CNNM2 is a membrane-associated magnesium-handling protein with separable evidence for magnesium export and TRPM7 regulation. This review preserves all 13 original annotations, three alternative-product objects and six original reference identities. Eleven original assertions are accepted and the two synaptic locations are retained as non-core. One nonredundant molecular-function annotation, ion channel regulator activity, is proposed by ortholog inference; no biological process is added.
The original mouse study cloned a 693-residue ACDP2 sequence from distal convoluted tubule cells and measured magnesium-dependent currents and reversal shifts after expression in Xenopus oocytes, with water and poly(A) controls. This historical construct is not equated with every current human product. Authentic original Methods and Results were recovered as author-uploaded article text, separately from the unchanged normal abstract (PMID:15899945(https://pubmed.ncbi.nlm.nih.gov/15899945/)). The complete 2010 abstract reports human splice-variant 1 complementation and magnesium uptake in deficient Salmonella, while splice-variant 2 and empty-vector controls were negative; its full body remains unavailable (PMID:20519162(https://pubmed.ncbi.nlm.nih.gov/20519162/)).
A substantive conflicting result must be retained. The 2016 study expresses human NM_017649.4 and NM_199076.2 in HEK293 cells and detects no expression-dependent magnesium currents, influx or efflux under its conditions. TRPM7 and SLC41A1 controls establish that the assays can detect transport. Its authentic institutional full PDF was inspected without modifying the normal abstract (PMID:27068403(https://pubmed.ncbi.nlm.nih.gov/27068403/)). This context-dependent negative result is not dismissed, but neither is it a universal disproof of the positive CNNM2-specific export results.
Mouse CNNM2 expression drives magnesium efflux after extracellular magnesium withdrawal, and CBS deletion or T568I disrupts that response. Later mouse CNNM2 experiments show magnesium lowering even in TRPM7-knockout cells, and ARL15 inhibits efflux in a binding-dependent manner (PMID:24706765(https://pubmed.ncbi.nlm.nih.gov/24706765/), PMID:34928937(https://pubmed.ncbi.nlm.nih.gov/34928937/), PMID:37449820(https://pubmed.ncbi.nlm.nih.gov/37449820/)). These observations support retaining the existing broad magnesium-transporter inference. They do not settle autonomous human permeation, a sodium-coupling stoichiometry or activity of every natural product. Transport is not retained merely because CNNM2 regulates TRPM7.
The original patient study shows renal magnesium wasting with segregating CNNM2 variants and human basolateral renal staining. Its electrophysiology uses corrected full-length mouse AK032006.1 and measures magnesium-sensitive sodium currents, with no observed magnesium current. HEK293 host identity does not make this a human-protein assay (PMID:21397062(https://pubmed.ncbi.nlm.nih.gov/21397062/)). Human NM_017649.5 expression and 25Mg-uptake measurements independently support cellular magnesium handling and variant effects; uptake alone is not isolated-protein translocation (PMID:38519529(https://pubmed.ncbi.nlm.nih.gov/38519529/)). Zebrafish loss/rescue and brain phenotypes support biological relevance without establishing an extra developmental process performed by CNNM2 (PMID:24699222(https://pubmed.ncbi.nlm.nih.gov/24699222/)).
Purified human CNNM2 residues 429–584 bind adenine nucleotides in NMR and crystallographic experiments, with magnesium favoring ATP binding and changing the CBS-domain conformation. ATPase activity remains speculation, and no ATPase annotation is proposed (PMID:25184538(https://pubmed.ncbi.nlm.nih.gov/25184538/)). The mouse CBS-domain binding study supplies an independent direct donor premise, not a human-domain experiment (PMID:24706765(https://pubmed.ncbi.nlm.nih.gov/24706765/)). ATP binding is integrated into the magnesium-handling narrative instead of being presented as a separate physiological core.
The proposed ion channel regulator activity is supported by physical association plus functional modulation. Wild-type mouse CNNM2 stimulates TRPM7-dependent zinc entry; loss of TRPM7, an inactive TRPM7 pore or an inhibitor blocks the response (PMID:34928937(https://pubmed.ncbi.nlm.nih.gov/34928937/)). The 2025 study maps mouse CNNM2 contacts with TRPM7 and uses export-disabled S269P or CBS-deletion constructs to separate channel regulation from magnesium lowering. The CNBH domain and the ARL15-binding site affect channel modulation, but the exact gating mechanism remains unresolved. The modest isolated-domain stimulation of TRPM7 kinase with a generic MBP substrate is not promoted to a separate physiological kinase-activator function (PMID:40962059(https://pubmed.ncbi.nlm.nih.gov/40962059/)).
Two cores retain these distinct activities: magnesium transport/homeostasis and TRPM7 regulation. The latter is an ISS inference from authenticated mouse Cnnm2/Q3TWN3, not human IDA. Human ARL15–CNNM2 fragment structures and binding experiments corroborate the regulatory interface; mouse full-length efflux assays remain separate (PMID:37449820(https://pubmed.ncbi.nlm.nih.gov/37449820/)). Current GO definitions and typed ancestry show that GO:0099106 adds coverage beyond the two seeded molecular functions. No suitable magnesium-specific regulator child is available among its current children. The target has no match in the selected GO-CAM index; this is not used as evidence of a missing process.
Official UniProt alternative sequences reconstruct the three source products exactly against separately retrieved FASTAs: Q9H8M5-1 is 875 residues, Q9H8M5-2 is 853 residues after deletion of canonical 721–742, and Q9H8M5-3 is 552 residues after replacement of canonical 541–552 and loss of 553–875. The short product interrupts the CBS region and lacks the distal CNBH domain. The original product labels and sequence-note fields are preserved. Engineered deletions, historical constructs and current natural products are not interchangeable.
Current official donor identities and complete GOA pages authenticate mouse Cnnm1/2/4, fly uex, worm cnnm-1/3 and zebrafish cnnm2a/b. Current PAINT-node and target records were inspected, but historical tree topology and IBD placement were not reconstructed. This is a source-resolution limit, distinct from a positive target-function judgment; target self-inclusion and donor number are not objections to an IBA. The InterPro homeostasis route is supported by actual IPR045095 membership and target functional evidence.
The synaptic source is specifically resolved. Current SynGO annotation 5563 records endogenous mouse Cnnm2 staining in cultured hippocampal neurons with antibody PA5-102013, partially overlapping SYP and PSD95. The same record supplies a glutamatergic-synapse extension. The official dataset has the exact CNNM2/Q3TWN3 row, independently joined to current mouse GOA. This supports non-core localization, without establishing pre- versus postsynaptic placement, a particular synaptic mechanism or direct endogenous human imaging. The record explicitly treats PMID:31171447 as an interim resource citation for a later experiment, so no target figure is attributed to the 2019 SynGO resource paper.
The compact source index links identifiers, URLs, hashes and selected reading ranges. Raw primary responses, complete literal donor records and full read ledgers remain in the temporary review packet. The index is an authored provenance aid, not independent experimental support, and its citations have no self-quoted supporting text. The small actual primary anchor is used once in the YAML.
Genuine isolated fetch-gene and publication caching completed. Default Falcon was attempted and timed out after 90 seconds; the configured Perplexity-lite fallback returned HTTP 401 quota failure. No provider output was fabricated. Immutable normal outputs, selected main-cache bytes and recovered external variants remain separate. Bounded independent transport and molecular-function consultations informed the review, with complete whole-review verification and ordinary validation/status/rendering required before publication.
The current-head review accepted the annotation actions and both core functions. Its mandatory request concerns access to PMID:15899945 and PMID:27068403. Both public URLs and capture hashes were already present in the published source-evidence file. The amended reference assessments now link directly to them and explain the distinction between external original-source access and the unchanged abstract-only normal caches.
ef3b67f1c48d73a8eb1112cb577f9eb22a834467f6e682ce66ba44d3d8e67481. Printed page 1225 describes human constructs; Table 3 is on page 1229, Figure 8/current controls on page 1233, and the efflux discussion on page 1234. The source-evidence entry preserves the separate extracted-text hash and read ranges. This bounded reread used text, not a new inspection of figure pixels.The two reference access flags remain false because original Methods/Results were actually accessible externally. The cached full_text_available: false values remain untouched. These are different statements, not conflicting versions of the same source.
The optional suggestions were assessed individually. Three short passages are now used once each in YAML: human CBS-domain ATP recognition, loss of efflux in the tested mouse T568I construct, and the authors' stated gating-versus-functional-channel-number uncertainty. Two biological knowledge gaps make the existing mechanistic questions machine-readable without adding annotation coverage. The description is biological prose, and three reasons state their evidence and limits more directly. The SynGO record remains a positive curated observation with its own public locator; quoting our own index as independent evidence would add no support. All original sources, actions, products, terms, findings and access flags are preserved.