Background
Spinal cord injury (SCI) triggers a CNS macrophage response consisting of pro-inflammatory, classically activated cells and anti-inflammatory, alternatively activated cells [
1,
2]. Pro-inflammatory macrophages are neurotoxic, while anti-inflammatory macrophages promote axon growth and remyelination without concurrent neurotoxicity. Unfortunately, macrophages are polarized toward a pro-inflammatory phenotype after human and rodent SCI [
1‐
4], and it is believed that these cells contribute to secondary injury processes. Indeed, TNF-α, iron, or age-related shifts toward pro-inflammatory macrophages are detrimental for SCI recovery [
5‐
7]. In contrast, increasing anti-inflammatory macrophage activation through transplantation, adoptive transfer, or selective monocyte recruitment improves functional recovery [
8‐
10]. These approaches illustrate the therapeutic potential of altering macrophage phenotypes on SCI recovery and repair; however, as a field, we are challenged to identify non-invasive, clinically viable, pharmacological techniques for altering SCI macrophage activation.
Macrophages are plastic and can adopt dynamic phenotypic and functional properties in response to new stimuli [
11]. The pro-inflammatory SCI environment potentiates a pathological macrophage phenotype [
3,
6]. However, through pharmacological interventions, it is possible to alter the way macrophages respond to pro-inflammatory stimuli. Azithromycin (AZM) is a macrolide antibiotic commonly used to treat infections in SCI individuals [
12,
13]. In addition to its antibiotic properties, AZM increases alternative macrophage activation in rodent models of lung infection, skin inflammation, and sepsis; in alveolar macrophage and human monocyte cultures when incubated with pro-inflammatory stimulants; and in humans with cystic fibrosis [
14‐
21]. Specifically, we previously observed that macrophages activated with pro-inflammatory stimuli adopt an anti-inflammatory phenotype in the presence of AZM [
14]. Despite its reported immunomodulatory effects and safe pharmacological properties, the neurotherapeutic potential of altering macrophage activation with AZM in CNS disorders and trauma has not been examined.
In the present study, we used AZM to increase pro-reparative, alternative macrophage activation in the injured mouse spinal cord. In vivo, AZM significantly increased gene expression indicative of anti-inflammatory macrophage activation and reduced macrophage pro-inflammation gene expression. AZM treatment resulted in significantly increased functional recovery and less long-term tissue damage. In vitro, AZM drove anti-inflammatory cytokine production in response pro-inflammatory stimuli and rendered pro-inflammatory macrophages non-toxic. Collectively, these data illustrate the therapeutic potential of pharmacologically manipulating macrophages in SCI and identify AZM as a novel tool and therapeutic for application in neuroinflammatory conditions.
Methods
Experimental design
Mice were treated with AZM (160 mg/kg/day) or vehicle for 3 days prior to a moderate-severe contusion SCI. Drug administration was continued daily up to 7 days post injury (dpi). At 1, 3, and 7 dpi, n = 3–5 animals/treatment group were sacrificed and spinal cord tissue harvested for cell phenotypic analysis using fluorescent-activated cell sorting (FACS). Cells sorted from 3 and 7 dpi were further phenotypically evaluated using quantitative real-time PCR (rtPCR). Locomotor analyses (Basso Mouse Scale (BMS) and gridwalk) were conducted on a separate set of animals (n = 8–10/group) over the course of 4 weeks. At 28 dpi, these animals were sacrificed and spinal cord sections generated for histological analyses of tissue sparing and macrophage phenotype. In vitro studies were conducted using bone marrow-derived macrophages (BMDM) from adult mice. BMDMs were stimulated with AZM and/or pro-inflammatory stimuli (LPS + interferon-gamma (INFγ)). Control cells were left unstimulated. Secreted interleukin (IL)-10 and IL-12 levels were determined in the BMDM supernatant through ELISA analysis. The neurotoxicity of the BMDM supernatants was determined using the MTT assay to quantify the viability of supernatant-treated Neuro-2a cells.
Animals
Experiments were performed using 4-month-old female C57BL/6 mice (Jackson Laboratory, Bar Harbor, Maine). Animals were housed in IVC cages with ad libitum access to food and water. All procedures were performed in accordance with the guidelines and protocols of the Office of Research Integrity and with approval of the Institutional Animal Care and Use Committee at the University of Kentucky.
Spinal cord injury
Animals were anesthetized via intraperitoneal (i.p.) injections of ketamine (100 mg/kg) and xylazine (10 mg/kg). Following a T9 laminectomy, a moderate-severe thoracic SCI was produced using the Infinite Horizon (IH) injury device (75-kdyn displacement; Precision Systems and Instrumentation) [
22]. Any animals receiving SCI with abnormalities in the force vs. time curve generated by the IH device were excluded from analysis. These abnormalities are indicative of bone hits or instability in the spinal cord at the time of injury and occurred <10 % of the time. Mice receiving a laminectomy without injury were used as sham controls. After injury, muscle and skin incisions were closed using monofilament suture. Post surgically, animals received one subcutaneous injection of buprenorphine-SR (1 mg/kg) and antibiotic (5 mg/kg, enroloxacin 2.27 %: Norbrook Inc., Lenexa, KS) in 2 ml of saline and were housed in warming cages overnight. Animals continued to receive antibiotic subcutaneously in 1 ml saline for 5 days. Azithromycin (160 mg/kg) or vehicle (1 % methylcellulose) was delivered in 0.1-ml volume via oral gavage daily beginning 3 days prior and continuing for 7 days post injury. Food and water intake and the incision site were monitored throughout the course of the study. Bladder expression was performed on injured mice twice daily.
Cell isolation and phenotyping for flow cytometry
Following i.p. injection of ketamine (120 mg/kg) and xylazine (10 mg/kg), mice were transcardially perfused with ice cold diethyl pyrocarbonate phosphate-buffered saline (DEPC-PBS) then 1 cm of spinal cord centered on the injury site was rapidly dissected and placed in ice cold DEPC-PBS. The tissue was dissociated on ice using a size 40 mesh cell dissociation kit (Sigma, S0770) and rinsed twice with PBS. The dissociated tissue was then passed through a 70-μm screen filter (BD:352350). Cells were centrifuged at 200×
g for 10 min at 4 °C, resuspended in fetal bovine serum (FBS) staining buffer (BD: 554656), and then cell numbers for each animal acquired using a hemocytometer. Cells were incubated with Fc block (BD:553142) for 15 min on ice and then were incubated with CD11b-APC, GR1-PE-Cy7, CD45-PerCP-Cy5.5, and CD206 (mannose receptor)-PE antibodies (BD Biosciences) as previously described [
17]. Cell were washed twice with FBS staining buffer and resuspended in appropriate volumes of FBS staining buffer for fluorescent-activated cell sorting (FACS) analysis. Expression of these surface receptors was determined using an iCyt Synergy sorter system (Sony) in the UK Flow Cytometry Core Facility. Microglia, macrophages, and neutrophils were identified by CD11b
+/CD45
lo/GR1
lo/neg, CD11b
+/CD45
hi/GR1
lo/neg, and CD11b
+/CD45
hi/GR1
hi expressions, respectively [
5,
6]. CD206 expression levels were used to determine M2-polarization states. For each antibody, gating was determined based upon appropriate negative isotype-stained controls. Flow data were analyzed using FlowJo software (Tree Star). Cell numbers for each animal were estimated from cell percentages and hemocytometer counts. All investigators involved in the flow/FACS analyses have been certified for flow research methods and applications through the completion of the Annual Course in Cytometry sponsored by the Cytometry Education Association and Verity Software House.
Gene expression from FACS-sorted cells
All FACS-sorted macrophages (CD11b+/CD45lo/hi/GR1lo/neg), which consisted of both microglia- and monocyte-macrophages, were collected in FBS staining buffer (BD:554656), and 0.75 ml TRIzol LS reagent (Life Technologies) was added per 0.25 ml of suspension. Total RNA was isolated based on the manufacturer’s protocol, with an additional phase separation using BCP, precipitation with isopropanol (Sigma-Aldrich, St. Louis, MO), and wash of the isolated RNA in 70 % ethanol. Then, 1 μg RNA was reverse-transcribed using the high-capacity complementary (cDNA) reverse transcription kit (Life Technologies). Real-time PCR amplification was performed on the mixture of 100 ng cDNA sample, Taqman Universal PCR Master Mix, and Taqman Probes (Life Technologies) using the Applied Biosystems Step One Plus Real-Time PCR System. Probes included Arg1 (Mm00475988), CD206 (Mm00485148), and CD86 (Mm00444543). Expression of genes was normalized to 18S mRNA for each sample, and reported values were calculated as 2-ΔΔCT relative to a sham reference sample.
Behavioral analysis
All experimental animals were assessed using the Basso Mouse Scale (BMS) to score hindlimb function as previously described [
23]. Mice were tested in an open field for 4 min before surgery and at 1, 3, 7, 14, 21, and 28 days post injury (dpi). Each hindlimb was scored separately based on movement (e.g., ankle placement and stepping), coordination, and trunk stability, and averaging both hindlimb scores generated a single score for each animal. A score of 0 indicated complete paralysis and a score of 9 indicated normal locomotion. Assessment of hindlimb function was also carried out using the gridwalk test [
24]. The gridwalk utilizes a horizontal ladder with stainless steel rungs 4 mm in diameter spaced 1.2 cm apart. All experimental mice were trained before injury. On 27 dpi, only mice that could support their own body weight were tested on the apparatus. Animals were videotaped and evaluated on 30 continuous rungs on the center of the ladder. Frame-by-frame video analysis was used to track the total number of hindlimb steps/footfalls.
Tissue processing and immunohistochemistry
Mice were anesthetized and then transcardially perfused with cold PBS (0.1 M, pH 7.4), followed by perfusion with cold 4 % paraformaldehyde (PFA). Dissected spinal cords (1 cm) were post-fixed for another 2 h in 4 % PFA and subsequently rinsed and stored in cold phosphate buffer (0.2 M, pH 7.4) overnight at 4 °C. On the following day, tissues were cryoprotected in 30 % sucrose for 3 days at 4 °C, followed by rapidly freezing and blocking in optimal cutting temperature (OCT) compound (Sakura Finetek USA, Inc.) on dry ice. Tissue was systematically randomized into blocks with equal group distribution to ensure uniformity of staining across groups, and blocked tissue was stored at −80 °C before sectioning. Tissue blocks were cut in serial coronal sections (10 μm) and mounted onto Colorfrost plus slides (Fisher #12-550-17). Spinal cord sections were stained for glial fibrillary acidic protein (GFAP) or neurofilament (NF) to measure tissue sparing. Slides were incubated with chicken anti-GFAP (1:200; Aves GFAP) or chicken anti-NF (1:1500, Aves NFH 0211) primary, followed by biotinylated goat anti-chicken (1:1000; Aves B-1005) then Alexa Fluor 488 (1:1000; Invitrogen S32354) secondary antibodies. Slides were coverslipped with Immu-Mount (Thermo Scientific, Waltham, MA). GFAP or NF fluorescent images were taken using a C2+ laser scanning confocal microscope (Nikon Instruments Inc., Melville, NY). To quantify spared tissue area, the regions of dense GFAP- or NF-positive staining were outlined and measured using the MetaMorph analysis program (Molecular Devices, Sunnyvale, CA).
Cell culture
Bone marrow-derived macrophages (BMDMs) were extracted from the femur and tibia of female C57BL/6 mice at 8–10 weeks of age as previously reported [
25,
26] and were plated at 0.8 ~ 1 × 10
6 cells/ml in differentiation media: Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 1 % penicillin/streptomycin, 1 % HEPES, 0.001 % β-mercaptoethanol, 10 % FBS, and 20 % supernatant from sL929 cells (a generous gift from Phillip Popovich, The Ohio State University). Supernatant collected from sL929 cells contains macrophage colony-stimulating factor, which helps to promote bone marrow cells’ differentiation into macrophages [
27]. The BMDMs were allowed to differentiate for 7 days in culture, and cells were then replated on day 7 at a density of 1 × 10
6 cells/ml in 12-well plates in differentiation media without L929 supernatant. On day 8, cells were stimulated to be M1 using LPS (50 ng/ml; Invivogen tlrl-eblps) plus IFNγ (20 ng/ml; eBioscience 14-8311-63) diluted in N2A growth medium. AZM (10 or 30 μM; Sigma PHR1088) or vehicle (DMSO; MP Biomedical 190186) was added at the time of stimulation. Unstimulated BMDMs were used as control. Six hours after incubation, the supernatant of the stimulated macrophages (macrophage-conditioned media (MCM)) was collected, filtered, then applied to cultured Neuro-2a cells or tested for IL-10 and IL-12p40 levels using standard ELISA kits (Thermo Scientific, Rockford, IL).
Mouse neuroblastoma cell lines (aka Neuro-2a or N2A, a gift from Chris Richards , University of Kentucky) were maintained in N2A growth medium containing 45 % DMEM, 45 % OPTI-MEM reduced-serum medium, 10 % fetal bovine serum (FBS), and 1 % penicillin/streptomycin. N2A were plated at a density of 1 × 10
5 cells/ml in 48-well tissue culture plates and allowed to proliferate for 48 h. The neurotoxicity of MCM was evaluated as reported previously [
7] using a MTT-based cell growth determination kit according to the manufacturer’s instructions (Sigma-aldrich). Briefly, on the day of testing, N2A growth media was replaced by fresh MCM, and the N2A cells were incubated in MCM for 24 h then thiazolyl blue tetrazolium bromide (MTT (5 mg/ml), 20 μl per well) was added to each well and the cells further incubated for 2 h. The tetrazolium ring of MTT can be cleaved by mitochondrial dehydrogenases of viable cells, yielding purple formazan crystals, which were then dissolved in acidified isopropanol solvent. The resulting purple solution was spectrophotometrically measured at 570 nm Epoch microplate reader (BioTek instruments, Inc., Winooski, VT) using 690 nm as a background absorbance. All measurements were done in triplicates, and at least three independent experiments were carried out.
Statistical analysis
Investigators blinded to experimental conditions performed all data acquisition and analysis. Statistical analyses were completed using GraphPad Prism 6.0 (GraphPad Software). Data were analyzed using one- or two-way ANOVA followed by Holm-Sidak’s test for multiple comparisons. F-values are reported for repeated measures. Chi square and independent sample t tests were used when appropriate. Results were considered statistically significant at p ≤ 0.05. All data are presented as mean ± SEM unless otherwise noted. Figures were prepared using Adobe Photoshop CS6 (Adobe Systems) and Prism 6.0.
Discussion
There is growing evidence that altering the phenotype of macrophages responding to SCI can improve recovery. Despite this, few safe pharmacological approaches have been identified that can manipulate SCI macrophages. Here, we show that in a mouse model, treatment with the macrolide antibiotic, AZM, results in increased macrophage expression of anti-inflammatory genes and facilitates significant improvements in SCI locomotor recovery and tissue sparing. Macrophages, purified from the injured spinal cord of mice treated with AZM, had increased expression of CD206 and arginase, indicators of an anti-inflammatory phenotype, with decreased expression of the pro-inflammatory marker CD86 (Fig.
1). AZM treatment significantly improved locomotor function compared to vehicle control, specifically in indices of locomotor coordination (Figs.
2 and
3). The improvements in locomotor recovery were associated with significant increases in tissue sparing (Fig.
2), presumably due to AZM reducing the neurotoxic potential of SCI-activated macrophages. Indeed, in vitro, AZM drove pro-inflammatory macrophages toward an anti-inflammatory phenotype with reduced neurotoxic properties (Fig.
4). Collectively, these data highlight the potential for an immunomodulatory, pharmacological therapy to be an effective treatment for SCI and identify AZM as a promising candidate for further translational development.
These key findings are consistent with observations of AZM-mediated changes in macrophage phenotype in models of lung infection, skin inflammation, and sepsis [
17,
19,
31]; however, the results reported here are the first to document that AZM can have a similar effect after traumatic CNS injury. While pro-inflammatory macrophage activation is reduced with AZM treatment in acute conjunctiva [
32], to the best of our knowledge, the current results are the first to demonstrate that AZM, or any other macrolide antibiotic, alters macrophage phenotype in response to spinal cord injury and reduces macrophage neurotoxicity. This is significant as neuroinflammation, and specifically pro-inflammatory macrophage activation, is a common feature of most neuropathological conditions including Alzheimer’s disease, stroke, aging, ALS, and traumatic brain injury [
33‐
36]. The ability of AZM to be widely distributed in brain tissue following oral administration [
37] makes it an intriguing candidate for manipulating macrophages in a variety of nervous system pathologies.
There is extensive data regarding the safety and dosing of AZM. Specifically, AZM is one of the antibiotics of choice for treating pneumonia in SCI individuals and is routinely administered at 10–45 mg/kg/day to treat infections in humans including community-acquired pneumonia, otitis media, and sinusitis [
12,
13,
38]. Accounting for allometric scaling, the dose of 160 mg/kg used in the current study is high but still clinically relevant, especially considering that higher AZM doses should be tolerated if necessary for neuroprotection due to the drug’s large therapeutic window and limited toxicity profile. Additionally, the recent results of the “COPD: influence of macrolides on exacerbation frequency in patients” (COLUMBUS) clinical trial report that AZM can be administered chronically (for 12 months), albeit at lower doses, with maintained immunomodulatory effects and no increased adverse effects [
39]. Ongoing work in our lab is examining the effect of lower doses of AZM on SCI inflammation and recovery.
One major limitation of the current work regarding the effectiveness of AZM for SCI treatment is that we utilize a combined pre- and post-SCI dosing strategy. We used this approach, based upon a previous dosing strategy we found to be effective for reducing inflammatory damage associated with acute lung infection [
17], to test the proof-of-concept that AZM can effectively alter inflammation in response to CNS perturbations. Indeed, our findings provide evidence that initiating treatment prior to CNS inflammation is effective. This approach may be beneficial for altering inflammation in chronic neurodegenerative disease, e.g., aging and Alzheimer’s. This is especially relevant in light of the effectiveness of chronic AZM administration reported in the COLUMBUS study [
39] and our previous observations that AZM treatment produces similar immunomodulary changes in both rodents and humans [
17,
40]. We have made preliminary observations that AZM is effective when administration begins after SCI (unpublished data, manuscript in preparation). In addition, in the current study, AZM had no effect on macrophage phenotype in the absence of an inflammatory stimulus in vitro (Fig.
4). This is consistent with our previous observations, and that of others, that AZM does not polarize unstimulated macrophages [
41]. Collectively, we therefore postulate that AZM treatment prior to SCI may not be required to facilitate improvements; however, additional studies are required to determine the therapeutic potential and required dosing for effective post-injury AZM treatment of SCI.
One observation in this study was that AZM treatment decreased monocyte-derived macrophages in the injured spinal cord. This is consistent with observations of reduced inflammation and macrophage accumulation with AZM treatment in models of acute conjunctiva, lung infection, and skin inflammation [
17,
19,
32]. In addition, decreasing macrophage accumulation at the site of SCI is neuroprotective and facilitates recovery [
42]. However, the mechanisms underlying the decreased macrophage recruitment we observed remain to be elucidated, as do the effects of AZM on monocyte- vs. microglia-derived macrophages.
Evidence that altering macrophage phenotypes or reducing M1 macrophage activation in SCI can be therapeutic comes from recent publications demonstrating that decreasing M1 macrophages in transgenic models of SCI leads to improved recovery, decreasing M2 macrophages or increasing M1 macrophages impairs SCI recovery, and increasing M2 macrophages using viral or transplantation approaches correlates with improvements in recovery [
5,
6,
9,
43‐
45]. Our data demonstrate that the underlying mechanism mediating improvements in SCI recovery with AZM treatment may be due to its ability to reduce the neurotoxic potential and pro-inflammatory activation state of SCI macrophages. We have previously observed an M1 to M2 macrophage shift with AZM treatment in vitro [
14]. The concept that AZM can shift macrophage phenotype is further supported by independent publications noting decreased IL-12, IL-6, IL-1β, TNF-α, and other pro-inflammatory mediators when macrophages are stimulated in the presence of AZM [
15,
21,
46‐
49]. Interestingly, these publications suggest that M1, but not M2, macrophage activation is affected by AZM. It is also possible, given the combined pre-and post-SCI dosing strategy, that AZM prevented M1 polarization in the current study rather than altering the M1 to M2 phenotype. This concept is support by the observation that AZM inhibits signaling cascades specific to pro-inflammatory stimuli [
50]. The specific mechanism responsible for the immunomodulatory potential of AZM and other macrolides are not well understood. Nonetheless, these collective observations indicate that AZM may selectively attenuate pro-inflammatory macrophage activation.
The neuroprotective effects we report with AZM are similar to the effects reported for SCI treatment with the antibiotic minocycline [
51]. Identifying neuroprotective, non-minocycline antibiotics has important implications for SCI therapeutic translation and treatment. There is an inherent risk of developing antimicrobial resistance with any antibiotic use, especially in SCI individuals who often undergo repeated antibiotic treatments to fight recurrent infections [
13,
52]. In addition, higher adverse reaction rates are associated with minocycline vs. other antibiotic treatments [
53]. Therefore, the identification of neuroprotective and antibiotic alternatives to minocycline increases the probability that these drugs can be used as neuroprotective strategies for treating SCI.
It is also important to identify the immunomodulatory mechanism of actions in order to develop more potent neuroprotectants. Due to structure differences between tetracycline (minocycline) and macrolide (azithromycin) antibiotics, it is difficult to imagine that both antibiotics are working through similar mechanisms of action. Attenuated pro-inflammatory microglial activation, however, has been reported with minocycline treatment in vitro [
54]. Interestingly, similar to AZM, minocycline selectively affects M1 but not M2 macrophages. Inhibition of the NF-κB pathway in pro-inflammatory macrophages has been observed for both antibiotics [
41,
54]. In addition, this is a somewhat common feature of other macrolide antibiotics [
50]. It is worth investigating the common structural elements of macrolides, and potentially other antibiotics, that mediate these immunomodulatory effects. Identifying the necessary structural components of macrolides that effect macrophage biology would provide insight into pro-inflammatory macrophage activation while facilitating development of more potent therapies that do not have the potential for causing antimicrobial resistance. There is evidence that modified macrolides retain their immunomodulatory properties [
55‐
58], and we are currently exploring the ability to use these or other novel macrolide compounds to facilitate CNS repair.
We recently reviewed the potential positive impact of M2 macrophages on SCI wound repair [
2]. Moving forward, it is important to determine mechanistically whether the benefits observed in SCI through AZM treatment are due primarily to the reduction of inflammatory factors and M1 macrophage activity or if there is a specific beneficial function of the M2 macrophage that AZM potentiates. A more regulatory battery of cytokine production, including IL-10 and TFGβ, is typical with M2 cells [
59,
60]. In addition, M2 macrophages can actively inhibit inflammatory processes through arginase-1 up-regulation, which competes with inducible nitric oxide synthase (iNOS) for
l-arginine [
61,
62]. These studies demonstrate the ability of M2 macrophages to suppress iNOS production, reduce inflammatory cytokine/chemokine secretion, and control neutrophil infiltration [
63]. However, in our own studies of cystic fibrosis patients, we observed that chronic AZM treatment significantly lowered inflammatory gene expression including iNOS and TNF-α, but did not significantly increase M2-associated gene expression [
40]. The future goal of our studies is to characterize the complicated role of anti-inflammatory macrophages over the entire post-injury time course.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
BZ, WMB, DJF, and JCG designed the research; BZ, WMB, TJK, and JCG performed the research; BZ, WMB, TJK, MBO, DJF, and JCG analyzed the data; and BZ, WMB, DJF, and JCG wrote the paper. All authors read and approved the final manuscript.