The E3 ligase TRIM31 regulates hematopoietic stem cell homeostasis and MLL-AF9 leukemia

Abstract

Hematopoietic stem cells (HSC) are kept in a quiescent state to maintain their self-renewal capacity. Proper regulation of cyclin-dependent kinases (CDK) and cyclin proteins is critical for the maintenance of HSC homeostasis. Here, we found that the E3 ligase, TRIM31, regulates HSC homeostasis and leukemia through the accumulation of CDK8. TRIM31 deficiency promotes hematopoietic stem and progenitor cell proliferation and long-term HSC exhaustion. Serial competitive transplantation assays showed that TRIM31-deficient HSC exhibit impaired reconstitution ability. TRIM31 loss led to a lower rate of survival of mice under conditions of stress (5-fluorouracil administration), which was correlated with a lower number of hematopoietic stem and progenitor cells. In a murine model of acute myeloid leukemia, the initiation of leukemia was significantly accelerated upon TRIM31 deletion. Mechanistically, we found that ubiquitin-mediated degradation of CDK8 was impaired by TRIM31 deletion, which further induced transcriptional expression of PBX1 and cyclin D1. Taken together, these findings reveal the function of TRIM31 in the regulation of HSC homeostasis and leukemia initiation, and indicate the physiological importance of TRIM31 in the early stage of the development of leukemia.

Introduction

Under steady conditions, hematopoietic stem cells (HSC) are in a quiescent state to prevent exhaustion of the cells and to restrict the occurrence of replication-associated mutations.¹ An uncontrolled self-renewal program disrupts stem cell maintenance, and over-proliferation of stem cells often leads to HSC exhaustion and leukemia.² Quality control of the cell cycle is achieved through cell-intrinsic regulators, including transcription factors, signal transducers, cell-cycle inhibitors, surface receptors, and cell-extrinsic factors, such as the bone marrow niche and cytokines. Notably, cell-intrinsic regulators can be manipulated by the ubiquitin-proteasome degradation system at the protein level.

Ubiquitin E3 ligases play a critical role in the hematopoietic system. Loss of E3 ligases (including c-Cbl, Itch and SCFF^bxw7) may affect HSC quiescence and lead to stem cell expansion, myeloid proliferative disorders and acute myeloid leukemia (AML).^3-5 As E3 ligases, TRIM proteins are widely involved in various cell processes, such as cell proliferation, differentiation, development and apoptosis. Some of them, notably TRIM19 and TRIM33, are vital for the maintenance and function of HSC.^6,7 Other members of the TRIM family, such as TRIM13 and TRIM24, have also been reported to be essential in various hematologic malignancies.^8,9 TRIM31, another TRIM family protein, is crucial for intracellular signaling, innate immunity, autophagy and carcinogenesis. TRIM31 regulates the immune response through MAVS and SYK, suppresses NLRP3-induced inflammasome activation and promotes autophagy in intestinal cells.^10-13 TRIM31 not only contributes significantly to cerebral ischemic injury by promoting degradation of TIGAR,¹⁴ but also contributes to hypertensive nephropathy by promoting degradation of MAP3K7.¹⁵ Through targeting Rhbdf2 in mouse hepatocytes, Trim31 alleviates non-alcoholic fatty liver disease.¹⁶ In cases of aggressive AML, the mixed lineage leukemia 1 protein (MLL) has frequently been found to be fused with a partner (e.g., AF9).¹⁷ Various investigations of MLL-AF9-induced AML have studied transcriptional regulators (e.g., HoxA9) and epigenetic modulators (e.g., Dot1L);^18,19 however, the study of post-translational regulation in MLL-AF9-mediated leukemogenesis remains scarce.

Cyclin-dependent kinase 8 (CDK8) is a cell-intrinsic regulator that functions conservatively in transcription, as a part of the CDK8-mediator complex.²⁰ The CDK8-mediator complex functions through releasing RNA polymerase II (RNAPII) from a paused state to start transcription.²¹ The CDK8-mediator complex also functions as a tethering module in enhancement of gene transcription regulated by noncoding RNA.²² During the innate immune response and inflammation, the mediator-associated kinase CDK8 plays a role as a negative regulator of interleukin-10.²³ CDK8 has also been reported to act as an oncogene in both colon cancer^24,25 and melanoma.²⁶ In addition, the growth of AML cells can be inhibited by repressing CDK8 activity with small-molecule drugs.²⁷

In this study, we found that TRIM31 deletion leads to HSC proliferation and functional decline, while accelerating the initiation of leukemia. Mechanistically, TRIM31 functions as an E3 ligase of CDK8, and its deletion causes blockage of ubiquitin-mediated CDK8 degradation. The accumulation of CDK8 leads to enhanced expression of PBX1 and cyclin D1. Potentially, an increase in E3 ligase activity of TRIM31 could be used as an anti-leukemia target because TRIM31 deletion promotes the initiation and development of MLL-AF9-related AML.

Methods

Mice

TRIM31^+/- mice¹² were a kind gift from Professor Chengjiang Gao of Shandong University and were produced by micro-injecting transcription activator-like effector nuclease (TALEN) mRNA into fertilized eggs of mice with a C57BL/6 background. The TRIM31^-/- mice were genotyped by sequencing polymerase chain reaction (PCR) fragments (250 bp) in the TALEN-targeting region, which was amplified from isolated genomic DNA from the mouse tail using the following primers: forward 5’-GGCCTTGGATTTCTGTACTTTCACATC-3’ and reverse 5’-TGGGCCTGAACGTATTCTTATTCACAG-3’. Wild-type (WT) and TRIM31^-/- mice aged 8-12 weeks were used in the experiments, except for those of marked age. The recipient mice, which were used in the competitive transplantation assays, were either CD45.1 mice or CD45.1/CD45.2 mice with a C57BL/6 background. The Animal Care and Ethics Committee at Jinan University approved all animal experiments in our study.

Flow cytometry and cell sorting

Bone marrow cells were freshly isolated from mice and incubated in a lineage cocktail of antibodies targeting CD4 (1:100, RM4-5), CD8 (1:100, 53-6.7), Ter-119 (1:100, TER-119), CD11b (1:150, M1/70), Gr-1 (1:150, RB6-8C5) and B220 (1:100, RA3-6B2) for 30 min. The cells were then washed and incubated in an antibody mix containing antibodies against CD34 (1:100, RAM34), and CD48 (1:200, HM48-1), CD45.2 (1:100, 104), IL-7R (1:100, A7R34), Flt3 (1:100, A2F10), CD150 (1:100, TC15-12F12.2), CD45.1 (1:100, A20), Sca1 (1:100, E13-161.7), c-Kit (1:100, ACK2), and CD16/32 (1:100, 93) and strep-tavidin. All antibodies were monoclonal and purchased from BD Biosciences. For cell sorting, the bone marrow cells were enriched with anti-antigen-presenting cell microbeads (Miltenyi Biotec) and then stained with antibodies for surface markers. Cell analysis and data acquisition were performed using an LSR Fortessa (BD Biosciences) cell analyzer, and cell sorting was carried out using an Aria 3 cell sorter (BD Biosciences). The data were analyzed using FlowJo software.

Cytokine stimulation of hematopoietic stem cells

HSC were sorted from the bone marrow of WT mice via flow cytometry, and then plated in SFEM medium supplemented with stem cell factor (10 ng/mL; Pepro Tech), thrombopoietin (10 ng/mL; Pepro Tech), interleukin-3 (10 ng/mL; Pepro Tech) and 100 U/mL penicillin/streptomycin. Cells were collected at 0 h, 24 h and 48 h time-points for the experiment.

MLL-AF9-mediated leukemia transformation assay

c-Kit⁺ cells were enriched from the bone marrow of WT and TRIM31^-/- mice and then cultured in Iscove-modified Dulbecco medium containing 10% fetal bovine serum, 50 ng/mL stem cell factor, 10 ng/mL interleukin-3 and 10 ng/mL interleukin-6 overnight to stimulate cell proliferation. The next day, cells were transduced with retrovirus encoding MLL-AF9. After 72 h, cells were harvested and GFP⁺ cells were sorted by an Aria 3 Sorter. GFP⁺ cells (5x10⁴) were transplanted together with 5x10⁵ bone marrow cells into lethally irradiated recipients. Mice were monitored for MLL-AF9 AML development. Five thousand GFP⁺ cells were sorted and plated in colony-forming units assay using MethoCult M3434 (STEMCELL Technologies) media for 10 days before colony counting.

Results

TRIM31 deletion impairs hematopoietic stem and progenitor cell homeostasis

E3 ligase regulates the cell cycle efficiently through precise control of cell cycle factors (e.g., CDK and cyclin proteins) at the protein level. An increasing number of studies have shown that the E3 ligase function of TRIM proteins plays an essential role in tumorigenesis;²⁸ however, its role in the hematopoietic system and leukemia initiation has rarely been studied. Real-time PCR revealed that the mRNA level of TRIM31 is reduced significantly after stimulation of HSC proliferation and differentiation (Online Supplementary Figure S1A), which indicates high expression of TRIM31 may play a potential role in the cell cycle control of HSC. To determine the function of TRIM31 in regulating HSC, TRIM31 knockout mice were created¹² (Online Supplementary Figure S1B, C) and analyzed by using fluorescence activated cell sorting (FACS). In the hematopoietic system, the cellularity of TRIM31^-/- bone marrow and Lin^- cells remained similar to that of WT mice (Online Supplementary Figures S1D-F); however, the number of LK (Lin^-Sca-1^-c-Kit⁺) progenitor cells decreased significantly (Online Supplementary Figure S1G, H). In the bone marrow of TRIM31^-/- mice, FACS analysis revealed higher numbers of common lymphoid progenitors (CLP) and lower numbers of common myeloid progenitor (CMP), but no variation in the number of granulocyte/monocyte progenitors (GMP) (Figure 1A, B; Online Supplementary Figure S1G, I-K). A more than 2-fold increase was also shown in both the absolute number and frequency of LSK (Lin^-

Figure 1.TRIM31 deficiency induces hematopoietic stem cell expansion and long-term hematopoietic stem cell exhaustion. (A) FACS plots of granulocyte/monocyte progenitor (GMP; CD34⁺CD16/32⁺LKS-), common myeloid progenitor (CMP; CD34⁺CD16/32-LKS-) and megakaryocyte/erythroid progenitor (MEP; CD34CD16/32LKS) populations in wildtype (WT) and TRIM31^-/- mice (n=5-6 per group) with the frequencies indicated. (B) The numbers of GMP, MEP and CMP in WT and TRIM31^-/- mice (n=5-6 per group). (C) Representative FACS plots of LSK cells (Lin-Sca-1⁺c-Kit⁺), long-term hematopoietic stem cells (LT-HSC; CD34-Flt3-LSK), short-term hematopoietic stem cells (ST-HSC; CD34⁺Flt3-LSK) and multipotent progenitors (MPP; CD34⁺Flt3⁺LSK) with the frequencies indicated. (D, E) The absolute numbers (D) and percentages (E) of LSK cells (among Lin^- cells) from WT and TRIM31^-/- mice (n=5-6 per group). (F) The numbers of LT-HSC, ST-HSC, and MPP in WT and TRIM31^-/- mice (n=5-6 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. NS=not significant. Eight- to 12-week-old WT and TRIM31^-/- mice were used in the experiments.

Sca-1⁺c-Kit⁺) cells in TRIM31^-/- bone marrow (Figure 1C-E). To further characterize the composition of the LSK population, CD34 and Flt3 markers were used to distinguish long-term HSC (LT-HSC; CD34^-Flt3^- LSK), short-term HSC (ST-HSC; CD34⁺Flt3^- LSK), and multipotent progenitors (MPP; CD34⁺Flt3⁺ LSK). In TRIM31^-/- bone marrow, the numbers of ST-HSC and MPP increased dramatically, while the absolute number of LT-HSC decreased significantly (Figure 1C, F), which is probably attributable to stem cell exhaustion.

In bone marrow, TRIM31 depletion resulted in a significant decrease of B lymphocyte number (Online Supplementary Figure S2A, B), while the difference in peripheral blood was minimal (Online Supplementary Figure S2C). During erythrocyte development, the percentage of E3 (TER119⁺CD71^mid) cells increased dramatically in TRIM31^-/- mice (Online Supplementary Figure S2D, E). In addition, the spleen was hypertrophic in TRIM31^-/- mice (Online Supplementary Figure S3A, B), and the hypertrophic spleen contained a higher percentage of LSK cells (Online Supplementary Figure S3C), indicating that TRIM31 deletion may cause malfunction of the spleen. Consistently, the composition of B cells decreased while the proportion of myeloid cells increased in TRIM31^-/- mouse spleens (Online Supplementary Figure S3D). Overall, TRIM31 deletion impaired HSC homeostasis, causing harm to the cellularity of LT-HSC, ST-HSC, MPP and LSK cells, which further resulted in abnormal regulation of progenitors, such as CMP and CLP.

Loss of TRIM31 impairs hematopoietic stem cell function of reconstitution capacity

To evaluate the self-renewal and differentiation capacity of LT-HSC in vivo, we performed a competitive LT-HSC transplantation experiment. As donor cells, 300 LT-HSC were sorted from WT and TRIM31^-/- mice and transplanted into lethally irradiated recipient mice together with 3x10⁵ bone marrow competitors (Online Supplementary Figure S3E). The percentage of donor-derived cells (chimerism) in peripheral blood was analyzed every 4 weeks. The chimerism of TRIM31^-/- cells in peripheral blood was much lower than that of WT cells (Figure 2A). After 16 weeks, bone marrow analysis of chimeric mice showed that donor-derived TRIM31^-/- LSK cells were also greatly reduced in comparison to donor-derived WT cells (Figure 2B). Further analysis revealed that the chimerism of TRIM31^-/- cells was significantly decreased at the LT-HSC, ST-HSC and MPP (MPP2 and MPP3) levels (Figure 2B). For consecutive transplantation, 1x10⁶ chimeric bone marrow cells were re-transplanted into secondary recipient mice (Online Supplementary Figure S3E). As in the first round of transplantation, the percentage of donor-derived TRIM31^-/- cells in peripheral blood dropped further, while that of donor-derived WT cells remained nearly the same. The peripheral blood chimerism of TRIM31^-/--derived cells was almost zero at 28 weeks after transplantation, while that of WT-derived cells remained approximately 50% (Figure 2A). Bone marrow analysis after the second transplantation showed that LSK chimerism was dramatically decreased in TRIM31^-/--derived cells in comparison to WT-derived cells (Figure 2C). Furthermore, chimerism analysis of LT-HSC, ST-HSC and MPP showed dramatic reductions in TRIM31^-/--derived cells (Figure 2C). The chimerism in T lymphocytes, B lymphocytes and myeloid cells in peripheral blood was also clearly decreased following both the first and second transplants (Figure 2D-F).

To confirm the results of LT-HSC transplantation, a competitive LSK transplantation experiment was performed using 4,000 LSK cells sorted from WT and TRIM31^-/- mice, together with 1x10⁶ competitive cells (Online Supplementary Figure S4A). As for the LT-HSC transplantation, peripheral blood chimerism was analyzed every 4 weeks, and the percentage of donor-derived TRIM31^-/- cells in peripheral blood was significantly decreased in comparison to that of donor-derived WT cells (Online Supplementary Figure S4B). Bone marrow analysis also indicated that chimerism of donor-derived LSK cells was significantly reduced in TRIM31^-/--derived cells (Online Supplementary Figure S4C). This reduction in chimerism was simultaneously observed at the LT-HSC, ST-HSC and MPP levels in TRIM31^-/--derived cells (Online Supplementary Figure S4D, E). Additionally, T lymphocytes, B lymphocytes, and myeloid cells all displayed drastic reductions in chimerism (Online Supplementary Figure S4F-H). To exclude the effect of TRIM31 deletion in the stem cell niche, 4,000 WT LSK cells were sorted and transplanted into WT and TRIM31^-/--recipient mice. No statistically significant difference of donor-derived LSK chimerism was shown between WT and TRIM31^-/--recipient mice (Online Supplementary Figure S4I, J). In summary, after serial competitive transplantation, TRIM31 deletion significantly reduced peripheral blood chimerism and bone marrow chimerism at the levels of LSK cells, LT-HSC, ST-HSC, and MPP, indicating that TRIM31 deletion impairs the reconstitution ability of HSC.

TRIM31 deficiency promotes the proliferation of hematopoietic stem cells and reduces their self-renewal capacity under stress

To ascertain the underlying reasons for hematopoietic stem and progenitor cell alterations in TRIM31^-/- mice, a BrdU incorporation assay was performed. TRIM31 deletion induced a significant increase of BrdU-positive cells both at the LSK cell and LT-HSC levels, indicating that TRIM31 deletion promotes proliferation of HSC (Figure 3A, B). This indication was also supported by the decrease of Ki67-negative staining in TRIM31^-/- cells (Online Supplementary Figure S5A-D). Conditions of stress, such as 5-fluorouracil (5-FU) treatment, could cause apoptosis of cycling HSC, whereas quiescent HSC might remain viable.²⁹ To detect the stress effect of TRIM31 deficiency on HSC, we challenged TRIM31^-/- mice with sequential 5-FU administration. TRIM31^-/- mice challenged with sequential 5-FU treatment died much earlier than WT controls (Figure 3C), demonstrating that TRIM31-null HSC were more activated.

5-FU treatment was also used to quantify cell self-renewal capacity. Analysis of the results at 6 days after 5-FU treatment revealed that TRIM31^-/- mice had considerably fewer LSK cells than WT mice. Subsequent analysis demonstrated that the numbers of LT-HSC, ST-HSC and MPP after 5-FU administration were all substantially reduced in TRIM31^-/- mice compared to the numbers in WT mice (Figure 3D, E). To further confirm the defect caused by TRIM31 deletion, a single-colony formation assay was used to quantify the ability of HSC to proliferate and differentiate in vitro. In TRIM31^-/- LT-HSC, the number of large colonies was significantly decreased, indicating that the self-renewal ability of LT-HSC was impaired after long-term activation (Figure 3F). In short, TRIM31 deletion induces HSC proliferation in vivo and leads to a functional decline of self-renewal ability under a condition of stress.

Figure 2.Loss of TRIM31 impairs long-term hematopoietic stem cell function. (A) The percentage of donor-derived peripheral blood cells at the indicated time-points in the serial competitive transplantation assay is shown (1st, for the first competitive transplantation, 4 recipient mice per group; 2nd, for the second competitive transplantation, 6 recipient mice per group). (B) Percentages of donor-derived LSK (Lin^-Sca-1⁺c-Kit⁺) cells, long-term hematopoietic stem cells (LT-HSC; CD48^-CD150⁺Flt3^-LSK), short-term hematopoietic stem cells (ST-HSC; CD48^-CD150^-Flt3^-LSK) and multipotent progenitors (MPP; CD48⁺Flt3^-LSK) 16 weeks after the first transplantation are shown (n=4 per group). (C) Percentages of donor-derived LSK cells, LT-HSC, ST-HSC and MPP cells 12 weeks after secondary transplantation are shown (n=6 per group). (D-F) Two-round serial transplantation was conducted using 300 purified LT-HSC cells along with 3x10⁵ fresh competitors each time. Chimerism of T, B and myeloid cells in peripheral blood is shown at the indicated time-points after transplantation. All results are presented as the mean ± standard deviation. **P<0.01; ***P<0.001; ****P<0.0001. Eight- to 12-week-old wildtype and TRIM31^-/- mice were used in the experiments. WT: wildtype; PB: peripheral blood; BM: bone marrow.

Figure 3.TRIM31 deletion leads to enhanced hematopoietic stem cell proliferation and reduced self-renewal capacity under stress. (A) Representative FACS plots showing the proliferation analysis in long-term hematopoietic stem cells (LT-HSC; CD48^-CD150⁺Flt3^- LSK) and LSK (Lin-Sca-1⁺c-Kit⁺) cells. The percentages of BrdU-positive cells are shown. (B) The percentages of BrdU-positive cells among LT-HSC and LSK cells from wildtype (WT) and TRIM31^-/- mice after long-term BrdU labeling (10 days) (n=4-5 per group). (C) Survival curves of WT and TRIM31^-/- mice following sequential 5-fluoruracil (5FU) treatment (n=13 per group). 5FU was injected into mice once a week for a total of two injections. (D) Representative FACS plots of LSK, LT-HSC, short-term hematopoietic stem cells (ST-HSC; CD48^-CD150^-Flt3^- LSK) and multipotent progenitors (MPP2; CD48⁺CD150⁺Flt3^- LSK and MPP3; CD48⁺CD150^-Flt3^- LSK) 6 days after 5FU treatment. (E) The absolute numbers of LT-HSC, ST-HSC, MPP2 and MPP3 cells in WT and TRIM31^-/- mice after 5FU treatment (n=4 per group). (F) Representative images of large, intermediate and small colonies are shown. The percentages of colonies formed after 14 days of culture of single LT-HSC sorted from WT and TRIM31^-/- mice are shown (n=3 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ****P<0.0001. Eight- to 12-week-old WT and TRIM31^-/- mice were used in the experiments.

TRIM31 deletion accelerates leukemia initiation in a model of MLL-AF9-induced acute myeloid leukemia

To evaluate whether the accelerated proliferation of TRIM31^-/- hematopoietic cells is related to leukemia, a database analysis was carried out and revealed that low expression of TRIM31 was associated with poor AML survival (Figure 4A, B; Online Supplementary Figure S5E). Among patients with various types of AML, TRIM31 expression was relatively lower in those with MLL rearrangements (Online Supplementary Figure S5F). Therefore, a murine AML model driven by the human oncogene MLL-AF9 was used to investigate the role of TRIM31 in leukemia initiation and development.³⁰ c-Kit⁺ cells were enriched from bone marrow of WT and TRIM31^-/- mice, and then transduced with MLL-AF9 retrovirus (Online Supplementary Figure S6A). In vitro, a surrogate functional examination using methylcel-lulose medium revealed that TRIM31-null MLL-AF9 primary transformed cells had significantly increased colony-forming capability (Online Supplementary Figure S6B). During primary plating and second plating, the numbers of colonies and total cells were markedly higher for MLL-AF9-converted TRIM31^-/- cells than for WT cells (Online Supplementary Figure S6C, D). To assess leukemo-genesis in vivo, MLL-AF9 retrovirus-transduced WT and TRIM31^-/- cells were sorted and transplanted into lethally irradiated recipient mice after which the development of AML was monitored (Online Supplementary Figure S6A). The frequency of L-GMP (Lin^-Sca1^-c-Kit⁺CD34⁺CD16/32⁺) cells was reported as leukemia-initiating cells in the MLL-AF9-induced leukemia.³¹ A higher frequency of L-GMP was detected in the bone marrow of TRIM31^-/- donors than in WT donors (Figure 4C). In peripheral blood, the frequency of GFP⁺ leukemic cells was significantly higher in TRIM31^-/- donors than in WT ones (Figure 4D). Meanwhile, recipient mice that received MLL-AF9-transduced TRIM31^-/-donor cells had a significantly shorter overall survival than WT mice (Figure 4E). Moreover, after the L-GMP cells from leukemia mice had been sorted and transplanted, the disparity in overall survival of mice receiving grafts from WT or TRIM31^-/- donors increased dramatically (Figure 4F). Taking into consideration the uniform expression, copies of MLL-AF9 fusion gene and mean fluorescent intensity of GFP in both WT and TRIM31^-/- donor cells (Online Supplementary Figure S6E-J), these data collectively support that TRIM31 deletion accelerated the initiation and development of MLL-AF9-induced AML.

TRIM31 functions as an E3 ligase of CDK8 and regulates its ubiquitin-mediated degradation

The selected TRIM31 interacting protein candidates after mass spectrometry analysis are summarized.

Through MAVS and NLRP3, TRIM31 plays a critical role in macrophages and intestinal cells.^15-17 In order to identify the underlying mechanism and the functional substrate of TRIM31 in hematopoietic cells, we performed mass spectrometry (MS) on WT cells, with TRIM31^-/- cells used as the negative control (Online Supplementary Figure S7A). Several TRIM31 interacting protein candidates in WT cells were selected after MS (Online Supplementary Table S1). Interestingly, FACS analysis showed that the median fluorescence intensity of CDK8 was dramatically increased in LT-HSC, LSK cells and Lin^- cells from TRIM31^-/- mice, whereas no difference in the mean fluorescence intensity of EIF5A and Cdkn2a was detected between WT and TRIM31^-/- mice (Figure 5A; Online Supplementary Figure S7B, C). The increased CDK8 level in cells from TRIM31^-/-mice was verified by western blot (Figure 5B). Moreover, the cell cycle regulator cyclin D1 was strongly upregulated in TRIM31^-/- cells (Figure 5B), and this upregulation was accompanied by an increase in its interaction partner CDK6 (Online Supplementary Figure S7D), which may be due to cell proliferation. To verify whether CDK8 is a direct ubiquitin substrate of TRIM31, we performed a co-immunoprecipitation assay of CDK8 antibody using WT and TRIM31^-/- cells, which were treated with a proteasomal inhibitor (MG132) before cell lysis. The level of ubiquitination of CDK8, but not of CDK6, was reduced dramatically in TRIM31^-/- cells (Figure 5C; Online Supplementary Figure S7E), indicating that CDK8 is a ubiquitin substrate of the E3 ligase, TRIM31. Furthermore, the successful pulling down of TRIM31 and CDK8 together confirmed a direct protein interaction between them (Figure 5D). Surprisingly, CDK8 could pull down and co-localize with cyclin D1 (Figure 5D; Online Supplementary Figure S7F), suggesting that cyclin D1 might function together with CDK8 in hematopoietic cells. Importantly, upregulation of CDK8 and cyclin D1 was observed in TRIM31^-/- mouse spleens (Figure 5E), and the level of CDK8 protein was also enhanced in TRIM31^-/- leukemia and L-GMP cells (Figure 5F). In brief, CDK8 accumulation in hematopoietic and leukemia cells was caused by deletion of its E3 ligase TRIM31 and eventually led to enhanced expression of the cell cycle regulator cyclin D1.

Genetic knockdown of CDK8 or cyclin D1 rescues the defects in TRIM31-/- hematopoietic stem cells

To confirm that the functional decline in stem cells in TRIM31^-/- mice was attributable to upregulation of CDK8, an experiment of inhibition of CDK8 was performed. LY2857785 is a type I reversible and competitive ATP kinase inhibitor of both CDK8 and CDK9.³² After the administration of LY2857785 to WT and TRIM31^-/- mice for 2 weeks, LT-HSC were sorted and transplanted together with competitors into recipient mice (Online Supplementary Figure S8A). Eight to 10 weeks later, FACS analysis showed that the treatment of TRIM31^-/- mice with a CDK8 inhibitor caused enhancement of peripheral blood LSK and LT-HSC chimerism, while no difference was shown in the control group (Figure 6A, B; Online Supplementary Figure S8B-D). To inhibit CDK8 more specifically, LT-HSC from TRIM31^-/- mice were infected with lentivirus containing GFP-labeled shRNA against CDK8 and transplanted into recipient mice. Twelve, 16 and 20 weeks later, peripheral blood analysis showed that the percentage of donor-derived cells was dramatically increased in the CDK8 knockdown group (Figure 6C). A similar phenomenon was also found in donor-derived LSK cells after bone marrow analysis (Figure 6D; Online Supplementary Figure S8E-G). The above results demonstrate that knockdown of CDK8 in TRIM31^-/- LT-HSC could lead to enhanced stem cell function, which might be due to the prevention of LT-HSC exhaustion (Figure 6E, F). In addition, knockdown of cyclin D1 in TRIM31^-/- LT-HSC also resulted in an increased percentage of donor-derived cells (Online Supplementary Figure S9A-F), which was similar to the effects of overexpression of TRIM31 (Online Supplementary Figure S10A-C), showing that TRIM31 and cyclin D1 may function in the same pathway (Online Supplementary Figure S9G). Meanwhile, overexpression of TRIM31 caused reduced colony formation in WT MLL-AF9 leukemia cells (Online Supplementary Figure S10D). In short, TRIM31 functions through the CDK8/cyclin D1 pathway, which also inhibits the development of MLL-AF9, since both pharmaceutical inhibition of CDK8 and genetic knockdown of CDK8 and cyclin D1 rescued the functional defects in TRIM31^-/- HSC.

Figure 4.TRIM31 deletion accelerated leukemia initiation in an MLL-AF9-induced mouse model. (A) Kaplan-Meier survival curve analysis of The Cancer Genome Atlas dataset (https://xenabrowser.net/) comparing the acute myeloid leukemia (AML) patients with high (n=33) versus low (n=99) TRIM31 expression. (B) Kaplan-Meier survival curve analysis of the GSE12417 dataset comparing AML patients with high (n=39) versus low (n=40) TRIM31 expression. (C) Representative FACS plots showing the percentages of L-GMP cells (Lin^-Sca1^-c-Kit⁺CD34⁺CD16/32⁺) in the bone marrow (BM) of wildtype (WT) and TRIM31^-/- recipient mice (n=4-6 per group). (D) Representative FACS plots showing the percentages of GFP⁺ cells and myeloid cells in the peripheral blood (PB) of WT and TRIM31^-/- recipient mice (n=4-6 per group). (E) Survival curves of mice transplanted with MLL-AF9 WT or TRIM31^-/- leukemic cells (n=5 per group). (F) Survival curves of mice transplanted with 5,000 WT or TRIM31^-/- L-GMP cells (n=5 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01. Eight- to 12-week-old WT and TRIM31^-/- mice were used in the experiment. GDC: Genomic Data Commons; TCGA: The Cancer Genome Atlas; SSC: side scatter; FSC: forward scatter; GFP: green fluorescent protein.

Figure 5.Accumulation of CDK8 and upregulation of cyclin D1 were caused by TRIM31 deletion. (A) Representative FACS plots of CDK8 mean fluorescence intensity (MFI) in Lin-cells from wildtype (WT) and TRIM31^-/- mice. The MFI of CDK8 in long-term hematopoietic stem cells (LT-HSC; CD48-CD150+Flt3- LSK), LSK cells (Lin-Sca-1+c-Kit+) and Lin- cells from WT and TRIM31^-/- mice is shown (n=5 per group). (B) Immunoblot (IB) of cyclin D1 and CDK8 in the whole cell lysate (WCL) of Lin- cells from WT and TRIM31^-/- mice. β-tubulin was used as a loading control (n=3 per group). IB of CDK8 in the WCL of LSK cells from WT and TRIM31^-/- mice. β-actin was used as a loading control (n=3 per group). (C) Ubiquitination assay of CDK8 in the whole cell lysate of Lin- cells from WT and TRIM31^-/- mice treated with a proteasomal inhibitor (MG132). IB against ubiquitin was performed with the immunoprecipitation (IP) assay precipitate of anti-CDK8 and CDK8 from WCL used as a loading control. (D) IP of anti-CDK8 and anti-TRIM31 was performed on the WCL of Lin- cells from WT mice; IP of anti-IgG was used as a control. IB against TRIM31, cyclin D1 and CDK8 was performed with the IP assay precipitate and the WCL of Lin- cells, which was used as a loading control. (E) IB of cyclin D1 and CDK8 in WCL of spleens from WT and TRIM31^-/- mice. β-actin was used as a loading control (n=2 per group). (F) IB of CDK8 in the WCL of c-Kit+ cells from recipient mice transplanted with MLL-AF9 WT and TRIM31-/- cells. β-actin was used as a loading control (n=3-4 per group). IB of CDK8 in the WCL of L-GMP cells from recipient mice transplanted with MLL-AF9 WT and TRIM31-/- cells. β-actin was used as a loading control (n=3 per group). All results are presented as the mean ± standard deviation. **P<0.01; ***P<0.001. Eight- to 12-weekold WT and TRIM31^-/- mice were used in the experiments. Ub: ubiquitin.

CDK8 regulates the expression of cyclin D1 through transcriptional factor PBX1

The CDK8-mediator is a large macromolecular complex comprising four modules, the head, middle, tail, and kinase modules, which consists of CDK8 and three other factors: Med12, Med13 and cyclin C.³³ CDK8, MED12 and RNAPII were found to bind the promotor region of the transcriptional factor PBX1 after analysis of chromatin immunoprecipitation (ChIP) sequencing data (GEO dataset: GSE128242) (Figure 7A).³⁴ PBX1 is a proto-oncogene in the hematopoietic system, and its loss results in exhaustion of LT-HSC and defects in HSC maintenance and function.³⁵ Meanwhile, the upregulation of PBX1 promotes cell proliferation.³⁶ Through ChIP-quantitative PCR, CDK8 was verified to be enriched at the promoter region of PBX1 (Figure 7B). To determine the underlying mechanism of CDK8 regulation, quantitative PCR analysis of cyclin D1 and PBX1 was performed. Both mRNA expression levels of cyclin D1 and PBX1 were enhanced in TRIM31^-/- hematopoietic and L-GMP cells (Figure 7C), which was verified at the protein level by western blot (Figure 7D). In order to further determine the relationship between CDK8, PBX1 and cyclin D1, CDK8 was knocked down and overexpressed in NIH/3T3 mouse fibroblasts. The transcriptional expression of PBX1 and the mRNA expression of cyclin D1 were regulated through the mRNA level of CDK8 (Figure 7E). Similarly, the downregulation of PBX1 caused by knockdown of CDK8 was also shown in hematopoietic cells (Figure 7F). Moreover, the changes in PBX1 directly regulated the expression of cyclin D1, indicating that PBX1 is located upstream of cyclin D1 (Online Supplementary Figure S11A). Functionally, knockdown of PBX1 in TRIM31^-/- LT-HSC resulted in an increased percentage of donor-derived cells (Online Supplementary Figure S11B-F), and knockdown of TRIM31 resulted in increased expression of PBX1 and cyclin D1 in THP1 cells (Online Supplementary Figure S11G), implying that TRIM31 and PBX1 may act in the same pathway. In concordance with our quantitative PCR data, the expression of PBX1 and cyclin D1 in primary AML patients was correlated with CDK8 expression (Online Supplementary Figure S12A, B). Furthermore, cyclin D1 was significantly increased in patients with higher PBX1 expression (Online Supplementary Figure S12C). In general, CDK8 regulates the expression of cyclin D1 through the transcriptional factor PBX1.

Discussion

Ubiquitin E3 ligases are vital regulators of the hematopoietic system, and their absence dramatically impairs HSC maintenance and function. For example, deletion of Huwe1, a ubiquitin ligase, leads to increased proliferation and stem cell exhaustion via upregulation of N-myc expression.³⁷ Conditional deletion of SCF^Fbxw7, another ubiquitin E3 ligase, causes loss of HSC quiescence and stem cell exhaustion via enhancement of c-Myc.³ In the present study, we found that the E3 ligase TRIM31 could regulate HSC homeostasis and function. TRIM31 deletion caused enhanced proliferation of LT-HSC and further led to their exhaustion. Furthermore, HSC function was significantly impaired in TRIM31^-/- mice. In a serial competitive transplantation assay and 5-FU challenge, TRIM31^-/- HSC showed dramatically reduced self-renewal capacity. TRIM31 was reported to be an E3 ligase in the immune system^11,12 and was also found to play a role in cancer cells.^38,39 Here, we discovered a novel role for TRIM31, which is crucial for the hematopoietic system. Prospectively, TRIM31 could be targeted to regulate HSC maintenance and function.

In human AML, MLL is mutated by translocation in about 4% of cases and MLL fusion partners include AF9, ENL, and AF4.³⁰ Sustaining proliferative signaling is a hallmark of cancer development; and deletion of TRIM31 promotes breast cancer progression through ubiquitination of p53.⁴⁰ Based on our findings of cell proliferation induced by TRIM31 deletion, we explored the function of TRIM31 in the initiation of AML. TRIM31 may promote leukemia progression through Wnt/β-catenin signaling in AML cell lines.⁴¹ However, by using a murine model of MLL-AF9 in leukemia initiation and development, we demonstrated that TRIM31 depletion increased the proportion of leukemia-initiating cells and accelerated the kinetics of leukemia development. These effects might be due to the functional mechanisms of TRIM31 varying at different disease stages. Consistent with our results, a database analysis in AML patients also showed that higher expression of TRIM31 correlated with significantly better survival. These data indicate that TRIM31 functions as a suppressor in AML initiation and development. Consolidating all the results of the TRIM31^-/- HSC analysis, TRIM31-mediated control of HSC proliferation is integral to suppress the vulnerability of HSC to leukemic transformation and disease initiation in AML.

Members of the CDK8 module are vital in various cancers and developmental diseases.^42,43 Loss of the CDK8 module member MED12 causes rapid bone marrow failure and acute lethality, showing that MED12 is essential for HSC homeostasis.⁴⁴ The kinase activity of CDK8 was reported to be inhibited by small molecules;⁴⁵ however, the mechanism of its upstream regulation remains unknown. In this study, we found that CDK8 was regulated by the E3 ligase TRIM31 through ubiquitin-mediated protein degradation. Upregulation of CDK8 in TRIM31^-/- hematopoietic and leukemia cells leads to enhanced stem cell proliferation, indicating that CDK8 may function as a regulator of cell cycle-related pathways in the hematopoietic system. Therefore, TRIM31 might be used as a pharmaceutical target in CDK8-enhanced tumors.

Figure 6.Inhibition and knockdown of CDK8 rescues TRIM31^-/- hematopoietic stem cell function. (A) Percentage of donor-derived peripheral blood (PB) cells at the indicated time-points after competitive transplantation in long-term hematopoietic stem cells (LT-HSC: CD48^-CD150⁺Flt3^- LSK) sorted from wildtype (WT) and TRIM31^-/- mice treated with a CDK8 inhibitor (LY2857785) (n=4 per group). (B) Percentage of donor-derived LSK (Lin^-Sca-1⁺c-Kit⁺) cells and LT-HSC 10 weeks after LT-HSC competitive transplantation. (C) The percentage of GFP⁺ cells among donor-derived PB cells after transplantation with shCDK8-1 and -2 virus-infected TRIM31^-/- LT-HSC (n=4 per group). (D) The percentage of GFP⁺ cells in donor-derived LSK cells after transplantation with shCDK8-1 and -2 virus infected TRIM31^-/- LT-HSC (n=4 per group). (E) The percentage of G₀ phase (Ki67-negative) cells from mice 12 weeks after transplantation with WT and TRIM31^-/- LT-HSC infected with shCDK8-1 lentivirus (n=4-5 per group). (F) The knockdown efficiency of shCDK8-1 in GFP⁺ LSK cells after transplantation with shCDK8-1 virus-infected TRIM31^-/- LT-HSC (n=5-6 mixed per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ***P<0.001. Eight- to 12-week-old WT and TRIM31^-/- mice were used in the experiment.

CDK8 is reported to be involved in several signaling pathways. In colon cancer, β-catenin hyperactivity promotes cell proliferation and drives tumor transformation which is inhibited by repression of CDK8 expression.²⁴ Through the downstream target Myc, CDK8 functions in regulating the pluripotent state of embryonic stem cells.⁴⁶ Here, in hematopoietic cells, we found that PBX1 acts as a direct down stream target of CDK8, which binds to the promoter region of PBX1; the mRNA expression levels of PBX1 and cyclin D1 were upregulated in TRIM31^-/- mice. CDK8 regulates transcription factors through phosphorylation and proteasomal degradation, suggesting that CDK8 may function as a general regulator of transcription factors.⁴⁷ CDK8 acts as a key mediator of BCR-ABL1-driven leukemia through transcriptional changes of the mTOR signaling pathway.⁴⁸ In this study, we found that CDK8 regulates cyclin D1 through the transcription factor PBX1. Meanwhile, the association of CDK8 with mediators is reversible and can dramatically alter the structure and function of the mediators.⁴⁹ In our context, the accumulation of CDK8 in TRIM31^-/- hematopoietic and leukemia cells may open the structure of the transcriptional complex and lead to modification of chromosome elements such as histones, further promoting transcription of downstream genes.

Figure 7.CDK8 regulates the expression of PBX1 and cyclin D1. (A) The binding sites of CDK8, MED12 and RNAPII in the region of the PBX1 gene are shown. Raw data (from GEO dataset: GSE128242) were analyzed. (B) Real-time polymerase chain reaction (PCR) analysis of CDK8 chromatin immunoprecipitated DNA. Two pairs of primers were designed based on the PBX1 promoter region (2,000 bp upstream of the start point). For each primer pair, amplifications of chromatin before immunoprecipitation and chromatin immunoprecipitated with preimmune serum were performed as input and negative controls, respectively. The value of bound DNA relative to input is shown as a percentage (n=3 per group). (C) The mRNA levels of PBX1 and cyclin D1 in LSK (Lin^-Sca-1⁺c-Kit⁺) cells from WT and TRIM31^-/- mice were measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). The mRNA levels of CDK8, PBX1 and cyclin D1 in L-GMP cells from recipient mice transplanted with MLL-AF9 WT and TRIM31^-/- cells were measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). (D) Immunoblot of PBX1 in the whole cell lysate of Lin^- cells from WT and TRIM31^-/- mice. β-actin was used as a loading control (n=3 per group). (E) The relative mRNA expression of CDK8, PBX1 and cyclin D1 in 3T3NIH cells with and without CDK8 knockdown (CDK8-shRNA1, CDK8-shRNA 2) or overexpression (CDK8-OE) was measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). (F) The relative mRNA expression of CDK8 and PBX1 in control and CDK8 knockdown (GFP⁺) Lin^- cells in vivo, which were isolated from mice 8 weeks after transplantation with TRIM31^-/- long-term hematopoietic stem cells (LT-HSC: CD48^-CD150⁺Flt3^- LSK) infected with lentivirus. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ***P<0.001; ns=not significant. Eight- to 12-week-old WT and TRIM31^-/- mice were used in the experiments.

PBX1, a proto-oncogene in childhood leukemia, is a typical homeodomain transcription factor belonging to the TALE family. PBX1 is very importantor the hematopoietic system and its downstream targets could be TGFβ and JAK2/Stat3.^35,36 In TRIM31^-/- mice, cyclin D1 upregulation was induced by enhanced expression of PBX1 (Online Supplementary Figure S11G), which may occur through the Jak2/Stat3 pathway.³⁶ In addition, cyclin D1 may function together with CDK8 in the hematopoietic system of TRIM31^-/- mice, although the underlying mechanism needs to be further investigated. In conclusion, the accumulation of CDK8 caused by TRIM31 deletion induces transcription of PBX1, which further leads to upregulation of cyclin D1 (Online Supplementary Figure S11H). Enhancement of CDK8 and PBX1 target genes is observed in human leukemia;⁵⁰ therefore, using TRIM31 as an anti-leukemia target is promising.

Footnotes

• Received August 18, 2022
• Accepted December 28, 2022

Correspondence

*KZ and DL contributed equally as first authors.

D. Diao diaodaojun@hotmail.com

Z. Ju zhenyuju@163.com

Disclosures

No conflicts of interest to disclose.

Contributions

DD and ZJ conceived and designed the experiments. KZ, DL, YL and ZS performed the experiments. KZ and DL analyzed the data. DD, KZ, DL and YL wrote the paper. CG provided TRIM31^-/- mice and JG and HW provided valuable suggestions.

Data-sharing statement

All plasmids and cell lines generated in this study are available from the authors upon request.

Funding

This work was supported by the National Natural Science Foundation of China (grants #81970096 and 91849128 to DD) and the Natural Science Foundation of Guangdong Province, China (#2020A1515010453 to DD); the National Natural Science Foundation of China (#92049304, 82030039), the National Key R&D Program of China (2021YFA1100103) and the Program for Guangdong Introducing Innovative and Entrepreneurial Teams (2017ZT07S347) to ZJ.

Acknowledgments

We thank Dr. H. Cheng for his kind gift of the MSCV-MLL-AF9-IRES-GFP encoding plasmid.