Macrophage to myofibroblast transition contributes to subretinal fibrosis secondary to neovascular age-related macular degeneration

Background Macular fibrosis causes irreparable vision loss in neovascular age-related macular degeneration (nAMD) even with anti-vascular endothelial growth factor (VEGF) therapy. Inflammation is known to play an important role in macular fibrosis although the underlying mechanism remains poorly defined. The aim of this study was to understand how infiltrating macrophages and complement proteins may contribute to macular fibrosis. Methods Subretinal fibrosis was induced in C57BL/6J mice using the two-stage laser protocol developed by our group. The eyes were collected at 10, 20, 30 and 40 days after the second laser and processed for immunohistochemistry for infiltrating macrophages (F4/80 and Iba-1), complement components (C3a and C3aR) and fibrovascular lesions (collagen-1, Isolectin B4 and α-SMA). Human retinal sections with macular fibrosis were also used in the study. Bone marrow-derived macrophages (BMDMs) from C57BL/6J mice were treated with recombinant C3a, C5a or TGF-β for 48 and 96 h. qPCR, Western blot and immunohistochemistry were used to examine the expression of myofibroblast markers. The involvement of C3a-C3aR pathway in macrophage to myofibroblast transition (MMT) and subretinal fibrosis was further investigated using a C3aR antagonist (C3aRA) and a C3a blocking antibody in vitro and in vivo. Results Approximately 20~30% of F4/80+ (or Iba-1+) infiltrating macrophages co-expressed α-SMA in subretinal fibrotic lesions both in human nAMD eyes and in the mouse model. TGF-β and C3a, but not C5a treatment, significantly upregulated expression of α-SMA, fibronectin and collagen-1 in BMDMs. C3a-induced upregulation of α-SMA, fibronectin and collagen-1 in BMDMs was prevented by C3aRA treatment. In the two-stage laser model of induced subretinal fibrosis, treatment with C3a blocking antibody but not C3aRA significantly reduced vascular leakage and Isolectin B4+ lesions. The treatment did not significantly alter collagen-1+ fibrotic lesions. Conclusions MMT plays a role in macular fibrosis secondary to nAMD. MMT can be induced by TGF-β and C3a but not C5a. Further research is required to fully understand the role of MMT in macular fibrosis. Graphical abstract Macrophage to myofibroblast transition (MMT) contributes to subretinal fibrosis. Subretinal fibrosis lesions contain various cell types, including macrophages and myofibroblasts, and are fibrovascular. Myofibroblasts are key cells driving pathogenic fibrosis, and they do so by producing excessive amount of extracellular matrix proteins. We have found that infiltrating macrophages can transdifferentiate into myofibroblasts, a phenomenon termed macrophage to myofibroblast transition (MMT) in macular fibrosis. In addition to TGF-β1, C3a generated during complement activation in CNV can also induce MMT contributing to macular fibrosis. RPE = retinal pigment epithelium. BM = Bruch’s membrane. MMT = macrophage to myofibroblast transition. TGFB = transforming growth factor β. a-SMA = alpha smooth muscle actin. C3a = complement C3a.


Background
Age-related macular degeneration (AMD) is a disease which results in loss of central vision in the elderly. It is estimated that around 288 million people will be diagnosed with AMD worldwide by the year 2040 [1]. Around 10% of patients with AMD suffer from the neovascular form of the disease (nAMD), which is characterized by the growth of abnormal blood vessels in the macula and leaves patients more at risk of severe vision loss [2]. If nAMD remains untreated, eventually, patients will lose sight and develop macular fibrosis [3,4]. The introduction of anti-vascular endothelial growth factor (anti-VEGF) therapy has revolutionized nAMD therapy [5,6]. However, around one-third of patients still develop macular fibrosis even with anti-VEGF therapy [7]. Macular fibrosis remains to be a major clinical challenge in nAMD management. With an ageing population and predicted increases in AMD patients, this challenge is set to become an increasing problem.
Macular fibrosis lesions contain blood vessels and hence are known as fibrovascular lesions [8][9][10]. The specific cues involved in the transition of the diseased vessels to a fibrovascular scar remain unknown. As well as blood vessels, macular lesions contain infiltrating immune cells, myofibroblasts and excessive amounts of extracellular matrix (ECM) proteins such as collagens, fibronectin and laminin [11][12][13]. Myofibroblasts are the active form of fibroblasts, which do not exist in the macula. It has been hypothesized that myofibroblasts in macular fibrosis originate from differentiation of either resident retinal cells (e.g., retinal pigment epithelial cells) or infiltrating inflammatory cells [8,14] although direct evidence is lacking.
A previous study has shown that 61% of human choroidal neovascularization (CNV) lesions contained macrophages [15]. Macrophages constitute 20% of all cells in experimental CNV, and~70% of infiltrating macrophages originate from the bone marrow [16]. Macrophage is believed to play an important role in macular fibrosis development although the underlying mechanism remains to be determined. Recent studies have shown that macrophages can transdifferentiate into myofibroblast-like cells after TGF-β stimulation (macrophage to myofibroblast transition; MMT) and that this process contributes to kidney fibrosis [17][18][19].
Complement system dysregulation has been identified as a key inflammatory pathway in AMD [20][21][22][23]. Our group has shown that patients with macular fibrosis in nAMD have increased levels of complement components including 3a, 4a and 5a [24]. Neutralising C3a and C5a during experimental laser-induced CNV reduced neovascular lesions [25]. It has been reported that C3a can induce epithelial to mesenchymal transition (EMT) differentiation in proximal tubule epithelial cells (PTECs) through the C3a receptor-C3aR1 [26]. In this study, we investigated the role of MMT in macular fibrosis. We further examined the influence of C3a and C5a in MMT in bone marrow-derived macrophages and in a mouse model of subretinal fibrosis.

Animals
C57BL/6J mice (male and female) aged between 2 and 3 months were used for this study. All animals were housed and bred in the Biological Service Unit of Queen's University Belfast and exposed to a 12-h light/ dark cycle with free access to food and water. All procedures were conducted under the regulation of the United Kingdom (UK) Home Office Animals Scientific Research Act 1986 and in accordance with the ARVO statement for the Use of Animals in Ophthalmic and Vision Research.

Two-stage laser model of subretinal fibrosis
A two-stage laser model was carried out as previously described [9]. Briefly, laser CNV was induced. The settings for the laser were as follows: laser power-250mv, duration-0.1 s, and spot size-100 μm. Three laser spots were delivered per eye. Seven days later, a second laser burn was applied to each CNV lesion, using the same laser configuration.

Inhibition of C3a in an in vivo model of subretinal fibrosis
Subretinal fibrosis was induced using the two-stage laser model as described above. Immediately after the second laser injury (day 0), mice received an intravitreal injection of a C3a blocking agent using the method described previously [27,28]. A second injection was performed 20 days later (day 20).
Fundus images were collected on day 20 using Micron IV system and the Discover 2.2 Programme (Phoenix Technology Group, USA). Fundus fluorescein angiography (FFA) images were collected using the same system. FFA was carried out 5 min after intra-peritoneal injection of 100 μl of 10% sodium fluorescein (Sigma-Aldrich, Gillingham, UK, Cat No: F6377). Exposure level was also kept consistent between animals. ImageJ was used to measure the leakage from each lesion, in a masked fashion. Mean grey value per lesion was compared between traditional laser CNV and two-stage laser groups. The findings were confirmed by a blinded, independent researcher.
Eyes were collected at 30 days post the second laser and fixed in 2% paraformaldehyde (PFA) (Sigma-Aldrich, Cat No: 158127). The size of fibrotic lesion was measured using area measurements of collagen-1 + and Isolectin-B4 + lesion size on retinal pigment epithelium (RPE) flatmounts.

Human samples
Human eye samples with nAMD were obtained from San Diego Eye Bank. This study was carried out within the parameters of the Declaration of Helsinki, and patient tissues were stored in accordance with the Human Tissue Act (2004). The research was approved by the Ethical Review Boards of Queen's University Belfast. The eyes were maintained in formalin. Upon arrival, the eyes were dissected and embedded in paraffin wax. Tissue sections were stored at 4°C in a fridge designated for human tissue samples until required for staining. The results in this paper are representative of a single human sample. Haematoxylin and eosin (H+E) staining was carried out using Sigma-Aldrich H3136 Haematoxylin and BDH/VWR 95057-848 Eosin. Masson's trichrome staining was performed on cryosections and wax sections using the Abcam trichrome staining kit according to the manufacturer's instructions (Abcam, Cambridge, UK, Cat No: ab150686).
Following antigen retrieval, retinal sections were blocked using 5% BSA (Sigma-Aldrich, Cat no: A3803-10G) for 1 h at room temperature. Samples were incubated in primary antibody (Table 1) overnight at 4°C, followed by incubations with secondary antibody (Table  1) in the dark at room temperature for 1 h. The sections were mounted in Vectashield with 4′,6-diamidino-2phenylindole (DAPI; Vector Laboratories, USA, Cat no: H-1200). RPE flatmounts were stained using our protocol, as previously described [29]. Lesion size measurement in RPE flatmount was carried out as previously described [9]. The measurement was confirmed by an independent, blinded researcher.

Immunohistochemistry staining
BMDMs were fixed for 20 min in 2% PFA at room temperature. After washing, the cells were blocked in 1% BSA (Sigma-Aldrich, Cat no: A3803-10G) diluted in PBS with 0.02% Triton-x for 1 h at room temperature, followed by 1-h room temperature incubation with primary antibody (Table 1). Secondary antibody (Table 1) was applied for 1 h room temperature after thorough washes. DAPI (Sigma-Aldrich, Cat No: D9542) was applied after further thorough washes. Antibody conditions are described in Table 1. Fibronectin was also stained in BMDMs using rabbit anti-mouse antibody (Abcam, Cat No: ab2413) followed by Alexa Fluor® 488 AffiniPure Donkey Anti-Rabbit IgG (H+L) (Jackson ImmunoResearch, Cambridge, UK, Cat No: 712-545-150). Incubation with secondary antibody alone was used as negative control. Total cells were counted using DAPI + cells, and myofibroblasts were defined as α-SMA + cells with a fibroblastic cell shape (elongated, stress fibres present). Each treatment group was performed in triplicate, in each independent experiment. Three images per well were taken at random. The data was confirmed by an independent, blinded researcher.

Data analysis
Graph Pad Prism (V6, GraphPad Software, San Diego, USA) was used to create graphs and conduct statistical analysis. Data was tested for normality, and variances were tested to ensure similarity. This was conducted by Shapiro-Wilks test and Bartlett's test. Analysis of statistical significance between two groups was conducted via an independent Student's t test. one-way or two-way ANOVA was used where appropriate. Bonferroni correction was used for multiple comparison testing.

F4/80 + macrophages express α-smooth muscle actin (α-SMA) in the subretinal fibrotic lesion
In the two-stage laser model of subretinal fibrosis, infiltrating F4/80 + macrophages were detected at the lesion site throughout the disease course (Fig. 1). The infiltrating cells were present on the lesion surface as demonstrated in RPE/choroidal/sclera flatmount (Fig. 1a) as well as inside the lesion (Fig. 1b-e). Interestingly, we found that approximately 30% of F4/80 + cells co-express α-SMA (arrows in Fig. 1f and high-magnification images in Fig.  1g). The co-localisation of F4/80 and α-SMA was further confirmed using the Pixel Intensity Spatial Correlation Analysis in ImageJ (Pearson's coefficient = 0.74, n = 3, Additional file 1 Figure S1A). Around the lesion area, some pigmented cells (likely RPE cells) were also α-SMA + (Additional file 2: Figure S2). Our results suggest the To understand if the phenomenon of MMT also exists in human macular fibrosis, we conducted dual staining of Iba-1 and α-SMA in nAMD eyes. H+E and trichrome staining were used to identify subretinal lesion (SRL; Fig. 2a, b). Trichrome staining showed a well-contained lesion with a large amount of collagen fibres (blue), blood vessels and infiltrating cells (Fig. 2b). The infiltrating Iba-1 + cells were detected in the junction between the retina and SRL (blue box 1 with high-magnification image, Fig. 2c) as well as inside the lesion (blue box 2 with high-magnification image, Fig. 2c). Immunofluorescence uncovered multiple α-SMA + cells in SRL (Fig. 2d with high magnification of boxed area in Fig. 2e). Dual staining for α-SMA (red) and Iba-1 + (green) identified several dual positive cells (Fig. 2g, boxed area with high magnification in Fig. 2h, negative control staining in Fig. 2f), making up approximately 20% of infiltrating Iba-1 + cells at the lesion. This data suggests that MMT is present in macular fibrosis secondary to nAMD.

C3a but not C5a induces MMT in bone marrowderived macrophages
Previously, we have found that higher plasma levels of C3a and C5a are related to macular fibrosis in nAMD [24] and C3a is known to be able to induce EMT [26]. We were then interested to know if C3a or C5a contributes to MMT. Flow cytometry confirmed that the BMDMs were double positive for CD11b and F4/80 (data not shown). BMDMs were treated with C3a or C5a for 48-96 h. TGF-β1 (10 ng/ml) was used as a positive control. Phase-contrast microscopic investigation uncovered several elongated stretched cells in the TGF-β1and C3a-treated groups (Additional file 3: Figure S3A, arrows) but not in the C5a-treated group (data not shown). qPCR showed a significant upregulation of α-SMA mRNA in TGF-β-and C3a-treated cells but not in C5a-treated cells (Additional Figure 3B). This was further confirmed by counting α-SMA + cells to identify the percentage of cells which were differentiated myofibroblasts (Additional file 3: Figure S3C, D). Therefore, in the rest of our study, we focused on the effect of C3ainduced MMT in the context of retinal fibrosis.
Together, these data suggest that TGF-β and complement C3a can induce MMT.

C3a induced MMT is mediated through the C3a-C3aR signalling pathway
Macrophages are known to express C3aR [31]. To understand if C3a-induced MMT is mediated through the C3aR signalling pathway, a C3aR-specific antagonist (C3aRA) was used in our study (Fig. 4a).

TGF-β, complement C3, C3a and C3aR are upregulated in the subretinal fibrotic lesion
To understand the molecular pathways involved in MMT in subretinal fibrosis, we examined TGF-β, C3a and C3aR expression in subretinal lesions of the two-stage laser model [9]. TGF-β mRNA was significantly increased in the retina 3 and 25 days post the second laser treatment (Additional file 4: Figure S4A). In addition, TGF-β mRNA was significantly upregulated in the RPE/choroid/sclera layer 25 days post the second laser treatment (Additional file 4 Figure S4B). Confocal microscopy detected TGF-β in normal mouse retina, including RPE cells (Additional file 4 Figure S4C). Increased TGF-β staining was seen in the retina and subretinal lesion 30 days after the second laser treatment (Additional Figure 4D, white arrow). We observed a significant increase in C3 gene expression in the RPE/Choroid/sclera 3 days after the second laser treatment, compared to tissues from matched nonlasered control mice (Fig. 5a). Thirty days post the second laser treatment, complement C3a is present in the retina (Fig. 5b, arrows) and within the subretinal lesion (Fig. 5b, yellow arrow). C3aR was also detected in the lesion (Fig. 5c) and some of the C3aR + cells were F4/80 + macrophages (Fig. 5c). Negative control staining is showed in Fig. 5d. Our results suggest that TGF-β, C3a and C3aR are present within the lesion in this model of subretinal fibrosis. C3a. Image has been cropped for clarity. Data presented as fold change in α-SMA expression compared to housekeeping protein (rab11), mean ± SEM, n = 5-6 samples from 2 independent experiments, **p < 0.01, ***p < 0.005 compared to control untreated cells, Student's t test

The effect of C3a blockade on subretinal fibrosis
To further understand the contribution of the C3a/ C3aR pathway in retinal fibrosis, we used C3a mAb and a C3aR antagonist (C3aRA) in our two-stage laser mouse model of subretinal fibrosis [9]. Clinical examination showed that fluorescence leakage from subretinal fibrovascular lesions was significantly reduced 20 days after C3a mAb treatment (Fig. 6a-c). C3aRA treatment did not significantly reduce fluorescence leakage compared to the vehicle (DMSO) group (Fig. 6a-c).

Discussion
Neovasculature in nAMD often progresses into fibrovascular lesions (macular fibrosis) even after anti-VEGF therapy [32]. Inflammation, in particular the innate immune responses including macrophage infiltration and complement activation, is critically involved in the development and progression of nAMD [21,23,33]. In this study, we Mean ± SEM. *p < 0.05, **p < 0.01. Two-way ANOVA, Bonferroni corrected. e Percentage of α-SMA + cells in BMDM cultures following 96 h of C3a, C3a + DMSO and C3a + C3aRa treatments. n = 3, representative of the two independent experiments. Mean ± SEM. One-way ANOVA with Bonferroni correction, The asterisk is indicated above the group, significance is compared to the control group (untreated BMDMs). Where the asterisk is indicated between two conditions, significance stated is that between the two groups. *p < 0.05, **p < 0.01. Representative images of α-SMA + cells (red) in different groups are shown in the left panel.
show that infiltrating macrophages may participate in fibrovascular lesion development through transdifferentiating into myofibroblasts (i.e. MMT) and TGF-β and complement C3a may contribute to this process.
A persistent low-grade inflammation in the vascular lesion site is believed to be a key driver of macular fibrosis [8]. The inflammatory response can not only damage macular cells (e.g., RPE and photoreceptors), but also lead to the release of various soluble mediators, including proinflammatory and profibrotic factors. These mediators can initiate a cascade within the cellular milieu that leads to the accumulation of ECM, rich in fibrillary collagens and fibronectin in the deposition of diseased blood vessels. Macrophages are known to actively participate in tissue fibrosis [34]. In the disease initiation stage, the classically activated macrophages promote inflammation and tissue damage, whereas in the disease resolving stage, alternatively activated macrophages suppress inflammation, promote tissue repair and wound healing [35]. Dysregulated macrophage function, either in the disease initiation or in the resolving stage may lead to persistent inflammation, pathogenic wound healing and fibrosis. Macrophages may also promote fibrosis through their transition to myofibroblasts. This phenomenon was previously observed in kidney fibrosis [17][18][19]. A key marker of MMT is the co-expression of macrophage marker F4/80 and myofibroblast marker α-SMA. We found around 20% of Iba-1 + macrophages coexpress α-SMA in human nAMD eyes and 30% (of F4/ 80 + macrophages) in our mouse model of subretinal fibrosis. Our results suggest that infiltrating macrophages may participate in macular fibrosis by converting themselves into myofibroblasts.
Besides macrophages, other cells may give rise to myofibroblasts in macular fibrosis. RPE cells have previously been shown to undergo EMT [36], and we observed α-SMA + RPE-like cells in the mouse model of subretinal fibrosis (Additional file 2 Figure S2), suggesting that RPE may contribute to macular fibrosis through EMT. Myofibroblasts may also originate from endothelial cells through endothelial to mesenchymal transition (EndoMT) [8]. Other cells of interest include glial cells, circulating fibrocytes and choroidal stromal cells [8]. Further research is required to fully understand the origin of α-SMA + myofibroblasts in macular fibrosis and the molecular pathways involved in their recruitment and activation.
TGF-β is a well-established profibrotic mediator that can induce mesenchymal transition in a variety of cells including epithelial cells [37], endothelial cells [38] and macrophages [19]. RPE cells are known to constitutively express TGF-β to maintain the immune suppressive microenvironment of the subretinal space [39]. Recently, using single-cell RNA sequencing analysis, we reported that TGF-β1 and TGF-β2 mRNA was expressed predominately by retinal microglia and Müller cells, respectively [40]. In the current study, we found that the expression of TGF-β1 in the retina and RPE was increased 3 and 25 days after the second laser treatment in our model of subretinal fibrosis. TGF-β may be a key driver of MMT in retinal fibrosis.
Another interesting observation of this study is that complement C3a but not C5a induced MMT. Approximately 5-10% of C3a-treated macrophages expressed α-SMA, fibronectin and collagen-1, and this C3a-mediated MMT can be prevented by C3aR antagonist C3aRA. C3a has been reported to induce mesenchymal transition in PTECs [26]. A previous study from our group has reported higher plasma levels of C3a, C4a and C5a in nAMD patients with macular fibrosis [24]. These data suggest that C3a may participate in macular fibrosis in nAMD through induction of MMT. However, in the mouse model of subretinal fibrosis, blocking of C3a only suppressed the vascular component of subretinal lesion, but did not significantly reduce subretinal fibrosis.
C3aRA also failed to suppress subretinal fibrosis. A previous study has shown that blocking C3a reduced CNV lesion size [25] and that C3a treatment increased VEGF production by RPE cells [25]. In addition, we previously observed higher plasma levels of C3a in nAMD patients who were partially responsive to anti-VEGF therapy compared with the full responders [24]. It is possible that the C3a-C3aR pathway may play a more important role in neovascularisation than in subretinal fibrosis. It is also possible that the lack of therapeutic effect is due to limitations of the treatment regime employed in our study. In this study, the intravitreal injections of C3a blockage were conducted at day 0 and day 20 post the second laser. In our experience, multiple intravitreal injection in mouse eyes with an interval time shorter than 2 weeks causes ocular trauma and inflammation that can be severe enough to affect retinal disease [27]. Indeed, the lesion size in our IgG treated group was significantly larger than that of intravitreal injection-free group (Fig.  6g). The 20-day interval between two injections, although minimized injection-related ocular trauma and Fig. 6 The effect of C3a blockade on a mouse model of subretinal fibrosis. Representative fundus (a) and FFA (b) images at 20 days post the twostage laser subretinal fibrosis model from mice that were untreated (control) or treated with rat IgG, or anti-C3a mAb or C3aRA or vehicle (DMSO). c Quantitative measurement of fluorescein leakage at 20 days p.l. expressed as mean of gray value. Mean ± SEM, n = 6 eyes per group, 11-18 lesions per group. One-way ANOVA, Bonferroni corrected, *p < 0.05, vs untreated controls. d-g RPE/chroid flatmounts from 30 days twostage laser model treated with C3a blocking agents were stained for isolectin B4 (d) or collagen-1 (f) and imaged by confocal microscopy. Scale bar = 100 μm. e Quantitative measurement of isolectin B4 + lesion area in different groups. g Quantitative measurement of collagen-1 + lesion area in different groups. Mean ± SEM. Collagen-1: n = 10-20 lesions per group, 3-5 eyes. Isolectin-B4: n = 16-22 lesions per group, 5-7 eyes. *p < 0.05. **p < 0.01. ns = non-significant.one-way ANOVA, Bonferroni corrected inflammation, may result in insufficient intraocular neutralisation of C3a or C3aR. Further studies using controlled drug release device or long-lasting drugs will be needed to address this question.

Conclusions
In conclusion, our study shows that MMT is involved in macular fibrosis secondary to nAMD. TGF-β and complement C3a (but not C5a) are potential inducers of MMT in macular fibrosis. Further studies on the importance of MMT in macular fibrosis, the key triggers of MMT, the macrophage subsets and molecular pathways involved in MMT will be essential to understand if MMT can be targeted for therapy.