Islet1 and Brn3 Expression Pattern Study in Human Retina and hiPSC-Derived Retinal Organoid

This study was conducted to determine the dynamic Islet1 and Brn3 (POU4F) expression pattern in the human fetal retina and human-induced pluripotent stem cell- (hiPSC-) derived retinal organoid. Human fetal eyes from 8 to 27 fetal weeks (Fwks), human adult retina, hiPSC-derived retinal organoid from 7 to 31 differentiation weeks (Dwks), and rhesus adult retina were collected for cyrosectioning. Immunofluorescence analysis showed that Islet1 was expressed in retinal ganglion cells in the fetal retina, human adult retina, and retinal organoids. Unexpectedly, after Fwk 20, Brn3 expression gradually decreased in the fetal retina. In the midstage of development, Islet1 was detected in bipolar and developing horizontal cells. As the photoreceptor developed, the Islet1-positive cone precursors gradually became Islet1-negative/S-opsin-positive cones. This study highlights the distinguishing characteristics of Islet1 dynamic expression in human fetal retina development and proposes more concerns which should be taken regarding Brn3 as a cell-identifying marker in mature primate retina.


Introduction
Islet1, also known as ISL1, is a LIM-homeodomain transcription factor that plays critical roles in differentiation, cell specification, and phenotype maintenance of horizontal cells as well as cholinergic amacrine and ganglion cells in the retina of different species including fish, reptiles, birds, amphibians, chickens, and mammals. Moreover, numerous studies have revealed that ISL1 plays a key role in multiple tissue types, such as the heart [1], kidneys [2], skeletal muscle, endocrine organs [3], and nervous system [4]. Additionally, previous studies suggested that ISL1 is highly expressed in gastric and breast cancer [5] and is associated with advanced tumor invasion, proliferation, migration, tumor stage, tumor size, metastasis, and poor overall survival. ISL1, as a crucial transcription factor, is required for retinal neuroblast differentiation during human retinogenesis. Various previous studies also demonstrated Islet1 dynamic expression shows a specific temporal and spatial pattern in the retina of multiple animal models [6][7][8][9][10]. However, the detailed Islet1 expression pattern during human retinal development remains unclear.
Since 2012 [11], studies using retinal organoids derived from human embryonic stem cells (hESCs) and humaninduced pluripotent stem cells (hiPSCs) have provided insight into developmental biology research, disease modeling, and stem cell replacement therapy. As the production of retinal organoids for studying differentiation has increased [12], this approach has begun to replace animal models because of the lower cost, fewer ethical concerns, and humanization properties.
An essential question regarding retinal organoids is whether they mimic human fetal eye development. If so, Islet1 may be useful as an indicator for determining the developmental process of retinal organoids compared to the human fetal retina.
To evaluate Islet1 dynamic expression in human fetal retina and hiPSC-derived retinal organoid development, we characterized subsets of Islet1-positive cells based on morphological features and performed coimmunostaining with specific markers of retinal neuron subtypes. We determine the expression profile of Islet1 during human fetal retina development and examined variations in the developmental process in retinal organoids. During the study on Islet1, Brn3 showed distinguished express pattern from a rodent model. Therefore, the Brn3/Pou4f family of transcription factors, well-known markers of RGC, is also investigated in the present study.

Ethics Statement and Tissue
Collection. Human fetal eyes ranged in age from 8 to 27 fetal weeks (Fwk) and were obtained from legal routine therapeutic abortions at the Third Affiliated Hospital, Sun Yat-sen University. Fetal age was determined by eye size and foot length. The human adult retina slices were collected by the pathology department of Zhongshan Ophthalmic Center, from the samples for pathology examination and choosing the peripheral normal retina. The retinoblastoma tissue was collected by Prof. Rong Lu. Rhesus slices were collected from the control eyes of a previous study [13]. All samples were collected with patient consents and in accordance with protocols approved by the institutional review boards at Zhongshan Ophthalmic Center, SYSU (approval number: 2013PRLL0029). Eyes for immunofluorescence (IF) and morphologic analysis were fixed in 4% (2-12 h) paraformaldehyde in 0.1 M phosphatebuffered saline (PBS). Only sections adjacent to the fovea were used for immunofluorescence.

Retinal
Organoids. The retinal organoids were differentiated, as described previously [12], from the BC1-GFP hiPSC line (gifted by Prof. Linzhao Cheng, Johns Hopkins University School of Medicine, Baltimore, MD, USA) and the SB hiPSC line (CA4002106; Cellapy Bio, Beijing, China). The day on which embryonic bodies (EBs) were formed was considered as Day 0, and the retinal organoid was cultured in suspension after Day 28 (differentiation week 4, Dwk 4).

Fixation and Sectioning. Human fetal eyeballs < Fwk 15
were fixed with 4% paraformaldehyde in 0.1 M PBS (pH 7.2-7.3) for 2-4 h according to the size of the eyeballs at room temperature (24-26°C). In elder eyes, the anterior segment (including the cornea, iris, and lens) and vitreous were removed to obtain posterior cups for fixation. Residual eyecups were fixed overnight at 4°C. Retinal organoids were collected and fixed in 4% paraformaldehyde for 0.5 h.
Primary antibody staining was followed by three washes with 1x PBS. Subsequently, cryosections were incubated for 1 h at room temperature with secondary antibodies that included the corresponding species-specific Alexa Fluor-488-, Alexa Fluor-555-, and Alexa Fluor-657-conjugated antibodies (1 : 500; Gibco, Grand Island, NY, USA). Next, 2 μL ProLong® Gold Antifade Reagent with DAPI was applied 10 minutes at room temperature to counterstain the nuclei. Subsequently, all samples were rinsed in PBS and mounted in Vectashield and coverslipped. H&E staining was performed as described previously [14,15].
2.5. Image Acquisition and Processing. Images were acquired on an Olympus BX53 microscope (Tokyo, Japan) and Zeiss (Oberkochen, Germany). Confocal images were acquired with a Zeiss LSM 510 confocal microscope. Immunofluorescence intensity was evaluated by ImageJ (NIH, Bethesda, Maryland), and for statistical comparisons, values were subjected to a two-tailed Student's t test Prism Software Version 7 (GraphPad Software, Inc., La Jolla, CA).
2.6. Real-Time PCR. Total RNA was isolated using TRIzol reagent (Sigma-Aldrich) from fresh tissue and RNAprep pure FFPE kit (TIANGEN, DP439) from formalin-fixed tissue, and RNA quality was evaluated with a NanoDrop1000 spectrophotometer (Thermo Fisher Scientific). The firststrand cDNA was synthesized with a PrimeScript RT Master Kit (RR036A; Takara Bio, Shiga, Japan) according to the manufacturer's instructions. Quantitative PCR was performed using a LightCycler 480 SYBR Green I Master (4887352001-1; Roche, Basel, Switzerland) on a LightCycler 480II system (Roche). Reactions were performed in triplicate, and Ct values were calculated using the 2 -ΔΔCt method. The expression levels of target genes were normalized to that of GAPDH. Primer sequences are listed in Table 1.    Fwk 8, Islet1 was intensively expressed in the GCL. During development, Islet1 expression slowly decreased in the GCL and translocated to the outer layer. Initially, Islet1 was on the most apical side of the retina, followed by the thickening neuroblast layer. Finally, Islet1 was mainly expressed on the apical side of INL. The Islet1 expression pattern in the hiPSC-derived retinal organoid was consistent with that in the fetal retina (Figures 1(a″)-1(f″)). However, at approximately differentiation week (Dwk) 16, lamination of the retinal organoid was not sufficient, compared to the contemporary fetal retina (Figure 1(c″)). In Dwk 24, the Islet1-positive GCL disappeared ( Figure 1(e ″ )). Whether this was because the RGCs had matured or disappeared was evaluated by double-staining with RGC markers.

RGC Specification and Islet1 Expression in Human Fetal
Retina and Retinal Organoid. To determine whether Islet1 colocalizes with early markers of RGC specification, coimmunolabeling with anti-Islet/-Brn3 and anti-Islet/-HuD was performed. In the early weeks, Islet1 was mainly expressed in Brn3-positive retinal ganglion cells. However, at the most basal side, Islet1 expression was higher, while on the outer side, clusters of Brn3-positive but Islet1-negative cells were detected ( Figure 2(a)). In the same period of the retinal organoid, Islet1 was colocalized with Brn3 ( Figure 2 , and Islet1 showed that Islet1 levels were higher on the basal side and the progenitor marker MCM2 was on the apical side (Figures 2(n1) and 2(n2)-2(n5)). According to the developmental axis oriented vitreosclerally, the earliest specified ganglion cells were intensively Islet1 positive, while the migrating and specifying RGC were Brn3 positive. Double staining with HuD, which is also an RGC marker as demonstrated by Romero-Aleman and colleagues [16], showed that Islet1 was also coexpressed in the early weeks (Figures 3(a)-3(e)). In Fwk 27, GCL was labeled with Islet1 and HuD, indicating that during this period, the ganglion cell existed but was no longer stained by Brn3 antibody (Figure 3 Costaining with another common RGC marker, RPBMS, it was shown that Islet1, HuD, and Brn3 could be coexpressed in RGC during Fwk 16 to Fwk 22 (Figures 4(a)-4(c)), as well as in retinal organoid (Figure 4(d)). But in the earlier stage (Fwk 8), those Brn3 (+)/HuD (+)/Islet1 (+) RGCs were RPBMS negative, and to Fwk 27, RGCs could only be stained with RPBMS, Islet1, and HuD antibody. To investigate whether the later stage of ganglion cells is still Brn3 positive, we stained the human adult retina with Brn3. As shown in Figures 4(e1), 4(e2), 4(f1), and 4(f2), Islet1-/HuD-positive GCL was Brn3 negative. In rhesus retina, which is one of the most widely used primate models, RGC was also stained by RBPMS, HuD, and Islet1, but not Brn3 (Figures 4(g1) and 4(g2)-4(g4)). Immunofluorescence intensities of Brn3 and RBPMS were evaluated for Fwk 9 to Fwk 27 fetal retina. To Fwk 27, the intensity of Brn3 was extremely low, while that of RBPMS was obviously high (Figure 4(h)).
On the mRNA level, Brn3a expression was significantly low in adult human and rhesus retina as referred to hiPSC. Brn3b and c were not detectable in qPCR in adult human retina. Interestingly, during retinal organoid development, Brn3b expression was the main expression subtype among Brn3a, b, and c, while in human fetal retina, Brn3a was the major (Figure 5(a)). Considering that the retinoblastoma (Rb) cells were regarded as the dedifferentiated cells [17], we included Rb cells as a positive control. In Rb cells, the Brn3a, b, and c mRNA expressions were significantly higher than those in mature retinal cells. During the fetal retina development, RGC population would decrease rapidly in the later stage. Therefore, the expression level of Brn3a, b, and c was then normalized with the expression level of pan-RGC marker, RBPMS. As

Amacrine and Horizontal Cell Development and Islet1
Expression in Human Fetal Retina and Retinal Organoid. Interestingly, in the early stage, HuD protein was expressed mainly in the cytoplasm (Figures 3(g) and 3(h)). From Fwk 15, a layer of HuD-positive/Islet1-negative cells was observed (Figure 3(b)) and HuD was in the nucleus in these cells. These cells were more obvious in the Dwk 16 retinal organoid (Figure 3(i)) and settled between the retinal ganglion cells and outer Islet1-single-positive cells. Based on the cells on the inner side of INL and previous studies showing that HuD protein was also expressed in amacrine cells [16], this cluster of cells may be a subtype of amacrine cells. Until Fwk 27, HuD-positive cells lined up on the inner side of the INL, apart from the Islet1-positive cells (Figure 3(f)). However, in the retinal organoids, these cells remained mixed with Islet1-positive cells in the INL (Figure 3(l)).
AP2α-positive cells were also located in the INL. According to a study by Bassett et al., AP2α is exclusively expressed in postmitotic amacrine cells during retinal development and in mature amacrine cells in the adult retina [18]. From Fwk 9, several postmitotic amacrine cells were observed in the fetal retina (Figure 6(a)), but not observed in the retinal organoid until Dwk 13 (Figure 6(h)). These subpopulations of amacrine cells shared similar location of HuD-positive/Islet1-negative cells (Figures 6(i)-6(k)).
Prox1, a developing horizontal cell marker, colocalized with Islet1 (Supplementary Fig 1A). Further, the absence of colocalization of Islet1 with Prox1 at Fwk 22 in the human fetal retina confirmed that Islet1 is not expressed in mature but rather in developing horizontal cells (Supplementary Fig 1B-C) (Figure 7(j)). However, in the retinal organoid, the colabeled cells did not develop until Dwk 26, which is delayed compared to in the fetal retina (Figure 7(j)). To confirm whether CHX10-positive cells were progenitor or bipolar, we double stained the cells with CHX10 and MCM2, another progenitor marker [21]. At Dwk 13 and 16, nearly all CHX10-positive cells were found to be progenitors (Figures 7(m)-7(o)), but the inner side population of cells had already exited the cell cycle. At Dwk 26, CHX10-/Islet1-positive bipolar cells were not stained with MCM2 ( Figure 7(p)). PKCα identified rod bipolar cells [22]. In the retinal organoid, the PKCα-/Islet1-positive bipolar cells were in the INL ( Supplementary Fig 2A-D).         colocalization was seen ( Supplementary Fig 3D-F). The peripheral retina was Islet1 positive in the most outer layer ( Supplementary Fig 3G), but in the central retina, when cone markers expressed, Islet1 expression decreases ( Supplementary Fig 3H-I). Developing photoreceptors are well-organized by Fwk 15, but the inner segments did not develop until Fwk 16. Until Fwk 20, the inner segment could easily be detected. However, in the retinal organoid, when the inner segment appeared in Dwk 16, a cluster of immature photoreceptors migrating from the basal side remained.

Discussion
We evaluated the dynamic expression of Islet1 in the human fetal retina and compared the hiPSC-derived retinal organoid with human fetal retina development. The results revealed that Islet1 temporal and spatial expression was generally similar to that in various vertebral animal models. However, we also found the Islet1 was specifically expressed in cone precursors of the human fetal retina. Further, the retinal organoid showed the same patterns of Islet1 expression as the fetal retina, indicating its potential as a development model and drug-screening model. However, in later stages of development, the retinal organoid showed a developmental delay compared to the human fetal retina in the specification of photoreceptors and retinal lamination.
A recent study by Bejarano-Escobar et al. [6] confirmed the presence of the LIM-domain transcription factor Isl1 in differentiating and mature ganglion, amacrine, bipolar, and horizontal cells in the retina of mammals, birds, reptiles, fish, and Xenopus laevis. In the present study, we examined the retina of human embryos. The Islet1 spatiotemporal expression in human fetal showed a very similar pattern to in mammals, from ganglion to amacrine and bipolar cells. Further, the "retina-in-dish" showed a similar pattern of expression.
Consistent with a study by Prasov and Glaser [23], human fetal retinal ganglion cells coexpressed Brn3 and Islet1 during retinal development. Interestingly, in the fetal retina, the signal intensity between Brn3 and Islet1 varied from the basal side to the apical side. In the most inner side, cells showed stronger expression of Islet1 than of Brn3. This layer of cells was postmigrated RGCs, which were the first specified retinal neurons. In the outer direction from this layer, cluster cells expressed the two proteins in the same quantities. Further towards the outer side of the neuroblast layer, migrating RGCs showed higher expression of Brn3. A similar pattern was reported mouse models [24]. Meanwhile, Li and colleagues' study also shown that the Islet1 and Brn3 (including Brn3a, b, and c) collaborated nonsynergistically in regulating RGC differentiation [25]. In the retinal  organoid, this phenomenon was also observed. However, the postmigrated RGC with higher Islet1 levels did not line up as clearly as in the fetal eye. Thus, Islet1 indicated more mature RGCs, while Brn3 indicated relatively immature RGCs.
We found that Brn3 expression gradually decreased during fetal development. The RGC in adult human retina could not be stained by Brn3 antibody, as in the adult primate retina. Then, we investigated the mRNA level of Brn3a, b, and c during the development of fetal retina and adult retina. In adult retina, Brn3b or c was not detectable, while Brn3a was 135-fold higher, consistent with Whitmore's study by RNA-seq that POU4F1 (Brn3a) was enriched in the macular part of retina [26]. After normalized with another pan-RGC marker, RBPMS, the Brn3a level was even higher in adult human retina. Costaining Brn3 and Brn3a in fetal retina indicates that the Brn3 antibody we used could cover those population of Brn3a-positive RGCs. Therefore, although the Brn3a transcript factor was high in adult retina, there is a possibility that only a few of them translated into protein, leading to the negative result in immunofluorescence. Another research performing RNA-seq with human retina showed that the Brn3b (POU4F2) was detectable [27], although the expression level was relatively much lower.    could be the reason that Brn3b was not detected in the present study. In fetal retinas, Brn3a expression was gradually decreasing along development. Those periods with a high level of Brn3a expression coincidently matched with the period of RGC redistribution in retinal development [28][29][30]. As a control, the dedifferentiated retinal cells, retinoblastoma cells expressed a high level of Brn3a, b, and c, indicating the relationship between low level Brn3 and retinal neuron maturation. Contrarily, the Brn3a level kept low in retinal organoid development and they mainly expressed a high level of Brn3b. Consistent with Langer et al.'s findings, hiPSC-derived RGC population had subtype profile varied from human RGCs [31]. This result indicated that although the stem cell-derived RGC was highly similar to human RGCs in morphology, even in electrophysiological characteristics, they were still in high heterogeneity. Considering the relation between Brn3a and the bcl-2-related apoptosis pathway [32,33], the varied transcript profile of Brn3a, b, and c in stem-cell RGC may also be one of the factor that related to the gradually RGC apoptosis in long-term culture of retinal organoid. Because most studies of RGC development [34][35][36][37][38] and RGC damage in glaucoma [39,40] used Brn3 as a marker in rodent models, caution should be used when extrapolating the results to primates and stem cellderived retinal neurons.
Moreover, HuD protein (also known as ELAVL4) is an RGC marker [15]. In contrast to Brn3, HuD is expressed in the GCL in the mature retina. Additionally, in the present study, after Fwk 15 and Dwk 16, HuD was expressed on the inner side of the GCL, which was found to contain amacrine cells. This agrees with the results of Ekstrom and Johansson [41].
Correspondingly, double labeling of Islet1 and amacrine-and bipolar-specific markers in the human fetal retina and retinal organoid showed the same pattern as in the mouse retina, chicken retina, etc. [7,[42][43][44]. Meanwhile, Islet1 also expressed in amacrine cells and bipolar cells in pig, and they appeared around midgestation [45], which was consistent with fetal retina. Although AP2αpositive amacrine cells in retinal organoids gained their fate during a similar period as in the human fetal retina, they failed to migrate and organize in later development. In contrast, the development of bipolar cell delayed, but the cells organized into a layer in INL.
Few studies have focused on Islet1 and photoreceptor specification. According to Wang et al., Islet1-/Recoverinpositive cells were cone precursors [46], which were labeled by L/M-opsin in later development. Moreover, a review by Bejarano-Escobar et al. described that most studies showed that Islet1 failed to colocalize with typical markers of rods and cones [6]. In the present study, we found that in the ONL of the human fetal retina, Islet1 labeled Islet1-/Recoverin-positive cone precursors. Moreover, in later development stages, when we stained cone cells with S-opsin and L/M-opsin antibody, no colabeling was seen with Islet1 in human fetal retina and rhesus retina. But in long-term cultured retinal organoid, S-opsin or L/M-opsin was colabeled with Islet1 in cone cells. Meanwhile, we also observed that from peripheral to central fetal retina, Islet1 downregulated in the most outer layer coincided with cone-opsin upregulation. Fischer et al.'s study demonstrated that the upregulation of cone-opsin coincided with the downregulation of Islet2 in photoreceptors in the far peripheral regions of retina [9]. This may be one of the reasons why no coexpression of cone-opsin and Islet1 was observed in fetal retina and rhesus retina.
In conclusion, this is the first study to describe the Islet1 and Brn3 expression pattern throughout development in the human fetal retina. The expression of Islet1 in subsets of retinal neurons was mainly consistent across species from fish to human, supporting that Islet1 plays a critical role in human retinal cell specification, differentiation, and maintenance. Additionally, we determined the full picture of development and cell specification in hiPSC-derived retinal organoid comparing to in the fetal retina.
There were some limitations to this study. First, although Brn3 and Brn3a did not stain RGCs in the adult primate retina, whether Brn3b and c proteins were all covered by present antibody was not clear. Due to the shortage of samples, quantification of protein was not able to be performed in the present study. Thus, it was unclear how many Brn3 transcript factors would translate into protein. Therefore, the role of Brn3 and Islet1 in RGC differentiation in primate requires further analysis. Second, a previous study demonstrated that the Islet coexpressed with cone markers was Islet2, rather than Islet1 [9]. Thus, although the Islet1 primary antibody used was monoclonal, further confirmation should be performed by RNA in situ hybridization. Finally, compared to the fetal retina, the hiPSC-derived retinal organoid has three main limitations: RGC was not preserved in later development, RGC transcript profile differed from those in fetal retina, and the lamination and organization were not distinct. Therefore, the differentiating protocol must be optimized, possibly by using additional transcription factors.

Data Availability
The immunofluorescence data used to support the findings of this study are included within the article and the supplementary information file.