Microglia experience dramatic molecular and functional changes when transferred from the central nervous system (CNS) to a cell culture environment. Investigators largely attribute these findings to the loss of CNS-specific microenvironmental cues that dictate the gene-regulatory networks specified by master regulator transcription factors such as V-maf musculoaponeurotic fibrosarcoma oncogene homolog B (MafB). MafB regulates macrophage differentiation and activation by activating or repressing target genes critical to these processes. Here, we show that basal MafB levels in the BV-2 microglial cell line depend on the availability of lipids in the cell culture environment. Depletion of lipids, either by serum deprivation or the use of lipid-depleted serum, reduced MafB protein levels in BV-2 cells. Using live imaging, we also observed the engulfment of apoptotic BV-2 cell debris by neighboring BV-2 cells, highlighting an additional potential source of lipids in the cell culture environment. This observation was supported by experiments showing reduced MafB protein levels in BV-2 cells cultured with various phagocytosis inhibitors (cytochalasin D, annexin V) and reduced BV-2 cell phagocytic activity with serum deprivation. In aggregate, our data suggest that serum exposure regulates the transcription factor MafB in BV-2 cells through direct and indirect mechanisms.

As the primary tissue macrophage population of the central nervous system (CNS), microglia perform critical functions throughout life by integrating ontogenetic and environmental influences^[2]. This integration occurs in a hierarchical and cooperative fashion such that lineage-dependent transcription factors (LDTFs; e.g., MafB) collaborate to select the genomic elements with which signal-dependent transcription factors (e.g., Nuclear Factor kappa-light-chain-enhancer of activated B cells [NF-kB]) can interact^[2]. Microglia lose their in vivo identity when transferred to a cell culture environment, an effect that has been attributed largely to the loss of instructive signals from the CNS and exposure to factors like serum that are absent in the healthy CNS^[3][1][4]. The relative contribution of the latter over the former has been complicated by the lack of defined medium formulations for comparisons. Because serum enhances the phagocytic and proliferative capacity of microglia in vitro, some of the molecular and functional changes that microglia undergo could be direct and indirect effects of serum exposure. Commonly used microglial cell lines (e.g., BV-2 cells) display a similar profile to primary microglia^[3], making it plausible that some of their phenotypic characteristics are also caused by serum exposure.

The LDTF MafB contributes to the core transcriptional identity of tissue macrophages in general^[5] and to the specification of the mature, homeostatic microglial phenotype in particular^[6]. MafB is also a critical determinant of macrophage identity, regulating many aspects of macrophage physiology by activating or repressing target genes related to these functions. One such function is phagocytosis, the process by which macrophages engulf cellular debris and pathogens, among other prey. MafB deficiency in macrophages impairs their ability to phagocytose apoptotic debris (i.e. efferocytosis) via reduced expression of the MafB target genes C1qa^[7] and Axl^[8]; damage-associated molecular patterns (DAMPs) via reduced expression genes Msr1 and Marco^[9]; and Fc-opsonized targets via reduced expression of Fcgr3^[10]. In addition, MafB participates in the formation of membrane protrusions by macrophages^[11][12] and is upregulated in peripheral macrophages exposed to apoptotic cells. Because microglia exhibit altered phagocytic and morphological profiles when transferred to serum-containing cultures^[1], MafB could contribute to the functional and transcriptomic consequences of introducing microglia to the cell culture environment.

The objective of the present study was to determine the effect of the cell culture environment on the transcription factor MafB in the BV-2 microglial cell line.

BV-2 cells are routinely cultured in a serum-containing medium. Given the roles that MafB plays during macrophage phagocytosis and actin organization, we characterized the effect of serum on MafB protein levels in BV-2 cell cultures. Culturing BV-2 cells in reduced (0–5%) serum for 24 h decreased MafB protein levels to less than half of those cultured in 10% serum, respectively (0.44 and 0.42–fold, respectively, vs 10% FBS; p-values <0.0001; Suppl. Fig. 1A). Ionized calcium-binding adaptor molecule 1 (Iba1), a key player in phagocytic cup formation and membrane ruffling in microglia^[13][14], was decreased as well in 5% serum cultures (0.70–fold vs 10% FBS; p = 0.044), but not serum-free cultures (0.83–fold vs 10% FBS; p = 0.156), relative to 10% serum cultures (Suppl. Fig. 1A, full blots shown in Suppl. Fig. 2). Because 5% of serum cultures showed significantly lower Iba1 levels compared to 10% serum cultures, we next asked whether the phagocytosis of phosphatidylserine-coated (PS) beads would be impaired in reduced serum cultures. To this end, we cultured BV-2 cells in 10% or 5% serum for 18 h before adding 3 μm diameter PS beads for another 6 h. We immunostained these cultures with the lysosomal marker Cluster of Differentiation 68 (CD68) and Iba1 to visualize particle engulfment (Suppl. Fig. 1B). For quantification, we measured the proportion of CD68 co-localized with fluorescent PS beads to the total volume of Iba1 using pseudo-confocal microscopy. BV-2 cells cultured in 5% serum showed reduced phagocytosis of fluorescent PS beads compared to those cultured in 10% serum (engulfment index 0.21 [5% serum] vs 0.27 [10% serum]; p = 0.0025, Suppl. Fig. 1C). As a negative control, we pretreated BV-2 cells with the actin polymerization inhibitor cytochalasin D to prevent phagocytosis. Although cytochalasin D-treated BV-2 cells were able to bind fluorescent PS beads (Suppl. Fig. 1B, Suppl. Fig. 3), BV-2 engulfment of them was dramatically impaired when compared to untreated 10% serum cultures (engulfment index 0.08; p <0.001) (Suppl. Fig. 1C). The lack of complete inhibition by cytochalasin D might reflect CD68’s ability to bind phosphatidylserine coating the bead surface^[15], which could account for the apposition of these signals (see Suppl. Fig. 3), particularly given the resolution limits of pseudo-confocal microscopy.

Pure, serum-treated primary microglia cultures are largely devoid of apoptotic cells because neighboring microglia rapidly clear them^[1]. Also, peripheral macrophages upregulate Mafb when exposed to apoptotic cells^[7]. However, it is unclear whether BV-2 cells phagocytose one another in pure cultures and, if so, whether this occurs frequently enough to elicit measurable changes in MafB at the population level. Having observed decreased MafB levels in BV-2 cells cultured in reduced serum conditions (Suppl. Fig. 1A), we next asked whether this effect could be explained by the phagocytosis of neighboring and perhaps apoptotic BV-2 cells or debris. To test this, we compared three phagocytosis inhibition strategies—reduced serum (Suppl. Fig. 1C), the actin polymerization inhibitor cytochalasin D, and the phosphatidylserine-binding protein Annexin V—for their ability to modulate MafB protein levels in BV-2 cultures. As expected, 5% serum significantly reduced MafB protein levels relative to 10% serum (0.72–fold vs 10% FBS; p = 0.042, Suppl. Fig. 1D), albeit to a lesser extent than we observed in supplementary figure 1A. This inconsequential discrepancy is likely due to the technical differences between these experiments (see Suppl. Experimental Methods) and the semi-quantitative nature of western blotting. Cytochalasin D treatment in the presence of 10% serum dramatically reduced MafB protein levels (0.36–fold vs 10% FBS; p = 0.0004; Suppl. Fig. 1D; full blots shown in Suppl. Fig. 4) versus untreated 10% serum cultures. Finally, Annexin V treatment in the presence of 10% serum mimicked the result of reduced serum by decreasing MafB relative to untreated 10% serum cultures (0.71–fold vs 10% FBS; p = 0.042; Suppl. Fig. 1D). Together, these results suggest that MafB protein levels in BV-2 cells could depend on the phagocytosis of neighboring, perhaps apoptotic BV-2 cells.

To further explore the possibility that BV-2 cells phagocytose neighboring BV-2 cells under standard cell culture conditions (see Methods), we visualized BV-2 cell apoptosis using a cell-permeable, activated caspase-3/7 dye. We observed activated caspase-3/7-positive BV-2 cells in all imaged cell culture wells. Live BV-2 cells frequently surrounded activated caspase-3/7-positive (called “apoptotic” hereafter; see Limitations for a caveat to this interpretation) BV-2 cells (Suppl. Videos 1–3). In such cases, activated caspase-3/7 puncta appeared to radiate from apoptotic cells into surrounding live BV-2 cells (Suppl. Videos 1 & 2). In one case, activated caspase-3/7 debris was engulfed by a distant BV-2 cell (Fig. 1A, Suppl. Video 1). We also observed the disappearance of engulfed, activated caspase-3/7 puncta in live BV-2 cells (Fig. 1B, Suppl. Video 1). These observations provide evidence of BV-2 cells’ phagocytosing one another in pure cultures. To determine the robustness of our initial phagocytosis inhibitor experiment (Suppl. Fig. 1D), we replicated the experiment (without the serum deprivation condition) in 3 independent experiments. These experiments generated similar results (Fig. 1C, Suppl. Fig. 5). Nevertheless, the interpretation that basal MafB protein depends on BV-2 cells’ phagocytosing one another remains speculative because it is unclear whether the engulfment of BV-2 cell debris we observed is sufficient to drive changes in MafB protein at the population level. Instead, MafB levels could depend on other factors present in the cell culture environment.

How might MafB be regulated by the phagocytosis of cells and cellular debris? This effect could be mediated by the activation of liver X receptors (LXR) by hydrolyzed cholesterol obtained from engulfed apoptotic bodies. After all, engulfed apoptotic bodies supply macrophages with a substantial cholesterol load^[16][17], and MafB is known to be transcriptionally induced by LXR agonists^[18]. This putative mechanism raises the possibility that other lipid sources in the cell culture environment—namely, serum, which is an abundant source of lipids and lipoproteins^[19]—could contribute to basal MafB protein levels in BV-2 cell cultures. To test this hypothesis, we compared BV-2 cells cultured in normal growth medium with those cultured in medium containing lipid-depleted (LD) serum. BV-2 cells exhibited lower MafB protein when cultured in medium containing LD serum (0.66-fold vs normal FBS, p = 0.0475; Fig. 1D). This effect was prevented by the addition of human low-density lipoproteins (LDL), another rich source of cholesterol^[20] (p = 0.0138; Fig. 1D). We replicated this result in three independent experiments (Suppl. Fig. 6). In aggregate, our results indicate that basal MafB levels in the BV-2 microglial cell line depending on the availability of lipids in the cell culture environment—either obtained directly from serum or via phagocytosis of apoptotic BV-2 cell debris.

When transferred to a cell culture environment, microglia rapidly downregulate transcripts that typify mature, homeostatic microglia in vivo. Such transcripts—e.g., P2ry12, Tmem119, and Sparc—exhibit a remarkably short half-life of <1 h in vitro^[1], consistent with previous estimates that show downregulation of thousands of genes within 6 h of transfer to culture^[4]. Microglia also transiently upregulate many inflammatory transcripts when transferred in vitro^[1][4]. However, with longer exposure (1–5 days) to serum-supplemented or defined media, microglia assume an alternative gene signature characterized by increased expression of Mafb (Fig. 1E) and phagocytosis-related genes (Suppl. Fig. 7)^[1]. These include known/putative MafB target genes C1qa^[7], Fcgr3^[10], Cd5l^[21], Msr1^[9], and Rab13^[22], as well as the lysosomal marker Cd68. We speculate that the upregulation of Mafb and putative MafB target genes by serum in this culture system may be a consequence of increased lipid content in the cell culture environment.

Basal MafB protein levels in BV-2 microglia are regulated by the lipid content of the cell culture environment.

Generated from primary microglia cultures infected with the J2 retrovirus, BV-2 cells express the v-raf and v-myc oncogenes^[23] and possess a hypertriploid karyotype^[24]. These problematic features aside, BV-2 cells overlap substantially with primary microglia in their expression of phagocytosis-related genes^[25]. Although the BV-2 cell line represents a convenient, albeit imperfect, an approximation of primary microglia, BV-2 cells almost certainly do not recapitulate accurately the transcriptional and signaling networks that operate in microglia in vivo. Further, our observation may not be generalizable to all microglia because BV-2 cells were derived from primary microglia isolated from female C57Bl/6 mice^[23]. In fact, neonatal microglia of the female sex phagocytose more neural progenitor cells in the hippocampus than their male counterparts^[26]. However, because these sex-specific effects often depend on circulating hormones^[26], it is unclear to what extent they would have been retained in the microglia that were used to generate the BV-2 cell line.

With regard to our bead phagocytosis assay (Suppl. Fig. 1B, C), there are two caveats worth mentioning: (1) our assay does not fully recapitulate the complex “find me” and “eat me” signals that direct target identification and phagocytosis in vivo; and (2) cultured macrophages have been shown to use somewhat different transcriptional machineries to phagocytose beads versus biological targets^[27].

Our caspase-3/7 live imaging assay (Fig. 1A, B; Suppl. Videos 1–3) may have detected non-apoptotic activation of caspase-3. Indeed, non-apoptotic roles for caspase-3 have been identified in macrophages^[28][29]. However, we also observed other signs of apoptosis that accompanied the activated-caspase-3/7 signal, including membrane blebbing and cell shrinkage. We, therefore, contend that most if not all activated caspase-3/7 signals in our live imaging assay labeled apoptosis instead of non-apoptotic processes.

Overall, our observations would be strengthened by the use of more quantitative methods than western blotting and immunocytochemistry (e.g. flow cytometry).

The phagocytosis inhibition strategies we used in this study could modulate MafB through divergent mechanisms or converge on MafB through a non-phagocytic pathway. Rather than preventing the phagocytosis of nearby apoptotic cells or debris, cytochalasin D and annexin V could instead inhibit the uptake of LDL or other lipid sources from the culture medium. In support of this hypothesis, cytochalasin D has been shown to modulate lipid droplet formation and cholesterol flux in macrophages^[30], and annexin V can directly bind negatively charged lipids in LDL particles^[31], perhaps preventing their recognition by LDL receptors. Macrophages have also been shown to shed cholesterol-rich particles as they navigate cell culture dish surfaces^[32], a property that may explain the aforementioned effect of cytochalasin D on cholesterol flux in macrophages. This could explain the potent effect of cytochalasin D in this study, as cytochalasin D both inhibits phagocytosis and cell movement. Nevertheless, these alternative explanations do not contradict our primary conclusion that basal MafB levels in BV-2 cell cultures depend on the lipid content of the cell culture environment.

Although the present study was limited to the cell culture environment, our results may nevertheless offer clues into the in vivo biology of macrophages, including microglia. Whereas BV-2 cells were observed engulfing apoptotic BV-2 cell debris in the present study, we speculate that the engulfment of, say, apoptotic neurons or focally apoptotic synapses by microglia in the CNS would increase MafB protein in these cells in vivo. Indeed, synapses contain high levels of cholesterol, an important structural molecule that supports diverse synaptic functions^[33].

Cell culture

BV-2 cells (RRID:CVCL_0182, sex: female) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM; Corning, cat no. 17-205-CV) supplemented with GlutaMAX (1% v/v; ThermoFisher, cat no. 35050061) and heat-inactivated fetal bovine serum (FBS; 10% v/v; Atlas Biologicals, cat no. F-0500-D, lot no. F31E18D1) in a humified cell culture incubator (37°C, 5% CO2). BV-2 cells were passaged 1:10 every other day and never reached greater than 80% confluency during passaging.

Serum lipid-depletion

Lipids were depleted from FBS using a Cleanascite lipid removal agent (Biotech Support Group, cat no. X2555-10). Cleanascite reagent was thoroughly resuspended before mixing with FBS at a volume ratio of 1:4 (Cleanascite: FBS). The mixture was gently and periodically inverted for 10 min to facilitate lipid binding. The solution was centrifuged at 1,000 x g for 15 min to pellet the removal agent. The resulting supernatant was used for experiments. Control FBS was prepared using the same protocol, except PBS was added instead of the Cleanascite reagent.

Western Blot

BV-2 were plated at a density of 300,000 cells per well in a 12-well plate and incubated overnight in a growth medium containing 10% FBS; 5% FBS; 10% FBS supplemented with 10 μM cytochalasin D (MilliporeSigma, cat no. C8273); 10% FBS supplemented with purified, recombinant Annexin V (BD Bioscience, cat no. 556416; 1 mg/mL final concentration); 10% lipid-depleted FBS; or 10% lipid-depleted FBS supplemented with human LDL (Stemcell, cat no. 02698; 100 mg/mL final concentration). 24 h later, cells were washed once with 1× PBS and lysed with RIPA buffer supplemented with protease inhibitors. Lysates were kept on ice and periodically vortexed for 30 min, after which they were centrifuged at 13,000 rpm for 10 min. Supernatants were collected and mixed with Laemmli sample buffer. Proteins were denatured at 100°C for 10 min and separated on 4–15% SDS-PAGE gel (Bio-rad, cat no. 4561084) at 100 V for 1 h. Proteins were transferred onto nitrocellulose membranes at 100 V for 1 h at 4°C. The resulting membranes with blocked in 5% (w/v) nonfat milk in TBS-Tween (0.05% v/v; TBST) for 1 h at RT or overnight at 4°C. Blots were incubated with primary antibodies overnight at 4°C. Primary antibodies included a recombinant, monoclonal rabbit anti-MafB (Bethyl, 1:2000; cat no. A700-46); β-actin (Santa Cruz, cat no. SC-47778, RRID:AB_626632); rabbit anti-Iba1 (Wako Biochemicals, cat no. 019-19741, RRID:AB_839504), and GAPDH (Millipore cat no. CB1001, RRID:AB_2107426; 1:4000). Blots were subsequently washed with TBST 3× and incubated with secondary antibodies for 1 h at RT. After another 3 washes with TBST, blots were incubated briefly (1-2 min) with Clarity ECL substrate (Bio-rad, cat no. 1705061) and imaged using an Azure C600 imager. When necessary, blots were stripped using ReBlot Plus Strong Antibody Stripping Solution (EMD Millipore, cat no. 2504) and reprobed as described above. Full western blots are shown in supplemental figure 2 (for Suppl. Fig. 1A), supplemental figure 4 (for Suppl. Fig. 1D), supplementary figure 5 (for Fig. 1C), and supplementary figure 6 (for Fig. 1D). Western blots quantifications were performed by normalizing the densitometry of MafB to that of GAPDH or β-actin within each lane. GAPDH was used as an endogenous control for experiments that used cytochalasin D because cytochalasin D itself alters the expression of β-actin (data not shown). The recombinant, monoclonal rabbit anti-MafB antibody was determined by the manufacturer (Bethyl Laboratories) to be specific to human and mouse MafB, but not other large Maf proteins such as MafA and c-Maf. We further validated this antibody in BV-2 cell lines using MafB siRNA knockdown and CRISPR-based activation of MafB, as described in the supplementary experimental procedures and shown in supplementary figure 8. Both bands labeled by the antibody were quantified in all analyses.

Live caspase-3/7 imaging

Activated caspase-3/7 was visualized in live BV-2 cell cultures using the Incucyte Caspase-3/7 Green Reagent. BV-2 cells were plated overnight in 96-well plates at a density of 12K cells per well. The next day, the culture medium was replaced with culture medium supplemented with Incucyte Caspase-3/7 Green Reagent (1:1000 dilution). Live imaging was performed on a Lionheart Fx Scope (Biotek, Winooski, VT) starting 30 min after the addition of the dye. Prior to imaging, the Lionheart Fx Scope was brought to “standard cell culture conditions” (humified, 37°C, 5% CO2). These conditions were maintained throughout live imaging. Images were acquired from the center of each cell culture well every 5 min using a 20x objective. The focal plane was determined using brightfield laser autofocus; there was no offset between the brightfield and fluorescence acquisitions. Time-lapse images were assembled into videos using various stack and annotation functions in ImageJ. Supplementary Videos 1 and 2 represent independent wells in a single 96-well plate acquired in the same imaging session. Supplementary Video 3 represents an independent experimental replication of the entire cell culture and imaging protocol.

Transcriptomic data from published datasets

The transcriptomic data displayed in figure 1E and supplementary figure 7 were derived from raw data published by Bohlen et al.^[1]. These data were obtained by Bohlen et al.^[1] using rat microglia freshly-isolated from P21 rats (Acute), P21 rat microglia cultured for 8 days in TIC medium (TIC), and P21 rat microglia cultured in TIC medium for 3-7 days followed by 1–5 days exposure to 10% fetal calf serum such that all conditions were cultured for 8 days total (FCS 1d, 3d, 5d). For example, the “FCS 3 d” condition was cultured in TIC medium for 5 d followed by 3 d exposure to FCS. TIC medium refers to a DMEM/F12-based medium supplemented with human TGF-β2 (2 ng/mL), murine IL-34 (100 ng/mL), and ovine wool cholesterol (1.5 mg/mL).

Statistics

Normality and homoscedasticity were tested using the Shapiro-Wilk and Brown-Forsythe tests, respectively. All data passed these tests (p-values >0.05) except for the DMSO-treated group presented in figure 1C (Shapiro-Wilk test, p = 0.03). Because the two additional experimental replicates for this particular experiment passed this test (p-values >0.05), we assumed normality and homoscedasticity for all data. Therefore, all data were analyzed by one-way analysis of variance (ANOVA) with Holm-Sidak’s multiple comparison test. All data is represented as mean ± standard deviation.

The study was supported by the NIH RO1 AG057525 and NIH F31 MH113504.

We thank Dr. Sanjay Maggirwar for generously supplying us with the BV-2 cell line. We also thank Kathleen Miller-Rhodes for her helpful comments on the initial draft of this manuscript and her assistance with the live imaging experiment during revision.

The authors declare no conflicts of interest.

Not Applicable.

No fraudulence is committed in performing these experiments or during processing of the data. We understand that in the case of fraudulence, the study can be retracted by ScienceMatters.