The fatty acid elongase Elovl6 is crucial for hematopoietic stem cell engraftment and leukemia propagation

TO THE EDITOR Lipid metabolism plays an essential role in regulating stem [1] and cancer [2] cell function. The biological significance of lipid species diversity, termed ‘lipid code’, has drawn increasing attention. Variations in the length and degree of desaturation of fatty acid (FA) chains contribute to this diversity. Among the rate-limiting enzymes catalyzing long-chain FA elongation [elongation of very long-chain fatty acid proteins (ELOVL1–7)], ELOVL6 converts C16 saturated and monounsaturated FAs to C18 species, the most important step for the de novo synthesis of endogenous long-chain FAs [3]. Elovl6 gene disruption (E6KO) in mice decreases the proportion of stearate (C18:0) and oleate (C18:1n-9) and increases that of palmitate (C16:0) and palmitoleate (C16:1n-7) in the liver [3, 4]. E6KO ameliorates metabolic and inflammatory diseases in mice such as insulin resistance [3] and non-alcoholic steatohepatitis [5]. Moreover, high ELOVL6 expression correlates with poor prognosis of patients with breast [6] and liver [7] cancer. Here, we report that ELOVL6 is essential for hematopoietic stem cell (HSC) engraftment after bone marrow (BM) transplantation and that it blocks acute myeloid leukemia (AML) development in a mouse model. These outcomes are at least in part attributed to the defective chemotaxis of HSCs or of transformed hematopoietic progenitor cells (HPCs) owing to diminished CXCL12 signaling through the PI3K-RAC pathway. We first found that Elovl6 transcripts levels were higher in HSC and HPC fractions than in most peripheral blood cell fractions (Fig. S1A, B). Then, we analyzed FAs in BM cell lysates and observed significantly increased palmitate levels and decreased stearate to palmitate ratio in E6KO cells, relative to that in wildtype (WT) cells, as seen in the liver of E6KO mice [3] (Fig. S1C). To assess HSC function following E6KO, we performed competitive repopulation assays at a 1:1 ratio between WT or E6KO donor cells (Ly5.2) and WT competitors (Ly5.1/Ly5.2). Notably, in the recipient mice, percentage of donor-derived BM cells was around 50% and <0.2% with WT and E6KO Ly5.2 cells, respectively, at 12 weeks after transplantation. Even at a ratio of 10:1, the percentage of donor-derived E6KO BM cells was similarly low while that of WT Ly5.2 cells was >90% (Fig. 1A(i), (ii), and S1D, E). Next, we explored the effect of E6KO on leukemia transformation and propagation using the HPCs from E6KO mice. We retrovirally transduced MLL::AF9 into lineage-negative and c-Kit cells from WT or E6KO mouse BM. The resulting cells (MA9 cells) of both genotypes proliferated exponentially in a comparable manner [8, 9] (Fig. S2A). Subsequently, we transplanted lethally irradiated syngeneic mice with either WT or E6KO MA9 cells (Fig. S2B). All mice receiving WT cells developed AML as previously reported [9]; however, AML propagation was blocked in E6KO MA9 cell-transplanted recipients (Fig. 1B). To determine the mechanisms underlying E6KO HSC and MA9 cell engraftment failure, we performed whole transcriptome analysis with CD34LSK and MA9 cells of WT or E6KO backgrounds. This analysis demonstrated enrichment of chemotaxis and CXCR4 pathway gene sets in E6KO MA9 cells, relative to those in WT cells (Fig. S3A). To substantiate these results, we tested in vitro chemotaxis of E6KO CD34LSK and MA9 cells toward CXCL12 in a chemotaxis assay (Fig. 1C). A greater fraction of WT than that of E6KO CD34LSK cells migrated to the CXCL12-containing chamber, with specific chemotaxis of ~20% and ~0%, respectively (Fig. 1C(i)). In WT and E6KO MA9 cells, the fraction of cells showing specific chemotaxis toward CXCL12 was also greater in WT cells than in E6KO MA9 cells (Fig. 1C(ii)), while CXCR4 expression levels on WT and E6KO MA9 cells were comparable (Fig. S3B). To determine whether inhibition of ELOVL6 affects the migratory ability of MA9 cells, we performed a chemotaxis assay using Compound B, an

To determine the mechanisms underlying E6KO HSC and MA9 cell engraftment failure, we performed whole transcriptome analysis with CD34 -LSK and MA9 cells of WT or E6KO backgrounds. This analysis demonstrated enrichment of chemotaxis and CXCR4 pathway gene sets in E6KO MA9 cells, relative to those in WT cells (Fig. S3A). To substantiate these results, we tested in vitro chemotaxis of E6KO CD34 -LSK and MA9 cells toward CXCL12 in a chemotaxis assay (Fig. 1C). A greater fraction of WT than that of E6KO CD34 -LSK cells migrated to the CXCL12-containing chamber, with specific chemotaxis of~20% and~0%, respectively ( Fig. 1C(i)). In WT and E6KO MA9 cells, the fraction of cells showing specific chemotaxis toward CXCL12 was also greater in WT cells than in E6KO MA9 cells (Fig. 1C(ii)), while CXCR4 expression levels on WT and E6KO MA9 cells were comparable (Fig. S3B). To determine whether inhibition of ELOVL6 affects the migratory ability of MA9 cells, we performed a chemotaxis assay using Compound B, an ELOVL6 inhibitor. Chemotaxis was decreased in a Compound B concentration-dependent manner in WT MA9 cells. By contrast, Compound B did not affect the migration of E6KO MA9 cells ( Fig. 1C(iii)). To verify these findings in HSCs in vivo, we connected Ly5.1 WT mice using parabiotic surgery one by one to WT or E6KO Ly5.2 mice (Fig. S3C). Four weeks later, 0.8% of LSK and 1.6% of CD34 -LSK cells, on average, in the BM of Ly5.1 mice, were replaced with Ly5.2 cells from WT mice, while almost no LSK or CD34 -LSK cells were replaced with Ly5.2 cells from E6KO mice (Fig. S3C). We then analyzed homing capacity of MA9 cells in the recipient BM at 48 h after transplantation, finding a proportion of WT MA9 cells detectable but E6KO MA9 cells undetectable (Fig. S3D).
CXCR4 signaling depends in part on PI3K activating the downstream cascades. To determine whether Elovl6 loss alters PI3K activation, we examined AKT phosphorylation, a representative readout of PI3K activity [10], in WT and E6KO MA9 cells before and after CXCL12 stimulation. CXCL12 stimulation increased AKT phosphorylation at Ser473 and Thr308 in MA9 cells of both genotypes, but these increases were significantly weaker in E6KO cells than in WT MA9 cells ( Fig. 2A(i), (ii)). ERK phosphorylation, which is induced by CXCR4 activation, via a signaling pathway different from PI3K [11], was also induced by CXCL12 stimulation but was unaffected by Elovl6 loss (Fig. S4A). Gene set enrichment analysis indicated that Rac1 pathway molecules were more enriched in E6KO cells than in WT MA9 cells (Fig. S3A(v)) and that Rac1 is reportedly activated by PI3K downstream [12]. We, thus, investigated whether Rac1 activation is blocked by Elovl6 loss. Levels of GTP-bound Rac1 markedly increased in WT MA9 cells following CXCL12 stimulation, but the increase was not significant in E6KO MA9 cells ( Fig. 2A(iii), (iv)).
We then examined the pharmacologic effects of inhibiting PI3K, AKT, or RAC pathways on chemotaxis. In a transwell assay, the presence of the pan-PI3K inhibitor copanlisib or pan-RAC inhibitor EHT1864 decreased the fraction of WT MA9 cells migrating to the CXCL12-containing chamber. However, E6KO MA9 cell migration, which was already compromised, was not significantly affected by copanlisib or EHT1864. By contrast, the pan-AKT inhibitor MK2206 did not alter WT or E6KO MA9 cell migration ( Fig. 2B(i)-(iii)), suggesting minimal effects of AKT activation on MA9 cell motility.
Given impaired chemotaxis and PI3K-Rac1 signaling in Elovl6deficient cells, we investigated whether these cells exhibited altered actin remodeling in response to CXCL12. CXCL12 stimulation increased the fraction of WT CD34 -LSK cells showing lamellipodia from 14 to 52% (Fig. S4F(i), (ii)), whereas the fraction of E6KO CD34 -LSK cells exhibiting lamellipodia remained unchanged (Fig. S4F(ii)). Thus, ELOVL6 is likely to regulate a pathway that is responsive to CXCL12, resulting in PI3K-Rac1mediated cytoskeletal remodeling and chemotaxis.
To determine whether exogenously expressed Elovl6 could rescue E6KO MA9 cell phenotypes, we generated E6KO MA9 cells expressing Flag-tagged ELOVL6 (E6KO MA9 Flag-E6 ) and mock controls (E6KO MA9 mock ) (Fig. S5A). First, we evaluated AKT phosphorylation (Fig. S5B) and Rac1 activation (Fig. S5C) following CXCL12 stimulation. Levels of both were higher in E6KO MA9 Flag-E6 cells than in E6KO MA9 mock cells and comparable to the levels seen in WT MA9 Flag-E6 and WT MA9 mock cells. In chemotaxis analysis, the migration ratio was increased in E6KO MA9 Flag-E6 cells compared to that seen with E6KO MA9 mock cells (Fig. S5D). Finally, to evaluate rescue of AML propagation, we transplanted lethally irradiated syngeneic mice with either E6KO MA9 mock or E6KO MA9 Flag-E6 cells and observed that among the recipient mice transplanted with E6KO MA9 mock cells, very few recipients developed AML, as with Elovl6 −/− MA9 cells (Fig. 2C). By contrast, 10 out of 14 recipients died due to AML following E6KO MA9 Flag-E6 Effect of ELOVL6 inhibitor on migration of WT or E6KO MA9 cells toward a CXCL12-containing chamber; n = 3 each (iii). Data are shown as the mean ± SD in all bar charts. *P < 0.05, **P < 0.01, ***P < 0.001.
To gain an insight into whether ELOVL6 expression levels show any clinical significance in AML patients, we used GEPIA2 (ref. [13]) for The Cancer Genome Atlas cohort. This analysis revealed the association between high ELOVL6 expression levels and worse overall survival (Fig. S6A). Similar results were obtained from the Beat AML data (Fig. S6B).
In summary, this study reports novel outcomes in normal HSCs and transformed hematopoietic cells following lipid metabolic changes induced by Elovl6 deficiency. Our finding that Elovl6 loss hampers AML propagation will facilitate the development of novel cancer treatments; ELOVL6 activity or pathways regulated by ELOVL6 are potential targets of anti-AML therapy.

DATA AVAILABILITY
Further information and requests for resources and reagents including cell lines and mouse models should be directed to and will be fulfilled by TK and SC.