Early cis-regulatory events in the formation of retinal horizontal cells

During retinal development, multipotent and restricted progenitor cells generate all of the neuronal cells of the retina. Among these are horizontal cells, which are interneurons that modulate the light-induced signal from photoreceptors. This study utilizes the identification of novel cis-regulatory elements as a method to examine the gene regulatory networks that direct the development of horizontal cells. Here we describe a screen for cis-regulatory elements, or enhancers, for the horizontal cell-associated genes PTF1A, ONECUT1 (OC1), TFAP2A (AP2A), and LHX1. The OC1ECR22 and Tfap2aACR5 elements were shown to be potential enhancers for OC1 and TFAP2A, respectively, and to be specifically active in developing horizontal cells. The OC1ECR22 element is activated by PTF1A and RBPJ, which translates to regulation of OC1 expression and suggests that PTF1A is a direct activator of OC1 expression in developing horizontal cells. The region within the Tfap2aACR5 element that is responsible for its activation was determined to be a 100 bp sequence named Motif 4. Both OC1ECR22 and Tfap2aACR5 are negatively regulated by the nuclear receptors THRB and RXRG, as is the expression of OC1 and AP2A, suggesting that nuclear receptors may have a role in the negative regulation of horizontal cell development.


Introduction 20
The vertebrate retina is comprised of six major classes of neuronal cells. have 2-4 HC types (Fischer et al., 2007;Mariani and Leure-DuPree, 1977), while only 1 HC 45 subtype has been identified in the mouse retina (Peichl and González-Soriano, 1994) and 3 HC 46 types have been reported in the human retina Kolb et al., 1992Kolb et al., , 1994. 47

48
The gene regulatory network (GRN) that underlies and directs the development of HCs has yet 49 to be fully elucidated. FOXN4, a winged helix/forkhead transcription factor (TF), has been 50 shown to be a key factor for the fate commitment of HCs, and accordingly, there is a complete 51 absence of HCs in the FOXN4 knockout (KO) (Li et al., 2004). A downstream target of FoxN4 is 52 the bHLH factor PTF1A, whose loss leads to the same HC phenotype (Fujitani et al., 2006). 53 These factors are both necessary and sufficient for the generation of HCs but are similarly 54 required for the formation of ACs, which are drastically decreased in the FOXN4 and PTF1A 55 knockouts (Fujitani et al., 2006;Li et al., 2004). An additional series of TFs are therefore 56 necessary to specify the HC fate. Several other genes have been reported to be important for 57 HC development, which include ONECUT1 (OC1), TFAP2A/B (AP2A), and LHX1 (LIM1). More 58 than 80% of HCs fail to develop upon the conditional KO of OC1 (Wu et al., 2013)  order to obtain a more accurate assessment of the cells with enhancer activity, we utilized a 115 lineage tracing system to convert reporter activity into a constitutive signal upon initial activity 116 of the DNA element and thereby provide a history of enhancer activity. This system has 117 previously been shown to have no leaky expression in control conditions (Schick et al., 2019). 118 The 10 elements with activity in a HC-like pattern were cloned into lineage tracing plasmids and 119 electroporated into E5 chick retinas alongside CAG::nβgal as an electroporation control ( Figure  120 2A). After 2 days in culture, we counterstained the retinas for Visinin or for LIM1, a specific 121 marker of some HCs. The lineage tracing revealed three patterns of enhancer-driven activity: 1) 122 some elements, such as Lhx1ACR1, drove reporter expression non-specifically throughout the 123 retina; 2) others, such as Tfap2aACR4, drove similar amounts of reporter expression in the 124 outer and inner limits of the retina; 3) the remaining elements, such as Tfap2aACR5, drove 125 reporter expression largely in the inner retina ( Figure 2B-C). This last category is comprised of 126 Ptf1aECR12, Tfap2aACR2, Tfap2aACR5, OC1ECR22, and OC1ACR8. These 5 elements drove 127 reporter expression in cells organized in the inner retina that did not have Visinin expression,128 and that showed some colocalization with LIM1 ( Figure 2B-C, Figure S3). This suggests that 129 these elements are specifically activated in cells that will become HCs and not photoreceptors. 130

Characterization of enhancer-driven activity 132
We further characterized the activity of these elements by electroporating the lineage tracing 133 plasmids into E5 chick retinas and dissociating the cells after three days in culture. We then 134 counterstained the cells for a variety of markers for all cell types in the retina, and using flow 135 cytometry, we quantified the percentage of cells with expression of each marker among the 136 population labeled with GFP ( Figure 3A). We first stained the dissociated cells for LIM1 and 137 AP2A. LIM1 is expressed in H1 HCs, which represent 52% of HCs in the chick, and AP2A marks 138 those HCs as well as ACs (Fischer et al., 2007). Therefore, the cells with expression of both LIM1 139 and AP2A were identified as HCs, while the cells with expression of AP2A alone were 140 considered to be ACs. All 5 enhancers are active in H1 HCs, which represent a range of 14-26% 141 of cells with enhancer activity ( Figure 3B). Ptf1aECR12 had substantial activity in ACs, while ACs 142 represent only 1-7% of the other enhancer populations ( Figure 3C). 143 144 Next, cells were stained for ISLET1/2 and OTX2. ISLET1 is expressed in a variety of cell types 145 including H2-H4 HCs, some ACs, bipolar cells, RGCs and photoreceptors, while ISLET2 marks a 146 subset of photoreceptors and RGCs (Edqvist et al., 2006). OTX2 is expressed in photoreceptors 147 and bipolar cells (Koike et al., 2007), and has also been shown to be expressed in at least two 148 types of restricted retinal progenitor cells (Emerson et al., 2013;Ghinia Tegla et al., 2020). 149 Therefore, cells that expressed ISLET1/2 alone were considered to be Islet1-expressing cells, 150 identified as H2-H4 HCs, ACs or RGCs. The cells with expression of ISLET1/2 and OTX2, as well as 151 those that only expressed OTX2, were identified as ISLET2-negative photoreceptors, bipolar 152 cells or restricted progenitor cells. All 5 enhancers had activity in ISLET1 cells, which represent a range of 14-37% of cells with enhancer activity ( Figure 3D). Only Tfap2aACR2 had substantial 154 activity in OTX2 cells, which was expected based on the qualitative assessment of Tfap2aACR2 155 lineage tracing ( Figure S3), while OTX2-positive cells represent 2-12% of the other enhancer 156 populations ( Figure 3E). Lastly, cells were stained for BRN3A which, in the mouse, marks 70% of 157 retinal ganglion cells (Xiang et al., 1995). All enhancers had minimal activity in these RGCs, 158 which represent a range of 0-3% of cells with enhancer activity ( Figure 3F). These experiments 159 were repeated and analyzed by confocal microscopy, yielding the same qualitative results 160 ( Figure S4). Upon these analyses, we have shown that Ptf1aECR12 is active almost entirely in 161 HCs and ACs and that Tfap2aACR2 is largely active in HCs and cells expressing OTX2, while 162 Tfap2aACR5 activity is more restricted and is primarily active in HCs. Most of OC1ACR8 activity 163 is in HCs, although there is also activity in ACs and cells expressing OTX2, and finally, the activity 164 of OC1ECR22 is nearly exclusive to HCs. 165 166 PTF1A and RBPJ regulation of the OC1ECR22 element and OC1 expression 167 We sought to assess whether the activity of OC1ECR22 is consistent with its serving as an 168 enhancer for OC1. In fact, 74.4% ±2.96 of cells with OC1ECR22 activity express OC1, as 169 compared to only 18% ±1.56 of a control population (mean ± SEM, n=3, p<0.0001) ( Figure 4A). 170 This indicates that OC1ECR22 is likely an enhancer for the OC1 gene, and we were therefore 171 interested in determining the transcription factors that activate the enhancer and thereby 172 regulate OC1 expression. We performed motif analyses on the OC1ECR22 sequence with the 173 MEME Suite to identify potential binding sites, and also searched manually through the 174 sequence for additional relevant sites. Among the TF consensus sites that we identified were those for PTF1A and RBPJ ( Figure 4B), two factors that have been shown to function as part of a 176 trimeric complex along with an E-protein (Beres et al., 2006). These sites are separated by 5 177 nucleotides, which is in accordance with previously described PTF1A (E-Box) and RBPJ (TC-Box) 178 binding sites separated by one helical turn of DNA (Masui et al., 2008). In addition, these factors 179 have previously been shown to work together to regulate HC development (Lelièvre et  Tfap2aACR5 is another element that is predominantly active in HCs. Furthermore, 40.62% ±2.73 220 of cells with Tfap2aACR5 activity express AP2A, compared to only 11.1% ±0.7 of a control 221 population (mean ± SEM, n=4, p<0.0001) ( Figure 5B). This indicates that Tfap2aACR5 is likely an 222 enhancer for the TFAP2A gene, and we were again interested in determining the factors that 223 activate the enhancer. We created truncated versions of the enhancer so that the required 224 portions of this 1.4 kb sequence could be identified. Five motifs of evolutionary conservation 225 were identified within this sequence, and are likely to contain the sites required for the activity 226 of Tfap2aACR5. We created a series of truncations within Tfap2aACR5 in the context of a GFP 227 reporter plasmid and individually co-electroporated each one into E5 chick retinas with a full-228 length Tfap2aACR5 TdTomato reporter and CAG::iRFP as an electroporation control, and 229 analyzed them by flow cytometry after two days in culture. We first deleted a region that 230 contained motifs 1, 2, and 3, and observed no change in Tfap2aACR5 activity. We next 231 extended this deletion to include motif 4, and observed a significant decrease in reporter (mean ± SEM, n=3, p<0.05) ( Figure 5A). Nevertheless, to test the requirement of motifs without 242 altering the distance between remaining motifs and the basal promoter, we created two 243 truncations of the reversed Tfap2aACR5. When a region that includes motifs 4 and 5 was 244 deleted, there was again a decrease in the activity of the enhancer. When a region that includes 245 motifs 1, 2, and 3 was deleted, we saw a significant increase in the activity of Tfap2aACR5, 246 potentially indicating that repression mediated through those motifs was relieved. The reversed 247 Tfap2aACR5 was active in 0.51% ± 1.3 of electroporated cells and the truncated version was 248 active in 1.8% ± 0.004 of electroporated cells (mean ± SEM, n=3, p<0.01) ( Figure 5A). These 249 analyses identified a 100 bp region necessary for Tfap2aACR5 activation, and have introduced 250 the possibility of repressors binding to negatively regulate enhancer activity. 251 252

Nuclear receptors negatively regulate HC enhancer activity and gene expression 253
Among the sites identified upon the initial motif analysis of Tfap2aACR5 were half-sites that 254 could be bound by several nuclear receptors, which include THRB, RXRG, ESRRG, and RORB. 255 However, the mutation of these sites did not result in any change in the activity of the enhancer 256 ( Figure S7). In parallel, we overexpressed THRB and RXRG in combination with Tfap2aACR5 or 257 OC1ECR22 in E5 chick retinas, which were then analyzed by flow cytometry after two days in 258 culture. We observed significantly decreased activity of both enhancer elements, with 259 Tfap2aACR5 activity decreasing from 3.14% ±0.48 to 0.19% ± 0.04 of electroporated cells, and 260 OC1ECR22 activity decreasing from 20.51% ±0.5 to 4.14% ±0.7 of electroporated cells upon 261 overexpression (mean ± SEM, n=3 or 4, p<0.001) ( Figure 5C). Furthermore, when we stained the 262 dissociated cells for AP2A and for OC1 we observed similar decreases. The overexpression of 263 THRB and RXRG resulted in 2.32% ±0.19 of cells expressing AP2A, compared to 6.67% ±0.59 of 264 cells in the control condition, and 5.83% ±0.99 of cells expressing OC1 compared to 13.52% 265 ±0.99 of cells in the control condition (mean ± SEM, n=3 or 4, p<0.01) ( Figure 5C). This decrease 266 in AP2A and OC1 expression is presumably mediated by the decreased enhancer activity. These 267 data introduced the possibility that these nuclear receptors are serving as negative regulators 268 of HC genes in developing photoreceptors, although this interaction could be indirect given that 269 the identified half-sites in Tfap2aACR5 were not required for this effect. 2007). Therefore, the strict screening measures described above are crucial to ensure that 308 these novel elements can be used to specifically mark developing HCs. Interestingly, we found 309 that several elements drove expression of reporters in cell types that are unexpected when 310 considering the genes that they are located near. This was the case for elements selected near 311 all four targeted genes and is especially striking in the case of LIM1 which is exclusively 312 expressed in HCs. It is of course possible that although the DNA elements in question were 313 selected based on proximity to PTF1A, OC1, TFAP2A or LHX1, they may not be involved in the 314 transcriptional regulation of those genes but of others that are more distally located. Another 315 potential explanation is that there are several elements for each gene that must be active in 316 combination to lead to transcriptional regulation in specific cell types, so that the pattern of hormone receptor binding sites within the Tfap2aACR5 sequence were dispensable to its 382 regulation. We were interested in observing the loss-of-function phenotype for THRB and 383 RXRG, and performed CRISPR/Cas9 gene editing of these genes via electroporation (data not 384 shown). We did not detect any changes in the photoreceptor or HC populations which may 385 indicate that the effects observed with ectopic expression of THRB and RXRG does not reflect 386 their in vivo function, but could also be due to a number of technical limitations. For example, 387 we are currently unable to validate loss of THRB and RXRG protein in cells using antibodies and 388 therefore cannot determine the efficacy of our gene editing strategy to create loss-of-function 389 alleles. Another possibility is that THRB and RXRG are not directly involved in HC gene 390 regulation, but their overexpression interferes with the normal function of a nuclear receptor, 391 such as RORB, that is critical for HC differentiation. Further experiments will be necessary to 392 clarify the role of THRB and RXRG in relation to HC gene expression during retinal development.

Methods 394
Animals 395 Fertilized chicken eggs were obtained from Charles River and were stored at 16 °C for 0-10 396 days. The eggs were then incubated in a 38 °C humidified incubator for 5 days.

Mutagenesis and Truncation Assays 438
To create mutations within the OC1ECR22 and Tfap2aACR5 sequences, scrambled DNA 439 sequences were first generated using the Shuffle DNA tool from the Sequence Manipulation 440 Suite (Stothard, 2000). These sequences were incorporated into a set of primers that spanned 441 the potential TF binding sites to create mismatches. A second set of primers included the 442 outermost portions of the CRM sequence as well as restriction sites to facilitate cloning. The 443 mutations were created using overlap extension PCR with the initial CRM plasmid serving as a 444 template, and the mutated sequences were cloned into Stagia3 or Statia plasmids. 445

446
To create truncated versions of Tfap2aACR5, primers were designed so that a portion of the 447 sequence would be excluded during amplification. The truncated sequences were PCR-448 amplified using the initial CRM plasmid as a template, and the amplicons were digested using 449 Ecor1 and Sal1 whose sites were appended to the primer sequences. The truncations were then 450 After fixation, retinas were incubated with 1 mL of NTM buffer (100 mM NaCl, 100 mM Tris pH 518 9.5, 50 mM MgCl2) shaking at room temperature. After 15 minutes, the NTM was replaced with 519 1 mL NTM with 0.25 mg/mL NBT (VWR, 97061-412) and 0.125 mg/mL BCIP. The retinas were 520 incubated in the dark with these substrates, shaking for 2-3 hours. A positive control was used 521 to determine that the development of the AP stain was complete. 522 523 524 Imaging and Image Processing 526 Images of AP stained retinas were acquired with a Zeiss AxioZoom V16 microscope using a 527 PlanNeoFluarZ1x objective. All confocal images of vertically sectioned retinas were acquired 528 with a Zeiss LSM710 inverted confocal microscope and ZEN Black 2015 21 Sp2 software. Images 529 were acquired at 1024 x 1024 resolution with an EC Plan-Neofluar 40X/1.30 Oil DIC M27 530 objective. Images were analyzed with FIJI (Schindelin et al., 2012). Cells were counted using the 531 Cell Counter plugin for ImageJ in Z-stacks from a minimum of three retinas in each condition. All 532 figures were assembled using the Affinity Designer vector editing program, and any 533 adjustments to brightness and contrast were applied uniformly across images. Schematics were 534 assembled with Adobe Illustrator. 535 536

Statistical Analysis 537
Graphs were made using GraphPad Prism8 software version 8.4.2, and error bars represent 538 standard error of the mean (SEM). Statistical analyses were also performed using GraphPad 539 Prism 8 software. Data was first tested for normality using the Shapiro-Wilk test. When data 540 was found to be distributed normally, an unpaired t-test with two tailed distribution or one-way 541 ANOVA with Dunnett's multiple comparison test was performed, as noted. 542 543 Acknowledgements 544 Support was provided by National Science Foundation grant CAREER 1453044 (to MME), 545 National Eye Institute grant R01EY024982 (to MME) and NIMHD 3G12MD007603-30S2 (CCNY). 546 The content is solely the responsibility of the authors and does not necessarily represent the 547