Background
Toll-like receptors (TLRs) are a subfamily of pattern recognition receptors (PRRs) essential for the initiation of inflammatory responses. Among 12 different TLRs that have been identified in mice, TLR4 is a main contributor to neuronal cell death, ischemic brain injury, and breakdown of the blood brain barrier [
1‐
4]. In the retina, TLR4 is mainly expressed by microglia cells [
5], Müller cells [
6], retinal pigment epithelial cells (RPEs) [
7,
8], and vascular endothelial cells [
9]. TLR4 is primarily activated by lipopolysaccharides (LPS); a bacterial endotoxin exposed on Gram-negative bacterial walls, allowing host cells to react to bacterial invasions by initiating rapid and robust inflammatory responses [
10‐
12].
LPS is the most commonly used molecule to trigger microglia and macrophages activation both in vivo and in vitro. In the central nervous system (CNS), microglia are involved in neuronal survival and function, they are the first responders to invading pathogens, and they actively participate in the inflammatory response together with monocyte-derived macrophages and other immune cells invading from the circulation after infection or injury [
13]. Microglia activation in the retina after systemic LPS exposure has been reported, and was correlated with exacerbation of retinal inflammation, increased photoreceptor death and worsening of retinal function in a genetic model of retinal degeneration, the RHO-P23H rat [
14]. Moreover, LPS-induced systemic inflammation in P4 mouse pups led to abnormalities in the development of retinal vasculature and permanent impairment of retinal function, both associated with microglia activation [
15]. On the other hand, LPS preconditioning has been shown to act in a neuroprotective manner in retina and brain disease [
16‐
18]. In the healthy CNS though, repetitive systemic challenges with LPS lead to widespread activation of microglia both in the retina and the brain [
16,
19,
20] and we have recently shown that they can also lead to blood retinal barrier (BRB) breakdown in albino mice [
19]. An intact BRB preserves retinal homeostasis acting as a physical barrier against the entry of molecules and pathogens from the periphery that could otherwise disturb retinal physiology [
21]. The involvement of TLR4 in blood brain barrier (BBB) disruption has been suggested in rodents [
4,
22,
23], while LPS-induced BRB breakdown suggests a role of TLR4 in BRB breakdown as well [
19].
RPE cells together with the underlying Bruch’s membrane and the fenestrated choriocapillaris form the outer BRB. The inner BRB is formed by endothelial cells’ interactions with pericytes, neurons, astrocytes, microglia and Müller cells, and insulates the inner retina from the deep, intermediate, and superficial vascular plexus. Endothelial cells are connected with tight junctions and create a tight monolayer that prevents the entry of foreign material into the retina, contributing to immune privilege [
24]. These cells, actively participate in immune regulation during inflammation, by upregulating the expression of cell adhesion molecules and chemokines that facilitates the rolling and influx of leukocytes into the CNS (for a recent review see Ref. [
25]).
Systemic LPS triggers activation of both endothelial cells and microglia in the CNS [
26‐
28]. However, whether LPS-induced microglia activation precedes endothelial activation or vice versa is unclear. For instance, vascular damage caused by methamphetamine or thiamine deficiency has been shown to attract microglia towards the affected vasculature, indicating that vascular damage is the event that triggers microglia migration and activation in an effort to repair the affected vessels [
29,
30]. However, disruption of the BBB by activated microglia has been proposed in vitro [
31]. Previous studies have shown that brain neurodegeneration and vascular damage are the cause and not the consequence of microglia activation [
32,
33]. Older studies also suggested that activation of microglia is mostly beneficial and could rarely initiate neurodegenerative cascades in the brain [
34]. We have previously shown that LPS-induced BRB disruption correlated with monocyte-derived macrophages influx and clustering of microglia/macrophages around the retinal endothelium in Balb/c mice, indicating microglia/macrophage–endothelium interactions following systemic LPS exposure [
19]. The influx of monocyte-derived macrophages into the central nervous system upon BRB and BBB breakdown has been described in several disease models (for a recent review see Ref. [
35]). These cells express TLR4 that allows their direct activation by LPS [
36] and their interaction with activated endothelial cells plays a major role in their ability to penetrate brain vessels [
37,
38].
In a previous study using bone marrow chimera mice, researchers have shown that TLR4 expressed on endothelial cells and microglia, rather than on cells originating in the blood, is required for systemic LPS-induced microglia activation in a murine brain injury model [
16]. Based on that study and since endothelial cells bear TLR4 and are in direct contact with blood, acting as a barrier that insulates the retina from the periphery, we hypothesized that systemic LPS would first bind to endothelial TLR4 before directly activating microglia TLR4. To test our hypothesis, we employed wild-type C57BL/6J mice and generated mice that lack
Tlr4 expression selectively on endothelial cells. First, we investigated the effect of systemic LPS exposure on the wild-type C57BL/6J pigmented retina and on the Tek
Cre−negTlr4
loxP/loxP retina, and secondly, the impact of endothelial
Tlr4 depletion on LPS effects. Our data support the hypothesis that upon systemic LPS exposure, activation of TLR4 located on endothelial cells is essential for subsequent microglia activation, monocyte-derived macrophages influx, microglia/macrophage interactions with endothelial cells and functional impairment.
Materials and methods
Animals
This study was approved by the local Animal Ethics Committee (Veterinärdienst des Kantons Bern: BE53/17 and BE3/2021) and conformed to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Adult 12–20 weeks male and female C57BL/6J mice, conditional knockout mice that are deficient for
Tlr4 expression in endothelial cells (Tek
Cre−posTlr4
loxP/loxP mice) and Tek
Cre−negTlr4
loxP/loxP mice were used in this study. The conditional knockout mice were generated by breeding B6.Cg-Tg(Tek-cre)1Ywa/J mice that express Cre recombinase under the direction of the receptor tyrosine kinase Tek (
Tie2) promoter/enhancer, which has been shown to provide uniform expression in endothelial cells [
39], with B6(Cg)-Tlr4
tm1.1Karp/J mice (The Jackson Laboratory strain #024872), which possess loxP sites flanking exon 3 of the
Tlr4 gene. Tek
CretdTomato mice were used to confirm the lack of tdTomato expression in Iba-1-positive cells (Iba-1
pos). Animals were housed under controlled temperature and humidity conditions in ventilated cages with a 12-h light–dark cycle. All mice had free access to food and water.
Systemic LPS challenge and in vivo imaging
Mice were injected via the tail vein once per day for a total of 4 days with 1 mg/kg LPS from
Escherichia coli serotype O111:B4 (Cat # L2630; Sigma-Aldrich, Buchs, Switzerland) as previously described [
19]. Fluorescein angiography (FA) and spectral-domain optical coherence tomography (SD-OCT) were used for the visualization of retinal vasculature and retinal structure, before the first LPS challenge and approximately 5 h after the last LPS challenge, using a Heidelberg Spectralis system (Heidelberg Spectralis HRA 2; Heidelberg Engineering GmbH, Heidelberg, Germany) as described elsewhere [
19]. Before and after in vivo imaging, the mice were anesthetized and the anesthesia was reversed, respectively, as described elsewhere [
19]. At the end of the experiments, mice were euthanized with CO
2 inhalation followed by decapitation.
Flow cytometry analysis (FACS)
Naïve and LPS-challenged C57BL6/J and Tek
Cre−posTlr4
loxP/loxP mice were euthanized approximately 16 h after the last LPS injection and their retinas were prepared for flow cytometry analysis as described elsewhere [
19]. Both retinas of each mouse were treated as one sample. Dead cells stained with Zombie Green™ Fixable Viability Kit (1:800; Biolegend, San Diego, CA, USA; Cat # 423111) were excluded from the analysis. Extracellular markers against the leukocyte common antigen CD45 (30-F11; 1:400; allophycocyanin-Cy7/APC-Cy7; Cat # 103115) and the CD11b antigen-like family member B (M1/70; 1:200; allophycocyanin/APC; Cat # 101212) were used for the labeling of microglia as CD11b positive and CD45 low or negative (CD11b
posCD45
low/neg) and macrophages as CD11b positive and CD45 high (CD11b
posCD45
hi) expressing cells [
40‐
43]. Fluorescein isothiocyanate (FITC) conjugated antibodies against CD19 (6D5; 1:200; Cat #115505), CD3 (17A2; 1:200; Cat #100203), NK1.1 (PK136; 1:200; Cat #108705) and Ly6G (1A8; 1:200; Cat #127605) were used for the exclusion of B-lymphocytes, T cells, natural killer cells and granulocytes, respectively. To test the selective ablation of TLR4 from endothelial cells in the conditional knockout mice, we used naïve retinas from Tek
Cre−posTlr4
loxP/loxP and Tek
Cre−negTlr4
loxP/loxP mice. Single-cell suspensions were stained with a TIE2 antibody (TEK4; 1:100; allophycocyanin/APC; Cat # 124009) and a primary mouse monoclonal TLR4 antibody (76B357.1; 1:400; Novus Biologicals, Littleton, CO, USA; Cat # NB100-56566SS), followed by incubation with a secondary goat F(ab′)2 anti-mouse IgG H&L (phycoerythrin, PE) pre-adsorbed antibody (1:100; Cat # ab7002; Abcam, Cambridge, United Kingdom). DAPI (1:800; Thermo Fisher Scientific, Waltham, MA, USA Cat # 62248) was used for exclusion of dead cells [
44]. Flow cytometry antibodies were purchased from Biolegend (San Diego, CA, USA) unless otherwise indicated above.
Immunohistochemistry
One eye per mouse was used for immunohistochemistry and one for retinal whole mounts. For immunohistochemistry, eyes were enucleated and fixed in 4% paraformaldehyde in 1× PBS (pH 7.4) overnight and embedded in paraffin. Slides with 5-μm sections were deparaffinized, rehydrated, and incubated for 15 min in Tris–EDTA buffer (pH 9.0) at 95 °C. Slices were blocked for 30 min with blocking buffer [5% normal goat serum (NGS, Lucerna-Chem AG, Luzern, Switzerland) in 0.2% Triton X-100 in 1× PBS] and incubated with the following primary antibodies: a rabbit polyclonal antibody against ionized calcium binding adapter molecule 1 (Iba-1, 1:1000; Cat # 0162-0001, FUJIFILM Wako Shibayagi, Japan), a mouse monoclonal antibody against c-terminal binding protein-2 (CtBP2, 1:250; Cat # 612044, BD Biosciences, Basel, Switzerland), and a rat monoclonal antibody against intercellular adhesion molecule 1 (ICAM-1, 1:200; Cat # ab119871, Abcam, Cambridge, United Kingdom). The sections were incubated with the primary antibodies overnight at 4 °C. All antibodies were diluted in blocking buffer. The slides were washed in 1× PBS and incubated with the secondary antibodies goat anti-rabbit Alexa Fluor 594 conjugate (1:1000; Cat # A11012, ThermoFisher scientific, Waltham, MA, USA), goat anti-mouse F(ab′)2 IgG (H + L) cross-adsorbed secondary antibody Alexa Fluor 488 conjugate (1:1000; Cat # A48286, ThermoFisher scientific, Waltham, MA, USA) and/or a goat anti-rat IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor™ 488 (1:1000; Cat # A-11006, ThermoFisher scientific, Waltham, MA,) for 1 h at RT. Slides were washed again and mounted in mounting medium with DAPI (Vector Laboratories, Reactolab SA, Servion, Switzerland).
Retinal whole mounts
For retinal whole mounts, eyes were isolated and processed as described elsewhere [
45]. Retinas were incubated with a primary antibody against Iba-1 diluted in blocking buffer (1:500; anti-rabbit Cat #019-19741, FUJIFILM Wako Shibayagi, Japan) overnight at 4 °C, for labeling of microglia/macrophages. Samples were washed in 0.2% TritonX-100 in 1× PBS and incubated with goat anti-rabbit Alexa Fluor 594 conjugate (1:200; Cat # A11012, ThermoFisher scientific, Waltham, MA, USA) for 2 h at RT. Retinas were flattened as described elsewhere [
45].
Microscopy
Retinal paraffin sections stained for Iba-1 and ICAM-1 were examined under a fluorescence microscope (Nikon Eclipse 80i microscope; Nikon, Tokyo, Japan). An inverted Zeiss LSM 710 fluorescence confocal microscope (Carl Zeiss, Oberkochen, Germany) was utilized for the imaging of the retinal whole mounts and for the CtBP2 staining. For Iba-1 staining in retinal whole mounts, Z-stacks with 0.85 μm interval were obtained using Zen system 2011 software (Carl Zeiss) with a 20× lens (Plan-Apochromat 20×/0.8 M27/a = 0.55 mm). For the CtBP2 staining, Z-stacks with 0.5 μm interval were obtained with a 100× lens (Plan-APOCHROMAT 100×/1.4 Oil Ph3). Figures were finalized in Coreldraw 2021.5 (Corel Corporation, Ottawa, ON, Canada).
Quantification of Iba-1pos cells morphology
For the quantification of branch length and endpoints per Iba-1
pos cell, z-stacks of retinal whole mounts were exported as individual images using the ZEN system 2011 software (Carl Zeiss, Oberkochen, Germany). Individual images corresponding to the area between the ganglion cell layer and the inner nuclear layer (GCL-INL) or outer plexiform layer (OPL) were combined for the generation of 2 new z-stacks. Z-stacks were z-projected and converted to grayscale. Skeleton analysis was performed in the z-stacks derived from the OPL using the ImageJ 1.52 h software according to a published protocol [
46]. The total branch length and the total number of endpoints per image were divided by the number of Iba-1
pos cells in each image. For measurements of Iba-1
pos cells volume, ImageJ 1.52 h software was used. Image threshold was adjusted in z-stacks containing the retinal thickness from the GCL to the OPL, and the Iba-1 occupied volume was measured using the “Measure stack” plugin. Iba-1 volume is expressed as a percentage of the total z-stack volume.
Retinal vasodilation measurements
Retinal vasodilation was measured in SD-OCT b-scans, measuring the diameter of the veins in the different layers of the retina, using the Heyex software version 1.9.17.0 (Heidelberg Engineering Inc., Franklin, USA). Three different veins from 6 b-scans were measured and averaged per eye.
Electroretinography (ERG) recordings
Retinal function of LPS-challenged mice was recorded at baseline and 4 days after the first LPS challenge, using electroretinography (ERG). The animals were recorded in photopic and scotopic conditions to measure the function of cone and rod photoreceptors, respectively, upon the LPS stimuli. To achieve cone saturation, mice were dark adapted overnight before all measurements. Mice were anesthetized and waked up as previously described [
47]. The pupils were dilated with tropicamide 0.5% and phenylephrine 2.5% eyedrops. A drop of Viscotears Liquid Gel was placed on the eyes to avoid mydriasis. Two reference electrodes with platinum needles were placed below the eyes and one ground electrode was placed above the tail. Gold rings contact electrodes were placed on the eyes of the animals and the animals were positioned on a custom-made platform. For measurements of
a- and
b-waves in scotopic conditions, ten different increasing luminances from − 4 to 1.5 log were used. A light adaptation period of 10 min was followed by eight different elevating intensities from − 2.0 to 1.5 log to measure both waves in photopic conditions. For the quantification of
b-waves in both conditions, a filter to remove the oscillatory potentials was applied.
Transmission electron microscopy (TEM)
Mice were deeply anesthetized with 150 mg/kg pentobarbital (Esconarkon, 300 mg/ml, Streuli Pharma AG, Uznach, Switzerland) diluted in 0.9% NaCl. Under deep anesthesia, mice were transcardially perfused with 50 ml 1× PBS supplemented with 10 U/ml heparin (in 0.9% NaCl; Inselspital, Bern, Switzerland), followed by 50 ml of Karnovsky solution (2.5% glutaraldehyde, and 2% paraformaldehyde in 0.1 M Na-cacodylate buffer pH 7.39). Mouse eyes were isolated and fixed with Karnovsky solution for at least 24 h before being further processed. Samples were washed with 0.1 M Na-cacodylate buffer (Merck, Darmstadt, Germany) and postfixed with 1% OsO4 (Electron Microscopy Sciences, Hatfield, USA) in 0.1 M Na-cacodylate buffer at 4 °C for 2 h. After washing, the samples were dehydrated in 70, 80, 96% and 100% ethanol (Alcosuisse, Switzerland), followed by acetone (Merck, Darmstadt, Germany), and finally samples were immersed in acetone–Epon (1:1) overnight. Samples were embedded in Epon (Sigma-Aldrich, Buchs, Switzerland) and left to harden at 60 °C for 5 days. Sections were produced with an ultramicrotome UC6 (Leica Microsystems, Vienna, Austria), first semithin sections (1 μm) for light microscopy which were stained with a solution of 0.5% toluidine blue O (Merck, Darmstadt, Germany) and then ultrathin sections (75 nm) for electron microscopy. The sections, mounted on Formvar® (Ted Pella Inc., USA) coated single slot copper grids, were stained with Uranyless (Electron Microscopy Sciences, Hatfield, USA) and lead citrate (Leica Microsystems, Vienna, Austria) with an Ultrostainer (Leica Microsystems, Vienna, Austria). Sections were examined with a transmission electron microscope (Tecnai Spirit, FEI, Brno, Czech Republic) equipped with a digital camera (Veleta, Olympus, Soft Imaging System, Münster, Germany).
Statistics
For the statistical analysis, GraphPad Prism software version 8.0 was used (GraphPad Software, Inc., San Diego, CA). Statistically significant differences between naïve and LPS groups were determined using unpaired two-tailed t-tests except for the vasodilation data, where paired two-tailed t-test was used. ERG data were compared with repeated measures 2-way ANOVA followed by Sidak’s post hoc analysis. All data are expressed as means ± SD. P values < 0.05 were considered statistically significant. Single dots in bar graphs represent individual samples (2 retinas from a single mouse for the FACS data and 1 eye per mouse for the rest of the data).
Discussion
The results of the present study demonstrate for the first time that activation of endothelial TLR4 is the first essential step for the sequel of events that lead to microglia activation, monocyte-derived macrophages infiltration into the retina and impaired retinal function upon systemic LPS exposure.
First, we generated mice that lack
Tlr4 expression selectively on endothelial cells to investigate the role of endothelial TLR4 in systemic LPS-induced retinal inflammation. To substantiate the validity of our conditional knockout mice, we utilized retinal whole mounts in Tek
CretdTomato reporter mice and we confirmed that these mice express Cre recombinase selectively in endothelial cells and not Iba-1-positive cells (Fig.
1). Additionally, flow cytometry analysis substantiated TLR4 depletion in TIE2-expressing retinal endothelial cells in our Tek
Cre−posTlr4
loxP/loxP mice (Fig.
1). The effect of repetitive systemic LPS exposures on the retina was investigated in C57BL/6J, Tek
Cre−negTlr4
loxP/loxP and Tek
Cre−posTlr4
loxP/loxP mice. Compared to our previous study in albino Balb/c mice; which are more susceptible to bacterial infection [
50,
51]; here, we did not detect major disruption of the blood retinal barrier as examined by fluorescein angiography in C57BL/6J mice (Fig.
2). However, this method would only detect large disturbances based on fluorescein size (376.7 kDa; [
52]). We did detect however, retinal vein dilation and monocyte-derived macrophages influx, which support BRB disruption (Figs.
2,
3). In agreement with our previous study and studies in the brain [
16,
19], we observed that microglia from C57BL/6J and Tek
Cre−negTlr4
loxP/loxP mice adopt an activated phenotype after the LPS challenge and they accumulate around retinal blood vessels (Fig.
2). At the same time, expression of ICAM-1 in the deep and intermediate vascular plexus correlated with elevated numbers of macrophages as shown by FACS analysis (Fig.
3). Interactions between ICAM-1 expressed by endothelial cells and leukocyte function-associated antigen-1 (LFA-1) expressed by leukocytes, contributes to the migration of leukocytes across the endothelium, partially through endothelial cell separation [
53,
54].
Next, based on previous observations that systemic LPS exposure leads to permanent impairment of retinal function in P4 mouse pups [
15], we performed ERG to detect possible LPS-induced functional impairments in the adult retina (Fig.
4). ERG showed reduced
a- and
b-wave amplitudes after the LPS exposure, reflecting the activity of photoreceptors and mainly ON bipolar cells, respectively. These effects were correlated with ribbon synapses damage as shown by immunohistochemistry and transmission electron microscopy (Fig.
5). Interestingly, similar impaired ERG responses and synapse dysfunction have been observed in microglia-depleted mice [
55]. Engulfing and elimination of synapses by microglia during development is necessary for synapse maturation in the brain [
56] and there is a hypothesis that microglia may influence synaptic function in the adulthood as well. In a rat model of retinitis pigmentosa, microglia cells were shown to phagocytize synaptic elements and this phagocytic activity was increased when synapses were destroyed [
57]. In the present study, we did not observe internalization of CtBP2-positive elements from Iba-1
pos cells (Additional file
2: Video S1) indicating that microglia may not directly affect synapses in our model. Interestingly, retinal microglia despite sharing a common developmental lineage, seem to have different functions depending on their location in the retina, determined by interleukin 34 (IL-34) dependency. Specifically, IL-34-dependent microglia that reside in the IPL seem to play an important role in retinal function since IL-34-deficient mice have reduced numbers of microglia in the IPL and reduced
b-wave responses, suggesting that they may play a role in visual processing [
58]. Thus, we speculate that microglia is not responsible per se for synapse elimination upon the LPS challenge. In contrast, our data in combination with the studies mentioned above, suggest that given the role of microglia in homeostatic synaptic regulation, disruption of microglia–neuronal interactions due to the diversion of microglia activity towards the vasculature may account for the observed LPS-induced impaired retinal function as it has been suggested in the brain [
59].
The results in our wild-type C57BL/6J mice correlate well with observations in the brain, where local LPS injection induced microglia and endothelium activation and recruitment of leukocytes from the periphery [
60]. Based on their location in the vessel lumen, endothelial cells are of the first encounters of pathogens present in the circulation, and they can modulate subsequent inflammatory responses [
61]. Additionally, leukocytes that circulate in the blood express TLR4 and can be directly affected by systemic LPS exposure. Systemic LPS exposure in mice that expressed TLR4 exclusively in endothelial cells led to neutrophil rolling and adhesion in brain vessels but failed to trigger their influx into the brain parenchyma. In the same study, expression of TLR4 by microglia rather than circulating cells was required for recruitment of neutrophils into the CNS. The authors suggested that direct activation of endothelial TLR4 by LPS is sufficient to initiate leukocyte–endothelial interactions, but TLR4 activation on brain microglia is required for the entry of leukocytes into the brain parenchyma [
62]. The same group has previously reported that inhibition of microglia activation with minocycline did not affect the activation of endothelial cells in vitro [
60], suggesting that microglia do not play a major role in endothelial activation.
Here, using conditional knockout mice, we show that in absence of endothelial TLR4, systemic LPS fails to induce any effect in the retina in terms of microglia activation, retinal vasodilation, monocyte-derived macrophages influx or synaptic impairment (Figs.
2,
3,
4,
5). A recent study showed that activation of microglia by systemic LPS is mediated by cytokines released from activated endothelial cells together with direct activation of TLR4 expressed by microglia in the perivascular space [
20]. Activated endothelial cells release chemokines and cytokines, such as the chemokine (C–C motif) ligands 2, 7 and 20 (Ccl2, Ccl7 and Ccl20, respectively) [
63] and inflammatory molecules, such as nitric oxide [
64], which attract microglia to the affected vasculature in an attempt to maintain the integrity of the barrier. This involves expression of tight-junction proteins and establishment of physical contact with the endothelial cells [
65]. Previous studies has shown that only a minimal amount of LPS (approximately 0.025% of the injected amount) can enter the brain parenchyma after intravenous LPS administration in mice, while the majority of LPS interacts with the BBB via binding to the luminal surface irreversibly [
66]. We did not detect any vasodilation or signs of microglia/macrophages activation in the absence of endothelial TLR4 (Fig.
2), suggesting that the LPS-induced microglia activation that was observed in the present study depends on the activation of TLR4 located on endothelial cells. Moreover, we did not detect differences in the ERG responses of our conditional knockout mice before and after the LPS challenge (Fig.
4). However, there were signs of retinal dysfunction in both Cre-negative and Cre-positive Tlr4
loxP/loxP mice, revealed by lower ERG responses at baseline compared to C57BL/6J mice, suggesting that irrespective of Tlr4 deficiency in endothelial cells, the genetic background of these mice may have an effect on retinal function under normal conditions.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.