DSF2 (YBR007C / P38213) — curation notes

Journal-style notes for the AI GO review of Saccharomyces cerevisiae DSF2. Provenance is
recorded inline as [PMID:xxxx "verbatim supporting text"] or [SGD]/[UniProt P38213].

Identity

• Standard name: DSF2 = "Deletion Suppressor of mptFive" (MPT5 = PUF5). The
  UniProt alt name is "Deletion suppressor of MPT5 mutation protein 2" [UniProt P38213].
• Systematic name: YBR007C (chromosome II). UniProt AC P38213; SGD S000000211.
• 736 aa, ~82.3 kDa, PE 1 (protein-level evidence: detected by mass spectrometry).
• Verified ORF in SGD, but SGD lists molecular function = Unknown and biological
  process = Unknown [SGD]. This is a genuinely understudied ("dark") gene.

CRITICAL attribution: DSF2 vs DSF1 are NOT homologs

The names DSF1 and DSF2 come from the same phenotypic screen, not from sequence homology.
Both dsf1 (YEL070W) and dsf2 (YBR007C) were recovered as deletion suppressors that rescue
the temperature-sensitivity of an mpt5Δ (puf5Δ) strain
PMID:16328373.

• DSF1 (YEL070W) is a mannitol dehydrogenase (alias MAN1), an oxidoreductase enzyme
  with a paralog MAN2 [SGD].
• DSF2 (YBR007C) is a Sel1-like / TPR-repeat protein with no known catalytic activity.

So DSF1 and DSF2 are functionally and structurally unrelated; the shared "Dsf" prefix is a
phenotype label. DSF2-specific evidence must not be conflated with DSF1 enzymology. DSF2's
closest characterized family member is instead the fission-yeast Nif1 (see Orthology).

Domain architecture (inline analysis of DSF2-uniprot.txt)

Reading DSF2-uniprot.txt (P38213):

• N-terminal ~1–461: extensively disordered / low-complexity. UniProt/MobiDB-lite marks
  disordered REGIONs at 1–46, 178–208, 229–410, 440–461, with many polar / low-complexity /
  basic-and-acidic COMPBIAS stretches [UniProt P38213]. This half of the protein is a large
  intrinsically-disordered region with no recognizable folded domain.
• C-terminal ~560–736: Sel1-like TPR/α-solenoid. UniProt records SMART SM00671 SEL1
  (Sel1-like repeat), InterPro IPR006597 (Sel1-like), IPR011990 (TPR-like helical domain
  superfamily), Gene3D 1.25.40.10 (TPR domain), and SUPFAM SSF81901 (HCP-like)
  [UniProt P38213]. Sel1/TPR repeats form α-helical solenoids that are generic
  protein–protein interaction / scaffolding modules — they do NOT specify a catalytic
  activity or a particular partner.
• Phosphoprotein. UniProt cites two large-scale phosphoproteomics/Cdk1-substrate studies
  (PMID:18407956, PMID:19779198) as evidence at protein level. DSF2 is not named in the cached
  text of either (the hits are in supplementary tables), so specific phosphosites are not
  extracted here; the safe statement is that DSF2 is a detected phosphoprotein.

Interpretation: a disordered N-terminal region plus a C-terminal Sel1/TPR solenoid is the
architecture of a scaffold / adaptor, but the architecture alone does not tell us which
complex or which partners, and there is no experimental interactor that pins down a molecular
function. Domain content is therefore consistent with, but not proof of, an adaptor role.

Orthology / family context (PANTHER PTHR43628)

• UniProt assigns DSF2 to PANTHER PTHR43628 (family), subfamily PTHR43628:SF11 "PROTEIN
  DSF2" [UniProt P38213].
• The reviewed members of PTHR43628 (interpro/panther/PTHR43628/PTHR43628-entries.csv) are
  DSF2 (yeast), S. pombe Nif1 (P87159, "Mitosis inhibitor nif1", SPBC23G7.04c), and several
  bacterial/viral Sel1-repeat proteins (E. coli YjcO/YbeT, mimivirus L18). The family is united
  by Sel1-like repeats, i.e. it is a structural family, not a family with a single shared
  biochemistry.
• The PANTHER family name "Mitotic Progression Regulator" and its description are
  LLM-generated and unreviewed (llm: true, checked: false in the metadata YAML) — treat
  as a hypothesis, not evidence.
• S. pombe Nif1 negatively regulates mitotic entry by binding and inhibiting the Nim1/Cdr1
  kinase (Nim1 normally inhibits Wee1, so Nif1 keeps Wee1 active → Cdc2/Cdk1 stays inhibited)
  [Wu & Russell 1997 EMBO J 16:1342, PMID:9135149 (Nif1 novel mitotic inhibitor); via WebSearch].
  This is the biological source of the DSF2 IBA annotations (GOA with field =
  PomBase:SPBC23G7.04c).

Caveat on the IBA propagation: the inference DSF2 → "negative regulation of G2/M transition"
+ "cell division site" is a phylogenetic (IBA) transfer from a distant fungal ortholog that
shares only Sel1/TPR repeats. There is no S. cerevisiae experimental evidence that DSF2
acts on Swe1 (the budding-yeast Wee1), Cdc28/Cdk1, or a Nim1-family kinase. The IBA is a
reasonable placeholder but is not independently corroborated for DSF2.

What is experimentally KNOWN about DSF2 (budding yeast)

1. Genetic: dsf2 deletion suppresses (rescues) the temperature-sensitivity of mpt5Δ/puf5Δ,
   and partially suppresses its hydroxyurea sensitivity
   PMID:16328373;
   [UniProt P38213 DISRUPTION PHENOTYPE "Rescues temperature-sensitivity of MPT5 deletion. Partially suppresses the hydroxyurea (HU) sensitivity of MPT5 deletion."].
   Note this is a suppressor genetic interaction, i.e. loss of DSF2 compensates for loss of
   the Puf5 RNA-binding protein; it does not define DSF2's own molecular function.
2. Localization: Dsf2-GFP localizes to the bud tip / bud neck in unperturbed cells and
   relocalizes to the cytoplasm upon DNA-replication stress (HU/MMS) in a genome-wide GFP
   screen [SGD; PMID:22842922]. This is the basis of the HDA "cellular bud tip" (GO:0005934)
   annotation. DSF2 is not discussed individually in the PMID:22842922 text (it is one of the
   254 relocalizing proteins scored in the supplementary tables), so the localization is a
   high-throughput imaging call, not a dedicated study of DSF2.
3. Expression: low-abundance protein (~377 molecules/cell log phase per UniProt
   [UniProt P38213 MISCELLANEOUS]; ~1,600/cell per SGD).
4. Post-translational: detected as a phosphoprotein in Cdk1-substrate / phosphoproteome
   screens (PMID:19779198, PMID:18407956) — DSF2 in the supplementary hit lists.
5. Deletion phenotypes (SGD systematic data): increased replicative lifespan; decreased
   competitive fitness; sensitivity to the calcium chelator BAPTA [SGD]. These are pleiotropic
   fitness readouts, not a mechanistic function.

What is NOT known (candidate knowledge gaps)

• Molecular function is unknown. No catalytic activity is expected (Sel1/TPR = binding
  module). No specific binding partner has been shown to define an adaptor/scaffold role.
  → MF_DARK.
• Biological process is unknown. The only BP annotation is the distant-ortholog IBA
  (negative regulation of G2/M). There is no direct evidence DSF2 regulates the budding-yeast
  cell cycle, and the suppressor and lifespan phenotypes could reflect an unrelated process.
• Mechanism of mpt5Δ suppression is unknown — how loss of a bud-tip Sel1/TPR protein
  rescues loss of the Puf5 RNA-binding protein is not established.
• Relationship to DNA-replication-stress relocalization is unexplained — why a bud-tip
  protein moves to the cytoplasm under HU/MMS, and whether this matters functionally, is unknown.
• Redundancy: DSF2 has NO close sequence paralog in S. cerevisiae (it is NOT a paralog of
  DSF1); the only same-family protein is the distant fission-yeast Nif1. So genetic redundancy
  is unlikely to explain the mild phenotypes; the weak phenotypes more likely reflect a
  conditional / accessory role.

GO annotation review plan

Five GOA annotations:

Rationale for keeping the IBAs rather than removing: per CLAUDE.md we do not REMOVE on
paralog/ortholog grounds alone; the Nif1 IBA is a legitimate phylogenetic inference and the
bud-tip/bud-neck localization is loosely consistent with a cell-division-site association. But
neither is corroborated by DSF2 experiment, so they are marked non-core and the honest position
(MF/BP unknown) is retained in core_functions/knowledge_gaps.

Falcon deep-research (DSF2-deep-research-falcon.md) — corroboration and cautions

The falcon report (2026-07-05, 16 citations) corroborates the core picture: Sel1-like/TPR
scaffold with no catalytic domain, bud-neck localization, relocalization to the cytoplasm under
DNA-replication stress, mpt5/puf5 deletion-suppressor origin, and Nif1 (S. pombe) as the
same-family mitotic inhibitor. It adds two inferences that I deliberately did NOT promote into
the review's supported claims because they are not verified from primary full text here:

• A proposed link to the Cbk1 / RAM signaling network (Dsf2 has Cbk1 phospho-consensus
  motifs but LACKS the canonical Cbk1 docking motif — a "non-docking putative substrate", per
  Gögl et al. 2015). This is an inference from motif presence, and even the report states the
  in-vivo phosphorylation evidence for Dsf2 specifically "remains indirect". Treated as a lead.
• Specific phosphosites (S391/S395) attributed to a 2025 GSK3 phosphoproteomics preprint.
  Unverified against cached full text; not cited in the review.

The falcon report's citations use internal keys (e.g. gogl2015thestructureof), not PMIDs, so
per project policy they are NOT used as supported_by in the review; the review is grounded only
on verified PMIDs (16328373, 22842922), UniProt P38213, and the PANTHER family file. The falcon
file is retained in the gene folder as the deep-research artifact.

core_functions position

Given MF and BP are genuinely unknown, core_functions will be minimal: a single entry
capturing the best-supported, domain-grounded statement — that DSF2 is a Sel1-like/TPR-repeat
protein consistent with a protein-interaction/scaffold module localizing to the bud tip/neck —
using the MF root (GO:0003674) since no specific MF term is defensible, and locations
(GO:0005934). The real deliverable is knowledge_gaps.