CERKL has 16 original annotations and eight alternative products. This review preserves every original source object, including the two PPM1A partners, electronic donors and PAINT nodes. One new molecular-function annotation records direct mRNA binding by the sequence-defined human CERKLa product. No new biological-process or NOT assertion is introduced.
The campaign's immutable curation instructions permit supported generic binding to remain non-core where no stronger MF is justified. Here both original PPM1A associations admit a partner-class refinement; neither is interpreted as CERKL phosphatase activity or activation. The ancestry/comparator restriction for new process terms is not extended into a categorical ban on new molecular functions.
PMID:24498393(https://doi.org/10.1371/journal.pone.0087898), Figure 4C-D and its protein-purification/RNA-EMSA Methods, directly tests bacterial His-MBP-tagged human CERKLa. Purified protein shifts mRNA from human retina or COS-7 cells with protein-negative and unlabeled-competition controls. The N-terminal 1–256 fragment retains binding; the 252–532 fragment does not. The paper's NM_201548.4 translation, NP_963842.1, exactly matches the 532-aa current Q49MI3-2 sequence after the documented VSP_016659 deletion from displayed Q49MI3-1. This is tested-product specificity, not proof that other products lack the function. Short exact source anchors are present in the NEW annotation and RNA-binding core.
The same paper's cap-affinity experiment uses a human-cell immunoprecipitate, so it does not establish purified direct cap recognition. Its microtubule and translation-factor associations also include complex/RNA-mediated interactions. No direct microtubule-binding, translation-factor, motor or RNA-transport activity is added. The cellular mRNP evidence supplies context for mRNA binding without proving that RNA binding causes protection from apoptosis.
PMID:23501591(https://doi.org/10.1016/j.exer.2013.03.003), Sections 2.6–2.8 and Figure 8, reports sphingomyelin, glucosylceramide and galactosylceramide association using transfected HEK293T protein lysates and immobilized lipids. Retinoic-acid signal also occurred with empty vector and was discarded. This is not a purified affinity or transport assay. Earlier PMID:19158957(https://pmc.ncbi.nlm.nih.gov/articles/PMC2628313/), Results following Figure 3, reported negative overlays with both purified GST-CERKLa and fresh HEK293T lysate. The retained sphingolipid-binding annotation is therefore bounded to the positive assay context, not a universal behavior or explanation of all retinal lipid phenotypes. The 2013 mouse model retained about 35% expression and should not be described as a complete null.
The kinase name/domain alone does not establish phosphorylation. Full selected Methods and Results of PMID:19158957 test four cloned human retinal a/b/c/d products, both cell lysates and purified GST formats, with functioning CERK controls, standard and photoreceptor-derived lipid preparations, and cellular labeling. PMID:23501591 extends the negative tests to very-long-chain ceramides and mixed-lipid phosphorus assays. PMID:22678504(https://pmc.ncbi.nlm.nih.gov/articles/PMC3394741/) provides additional C6-NBD-ceramide/CERK-control assay text, although its exact full-length construct mapping remains limited.
These are target-specific reasons to reject positive lipid-kinase and kinase inferences. They do not prove inactivity of all eight products under every substrate/cofactor condition. No NOT annotation is manufactured. The current InterPro IPR001206 record really does map to GO:0016301: the issue is transfer to this target, not a nonexistent mapping or fabricated lost residue. Historical PAINT node PTN002285910 remains explicitly unresolved; donor count and target self-evidence are not used as objections.
PMID:19158957 Figure 7 measures PARP cleavage in human HeLa cells expressing human 532-aa CERKLa. Protection is seen with 300 micromolar hydrogen peroxide over 12–24 hours, with important negative boundaries at the higher dose, other stimuli and later time point. The COS-7 propidium-iodide experiment separately measures both necrotic and apoptotic death. The existing negative-apoptosis BP is retained, without inventing intrinsic antioxidant or caspase-inhibitor molecular activity.
The authentic abstract of PMID:24735978(https://doi.org/10.1016/j.bbadis.2014.04.009) reports TRX2 association, redox-state maintenance and protection; its full Methods were not recovered, so biochemical species/product assignments and direct redox catalysis are not inferred. PMID:30205735(https://doi.org/10.1080/15548627.2018.1520548), Figures 5–9 and the cloning/RNAi Methods, supports CERKL-dependent SIRT1 stability and autophagy in human ARPE-19 cells through depletion/rescue and tagged co-IP. It does not show CERKL deacetylase or kinase activity.
The newer PMID:41288005(https://pubmed.ncbi.nlm.nih.gov/41288005/), Results 3.6–3.7 and the lipid/PI3P/LC3 Methods, instead finds lower PI3P and autophagy with CERKL expression in human pancreatic cancer lines, including bafilomycin controls. Its Discussion explicitly distinguishes the ARPE-19 result. No universal autophagy direction, new cancer process or direct lipid-catalytic activity is asserted. Product identity in these two autophagy papers remains incompletely resolved.
The mouse donor A2AQH1 has exact IDA segment-localization records from PMID:23142158(https://pubmed.ncbi.nlm.nih.gov/23142158/). Its Figure 3 and immunofluorescence Methods show mouse and rat inner/outer-segment labeling with omitted-primary controls. Those are ortholog observations, not direct human-segment experiments.
PMID:21151604(https://pubmed.ncbi.nlm.nih.gov/21151604/), Figures 4–6 and antibody Methods, directly shows endogenous human ARPE-19 nuclear, cytoplasmic and perinuclear CERKL, with antigen-pre-absorption controls. It also finds exclusion from fibrillarin-positive nucleoli in human ARPE-19 and mouse 661W cells. This differs from expressed-product nucleolar accumulation seen by two independent groups in PMID:19158957 and PMID:19501188. The nucleolus assertion is retained as a non-core expression-context observation; its physiological relevance remains a question. In contrast, the ER and Golgi electronic locations are over-annotations: the 2009 imaging/fractionation is overexpressed and non-exclusive, and the later PMID:24498393 Figure S1 reports no colocalization with those organelle markers. No uniform localization of all eight products is claimed.
CERKL-source-evidence.json contains literal source objects, exact interaction records, donor annotations, source URLs/query receipts and SHA-256 hashes. It is a curated source extract, not a custom sequence-analysis pipeline or a claim that every full supplement was inspected.
For PMID:33961781(https://doi.org/10.1016/j.cell.2021.04.011), three exact IntAct records link Q49MI3 to P35813 in human HEK293T/HCT116 AP-MS experiments. For PMID:40205054(https://doi.org/10.1038/s41586-025-08878-3), the publication-linked author AP-MS network contains CERKL–PPM1A edge 609146. UniProt independently establishes P35813 as PPM1A, a protein Ser/Thr phosphatase. These observations support protein phosphatase binding at physical-association resolution; they do not establish a purified binary interface, activation, substrate or individual CERKL product. Database records and GOA share source provenance; repeated records are not automatically biological replication.
The genuine normal seed and source caches were preserved. Existing main caches were selected exactly; missing references were obtained by the unchanged normal fetch command in isolation. Falcon was attempted with a 90-second bound and timed out; the explicit Perplexity-lite fallback returned HTTP 401 for insufficient quota. No provider report was fabricated. Manual primary research continued. The normal abstracts of PMID:15708351, PMID:19501188 and PMID:24735978 remain explicitly abstract-only; selected complete primary papers carry the actual load-bearing claims.
Independent annotation-review and core-synthesis consultations assessed catalytic transfer and positive mRNA binding separately. GO definitions were checked through the official service. The authenticated GO-CAM index had no CERKL/Q49MI3 entry. This absence was not used to manufacture a new process. All newly authored source excerpts, including repeats, are counted per publication; notes introduce no additional verbatim literature excerpts.
This dated entry supersedes only the ER/Golgi over-annotation interpretation in the preceding localization paragraph. The entire notes file at head 478d1237f73e3e33e5c103bd2a4d8584bbb5b30c is retained as an exact prefix. That head incorporated an intervening localization change after 447d901981a2015ff392eeeda0e27e09ecbd9fb2; its independent nucleolar citation and apoptosis dose/time clarification remain valid. The current source objects and all eight product records remain unchanged.
The full PMID:19158957(https://pmc.ncbi.nlm.nih.gov/articles/PMC2628313/) localization Results and Figure 4E explicitly report calnexin and GM130 overlap for expressed human retinal products in COS-7 cells. The subsequent negative comparison concerns endosomal and mitochondrial markers. Figure 5 and its sucrose-gradient Methods separately place human HeLa CERKLa-HA in ER/Golgi marker fractions, with stronger trans-Golgi/Golgi overlap. These positive observations support conditional association with both compartments; distribution over several compartments is not evidence against either individual location. They do not demonstrate endogenous retinal residence, exclusive compartment identity or membrane topology. The later PMID:24498393(https://doi.org/10.1371/journal.pone.0087898) Results/Figure S1 locate aggregates partly around the ER and find no corresponding ER/Golgi marker colocalization. Accordingly both source locations are retained as non-core with these differing expression contexts explicit, not as universal steady-state locations.
The independent PMID:19501188(https://pubmed.ncbi.nlm.nih.gov/19501188/) abstract corroborates expressed CERKL nucleolar accumulation. Only the authentic abstract was recovered, so its host, construct and product mapping remain unassigned. Together with the positive tagged-product assay in PMID:19158957, this contrasts with endogenous ARPE-19 and 661W nucleolar exclusion in PMID:21151604(https://pubmed.ncbi.nlm.nih.gov/21151604/). The two positive reports are retained without conflating overexpression and endogenous distributions.
The PPM1A refinements remain at AP-MS complex-association resolution: two of the three BioPlex records are spoke-expanded, and the U2OS network is another AP-MS context. No separate low-throughput orthogonal binary assay was verified. The original machine records remain literal in CERKL-source-evidence.json; no phosphatase activation, substrate assignment or intrinsic CERKL catalysis is inferred.
For sphingolipid binding, PMID:23501591(https://pubmed.ncbi.nlm.nih.gov/23501591/) Figure 8 actually uses human CERKLa-containing HEK293T lysate, despite the retained mouse-donor IEA. Positive sphingomyelin and monohexosylceramides are distinguished from negative ceramide, lactosylceramide, GM1 and free fatty acids; retinoic-acid signal was also present with empty-vector lysate. A curator question records the human-source opportunity without fabricating a source annotation or a purified binary interaction. The survival core preserves the intervening clarification of 300 micromolar hydrogen peroxide over 12–24 hours; the annotation reason retains the stronger-dose, later-time and other-stimulus limitations. No additional core activity or process annotation is proposed.
Row: GO:0005783 IEA (GO_REF:0000044, SL-0095), KEEP_AS_NON_CORE (unchanged).
Checks:
- UniProt location for the displayed isoform is "Cytoplasm. Nucleus, nucleolus."; ER appears only in "[Isoform 2]: Cytoplasm. Nucleus, nucleolus. Golgi apparatus, trans-Golgi network. Endoplasmic reticulum." with no evidence tag [file:human/CERKL/CERKL-uniprot.txt].
- PMID:19158957 (full text) tested several isoforms: "neither a unique localization pattern nor a restricted subcellular localization could be ascribed to any CERKL isoform"; Figure 4E calnexin overlap in transfected cells ("mainly ER and Golgi"); HA-tagged CERKLa in PDI-containing sucrose fractions that also carry Golgi markers. So the isoform-2 restriction is UniProt's, and no source_defects entry was recorded.
- No signal peptide, TM or retention motif; all ER data are on overexpressed tagged constructs; PMID:24498393 sees aggregates partly around the ER without marker colocalization.
Outcome: KEEP_AS_NON_CORE kept (precedent AHCY, ER batch 1), with a dated ORGANELLES sentence; close call with MARK_AS_OVER_ANNOTATED stated in the reason. No sub-compartment (cytosolic-face pool).