Structural mechanism of synergistic targeting of the CX3CR1 nucleosome by PU.1 and C/EBPα

Pioneer transcription factors are vital for cell fate changes. PU.1 and C/EBPα work together to regulate hematopoietic stem cell differentiation. However, how they recognize in vivo nucleosomal DNA targets remain elusive. Here we report the structures of the nucleosome containing the mouse genomic CX3CR1 enhancer DNA and its complexes with PU.1 alone and with both PU.1 and the C/EBPα DNA binding domain. Our structures reveal that PU.1 binds the DNA motif at the exit linker, shifting 17 bp of DNA into the core region through interactions with H2A, unwrapping ~20 bp of nucleosomal DNA. C/EBPα binding, aided by PU.1’s repositioning, unwraps ~25 bp entry DNA. The PU.1 Q218H mutation, linked to acute myeloid leukemia, disrupts PU.1-H2A interactions. PU.1 and C/EBPα jointly displace linker histone H1 and open the H1-condensed nucleosome array. Our study unveils how two pioneer factors can work cooperatively to open closed chromatin by altering DNA positioning in the nucleosome.

binding domain of C/EBP⍺ fused to maltose binding protein at the N-terminus to increase its solubility (Extended Data Fig. 1a). EMSA experiments showed that PU.1 binds the nucleosome with an apparent Kd app of ~0.65 µM obtained by fitting the data to the Hill equation (Extended Data Fig. 1). In contrast, MBP-DBD C/EBP⍺ binds the nucleosome with a lower affinity (Kd app of ~1.7µM). Using the single-chain antibody (scFv)-assisted cryo-EM method, 33 we obtained the density map of the nucleosome-scFv2 complex at an overall resolution of 2.6 Å (Fig. 1c, Extended Data Figs. 2, and Table 1). We also solved the structure of the nucleosome-scFv2-PU.1 complex at an overall resolution of 2.9 Å (Extended Data Figs. 3 and Table 1). However, we could not obtain a density map for the nucleosome-scFv2-MBP-DBD C/EBP⍺ complex, possibly due to its lower affinity. To investigate whether scFv affects the nucleosome binding by PU.1 and MBP-DBD C/EBP⍺, we conducted EMSA experiments, showing that scFv had little effect on the apparent binding affinity of PU.1 and MBP-DBD C/EBP⍺ to the nucleosome (Extended Data Fig. 1g, h).
The high-resolution density maps for the free CX3CR1 nucleosome and its complex with PU.1 allow us to define the DNA base pairs unanimously (Fig. 2a, b), showing that the nucleosomes are uniquely positioned with nucleotides 79 and 96 at the dyad (Fig. 1c, d), respectively. Thus, PU.1 binding repositions the nucleosome by 17 bp. However, the density map for the PU.1 region has a local resolution of ~6 Å (Extended Data Fig. 3), likely due to its dynamic motion. Nevertheless, the crystal structure of the DNA binding domain (DBD or ETS) of PU.1 bound to the DNA fragment (PDB ID: 1PUE) can fit the density map well, leading to a structure model showing that PU.1 G220 is close to H2A T77. To better define the structure, we engineered a disulfide bond by mutating these two 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08. 25.554718 doi: bioRxiv preprint residues to Cys to restrict the dynamic motion and used a longer DNA (167 bp, Fig. 1a) for PU.1 binding, which improved density map to an overall resolution of 2.7 Å and ~4 Å for the PU.1 ETS binding region, respectively (Figure 1d and Extended Data Fig. 4). The new map shows densities for all core histones and backbone structures of the PU.1 ETS and residues interacting with DNA motif (Figure 1d, Extended Data Fig. 5a-c). The structural model derived from the density map of the mutated proteins fits the density of the wild-type (WT) complex very well (Extended Data Fig. 5d), indicating that the disulfide bond only restricts its dynamics without altering the structure of the complex.  Fig. 6c and h). This result suggests that the C/EBP⍺ dimer binds a region including a non-canonical motif in the unwrapped DNA at the entry side (corresponding to SHL -6.0) (Fig. 1a).
To confirm the binding of C/EBP⍺ at this location, we introduced a mutation in the associated DNA region, changing it from CAGCTGGTTG to CAGCAACTTG (highlighted 105 and is also made available for use under a CC0 license. (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08. 25.554718 doi: bioRxiv preprint in the dashed box in Fig. 1a). The mutation resulted in a decrease in the apparent binding affinity between C/EBP⍺ and the nucleosome, with the Kd app value changing from 1.7 to 2.4 µM (Extended Data Fig. 7a, b). To further validate that the observed density on the DNA originated from C/EBP⍺, we incorporated the CAGCTGGTTG sequence to the '601' nucleosome at the location corresponding to the one found in the CX3CR1 nucleosome bound to PU.1 and C/EBP⍺ (Extended Data Fig. 7d). EMSA experiments demonstrated that C/EBP⍺ bound the chimeric nucleosome better than to the '601' nucleosome (Extended Data Fig. 7c, d). Furthermore, we discovered that C/EBP⍺ also exhibited binding to a 20 bp double-stranded DNA fragment containing the CAGCTGGTTG sequence in the middle, while showing little binding to the fragment with the CAGCAACTTG mutation (Extended Data Fig. 7e). Importantly, we established that MBP did not bind the free CX3CR1 DNA and the CX3CR1 nucleosome (Extended Data Fig.   7f, g), ruling out the possibility that the observed extra density originated from MBP.
Based on our structural findings, we propose that PU.1 binds at the exit site and C/EBP⍺ binds at the entry site of the CX3CR1 nucleosome, which agrees with the ChIP-seq and MNase-seq results. 25

Interactions between PU.1 and nucleosomal DNA
In the structure of the nucleosome-PU.1 complex, PU.1 recognizes the AAATAGGAA sequence in the canonical motif near the exit site at SHL 5.5 (Fig. 1d), 34 leading to the unwrapping of ~20 bp nucleosomal DNA. The DNA unwrapping does not lead to the H2A-H2B dissociation (Extended Data Fig. 3i). PU.1 interacts with the nucleosomal DNA similarly as with the DNA fragment in the crystal structure (Fig. 2c). Notably, in addition 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC  to the formation of hydrogen bonds with the GGAA motif, PU.1 ETS residues K171, R222, K245, and K249 also interact with the backbone phosphates beyond the DNA motif and bend the DNA (Fig. 2c), bending the DNA towards itself. Similarly, PU.1 also bends the unwrapped nucleosomal DNA (Fig. 2d).
The CX3CR1 DNA also includes GGAA core sequences at SHLs -1.5 and -1.0 in the structure of the nucleosome bound to PU.1 (Fig. 1a), which are accessible to PU.1. To investigate the cause for the absence of PU.1 at these two locations in our structures, we conducted EMSA experiments using the free CX3CR1 DNA and its mutant with the GGAA sequence mutated to GGGG in the two regions. The mutation only slightly reduced the apparent binding affinity of PU.1 (Extended Data Fig. 8). These results explain the absence of PU.1 at these two regions in our structure and reveal that the GGAA motif alone without considering flanking sequences is not a good indicator to predict PU.1 binding although it is the major component of the canonical motif of PU.1. 34

Interactions between PU.1 and H2A
In the structural model of the PU.1-nucleosome complex ( Fig. 3a and Extended Data Fig.   4), PU.1 residue Q218 and H2A residues K75, K76, and T77 in the nucleosome at the interface are close, suggesting that they may form interactions. Notably, PU.1 Q218H mutation is associated with acute myeloid leukemia. 17,18 To test this hypothesis, we mutated Q218 to His and H2A residues K75, K76 and T77 to Ala (termed 3A H2A) and measured the apparent binding affinity changes. We performed EMSA experiments by titrating the nucleosome with PU.1. Fitting the binding data to the Hill equation showed 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08.25.554718 doi: bioRxiv preprint that the mutations of PU.1 Q218H and 3A H2A reduced the binding affinity by ~2-and 3fold, respectively (Fig. 3b, c). In addition, EMSA experiments showed that Q218H has little effect on PU.1 binding to the DNA (Fig. 3d,e), further supporting that PU.1 interacts with H2A.
To investigate the role of these interactions in nucleosome repositioning by PU.1 binding, we performed restriction enzyme digestion using Sau96I. Based on our structures, the cutting site of Sau96I is inside the nucleosome core particle of the free 162 CX3CR1 nucleosome. However, it will shift to the exposed linker DNA region near the entry site after PU.1 binding to the nucleosome and subsequently increase its accessibility by the restriction enzyme (Fig. 4a). Indeed, we found that PU.1 binding increased the cutting efficiency of the 162 bp CX3CR1 nucleosome (Fig. 4b, c). We next examined the effects of PU.1 Q218H and 3A H2A mutations on digestion efficiency. PU.1 Q218H and 3A H2A mutations decreased digestion efficiency compared with wild-type (WT) PU.1 and H2A (Fig. 4b, c). These results are consistent with the free nucleosome and the nucleosome-PU.1 structures, and reveal that the interactions between PU.1 and H2A play essential roles in the shift of the nucleosomal DNA. To further confirm the DNA repositioning at different temperatures and buffer conditions, we conducted FRET experiments by labeling the H2A residue 116 with Cy3 and the entry nucleotide 1 with Cy5 in the presence of 140 mM KCl or 150 mM NaCl at both 4 and 20 o C (Extended Data Fig. 9). PU.1 binding resulted in a significant decrease in FRET signals under all these conditions. FRET assay on nucleosomal DNA unwrapping 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC To validate the unwrapping of the nucleosomal DNA by PU.1 at the exit side, we reconstituted the 146 bp CX3CR1 nucleosome consisting of nucleotides 23-168. The shorter DNA forces the nucleosome core particle to be in a position mimicking that in the 162 bp nucleosome repositioned by PU.1. Therefore, binding of PU.1 to this 146 bp DNA nucleosome will not cause translation of the DNA, allowing the FRET to report the anticipated DNA unwrapping at the exit site only. We labeled H2A residue 116 with Cy3 and the DNA at the exit site with Cy5. We conducted a FRET assay by titrating PU.1 to the nucleosome (Fig. 4d). PU.1 binding decreased the FRET signal, supporting its role in DNA unwrapping at the exit side. We applied a similar assay with the Cy5 labeling at the DNA entry site to investigate the effect of C/EBP⍺ binding on the unwrapping of DNA. We found that the FRET signal decreased with the presence of MBP-DBD C/EBP⍺, confirming that MBP-DBD C/EBP⍺ binding unwrapped the DNA at the entry side (Fig. 4e).

PU.1 facilitates C/EBP⍺ binding to the nucleosome
Our data show that PU.1 binding shifts the C/EBPa binding site from the inner nucleosome core region to the region near the entry site. According to the site-exposure model, 35 PU.1 binding would facilitate the binding of C/EBP⍺ (Fig. 5a). To verify it, we conducted FRET experiments by labeling C/EBP⍺ with Cy5 and the nucleotide 19 with Cy3, which is close to the C/EBP⍺ binding site. Titration of C/EBP⍺ to nucleosome led to a gradual increase of FRET signals, confirming that C/EBP⍺ binds on the unwrapped DNA at the entry side ( Fig. 5b and Extended Data Fig. 9e). Addition of PU.1 led to the increase of FRET signals more rapidly, suggesting that PU.1 facilitates the binding of C/EBP⍺ to the nucleosome, consistent with the earlier in vivo result that C/EBP⍺ binding is partially dependent on PU.1. 22,23 We also conducted a FRET experiment by labeling the DNA at the exit site with Cy3 and PU.1 with Cy5. Titration of PU.1 to nucleosome increased the FRET signal (Fig 5c and Extended Data Fig. 9f). However, the addition of C/EBPa did not facilitate PU.1 binding to the nucleosome.

H1 eviction by PU.1 and C/EBP⍺
The unwrapping of nucleosomal DNA by PU.1 and C/EBP⍺ at both exit and entry sides suggest that the two pioneer factors together might evict linker histone in the chromatosome (Fig. 6a). We tested this hypothesis using FRET by labeling Cy3 on H2A and Cy5 on K26C H1.4. Neither PU.1 nor MBP-DBD C/EBP⍺ caused a significant change in the FRET signal. In contrast, together they led to a large decrease in the FRET signal ( Fig.   6b), indicating that the synergistic binding of PU.1 and C/EBP⍺ unwraps nucleosomal DNA from both sides of the nucleosome, causing H1 eviction. Moreover, PU.1 Q218H mutant showed a much weaker H1-eviction function, consistent with our earlier results that the interactions between PU.1 and H2A play essential roles in PU.1 binding and nucleosome repositioning (Fig. 6c).

Opening of H1-condensed nucleosome array by PU.1 and C/EBP⍺
Pioneer transcription factors can open closed chromatin without using ATP. To verify it, we engineered XhoI and EcoRI cutting sites near the binding sites of PU.1 and C/EBP⍺, respectively, in the CX3CR1 nucleosome (Fig. 7a, Extended Data Fig. 10a). We then reconstituted a 12 x 197 bp nucleosome array (NA) with five and six W601 nucleosomes on each side of the engineered CX3CR1 nucleosome (Fig. 7b, Extended Data Fig. 10). 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC We condensed the nucleosome array using H1. Finally, we performed the restriction enzyme digestion experiments with an increasing amount of PU.1 or C/EBP⍺ or both.
Under the same experimental conditions, we found that PU.1 and C/EBP⍺ facilitate the restriction enzymes to cut the nucleosome array (Fig. 7c, d), suggesting they can open the closed chromatin. In contrast, the nucleosome array with the CX3CR1 DNA replaced with W601 showed resistance to the restriction enzymes, indicating that the opening only occurred locally at the CX3CR1 nucleosome (Extended Data Fig. 10f, g). The enzyme cutting efficiency did not increase using the nucleosome array with mutations at PU.

Discussion
Nucleosomes were thought inhibitory for binding transcription factors due to potential steric obstructions from core histones and the neighboring DNA gyres. To explain how transcription factors may bind the nucleosome, the site-exposure model hypothesizes that nucleosomal DNA may spontaneously unwrap and expose the DNA motifs for transcription factor binding. 35 On the other hand, functional genomics and structural modeling studies suggest that the pioneer transcription factor functionality may be determined by how they can avoid the steric clash with the nucleosome when recognizing the exposed DNA motif. 36 Recent structural studies using nucleosomes with non-genomic 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08.25.554718 doi: bioRxiv preprint DNA as models suggest that the DBDs of transcription factors can bind the exposed DNA motifs in the nucleosome core region with or without DNA unwrapping. 10,11 Our results show that PU.1 and C/EBP⍺ each can bind the CX3CR1 nucleosome and unwrap nucleosomal DNA, suggesting that they may use the site exposure mechanism for targeting the nucleosome in vivo. In addition, the interactions between PU.1 ETS and H2A loop indicate that PU.1 may stabilize the PU.1-nucleosome complex and possibly decrease its dissociation rate, 8,37 which would increase the residence time of pioneer factors on the nucleosome. For C/EBP⍺ binding at the entry site, structural modeling suggests that C/EBP⍺ may only bind the unwrapped DNA as its binding to fully wrapped nucleosome would cause steric clashes with the neighboring DNA gyre and core histones (Extended Data Fig. 6i). Notably, EMSA experiments revealed that when the ratio of pioneer factor to nucleosome is high, multiple PU.1 and C/EBP⍺ molecules can bind to the CX3CR1 nucleosome (as shown in Extended Data Figs. 1b, d). These additional bindings are likely the result of C/EBP⍺ associating with the nucleosome at weaker binding sites. The investigation of these higher-order bindings and the relevance of our results to the in vivo function using genome editing is a topic for future studies. However, it is important to note that these findings should not impact the roles of PU.1 and C/EBP⍺ as identified in our structures. This is because the mutations observed in the DNA sites bound to the pioneer factors in our structures can effectively eliminate the ability of the pioneer factors to open the H1-condensed nucleosome arrays (as depicted in Fig. 7c, d).
Our study reveals a mechanism for the cooperative binding of two pioneer factors whereby one pioneer factor (PU.1) binds the DNA motif and core histones, leading to the repositioning of the nucleosome, which in turn shifts the DNA motif of the other transcription factor (C/EBP⍺) from the inner nucleosome core region to the location near the entry site ( Previous studies using restriction enzymes and DNA unzipping have shown that the location near the entry/exit site has a higher probability of unwrapping spontaneously than in the inner region of the nucleosome core particle. 40,41,35 Thus, PU.1 appears to have taken advantage of this physical mechanism to facilitate the binding of C/EBP⍺ by repositioning the nucleosome for their pioneer functions. Furthermore, this mechanism could explain the in vivo results that PU.1 alone can target compacted chromatin and induce its accessibility, whereas C/EBP⍺ binding to nucleosome-enriched regions partially depends on PU.1. 22,23 Finally, the synergistic binding mechanism for the two transcription factors revealed from our study is distinct from known models for cooperative binding of transcription factors to DNA, 42,43 including direct interactions between transcription factors, DNA-mediated interactions, and nucleosome displacement.

Acknowledgments
We thank Dr. Rick Huang and Ms. Allison Zeher for assistance in cryo-EM Data collection.
The cryo-EM work utilized NCI-NIH IRP Cryo-EM Consortium (NICE) microscopy resource and NIH high performance computing Biowulf system for data processing.

Competing interests
Authors declare no competing interest 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC     105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC  105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC  105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC  105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC  105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC

Recombinant human histones and their mutants were expressed individually in
Escherichia coli BL21(DE3) cells. All mutations were generated using QuikChange kit (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08.25.554718 doi: bioRxiv preprint Data Table 2). The plasmids were transformed into BL21(DE3) cells and grown in LB medium. When OD600 reached 0.8, 0.5 mM IPTG was added at 37 ºC for 2 hrs. Expression  Table 1) were prepared by PCR amplification, followed by ethanol precipitation and purification using the POROS column. Briefly, The PCR products were pelleted by 75% ethanol containing 0.3 M NaAc at pH 5.2. The sample was incubated for 120 min at -20 ºC, followed by centrifugation. The pellet was resuspended by TE buffer, loaded to POROS column chromatography (GE Healthcare), 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. Cryo-EM data were collected using a Titan Krios G3 electron microscope (Thermo-Fisher) operated at 300kV. Micrographs were acquired in super-resolution mode at the nominal magnification of 81,000x with 0.528 Å image pixel size using a 20-eV slit post-GIF Gatan K3 camera. The dose rate on the camera was set to 15 e -/pixel/s. The total exposure time of each micrograph was 4 s fractionated into 50 frames with 0.08 s exposure time for each frame. The data collection was automated using the SerialEM software package. 46 A total of 2,986 micrographs were collected for free CX3CR1 nucleosome sample, and 4,503 micrographs were collected for the wild-type nucleosome-PU.1 complex sample. 7,014 micrographs were collected for the disulfide bond mutated nucleosome-G220C PU.1 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC The copyright holder for this preprint this version posted August 26, 2023. ; https://doi.org/10.1101/2023.08.25.554718 doi: bioRxiv preprint sample and 5,204 micrographs were collected for the nucleosome-G220C PU.1-MBP-DBD C/EBP⍺ complex sample.
All datasets were processed in RELION/3.1.3 and cryoSPARC v3.2 following the standard procedures. 47,48 The beam-induced image drift was corrected using MotionCor2. 49 The averaged images without dose weighting were used for defocus determination using CTFFIND4.1 50 , and images with dose weighting were used for particle picking and extraction. Particles were automatically picked using Gautomatch (https://www.mrc-lmb.cam.ac.uk/kzhang/Gautomatch/). 1,274,643 particles were picked for the free CX3CR1 nucleosome dataset. Bad particles were removed by 2D classification and 3D classification in RELION using 4x binned particles. 674,406 particles were selected and re-extracted without binning, followed by two more rounds of 3D classification using 7.5° and 3.7° angular sampling rate. 1 class containing 143,068 particles with good density of DNA was selected for auto-refine. Bayesian Polishing, followed by importing to CryoSPARC for CTF-refinement and nonuniform refinement, further improved the resolution to 2.6 Å.
For Nuc-ScFv-PU.1 complex dataset, 2,506,660 particles were picked from 4,503 micrographs. Bad particles and free nucleosome particles were removed from 2D classification and 3D classification using 2x binned particles. 581,016 particles with one blurry DNA end were selected and re-extracted without binning. Two more rounds of 3D classification with angular sampling rate 7.5° and 3.7° were performed. one class with good density of unwrapped DNA end was selected for focused 3D classification without alignment and with a small mask only covering the unwrapped DNA part. Another two classes with blurry DNA end density were also selected for focused 3D classification without alignment. The two focused 3D classification jobs generate two classes with clearly unwrapped DNA and also the extra bulb density on the unwrapped DNA. The two classes were combined and submitted for Bayesian Polishing. After CTF-refinement and 3D-auto refinement in RELION, a 2.9 Å map was generated for model building. The PU.1 ETS domain (PDB: 1PUE) can fit into the extra density on the unwrapped DNA well.
2,000,352 particles were selected for 3D classification using the above wild type nucleosome-scFv-PU.1 map lowpass filtered into 60 Å as the initial model. 622,795 particles with both good nucleosome density (contained one side unwrapped DNA) and also good ETS PU.1 density were selected. Particles were then re-extracted without binning and submitted to 2 more rounds of 3D classification. One class with high-resolution feature and good ETS PU.1 density were selected. Bayesian Polishing was performed, followed by importing into CryoSPARC. CTF refinement and non-uniform refinement were used to further improve the resolution to 2.7 Å.
For the nucleosome ( T77C H2A)-scFv-G220C PU.1-MBP-DBD C/EBP⍺ complex dataset, 2,888,287 particles were picked, and 1,411,125 particles were selected after 2D classification using 4x binned particles. 3D classification using the above nucleosome ( T77C H2A)-scFv-G220C PU.1 lowpass filtered to 60 Å as the initial model generated a class with good nucleosome and PU.1 density. 468,250 particles in this class were then reextracted without binning, and two more rounds of 3D classification were performed. 1 class with good PU.1 density and clearly unwrapped entry site DNA density was selected for 3D auto-refine, generating a 3.9 Å map with 116,597 particles. Focused refinement with mask only covering the nucleosome core and PU.1, excluding the unwrapped entry site DNA, generated a 2.8 Å map. The focused 3D classification without alignment and with mask only covering the unwrapped entry site DNA generated three classes all around 4.1 Å. For these 3 classes, the entry site DNA was all unwrapped.

Model building and structure analysis
For the free CX3CR1 nucleosome, an initial model of the nucleosome histone octamer and scFv was generated using the nucleosome structure (PDB: 7K61). The model was fitted into the cryo-EM density map of the CX3CR1 nucleosome-scFv complex. The DNA sequence was built into the map from scratch in COOT 51 and the histone octamer and scFv were optimized by manual rebuilding. The whole complex was refined using realspace refinement in PHENIX. 52 For the nucleosome-scFv-PU.1 complex, the free  (Table S2)  (which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available for use under a CC0 license.
(which was not certified by peer review) is the author/funder. This article is a US Government work. It is not subject to copyright under 17 USC Restrict enzyme digestion of H1.4-condensed nucleosome array 3 mg wild-type CX3CR1 nucleosome array or mutated nucleosome array or 601 nucleosome array in 20 μL was mixed with PU.1 or DBD C/EBP⍺ with a molar ratio to the nucleosome array ranging from 0:1 to 16:1, respectively, in digestion buffer (10 ul, 10 mM Tris-HCl, pH 8.0, 60 mM NaCl, 1 mM magnesium chloride, 2 mM DTT) at room temperature. 10 units of restriction enzyme XhoI (NEB) were added to the nucleosome array with and without PU.1. 5 units of restriction enzyme EcoRI (NEB) were added to the nucleosome array in the absence and presence of DBD C/EBP⍺. Samples were incubated at 37 °C for 30 min, and the enzyme was inactivated by NEB purple loading dye. Samples were then incubated with proteinase K at 50 °C for 60 min. After centrifugation, the top solution was harvested and loaded into a 1 % agarose gel stained with SYBR Safe dye (Invitrogen). Electrophoresis was performed at 130 V in 1 x TBE buffer for 25 min. Band intensities for the digestion product and input were measured using ImageJ. The relative digestion efficiency was calculated using the intensity of the product in the presence of transcription factors divided by the product intensity without transcription factors.

Data availability
The cryo-EM reconstructions and atomic models of the CX3CR1 nucleosome and the