Graphene Triggers Inflammation in Murine Microglia via Phagocytosis

Abstract

Graphene and related materials are increasingly used in biomedical technologies, including neural interfaces, but their impact on brain immune cells remains poorly understood. The present study investigates the acute response of murine microglial SIM‐A9 cells to single‐layer graphene, with carbon nanotubes and graphene nanoplatelets included as comparative controls. Short‐term culture for 3 h on graphene‐coated substrates did not induce cytotoxicity but promoted inflammatory activation, reflected in increased release of the cytokine tumor necrosis factor alpha (TNF‐α). Raman spectroscopy revealed partial removal of the graphene layer, indicating phagocytic uptake by microglia. Supporting this mechanism, small carbon nanotubes elicited a similar inflammatory response, whereas larger graphene nanoplatelets, which are less readily internalized, did not. Potential contamination by bacterial endotoxin could be excluded using the lipopolysaccharide (LPS) inhibitor polymyxin B. Comparative proteomic analysis demonstrated that ingestible graphene alters pathways related to inflammation, cytoskeleton organization, and cell proliferation. These findings indicate that graphene can affect microglia through phagocytosis and highlight the importance of preventing delamination from biomedical devices to ensure safe use.

Phagocytosis of single‐layer graphene by microglia triggers inflammatory activation without cytotoxicity. Small carbon nanotubes elicit similar responses, whereas larger graphene nanoplatelets do not. Proteomic analysis reveals effects on inflammation, cytoskeleton, and proliferation, highlighting the importance of graphene stability for safe neural applications.

1. Introduction

Graphene is a two‐dimensional nanomaterial composed of a monolayer of carbon atoms arranged in a honeycomb lattice [1, 2]. According to the EU Graphene Flagship classification, graphene‐based materials (GBMs) include monolayers or few‐layers graphene, graphene oxide, and nano‐ and micro‐platelets [3]. These materials and their derivatives possess distinct chemical, mechanical, and physical properties, enabling diverse applications in biotechnology, food technology [4], and environmental fields such as wastewater treatment [5]. Biomedical applications are also expanding, including incorporation into antibiotics [6], drug screening and delivery systems [2, 7], biosensing [8, 9, 10, 11, 12, 13, 14, 15, 16], and surgical or therapeutic devices [17, 18]. Such uses can lead to passive human exposure through water, air, and food, or active introduction into the body, making it essential to understand the impact of graphene on tissue physiology and pathology.

Cytotoxic effects of graphene have been reported in cells [19] with high environmental exposure, including lung [20, 21], intestinal [22, 23], and liver cells [24, 25]. Peripheral immune cells also respond to graphene, as illustrated by increased cytokine production in monocyte and T‐cell models after short‐term exposure [26]. In contrast, brain immune cells might be seen as largely protected by the blood–brain barrier (BBB) [27], although its permeability may increase with age, disease, or targeted delivery of carbon‐based materials [28, 29].

Graphene is emerging as a transformative candidate for next‐generation neural interfaces or implantable electrodes [30, 31, 32, 33, 34]. Traditional metals and nanomaterials such as Ag, Pt, and Au, although widely used for neural devices, face issues such as poor signal sensitivity [33, 35], oxidative degradation, and potential toxicity [36]. Polymers like poly(3,4‐ethylenedioxythiophene) (PEDOT), though offering better charge transfer and lower impedance [37], suffer from in vivo degradation and adhesion problems [33, 35, 38, 39], while inorganic material such as iridium oxide lacks optical transparency, flexibility, and bio inertness [40, 41].

Graphene, on the other hand, has an exceptional electrical conductivity, optical transparency, and mechanical strength [40]. Its ability to conform to the soft, dynamic neural environment due to the excellent flexibility minimizes neural tissue damage while enabling highly sensitive and accurate neural signal recording [39, 40, 42]. The advancement and significance of graphene in the development of neural devices have been summarized in various review articles [33, 34, 40]. The first clinical trial of graphene‐based electrodes in the human brain has confirmed their superior performance, underscoring their promise in neural sensing, therapy, and drug delivery applications—envisioning a wider usage in the future [42]. Furthermore, GBMs are capable of crossing the BBB and coming in contact with the brain [43, 44]. The passage of reduced graphene oxide (rGO) in the brain of mice through BBB by weakening the paracellular pathway or by suppressing the expression of its tight junction proteins has been already reported [45].

The use of GBMs as neural interface/implants directly in the brain or at the retina, as targeted drug delivery material in central nervous system (CNS) or via introduction from environmental sources via a compromised BBB, can plausibly place microglia in direct or indirect contact with these materials [34, 46] as these cells are the first responders in CNS to external stimuli and against almost all pathological conditions [46, 47]. Direct contact might influence the morphology of the microglia, triggering their activation, and inducing inflammatory responses [46]. Conversely, GBMs may indirectly affect microglia by altering the neural microenvironment. Therefore, it is necessary to understand how these cells react to or tolerate GBMs and their bio‐integrative interfaces.

Previous studies showed that murine BV2 microglia remain viable after exposure to two‐dimensional graphene films or three‐dimensional graphene foams, with foam even exhibiting anti‐inflammatory effects [48]. Graphene oxide, however, enhanced autophagy and phagocytic capacity toward neurotoxic Aβ peptides [49, 50], highlighting the importance of exposure duration and microglial state. Given the existence of diverse microglial subtypes in vivo, the timing and context of graphene exposure may strongly influence immune responses. This study investigated the immediate response of unstimulated murine microglia to single‐layer graphene and employed proteomic analyses to identify early pathways potentially shaping long‐term outcomes.

2. Results

2.1. Short‐Term Cultivation of the Murine Microglia Cell Line SIM‐A9 on Graphene‐Coated Surfaces Leads to Activation

SIM‐A9 cells were passaged on pure glass or polyethylene terephthalate (PET) plates (0.5 cm edge length), and adenosine triphosphate (ATP) levels were measured in comparison to cells cultured for 3 h on standard plastic tissue culture plates. Exposure to uncoated glass or PET surfaces resulted in ATP levels comparable to control cells (Figure 1A). In contrast, cells cultured on graphene‐coated materials showed a fourfold increase in ATP on glass and a twofold increase on PET. ATP levels reflect various cellular states, including proliferative activity, cytotoxicity, or cellular stimulation [19, 51]. Imaging of cells immediately prior to lysis demonstrates that cell numbers are comparable between graphene‐coated and uncoated substrates, for example on glass (Figure 1B).

FIGURE 1.

Graphene‐dependent activation of SIM‐A9 microglia. (A) ATP content of SIM‐A9 cells cultured for 3 h on glass or PET surfaces with (+) or without (−) monolayer graphene‐coating in PBS. (B) Representative images of cells on glass (Gl) and graphene‐coated glass (Gr/Gl); scale bar: 200 µm. (C) TNF‐α secretion measured by ELISA. Lipopolysaccharide (LPS) treatment served as a positive control. (D) Cells were cotreated with surfaces or LPS in combination with polymyxin B for 3 h. Data are presented as percentage of control (untreated cells on tissue culture plastic) and shown as mean ± SD from at least two independent experiments (n ≥ 3 each). Statistical analysis was performed using one‐way ANOVA with Sidak's post hoc test (*p < 0.05, **p < 0.01, ***p < 0.001).

ATP plays a key role in microglial energy metabolism and activation. ATP consumption is associated with increased phagocytic activity, while lysosomal accumulation can lead to ATP excretion [52], and extracellular ATP can itself activate microglia [53]. The combination of altered ATP levels with unchanged cell counts suggests a change in activation state. Tumor necrosis factor alpha (TNF‐α) secretion was quantified from cell supernatants using an enzyme‐linked immunosorbent assay (ELISA) (Figure 1C). Cells cultured on pure glass and treated with lipopolysaccharides (LPS) exhibited a 200% increase in TNF‐α compared to untreated controls. A comparable increase was observed in cells cultured on graphene‐coated glass without additional stimuli.

To exclude the possibility of LPS contamination [54], polymyxin B, which neutralizes LPS by arranging it into inactive crystalline structures [55], was added during the 3‐h cultivation period. Polymyxin B completely abolished the LPS‐induced ATP increase but had no effect on ATP levels in cells cultured on graphene‐coated glass (Figure 1D) [55]. These results indicate that short‐term activation of SIM‐A9 cells on graphene‐coated surfaces is independent of LPS contamination. The absence of bacterial contamination was further confirmed by measuring bacterial ATP and failure of attempting to cultivate bacteria from the used materials (data not shown).

2.2. Number of Graphene Layers Does Not Influence the Activation Observed in SIM‐A9 Cells

To investigate how graphene may induce microglial activation, three potential mechanisms are considered: (1) surface topography, (2) material properties, and (3) activation via ingestion (phagocytosis). Previous studies report that three‐dimensional graphene foams exhibit anti‐inflammatory effects [48], suggesting that dimensionality may influence microglial responses. To test this, monolayer, bilayer, and trilayer graphene films were fabricated on glass, and their surface properties were characterized using Raman spectroscopy, contact angle measurements, and atomic force microscopy (Figure 2A–D).

FIGURE 2.

Multilayering does not affect graphene‐induced activation of SIM‐A9 cells. (A) Raman spectra of glass substrates (Gl) coated with 1–3 layers of graphene (1×–3× Gr/Gl); exemplary spectra from single points are shown. (B) Contact angle measurements (n = 3 per group). (C,D) Atomic force microscopy (AFM) images of graphene surfaces; representative images are shown (n = 5 per group). (E) ATP content of SIM‐A9 cells cultured for 3 h on glass coated with 1–3 layers of graphene. Data are presented as percentage of control and shown as mean ± SD from three independent experiments (n = 3 per group). Statistical analysis was performed using one‐way ANOVA with Sidak's post hoc test (ns, p > 0.05; *p < 0.05; ***p < 0.001).

Raman spectra confirmed the identity of the graphene substrates, with the expected G and 2D peaks observed (Figure 2A). The position and intensity of these bands indicate the number of graphene layers [56]. As the number of layers increases, the G and 2D bands shift, and the G‐band intensity rises, consistent with previous reports (Figure 2A) [57]. Peak intensities increased proportionally with the number of layers. Contact angle measurements showed no significant difference between the layers (Figure 2B), while surface roughness increased with the number of layers (Figure 2C,D). Despite these changes, the number of graphene layers did not affect the activation state of SIM‐A9 cells, as indicated by intracellular ATP levels (Figure 2E).

2.3. Carbon Nanotubes Activate SIM‐A9 Cells While Nanoplatelets Have No Effect

Monolayer and multilayer graphene coatings activated the SIM‐A9 microglial cell line, suggesting that activation may be related to general material properties, such as the hexagonal carbon lattice. To further explore this, two graphene‐derived nanomaterials—carbon nanotubes and graphene nanoplatelets—were tested. The amount of each material applied was calculated to match the graphene‐covered surface area of the monolayer films during a 3 h exposure.

Exposure to both nanomaterials resulted in a decrease in intracellular ATP levels, which reached statistical significance for graphene nanoplatelets (Figure 3A). Given the multifaceted role of ATP in microglia, tumor necrosis factor alpha (TNF‐α) secretion was measured in cell supernatants (Figure 3B). Lipopolysaccharide (LPS) served as a positive control and induced strong TNF‐α release [58]. While graphene nanoplatelets showed no effect, carbon nanotubes induced a tenfold increase in TNF‐α secretion. Coadministration of polymyxin B did not alter this response, confirming that the effect was independent of a potential LPS contamination of the used graphene material.

FIGURE 3.

Carbon nanotubes (NT) but not nanoplatelets (NP) induce stimulation of SIM‐A9 cells. (A) ATP content of SIM‐A9 cells incubated for 3 h with graphene nanomaterials (amount adjusted to match a graphene monolayer). (B) TNF‐α secretion measured by ELISA; lipopolysaccharide (LPS) served as a positive control. (C,D) Prostaglandin and nitric oxide levels measured in cell supernatants. Data are presented as percentage of control and shown as mean ± SD from three independent experiments (n = 2 each). Statistical analysis was performed using Kruskal–Wallis test with Dunn's multiple comparisons (ns, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001).

Additional indicators of microglial activation were assessed. Cyclooxygenase‐2 (COX2) activity, as measured by prostaglandin production, and nitric oxide (NO) release were both elevated following short‐term exposure to carbon nanotubes, whereas graphene nanoplatelets did not affect these pathways (Figure 3C,D). These results indicate that phagocytizable graphene nanotubes, but not nanoplatelets, trigger a robust inflammatory response in microglia.

2.4. Activation of SIM‐A9 Cells Due to Phagocytosis of the Graphene Material

Although both graphene nanoplatelets and carbon nanotubes are composed of graphene [59], only nanotubes induced microglial activation, suggesting that intrinsic material properties and their behavior during incubation drive this effect. To investigate this, both nanomaterials were incubated in phosphate‐buffered saline (PBS) for 3 h in the absence of cells (Figure 4A), and particle size distribution and total particle number were analyzed.

FIGURE 4.

Aggregation behavior of carbon nanotubes might contribute to their immune stimulatory potential. (A) Representative images of nanoplatelets (NP) and nanotubes (NP) incubated for 3 h in PBS. (B) Particle distribution analyzed using ImageJ. (C) Quantification of particles in size groups > 6 µm (large), 6–4 µm (medium), and 2–3 µm (small) at baseline and after 3 h. Data are presented as percentage of total particles and shown as mean ± SD from three independent experiments (n = 3 per group). Statistical analysis was performed using one‐way ANOVA with Sidak's multiple comparisons (ns, p > 0.05; *p < 0.05; ***p < 0.001). Scale bars: 100 µm (full images) and 5 µm (magnifications).

The nanoplatelets used in these experiments consisted predominantly (∼90%) of particles larger than 6 µm, which are too large for microglial phagocytosis. Smaller particles capable of being phagocytosed (4–6 µm) or optimal for phagocytosis (2–3 µm) constituted approximately 7% and 2% of the population, respectively (Figure 4B). After 3 h of incubation, the proportion of larger particles increased slightly at the expense of smaller species, but changes were not statistically significant.

In contrast, graphene nanotubes exhibited a shift in particle distribution over 3 h: large particles decreased by approximately 10%, while smaller, phagocytosis‐optimal particles increased from 1% to 6% (p = 0.068). This behavior suggests that nanotubes provide a higher proportion of particles suitable for microglial ingestion. Total particle number decreased for nanoplatelets, consistent with aggregation, whereas nanotube particle numbers remained relatively stable (Figure 4C).

All experiments described above were performed in phosphate‐buffered saline (PBS), which does not affect SIM‐A9 cell viability during short‐term culture (data not shown) and avoids interference from serum proteins, such as the formation of a protein corona on graphene [60, 61, 62]. To visualize and assess phagocytosis of graphene, a longer incubation period was required. Therefore, cells were exposed to nanomaterials for 24 h in a minimal pH‐balanced medium containing glucose and amino acids, but without serum.

Under these conditions, both graphene nanoplatelets and carbon nanotubes were internalized by SIM‐A9 cells (Figure 5A), with nanotubes taken up to a greater extent (Figure 5B). Ingested material showed a preference for small particles or particle aggregates over medium or large particles (Figure 5C). The proportion of cells exhibiting morphological alterations increased in cultures exposed to either nanoplatelets or nanotubes; however, the effect was significantly more pronounced in cells that internalized nanotubes. (Figure 5D). Notably, even cells that did not visibly internalize graphene displayed morphological changes, particularly in nanotube‐exposed cultures. This observation suggests that damaged microglia may release soluble factors that propagate the response to neighboring cells.

FIGURE 5.

Uptake of graphene nanomaterials and morphological changes in SIM‐A9 cells after 24 h of exposure. (A) Representative microscopic images of cells incubated with graphene nanoplatelets (NP) or nanotubes (NT) in serum‐free medium for 24 h; serum‐free medium alone served as control. (B) Percentage of cells containing ingested particles (mean ± SD); quantified from all evaluated cells. Statistical analysis: unpaired t‐test (*p < 0.05). (C) Size distribution of ingested particles, assessed using the Single Cellome System SS2000. (D) Morphological alterations in SIM‐A9 cells, including blebbing (upper, indicated by white arrow) and mitochondrial fragmentation (lower; mitochondria stained with LumiTracker Red). Percentage of damaged cells (cells were designated “damaged” when showing blebbing and/or mitochondrial defects) presented as mean ± SD. Cells were evaluated in three independent experiments (n = 8). Statistical analysis: one‐way ANOVA with Sidak's multiple comparisons (***p < 0.001; **p < 0.01). Scale bar: 10 µm.

To further assess nanotube‐induced cellular impairment, viability was measured 24 h after exposure, and inflammasome activity was evaluated via caspase‐1 catalytic function (Figure S1). Nanotube treatment significantly reduced both cell viability and caspase‐1 activity at 24 h. These effects were not reversed after an additional 24 h (48 h total exposure). However, due to strong adhesion of nanotubes to the culture vessels, complete removal by medium exchange was not feasible, resulting in continued exposure. Future studies should determine whether short‐term exposure followed by a graphene‐free recovery phase allows restoration of cellular function.

Since activation of SIM‐A9 cells appeared to depend on a higher rate of phagocytosis, the integrity of the graphene coating on glass plates used in Figure 1 was examined. Surprisingly, Raman spectroscopy revealed that the coating was largely removed in the areas measured after culturing the cells (Figure 6A). In five out of six samples, no graphene signal was detected at any measurement point, whereas only one sample showed residual signal in limited regions  (Figure 6B).

FIGURE 6.

Raman spectra of previously graphene‐coated glass after colonization with SIM‐A9 cells. Graphene‐coated glass was incubated with SIM‐A9 cells for 3 h. (A) Raman spectra were recorded from a 20 × 20 µm² area at the center of the sample before and after cell culture (six independent samples). Exemplary point spectra from one sample each are shown. (B) Quantification of points showing graphene‐derived Raman signals before and after cell contact. Data are presented as mean ± SD. Statistical analysis: unpaired t‐test (***p < 0.001).

To exclude the possibility that coating loss resulted from transport or exposure to lysis reagents, graphene‐coated slides were subjected to identical handling procedures in the absence of cells (Figure S2). None of these cell‐free control conditions resulted in coating loss, indicating that graphene detachment occurred specifically in the presence of cells, most likely as a consequence of phagocytic activity.

2.5. Proteomic Analysis

While graphene nanoplatelets did not induce microglial activation, culture on a graphene monolayer and exposure to carbon nanotubes elicited a similar inflammatory state, as indicated by increased tumor necrosis factor alpha (TNF‐α) secretion. However, ATP responses differed between the two conditions, suggesting that distinct underlying pathways may mediate the inflammatory response. To investigate this without bias, a proteomic analysis was performed for all conditions: control (glass), graphene‐coated glass, graphene nanoplatelets, and carbon nanotubes (Figure 7).

FIGURE 7.

Identification of molecular pathways involved in graphene‐based microglial stimulation via proteome analysis. Proteomic analysis was performed on SIM‐A9 cell lysates after 3‐h incubation with graphene on glass (Gl/Gl), graphene nanoplatelets (NP), or carbon nanotubes (NT) in PBS, compared to cells on pure glass. Only proteins with log₂ fold change >1.5 or <−1.5 and adjusted p‐value < 0.05 were considered. (A) Volcano plot showing differential protein expression relative to glass control. The x‐axis represents log₂ fold change, and the y‐axis represents ‐log₁₀ ( p‐value) from a modified t‐statistic (t(SAM)). (B) Venn diagrams showing overlap of significantly upregulated (blue, left) and downregulated (red, right) proteins. Circle size does not correspond to protein quantity. (C) Gene ontology analysis of up‐ and downregulated proteins, highlighting the 10 most strongly regulated pathways. Color scale indicates pathway strength: blue = upregulation, red = downregulation. (D,E) Protein–protein interaction networks of the top 50 significantly upregulated (D) and downregulated (E) proteins, visualized using STRING. Nodes represent proteins; edges represent predicted functional associations, with edge thickness reflecting interaction confidence. Three central networks were identified for both up‐ and downregulated proteins using k‐means clustering (minimum interaction score: 0.400). White spheres indicate unclustered proteins. Proteomics were performed with n = 4 per condition from four independent experiments.

Out of 8,023 proteins identified, approximately 15% were significantly differentially expressed in response to graphene‐coated glass (1,273 proteins) or nanotube treatment (1,140 proteins), whereas only 15 proteins were differentially expressed following graphene nanoplatelet exposure compared to glass (Figure 7A). In both graphene film and nanotube conditions, more proteins were downregulated than upregulated: graphene‐treated cells showed 247 upregulated and 1,026 downregulated proteins, while nanotube‐treated cells exhibited 263 upregulated and 877 downregulated proteins (Figure 7B). In nanoplatelet‐treated cells, the differentially upregulated proteins included Mapre3, Gripap1, Dpy30, Pea15, Trim8, Slc48a1, Trim3, and Wdr72, and the downregulated proteins include Chd6, Gm550, Bloc1s3, Fubp3, Ift122, Ttyh2, and Grhl3, with only five proteins being unique to nanoplatelet‐treatment. The majority of upregulated (226) and downregulated (802) proteins were shared between incubation with graphene film and nanotubes, with smaller numbers being unique to each treatment.

Gene ontology analysis of the top 10 upregulated pathways for both graphene film and nanotube treatments revealed enrichment in ADP metabolic process, cellular response to oxidative stress, carbohydrate catabolic process, vacuolar transport, macro‐autophagy, cellular lipid catabolic process, autophagy, and cellular response to chemical stress (Figure 7C). Notably, ADP metabolic process and cellular response to oxidative stress were most strongly elicited by graphene film, whereas cellular response to chemical stress and vacuolar transport were strongest affected in nanotube‐treated cells. Downregulated pathways included carboxylic acid metabolic process, chromosome condensation, DNA primase and polymerase activity, DNA methylation, sister chromatid cohesion, mitotic spindle midzone assembly, and regulation of mitochondrial ATP synthesis‐coupled electron transport. The most strongly downregulated pathways differed slightly between graphene film and nanotube treatments.

Ranking of the top 50 significantly up‐ and downregulated proteins showed largely overlapping proteins but in different orders. Oxoglutarate dehydrogenase L (Ogdhl) was the most strongly upregulated protein in both conditions, and Ubiquitin Like With PHD And Ring Finger Domains 1 (Uhrf1) was the most downregulated in graphene film‐treated cells, whereas Kinesin Family Member 20B (Kif20b) was the most downregulated in nanotube‐treated cells. Protein–protein interaction analysis of the top 50 upregulated proteins revealed a single strongly connected network (Figure 7D), whereas downregulated proteins formed three highly interconnected networks (Figure 7E). K‐means clustering of both up‐ and downregulated proteins identified three central network clusters for each: upregulated clusters related to cytoskeleton organization, carboxylic acid metabolism, and inflammation response (seven proteins unclustered), and downregulated clusters related to cell division, chromatin remodeling, and DNA replication (six proteins unclustered).

To validate the proteomics results, Western blot analysis was performed using independently treated cell cultures to assess Kif20B protein levels, one of the most strongly downregulated proteins following nanotube exposure. Consistent with the proteomic data, nanotube‐treated cells exhibited a marked reduction in Kif20B protein levels, with only 32% remaining compared to control and nanoplatelet‐treated cells (Figure S3). No significant difference was observed between control and nanoplatelet‐treated cells.

3. Discussion

Biomaterials, including lipids, polymers, and inorganic materials such as graphene, are widely used in medicine and other technical applications. Despite their beneficial properties, it is important to assess their effects on specific cell types to prevent unforeseen side effects. Cells of the immune system, which primarily recognize foreign materials and initiate defensive responses, require particular attention.

This study analyzes the effects of graphene on murine microglial cells, the immune cells of the brain. Although microglia might not typically be exposed to ingested graphene‐based materials, such materials are increasingly used in bone reconstruction [63], peripheral nerve injury repair [17], and neural interfaces [64] due to their mechanical and electrical properties. Even as diagnostic or therapeutic tools, graphene‐based materials show efficacy in preclinical models [65]. However, careful evaluation of single‐cell type responses remains critical.

SIM‐A9 microglial cells exhibited inflammatory responses to graphene coated on glass or polyethylene terephthalate substrates. Physicochemical characteristics of the material, including pore size, roughness, rigidity, topography, hydrophobicity, and surface charge, can influence immune responses [62, 66]. However, varying the number of graphene layers did not affect activation, suggesting that surface properties alone do not determine the response. Extrinsic factors, such as bacterial lipopolysaccharide contamination, were excluded using polymyxin B, indicating that activation was not due to endotoxins.

Given their phagocytic capacity, microglia are likely to respond to graphene through ingestion rather than surface contact alone. Supporting this, small carbon nanotubes, which are efficiently internalized, induced stronger inflammatory responses than larger graphene nanoplatelets [67]. Raman spectroscopy confirmed that microglia actively remove graphene from coated surfaces, explaining the observed activation.

Proteomic analysis revealed that ingestion of graphene nanotubes alters pathways related to inflammation, metabolism, cytoskeleton organization, and DNA maintenance, whereas graphene nanoplatelets had minimal effects. These findings are consistent with previous studies showing transient inflammatory responses in primary microglia and cell line models exposed to carbon nanomaterials [49, 68]. It should be noted that microglia can exert both detrimental and beneficial effects in the injured or diseased brain. Together with the growing field of subcategorization of these cells, it is sometimes difficult to finally evaluate an observed phenotype regarding its (patho‐)physiological role (nurturer, sentinel, or warrior as stated in [69]). There are some immunological checkpoints, preventing an overshooting reaction due to external stimuli such as TREM2, CX3CR1‐fractalkine sensing, or the progranulin pathway. Scavenger receptors, in contrast, promote clearance of potentially harmful substances. Graphene exposure did not result in alteration of TREM2, CH3CR1, or progranulin; however, Scarb1 (SR‐B1, class B scavenger receptor) was downregulated, a receptor that has also been shown to be involved in microglial interaction with amyloidogenic peptides and their clearance [70, 71]. Together with the reduced expression of Kif20B, a protein involved in cytoskeletal rearrangement, cell migration, and adhesion [72], these findings may indicate microglial dysfunction, which in vivo could result in reduced clearance of deleterious substances from the brain parenchyma.

Finally, to understand the in vivo relevance of these findings, future studies should evaluate the persistence and attenuation of inflammatory responses [73] and apply single‐cell sequencing approaches to identify microglial subtypes that respond in physiological contexts, including homeostatic and disease‐associated populations [74].

4. Conclusion

This study investigated the effects of graphene on murine microglial cells and demonstrated that their reactivity depends on phagocytic engulfment. Particles with sizes optimal for phagocytosis elicited an inflammatory response, as indicated by multiple markers, including tumor necrosis factor alpha (TNF‐α), nitric oxide, and prostaglandins. The results provided evidence for active removal of graphene from coated surfaces by the cells. In contrast, graphene nanoplatelets did not induce microglial activation. Proteomic analysis identified cellular pathways that are up‐ or downregulated following short‐term, 3 h exposure to graphene. These findings highlight the importance of particle size in microglial responses and underscore the need for further studies to explore the in vivo reactivity of brain immune cells to graphene‐based materials.

5. Materials and Methods

5.1. Model System

SIM‐A9 cells (Applied Biological Materials Inc.; Richmond (BC), Canada) were cultured at 37°C, 95% humidity, and 5% CO₂ using Dulbecco's Modified Eagle Medium/Nutrient Mixture F‐12 (DMEM:F‐12), supplemented with 10% v/v heat‐inactivated fetal bovine serum (Gibco, Carlsbad, CA, USA), 5% v/v heat‐inactivated donor horse serum (Gibco, Carlsbad, CA, USA), 1% v/v Penicillin–Streptomycin (Sigma–Aldrich; Steinheim, Germany), and 1% v/v L‐Glutamine (Sigma–Aldrich; Steinheim, Germany). Upon reaching 80% confluency, the cells were passaged by adding Trypsin/EDTA‐solution (0.05% trypsin (0.54 mM EDTA in PBS) (Sigma–Aldrich; Steinheim, Germany)) for 5 min at 37°C, 95% humidity, and 5% CO₂.

5.2. Cleaning of Substrates

The quality of graphene film depends on the cleanliness of the surface. Therefore, all substrates including glass (Borofloat 33, Ra < 1.2 nm, SIEGERT WAFER GmbH; Aachen, Germany) and PET (bleher Folientechnik GmbH; Ditzingen‐Heimerdingen, Germany) were first chemically cleaned prior to graphene transfer to remove any organics or contaminants from the surface. The substrates were cleaned with detergent (Hellmanex; Hellma GmbH & Co. KG, Müllheim, Germany), acetone (min. 99.8%, CHEMSOLUTE; TH. Geyer GmbH & Co. KG, Rinningen, Germany), and 2‐propanol (min. 99.8%, CHEMSOLUTE; TH. Geyer GmbH & Co. KG, Rinningen, Germany) in an ultrasonic bath, for 5 min each at 70 W and a frequency of 50 Hz. After cleaning, the samples were dried using N₂. The glass substrates were then further treated in oxygen plasma (Diener electronics: low‐pressure plasma system, 100% power) for 15 min. The cleaned substrates were then immediately coated with graphene film.

5.3. Preparation of Graphene Samples

Monolayer graphene films grown by chemical vapor deposition (CVD) were purchased from Graphenea Semiconductor S.L. (San Sebastián, Spain) and transferred onto the substrates via a wet‐transfer process following the manufacturer's standard protocol. Briefly, the graphene film supported on a polymer layer and protected by a sacrificial layer was cut into the desired sizes and slowly immersed in deionized water. The floating graphene film, together with the protective layer detached from the polymer support, was then transferred onto the substrates (5 × 5 mm²). After transfer, the samples were air‐dried for 30 min. Graphene on glass substrates was subsequently annealed in a laboratory oven at 150°C for 1 h, whereas graphene on PET substrates was annealed at 50°C for 2 h. Prior to removal of the sacrificial layer, the samples were stored under vacuum overnight to prevent detachment of the graphene film from the substrate. For removal of the sacrificial layer, the samples were immersed in hot acetone (50°C, 1 h), followed by 2‐propanol (room temperature, 1 h). The resulting monolayer graphene samples were dried under a stream of N₂ and used without further modification. The quality of the graphene films was assessed by Raman spectroscopy to detect potential contamination‐related peaks (e.g., copper oxides) (Figure S4A) and by scanning electron microscopy (SEM) to examine surface morphology (Figure S4B,C). Bi‐ and trilayer graphene films were prepared by repeating the above‐described transfer procedure, including intermediate graphene treatment and transfer steps.

5.4. Exposure of SIM‐A9 Cells to Graphene and Cell Stimulation

Glass and PET surfaces, both with and without graphene‐coating, were positioned at the base of a 48‐well plate. As a fully untreated control, a 5 × 5 mm² square was demarcated on the plate floor plastic to obtain the same amount of surface covered by cells as with the samples. Subsequently, 200 µL of PBS cell mixture, featuring a concentration of 20,000 cells/50 µL (counted with Scepter 3.0, Merck Millipore, Darmstadt, Germany), were carefully added. Carbon nanotubes (carbon purity: min. 95%; number of walls: 3–15; outer diameter: 5–20 nm; inner diameter: 2–6 nm; length: 1–10 μm; apparent density: 0.15–0.35 g/cm³; loose agglomerate size: 0.1–3 mm; specific surface: ca. 240 m²/g (PlasmaChem GmbH; Berlin, Germany)) and graphene nanoplatelets (thickness: 1–4 nm; particle size: up to 2 μm; specific surface area: 700–800 m²/g; purity: 91 at.%; other elements: 0 < 7 at.% & N < 2 at.% (PlasmaChem GmbH; Berlin, Germany)) were diluted in sterile PBS to obtain a standardized ratio of SIM‐A9 cells to the graphene surface area to the volume of PBS of 400 cells/0.125 mm²/1 µL. For the incubation with graphene nanoplatelets or nanotubes, SIM‐A9 cells were resuspended in a PBS particle mixture and subsequently added to the wells. For experiments involving surfaces, the designated surface was placed into the corresponding well plate. The cells were then resuspended in PBS and added onto the surface. The cells were incubated for 3 h at 37°C (95% humidity and 5% CO₂). Before and after incubation, images were captured using an EVOS XL Core microscope (Life Technologies; Carlsbad, USA).

To induce microglial activation, SIM‐A9 cells were incubated with lipopolysaccharide (LPS) from E. coli K12 (Invivogen; San Diego, CA USA) at a concentration of 10 ng/mL. To block the biological effect of lipopolysaccharide (LPS), 10 µM polymyxin B‐sulfate (Roth, Karlsruhe, Germany) was applied.

5.5. Harvesting of Cellular Material and Microscopy

The supernatant of cells was aspirated and stored at −80°C, while the cells were rinsed thrice with PBS. Cells originating from the designated control area were scraped off using a customized cell scraper (Sarstedt AG & Co. KG; Nümbrecht, Germany) (5 mm edge) and collected. Excess liquids were removed from the surface of the samples by gently placing a corner of the sample on a tissue (WEPA Deutschland GmbH & Co. KG; Arnsberg‐Müschede, Germany). Subsequently, the samples were transferred to 1.5 mL reaction tubes for following experimental procedures. The cells attached to the samples were lysed by adding 30 µL 1:5 diluted passive Lyse‐Puffer, 5x (Promega Corporation; Madison (WI), USA).

To visualize the mitochondria, cells were stained with LumiTracker Mito Red CMXRos (10 mM, Lumiprobe GmbH), diluted in CO₂‐independent medium to 25 nM. 100 μL of this solution was added to each well, and the plate was incubated at 37°C for 15 min. Microscopy was then performed using the Single Cellome System 2000 (Yokogawa) as described in a previous study [75].

5.6. Assessment of ATP Concentration

50 µL of CellTiter‐Glo reagent (Promega Corporation; Madison (WI), USA) were added to 50 µL of sample and agitated for 2 min at 500 rpm. The tubes were subsequently incubated at room temperature, shielded from light, for 10 min. Samples were transferred into a white 96‐well plate (Greiner Bio‐One International GmbH; Kremsmünster, Austria). The luminescent signal was quantified using a Microplate Reader FluoStar Omega (BMG Labtech) with a gain value of 4095 with 10 measurements every 0.5 s. Data analysis was carried out using the OMEGA analyzer software, version 2.40 (BMG Labtech).

5.7. TNF‐α ELISA

For the detection of TNF‐α, the ABTS ELISA Kit (peprotec; Hamburg, Germany) was used with the detection and capture antibody from the Murine TNF‐α Mini ABTS ELISA Development Kit (peprotec; Hamburg, Germany) and developed according to the manufacturer's ABTS ELISA‐protocol. Per sample, 25 µL cell supernatant was combined with 75 µL PBS. Sample measurement was performed at 405 nm with a wavelength correction set at 650 nm using the Sunrise Microplate reader (Tecan; Männedorf, Switzerland). Data analysis was carried out using the Magellan 7.3 software (Tecan; Männedorf, Switzerland).

5.8. COX‐2 ELISA

The level of COX‐2 was detected using the COX Inhibitor Screening Assay Kit (Cayman Chemicals; Ann Arbor (MI), USA) according to the manufacturer's protocol and using 50 µL of cell supernatant diluted 1:6 in PBS. Sample measurement was performed at 410 nm using the Sunrise Microplate reader (Tecan; Männedorf, Switzerland). Data analysis was carried out using the Magellan 7.3software (Tecan; Männedorf, Switzerland).

5.9. Caspase‐1 Activity Assay

The activity of Caspase‐1 was assessed by using the Caspase‐Glo 1 Inflammasome Assay (Promega Corporation; Madison (WI), USA), following the manufacturer's instructions.

5.10. Protein Determination According to Bradford

In order to determine the protein content of samples, a Bradford assay was performed using 25 µL of 1:5 diluted cell lysate. The assay was conducted by incubating the samples and standards with Roti‐Nanoquant (Carl Roth GmbH & Co. KG Karlsruhe, Germany) in a 1:5 dilution for 5 min at room temperature. Sample measurement was performed at 595 nm with a wavelength correction set at 450 nm using the Sunrise Microplate reader (Tecan; Männedorf, Switzerland). Data analysis was carried out using the Magellan 7.3 software (Tecan; Männedorf, Switzerland).

5.11. NO Assay

For detection of NO, Griess reagent (ENZO; Farmingdale, NY, USA) was used according to the manufacturer's recommendation. For each sample, 50 μL of cell supernatant combined with 50 μL of PBS was used. The nitrite concentration was measured at 540 nm using the Sunrise Microplate reader (Tecan; Männedorf, Switzerland) at a rate of 1 measurement per second. Subsequent data analysis was conducted using version 7.3 of the Tecan analysis software.

5.12. Atomic Force Microscopy (AFM)

The surface of graphene layers (mono‐, bi‐, and tri‐layers) on glass was measured by AFM (Dimension Icon, Bruker Corporation) in tapping mode. 0.01–0.025 Ohm‐cm antimony (n) doped Si tips (Bruker, RTESPA‐525, f0 = 525 kHz) were used. During the sample scan, the height and the phase diagrams were recorded (scan size: 800 nm, aspect ratio: 1.00, sample lines: 496). Three AFM images were taken from different positions of the sample. The AFM image analysis for calculating the mean RMS roughness using the line profile (thickness: 20 pixels, length: 1 µm, cutoff: 0.05) was performed using Gwyddion software (Version 2.63).

5.13. Raman Spectroscopy

Raman spectroscopy measurements of the samples were performed using a high‐resolution Raman microscope system (LabRAM HR Evo‐Nano, HORIBA Jobin Yvon GmbH, Bensheim, Germany). A microscope objective (100×) and a laser light (660 nm) for excitation were used. Raw data were baseline corrected. For mapping, the size (20 × 20 µm²) at the center of the sample was selected, and the D, G, and 2D peaks of graphene were analyzed.

5.14. Scanning Electron Microscopy (SEM)

The SEM images of the graphene on Au surface were taken in vacuum using a Zeiss Supra 40 SEM equipped with a through‐the‐lens detector at accelerating voltage of 3.0 kV. Different measurement parameters used are pixel sizes: 22.33 nm and 111.6 nm; aperture size: 30 µm; working distance (WD) = 3.1 mm.

5.15. Contact Angle Measurements

Contact angle measurements were performed using the “sessile drop method” with an OCA15 system (DataPhysics Instruments GmbH, Filderstadt, Germany). The image of a water droplet (5 µL volume) was taken with an automatic digital camera within 5–10 s. The software (SCA20_U) was used to automatically fit the droplet on both sides to obtain contact angles.

5.16. Proteomic Analysis

5.16.1. Sample Preparation

After seeding SIM‐A9 cells in a 48‐well plate on graphene/glass surfaces or with graphene nanoparticles, the cells were incubated for 3 h in PBS. Subsequently, the supernatant was aspirated, and the cells were mechanically detached from the surfaces by pipetting 25 µL of RIPA buffer onto the cells. Four wells were used per sample, and the collected samples were pooled into a single 1.5‐mL reaction tube.

5.16.2. Enzymatic Protein Digestion

All samples were processed using the SP3 approach [76]. The proteins were then digested using trypsin overnight at 37°C. The resultant peptide solution was purified by solid‐phase extraction in C₁₈ StageTips [77].

5.16.3. Liquid Chromatography Tandem Mass Spectrometry

Peptides were separated via an in‐house packed 45 cm analytical column (inner diameter: 75 μm; ReproSil‐Pur 120 C₁₈‐AQ 1.9 μm silica particles, Dr. Maisch GmbH) on a Vanquish Neo UHPLC system (Thermo Fisher Scientific). The online reversed‐phase chromatography separation was conducted through a 100‐min nonlinear gradient of 1.6%–32% acetonitrile with 0.1% formic acid at a nanoflow rate of 300 nL/min. The eluted peptides were sprayed directly by electrospray ionization into an Orbitrap Astral mass spectrometer (Thermo Fisher Scientific). Mass spectrometry measurement was conducted in data‐dependent acquisition mode using a top50 method with one full scan in the Orbitrap analyzer (scan range: 325–1,300 m/z; resolution: 120,000, target value: 3 × 10⁶, maximum injection time: 20 ms) followed by 50 fragment scans in the Astral analyzer via higher energy collision dissociation (HCD; normalized collision energy: 26%, scan range: 150–2,000 m/z, target value: 1 × 10⁴, maximum injection time: 5 ms, isolation window: 1.4 m/z). Precursor ions of unassigned, +1, or higher than + 6 charge state were rejected. Additionally, precursor ions already isolated for fragmentation were dynamically excluded for 20 s.

5.16.4. Mass Spectrometry Data Processing

Raw data files were processed by MaxQuant software (version 2.1.3.0) [78] using its built‐in Andromeda search engine [79]. MS/MS spectra were searched against a target‐decoy database containing the forward and reverse protein sequences of UniProt M. musculus reference proteome (release 2023_05; 63,458 entries) and a default list of common contaminants. Trypsin/P specificity was assigned. Carbamidomethylation of cysteine was set as fixed modification. Methionine oxidation and protein N‐terminal acetylation were chosen as variable modifications. A maximum of 2 missed cleavages were tolerated. The “second peptides” options were switched on. “Match between runs” was activated. The minimum peptide length was set to be 7 amino acids. False discovery rate (FDR) was set to 1% at both peptide and protein levels.

The MaxLFQ algorithm [80] was employed for label‐free protein quantification using its default normalization option. Minimum LFQ ratio count was set to one. Both the unique and razor peptides were used for quantification. Differential expression analysis was performed in R statistical environment. Reverse hits, potential contaminants, and “only identified by site” protein groups were first filtered out. Proteins were further filtered to retain only those detected in at least three out of the four replicates in either group of each comparison. Following imputation of the missing LFQ intensity values, the statistical significance of the difference between the two groups was assessed using a modified t‐statistic (t(SAM, statistical analysis of microarrays) [81] and visualized in a volcano plot. The combined significance threshold (hyperbolic curve) was defined as t ₀ = 1.2 and s ₀ = 1.5.

5.16.5. Evaluation by Western Blotting

In brief, cells were seeded according to the proteomic experiments by using cell numbers and graphene particle amounts adopted to the surface of the wells of 12‐well plates. Cells were harvested as described and protein content assessed via Bradford (see above). Ten micrograms of protein were loaded and separated on 8% Tris glycine PAA gels together with the SeeBluePlus2 protein marker (Invitrogen). After tank blot transfer to nitrocellulose, membranes were blocked by using I‐Block overnight at 4°C. Incubation with primary antibody (1:1000, Proteintech) took place for 1.5 h at RT under constant shaking; secondary antibody (HRP‐labelled) was from Thermo Fisher Scientific (diluted 1:3000). Development of the signal was performed by SuperSignal West Femto chemiluminescent substrate (Thermo Fisher Scientific) and a CCD camera (LumiImager F1, Roche).

5.17. Statistics

Statistical significance for the comparison of two groups was determined by two‐sided Student's t‐test. For results with more than two groups, one‐way ANOVA was performed with appropriate post‐test LSD post hoc test (GraphPad Prism 6 and 8 (GraphPad Software, Suite, CA, USA)). Data are presented as mean ± SD. Outliers were detected by performing ROUTs test with a 1% limit and values were excluded if identified as outliers.

Supporting Information

Additional supporting information can be found online in the Supporting Information section. Supporting Fig. S1: Nanotube exposure reduces SIM‐A9 cell viability and caspase‐1 activity after 24 and 48 h, indicating sustained cellular impairment. Supporting Fig. S2: Raman control experiments demonstrate that graphene coating integrity is unaffected by non‐cell‐based handling and treatment procedures. Supporting Fig. S3: Independent Western blot analysis confirms significant downregulation of Kif20B protein levels in nanotube‐treated SIM‐A9 cells. Supporting Fig. S4: Raman spectroscopy and SEM imaging verify the structural quality and contamination‐free transfer of graphene films.

Funding

This study was supported by the Ministerium für Wissenschaft und Gesundheit Rheinland‐Pfalz (Ministry of Science and Health of the State of Rhineland‐Palatinate, Forschungskolleg MultiSensE), the European Union's Horizon Europe research and innovation programme under Grant Agreement No. 101119473 (MUNASET, Graphene Flagship Initiative), and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; Grant Nos. 524805621 and 520334485).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Material

Rai Pratika, Slotos Robert Subirana, Hasnain Ahmar, Walter Emma, Grigoli Marina Mantellatto, Petrou Irini, Dejung Mario, Chen Jia‐Xuan, Tüscher Oliver, Tarasov Alexey, Endres Kristina, Small Science 2026, 6, e202500531. 10.1002/smsc.202500531

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.