How do RUNX1 aberrations drive SPATS2L and stress granules in blast-phase CML?

RUNX1 loss rewires the RNA-binding proteome

The study starts from a clinical pattern: somatic RUNX1 mutations appear in roughly 10%–30% of BP-CML and are seldom found at diagnosis, and the RUNX1::MECOM/EVI1 fusion from t(3;21)(q26.2;q22) is likewise linked to tyrosine kinase inhibitor (TKI) resistance and poor survival. Both lesions are treated here as ways RUNX1 function is disturbed, not as a single mutation type.

To see what happens downstream of that disturbance, the authors used orthogonal organic phase separation (OOPS), a method that crosslinks RNA to proteins with UV-C and then isolates the RNA-bound proteins from the interphase of a phase separation. Applied to RUNX1-depleted KU812 and K562 BP-CML cells, it produced a catalogue of RNA-binding proteins (RBPs) whose RNA association changed.

The altered RBPs clustered in stress-response pathways — including EIF2AK4 (GCN2) responses to amino acid deficiency, oxidative phosphorylation, NRF2-mediated oxidative stress and the unfolded protein response — and in translation and ribosome biogenesis. This is a proteome-level association result: it shows which RNA-bound proteins shift, not yet which one drives the phenotype.

Ribosome output rises without more active translation

One concrete follow-up concerned 5′TOP motifs, short sequence elements in the 5′ untranslated region of many translation-related mRNAs that cells use to stall translation under stress. In RUNX1-depleted KU812 cells, 22% (17 of 78) of 5′TOP-containing RBPs changed significantly, and 58% of those changes were increases in ribosomal subunit proteins.

Polysome profiling in KU812 and K562 cells added a nuance that is easy to misread. Area-under-the-curve analysis showed more monosomes and more polysomes after RUNX1 depletion, but no consistent rise in the polysome-to-monosome ratio. In the authors' reading, ribosome abundance increased without a shift toward actively translating polysomes, so global translation is not simply upregulated.

That distinction matters because more ribosome material and more active translation are different measurements. The paper's interpretation is that RUNX1 loss may promote translational reprogramming, a stress adaptation, rather than a blanket increase in protein synthesis.

SPATS2L rises under two different RUNX1 aberrations

Among the RBPs, the authors focused on SPATS2L (spermatogenesis associated serine rich 2-like), a stress-granule and nucleolar protein that also regulates rRNA processing. It was enriched 1.65 log2-fold in RNA from RUNX1-depleted K562 cells and 0.52 log2-fold in KU812 cells, and multiple SPATS2L isoforms increased in whole-cell lysates from both lines.

The induction was not limited to knockdown. Lentiviral expression of RUNX1::EVI1 in K562 and KU812 cells raised SPATS2L protein to levels comparable with SKH1, a BP-CML line that carries t(3;21) endogenously, and both RUNX1 depletion and RUNX1::EVI1 expression significantly increased SPATS2L mRNA. RUNX2 and RUNX3 expression was low, unchanged or inconsistent across the two lines, which the authors use to argue these paralogues are unlikely to explain the induction.

Primary-sample support comes from re-analysis of an existing RNA-seq dataset: SPATS2L was significantly higher in RUNX1-mutated (all runt-homology-domain mutant) than RUNX1-wild-type primary BP-CML samples. The authors note it remains unclear whether RUNX1 regulates SPATS2L directly or indirectly, since RUNX1 ChIP peaks are not consistently enriched at the SPATS2L promoter and oxidative stress alone was enough to induce SPATS2L.

SPATS2L depletion slows growth and cuts stress granules

To test whether SPATS2L matters for the cells rather than only marking them, the authors used two SPATS2L-targeting shRNAs in K562, KU812 and SKH1 cells. Depletion inhibited growth and viability in all three BP-CML lines, measured by cell counts and by 7AAD staining of dead cells at 24, 48 and 72 hours, with apoptosis assessed by Annexin V at 48 hours.

Stress granules are membraneless cytoplasmic condensates that sequester mRNAs and proteins during acute stress and are linked to survival and therapy resistance. After a one-hour sodium arsenite exposure to induce them, SPATS2L depletion reduced the average number of stress granules per cell — but not their size — in K562 and KU812 cells, scored with the G3BP1 marker.

So SPATS2L depletion affects two separable readouts: it lowers cell growth and viability, and it lowers granule number. The paper does not claim these are the same effect, and the size result stayed flat.

RUNX1 loss builds granules that HHT erases

Consistent with the proteome data, RUNX1 depletion raised basal SPATS2L, and sodium arsenite raised it further; arsenite alone was sufficient to induce SPATS2L, which the authors read as evidence that this induction can be a general stress response. RUNX1-depleted K562 cells assembled significantly more stress granules per cell, with a similar trend in KU812.

Homoharringtonine (HHT), a translation inhibitor used historically in TKI-resistant CML, binds the ribosomal A site and blocks new ribosomes from entering elongation. In K562 cells with RUNX1 depletion, HHT at 50 ng/mL for 48 hours nearly eradicated arsenite-induced stress granules compared with DMSO.

The clinically relevant comparison is in primary cells: BP-CML samples with loss-of-function or hypomorphic RUNX1 mutations were preferentially sensitised to HHT. The authors group RUNX1 mutations into loss-of-function, hypomorphic and wild-type-like classes, with hypomorphic mutants showing defective CBFβ interaction, and it is the first two classes that carry this sensitivity.

SPATS2L loss restores HHT sensitivity

The mechanistic test combined two knockdowns. In K562 cells already depleted of RUNX1, concurrent SPATS2L depletion further increased HHT sensitivity, measured as fold change because the double-knockdown cells already had higher basal death, and also increased imatinib sensitivity.

The same double-knockdown design brought stress-granule numbers back to control levels: RUNX1 depletion alone significantly raised granules per cell, but adding SPATS2L depletion removed that increase. This links the granule phenotype to SPATS2L specifically rather than to RUNX1 loss in general.

The authors are careful about causality here. Because SPATS2L depletion alone reduces growth and viability, they state they cannot exclude an additive effect of RUNX1 suppression acting through cellular stress or reduced fitness. The claim is that SPATS2L contributes to therapeutic resistance in RUNX1-mutated CML, not that it is the sole cause.

Granule inhibition re-sensitises TKI-resistant cells

A separate experiment asked whether stress granules matter beyond RUNX1-mutant cells. The authors used a commercially available imatinib-resistant K562 line (K562-R), first confirming it was significantly more resistant than parental K562, then inhibiting granule formation with FAZ-3532, a G3BP1 inhibitor described as previously untested in leukaemia.

At 5 µM, FAZ-3532 markedly reduced stress-granule assembly and was not cytotoxic to K562 cells on its own. Combined with imatinib, it re-sensitised K562-R cells, with significantly enhanced cell death at 24, 48 and 72 hours.

This is the paper's basis for suggesting stress granules may contribute more widely to TKI resistance in BP-CML. It is a cell-line result with a pharmacological inhibitor, and the authors present it as an indication rather than a demonstrated clinical strategy.

Where the RUNX1–SPATS2L axis changes the decision

The paper's most decision-relevant split is mutational, not simply diagnostic. BP-CML cells with loss-of-function or hypomorphic RUNX1 mutations were the ones preferentially sensitised to HHT, and the SPATS2L/SG axis is proposed as the reason those cells remain vulnerable even though the same axis may also underlie their resistance.

That creates a specific therapeutic logic: HHT or its semisynthetic derivative omacetaxine mepessucinate targets the granule-dependent survival state, while sustained SPATS2L-driven granule formation could restore resistance. The double-knockdown result — more HHT sensitivity and granule numbers back to control — is the evidence that this axis is adjustable rather than fixed.

The authors describe the work as preliminary evidence for a mutationally targeted approach, and they flag a small number of primary samples as a limitation, given how rare BP-CML is. The practical priority the data support is therefore narrow and testable: stratify BP-CML by RUNX1 mutation class and by SPATS2L/stress-granule status when considering translation-inhibitor strategies, rather than treating RUNX1 aberration as one uniform marker.

Sources used in this answer