Cyclic AMP is a key regulator of M1 to M2a phenotypic conversion of microglia in the presence of Th2 cytokines

Cytokines were purchased from PeproTech (Rocky Hills, NJ). Cyclic AMP analogs were obtained from the BIOLOG Life Science Institute (Bremen, Germany) except dioctynyl cyclic AMP, which was obtained from Santa Cruz Biotechnology Inc. (Dallas, TX). Lipopolysaccharide (LPS) was purchased from Sigma-Aldrich (St. Louis, MO).

The immortalized BV2 microglial cell line employed for in vitro experiments was derived from the C57BL/6 mouse [13] and exhibits phenotypic and functional properties of primary microglia [14]. BV2 microglia were cultured at 37 °C, 5 % CO₂ in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Life Technologies Corporation, Grand Island, NY) supplemented with 10 % heat-inactivated fetal bovine serum (FBS; HyClone, Logan, UT) and 100 units/mL each of penicillin and streptomycin (Sigma-Aldrich). Prior to experimental use, BV2 microglia were seeded on either six-well culture plates (5 × 10⁴ cells/well for biochemistry experiments) or eight-well chamber slides (5 × 10³ cells/well for immunocytochemistry) and grown to 70–80 % for experimental use. BV2 microglia were treated with various cytokines or pharmacological agents based on the experimental schema described below.

Primary cultures of microglia were prepared from the cerebral cortex of neonatal (P1) wild type C57Bl/6J mice (JAX® Mice and Services, Bar Harbor, ME) using previously published methods [15]. Briefly, the cerebral cortices were dissected and cut into 2-mm small pieces. Tissue was suspended in cold DMEM (Gibco, Life Technologies Corporation) and triturated using a P-1000 plastic tip. The resulting cell suspension was passed through a 100-μm cell strainer and then centrifuged at ×1000 rpm for 10 min. The supernatant was removed and the pellet re-suspended in DMEM containing 10 % FBS and 10 % horse serum (Gibco, Life Technologies Corporation). The cell suspension was plated on polylysine (PLL)-coated T75 flasks and incubated at 37 °C for 7 to 10 days. Culture flasks were then placed on an orbital shaker at X230 rpm for 3 h, and the media, containing detached microglia, was harvested. Following centrifugation for 10 min at ×1000 rpm, a pellet containing the microglial cell fraction was acquired and re-suspended in the same media as described above. Highly pure microglial cell cultures were obtained using this method (>95 % as assessed using immunocytochemistry for CD11b (AbD Serotec; Raleigh, NC) and iba1 (Wako Pure Chemicals; Richmond, VA)). Primary microglia were diluted to the desired cell concentration and plated for at least 24 h before experimental use.

M1 activation was achieved with LPS (100 ng/ml) or TNF-α (10 ng/mL). For M1 to M2a phenotype conversion, IL-4 (10 ng/mL) was used in combination with cyclic AMP analogs (dibutyryl cyclic AMP; (1 mM), 6-Phe-cyclic AMP; (100 μM), CPT-2′O cyclic AMP; (100 μM)) beginning at 15 min prior to M1 activation. Cell supernatants were analyzed for cytokine levels, cell lysates prepared, or the cells fixed with 4 % paraformaldehyde 24 h later to measure the expression of M1 and M2a phenotype markers. All data are representative of at least three independent experiments.

Phagocytic function was evaluated based on the uptake of phycoerythrin (PE)-conjugated latex beads (Phagocytosis Assay Kit; Cayman Chemical, Ann Arbor, MI) as per the manufacturer’s protocol. Percent phagocytic cells was quantified in relation to the total number of Hoechst-positive nuclei in four random fields per treatment well and averaged across five independent experiments by fluorescence microscopy.

Immunoblotting was performed on culture supernatants, total cell lysates, or spinal cord (T7–9) tissue homogenates to quantify the expression of M1 and M2a markers according to previously published methods [15]. Nitrocellulose membranes (Bio-Rad Laboratories) were probed with specific primary antibodies; Arg-1 (1:2,000; GeneTex Inc., Irvine, CA), iNOS (1:1,000; Cell Signaling Technology, Boston, MA), TNF-α (1:500; Life Technologies Corp.), and β-actin (Sigma-Aldrich). The optical density of the bands (arbitrary units) was measured with an imaging densitometer (Bio-Rad Laboratories) and normalized to β-actin levels. The data represents values from three independent experiments.

Levels of the pro-inflammatory cytokines in culture supernatants was quantified using mouse TNF-α, IP-10, and IL-1β enzyme-linked immunosorbent assays (ELISA; Pepro Tech, Rocky Hill, NJ) according to the manufacturer's protocols. Values were obtained as picogram per milliliter of the culture supernatant and expressed as absolute values and compared between each of the treatment conditions. The data presented are the mean ± SEM of four independent experiments.

The presence of reactive oxygen species (ROS) was measured using the non-fluorescent dye 2′,7′-dichlorofluorescein diacetate (DCFH-DA; Molecular Probes, Eugene, OR) as per the manufacturer’s instructions. DCF fluorescence measurements were made in 96-well plates using a SpectraMax 5 microplate reader at 485 nm for excitation and 530 nm for emission. The data obtained were representative of three independently conducted experiments.

Microglia nitrite production was assessed by measuring total nitrite concentration, a stable oxidation product of NO by the Griess reaction. Briefly, culture supernatants (50 μl) were mixed with an equal volume of Griess reagent (Life Technologies Corp.) in 96-well plates for 10 min at room temperature in the dark, and the absorbance at 570 nm was determined using a SpectraMax 5 microplate reader. Sodium nitrite at concentrations of 0 to 100 μM was used as a standard. Each value indicates the mean ± SEM and is representative of results obtained from three independent experiments.

M1 and M2 marker expression in microglia was examined using immunocytochemistry (ICC) according to previously published methods [15]. Fixed cells were probed with primary antibodies for the M1 phenotype, iNOS (1:200; BD Transduction Laboratories; Franklin Lakes, NJ), and COX-2 (1:200; Thermo Scientific, Rockford, IL) or the M2 phenotype, Arg-1 (1:200; GeneTex, Irvine, CA), RELMα/Fizz1 (1:100; PeproTech, Rocky Hill, NJ), or transglutaminase-2 (1:100; Thermo Fischer Scientific). In addition, the microglial cell markers anti-iba1 (1:5,000; Wako Pure Chemical Industries, Ltd., Tokyo, Japan) and CD68/ED1 (Macrosialin; 1:200; AbD Serotec) were used. For visualization, the secondary antibodies, goat anti-mouse Alexa-594 or anti-rabbit Alexa-594 (1:200, Life Technologies Corp.) were used along with the nuclear marker Hoechst 33342 (1:1,000; Life Technologies Corp.). The morphology of the cells was demarcated by staining with Phalloidin-488 (1:100; Life Technologies Corp). Images were acquired by sequential scanning of the immunostained cells with an Olympus Fluorescence confocal microscope (Olympus, Fluoview FV 1000). For each treatment condition, three to four randomly selected fields were imaged per well. Images obtained were representative of three independently conducted experiments.

Confocal images were acquired by sequential scanning of the immunostained cells with an Olympus Fluorescence Microscope (Olympus, FluoView FV1000) at laser lines based on the specific Alexa-fluor secondaries used. In Adobe Photoshop CS6 (Adobe Systems, San Jose, CA) fluorescent images had the same universal adjustments applied: brightness (+50), contrast (+30), and smart sharpen (1.0 pixels).

The per cell immunostaining fluorescence intensity of specific markers was quantified using Image J software (http://​imagej.​nih.​gov/​ij/​). For each treatment condition, randomly selected cells [10, 11] per well/plate from independent experiments were identified by Hoechst-positive nuclei and imaged. These images were converted to grayscale, and the average per pixel intensity of the signal per cell (arbitrary units (a.u.) was recorded after background subtraction. These measurements were averaged across the cells analyzed per plate and then among group replicates for comparison of signal change across treatments.

Two models of spinal cord injury (SCI) were used in this study to examine the effects of IL4 and db-cyclic AMP on microglia and macrophage phenotypes in rodents. The first model employed adult Lewis rats (female, 180–200 g, n = 3), the second C57BL/6 (female, 25 g) mice (both from Charles River Laboratories International Inc., Wilmington, MA). Animals were housed according to NIH guidelines and The Guide for the Care and Use of Animals. All animal procedures were approved by the University of Miami Institutional Animal Care and Use Committee (IACUC). The rats were subjected to a moderate (25.0 mm) thoracic (T8) spinal cord contusion induced by the MASCIS weight drop device developed at New York University [16] as described previously [17]. For the mouse model (n = 3 per group) animals were subjected to a 50 kdyne moderate spinal cord contusion at thoracic (T8) level using the Infinite Horizons (IH) impactor as described previously [18]. Post-operative care was provided as described in Patel et al. [19].

With the Lewis rat model, IL-4 and db-cyclic AMP were delivered by intravenous (i.v.) administration through a jugular vein catheter prepared as described previously [20]. A mix of db-cyclic AMP (50 mg/kg) and IL-4 (30 μg/kg) in 100 μl of sterile, deionized water was administered as a single bolus injection with a 0.5-ml tuberculin syringe attached to the i.v. catheter over a period of 2 min within 15 min of SCI, followed by a 250-μl physiological saline flush. For the mouse model, the animals received daily, intraperitoneal injections (100 μl) of the agents at the same concentrations beginning at 30 min post-SCI for 7 days until end-point.

Animals were deeply anesthetized (150 mg/kg ketamine, 10 mg/kg xylazine) and transcardially perfused and the extracted CNS tissue post-fixed and cryoprotected as described elsewhere [21]. The T7–9 spinal cord was cryosectioned at a thickness of 20 μm (coronal) on a Leica CM3050S Cryostat (Leica Microsystems Inc., Buffalo Grove, IL). For biochemistry, animals were deeply anesthetized at 24 h post-SCI and decapitated. A spinal cord segment (5 mm) encompassing the injury epicenter dissected and snap frozen in liquid nitrogen for storage at −80 °C until processed for analysis.

Coronal, 20-μm cryosectioned sections of injured spinal cord tissue (400-μm intervals) were immunohistochemically (IHC) stained for specific markers using methods as described elsewhere [20]. For IHC, the Arg-1 antibody (1:200) was combined with anti-CD68 (ED1; 1:200; AbD Serotec), Alexa-594 tagged Isolectin-IB4 (1:100; Life Technologies Corp.) or anti-iba1 (1:5,000; Wako) along with appropriate AlexaFluor conjugated secondary antibodies (1:200; Life Technologies Corp.) and Hoechst 33342 (Sigma-Aldrich).

For quantifying the conversion of microglia and macrophages to M2a, cells displaying double immunoreactivity for Arginase-1 (Alexa-488 secondary antibody) and Alexa-594-tagged Isolectin-IB4 were quantified from projection images obtained using a confocal microscope (Olympus, Fluoview 1000) at ×40 magnification. These images were taken from within the injury epicenter, from the dorsal funiculus region of the spinal cord (n = 3 mice per group of IL-4 and db-cyclic AMP-treated and SCI-only controls). Images were converted to 8-bit gray scale separately for each marker using Image J software, and a threshold value was obtained. The relative percent of cells that were displaying the M2a phenotype were analyzed by measuring the fluorescent intensity of the number of cells showing Arginase-1 immunoreactivity and the total number of Isolectin-IB4⁺ cells between the treated and the untreated groups.

Significant differences between groups were ascertained by a one-way analysis of variance (ANOVA) or a t test, followed by a Bonferroni post-hoc analysis using Graph Pad Prism 4.0 (Graphpad Software, La Jolla, CA). All data was analyzed at the 95 % confidence interval. Graphed data was expressed as the mean ± standard error of the mean (SEM) according to the number of replicates or independent experiments. Asterisks or hashes included on the graphs indicate statistical differences between the treatment and control condition(s) with significance indicated at ***/^### p < 0.001, **/^## p < 0.01, or */^# p < 0.05.

BV2 microglia exposed to a pro-inflammatory stimulus, LPS, showed M1 phenotype induction as evidenced by pronounced expression of the characteristic markers iNOS and COX-2 and an absence of the M2a markers, Arginase-1, transglutaminase-2, and RELM-α (Fig. 1a–e). Concurrent treatment of LPS-stimulated BV2 microglia with IL4 and db-cyclic AMP but, neither independently (data not shown), antagonized M1 marker expression while dramatically augmenting an M2a phenotype as demonstrated by robust expression of M2a markers (Fig. 1f–j). M1 to M2a phenotype conversion in BV2 microglia was demonstrated immunocytochemically also with another pro-inflammatory mediator, TNF-α, in which, compared to untreated (Fig. 1k–l) or TNF-α-stimulated controls (Fig. 1m, n), combined IL-4 and db-cyclic AMP induced robust Arginase-1 production and the antagonism of iNOS expression (Fig. 1s, t). Only minor changes in Arginase-1 and iNOS were observed in TNF-α-stimulated BV2 microglia treated with IL-4 (Fig. 1o–p) or db-cyclic AMP (Fig. 1q–r) independently at the doses employed. Similarly, modest expression of transglutaminase-2, and RELM-α were detected with IL-4 treatment, but the presence of db-cyclic AMP dramatically potentiated their expression and the alternatively activated M2a phenotype (data not shown).

The necessity of cyclic AMP for robust M1 to M2 phenotype conversion was further demonstrated biochemically by immunoblotting for the prototypical M2 marker Arginase-1 (Fig. 1u, v). In unstimulated, resting microglia (M0 phenotype), exposure to IL-4 alone, but not db-cyclic AMP, elevated levels of Arginase-1, indicative of a transition to an M2 phenotype. However, as observed with LPS- or TNF-α-induced M1 microglia, exposure of the resting M0 microglia to the combination of IL-4 and db-cyclic AMP showed a much more pronounced expression of Arginase-1 (3.6-fold higher, p < 0.001). In M1 activated microglia exposed to either LPS or TNF-α, IL-4 and db-cAMP when given independently at the dose used did not induce robust Arginase-1 compared to concurrent application of IL-4 and db-cyclic AMP (Fig. 1u, v).

To examine the dose-dependent effects of IL-4 and db-cyclic AMP on the phenotype conversion of M1 microglia (iNOS⁺) to M2 phenotype (Arginase-1⁺), immunoblotting studies with titrated concentrations of these agents were performed. When the concentration of db-cyclic AMP was kept constant at 1 mM, IL-4 concentrations at and above 10 ng/mL were required to produce a significant induction of Arginase-1 expression and concomitant reduction in iNOS expression in LPS-stimulated BV2 microglia (Fig. 1w, x). Similarly, maintaining IL-4 at 10 ng/mL demonstrated that db-cyclic AMP concentrations at and beyond 0.1 mM could trigger a significant induction of Arginase-1 along with the parallel inhibition of iNOS expression during the phenotype conversion of M1 microglia to M2 (Fig. 1y, z).

The cytokine secretory profile of untreated or resting (M0), M1 (induced by LPS or TNF-α stimulation), or M2a (converted from M1 with IL-4 and db-cyclic AMP) BV2 microglia was investigated in culture supernatants at 24 h using specific ELISAs for TNF-α, IL-1β, or IP-10. ELISAs for each of these major pro-inflammatory cytokines were performed from three independent experiments. Compared to M0, the LPS-treated form showed a significant increase in the levels of TNF-α (TNF-α, 7.4-fold increase, p < 0.001; Fig. 2a). Similarly, TNF-α stimulation of the cells produced elevated concentrations of IP-10 in the culture supernatant at 24 h following treatment (IP-10, 4.9-fold increase, p < 0.001; Fig. 2b). IL-4 addition alone produced a modest, though significant reduction in the levels of TNF-α and IP-10 compared to stimulated controls; however, a major abrogation in the secretion of these cytokines was achieved with db-cyclic AMP or its combination with IL-4 (Fig. 2a, b). In contrast, IL-1β production following a 24-h TNF-α stimulus did not show any significant differences between the M0, M1, or M2a phenotypes, though M1 microglia exposed to db-cyclic AMP alone did show significantly enhanced levels of IL-1β (1.7-fold increase, p < 0.05) compared to stimulated controls (Fig. 2c).

The synergistic effects of cyclic AMP and IL-4 in regulating expression of Arginase-1 and iNOS was examined within single cells of primary cortical microglia challenged with TNF-α (Fig. 2d–q). While levels of Arginase-1 and iNOS were low in M0 microglia and not significantly different from one another (Fig. 2d–f), upon TNF-α stimulation, a profound increase in iNOS over Arginase-1 was observed in single microglial cells (Fig. 2g–i). The exposure of M1 microglia to either IL-4 (Fig. 2j–l) or db-cyclic AMP (Fig. 2m–o) was not able to reverse the balance of Arginase-1 to iNOS expression that significantly favored the later in single cells. In contrast, the concurrent exposure of M1 microglia to IL-4 and db-cyclic AMP switched the balance of Arginase-1 to iNOS expression to significantly favor Arginase-1 (Fig. 2p–r), indicative of an M1 to M2 phenotype conversion.

The main effector pathways downstream of cyclic AMP are coordinated through PKA and the exchange protein directly activated by cyclic AMP (EPAC), which have been shown to regulate often distinct cellular processes in neural cells [22]. To probe the involvement of PKA and EPAC in M2a phenotype conversion, selective cyclic AMP analogs that activate PKA or EPAC were employed in combination with IL-4 in cultures of primary cortical microglial cells. Similar to db-cyclic AMP, the PKA-specific cyclic AMP analogs, dioctanoyl-cyclic AMP, and 6-phenyl-cyclic AMP, but not the EPAC-selective analogs, CPT-2′O methyl-cyclic AMP or 8-bromo-2′O methyl-cyclic AMP, induced a significant increase in Arginase-1 expression when combined with IL-4 treatment in M1 microglia (Fig. 2s–t). These results were then confirmed using immunocytochemistry for Arginase-1 in primary cultures of microglia (Fig. 2u–x). LPS-stimulated primary microglia exhibited little immunoreactivity for Arginase-1 and robust expression of macrosialin (ED-1), a characteristic marker of activated microglia and macrophages (Fig. 2u ₁–u₂). Strong Arginase-1 immunoreactivity was observed in M1 microglia exposed to IL-4 when used in combination with the PKA activators db-cyclic AMP (Fig. 2v ₁–v₂) and phenyl-cyclic AMP (Fig. 2w ₁–w₂), but not with the EPAC-selective cyclic AMP analog CPT-2′O methyl-cyclic AMP (Fig. 2x ₁–x₂).

The phagocytic ability of M2a converted microglia following treatment with LPS and IL4, with or without db-cyclic AMP, was comparatively examined to LPS-stimulated alone (M1) and M0 (no treatment) microglia through quantification of their capacity to phagocytose PE-conjugated latex beads. Compared to untreated cells (Fig. 3a), LPS stimulation of BV2 microglia to an M1 phenotype enhanced their ability to phagocytose PE-conjugated beads (3.3-fold higher than untreated, p < 0.001; Fig. 3b, f). In contrast, IL-4 produced a significant impairment in the phagocytic capacity of M1-activated microglia (2.7-fold lower than LPS, p < 0.001; ns compared to untreated, Fig. 3c, f). Conversely, the addition of db-cyclic AMP alone significantly increased the phagocytic activity of the M1 phenotype (Fig. 3d) and additionally restored the deficits observed in the phagocytic properties conferred in IL-4 driven M2a microglia (Fig. 3e, f), conferring a phagocytic activity significantly greater than that of LPS stimulation alone (1.4-fold and 1.3-fold higher than LPS, p < 0.01 and p < 0.05, respectively; Fig. 3f). These results demonstrate that an important role of classically activated M1 microglia, phagocytosis and debris clearance, is retained in the converted M2a phenotype.

We examined whether M2a converted microglia, obtained through the combined exposure of M1 microglia to IL-4 and db-cyclic AMP, produced reactive oxygen (ROS) and/or nitrogen species (RNS), cytotoxic molecules that are associated with the M1 microglial cell phenotype. The ability of M1 microglia to produce RNS, measured by nitrite concentration, was detected based on the Griess reaction. Compared to untreated controls, LPS stimulation of BV2 microglia to the M1 form produced a significant increase in cell nitrite concentration (1.5-fold increase, p < 0.001; Fig. 3g). Concurrent use of IL-4 with LPS significantly attenuated this increase (Fig. 3g). The employment of either db-cyclic AMP alone or M2a phenotypic conversion with the combination of IL-4 and db-cyclic AMP, further reduced nitrite concentrations to levels significantly lower than that of untreated controls (2.3-fold and 5.9-fold reduction compared to untreated, p < 0.001 for both, respectively; Fig. 3g). Levels of ROS detected using DCFH-DA showed that LPS stimulation of BV2 microglia produced a significant increase in ROS production compared to untreated controls (1.4-fold increase, p < 0.05; Fig. 3h). The concomitant use of IL-4 or db-cyclic AMP alone did not significantly reduce the increase in ROS produced by LPS stimulation. Only when db-cyclic AMP and IL-4 were employed in combination to promote M2a phenotype conversion was the LPS-induced generation of ROS in microglia completely perturbed (1.5-fold decrease, p < 0.01 compared to LPS stimulation; ns compared to untreated; Fig. 3h).

To investigate whether M1 to M2a phenotypic conversion of microglia could be accomplished in vivo using cyclic AMP elevation and IL-4 supplementation, their concurrent delivery was investigated after contusive spinal cord injury (SCI) in two different experimental paradigms. First, in a rodent model, adult Lewis rats were subjected to injury and the co-administration of IL-4 and db-cyclic AMP systemically during the acute phase (within 15 min) of SCI. Using the prototypical M2a marker Arginase-1, at 24 h after SCI an increase in Arginase-1 expression was identified in the total spinal cord tissue homogenates of the lesion site compared to naïve controls by immunoblotting (5.9-fold increase compared to naïve, p < 0.05 Fig. 4a, b); indeed, an increase in both M1 and M2 microglia and macrophage populations has been reported previously in the acute phase of SCI [10]. The use of M1 to M2a phenotype conversion with db-cyclic AMP and IL-4 resulted in a significant enhancement of Arginase-1 production at 24 h after SCI (11.1-fold increase compared to naïve, p < 0.01 and p < 0.05 versus SCI; Fig. 4a, b). Histological examination of tissue sections from the injury epicenter showed that, compared to SCI controls (Fig. 4c, d), the combination of db-cyclic AMP and IL-4 after SCI produced a marked reduction in numbers of Iba1⁺ (Fig. 4e) and ED1⁺ (Fig. 4f) microglia and macrophages within the lesion site. Co-staining of ED1⁺ microglia or macrophages with Arginase-1 revealed that at 24 h after SCI alone, only a small fraction of ED1⁺ cells were immunoreactive for Arginase-1 (Fig. 4d–f; white arrows). In contrast, when M2a phenotype conversion with db-cyclic AMP and IL-4 was used acutely after SCI, almost all the ED1⁺ microglia and macrophages were immunoreactive for Arginase-1 (Fig. 4h–j; white arrows), indicating that they were of the M2a phenotype. To assess the effects of M2 microglia and macrophage conversion on the pathological processes of injury, we assessed at 24 h the generation of total free ROS/RNS radicals within the lesion site by employing the fluorogenic probe dichlorodihydrofluorescin DiOxyQ (DCFH-DiOxyQ), which is a specific ROS/RNS probe [23]. Compared to SCI controls, the combination of db-cyclic AMP and IL-4 to induce M1 to M2a phenotypic conversion produced a significant reduction in the amount of free radicals generated within the injured spinal cord (57.6 % reduction, p < 0.05; Fig. 4k).

In subsequent work, we employed a murine model of moderate SCI using c57/Bl6 mice. These mice received either SCI alone or SCI with the co-administration of IL-4 and db-cyclic AMP acutely post-SCI and then daily for 7 days. Immunohistochemical analysis of spinal cord tissue at the lesion site at 7 days post-SCI showed that compared to SCI controls (Fig. 4l–n), a significant increase in the number of Isolectin-IB4⁺ microglia and macrophages immunoreactive for Arginase-1 occurred with the concurrent delivery of IL-4 and db-cyclic AMP (Fig. 4o–q). Arginase-1 immunoreactivity at the center of the lesion appeared largely restricted to Isolectin-IB4⁺ cells. Quantitative assessment of the number of Isolectin-IB4⁺ microglia and macrophages that co-stained with Arginase-1 after SCI showed a significant increase with IL-4 and db-cyclic AMP over SCI controls (3.0-fold increase; p < 0.001; Fig. 4r).