The multifaceted role of vitreous hyalocytes: Orchestrating inflammation, angiomodulation and erythrophagocytosis in proliferative diabetic retinopathy

Abstract

Background

Despite great advances in proliferative diabetic retinopathy (PDR) therapy over the last decades, one third of treated patients continue to lose vision. While resident vitreous macrophages called hyalocytes have been implicated in the pathophysiology of vitreoretinal proliferative disease previously, little is known about their exact role in PDR. In this study, we address molecular and cellular alterations in the vitreous of PDR patients as a means towards assessing the potential contribution of hyalocytes to disease pathogenesis.

Results

A total of 55 patients were included in this study encompassing RNA-Sequencing analysis of hyalocytes isolated from the vitreous of PDR and control patients, multiplex immunoassay and ELISA analyses of vitreous samples from PDR and control patients, as well as isolation and immunohistochemical staining of cultured porcine hyalocytes. Transcriptional analysis revealed an enhanced inflammatory response of hyalocytes contributing to the cytokine pool within the vitreous of PDR patients by expressing interleukin-6, among others. Further, increased angiopoietin-2 expression indicated that hyalocytes from PDR patients undergo a proangiogenic shift and may thus mediate the formation of retinal neovascularizations, the hallmark of PDR. Finally, RNA-Sequencing revealed an upregulation of factors known from hemoglobin catabolism in hyalocytes from PDR patients. By immunohistochemistry, cultured porcine hyalocytes exposed to red blood cells were shown to engulf and phagocytose these, which reveals hyalocytes’ potential to dispose of erythrocytes. Thus, our data suggest a potential role for vitreous macrophages in erythrophagocytosis and, thereby, clearance of vitreous hemorrhage, a severe complication of PDR.

Conclusion

Our results strongly indicate a critical role for vitreous hyalocytes in key pathophysiological processes of proliferative diabetic retinopathy: inflammation, angiomodulation and erythrophagocytosis. Immunomodulation of hyalocytes may thus prove an essential novel therapeutic approach in diabetic vitreoretinal disease.

Graphical Abstract

In the left-hand side panel, a human eye of a patient with proliferative diabetic retinopathy with retinal neovascularizations (RNV) growing in the preretinal vitreous is schemed. Hyalocytes, the macrophages of the vitreous, are located mainly in the posterior part of the eye, close to the vitreo-retinal interface and, thus, RNV. Diabetic hyalocytes (on the right) modulate angiogenesis by expressing factors such as interleukin-6 (IL-6) and angiopoietn-2 (ANGPT2), contribute to fibrosis by transdifferentiation in myofiboblasts secreting extracellular matrix (ECM) components, and participate in erythrophagocytosis, as shown in this study

Background

Diabetic retinopathy (DR) is a leading cause of blindness worldwide, with a global prevalence estimated to 93 million people in 2012 [1] projected to nearly triple in the USA by 2050 [2]. Proliferative DR (PDR), the advanced stage of the disease, is marked by areas of reduced retinal perfusion that generate an uncontrolled release of proangiogenic growth factors eventually leading to the formation of vulnerable new blood vessels called “retinal neovascularizations” (RNV, [3]). The compensatory attempts at revascularization of the ischemic retina generally result in unfavorable angiogenesis towards vitreous [4], the stimulus for this pathological preretinal formation of contractile vascular membranes remaining the matter of debate. Evidentially, a major driver of neovascularization in PDR is the vascular endothelial growth factor (VEGF) targeted effectively in routine treatment of diabetic macular edema (DME) and PDR by the application of anti-VEGF antibodies or decoy antibody receptors [5]. More recently, faricimab, a bispecific antibody acting as an inhibitor of both VEGF and angiopoietin-2 (ANGPT2), was introduced to enhance treatment for neovascular eye disease [6]. Those successes notwithstanding, other pathways likely contribute to PDR pathogenesis, as the disease often progresses even under continuous anti-VEGF therapy [7] and addition of ANGPT2 inhibition has improved outcomes in only a minority of DME patients switched from aflibercept [8]. A deeper understanding of additional mechanisms underlying PDR development is therefore critical for unveiling alternative therapeutic options, given the disease’s growing socio-economic impact.

Clinical evidence clearly demonstrates that the vitreous is critically involved in PDR, as florid preretinal neovascularization in PDR is mitigated following posterior vitreous detachment (PVD, [9]) and generally does not recur post-vitrectomy [10]. It appears likely that additional components of the vitreous, other than the supporting structure of vitreal collagen fibers, are involved. Recent evidence suggests that hyalocytes, the vitreous resident myeloid cells [11, 12], may be essential participants in the course of proliferative vitreoretinal disease. Hyalocytes are distributed largely within the vitreous cortex abutting the retinal surface [13], in immediate proximity to RNV sites. We have previously demonstrated in a RNV mouse model that myeloid cells cluster in the vicinity of RNV [14] and may influence their formation [15]. Moreover, transcriptional and single-cell protein analysis of human RNV have revealed an abundance of antigen-presenting cells, most likely hyalocytes, which bear the potential of myofibroblastic transdifferentiation [16]. Finally, advanced in vivo imaging techniques have identified macrophage-like cells (MLC) accumulating around RNV in the vitreoretinal interface (VRI) of DR patients [17]. While these observations are suggestive, the exact role of vitreous hyalocytes in the pathophysiology of PDR remains largely unknown.

In this study, we report a high-dimensional molecular characterization of hyalocytes from the diabetic vitreous compared to controls performed with the aim to better assess their potential role in PDR pathophysiology. Our transcriptional analysis reveals enhanced inflammatory responses by hyalocytes, which appear to contribute, as a so far underestimated factor, to the cytokine pool within the PDR vitreous. The data further indicate an expression shift in hyalocytes in PDR towards a proangiogenic phenotype, suggesting they may mediate RNV formation. This could explain the misdirected growth of RNV towards the preretinal vitreous and alleviated RNV formation following vitrectomy. Finally, our data suggest a role of vitreous macrophages in erythrophagocytosis and removal of vitreous hemorrhage (VH) debris. The results of our study may pave the way for novel immunomodulatory approaches in the treatment of end-stage diabetic vitreoretinal disease.

Materials and methods

Patients’ characteristics

A total of 55 patients were included in this study. Only patients with no history of previous vitreoretinal surgery, concurrent vitreoretinal disease, or anti-VEGF therapy within the last three months were enrolled. Three control patients had type II diabetes without evidence of DR.

Adaptive optics scanning light ophthalmoscopy (AOSLO) imaging was performed on two patients (a 32-year-old healthy control [18] and a 26-year-old PDR patient [17]), as described previously [18]. The other 53 patients underwent vitrectomy for complications of PDR, macular pucker (MP) or macular hole (MH) between 2018 and 2021 (demographics and clinical characteristics summarized in Tables 1 and 2). RNA-Sequencing (RNA-Seq) analysis was performed on 21 samples from 30 patients (Tbl. 1). In order to exclude possible confounding effects of Red Blood Cell (RBC) lysis buffer treatment on the expression profile of vitreous hyalocytes, four pooled samples were processed for a preliminary analysis. For each final sample, the vitreous specimens of three to four MP or MH patients were pooled, divided in two equal halves and either subject to lysis (“+ lysis”, similar to diabetic samples in the main analysis) or analyzed without lysis treatment (“- lysis”, Tbl. 1). Eight samples from eight patients with PDR (mean age 57 ± 15.4 years) were compared to nine control samples from nine patients (MP and MH, mean age 75 ± 6.2 years) in the main analysis.

Table 1 Characteristics of patient samples included in RNA-Sequencing analysis

Table 2 Characteristics of patient samples included in protein analysis

For protein analysis by enzyme-linked immunosorbent assay (ELISA) or multiplex immunoassay, 31 undiluted vitreous samples were obtained at the start of vitrectomy (before intraocular fluid infusion) from the mid-vitreous of 23 patients (samples from eight patients were processed for both readouts) and centrifuged at 500 x g for 20 min at 4 °C. Corresponding plasma specimens were collected before or during surgery by peripheral venous puncture, centrifuged (3000 x g for 15 min at 4 °C) and, like vitreous samples, stored at -80 °C until processing.

Ethics approval was granted by the local Ethics Committee and a written informed consent was obtained from each patient prior to surgery. All research adhered to the tenets of the Declaration of Helsinki.

Fluorescence-activated cell sorting (FACS)

Vitreous tissue samples were collected in vitrectomy bags during surgery and processed for cell isolation within two hours of resection, according to a previously published protocol [11] adapted for the specificities of diabetic vitreous. For half of the samples in the preliminary analysis (see Table 1) and for diabetic samples, 1 ml of RBC lysis buffer (Thermo Fisher Scientific) was added for erythrocyte lysis. Samples were stained for cluster of differentiation (CD) 45 (BV421, anti-human, 1:100, BioLegend), CD11b (FITC, anti-human, 1:100, BioLegend), CX₃C motif chemokine receptor 1 (CX₃CR1, PE-Cy7, anti-human, 1:200, BioLegend), and with the anti-human Mature Macrophages (MatMac) antibody, an ED2-like (ectodermal dysplasia 2) marker for resident macrophages (eFluor660, anti-human, 1:100, eBioscience). Finally, cells were processed for sorting on the MoFlo Astrios EQ Cytometer (Beckman Coulter). Hyalocytes were isolated as CD45⁺CD11b⁺CX3CR1⁺MatMac⁺ cells.

Total RNA extraction and RNA-Seq library preparation

RNA extraction, library preparation and RNA-Seq were conducted at the Genomics Core Facility “KFB - Center of Excellence for Fluorescent Bioanalytics” (University of Regensburg, Germany), according to previously published protocols [11, 16].

Bioinformatics

Sequencing data were analyzed on the Galaxy web platform (usegalaxy.eu [19]), as previously described by our group [20]. Transcripts with log2 fold change (log2FC) > 2 or < -2 and adjusted p-value < 0.05 were considered differentially expressed genes (DEG). Gene ontology (GO) analysis for clusters related to biological processes (BP) was performed based on all DEG in hyalocytes from PDR patients vs. controls.

Multiplex immunoassay

Cytokine levels were measured in vitreous and corresponding plasma samples from nine PDR (mean age 45 ± 14.7 years) and 10 MH (control) patients (mean age 62 ± 14.8 years, Tbl. 2). A multiplex electrochemiluminescence assay (V-Plex Human Biomarker 54-Plex Kit, Meso Scale Discovery) was used according to manufacturer’s instructions. By combination of electrochemiluminescence and multi-array technologies, the levels of 54 cytokines were simultaneously measured including interleukin (IL)-6, IL-8, IL-15, monocyte chemoattractant protein 1 (MCP-1), placental growth factor (PlGF), VEGF-A. This panel was selected to assess the expression of factors previously identified as relevant in the progression of PDR or DEG in hyalocytes from PDR patients based on the RNA-Seq results of this study. For statistical analysis, all values below the detection limit were assigned to half the respective values of the detection limit.

Enzyme-linked immunosorbent assay (ELISA)

Vitreous and plasma levels of ANGPT2 were measured in samples from six PDR (mean age 44 ± 18.2 years) and six MH (control) patients (mean age 66 ± 8.5 years, Tbl. 2) with the Human Angiopoietin-2 Quantikine ELISA (R&D Systems) according to manufacturer’s instructions.

Cultivation and staining of porcine hyalocytes

Porcine eyes were obtained from a local abattoir. In terms of development, anatomy and morphology, these are considered representative for the human situation [21]. Vitreous was extracted from six eyes and placed in Dulbecco’s Modified Eagle Medium (DMEM, Gibco). On the next day, peripheral blood was obtained from healthy human donors by venous puncture and centrifuged at 3000 rpm for 10 min at 4 °C to obtain erythrocytes. One drop of erythrocyte concentrate was added to each vitreous sample. On day 3, specimens were fixed in methanol for 10 min at -20 °C prior to blocking for one hour at room temperature (RT) in a solution of 1% bovine serum albumin (Roth) and 5% normal donkey serum (Biozol) in phosphate-buffered saline Triton-X 0.1% (Gibco). For immunohistochemistry (IHC), cells were incubated with ionized calcium-binding adapter molecule 1 (IBA-1, abcam) and CD235a (Thermo Fischer Scientific) primary antibodies for one hour at RT. Primary antibodies were omitted for negative control. Incubation with secondary antibodies (Alexa 647, Life Technologies; Alexa 488, Invitrogen) and Phalloidin 5(6)-Tetramethylrhodaminisothiocyanate (TRITC, Sigma-Aldrich) was performed for one hour at RT. Nuclei were counterstained with 4′,6-Diamidin-2-phenylindol (DAPI) prior to embedding in Fluorescence Mounting Medium (Agilent Dako). Representative images were taken on a Leica TCS SP8 Confocal System coupled to a Leica DMi8 inverted microscope.

Statistical analysis

Statistical analysis was performed using GraphPad Prism (GraphPad Software, Version 10.2.3). For multiplex immunoassay analysis, one-way ANOVA and subsequent multiple comparison testing was performed. For ANGPT2 expression analysis by ELISA, a Mann-Whitney test was applied. The following significance levels were considered: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Results

Analysis workflow

Diagnosis for the 55 included patients was based on a thorough funduscopic exam (Fig. 1A), spectral domain optical coherence tomography (OCT) and fluorescence angiography (FA, HRA2, Heidelberg Engineering) for PDR patients (Fig. 1A’). Samples from 30 patients undergoing vitrectomy for PDR, MP or MH were processed for RNA-Seq analysis (Tbl. 1). Hyalocytes were isolated from vitreous tissue by flow cytometry (Figs. 1B, [11]) specifically targeting resident immune cells by utilizing, among others, the MatMac marker for tissue-specific macrophages. This antibody was applied to exclude any potential contamination with blood-derived monocytes owing to possible surgically induced microbleeds, as infiltrating leukocytes generally lack the antigen [22, 23]. In the following, sorted hyalocytes were analyzed by RNA-Seq (Fig. 1C). Staining by IHC was performed on hyalocytes (Fig. 1D) isolated from porcine vitreous tissue. MLC in the VRI of a healthy control and a PDR patient imaged by AOSLO are shown in Fig. 2 and Additional File 1.

Fig. 1

Experimental workflow. Diagnosis was based on a thorough funduscopic exam documented by fundus photography (CF, color fundus, A.), spectral domain optical coherence tomography (OCT) and fluorescence angiography (FA, A’.), for proliferative diabetic retinopathy (PDR) patients (black areas are not sufficiently supplied with blood and therefore ischemic). B. Vitreous samples were obtained by vitrectomy from PDR and control patients. The black rectangle is to illustrate the part of the bulb imaged by CF in A. and FA in A’. Hyalocytes were isolated from vitreous tissue by flow cytometry (fluorescence-activated cell sorting, FACS) as CD45⁺CD11b⁺CX3CR1⁺MatMac⁺ cells and were further processed for C. RNA extraction and library preparation. RNA-Sequencing (RNA-Seq) data were analyzed bioinformatically. D. For visualization of hyalocytes, porcine vitreous tissue was cultured and stained by immunohistochemistry (IHC). ONH. Optic nerve head. RNV, retinal neovascularization

Fig. 2

Adaptive optics scanning light ophthalmoscopy (AOSLO) imaging of macrophage-like cells in the vitreoretinal interface. Two adjacent hyalocytes with different movement behaviors over 30 min of AOSLO imaging in (top row, A1-A4; modified from [18]) a 32-year-old healthy control and (bottom row, B1-B4) a 26-year-old proliferative diabetic retinopathy (PDR) patient. A5 & B5. Chromo-temporal map composites of the four time points over 30 min, which indicate movement of cell bodies and processes across the imaging period. Time of acquisition in the lower-left corner is displayed in mins:secs. A video of the entire imaging region of interest, which includes these four cells, is shown in Additional File 1

Effect of RBC lysis on the transcriptional profile of human hyalocytes

In our study, RBC lysis proved indispensable for isolation of hyalocytes from the blood-tinged diabetic vitreous, but unnecessary to process control vitreous. Furthermore, a lysis-induced reduction of cell count in single non-diabetic samples under the limit of feasible transcriptional analysis was observed in preparation for this study.

Although lysed samples in our preliminary analysis showed a trend to a lower cell count (mean count: 100.5 ± 26.4 cells vs. 159.0 ± 75.1 cells, p = 0.22) and lower RNA concentration (mean concentration: 75.5 ± 25.6 pg/µl vs. 91.5 ± 60.6 pg/µl, p = 0.65), no distinct effect of lysis on the RNA profile of hyalocytes could be inferred, as highlighted by the conducted principal component analysis (PCA, Additional File 2A) and heatmap visualization (Additional File 2B). Based on the negligible effect of RBC lysis treatment on hyalocyte transcriptional data, further analysis was performed comparing lysed diabetic samples and non-lysed control samples (MP and MH). MP and MH hyalocytes are considered comparable in their transcriptional signature [11], which prompted us to process them together.

Transcriptional profiling of human diabetic hyalocytes

Diabetic samples tended to yield a higher amount of hyalocytes in comparison to control samples (1015.8 ± 703.6 vs. 454.8 ± 246.8 cells, p = 0.063), while mean concentration of RNA extracted from isolated hyalocytes was comparable (115.0 ± 99.3 vs. 135.7 ± 72.5 pg/µl, p = 0.636). A mean total number of 43.7 million (± 4.3) raw reads per sample was obtained from diabetic hyalocytes, while 37.8 million (± 6.5) raw reads per sample were detected for control hyalocytes (p = 0.045).

A total of 43,278 transcripts with at least one read in at least one sample were obtained. PCA revealed clear separation of diabetic and control hyalocytes’ transcriptomes on the first two principle components indicating substantial differences between the two groups (Fig. 3A). Despite within-group heterogenity visible in the generated DEG heatmap, a distinct pattern of transcriptome differences was evident between both entities (Fig. 3B). Comparative analysis of the transcriptome of diabetic and control hyalocytes revealed 126 differentially upregulated and 222 differentially downregulated genes in diabetic samples relative to controls (Fig. 3C). Notably, among the most strongly upregulated genes in diabetic hyalocytes ranked a number of factors implicated in inflammatory (cathepsin (CTS) B (CTSB), CTSL, CTSD and S100 calcium-binding protein A8 (S100A8)), but also anti-inflammatory responses (ferritin light chain, FTL), or both (legumain (LGMN)).

Fig. 3

Transcriptional signature of diabetic hyalocytes. (A) A Principal Component Analysis (PCA) plot demonstrates the distribution of hyalocyte samples isolated from the vitreous of patients with PDR (red dots) and control samples (green dots). (B) Heatmap of differentially expressed genes (DEG, log2FC > 2, adjusted p value < 0.5, sorted according to the logFC) between hyalocytes isolated from the vitreous of PDR patients and control hyalocytes. Color coding of the transcripts according to the z-score (deviation from a gene’s mean expression in standard deviation units). (C) Readplot of DEG in diabetic hyalocytes (upregulated genes in red, downregulated genes in green, not differentially expressed genes in grey; the most strongly expressed genes from each group are labeled; also labeled are the biliverdin reductases A and B, BLVRA and BLVRB). (D) Top 10 most significantly upregulated Gene Ontology (GO) biological process (BP) clusters on the basis of all DEG in diabetic hyalocytes. Color coding of the dots according to the adjusted p value, size of the dots according to the count of transcripts associated with the respective GO term. (E) Top five most highly expressed transcripts in the three most enriched GO terms from D

To gain more insight into BP diabetic hyalocytes are involved in, we performed a GO cluster analysis of all DEG (Fig. 3D). According to our data, DEG mostly contributed to processes like “chemotaxis” (GO:0006935) and “leukocyte migration” (GO:0050900), but also “regulation of cell adhesion” (GO:0030155) and “collagen metabolic process” (GO:0032963). S100A8 was the most highly expressed factor in the cluster “chemotaxis” and the second-most highly expressed factor in “leukocyte migration” (Fig. 3E).

Since angiogenesis and immune response are key processes in the pathogenesis of PDR [24], we took a closer look at genes involved in both processes. We found heme oxygenase 1 (HMOX1) involved in hemoglobin catabolism, but also IL6 and angiopoietin-2 (ANGPT2) to be among the most strongly expressed factors in “angiogenesis” in hyalocytes from PDR patients (Fig. 4A). For “immune response”, besides the already established factors FTL, CTSB, CTSL, CTSD and S1008, matrix metalloproteinase 9 (MMP9) was among the PDR-associated factors in hyalocytes that stood out (Fig. 4B).

Fig. 4

“Angiogenesis” and “immune response”-associated genes expressed in hyalocytes from proliferative diabetic retinopathy (PDR) patients. (A) Readplot of differentially expressed genes (DEG) in diabetic hyalocytes (upregulated genes in red, downregulated genes in green, not differentially expressed genes in grey; factors associated with the Gene Ontology (GO) biological process (BP) “angiogenesis” are highlighted in yellow; the most strongly upregulated “angiogenesis”-related genes are labeled). (B) Readplot of DEG in diabetic hyalocytes (upregulated genes in red, downregulated genes in green, not differentially expressed genes in grey; factors associated with the GO BP “immune response” are highlighted in yellow; the most strongly upregulated “immune response”-related genes are labeled). (C) Left-hand side panels show boxplots for multiplex immunoassay analysis of interleukin (IL)-6, placental growth factor (PlGF), IL-8, monocyte chemoattractant protein-1 (MCP-1) and IL-15 protein expression or enzyme-linked immunosorbent analysis (ELISA) of angiopoietin-2 (ANGPT2) protein expression in PDR vitreous and plasma and control (Ctrl) vitreous and plasma samples. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (one-way ANOVA and consequent analysis of multiple comparisons conducted for multiplex immunoassay analysis and Mann-Whitney-Test performed for ELISA analysis). Corresponding gene expression analysis for IL6, PlGF, IL8, MCP1 (= CCL2), IL8 (= CXCL8) and ANGPT2 in PDR and control hyalocytes is shown on the right, respectively

Protein analysis

According to the multiplex immunoassay, IL-6, which was overexpressed in PDR hyalocytes on the RNA level (log2FC = 4.24, adjusted p = 0.01), was also strongly enriched in PDR vitreous in comparison to control vitreous and when compared to corresponding plasma on the protein level (Fig. 4C). PlGF was also significantly upregulated in PDR vitreous in comparison to both control vitreous and corresponding plasma. However, these changes did not correspond to a significant enrichment of the respective gene in diabetic hyalocytes (log2FC = 0.83, adjusted p = 0.74, Fig. 4C). The same was the case for IL-8 (log2FC = 0.17, adjusted p = 0.97), MCP-1 (log2FC = 0.25, adjusted p = 0.95) and IL-15 (log2FC = -1.00, adjusted p = 0.72, Fig. 4C). VEGF was detected on the protein level by two assays, which both demonstrated an upregulation in the diabetic vitreous when compared to control vitreous (**p < 0.01). In one of the assays, an upregulation against corresponding plasma of PDR patients was detected (**p < 0.01). Of note, gene expression of VEGFA did not differ significantly between hyalocytes isolated from the PDR and control vitreous, indicating that other ocular cell types are likely the source of increased VEGF protein levels in the diabetic vitreous.

As assessed by ELISA, ANGPT2, a DEG in PDR hyalocytes when compared to control hyalocytes (log2FC = 3.57, adjusted p = 0.01), was significantly upregulated in the PDR vitreous in comparison to control vitreous, too (Mann-Whitney test, Fig. 4C).

Erythrophagocytosis in hyalocytes

According to our transcriptional analysis, HMOX1, an important factor in heme catabolism [25, 26], was one of the most highly expressed genes in diabetic hyalocytes when compared to control hyalocytes. This prompted us to elucidate further marker genes of erythrophagocytosis and iron regulator genes [25, 27] in the data set. Beside FTL and HMOX1, ferritin heavy chain 1 (FTH1) was also significantly upregulated in diabetic hyalocytes. Other factors, such as the biliverdin reductases A and B (BLVRA and BLVRB), showed a trend (BLVRA: log2FC = 1.78, adjusted p = 0.30, BLVRB: log2FC = 1.42, adjusted p = 0.32) towards an enrichment in PDR hyalocytes (Figs. 2C and 5A).

Fig. 5

Erythrophagocytosis in hyalocytes. (A) Boxplots of the expression of marker genes of erythrophagocytosis and iron regulator genes in proliferative diabetic retinopathy (PDR, red) and control (green) hyalocytes. *p < 0.05. (B) Immunohistochemical staining of cultured porcine hyalocytes for ionized calcium-binding adaptor molecule 1 (IBA-1), cluster of differentiation 235a (CD235a) and Phalloidin. Nuclei are counterstained with 4′,6-Diamidin-2-phenylindol (DAPI). Within the cytoplasm of hyalocytes, vesicles containing CD235a- (= erythrocyte debris), but also DAPI-stained particles (= digested nuclei) were observed. Scale bar corresponds to 50 μm

In order to assess the potential of hyalocytes to phagocytose erythrocytes, as suggested by our sequencing results, we next examined cultured porcine hyalocytes. Hyalocytes seemed to engulf erythrocytes, which could clearly be distinguished within the Phalloidin-stained cytoskeleton of the immune cells. Within the cytoplasm of IBA1-positive hyalocytes, we observed vesicles containing CD235a-stained, most probably erythrocyte debris, but also DAPI-positive particles to be considered digested nuclei (Fig. 5B). Negative controls are shown in Additional File 3 (for IBA-1: A., for CD235a: B.). These data suggest that hyalocytes have the potential to remove damaged erythrocytes, which may be of critical importance for diabetic VH clearance.

Discussion

More than one third of PDR cases deteriorate despite complete panretinal photocoagulation or continuous anti-VEGF treatment [28]. Thus, investigation of the molecular and cellular mechanisms underlying end-stage DR pathophysiology is indispensable to prevent severe vision loss. In this study, we performed an in-depth transcriptional characterization of hyalocytes isolated from the vitreous of PDR patients to assess the role of this specialized immune cell population in disease progression. We demonstrate that hyalocytes contribute to a proinflammatory and proangiogenic milieu in the diabetic vitreous by expressing several cytokines but may also play a role in VH clearance. In addition to a presumed role in tractional retinal detachment (TRD) by hyalocyte-to-myofibroblast transdifferentiation, as suggested by us previously [16], our current data implicate the importance of hyalocytes in angiogenesis and red blood cell cleanup, two further hallmarks of vision-threatening PDR.

The transcriptional analysis of 43,278 genes revealed a distinct expression signature of diabetic hyalocytes when compared to control hyalocytes. GO cluster analysis demonstrated that DEG in diabetic hyalocytes mainly contribute to processes like “chemotaxis” and “leukocyte migration” highlighting the importance of innate immune responses in the course of severe vitreoretinal disease. The more active immunological state of PDR hyalocytes was supported by AOSLO imaging of hyalocytes, which in PDR appeared amoeboid-shaped in comparison to the slender, ramified and quiescent MLC in a healthy patient. Besides, DEG were enriched in processes like “regulation of cell adhesion” and “collagen metabolic process”, which suggests an involvement of diabetic hyalocytes in wound repair and would be in line with previous findings on hyalocyte-to-myofibroblast transdifferentiation in PDR [16].

Next, we examined closely two key pathways in PDR pathophysiology – angiogenesis and immune response. In contrast to VEGFA, IL6 and ANGPT2 were among the most strongly enriched factors for the term “angiogenesis” in hyalocytes from PDR patients suggesting these cells as an origin for proinflammatory and proangiogenic cytokines in the diabetic vitreous. IL-6 has been previously shown to be enriched in the diabetic vitreous [29], which may correspond to disease severity [30]. In our study, IL-6 was strongly overexpressed in PDR vitreous samples in comparison to control vitreous and corresponding plasma, which implies a local production by vitreal and/or retina cells. The potential of IL-6 signaling inhibition is currently being explored in experimental and clinical trials for the treatment of DR-related complications [31]. ANGPT2 also lined among the “angiogenesis”-related DEG in diabetic hyalocytes. The factor is known from routine clinical practice with the application of the faricimab antibody [32]. An overexpression of ANGPT2 in the diabetic vitreous on the protein level has been previously demonstrated [33] and was confirmed here. For other analytes, too, such as PlGF, IL-8, MCP-1, and IL-15, a strong upregulation in PDR vitreous samples when compared to control vitreous and corresponding plasma was detected, which is in line with previously published data [29, 34, 35]. However, these factors were not significantly enriched in diabetic hyalocytes implying alternative cell populations as the source for their abundance in the vitreous. It should further be critically considered that various cells of the retina known to express IL-6 (fibroblasts and pericytes) and ANGPT2 (endothelial cells, [36]) in the steady state more pronouncedly than hyalocytes may be the major contributors of both factors to the diabetic vitreous.

Among the 126 differentially upregulated genes in hyalocytes from PDR patients, several proinflammatory factors, such as CTSB, CTSL, CTSD and S100A8, stood out. Cathepsins have previously been implied in mediating an inflammatory response in macrophages, e.g. in carotid plaques [37], adipose tissue [38] and inflammatory bowel disease [39]. S100A8, together with S100A9, encodes for the S100A8/A9 protein complex calprotectin and its expression is upregulated in human choroidal neovascularization tissue [40]. Myeloid cells including macrophages express S100A8/A9 constitutively, while in inflammation, secretion of the protein is enhanced and known to stimulate cytokine release [41]. Inhibition of S100A8/A9 alleviates excessive cytokine production, which reveals its potential as a therapeutic target [42]. In addition, FTL, which has previously been determined as the second-most prominent transcript in human control hyalocytes [11], ranked among the top expressed factors in diabetic hyalocytes, also, with a much higher expression in the disease state. FTL encodes the light subunit of the ferritin protein, the main iron storage in the human body previously associated with anti-inflammatory responses in murine macrophages [43]. Since an iron overload and subsequent susceptibility to oxidative damage has been described for the PDR vitreous [44], it is tempting to speculate about a role of hyalocyte-derived FTL in neuroprotective iron reduction in the vitreous.

On another note, potential deleterious effects of activated hyalocytes in PDR conveyed by the expression of inflammatory and angiogenic factors such as IL-6 and ANGPT2 may well be in contrast to beneficial features of this cell population known, similarly to microglia, for its dual nature [45]. Along these lines, our transcriptional analysis revealed an upregulation of factors known from hemoglobin catabolism, such as HMOX1, BLVRA and BLVRB, in diabetic hyalocytes. The notion of hyalocytes’ phagocytic activity has been suggested as early as in the middle of the 20th century by Hamburg, who assumed that hyalocytes contribute to clearance of metabolic products [46]. By state-of-the-art imaging, hyalocytes have been shown here and previously [18] to continuously scan the environment with their protrusions in anticipation of harmful signals. Upon detection of danger, hyalocytes are likely to migrate and phagocytose the pathogen, akin to microglia of the central nervous system and in accordance with recent evidence that hyalocytes express factors important for phagocytosis, such as MERTK (MER proto-oncogene, tyrosine kinase), CD74 and MHCII-associated genes (major histocompatibility complex class II, [11]). To assess hyalocytes’ potential to phagocytose erythrocytes, we exposed cultured porcine hyalocytes to erythrocytes and subsequently stained them for CD235a, an abundant protein in erythrocyte membranes. Our data demonstrate the capacity of hyalocytes to engulf and dispose of erythrocytes, a process known as erythrophagocytosis. Hereby, damaged/senescent erythrocytes are removed from circulation primarily by macrophages in the spleen, liver, and bone marrow [47]. For erythrocytes, CD47 is known as a “marker of self” preventing their premature clearance [48]. Further studies are needed, in order to elucidate if “eat me/don’t eat me” signals [49] to hyalocytes are also conveyed CD47-dependently and how hyalocytes decide whether they are going to embrace their advantageous functions with erythrophagocytosis, or involve in pernicious cytokine release in vitreoretinal disease.

Interpretation of our data is limited by several factors, for instance the use of samples from patients with vitreoretinal disease, namely MP or MH, as controls. However, MP and MH represent the most physiological fresh samples that can be obtained in clinical routine. Another important limiting factor in our study is the age difference between PDR and control patients. While MP and MH are regarded age-related disorders of the VRI, severe PDR commonly affects younger patients. Differences in age distribution in this study should therefore be critically considered. Further, it was not possible to obtain detailed information on the preoperative extent of avascularity and degree of neovascularization for PDR patients, since FA of sufficient quality was not feasible for the majority of patients due to view obscuration by VH. Lastly, the young age of pigs serving as a source for our in vitro model could also represent a confounder. However, since the vitreous of 8-10-month-old pigs as examined in our study, an age that in theory corresponds to human 18 years, has been reported to theoretically possess the viscoelastic properties of adult human vitreous [50], we consider this limitation negligible.

Conclusions

In conclusion, transcriptional analysis of diabetic hyalocytes, protein analysis of PDR vitreous and immunohistochemical studies on cultured hyalocytes in our work reveal an enrichment of proinflammatory and proangiogenic factors, such as IL6 and ANGPT2, in hyalocytes from PDR patients, which is conveyed in an abundance of both factors in the diabetic vitreous on the protein level. Our data further suggest an important role of hyalocytes in erythrophagocytosis, which may be critical in VH clearance in PDR, especially in non-vitrectomized eyes. As hyalocytes have previously been shown to transdifferentiate to myofibroblasts, this unique cell population of the vitreous may play a role in both critical complications of PDR: vitreous hemorrhage and tractional retinal detachment. Immunomodulation of hyalocytes may thus prove an essential novel therapeutic approach in diabetic vitreoretinal disease.