Single-nucleoid architecture reveals heterogeneous packaging of mitochondrial DNA

Abstract

Cellular metabolism relies on the regulation and maintenance of mitochondrial DNA (mtDNA). Hundreds to thousands of copies of mtDNA exist in each cell, yet because mitochondria lack histones or other machinery important for nuclear genome compaction, it remains unresolved how mtDNA is packaged into individual nucleoids. In this study, we used long-read single-molecule accessibility mapping to measure the compaction of individual full-length mtDNA molecules at near single-nucleotide resolution. We found that, unlike the nuclear genome, human mtDNA largely undergoes all-or-none global compaction, with most nucleoids existing in an inaccessible, inactive state. Highly accessible mitochondrial nucleoids are co-occupied by transcription and replication components and selectively form a triple-stranded displacement loop structure. In addition, we showed that the primary nucleoid-associated protein TFAM directly modulates the fraction of inaccessible nucleoids both in vivo and in vitro, acting consistently with a nucleation-and-spreading mechanism to coat and compact mitochondrial nucleoids. Together, these findings reveal the primary architecture of mtDNA packaging and regulation in human cells.

Extended Data Figure 1

(Related to Figure 1) (a) Histogram showing the distance to the next neighboring A/T nucleotide from the 9,218 A/T nucleotides present in the human mitochondrial genome. (b) Histogram showing the distance to the next GC dinucleotide from the 711 present in the human mitochondrial genome. (c) Bar plot showing the percent of reads binned by the number of m6A modifications per read comparing untreated samples and those treated with 500 U of the m6A-MTase Hia5. Individual dots represent three and six biological replicates for 0 U Hia5 and 500 U Hia5, respectively (d) Bar plot showing the percent of reads binned by the number of m6A modifications per read for samples treated with 200, 500, 750, and 1000 U of the m6A-MTase Hia5. (e) Bar plot showing the percent of reads binned by the number of m6A modifications per read for samples treated with 500 U of the m6A-MTase Hia5 for 10, 30, 45, 60, and 120 minutes. The 120 minute sample received an additional SAM spike-in after 60 minutes.

Extended Data Figure 2

(Related to Figures 1, 2) (a) Correlation scatter plots comparing six mtFiber-seq samples from HeLa S3 cells for (bottom left) fraction of reads methylated at each adenine and (top right) fraction of reads with a footprint at each genomic position. PacBio chemistry version is indicated for each replicate. Pearson’s correlation coefficient is shown for each correlation. (b) Correlation scatter plots comparing two mtFiber-seq samples from U2-OS cells for (bottom left) fraction of reads methylated at each adenine and (top right) fraction of reads with a footprint at each genomic position. PacBio chemistry version is indicated for each replicate. Pearson’s correlation coefficient is shown for each correlation. (c) Correlation scatter plots comparing three mtFiber-seq samples from undifferentiated human skeletal muscle myoblasts for (bottom left) fraction of reads methylated at each adenine and (top right) fraction of reads with a footprint at each genomic position. PacBio chemistry version is indicated for each replicate. Pearson’s correlation coefficient is shown for each correlation.

Extended Data Figure 3

(a) Schematic depicting experimental design. Tn5 was loaded with ATTO488-labeled oligos to form active transposomes. U2-OS cells were treated with transposome and imaged by confocal fluorescence microscopy (b) Schematic depicting the ATAC-see segmentation and analysis pipeline. Background corrected images were masked and segmented in Arivis. Features of assigned objects were extracted and analyzed. (c) Representative image of a U2-OS cell showing ATAC-see and DNA signals. DNA was labeled with an anti-ss/dsDNA antibody that shows preferential labeling of mtDNA^– (Scale bars, 5 µm for single cell, 1 µm for zoom) (d) Histogram and violin plot showing the distribution of ATAC-see and DNA signal. Shown are the min-max normalized mean intensities from 27,079 segmented objects (e) Distribution of ATAC-see and DNA signal from 6 individual U2-OS cells. Shown are the min-max normalized mean intensities from each segmented object.

Extended Data Figure 4

(a) Tn5 in vitro activity in four different reaction buffers measured by DNA fragment analysis. Assembled transposome was mixed with 50 ng plasmid DNA for 30 minutes at 37°C and DNA fragments were assessed by Agilent TapeStation D1000. Tn5 activity fragments the plasmid DNA, resulting in the appearance of smaller (<500 bp) bands. Four buffers were tested: B1 (50 mM Tris, pH 7.4, 10 mM potassium chloride, 75 µM disodium phosphate, 274 mM sodium chloride), B2 (33 mM Tris, pH 7.8, 66 mM potassium acetate, 11 mM magnesium acetate, 16% N,N-dimethylformamide), B3 (20 mM Tris, pH 7.6, 10 mM magnesium chloride, 20% N,N-dimethylformamide), and B4 (50 mM TAPS, pH 8.5, 25 mM magnesium chloride, 40% PEG8000). Buffer B2 was used for ATAC-see reactions performed in Extended Data Figures 2 and 3. (b) Z-projection of the max intensities of a background corrected field-of-view of U2-OS cells treated with Tn5 transposomes in the presence and absence of EDTA. DNA was labeled with an α-ss/dsDNA antibody. (Scale bar, 10 µm) (c) Representative images of U2-OS cells sowing ATAC-see signal after treatment with Tn5 over a 60 minute time course. Two intensity ranges are shown to highlight the nuclear and mitochondrial signals .(Scale bar, 10 µm) (d) Confocal fluorescence microscopy showing mtDNA labeling throughout the mitochondrial network. A single Z-plane (0.3 µm) is shown. The mitochondrial network is labeled with an α-TOM20 antibody, mtDNA with an α-ss/dsDNA antibody, and chromatin with DAPI (Scale bars, 10 µm, 2 µm for zoom).

Extended Data Figure 5

(Related to Figure 2) (a) UpSet plot showing the co-occurence of footprints on the same molecule at the Termination Associated Sequence (TAS), Conserved Sequence Box I (CSBI), and MTERF1 binding site. Maximum footprint sizes were set for each footprint based on the footprint size distribution at these loci: 60 bp for TAS, 140 bp for CSBI, and 35 bp for MTERF1. Reads with larger footprints at these loci were not considered. Paired Student t-tests were used to compare the frequency of footprint co-occurrence, n.s. signifies p-value > 0.05, *** signifies p-value < 0.001. The categories in the plot represent 1.74%, 1.94%, 3.75%, 1.66%, 4.05%, and 2.08% of the total molecule population for replicates 1–6, respectively. (b) Hia5 MTase activity on single-stranded (ssDNA) and double-stranded DNA (dsDNA) substrates. The K_m for dsDNA is 0.233 µM. A lower limit of 3.48 µM was set for the K_m for ssDNA as the reaction never reached saturation even at 5 µM substrate. Results shown are the mean with s.d. from three replicates. (c) mtFiber-seq methylation strand bias from genomic positions 1,000 to 3,000 in untreated HeLa cells and cells treated with 2CMA. Methylation bias is calculated as the number of methylations on the light strand and heavy strand, averaged using a 150 nt window and normalized against the region’s AT content. Each window was required to have at least 2,250 methylations across all reads combined. (d) mtFiber-seq methylation strand bias at the NCR from three biological replicates of HeLa cells treated with DMSO or 2CMA. Methylation bias is calculated as the number of methylations on the light and heavy strands, averaged over a 150 nt sliding window and normalized against the region’s AT content. Each window was required to have at least 2,250 methylations across all reads combined (e) Log₂ fold-change in the methylation strand bias score between 2CMA treated and control samples at the D-loop and three alternate genomic loci. Individual dots represent four biological replicates. Samples were compared with a Student’s t-test, * signifies p-value < 0.05, n.s. Signifies p-values > 0.05. Results from 4 biological replicates shown. (f) Heatmap showing RNA levels in HeLa cells after treatment with 2CMA as measured by NanoString. RNA counts were internally normalized to GAPDH. Shown are levels for different treatment times with 2CMA relative to the DMSO control. All RNAs shown are mitochondrially encoded except for NDUFA7, which is nuclear encoded. Shown is the mean from three biological replicates. (g) Bar plot showing the percent of total nucleoids containing D-loops in five cell types. A nucleoid was defined as having a D-loop if it had at least 7 methylation events within the region and a Light:Heavy strand methylation ratio of 3.01 or greater. Individual dots represent 6 biological replicates for HeLa S3, two for U2-OS, and 3 for each HSMM differentiation time point.

Extended Data Figure 6

(Related to Figure 2) (a) Histogram of the natural log of the ratio of light strand methylation to heavy strand methylation for reads with a minimum of 7 methylations in the D-loop. A Gaussian mixture model was applied and a threshold was identified based on a GMM posterior probability of 0.99. Red and green lines indicate each Gaussian fit. The blue and orange lines indicate the posterior probability of each population. A threshold of 3.01 was determined and used to identify reads with D-loop as shown in main Figure 2h,i and Extended Data Figure 6b,c (b,c) Heatmap of the footprint size enrichment at the D-loop region in (b) reads containing a D-loop and in (c) reads lacking a D-loop in HeLa cells. Each row represents a footprint size, and each column shows a position in the genome. Presence of a D-loop was calculated using a GMM with a threshold of 3.01 from the log distribution of the ratio of Light Strand and Heavy Strand methylation.

Extended Data Figure 7

(Related to Figure 3) (a) Volcano plot showing differential expression analysis of human skeletal muscle myoblasts in differentiation media for 3 days compared to 0 days (left) and 6 days compared to 0 days (right). Red dotted lines are shown for a p_adj value of 0.01 and a fold-change value of 1.5. OXPHOS genes are shown in yellow and key nucleoid-associated proteins are labeled in light blue. (b) Western blot of nuclear-encoded (NDUFS1) and mitochondrial-encoded (ATP6) OXPHOS subunits from human skeletal muscle myoblasts in differentiation media for 0, 3, and 6 days. GAPDH was used as a loading control. Shown are two biological replicates. (c) mtFiber-seq methylation strand bias at the NCR from three biological replicates of human skeletal muscle myoblasts after 0, 3, and 6 days in differentiation media. Methylation bias is calculated as the number of methylations on the light and heavy strands, averaged over a 150 nt sliding window and normalized against the region’s AT content. Each window was required to have at least 2,250 methylations across all reads combined. (d) Western blot of TFAM levels from human skeletal muscle myoblasts in differentiation media for 0, 3, and 6 days. GAPDH was used as a loading control. Shown are three biological replicates. (e) Quantification of relative mtDNA levels by qPCR from human skeletal muscle myoblasts in differentiation media for 0, 3, and 6 days. Shown are mtDNA levels relative to day 0 from three biological replicates. Separate replicates are indicated by circle, triangle, and square shapes. (f) Quantification of TFAM levels by western blot. TFAM bands were quantified and normalized against GAPDH. Shown are the TFAM levels relative to day 0. Shown are three biological replicates. Separate replicates are indicated by circle, triangle, and square shapes.

Extended Data Figure 8

(Related to Figure 3) (a) Western blot of Proteinase K protection assay in the presence or absence of 0.5% Tween-20 and NP-40/Igepal-630 to validate overexpressed TFAM-HA localization to mitochondria. TFAM-HA appears as a third upper band and was protected from digestion except in the presence of 0.5% detergent. (b,c) Tables of Pearson correlation coefficients from three mtFiber-seq replicates from HeLa S3 TetOn-TFAM-HA cells treated with (b) DMSO or (c) doxycycline for 48 hours comparing (bottom left) fraction of reads methylated at each adenine and (top right) fraction of reads with a footprint at each genomic position. PacBio chemistry version is indicated for each replicate. Pearson’s correlation coefficient is shown for each correlation. (d) Quantification of relative mtDNA levels by qPCR from HeLa S3 TetOn-TFAM-HA cells treated with DMSO or doxycycline for 48 hours. Shown are mtDNA levels relative to the DMSO control from three biological replicates. (e) UpSet plot showing the co-occurrence of footprints on the same molecule at the Termination Associated Sequence (TAS) and Conserved Sequence Box I (CSBI) with footprints at the MTERF1 binding site from HeLa S3 TetOn-TFAM-HA cells treated with DMSO or doxycycline for 48 hours. Maximum footprint sizes were set for each footprint based on the footprint size distribution at these loci: 60 bp for TAS, 140 bp for CSBI, and 35 bp for MTERF1. A maximum footprint size was set at 170 bp. Reads with larger footprints at these loci were not considered. Individual dots represent three biological replicates. The categories in the plot represent 0.59%, 0.98%, and 1.67% of the total molecule population for DMSO replicates 1–3, respectively, and 0.44%, 0.54% and 1.31% of the total molecule population for doxycycline replicates 1–3, respectively. (f) UpSet plot showing the co-occurrence of footprints on the same molecule at the TAS and CSBI with containing a D-loop in HeLa S3 TetOn-TFAM-HA cells treated with DMSO or doxycycline for 48 hours. Maximum footprint sizes of 60 bp and 140 bp were used for TAS and CSBI, respectively. Individual dots represent three biological replicates. The categories in the plot represent 1.23%, 1.85%, and 3.37% of the total molecule population for DMSO replicates 1–3, respectively, and 1.00%, 1.14%, and 2.61% of the total molecule population for doxycycline replicates 1–3, respectively. (g) Log₂ fold-change in the fraction of reads containing a D-loop from all nucleoids and accessible nucleoids (>1% adenines methylated per read) in HeLa S3 TetOn-TFAM-HA cells treated with doxycycline for 48 hours relative to the DMSO control. Data were subsampled to each other to match methylation distributions.

Extended Data Figure 9

(Relative to Figure 4) (a) TFAM binding to a 28mer corresponding to the HSP TFAM binding site measured by fluorescence polarization. The K_D was determined to be 6.2 nM. Results shown are the mean with s.d. from three replicates. (b) (Left) 0.5% agarose gel showing mtDNA LR-PCR product before and after column cleanup. (Right) Genomic DNA ScreenTape analysis showing mtDNA LR-PCR product. (c) Two replicates of a dot blot assessing methylation of mtDNA with increasing concentrations of TFAM. A dilution series of DNA amounts were adsorbed onto a nitrocellulose membrane, crosslinked, and detected with an anti-m6A antibody. (d,e) Tables of Pearson correlation coefficients between replicates and TFAM concentrations for (d) the fraction of reads methylated at each adenine and (e) the fraction of reads with a footprint at each genomic position from two replicates with each TFAM concentration. (f) Hill coefficients calculated from genomic position 2,500 to 14,000 from a four parameter logistics regression based on the fraction of reads protected with a footprint at least 20 bp long. Results are from the average of two replicates. (g) Meta density plots of footprint sizes from (top) 26 high affinity sites and (bottom) 64 low affinity sites at each TFAM concentration with the 95% confidence intervals shaded. Distributions between high and low affinity sites are significantly different at 5 µM TFAM (KS test, p << 0.05, D = 0.32) and 10 µM TFAM (KS test, p << 0.05, D = 0.29). (h) Heatmaps of footprint size enrichment at the light strand promoter (LSP) (top) and heavy strand promoter (HSP) (bottom) from HeLa cells subsampled to each TFAM concentration. Each heatmap row represents a footprint size, and each column shows a position in the genome.

Extended Data Figure 10

(Relative to Figure 4) (a,b) Heatmaps of the footprint size enrichment from two high affinity sites for each concentration of TFAM and from HeLa cells subsampled to each concentration: (a) from position 5,040 to 5,540 and (b) from position 11,900 to 12,400. Each heatmap row represents a footprint size, and each column shows a position in the genome. Line plots indicating the 1/K_1/2 across these loci are shown above the heatmaps. (c,d) Density plots showing the footprint size distribution between 20 and 180 bp at (c) position 5,290 and (d) position 12,150 as a function of TFAM concentration. (e,f) Heatmaps of the footprint size enrich from two low affinity sites for each concentration of TFAM and from HeLa cells subsampled to each concentration: (a) from position 5,991 to 6,491 and (b) from position 10,332 to 10,802. Each heatmap row represents a footprint size, and each column shows a position in the genome. Line plots indicating the 1/K_1/2 across these loci are shown above the heatmaps. (g,h) Density plots showing the footprint size distribution between 20 and 180 bp at (g) position 6,241 and (h) position 10,552 as a function of TFAM concentration.

Fig. 1.. The majority of human mitochondrial nucleoids are inaccessible

(a) Schematic of the human mitochondrial genome. (Top) The non-coding region (NCR) of the mitochondrial genome, highlighting the Termination Association Sequence (TAS), the D-loop and 7S DNA, the heavy strand origin of replication (O_H) and Conserved Sequence Box I (CSBI), and the light and heavy strand promoters (LSP and HSP, respectively). (Bottom) The 16.5-kb circular human mitochondrial genome encodes 13 polypeptides (blue), 22 tRNAs (green), and 2 rRNAs (orange). (b) Schematic depicting the experimental design. Human cells were subjected to cellular fractionation to isolate mitochondria, which were then permeabilized and treated with Tn5 Transposase for ATAC-seq or the Hia5 m6A-MTase for mtFiber-seq. (c) Permeabilization of mitochondria with increasing concentrations of detergent was assessed by Proteinase K digestion of the outer mitochondrial membrane (OMM) protein TOM40, the inner mitochondrial membrane (IMM) protein COX1, and the matrix-soluble protein HSP60, followed by western blot analysis. Representative blot of two independently repeated experiments. (d) Mitochondrial genome comparing the signal between ATAC-seq, aggregated methylation from mtFiber-seq, and 80 randomly sampled mtFiber-seq reads. Individual reads are indicated by horizontal black lines, and m6A-modified bases are marked by purple vertical dashes. (e) Zoom-in of positions 13,000–16,000 in the mitochondrial genome showing 20 sampled reads. (f) Bar plot showing the percent of HeLa mtFiber-seq reads per bin based on the number of m6A modifications per read. The majority of reads (80%) have fewer than 1% As methylated per molecule. Reads with greater than 5% As methylated were aggregated into a single bin. N = 6 biological replicates. (g) Bar plot showing the fraction of nucleoids with >1% As methylated in five cell types: HeLa, U2-OS, undifferentiated skeletal muscle myoblasts, and 3- and 6-day differentiated skeletal muscle myocytes. N = 6 biological replicates for HeLa, 2 biological replicates for U2-OS, and 3 biological replicates for each time point across HSMM differentiation. (h) Bar plot showing the fraction of nucleoids with >1% As methylated in HeLa cells after treatment with antimycin A (AMA) or vehicle control for 24 hours. N = 2 biological replicates. For all plots, error bars represent s.d. of the mean.

Fig. 2.. Accessibility patterns reveal the mtDNA architecture

(a) HMM used to identify accessible and inaccessible regions of the genome using probabilities of methylation in hexamer sequence contexts. (b) Heatmaps showing the identified accessible and inaccessible regions of the genome from six biological replicates from HeLa cells. Each row represents an individual read. Accessible regions are colored purple, and protected regions are colored according to size. (c) (Top) Metaplot of footprint enrichment showing the fraction of reads with >10% m6A. (Bottom) Zoom-in of the MTERF1 binding site for reads containing >10% m6A. The MTERF1 binding site is indicated by a blue line. (d) (Top) The crystal structure of MTERF1 bound to its 22 bp binding site (PDB code 3MVA). (Bottom) Heatmap of the footprint size enrichment at the MTERF1 site. The binding site is indicated by a blue line. (e) (Top) ChIP-seq tracks showing the fold change of signal over control for Polɣ and TWINKLE. Metaplot of the footprint enrichment showing the fraction of reads with >10% m6A. (Bottom) Zoom-in showing the region surrounding the D-loop. Heatmap shows reads containing >10% m6A. (f) Cartoon depicting D-loop formation. POLRMT transcribes the RNA primers for POLɣ to synthesize the 7S DNA, forming a D-loop. 2’-C-methyladenosine (2CMA) inhibits POLRMT, resulting in no 7S DNA synthesis or D-loop formation. (g) mtFiber-seq methylation strand bias at the NCR in HeLa cells. Bias is calculated as the number of methylations on the light and heavy strands, averaged over a 150-nt sliding window and normalized against the region’s AT content. (h) Heatmap of the footprint size enrichment at the D-loop from HeLa cells in reads containing a D-loop. (i) UpSet plot showing the frequency of nucleoids containing combinations of footprints at TAS and CSBI and containing a D-loop. Maximum footprint sizes of 60 bp and 140 bp were used for TAS and CSBI, respectively. Groups were compared by paired two-sided Student’s t-test, n.s. P>0.05, *P<0.05, **P<0.01. Exact p-values: 0.0097, 0.0349, and 0.0045. N = 6 biological replicates. Error bars represent s.d. of the mean. The categories in the plot represent 2.10%, 2.31%, 1.89%, 1.01%, 2.81%, and 2.67% of the total molecule population for replicates 1–6, respectively.

Fig. 3.. Altered TFAM levels shift the population of accessible nucleoids

(a) Methylation strand bias at the NCR in HSMM throughout differentiation. (b) Log₂ fold change in the strand bias score after 3 and 6 days in differentiation media (DM) across three regions. Samples were compared with two-sided Student’s t-tests: *P<0.05; n.s. P>0.05. Exact p-values: 0.034 (3d vs 0d) and 0.022 (6d vs 0d). N = 3 biological replicates. (c) Log₂ fold change of D-loop containing nucleoids from HSMMs after 3 and 6 days in DM relative to undifferentiated myoblasts. Datasets were subsampled to match methylation distributions. Samples were compared by Fisher’s exact tests: *P<0.05. Exact p-values: 0.066 and 0.244 (3d vs 0d replicates) and 0.0008 and 0.027 (6d vs 0d replicates). N = 2 biological replicates. (d) Bar plot showing the TFAM:mtDNA ratio from HSMMs after 0, 3, and 6 days in DM. N = 3 biological replicates. (e) Overexpression of TFAM. WT HeLa cells and TetOn-TFAM-HA HeLa cells were treated with DMSO or doxycycline for 48 hours. (Top) TFAM levels were assessed by western, with ACTB as a control. Treatment of TetOn-TFAM-HA cells with doxycycline results in the appearance of a band corresponding to the HA-tagged TFAM. (Bottom) TFAM bands were quantified and normalized against ACTB. N = 5 biological replicates. Error bars represent s.d. of the mean. (f) Confocal microscopy showing the TFAM-HA construct localized to mitochondria and nucleoids. Cells were treated with doxycycline for 48 hours and labeled for TOM20, ss/dsDNA, and HA (Scale bars: 5 µm and 1 µm for the zoom-in). (g) Bar plot showing the fraction of accessible nucleoids (>1% m6A). Individual replicates were compared with two-sided chi-squared tests: ***P<0.001. Exact p-values: 2.265*10⁻⁹, <2.2*10⁻¹⁶, and 8.022*10⁻¹³ for replicates 1–3. N = 3 biological replicates. (h) Bar plot showing the TFAM:mtDNA ratio after treatment with DMSO or doxycycline for 48 hours. N = 3 biological replicates. (i) (Left) Line plot and (Right) scatterplot showing the footprint enrichment with TFAM overexpression relative to the control (Pearson’s r=0.97). Datasets were subsampled to match methylation distributions. Error bars represent s.d. of the mean for all plots.

Fig. 4.. In vitro reconstituted nucleoids reveal preferential TFAM binding and nucleation sites throughout the genome

(a) Nucleoids were reconstituted with PCR-amplified mtDNA and recombinant TFAM and methylated with Hia5 (b) (Left) Dot blot assessing methylation with increasing TFAM. DNA was adsorbed onto a nitrocellulose membrane and detected with an anti-m6A antibody. (Right) Quantification of the dot blot. Intensities were quantified and normalized against the 0 µM sample. N = 2 replicates. (c) Five randomly sampled mtFiber-seq reads for each TFAM concentration. Reads are indicated by black lines and m6A bases are marked by purple dashes. (d) Box plot showing the percent of As methylated per read for each TFAM concentration. Box-plot elements are: center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range. N = 29,799, 47,300, 68,377, 80,562, and 82,809 reads for 0, 5, 10, 20, and 30 µM TFAM, respectively (e) Scatterplot showing the footprint enrichment across the genome in HeLa cells and in vitro with 5 µM TFAM for reads subsampled to match methylation distributions (Pearson’s r=0.665). (f) Example binding curves from in vitro mtFiber-seq data. Reads were classified as bound at a site if a ≥20 bp footprint overlapped the position. Data were fit to a four-parameter logistic regression. N = 2 replicates. (g) Affinities were determined from position 2,500 to 14,000. 1/ K_1/2 constants are shown. Results are from the average of two replicates. (h) Meta density plots of footprint sizes from (top) 26 high affinity sites and (bottom) 64 low affinity sites at each TFAM concentration. The center line represents the kernel density estimate with the 95% confidence intervals shaded. Distributions between high and low affinity sites are significantly different at 5 µM TFAM (two-sample KS test, P<2.2*10⁻¹⁶, D=0.32) and 10 µM TFAM (two-sample KS test, P<2.2*^–, D=0.29). (i) Heatmaps of footprint size enrichment at two loci from in vitro reconstituted nucleoids (top) and for HeLa nucleoids, subsampled to match the methylation distribution of the 20 µM TFAM dataset. Line plots indicating the 1/K_1/2 are shown above each heatmap (j) Density plots showing the footprint size distribution at positions 5,290 and 6,241 as a function of TFAM concentration. (k) TFAM binds and nucleates from preferred sites (blue) located throughout the genome.