Bisphosphonate drugs have actions outside the skeleton and inhibit the mevalonate pathway in alveolar macrophages

Bisphosphonates drugs target the skeleton and are used globally for the treatment of common bone disorders. Nitrogen-containing bisphosphonates act by inhibiting the mevalonate pathway in bone-resorbing osteoclasts but, surprisingly, also appear to reduce the risk of death from pneumonia. We overturn the long-held belief that these drugs act only in the skeleton and show that a fluorescently-labelled bisphosphonate is internalised by alveolar macrophages and peritoneal macrophages in vivo. Furthermore, a single dose of a nitrogen-containing bisphosphonate (zoledronic acid) in mice was sufficient to inhibit the mevalonate pathway in tissue-resident macrophages, causing the build-up of a mevalonate metabolite and preventing protein prenylation. Importantly, one dose of bisphosphonate enhanced the immune response to bacterial endotoxin in the lung and increased the level of cytokines and chemokines in bronchoalveolar fluid. These studies suggest that bisphosphonates, as well as preventing bone loss, may boost immune responses to infection in the lung and provide a mechanistic basis to fully examine the potential of bisphosphonates to help combat respiratory infections that cause pneumonia.


INTRODUCTION
Nitrogen-containing bisphosphonates (N-BPs) are a class of bone-seeking drugs used worldwide as the front-line treatment for disorders of excessive bone resorption such as postmenopausal osteoporosis and cancer-associated bone disease (Russell, 2011). By virtue of their avidity for calcium ions, N-BPs bind rapidly to the skeleton, where they are internalised by bonedegrading osteoclasts (Russell et al., 2008, Rogers et al., 2020. Intracellularly, N-BPs disable osteoclast function by inhibiting the enzyme farnesyl diphosphate (FPP) synthase in the mevalonate pathway (van Beek et al., 1999, Dunford et al., 2001, thereby preventing the post-translational prenylation of small GTPase proteins necessary for osteoclast function (Luckman et al., 1998, Fisher et al., 1999. There is increasing evidence that N-BP drugs, such as zoledronic acid (ZOL), have benefits beyond preventing bone loss (Center et al., 2020) and, unexpectedly, N-BP therapy has recently been linked to reduced risk of mortality from pneumonia (Colon-Emeric et al., 2010, Sing et al., 2020. In a randomised, controlled trial of >2,000 hip fracture patients, ZOL therapy reduced the risk of death by 28% compared to placebo infusion (Lyles et al., 2007). Retrospective analysis also suggested that ZOL-treated patients were less likely to die from pneumonia than placebo-treated subjects (Colon-Emeric et al., 2010). Recently, a "real-world" population-based, observational study of hip fracture patients aged 50 years or above also showed a significant reduction in risk of pneumonia and pneumonia mortality in hip fracture patients that had received N-BP therapy, compared to no treatment or other osteoporosis medications (Sing et al., 2020). Similar findings were reported in the post hoc analysis of a randomised controlled trial of ZOL in women over the age of 65 years (Reid et al., 2021). Pneumonia is the most frequent cause of admission to ICU and a study of long term patients in respiratory ICU revealed a significant reduction in mortality in people treated with the N-BP pamidronate compared to those without treatment (Schulman et al., 2016). Furthermore, in a retrospective cohort study of ICU subjects, we showed a 59% reduction in mortality in patients treated with bisphosphonate prior to hospitalisation (Lee et al., 2016). However, the mechanisms underlying the surprising beneficial effects of these drugs on pneumonia and ICU patients are unknown.
Globally, respiratory diseases constitute the most common cause of death and thus, therapies that boost the immune response to common lung infections are urgently needed. Bacterial infections such as Streptococcus pneumoniae, Haemophilus influenza, Chlamydia pneumoniae and Staphylococcus aureus are the main cause of community-acquired pneumonia. Importantly, these pathogens also underlie severe complications of viral respiratory disease that can significantly increase morbidity and mortality. For example, influenza-related mortality is often associated with pneumonia caused by co-or secondary bacterial infection (Morris et al., 2017). In this study we debunk the long-held view that N-BP drugs act only in the skeleton and show that even a single dose of N-BP in mice is sufficient to affect tissue-resident macrophages, including lung alveolar macrophages (AMf), boosting their response to bacterial endotoxin.

Systemically administered N-BP is internalised by tissue-resident macrophages outside the skeleton.
We previously reported that cultured macrophages and tumour-associated macrophages in vivo, like osteoclasts, have the ability to internalise N-BP by endocytosis (Thompson et al., 2006, Junankar et al., 2015. In this study we used a fluorescently-labelled analogue of ZOL (AF647-ZOL) to determine whether tissue-resident macrophages are capable of internalising systemically administered N-BP in mice. Given the important role of AMf in lung homeostasis and the initial immune response to respiratory infection (Byrne et al., 2015, Crane et al., 2018, we focused on whether these cells are targeted by N-BP. To answer this question, animals were injected with a single intravenous (i.v.) dose of AF647-ZOL and cells collected by bronchoalveolar lavage (BAL) were analysed by flow cytometry. In the absence of immune challenge in mice, approximately 90% of cells in BAL were AMf (Suppl Fig 1a). Importantly, >98% of AMf (TCRb -B220 -CD11b lo/-CD11c hi F4/80 + ) in BAL samples were clearly labelled with AF647-ZOL 4 hours after i.v. administration (Fig 1a-c).

Similar to BAL cells, approximately 80% of peritoneal cells incorporated AF647-ZOL after a single
i.v. dose (Fig 1e), the majority of which (99%) were CD11b hi F4/80 hi large PMf (Fig 1f,g). In contrast, the CD11b + F4/80 int small PMf incorporated negligible amounts of fluorescently-labelled ZOL ( Fig   1h). In addition to the well-described uptake of N-BP by osteoclasts in bone (Rogers et al., 2020, Coxon et al., 2008, these findings clearly demonstrate that N-BP can also be efficiently internalised in vivo by tissue-resident macrophages outside the skeleton, including AMf in the lung and LPMf in the peritoneal cavity.

A single i.v. dose of N-BP is sufficient to inhibit the mevalonate pathway in alveolar and peritoneal macrophages
To examine whether tissue-resident macrophages can incorporate sufficient N-BP in vivo to have a pharmacologic effect, we analysed two biochemical outcomes that, together, are reliable features of intracellular N-BP action in cells (Rogers et al., 2020): (i) the cytoplasmic build-up of the upstream metabolite isopentenyl diphosphate (IPP) and its isomer dimethylallyl diphosphate (DMAPP) (Raikkonen et al., 2009); and ii) reduced production of the isoprenoid lipid geranylgeranyl diphosphate (GGPP), with the consequent accumulation of unprenylated small GTPase proteins including those of the Rab and Rho superfamilies (Luckman et al., 1998) (Fig 2a). To address whether N-BP has pharmacological effects on AMf and PMf, we used liquid chromatography tandem mass spectrometry (LC-MS/MS) to examine the accumulation of IPP/DMAPP and a sensitive biochemical in vitro assay to detect changes in the level of unprenylated Rab proteins , Rogers et al., 2020.
LC-MS/MS analysis showed that IPP/DMAPP was undetectable in extracts of BAL or PL cells from saline-treated mice, but there was a clear increase in the level of IPP/DMAPP in BAL and PL cells collected 48 hours after a single i.v. injection of the N-BP ZOL (Fig 2b). Importantly, ZOL treatment also resulted in a marked accumulation of unprenylated Rab proteins in BAL and PL cell samples (Fig 2c). It is unlikely that such an effect of N-BP on protein prenylation in macrophages in vivo could have been detected using the relatively insensitive western blot approach previously employed to study bone-resorbing osteoclasts, which engulf large amounts of N-BP (Frith et al., 2001, Rogers et al., 2020. However, the development of a much more sensitive in vitro prenylation assay  now allows the detection of subtle effects on protein prenylation in cells outside the skeleton that may internalise much smaller quantities of N-BP. Together with the evidence for uptake of N-BP by large PMf and AMf (Fig 1), these findings demonstrate unequivocally for the first time that systemic administration of N-BP has pharmacological activity outside the skeleton. We show that a single dose of ZOL is sufficient to inhibit the mevalonate pathway in tissue-resident macrophages (AMf and PMf), causing a build-up of IPP/DMAPP metabolites and an accumulation of unprenylated small GTPase proteinscharacteristic hallmarks of N-BP action (Rogers et al., 2020). Although still to be confirmed in humans, our findings overturn the longstanding textbook paradigm that N-BP drugs, which have been in clinical use for several decades (Russell, 2011), act only in the skeleton.

Treatment with N-BP in vivo enhances the production of cytokines and chemokines in response to immune challenge
We recently reported that loss of protein prenylation in cultured monocytes promotes the formation of the NLRP3 inflammasome, resulting in increased caspase-1-mediated processing of pro-IL-1b following bacterial endotoxin (lipopolysaccharide/LPS) stimulation (Skinner et al., 2019).
Therefore, we next examined whether N-BP-mediated inhibition of protein prenylation in tissueresident macrophages alters the response to LPS in vivo, particularly in the lung. Mice were challenged intranasally (i.n.) or intraperitoneally (i.p.) with LPS 48 hours after i.v. administration of ZOL or saline (Fig 3a). ZOL treatment alone did not alter cell viability, the ratio or total number of macrophages recovered in BAL samples (Suppl Fig 1b,c), nor had any effect on the levels of cytokines/chemokines in BAL fluid (Fig 3b). However, i.n. LPS administration in ZOL-treated mice resulted in a significant increase (2.5-5.0 fold) in the production of proinflammatory cytokines IL-1b, IL-6, TNFa, G-CSF, GM-CSF and chemokines CXCL1, CCL2, CCL3, CCL4 and CCL5 in BAL fluid compared to control mice (Fig 3b). This increase in cytokine and chemokine release was not associated with changes in cell viability, total or relative numbers of BAL cells or AMf (Suppl Fig   1b,c).
I.p. LPS challenge of ZOL-treated mice also resulted in a significant elevation in IL-1b in peritoneal fluid, and CCL2 and CCL5 in serum, compared to controls (Fig 3c). Furthermore, peritoneal cells from ZOL-treated mice produced almost 3-times more IL-1b upon ex vivo stimulation with LPS and nigericin (a well-described NLRP3 activator) (Fig 3d), than cells from control mice.
IL-1b is primarily produced by monocyte/macrophages (Kany et al., 2019), the predominant cell type in PL (Suppl Fig 1a). Thus, our results strongly suggest that macrophages are the most likely source of IL-1b. Importantly, treating the mice with ZOL did not cause IL-1b release from peritoneal cells stimulated ex vivo with LPS in the absence of nigericin (Fig 3d), and this is in accord with our previous finding that inhibition of the mevalonate pathway alone does not trigger NLRP3 inflammasome assembly but enhances its activation (Skinner et al., 2019).
The observations described here begin to provide a plausible mechanistic explanation for the decreased risk of pneumonia mortality associated with N-BP treatment (Colon-Emeric et al., 2010, Sing et al., 2020, Reid et al., 2021. AMf are the initial line of defence against common respiratory tract infections (Byrne et al., 2015), and inhibition of the mevalonate pathway in these cells may help boost the initial response to bacterial as well as viral lung infections by a variety of routes (Fig 4).
First, uptake of N-BP into cells causes the accumulation of IPP/DMAPP (Fig 2b), which activates human Vg9Vd2-T cells (Thompson and Rogers, 2004). Vg9Vd2 are non-conventional T cells with potent anti-bacterial and anti-viral activity that recognise phosphoantigens, including IPP, derived from the mevalonate or DOXP pathways in bacterial pathogens (Tanaka et al., 1995, Jomaa et al., 1999. There is considerable interest in the use of N-BP-expanded g,d-T cells as an immunotherapy for cancer (Clezardin andMassaia, 2010, Tanaka, 2020) as well as viral diseases (Juno and Kent, 2020). Indeed, N-BP-expanded Vg9Vd2-T cells reduce disease severity and mortality from influenza A virus (IAV) infection in humanised mice (Tu et al., 2011, Zheng et al., 2015. Second, the genome of some pathogens such as IAV encodes proteins with a prenylation motif, which require the host cell's mevalonate pathway to enable prenylation and allow pathogen propagation (Marakasova et al., 2017). Agents that block the mevalonate pathway (such as statins) or that inhibit prenylation (such as lonafarnib, currently in clinical trials for hepatitis delta virus infection), have well-described anti-viral or anti-microbial effects (Parihar et al., 2019, Einav andGlenn, 2003). Intriguingly, simvastatin was shown to improve outcomes in hospitalised older adults with community-acquired pneumonia (Sapey et al., 2019). Third, inhibition of FPP synthase by N-BP in AMf mimics the decreased flux through the mevalonate pathway in macrophages in response to endogenous IFN signalling, which serves to limit viral uptake and replication by several mechanisms including the synthesis of 25-hydroxycholesterol Ghazal, 2016, Cyster et al., 2014). Fourth, lack of protein prenylation (Fig 2c) enhances NLRP3 inflammasome activation and promotes the release of IL1b (Skinner et al., 2019). IL1-b is a central mediator of the innate immune response that orchestrates the production of a cascade of cytokines and chemokines (Garlanda et al., 2013). Our observation that systemic ZOL treatment significantly enhanced the release of IL-1b and several other cytokines and chemokines in lung, peritoneum and serum after LPS challenge (Fig 3), is consistent with increased inflammasome activation. To our knowledge there is no evidence that enhanced cytokine production caused by N-BP therapy worsens lung inflammation in pneumonia patients -on the contrary, N-BP treatment appears to have a beneficial effect on pneumonia risk and mortality (Colon-Emeric et al., 2010, Sing et al., 2020, Reid et al., 2021. Finally, it is noteworthy that ZOL was recently identified, using computational biology approaches, as one of 200 clinically-approved drugs that are predicted to target pathways induced by SARS-CoV-2 and could be suitable for drug repurposing against COVID-19 (Han et al., 2021).
Epidemiological studies are therefore urgently needed to determine whether N-BP therapy alters the incidence, severity or risk of mortality from SARS-Cov-2 infection.

CONCLUSIONS
We show that systemically administered N-BP drug can inhibit the mevalonate pathway and prevent protein prenylation in tissue-resident macrophages beyond the skeleton, and this in turn enhances macrophage responsiveness to bacterial endotoxin. Our observations in mice, together with data from clinical studies in humans (Colon-Emeric et al., 2010, Sing et al., 2020, Reid et al., 2021, suggest that the beneficial effects of these drugs against pneumonia infection and mortality are, at least in part, mediated by targeting AMf, thereby boosting early immune responses in the lung. These findings are particularly relevant to the elderly population, and clinical trials are warranted to determine whether N-BP drugs, aside from preventing bone loss, can provide protection to pneumonia infection in vulnerable individuals, for instance, patients in aged care homes.  Aliquots (850µL)  Poroshell 120 EC-C18 UHPLC guard column (3.0×150 mm, 2.7 µm), maintained at 20 o C. The mobile phases were 10 mM ammonium acetate in water (A) and methanol (B), both containing 5 µM medronic acid to chelate metal ions (gradient 98% A from 0 to 3 minutes, decreased to 2% A from 3.5 to 6.5 minutes at 0.5 mL/min, then increased to 98% A at 0.4 mL/min from 6.5 to 12 minutes (total run time 12 minutes). Autosampler temperature was 4 o C. The mass spectrometer was operated in negative electrospray ionisation mode: source gas temperature was 250 o C with flow at 17 L/min, sheath gas temperature was 400 o C with flow at 12 L/min, and nebuliser pressure was 45 psi. Data were acquired in MRM (Multiple Reaction Monitoring) mode and was processed using Agilent

Animals and tissue collection
MassHunter Quantitative Analysis software version B08.00.00. By comparison with pure standard compounds (Sigma Aldrich), the isomers IPP and DMAPP eluted at the same retention time (approximately 2.4 minutes) and were calculated as total area under the curve. The limit of detection in cell extracts was 20nM.

Detection of unprenylated Rab proteins
To assess the accumulation of unprenylated Rab GTPase proteins we used an in vitro prenylation assay as previously described . Mice were treated with i.v. ZOL or saline as described above, then cell pellets were obtained by BAL or PL (each pooled from n=5 mice) and lysed by sonication in prenylation buffer (50 mm HEPES, pH 7.2, 50 mm NaCl, 2 mm MgCl2, assay, 10 µg of protein were incubated with recombinant GGTase II, REP-1 and biotin-conjugated GPP (a synthetic isoprenoid lipid) for the labelling of unprenylated Rab proteins .
The resulting biotinylated Rabs were then detected on PVDF blots using streptavidin-680RD (LiCOR). A narrow doublet (often appearing as a broad singlet) of endogenous biotinylated 75 kDa proteins were used as a sample loading control.

LEGEND TO SOURCE DATA
Files contain TIF images of protein blots, after in vitro prenylation of BAL or PL cell lysate samples to detect unprenylated Rab GTPases (uRabs), or western blot of conditioned media to detect IL-1b: Following the in vitro prenylation assay (described by Ali et al 2015, Small GTPases 6, 202-211) blots typically show a cluster of three bands around 23-27 kDa (unprenylated Rab proteins) and several bands of endogenous biotinylated proteins including around 40 kDa, 120 kDa, and a band around 75 kDa of an endogenous, ubiquitously expressed protein used as a loading control (Ali et al, Small GTPases 2015). Brightness and contrast of the entire blot was adjusted in Image Studio software (LiCOR) before cropping areas of uRabs and the loading control band.

Figure 3d_source data 1,2
Western blot of conditioned media using anti-IL-1b shows a single 17 kDa band of cleaved IL-1b.