then blocked for 1 hr in 5% NGS and 0.25% Triton-X-100 in 1X TBS at RT. Sections were incubated overnight at 4°C in primary antibodies diluted in blocking solution. The following primary antibodies were used: rabbit anti-GFAP (1:1000; Abcam ab7260), rabbit anti-Iba1 (1:250; Wako Chemical 019-19741), mouse anti-NeuN (1:500; Abcam ab104224), and chicken anti-TH (1:1000; Abcam ab76442). The following day sections were washed 3x for 10 min each in 1X TBS and incubated with appropriate secondary antibodies in blocking solution for 2 hr at RT. The following secondary antibodies were used: goat anti-rabbit (1:1000; Abcam ab150077), goat anti-mouse (1:1000; Abcam ab150120), and goat anti-chicken (1:1000; Abcam ab150176). The sections were rinsed 3x for 10 min in 1X TBS and then mounted on microscope slides in Fluoromount (Diagnostic Biosystems; K024) and coversliped for imaging.
Tissue Imaging
Images were obtained using a FV 1200 Olympus inverted confocal microscope equipped with 20x, 0.85 NA oil immersion objective, 473 nm, and 561 nm laser lines to excite appropriate Alexa Fluor secondary antibodies. Images were obtained at 1x digital zoom. HV, gain, and offset was adjusted so that fluorescent signals from images were just below saturation. Laser power for 473 and 561 excitation lines were maintained between 2-3% of maximum. All images were acquired as z-stacks with a 1 µm step size and stack sizes ranged between 25-30 µm. Parameters were kept constant for all mice in an experimental group. We define experimental groups based on the mouse genotype, infection status and the timepoint at which the brain was extracted after infection or sham infection.
Tissue Image Analysis
Images were processed using ImageJ. For image analysis, maximum intensity projections of z-stacks were first obtained. Projected images were thresholded such that GFAP staining in astrocytic cell bodies or Iba-1 staining in microglial cell bodies along with branches (1° and 2°) were masked and ROIs were obtained in this way. In each case, corresponding NeuN labeled or TH labeled sections were processed in a similar manner to astrocytic and microglial staining. Integrated density values were extracted from astrocytic, microglial, and corresponding neuronal components of each slice. Ratios of astrocytic or microglial integrated density to respective neuronal integrated density (NeuN/TH) were obtained. Ratios obtained in this way were averaged across each brain region and all slices for each mouse. By utilizing ratios of glial signal to neuronal staining intensity, we controlled for differences between individual sections that occur due to variations in the efficiency of antibody binding or tissue quality. Data are presented as averages for each mouse. Mean values ± s.e.m. from the averages are presented.
M. Leprae Infections
M. leprae cultivated in the footpads of nude mice was generously provided by the National Hansen’s Disease Program. Bacilli were recovered overnight at 33° C, mixed to disperse clumps and resuspended in DMEM supplemented with 10% horse serum. Media were removed from BMDM cells and monolayers overlaid with the bacterial suspension and centrifuged for 10 min at 1000 RPM. Cells were washed twice in PBS and returned to complete media.
Primary Cell Culture
Bone marrow derived macrophages (BMDMs) were differentiated from BM cells isolated by washing mouse femurs with 10ml DMEM. Cells were then centrifuged for 5 min at 1000 rpm and resuspended in BMDM media (DMEM 20% FBS (Millipore) 1mM Sodium pyruvate 10% MCSF conditioned media (Waston lab)). BM cells were counted and plated at 5×106 in 15cm non-TC treated dishes in 30ml complete media. Cells were fed with an additional 15ml of media on day 3. Cells were harvested on day 7 with 1XPBS EDTA. Mouse embryonic fibroblasts (MEFs) were isolated from embryos. Briefly, embryos were dissected from yolk sacs, washed 2 times with cold 1XPBS, decapitated, and peritoneal contents were removed. Headless embryos were disagreggated in cold 0.05% trypsin-EDTA (Lonza) and incubated on is for 20min. followed by incubation at 37°C for an additional 20min. Cells were then DNAse treated with 4ml dissagregation media (DMEM 10% FBS 100ug/ml DNAse) for 20min at 37C. Supernatants were removed and spun down at 1000rpm for 5min. Cells were resuspended in DMEM 10% FBS, 1mM sodium pyruvate, and plated in 15cm TC treated dishes 1 dish per embryo. MEFs were allowed to expand for 2-3 days before harvest with Trypsin 0.05% EDTA. Mixed glial cultures were differentiated from the brains of neonatal mice as described (Lian 2016 Bio Protoc). Microglial cells were differentiated using complete media DMEM 10%FBS 1mM Sodium pyruvate with 10% MCSF conditioned media. Peritoneal macrophages (PEMS), were elicited by intraperitoneal injection of 1ml 3% Thioglycollate broth (BD) for 4 days prior to harvest. For harvest PEMs were isolated from mice by lavage (1X PBS 4°C) and resuspended in RPMI 1640 media with 20% FBS, 1mM Sodium pyruvate(Lonza) and 2mM L-Glutamine(Lonza). Following overnight incubation at 37°C, cells were washed twice (1X PBS 37°C) to remove non-adherant cells (25%).
Cell lines and Treatments
RAW264.7 LRRK2 KO cells (ATCC® SC-6004™) generated by the MJFF, were obtained from the ATCC and used with wild type control LRRK2 parental RAW 264.7 (ATCC® SC-6003™). To deplete mtDNA RAW264.7 cells were seeded at 2×106 cells/well in 10cM non-TC treated dishes and cultured for 4 days in complete media (DMEM 10%FBS 1mMSodium pyruvate) with 300ng/ml Ethidium bromide or 10uM ddC. Cells were split and harvested with 1X PBS EDTA.
Cell stimulations
BMDMs were plated in 12 well plates at 5×105 cells/well, or 6well plates at 1×106 cells/well. MEFs were plated in 12 well dishes at 3×105 cells/well. PEMs were plated in 24-well flat-bottomed plates (Corning) at 1×106 cells/ well. RAW264.7 cells were plated at 7.5×105 cells/well. Astrocyte cultures were plated at 2.5×104 cells/well in 12 well dishes. Microglia were plated at 5×105 cells/well in 12well plates. Cells were stimulated for 4 hrs with 1µM CLO97, 100ng/ml LPS, 10µM ABT737/10uM OVD-PH, or transfected 1ug/ml ISD, 1ug/ml PolyI:C, 1ug/ml cGAMP with lipofectamine. Cells were transfected for 4hrs with 1uM CpG 2395 with Gene Juice. Cells were stimulated for 2hrs with 10µM DMXAA (RAW cells) 50nM (BMDMs), or 200IU IFNB (BMDMs). Glial cells were stimulated for 8 or 24 Hrs with 800IU IFNB, or supernatants from Mtb infected control BMDMs at a 1:5 dilution.
mRNA Sequensing
RNA was isolated using PureLink RNA mini kits (Ambion) quantified on a Bioanalyzer 2100 (Agilent). PolyA+ PE 100 libraries were sequenced on a HiSeq 4000 at the UC Davis Genome Center DNA Technologies and Expression Analysis Core. Heatmaps were generated by performing Cluster analysis (Cluster3) followed by Java TreeView. Transcriptome analysis was performed using IPA analysis to generate GO term and disease pathway lists. Instant Clue was used to generate scatter plots and volcano plots.
qRT-PCR
RNA was isolated using Directzol RNAeasy kits (Zymogen). cDNA was made with iScript direct synthesis (BioRad) per manufacturers protocol. qRT-PCR was performed in triplicate using Sybr green Power-up (ThermoFisher). Data was analyzed on a ViiA 7 Real-Time PCR System (Applied Biosystems). See extended methods for list of primer sequences.
Cytosolic DNA isolation
MEFs were plated in 10cm dishes at 3×106. The next day, confluent plates were treated as indicated with inhibitors. To harvest, cells were lifted with PBS+EDTA. To determine total DNA content, 1% of the input was saved and processed by adding NaOH to 50 mM, boiling 30 min, and neutralizing with 1:10 1 M Tris pH 8.0. To isolate cytosolic DNA, the cells were pelleted and resuspended in digitonin lysis buffer (150 mM HEPES pH 7.4, 50 mM NaCl, 10 mM EDTA, 25 µg/ml digitonin). Cells were incubated for 15 at 4°C on an end-over-end rotator. Cells were spun at 980 x g for 3 min, and the DNA from the supernatant (cytosolic fraction) was then extracted via phenol:chloroform (1:1 supernatant:phenol/chloroform). The DNA from the aqueous layer was precipitated in 0.3M sodium acetate, 10 mM magnesium chloride, 1µg/ml glycogen, and 75% ethanol. After freezing overnight at −20°C, the DNA was pelleted, washed in 70% ethanol, dried, resuspended in TE, and solubilized at 50°C for 30 min. qPCR was performed on the input (1:50 dilution) and cytosolic (1:2 dilution) samples using nuclear (Tert) and mitochondrial (16s and cytB) genes. The total and cytosolic mitochondrial DNA was normalized to nuclear DNA in order to control for variation in cell number.
Western blot
Cells were washed with PBS and lysed in 1X RIPA buffer with protease and phosphatase inhibitors (Pierce). DNA was degraded using (units) benzonase (EMD Millipore). Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked for 1hr RT in Odessy blocking buffer (Licor) Blots were incubated o/n at 4°C with the following Abs. IRF3 (Cell Signaling) 1:1000, pIRF3 Ser396 (Cell Signaling 4947) 1:1000, Iba1 (Wako Chemical 019-19741) 1:2000, Beta Actin 1:5000 (abcam), tubulin 1:5000 (abcam). Membranes were incubated with respective secondary Ab for 2hrs at RT prior to reading on an Odyssey Fc Dual-Mode Imaging System (Licor).
Seahorse metabolic assays
Seahorse XF mito stress test kits and cartridges (Agilent) were prepared per manufacturers protocol as described here (Bossche 2015) and analyzed on a Seahorse XF 96well Analyzer(Agilent). BMDMs were seeded at 5×104 cells/well o/n and treated with 200µM mitoTEMPO, IFNB neutralizing Ab, or Sodium pyruvate at 0mM, 1mM, or 2mM final concentration.
Immuno-multiplex assay
Sera was analyzed by Eve Technologies: Mouse Cytokine Array / Chemokine Array 13-plex Secondary Panel (MD13). Briefly, sera was isolated following decapitation in Microtainer serum separator tubes (BD Biosciences) followed by 2x sterile filtration with Ultrafree-MC sterile fiters, 10min at 10,000rpm (Millipore Sigma). For analysis sera was prediluted 1:1 to a final volume of 100ul in 1xPBS pH 7.4 and assayed/analyzed in duplicate per manufacturer.
Flow cytometry
JC-1 assay to assess mitochondrial membrane potential
Cells were lifted off culture plates with 1X PBS EDTA (BMDMs, RAW264.7 and microglia) or Accutase (Biolegend) (MEFs and Astrocytes). Single cell suspensions were made in 1X PBS 4% FBS. JC-1 dye (ThermoFisher) was sonicated for 5 minutes with 30 second intervals. Cells were stained for 30min at 37°C in 1µM JC-1 dye and analyzed on an LSR Fortessa X20 (BD biosciences). Aggregates were measured under Texas Red (610/20 600LP). monomers were measured under FITC (525/50 505LP). To assess mitochondrial membrane potential under stress, cells were treated for 3 hrs with rotenone 2.5uM prior to being lifted of the culture plates followed by 5uM ATP for 5, 15, and 30 min. For rescue assays cells were treated o/n with mitoTEMPO (Sigma Aldrich) 200uM, or urate (Sigma Aldrich) 100uM.
Phospho-DPR1 assay
Cells were washed once in 1X PBS and Fixed in 4% cold PFA for 10 min cells were then permeabilized with 0.3% Triton-X for 15 minutes followed by 30 min block in 0.1% Triton-X 5% normal rat serum (Stem Cell Technologies). Cells were incubated in pDrp1 Ab o/n at 4°C in 0.1% triton-X 1% BSA. Ab AF488 Goat anti Rabbit were used as secondary antibodies, Cells were analyzed on an LSR Fortessa X20 (BD biosciences). Cells were measured under FITC (525/50 505LP). For rescue and exacerbation assays Cells were treated with H2O2 100µM for 1hr at 37°C or for 12 hrs in the presence of The Drp1 inhibitor Mdivi-1 (50uµM).
LCM/MS/MS
Sample Extraction
Samples were weighed and extracted with a methanol:chloroform:water based extraction method. Briefly 800 uL ice cold methanol:chloroform (1:1, v:v) was added to samples in a bead based lysis tube (Bertin, Rockville, MD). Samples were extracted on a Precyllys 24 (Bertin) tissue homogenizer for 30 seconds at a speed of 6000. The supernatant was collected and samples were homogenized a second time with 800 µL ice methanol:chloroform. 600 µL ice cold water was added to the combined extract, vortexed and centrifuged to separate the phases. The upper aqueous layer was passed through a 0.2 µm nylon filter (Merck Millipore, Burlington, MA). 500 µL of the filtered aqueous phase was then passed through a 3 kDa cutoff column (Thermo Scientific) and the flow through was collected for analysis.
Sample Analysis
Untargeted liquid chromatography high-resolution accurate mass spectrometry (LC-HRAM) analysis was performed on a Q Exactive Plus orbitrap mass spectrometer (Thermo Scientific, Waltham, MA) coupled to a binary pump HPLC (UltiMate 3000, Thermo Scientific). For acquisition the Sheath, Aux and Sweep gasses were set at 50, 15 and 1 respectively. The spray voltage was set to 3.5 kV (Pos) or 2.8 kV (Neg) and the S-lens RF was set to 50. The source and capillary temperatures were maintained at 350ᵒC and 350ᵒC respectively. Full MS spectra were obtained at 70,000 resolution (200 m/z) with a scan range of 50-750 m/z. Full MS followed by ddMS2 scans were obtained at 35,000 resolution (MS1) and 17,500 resolution (MS2) with a 1.5 m/z isolation window and a stepped NCE (20, 40, 60). Samples were maintained at 4 °C before injection. The injection volume was 10 µL. Chromatographic separation was achieved on a Synergi Fusion 4µm, 150 mm x 2 mm reverse phase column (Phenomenex, Torrance, CA) maintained at 30 °C using a solvent gradient method. Solvent A was water (0.1% formic acid). Solvent B was methanol (0.1% formic acid). The gradient method used was 0-5 min (10% B to 40% B), 5-7 min (40% B to 95% B), 7-9 min (95% B), 9-9.1 min (95% B to 10% B), 9.1-13 min (10% B). The flow rate was 0.4 mL min−1. Sample acquisition was performed Xcalibur (Thermo Scientific). Data analysis was performed with Compound Discoverer 2.1 (Thermo Scientific).
Statistical analysis
All data are representative of 2 or more independent experiments with an n=3 or 4. Graphs were generated using Prism (GraphPad). Significance for assays were determined using a student’s two-tailed t test, or a one way ANOVA followed by a Bonferroni’s multiple comparisons test for more than two variables, unless otherwise noted. Error bars represent SEM. Statistical tests for brain sections were run in OriginPro 2019. For each experimental group 5-8 mice were utilized. Eight sections per mouse were used for respective antibody combinations 1) GFAP/NeuN 2) Iba-1/NeuN 3) GFAP/TH and 4) Iba-1/TH, for 2 sections per combination. Resulting in a total of 10-16 brain sections, which represent all of the mice within an experimental group. Utilizing the total number of sections as the sample size for each experimental group, we obtained a power of 1. For statistical comparison each experimental group was tested for normal distribution. Normally distributed sets of data were compared using student’s two-tailed t test, while non-normally distributed data were tested using a two-tailed Mann Whitney’s test. Difference were considered statistically significant if p < 0.05.
ACKNOWLEDGEMENTS
We would like to thank Cory Klemashevich at the TAMU Integrated Metabolomics Analysis Core for his help with the metabolomics analysis. We would also like the acknowledge Monica Britton at the University of California, Davis DNA Technologies & Expression Analysis Core Library for her help with the RNA-seq analysis. We would like to thank A. Phillip West and the West lab for giving suggestions in experimental design regarding mitochondria and for providing us with Tfam Het and cGAS ko mice. We’d lastly like to thank Nevan Krogan at University of California, San Francisco for his help with the conceptual design of this manuscript and other members of the Patrick and Watson labs for discussions and feedback.