Generation of a rod-specific NRL reporter line in human pluripotent stem cells

Reporter lines generated in human pluripotent stem cells can be highly useful for the analysis of specific cell types and lineages in live cultures. We created the first human rod reporter line using CRISPR/Cas9 genome editing to replace one allele of the Neural Retina Leucine zipper (NRL) gene with an eGFP transgene in the WA09 human embryonic stem cell (hESC) line. After confirming successful targeting, three-dimensional optic vesicle structures were produced to examine reporter specificity and to track rod differentiation in culture. The NRL+/eGFP hESC line robustly and exclusively labeled the entirety of rods throughout differentiation, eventually revealing highly mature structural features. This line provides a valuable tool for studying human rod development and disease and testing therapeutic strategies for retinitis pigmentosa.

constructed by fusing the eGFP coding sequence to the NRL start codon followed by the rabbit beta-globin polyA terminator and a loxP-flanked puromycin resistance (PuroR) selection cassette (Fig. 1A). Following electroporation of the donor plasmid, a Cas9 D10A nickase plasmid, and plasmids containing NRL targeting sgRNAs, clones that incorporated the donor sequence were identified by puromycin resistance. Surviving colonies were screened by PCR and one clone was selected for further analysis and removal of the PuroR cassette to optimize eGFP expression. Genotyping was performed with primers that distinguished the unedited NRL allele from successfully An eGFP coding sequence was fused to the NRL start codon, followed by the rabbit beta-globin polyA terminator and a loxP-flanked puromycin resistance (PuroR) selection cassette. NRL homology arms flanked the genomic sequence on either side of the NRL start. (B) Schematic of an unedited (wildtype) NRL allele (top), an NRL allele with the eGFP transgene and PuroR selection cassette inserted (NRL-eGFP+ PuroR reporter; middle), and an NRL allele with the eGFP transgene following CRE recombinase-mediated removal of the PuroR selection cassette (NRL-eGFP− PuroR; bottom). Primers that amplify DNA located between the left and right homology arms of NRL ("homology arm primers"; blue arrows) or DNA spanning the junction between the left homology arm and the GFP transgene ("left junction primers"; red arrows) are shown. (C,D) Images of agarose gels of genomic PCR products obtained using the homology arm primers (C: blue primer set) or the left junction primers (D; red primer set). Numbered lanes in panels C and D gels correspond to the same control or clone. Lane  (Note that under the PCR conditions used, the 3.2 kb fragment predicted for the NRL-eGFP+ PuroR is not to be amplified. Gels in (C,D) were cropped for space. The full-length gel is available in Supplementary Information). (E) G-banding analysis demonstrating maintenance of a normal karyotype in the NRL +/eGFP line. targeted alleles with and without the PuroR cassette (Fig. 1B). The selected clone demonstrated targeted incorporation of the eGFP reporter transgene at a single NRL locus prior to and after successful excision of the PuroR cassette (lanes 2 and 3, respectively, in Fig. 1C and D). The NRL +/eGFP clone was expanded and fully sequenced to rule out unintended genetic modifications. Thereafter, to ensure that no off-target mutations were introduced, the ten highest-likelihood sites predicted by the crispr.mit.edu algorithms for each sgRNA were amplified and sequenced from the original clone and no off-target indels were detected (see Supplemental Information). Finally, the gene-edited NRL +/eGFP line was karyotyped to confirm that the multi-step clonal selection process did not result in chromosomal abnormalities (Fig. 1E).
Production and isolation of OVs from NRL +/eGFP hESCs. WA09 NRL +/eGFP hESCs were lifted on day (d) 0 to form free-floating embryoid bodies (EBs) ( Fig. 2A). EBs were subsequently plated on d7 and maintained in retinal differentiation medium (RDM; see Materials and Methods) until d20-30 when early OV colonies were clearly identifiable by light microscopy (Fig. 2B). Following manual dissection, the isolated OVs were maintained in suspension ( Monitoring rod production in live NRL +/eGFP OVs. NRL-eGFP expression was first observed in patches of cells in cultured OVs at d68 of differentiation (Fig. 3A,B), which increased dramatically over time ( Fig. 3C-F). By d89, all OVs displayed robust NRL-eGFP expression (Fig. 3E,F) that persisted at least to d230, the latest time point tested (Fig. 2G,H). Outer segment-like protrusions from the surface of OVs were first noted at d140 (data not shown) and remained readily apparent by light microscopy until at least d230 (Fig. 3G,I). To determine the percentage of NRL-eGFP+ and NRL+ cells generated over time, we dissociated NRL +/eGFP hESC-OVs at multiple stages of differentiation, immunostained for NRL, and quantified labelled cells using high content image analysis (HCIA). The percentage of NRL-eGFP+ cells increased between d70-130 from 0.3 ± 0.3% to 47.2 ± 3.3% (n = 6; mean ± standard deviation); similarly, the percentage of cells labeled with an NRL primary antibody increased from 5.7 ± 2.3% to 50.3 ± 1.9% (Fig. 3J). Within the d70-130 time window, the most dramatic increase in NRL-eGFP and NRL expression occurred between d70 and d95 (Fig. 3J), which ultimately led to approximately half of the entire cell population of NRL +/eGFP hESC-OVs adopting an early rod identity. Of note, the difference between the percentage of cells at d70 expressing NRL-eGFP vs. NRL likely reflects a relative delay in visible expression of the transgene in live cultures.
Specificity of eGFP expression during early rod production in NRL +/eGFP OVs. To evaluate the specificity of expression of the NRL +/eGFP line, we first compared localization of eGFP to that of the pan-photoreceptor transcription factor CRX (cone rod homeobox) at d90 of differentiation, a time point coinciding with early rod production ( Fig. 4A-C). All eGFP+ cells had CRX+ nuclei and well-defined eGFP-labeled projections that extended to the OV surface ( Fig. 4A-C). Most of the eGFP+/CRX+ cell bodies were located within a thin band that was separated from the OV surface by a similarly thin row of eGFP−/CRX+ cells. As expected, eGFP+ cells were also immunopositive for the rod-specific transcription factors NRL and NR2E3 (Fig. 4D-I). To confirm co-expression of eGFP and NRL, manual counts were performed on OV sections. At this stage, 95.7 ± 0.02% of eGFP+ cells had pronounced NRL expression (n = 3; mean ± standard deviation). Conversely, 98.8 ± 0.02% eGFP labeling of rods in mid-stage NRL +/eGFP OVs. Markedly increased eGFP expression was noted by d145 of OV differentiation (Fig. 5), coincident with an increase in the thickness of the NRL+ and NR2E3+ rod layer (6-8 nuclei) in the developing outer nuclear layer (ONL). All eGFP+ cells were immunopositive for NR2E3 and immunonegative for the cone-specific marker ARR3 ( Fig. 5A-E). To further examine eGFP expression in maturing rods, rhodopsin (RHO) expression was analyzed via immunocytochemistry (ICC). At d145, RHO was expressed in some, but not all NRL+ rods, but all RHO+ cells co-expressed eGFP ( Fig. 5F-H). Manual counts confirmed that all eGFP+ cells co-expressed NRL, while 98.8 ± 0.01% NRL+ cells co-expressed eGFP+ (n = 3 for each). RHO expression was more variable, with eGFP+ cells co-expressing RHO in 67.7 ± 0.24% of cells (n = 3). Cone opsin ICC was also performed on NRL +/eGFP OV sections ( Fig. 5I-P), with medium/long-wavelength cones (OPN1MW/OPN1LW) found primarily in a highly organized layer along the apical portion of OVs (Fig. 5K). OPN1MW/OPN1LW+ cones were also present deep to the eGFP+/NRL+ rod layer where they lacked the organized structure found in the overlying ONL (Fig. 5K,L). Similarly, short wavelength cones (OPN1SW) were found predominantly in the apical ONL ( Fig. 5O,P). In all sections, eGFP co-localized with NRL and NR2E3 but not with cone opsins. eGFP labeling of rods in late stage NRL +/eGFP OVs. Lastly, eGFP fluorescence in NRL +/eGFP OVs was examined relative to the expression of markers of retinal maturity at time points ≥d185 (Fig. 6). At day 185, 98.9 ± 0.01% of eGFP+ cells co-expressed NRL while 100% of NRL+ cells co-expressed eGFP (n = 4). Immunolabelling for the rod bipolar cell (BPC) marker PKCα ( Fig. 6A-C) and the cone and rod BPC marker VSX2 marker (Fig. 6I) was found directly beneath the NRL-eGFP+ ONL, reminiscent of the inner nuclear layer (INL) observed in vivo. Of note, retinal ganglion cells (RGCs) were largely absent by this time in culture, possibly due to lack of diffusion and metabolic support to the innermost portion of the OV where RGCs reside (see Fig. 2H). To look for synapse formation and possible development of an outer plexiform layer (OPL) in NRL +/eGFP OVs, VGLUT1 ICC was performed at d185. VGLUT1 labeling was found on eGFP+ rod terminals (Fig. 6D-F) within an OPL-like band between the ONL and INL. VGLUT1+ puncta were also present that did not co-localize with eGFP, consistent with its expression in both cone and rod terminals. At d230 of differentiation, RHO immunostaining was present primarily within elongated outer segments that extended from the OV surface, where it also co-localized with eGFP ( Fig. 6G-I; compare to Fig. 5G). At this stage, 88.1 ± 0.11% (n = 5) of eGFP+ rods co-expressed RHO, and eGFP and NRL remained co-expressed in virtually all cells (percentage of cells eGFP+ that are also NRL+ = 99.6 ± 0.01%; percentage of NRL+ cells that are also eGFP+ = 98.1% ± 0.04%; n = 4). An INL was maintained at d230 as shown by the presence and location of VSX2 + BPCs (Fig. 6I). To confirm that the OV surface projections corresponded to outer segments, transmission electron microscopy was performed at d230. Mitochondria-rich inner segments were found along the OV periphery (Fig. 6J-L), many of which were attached to developing outer segments via a connecting cilium (Fig. J-M).

Discussion
In this study, we produced the first rod-specific hPSC reporter line by knocking the eGFP transgene into an endogenous NRL locus, a strategy that increases the likelihood that reporter expression will faithfully mirror that of NRL. This targeted knock-in strategy is also novel among mammalian rod reporter lines, since the widely used Nrl-GFP mouse was created by inserting multiple copies of an Nrl-GFP transgene randomly in the murine genome 1 . While our approach does create a nonfunctional NRL allele, no phenotypic consequences are observed in human patients heterozygous for NRL loss of function mutations 29,30 .
The WA09 NRL +/eGFP line strongly labels NRL+ rods throughout OV differentiation, demarcating the entire cell from axon terminal to outer segment. The first committed NRL-eGFP rods arose at d68 in differentiating OVs, consistent with the time course of human prenatal rod development 31 . Thereafter, the NRL +/eGFP rod population expanded to account for the bulk of the OV cell population, with approximately half of all cells expressing eGFP and other rod markers by d130. Mature eGFP+ rods developed pronounced RHO + outer segments that populated the apical surface of OVs and became more prominent with time. These findings are consistent with previous reports of 3D retinal organoids produced from hPSCs [24][25][26][27][28] , and are indicative of a well-organized ONL. Unlabeled cones were also readily generated in NRL +/eGFP OVs, typically near the apical surface of the ONL where they are also located in vivo. Rods produced in NRL +/eGFP OVs possessed VGLUT1+ axon terminals that aligned themselves within an OPL-like layer, further demonstrating the capacity of NRL +/eGFP OVs to generate polarized photoreceptor progeny within highly organized laminae. However, not all cells were correctly positioned within laminar structures, as occasional clusters of cone and rod photoreceptors were also present within the INL-like layer.
Given these findings, the NRL +/eGFP reporter line has potential to play a valuable role in facilitating basic and translational studies using hPSC-OV technology. For example, the capacity to specifically label all rods provides a platform to study human rod development or test protocols that bias toward or against rod production relative to cones. In addition, mutations in genes known to cause primary rod dysfunction and death (e.g., retinitis pigmentosa) could be edited into the NRL +/eGFP line, allowing rapid and accurate assessment of rod-specific outcomes relative to the unedited, isogenic NRL +/eGFP line.
Access to an unlimited supply of fluorescently labeled, hPSC-derived rods may also accelerate the development of rod replacement strategies for patients with retinitis pigmentosa and other rod degenerative diseases. While labeled cells generated by reporter lines would presumably not be used directly in therapies, they could serve as a powerful tool for proof-of-concept studies and for tracking rod fate post-transplantation in animal models. Caution would need to be exercised in such preclinical studies to distinguish truly integrated donor cells  from host cells taking up eGFP by biomaterial transfer [32][33][34][35] . Toward this end, the ability to use human-specific primary antibodies to identify hPSC-derived donor cells in xenografts would offer an added level of confidence when interpreting results.

Methods
Generation of the WA09 NRL +/eGFP reporter line. The following fragments were amplified to create the donor plasmid for Cas9D10A-mediated homologous recombination: (1) an 800 bp fragment 5′ of the NRL start codon (amplified from WA09 hESC genomic DNA) engineered with flanking KpnI and SalI restriction enzyme sites, (2) an 837 bp fragment 3′ of the NRL start codon (amplified from WA09 hESC genomic DNA) engineered with flanking BamHI and NotI sites, and (3) eGFP coding sequence with the rabbit beta-globin polyA terminator flanked by SalI and MfeI sites (amplified from the plasmid AAVS1-Pur-CAG-EGFP; Addgene #80945). The three amplicons were ligated into the PL552 plasmid backbone (Addgene #68407) containing a floxed puromycin resistance (PuroR) gene expression cassette, and the SalI cloning site located between the ATG and the eGFP coding sequence was eliminated by site-directed mutagenesis using the NEB Q5 Site-directed Mutagenesis kit. The resulting plasmid (Fig. 1A) was fully sequenced to confirm correct assembly and sequence. Two sgRNAs targeting opposite strands flanking the NRL start codon (TGACATATTCCATGGCCAGG and GTAAAGCGGGAACCCTCTGA) were identified using the online crispr.mit.edu tool and cloned into the Cas9 sgRNA plasmid (Addgene #68463) according to the published protocol 36 . A full list of cloning primers can be found in Supplemental Table 1 . Following electroporation, cells were plated on MEF feeder cells in 0.5 μM ROCK inhibitor and 5 µM L-755,507, a β3-adrenergic receptor agonist that has been shown to bias cells toward homologous repair over non-homologous end joining 37 . At 72 hours post-electroporation, cells were treated with puromycin (0.5 µg/mL, Invivogen, ant-pr-1) in MEF-conditioned hESC media to select for cells incorporating the plasmid. Puromycin treatment was increased to 1 µg/mL on day (d) 16 post-electroporation and maintained at this level until colonies were large enough for passaging. Puromycin was removed and 0.5 µM ROCK inhibitor was added 24 hours prior to passaging and screening clones. Individual colonies were manually selected, genomic DNA was isolated from a portion of each colony using QuickExtract DNA Extraction Solution 1.0 (Epicentre), and the presence of a single insertion event at only one NRL locus was confirmed by PCR analysis (discussed in the Results section below). One clone was selected and the NRL loci were fully sequenced to confirm targeted insertion. In addition, the top ten predicted off-target sites (identified using the online crispr.mit.edu tool) for each sgRNA were PCR-amplified and sequenced to confirm that no non-targeted editing occurred. Genotyping, off-target primer sequences, off-target sequences, and off-target amplicons can be found in Supplemental Information.

CRE/LoxP recombination.
To enhance reporter gene expression, the floxed PGK-PuroR cassette was removed using CRE-mediated recombination. Briefly, hESCs were digested with TrypLE Express enzyme (Life Technologies) and harvested in 50 µl DMEM/F12 with 0.5 µM ROCK inhibitor. 10 µL of TAT-CRE recombinase (Excellgen) was then added to the cells and incubated for 20 minutes at 37 °C. Subsequently, cells were triturated to ensure single-cell dispersion and plated in 6-well plates for clonal selection in hESC media with 0.5 µM ROCK inhibitor. Individual colonies were manually selected and PCR verified for loxP recombination and precision removal of the PuroR cassette.
Immunocytochemistry. NRL +/eGFP OVs were processed for immunocytochemistry (ICC) as previously described 38,[40][41][42] . Briefly, OVs were fixed in 4% paraformaldehyde for 35 min, washed in PBS, cryoprotected for 1 hr at RT in 15% sucrose followed by overnight at 4 °C in 30% sucrose, and cryosectioned. For immunolabeling, frozen sections were first blocked (10% NDS, 5% BSA, 1% fish gelatin, and 0.5% Triton-X in PBS) for 1 hr, followed by incubation with primary antibodies overnight at 4 °C (Supplemental Information). On d2, primary antibodies were removed with PBS washes and sections were incubated with secondary antibodies at 1:500 (Alexa Fluor, Thermo Fisher) for 30 min. After washing and mounting with Prolong Gold (Thermo Fisher), samples were imaged on an A1R-Si confocal microscope (Nikon). Manual counts were performed using single optical plane confocal micrographs to determine co-expression of eGFP, NRL, and RHO (n ≥ 3 per time point).
High content imaging analysis (HCIA). NRL +/eGFP hESC-OVs were dissociated with papain (Worthington Biochemical Corp.) and plated on 96 well plates coated with poly-L-lysine at 40,000 cells/well. ICC was performed and cell counting was done using unbiased HCIA (Operetta, Perkin Elmer). 20 fields were captured in each well of a 96 well plate at 20X magnification (n = 6 per time point) and images were imported into Columbus software for analysis (Perkin Elmer). Pyknotic nuclei were excluded based on size, shape, and DAPI fluorescence intensity, as were all cells touching borders. Positive cells were determined by scatter plot gating of fluorescence intensity. Identical parameters for positive cell identification were used across samples for consistency.
Transmission electron microscopy. Whole OVs were fixed overnight at 4 °C in 3% glutaraldehyde/1% paraformaldehyde in 0.08 M sodium cacodylate buffer, washed in 0.1 M cacodylate, and post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate for 2 hrs at RT. Samples were subsequently dehydrated in a graded ethanol series, further dehydrated in propylene oxide, and embedded in Epon eposy resin. Semi-thin (1 µm) sections were cut with a Leica EM UC6 Ultramicrotome and examined under a light microscope to establish proper orientation. Ultra-thin sections were cut with the same microtome and collected on pioloform-coated 1 hole slot grids (Ted Pella Inc, cat # 19244). Sections were contrasted with Reynolds lead citrate and 8% uranyl acetate in 50% EtOH. Ultrathin sections were imaged with a Philips CM120 electron microscope equipped with an AMT BioSprint side mounted digital camera using AMT Capture Engine software.
Data availability statement. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.