RGCs are particularly susceptible to SARS-CoV-2 infection

To identify the retinal cells that are infected, day 74 infected organoids were co-stained with antibodies directed against SARS-CoV-2 N and retinal cell-type-specific markers (Figures 3A–3D). Interestingly, RGCs appeared to be the most commonly infected cell type, as approximately 40% of the N⁺ cells also stained positive for the RGC marker SNCG (Figures 3A and 3D). A smaller number of OTX2⁺ photoreceptors (and perhaps a few bipolar cells) and AP2A⁺ amacrine cells (Figures 3B and 3D) also stained positive for N, indicating the permissibility of these cells. An even smaller number of VSX2⁺ progenitors (and perhaps a few bipolar cells) also stained positive for N (Figures 3C and 3D), suggesting that progenitor cells are less susceptible to infection than neurons but may still be infected. As MGs are still rare at this stage of retinal organoid development, we did not attempt to stain the organoids with an MG marker.

Open in a separate window

Figure 3

SARS-CoV-2 infects retinal ganglion cells at a higher rate than other retinal cell types

(A–C) Retinal organoids infected with SARS-CoV-2 on day 74 and incubated for 96 h before fixation were stained with antibodies against SARS-CoV-2 nucleocapsid (N, green) and the retinal ganglion marker SNCG (red) (A), the photoreceptor marker OTX2 (white) (B), the amacrine and horizontal marker AP2α (red) (B), and the retinal progenitor marker VSX2 (red) (C).

(D) The percentage of cells co-stained with each of the markers was compared. N = 6 organoids taken from 2 different wells, ANOVA p < 0.0001, F = 21.69; p = 0.032; p = 0.0008; p < 0.0001. Error bars represent standard deviation.

(E–H) Day 74 control organoids were stained against OTX2 (white), HUC/D (red), and ACE2 (green) (E and G) or rabbit polyclonal isotype control (green) (F and H). Scale bars: 10 μm.

Angiotensin-converting enzyme 2 (ACE2) is the receptor most commonly associated with the SARS-CoV-2 infection pathway. While ACE2 mRNA expression in the human retina is thought to be low, ACE2 protein is present in the adult human and rat retinas (Senanayake et al., 2007; Tikellis et al., 2004). Accordingly, an IF analysis identified an ACE2 signal concentrated in OTX2⁺ photoreceptors and HUC/D⁺ RGC and amacrine cells in our retinal organoids (Figures 3E and 3G). These results are consistent with the retinal infection profile of SARS-CoV-2 shown here.

Day 90 infected organoids seemed to show (although reduced) infection patterns similar to that of day 74 (Figure S2B). Corresponding with the known depletion of RGCs in late-stage retinal organoids, the majority of N⁺ cells in day 130 infected organoids are co-stained with the photoreceptor/bipolar marker OTX2 (Figures S2C and S2D).

A transcriptomic analysis of SARS-CoV-2-infected retinal organoids

To study the effect of SARS-CoV-2 infection on retinal organoids, we conducted an RNA sequencing (RNA-seq) analysis that compared the transcriptome of noninfected retinal organoids to that of retinal organoids that were infected with SARS-CoV-2 on day 80 of differentiation and incubated for either 24 or 96 h after infection. Unsurprisingly, 10 SARS-CoV-2 transcripts were highly enriched within the infected samples, although no difference was seen between 24 h post-infection incubation and 96 h post-infection incubation (Figure 4A). This is consistent with the reduction in plaque-forming units (PFUs) observed in the plaque assay after 72 h (Figure 1E).

Open in a separate window

Figure 4

A transcriptomic analysis of SARS-CoV-2-infected retinal organoids

(A) A transcriptomic analysis was conducted using infected retinal organoids, incubated for 24 or 96 h following infection, and their respective controls. N = 3 per group. Each repeat included 5 separately treated organoids. SARS-CoV-2 transcripts were highly expressed in the infected samples.

(B) Genes that were significantly differentially expressed (DE) with a fold change of at least 2 (or −2) among all of the samples together were used for Gene Ontology (GO) analysis. A heatmap of genes related to the top enriched categories identified using GO is presented.

(C) The number of normalized reads of the ACE2 gene in noninfected and infected organoids was extracted from the analysis. A Student’s t test did not identify significant differences between the groups (p = 0.83 [24 h] and 0.48 [96 h]). Error bars represent standard deviation.

(D)The DE genes identified in the retinal organoids 24 h post-infection were compared to those identified in SARS-CoV-2-infected choroid plexus after 24 h. The comparison is presented as a Venn diagram.

(E) Control and retinal organoids infected on day 74 of differentiation and incubated for 24 h before fixation were stained against IL-33 (green) and OTX2 (red). Scale bars: 50 μm.

ACE2 transcripts were identified in all of the samples, with no change in expression due to virus infection at either incubation time (Figure 4C), suggesting that the infection of retinal cells does not result in ACE2 upregulation. This result contradicts studies suggesting ACE2 upregulation upon SARS-CoV-2 infection of other types of host cells (Butler et al., 2021).

Next, we compared the transcriptome of all of the infected samples to that of the control samples to identify genes that are differentially expressed (DE) upon infection. This analysis identified 759 DE genes (false discovery rate [FDR] adjusted p- ≤ 0.05), 366 of which were upregulated and 393 were downregulated (Figure S3B).

To gain further information regarding the processes occurring in the retinal organoids upon infection, a Gene Ontology (GO) analysis of the most highly upregulated genes (fold change [FC] ≥ 2) was performed. The top 5 most enriched biological process categories in this analysis were “potassium ion transport,” “cellular response to cytokine stimulus,” “cellular response to bone morphogenetic protein (BMP) stimulus,” “metal ion transport,” and “cytokine-mediated signaling pathway” (Figure 4B; Table 1).

The enrichment of cytokine-related genes indicates the mounting of an immune response upon SARS-CoV-2 infection. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis performed using the same genes found that SARS-CoV-2 infection also led to the enrichment of genes related to the Janus kinase-signal transducer and activator of transcription (JAK-STAT) signaling pathway, a signaling pathway involved in inflammation and the innate antiviral interferon response (data not shown). The enrichment of potassium ion transport-related genes in the data may correspond with evidence suggesting that hypokalemia is relatively common in COVID-19 (Mabillard and Sayer, 2020). While data from other tissues and organoids suggested a strong chemokine response upon SARS-CoV-2 infection (Blanco-Melo et al., 2020; Yang et al., 2020), our analysis identified only 2 chemokines, C-X-C motif chemokine ligand 2 (CXCL2) (FC = 1.83, FDR adjusted p value = 7.94 × 10E−5) and CXCL10 (FC = 3.82, FDR adjusted p value = 0.0003), to be significantly upregulated in retinal tissue.

Comparison of the transcriptome changes of SARS-CoV-2-infected retinal organoids to those of other organoid systems may allow for a better understanding of the unique features of the retinal response to SARS-CoV-2 infection. To this end, we have used a dataset of genes altered in SARS-CoV-2-infected choroid plexus organoids (Jacob et al., 2020). While the choroid plexus is not a neuronal tissue like the retina, it does develop from neural progenitors (Liddelow, 2015) and may share some of the response to SARS-CoV-2 infection with the retina.

Comparison of the DE genes of infected retinal organoids after incubation for 24 h to those of infected choroid plexus organoids after a similar incubation identified 35 genes that are upregulated and 12 genes that are downregulated in both organoid systems (Figure 4D). Of the genes upregulated in both systems, 17 were highly upregulated in the retinal organoids (FC ≥ 2). GO analysis identified enriched categories related to megakaryocyte differentiation, negative regulation of SMAD (suppressor of mothers against decapentaplegic) signaling and serine-threonine phosphorylation, peptidyl-tyrosine phosphorylation, and cell proliferation (Table 2). Interestingly, GO of genes that are specifically highly upregulated in the infected retinal organoids (and are not upregulated in choroid plexus organoids) identified several enriched categories related to transforming growth factor β (TGF-β) and BMP signaling (Table 2). Such a result suggests a difference in TGF-β response to SARS-CoV-2 infection between the retina and the choroid plexus.

Unlike the choroid plexus organoids, retinal organoids show upregulation of the cytokine interleukin-33 (IL-33), the most upregulated (FC = 8.11, FDR adjusted p value = 4.27 × 10E−7) cytokine in these organoids. Accordingly, an IF analysis identified several IL-33⁺ cells in infected organoids, but hardly any in the controls (Figure 4E). Thus, SARS-CoV-2 retinal infection appears to induce IL-33 expression. IL-33 signaling is thought to play a role in COVID-19 pathology (Zizzo and Cohen, 2020), as well as in retinal inflammation and photoreceptor degradation following injury (Xi et al., 2016).

Another interesting gene found to be upregulated in SARS-CoV-2-infected retinal organoids is the inflammasome gene NLR family pyrin domain containing 1 (NLRP1) (FC = 2.2, FDR adjusted p value = 0.0346). NLRP1 inflammasomes are thought to be involved in RGC death in acute glaucoma (Yerramothu et al., 2018).

Interestingly, the top 5 categories enriched in the genes most significantly downregulated in infected retinal organoids (FDR adjusted p-value ≤ 0.05, FC ≤ −2) are related to DNA metabolism, repair, and recombination (Figure 4B; Table 1). This may suggest an effect of SARS-CoV-2 on proliferation or DNA damage repair mechanisms. Other coronaviruses have been shown to interact with the cell cycle and DNA damage response in different ways (Surjit et al., 2006; Xu et al., 2011; Zhou et al., 2008).

SARS-CoV-2 infection of retinal organoids induces inflammatory genes

The pathology of COVID-19 is highly attributed to uncontrolled hyperinflammation, referred to as a cytokine storm (Burke et al., 2020; Mehta et al., 2020). Our data suggest that the infection of retinal organoids results in the upregulation of several inflammatory genes. Among these are several genes involved in the immune response to viral infection, such as CXCL10 (Trifilo et al., 2004) and interferon-induced transmembrane protein 1 (IFTIM1) (Brass et al., 2009).

The most significantly upregulated cytokine in infected retinal organoids was IL-33. Several studies have identified a strong correlation between COVID-19 severity and the levels of IL-33 in the blood serum of patients (Burke et al., 2020; Munitz et al., 2021). In the mature retina, IL-33 is expressed mainly by MG and RPE cells (Xi et al., 2016). The elevated expression and secretion of IL-33 in the rodent retina were shown to be the result of different harmful stimuli such as infection with Toxoplasma gondii, retinal detachment, and light-induced damage (Augustine et al., 2019; Tong and Lu, 2015; Xi et al., 2016). While suggested to have a protective role in a mouse model of retinal detachment (Augustine et al., 2019), IL-33 was shown to pathologically amplify the innate immune response in the rat retina following light-induced damage or RPE disruption, promoting the expression of inflammatory chemokines and cytokines from MGs, the accumulation of myeloid cells in the ONL, and RGC and photoreceptor cell death (Xi et al., 2016).

Thus, our data suggest that SARS-CoV-2 infection of retinal cells results in an inflammatory response of immune system factors such as IL-33 and NLRP1, which are involved in retinal degenerative diseases that cause irreversible blindness. Cytokines that were upregulated in the infected organoids may prove to be worthy candidates for further studies regarding COVID-19 retinal pathology.

Comparison of the DE genes in retinal organoids and choroid plexus organoids identified the enrichment of TGF-β response genes among genes that were induced in the retinal organoids but not in the choroid plexus organoids. TGF-β signaling plays an important role in maintaining the immune-privileged status of the eye (Masli and Vega, 2011; Zhou and Caspi, 2010). For example, both RPE cells and RGCs suppress the activation of T cells via the secretion of TGF-β molecules (Edo et al., 2020; Sugita et al., 2011). Unlike most of the CNS, the choroid plexus is not an immune-privileged structure (Galea et al., 2007). Thus, differences in the TGF-β response of retinal and choroid plexus organoids to SARS-CoV-2 infection may be related to the immunosuppressive role of TGF-β.

Virus production and organoid infection

SARS-CoV-2 was isolated from a patient in South Tyrol (hCoV19/Germany/FI1103201/2020) and identified as type “Ischgl.” The virus sequence is available in GSAID: EPI-ISL_463008.

All SARS-CoV-2-containing experiments were conducted under biosafety level (BSL) 3 conditions. The virus was propagated on VeroE6-TMPRSS2 cells using DMEM (Sigma-Aldrich) supplemented with 2% FBS (Capricorn Scientific), 1% penicillin/streptomycin (Sigma-Aldrich), 1% sodium pyruvate solution (Sigma-Aldrich), 1% nonessential amino acid (NEAA) solution (Sigma-Aldrich), and 1% HEPES solution (Sigma-Aldrich) using an MOI of 0.01. Two days post-infection, the virus titer was determined.

SARS-CoV-2 was diluted to the desired amounts in RDM supplemented with 100 μM taurine (Sigma-Aldrich) without FBS. The organoids were washed twice with PBS and incubated for 1 h at 37 °C with the virus dilutions. Afterward, organoids were washed once with PBS and incubated in RDM with 100 μM taurine for specific time points.

Plaque assay

VeroE6 cells were grown to a confluent monolayer in 6-well plates. A 10-fold dilution series of viral solutions was prepared in PBS (Sigma-Aldrich) containing 1% (v/v) penicillin/streptomycin, 0.6% (v/v) BSA (35%) (Sigma-Aldrich), 0.01% (w/v) CaCl₂ (1%), and 0.01% (w/v) MgCl₂ (1%) from the medium in which the organoids were incubated. The VeroE6 cells were incubated with the dilution series for 1 h at 37 °C. Afterward, the inoculum was replaced by plaque medium: 63% (v/v) 2× MEM 20% (v/v), 10× MEM (Gibco), 3.2% (v/v) NaHCO₃ (Lonza), 2% (v/v) HEPES (1 M; pH 7.2) (Sigma-Aldrich), 1.2% (v/v) BSA (35%), 1% (v/v) 100× penicillin/streptomycin/l-glutamine solution (10,000 U/mL penicillin; 10,000 μg/mL streptomycin, 29.2 mg/mL l-glutamine) (Gibco), 2% (v/v) FBS, and 35% (v/v) Agar (2%) (Oxoid). After 24–96 h, the plaques were counted.

Immunohistochemistry

Organoids were fixed in 4% paraformaldehyde (PFA) for 30 min (organoids used for ACE2 IF were fixed for 2 h), washed 3 times in PBS, and incubated overnight at 4°C in 30% sucrose in PBS. The organoids were then cryopreserved in optimum cutting temperature (O.C.T.; Tissue-Tek) and sectioned to a thickness of 20 μm. The sections were blocked and permeabilized with 10% normal donkey serum (Abcam) and 0.3% Triton X-100 in PBS for 1 h at room temperature (RT), and then incubated with primary antibodies diluted in a blocking medium overnight at 4°C. The primary antibodies used for the study can be found in Table S1. Sections were further washed and incubated with a species-specific secondary fluorescent antibody, with DAPI, for 1 h in the dark at RT and then mounted. Imaging was done using an LSM 780 scanning confocal microscope (Zeiss). All of the images presented are maximal projection images. Cell counting was performed manually on random sections from different organoids, with a similarly sized area.

RNA extraction and cDNA library preparation and RNA-seq

RNA from untreated and infected organoids was extracted with TRIzol reagent according to the manufacturer’s protocol, with minor modifications. Briefly, washed retina organoids were resuspended in 700 μL TRIzol reagent. The tissue was disrupted mechanically by vortexing for 30 s and pipetting 20 times. Chloroform 140 μL was added and samples were mixed vigorously for 15 s. Following centrifugation at 4°C, the aqueous phase was carefully separated and RNA was further precipitated by 350 μL 2-propanol. RNA pellets were washed twice with 75% ethanol and resuspended in diethyl pyrocarbonate (DEPC)-treated water for further library preparation.

Total RNA 600 ng was used for RNA isolation, fragmentation, and cDNA synthesis using the NEBNExt poly(A) mRNA magnetic isolation module (E7490). cDNA libraries were amplified and index labeled using the NEBNext Ultra TM II RNA Library Prep kit for Illumina (E7770, E7775). DNA library quality and concentration were analyzed on an Agilent Bioanalyzer DNA Chip. cDNA samples were loaded on a NextSeq 2000 system (Illumina) at a concentration of 800 pmol, and the sequencing was performed with single-end 100-bp reads.

Bioinformatic analysis

The sequencing data were first demultiplexed using the Illumina software bcl2fastq version 2.20.0. The quality of the resulting FASTQ files was then evaluated with the program FastQC version 0.11.8 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/).

To perform the alignment, a customized combination of the human genome and the SARS-CoV-2 viral genome was created. The reference sequence and gene annotation for SARS-CoV-2 were obtained from the NCBI repository: NC_045512.2 “Severe acute respiratory syndrome coronavirus 2 isolate Wuhan-Hu-1, complete genome.” The human genome sequence and gene annotation were also obtained from NCBI: GRCh38.p13. Both viral and human genomes were concatenated using a customized bash script to create the alignment reference sequence and annotation files. Finally, the alignment was performed using STAR version 2.7.7a (Dobin et al., 2013), with default parameters and the extra instruction “--quantMode GeneCounts” to create additional gene count tables.

The following secondary analysis was performed in the statistical environment R version 4.0.3 (https://www.r-project.org/). After importing the gene count tables, only genes that had a minimum of 5 reads in at least 3 samples were kept as a quality filter. Then, a dispersion trend among the samples was assessed with a principal-component analysis (PCA) applied over the normalized read counts, transformed through the regularized log transformation (Figure S3A).

Differential gene expression was assessed based on a negative binomial distribution through the package DEseq2 version 1.30.1 (Love et al., 2014). The Benjamini-Hochberg FDR method was used for multiple comparisons correction. Genes were considered DE if they presented an absolute value of FC of >1 and a Benjamini-Hochberg FDR of ≤5%. A very small number of SARS-CoV-2 transcripts was identified in 1 of the 24-h control samples; however, considering the vast difference in scale between the levels of SARS-CoV-2 transcripts in this sample and the levels found in the infected samples (Figure 4A, FC > 200), these transcripts most likely do not represent a SARS-CoV-2 infection. In addition, this sample clustered together with the control samples in a PCA analysis (Figure S3A). Thus, we decided to include this sample in the analysis.

Heatmap representation of the expression of selected genes was created with the package pheatmap version 1.0.12 using the values obtained after applying the regularized log transformation over the raw counts. GO and KEGG pathway analyses were performed by separately collecting all of the upregulated genes with an FC of ≥2 and all of the downregulated genes with an FC of ≤−2 and uploading the gene lists to the Enrichr software (Chen et al., 2013; Kuleshov et al., 2016).

Quantitative PCR (qPCR)

cDNA synthesis was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR was performed using iTaq SYBRGreen Supermix with ROX (Bio-Rad) and a QuantStudio3 machine (Applied Biosystems). Gene expression was normalized according to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and calculated using the delta Ct method. qPCR primers can be found in Table S2.

ACE2 blocking

Retinal organoids on day 90 of differentiation were incubated at RT for 1 h in PBS with either a goat anti-ACE2 antibody (AF933, R&D Systems) or a normal goat IgG control (AB-108-C, R&D Systems), at a concentration of 100 μg/mL before being infected with SARS-CoV-2.

Statistical analysis

Statistical analysis was done using GraphPad Prism 9. An analysis of experiments with multiple groups was done by a 1-way ANOVA with Tukey’s multiple comparison test. Statistical analysis of experiments comparing 2 groups was done using an unpaired, 2-tailed Student’s t test. No statistical tests were used to predetermine the sample size.

This research was supported by the Max Planck Society’s White Paper-Project: Brain Organoids: Alternatives to Animal Testing in Neuroscience. Further financial support was received from the German Research Foundation (DFG), grants SFB1009 B13 and CRU342 P6; the German Ministry of Education and Research (BMBF); project OrganoStrat (01KX2021); SFB944 (Z-Project); iBiOs (PI 405/14-1); and the IZKF, grant Bru2/015/19. We thank Ingrid Gelker and Manuela Haustein for their technical help. We also thank the Core Facility Genomik in the medical faculty in the Münster team for RNA-seq and Areti Malapetsas for editing the manuscript.