FUN19 (P28003 / YAL034C) — Transcription Coactivation After Ada2/SAGA Divergence

Focus type: function_assignment
Hypothesis slug: transcription-coactivation-after-ada2-saga-divergence
Target: Saccharomyces cerevisiae FUN19 (UniProt P28003; systematic name YAL034C)

Executive Judgment

Verdict: REFUTED / OVER-ANNOTATED.

The seed hypothesis proposes that FUN19 has transcription coactivator activity (GO:0003713) and positively regulates DNA-templated transcription (GO:0045893), reasoning that although FUN19 lost the Ada2 ZZ/SANT domains and SAGA-complex membership, it retained SWIRM and therefore may keep "some activation mechanism." The evidence assembled across three investigative iterations does not support this. Every strand of evidence — domain architecture, existing GO annotation set, and both aggregated (STRING) and curated (IntAct) interaction data — points away from positive coactivation and, where it points anywhere, points toward a repressive (Rpd3L/Sin3 histone-deacetylase) context or toward chaperone/nuclear-import associations that have nothing to do with transcriptional activation.

The core logical flaw in the seed hypothesis is that it treats the retained SWIRM domain as a residual activation module. Primary structural work on the SAGA HAT module shows the opposite: SWIRM is a tethering/interaction module, not a catalytic or intrinsically coactivating element. The domains that actually confer Ada2's coactivator function — the ZZ-type zinc finger and the SANT domain, which stabilize Gcn5 and the histone-acetyltransferase reaction — are exactly the domains FUN19 has lost. Retaining the "docking" piece while losing the "business end" is not a conservative basis for inferring retained activation.

The most important caveats: (1) FUN19 is an essentially uncharacterized, non-essential gene with no experimental FUNCTION or SUBUNIT annotation in UniProt, so the case is built on comparative/interaction inference rather than a direct assay that refutes coactivation; (2) all four current GO annotations are IBA (phylogenetic inference, GO_Central) with no experimental backing, so the annotation baseline itself is weak in both directions; and (3) the STRING SIN3 signal, while the strongest experimental partner, is aggregated high-throughput co-occurrence rather than a validated binary complex membership. The recommendation is therefore to reject the coactivator / positive-regulation assignment and retain only the neutral IBA-level chromatin terms pending direct experimental evidence.

Key Findings

F001 — FUN19 retained only SWIRM and lost the Ada2 domains that confer coactivator function

FUN19 (UniProt P28003; YAL034C; 413 aa) carries a single annotated structured domain — SWIRM (residues 316–413; Pfam PF04433; PROSITE PS50934) — plus three disordered/low-complexity regions. It has no HAT domain, no bromodomain, no ZZ-type zinc finger, and no SANT domain. This is in direct architectural contrast to yeast ADA2 (UniProt Q02336; YDR448W; 434 aa), which carries a ZZ-type zinc finger (residues 2–58), a SANT domain (60–112), and a SWIRM domain (349–434). ADA2 is one of the ~19 subunits of the SAGA coactivator complex (TRA1, SPT7, TAF5, NGG1/ADA3, ADA2, GCN5, …). FUN19 is not in the SAGA subunit list. Both proteins fall within PANTHER family PTHR12374 (ADA2-related), but FUN19 sits in the distinct subfamily SF21 (FUN19-related), consistent with an ancient duplication/divergence event rather than orthology to functional Ada2.

The mechanistic significance of this domain loss is established by primary structural work on the SAGA HAT module (PMID: 36965704):

"A minimal HAT module complex containing Gcn5 bound to Ada2 and Ada3 is required for full Gcn5 activity on nucleosomes."

This establishes that Ada2's coactivator (histone-acetyltransferase) role depends on the assembled Ada2/Ada3/Gcn5 module. The domains Ada2 uses to build and stabilize that module — beyond SWIRM — are precisely the ZZ and SANT domains FUN19 lacks. The same paper further clarifies what SWIRM actually does:

"the SWIRM domain is required for incorporation of the HAT module into the yeast SAGA complex, but not the ADA complex."

SWIRM is therefore a tethering/interaction module that helps dock the HAT module into SAGA — it is not a catalytic or intrinsically activating element. Its presence in FUN19 does not by itself imply any coactivator activity. The seed hypothesis's premise ("retains SWIRM; those losses may change complex context without eliminating every activation mechanism") is thus undermined at the structural level: FUN19 kept the docking peg and discarded the machine.

F002 — Existing FUN19 GO annotations are neutral or repressive in context and are entirely IBA

All four current FUN19 (P28003) GO annotations are IBA:GO_Central — phylogenetic inference from GO Central, with no experimental support:

Critically, neither GO:0003713 (transcription coactivator activity, MF) nor GO:0045893 (positive regulation of DNA-templated transcription, BP) is currently annotated to FUN19. The seed hypothesis is proposing to add activation-directional terms that the existing curated set does not contain. Moreover, the one complex context that was propagated — Rpd3L-Expanded (GO:0070210) — is a histone-deacetylase (repressive) assembly, i.e., opposite in sign to "positive" coactivation. UniProt lists no experimental FUNCTION or SUBUNIT comment for FUN19 (only that it is a non-essential gene and a phosphoprotein). The existing annotation baseline therefore offers no support for adding activation directionality and, if anything, its only complex-level context leans repressive.

F003 — FUN19's strongest experimental interaction is SIN3 (co-repressor scaffold); SAGA links are database carry-over

STRING (S. cerevisiae, YAL034C) interaction partners were examined by evidence channel. The highest experimental-evidence (escore) partners are the co-repressor scaffold SIN3 (escore = 0.755), KAP123 (0.727), TAF1 (0.668), ADH1 (0.625), and TAF5 (0.608). By contrast, all recovered SAGA subunits — GCN5, ADA2, NGG1/ADA3, SPT7, SPT20, SPT3, SPT8, TRA1, SGF29, SGF73, SUS1, HFI1 — have escore = 0.000 and a uniform dscore = 0.540, meaning they derive from database/complex co-annotation, not from any experiment. In other words, the apparent "SAGA membership" that seeds the coactivator hypothesis is a database artifact: it is co-annotation transfer, not experimental co-purification. Among curated complex subunits, FUN19's only genuine experimental interaction signal maps to SIN3, the scaffold of the Rpd3L/Rpd3S co-repressor — again a repressive, not activating, context.

F004 — Curated IntAct interactions place FUN19 with chaperones and nuclear-import factors, not any coactivator complex

Curated IntAct/PSICQUIC experimental interactions for P28003 are dominated by molecular chaperones and folding machinery: Ssa1 (Hsp70), Hsc82 (Hsp90), Hsp104, Sse2, Gim4 (prefoldin), Aha1, and Csa1 — recovered via tandem affinity purification in the chaperone interaction network (PMID: 19536198) — plus KAP123 (karyopherin/nuclear import; TAP, PMID: 16429126) and a two-hybrid self-association (PMID: 14690591). There is no curated FUN19–Sin3 or FUN19–SAGA binary interaction in IntAct. This has two implications. First, it further confirms that FUN19 is not a validated member of any transcription coactivator complex. Second, it tempers the STRING SIN3 signal: the escore of 0.755 reflects aggregated high-throughput co-occurrence rather than an established stable complex, so even the repressive-context lead should be treated as suggestive, not proof of Rpd3L subunit status. The chaperone-heavy interactome is more consistent with a protein that is metastable/poorly folded on its own, or that is captured non-specifically by folding machinery, than with a stably assembled transcriptional coactivator.

Mechanistic Model / Interpretation

The seed hypothesis and the evidence can be summarized in a single comparative picture:

 ZZ finger   SANT      SWIRM     HAT module role
 --------    ------    ------    ----------------------------
  ADA2 (Q02336)    YES        YES       YES      Coactivator: stabilizes Gcn5,
   (SAGA subunit)                                 enables full HAT activity on
                                  nucleosomes; SWIRM docks HAT
                                  module into SAGA
  --------------------------------------------------------------------------
  FUN19 (P28003)   NO         NO        YES      Docking peg only; no catalytic
   (subfamily SF21)                              or activation domains retained

The logical chain the seed hypothesis relies on is: SWIRM retained → some activation mechanism survives Ada2/SAGA divergence → assign coactivator + positive regulation. Each link fails:

1. SWIRM is not an activation module. Structural work shows SWIRM tethers the HAT module into SAGA; it does not carry out or intrinsically drive acetylation/activation (PMID: 36965704). Retaining it does not retain activation.
2. The activating domains were lost. The ZZ and SANT domains — those that stabilize Gcn5 and enable full HAT activity — are absent from FUN19. Loss of the effector domains is the functionally decisive change, not the retained docking module.
3. The complex context that survived is repressive. The only propagated complex term is Rpd3L-Expanded (an HDAC assembly), and the only genuine experimental interaction among transcription factors is SIN3 (the Rpd3 co-repressor scaffold). If FUN19 has any residual chromatin role, the vector of evidence points repressive, not activating.
4. The "SAGA membership" is an artifact. STRING SAGA links have experimental score 0.000; they are co-annotation transfer. IntAct records no SAGA/Sin3 binary interaction at all.

The coherent interpretation is that FUN19 is a divergent, largely uncharacterized paralog within the ADA2 (PTHR12374) family that lost the coactivator effector apparatus. Its curated interactome (chaperones + a karyopherin) is that of a protein whose stable partners are folding/transport machinery, not a transcriptional activation module. Any transcription-related role is at best neutral chromatin association, and the strongest specific lead is a repressive Sin3/Rpd3 context — the opposite sign from the hypothesis.

GO Decision Table (lead requiring curator verification)

Evidence Matrix

Evidence Base (Literature)

PMID: 36965704 — The SAGA HAT module is tethered by its SWIRM domain and modulates activity of the SAGA DUB module. This is the pivotal mechanistic reference. It directly establishes (1) that full Gcn5/HAT activity on nucleosomes requires the assembled Gcn5–Ada2–Ada3 module, and (2) that the SWIRM domain is a tethering element that incorporates the HAT module into SAGA rather than a catalytic/activating element. Both quotes were verified against the abstract. Together they dismantle the seed hypothesis's central premise that retention of SWIRM implies retention of coactivation: the domain FUN19 kept is the docking module, while the domains it lost (ZZ, SANT) are those coupled to Gcn5 stabilization and HAT activity.

PMID: 19536198 (chaperone interaction network, TAP-MS) — source of the curated FUN19 interactions with Hsp70/Hsp90/Hsp104/prefoldin. Supports the interpretation that FUN19's stable curated partners are folding machinery, not a coactivator complex.

PMID: 16429126 (TAP) — FUN19–KAP123 (nuclear import) interaction. Consistent with nuclear localization but not with transcriptional activation.

PMID: 14690591 (Y2H) — FUN19 self-association; low functional weight but part of the curated interaction record.

Database-level orientation (STRING, IntAct/PSICQUIC, UniProt, PANTHER PTHR12374) is used as orientation, not as primary support, per the research directive — the STRING escore breakdown is treated as evidence of artifactual SAGA co-annotation rather than as positive interaction evidence.

Conflicts and Alternatives

1. Paralog over-annotation (most likely explanation). FUN19 sits in PANTHER PTHR12374 (ADA2 family) but in a distinct subfamily (SF21). IBA propagation from the ADA2-containing family, plus STRING/complex co-annotation, generates the appearance of SAGA/coactivator association. The escore=0.000 for every SAGA subunit is the signature of database carry-over, not biology.

2. Repressive alternative. If FUN19 has any real chromatin function, the evidence sign is repressive: the only propagated complex is Rpd3L-Expanded (HDAC), and the top experimental interactor is the Sin3 co-repressor scaffold. This is the opposite of the seed hypothesis and is the more parsimonious residual-function lead.

3. Non-transcriptional / chaperone-client alternative. The curated IntAct interactome (Hsp70/Hsp90/Hsp104/prefoldin/karyopherin) is equally consistent with FUN19 being a poorly characterized, possibly metastable protein whose principal validated partners are folding and transport machinery — not a transcription factor at all.

4. Weak baseline in both directions. Because every GO annotation is IBA and there is no experimental FUNCTION comment, no interpretation is anchored on direct assay. The refutation is strong relative to the added claim (coactivation) but should not be overstated into a positive assignment of repressive function.

Limitations and Knowledge Gaps

• No direct functional assay for FUN19. There is no reporter, ChIP, or transcriptome study establishing FUN19's effect on transcription in either direction. The case against coactivation is comparative and interaction-based, not a direct refuting experiment. Resolution: an inducible-reporter or nascent-transcription (e.g., 4tU-seq) assay in a fun19Δ strain.
• STRING SIN3 signal is aggregated, not binary. The escore of 0.755 is high-throughput co-occurrence; IntAct records no FUN19–Sin3 binary interaction. Whether FUN19 is a genuine Rpd3L/Sin3 subunit is unresolved. Resolution: co-IP / reciprocal TAP of Sin3 vs FUN19; size-exclusion co-migration.
• SWIRM functional test in FUN19 context. SWIRM tethers the HAT module in ADA2; whether FUN19's SWIRM binds any chromatin-modifying module is untested. Resolution: pulldown of FUN19-SWIRM against HAT/HDAC modules.
• Localization confirmation. KAP123 interaction suggests nuclear import but a direct GFP localization for FUN19 was not established in this analysis. Resolution: FUN19-GFP microscopy (or consult existing SGD localization data).
• PANTHER IRD / ancestral judgment. The seed cites an IRD loss of SAGA-type-complex membership (GO:0070461) at the fungal node. This was interpreted from the seed context, not independently re-derived from the PANTHER tree in these iterations. Resolution: directly inspect the PTHR12374 tree/IRD annotations.

Discriminating Tests (most efficient)

1. fun19Δ nascent-transcription / reporter assay. A directional readout: if FUN19 were a coactivator, its deletion should decrease activated transcription. This is the single most decisive test to distinguish coactivation from a repressive or null role.
2. Reciprocal co-IP of Sin3 and Gcn5/Ada2 with FUN19. Distinguishes genuine Rpd3L/Sin3 (repressive) membership from artifactual SAGA (activating) co-annotation. Predicts Sin3 pulls down FUN19 while SAGA does not.
3. FUN19-SWIRM biochemical pulldown vs HAT and HDAC modules. Tests whether the retained SWIRM tethers FUN19 to any chromatin-modifying complex, and if so which sign.
4. FUN19-GFP localization + KAP123-dependence. Confirms nuclear vs cytoplasmic partitioning and whether the karyopherin interaction is functional.
5. Comparative PANTHER/IRD re-derivation. Independently confirm the ancestral judgment (SAGA-membership loss at the fungal node) rather than relying on the seed's summary.

Curation Leads (require curator verification)

Primary lead — reject the added activation terms.
- Do NOT add GO:0003713 (transcription coactivator activity, MF). No experimental support; Ada2 effector domains (ZZ, SANT) are lost; SWIRM is a tether, not an activator; interactome is repressive/chaperone-dominated.
- Do NOT add GO:0045893 (positive regulation of DNA-templated transcription, BP). Same reasoning; the only propagated complex context is the repressive Rpd3L/Sin3 HDAC assembly.

Secondary leads — annotation hygiene.
- Retain the existing IBA terms as non-core / unverified: GO:0070210 (Rpd3L-Expanded), GO:0003682 (chromatin binding), GO:0006338 (chromatin remodeling), GO:0006357 (regulation of transcription by Pol II). Flag each as IBA-only, no experimental support.
- Flag SAGA-membership propagation as artifactual. Any pipeline that would transfer SAGA/coactivator terms to FUN19 from PTHR12374 should be blocked; STRING SAGA escore = 0.000 confirms these are co-annotation carry-over, and PANTHER records an IRD loss of SAGA-type-complex membership at the fungal node.
- Note the repressive lead, but do not annotate it as fact. The SIN3 (escore 0.755) association is the strongest experimental transcription-factor link and, with GO:0070210, suggests a repressive context — but IntAct shows no binary Sin3–FUN19 interaction, so this is a lead, not an assignment.

Candidate references with snippets to verify:
- PMID: 36965704: "the SWIRM domain is required for incorporation of the HAT module into the yeast SAGA complex, but not the ADA complex" — supports SWIRM = tether, not activator.
- PMID: 36965704: "A minimal HAT module complex containing Gcn5 bound to Ada2 and Ada3 is required for full Gcn5 activity on nucleosomes." — supports that coactivation requires the effector module FUN19 lacks.

Suggested curator questions:
- Is there any experimental (non-IBA) evidence anywhere in the literature for FUN19 affecting transcription in either direction?
- Should FUN19's IBA Rpd3L-Expanded annotation be retained given no binary Sin3 interaction in IntAct?

Conclusion

The seed hypothesis that FUN19 has transcription coactivator activity and positively regulates transcription is refuted / best treated as paralog-driven over-annotation. FUN19 retained only Ada2's SWIRM tethering module while losing the ZZ and SANT domains that underpin coactivation; it is not a SAGA subunit; all current annotations are IBA; the only propagated complex context and the strongest experimental interactor are repressive (Rpd3L / Sin3), and curated interactions place FUN19 with chaperones and a karyopherin rather than any coactivator complex. Curators should not add GO:0003713 or GO:0045893 and should retain only the neutral IBA-level chromatin terms pending direct experimental validation.

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist FUN19 GO decision table
• OpenScientist FUN19 IntAct curated interactions
• OpenScientist FUN19 STRING partners