Molecular basis of CRX/DNA recognition and stoichiometry at the Ret4 response element

SUMMARY

Two retinal transcription factors, CRX and NRL, cooperate functionally and physically to control photoreceptor development and homeostasis. Mutations in CRX and NRL cause severe retinal diseases. Despite the roles of NRL and CRX, insight into their functions at the molecular level is lacking. Here, we have solved the crystal structure of the CRX homeodomain in complex with its cognate response element (Ret4) from the rhodopsin proximal promoter region. The structure reveals an unexpected 2:1 stoichiometry of CRX/Ret4 and unique orientation of CRX molecules on DNA, and it explains the mechanisms of pathogenic mutations in CRX. Mutations R41Q and E42K disrupt the CRX protein-protein contacts based on the structure and reduce the CRX/Ret4 binding stoichiometry, suggesting a novel disease mechanism. Furthermore, we show that NRL alters the stoichiometry and increases affinity of CRX binding at the rhodopsin promoter, which may enhance transcription of rod-specific genes and suppress transcription of cone-specific genes.

eTOC Blurb

Srivastava et al. determined the crystal structure of the CRX homeodomain in complex with its Ret4 response element from the rhodopsin promoter. This study provides structural insights into the recognition of DNA by CRX, its function and the cooperative DNA binding by CRX and NRL.

Graphical Abstract:

INTRODUCTION

Two key transcription factors, cone–rod homeobox (CRX) and neural retina leucine zipper (NRL), play fundamental roles in photoreceptor cell differentiation and functional maintenance ^1,2. CRX, a homeodomain transcription factor (TF) of the paired type, is expressed in postmitotic photoreceptor precursors, where it induces expression of hundreds of photoreceptor genes involved in photoreceptor morphogenesis and function ^3–5. In the Crx knock-out (KO) mouse retina, photoreceptors are specified but are nonfunctional, fail to form outer segments (OS), and undergo progressive degeneration ⁶. NRL, a basic motif–leucine zipper (bZIP) TF of the Maf family, is expressed soon thereafter during photoreceptor development and determines the rod cell fate ^1,7–9. In coordination with CRX and other TFs, NRL induces the expression of a majority of rod-specific genes while simultaneously suppressing cone-specific gene expression in coordination with NR2E3 ^1,9,10. Mouse retina lacking NRL has no rods and instead contains S cone-like photoreceptors ¹⁰. Inversely, expression of NRL in photoreceptor precursors produces rods ¹¹. An essential aspect of the control of photoreceptor development by CRX and NRL is transcriptional synergy between these two TFs that has been demonstrated for several genes, including rhodopsin (2). Integrated bioinformatic analyses of RNA-seq, ChIP-seq, ATAC-seq, and HiC datasets have identified hundreds of human retina-expressed genes that are potentially controlled by synergistic action of CRX and NRL ^12–18. CRX and NRL cooperate not just functionally but physically as well via direct protein-protein interaction ⁹.

Defects in transcriptional regulation in developing and mature photoreceptors lead to retinal degeneration and blindness ^1,5,19,20. Given the key roles of CRX and NRL in photoreceptor development and homeostasis, it is not surprising that mutations in the genes encoding these TFs are responsible for a broad range of visual disorders. Mutations in the gene encoding CRX are associated with retinitis pigmentosa (RP), cone-rod dystrophy (CORD), and Leber congenital amaurosis (LCA) ^5,19,21–27, whereas NRL mutations are associated with retinopathies and enhanced S-cone syndrome ^28–35. The significance of the biology and the clinical relevance of CRX and NRL has led to intense investigation of the genomic binding profiles, transcriptional activities, and gene-regulatory networks of these TFs ^{1,12,13,36–39}. Many human disease-associated mutations in CRX and NRL have been characterized biochemically, and the phenotypic consequences are documented in studies of mouse models ^40–42. However, atomic structural information on the interaction of these TFs with respective cis-regulatory elements (CREs) in photoreceptor promoters is lacking, leading to poor understanding of molecular mechanisms underlying the function and synergy between CRX and NRL.

Here, we have solved the crystal structure of the human CRX homeodomain (HD) in complex with its cognate response element (Ret4) from the rhodopsin proximal promoter region (RPPR). The crystal structure shows that CRX interacts with Ret4 with an unexpected stoichiometry of 2 to 1 and causes DNA bending. We validated this unusual stoichiometry of CRX binding by Size exclusion chromatography-Multi angle light scattering-Small Angle X-ray scattering (SEC-MALS-SAXS). Importantly, the structure helps to elucidate the mechanisms of pathogenic mutations in the CRX HD. In particular, we examined the R41Q, E42K, and R90W mutations of CRX that underlie CORD, LCA and RP ^9,19,43,44. Interestingly, we found that the R41Q and E42K mutations disrupt the stoichiometry of CRX HD binding to Ret4. In agreement with the structure-based prediction, we also show the R90W mutation causes severe impairment of the protein stability rather than it is affecting CRX binding to DNA. Our analysis of the interplay of NRL and CRX in vitro revealed binding of NRL to its response element (NRE) in rhodopsin promoter changes the stoichiometry of CRX binding to Ret4 to 1:1. Finally, we examined CRX/DNA interaction and the synergy with NRL using single-molecule total internal reflection fluorescence microscopy (smTIRFM). This analysis provided direct evidence for the increased affinity and “ON’ DNA dwell times of CRX in the presence of NRL. Our study establishes a framework for future structural studies of the CRX/NRL cooperativity.

RESULTS

Crystal structure of the CRX HD bound to Ret4 response element.

The structure of the CRX HD/Ret4 complex was solved at a 2.9 Å resolution by molecular replacement (MR) using single HD with cognate DNA from the structure of the HD of Drosophila Paired protein (PDB 1FJL) as a search template (Suppl. Table 1). Ret4 was used because it is adjacent to and immediately follows the NRE site in RPPR. Rhodopsin is the major photoreceptor protein, and its promoter including the NRE/Ret4 sites is the classical model for transcriptional regulation ^4,8,45,46. There are four molecules of CRX HD and two molecules of duplex Ret4 DNA in the asymmetric unit with two molecules of CRX HD bound to each Ret4 DNA (Fig. 1A, Suppl. Fig. 1). Limited interactions between the two (CRX)₂/DNA complexes do not appear to affect either protein or DNA conformation. The CRX HD is a globular monomer in solution ¹⁵. Thus, dimerization of CRX HD in the CRX-Ret4 complex is primarily mediated by DNA binding. Unlike the paired homeodomains bound to palindromic P2 or P3 sites, the Ret4 site is not palindromic and the two CRX molecules are bound to two different sites in a unique head-to-head orientation. Binding of CRX to Ret4 site results in the significantly more DNA bending (23 degrees) compared to the bending by Pax 3 (11 degree) and Airstaless/Clawless (9 degrees) (Suppl. Fig. 2). DNA bending has previously been suggested as a regulatory mechanism to control CRX binding to DNA ⁴⁷.

Figure 1. Structure of the CRX HD-Ret4 complex.

A. An overview. The structure features two molecules of CRX HD bound to DNA. CRX HD is rainbow colored with the N-terminus in blue and the C-terminus in red. α3-helix of the CRX HD binds to the major DNA groove while the N terminal arm inserts into the minor groove. B. Interactions of CRX HD chain A with nucleotide bases. R43 makes hydrogen bonding interactions with adjacent T and A (TAAG), whereas N89 makes hydrogen bonding interaction with A (TAAG) in the antisense strand. C. Interactions of CRX HD chain B with nucleotide bases. R43 makes contact with TT (bold, TTAG) of sense strand and G of antisense strand (bold, TAAG), whereas N89 makes hydrogen bonding interaction with A (TAAG) in the sense strand.

The two (CRX)₂/DNA complexes in the asymmetric unit are very similar (RMSD of 0.31 Å), except the density for the HD N-termini is slightly better defined in the complex with chains A and B compared to that with chains C and D (Suppl. Fig.1). Accordingly, we describe below the protein/DNA contacts for chains A and B. Both CRX HD molecules in (CRX)₂/DNA complex adopt a classical HD fold and are closely superimposable (RMSD 0.43 Å) (Fig. 1). Similarly, to other HDs, α3-helix of the CRX HD binds to the major DNA groove while the N terminal arm inserts into the minor groove. The first molecule of CRX HD (chain A) buries ~777 Ų of the protein/DNA interface and makes 22 direct hydrogen-bonding contacts with DNA, whereas the second CRX HD (chain B) buries ~832 Ų of the protein/DNA interface and makes 18 direct hydrogen-bonding contacts with DNA. In Chain A, R43 makes hydrogen bonding interactions with adjacent T and A in the antisense strand (bold, TAAG) while in chain B, R43 makes contact with two T (bold, TTAG) of sense strand and G of antisense strand (bold, TAAG) (Fig. 1BC). N89 in both chains A and B makes hydrogen bonding interaction with A in antisense strand (TAAG) (Fig. 1BC). Also in both chains, K88 side chain is dynamic and its specific interaction with nucleotide bases cannot be deduced from the structure. K88 counterpart in Bicoid HD (K50 using the HD numbering convention) is dynamic and it is important for adaptive DNA recognition ⁴⁸. The recognition sequences for the both CRX HD molecules somewhat differ from the consensus for the CRX binding site in the promoter regions of photoreceptor proteins (^C/_TTAATCC) ⁴. Besides the base pair specific interactions, CRX homeodomain makes several hydrogen bonding and ionic interactions with the phosphate and deoxyribose backbone (Suppl. Fig. 3) ⁴⁹. Side chains of T44, Y63, and Q84 of both chains A and B interact with the phosphate backbone of DNA through hydrogen bonding interactions while R82, R91, and R95 make ionic interactions with the phosphate backbone. In addition to making base specific contacts, R43 of chains A and B interacts with respective sugar backbone (Suppl. Fig. 3).

In both CRX₂/DNA complexes in the asymmetric unit, the N termini of the two CRX HD molecules point towards each other. Furthermore, R41 in chain B is within hydrogen-bonding distance of E42 in chain A (Fig. 2). The B factor for E42 from chain A is 92 and R41 from chain B is 85. These B factors are slightly higher than the average B factor (73) of the entire molecule (Table 1, Suppl Fig. 4). The omit map supports the model where R41 of chain B and E42 of chain A come in close contact and make ionic interaction with each other (Fig. 2). Besides making ionic interaction with the E42 of other chain, R41 also makes ionic interactions with the phosphate backbone. The R41/E42 interaction precludes the insertion of R41 into the minor groove and prevents specific interaction with nucleotide base at Ret4 site. However, this allows R41 to contribute to CRX HD dimerization on DNA at this or similar sites.

Figure 2. Interaction between the N-termini of two CRX HD molecules bound to Ret4.

The N termini of the two CRX molecules come into close proximity. R41 in chain B is within hydrogen-bonding distance of E42 in chain A (2.6 Å). An omit map (Fo−Fc) for N terminal CRX residues (residue 41–43 of chain A and 39–43 of chain B) is contoured at 2.2σ and shown for 1.6 Å around the omitted residues (blue mesh). The omit map supports the interaction.

Table 1.

Single-molecule TIRF microscopy summary

S.No.	Complex (Protein-DNA)	DNA	No. of traces taken	No. of Events (On Events)	State 1 (von) per sec	State 1 (kon) s −1 M−1	State 2 (koff or ) s−1	State 2 (tau) s	smTIRF Kd=koff/kon
1.	1 nM CRX	500 pM	146	285	0.01	0.01 s−1 nM	0.303 ± 0.011	2.286	55.1 nM
2.	1 nM CRX 2.5 nM NRL	500 pM	101	198	0.01	0.01 s−1 nM	0.218 ± 0.024	4.854	39.63 nM
3.	1 nM CRX 5 nM NRL	500 pM	124	215	0.01	0.01 s−1 nM	0.154 ± 0.013	6.513	28 nM

Validation of the stoichiometry of CRX binding to Ret4.

Binding of two molecules of the CRX HD to Ret4 has not been described or anticipated. Therefore, we sought to confirm that the 2:1 CRX/DNA stoichiometry observed in the crystal structure is not an artifact of crystallization. We examined the stoichiometry of CRX binding to Ret4 by two independent techniques, SEC-MALS and SEC-SAXS. The MALS-calculated molar mass of the major peak (~30.5 kDa) corresponds well to the predicted mass of the DNA-bound complex with 2 CRX HD molecules (Fig. 3A). Using the second approach, the experimental scattering profile obtained by SEC-SAXS is in agreement (χ² =1.53) with the theoretical SAXS profile calculated by CRYSOL for the crystal structure of the CRX HD/Ret4 complex (Fig. 3B, Suppl. Table 2) ⁵⁰. Guinier analysis suggested the R_g value of 22.0± 0.1 and pair distribution function indicated the maximum dimension of 75 Å (Suppl. Fig. 5). Thus, the unexpected 2:1 CRX/Ret4 stoichiometry is also characteristic of the complex in solution.

Figure 3. Analysis of the stoichiometry of the CRX HD-Ret4 complex by SEC-MALS-SAXS.

A. SEC MALS profile of the CRX HD/Ret4 complex. Light-scattering (red) and refractive index (blue) curves are shown together with the molar mass (black) calculated by MALS. The molar mass of the major peak (~30.5 kDa) corresponds well to the predicted mass of the DNA-bound complex with 2 CRX HD molecules. (B) The theoretical SAXS profile for the crystal structure of the CRX HD/Ret4 complex (red) is in agreement (χ² =1.53) with the experimental scattering profile obtained by SEC-SAXS (black).

The CRX HD is most homologous (~86% identity) to the HD of the OTX2, a TF with a role in brain and retina organ development ⁴. OTX2 is proposed to activate NRL in neonatal retina ⁵¹. We have tested the possibility that OTX2 may also bind Ret4 with a 2:1 stoichiometry. Our SEC-MALS analysis demonstrated that this is indeed the case (Suppl. Fig. 6).

The N-terminal mutations in CRX-HD linked to retinal diseases disrupt the stoichiometry of its binding to Ret4.

Mutations at the N-terminus of the CRX HD have been linked to several retinal diseases. The E42K mutation was found in patients with LCA7⁴⁴, whereas the R41W and R41Q mutations were associated with CORD2 and RP, respectively ^9,19,43. Since according to our structure, R41 and E42 may influence the stoichiometry of the CRX/DNA binding, we examined the DNA-bound complexes of the mutant CRX HD proteins, R41Q and E42K, by MALS. This analysis demonstrated that both CRX mutations reduced the CRX HD/Ret4 stoichiometry from 2:1 to 1:1 (Fig. 4A,B). The change in binding stoichiometry was not simply because of the impaired CRX HD/DNA binding affinity. We conducted the BLI binding assays employing biotinylated duplex primers encompassing Ret4 and attached to streptavidin biosensors (Fig. 4C, Suppl. Fig. 7). According to the steady state analysis of the BLI data, the K_D for the CRX HD binding to Ret4 was 0.43 μM. The two mutations had opposite effects on the DNA binding; the R41Q substitution moderately reduced the binding affinity, whereas E42K significantly strengthened the interaction relative to that for WT CRX HD (Fig. 4C, Suppl. Fig. 7). Thus, the altered CRX/DNA binding stoichiometry may underlie the pathogenic effects of the mutations at the N-terminus of the CRX HD.

Figure 4. The N-terminal mutations in CRX-HD disrupt the stoichiometry of its binding to Ret4.

A,B. SEC MALS profiles of the mutant CRX HD proteins in complex with Ret4. Refractive index (blue) curves are shown together with the molar mass (red) calculated by MALS. The molar masses of the major peaks correspond well to the predicted mass of the CRX HD/DNA-bound complexes with 1:1 stoichiometry. C. Steady-state Ret4 binding data for CRX HD, the R41Q and E42K mutants are from on three independent BLI assays for each of the proteins (Mean±SD). The hyperbolic fits yield the following K_D values: CRX HD, 0.43±0.03 μM; R41Q, 1.25±0.03 μM; E42K, 0.03±0.01 μM). Unpaired t test WT versus R41Q, p****<0.0001; WT versus E42K, p***=0.0002. Note that cooperative binding of two CRX HD to Ret4 cannot be excluded on the basis of these BLI data.

Reduced protein stability underlies the pathogenic mechanism of R90W CRX mutation.

Missense mutation R90W in the CRX HD is linked to late-onset autosomal dominant cone-rod dystrophy (adCORD) ^19,20,26. Homozygous R90W causes a more severe LCA phenotype ^20,24,40. Biochemically, this mutation revealed reduced DNA binding and reduced transactivation activity, and the mutant CRX did not appear to interfere with activity of WT CRX ⁴⁰. The crystal structure of the Ret4-bound CRX HD shows that R90 does not directly interact with DNA. Modeling of the R90W mutation into the structure of CRX suggests that it would break the bond between R90 and E55 and cause a steric clash between helices α1 and α3 potentially leading to protein instability (Fig. 5A). We examined the effects of the R90W mutation on the CRX binding to Ret4 and protein thermal stability using Bio-Layer Interferometry (BLI) and dynamic light scattering (DLS), respectively. It revealed an affinity of R90W for the response element at 26°C is comparable to that of WT CRX HD (Fig. 5B). On the other hand, the thermal stability of R90W was reduced by ~20°C compared to that of the WT protein (Fig. 5C). Thus, the origin of the CRX mutant dysfunction is protein instability that leads to protein aggregation at temperatures above 25°C.

Fig. 5. Reduced protein stability of the CRX R90W mutant.

A. The R90W mutation modelled into the structure of the CRX-Ret4 complex. This mutation is breaking the bond between R90 and E55 and causes a steric clash between helices α1 and α3 potentially leading to protein instability. B. Steady-state Ret4 binding data for CRX R90W from three independent BLI assays (Mean±SD). The calculated K_D for R90W/Ret4 interaction is 0.9±0.03 μM. Unpaired t test WT versus R90W, p***=0.0002. C. DLS thermal stability curves. Thermal stability of R90W CRX31–107 is significantly lower than the stability of WT CRX31–107.

NRL alters the stoichiometry of CRX binding to Ret4 site.

To investigate if the 2:1 stoichiometry of CRX binding to Ret4 is maintained when NRL and CRX are simultaneously bound to RPPR, we analyzed the NRL/CRX/DNA complex by SEC-MALS. The molar mass of the major peak (~60 kDa) corresponds well to the predicted mass of the DNA-bound complex with the NRL DNA-binding domain dimer (2 molecules) and one molecule of CRX HD (Fig. 6). This finding suggests that NRL binding to its CRE in RPPR results in binding of only one molecule of CRX to Ret4.

Fig. 6. NRL alters the stoichiometry of CRX binding to Ret4 site.

SEC-MALS profile of the complex of CRX HD and NRL DBD with Ret4-NRE DNA. Refractive index (black) curve is shown together with the molar mass (red) calculated by MALS. The molar mass of the major peak (~60 kDa) corresponds well to the predicted mass of the DNA-bound complex with the NRL dimer (2 molecules) and one molecule of CRX HD.

NRL modulates the CRX binding affinity to NRE/RET4 from the rhodopsin proximal promoter.

To examine the interaction of CRX with RPPR and its cooperativity with NRL we utilized single-molecule TIRF Microscopy (smTIRFM) ^52,53. In the smTIRFM experiments, the biotinylated DNA corresponding to NRE/Ret4 sites from RPPR was tethered to the surface of the reaction chamber, whereas the Cy3-labled CRX HD was infused into the chamber. The time-dependent changes in Cy3 fluorescence in a specific location in the reaction chamber indicated Cy3-CRX binding to and dissociation from individual surface-tethered DNA molecules. These changes were recorded, and the fluorescence trajectories were analyzed using hFRET and KERA ^54–56. Representative trajectories overlaid with the corresponding idealized trajectories are shown in Fig. 7. The trajectories showed direct binding of CRX to the DNA that typically lasted for a few seconds (Fig. 7). No binding was observed in control experiments in the absence of surface-tethered DNA or with random dsDNA. The best model for all data was the two-state model consisting of unbound state 1 and bound state 2. Although a fraction of trajectories clearly contained CRX binding events corresponding to two molecules of CRX HD binding (Suppl. Fig. 8), we did not use a three-state model since our assay would not distinguish the individual dwell times of the two bound CRX molecules, and the CRX-binding events were not correlated with NRL binding. The “ON” dwell times (τ_on) were calculated and the dissociation rate constant k_off of 0.3 s⁻¹ for the CRX/DNA interaction was obtained from a single-exponential fit of the dwell-time distribution constructed from all “ON” dwell times (Fig. 7, Table 1). The association rate constant k_on of 0.01 s⁻1 nM⁻¹ was obtained considering the association rate V_on calculated from a total number of events in all trajectories over a period of 200 s and the concentration of CRX (1 nM). Thus, our analysis yields the equilibrium dissociation constant K_D (k_off/k_on) of 55 nM indicating relatively strong binding of CRX to Ret4 even in the absence of NRL. This K_D indicates a tighter CRX/DNA binding than it was estimated from the BLI assay. Such an observation is not uncommon for single-molecule experiments that allow direct access to on and off dwell times and can detect very transient events, whereas BLI may underestimate the affinity of binding interaction. Also, the concentrations of the DNA and CRX-HD in the reaction chamber were much lower than the K_D, which may explain the relatively rare occurrence of the two-molecule binding events. Similar analyses of CRX binding to NRE/Ret4 were performed in the presence of 2.5 or 5 nM NRL. We observed a trend of increasing affinity of CRX/DNA interaction (decreasing K_D values) with the rising concentration of NRL, revealing a cooperative effect of NRL on CRX/DNA binding. This was accompanied by increase in “ON” dwell times, from ~2.3 s in the absence of NRL to 6.5 s in the presence of 5 nM NRL.

Fig. 7. NRL modulates CRX binding affinity to the rhodopsin proximal promoter.

A. Representative raw and idealized traces from smTIRF experiments where Cy3-labelled CRX HD binds to a surface tethered Ret4-NRE DNA molecule. B, C, D, The dwell-time distributions were constructed from all “ON” dwell times for Ret3-NRE DNA binding of 1 nM of CRX HD alone (B), or in the presence of 2.5 nM (C) and 5 nM of NRL DBD (D). The dissociation rate constants (k_off) from a single-exponential fit of the distributions were (s⁻¹): B, 0.303± 0.011; C, 0.218 ± 0.024 D. 0.154 ± 0.013.

DISCUSSION

Given the fundamental roles of CRX and NRL in transcriptional regulation of developing and mature photoreceptors, the genomic binding profiles, transcriptional activities, and gene-regulatory networks of these TFs have been intensely investigated ^{1,12,13,36–39}. However, with no atomic structures of either CRX or NRL reported to date, our understanding of the biological functions and cooperativity of CRX and NRL from the structural perspective has been poor. Both CRX and NRL have homologs for which atomic structures of their respective DNA-binding domains have been elucidated ^57–59; nonetheless, the transcriptional activity profiles of CRX and NRL are unique and cannot be fully replicated with related TFs. The CRX HD is most homologous (~86% identity) to the HD of OTX2 ⁴. An NMR structure of the HD of mouse OTX2 in the absence of bound DNA was solved by the RIKEN initiative structural pipeline (PDB 2DMS). Structures of other more distant HDs from homeodomain TFs have been solved in complexes with CREs, but the homology of these HDs with CRX is significantly lower (e.g., the HD of Pax3, PDB 3CMY, 66% identity) ⁵⁷. To begin addressing the gap in our knowledge regarding the structure of CRX and the cooperativity with NRL, we have solved the structure of the human CRX HD in complex with the biologically important Ret4 CRE from the RPPR.

Unexpectedly, the structure reveals two molecules of the CRX HD bound to Ret4. This mode of interaction is not an artifact of crystallization, and it was confirmed by SEC-MALS and SEC-SAXS. The two-molecule binding mode of the CRX HD to Ret4 is different from the known binding of the paired class homeodomains to palindromic DNA sites with two inverted TAAT sites ^58,59. The cooperative TF binding at these sites is mediated by the interface between two homeodomains ⁵⁹. Furthermore, the binding of a single HD introduces DNA distortions that prepare a template for the cooperative interaction with a second HD molecule⁵⁹. One of the CRX-binding sites in RPPR, BAT1, contains a P3-type site with two potential sites for CRX ⁴. Both sites of BAT1 apparently participate in binding to CRX, but it is unclear if the binding is simultaneous and cooperative ⁴. The two CRX HD molecules bind to non-palindromic Ret4 DNA sites in a unique “head-to-head” orientation whereby the N-terminal sequences are in proximity to each other (Suppl. Fig. 9). Interestingly, a recent study of CRX and NRL interaction in HEK293T cells using FRET reported formation of CRX homodimers in parallel orientation where the N-termini (i.e. the HD domains) are closer than C-termini ⁶⁰. Although such a dimerization was attributed to the protein-protein interaction, our data suggest that it is mediated by both CRX/DNA and CRX-CRX interactions. Importantly, we found that the impaired stoichiometry of CRX/DNA binding may contribute to the pathogenicity of mutants CRX proteins with substitutions of residues N-terminal to the CRX HD. Our analysis demonstrates that the disease-linked R41Q and E42K disrupt the 2:1 CRX/DNA binding stoichiometry. Besides uncovering a novel mechanism of CRX mutations in retina disease, the CRX/Ret4 structure also may help to re-evaluate the mechanisms of already characterized pathogenic CRX mutants. To illustrate this notion, we examined the R90W mutant CRX HD proteins and determined that the primary deficiency of this mutant stems from its structural instability.

Although we have not determined directly if the binding of two CRX molecules to Ret4 is cooperative, our data indicate positive cooperativity. If the binding of two CRX molecules is independent of each other, the distribution of the 1-molecule and 2-molecule binding events in smTIRFM trajectories would be binominal. Accordingly, the probability of a single site CRX occupancy is 2p(1-p) and the probability of both CRX sites being occupied simultaneously is p², where p is a probability of CRX binding to each site. For 33 trajectories showing both single and double occupancy events (total time of 6600 s), the single- and two-molecule binding events lasted 262 s and 114 s, respectively. Thus, the p value estimated as (114/6600)⁻² is 0.13, which would correspond to the probability of a single-site occupancy of ~0.23 if the binding is not cooperative. However, the observed probability of a single-site occupancy (262/6600 or ~0.04) is markedly lower, suggesting positive cooperativity of CRX binding. This cooperativity is likely underlined by DNA distortions caused by the initial binding of monomeric CRX. We hypothesize that similarly to paired class HD binding to palindromic DNA sites, the two-molecule CRX/Ret4 binding enhances the affinity and specificity of DNA recognition by CRX in the absence of NRL ^58,59. Increased selectivity of CRX is potentially critical for tight control of transcription during photoreceptor cell differentiation and functional maintenance of cones. It may also be necessary to ensure that homologs of NRL, such as MafB and c-MAF, that are expressed in rods, do not cooperate with CRX to induce expression of rod-specific genes in the absence of NRL.

Another critical finding in this study is the demonstration that NRL alters the stoichiometry of CRX binding to the Ret4 site from 2:1 to 1:1. This conclusion is supported by SEC-MALS experiments showing the MW of the CRX/NRL/DNA complex consistent with the stoichiometry of 1:1:1. Furthermore, using a smTIRFM approach, we directly show that NRL increases DNA affinity and dwell times of CRX at the RPPR. While the mechanism of NRL-CRX cooperativity whereby the affinity of CRX for its CREs is increased apparently serves to enhance transcription of rod-specific genes, the alteration of stoichiometry of CRX/DNA binding by NRL may hypothetically contribute to repression of transcription of cone-specific genes. The present work sets the groundwork for future studies aimed at structural characterization of the ternary CRX-NRL-DNA complexes on promoters of different photoreceptor genes.

STAR Methods.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled upon reasonable request by the lead contact, Nikolai O. Artemyev (nikolai-artemyev@uiowa.edu).

Materials availability

All the newly generated and stable materials used in this study are available from lead contact upon request.

Data and code availability

• Structure data for the CRX HD bound to Ret4 response element (PDB 9B8U) and the SAXS data on the CRX HD-Ret4 complex (SASBDB ID SASDU29) have been deposited and are publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and Study Participant Details

E. coli strain BL21-AI (Invitrogen), and NEB5α (New England Biolabs) were used for protein expression and cloning. Cells were cultured in 2XTY and LB media supplemented with appropriate antibiotic.

Method details

Protein expression and purification.

CRX homeodomain residues 31–107 (CRX HD) was cloned between NcoI and XhoI sites of modified pET21a vector with N-terminal TEV protease (TEVP) cleavable His tag. Mutants were created using site-directed mutagenesis. The NRL DNA-binding domain residues 128–237 (NRL DBD) was cloned into NcoI-XhoI site of modified pET21a vector with N-terminal His- and thioredoxin-tag followed by TEVP site. Both proteins were expressed in BL21-AI strain. Cells were grown in LB media to OD₆₀₀ =0.6 and induced with 0.05 mM IPTG/0.1 % Arabinose at 18 degree C overnight. Cells were harvested and resuspended in 50 mM HEPES, 800 mM NaCl, 5 % glycerol, pH 7.5 (Buffer N) supplemented with 5 mM β-mercaptoethanol and cOmplete, EDTA-free Protease Inhibitor Tablet. Cells were lysed by sonicated, cell debris cleared by centrifugation and supernatant were loaded onto Ni-NTA His•Bind Resin charged with Ni⁺⁺. Resin was washed with buffer N followed by Buffer N + 30 mM Imidazole. Proteins were eluted with buffer N + 300 mM imidazole. For the crystallization and Small-angle X-ray scattering (SAXS) studies of CRX HD, His tag was removed by TEVP, sample was dialyzed in 20 mM HEPES, 800 mM NaCl, 5 % glycerol, 0.5 mM TCEP, pH 7.5 and untagged protein was separated by passing sample through Ni-NTA His•Bind Resin. For all other experiments, His-tag was not removed. CRX HD was further purified by HiLoad 26/600 Superdex 75 equilibrated with 20 mM Tris, 800 mM NaCl, 5 % glycerol, 1 mM TCEP, pH 7.5 (buffer S). His- and Thioredoxin-tag was removed from NRL DBD after purification by Ni-NTA His•Bind Resin using TEVP. Sample was dialyzed in 20 mM Tris, 800 mM NaCl, 5 % glycerol, 2 mM TCEP, pH 7.5 and further purified by SP Sepharose Fast Flow resin. The resin was equilibrated with 20 mM Tris, 5 % glycerol, 2 mM TCEP, pH 7.5 (Buffer I) and loaded with sample containing NRL DBD. The resin was washed with Buffer I + 600 mM NaCl. Free His-thioredoxin tag comes in flowthrough. Bound NRL DBD was eluted with Buffer I + 1.2 M NaCl.

Crystallization.

For the crystallization experiments, complementary nucleotides containing the RET4 site were purchased from Integrated DNA Technologies, Inc. Oligonucleotides were dissolved in 10 mM Tris-HCl, pH 8.5, and complementary oligonucleotides were mixed in equimolar ratio. 100 mM NaCl was added to the oligonucleotide mixture and annealed by heating to 95°C followed by slow cooling to room temperature. CRX HD was mixed with annealed oligonucleotides in 4:1 molar ratio. The complex was formed by stepwise dialysis in low salt buffer from 800 to 600 to 300 to 150 mM KCl in 20 mM Tris, 5 % glycerol, 1 mM TCEP, pH 7.5. Excess, free-CRX HD were precipitated and were removed by centrifugation. Samples were concentrated to 10 mg/ml of final complex concentration and screening for initial crystals were performed using mosquito TTP labtech crystallization robot and crystal screen HT and natrix HT random sparse matrix screen. Initial crystals were obtained with 21 base pair oligonucleotides with hoogstein base pair end joining (5’-CCAGGAGCTTAGGAGGGGGAG-3’ and 5’-CCTCCCCCTCCTAAGCTCCTG-3’) in several conditions in natrix HT screen with MPD as precipitant and cacodylate buffer pH 6.0–7.0. Crystals grown in 12 mM NaCl, 80 mM KCl, 40 mM Sodium cacodylate pH 6.0, 50 % MPD, 12 mM Spermine tetrahydrochloride were optimized for pH and MPD concentration and were used for microseeding and matrix microseeding in Natrix HT screen. Crystals obtained in 80 mM KCl, 40 mM cacodylate pH 6.0, 55% MPD, 12 mM Spermine tetrahydrochloride were looped and flash frozen in liquid nitrogen. Data were collected at 100K using beamline 24ID (NE-CAT) at Advanced Photon Source (APS,Argonne National Laboratory, Argonne, IL). Data were processed using XDS and initial phases were determined by molecular replacement using phaser and PDB 1FJL as search model ⁶⁴. Only one polypeptide chain and nucleotide bases covering one polypeptide chain was used in the search model. Manual model building was performed with coot ⁶² and structure was refined using phenix ⁶³. Initial round of refinement includes rigid body, TLS (Translation-Libration-Screw-rotation), individual isotropic ADP (atomic displacement parameters), restrained coordinate and simulated annealing refinement. At later stage, individual isotropic ADP, TLS, and restrained coordinate refinement were performed without rigid body and simulated annealing refinement. Occupancies were refined for residues with alternative conformations. The weight between X-ray target and stereochemistry/ADP restraints were determined automatically by phenix. For the Omit map, residues 41–43 of chain A and 39–43 of chain B were deleted and structure was refined for individual isotropic ADP, TLS, and restrained coordinate refinement using phenix. All the figures showing structures were made using PyMol.

SEC-MALS-SAXS and SEC-MALS.

SAXS was performed at BioCAT (beamline 18ID at the Advanced Photon Source, Chicago) with in-line size exclusion chromatography (SEC) to separate sample from aggregates and other contaminants thus ensuring optimal sample quality and multiangle light scattering (MALS), dynamic light scattering (DLS) and refractive index measurement (RI)) for additional biophysical characterization (SEC-MALS-SAXS). The complex between untagged CRX homeodomain and Ret4 oligonucleotide were premade as described for crystallization. The complex with Ret4 oligonucleotide was loaded (4.6 mg/ml) on a Superdex 75 Increase 10/300 GL column (Cytiva), equilibrated with 20 mM Tris, 150 mM KCl, 5 % glycerol, 1 mM TCEP, pH 7.5 and run using a 1260 Infinity II HPLC (Agilent Technologies) at 0.6 ml/min. The flow passed through (in order) the Agilent UV detector, a MALS detector and a DLS detector (DAWN Helios II, Wyatt Technologies), and an RI detector (Optilab T-rEX, Wyatt). The flow then went through the SAXS flow cell. Scattering intensity was recorded using an Pilatus3 X 1M (Dectris) detector which was placed 3.6 m from the sample giving us access to a q-range of 0.003 Å−1 to 0.42 Å−1. 0.5 s exposures were acquired every 1 s during elution and data was reduced using BioXTAS RAW 2.1.1 ⁶¹. Buffer blanks were created by averaging regions flanking the elution peak and subtracted from exposures selected from the elution peak to create the I(q) vs q curves used for subsequent analyses. Molecular weights and hydrodynamic radii were calculated from the MALS and DLS data respectively using the ASTRA 7 software (Wyatt).

For the mutant CRX HD-Ret4 complexes as well as the NRL-CRX-rhodopsin promoter complex, only SEC-MALS data were collected. Untagged NRL (aa 128–237) was mixed in 4:1 ratio with the rhodopsin promoter duplex oligonucleotide (NRE-Ret4, made by annealing complimentary oligonucleotide 5’- CCAGATGCTGATTCAGCCAGGAGCTTAGGAGGGGGAGG-3’ and 5’- CCTCCCCCTCCTAAGCTCCTGGCTGAATCAGCATCTGG-3’) and dialyzed in buffer containing 600 mM NaCl. Untagged CRX HD was added to the NRL-rhodopsin promoter complex in 4:1 molar ratio and sample was dialyzed in buffer containing 300 mM NaCl, followed by dialysis in final buffer, 20 mM Tris, 150 mM KCl, 5 % glycerol, 1 mM TCEP, pH 7.5. Complexes between the His-tagged CRX HD R41Q, E42K and E80A and Ret4 oligonucleotide were prepared similar to WT CRX HD, loaded onto superdex 75 increase 10/300 GL column (Cytiva) (at 4.5, 3.3, and 3.6 mg/ml respectively) run using a 1260 Infinity II HPLC (Agilent Technologies) at 0.6 ml/min. The flow passed through (in order) the Agilent UV detector, a MALS detector and a DLS detector (DAWN Helios II, Wyatt Technologies), and an RI detector (Optilab T-rEX, Wyatt). Molecular weights and hydrodynamic radii were calculated from the MALS and DLS data respectively using the ASTRA 7 software (Wyatt).

Biolayer Interferometry.

Binding studies of CRX and mutants with RET4 oligonucleotide were performed by biolayer inferometry (BLI). Oligonucleotide containing the RET4 site was biotin labeled at 5’ end using Biotin-PEG6-Maleimide, T4 PNK and ATPγs ⁶⁵. Briefly, 5’ end of single stranded oligonucleotide (antisense strand, 5’-CTCCCCCTCCTAAGCTCCTG-3’) was phosphorothioate labeled using T4 polynucleotide kinase and ATPγs overnight at 37°C. phosphorothioate labeled oligonucleotide was annealed with sense strand (5’-CAGGAGCTTAGGAGGGGGAG-3’), reduced with THP, ethanol precipitated, and labeled with Biotin-PEG6-Maleimide in 1:5 molar ratio for 3 hours. Labeled oligonucleotide was ethanol precipitated (three times) to remove excess unreacted Biotin-PEG6-Maleimide. An Octet RED96 system and streptavidin (SA)-coated biosensors (FortéBio, Menlo Park, CA) were used to measure association and dissociation kinetics for Biotinylated oligonucleotide with CRX HD and mutants. Binding studies were performed in 20 mM Tris, 150 mM KCl, 5% glycerol, 1 mM TCEP, 0.5 mg/ml BSA, pH 7.5. All steps were performed at 26 °C, with biosensors stirred into 0.2 ml of sample in each well at 1000 rpm, and at a data acquisition rate of 5.0 Hz. Biotinylated oligonucleotide was loaded onto SA sensors at a concentration of 2.5 μg/ml for 90 seconds. Data for association and dissociation phases of the assay were collected as shown in Suppl Fig. 4. To correct for baseline drift and non-specific binding, reference sensors lacking bound oligonucleotide were used in the BLI assays with CRX HD proteins at each concentration. Kinetic data fitting was performed using FortéBio Data Analysis software 10.0. Steady-state data fitting was performed using the GraphPad Prism 7 software with the equation for one site specific binding.

Single-Molecule Total Internal Reflection Microscopy.

To label CRX HD with cyanine3 (Cy3) dye, 2.5 mM of Tris(hydroxypropyl)phosphine (THP) was added to pure CRX 31–107 and incubated for 1hr in a desiccator to reduce disulfide bonds. Excess THP was removed by subjecting reduced CRX-HD to analytical gel filtration equilibrated with buffer containing 20 mM HEPES, 600 mM NaCl, 5 % glycerol (pH 7.5). Sulfo-Cy3 maleimide (Lumiprobe) dissolved in DMSO was added to reduced CRX HD (60 μM) with 3-fold excess and incubated in a dark desiccator for 20 min. The reaction was quenched by addition of 4 mM β-mercaptoethanol followed by removing free dye using a PD-10 column and determining labeling efficiency.

A custom-built prism TIRF microscope was used to perform single-molecule TIRF experiments ⁵². The microscope is built on an Olympus IX71 microscope frame and combines 532 nm (Compass 215M-50, Coherent Inc., Santa Clara, CA, USA) and 641 nm (Coherent, Cube 1150205/AD) laser beams using a polarizing beam splitting cube (CVI Melles Griot, PBSH-450-700-050), which are directed to the microscope objective at a 30° angle. TIR is achieved through a UV fused silica pellin–broca prism (325–1206, Eksma Optics, Vilnius, Lithuania) and an uncoated N-BK7 plano–convex lens (LA1213 Thorlabs Inc., Newton, NJ, USA). Photons are collected using a 60X, NA 1.20 water immersion objective (UPLSAPO60XW Olympus Corp., Shinjuku City, Tokyo, Japan), and spurious fluorescent signal is removed using a dual bandpass filter (FF01–577/690–25 Semrock Inc., Rochester, NY, USA). Cy3 and Cy5 emissions are separated using a dual-view housing (DV2 Photometrics, Tucson, AZ, USA) containing a 650 nm longpass filter (T650lpxr Chroma Technology Corp., Bellows Falls, VT, USA), and fluorescent images are captured using an Andor iXon 897 EMCCD (Oxford Instruments, Abingdon, UK).

Surface Tethered DNA Single-Molecule Experiments.

Prior to surface tethering, a 5’ biotin labeled oligonucleotide containing NRE and Ret4 sites (biotin-5’-CCAGATGCTGATTCAGCCAGGAGCTTAGGAGGGGGAGG-3’, sense strand, synthesized by IDT) was annealed with complementary unlabeled antisense strand (5’-CCTCCCCCTCCTAAGCTCCTGGCTGAATCAGCATCTGG-3’). The two oligonucleotide were mixed in equimolar ratio to the final concentration of 50 μM in 20 mM Tris, 50 mM NaCl, pH 8.0 and the mixture washeated at 95 °C for 5 min, slowly cooled down to allow for annealing, and diluted to working concentrations. To extend the lifespan of fluorophores in single-molecule experiments, an oxygen scavenging system is necessary to reduce reactive oxygen species (ROS) that cause rapid photobleaching. We utilized 12 mM Trolox (6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid) and Gloxy (catalase and glucose oxidase solution) to reduce ROS effects. Trolox is prepared by adding 60 mg of Trolox powder (238813–5G, Sigma-Aldrich) to 10 mL of water with 60 μL of 2 M NaOH, mixing for 3 days, filtering, and storing at 4 °C. Gloxy is prepared as a mixture of 4 mg/mL catalase (C40–500MG, Sigma-Aldrich) and 100 mg/mL glucose oxidase (G2133–50KU, Sigma-Aldrich) in wash buffer (25 mM Tris-HCl pH 7.0, 140 mM KCl); special KCl of Spectroscopy grade (#39795, Alfa Aesar, Haverhill, MA, USA) was used for all single molecule experiments.

Before the starting of the experiment, quartz slides (25 mm × 75 mm × 1 mm #1×3×1MM, G. Frinkenbeiner, Inc., Waltham, MA, USA) and cover glass (24 mm × 60 mm−1.5, Fisherbrand, Fisher Scientific, Hampton, NH, USA) are treated with passivation and PEGylation. The flow cell is treated with 0.2 mg/mL NeutrAvidin (#3100, ThermoScientific, Thermo Fisher Scientific, Waltham, MA, USA) to tether biotinylated molecules to the flow cell surface. Excess neutravidin is removed with flow of 1mL wash buffer. To prepare the flow cell for imaging, 500 pM of biotinylated-DNA substrates are added and incubated for 3 min. Excess DNA is then removed using a wash buffer.

For image collection, 1 nM of Cy3-CRX protein alone and in combination of 2.5/5 nM NRL protein was flowen in chamber in a imaging buffer (containing 25 mM Tris-HCl pH 7.0, 140 mM KCl, 10 mM MgCl₂, 1 mg/mL BSA, 1 mM DTT, 0.8% w/v D-glucose, 12 μM glucose oxidase, 0.04 mg/mL catalase and TROLOX) is added to the flow cell. The proteins in imaging buffer were incubated in dark and in chamber for 5 min followed by recording using a custom software, single.exe (generously provided by the Taekjip Ha Lab, JHU), with 532 nm laser power set to 45 mW. Image collection begins using 100 ms time resolution, gain of 290, background set to 400 and correction set to 1200. Once 300 frames have been collected. Images are collected for a total of 2000 frames (200 s). Fluorescence trajectories, showing the time-based changes in Cy3 fluorescence in a specific location in the TIRFM reaction chamber, in each experiment were collectively analyzed using hFRET ⁵⁴. All the data were described by the two-state model (bound and free). The dwell times were extracted and sorted using kinetic event resolving algorithm (KERA) ^55,56. The dwell time were obtained for both the “ON” dwell time (τ_on) and “OFF” dwell time (τ_off) for each data set. Using a single-exponential function and considering durations of all bound events binned in different time intervals. the dissociation rate constants (k_off) were obtained from a single-exponential fit of the dwell-time distributions constructed from all “ON” dwell times ⁵³.

Quantification and statistical analysis.

Statistical analyses were performed using GraphPad Prism 9. Measurements were compared using t-test. X-ray data collection and refinement statistics are reported in Table S1.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
E.coli BL21-AI	Invitrogen	Cat # C6070-03
Chemicals, peptides, and recombinant proteins
sulfo-Cyanine3 maleimide	Lumiprobe	Cat # 21380
Biotin-PEG6-Maleimide	Thermo Fisher Scientific	Cat # B556350MG
Octet Streptavidin (SA) Biosensor	Sartorius	Item # 18-5019
Trolox	Sigma-Aldrich	Cat # 238813-5G
Deposited data
Crystal structure of the CRX HD bound to Ret4 response element.	This study	PDB 9B8U
CRX HD-Ret4 oligonucleotide complex SAXS data	This study	SASBDB: SASDU29
Oligonucleotides
RET4.f: CCAGGAGCTTAGGAGGGGGAG	This study	N/A
RET4.r: CCTCCCCCTCCTAAGCTCCTG	This study	N/A
NRE-RET4.f: CCAGATGCTGATTCAGCCAGGAGCTTAGGAGGGGGAGG	This study	N/A
NRE-RET4.r: CCTCCCCCTCCTAAGCTCCTGGCTGAATCAGCATCTGG	This study	N/A
CRX_R41Q.f: AAGCAGCGGCAGGAGCGCACCACCTTCACC	This study	N/A
CRX_R41Q.r: GGTGCGCTCCTGCCGCTGCTTCCTGGGGGC	This study	N/A
CRX_E42K.f: CAGCGGCGGAAACGCACCACCTTCACCCGG	This study	N/A
CRX_E42K.r: GGTGGTGCGTTTCCGCCGCTGCTTCCTGGG	This study	N/A
CRX_R90W.f: TTCAAGAACTGGAGGGCTAAATGCAGGCAG	This study	N/A
CRX_R90W.r: TTTAGCCCTCCAGTTCTTGAACCAAACCTG	This study	N/A
Recombinant DNA
CRX HD in pET-21a	This study	N/A
NRL DBD in pET-21a	This study	N/A
Software and algorithms
BioXTAS RAW 2.1.1	Hopkins et al 61	https://bioxtas-raw.readthedocs.io/en/latest/
COOT	Emsley et al 62	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PHENIX	Liebschner et al 63	https://phenix-online.org/
Other
His-Bind resin	Novagen	Cat #69670
HiLoad 16/600 Superdex 75 column	GE Healthcare	Cat #28989333
SP sepharose fast flow resin	GE Healthcare	Cat #17072901
Catalase	Sigma-Aldrich	Cat # C40-500MG
Glucose oxidase	Sigma-Aldrich	Cat # G2133-50KU
NeutrAvidin	Thermo Fisher Scientific	Cat # 31000

Highlights.

• Crystal structure of the CRX homeodomain in complex with its Ret4 response element

• Complex displays a 2:1 stoichiometry and a unique orientation of CRX on the DNA

• The CRX/Ret4 structure helps to elucidate the mechanisms of pathogenic CRX mutations

• NRL alters the stoichiometry and strengthens CRX binding to the rhodopsin promoter