Responses of rat and mouse primary microglia to pro- and anti-inflammatory stimuli: molecular profiles, K⁺ channels and migration

All procedures on animals were approved by the University Health Network Animal Care Committee (Animal Use Protocols 914 and 1573) and adhered to the Canadian Council on Animal Care guidelines for humane animal use. Pure neonatal microglia cultures were prepared from Sprague-Dawley rat pups (P1–P2) and C57BL/6 mouse pups (P0–P2). We selected this outbred rat strain and inbred mouse strain because they are both widely used in biomedical research, and specifically because C57BL/6 mice are the primary strain used in transgenic studies. Animals were purchased from Charles River (St-Constant, PQ, Canada). As we recently described for rat microglia [18‐20, 30, 31], brain tissue (excluding the cerebellum and meninges) was mashed, strained, and centrifuged (300×g, 10 min) in cold Minimal Essential Medium (MEM; Invitrogen, Carlsbad, CA). The pellet was re-suspended in MEM and seeded in 75-cm² flasks containing 20 mL of MEM supplemented with 10% fetal bovine serum (FBS; Wisent St-Bruno, PQ) and 0.05 mg/mL gentamycin (Invitrogen). Cells were incubated at 37 °C with 5% CO₂, and after 48 h, the medium was changed to remove cellular debris and non-adherent cells. After 5–6 days (rat) or 10–14 days (mouse), microglia were harvested by shaking the flasks (5 h, 65 rpm) on an orbital shaker in the incubator (37 °C, 5% CO₂). The supernatant containing microglia was collected, centrifuged, and re-suspended in fresh MEM (2% FBS, 0.05 mg/mL gentamycin). Microglia were seeded onto UV-irradiated 15-mm glass coverslips (Fisher Scientific, Ottawa, ON) at different densities based on the experiment, as detailed below. For mouse microglia, it is difficult to obtain the large numbers of cells needed to perform the multiple treatments, functional assays, and Western blot analyses. Rather than using a cell line (e.g., BV2, which does not necessarily respond the same as primary microglia [32, 33], we grew the cells longer to expand the population. Importantly, we confirmed that expression of numerous genes (see “Results” below) shows that their initial “resting” state was similar to rat microglia, and that many activation responses were similar between the species. Thus, where specific differences were seen between the species, they are unlikely to reflect culturing times.

After seeding, microglia were allowed to settle for 2–3 days (37 °C, 5% CO₂), and then were left unstimulated (control; CTL) or stimulated with 20 ng/mL IFN-γ plus 50 ng/mL TNF-α to induce a pro-inflammatory state [M(I + T)], or with 20 ng/mL IL-4 [M(IL-4)] or 20 ng/mL IL-10 [M(IL-10)] to induce anti-inflammatory states. The recombinant cytokines (R & D Systems Inc., Minneapolis, MN) were specific to the rodent species; e.g., mIL-4 was used on mouse microglia. For messenger RNA (mRNA), protein, and functional analyses (nitric oxide production, migration), the cells were stimulated for 24 h, which was chosen to facilitate comparisons with our previous studies of rat microglia [17‐20, 30, 34, 35], which show that many genes respond at 24 h. For electrophysiological analysis, cells were used 30 h after stimulation to provide additional time for channel trafficking and post-translational modifications.

Microglia were seeded at 5 × 10⁵ cells/coverslip in a 12-well culture plate, allowed to settle for 1–2 days (37 °C, 5% CO₂), and then stimulated with cytokines for 24 h. Total RNA was extracted using TRIzol reagent (Invitrogen) and purified using an RNeasy Mini Kit (QIAGEN, Mississauga, ON, Canada), as previously described [17‐19, 30]. Samples were stored at −80 °C and used for NanoString and real-time qRT-PCR assays.

For NanoString analysis, extracted RNA (200 ng per sample) was sent to the Princess Margaret Genomics Centre (https://​www.​pmgenomics.​ca/​pmgenomics/​; Toronto, Canada), where sample purity was assessed (using Nanodrop 1000), and the assay was conducted (hybridization, detection, scanning) using samples from each rodent species. We analyzed the data using nSolver Analysis Software (ver3.0). The methods were similar to our recent studies of primary rat microglia [17, 20, 30]; nevertheless, it is useful to more fully describe the controls and normalization procedures. NanoString is a medium-throughput method that can analyze many genes in a single sample with comparable sensitivity and accuracy to quantitative real-time RT-PCR (qRT-PCR) [36]. Moreover, by eliminating the need for amplification, it reports mRNA counts in a given sample, is more sensitive and accurate than microarrays [36], and in fact, is sometimes used to validate microarray data [37, 38]. Many NanoString studies have not directly compared the results with qRT-PCR (e.g., [39‐44]) but a recent study directly compared qRT-PCR with NanoString and the ABI OpenArray System (another medium-throughput platform) [45]. While overall trends in mRNA expression were similar using all three platforms, NanoString and OpenArray results were better correlated.

Separate plates had to be used for each species, and different probe sets were designed and synthesized by NanoString nCounter technologies (rat, Table 1; mouse, Table 2. Note that gene names sometimes differ slightly between species.) Each transcript of interest was recognized by a capture probe and a reporter probe, each containing 30–50 bases complementary to the target mRNA. To minimize assay variability, the code sets also included negative and positive control reporter probes that were developed by the External RNA Control Consortium (ERCC). The eight negative-control reporter probes representing foreign sequences (not homologous to any organism) are not expected to detect the foreign transcripts in the samples. These background levels were calculated for each sample (geometric mean counts) and subtracted from the raw counts for each gene. Six positive control reporter probes (ERCC-selected mRNA targets) were pre-mixed with (Spike-Ins) the code set at a concentration range (0.125–128 fM), a range corresponding to the expression levels of most mRNA of interest, to control for overall efficiency of probe hybridization and determine the detection range for transcripts of interest in each assay. A scaling factor was determined, as follows. For each positive control probe, the geometric mean was calculated from counts obtained from each microglia sample, and the geometric means of all six positive controls were then averaged. This mean was divided by the geometric mean of the positive controls within a given microglia sample to obtain a sample-specific positive control scaling factor. A scaling factor outside the range of 0.3 to 3 indicates suboptimal hybridization. In our samples, the scaling factor always fell within the optimal range and was thus applied to all counts in the sample. Next, a reference gene scaling factor was calculated in the same manner using two housekeeping genes (Hprt1, Gusb), and used to adjust the counts for each sample (unstimulated or stimulated microglia). Sometimes, expression of a gene was very low (<20 mRNA counts/200 ng sample), which approaches the detection limit and should be treated with caution.

We analyzed over 50 genes in rat and mouse microglia under different activation states. To assess microglial activation, markers, pro- and anti-inflammatory mediators, receptors and signaling molecules were analyzed. We also assessed several immunomodulators, nicotinamide adenine dinucleotide phosphate-oxidase (NOX) enzymes, purinergic and phagocytic receptors, and potassium (K⁺) channels that play roles in microglial functions. For inter-species comparisons, we converted relative mRNA counts to fold changes relative to unstimulated (control) levels and then compared fold changes in response to cytokine stimulation.

Microglia were seeded at 8 × 10⁴ per glass coverslip and were either left unstimulated or treated with I + T, IL-4 or IL-10 for 24 h. The colorimetric Griess assay (Invitrogen) was used to measure nitrite levels as an indirect measure of nitric oxide production. For the Griess reaction, 200 μl of conditioned medium from microglia samples was added to wells of a 96-well plate containing 25 μl sulfanilic acid. Then, 25 μl 0.1% N-(1-naphthyl) ethylenediamine was added, and the medium was kept in the dark at room temperature for 30 min to allow the reaction to occur. The color change in the samples was quantified using a multi-label plate reader (Victor³ 1420, Perkin Elmer, Woodbridge, ON, Canada) set at an absorbance wavelength of 570 nm. Nitrite concentrations in the samples were determined by interpolation on a standard curve generated from a series of NaNO₂ samples of known concentration. Results are expressed as fold change relative to untreated (CTL) samples.

Microglia were seeded at 7–9 × 10⁴ cells/coverslip, and stimulated with cytokines for 24 h. They were briefly washed in phosphate-buffered saline (PBS), and fixed in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) at room temperature for 15 min. After permeabilizing the cells with 0.2% Triton X-100 for 5 min, they were washed in PBS (3×, 5 min), and labeled with Alexa Fluor 488-conjugated phalloidin (1:50 in PBS for 1 h; Invitrogen) to visualize filamentous (F-) actin, and with 4′,6-diamidino-2-phenylindole (DAPI; 1:3000 in PBS for 10 min; Invitrogen) to label nuclei. After washing (3×, 5 min), coverslips were mounted on glass slides using Dako mounting medium (Dako, Glostrup, Denmark) and stored in the dark at 4 °C. Images were acquired using an Axioplan 2 wide-field epifluorescence microscope equipped with an Axiocam HR digital camera (both from Zeiss, Toronto, ON, Canada).

To quantify migration, microglia were seeded at 3 × 10⁴ cells per insert filter (which bore 8-μm-diameter holes), and placed in the upper well of a Transwell migration chamber (VWR, Mississauga, ON, Canada) containing 500 μL MEM with 2% FBS, as recently described [18, 20, 30]. After 30 min, 500 μL of MEM with 2% FBS was added to the lower well, and microglia were left unstimulated or stimulated for 24 h (37 °C, 5% CO₂) with I + T, IL-4, or IL-10, as above. When used, a channel blocker (ML133 or AgTx-2) was added at the time of cytokine addition. Transwell inserts were then briefly washed with PBS, fixed for 10 min in 4% paraformaldehyde, and washed again in PBS (3×, 5 min). A Q-tip was used to remove microglia from the top of the Transwell inserts. Cells that had migrated to the underside of the membrane were stained with 0.3% crystal violet in methanol (~1 min) and washed with PBS. Cells from five random fields at 40× magnification were counted using an Olympus CK2 inverted microscope (Olympus, Tokyo), summed and normalized to the unstimulated (CTL) group.

We used the CyQuant NF assay (Invitrogen) to measure cell proliferation, as previously described [18, 30]. Microglia were seeded at 2–3 × 10⁴ cells per well of a 96-well flat-bottom plate and cultured in MEM with 2% FBS for 1–2 days. Then, cells were unstimulated or stimulated with I + T, IL-4, or IL-10 in the presence or absence of a channel blocker (ML133 or AgTx-2). After 24 h, the CyQuant dye solution was added to each well and incubated for 30 min (37 °C, 5% CO₂). The fluorescence intensity was measured using a multi-label plate reader (Victor³ 1420, Perkin Elmer, Woodbridge, ON, Canada), with excitation at 485 nm and emission at 535 nm. Readings were taken for 0.1 s at 3 mm from the bottom of the plate in duplicate and were averaged, and background was subtracted before normalizing to the unstimulated (CTL) group.

All graphical data are presented as mean ± SEM for the number of replicates indicated. The statistical significance was analyzed in GraphPad ver 6.01 (GraphPad Software, San Diego, CA) using either a one-way analysis of variance (ANOVA) with Dunnett’s post hoc analysis or two-way ANOVA with Bonferroni post hoc analysis (electrophysiology, Western blotting, Griess and migration assays). For NanoString analysis, the mRNA counts acquired after normalization were expressed as fold changes relative to control cells to compare the effects of activation responses between rat and mouse microglia. A two-way ANOVA with Fisher’s LSD test was then conducted, and the p values for differences in activation state or species were adjusted using a 5% false discovery rate correction for multiple comparisons [54] in R (version 3.3.1).

The terminology for microglial activation is evolving; thus, for clarity, activation states are denoted by the stimulus used, as follows. M(I + T). Microglia were treated with a combination of IFN-γ and TNF-α to evoke a pro-inflammatory state (also called classical or M1 activation), as before [17]. M(IL-4). IL-4 was applied to skew them toward an anti-inflammatory state (also called alternative or M2a activation). M(IL-10). IL-10 was used to skew them toward an acquired deactivation state (sometimes called M2c).

Those results showed increases in pro-inflammatory markers after LPS treatment (e.g., NOS2, TNFα, IL-1β) versus increases in anti-inflammatory markers after IL-4; e.g., arginase 1 (ARG1), CD163, mannose receptor (MRC1/CD206), IL-4 receptor α (IL-4RA), IL-10, and TGF-β1 [19, 30, 35]. In addition, our previous NanoString analysis showed that hallmark M1 and M2a responses could be detected in vitro and in vivo [17, 55]. Here, to create a comparison profile of responses of rat and mouse microglia, transcript expression was quantified by NanoString for 58 genes encompassing pro- and anti-inflammatory mediators and their receptors, other immunomodulators (nicotinamide adenine dinucleotide phosphate-oxidase (NOX) enzymes), and purinergic and phagocytic receptors. The gene nomenclature is indicated for rat (Table 1) and mouse (Table 2); however, for simplicity, the rat names will be used.

To illustrate differences in basal transcript abundance, the left-hand columns of Tables 3, 4, 5, and 6 show results for each gene, stated as the number of mRNA counts per 200 ng sample under control (unstimulated) conditions. Then, for M(I + T), M(IL-4), and M(IL-10) stimulation paradigms, results are shown as fold changes with respect to the control values. Significant differences within a species are indicated by up arrows for increases and down arrows for decreases. In addition, these tables indicate species differences within an activation paradigm (bold numbers and asterisks). [Additional files 1, 2, 3, 4, and 5: Fig. S1–S5, graphically show the complete mRNA data for both species.]

We next examined several molecules routinely used to identify “activated” microglia after acute brain injury, and several immunomodulatory molecules. Unstimulated. Microglia of both species expressed low transcript levels of Socs1 (suppressor of cytokine signaling 1) and Socs3 (Table 5; Additional file 4: Fig. S4). Both species showed moderate to very high levels of the other molecules examined: Itgam (CD11b), Cd68, Cx3cr1, toll-like receptor 2 (Tlr2), Tlr4, Nfkbia (IκBα; endogenous inhibitor of NFκB), Tspo (translocator protein), Nr3c1 (glucocorticoid receptor, GR), and Ptpn6 (Src homology region 2 domain-containing phosphatase-1; SHP-1). Interestingly, most control transcript levels were higher in rat than in mouse microglia, except for Cx3cr1, which was higher in mouse. Surprisingly, mRNA for Aif1 (which codes for ionized Ca²⁺ binding adapter molecule 1, Iba1) was highly expressed in rat cells (>20,000 mRNA counts) but very low in mouse. This species difference was confirmed at the protein level, where Iba1 was much higher in unstimulated rat microglia (Fig. 6). M(I + T). Both species showed increased transcript expression of Nr3c1, Nfkbia, Socs1, and Tspo, and decreased Cd68 and Cx3cr1. The main species differences included the higher increase in mouse cells for Socs1 (5.7-fold) and Tspo (4-fold), and the mouse-specific decrease in Itgam and increases in Aif1, Ptpn6, Socs3 and Tlr4 (which decreased in rat). Although both species showed modest increases in Iba1 protein (1.6-fold in mouse, 1.4-fold in rat), they did not reach statistical significance (Fig. 6c). M(IL-4). Both species showed increased expression of Tlr4 and decreased Tlr2. [A pilot study of rat microglia also showed an increase in Tlr4 at 6 h (not shown)]. Socs1 increased in rat, and despite a 33-fold increase in mouse, it did not reach significance (p = 0.08). Notable species differences were increased Cx3cr1 in rat but a dramatic decrease in mouse, decreased Cd68 in rat but an increase in mouse, lower Itgam in rat, and lower Ptpn6 in mouse. M(IL-10). Both species showed increased expression of Itgam and Tlr4, and decreased Cx3cr1. Species differences were increases in Cd68 and Tspo (rat only), and increase in Nc3r1 (mouse only). Mouse had increased Socs3, but the 9.4-fold increase in rat was not significant. Interestingly, Itgam was the only gene selectively upregulated by IL-10 treatment in both species.

Overall, the pro-inflammatory state in both species was marked by increased mRNA expression of Nfkbia. Several genes increased with I + T but were not selective pro-inflammatory markers. I + T increased Nc3r1 and Socs3 but they were also slightly increased with IL-10 in mouse; Socs1 and Tspo were elevated in both species but also increased by IL-4 or IL-10 in rat. Several genes were not selective activation state markers in either species at the mRNA level: Aif1, Itgam, Cd68, Cx3cr1, Tlr2, Tlr4, and Ptpn6. Of note, I + T treatment decreased Cx3cr1, which is often used to identify microglia, and Cd68, which is often used to identify their phagocytic state.

We conducted a species comparison of transcript levels of several molecules to follow up on our recent study of myelin phagocytosis and consequent production of reactive oxygen species (ROS) in rat microglia [17]. We also examined several purinergic receptors that can modulate microglial phagocytosis, ROS production, cytokine secretion, and migration [58]. Unstimulated. Microglia from both species had modest transcript levels (<500 mRNA counts/200 ng RNA) of Nox1, Nox4, P2rx7, P2ry2, and P2ry12; and higher levels (>2500 mRNA counts) of Cybb (NOX2), Fcgr2b (CD32), Fcgr3a (CD16), and Msr1 (SR-A) (Table 6; Additional file 5: Fig. S5). The main species differences were the higher control levels in rat of Fcgr1a (CD64; 4.9-fold), Ncf1 (9.3-fold), Trem2 (31.2-fold), and P2ry12 (3.4-fold). M(I + T). Both species had decreased transcript levels of Trem2, and increased Ncf1 (although 1.4-fold greater in rat), Cybb (1.8-fold greater in mouse), and Fcgr3a (1.5-fold greater in rat). Other species differences were the opposite changes in Fcgr1a expression, the increase in P2ry2 in rat only, and decreases in Msr1 and P2ry12 in rat only. M(IL-4). Both species showed decreased expression of Cybb. All other changes were species dependent, with most genes altered in rat cells but few in mouse. In rat, Fcgr2b, P2rx7, and P2ry12 increased, while Trem2 and Msr1 decreased. Mouse-specific changes were decreased P2ry12 and increased Nox1, which were opposite to the trends in rat. M(IL-10). In both species, Fcgr2b and Fcgr3a increased. Species differences were increases in P2ry2 and Trem2 in rat only, increased Msr1 and P2ry12 in mouse only, and opposite changes in Cybb (up in rat, down in mouse). Nox4 transcripts remained low in both species and were unaffected by any treatment tested.

Primary rat and mouse microglia express Kir2.1 mRNA and protein [18, 24] but other Kir2-family members have not been assessed. Because these channels can function as homotetramers or heterotetramers [59], we first compared expression of Kcnj2 (Kir2.1), Kcnj12 (Kir2.2), Kcnj4 (Kir2.3), and Kcnj14 (Kir2.4). [We omitted Kir2.5 because it is electrically silent and Kir2.6 because it is expressed primarily in skeletal muscle [60, 61]]. In both rodent species, transcript expression of Kir2.1 (Kcnj2) predominated (Fig. 7a). In a pilot study, NanoString showed that expression of Kcnj12, Kcnj4, and Kcnj14 in rat microglia was not changed in M(I + T), M(IL-4), or M(IL-10) states, and real-time RT-PCR corroborated these results (data not shown). Thus, homomeric Kir2.1 channels likely produce the inward-rectifying K⁺ current in both species, which is important because the blocker we used (ML133) can affect other Kir2 members [62].

Kv1.2 (Kcna2), Kv1.3 (Kcna3), and Kv1.5 (Kcna5) mRNA and protein have been detected in primary rat microglia [28, 29, 63, 64]. In primary mouse microglia, Kv1.3 and Kv1.5 have been detected [65], but we found no reports of Kv1.2 expression. Here, we compared transcript expression of Kcna2, Kcna3, and Kcna5, and Kcnj2 (Kir2.1) in different activation states in both species (Fig. 7b). Unstimulated . The main species difference was the 5.2-fold higher Kir2.1 mRNA expression in rat microglia (1984 ± 656 (SD) mRNA counts in rat vs 382 ± 82 in mouse). Both species expressed relatively low transcript levels of Kv1.2 (77 ± 50 mRNA counts in rat vs 9 ± 5 in mouse), Kv1.3 (113 ± 19 mRNA counts in rat vs 63 ± 13 in mouse), and Kv1.5 (4 ± 2 mRNA counts in rat vs 3 ± 1 in mouse). M(I + T). Two channels showed opposite changes. Kir2.1 expression increased 4.8-fold in rat but decreased 2.8-fold in mouse; Kv1.2 decreased 10.8-fold in rat but increased 11.7-fold in mouse. Kv1.3 expression increased in both species but to a greater degree in mouse. Kv1.5 was unchanged in both species. M(IL-4). In mouse, the only effect was a 3.8-fold increase in Kv1.5 but the mRNA level remained very low (~10 counts/200 ng RNA). In rat cells, Kv1.3 increased 2.2-fold. [In a pilot study on rat cells, Kv1.3 mRNA also increased at 6 h (data not shown).] M(IL-10). There were no significant changes. Overall, the most notable species differences were the opposite changes evoked by I + T in Kir2.1 and Kv1.2, and the IL-4-evoked increase in Kv1.3 in rat only and Kv1.5 in mouse only.

The next step was to use electrophysiology to compare the currents under each activation state. This is a more accurate readout than simply measuring protein levels because ion channel function can be affected by post-translational modulation and trafficking to the surface membrane; for instance, as we have shown for Kv1.3 in microglia [28, 50]. Here, we measured total inward and outward currents, and then used specific voltage protocols and channel blockers to isolate and quantify Kir2.1 and Kv1.3 currents. Because microglial morphology changes with M1 activation [19, 20, 56] and potentially affects cell size, we first determined that the cell capacitance, which is proportional to cell size, did not differ under any activation condition or between species (Table 7). Subsequently, we recorded currents from microglia with the most prevalent morphologies; i.e., unipolar for unstimulated, M(IL-4) and M(IL-10) cells; rounded or amoeboid for M(I + T) cells.

Microglia Kir currents displayed the stereotypical features of Kir2.1 (Fig. 8a). This current activates at negative potentials due to relief of channel block by internal Mg²⁺ and polyamines [59, 66], and then the current relaxes at very negative potentials due to time-dependent block by external Na⁺ [52, 67].

To quantify the Kir2.1 component of the whole-cell current, we used ML133, a membrane-permeant blocker [62] that acts in a time-dependent manner with an IC₅₀ ~ 3.5 μM in rat microglia [18]. Regardless of the microglial activation state, most of the Kir current (82–86% in rat, 85–95% in mouse) was blocked by 20 μM ML133 (Fig. 8b), and the small remaining current had a linear current-versus-voltage (I-V) relation (not shown). Thus, to examine whether microglial activation states affect channel activity, we compared the total Kir current density. In both species, the I-V relations showed inward rectification and reversal at about − 82 mV after junction potential correction (Fig. 8c), which is close to the K⁺ Nernst potential. Unstimulated. Both rodent species displayed a similar magnitude of Kir2.1 current (Fig. 8a, c) despite the differences in mRNA counts (Fig. 7b). Species differences were seen under all activation states examined (Fig. 8c). M(I + T). The Kir2.1 current decreased in mouse cells but was unchanged in rat. M(IL-4). The Kir2.1 current decreased substantially in rat cells but was unchanged in mouse. M(IL-10). The Kir2.1 current decreased in mouse cells but was unchanged in rat. Overall, the Kir2.1 current amplitude is not a reliable indicator of the microglial activation state, but instead depends on both the rodent species and activating stimulus.

Very little is known about migration of mouse microglia in different activation states or roles of Kir2.1 and Kv1.3 channels in migration in either rodent species (see “Discussion”). Therefore, we next compared these aspects under all four activation states, starting with morphology. We had previously shown that unipolar rat microglia migrate in the direction of the lamellum, which usually contains a prominent ring of F-actin-rich podosomes that we called a “podonut” [19, 31, 68]. Podosomes are subcellular structures involved in migration through roles in cell adhesion and matrix degradation [69]. Unstimulated. Both species showed a similar initial morphology. Many were unipolar, with a uropod and a large lamellum that often contained a podonut (Fig. 11a). There were both similarities and differences in their morphological responses to activating stimuli. Broadly speaking, unipolar microglia were common in highly migratory, unstimulated and M2 states, while round or amoeboid cells were common for M1 microglia of both species. M(I + T). The shape of the cells changed dramatically, but with subtle species differences. Rat microglia were clustered into chains of cells, while mouse microglia were more evenly distributed. For both species, but especially in mouse microglia, individual cells often appeared flat and round or amoeboid. There was no apparent cell polarity but the cells often bore short, spiky processes. M(IL-4). Both species showed many unipolar cells, and many of them contained a podonut. The morphology of rat microglia was more heterogeneous and lamellae were often smaller than in unstimulated rat cells or M(IL-4) mouse cells. M(IL-10). For rat microglia, we previously reported that IL-10 increases the proportion of cells containing podonuts [20]; however, here, the mouse microglia continued to resemble unstimulated cells.

Next, we quantified migration in each activation state. Rat. M(I + T) cells migrated 2.9-fold less than unstimulated cells, while M(IL-4) and M(IL-10) migrated more, by 1.8-fold and 2.4-fold, respectively (Fig. 11b). Blocking Kir2.1 channels (20 μM ML133) reduced migration by 2.6-fold in M(IL-4) cells and 3.7-fold in M(IL-10) cells (Fig. 11c). Although ML133 reduced migration by 2.8-fold in unstimulated cells and 4-fold in M(I + T) cells, ANOVA did not show statistical significance, likely because migration was not as high as with IL-4 or IL-10 treatment. Blocking Kv1.3 (5 nM AgTx-2) increased migration of unstimulated microglia (2.9 fold), M(IL-4) cells (1.7 fold), and M(IL-10) cells (1.6 fold) (Fig. 11d). Mouse. Effects of the activation state on migration were similar to rat. Migration decreased 1.6-fold in M(I + T) cells, increased 1.9-fold in M(IL-4) cells, and 3.3-fold in M(IL-10) cells. Blocking Kir2.1 also inhibited migration of mouse microglia: by 7.2-fold in M(IL-4) cells and 14.2-fold in M(IL-10) cells. Again, while ML133 apparently reduced migration by 5.8-fold in unstimulated mouse microglia and by 20.8-fold in M(I + T) cells, ANOVA did not show statistical significance. Similar to rat cells, blocking Kv1.3 with AgTx-2 increased migration of unstimulated mouse microglia (2.9 fold), M(IL-4) cells (1.9 fold), and M(IL-10) cells (1.4 fold). While AgTx-2 appeared to increase migration of M(I + T) cells (3.2 fold), it was not statistically significant. For both species, we ruled out changes in the number of cells available to migrate. That is, a CyQuant NF assay showed that neither channel blocker altered the cell density over the 24-h migration period, regardless of activation state (data not shown).

Together, our results suggest that in microglia of both species, Kir2.1 activity facilitates migration and Kv1.3 activity inhibits it, regardless of the activation state.

Transcriptomics is increasingly used to profile responses of cells or tissues (e.g., after damage or disease). The information obtained can stand on its own or be used to decide which genes are interesting candidates for further study, including protein analysis by immunohistochemistry or Western blots. Most DNA is transcribed [71], and thus, differences in mRNA generally produce differences in protein. Exceptions to this can occur if a protein is either unusually stable or unstable, or if specific miRNAs interfere with mRNA translation or lead to mRNA degradation. For the proteins we examined by Western analysis (iNOS, COX-2, PYK2, CD206, ARG1, Iba1), the results correlated well with the mRNA data and further support the observed species differences.

While most studies of microglial activation have used primary rat or mouse cells, very few direct species comparisons have been made and the cell activation state was often not determined. We found one older study that directly compared primary rat and mouse microglia but it was restricted to glutamate secretion from unstimulated cells [9]. An interesting recent in vitro study found differences between responses of rat, mouse, and human microglia to oxygen glucose deprivation (OGD) but the resulting activation state was not determined [8]. In that study, rat and human microglia were similar but mouse cells differed in cytokine mRNA expression (IL-1α, IL-1β, IL-6, IL-10, and TNF-α) both at baseline and after OGD. In contrast, mouse and human levels of several chemokines (CX3CL1, CXCL10, CXCL12, CCL2, CCL3) were unaffected by OGD, but increased in rat microglia. A small number of in vivo studies have directly compared inflammatory responses in rat and mouse CNS but the microglial activation state was not determined. For instance, after focal cerebral ischemia in mice, induction of pro-inflammatory mediators (IL-1β, iNOS, TNF-α) in the infarcted tissue was lower and more delayed than in rat [72]. In another study, following intracortical microelectrode implantation, CD68 immunoreactivity declined over several weeks in rat but not in mouse [73]. Following spinal cord contusion, rats and mice had similar microglia/macrophage accumulation in the lesion (maximal infiltration by 7 days); however, mice had delayed and/or protracted infiltration of T lymphocytes and dendritic cells, and a unique cellular response consisting of clusters of fibrocytes forming a clear fibrous scar [74].

Previously reported changes in Kir and Kv currents with microglial activation have been inconsistent and hinted at species differences.

The first reports of Kir2.1 mRNA transcripts in microglia were in the mouse cell lines, BV-2 and C8-B4 [25, 88], and their activation state was not addressed. Recently, we showed that primary rat microglia express Kir2.1 mRNA transcripts [18]. The present study is apparently the first to show that microglia from both species express transcripts for several members of the Kir2 family (Kir2.1, Kir2.2, Kir2.3, Kir2.4) and that Kir2.1 transcript expression was much higher and was the only subtype that responded to the activating stimuli.

While the mRNA results suggest that the Kir current in rodent microglia is produced by Kir2.1, numerous patch-clamp studies have not confirmed its identity. It is useful to compare electrophysiological and pharmacological properties of the microglial current with cloned Kir2.1 channels. Like cloned Kir2.1, the microglial current shows fast activation kinetics, Na⁺-dependent relaxation at very hyperpolarized potentials, a single channel conductance of 25–30 pS [25, 52, 89], and a small outward current above the reversal potential [18, 90]. There is no selective inhibitor of Kir2.1 channels, and most studies have used the pan-Kir blocker, Ba²⁺. It is important to note that Ba²⁺ is a poor blocker at depolarized potentials [18], and this will compromise its value in cell function assays that address roles of Kir channels. A better blocker is the Kir2 family-specific blocker, ML133 (20 μM), which is not voltage dependent and blocks most of the Kir current in primary rat microglia within several minutes [18]. In the present study, a similar block by 20 μM ML133 was seen in rat (86%) and mouse (95%) microglia. Because ML133 takes time to enter the cell and act on an internal site on the channel [62], its efficacy depends on the duration of drug exposure. For mouse microglia, it was recently reported that 20 μM ML133 blocked only ~ 48% of the Kir current [91] but the exposure time was not stated.

Published responses of the Kir current to microglial activating stimuli have been quite variable. While the Kir current is present in most unstimulated rat and mouse microglia, effects of pro-inflammatory stimuli in rat microglia range from no change to a small increase (with LPS) or a large increase (with IFN-γ) [27, 52, 92, 93]. In contrast, in primary mouse microglia, LPS or IFN-γ reduced the Kir2.1 current [27, 94, 95]. Our direct species comparison showed that M(I + T) decreased the Kir2.1 current in mouse, with no change in rat. For anti-inflammatory activation states, there are even fewer publications. When TGF-β1 or TGF-β2 was used as a deactivating treatment on primary mouse microglia, there was no change in Kir current [25]. Here, we found that the Kir2.1 current was reduced by IL-4 in rat or IL-10 in mouse; changes that were consistent with the Kir2.1 transcript changes. For rat microglia, we previously reported a trend toward a decrease in current with IL-4 or IL-10 treatment but it did not reach statistical significance [18].

Kv1.2, Kv1.3, and Kv1.5 transcripts and protein have been detected in primary rat and mouse microglia [28, 29, 63‐65]. Here, we detected all three channel transcripts in unstimulated rat and mouse microglia but Kv1.5 was very low. Some studies have suggested that elevated Kv1.3 indicates a pro-inflammatory state; e.g., increases in mRNA and channel expression in M(LPS) rat microglia [29, 96], and the present increase in Kv1.3 transcript expression only in M(I + T) mouse microglia. However, for rat microglia, we now show that Kv1.3 expression increased in both M(I + T) and M(IL-4) cells; thus, this channel gene is not a reliable marker of a pro-inflammatory state. Kv1.2 was striking in showing opposite responses to M(I + T): an increase in mouse but a decrease in rat microglia. The only other related studies on Kv transcript expression we found were on mouse cell lines. Kv1.2 mRNA expression increased in M(LPS) BV-2 cells [64], and Kv1.3 increased after TGF-β1 or TGF-β2 treatment in BV-2 and C8-B4 cells [25, 88].

There is considerable variability in the prevalence of Kv currents reported in primary rodent microglia. For unstimulated microglia, early reports showed an absence in rat [89, 97], presence in a small proportion (<10%) of rat and mouse cells [27, 95] or moderate prevalence (30%) in rat cells [93], while several later studies found a Kv current in most to all rat microglia [28, 29, 50, 90, 96, 98]. Most studies have not used selective blockers to isolate the Kv1.3 portion of the current. Here, we observed an AgTx-2-sensitive Kv1.3 current in all unstimulated rat microglia, which were predominantly unipolar, and this is consistent with our earlier studies on unstimulated bipolar or amoeboid rat microglia [26, 28, 50]. In contrast, despite the similar morphology and molecular profile of unstimulated mouse microglia, the Kv current prevalence was much lower (~56% of cells). A margatoxin (MgTx)-sensitive Kv current was seen in <10% of cultured mouse microglia [27] but that current was not proven to be Kv1.3 because MgTx can block other Kv1-family channels [99]. There is some evidence that the Kv prevalence changes with age in mouse but this has not been examined in rat. In mice, a MgTx- and AgTx-2-sensitive Kv current was seen in 20–60% of postnatal (P5–9) microglia in acutely isolated brain slices [100]. Little or no Kv current was detected in acutely isolated adult mouse brain slices [100‐102] or microglia isolated from “control” mouse brain [103], but the prevalence increased to 29% in dystrophic microglia from aged mice [102]. A further complication is that the Kv prevalence and magnitude increased when mouse hippocampal slices were cultured ex vivo [101].

Some variability in the prevalence and amplitude of Kv currents is very likely due to differences in voltage protocols; e.g., using a depolarized holding potential, which is well known to inactivate Kv1.3 [26, 28, 93] and Kv1.5 channels [28, 104]. One study that did not detect Kv current in rat microglia used a holding potential of −20 mV [21]. Because Kv1.3 undergoes pronounced cumulative inactivation [28, 50‐52], the current amplitude can be substantially underestimated if the interval between voltage pulses is too short. Most studies do not state the interpulse interval [21, 27, 89, 95]. Additional variability in Kv current might also reflect a different initial activation state but this has rarely been determined. For instance, both Kv1.3 and Kv1.5 currents have been identified in rat and mouse microglia [28, 65] but, in rat microglia from hippocampal tissue prints, Kv1.5 initially produced the current and was replaced by Kv1.3 as the cells became more proliferative [28].

Both Kv1.3 and Kv1.5 currents have been reported in ex vivo rat microglia [28] and in M(LPS) mouse microglia [65]. We previously provided the first evidence for a role for Kv1.3 in microglia, where it contributed to proliferation shortly after preparing ex vivo tissue prints from rat brain slices [28]. Changes in Kv current with activating stimuli suggest that it is induced or increased in pro-inflammatory states. LPS or IFN-γ increased Kv currents in rat [29, 52, 92, 93] and mouse microglia [27, 65, 95, 105]. Here, we found that the pro-inflammatory M(I + T) state increased the total Kv current and Kv1.3 component in both species. However, for rat microglia, the Kv1.3 increase was not specific to the pro-inflammatory state, as IL-4 also increased the current. Consistent with Kv1.3 mRNA expression, the pro-inflammatory response was more specific in mouse cells, where only I + T increased Kv and Kv1.3 currents. Kv currents were also prevalent in “activated” microglia in damage models but the specific activation state and channel type were not determined; for instance, in the denervated rat facial nucleus [21], and in the ischemic cortex [22]. MgTx- and AgTx-2-sensitive Kv1.3 currents were found in microglia within the hippocampus of adult mice after status epilepticus but the prevalence and cell activation state were not determined [23]. After LPS injection into the mouse brain, the PAP-1-sensitive Kv1.3 current in acutely isolated adult microglia was fivefold larger [103]. The present results show that additional Kv channel types were active in mouse microglia; i.e., less than half the current was blocked by AgTx-2 in unstimulated cells and slightly more than half in M(I + T) cells.

Little is known about Kv currents in anti-inflammatory states. Mouse microglia stimulated with TGF-β1 or TGF-β2 had increased Kv currents that were blocked by kaliotoxin, charybdotoxin, or MgTx [25, 88] but the cell activation state was not characterized. In both species, we found that the total Kv and isolated Kv1.3 currents were unchanged in M(IL-4) and M(IL-10) states.

For rat microglia, we previously reported that M(LPS) cells migrate less than unstimulated cells, while M(IL-4) and M(IL-10) cells migrate more [18‐20, 30], and that blocking Kir2.1 channels reduced migration in unstimulated, M(IL-4) and M(IL-10) microglia [18]. We now report similar changes in migratory capacity with activation state in both species, and that, regardless of the activation state, blocking Kir2.1 inhibited migration while blocking Kv1.3 increased it. In the only directly relevant paper that we found, blocking Kv1.3 (and possibly other K⁺ currents) with MgTx in rat microglia reduced chemotaxis induced by monocyte chemoattractant protein 1 (CCL2) or ADP but, again, the activation state was not identified [106].

One potential role for Kv and Kir channels is to regulate the membrane potential and subsequent Ca²⁺ signaling, which is involved in cytoskeletal remodeling, adhesion and migration [107, 108]. For rat microglia, we previously showed that Ca²⁺ entry through Ca²⁺ release-activated Ca²⁺ (CRAC) channels is increased with hyperpolarization [109], and that blocking Kir2.1 channels with ML133 reduced CRAC-mediated Ca²⁺ entry and migration [18]. Similarly, blocking Kir2.1 in rat microglia with Ba²⁺ prolonged depolarization and reduced the amplitude of ATP-induced Ca²⁺ transients [110]. Blocking Kv1.3 in rat microglia disrupted membrane potential oscillations, showing a role in repolarization after depolarizing events [90]. This is apparently the first report of a role for Kv1.3 in microglia migration in the absence of a chemoattractant, and it was surprising that blocking Kv1.3 and Kir2.1 had opposite effects. Further evidence that the two channels do not always play the same functional roles is that Kir2.1 (but not Kv1.3) is involved in myelin phagocytosis and ROS production in activated rat microglia [17]. One possibility is that the channels have roles other than regulating K⁺ flux and membrane potential. For instance, cell migration requires integrins, which regulate adhesion to the extracellular matrix, and there is evidence that integrins and K⁺ channels can co-exist in signaling complexes [111]. A physical link between Kv1.3 and the β₁ integrin moiety was reported in T lymphocytes and melanoma cells [112, 113] but it is not known whether a similar association occurs in microglia. Future studies will be needed to determine the mechanisms by which Kir2.1 and Kv1.3 channels regulate microglia migration.

1. Many molecules were analyzed using gene profiling to assess microglial activation responses. While mRNA levels are often well linked to protein levels, this is not always the case. Several interesting changes were further assessed at the protein and functional level. While it is important that channel proteins were assessed at the level of functional expression (ion currents), their role in only one cell function was examined (migration). In future, it would be valuable to assess roles of proteins in other microglial functions in the different activation states; e.g., relating ECM-degrading enzymes to invasion (as in [19]), relating phagocytosis receptors and NOX enzymes to phagocytosis and ROS production ([17]). 2. We assessed one time point (24 h for gene expression), which was chosen to facilitate comparisons with published gene analyses that include 24 h in vitro. For rat microglia, we knew that many genes respond at this time and, here, we found that some genes responded similarly in both species. However, it is possible that other responses, such as some non-responding genes in mouse cells reflect a different timing, and in future, one could examine the time course. 3. It is important to compare responses of both species to stimuli that are physiologically relevant to acute CNS damage in vivo. We used single cytokines or I + T but did not examine repolarization between M1 and M2 states, as we had previously done ([17]). The ability to compare with the literature is reduced because many studies have used LPS from gram-negative bacteria as the stimulus. 4. Although this study used primary cultured microglia (rather than cell lines), it was an in vitro study. We believe the results will at least aid in selecting and interpreting markers and molecules in animal models of CNS disease. Many rodent in vivo studies have presented evidence for M1 and M2 states of microglia (and macrophages) in both acute damage and chronic disease scenarios. For example, recent in vivo profiling shows gene expression and immunohistochemical evidence for M1 and M2a states in injured brain tissue [114‐120]. Based on in vitro responses to cytokine stimulation, studies are increasingly using cytokines or their inhibitors in vivo; e.g., to promote the M2 phenotype [55, 121, 122]. Ultimately, it will be important to assess how rodent in vivo studies relate to human disease. 5. This study examined expression of numerous genes, verified changes of several at the protein level, and examined protein function for iNOS, Kv and Kir channels. However, in future, it would be valuable to examine the functionality of other proteins, such as the ROS-producing NOX enzymes.

Studies of disease mechanisms rely on animal models, especially rodents. To fulfill the larger goal of translating such results to humans, it is crucial to understand whether rats and mice respond the same and, if not, which species is a more reliable model. There is considerable debate as to how closely mouse models resemble human responses in inflammatory diseases [123, 124]. While some in vivo results suggest that CNS inflammation differs between rats and mice, it is premature to ascribe the differences to microglial responses. For peripheral immune cells, rat and mouse responses are sometimes assumed to be comparable [125]. However, very few species comparisons have been published, and our findings illustrate important differences in molecular activation profiles between rat and mouse microglia. This might help explain why, after CNS injury/disease, some “hallmark” microglial activation state markers have been detected by immunohistochemistry in mouse but not in rat. While it is sometimes assumed that the antibodies do not work in rat (despite manufacturers’ claims), it might be that the species responses actually differ.

Microglial properties are increasingly being investigated in human tissue—often in surgical biopsies from epileptic patients—but the cell activation state is usually not determined and they cannot be assumed to be normal, resting cells. Further complications include the potential for strain differences in rodents [126], and genetic polymorphisms and epigenetic changes in humans [127]. Increasingly, in vivo studies include injection of a single stimulus (e.g., LPS, IFNγ, IL-4, IL-10, IL-13) in an attempt to skew the brain toward one activation state; however, measured brain responses reflect cell-cell interactions and numerous cell types. In vitro studies are the only way to stimulate just the microglia in order to elucidate similarities and differences in how different species respond. The information gained will be useful to help interpret results of previous and future in vivo studies. Many potential treatments identified in rodents have failed in human clinical trials. To narrow this translational gap, it is essential to investigate and acknowledge species similarities and differences in immune responses.