Leukemia-Mutated Proteins PHF6 and PHIP Form a Repressive Chromatin Complex That Suppresses AML Stemness

PHF6 (Plant Homeodomain Finger protein 6) is a chromatin-binding protein that is mutated in 3-5% of myeloid malignancies. About 2/3rd of PHF6 mutations are frameshift or nonsense mutations distributed throughout the gene body, presumed to produce null alleles, while 1/3rd are missense mutations clustered in its ePHD2 (extended PHD2) domain. The molecular consequences of PHF6 missense mutations are unknown. Mouse hematopoietic knockout of Phf6 leads to increased hematopoietic stem cell (HSC) self-renewal, and we recently reported that this ability is co-opted in Phf6-knockout mouse acute myeloid leukemia (AML) to drive leukemia stem cell self-renewal (Jalnapurkar, Pawar et al, Leukemia 2024). However, the molecular partners of PHF6 and its function on chromatin remain unknown.

We generated PHF6 knockout and rescue systems in the human THP-1 AML cell line and observed that similar to mouse AML, PHF6 loss de-represses a stemness gene network and reduces surface expression of myeloid differentiation markers CD11b, CD11c, and CD14. ChIP-Seq identified ~8600 high-confidence PHF6 peaks at open and active regions of chromatin, with approximately half of them overlapping promoters. HOMER motif analysis showed enrichment of ETS motifs at PHF6 peaks, indicating that PHF6 occupies loci bound by hematopoietic ETS factors like PU.1 and ERG. ATAC-Seq and RNA-Seq showed that promoters bound by PHF6 showed increased accessibility and expression following PHF6 knockout. Thus, PHF6 directly represses gene expression through occupancy at promoters bound by ETS factors.

We used CRISPR/Cas9-based knockout and homology-directed repair to engineer, at the endogenous PHF6 locus, the six most common ePHD2-domain missense mutations seen in AML patients (C242Y, D262V, R274Q, G287V, C297Y, and I314T). Most of these mutations are currently annotated as VUS (variants of unknown significance) by clinical laboratories. We observed that C242Y, G287V, and C297Y led to a dramatic reduction in PHF6 protein stability, while D262V, R274Q, and I314T led to relatively conserved PHF6 protein levels with reduced chromatin occupancy. All mutants converged on effects on myeloid surface markers similar to those seen with knockout. Thus, PHF6 somatic missense mutations seen in AML are either functionally knockout or hypomorphic.

Analyses of pan-cancer CRISPR screen results from the DepMap Portal indicated a high correlation of PHF6 knockout effects with knockout of the gene PHIP, that encodes a protein associated with the CRL4-E3 ubiquitin ligase complex. Rare germline mutations in both PHF6 and PHIP produce neurodevelopmental disorders with similar symptoms such as intellectual disability, obesity, and craniofacial abnormalities along with similar epigenetic profiles of DNA methylation (Vos et al, Human Genetics 2024). PHIP mutations had been uncommonly reported in myeloid malignancies and T-ALL but were recently observed in 7% of black patients with AML (Stiff et al, ASH Abstracts 2023). To explore a convergent biological function for these two proteins, we generated single and double knockouts of PHF6 and PHIP in THP-1 cells. RNA-Seq and flow cytometry showed that knockouts of either or both had highly similar effects on the transcriptome and surface markers, shifting the cells to a more stem-like state. PHF6 and PHIP KO clones overlapped on RNA-Seq principal component analysis (PCA), away from WT clones. ChIP-Seq showed that virtually all PHF6 peaks overlapped with PHIP peaks. Strikingly, PHIP knockout led to the complete inability of PHF6 protein to occupy chromatin, even though PHF6 protein level and nuclear localization remained unchanged. On the other hand, PHF6 loss did not affect PHIP occupancy on chromatin. Thus, PHF6 requires PHIP for its chromatin occupancy, and they collectively repress a stemness gene network.

Overall, our work unifies the molecular roles of two leukemia-mutated proteins PHF6 and PHIP, demonstrating that they suppress AML stemness by forming a repressive complex at ETS-factor-bound promoters.

No relevant conflicts of interest to declare.