EZH2 source and evidence guide

This guide distinguishes target experiments, ortholog evidence and database propagation for all 186 imported annotation contexts. The five source-described products remain separate; cell species, expression host and historical construct names do not establish equivalent natural-product functions. The campaign curation instructions govern supported generic binding and reference-only support without new publication quotations.

The annotation checkpoint contains 81 ACCEPT, 69 KEEP_AS_NON_CORE, 21 MODIFY and 15 UNDECIDED decisions. These are evidence assessments, not a claim that every biochemical or product-specific question is resolved.

Functional experiments and source access

This section covers the 67 functional source contexts in 30 intact PMID groups. Binding and database evidence are addressed in the following sections. Human proteins, variant preparations, heterologous hosts, native cell assays and ortholog evidence remain distinguished. No NEW annotation or quotation is added.

Two correction notices are included as genuine normal caches: PMID:38383797(https://pubmed.ncbi.nlm.nih.gov/38383797/) and PMID:38347304(https://pubmed.ncbi.nlm.nih.gov/38347304/). Their full publisher notices were inspected separately because the normal caches are abstract-only. The corrections are limited to the specified control panels, not whole-paper retractions.

Broad protein-lysine methyltransferase annotations retain H1.4 and JARID2 substrate chemistry. H3K27-specific sources retain their narrower scope. Histone reading by EED/PHF proteins, O-GlcNAc transfer by OGT and RBBP4-dependent association are not recast as EZH2 catalytic or direct binary functions.

Promoter-specific chromatin binding means association with promoter chromatin; it does not require intrinsic sequence-specific DNA recognition. Chromatin architecture, transcription, downstream signaling and cell phenotypes are assessed separately. The circadian and lymphoma papers include actual human EZH2 experiments despite mouse-focused abstracts.

The reference full_text_unavailable flags describe the selected normal caches. A separately inspected original can support a VERIFIED manual assessment while its normal cache remains abstract-only.

Database and conserved-function evidence

EZH2 is a catalytic component of PRC2. The experiments discussed here distinguish its contacts with chromatin and RNA from the targeting contributions of accessory proteins, and distinguish enzymatic chemistry from the downstream developmental contexts in which that chemistry acts.

Human substrate contacts and chemistry

PMID:33479123(https://pubmed.ncbi.nlm.nih.gov/33479123/) was read in its operative structural Results and expression, purification, nucleosome and activity Methods. The preparation contains full-length human EZH2, engineered human EED/SUZ12 and JARID2 components, and additional PRC2 partners expressed in High Five insect cells. EZH2 SET/CXC and bridge-helix regions contact nucleosomal DNA and the histone tail. These are direct target contacts, not evidence that EZH2 independently recognizes a promoter nucleotide motif. The study uses Xenopus histone preparations and separate commercial human modified nucleosomes; the cellular origin of the enzyme and substrate should not be conflated. Its SAH assay integrates methyl-transfer events rather than independently resolving each methylation state.

Broader substrate terms remain meaningful. In PMID:19144645(https://pubmed.ncbi.nlm.nih.gov/19144645/), anti-EZH2 immunoprecipitates from a HeLa nuclear-pellet fraction methylate H1.4, with WT/K26A substrate comparisons. The paper also studies G9a, but the EZH2 experiment is separately positive. The preparation is not isolated EZH2. PMID:25620564(https://pubmed.ncbi.nlm.nih.gov/25620564/) demonstrates PRC2 methylation of nonhistone JARID2 K116 and cellular dependence on Ezh2/Eed. Its peptide-binding and structural experiments identify EED as the methyllysine reader; that binding role is not reassigned to EZH2. Mammalian and fly PRC2 preparations and mouse ES-cell experiments are distinguished; all construct details delegated to supplements were not independently inspected.

The official EC 2.1.1.356 reaction and the target UniProt RHEA:60292 mapping support H3K27 trimethylation. The current ARBA00098061 endpoint returned 404, and its unavailable condition sets are not reconstructed. A source-specific H3K27 refinement does not erase independent H1.4 or JARID2 evidence.

Chromatin occupancy is not automatically sequence recognition

The mouse donor experiments in PMID:24137001(https://pubmed.ncbi.nlm.nih.gov/24137001/), 23942234 and 20956546 contain positive Ezh2 chromatin occupancy. However, the inspected synthetic RNA–DNA hybrid assay in the first does not contain EZH2 protein; motif experiments in the second concern YY1; and motif/EMSA experiments in the third concern STAT6. Figure 2D in PMID:20798045(https://pubmed.ncbi.nlm.nih.gov/20798045/) measures H3K27me3 ChIP. None of those observations alone demonstrates EZH2 sequence discrimination.

PMID:18483221(https://pubmed.ncbi.nlm.nih.gov/18483221/) includes mouse ES-cell chromatin measurements and separate human 293T association/artificial Gal4 recruitment experiments. PMID:20064375(https://pubmed.ncbi.nlm.nih.gov/20064375/) includes mouse ES-cell work and a separate positive human ES-cell ChIP-qPCR experiment. Its direct JARID2 interaction maps to SUZ12, not EZH2. Selected publisher Results from PMID:26658965(https://pubmed.ncbi.nlm.nih.gov/26658965/) support Linc-YY1-dependent changes in PRC2 occupancy in mouse C2C12 cells; the motif/EMSA experiments concern YY1. Full operative Methods, underlying supplementary datasets and figure pixels from that paper were not read.

GO:0000976 deliberately permits non-sequence-specific regulatory-region binding, and GO:1990841 describes promoter-associated chromatin binding. These are distinct from the sequence-specific terms. The positive evidence supports chromatin DNA binding where the original assertion implied EZH2 nucleotide-sequence recognition.

RNA ligands and PRC2 targeting

PMID:21172659(https://pubmed.ncbi.nlm.nih.gov/21172659/) explicitly tests human FLAG-EZH2 and human PRC2 purified from Sf9 cells, including heparin and size-exclusion steps, with RNA EMSA. Cellular mouse RIP and UV-crosslink experiments are separate. The direct ligands include RepA, Hes1 antisense and Gtl2-derived RNAs, supporting the more specific lncRNA-binding activity. The purified ligand is RNA, not an assembled RNA–protein complex; no lncRNA sequence exclusivity is inferred.

In PMID:27618650(https://pubmed.ncbi.nlm.nih.gov/27618650/), the EMSA uses a mouse Ezh2 insert in a human in-vitro translation system. The host system does not make the product human. Rat/mouse Chaer experiments, reported human iPSC-cardiomyocyte RIP, and purified human RNA binding from the independent paper are kept separate. The human supplementary RIP figure was not viewed. PMID:21874018(https://pubmed.ncbi.nlm.nih.gov/21874018/) uses mouse ES-cell formaldehyde RIP for lincRNA associations; this is not an isolated-protein binding assay for every recovered RNA. The pri-miR-208b result in PMID:24137001 is a particular mouse cardiac RNA-ChIP context, retained as non-core.

Conserved repression and developmental settings

PMID:19026781(https://pubmed.ncbi.nlm.nih.gov/19026781/) distinguishes EZH2-associated histone methylation from stronger EZH1 SAM-independent compaction. The latter is not transferred to EZH2. PMID:16717091(https://pubmed.ncbi.nlm.nih.gov/16717091/) contains positive human EZH2 constructs: enhancement of CRY-dependent repression of mouse promoter reporters in HeLa and rescue of mouse-specific siRNA effects in NIH3T3 cells. Mouse liver ChIP/co-IP and mouse fibroblast oscillations remain distinct from those human-construct results. No native human circadian time course is claimed.

PMID:23531880(https://pubmed.ncbi.nlm.nih.gov/23531880/) perturbs YY1 and rescues with WT versus deltaREPO; it does not knock out Ezh2. Independent actual PMID:12496962(https://pubmed.ncbi.nlm.nih.gov/12496962/) deletes the mouse Ezh2 SET region and demonstrates a cell-autonomous pro-B to pre-B defect with H3K27 changes. EZH2 is not thereby assigned the recombinase step. The inspected author original for PMID:12900441(https://pubmed.ncbi.nlm.nih.gov/12900441/) provides mouse pronuclear Ezh2 localization and maternal-depletion controls. PMID:27892467(https://pubmed.ncbi.nlm.nih.gov/27892467/) shows conditional pericentric recruitment in Dnmt1-deficient mouse ES cells and separate engineered MBD–Ezh2 targeting. These do not establish constitutive native human pericentric occupancy.

Operative originals remain unavailable for the specific rat contexts in PMIDs 22821416, 23932971, 25088689, 27690696 and 18467665. Registry/abstract access is explicitly distinguished from actual experiment reading. Current rat GOA assigns IEP to hippocampus development, estradiol response and kidney regulation; other contexts use IMP. Expression can be a valid response readout, but it does not automatically establish participation in a developmental step. The unusual title of PMID:27690696 is not grounds to reject its curator-supported GABAergic or inflammatory assertions.

PAINT, rules and localization

Actual PANTHER v19 PTHR45747 tree and MSA place human Q15910 at PTN002505151; its ungapped 746-residue sequence exactly matches the displayed target. Positive IBDs at PTN000768863 support chromatin binding, nucleus, heterochromatin formation, negative transcription regulation and H3K27 methyltransferase activity. PTN005243577 supports ESC/E(Z) complex membership. Both nodes are ancestral to the target; no target-path NOT assertion was found. Human self-WITH entries are legitimate experimental descendant support. Refining nucleus to nucleoplasm or heterochromatin formation to the facultative process does not dispute those placements.

Current ARBA00085892 matches FunFam 2.170.270.10:FF:000001 plus Primates; ARBA00089504 matches that FunFam plus Mammalia. Current target InterPro records include IPR045318 and the IPR021654/IPR044439 domains. Their mappings and actual human assays are assessed separately. GO:0006338 includes histone-state changes and is not restricted to ATP-driven sliding. An incidental H3K37 spelling in InterPro prose is not propagated as new EZH2 chemistry.

The Human Protein Atlas EZH2 record reports nucleoplasmic localization. Selected supported CAB009589 U-251MG and HPA029131 SiHa composites were actually viewed; neither selected composite has a visible blue overlay. Green central/nuclear-looking signal is interpreted together with official antibody/cell annotations, without quantitative localization claims. The complete HPA029131 antigen matches all five source products; the CAB009589 antigen sequence is not supplied. The images therefore do not identify a particular natural product.

Reactome target identity and performing roles

All imported Reactome groups have independently verified representative paths to native human EZH2 entity 212298: reference entity 54556 identifies Q15910, coordinates are modeled 1–746, and the entity’s own compartment is nucleoplasm. The shared PRC2 complex is 212285. EZH1-containing alternatives are not substituted for the explicit EZH2 route.

R-HSA-212263, R-HSA-5638332 and R-HSA-9764932 explicitly identify EZH2 as an active unit for histone chemistry. R-HSA-3240783 instead produces EZH2 protein from a regulated expression event; this supports a modeled location without assigning EZH2 the act of its own transcription. In DNMT-associated pathways, the DNA methyltransferases perform DNA cytosine methylation. The location and neighboring enzyme functions are not conflated. Representative target routes are verified; unrelated branches with retained access failures are not claimed to form a complete graph survey.

The five UniProt products remain distinct. Exact antigen sharing, a displayed-sequence PAINT match and a modeled full-length Reactome participant do not establish equivalent activity or localization for all products. No new figure-pixel or supplementary-file reading is implied beyond the two explicitly viewed HPA images.

Interaction-source evidence

This source-specific assessment covers 53 protein-interaction contexts from 31 publications. Cellular co-IP/AP-MS associations, two-hybrid tests, purified or translated-protein assays, proximity measurements and computational network context are distinguished. Supported generic interactions are retained as non-core under the campaign instruction; seven partner-specific molecular-function refinements use current ontology definitions. No new publication quotations or annotation rows are added.

The five source-declared human products (746, 751, 707, 737 and 695 residues) remain distinct. A human host or a database isoform label does not by itself establish a historical insert sequence or functional equivalence of all products. Negative tests, mutant effects and disease-fusion partners retain their experimental context.

An unavailable original does not overturn an experimental curator. Where the same human pair is independently established by inspected primary experiments, the reason names that independent source and preserves the original-access limitation. The ATRX context remains unresolved because the human EZH2/ATRX reagent provenance is not established; the listed corrections are unread, without an asserted scientific impact.

PMID: 11571280 — contexts 20

Polycomblike PHD fingers mediate conserved interaction with enhancer of zeste protein. (PubMed record).

Official metadata/abstract and curated target interaction; bounded original routes failed.

The abstract explicitly reports human PHF1–EZH2 association alongside Drosophila PCL/E(Z) experiments. Original construct/Methods boundaries remain unread. Independent human PHF1-complex evidence was actually read in PMID23142980 and the BioPlex target tables; no fly assay is silently presented as a human assay.

Public read scope: The publication abstract and the curated IntAct EZH2–PHF1 interaction record; original Methods remain unread. Public source: https://pubmed.ncbi.nlm.nih.gov/11571280/; https://www.ebi.ac.uk/intact/search?query=pubmed%3A11571280. No figure-pixel read by this lane is claimed for this source.

PMID: 16357870 — contexts 22, 23, 24

The Polycomb group protein EZH2 directly controls DNA methylation. (PubMed record).

Publisher abstract, original supplementary Methods and actual 2007 corrigendum; main remains unavailable.

The corrigendum restores omitted lanes in Figures 2a, 2c, 4a and Supplementary Figure 1 and states that conclusions are unaffected. Corrected Figure 2a tests GST-EZH2 fragments against in-vitro-translated DNMT1/3A/3B; Figure 2c shows endogenous HeLa EZH2/EED IP recovery of DNMTs. GST-EZH2 1–746 and 1–340 interact, whereas the tested C-terminal fragment does not. Supplementary Methods distinguish bacterial GST fusion, reticulocyte translation and HeLa nuclear co-IP. This is binding evidence, not DNA-methyltransferase chemistry performed by EZH2; coordinates are the paper constructs.

Actual figure pixels inspected: corrected Figure 2 panels, corrigendum page 2.

Public read scope: Corrected Figures 2a and 2c and S1; original Supplementary Methods: GST pull-down and HeLa nuclear-extract immunoprecipitation. Public source: https://doi.org/10.1038/nature05710. Corrected Figure 2 panels were inspected as pixels by this lane.

PMID: 17560333 — contexts 25, 26

Role of the polycomb repressive complex 2 in acute promyelocytic leukemia. (PubMed record).

Actual author-hosted original PDF.

Human NB4/U937-PR9 experiments establish pathological PML-RARA/PRC2 association. Transfected 293T native RARA shows weaker association, while endogenous native RARA is not recovered in the tested U937/HL60 contexts. SUZ12 contributes to the fusion association. Q15156 is the PML-RARA entry in this source context, not a native PML interaction inferred by gene-name shorthand. No purified binary affinity or universal natural-product equivalence is established.

Public read scope: Figure 1D and accompanying Results, printed pp. 514–515; Experimental Procedures: Plasmids and Antibodies, Cell Lines and Immunoprecipitation, printed p. 523. Public source: https://doi.org/10.1016/j.ccr.2007.04.009. No figure-pixel read by this lane is claimed for this source.

PMID: 20543829 — contexts 30

Role for the MOV10 RNA helicase in polycomb-mediated repression of the INK4a tumor suppressor. (PubMed record).

Actual original main body in immutable normal record.

Human 293T Flag-PRC2 co-IP controls show EZH2–SUZ12 association in Figure 6a. The paper distinguishes this positive control from negative MOV10 association; the latter does not invalidate the source pair. This is cellular assembly evidence, not purified binary affinity or a mapped test of all EZH2 products.

Public read scope: Figure 6a and accompanying PRC2-control Results; original co-immunoprecipitation Methods. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC2929459/. No figure-pixel read by this lane is claimed for this source.

PMID: 22009739 — contexts 31

Corepressor protein CDYL functions as a molecular bridge between polycomb repressor complex 2 and repressive chromatin mark trimethylated histone lysine 27. (PubMed record).

Actual original main body in immutable normal record.

Endogenous human MCF7 CDYL–EZH2 co-IP and GST-CDYL/in-vitro-translated EZH2 mapping establish the interaction. The source maps paper-coordinate CDYL 61–309 and EZH2 329–522 regions and separately uses human ORFs in Sf9. CDYL is the chromatin reader/bridging corepressor; its reader or adaptor function is not assigned to EZH2.

Public read scope: Figures 2A–B and 3; Results on CDYL interaction with PRC2; Experimental Procedures: GST Pull-down and Baculovirus Expression. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC3234934/. No figure-pixel read by this lane is claimed for this source.

PMID: 22094255 — contexts 32, 33, 34

Oxidative damage targets complexes containing DNA methyltransferases, SIRT1, and polycomb members to promoter CpG Islands. (PubMed record).

Official BioC full main; normal cache is truncated despite its full-text flag.

Human NCCIT DNMT3B IP recovers EZH2/EED2/SUZ12 before and after oxidative stress; human HCT116 hypomorphic FLAG-DNMT1 experiments supply a separate DNMT1 context. EED2 describes the partner, not an EZH2 isoform. Co-IP/fractionation and stress-dependent redistribution do not measure purified binary affinity. Complete supplementary Methods and figure pixels were not inspected.

Public read scope: Results describing NCCIT DNMT3B immunoprecipitation and HCT116 FLAG-DNMT1 experiments; original immunoprecipitation and fractionation Methods. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC3220885/. No figure-pixel read by this lane is claimed for this source.

PMID: 22439931 — contexts 35, 36

Deconstruction of the SS18-SSX fusion oncoprotein complex: insights into disease etiology and therapeutics. (PubMed record).

Official BioC main plus actual supplementary PDF.

Reciprocal EED, EZH2 and SUZ12 IPs from human SYO-1 cells recover the other PRC2 components in Supplementary Figure S6D, with IgG controls. Keep these cellular associations separate from mouse models and SS18-SSX disease-fusion biology.

Actual figure pixels inspected: Supplementary Figure S6D.

Public read scope: Supplementary Figure S6D, accompanying PRC2 Results and immunoprecipitation Methods. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC3734954/. Supplementary Figure S6D was inspected as pixels by this lane.

PMID: 22897849 — contexts 37, 38

ASXL1 mutations promote myeloid transformation through loss of PRC2-mediated gene repression. (PubMed record).

Actual original main body in immutable normal record.

Human FLAG-ASXL1/EZH2 assays and endogenous SET2/UKE1 co-IP support ASXL1 association, including nuclease treatment controls. EED is present in the curated source-pair record; the independent human reciprocal PRC2 IP in PMID22439931 positively grounds that pair without claiming an unread panel was inspected. Neither ASXL1 transcriptional role nor all-product activity is inferred solely from association.

Public read scope: Results and Methods describing FLAG-ASXL1 and endogenous SET2/UKE1 immunoprecipitation, including nuclease treatment; the EED source-pair record is retained separately from the independently read reciprocal PRC2 experiment in PMID:22439931. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC3422511/. No figure-pixel read by this lane is claimed for this source.

PMID: 23142980 — contexts 39

Molecular basis for H3K36me3 recognition by the Tudor domain of PHF1. (PubMed record).

Actual original main body and Online Methods in immutable normal record.

Human PHF1-containing PRC2 is purified from HEK293T cells; PHF1 Tudor mutants W41A/Y47A retain complex association. The H3K36me3-reader function belongs to PHF1. Mouse ESC and yeast-nucleosome experiments are separate from this human complex evidence.

Public read scope: Results on PHF1 Tudor mutants and PRC2 association; Online Methods for purification of human PHF1-containing PRC2. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC3603146/. No figure-pixel read by this lane is claimed for this source.

PMID: 23238014 — contexts 40

CRL4B catalyzes H2AK119 monoubiquitination and coordinates with PRC2 to promote tumorigenesis. (PubMed record).

Official abstract/metadata and curated ChIP-associated source context; original full route failed.

The imported EZH2–SUZ12 context is retained using independent human reciprocal/co-IP evidence from PMID20543829 and PMID22439931. The original CRL4B paper was not fully inspected, so no claim is made about its exact protein interaction versus chromatin co-occupancy experiment, nor is CRL4B ubiquitin-transfer chemistry assigned to EZH2.

Public read scope: The publication abstract and curated IntAct source-pair record; complete original Methods remain unread. Public source: https://pubmed.ncbi.nlm.nih.gov/23238014/; https://www.ebi.ac.uk/intact/search?query=pubmed%3A23238014. No figure-pixel read by this lane is claimed for this source.

PMID: 23455924 — contexts 41

A Y2H-seq approach defines the human protein methyltransferase interactome. (PubMed record).

Original supplementary PDF and XLS tables recovered; main remains publisher abstract/paywall.

Y2H-seq reports EZH2–SUV39H1 with retest fraction 0.5. The orthogonal luminescence co-IP table reports Firefly-SUV39H1 and Protein A-EZH2 as positive (Z-score 14.5). The PDF defines the columns and positive/negative coding. This supports binding to a histone methyltransferase, not transfer of SUV39H1 H3K9 chemistry to EZH2. The complete main Methods/host details remain unavailable; no figure pixels were inspected.

Operative original tables: Supplementary Table S3 row 123 (Y2H) and Table S5 row 133 (positive luminescence co-IP); the supplemental PDF defines these fields.

Public read scope: Supplementary Figure 1; Supplementary Tables 3 and 5 (separate XLS), EZH2/SUV39H1 row and reciprocal luminescence co-IP record; PDF definitions pp. 15–17. Public source: https://doi.org/10.1038/nmeth.2397. No figure-pixel read by this lane is claimed for this source.

PMID: 23684459 — contexts 42, 43

Phosphorylation of EZH2 activates STAT3 signaling via STAT3 methylation and promotes tumorigenicity of glioblastoma stem-like cells. (PubMed record).

Official BioC original main.

Human patient-derived glioblastoma stem-like cells show EZH2–STAT3 co-IP; association is weaker in differentiated cells and tested established lines. STAT3 IP also recovers PRC2 components. STAT3 K180 and EZH2 S21/catalytic-mutant experiments establish context dependence. The cellular methylation experiments are not a purified turnover assay, and no uniform natural-product equivalence is established.

Public read scope: Figure 1A–D and Results on EZH2–STAT3 interaction in glioblastoma stem-like cells; Experimental Procedures: human GBM specimens and immunoprecipitation. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC4109796/. No figure-pixel read by this lane is claimed for this source.

PMID: 24074865 — contexts 44, 45, 46

Dnmt3L antagonizes DNA methylation at bivalent promoters and favors DNA methylation at gene bodies in ESCs. (PubMed record).

University repository author-proof PDF including supplementary Methods.

The mouse Dnmt3L construct is explicitly Addgene 13356. Mouse ESC endogenous IPs, FLAG-Dnmt3L in human HEK293, and recombinant GST-EZH2/FLAG-Dnmt3L tests are distinguished. The paper names GST-EZH2 reagent H00002146-P01; the historical lot sequence is not verified. HEK293 experiments show endogenous DNMT3A/3B co-IP reduced by increasing mouse Dnmt3L. The partner Q9CWR8 is correctly mouse; human host does not change its species. This is an author proof, not an assertion of final-version byte identity.

Public read scope: Author-proof Results on Dnmt3L–PRC2 association and Supplementary Methods describing the mouse Dnmt3L construct, immunoprecipitation and recombinant GST-EZH2 assays. Public source: https://doi.org/10.1016/j.cell.2013.08.056. No figure-pixel read by this lane is claimed for this source.

PMID: 24183680 — contexts 47

Reduced H3K27me3 and DNA hypomethylation are major drivers of gene expression in K27M mutant pediatric high-grade gliomas. (PubMed record).

Official abstract/metadata; bounded main PDF route failed.

The precise K27M glioma experiment remains unread. Retention of the EZH2–SUZ12 association rests on independently inspected human PRC2 co-IP evidence (PMID20543829/22439931), without inferring a universal change in association from a histone mutant or the title.

Public read scope: The publication abstract; the complete original and precise K27M interaction experiment remain unread. Public source: https://pubmed.ncbi.nlm.nih.gov/24183680/. No figure-pixel read by this lane is claimed for this source.

PMID: 24981860 — contexts 49, 50, 51

Human-chromatin-related protein interactions identify a demethylase complex required for chromosome segregation. (PubMed record).

Official abstract/metadata; publisher route not recovered as full main.

The abstract describes tagged human chromatin-protein AP-MS. Exact target rows and complete Methods remain unread. Independent human PHF1, EED and SUZ12 evidence from PMID23142980,22439931 and33961781 supports retaining the corresponding generic associations. The demethylase discovery elsewhere in the study is not an EZH2 catalytic claim.

Public read scope: The publication abstract describing tagged human chromatin-protein AP-MS; exact target rows and complete Methods remain unread. Public source: https://pubmed.ncbi.nlm.nih.gov/24981860/. No figure-pixel read by this lane is claimed for this source.

PMID: 25416956 — contexts 52

A proteome-scale map of the human interactome network. (PubMed record).

Official BioC main plus genuine supplementary Methods DOC and Table S2 XLSX.

Human ORFeome Y2H screen with pairwise retesting, autoactivation controls and sequence identification reports EZH2/CEP63. Table S2G records one source screen; S2C provides MAPPIT/PCA/wNAPPA scores, without a new claim that every orthogonal assay is positive. The tested clones have not been independently aligned to all current EZH2 products. Earlier reused XML was empty and reused XLSX was not a workbook; those failed inputs remain archived and are not evidence.

Operative original tables: S2G row 1667 and S2C row 189; full supplementary Y2H Methods describe retesting and controls.

Public read scope: Original main and Supplementary Methods on Y2H retesting, controls and sequence identification; Table S2G EZH2–CEP63 row 1667 and Table S2C row 189 with orthogonal-assay scores. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC4266588/. No figure-pixel read by this lane is claimed for this source.

PMID: 25417162 — contexts 53

ATRX directs binding of PRC2 to Xist RNA and Polycomb targets. (PubMed record).

Actual original full main in immutable normal record.

Mouse Xist biology is separate from purified ATRX/PRC2 binding and sequential ATRX then EZH2 IP of the RNA-containing assembly. The Methods describe FLAG/HA full-length ATRX produced in Sf9 and refer PRC2 preparation to prior work; the human EZH2/ATRX reagent provenance is not resolved by these passages. Cellular nuclear IP from human 293F is a macroH2A experiment, not automatically this exact pair. Preserve the source assertion without claiming a verified human binary assay or all-product specificity. Official metadata additionally identifies a correction DOI 10.1016/j.cell.2014.11.010; its operative text and the second cache-listed correction were not recovered in this bounded pass, so no claim about their scientific impact is made.

Public read scope: Figure 4A–B and corresponding Results; Experimental Procedures: Protein Expression and Purification; In Vitro Protein-RNA Interaction Assays; Immunoprecipitation and Mass Spectrometry. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC4379047/. No figure-pixel read by this lane is claimed for this source. VERIFIED concerns intended paper identity and the actual positive assembly experiments; exact human recombinant attribution remains unresolved for the annotation.

PMID: 26496610 — contexts 54, 55, 56

A human interactome in three quantitative dimensions organized by stoichiometries and abundances. (PubMed record).

Author thesis containing a 2014 under-review manuscript; published exact pair records in IntAct.

The author manuscript describes BAC-GFP AP-MS in HeLa Kyoto and explicitly accommodates mouse BAC sequences in its search database. Thus a human host alone does not establish every bait species. The final 2015 paper and target BAC metadata are not fully inspected. Independent human PHF1/EED/SUZ12 assays and BioPlex tables support the generic pairs without treating the earlier manuscript as the final published version.

Public read scope: The 2014 under-review manuscript within the author thesis, BAC-GFP AP-MS Methods and species-inclusive database search; separately curated final-paper IntAct EZH2–PHF1/EED/SUZ12 records. The final 2015 main and target BAC metadata remain incompletely inspected. Public source: https://pubmed.ncbi.nlm.nih.gov/26496610/; https://www.ebi.ac.uk/intact/search?query=pubmed%3A26496610. No figure-pixel read by this lane is claimed for this source.

PMID: 27705803 — contexts 57, 58, 59

A High-Density Map for Navigating the Human Polycomb Complexome. (PubMed record).

Actual July 7, 2016 author preprint recovered by the independent source helper; final journal PDF remains inaccessible.

The author preprint describes human ORFeome or HEK293-cDNA N-terminal Strep–HA baits, human Flp-In HEK293 AP-MS, tandem purification with a low-yield single-step exception, and the EZH2/EED/SUZ12 core with PCL-associated complexes. Some initial threshold-failing edges were rescued by network proximity. No exact target score table, historical clone sequence, purified binary binding or equivalence to the final published version is established. Independent human PRC2/PHF1 evidence corroborates the retained pairs.

Accessible original version: July 7, 2016 author preprint, PRC2 Results embedded lines 233–255 and Methods 427–485. An independent source helper also inspected Figure 2 and supplementary filtering Methods; this does not establish identity to the final published version.

Public read scope: July 7, 2016 author preprint, PRC2-complex Results and main purification Methods reread by this lane. The independent helper additionally read the AP-MS overview and Supplementary Methods for controls/scoring and viewed the title page and Figure 2. Final-paper IntAct EZH2–PHF1/EED/SUZ12 records were separately inspected. Public source: https://doi.org/10.1101/059964; https://www.ebi.ac.uk/intact/search?query=pubmed%3A27705803. No figure-pixel read by this lane is claimed for this source.

PMID: 27728805 — contexts 60

N-Myc Induces an EZH2-Mediated Transcriptional Program Driving Neuroendocrine Prostate Cancer. (PubMed record).

Actual original main and supplementary Methods PDF.

N-MYC IP in the human prostate-cell model recovers EZH2 and SUZ12, and catalytic inhibitors alter the assembly. The SET-deleted EZH2 construct is an engineered mutant. Co-recovery via N-MYC is not purified EZH2–SUZ12 binary binding; independent reciprocal human PRC2 IP supports this source pair. Mouse prostate tumor measurements are separate.

Public read scope: Results describing N-MYC-associated PRC2 in the human prostate-cell model and Supplementary Methods for constructs and immunoprecipitation. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC5540451/. No figure-pixel read by this lane is claimed for this source.

PMID: 27728808 — contexts 61

Enhancing the Cytotoxic Effects of PARP Inhibitors with DNA Demethylating Agents - A Potential Therapy for Cancer. (PubMed record).

Official BioC main and complete supplementary PDF.

Supplementary Figure S1A fractionates untreated/H2O2-treated human NCCIT nuclear extracts, then performs DNMT3B IP with IgG controls. The figure includes EZH2 recovery (pixel inspection recorded separately). The main PARP1/DNMT1 experiments are not silently substituted for the DNMT3B target pair. This is chromatin-associated complex evidence, not intrinsic DNA methylation by EZH2.

Actual figure pixels inspected: Supplementary Figure S1A.

Public read scope: Supplementary Figure S1A and its caption; original main and Supplementary Methods for nuclear fractionation and DNMT3B immunoprecipitation. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC5201166/. Supplementary Figure S1A was inspected as pixels by this lane.

PMID: 29909548 — contexts 65

Molecular heterogeneity and CXorf67 alterations in posterior fossa group A (PFA) ependymomas. (PubMed record).

Actual original main in immutable normal record.

Human Daoy/U2OS anti-EZHIP IP-MS and repeated Daoy anti-SUZ12/anti-EZH2 IPs establish a PRC2 assembly containing EZHIP; EPD210FH cells provide an additional cellular association. Row65 carries both EZHIP and SUZ12 and is not reduced to a fabricated binary stoichiometry. Detailed supplementary IP tables remain unread.

Public read scope: Results and Methods describing human Daoy/U2OS anti-EZHIP IP-MS, reciprocal PRC2 immunoprecipitation and EPD210FH cellular association. Detailed supplementary IP tables remain unread. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC6105278/. No figure-pixel read by this lane is claimed for this source.

PMID: 30510991 — contexts 67

Disruption of FOXP3-EZH2 Interaction Represents a Pathobiological Mechanism in Intestinal Inflammation. (PubMed record).

Actual original main in immutable normal record.

Fresh human Treg PLA and tagged human FOXP3/EZH2 co-IP/PLA in HEK293T provide positive association, distinct from mouse Treg assays. FOXP3 was cloned from human PBMC and EZH2 includes Origene RC202054; complete historical plasmid-to-current-product identity is not established. Specific FOXP3 mutations have differing effects. PLA is proximity, not purified binary affinity.

Public read scope: Figure 1E–F human Treg PLA; Figure 2C and surrounding tagged human-protein co-IP/PLA; Methods: PLA, co-IP, transfection and plasmid generation. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC6260395/. No figure-pixel read by this lane is claimed for this source.

PMID: 31515488 — contexts 69

Extensive disruption of protein interactions by genetic variants across the allele frequency spectrum in human populations. (PubMed record).

Actual original main and official supplementary workbooks reused byte-exact.

The Y2H design compares mutant proteins with their WT interaction. EZH2–CEP63 is disrupted by the source-labeled M439T/T1331C allele, while N640K and R288Q are scored non-disruptive. SD6 carries a PCA retest for the disrupted allele. Paper coordinates/accession labels are retained without reassigning them to all current products. The variant comparison does not negate the WT interaction.

Operative original tables: Supplementary Data 2 row 1501, Data 3 rows 113–114 and Data 6 row 103. These are source coordinates, not a remapping to current products.

Public read scope: Original Methods on mutant-versus-WT Y2H and orthogonal retesting; Supplementary Data 2 row 1501, Data 3 rows 113–114 and Data 6 row 103. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC6742646/. No figure-pixel read by this lane is claimed for this source.

PMID: 32051553 — contexts 70

The EGFR-ZNF263 signaling axis silences SIX3 in glioblastoma epigenetically. (PubMed record).

Actual original main in immutable normal record.

Human U251 ZNF263 IP/western detects EZH2 even though EZH2 was not in the initial mass-spectrometric hit list. SUZ12 is negative in the tested co-IP and is not inferred positive from chromatin occupancy. ZNF263 is the DNA-binding regulator at the SIX3 promoter; this association supports transcription-factor binding, not transfer of ZNF263 DNA recognition to EZH2.

Public read scope: Figure 3e and Results on ZNF263-associated proteins; Methods: Co-immunoprecipitation. Supplementary Figure S4D is described in the main but was not viewed here. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC7142014/. No figure-pixel read by this lane is claimed for this source.

PMID: 33961781 — contexts 71, 72, 73

Dual proteome-scale networks reveal cell-specific remodeling of the human interactome. (PubMed record).

Actual complete PMC main reused byte-exact; original BioPlex directed-edge/bait files reused byte-exact.

Human ORFeome FLAG-HA AP-MS identifies all three PHF1/EED/SUZ12 pairs in 293T and HCT116. Literal EZH2 Q15910-2 and EED O75530-2 labels are recorded; HCT116 also contains PHF1 bait O43189-2. Dataset labels do not independently verify current protein sequence or all five natural EZH2 products. AP-MS is cellular association, not proof that each pair contacts directly.

Target data: BioPlex downloads, directed 293T and HCT116 files. Both contain all three requested pairs.

Public read scope: Original affinity-purification/mass-spectrometry Methods and published directed-edge/bait datasets for 293T and HCT116, target EZH2–PHF1/EED/SUZ12 records. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC8165030/; https://bioplex.hms.harvard.edu/downloadInteractions.php. No figure-pixel read by this lane is claimed for this source.

PMID: 34591612 — contexts 74, 75, 76

A protein interaction landscape of breast cancer. (PubMed record).

Official BioC main, complete supplementary Methods PDF and author-linked NDEx network.

The union network contains EZH2 bait Q15910 and PHF1/EED/SUZ12 prey edges marked public_ppi=false, distinguishing measured AP-MS edges from imported public associations among other nodes. MCF7/MDA-MB-231/MCF10A were assayed with tagged human ORFs and biological triplicates. The network merges WT and mutant baits globally, so no unverified target clone or all-product identity is asserted. Its differential scores are not binding affinities.

Exact measured edges in the author-linked NDEx network: 1489, 1491 and 1496. Imported public associations among other nodes are separately marked.

Public read scope: Original AP-MS Results and Supplementary Methods for cell lines, constructs and replicate analysis; author-linked NDEx network edges 1489, 1491 and 1496, including public_ppi attributes. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC9040556/; https://doi.org/10.18119/N9BS4B. No figure-pixel read by this lane is claimed for this source.

PMID: 35016035 — contexts 77

Identification and functional characterization of transcriptional activators in human cells. (PubMed record).

Official abstract/metadata and data-availability statement; bounded preprint/full routes failed.

The exact EZH2–SUZ12 target proteomics result and full Methods were not recovered. The association is independently established in actual human PRC2 co-IP and BioPlex data. No claim that EZH2 is a transcriptional activator follows from the title or from its occurrence as a partner.

Public read scope: Publication abstract and data-availability statement; the exact target proteomics result and complete Methods remain unread. Public source: https://pubmed.ncbi.nlm.nih.gov/35016035/. No figure-pixel read by this lane is claimed for this source.

PMID: 35512704 — contexts 78

Systematic discovery of mutation-directed neo-protein-protein interactions in cancer. (PubMed record).

Official BioC main plus original Tables S3/S4 and Data S1.

Human HEK293T qHT-dSBRET/GST pulldown and MCF7 NanoPCA tests report an AKT1-E17K gain of association with EZH2. Table S3 distinguishes positive GST/NanoPCA from untested semi-endogenous/endogenous validation; Table S4 gives the target result. Data S1 Figureiii shows EZH2 recovery with WT and E17K AKT1. This is not an EZH2 mutant, purified affinity measurement or all-natural-product comparison.

Actual figure pixels inspected: Data S1 Figure iii (AKT1 WT/E17K), page 3.

Public read scope: Results on comparative WT/mutant interaction screening; STAR Methods: qHT-dS, NanoPCA and GST pull-down/co-IP; Data S1 Figure iii; Table S3B EZH2 target row; Table S4 AKT1-E17K MCF7 EZH2 target row. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC9597701/. Data S1 Figure iii (page 3) was inspected as pixels by this lane.

PMID: 37553330 — contexts 79

SNIP1 and PRC2 coordinate cell fates of neural progenitors during brain development. (PubMed record).

Actual original full main in immutable normal record.

Mouse neural progenitor nuclear IPs recover PRC2 components with SNIP1/JARID2 and distinguish RBBP5 negative controls. This is not a human cell experiment. Independent human EZH2–SUZ12 co-IP and BioPlex data establish the retained generic association without transferring the mouse developmental phenotype or all-product scope.

Public read scope: Results and Methods on mouse neural-progenitor nuclear immunoprecipitation and PRC2-associated controls. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC10409800/. No figure-pixel read by this lane is claimed for this source.

PMID: 40205054 — contexts 80, 81, 82

Multimodal cell maps as a foundation for structural and functional genomics. (PubMed record).

Actual original main and publisher correction, plus exact existing-archive NDEx extraction independently replayed by source helper.

Human U2OS C-terminal FLAG–HA ORFeome AP-MS supports all three pairs. Exact archived edges and literal node/attribute records are retained. EZH2 and PHF1 both have bait=true, but node inventory/edge ordering does not prove reciprocal individual IPs. PHF1 has three literal Ensembl identifiers, not a silently normalized unique product. Imaging, SEC and predicted structures are distinct modalities. The correction replaces a duplicated loss equation and does not state a revision of these interaction data. No target clone sequence or all-product equivalence is established.

Exact AP-MS network edges: PHF1 553406, EED 553410 and SUZ12 553416, with EZH2 node 553402; the original article links its public NDEx data.

Public read scope: Original AP-MS Methods and multimodal-interaction Results, plus publisher correction. Author-network target records identify EZH2 node 553402 and PHF1 553404/edge 553406, EED 553408/edge 553410, SUZ12 553414/edge 553416. The complete target projection was inspected; the separate helper independently replayed these three edges. Public source: https://pmc.ncbi.nlm.nih.gov/articles/PMC12137143/. No figure-pixel read by this lane is claimed for this source.

Term specificity

The authored refinements are nuclear receptor binding, transcription corepressor binding, histone methyltransferase binding, DNA-binding transcription factor binding and protein kinase B binding. Current definitions and typed ancestor closures place each under molecular function and protein binding. Binding to an enzyme or regulator does not transfer the partner’s catalytic, DNA-recognition or adaptor activity to EZH2.

No normal publication cache was edited. Cache full-text flags are not treated as proof of complete original access; recovered original versions, supplementary tables, corrections and remaining access gaps are recorded separately.

Cache availability and independent original access

The reference full_text_unavailable values describe the metadata of the selected normal publication caches. They do not erase verified external original-text, supplementary-table or figure access recorded above. In particular, a source can have a VERIFIED manual reference assessment while its selected cache contains only an abstract. Metadata marked as full text can also contain incomplete extraction; those scientific access limitations remain explicit in the individual assessments.

Five donor publication files (PMIDs 24137001, 20064375, 12900441, 27892467 and 21874018) use the exact already-committed cache bytes for delivery. The privately fetched versions differ only in per-line trailing whitespace and, in four records, publication-type metadata. Their identifiers, titles, full-text availability and scientific body text agree. Both versions are preserved in the evidence record; no publication cache is edited or synthesized.

Public paper and original table locators above replace private extraction filenames in reference prose. IntAct PMID links are public search locators, not a claim of a new retrieved query. Exact private source captures and read ledgers remain separately pinned.

Core-function synthesis

Two activities summarize the accepted evidence: H3K27 methylation by the catalytic EZH2 subunit of PRC2, and direct binding to tested long noncoding RNA ligands. The current definition of GO:0140951 includes successive methyl addition to form H3K27me3; it does not mean that EZH2 acts only on a dimethylated substrate. WT PRC2 acts on unmethylated, mono- and dimethylated H3K27 substrates with different efficiencies (PMID:22323599, Results accompanying Figure 3 and Materials and Methods). Human PRC2 reconstitution and native human-cell biochemical evidence establish the complex requirement (PMID:15385962, Figures 6–7 and Biochemical and enzymatic analyses). The direct molecular-function slot identifies EZH2's catalytic site within that complex; it does not assert an active isolated enzyme. A separate broad protein-lysine methyltransferase unit would duplicate an ancestor of this core activity. H1.4 methylation by human anti-EZH2 immunoprecipitates (PMID:19144645, Figure 1 and Experimental Procedures) and PRC2-mediated JARID2 K116 methylation (PMID:25620564, Figure 1) remain positive substrate-breadth evidence in the biological description and consistently retained protein-lysine methyltransferase reviews.

EZH2 contacts nucleosomal DNA and the H3 tail, making chromatin the location of the catalytic unit (PMID:33479123, nucleosome-engagement Results and Protein purification/Nucleosome reconstitution). The experiment uses full-length human EZH2 with engineered PRC2/accessory constructs; Xenopus histones and separate purchased human modified nucleosomes are distinguished. Total SAH production is cumulative methyltransferase output, while Western blots and reported mass spectrometry separately address H3K27 methylation states. EED reads methylated ligands; JARID2/AEBP2 and Polycomb-like proteins add distinct targeting functions. These are not transferred to EZH2. The two selected process facets, facultative heterochromatin formation and negative regulation of Pol II transcription, are not ancestors of one another in the captured ontology. EZH2 performs the methyl-transfer step; the process attribution is supported by chemistry together with human catalytic-mutant rescue, repression and chromatin-accessibility/FISH experiments (PMID:26581166, Results accompanying Figures 4–6 and the corresponding main Methods), rather than necessity alone. Detailed supplemental construct protocols and new figure-pixel inspection are not claimed.

The lncRNA-binding unit rests on purified human EZH2 and human PRC2 EMSAs (PMID:21172659, Figure 5 and protein-purification/RNA-EMSA Methods) and the independent isolated-loop experiment (PMID:35682829, Results 2.4 and Methods 4.1–4.3). Human recombinant binding is distinct from the mouse ES-cell RIP and locus-regulation experiments. The tested direct ligands are lncRNA-derived; the refinement specifies those ligands without asserting exclusive recognition of lncRNAs. The isolated loop binds the HOTAIR fragment weakly with low specificity; its phosphomimetic mutant is not a measurement of phosphorylation. Neither experiment establishes that every PRC2-associated transcript directly binds EZH2, that an RNA G-quadruplex is universally required, or that RNA binding always recruits, activates or inhibits PRC2. Accordingly, this unit has no separately asserted process term. Nucleoplasm is a supported broad compartment, not a claim that the in-vitro binding assay localizes the interaction. RNA binding and nucleosomal substrate engagement are retained without extra generic chromatin- or protein-binding core units.

The five source-described EZH2 products remain separate. Historical residue numbering, full-length assay labels and engineered fragments are not universal current-product assignments. The new questions address product-specific assembly/catalysis and RNA-binding consequences. All earlier source-access limits and cache-versus-original distinctions remain in force. The feedback clarifications below change authored judgments and add selected reference-only support and findings while preserving every imported source object, reference identity and product.

Mechanistic clarification after external review

Ras signaling and hydrolysis. PMID:20154697(https://pubmed.ncbi.nlm.nih.gov/20154697/), Figure 5 and the corresponding Results/Methods, tests human prostate epithelial cells, promoter-associated repression of DAB2IP, Ras-GTP and downstream signaling, and reversal by DAB2IP reconstitution. Suppression of a Ras GAP does not increase GTPase catalytic activity. GO:0046579 captures the positive signaling outcome; EZH2 is the upstream transcriptional regulator. The 2024 correction concerns a Figure 1g Ki-67 control micrograph, not these Figure 5 assays.

Chromatin and substrate breadth. PMID:26581166(https://pubmed.ncbi.nlm.nih.gov/26581166/), Figures 4–6 and chromatin-accessibility/FISH Methods, supports a regulated H3K27-dependent facultative process in human MCF7 cells. The four broad heterochromatin contexts are refined consistently without assigning H1.2 reader/structural activity to EZH2. The broad protein-lysine methyltransferase contexts are retained because the H1.4 and JARID2 experiments supply real additional substrate classes. Reactome R-HSA-212263 supplies narrower H3K27-trimethylation evidence for one of those broad assertions. R-HSA-5638332 instead describes mono-, di- and trimethylation and the MeK28-histone H3 output set; the H3K27-site refinement does not assert that this event is mono-only. The source's histone K28 numbering is distinguished from the conventional H3K27 name.

Catalytic granularity and disease. GO:0140951 explicitly describes three successive methyl transfers, so retaining it includes intermediate methylation steps. Adding its broader ancestor as another core would duplicate that activity. PMID:22323599(https://pubmed.ncbi.nlm.nih.gov/22323599/), Figures 3–4 and main Materials and Methods, directly compares wild-type and lymphoma-associated variant preferences. PMID:26694085(https://pubmed.ncbi.nlm.nih.gov/26694085/), Results and methyltransferase-assay Methods, tests Weaver-associated variants in reconstituted human PRC2. Several tested variants reduce activity; neither this experiment nor the separate EED/SUZ12 variant study establishes a uniform mechanism for every EZH2 allele.

Evidence limits. No new original figure-pixel inspection or inaccessible supplement recovery is claimed by this feedback pass. PMID:9214638 remains abstract-only. The lncRNA refinement rests on actual human biochemical assays, not on mouse RNA-ChIP or a generic claim that every PRC2-associated RNA is a ligand. Existing historical construct and natural-product boundaries remain unchanged.

Consistent breadth and localization

PMID:26694085(https://pubmed.ncbi.nlm.nih.gov/26694085/), Materials and Methods and Figure 2, tests human wild-type and variant PRC2 with mixed core histones, alongside separate H3-tail peptide assays. PMID:19144645(https://pubmed.ncbi.nlm.nih.gov/19144645/), Experimental Procedures and Figure 1E, independently measures H1.4-directed methylation in anti-EZH2 immunoprecipitates from HeLa nuclear pellets. These scopes support retaining broad histone methyltransferase activity without changing the specific H3K27 catalytic core or claiming isolated EZH2 activity on H1.4.

The seven nucleus assertions are consistently refined using the human HPA nucleoplasm annotation and native human Reactome participant 212298. Nuclear fractionation and promoter ChIP retain their actual resolutions; neither is relabeled as a new nucleoplasm imaging assay. Nucleoplasm is a finer nuclear compartment, not an assertion that EZH2 is absent from chromosomes or other supported pools. The existing antibody/product and structural-source access limits remain unchanged. No additional image pixels, original figures or supplementary experiments were inspected for this clarification.