Comparing Effects of Transforming Growth Factor β1 on Microglia From Rat and Mouse: Transcriptional Profiles and Potassium Channels

The cytokine, transforming growth factor β1 (TGFβ1), is up-regulated after central nervous system (CNS) injuries or diseases involving microglial activation, and it has been proposed as a therapeutic agent for treating neuroinflammation. Microglia can produce and respond to TGFβ1. While rats and mice are commonly used for studying neuroinflammation, very few reports directly compare them. Such studies are important for improving pre-clinical studies and furthering translational progress in developing therapeutic interventions. After intracerebral hemorrhage (ICH) in the rat striatum, the TGFβ1 receptor was highly expressed on microglia/macrophages within the hematoma. We recently found species similarities and differences in response to either a pro-inflammatory (interferon-γ, IFN-γ, +tumor necrosis factor, TNF-α) or anti-inflammatory interleukin-4 (IL-4) stimulus. Here, we assessed whether rat and mouse microglia differ in their responses to TGFβ1. Microglia were isolated from Sprague-Dawley rats and C57BL/6 mice and treated with TGFβ1. We quantified changes in expression of >50 genes, in their morphology, proliferation, apoptosis and in three potassium channels that are considered therapeutic targets. Many inflammatory mediators, immune receptors and modulators showed species similarities, but notable differences included that, for some genes, only one species responded (e.g., Il4r, Il10, Tgfbr2, colony-stimulating factor receptor (Csf1r), Itgam, suppressor of cytokine signaling 1 (Socs1), toll-like receptors 4 (Tlr4), P2rx7, P2ry12), and opposite responses were seen for others (Tgfb1, Myc, Ifngr1). In rat only, TGFβ1 affected microglial morphology and proliferation, but there was no apoptosis in either species. In both species, TGFβ1 dramatically increased Kv1.3 channel expression and current (no effects on Kir2.1). KCa3.1 showed opposite species responses: the current was low in unstimulated rat microglia and greatly increased by TGFβ1 but higher in control mouse cells and decreased by TGFβ1. Finally, we compared TGFβ1 and IL10 (often considered similar anti-inflammatory stimuli) and found many different responses in both species. Overall, the numerous species differences should be considered when characterizing neuroinflammation and microglial activation in vitro and in vivo, and when targeting potassium channels.


INTRODUCTION
In the central nervous system (CNS), transforming growth factor β1 (TGFβ1) is important for development of microglia (Butovsky et al., 2014), and is thought to help maintain their ''resting'' state (Spittau et al., 2013). TGFβ1 is up-regulated after CNS damage or in disease states that involve microglial activation, including stroke and neurodegenerative diseases (Lehrmann et al., 1998;Lively and Schlichter, 2012;Lively et al., 2016;Chen et al., 2017;Taylor et al., 2017). TGFβ1 is expressed in microglia in the healthy brain (De Groot et al., 1999) as is the TGFβ1 receptor (De Groot et al., 1999;Pál et al., 2014). After experimental stroke, TGFβ1 (Wiessner et al., 1993;Lehrmann et al., 1998) and the TGFβ1 receptor (Pál et al., 2014) are up-regulated, especially in microglia/activated macrophages. Brain TGFβ1 also rapidly increased in a mouse model of intracerebral hemorrhage (ICH; Taylor et al., 2017). To improve the use of animal models for translation to human CNS diseases, it is important to determine whether immune responses are conserved among species. As a prelude to this in vitro study, we first determined that TGFβ1 mRNA  and the TGFβ1 receptor protein (present study) increased in a rat model of ICH.
There are surprisingly few studies comparing microglial responses in these rodent species. Our recent study showed many differences in responses of rat and mouse microglia to pro-and anti-inflammatory cytokines (Lam et al., 2017). Based on microglial responses to TGFβ1 in vitro and in vivo, it is thought that this cytokine promotes a homeostatic phenotype (Abutbul et al., 2012;Spittau et al., 2013;Butovsky et al., 2014;Krasemann et al., 2017). The microglial reaction after CNS damage is often followed by resolution and repair phases, and many authors have defined a microglial ''acquired deactivation'' state, similar to macrophage responses to TGFβ1, interleukin-10 (IL-10) or glucocorticoids (Mantovani et al., 2004;Boche et al., 2013;Cherry et al., 2014;Orihuela et al., 2016). Recently, some transcriptional profiling of TGFβ1 responses has been published for mouse microglia (Butovsky et al., 2014;Taylor et al., 2017), and two studies compared a small number of molecular responses to TGFβ1 and IL-10 in mouse (Ledeboer et al., 2000) and rat microglia (Lodge and Sriram, 1996). However, more detailed comparisons of both species are needed. Here, we compared transcriptional profiles of rat and mouse microglia in response to TGFβ1 application in vitro by quantifying >50 genes that included microglial activation markers, proand anti-inflammatory mediators and receptors, signaling molecules, nicotinamide adenine dinucleotide phosphate (NADPH) oxidases (NOX enzymes), purinergic and phagocytic receptors, as well as several ion channels and regulators. At the protein level, we also assessed the TGFβ1 receptor (TGFβR1), the interferon-γ (IFNγ) receptor (IFNGR1), the mannose receptor (CD206/MRC1) and the Ca 2+ -activated K + channel, KCa3.1. While there were some species similarities in responses to TGFβ1, the observed differences reinforce the importance of analyzing both species in pre-clinical studies. Moreover, in comparing the TGFβ1 responses with our recently published analysis of IL-10 responses of primary rat and mouse microglia, substantial differences were seen between these two anti-inflammatory cytokines.
When considering microglial contributions to CNS damage and recovery, numerous recent studies using specific K + channel blockers have shown roles for Kv1.3, Kir2.1 and KCa3.1 channels, leading to their consideration as targets for regulating neuroinflammation (for recent reviews and further references, see Lam and Schlichter, 2015;Lam et al., 2017;Nguyen et al., 2017a,b). In principle, K + channels hyperpolarize cells and increase the driving force for Ca 2+ entry. Previous work from our lab (for instance, Jiang et al., 2003;Ohana et al., 2009;Ferreira et al., 2014;Siddiqui et al., 2014;Lam and Schlichter, 2015) and others (e.g., Heo et al., 2015;Michaelis et al., 2015) shows that microglia use Ca 2+ entry through store-operated channels (SOC) and Ca 2+ -release-activated Ca 2+ (CRAC) channels to regulate many microglial functions, including proliferation, migration, phagocytosis and cytokine production (reviewed in Möller, 2002;Färber and Kettenmann, 2006). Therefore, we compared transcript expression between rat and mouse microglia, with and without TGFβ1 treatment, for several Ca 2+ and K + channels and regulators; as well as current amplitudes (the final outcome of functioning channel proteins) for Kir2.1, Kv1.3, and KCa3.1. Again, despite some similar responses, there were species differences that should be considered for future translational studies.

MATERIALS AND METHODS
All procedures on animals adhered to the Canadian Council on Animal Care guidelines for humane animal use and were approved by the University Health Network Animal Care Committee (Animal Use Protocols 914, 1573 and 1132).

Isolating Microglia and Treating Them With TGFβ1
Neonatal Microglia Microglia were isolated from Sprague-Dawley rat pups (P1-P2) and C57BL/6 mouse pups (P0-P2; both from Charles River), as recently described (Lam et al., 2017). These strains were selected because they are widely used, and C57BL/6 is the main strain for transgenic studies. In brief, mashed brain tissue was strained and centrifuged (300× g, 10 min) in cold Minimal Essential Medium (MEM; Thermo Fisher Scientific, RRID:SCR_008452; Cat# 11095080). Mixed brain cell cultures were established by re-suspending the pellet in MEM, incubating in 75 cm 2 flasks with 20 ml MEM supplemented with 10% fetal bovine serum (FBS; Wisent; Cat# 080-150) and 0.05 mg/ml gentamycin (Thermo Fisher Scientific; Cat# 15710072). The medium was changed after 2 days, and microglia (≥99% pure) were harvested a few days later by shaking the flasks (5 h, 65 rpm) on an orbital shaker in the incubator, centrifuging the supernatant, and re-suspending the cells in fresh MEM (with 2% FBS, 0.05 mg/ml gentamycin). Microglia were seeded onto UV-irradiated 15 mm coverslips (Fisher Scientific,Ottawa,ON,Canada; at densities that are noted under each assay. Their initial state was similar for both species, as confirmed from the expression of numerous genes (see ''Results'' section).

Proliferation
Proliferation was quantified using the CyQUANT TM NF assay (Thermo Fisher Scientific; Cat # C35006), as before (Lam and Schlichter, 2015). Microglia were seeded at 2-3 × 10 4 cells/well of a 96-well flat-bottom plate in MEM with 2% FBS and allowed to settle for 24 h, then left untreated (control) or stimulated with TGFβ1 for 24 h. The detection dye was added to each well for 30 min (37 • C, 5% CO 2 ), and the fluorescence intensity was measured using a multi-label plate reader (Victor 3 1420, Perkin Elmer, Woodbridge, ON, Canada), with excitation at 485 nm and emission at 535 nm. Duplicate readings were taken for 0.1 s at 3 mm from the bottom of the plate, averaged, and then the background was subtracted before normalizing to the control group.

Apoptosis
Cells were seeded at 4 × 10 4 cells/well in 96-well plates (as for proliferation, above), treated with TGFβ1 for 24 h, and fixed in 4% PFA for 7 min, washed in PBS and post-fixed in methanol for 20 min. Apoptotic cells were detected using the TiterTACS TM Colorimetric Apoptosis Detection Kit (Trevigen Cell Assays, Gaithersburg, MD, RRID:SCR_012449; Cat# 4822-96-K), with positive controls (TACS-Nuclease TM added for 10-20 min to generate DNA breaks) and negative controls (omitting TdT enzyme). Absorbance was measured at 450 nm using a SPECTROstar Nano plate reader (BMG LABTECH, Offenburg, Germany).

NanoString Gene Expression Analysis
Microglia were seeded at 5 × 10 5 cells/coverslip in a 12-well culture plate, allowed to settle for 1-2 days (37 • C, 5% CO 2 ), and then left untreated (control) or stimulated with TGFβ1 for 24 h. In a pilot study on primary rat microglia, some genes were also analyzed at 6 h after TGFβ1 treatment and the relevant results are mentioned in the text. The sample preparation and transcriptional profiling were conducted as before Siddiqui et al., 2014Siddiqui et al., , 2016Lam et al., 2017). Microglia were stimulated with TGFβ1 for 24 h, and then total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific; Cat# 15596018), purified using an RNeasy Mini Kit (QIAGEN, Mississauga, ON, Canada; Cat# 74104), and stored at −80 • C. Extracted RNA (200 ng per sample) was sent to the Princess Margaret Genomics Centre (Toronto, ON, Canada 2 ) where the assay was conducted (hybridization, detection, scanning). Separate plates had to be used for each species, with different probe sets that were designed and synthesized by NanoString nCounter technologies (see Supplementary Tables S1, S2 for probe sequences). Because gene names sometimes differ between species, for clarity, rat names are used.
Data were analyzed with nSolver Analysis Software (ver3.0; RRID:SCR_00342). Each transcript of interest was recognized by a capture probe and a reporter probe, and assay variability was minimized using negative reporter probes for background subtraction. Positive control reporter probes were included to assess the hybridization efficiency and detection range, and to calculate a scaling factor that was applied to all mRNA counts in each sample. Finally, a reference gene scaling factor was calculated in the same manner using the housekeeping genes, Hprt1 and Gusb, and used to adjust the counts for each sample (control, TGFβ1 treated).

Western Blot Analysis
Protein changes generally map well onto mRNA changes, which we previously validated for several molecules investigated here (iNOS, COX-2, Pyk-2, CD206, Arg1, Iba1; Lam et al., 2017), and here, we examined four proteins using the same methods. Microglia were seeded on 25 mm coverslips (Fisher Scientific, Ottawa, ON, Canada; Cat# 12-545-86) in 35 mm culture dishes at 1-3 × 10 6 cells (4-7 independent cell cultures for rat; 4-8 for mouse). Cells were incubated for 24 h with or without TGFβ1, and then briefly washed with PBS, lysed for 30 min in ice-cold RIPA lysis buffer with a mammalian protease inhibitor cocktail (Sigma-Aldrich; Cat# P3840), and spun down to remove insoluble material. Protein concentration was determined with a Pierce TM BCA protein assay (Thermo Fisher Scientific; Cat# 23225), and lysates were stored at −80 • C until used. Proteins were denatured (100 • C for 5 min in a dry-bath incubator) in NuPage LDS sample buffer (Thermo Fisher Scientific; Cat# NP0007) with 5% 2-β-mercaptoethanol. Samples were subjected to SDS-PAGE on 8% or 12% acrylamide gels at 10 µg protein/lane, and electrophoresed for 1.5-2 h at 80 mV (stacking gel) and 120 mV (resolving gel). Proteins were then transferred to a PVDF membrane and blocked (2-3 h) in 5% non-fat dry milk in Tris-Tween buffered saline (TTBS).
We used total protein normalization of Coomassie-stained immunoblots (Welinder and Ekblad, 2011) by adding 0.1% Coomassie Brilliant Blue G (Sigma-Aldrich; Cat# B8522) for 1 min, de-staining for 2 min in acetic acid/methanol/water (1:5:4), air-drying, and imaging with a ChemiDoc TM XRS System. Using Image Lab (ver.5.2.1, RRID:SCR_014210), gel lanes and bands were user-delineated, background was subtracted using the software's rolling disk algorithm, and signal intensities of identified bands were determined. Bands of interest were then normalized to the total Coomassie blue staining intensity of a given lane, and expressed as fold-changes relative to unstimulated (control) cells. Supplementary Figure S1 shows uncropped images of representative blots used for quantification.

Whole-Cell Patch-Clamp Recordings
Microglia were plated at 7-9 × 10 4 cells/coverslip, mounted in a 300 µL volume perfusion chamber (Model RC-25, Warner Instruments, Hamden, CT, USA) at room temperature, and recorded in the whole-cell configuration. The bath solution consisted of (in mM): 125 NaCl, 5 KCl, 1 CaCl 2 , 1 MgCl 2 , 10 HEPES, 5 D-glucose, adjusted to pH 7.4 and 290-300 mOsm/kg H 2 O. Pipettes were pulled from thin wall borosilicate glass (World Precision Instruments, Sarasota, FL, USA, RRID:SCR_008593) on a Narishige puller (Narishige Scientific, Setagaya-Ku, Tokyo) and fire polished with a microforge (MF 900; Narishige) from about 2 MΩ to 5-8 MΩ. The pipette (internal) solution was 40 KCl, 100 K aspartate, 1 MgCl 2 , 10 HEPES, 2 MgATP (pH 7.2; 290-300 mOsm/kg H 2 O), and with 0.5 CaCl 2 and 1 EGTA to buffer internal free Ca 2+ to ∼120 nM. The same bath and pipette solutions were used to record all currents. Data were acquired using an Axopatch 200A amplifier, filtered at 5 Hz with a DigiDATA 1322A board, and analyzed with pCLAMP 10 software (all from Molecular Devices, Sunnyvale, CA, USA). Junction/offset potentials were calculated with the pCLAMP utility and used to correct each figure. Voltage protocols are indicated in the Figure Legends and the rationale briefly explained in the ''Results'' section text.

Statistics
Quantified data are expressed as mean ± SEM for the number of biological replicates indicated in figure and table legends. For Nanostring mRNA analysis, a one-way analysis of variance (ANOVA) with Dunnett's post hoc test was used to assess TGFβ1-induced changes within a rodent species. To compare effects of TGFβ1 and IL-10 between rat and mouse microglia, mRNA counts were normalized to housekeeping genes, and then results were expressed as fold-changes relative to control cells. A 2-way ANOVA with Fisher's LSD test was conducted in Graphpad Prism ver 6.01 (RRID:SCR_002798). The p values for differences in treatment or species were adjusted using a 5% false discovery rate correction for multiple comparisons (Benjamini and Yekutieli, 2001) in the program R (version 3.3.1; R Project for Statistical Computing, RRID:SCR_001905). Proliferation, apoptosis, Western blotting, and electrophysiological data were analyzed in GraphPad Prism using a 2-way ANOVA with Bonferroni or Tukey's post hoc analysis.

RESULTS
As mentioned in the ''Introduction'' section, TGFβ1 is often increased after acute CNS damage, and microglia are important responders to this cytokine. We previously reported an increase in TGFβ1 mRNA in the damaged rat striatum after ICH was induced by collagenase injection . Before assessing TGFβ1 effects on isolated microglia, we wanted to ensure that the TGFβR1 receptor is well expressed on microglia in this model of acute CNS injury. We previously showed that at 3-7 days after ICH in the rat striatum, the number of microglia/activated macrophages dramatically increases in the hematoma Moxon-Emre and Schlichter, 2011;Lively and Schlichter, 2012). Here, we show that at 7 days after ICH, there was substantial TGFβR1 staining in microglia/macrophages in the hematoma ( Figure 1A; CD11b staining; the hematoma was readily identified by the presence of blood). More distant from the hematoma (toward the right side of each image), and in the undamaged contralateral striatum, there was ubiquitous, diffuse TGFβR1 staining. The remainder of this study addresses the questions: what are the effects of TGFβ1 exposure on isolated microglia, and do the outcomes differ between rat and mouse?

Effects of TGFβ1 on Microglial Morphology, Proliferation and Apoptosis
To assess the general health of isolated rat and mouse microglia and their response to TGFβ1, we first examined morphology, proliferation and apoptosis in a pilot study using 5, 10 and 20 ng/ml of TGFβ1 (Figures 1B-D). As seen in Figure 1B, untreated cells had a similar morphology to our recent studies of primary rat (Siddiqui et al., 2012;Lively and Schlichter, 2013;Lam et al., 2017) and mouse (Lam et al., 2017) microglia. That is, about half the cells had a distinct unipolar, migratory phenotype with a lamellum at the leading edge and a tail/uropod at the trailing end (50 ± 2%, 12 individual cultures for rat; 53 ± 2%, eight individual cultures for mouse). For control cells, the main species differences are the wider lamellum in rat (32.5 ± 1.5 µm) than mouse (25.4 ± 0.9 µm; p = 0.0025, unpaired t-test), and greater cell length in mouse (79.2 ± 3.9 µm) than rat (68.0 ± 2.0 µm; p = 0.0273, unpaired t-test). Consequently, the aspect ratio (lamellum width ÷ cell length) was 0.54 ± 0.02 for rat and 0.35 ± 0.01 for mouse (p < 0.0001). Moreover, only in rat cells was the lamellum ever wider than the cell length.
TGFβ1-treated rat cells showed increased CD11b staining and retracted their lamellae and uropods at all concentrations tested. However, visual inspection cannot determine whether FIGURE 1 | Transforming growth factor β1 (TGFβ1) receptor in microglia after Intracerebral hemorrhage (ICH), and general effects of TGFβ1 on microglia in vitro. (A) Microglia/macrophages increase in the hematoma (H) after ICH and express TGFβR1. Representative confocal micrographs taken at 7 days in the damaged ipsilateral rat striatum (left) and the undamaged contralateral striatum (right). The inset is a higher magnification image of the boxed area (scale bar, 20 µm) showing microglia and infiltrating macrophages in the hematoma labeled with CD11b (OX-42 antibody; red) and TGFβR1 (green). Cell nuclei are labeled with 4'-6-diamidino-2-phenylindole (DAPI; blue). Scale bar for main images is 100 µm. (B) The morphology of isolated microglia (stained for CD11b, red, as in (A) was assessed using phalloidin to stain F-actin (green) and DAPI (blue) to label nuclei. Images are representative of results from six separate cell cultures for each species. Scale bar is 100 µm. (C) Microglia proliferation was determined 24 h after TGFβ1 treatment using the CyQUANT TM assay. Results are shown as fold-change with respect to untreated (control, CTL) microglia (blue dashed line) and expressed as mean ± SEM for 9-11 individual cell cultures. (D) Apoptosis was monitored using the TiterTACS TM colorimetric assay at 24 h after TGFβ1 treatment. Results are shown as mean absorbance at 450 nm (mean ± SEM, 8-9 individual cultures). The blue dashed line indicates values for the negative controls, and the red dashed line indicates the positive controls. Significant differences are shown between control and TGFβ1 treated cells ( * ). One symbol of indicates p < 0.05; 2, p < 0.01; 3, p < 0.001. the rounding up reflects a change in size or just a change in morphology at a constant size. The measured membrane capacitance can be used as a direct measure of total cell membrane area, regardless of 3-dimensional cell shape. For instance, capacitance accounts for retraction of cell processes, membrane ruffling and changes in cell height. This is because the bimolecular lipid membrane is a remarkably consistent dielectric, and specific membrane capacitance is, in essence, a biological constant (∼1 µF/cm 2 ). We measured cell capacitance during patch-clamp experiments (below) and showed no differences between species or with TGFβ1 treatment. For control microglia, the values (mean ± SEM) were 25.3 ± 1.8 pF (n = 32) for rat, and 24.2 ± 1.4 pF (n = 45) for mouse. For mouse microglia, 24 h treatment with TGFβ1 did not obviously affect the

Control TGFβ1
Relative mRNA counts ± SD Fold change wrt control Rat and mouse microglia were stimulated for 24 h with transforming growth factor β1 (TGFβ1). To assess species differences in basal expression, unstimulated (control) mRNA counts/200 ng RNA sample are expressed as mean ± SEM (n = 4-6 individual cultures). For clarity, protein names are included for some genes. Effects of TGFβ1 on a given gene are expressed as fold changes with respect to species-matched control levels. Arrows indicate statistically significant increases (↑) or decreases (↓) in expression within a species. Species differences are indicated by bold numbers and asterisks. One symbol (arrow or asterisk) indicates p < 0.05; two, p < 0.01; three, p < 0.001. morphology or intensity of CD11b staining, and the capacitance was 22.1 ± 1.0 pF (n = 41; p = 0.2206, unpaired t-test).
Despite rat cells rounding up, their capacitance was not affected (22.3 ± 1.5 pF; n = 29; p = 0.2125, unpaired t-test). Thus, the shape changes were due to rounding up, not to cell size changes. The longest TGFβ1 treatment time used for gene and protein analysis in this study was 24 h. The proliferation assay showed a modest decrease (20%-25%) at 24 h in the number of rat microglia at all TGFβ1 concentrations; while mouse cell numbers were unaffected ( Figure 1C). Despite the decrease in rat cell numbers, TGFβ1 treatment did not evoke apoptosis at 24 h in either species at any concentration tested ( Figure 1D). This suggests that the decrease in rat cell numbers was an effect on cell cycle, although this was not directly examined. Based on the morphological changes and slightly decreased proliferation of rat microglia, we chose the lowest effective TGFβ1 concentration (5 ng/ml) for subsequent rat experiments, while 20 ng/ml was used for mouse.

Transcriptional Profiling to Compare Rat and Mouse Microglial Response to TGFβ1
The next goal was to compare responses of rat and mouse microglia to TGFβ1 by profiling transcript expression of >50 selected genes. We have organized the results into functional groupings, including microglial markers and immunomodulators, pro-and anti-inflammatory mediators and their receptors, and some molecules related to specific microglial physiological functions. The gene nomenclature is indicated for rat (Supplementary Table S1) and mouse (Supplementary Table S2), and where different, the rat names are used.

Microglia Markers and Immune Modulators
First, we quantified expression of several genes that are often used to identify microglial ''activation'' after acute brain injury, as well as several immunomodulatory molecules (Table 1;  Supplementary Table S3).

TGFβ1 Treated
Of genes in this category, 6/12 were altered in one or both species but there were several species differences in the magnitude of responses (Table 1). For instance, rat microglia had greater up-regulation of Cx3cr1, Itgam and Csf1r; mouse cells had reduced Nr3c1 and Tlr4 ; and rat cells had reduced Socs1. Several genes in this category were not significantly altered in either species: Aif1, Cd68, Nfkbia, Socs3, Tlr2 and Tspo (Supplementary  Table S3). There were several similarities in a pilot study of 6 h TGFβ1 treatment of rat cells, including up-regulated Cx3cr1 and Itgam but no changes in Tlr2 or Tlr4 (not illustrated).

TGFβ1 Treated
Changes were seen in 12/18 genes examined in one or both species. Both species had decreased Mrc1 and Pparg, and increased Tgfbr1 ( Table 2). We previously showed that mRNA changes in several hallmark genes corresponded well with protein changes in both rat and mouse microglia (Lam et al., 2017). Here, we corroborated protein changes for two exemplary molecules. Figure 2 shows representative Western blots and summarized protein changes for CD206 (Mrc1; Figure 2A) and TGFβR1 ( Figure 2B). In accord with the mRNA changes, CD206 protein decreased in both species (Figure 2A), and TGFβR1 protein increased, particularly in rat microglia ( Figure 2B). With respect to mRNA levels, there were several species differences in the magnitude or even in the direction of responses. For instance, Tgfb1 and Myc increased in rat but decreased in mouse. Only rat cells showed increases in Tgfbr2 and Trem2, and only mouse cells showed increases in Il4, Il10, Il4r and Il13ra1. Interestingly, neither species showed a change in several alternative activation-associated genes: Arg1, Ccl22, Cd163, Retnla or the IL10 receptors: Il10ra and Il10rb (Supplementary  Table S3). A pilot study of 6 h TGFβ1 treatment of rat cells showed several similarities: decreased Mrc1 and Pparg, and no change in Ccl22 (not illustrated).

TGFβ1 Treated
There were changes in 8/17 in one or both species, but again, there were species differences. Ifngr1 increased in rat but decreased in mouse. Representative Western blots and summarized protein changes for IFNGR1 ( Figure 2C) corroborate this species difference in response to TGFβ1 treatment. In rat, five genes were decreased (Ccl3, Ccr5, Il1b, Tnfa, Tnfrsf1b) and Trem1 was increased. There were some similarities in mouse (decreased Ccl3 and Ccr5, increased Trem1) but a unique response was decreased Casp1, an enzyme that contributes to IL1β activation. Hallmark pro-inflammatory genes that did not change in either species included Ccr2, Ifng, Ifngr2, Il1r1, Il6, Nos2, Ptgs2, Ptk2b and Tnfrs1a (Supplementary  Table S3). It appears that many transcriptional responses to TGFβ1 were rapid; i.e., in a pilot study of 6 h treatment of rat microglia (not illustrated), 17/25 genes showed qualitatively similar responses to those seen at 24 h; e.g., increases in Cx3cr1, Itgam, Ncf1 and P2rx7; decreases in Tnfa, Mrc1 and Pparg, and no changes in Nos2, Il4, Il4ra, Tlr4, Cd163 and Retnla.

Additional Genes Related to Microglial Physiology
Our earlier qRT-PCR (Sivagnanam et al., 2010) and recent Nanostring analyses (Siddiqui et al., 2016;Lam et al., 2017) showed that unstimulated rat microglia express several phagocytosis-related receptors, oxygen radical-producing NOX enzymes, as well as purinergic receptors that can respond to adenosine triphosphate (ATP) released from damaged and dying cells, modulate microglial phagocytosis, reactive oxygen species production and migration. Expression of some of these molecules changed according to the activation state (Sivagnanam et al., 2010;Siddiqui et al., 2016;Lam et al., 2017). For instance, those studies showed that pro-inflammatory stimuli elevated Cybb (NOX2), Ncf1, Fcgr3a; while IL-4 treatment reduced Cybb and increased Fcgr2b. Cybb is the main contributor to the respiratory burst in microglia (Siddiqui et al., 2016), and its activation requires (neutrophil cytosolic factor 1, Ncf1/p47phox) relocation to the plasma membrane (reviewed in El-Benna et al., 2009). The voltage-gated proton channel, Hvcn1 (Hv1), can promote NOX enzyme activity by allowing H + efflux as charge compensation for NADPH-generated electrons (Ramsey et al., 2009;DeCoursey, 2013).

TGFβ1 Treated
Several genes were affected by TGFβ1 (Table 4), including up-regulated Fcgr2b, Ncf1 (higher in rat) and Nox1 in both species. Species differences were decreased Fcgr3a and increased P2rx7 (also seen at 6 h, not illustrated) and P2ry12 in rat cells only. In both species, Adora1, Adora2a, Cybb, Fcgr1, Hvcn1, Msr1, Nox4 and P2ry2 were unchanged (Supplementary Table S3).  2) 73 ± 47 in rat, 10 ± 6 in mouse (p = 0.011; n = 6 each); Kcna3 (Kv1.3) 109 ± 18 in rat, 70 ± 14 in mouse (p = 0.013); Kcna5 (Kv1.5) 4 ± 2 in rat, 3 ± 2 in mouse. TGFβ1 treatment did not alter Kv1.2 mRNA, but significantly increased Kv1.3 mRNA: about 7.4-fold at 24 h in rat (∼10 fold at 6 h, data not shown), and 5.6-fold in mouse ( Figure 3A). Kv1.5 mRNA also increased about 4.2-fold in rat and 3.4-fold in mouse cells. Low levels of mRNA for K + channels can nonetheless result in robust currents, as we previously observed for Kv1.3 and Kir2.1 (Lam and Schlichter, 2015;Lam et al., 2017), possibly because channel proteins can have very slow turnover (Jugloff et al., 2000;Balut et al., 2012). Thus, it is important to measure the currents as a direct measure of functional ion channels. We used whole-cell patch-clamp recordings to compare the total Kv and Kv1.3 currents with and without TGFβ1 treatment. Cells were treated for 30 h to provide additional time for channel protein expression, trafficking and post-translational modifications, as before (Lam et al., 2017). The cell capacitance (shown in the first section of the ''Results'') was used to normalize each current to the cell size.

Kv and Kv1.3 Currents
The voltage protocols used to quantify Kv currents were essentially as in our earlier work (Kotecha and Schlichter, 1999;Cayabyab et al., 2000;Newell and Schlichter, 2005;Lam et al., 2017). Importantly, they were designed to consider that the current amplitude depends on the holding potential, test potential, and frequency of depolarization (Grissmer et al., 1994;Nörenberg et al., 1994;Kotecha and Schlichter, 1999;Cayabyab et al., 2000). The holding potential (−105 mV) was used to relieve channel inactivation, and 60 s intervals between successive depolarizing steps allowed recovery from use-dependent (cumulative) inactivation. The rat protocol required ∼20 min, and because mouse recordings did not usually last long enough, the protocol was modified to a voltage ramp from −75 mV to +45 mV from the −105 mV holding potential. This protocol was first validated by ensuring that the amplitude at +45 mV was the same as for a single voltage step.

Neonatal Rat
Kv currents were observed in every rat microglial cell examined, with or without TGFβ1 treatment (examples shown in Figure 3B), and a similar proportion of the total current was Kv1.3 (i.e., blocked by AgTx-2) in both groups: 66 and 75% ( Figure 3C). Both the total Kv current and AgTx-2 sensitive Kv1.3 component activated in a time-and voltagedependent manner above about −40 mV (corrected for junction potential offsets), and the currents increased with TABLE 4 | Transcript expression of additional genes related to microglial physiology and ion signaling.

Control TGFβ1
Relative mRNA counts ± SD Fold change wrt control  depolarization (Figures 3B,D). Time-dependent inactivation was also apparent during the 1-s depolarizing test pulses. TGFβ1 treatment increased the total Kv current and AgTx-2 sensitive Kv1.3 component (Figures 3D,E), and is consistent with the increase in Kv1.3 mRNA ( Figure 3A).

Neonatal Mouse
Two species differences in control mouse microglia were the lower prevalence of observing a Kv current (18/31 cells) and that, among cells with a detectable current, less was AgTx-2-sensitive compared to control rat microglia (∼40%; Figure 3C). This implies that unstimulated mouse microglia have an additional, unidentified Kv current type. Of note, all TGFβ1-treated mouse cells expressed a Kv current (20/20), and a greater proportion of the current was Kv1.3 (65%; p < 0.001). TGFβ1 increased the amplitudes of total Kv and Kv1.3 currents (Figures 3B,D), consistent with the increase in Kv1.3 mRNA. In both species, control and TGFβ1-treated cells produced substantial currents despite the relatively low Kcna3 mRNA levels. This is consistent with our previous studies on neonatal microglia from rat (Kotecha and Schlichter, 1999;Cayabyab et al., 2000;Newell and Schlichter, 2005;Lam et al., 2017) and mouse (Lam et al., 2017). Figure 4 shows Kv currents in adult rat and mouse microglia. In both species, a Kv current was always observed in untreated and TGFβ1-treated cells (>100 cells). In control cells, the proportion of Kv current contributed by Kv1.3 was similar in rat and mouse cells (∼62 and 71%, respectively; Figure 4A). After TGFβ1 treatment, the Kv1.3 component was 78% in mouse but only 56% in rat because two cells appeared to express an AgTx-2insensitive Kv current ( Figure 4A). Effects of TGFβ1 on current amplitudes were similar to neonatal cells; i.e., both total Kv and Kv1.3 currents increased (Figures 4B,C).

Kir2.1 mRNA and Current Were Unchanged By TGFβ1 in Neonatal and Adult Rat and Mouse Microglia
Primary neonatal rat and mouse microglia express Kcnj2 mRNA and Kir2.1 protein, and our recent studies strongly suggest that homomeric Kir2.1 channels produce the inward-rectifying K + current in both species (Lam and Schlichter, 2015;Lam et al., 2017). In unstimulated microglia, there was a species difference with 4.5-fold higher Kcnj2 mRNA expression in rat; i.e., mRNA counts/200 ng RNA were 1913 ± 657 in rat and 422 ± 90 in mouse (p < 0.001). TGFβ1 treatment did not significantly alter Kcnj2 mRNA expression in either species (Figure 5A).

Kir Current
Inward currents were recorded in response to 500 ms-long voltage steps between −170 mV and −70 mV in 10 mV increments from a holding potential of −20 mV. A Kir current was detected in every rat and mouse microglial cell examined with and without TGFβ1 treatment. As previously described (Lam and Schlichter, 2015;Lam et al., 2017), microglia Kir currents display the stereotypical features of Kir2.1, with rapid activation at negative potentials, and current relaxation at very negative potentials (Figure 5B). The blocker, ML133 was used to quantify the Kir2.1 component of the whole-cell current, as before (Lam and Schlichter, 2015;Lam et al., 2017). In both species, almost all of the Kir current was blocked by 20 µM ML133 in both control and TGFβ1 conditions (86% and 91% in rat, 95% and 96% in mouse, respectively; Figure 5C); therefore, we subsequently 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 (Figure 5D), which is close to the K + Nernst potential in these solutions (−84 mV). In control microglia, there was a similar magnitude of Kir2.1 current in both species, and TGFβ1 treatment had no effect (Figures 5D,E).
Results from adult rat and mouse microglia are shown in Figure 6. In both species, Kir-like currents were always observed with or without TGFβ1 treatment (n > 70 cells), and almost all the current was contributed by Kir2.1 (i.e., blocked by Ba 2+ ; Figure 6A). Similar to neonatal cells, TGFβ1 did not change the Kir2.1 current (Figures 6B,C).
KCa3.1 is the only Ca 2+ -activated K + current we have detected in primary rat microglia (Khanna et al., 2001;Ferreira and Schlichter, 2013;Ferreira et al., 2014Ferreira et al., , 2015Wong and Schlichter, 2014). To activate KCa3.1 currents during whole-cell recordings, we used a positive-gating modulator, either NS309 or SKA-111 (Coleman et al., 2014;reviewed in Balut et al., 2012). First, total inward and outward K + currents were measured from a holding potential of −90 mV, by stepping to +30 mV followed by a ramp from −120 mV to +60 mV ( Figure 7B). Then, TRAM-34 was added to the bath, and the KCa3.1 component was isolated by subtracting the unblocked current (almost entirely Kv at depolarized voltages). Figure 7C illustrates an example of the time course of experiments. Figure 7D shows that SKA-111 activated a very small KCa3.1 current in control rat microglia (1.2 ± 0.8 pA/pF); whereas, after TGFβ1 treatment, a large KCa3.1 current was activated in 12/19 cells (33.6 ± 1.5 pA/pF for responders; p < 0.0001). In a pilot study on rat microglia (not illustrated), 500 nM NS309 failed to activate a current (3/3 cells) but after TGFβ1 treatment there was a robust current in 6/8 cells (33.6 ± 5.3 pA/pF; p < 0.01). Surprisingly, the opposite was seen for mouse cells. SKA-111 activated a large KCa3.1 current in control mouse microglia (17.9 ± 2.4 pA/pF) but after TGFβ1 treatment, almost no current was evoked (0.8 ± 0.2 pA/pF; p < 0.0001).

Comparing Microglial Responses to TGFβ1 and IL-10
As anti-inflammatory or resolving cytokines, TGFβ1 and IL-10 are often assumed to have similar effects (see ''Discussion'' section). In Figure 8, we compare transcriptional responses of rat and mouse microglia to TGFβ1 (present study) with responses to IL-10 from our recent study (Lam et al., 2017) and some additional genes (Ccl3, Ccr5, Csf1r, Itgb2, Kcnn3, Kcnn4, Orai3).
[Note: expression of 33/80 genes was unchanged; all the data are shown in Supplementary Table S4]. For ease of comparison, the bubble chart visually highlights both the magnitude and direction of changes. There were some intriguing differences in TGFβ1 and IL-10 responses, and we will highlight some examples under each category.

Microglial Markers and Immune Modulators
In both species, TGFβ1 increased Cx3cr1 expression; whereas, IL-10 decreased it. TGFβ1 decreased Socs1 only in rat. IL-10 increased Socs3 only in mouse. IL-10 increased Tlr4 in both species but TGFβ1 did not; indeed, Tlr4 decreased in mouse. Both cytokines increased Itgam in rat; only IL-10 increased it in mouse. TGFβ1 increased Csf1r in rat only; IL-10 did not affect either species. IL-10 increased Cd68 and Tspo in rat only. In mouse, TGFβ1 decreased Nr3c1 but IL-10 increased it.

Anti-inflammatory Genes and Receptors
In both species, TGFβ1 decreased Pparg and increased Tgfbr1; both were unaffected by IL-10. TGFβ1 decreased Mrc1 in both species; IL-10 increased it in mouse. TGFβ1 increased Tgfb1 and Myc in rat but decreased them in mouse. However, TGFβ1 decreased Il1rn in rat. In mouse only, TGFβ1 increased Il4 and Il10 and both cytokines increased Il4r and Il13ra1. Both cytokines increased Trem2 and Tgfbr2 in rat; whereas, IL-10 only elevated Tgfbr2 in mouse. IL-10 increased Il10rb only in rat.

Pro-inflammatory Genes and Receptors
In both species, TGFβ1 increased Trem1 but decreased Ccl3 and Ccr5. TGFβ1 decreased Tnfa and Tnfrsf1b only in rat and Casp1 only in mouse. Ifngr1 was oppositely affected by TGFβ1; increased in rat, decreased in mouse. IL-10 increased Il1b in mouse, while TGFβ1 decreased it in rat.

Additional Genes Related to Microglial Physiology
In both species, both cytokines increased Fcgr2b, and only TGFβ1 increased Ncf1 and Nox1. In mouse, there were no other effects of TGFβ1; however, in rat, P2rx7, Itgb2 and P2ry12 were increased and Fcgr3a was decreased. In contrast, IL-10 increased Fcgr3a in both species. IL-10 also elevated P2ry2 in rat, and P2ry12 and Msr1 in mouse.

Ion Channels and Regulators
TGFβ1 increased Kcna3, Kcna5 and Rest in both species; Kcnn4, Orai3 and Trpm7 in rat; and Kcnn3 in mouse. The only effect of FIGURE 6 | In adult rat and mouse microglia, TGFβ1 had no effect on the Kir2.1 current. Treatments and recording conditions were as in Figure 5. (A) Scatterplot of individual cells showing the proportion of the peak current (at −140 mV) that was blocked by 100 µM Ba 2+ . (B) Peak Kir2.1 current density (pA/pF) as a function of voltage (mean ± SEM for the number of cells indicated). (C) Summary of peak inward current density (pA/pF) measured at −140 mV. There were no significant TGFβ1 effects at the p < 0.05 confidence level.
IL-10 was reduced Kcnj2 in mouse. For the K + currents, there were some notable differences. In both species, the Kv1.3 current was unaffected by IL-10 (Lam et al., 2017) but increased by TGFβ1 in both neonatal and adult microglia (present study). Here, TGFβ1 did not affect the Kir2.1 (inward-rectifier) current in neonatal or adult microglia of either species; whereas, it was decreased by IL-10 in neonatal mouse microglia (Lam et al., 2017; KCa currents were not assessed in our recent study of IL-10).

DISCUSSION
We will summarize the present findings and relevant literature on rat and mouse under several topics. The emphasis will be on studies using primary rodent microglia because cell lines (e.g., BV-2 cells) differ in transcriptional responses (Butovsky et al., 2014;Das et al., 2016) and ion channel expression (Cayabyab et al., 2000;Ferreira and Schlichter, 2013;Liu et al., 2013).

TGFβ1 in the Brain
TGFβ1 is a pleiotropic cytokine with immunosuppressive and anti-inflammatory properties (Colton, 2009;Cherry et al., 2014;Orihuela et al., 2016). TGFβ1 levels are often increased with brain damage or disease (see ''Introduction'' section), and it can act on microglia and other cells. We previously found that TGFβ1 expression is up-regulated in the damaged rat striatum in the first week after ICH or ischemia Lively et al., 2016). It was also increased by intracerebral  injection of IL-4 in vivo  or by adding IL-4 to rat microglia in vitro (Liu et al., 2013). A recent study of ICH in mice showed an increase in TGFβ1 in vivo (days 1-10), and an important role in reducing microglial inflammatory cytokine and chemokine expression in vitro (Taylor et al., 2017). In that study, microglia assumed a pro-inflammatory phenotype in vivo (expressing Tnf, Il6, Ccl2, Cxcl10) up to 3 days after ICH, and then during the recovery period (days 7-14), they up-regulated Tgfb1, Tgfbr1 and Tgfbr2. Here, we found that 7 days after ICH in the rat, there was strong expression of the TGFβ1 receptor in microglia/activated macrophages within the hematoma.

Effects of TGFβ1 on Microglia in Vitro
In vitro, microglia can be skewed to multiple anti-inflammatory phenotypes, which are usually assessed by molecular profiles and functional outcomes (reviewed in Colton, 2009;Cherry et al., 2014;Orihuela et al., 2016). Such states are increasingly defined by the stimulus (e.g., IL-4, IL-10, TGFβ1, glucocorticoids); however, it is not clear to what degree microglial responses to these stimuli differ and whether there are species differences.

Effects of TGFβ1 on Microglial Health
TGFβ1 is required for production of microglia in the mouse (Butovsky et al., 2014). Only a small number of in vitro studies have addressed whether TGFβ1 affects microglial health and survival. The results vary, but whether this reflects differences in protocols or species is not known. Stimulating primary mouse microglia with a wide range of TGFβ1 concentrations (0.001-1 ng/ml) did not affect proliferation 3 days later; but if proliferation was first stimulated by a mitogen (conditioned medium or granulocyte-macrophage colony-stimulating factor; GM-CSF), it was inhibited by as little as 0.01 ng/ml TGFβ1 (Suzumura et al., 1993). TGFβ1 inhibited proliferation when it was added to mouse microglia grown on a rat astrocyte bed (Jones et al., 1998). We found that 24 h treatment with 5, 10 or 20 ng/ml TGFβ1 did not affect primary mouse microglia proliferation (in agreement with Suzumura et al., 1993) but reduced proliferation of rat microglia by 20%-25%. We did not observe microglial apoptosis in either species at 24 h. In the only relevant study that we found of primary rat microglia, 10 ng/ml hTGFβ1 increased apoptosis at 48 h (Xiao et al., 1997); the 24 h time point was not mentioned.

TGFβ1-Induced Inflammatory Profile
Based mainly on mouse studies, it is thought that TGFβ1 can maintain a resting microglial state and can also contribute to alternative activation. IL-4-mediated up-regulation of Ym1 and Arg1 mRNA depended on TGFβ1 autocrine actions on Ras/MEK/ERK signaling (Zhou et al., 2012) and, conversely, inhibition of TGFβ1 signaling decreased Ym1 and Arg1 protein levels and up-regulated Il6 and iNOS mRNA (Spittau et al., 2013). Broader transcriptional profiling of responses to TGFβ1 has recently begun for mouse microglia. An extensive microarray study applied TGFβ1 with CSF-1 (48 h), characterized the resulting state, which they called M0 because it was similar to freshly isolated adult microglia (Butovsky et al., 2014), and observed increases in 60 microglial genes, including Tgfbr1, Tgfbr2, Kcnd1 (Kv4.1) and P2ry12. There were very few genes in common with the present study but qualitatively similar results following TGFβ1 treatment were seen for Cx3cr1, Myc, Tgfbr1 and Ccr5 in mouse microglia. Applying the blood product, thrombin, to cultured mouse microglia to model ICH, up-regulated Il6, Tnf and Ccl2 and reduced Tgfbr1, and these changes were reversed by adding TGFβ1 (Taylor et al., 2017). Numerous in vitro studies have shown that TGFβ can interfere with pro-inflammatory signaling pathways induced by lipopolysaccharide (LPS; Kim et al., 2004;Le et al., 2004;Herrera-Molina and von Bernhardi, 2005;Naiki et al., 2005), and it inhibited LPS-, IFNγ-and GM-CSF-induced expression of major histocompatibility complex class II molecules (Loughlin et al., 1993;Suzumura et al., 1993;Lee et al., 1997;Hailer et al., 1998;O'Keefe et al., 1999). Such studies have led to TGFβ1 being proposed as a therapeutic strategy for treating CNS disease (reviewed in Cherry et al., 2014); however, few studies have compared microglial responses in rat and mouse (Patrizio and Levi, 1994;Du et al., 2017;Lam et al., 2017).
Differences in responsiveness to TGFβ1 and IL-10 were not accompanied by a deficiency or down-regulation of any of the cognate receptors in either species. Receptor transcript levels (Il10ra, Il10rb, Tgfbr1, Tgfbr2) were moderately high in unstimulated rat and mouse microglia, and not reduced by either of the cytokine treatments (Lam et al., 2017;present study). The finding that TGFβ1 treatment increased Tgfbr1 and Tgfbr2 expression suggests an increased capacity for autocrine anti-inflammatory signaling, which was not the case for IL-10 treatment. Instead, IL-10 increased expression of some pro-inflammatory receptors (Tlr4, Tnfrsf1a) as well as anti-inflammatory receptors (Il4r, Il13ra1, Tgfbr2), suggesting an overall increased ability to respond to subsequent inflammatory stimuli.

Potassium Channel Expression and Function Is Affected By TGFβ1
Numerous microglial functions depend on Ca 2+ signaling through CRAC (produced by Orai + Stim subunits) and SOC channels (Möller, 2002;Färber and Kettenmann, 2006), signaling that can be evoked by metabotropic purinergic (P 2 Y) receptors. K + channels affect the membrane potential (Chung et al., 1998(Chung et al., , 1999Franchini et al., 2004;Newell and Schlichter, 2005) and regulate numerous microglia functions (Kettenmann et al., 2011;Stebbing et al., 2015). Here, TGFβ1 treatment did not affect the CRAC channel pore-forming subunit, Orai1, the CRAC regulator, Stim1, or P2ry2 in either species. Orai3 and P2ry12 were up-regulated by TGFβ1 only in rat microglia. However, there were important effects on the K + channels, each of which will be discussed separately with an emphasis on literature from primary rodent microglia. Even within studies on primary rodent microglia, there are considerable discrepancies in which K + channel types are expressed, which currents are active, and what changes occur with different activation states. Recent patch-clamp studies of microglia in pro-and anti-inflammatory states have examined Kir2.1 in neonatal rat microglia (Lam and Schlichter, 2015), Kv1.3, Kir2.1 and KCa3.1 in neonatal mouse microglia (Nguyen et al., 2017a,b), and have compared Kv1.3 and Kir2.1 between neonatal mouse and rat microglia (Lam et al., 2017). From these and earlier studies, it is becoming clear that there are substantial qualitative and quantitative differences between the two species.

Kv1.3 Channels
An early hypothesis was that Kv current is induced by ''activation'' (Nörenberg et al., 1994;Fischer et al., 1995;Visentin et al., 1995;Illes et al., 1996) but the diversity of microglial states was not recognized nor were potential species differences. Kv currents can be up-regulated by pro-inflammatory stimuli (LPS, IFNγ, IFNγ+TNFα) in rat (Nörenberg et al., 1992(Nörenberg et al., , 1994Visentin et al., 1995;Fordyce et al., 2005;Lam et al., 2017) and mouse microglia Prinz et al., 1999;Pannasch et al., 2006;Lam et al., 2017;Nguyen et al., 2017a). More than one Kv channel type has been detected in rodent microglia; e.g., Kv1.3 and Kv1.5 currents in ex vivo rat microglia (Kotecha and Schlichter, 1999) and LPS-stimulated mouse microglia (Pannasch et al., 2006). Thus, it is important to use selective blockers to identify Kv1.3. Using such blockers, we almost always detect robust Kv1.3 currents in unstimulated rat microglia (Schlichter et al., 1996;Cayabyab et al., 2000;Fordyce et al., 2005;Newell and Schlichter, 2005;Lam et al., 2017;present study). In unstimulated neonatal mouse microglia, we detected Kv1.3 currents less often, and they were small and accompanied by an additional Kv current (Lam et al., 2017;present study) that might be Kv1.5. In contrast, adult mouse microglia always had a Kv current, and it was predominantly Kv1.3, suggesting a developmental change. Of note, anti-inflammatory stimuli can also increase Kv1.3 expression and current in microglia; i.e., in response to IL-4 in rat (Lam et al., 2017), and TGFβ1 in neonatal mouse (Schilling et al., 2000) and neonatal and adult rat and mouse microglia (present study). Thus, Kv1.3 current is not specific to either a pro-or anti-inflammatory state and, instead, its up-regulation appears to signal a change from the resting, surveillance state. This might not be surprising, given that Kv1.3 is involved in numerous microglial functions that might occur in pro-or anti-inflammatory milieu, including proliferation (Kotecha and Schlichter, 1999), chemotactic migration (Nutile-McMenemy et al., 2007), chemokine and cytokine production (Liu et al., 2012;Charolidi et al., 2015), respiratory burst (Khanna et al., 2001), and even, neurotoxicity (Fordyce et al., 2005). The observed expression and roles need to be considered when targeting this channel in vivo to affect CNS inflammation.

Kir2.1 Channels
From the earliest recordings onward, a Kir2.1-like inwardrectifying current has usually been seen in unstimulated rat (Kettenmann et al., 1990;Korotzer and Cotman, 1992;Visentin et al., 1995;Schlichter et al., 1996) and mouse cells Prinz et al., 1999). But there are inconsistencies concerning its prevalence and amplitude in stimulated cells, and it is possible that they reflect heterogeneity in culture, stimuli or activation states. In response to LPS, IFNγ or IFNγ+TNFα, mouse microglia had smaller Kir currents Prinz et al., 1999;Boucsein et al., 2003;Lam et al., 2017) but in a recent study, LPS did not affect the current (Nguyen et al., 2017b). Rat cells varied from no change to a small increase with LPS, a large increase with IFNγ (Nörenberg et al., 1992(Nörenberg et al., , 1994Visentin et al., 1995;Draheim et al., 1999), and a decrease with IFNγ+TNFα (Lam et al., 2017). The contrast with earlier studies using IFNγ might relate to our additional use of TNFα. While less is known about Kir2.1 responses to anti-inflammatory stimuli, the results also vary. In primary mouse microglia, IL-4-treated cells had large currents (Lam et al., 2017;Nguyen et al., 2017b), and the current was reduced by IL-10 (Lam et al., 2017) but not affected by TGFβ1 (Schilling et al., 2000;present study). In primary rat microglia, Kir2.1 was reduced by IL-4 (Lam et al., 2017) but unaffected by TGFβ1 in either neonatal or adult cells (present study). Roles of Kir2.1 are not well understood, possibly because the earlier blockers were not sufficiently potent or selective. In rat microglia, there is evidence that Kir2.1 regulates proliferation (Schlichter et al., 1996), and using a better blocker (ML133) we found that it regulates CRAC-mediated Ca 2+ entry and ATP-induced chemotaxis (Lam and Schlichter, 2015). The observed decreases in Kir2.1 current by some pro-and anti-inflammatory stimuli, and possible species differences might make it difficult to target this channel for controlling microglial functions.

KCa3.1 Channels
KCa3.1 is considered a therapeutic target in the CNS because selective KCa3.1 blockers have improved the outcome in animal models of trauma, spinal cord injury, stroke and neurodegeneration (reviewed in Maezawa et al., 2012;Feske et al., 2015;Chen et al., 2016;Dale et al., 2016). Although Kcnn4 transcripts have been reported in primary microglia both from rats (Khanna et al., 2001;Kaushal et al., 2007) and mice (Bouhy et al., 2011), the presence of a KCa3.1 current appears to be quite variable and dependent on the activation state. In rat microglia, small TRAM-34-sensitive KCa3.1 currents were recorded in a minority of unstimulated cells but transcripts and current were dramatically up-regulated by IL-4  or TGFβ1 (present study). IL-4 up-regulated KCa3.1 through Ras/MEK/ERK signaling, and the channel then contributed to migration, invasion and Ca 2+ signaling . However, KCa3.1 also regulates LPS-mediated activation (through p38 MAPK), leading to production of pro-inflammatory mediators and neuron killing (Kaushal et al., 2007;Liu et al., 2013), and UTP-stimulated chemotaxis and expression of the pro-inflammatory genes, IL-1β and iNOS . In mouse microglia, KCa3.1 currents are also variable, and apparently susceptible to culturing conditions and activation states but effects of TGFβ1 were not studied. In early studies, moderate charybdotoxin-sensitive KCa currents were seen when microglia from NMRI mice were cultured with M-CSF and astrocyteconditioned medium (Eder et al., 1997;Schilling et al., 2002). Later studies using the more selective blocker, TRAM-34, showed that unstimulated cells had small (Maezawa et al., 2011;Nguyen et al., 2017b) to moderate KCa3.1 currents (present study). Intriguingly, the current was unaffected by IL-4 or LPS alone, and moderately increased by IFNγ but not by IFNγ + LPS (Nguyen et al., 2017b), while we found that it was greatly decreased by TGFβ1 (present study). Variable KCa3.1 expression was seen in ex vivo tissue from damage models. For instance, no KCa3.1 current was seen in juvenile mouse microglia within acute brain slices (Schilling and Eder, 2007); whereas in adult microglia isolated using CD11bcoated magnetic beads, there were KCa3.1 currents in some microglia at 8 days after ischemia, and in most microglia isolated from LPS-injected mouse brains (Chen et al., 2016). In adult human microglia from epilepsy patients, most cells expressed a KCa3.1 current, which was unaffected by LPS and slightly increased by IL-4 (Blomster et al., 2016). Thus, a clear picture of KCa3.1 expression in microglia has not emerged.

CONCLUSIONS AND FUTURE DIRECTIONS
Studies of disease mechanisms rely heavily on rodent models and translation to humans is a continuing challenge. Transcriptional profiling is increasingly used to investigate microglia in CNS damage and disease states, and an important question is whether differences reported in microglia responses in vivo and in vitro instead reflect species differences. Our in vitro study of >50 targeted molecules showed significant species similarities as well as important differences in microglial responses to TGFβ1. Some genes can be compared with recent high-content studies analyzing microglia isolated from both naive and damaged or diseased brains. Most of those studies were on mice, with some exemplary molecules validated in human brain tissue. Molecular signatures include homeostatic genes expressed in normal microglia (Gautier et al., 2012;Orre et al., 2014;Zhang et al., 2014), a microglial ''sensome'' in adult and aged mice (Hickman et al., 2013), and a unique plaque-associated phagocytic subtype in Alzheimer's (''Disease-associated microglia''/DAM; Keren-Shaul et al., 2017;Krasemann et al., 2017). Several molecules that we found were well expressed in control microglia and up-regulated by TGFβ1 are in a homeostatic fingerprint that is maintained by TGFβ1 signaling (Butovsky et al., 2014;Gosselin et al., 2014); e.g., Csf1r, Cx3cr1, P2ry12. Several microglial signature molecules were also restricted to microglia in humans; i.e., C1qa, Gas6, Mertk, P2ry12, and Pros1 (Butovsky et al., 2014). Some molecules we examined (including P2ry12, Csf1r, Cx3cr1, Itgam (CD11b), Cd68, Tlr2, Itgb2, and various Fc receptors) are part of the ''sensome'', which is a large group of molecules and receptors at the cell surface involved in microglia sensing environmental perturbations, as in damage or disease (Hickman et al., 2013). Some sensome-related molecules that were increased by TGFβ1 in our study are commonly down-regulated in AD-associated microglia and DAM (Keren-Shaul et al., 2017;Krasemann et al., 2017); e.g., Csf1r, Cx3cr1, P2ry12, Tgfbr1.
Goals of molecular fingerprinting studies include: selecting and interpreting markers and molecules in animal models of CNS disease; identifying a reliable panel of human markers to help define disease progress and determine whether treatments are effective; and identifying molecular targets to control neuroinflammation. To this end, it is essential to compare human microglia with rodent models; however, there is very little published data comparing human responses to TGFβ1. A recent article raises concerns because a hallmark TGFβ1 response in rodents was not seen in human microglia (Smith et al., 2013). The authors investigated human leukocyte antigen (HLA) molecules, which are human analogs of the MHC II complex. IFNγ increases HLA expression and TGFβ1 blocks this response in rodent microglia; however, in adult human microglia, TGFβ1 (10 ng/ml) was completely ineffective. In future, it will be crucial to compare a wide range of TGFβ1 responses between human and rodent microglia.

DATA AVAILABILITY
The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

AUTHOR CONTRIBUTIONS
The project was mainly conceived and spearheaded by LCS, with important contributions from SL, DL and RW. SL carried out the analyses of ICH tissue, Nanostring, Western blotting and functional studies (morphology, proliferation, apoptosis; Tables 1-4, Figures 1, 2, 8, Supplementary Tables S3, S4). DL and RW conducted all the patch-clamp electrophysiology and analysis (Figures 3-7). LCS and SL wrote the manuscript, with input from DL and RW.

FUNDING
This work was supported by operating grants to LCS from the Canadian Institutes of Health Research (CIHR; MOP 119578), the Canadian Stroke Network (CSN 1007005), and by the Toronto General/Toronto Western Hospital Foundation.