SMIM22 (CASIMO1) review notes

2026-10-03 — initial review (claude-code, MICROPROTEINS Tier 2)

Identity

• UniProt K7EJ46, 83 aa (isoform 3, displayed); isoform 2 (VSP_060203) inserts 5 residues
  after R69 (R -> RKVSPW). PE1. One predicted helical TM segment (residues 32-52,
  ECO:0000255); C-terminal disordered, basic/acidic region 60-83. Cys-rich juxtamembrane
  stretch (VAHCCCCSS) just after the TM helix.
• Canonical gene symbol SMIM22 (HGNC:48329), not an alternative ORF of another protein-coding
  host, so the standard genes/human/SMIM22/ folder is correct.
• Alias CASIMO1, "Cancer-Associated Small Integral Membrane Open reading frame 1"
  PMID:29765154.

Literature

PubMed search "CASIMO1 OR SMIM22" (2026-10-03) returns 7 records. Only two study the protein:

1. PMID:29765154 (Polycarpou-Schwarz et al., Oncogene 2018) — discovery and sole source of all
   GOA experimental rows. Abstract only in cache; not OA (Unpaywall/EuropePMC: no free copy).
2. Overexpressed in hormone-receptor-positive breast tumours PMID:29765154.
3. Knockdown reduces proliferation PMID:29765154.
4. Loss disturbs actin, motility, causes G0/G1 arrest PMID:29765154.
5. Protein (not RNA) responsible PMID:29765154.
6. Interacts with SQLE PMID:29765154.
7. Raises SQLE protein, not mRNA; LD clustering; knockdown lowers SQLE and pERK PMID:29765154.
8. SQLE epistasis PMID:29765154.
9. UniProt (from full text, which I cannot see): interacts with CANX and DDOST; partially
   co-localises with LAMP1 in late endosome. The IPI row lists CANX (P27824), DDOST (P39656)
   and SQLE (Q14534). CANX and DDOST are abundant ER-membrane proteins (calnexin, OST
   subunit) — common co-purifying partners of any ER-inserted TM protein.
10. PMID:39215037 (Zhao et al., NPJ Precis Oncol 2024) — NSCLC, NORAD/miR-520g-3p/SMIM22/GALE
    axis. Full text. Claims SMIM22 overexpression raises glycolysis and proliferation, co-IP with
    GALE, and nuclear co-localisation PMID:39215037.
    The nuclear localisation of a single-pass TM protein is implausible and not explained;
    the paper never tests the protein vs RNA distinction. Treated as LOW_QUALITY, used only as
    corroboration of a proliferation phenotype.

Other hits (ALDH7A1 methylation, Li-7 HCC gene set, TiO2 lncRNA profile, prostate cancer
recurrence gene set, a 2026 microprotein review) are expression-list mentions only; not cached.

Assessment

• Mechanistic picture is thin: one study, one interaction (SQLE) with epistasis, protein-level
  effect on SQLE (stabilisation vs translation not distinguished in abstract).
• All BP rows are cancer-cell-line knockdown phenotypes; plausibly downstream of SQLE /
  cholesterol synthesis / ERK signalling. Keep as non-core.
• No GO "regulation of lipid droplet formation" term exists (QuickGO search 2026-10-03), so the
  lipid droplet formation row cannot be refined to a regulation term; keep as non-core.
• protein binding IPI: modify to enzyme binding (GO:0019899) on the strength of SQLE, which is
  the functionally validated partner (epistasis). CANX/DDOST hits are not informative.
• No NEW annotation: protein stabilisation (GO:0050821) toward SQLE is tempting but the
  abstract does not show a half-life experiment; left as a question.
• Subcellular localisation: UniProt says late endosome (partial LAMP1 overlap); ER is
  predicted by CANX/DDOST association and by SQLE partner (SQLE is ER-resident), but no
  ER location row exists and I cannot verify from the abstract — not added.