Main

Thrombocytosis and erythrocytopenia are common consequences of abnormal hematopoiesis in individuals with advanced cancer1,2. Enhanced megakaryopoiesis increases platelet (PLT) numbers, which potentially induces thrombogenesis and tumor cell metastasis in individuals with cancer3,4,5. Simultaneously, erythrocytopenia occurs, which causes anemia and exacerbates tumor hypoxia2,6. Both anemia and thrombogenesis mark poor prognosis in various types of cancers2,7,8. Despite the important roles of erythrocytopenia and thrombocytosis in tumor progression, metastasis and treatment resistance, the molecular mechanisms of erythrocytopenia and thrombocytosis formation remain largely unclear. Notably, PLTs and erythrocytes are differentiated from the common megakaryocytic–erythroid progenitor cells (MEPs)9,10. Many transcription factors, such as GATA binding protein 1 (GATA1), GATA2, Runt-related transcription factor 1 (RUNX1), T cell acute lymphocytic leukemia 1 (TAL1), friend leukemia integration 1 (FLI1), MYB and Krueppel-like factor 1 (KLF1), are involved in megakaryocytic–erythroid differentiation11,12,13,14. RUNX1 and GATA1 are pivotal for megakaryocytic and erythroid differentiation, respectively15. Despite physiological regulation, little is known about how tumor-derived factors disrupt normal MEP differentiation and cause erythrocytopenia and thrombocytosis.

Aryl hydrocarbon receptor (AhR), a member of the basic helix-loop-helix family of transcription factors, plays an essential role in a wide variety of cells, including hematopoietic progenitor cells16,17. Notably, defects in PLT number and function have been observed in Ahr-null mice18; however, the AhR inhibitor stemregenin 1 (SR1) may promote erythropoiesis from human embryonic stem cells19, suggesting that AhR possibly regulates megakaryocytic–erythroid differentiation. As a cytoplasmic transcription factor, AhR can be activated by exogenous xenobiotics, such as 2,3,7,8-tetrachlorodibenzo-p-dioxin, and endogenous metabolites, including kynurenine (Kyn)20,21. Our previous studies have demonstrated that both tumor cells and immune cells mobilize the tryptophan–Kyn–AhR circuitry to dampen antitumor immunity22,23, and high levels of Kyn are present in the peripheral blood of individuals with cancer23,24. Based on these lines of information, we hypothesized that in individuals with cancer, Kyn is mobilized to trigger thrombocytosis and erythrocytopenia by activating AhR.

Results

Abnormal MEP differentiation occurs in advanced tumors

Given the frequency of anemia and thrombocytosis in individuals with advanced cancer1,2, we analyzed the number of red blood cells (RBCs) and PLTs in individuals with stage III or IV colon (n = 100), lung (n = 99) or breast (n = 60) cancer. We found that numbers of RBCs and PLTs were reduced and elevated (Fig. 1a), respectively, in the blood compared to numbers observed in healthy control individuals. Analysis of P-selectin and integrin GPIIb/IIIa of PLTs revealed that these PLT-activating markers were upregulated in individuals with cancer (Extended Data Fig. 1a,b), suggesting that increased PLT activation contributes to tumor-associated events, such as metastasis and thrombosis. We then validated these clinical results in mice. After inoculating mouse MC38 colon cancer cells or E0771 breast cancer cells into C57BL/6 mice, we observed that PLTs and RBCs gradually increased and decreased, respectively, in the blood of mice bearing large tumors (approximately 2,000 mm3; Fig. 1b and Extended Data Fig. 1c). Hematopoietic stem cells (HSCs) differentiate from long-term HSCs (LT-HSCs) to short-term HSCs (ST-HSCs) and multipotent progenitors (MPPs); the MPPs then differentiate into common myeloid progenitors (CMPs) and downstream MEPs25,26 (Extended Data Fig. 1d). Flow cytometric analysis of bone marrow (BM) HSCs and progenitor cells (Extended Data Fig. 1e–g) revealed that the numbers of LT-HSCs, ST-HSCs, MPPs, CMPs and MEPs were not altered in tumor-bearing mice compared to in tumor-free control mice (Fig. 1c and Extended Data Fig. 1h). We thus postulated that MEPs biased their differentiation into megakaryocytes for thrombocytosis and erythrocytopenia. Megakaryocyte progenitor cells (MkPs) and erythrocyte progenitor cells (EryPs) are directly differentiated from MEPs to generate megakaryocytes and erythrocytes, respectively. Analysis of MkPs and EryPs by flow cytometry (Extended Data Fig. 1i) showed that MkPs increased and EryPs decreased in mice bearing colon or breast cancer (Fig. 1d and Extended Data Fig. 1j). By conducting a colony-forming unit (c.f.u.) assay27, we found that seeding normal MEPs in 10% collagen medium mainly formed CD41+CD71+ megakaryocyte/erythrocyte, CD71+ erythroid-only and CD41+ megakaryocyte-only colonies (Fig. 1e), whereas MEPs from tumor-bearing mice formed more megakaryocyte-only but less erythrocyte-only colonies (Fig. 1f and Extended Data Fig. 1k). Together, these results suggest that tumor growth gradually biases MEP differentiation into megakaryocytes, leading to thrombocythemia and anemia.

Fig. 1: Advanced tumors induce abnormal MEP differentiation.
figure 1

a, Peripheral RBC and PLT analysis of healthy individuals (n = 50 males, n = 50 females) and individuals with untreated colon cancer (n = 50 males, n = 50 females), breast cancer (n = 60 females) and lung cancer (n = 50 males, n = 50 females). b–d, C57BL/6J mice were inoculated with MC38 colon cancer cells. RBCs and PLTs were counted once per week until MC38 tumors were approximately 2,000 mm3 at week 4 (b), then LT-HSCs, ST-HSCs, MPPs, CMPs and MEPs (c) and EryPs and MkPs (d) were analyzed after the last RBC and PLT count; Ctrl, control. e,f, The same as b, except that control and MC38 tumor-bearing mouse-derived MEPs were isolated and cultured in 10% collagen-based medium for 7 d. Colonies were labeled with anti-CD41 and anti-CD71 and were assessed by fluorescence microscopy (e). Representative images of mouse MEP-derived colonies (e) and results of a c.f.u. analysis of MEPs derived from control or MC38 tumor-bearing mice (f) are shown. Scale bar, 100 μm. BF, brightfield; E, erythrocyte; Mk, megakaryocyte. In b–d and f, n = 10 mice. Data were analyzed by two-tailed Student’s t-tests (b–d and f) or one-way analysis of variance (ANOVA) followed by a Bonferroni post hoc test (a). The data represent mean ± s.d. (a, c, d and f) or mean ± s.e.m. (b); NS, not significant.

Source data

Full size image

AhR promotes MEP differentiation into megakaryocytes

Next, we determined whether AhR plays a role in MEP differentiation. The genes encoding cytochrome P450 family 1 subfamily A polypeptide 1 (CYP1A1) and cytochrome P450 family 1 subfamily B polypeptide 1 (CYP1B1) are AhR-targeted genes28,29. The expression of both Cyp1a1 and Cyp1b1 was upregulated in MEPs isolated from MC38 or E0771 tumor-bearing mice compared to in control healthy mice (Fig. 2a and Extended Data Fig. 2a). In line with this finding, the active form of AhR in the nucleus of MEPs was verified by immunofluorescence staining (Fig. 2b and Extended Data Fig. 2b). However, tumor growth in Ahr–/– mice did not affect MEP differentiation relative to that in wild-type (WT) mice, as evidenced by the consistent number of peripheral RBCs and PLTs (Fig. 2c and Extended Data Fig. 2c) and BM MkPs and EryPs (Fig. 2d and Extended Data Fig. 2d), suggesting that AhR regulates MEP differentiation into megakaryocytes in tumor-bearing mice. To verify this finding, we adoptively transferred BM cells to lethally irradiated (9 Gy) tdTomato-transgenic mice. Six weeks later, when the increased number of PLTs and RBCs stabilized, mice were inoculated with MC38 or E0771 tumor cells (Extended Data Fig. 2e). We found that TER119+tdTomato− RBCs and CD41+tdTomato− PLTs gradually decreased and increased, respectively, in tumor-bearing mice compared to in tumor-free control mice (Fig. 2e and Extended Data Fig. 2f,g). Intriguingly, the adoptive transfer of Ahr–/– BM cells led to more RBCs and fewer PLTs in the periphery (Fig. 2f), which was not affected by tumor growth (Fig. 2g and Extended Data Fig. 2h). Kyn, a typical ligand for AhR, has been reported to be elevated in individuals with cancer and in tumor-bearing mice20,24,30. Consistently, higher serum Kyn levels were found in MC38 or E0771 tumor-bearing mice on day 14 after inoculation than in control mice (Fig. 2h and Extended Data Fig. 2i). Kyn treatment led to the entry of AhR into the nucleus and upregulation of Cyp1a1 and Cyp1b1 expression in MEPs (Fig. 2i,j). A c.f.u. assay showed that Kyn treatment biased MEP differentiation into megakaryocytes (Fig. 2k). To validate these results in vivo, we injected 400 μg of Kyn intraperitoneally (i.p.) into mice once every 12 h. This dosage imitated the Kyn content in tumor-bearing mice (Extended Data Fig. 2j). Two weeks later, we found that erythrocyte numbers were reduced, but the PLT count was elevated in the blood of treated mice, concomitant with increased MkPs but reduced EryPs (Fig. 2l). Together, these results suggest that increased AhR activity may unbalance the differentiation of MEPs into MkPs and EryPs, thus leading to thrombocytosis and erythropenia in tumor-bearing hosts.

Fig. 2: MEP differentiation is biased toward megakaryocytes by AhR activation.
figure 2

a,b, MEPs were isolated from control and MC38 tumor-bearing mice at week 4. Quantitative PCR with reverse transcription (RT–qPCR) detected Ahr, Cyp1a1 and Cyp1b1 expression (a). Cells were also stained for AhR and assessed by confocal microscopy (b). Scale bar, 10 μm. c,d, Ahr–/– mice were inoculated with MC38 cells, and RBCs and PLTs were counted at a tumor size of approximately 2,000 mm3 (c). EryPs and MkPs were analyzed (d). e, tdTomato-transgenic mice were irradiated with 9 Gy by X-ray with 4-h intervals, and 106 BM cells from WT mice were transferred into the irradiated mice. Mice were then inoculated with MC38 cells at week 6. Donor-derived CD41+tdTomato− PLTs and ter119+tdTomato− RBCs were counted at the same time point as in c and d. f,g, The same as e, except that WT or Ahr–/– BM cells were transferred. Reconstituted peripheral blood RBCs and PLTs were analyzed at week 6 (f). Mice transferred with Ahr–/– BM cells were inoculated with MC38 cells, and donor-derived PLTs and RBCs were analyzed at the same time point as in c and d (g). h, C57BL/6J mice were inoculated with MC38 cells, and serum Kyn levels from control and tumor-bearing mice were measured by high-performance liquid chromatography (HPLC) once per week for 4 weeks. i,j, MEPs were isolated from healthy C57BL/6J mice and treated with PBS or 5 μM Kyn. Treated MEPs were stained for AhR and analyzed by confocal microscopy at 48 h (i). Scale bar, 10 μm. Expression of Ahr, Cyp1a1 and Cyp1b1 was determined by RT–qPCR at 24 h (j). k, Analysis of c.f.u. for PBS- or Kyn-treated MEPs isolated from healthy C57BL/6J mice. l, C57BL/6J mice were treated with Kyn (400 μg i.p. every 12 h) for 2 weeks, followed by RBC, PLT, EryP and MkP analysis. In a, b, h and l, n = 10 mice. In c–g, n = 6 mice. In i–k, n = 3 independent experiments. Data were analyzed by two-tailed Student’s t-tests (a–l) and represent mean ± s.d (a–g and i–l) or mean ± s.e.m (h).

Source data

Full size image

AhR upregulates RUNX1 to bias MEP differentiation

Next, we investigated the molecular basis by which AhR-biased MEPs differentiate into megakaryocytes. A panel of transcription factors, such as GATA1, GATA2 and RUNX1, regulates MEP differentiation11,12,13,14. By examining these transcription factors in Kyn-treated MEPs, we found that expression of RUNX1, rather than the other transcription factors, was upregulated at the mRNA and protein levels (Fig. 3a,b), which was also verified by immunofluorescence (Extended Data Fig. 3a). Using a RUNX1 inhibitor, we found that Kyn-biased MEP differentiation was abrogated, but erythrocyte differentiation was augmented (Fig. 3c), prompting us to hypothesize that AhR regulates MEP differentiation via RUNX1. The Runx1 gene possesses a distal P1 promoter and a proximal P2 promoter31,32,33 (Extended Data Fig. 3b), which results in two major isomers RUNX1c and RUNX1b with differences between their N-terminal amino acid sequences. Analysis of the JASPAR database revealed the AhR-binding core sequence (5′-GCGTG-3′) in the Runx1 P1 and P2 promoters. By performing CUT&RUN-qPCR, we found that more DNA fragments were enriched in the P1 promoter than in the P2 promoter in Kyn-treated MEPs (Fig. 3d). A dual-luciferase assay also showed that AhR promoted reporter gene expression via the Runx1 P1 promoter (Fig. 3e); however, this effect was abrogated by the mutation of 5′-GCGTG-3′ to 5′-GCATG-3′ in the P1 promoter (Extended Data Fig. 3c). In line with this finding, Kyn treatment upregulated the expression of Runx1c but not Runx1b (Fig. 3f). In addition, Kyn treatment did not alter RUNX1 expression in Ahr–/– MEPs (Fig. 3g and Extended Data Fig. 3d). Consistently, pretreatment of MEPs with the AhR inhibitor SR1 abrogated the effect of Kyn on RUNX1 expression (Fig. 3h and Extended Data Fig. 3e). RUNX1 regulates a panel of target genes, such as KLF1, ITGA2B and GP1BA, to exert its function34,35,36. CUT&RUN-qPCR and dual-luciferase assays showed that Kyn treatment enables RUNX1 to bind and transactivate Itga2b and Gp1ba in MEPs (Fig. 3i,j). In addition, Kyn treatment upregulated the expression of Itga2b and Gp1ba, which, however, was abrogated by the AhR inhibitor SR1 and the RUNX1 inhibitor Ro5-3335 (Extended Data Fig. 3f), suggesting that AhR–RUNX1 regulates Itga2b/Gp1ba expression. Notably, Klf1, another target gene of RUNX1 (Extended Data Fig. 3g), although being not affected in Kyn-treated MEPs (Fig. 3a,b), was downregulated in MkPs, while Itga2b and Gp1ba were upregulated in MkPs, compared with in MEPs or EryPs (Extended Data Fig. 3h). Together, these data suggest that AhR transcriptionally activates RUNX1 to promote MEP differentiation into megakaryocytes.

Fig. 3: AhR regulates MEP differentiation bias by upregulating RUNX1.
figure 3

a, MEPs were isolated from healthy C57BL/6J mice and treated with Kyn. The expression of Gata1, Gata2, Runx1, Tal1, Klf1, Fli1 and Myb was determined by RT–qPCR after 24 h. b, Protein expression was detected by digital western blotting after 48 h. c, The same as a, except that MEPs were treated with Kyn or Kyn combined with the RUNX1 inhibitor Ro5-3335 in collagen medium for 7 d, and a c.f.u. analysis was performed. d, Mouse MEPs were treated with PBS or Kyn for 48 h. CUT&RUN-qPCR analysis was performed with an antibody to AhR and Runx1 P1 or P2 promotor-specific primers. e, NIH-3T3 cells were cotransfected with a Runx1 P1 promoter-luciferase reporter PGL4.10 and pCMV6-Ahr plasmid for 24 h. Cells were then treated with Kyn for another 48 h, followed by analysis of luciferase activity. f, The same as d. Runx1b and Runx1c expression levels were detected by RT–qPCR. g, MEPs from Ahr–/– mice were treated with Kyn for 48 h, and RUNX1 expression was determined by digital western blotting. h, Mouse MEPs were treated with Kyn or Kyn combined with SR1 for 48 h, and RUNX1 expression was determined by digital western blotting. i, Mouse MEPs were treated with PBS or Kyn for 72 h. CUT&RUN-qPCR analysis was performed with an antibody to RUNX1 and Itga2b or Gp1ba promotor-specific primers. j, NIH-3T3 cells were cotransfected with Itga2b or Gp1ba promoter-luciferase reporter PGL4.10 and pCMV6-Runx1 plasmid for 24 h. Cells were then treated with Kyn for another 48 h, followed by analysis of luciferase activity; Vec, vehicle; CMV, cytomegalovirus. In a–j, n = 3 independent experiments. Data were analyzed by two-tailed Student’s t-tests (a and f) or one-way ANOVA followed by a Bonferroni post hoc test (c–e, i and j). Data represent mean ± s.d.

Source data

Full size image

AhR upregulates SLC7A8 in MEPs to take up exogenous Kyn

Next, we investigated the manner by which AhR was activated in MEPs. Kyn is a typical endogenous ligand, which is produced by the catalysis of tryptophan via indoleamine 2,3-dioxygenase 1 (IDO1), IDO2 or tryptophan 2,3-dioxygenase 2 (TDO2)20,30. Surprisingly, although Ido1, Ido2 and Tdo2 were barely expressed in MEPs (Fig. 4a and Extended Data Fig. 4a), a high level of Kyn was found in the MEPs of tumor-bearing mice and a very low Kyn level was present in the MEPs of tumor-free mice (Fig. 4b and Extended Data Fig. 4b), suggesting that MEPs from tumor-bearing hosts might use exogenous Kyn to activate AhR. In fact, tumor tissues release large amounts of Kyn to the periphery, which can be used by MEPs to activate AhR. To achieve this, MEPs have to express transporters that transport Kyn across the plasma membrane. SLC1A5, SLC7A5, SLC7A8 and PAT4 have shown the ability to mediate Kyn transportation23,37. We found that although MEPs expressed the four transporters, only SLC7A8 was upregulated in the MEPs of tumor-bearing mice (Fig. 4c,d and Extended Data Fig. 4c,d). The use of Slc7a8 short interfering RNAs (siRNAs) or the inhibitor 2-amino-2-norbornanecarboxylic acid (BCH) blocked the entry of Kyn into MEPs (Fig. 4e and Extended Data Fig. 4e,f), concomitant with a reduction in AhR nuclear localization (Fig. 4f and Extended Data Fig. 4g). As expected, BCH treatment reversed the biased differentiation of MEPs (Extended Data Fig. 4h). However, knockdown of Slc1a5, Slc7a5 or Pat4 had minor effects on the transport of Kyn into MEPs (Fig. 4g and Extended Data Fig. 4i). Notably, in the in vitro assay, we found that the addition of Kyn markedly upregulated SLC7A8 expression in MEPs (Extended Data Fig. 4j), which was also verified in Kyn-treated mice (Fig. 4h), prompting us to postulate a positive feedback mechanism through which SLC7A8 was upregulated by Kyn-activated AhR. Indeed, SR1 effectively inhibited Kyn-induced SLC7A8 upregulation in MEPs both in vitro and in vivo (Fig. 4i and Extended Data Fig. 4k). A CUT&RUN assay also showed that AhR bound to the Slc7a8 promoter (Fig. 4j), and a luciferase assay verified that AhR promoted Slc7a8 expression (Fig. 4k and Extended Data Fig. 4l). Together, these data suggest that AhR transactivates the Kyn transporter gene and promotes the uptake of Kyn by MEP.

Fig. 4: AhR promotes the uptake of Kyn in MEPs by upregulating SLC7A8.
figure 4

a–d, C57BL/6J mice were inoculated with MC38 cells, and MEPs were isolated from control and tumor-bearing mice at week 4. RT–qPCR cycling threshold (Ct) values of Actb, Ido1, Ido2 and Tdo2 were determined (a), and intracellular Kyn levels of MEPs were determined by liquid chromatography–mass spectrometry (LC–MS; b). Expression of Slc1a5, Slc7a5, Slc7a8 and Pat4 in MEPs was detected by RT–qPCR (c), and MEPs were stained for SLC7A8 (d). Scale bar, 10 μm. e,f, MEPs transfected with negative control (NC) and Slc7a8 siRNA (si1 or si2) were treated with PBS or Kyn for 48 h, intracellular Kyn levels were determined (e), and cells were stained for AhR (f). Scale bar, 10 μm. g, NC, Slc1a5, slc7a5 and Pat4 siRNA-transfected mouse MEPs were treated with PBS or Kyn for 48 h, and intracellular Kyn levels were determined. h, C57BL/6J mice were treated with PBS or Kyn for 2 weeks, and isolated MEPs were stained for SLC7A8. Scale bar, 10 μm. i, C57BL/6J mice were treated with Kyn or Kyn combined with SR1 (50 μg i.p. every day) for 2 weeks, and isolated MEPs were stained for SLC7A8. Scale bar, 10 μm. j, Mouse MEPs were treated with PBS or Kyn for 48 h, and CUT&RUN-qPCR was performed with antibody to AhR and promoter-specific primers against Slc7a8. k, NIH-3T3 cells expressing Slc7a8 promoter-luciferase reporter PGL4.10 were cotransfected with pCMV6-Ahr. Cells were treated with Kyn for 48 h and analyzed by luciferase assay. In a–d, h and i, n = 10 mice. In e–g, j and k, n = 3 independent experiments. Data were analyzed by two-tailed Student’s t-tests (b–d and h) or one-way ANOVA followed by a Bonferroni post hoc test (e–g and i–k). Data represent mean ± s.d.

Source data

Full size image

Kyn–AhR–RUNX1 biases MEP differentiation in vivo

Next, we investigated whether our experimental results could be confirmed in vivo. In addition to AhR, which was highly active in the MEPs of tumor-bearing mice, we determined RUNX1 levels. We found that tumor-bearing mice displayed higher levels of RUNX1 in MEPs (Fig. 5a and Extended Data Fig. 5a). Injection of Kyn (i.p.) augmented RUNX1 expression in both healthy and tumor-bearing mice (Fig. 5b,c and Extended Data Fig. 5b). Given Kyn production via IDO1, we used 1-MT, an IDO1 inhibitor, to treat tumor-bearing mice. As expected, 1-MT effectively blocked AhR nuclear translocation (Fig. 5d and Extended Data Fig. 5c) and downregulated RUNX1 expression in MEPs (Fig. 5e and Extended Data Fig. 5d), concomitant with decreased Kyn levels in peripheral blood (Fig. 5f and Extended Data Fig. 5e). In addition, AhR inhibition by SR1 downregulated Cyp1a1, Cyp1b1 and Runx1 expression in MEPs in tumor-bearing mice (Fig. 5g,h and Extended Data Fig. 5f,g). However, RUNX1 downregulation was not detected in Ahr–/– tumor-bearing mice (Fig. 5i and Extended Data Fig. 5h). Given the role of RUNX1 in the regulation of megakaryocyte differentiation, we analyzed MEP differentiation under conditions of IDO1 or AhR inhibition. The results showed that both 1-MT and SR1 hindered MkP generation but favored EryP production in tumor-bearing mice (Fig. 5j,k and Extended Data Fig. 5i,j). Together, these results suggest that Kyn, via the AhR–RUNX1 pathway, leads to abnormal MEP differentiation, thrombocytosis and anemia in vivo.

Fig. 5: The Kyn–AhR–RUNX1 pathway regulates MEP differentiation into megakaryocytes in vivo.
figure 5

a, C57BL/6J mice were inoculated with MC38 cells, and MEPs were isolated and stained for RUNX1 at week 4. Scale bar, 10 μm. b, C57BL/6J mice were treated with PBS or Kyn for 2 weeks, and MEPs were isolated and stained for RUNX1; scale bar, 10 μm. c, The same as a, except that tumor-bearing mice were treated with PBS or Kyn for 2 weeks. Isolated MEPs were stained for RUNX1. Scale bar, 10 μm. d–f, The same as a, except that tumor-bearing mice were treated PBS or 1-MT (5 mg ml−1 in drinking water, 3–4 ml per mouse per day) for 2 weeks. MEPs were isolated and stained for AhR (d) and RUNX1 (e). Scale bar, 10 μm. Serum Kyn levels of PBS- or 1-MT-treated mice were measured by HPLC (f). g,h, The same as a, except that MC38 tumor-bearing mice were treated with PBS or SR1 for 2 weeks. Cyp1a1 and Cyp1b1 expression in MEPs was determined by RT–qPCR (g), and MEPs were stained for RUNX1 (h). i, Ahr–/– mice were inoculated with MC38 cells, tumor-bearing mice were treated with PBS or SR1, and MEPs were isolated and stained for RUNX1. Scale bar, 10 μm. j, The same as d, except that RBCs and PLTs or EryPs and MkPs in BM were analyzed. k, The same as g, except that RBCs and PLTs or EryPs and MkPs were analyzed. In a–h, j and k, n = 10 mice. In i, n = 6 mice. Data were analyzed by two-tailed Student’s t-tests (a–k) and represent mean ± s.d.

Source data

Full size image

Kyn–AhR–RUNX1 mediates thrombocytosis in humanized mice

Next, we investigated whether Kyn-biased mouse MEP differentiation could be translated into human MEPs. To this end, we isolated CD34+CD38+IL-3RA−CD45RA− MEPs from human umbilical cord blood (UCB; Extended Data Fig. 6a) and treated the cells with 10% collagen medium, which resulted in c.f.u. formation (Fig. 6a), consistent with previous reports38,39. After Kyn treatment, we found that Kyn facilitated differentiation toward MkPs in the c.f.u. assay (Fig. 6b), consistent with the differentiation of mouse MEPs by Kyn. Moreover, immunofluorescence staining showed that Kyn treatment resulted in the translocation of AhR into the nucleus of human MEPs (Fig. 6c) and upregulation of RUNX1 and SLC7A8 expression (Fig. 6d and Extended Data Fig. 6b). To further verify that the Kyn–AhR pathway is used by human MEPs to upregulate RUNX1 and SLC7A8 expression, we performed CUT&RUN analyses of Kyn-treated human MEPs. As expected, AhR-enriched DNA fragments from the RUNX1 P1 promoter were detected and enhanced by Kyn treatment (Fig. 6e). Luciferase assays showed that AhR bound to the RUNX1 P1 promoter and promoted reporter gene expression (Fig. 6f). Similar experiments showed that SLC7A8 expression was regulated by AhR (Extended Data Fig. 6c,d). Consistent with this, Kyn had little effect on the expression of RUNX1 in AhR-knockdown human MEPs (Fig. 6g and Extended Data Fig. 6e). In addition, Kyn did not affect SLC7A8 expression in these MEPs (Extended Data Fig. 6f), consistent with the positive feedback between AhR and SLC7A8 (refs. 23,40) by which SLC7A8 transports Kyn to activate AhR and the latter in turn promotes SLC7A8 expression. By conducting a c.f.u. assay, we found that although Kyn facilitated differentiation toward MkPs, Ahr knockdown switched differentiation toward EryPs (Fig. 6h). To validate the in vitro human results in vivo, we constructed a humanized mouse model by transplanting CD34+ UBCs into severely immunodeficient NCG-X mice, which lack the Kit gene and support human HSC transplantation without radiation41,42. These humanized mice were treated with Kyn or Kyn plus SR1 in combination with chlodronate (Extended Data Fig. 6g), which can delete macrophages to avoid phagocytosis of the generated human PLTs and RBCs43,44. As expected, Kyn treatment increased and decreased the amount of human PLTs and RBCs, respectively, which was blunted by SR1 (Extended Data Fig. 6h). By inoculating HCT116 human colon cancer cells into mice with treatment (Extended Data Fig. 6i), we found that the amounts of PLTs and RBCs also increased and decreased, respectively, in conjunction with tumor growth and elevated serum Kyn levels, which were also blunted by SR1 treatment (Fig. 6i,j). Together, these results suggest that the AhR–RUNX1 pathway triggered by Kyn decreases erythrocyte counts but increases PLT counts in cancer-bearing hosts.

Fig. 6: Thrombocytosis and erythropenia are induced by Kyn in humanized mice.
figure 6

a, Representative images of human MEP-derived colonies. Scale bar, 100 μm. b, MEPs from human UCB were cultured in collagen-based medium and treated with PBS or Kyn for 12 d, followed by c.f.u. analysis. c,d, Human MEPs were treated with PBS or Kyn for 48 h, and cells were stained for AhR (c) or RUNX1 (d). Scale bar, 10 μm. e, CUT&RUN-qPCR analysis was performed with anti-AhR and RUNX1 P1 promoter-specific primers. f, HEK 293T cells were cotransfected with RUNX1 P1 promoter-luciferase reporter PGL4.10 and AhR-overexpressing plasmid for 24 h. Cells were then treated with Kyn for another 48 h, followed by analysis of luciferase activity. g, Human MEPs transfected with short hairpin RNA (shRNA) targeting Ahr (Sh1 and Sh2) were treated with Kyn for 48 h and stained for RUNX1. Scale bar, 10 μm. h, Human MEPs were transfected with scrambled shRNA and shRNA targeting Ahr (Sh1 and Sh2), cultured in collagen medium and treated with Kyn for c.f.u. analysis at day 12. i, HCT116 tumor-bearing humanized mice were treated with PBS and SR1 for 2 weeks, and human-derived RBCs and PLTs in peripheral blood were analyzed. j, The same as i, except that serum Kyn levels of healthy control and tumor-bearing humanized mice were measured by HPLC. In i and j, n = 6 mice. In b–h, n = 3 independent experiments. Data were analyzed by two-tailed Student’s t-tests (b–d and j) or one-way ANOVA followed by a Bonferroni post hoc test (e–i). Data represent mean ± s.d.

Source data

Full size image

Kyn induces abnormal MEP differentiation in individuals with cancer

Finally, we attempted to validate our findings by using clinical samples. BM cells were collected from individuals with T cell acute lymphoblastic leukemia (T-ALL; n = 12) and B cell acute lymphoblastic leukemia (B-ALL; n = 12) who had not received treatment before or who had received HSC transplantation (HSCT; n = 6 for each type). Participant information is presented in Supplementary Table 5. Numbers of both RBCs and PLTs of newly diagnosed individuals were lower than those of individuals who had received HSCT (Extended Data Fig. 7a). Notably, although the numbers of MEPs decreased in newly diagnosed individuals compared to that observed in individuals who had received HSCT (Extended Data Fig. 7b), the ratio of PLTs to RBCs was higher before HSCT (Extended Data Fig. 7c). Using isolated MEPs to perform a c.f.u. assay, we found that MEPs from untreated individuals biased the differentiation to megakaryocytes more than those from HSCT-treated individuals (Fig. 7a); however, this was disrupted by the addition of SR1 (Fig. 7b). In line with this finding, Kyn levels were much lower in the BM of individuals who had received HSCT (Fig. 7c), concomitant with the downregulation of IDO1 in BM cells (Fig. 7d). To validate that the effect of Kyn is mediated by AhR, we determined AhR activity in MEPs from individuals with T-ALL or B-ALL. As expected, we found that HSCT resulted in the downregulation of AhR, CYP1A1 and CYP1B1 expression in MEPs, concomitant with decreased nuclear localization of AhR (Fig. 7e and Extended Data Fig. 7d), whereas the addition of Kyn blocked the HSCT-induced effects (Fig. 7f and Extended Data Fig. 7e). In line with this finding, RUNX1 expression was downregulated in MEPs of individuals who had received HSCT (Fig. 7g). In addition, expression of the gene encoding the Kyn transporter SLC7A8 was downregulated in MEPs (Extended Data Fig. 7f). However, Kyn treatment upregulated RUNX1 and SLC7A8 expression in MEPs from individuals who had received HSCT (Fig. 7h and Extended Data Fig. 7g) and biased MEP differentiation into megakaryocytes (Fig. 7i). In addition, BM samples were collected from individuals with lymphoma (n = 6) without tumor cell infiltration. We found that more MkPs and fewer EryPs were present in the BM of individuals with lymphoma than in the BM of healthy donors (n = 8; Extended Data Fig. 7h,i). Consistently, BM Kyn levels were higher (Extended Data Fig. 7j), and the differentiation of MEPs was more biased toward megakaryocytes than observed in healthy donors (Extended Data Fig. 7k). In addition, the expression of AhR, CYP1A1, CYP1B1, RUNX1 and SLC7A8 was upregulated in MEPs of individuals with lymphoma (Extended Data Fig. 7l–n), concomitant with enhanced nuclear localization of AhR (Extended Data Fig. 7o). Together, these results suggest that the Kyn–AhR–RUNX1 pathway also mediates abnormal differentiation of MEPs in individuals with cancer (Extended Data Fig. 7p).

Fig. 7: AhR mediates abnormal MEP differentiation in individuals with cancer.
figure 7

a, MEPs were isolated from the BM of individuals with T-ALL (n = 12; without HSCT (WH) n = 6, after HSCT (AH) n = 6) and B-ALL (n = 12; without HSCT n = 6, with HSCT n = 6), followed by c.f.u. analysis; c.f.u.-E, erythroid only; c.f.u.-Mk, megakaryocyte only. b, MEPs from individuals without HSCT were treated with PBS or SR1, followed by c.f.u. analysis. c, BM Kyn levels in individuals with cancer were detected by HPLC. d, BM-derived mononuclear cells from individuals with cancer were isolated, and IDO1 expression was detected by RT–qPCR. e, MEPs were isolated from the BM of individuals with T-ALL and B-ALL and were stained for AhR. Scale bar, 10 μm. f, MEPs from individuals who had received HSCT were treated with Kyn for 48 h, and cells were stained for AhR. Scale bar, 10 μm. g, MEPs were isolated from the BM of individuals with T-ALL and B-ALL, and RUNX1 expression was detected by RT–qPCR. h, MEPs isolated from individuals who had received HSCT were treated with Kyn for 24 h, and RUNX1 expression was detected by RT–qPCR. i, MEPs isolated from individuals after HSCT (T-ALL, n = 6; B-ALL, n = 6) were treated with PBS or Kyn in collagen medium, followed by c.f.u. analysis. Data were analyzed by two-tailed Student’s t-tests (a–i) and represent mean ± s.d.

Source data

Full size image

Discussion

Thrombocytosis and anemia are common in individuals with cancer1,45. Here we identified a tumor release factor-mediated pathway that causes thrombocytosis and anemia. During tumor growth, large amounts of Kyn are released to the periphery, enter MEPs and bias their differentiation into megakaryocytes by activating AhR. Tryptophan is an essential amino acid widely used by tumor cells, at least partially due to the Myc oncoprotein, which can induce the expression of the tryptophan transporters SLC7A5 and SLC1A5 (ref. 46). Thus, tumor cells take up and catalyze tryptophan for Kyn production via IDOs. The released Kyn then activates AhR in MEPs. In line with our study, high levels of the AhR ligand TCDD increase the number of PLTs in the blood of exposed individuals47 and Kyn-induced anemia during inflammation48.

MEPs are bipotential with a fate decision to the megakaryocyte versus erythrocyte lineage, which, however, is not completely understood. In this study, we identified that exogenous Kyn is critical for regulating MEP fate decisions in individuals with cancer. However, MEPs do not seem to generate endogenous Kyn, in which the tryptophan-degrading enzymes are rarely detected, raising the question of how MEPs effectively take up Kyn. We demonstrated that SLC7A8 is the main transporter that transports Kyn into MEPs. Notably, this uptake pathway is regulated by a positive feedback loop, because Kyn-activated AhR upregulates SLC7A8 expression in MEPs. Thus, activated AhR is present in MEPs in individuals with cancer, which provides new insights into MEP fate decisions. After activation by GATA1, the transcription factor KLF1/EKLF promotes erythroid differentiation and suppresses megakaryocytic differentiation. By contrast, FLI1 directly activates genes, such as ITGA2B, GP1BA and GPIX, for megakaryocyte maturation. Notwithstanding this transcriptional regulation, Kyn–AhR was shown to transactivate RUNX1, another key transcription factor for megakaryocytic lineage differentiation. By performing single-cell c.f.u. assays, we clearly demonstrate that exogenous Kyn biases MEP differentiation to the megakaryocyte lineage through the AhR–RUNX1 pathway, which is likely to play an instructive role. This is unlike thrombopoietin (TPO) and erythropoietin (EPO), which play permissive roles in megakaryopoiesis and erythropoiesis, respectively. In addition, AhR seems to have a permissive effect, considering its regulating the polyploidization of megakaryocytes18. Whether Kyn prejudicially targets MEPs rather than other hematopoietic progenitors due to the differential expression of SLC7A8 is worthy of investigation in the future.

The present study has several important clinical implications. Kyn concentration may range from 0.1 to 0.5 μM in normal blood but increases to up to 1 μM in individuals with cancer, which has pathophysiological effects by activating AhR of target cells23. To recapitulate this, we used a humanized mouse model to conduct MEP differentiation experiments with or without Kyn treatment. Moreover, we verified the Kyn–AhR–RUNX1 pathway in MEPs freshly isolated from individuals with leukemia. It is feasible to obtain BM from individuals with leukemia; however, we cannot obtain BM from individuals with solid tumors because of ethical issues. Nevertheless, leukemic and solid tumor cells highly mobilize the Kyn–AhR pathway, leading to the fidelity of MEP differentiation among individuals with various tumors. Apart from individuals with cancer, our findings might be useful in explaining thrombocytosis in individuals without cancer with the JAK2-V617F mutation49. This mutation leads to the activation of JAK2. However, activated JAK2 may result in the upregulation of IDO1 expression via activated STAT1/STAT2, thus producing more endogenous Kyn to bias MEP differentiation into megakaryocytes. Whether this JAK2 mutation drives thrombocytosis through the IDO–Kyn pathway is currently under study. Overall, this study identified that AhR activation drives abnormal erythrocyte and PLT differentiation in individuals with cancer, and targeting the Kyn–AhR pathway could be explored to treat anemia and thrombocytosis in individuals with cancer.

Methods

Cell lines

Mouse tumor cell lines MC38 (colon cancer) and E0771 (breast cancer), the mouse embryonic fibroblast cell line NIH-3T3, the human tumor cell line HCT116 (colon cancer) and the human embryonic kidney cell line HEK 293T were purchased from the China Center for Type Culture Collection and were cultured in RPMI 1640 medium (Gibco) or DMEM medium (Gibco) with 10% fetal bovine serum (Gibco). Before the study, the cells were examined for the presence of Mycoplasma, checked for interspecies contamination and validated using isoenzyme and short tandem repeat studies in the Cell Resource Centre of Peking Union Medical College.

Mouse transplant tumor models

C57BL/6J mice were obtained from the Chinese Academy of Medical Science’s Center of Medical Experimental Animals. NCG-c-Kit-Cas9-TM (NCG-X) mice were purchased from Gempharmatech. tdTomato-transgenic mice were purchased from Shanghai Model Organisms Center. Ahr–/–C57BL/6J mice were gifted by J. Yan (Third Military Medical University). The experiments used 6- to 8-week-old mice, except for NCG-X mice, which were 4 to 5 weeks old. Animals were maintained in pathogen-free environments at 20–21 °C with 60–70% relative humidity on a 12-h light/12-h dark cycle in the Chinese Academy of Medical Science’s animal facilities. The Chinese Academy of Medical Science’s Animal Care and Use Committee gave its approval for all investigations that used mice.

For colorectal carcinoma models, WT, Ahr–/– or tdTomato-transgenic mice were inoculated subcutaneously with 5 × 105 MC38 cells into the right flank. Humanized mice were inoculated subcutaneously with 5 × 106 HCT116 cells into the right flank. For breast carcinoma models, WT, Ahr–/– or tdTomato-transgenic mice were implanted with 5 × 105 E0771 cells into the inguinal mammary gland (sample size of n = 6 or 10).

Human information and samples

Peripheral blood samples were obtained from the First Affiliated Hospital of Zhengzhou University. UCB samples were obtained from Women and Children’s Hospital, Xiamen University. Human BM samples were obtained from Peking University People’s Hospital. Ethical permission was granted by the Clinical Trial Ethics Committee of First Affiliated Hospital of Zhengzhou University (2019-KY-256), Women and Children’s Hospital, Xiamen University (KY-2019-073) and Peking University People’s Hospital (NKRDP2021005-EC-2). Every participant signed a written informed consent form for their participation in the trial. Detailed information is presented in Supplementary Tables 1–6.

Cell preparation and in vitro treatment

For mouse MEPs, cells were isolated from BM by flow cytometry. Except for the c.f.u. assay, cells were cultured in StemSpan SFEM II (STEMCELL Technologies) with 50 μM β-mercaptoethanol and mouse cytokines (2.5 U ml−1 EPO, 10 ng ml−1 interleukin-3 (IL-3), 10 ng ml−1 IL-6, 25 ng ml−1 stem cell factor (SCF), 50 ng ml−1 TPO, 10 ng ml−1 IL-11 and 25 ng ml−1 FLT3L). Human MEPs were isolated from UCB or BM by flow cytometry. Except for c.f.u. assays, human MEPs were cultured in StemSpan SFEM II containing human cytokines (2.5 U ml−1 EPO, 10 ng ml−1 IL-3, 10 ng ml−1 IL-6, 100 ng ml−1 SCF, 25 ng ml−1 TPO, 10 ng ml−1 IL-11 and 100 ng ml−1 FLT3L). Mouse MEPs were treated with PBS or Kyn (5 μM), Kyn or Kyn combined with Ro5-3335 (2 μM) or SR1 (1 μM). Human MEPs were treated with PBS or Kyn (5 μM). After the indicated amounts of time, mouse or human MEPs were collected for analysis. Detailed reagent information is available in Supplementary Table 7.

Animal in vivo treatment

C57BL/6J mice were inoculated with MC38 or E0771 cells, and RBCs and PLTs were counted once a week with a Sysmex XN-1000V until tumors were approximately 2,000 mm3. Different doses of Kyn (100 to 400 μg) were injected (i.p.) into C57BL/6J mice, and serum Kyn levels were detected at 1, 3, 8, 12 and 24 h by HPLC to obtain time–concentration curves. C57BL/6J mice were injected with Kyn (i.p.; 400 μg every 12 h) for 14 d, and blood samples were taken for RBC and PLT counting. BM cells were used for EryP and MkP analysis, and MEPs were isolated and stained for RUNX1. MC38 or E0771 tumor-bearing mice were treated with PBS or Kyn, Kyn or Kyn combined with SR1 (i.p.; 50 μg every day), PBS or SR1 or PBS or 1-MT (5 mg ml−1 in drinking water, 3–4 ml per mouse per day) for 2 weeks, and the numbers of RBCs, PLTs, EryPs and MkPs were quantified. MEPs were isolated for AhR, SLC7A8 or RUNX1 staining.

For the in vivo study of humanized mice, human CD34+ cells were isolated from human UCB with a CD34 microbeads kit (Miltenyi Biotec). NCG-X mice were transferred (intravenously) with 5 × 104 CD34+ cells, and HCT116 cells were inoculated subcutaneously in these mice after 14 weeks. Concurrently, mice were injected (i.p.) with recombinant human EPO (30 IU per mouse), recombinant human TPO (250 ng per mouse) and recombinant human IL-3 (25 ng per mouse) every 3 d to promote RBC and PLT production. Two weeks later, mice were treated with chlorophosphate liposome (i.p.; 200 μl per mouse) every 2 d three times. Peripheral human-derived RBCs and PLTs of the treated mice were detected by flow cytometry at week 18. For Kyn and SR1 treatment, humanized NCG-X mice were treated with Kyn or Kyn combined with SR1, as previously described, after CD34+ cells were transferred at 14 weeks. Similarly, recombinant human EPO, recombinant human TPO, recombinant human IL-3 and chlorophosphate liposomes were administrated for the indicated lengths of time, and peripheral human-derived RBCs and PLTs were detected by flow cytometry.

For BM transplantation, tdTomato-transgenic mice were exposed to a split dose of 9 Gy by X-ray with 4-h intervals. Fresh BM cells (1 × 106) from WT or Ahr–/– mice were transplanted (intravenously) into each irradiated recipient. After 6 weeks, these recipient mice were injected with MC38 or E0771 cells. All donor-derived tdTomato– RBCs or PLTs were further counted by flow cytometry at the indicated times after MC38 or E0771 inoculation.

Flow cytometry and cell sorting

Single-cell suspensions from mouse BM or human UCB or BM were prepared. All samples were stained with a Zombie Aqua Fixable Viability kit (Biolegend) for 15 min at room temperature before they were stained with antibodies. All antibodies were diluted 1:100 with cell sorting buffer. An Invitrogen Attune NxT flow cytometer and Invitrogen Attune NxT software were used for data collection. Cell sorting was performed with a Sony MA900 flow cytometer. Data were analyzed with FlowJo software. For detailed antibody information, please see Supplementary Table 8.

Digital western blotting

MEPs isolated from C57BL/6J mice were treated with PBS, Kyn or Kyn combined with SR1 for 48 h, and MEPs isolated from Ahr–/– mice were treated with PBS or Kyn for 48 h. Equal numbers of MEPs from each group were collected, lysed in M2 lysis buffer and sonicated. Proteins were probed with different antibodies (Supplementary Table 8) by using a Digital Western Blot system (Jess, Bio-Techne). Digital western blot data were captured using Compass for SW.

Immunofluorescence

Cells were fixed in 4% paraformaldehyde and permeabilized with 0.2% Triton X-100 on slides. Fixed cells were blocked in 5% bovine serum albumin and incubated with different antibodies (1:200 dilution; Supplementary Table 8) at 4 °C overnight, and cells were then washed and incubated with secondary antibodies (1:500 dilution) for 1 h at room temperature. Slides were counterstained with DAPI. Images were collected on a Nikon A1 HD25 microscope using NIS-Elements AR software. Immunofluorescence intensity was analyzed with ImageJ v1.52 software. Mean fluorescence intensity within the AhR+, RUNX1+ or SLC7A8+ region of interest was calculated from 15 random cells per slide.

Transfection of siRNAs into mouse MEPs

Mouse MEPs were resuspended in 20 μl of cell-specific nuclear transfection solution (P3 Primary Cell kit, Lonza) with 50 nM Slc1a5, Slc7a5, Slc7a8 or Pat4 siRNA and transfected with a Lonza 4D X-Unit with the DK-100 program. After transfection, cells were cultured with the appropriate aforementioned medium and cytokines. siRNA sequences are provided in Supplementary Table 9.

Stable transfection of shRNAs

Human MEPs were infected with lentivirus for stable transfection. Lentiviruses containing pLKO.1-AhR-SH green fluorescent protein (GFP) or scrambled control were packaged in HEK 293T cells with psPAX2 (Addgene, 12260) and pMD2.G packaging plasmid (Addgene, 12259). Human MEPs were then infected by adding a ratio of 1:2 (vol:vol) viral supernatant:fresh medium containing 6 μg ml−1 polybrene (Sigma) and were centrifuged at 1,200g for 60 min at 32 °C. Four hours after spinning inoculation, the supernatant was replaced with fresh medium and cultured for another 20 h. Cells with GFP fluorescence were then isolated by flow cytometry cell sorting. shRNA sequences are provided in Supplementary Table 9.

Luciferase assays

One hundred nanograms of Renilla luciferase plasmid pRL-SV40, 1 μg of firefly luciferase plasmid pGL4.10-Runx1 P1 promoter-luciferase or pGL4.10-Slc7a8 promoter-luciferase and 1 μg of pCMV6-Ahr plasmid were transfected into NIH-3T3 cells for 12 h. Transfected cells were then treated with 100 μM Kyn for 48 h. One hundred nanograms of Renilla luciferase plasmid pRL-SV40, 1 μg of firefly luciferase plasmid pGL4.10-Itgb2b promoter-luciferase, pGL4.10-Gp1ba promoter-luciferase or pGL4.10-Klf1 promoter-luciferase and 1 μg of pCMV6-Runx1 plasmid were transfected into NIH-3T3 cells for 12 h. Transfected cells were then treated with 100 μM Kyn for 48 h. One hundred nanograms of Renilla luciferase plasmid pRL-SV40, 1 μg of firefly luciferase plasmid pGL4.10-RUNX1 promoter-luciferase or pGL4.10-SLC7A8 promoter-luciferase and 1 μg of pCMV6-AhR plasmid were transfected into HEK 293T cells for 12 h, and transfected cells were treated with 100 μM Kyn for 48 h. Cell lysis was performed with a Dual Luciferase Reporter Assay (Promega), and the ratio of firefly luciferase activity to Renilla luciferase activity was analyzed and calculated using GloMax Multi Plus (Promega).

RT–qPCR

TRIzol (Invitrogen) was used to extract total RNA from cells. A high-capacity cDNA reverse transcription kit (Applied Biosystems) was used to transcribe total RNA into cDNA. The primer sequences are shown in Supplementary Table 9. An ABI QuantStudio 3 (Applied Biosystems) and QuantStudio Design & Analysis were used for RT–qPCR. Values are shown as mean ± s.d. from three independent experiments.

CUT&RUN-qPCR assay

CUT&RUN-qPCR was performed by using a CUT&RUN assay kit (Cell Signaling Technology) according to the manufacturer’s protocol. Briefly, mouse MEPs were treated with Kyn for 48 or 72 h. Cells were permeabilized with digitonin, and antibody to AhR or RUNX1 and a Protein A-Protein G-Micrococcal Nuclease (pAG-MNase) were used to isolate specific protein–DNA complexes. DNA fragments were collected using DNA spin columns (Zymo Research), and enriched DNA was identified and quantified using qPCR. The primer sequences are shown in Supplementary Table 9.

HPLC analysis of serum Kyn levels

Serum Kyn determination was performed by Dionex UltiMate3000. Mouse serum was mixed with 10% trichloroacetic acid (1:1) for 30 min, and the mixture was centrifuged at 14,000g for 10 min at 4 °C. Then, 20 µl of supernatant was injected into the HPLC instrument for quantification. The mobile phase consisted of 15 mM sodium acetate containing 8.0% acetonitrile (pH 4.0), which was pumped at a flow rate of 1.0 ml min−1. The analytical column was a C18 chromatographic column (Thermo Fisher Scientific, 250 mm × 4.6 mm, 5 μm) and kept at 30 °C. Kyn was measured at a wavelength of 360 nm.

LC–MS analysis of MEP intracellular Kyn level

Intracellular Kyn determination was performed with an Agilent 6495. Mouse MEPs were collected and washed twice in PBS and lysed with 50 μl of 80% (vol/vol) methanol/water and centrifuged at 15,000g for 15 min at 4 °C. The supernatant was collected for direct detection or was stored at −80 °C. An Agilent 6495 triple quadrupole LC–MS system and C18 chromatographic column (Thermo Fisher Scientific, 100 mm × 2.1 mm, 1.7 μm) were used. For LC separation, gradient elution was performed using 0.1% formic acid as solvent A and pure acetonitrile as solvent B. The flow rate was 0.3 ml min−1, and the injection volume was 10 μl. The following gradient program was used: 0–1 min, 95% A; 1–6 min, 95% A to 5% A; 6–7 min, 5% A; 7–7.2 min, 5% A to 95% A and 7.2–11 min, 95% A. The total run time was 11 min for each sample, and the data were collected by Data Acquisition.

C.f.u. assays

Mouse MEPs were cultured in collagen-based medium with β-mercaptoethanol and mouse cytokines EPO, TPO, SCF, IL-3, IL-6 and IL-11, as described above. Cells were cultured with the addition of 2 μM Ro5-3335 (MedChemExpress), 100 μM BCH (MedChemExpress), 5 μM Kyn or PBS for 7 d. Human MEPs were cultured in collagen-based medium with recombinant human EPO, TPO, SCF, IL-3, IL-6 and IL-11, as described above. Cells were cultured with the addition of 5 μM Kyn or PBS for 12 d. A double-chamber slide kit (STEMCELL Technologies) was used for dual megakaryocyte/erythroid c.f.u. assays. Collagen-based medium in double-chamber slides was dehydrated and fixed according to the manufacturer’s protocol. All antibody incubations were performed for 30 min at room temperature. Cultures were rehydrated in Tris/NaCl buffer containing 10% goat and 2.5% mouse serum. For mouse and human c.f.u. megakaryocyte staining, we used rabbit anti-CD41 (1:100 dilution) and then goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody AF488 (1:500 dilution). For mouse and human c.f.u. erythroid staining, we used mouse anti-CD71 (1:100 dilution) and then donkey anti-mouse IgG (H + L) highly cross-adsorbed secondary antibody AF594 (1:500 dilution). Colonies were scored based on CD41 and CD71 staining as megakaryocyte only, erythroid only or megakaryocyte plus erythroid. The c.f.u. values were observed with a Leica DMi8 by using LAS X Life Science Software.

Quantification and statistical analysis

All experiments were performed with at least three biological repeats. Results are expressed as mean ± s.d. or mean ± s.e.m., as indicated, and were analyzed by one-way ANOVA followed by a Bonferroni post hoc test or by a two-tailed Student’s t-test. A P value of <0.05 was considered statistically significant. Analyses were conducted using GraphPad 8.0 software.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.