Mathematical image analysis based cell reprogramming with applications for epigenetic and non-epigenetic base induced pluripotent stem cell derivation

[0018] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0019] FIG. 1 illustrates how epigenetic modification of histones and DNA convert open euchromatin to condensed heterochromatin. The multistep process requires coordination of several enzymatic post translational modifications of chromatin including deacetylation of histones (by multiple HDACs), and methylation of histones (methyltransferases, HMT notably target H3 lysine 9 and 27) and DNA on Cytosine (by DNA methyltransferases). Figure from (Egger et al., 2004).

[0020] FIG. 2 shows that epigenetic enzymes are found in complexes bound to heterochromatin in differentiated cells. Studies indicate that HDAC 1, HDAC 2, SUV39H1 and DNMT 3b associate in complexes in somatic cells (right) that bind to heterochromatin and methylate H3k9 and 5′MeC (right). Binding constants are not equal among enzymes. In pluripotent cells, our preliminary immunofluorescence data (below) suggests that enzymes are not bound tightly (left).

[0021] FIG. 3 shows the identification of existing iPSC and hESC colonies (A) Established, mature iPSC colony containing over 300 cells on exterior extracellular matrix proteins. (B) Computer image analysis of colony where the image exterior is marked red and the pluripotent colony marked blue. The multiresolution texture algorithm was applied to obtain results. (C) A similar algorithm is also able to distinguish between a very small colony of 30 hESCs grown on feeders 24 hours after passaging, a situation similar to nascent iPSCs derived from feeders. (D) Discrimination of the colony from surrounding feeders.

[0022] FIG. 4 illustrates nascent iPSCs. Left panels, brightfield images at 0, 23 and 59 hrs after transfection with MKOS transcription factors. Right panels, rtTA-IRES-GFP knocking reveals newly forming iPSCs at 59 hrs, but not with certainty at earlier times.

[0023] FIG. 5 provides an example of time-lapse analysis. (A-E) GFP-H2B labeled nuclei shown in single confocal slices by color overlays blue (t=0 min), green (t=1 min) and red (t=2 min). Pluripotent nuclei (A, B) show more color when compared to differentiated (C, D) or control somatic endothelial nuclei (E). Quantitative texture analysis (second row, F) is shown for nuclei (color squares), each square containing 10 images over a 10 min interval. Color bar is measure of the Kullback-Leibler (K-L) divergence between textures (M1-M4) (Blue: min. distance, red: max distance). Pluripotent nuclei (A, B) differ from somatic nuclei (E) with a large K-L divergence (red and yellow in (F). Distance from somatic cell E decreases as cells differentiate.

[0024] FIG. 6 demonstrates performance evaluation of classification scheme applied to stem cell chromatin. Curves represent probability of correctly identifying a pluripotent nucleus vs. prob. of a false positive from partially or totally differentiated nuclei. Red curve indicates one can correctly identify a pluripotent nucleus (A) from a somatic (D) with prob. 0.99 for a maximum false positive prob. of 0.003. When compared to a 2 day differentiated nucleus (B, blue curve), a tradeoff is 0.96 vs. 0.02. For a 5 week differentiated nucleus (C, black curve), it is 0.094 vs. 0.02.

[0025] FIG. 7 illustrates the Matrix Edge concept. (a) Noise free image of a square, (b) image corrupted by spatially varying high intensity noise (noise in upper left corner of square much stronger than in other corners), processed (c1) with conventional segmentation, the edges are blurred as (c2) neighborhoods over which local smoothing takes place vary locally in size only, while with (d1) Matrix Edge segmentation, the edges are much improved thanks to (d2) the ability to vary, in addition to size, the shape and orientation of local neighborhoods.

[0026] FIG. 8 shows the application of Matrix Edge and Onion Peeling to stem cell nuclei. (a) Raw input (left column), smoothed (middle), and edge field (right) output images of pluripotent (top) and differentiated (bottom) nuclei. Matrix edge field output (right column) is rich in spatially varying details. Layers of the edge field image are to be extracted as if (b) peeling an onion and quantitated to (c) extract numerical features whose values are different in pluripotent (top curve) and differentiated nuclei (bottom curve).

[0027] FIG. 9 shows a stem cell nucleus. (a) segmentation of the nucleus, blue contour indicates identified nuclear region; (b) the image normalized, note that the brightest region is to the bottom right; (c) locus of internal window centers at each decomposition level, red are the internal windows at level 3, orange at level 2, green at level 1; (d) Parzen vs. GGD estimate of the coefficient distribution in the third horizontal subband.

[0028] FIG. 10 shows that antigen retrieval for H3K9me3 and HDAC1 improve uniformity of immunodetection. Chromatin reorganization during differentiation was evaluated in hESC colonies by immunofluorescence. Observations in control cells, including HeLa and MEFs, demonstrated inconsistent staining with heterochromatin antibodies using PFA fixation or MeOH fixation. Inconsistent antibody staining is not due to inconsistent penetration of antibody since all cells were detergent extracted and could be stained for other nuclear proteins such as transcription factors (not shown). We evaluated immunostaining of H3K9me3 (green) and HDAC1 (red) staining under different post fix antigen retrieval conditions. Optimal HDAC staining was obtained with 95 C citrate, pH 6.0, for 30 min© but this condition produced poor immunostaining of H3K9me3. Better, though suboptimal, H3K9me3 staining was observed in room temperature citrate, pH 7.4, for 5 min as not all cells stained positively for H3K9me3. Better consistency of both antigen staining was produced by alkaline conditions (C) or post paraformaldehyde treatment with cold methanol (D). Methanol is modestly better than alkaline conditions. These methods produce consistent staining for all antibodies studied to date including H3K9-3Me, H3K9ac, HDAC1, HDAC2, Crest, Oct-4, Nestin, HP1 alpha, HP1 beta, with almost 100% positive antigen staining and low cytoplasmic background staining.

[0029] FIG. 11 shows adaptive windowing and texture analysis. (a) Conceptual flow diagram for three-state textural classification. (b)-(e). Textural analysis of early differentiated nucleus. (b) Parzen vs. GGD estimate of the coefficient distribution in the first vertical subband of d; (c) segmentation of the nucleus, blue contour, indicates the identified nuclear region; (d) the normalized image, note that the brightest region is to the bottom right; (e) modeling coefficients selected at each decomposition level, red are the coefficients for level 4, orange at level 3, and yellow at level 2, and cyan at level 1; light blue are interior to the cell but exterior to level 1; dark blue is exterior to the cell. (f)-(g) as (c)-(d) but for a pluripotent cell.

[0030] FIG. 12 illustrates markers for pluripotent and early-differentiated cells Oct4 colocalizes in nuclei with additional markers for pluripotency including the transcription factors, Hnf3b (A) Nanog (B) and Foxd3 (C) in colonies grown in normal density feeders (A-C). On low-density feeders (D) Oct4 is absent and the trophectoderm marker, Cdx2 is present in rare cells. Neurectodermal lineages are preferred when hESC are differentiated on low density feeders in our hands.

[0031] FIG. 13 illustrates a pluripotency panel for molecular confirmation of early developmental stages of pluripotent lineages. H7's grown in 2 standard conditions—on feeders (in DSR and on gelatin) and in feeder free conditions (in StemPro and on Geltrex (matrigel)) show different levels of markers of pluripotency especially Oct, Sox2, HNF3b FoxD3 and Rex1. Rex 1 in particular is a marker of ICM-like cells (StemPro, B) while FGF5R is characteristic of post-implantation epiblast (feeders+DSR, C). Both ICM and epiblast are pluripotent.

[0032] FIG. 14 presents nuclei classification and KLD nearest neighbor classification. Class 1 (deforming pluripotent nuclei): images 1-11, class 2 (translocating pluripotent nuclei): 12-22, class 3 (streaming two day differentiated nuclei): 23-33, class 4 (gel-like five week differentiated nuclei): 34-43, class 5 (stationary endothelial cell): 44-53. For each nuclear image on the y-axis, the five nearest neighbors (lowest summed KLD) have been indicated by a white square along the x-axis.

[0033] FIG. 15 shows morphological comparison of pluripotent hESC and differentiated trophectoderm colony images. Pluripotent hESC cultures, line WA07 (a, d, g, i) were compared to hESC that were differentiated in vitro for 4 days in culture with 100 ng/ml BMP4 (b, e, h, j). The squamous epithelial monolayer that was formed is trophectoderm, found in the outer shell of blastocysts and destined to form the placenta. Fixed colonies were stained with the DNA dye, Hoechst 33342 (a, b) and with antibodies to the pluripotency marker OCT4 (d) and the trophectoderm marker, CDX2 (e). Morphological differences that are characteristic of the spreading colonies include an increase in cell (c) and nuclear size (f). The Hoechst image of 4 colonies were analyzed for nuclear size and number of nuclei per colony area (average cell area) and differences between pluripotent and trophectoderm colony morphology were significant at p<0.05 by Students T test. Colony morphology in live cells (g, h), not fixed and immunostained as was performed in (a-f), is provided by the distribution of spatial frequencies in the image (i, j). The 1D line profile along a 1 mm white line in (g, h) provides an integrated measure of colony morphology that depends on both nuclear and cell size. Spatial frequencies are higher in pluripotent than in TE colonies and can be more completely measured using wavelet analysis, a generalized 2D, multiscalar measure of image texture. Bar in (e) is 100 μm and bars in (g) and (h) are 1,128 μm. Empirical probability density functions (k-l) for three detail subbands of the line scans in (i) and (j) quantify the concept of spatial frequency. In the second (k) and third (l) subbands (scales 2² and 2³ pixels, respectively), the thicker distribution and wider tails in the pluripotent density indicate comparatively more edges at finer scales, and hence more and smaller cells in the pluripotent image. The situation is reversed in the sixth (k) detail subband (scale 2⁶=64 pixels), which is about the scale of the average diameter of pluripotent and differentiated stem cells (69±30 pixels and 82±40 pixels, respectively, mean±SD, 400 cells in 5 independent colonies).

[0034] FIG. 16 shows classification of nuclei. (a) KLD nearest neighbor classification. Class 1 (early differentiated nuclei): images 1-25, class 2 (pluripotent nuclei): 26-53. For each nuclear image on the y-axis, the ten nearest neighbors (lowest summed KLD) have been indicated by a white square along the x-axis; (c)-(d) classification accuracy with two and three decomposition levels for varying wavelet and pdf/KLD estimators; classification accuracy is highly robust to the number of decomposition levels and choice of wavelet, and parametric (GGD, SαS) methods exhibit comparable accuracy to non-parametric (A-L, L-Q) ones.

[0035] FIG. 17 illustrates hESC colony classification. (a) the statistical dissimilarity measure (KLD) between two windows is visualized via color in each entry of the matrix, from blue (similar) to red (very dissimilar); thus, the diagonal elements are each dark blue as each window is texturally identical to itself, and classes containing similar models appear as bluish blocks along the diagonal; (b) window by window classification of a pluripotent hESC colony; three passes are used to classify the original image, with intermediate results shown after the classification and border detection phases; blue is pluripotent, green is differentiated, red is exterior, and gold is unknown; (c) differentiated colony, clearly distinguished from pluripotent; (d) algorithm robustness to changes in classifier parameters; (d-i) is constant wavelet (Daubechies-4), classifier parameters (k, k_n) vary; blue is pluripotent, green is differentiated, red is exterior; bar indicates extent of 90% confidence interval, diamond is maximum a posteriori (MAP) estimate of accuracy; (d-ii) is pluripotent, (d-iii) is differentiated, and (d-iii) is exterior, these with constant classifier parameters (k, k_n)=(7, 5) and varying wavelet; (e) comparison between GGD and SαS models, Daubechies-4 wavelet, (k, (7, 5); performance is robust to choice of wavelet, classifier parameters; GGD performance equals or exceeds SαS performance.

[0036] FIG. 18 shows confocal images of GFP-H2B labeled nuclei from pluripotent hESC (A, B), two-day differentiated (C), and five-week (D) differentiated hESC, and a control fully differentiated endothelial cell (E); chromatin granularity increases during differentiation, approaching the control cell. Each color image is the overlay of images at time 0 min. (blue), 5 min. (green), and 10 min. (red). Color bar visually presents whether nuclear change over time (brighter color in pluripotent cells) or are unchanging over time (gray in differentiated cells). Bar in A is 10 μm.

[0037] FIG. 19 presents a parametric vs. non-parametric approach. Summed KLD showing dissimilarity between cell 11 (part of class 1) and classes 1 and 2. Class 1: images 1-11, class 2: 12-22. GGD statistical model is top; Parzen is bottom. Note the Parzen's improved class separation.

[0038] FIG. 20 shows further applications. (a, b) proof of concept on mixed, spontaneously differentiated stem cell colonies (line UC06) (a) showing differentiated epithelial cells (cyan, library windows 1-4), dark pigmented differentiated cells (purple, library windows 5-8), exterior feeder cells (orange, library windows 9-16), and pluripotent cells (light blue, library windows 17-20), bar is 100 μm; pluripotent cells may be distinguished from other textures with a KLD of 6 as shown in (b), the KLD plot for the corresponding colored regions in the mixed image (a). In (c, d), a discarded metaphase II egg from a Rhesus macaque was imaged at 40× with a Hoffman Modulation Contrast objective and observed for 10 hrs in an environmental chamber (c). Reducing CO2 from 5% to atmospheric concentrations produced a shift in pH before frame 30. The egg became atretic and cytoplasm granularity changed from fine-grained to coarse-grained as it died. The KLD is computed between each frame and the first (blue curve, live egg) and last (red curve, dead egg) frames. Initially the blue curve is low while the red is high, indicating textural similarity to the healthy cell and dissimilarity to the dead one, respectively; this changes starting at frame 30, when the egg is damaged.

[0039] FIG. 21 illustrates small hESC colonies after passaging appear similar to nascent iPSCs. Colonies are easily detectable on day 3 (C, F) at 4× and 10×, respectively. However, on day 1 colonies are distinct only at 10× (D) not 4× (A). Therefore, 2 variables for detection of nascent iPSCs are the time after formation and the scale and resolution of the image detection. The 4× field of view is 2 mm×3 mm.

[0040] FIG. 22 demonstrates identification of existing iPSC and hESC colonies (A) Established, mature iPSC colony containing over 300 cells on exterior extracellular matrix proteins. (B) Computer image analysis of colony where the image exterior is marked red and the pluripotent colony marked blue. The multiresolution texture algorithm was applied to obtain results. (C) A similar algorithm is also able to distinguish between a very small colony of 30 hESCs grown on feeders 24 hours after passaging, a situation similar to nascent iPSCs derived from feeders. (D) Discrimination of the colony from surrounding feeders.

[0041] FIG. 23 shows a brightfield time-lapse series of mESC colony growth and death. At higher magnification, cell growth and death can be measured cell-by-cell, in addition to morphological changes characteristic of differentiation. Mouse embryonic stem cell (mESC) division and death occur at high frequency during colony expansion. A small colony of adherent mESC cells were observed by time-lapse Hoffman modulation contrast microscopy for 41 hrs. An excerpt from 34 min is shown above where division of rounded cells can be observed. Mitotic cells from chromosome condensation to midbody formation progressed rapidly (28±7 min n=29 among 5 time lapse series of 6 hr duration). Data for division and cell death are pooled at 6 hr intervals in the bar graphs, above. Total number of cells increased for 35 hours until the culture media is exhausted. Although mitotic indices were high, ranging from 11 to 36%, death rates were also high, limiting colony expansion. A majority of cell deaths occurred just after telophase.

[0042] FIG. 24 illustrates how heterochromatin markers coincide, while euchromatin markers separate during differentiation. Chromatin markers (green or red) are co-localized if the overlay produces yellow. Confocal channels were adjusted for equal luminance in each channel. DNA is blue in all figures. (A-C) Heterochromatin markers (H3K9me3 and H3K27me3) do not overlap in pluripotent cells while overlap increases in neural stem cells and endothelial cells. (D-F) euchromatin markers H3K9Ach and H3K4me2 decrease overlap during differentiation. (G-I) Euchromatin and Heterochromatin markers do not overlap. Pluiripotent stem cells show different organization of histone pos-translational markers than differentiated cells. Different markers of heterochromatin (H3K9me3 green and H3K27me3 red, Figure C5, A-C) have different distributions in pluripotent cells but are more colocalized (yellow) in neurally differentiated cells and in somatic cells. Conversely, markers of euchromatin (H3K9ac, green, H3K4me2, red, Figure C5, D-E) are more colocalized in pluripotent cells. Euchromatin and heterochromatin markers are spatially distinct (Figure C5 G-I) at all stages.

[0043] FIG. 25 shows Western Blot analysis of Oct4 and modification of H3K9. Cells on feeders have lower levels of Oct4 (black) and H3K9ac than in the feeder free conditions, conditioned DSR media and mTeSR. H3K9me levels are higher in hESC on feeders or feeder conditioned media in hESC in mTeSR on matrigel.

[0044] FIG. 26 shows that iPSC chromatin is similar to hESC chromatin. (A) The distribution of H3K9ac (red) is diffuse throughout the nucleus and H3K9me3 is localized in large chromocenters (green), similar to hESCs (FIG. 6). (B) 50% of iPSCs are in S phase (green, EdU, a deoxyuridine analog) similar to hESCs (not shown).

[0045] FIG. 27 illustrates how lysis of iPSCs removes histone H2B, similar to hESCs. Unlysed (A) and Lysed (B) iPSCs show a dramatic loss of H2B (red), with no discernable loss of DNA (Hoechst, blue) or recently incorporated EdU nucleotide in S phase cells (green). Cells were extracted for 5 min at 0 C in Ca2+ free buffer. Similar results were obtained in hESC (not shown). Extraction is a measure of histone exchange, complementary to FRAP.

[0046] FIG. 28 shows how DNA methyltransferases and histone deacetylases change expression and location upon neural differentiation. DNMT3b (A) and DNMT 3a (B) (red nuclei, top right) are present in pluripotent hESC expressing Oct 4 (green, top right), but absent in differentiated hESC (Hoechst, blue nuclei, bottom left). HDAC1 (red) is present in widely varying levels in pluripotent hESC compared to DNA (blue) and Oct4 (green) levels (C). (D) The ratio of HDAC1, HDAC2 and DNA also varies from location to location in pluripotent cells (top right, D) but is constant throughout the nucleus except the nucleolus in differentiated cells (bottom left, D). Despite the presence of DNMTs in pluripotent cells, activity is low as shown by low levels of 5′me Cytosine (E) Also, the HDAC1 selective target, H2BK5, is still acetylated in pluripotent hESCs (F, top right), but not in neuronal lineages (F, bottom left), despite a constant level of HDAC expression in pluripotent and differentiated cultures of hESC and mESC (G-I). Bar in D, 10 μm.

[0047] FIG. 29 demonstrates that shRNA HDAC1 knockdown in hESC is as effective as 1 nM TSA, the highest concentration without toxic effects. Cells in B were nucleofected with Open Biosystems HDAC1 shRNAmir in a retroviral pSM2 backbone. Cells were cotransfected with pMAXGFP and 70-80% of cells expressed GFP. Three days after nucleofection a control colony was identified by lack of cytoplasmic GFP (A) while pSM2 HDAC1b transfection was identified within small colonies containing cytoplasmic GFP (B). Quantitative measurement of the nuclear HDAC1 staining (red) is shown in the graph in (D), HDAC1 knockdown reduces HDAC1 staining 30 fold, to a level comparable to 1 nM TSA.

[0048] FIG. 30 illustrates how HDAC1 or HDAC2 knockdown delays TE differentiation in BMP4 for 48 hours. HDAC1 or HDAC2 knockdown prevents epithelialization, as shown in the panels (bars, smaller internuclear distance) as measured by epithelial morphology. Internuclear distance (bars, top panel) is significantly greater in controls (−knockdown) Significant at P<0.05.

[0049] FIG. 31 presents a characterization of three stages of neural differentiation (1, pluri) pluripotent hESC (2, early diff) neurectoderm, and (3, late diff) neural rosette. hESC, lines H7 (a-b) and HSF6 (c-g) were prepared on MEF feeders in pluripotent conditions (pluri, a) and differentiation conditions on low-density feeders (b-g). Flattened, differentiated cells in monolayers were identified at colony edges after week on low-density fibroblasts (early diff, b), while the interior of large multilayer colonies formed rosettes after 2-4 weeks (late diff, c). Pluripotent cells were positive for Oct4 and negative for nestin (a) while early-differentiated cells were positive for nestin and had reduced levels of Oct4 (b), characteristic of neurectoderm. In late differentiated cultures, rosettes of neural stem cells were common between 2 and 4 weeks as shown by multiple astral arrays of NCAM staining (green), within the thicker portions at the center of the colony (cell surface labeled with WGA, red, and Hoechst, blue) (c). Brightfield image of toluidine blue-stained epon-embedded thick section of a differentiated colony show multiple rosettes containing clear central lumens and radial arrays of cells (arrows, d). Rosettes contain a radial array of mitotic and interphase nuclei (e). At the core of the rosette, a clear zone is found (box, enlarged in f) that contains villar projections. The apex of these cells surrounding the lumen shows tight plasma membrane apposition, with an osmophillic mordant in the intercellular space adjacent to the lumen (f). Bundles of intermediate filaments and microtubules radiate from the core and centrioles can be found (arrows) at the apex near the lumen, characteristic of a functionally polarized neuropithelium (g). Bar in b, 10 μm; bar in c and d, 100 μm; bar in f, 2 μm; Bar in g, 1 μm.

[0050] FIG. 32 shows that chromatin condensations increases by accretion during differentiation Pluripotent hESC, line HSF-6, have nuclei with uniform, smooth chromatin that becomes progressively more granular between early and late differentiated stages. Pluripotent (column a-j), early differentiated (column b-k), and late differentiated (column c-l) colonies show increasing granularity as detected by electron microscopy (d-f are enlargements of the box in a-c) and light microscopy with the dye Yoyo-1 (g-i). In pluripotent cells, uniform, fine grained chromatin has no aggregates larger than the limit of detection by light microscopy (g), in neurectoderm, fine aggregates of chromatin are formed by accretion of osmophillic aggregates that are separated by 0.2 □m, and are resolvable by light microscopy (h) and coarse aggregates of chromatin in neural rosettes (0, detectable as larger aggregates by light microscopy (i). Line plots of confocal sections of Yoyo-1 stained DNA show slowly varying intensities (low spatial frequencies) in pluripotent cells (j), high frequency, small amplitude variations in early-differentiated cells (k), and longer frequency, but large amplitude variations in late-differentiated cells (i) (graphs have uniform scaling for intensity and size). Bar in a, 0.5 μM Bar in d, 0.125 μm, Yellow bars in g-i, 10 μm.

[0051] FIG. 33 indicates that texture analysis of heterochromatin condensation during neural differentiation. Chromatin texture is a 2 dimensional, multiscalar measure of object size in the image and is a generalization of linear spatial frequency represented in line plots. Pluripotent (a, d), early differentiated (b-e), and late differentiated (c-f) colonies were stained with the Yoyo1 (green) for DNA. Pluripotent cells were additionally stained with wheat germ agglutinin (a, d, red) to reveal cell and colony surfaces. Confocal images of nuclei (d, e, f are enlargements of the white rectangles in a, b, c, respectively) show that chromatin in pluripotent cells is relatively smooth with gradual variations in intensity across the nucleus, which progressively condenses into fine grain, low contrast patterns (e) and finally into larger grained high contrast domains that are distinctly bright or dark (f), characteristic of heterochromatin and euchromatin, respectively. Quantitative multiscalar wavelet analysis in 2 dimensions in (P) was performed on 10 cells in each category, excluding nuclear borders. The Kullback-Leibler—distance between texture coefficients for every pair of cells is reported. The diagonal KL distances, representing self-similarity, are set to 0 (dark blue). Cross comparison shows that each class of cells has high self-similarity (blue) and high cross-class dissimilarity (red). Fine grained early differentiated nuclei are most different from pluripotent and late differentiated cells (red, KL distance>20), while, pluripotent and late differentiated cells have a KL distance averaging 8.2, showing large texture differences between pluripotent and early differentiated cells. During development, higher order chromatin organization proceeds through accretion from uniform densities, to fine grained, low contrast variations to large grained, high contrast aggregates characteristic of heterochromatin and euchromatin. Bars in a, d; 10 μm.

[0052] FIG. 34 shows progressive histone H3K9 methylation of centromeres during differentiation without a net change in H3K9 acetylation Confocal sections of HSF-6 in pluripotent colonies (column a) and 5 day differentiated monolayers (column b) were immunostained for histone H3K9me3 (red, a-d), centromeres (CREST serum, green, a-d), H3K9ac (green, e-h) and widefield images of DNA (blue, Hoechst 33342, a-h). Most centromeric spots are surrounded by pericentromeric H3K9me3 in differentiated cells (b, d) but not pluripotent cells (a, c). In prometaphase cells (c, d), centromeres on some chromosomes show undetectable methylation (c) while in early-differentiated cells, the interstitial zone between centromeres is methylated in nearly all chromosomes (inset, c, d). Centromeres are associated with the nuclear periphery in some, but not all differentiated cells (b), but rarely in pluripotent cells (a). H3K9ac levels are more granular after differentiation but total levels are unchanged (e-h). Total nuclear H3K9me3 increased 4 fold after 2 days of differentiation while the area of CREST staining is constant (i). The fraction of centromeres colocalized with H3K9me3 increases from 44% to 80% after 2 days of differentiation (j). Methylation and acetylation of H3 and H4 of human hESC during differentiation (shown by Oct4 levels) were detected by western blot (k). The mean, SD and P value for 4 independent experiments are shown in l, m, demonstrating a significant increase in methylation in mouse and human ESC (l). Acetylation levels measured by western blot (k) are constant by intensity measures in immunostained cells (e-h) and are reduced in mouse but not human ESC (m). Bar in d, 5 p.m.

[0053] FIG. 35 demonstrates that formation of heterochromatin during hESC differentiation includes methylation of histone H3K9 and DNA at cytosine and increasing condensation Methylated histone and DNA are increasingly colocalized during differentiation. Maximum Projection, contrast stretched images (a, b) and linear confocal sections (c-h) of hESCs line HSF-6 are shown with staining for methylated H3K9me3 (red) and 5meC (green). At the edge of a differentiating colony (a), 5meC (green) becomes detectable only outside the colony and is particularly elevated at the periphery of each nucleus. By 10 days after differentiation, methylated DNA and histone coincide in spots (white, c-h), although DNA methylation particularly in the nuclear periphery is not limited to areas of H3K9me3 staining (b). Interphase (c, d, e) and mitotic (f, g, h) cells are shown with H3K9me3 (red), 5meC (green) and colocalization above threshold (white). In interphase nuclei, pluripotent cells contain H3K9me3 while 5meC appears in cells at the colony edge, first appearing at the nuclear periphery (d) and finally colocalizing with H3K9me3 (e). In mitotic cells, centromeric H3K9me3 appears without significant 5meC (f). At colony edges, 5meC preferentially appears at distal chromosome arms (g), while overlap with H3K9 m3 is restricted to proximal regions of chromosomes. Methylated histone and DNA colocalize along entire chromosome arms in differentiated cells (h). Progressive correlation of methylated histone and DNA in confocal slices of the colonies shown in panel A, B is shown in cytofluorograms (i) that show the pixel-by-pixel distribution of intensity of DNA and histone methylation. The pluri cells show no colocalization (left side, A) with very little DNA methylation (K). The red lines (i) show the median of this distribution. The early differentiated cells (right side, a) show a broad distribution about the median value showing partial colocalization, while the late differentiated cells (b) show a tight distribution about the median showing high correlation with a stoichiometry of 1:1 (i). The Pearson's coefficient (white bar, j) shows increasing correlation during differentiation (mean, SD and P value from T-test are shown). Van Steensel's cross correlation is the Pearson's Coefficient after a 20-pixel shift of one of the images, measures the spatial sharpness of the co-distribution. The Van Steensel's coefficient (black bar, j) increases during differentiation, showing less diffuse, more particulate distribution after differentiation. Bar in b and h, 10 μm.

[0054] FIG. 36 shows expression of epigenetic enzymes in pluripotent hESCs. In pluripotent hESC, the epigenetic enzymes, DNMT1, 3a &, 3b have elevated protein levels but low activity on DNA and histone substrates until differentiation is initiated In hESC, line H7, DNMTs 1, 3a, 3b (red nuclei, a-c, respectively) are present in pluripotent hESC (left side of each panel) expressing Oct4 (green, a-c), but absent in differentiated hESC (right side of each panel, Hoechst, blue nuclei). Pluripotent nuclei positive in the red, green and blue channel appear mauve. DNMTs 3b and 3a follow Oct 4 intensity among pluripotent cells (constant mauve hue) while DNMT1 appears to vary in pluripotent cells independently of Oct4 (varying hues from green to red). The DNMT levels after differentiation are clearly above background for DNMT1 (a, e) but are nearly undetectable for DNMT3a (b, f) and DNMT3b (c, g). Despite the presence of DNMTs in pluripotent cells, the product of DNMTs (5meC) was present at very low levels and only increased in early-differentiated cells at the periphery of colonies (d). Quantitation of 5meC in HSF6 cells show progressive increases in global DNA methylation from pluri, early to late diff stages, h). Bar in c, 10 μm. Bar in d, 100 μm.

[0055] FIG. 37 shows that HDAC 1 and 2 colocalization and deacetylases activity increases but protein levels do not change during neuronal differentiation The ratio of HDAC1, HDAC2 and DNA varies between different nuclear domains as illustrated by polychrome pluripotent cell nuclei (a, b, Oct4 positive, not shown) but is uniform in differentiated cells (a, c). Cytofluorograms show that HDAC 1 and 2 distributions are uncorrelated in pluripotent cells and become more tightly correlated in early-differentiated cells (f). A selective substrate for HDAC1 (H2BK5ac) remains acetylated in pluripotent cells and become deacetylated in differentiated cells (e, g). Western blots (h) were probed for HDAC1 and reprobed for HDAC 2 in mouse and human ESC and Ponceau S staining of pluripotent (lane 1 and 2) and differentiated (lane 3 and 4) ESC. Normalized mean and SD of 3 samples of HDAC1 (i) and 2 (j) are not statistically different between pluripotent and differentiated cells in mouse and human ESC. Bar in a & e, 10 μm.

[0056] FIG. 38 demonstrates how HP1β increasingly colocalizes with H3K9me3, binds mitotic chromosomes, and is less extractable after lysis of differentiated cells Pluripotent (a, d, c) hESC, line H7, show diffuse HP1β (a, red), while early differentiated cells show increasingly focal distribution of HP1β (b) and HP1α (c, red). H3K9me3 (d-f, green) distribution is both diffuse and focal, and colocalization with HP1β occurs with diffuse H3K9me3 in pluripotent and early differentiated cells (g, h, yellow and white) and with H3K9me3 focal spots only in early differentiated cells (b, white). HP1α is less colocalized than HP1β with H3K9me3 positive chromocenters (c, i, reduced yellow and white). Cytofluorograms of intensity distributions show that HP1β has partial colocalization with H3K9me3 in pluripotent cells with a low stoichiometry (median line, red, j) a partial colocalization with a near 1:1 stoichiometry (median line, red, k), and HP1α has an intensity distribution that is mostly due to noise (low intensity near the origin, l). The Pearson's and Van Steensel's correlation coefficients (m) show that H3K9me3 is selectively colocalized with HP1β and the distribution becomes sharper in early-differentiated cells, and that HP1α has a lower correlation with H3K9me3. In mitotic cells, HP1β has a cytoplasmic distribution and is absent from mitotic chromosomes in pluripotent cells (n) while in early-differentiated cells, HP1β is bound to chromosomes and is absent from the cytoplasm (ob). Pre-fix lysis in 0.5% Tx-100 reveals complete extractability of HP1β from pluripotent hESC (pc) but not early differentiated cells (q). Western Blot levels in hESC line H7 show that levels of HP1α are unchanged in human but not mouse ESC during differentiation (r, t). HP1γ is expressed at low levels in hESC (s). Bar in i (for a-i, p, q) and o (for n, o), 5 μm.

[0057] FIG. 39 shows results obtained from adaptive windowing of iPSC images. Each square in the table compares two colonies from the sample of six colonies. (a)-(f) corresponds to colonies #1-6, which are iPSC colonies; (g)-(h) corresponds to colonies #7-8, which is a differentiated colony (top and bottom halves); (i) #9=small fragment of iPSC colony #4). (j) shows a coefficient mask for iPSC colony #9. (k) illustrates the KLD for iPSC and differentiated colonies. The more similar the colony pair, the smaller the divergence. Comparing colony #9 with any of colonies #1 thru #6 shows that colony #9 is closer to them, than it is to colonies #7 and #8.

[0058] FIG. 40 is a flow chart of a method for identifying induced pluripotent stem cells, in accordance with an illustrative implementation of the present disclosure.