Incorporating microglia‐like cells in human induced pluripotent stem cell‐derived retinal organoids

Abstract Microglia are the primary resident immune cells in the retina. They regulate neuronal survival and synaptic pruning making them essential for normal development. Following injury, they mediate adaptive responses and under pathological conditions they can trigger neurodegeneration exacerbating the effect of a disease. Retinal organoids derived from human induced pluripotent stem cells (hiPSCs) are increasingly being used for a range of applications, including disease modelling, development of new therapies and in the study of retinogenesis. Despite many similarities to the retinas developed in vivo, they lack some key physiological features, including immune cells. We engineered an hiPSC co‐culture system containing retinal organoids and microglia‐like (iMG) cells and tested their retinal invasion capacity and function. We incorporated iMG into retinal organoids at 13 weeks and tested their effect on function and development at 15 and 22 weeks of differentiation. Our key findings showed that iMG cells were able to respond to endotoxin challenge in monocultures and when co‐cultured with the organoids. We show that retinal organoids developed normally and retained their ability to generate spiking activity in response to light. Thus, this new co‐culture immunocompetent in vitro retinal model provides a platform with greater relevance to the in vivo human retina.


| INTRODUC TI ON
Microglia are the tissue-resident macrophages of the central nervous system, including the retina. They arise from a distinct population of primitive myeloid progenitors in the yolk sac prior to the onset of blood circulation. 1 Histological and more recently transcriptomic analyses of human foetal retina show that microglia migrate into the eye between 6 and 17 weeks of gestation. [2][3][4] Microglial function in the retina extends beyond the concept of immune defence to their role in development, homoeostasis and disease. In the developing retina, they are involved in shaping neuronal organization, formation and remodelling of synapses and development of vasculature. 5 This is achieved by maintaining a fine balance between phagocytosis of dying cells and developing neurons, and trophic support to induce cell proliferation and survival. [6][7][8] In homoeostasis and disease microglia use cytokines as a mode of communication in the microglia-Muller glia crosstalk, which is essential for mediating adaptive responses following retinal injury. 9 Disruption to homoeostasis could be associated with ageing and may lead to neurodegeneration. 10 Changes in microglial cells have also been shown to be a contributing factor to pathophysiological changes associated with several diseases affecting the retina, including diabetic retinopathy, 11 inherited retinal degeneration 12,13 and age-related macular degeneration. 14 There are currently no adequate human in vitro models that allow the study of the interaction of neural retinal cells and microglia and animal models do not fully recapitulate human retinal physiology. Retinal organoids derived from human induced pluripotent stem cells (hiPSCs) offer a platform to study disease mechanisms and provide a testing tool for drug discovery. [15][16][17][18][19][20] A potential limitation of the current retinal organoid technology is the lack of the immune cell component, which has been suggested as an underlying factor for the compromised inner retinal lamination. 21 Since microglia arise from non-neural lineage, they are unlikely to develop in situ under the existing differentiation conditions. A recent paper has described the development of innate microglia within cerebral organoids; however, their culture conditions included matrigel embedding and reducing the levels of some neuroectodermal stimulants, which encouraged the emergence and maturation of mesodermal progenitors and their differentiation to microglia-like cells. 22 Our culture conditions are optimized to direct differentiation of hiPSCs to retinal lineages and are not conducive to microglia emergence, hence, to improve the current system, 20 we generated a co-culture model of hiPSC-derived retinal organoid and microglia-like cells and evaluated its effect on developmental potential.
After day 120, retinoic acid was removed from culture medium. . The data were analysed using the QuantStudio™ software (Life Technologies) and relative gene expression was determined using the 2 −ΔΔC t method using GAPDH as a housekeeping gene. The list of oligonucleotides is shown in Table S1.

| Transmission electron microscopy
Retinal organoids were fixed with 2% glutaraldehyde in 0.1 M sodium cacodylate buffer. Transmission electron microscopy including all cell processing was performed at Newcastle University Electron Microscopy Research Services. Samples were post fixed in 1% osmium tetroxide, dehydrated in gradient acetone, and embedded in epoxy resin. Ultrathin 70 nm sections were picked up on copper grids, stained with uranyl acetate and lead citrate and imaged using a Philips CM100 transmission electron microscope with highresolution digital image capture.

| Single cell (sc) RNA-Seq
Retinal organoids were dissociated to single cells using the Uniform manifold approximation and projection (UMAP) was used to visualize the clusters. The Wilcoxon test was used to identify marker genes for each cluster.

| Electrophysiological recordings
Electrophysiological recordings were performed as described in Dorgau et al. 25  Extraction of spikes from raw traces was done using a quantilebased event detection 26 and single unit spikes were sorted by an automated spike sorting method for dense, large-scale recordings. 27 Firing rate analyses and statistical significance (Mann-Whitney test) were evaluated using Matlab (Mathworks) and Prism (GraphPad).
Retinal ganglion cells were considered responsive if they changed their spiking activity by at least 25% (increase or decrease) during 30 s after WLP onset compared to the similar time window before the light stimulus (dark condition). The mean % change (±SEM) in activity between windows was calculated and plotted using Prism (GraphPad).

| Differentiation and functional characterization of hiPSC-derived microglia-like cells
Microglia-like cells (iMG) were generated from hiPSCs by first deriving cells of mesodermal lineage following myeloid differentiation towards microglial progenitors ( Figure 1A). 23  to display ramified morphology and were positive for the microglial marker IBA1 ( Figure 1C). Although microglial transcriptomic signature can be region-specific and dynamic, a unique microglial signature was defined by Butovsky et al. 28 We compared our iMG cells to a primary microglial cell line and found that the cells had comparable or higher expression levels of TAM (TYRO3, AXL and MERTK)related genes MERTK, PROS1 and GAS6, as well as others including TMEM119 and TREM2 ( Figure 1D). We performed functional characterization of iMG cells and found that they were able to phagocytose fluorescent beads as illustrated in Figure 1E, where after exposure to the beads, cells changed their morphology from semi-activated state in vitro to amoeboid suggesting adaptation of an activated state. Quantification analyses by flow cytometry showed that ~90% of cells had phagocytic capacity. In addition, we used endotoxin challenge to assess the cytokines released upon stimulation with LPS over 24 h. Twenty-nine targets were assayed and iMG robustly responded to LPS and upregulated all cytokines tested apart from IL-1α, and to a higher level than primary microglia ( Figure 1F).

| Generation of retinal organoids containing iMG cells
Two control hiPSC lines were differentiated to retinal organoids using an established protocol. 20 Since microglial progenitors arise from mesodermal lineage, unlike the retina which develops from neuroectoderm, not allowing for microglial development in situ, we added iMG to retinal organoids after 13 weeks in culture (Figure 2A Figure 2C). Ultrastructural morphology assessment with transmission electron microscopy indicated that organoids with and without iMG developed connecting cilia, inner segment, and in some cases primitive outer segments. We were also able to detect presumptive iMG cells, which contained lipidic inclusions and resembled a morphology of a stimulated macrophage 29 ( Figure 2D).

| Transcriptomic analysis of iMGretinal organoid co-cultures
To assess the development of retinal organoids in the absence and presence of iMG, scRNA-Seq was carried out at 22 weeks of differentiation. Following quality control, ~51,000 cells were obtained and merged using the Seurat package. Transcriptionally similar cells were grouped together and visualized using UMAP, which revealed the presence of 21 cell clusters ( Figure 3A). The highly and differentially expressed genes (Table S3)

| Functional characterization of retinal organoids containing iMG
The impact of incorporating iMG on physiological function of retinal organoids was assessed using multielectrode ganglion cell recordings. Retinal organoids with and without iMG showed a decrease in their spiking activity after WLPs ( Figure 4A), indicating putative OFF centre-like responses from RGCs without any significant differences either in number of responding RGCs or their mean spiking activity change (11.8% in RO + iMG and 12.4% in RO) ( Figure 4A, p = 0.34).
Similarly, putative ON centre-like responses of RGCs were found in both conditions, revealing a firing rate increase after WLP and showing a statistical trend in % firing rate change in the presence of iMG, with a slight increase in the number of responding RGCs in the retinal organoids with iMG (10.5% in RO + iMG and 7.5% in RO) ( Figure 4A, p = 0.065).
Next, we asked whether LPS elicited an immune response in situ by challenging the organoids at 15 weeks for 24 h and used the same panel of human cytokines as we did for iMG cultured on their own ( Figure 1). After 24-h treatment, we saw an upregulation in the levels of most cytokines, including levels of ILα, IL-12/IL-23p40, IL-15,

F I G U R E 4 Functional characterization of retinal organoids and iMG co-cultures. (A) Spike raster plots from putative
Off-centre retinal ganglion cells (RGCs, top) and ON-centre RGCs (bottom) of retinal organoids with (RO + iMG) and without microglia (RO) revealed either a decreasing firing rate for Off-centre RGCs or an increasing spiking activity for ON-centre RGCs after WLPs. Each row in raster plots represent a different RGC and each vertical bar represents a spike from the corresponding RGC. Numbers of responding RGCs and the total amount of recorded cells are stated on the top right corner of each spike raster plot. The red line illustrates the pulsed stimulus onset whereas the left half before indicates the spontaneous activity before WLP exposure and the right half when exposed to WLP. Box plots indicate no statistical differences in percentage (%) firing rate change of putative ON RGC (left) and OFF RGCs (right) after WPL in RO and RO + iMG condition (Mann-Whitney test; p = 0.065 for ON RGCs and p = 0.34 for OFF RGCs). The box plot shows the median and interquartile ranges with Tukey whiskers. n = 1 experiment, 7-8 retinal organoids per condition. (B) Cytokine release measurements showed increased levels of cytokines released in the media after exposure to LPS in retinal organoids containing iMG comparing to organoids alone (mean ± SEM; n = 3; unpaired t-test). iMG, hiPSC-derived microglia-like cells; LPS, lipopolysaccharide; RGC, retinal ganglion cell; RO, retinal organoid IL-16, TNFβ, IL-1β, IL-8, IL-10, TNFα and IL-13 being significantly higher in organoids containing iMG ( Figure 4B).
Overall, the cytokine response to LPS was lower than the results obtained from isolated cultures. This is in line with published literature on cerebral organoids containing microglia and can be explained by iMG representing a low proportion of cells comparing to other cell types within the organoid. 22

| DISCUSS ION
Increasingly, it is recognized that there is a need for more physiologically relevant in vitro systems including an immune component.
Several studies described invasion of cerebral organoids with hiPSCderived microglial cells, [31][32][33] including a report that demonstrated utility of the co-culture system in the study of Alzheimer's disease. 34 However, to our knowledge there are no current immunocompetent models of retina derived from hiPSCs. Retinal organoids are increasingly being used for a variety of applications, including the study of basic principles of development, 35,36 disease modelling, 17,19,37 development of novel therapeutic strategies, 38,39 regenerative medicine 40,41 and drug safety assessment. 20,42 Their use in the preclinical setting could be extended and enhanced by increasing several characteristics that mimic the retina in vivo. Current models lack some of the key features, including vascularization, endothelial cells and immune cells, such as microglia.
In this study we aimed to address one of these limitations by incorporating microglial-like cells derived from hiPSCs into retinal organoids. Here, we show that functional human microglialike cells can be derived from hiPSCs and integrated into retinal organoids. The key findings were that microglia-like cells were able to integrate into the organoids and were able to respond to endotoxin challenge demonstrating retention of at least some of their functionalities. Our data showed a significant upregulation of several pro-inflammatory markers in response to endotoxin challenge, including ILα, IL-12/IL-23p40, IL-15, IL-16, TNFβ, IL-1β, IL-8 and TNFα. Interestingly, we also observed increased levels of an anti-inflammatory cytokine IL-10, which is known to have a neuroprotective function and prevent inflammation-mediated neurodegeneration, and reduce retinal microglial migration in response to LPS. 43 In addition, we also saw an upregulation in another antiinflammatory mediator IL-13, which has been previously shown to reduce ocular inflammation in response to LPS 44 and as a modulator of inflammation associated with uveitis. 45 This demonstrates that our system is able to replicate the balance between pro-and anti-inflammatory conditions. Using hiPSC-derived microglial-like cells as opposed to primary cells has several advantages, the major ones being the ability to generate cells at scale and reduce the effect of donor variability. Future work would require further investigating optimal co-culture conditions to allow long-term survival of iMG, including further co-culture medium optimization and potentially the addition of using a microfluidics platform. Regional heterogeneity of microglial gene signatures is one of their most distinctive features. Future work should also address the possibility of generating brain/retina specific iMG via genetic and epigenetic tools.
There are multiple applications where immunocompetent model of human retina would be of value. There are increasing reports of intraocular inflammation following adeno-associated virus (AAV) therapy, which leads to the reduction of therapeutic efficacy.
Although the pathways involved are complex, retinal microglia is one of the components of the innate immune response to AAV-mediated retinal gene therapy. 46 Studying the immune cells in a relevant microenvironment to human can provide advantages to using animal models which have physiological and functional differences com-

ACK N O WLE D G E M ENTS
This work was funded by CRACKIT23 challenge award (NC/ CO16206/1), BB/T004460/1 and MR/S035826/1 grants. The authors are grateful to Dr. Kathryn White and Tracey Davey from Newcastle University for help with the TEM analysis. The authors thank Rhea Guerra for assistance with microglial differentiations.

CO N FLI C T O F I NTE R E S T
The authors confirm that there is no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
scRNA-Seq data files were uploaded to the Gene Expression Omnibus (GEO) database (submission GSE173180).