Repression of TRIM13 by chromatin assembly factor CHAF1B is critical for AML development

Acute myeloid leukemia (AML) is an aggressive blood cancer characterized by blasts in the bone marrow (BM). Though AML tumors generally harbor a low mutational load, the mutations that they possess frequently occur in genes that encode epigenetic regulatory proteins.^1,2 This has led to the categorization of AML a disease of epigenetic dysregulation, and therapeutic compounds that target epigenetic regulators are some of the most exciting new treatment strategies for AML.^3,4 The chromatin assembly factor (CAF) CHAF1B is a master epigenetic regulator of transcription associated with cell state and self-renewal.^5-7 CHAF1B is the p60 subunit of the heterotrimeric CAF1 complex also containing p150 CHAF1A and p48 RBBP4, responsible for facilitating histone H3/H4 heterodimer assembly at the replication fork during the S-phase of cell cycle.⁸ In addition to its role in nucleosome assembly, CHAF1B has a regulatory role in transcription and is linked to cell state–directed transcriptional regulation in many metazoan organisms.^6,9-11 CHAF1B is upregulated in a wide variety of cancers, and enhanced expression correlates with poor prognosis in almost every cancer studied.^12-17 We recently reported that CHAF1B is upregulated in AML, in which it binds to chromatin at promoters and enhancers of differentiation genes, resulting in the repression of the transcription of differentiation genes through the displacement of transcription factors (such as C/EBPα).⁵ Reduction of CHAF1B to expression levels in healthy tissues is sufficient to release the transcriptional repression on differentiation genes and resolve the tumors in vitro and in vivo.⁵ This led us to hypothesize that CHAF1B promotes tumorigenicity by repressing tumor suppressors and differentiation genes. Because there are no therapeutic interventions that can directly target CHAF1B or the CAF1 complex, we are interested in understanding the transcriptional targets of CHAF1B so that we can leverage those as potential therapies.

TRIM13 is an E3 ubiquitin ligase (tripartite motif protein), characterized by a RING motif located near the N-terminus, zinc-binding B-box motif, and coiled-coil B-box motif.¹⁸ TRIM13 also has a transmembrane domain that is responsible for anchoring it to the endoplasmic reticulum (ER) in some cell types, such as those in samples from patients with multiple myeloma.¹⁹ TRIM13 is located on human chromosome 13q14, an area frequently deleted in B-cell chronic lymphocytic leukemia (CLL).^20-23 Because of its frequent deletion in CLL, TRIM13 was originally referred to as ret finger protein deleted in leukemia “LEU5/RFP2.”^24,25 Although not every patient with CLL has an accompanying loss of TRIM13,²⁶ TRIM13 expression is reduced in CLL tumors upon progression when compared with initial diagnosis.²⁷ Because of this relationship, TRIM13 is considered a tumor suppressor in CLL.^25,28 Chromosome 13 deletions are also present in myeloma, mantle cell lymphoma, diffuse large B-cell lymphoma, prostate cancer, and head and neck cancers,^29-34 and low expression of TRIM13 is a marker of poor prognosis in breast cancer.³⁵ Furthermore, elevated expression of TRIM13 is considered a marker of good prognosis in non–small cell lung carcinoma through the initiation of apoptosis in the tumor tissue in vivo.³⁶ In contrast, 1 study so far suggested that TRIM13 may act as an oncogene in the MMS1 multiple myeloma cell line by activating NF-κB through targeted the degradation of the NF-κB inhibitor IκBα.¹⁹ We demonstrate that TRIM13 represses leukemogenesis by localizing at the nucleus and ubiquitinating the cell cycle gene cyclin A1 (CCNA1), prompting premature entry into the cell cycle resulting in loss of self-renewal. CHAF1B, in turn, promotes leukemic transformation by repressing TRIM13, enforcing self-renewal of leukemic blasts, and promoting progression.

AML cell lines (U937, THP1, K562, NB4, and MOLM13) were maintained in RPMI-1640 supplemented with 10% heat-inactivated fetal bovine serum (HI-FBS), Pen-Strep, and L-glutamine. CMK cells were maintained in RPMI-1640 supplemented with 20% HI-FBS, Pen-Strep, and L-glutamine. The AML cell line, HL60, was maintained in Iscove modified Dulbecco medium supplemented with 10% HI-FBS, Pen-Strep, and L-glutamine. Mouse MLL-AF9 leukemic cells were maintained in RPMI-1640 supplemented with 10% HI-FBS, Pen-Strep, L-glutamine, and cytokines (100 ng/mL of stem cell factor, 50 ng/mL of interleukin-6 [IL-6], and 20 ng/mL of IL-3). Human CD34⁺ peripheral blood stem cells (PBSCs) were maintained in Stemspan II serum-free expansion medium, with CD34 supplement and primocin. Patient-derived xenograft (PDX) samples were maintained for a short-term in Iscove modified Dulbecco medium supplemented with 20% FBS, Pen-Strep, and 10 ng/mL each of IL-3, IL-6, stem cell factor, FLT3L, and thrombopoietin. 293T cells were maintained in Dulbecco’s modified Eagle medium supplemented with 10% HI-FBS, Pen-Strep, and L-glutamine.

Cells were pulsed with 10 μM 5-ethynyl-2'-deoxyuridine (EdU) for 1 hour at 37°C. After EdU pulse, cells were stained with fixable viability dye–red at 4 °C for 30 minutes. After washing, cells were fixed in 4% paraformaldehyde for 15 minutes and permeabilized in saponin buffer. Allophycocyanin-azide was clicked onto EdU for 30 minutes at room temperature, and DNA content was stained using FxCycle Violet for 30 minutes at 4°C. Cells were read on either a CantoII or Fortessa flow cytometer. Data were analyzed using FlowJo.

Complete single guide RNA targeting early in the coding sequence of TRIM13 were ordered from Synthego. These single guide RNAs were complexed with HiFi Cas9 V3 (IDT) for 20 minutes at room temperature before being resuspended in Neon Buffer R (Thermo Fisher) and Cas9 Electroporation Enhancer (IDT). A total of 1 × 10⁵ cells were electroporated in a 10 μL Neon Tip (human AML cells = 1350 V, 35 ms, 1 pulse; mouse AML cells = 1700 V, 20 ms, 1 pulse) and plated in complete media without antibiotics for 24 hours. After 48 hours of electroporation, CRISPR efficiency was measured via Sanger sequencing and inference of CRISPR edits analysis. Cells were then plated in methylcellulose, and single colonies were selected for expansion. TRIM13 knockout was confirmed in candidate colonies, first via Sanger sequencing, followed by western blot. Knockout clones were expanded. In-frame deletions (RINGdel) or out-of-frame knockout (TRIM13 knock out) were confirmed via Sanger and western blot, using expanded clones.

Oligonucleotides, antibodies, and expanded methods are presented in the supplemental figures and methods.

To identify mechanisms by which CHAF1B regulates transcription associated with leukemogenesis, we knocked out Chaf1b (Chaf1b^Δ/Δ) in mouse MLL-AF9 leukemic cells (Figure 1A-B), followed by RNA sequencing (supplemental Figure 1A). As previously shown,⁵ CHAF1B loss leads to transcriptional upregulation of many genes, including differentiation genes (supplemental Figure 1A; supplemental Table 1). We found that many genes were expressed from alternative promoters in the absence of CHAF1B (supplemental Figure 1B). Because our previous work demonstrated that CHAF1B occupies promoters of differentiation genes resulting in their transcriptional repression,⁵ we focused on this subset of genes for our analysis. We ranked these genes in the order of most to least exclusive use of alternative promoters using AltAnalyze³⁷ and found that the top ranked gene transcribed by an alternative promoter, after CHAF1B loss, was TRIM13. In Chaf1b^+/+ leukemic cells, TRIM13 is transcribed exclusively from the proximal exon (Trim13-201 isoform). Upon Chaf1b deletion, TRIM13 can be transcribed from either the proximal or a distal (Trim13-202) exon (supplemental Figure 1C). This matched with POLR2A chromatin occupancy at the proximal promoter associated with Trim13-201 but redistributed to the distal exon after Chaf1b loss (supplemental Figure 1D).⁵ This was also associated with an increase in H3K4me3 and assay for transposase-accessible chromatin signal at the distal exon, as measured via chromatin immunoprecipitation sequencing (supplemental Figure 1E).⁵ We confirmed the increased expression and use of the alternative exon from our RNA sequencing data via Taqman quantitative polymerase chain reaction, with probes that exclusively recognize the 201 or 202 isoforms of TRIM13 (supplemental Figure 1F-G).

To interrogate the function of this alternative promoter, we used CRISPR to delete the proximal (Trim13^Δ201) or the distal exon (Trim13^Δ202) (supplemental Figure 1H). Using isoform-specific Taqman probes, we found that in Chaf1b^+/+ MLL-AF9 leukemic cells, the proximal exon appeared necessary for TRIM13 expression. However, in Chaf1b^Δ/Δ MLL-AF9 leukemic cells, TRIM13 expression could be restored entirely from the distal promoter (supplemental Figure 1I). Although the proximal promoter was necessary for both messenger RNA (mRNA) transcription and protein production of TRIM13 in Chaf1b^+/+ MLL-AF9 leukemic cells (supplemental Figure 1J), this alternative promoter appeared to be associated with a significant increase in TRIM13 protein production after CHAF1B loss (Figure 1C). Reduction of TRIM13 expression, either by CRISPR-mediated Trim13 deletion or loss of the upstream exons leads to a slight but significant increase in the colony forming capacity of MLL-AF9 leukemic cells (supplemental Figure 1K).

Elevated levels of TRIM13 are associated with enhanced survival (Figure 1D). Analysis of the Bloodspot³⁸ database helped confirm our finding that TRIM13 was repressed in AML but upregulated in mature myeloid cells (Figure 1E). To determine whether CHAF1B/TRIM13 inverse expression also occurs in human AML, we first analyzed the Cancer Genome Atlas data and confirmed our observation that increased CHAF1B transcription correlates with decreased TRIM13 transcription (Figure 1F). Next, we compared BM aspirates from healthy donors to matched samples taken from patients either at diagnosis and relapse (initial/relapse) or diagnosis and remission (initial/remission) and observed the upregulation of CHAF1B in the leukemic marrows that coincided with the repression of TRIM13 (Figure 1G). In 1 patient with a matched initial diagnosis and remission sample AML3, CHAF1B returned to lower levels, whereas TRIM13 was re-expressed upon remission (Figure 1G). We also observed a similar pattern of high CHAF1B and low TRIM13 expression in unmatched AML samples (supplemental Figure 2). These findings suggest that concurrent upregulation of CHAF1B and repression of TRIM13 may be proleukemic. To test this, we knocked down TRIM13 (Figure 1H) and caused the overexpression of CHAF1B (Figure 1I) together in CD34+ PBSCs from healthy donors and measured the effects on colony formation in vitro. We found that colony formation capacity was significantly enhanced in PBSCs when CHAF1B was overexpressed and TRIM13 was repressed (Figure 1J).

We confirmed, via chromatin immunoprecipitation sequencing analysis, that CHAF1B occupies the human TRIM13 promoter in multiple AML cell lines (Figure 2A).⁵ We noted that the exon distribution around the promoter of the Homo sapiens TRIM13 was different compared with that of M. musculus Trim13 (supplemental Figures 1C and 2A), so we continued the examination of TRIM13 function in human models. U937 cells showed increases in TRIM13 expression at both the protein level (Figure 2B) and the mRNA level (Figure 2C) after knockdown of CHAF1B. CHAF1B overexpression in U937 cells led to the repression of TRIM13 mRNA production (Figure 2D), which was dependent on the binding ability of CHAF1B chromatin because overexpression of the CHAF1B^RR482/483AA (CHAF1B^RRAA) chromatin–binding-deficient mutant³⁹ was unable to repress TRIM13 transcription (Figure 2E). We confirmed this relationship in the MOLM13 AML cell line (Figure 2F-G). We noted that all changes in TRIM13 expression occurred before differentiation-driven changes in the cell cycle after CHAF1B overexpression or knockdown at 72 hours (Figure 2H). Because CHAF1B (working through the CAF1 complex) exerts its transcriptional repressive function at the chromatin, we hypothesized that CHAF1B represses TRIM13 by occupying its promoter in the chromatin. To test this, we cloned the promoter sequence bound by CHAF1B (Figure 2A) into a promoterless lentivirus to drive firefly luciferase (Figure 2I), with which we cotransduced cells, lentiviruses encoding a doxycycline-induced CHAF1B-T2A-CHAF1A (Figure 2J) lentivirus, and a Renilla luciferase control. We observed a slight but significant reduction in firefly luciferase activity upon CAF1 overexpression (Figure 2K), suggesting that promoter binding is likely responsible for some of the transcriptional repression of TRIM13.

TRIM13 expression is a marker of good prognosis in the the Cancer Genome Atlas AML data set (Figure 1D-E). Conversely, CHAF1B is a marker of poor prognosis in AML.⁵ To test the negative effects of enhanced TRIM13 expression on AML leukemogenicity, we caused the overexpression of TRIM13 in several AML cell lines (Figure 3A). This proved detrimental for long-term productivity (Figure 3B), as demonstrated by the reduced colony forming capacity (Figure 3C) and a small increase in apoptosis (Figure 3D; supplemental Figure 3A). To test whether this was because of differentiation, we used a panel of potential differentiation markers (Figure 3E-F; supplemental Figure 3B-E) and found that the overexpression of TRIM13 was able to induce CD14 expression in both U937 and MOLM13 AML cell lines (Figure 3G). We then demonstrated that the overexpression of TRIM13 repressed the colony formation capacity in 5 different primary samples of patients with AML (Figure 3H-I). We also caused TRIM13 to be overexpressed in U937 and MOLM13 cell lines and engrafted them into recipient NSG mice (Figure 3J-K). BM aspirates 26 days after the transplant revealed a significant reduction in TRIM13-overexpressing AML cells in the BM (Figure 3L), which coincided with a slight but significant extension of survival (Figure 3M). Together, these results suggest that upregulation of TRIM13 is detrimental to leukemic progression.

Because TRIM13 overexpression represses leukemogenesis, and low levels of TRIM13 are associated with poor outcome in AML, we hypothesized that TRIM13 may act as a putative tumor suppressor in AML. To test this, we knocked down TRIM13 in a pediatric PDX of AML (CBFA2T3-GLIS2 driven AML).⁴⁰ We were able to expand TRIM13 knockdown cells (Figure 4A) and found that TRIM13 knockdown drove the enhanced expansion of leukemic cells within the BM of recipient mice (Figure 4B) and contributed to a more aggressive disease (Figure 4C). We then confirmed the role of TRIM13 in 7 different pediatric AML samples (supplemental Table 2) by knocking down TRIM13 and assaying the colony forming capacity in vitro (Figure 4D), which led to an increase in colony formation rates in all samples (Figure 4E).

We used CRISPR editing to generate clonally derived populations of in-frame RING domain deletion mutants (T13RINGdel) to inactivate the catalytic activity of the endogenous TRIM13 loci (Figure 4F). We generated 2 independent clones of U937, HL60, and THP1 cell lines each with in-frame RING domain deletions (Figure 4G). T13RINGdel did not appear to have a negative effect on the truncated protein stability (Figure 4G). Mutant clones had increased colony forming capacity (Figure 4H) and reduced cell death (Figure 4I). Loss of TRIM13 catalytic activity was also sufficient to reduce the surface CD14 expression levels in the AML cells lines, suggesting a slightly less differentiated cell state (Figure 4J).

CHAF1B loss is associated with both enhanced TRIM13 expression (Figures 1C and 2C,F) and loss of colony formation capacity (supplemental Figure 4B-C). To test whether TRIM13 upregulation contributed to the colony formation loss phenotype of CHAF1B-depleted AML cells (supplemental Figure 4A), we performed a pathway suppressor experiment to determine whether ablation of TRIM13 function could synthetically rescue the loss of the colony formation capacity from CHAF1B-depleted AML cell lines. We found that inactivation of TRIM13 led to a partial rescue of the colony formation capacity in CHAF1B-depleted cell lines (supplemental Figure 4B). Despite the different promoter organization between mouse and human TRIM13, we were able to confirm thefunctional conservation of TRIM13 between the 2 species using Chaf1b deletion in our MLL-AF9 AML cell lines with either TRIM13 knocked out or upstream exons deleted (supplemental Figure 4C). These data, when taken together with our finding that TRIM13 upregulation is detrimental to leukemogenesis (Figure 3), suggest that TRIM13 plays a putative tumor suppressor role in AML.

TRIM13 has previously been reported to localize to the ER through a transmembrane domain on TRIM13 C-terminus.⁴¹ To determine its localization in AML, we used imaging cytometry to assay TRIM13 localization in wild-type T13 (T13WT), T13RINGdel U937, HL60, and THP1 cells. We first confirmed the specificity of our antibody for endogenous TRIM13 (Figure 5A). We found that TRIM13 was predominantly localized at the nucleus in the U937 cells, a localization pattern that was not dependent on presence of the RING domain (Figure 5B-C). We did note some perinuclear staining that may correspond to previously described ER localizations,^42,43 but this was found in a minority of the cells. This localization pattern was almost identical in THP1 and HL60 cell lines (Figure 5D), PDX models, and healthy mobilized CD34⁺ PBSCs (Figure 5E), suggesting that this may be a feature of immature hematopoietic cells. To determine whether nuclear localization of TRIM13 was necessary for its role as an antileukemic repressor, we caused a TRIM13 allele tagged with a nuclear exchange signal to be overexpressed to promote cytoplasmic expression. We confirmed the activity of the nuclear exchange signal tag via imaging cytometry (Figure 5F-G). Cytoplasmic localization of TRIM13 did not lead to a loss in clonogenic capacity of the AML cell lines, which led us to conclude that it is indeed the nuclear localization of TRIM13 that is responsible for its repressive phenotype in AML (Figure 5H). We then found that the most intense TRIM13 staining occurred in the nucleus (Figure 5I) and that TRIM13 staining intensity increased with progression through the cell cycle (Figure 5J), suggesting that it may play a role in cell cycle regulation.

Previous studies have linked TRIM13 to driving the degradation of target proteins via ubiquitin-linked proteasomal degradation.^19,36,44,45 We chose to focus on U937 cells for the mechanistic study of TRIM13 in AML because these cells responded most similarly to the PDX lines and displayed the greatest regulatory effect of TRIM13 by CHAF1B of the cell lines we tested. To identify potential degradation targets of TRIM13, we assayed the global proteome of T13WT and T13RINGdel U937 cells (supplemental Figure 5A; supplemental Table 3). To increase our depth of analysis, we also measured the nuclear/cytoplasmic fractions (Figure 6A; supplemental Figure 5B-C). To identify the specific degradation targets of TRIM13 because of ubiquitination, we also compared the differentially ubiquitinated peptides in T13RINGdel U937 cells vs T13WT (Figure 6B; supplemental Table 4). This yielded a substantial number of peptides with a loss of ubiquitination in T13RINGdel cells, suggesting that TRIM13 may have a wide variety of catalytic targets. To find potential degradation targets of TRIM13, we searched the overlap of proteins with both increased expression and loss of a ubiquitinated peptide in T13RINGdel U937 cells (25 total proteins; Figure 6C; supplemental Figure 5D). To attempt the confirmation of a target most likely to be proleukemic based on expression in adverse outcomes in AML, we selected the premRNA splicing factor WTAP. WTAP expression is increased in adverse outcome AML (supplemental Figure 5E), but we were not able to detect a significant change in expression in our nuclear or cytoplasmic fractions of the U937 AML cell line (supplemental Figure 5A).

Because of the small number of proteins that fit our original criteria, this led us to consider that TRIM13 may have a novel function in AML. Gene ontology analysis of our ubiquitination screen revealed a strong enrichment in proteins associated with cell cycle regulation (Figure 6D). We confirmed that T13RINGdel repressed cell cycle entry and TRIM13 overexpression enhanced cell cycle entry in U937 and HL60 cell lines (Figure 6E; supplemental Figure 5F). This led us to hypothesize that TRIM13 may be stabilizing cell cycle proteins through ubiquitination, and we found several cell cycle genes (CCNA1, CCNB1, and CDK1) containing peptides with high levels of ubiquitination in T13WT and undetectable ubiquitination in T13RINGdel (supplemental Table 4). We focused on CCNA1 because of its high degree of differential ubiquitination in T13RINGdel cells and its role in driving cell cycle progression and found that colocalization of CCNA1 and TRIM13 (Figure 6F) increased in a cell cycle stage–dependent manner in AML samples and cell lines (Figure 6G; supplemental Figure 5G).

We also found that the protein products of these key cell cycle regulatory genes are repressed in T13RINGdel U937 cells and TRIM13 knockdown PBSCs (and U937), compared with the mRNA products (Figure 6H; supplemental Figure 5H-I). These data suggest that TRIM13 may stabilize these cell cycle targets through its catalytic capacity as a ubiquitin ligase. To test this hypothesis, we caused the expression of CCNA1 using a doxycycline inducible lentivirus and found that T13RINGdel U937 cells were unable to sustain the levels of CCNA1 expression in T13WT (Figure 6I). In addition, we were able to functionally reverse the increase in clonogenic capacity of T13RINGdel U937 cells by re-expressing CCNA1 (Figure 6J).

To test whether TRIM13 was involved in stabilizing CCNA1 through ubiqutination, we caused the epitope-tagged TRIM13 and CCNA1 to be overexpressed in 293T cells and demonstrated that these 2 proteins could form a precipitable complex (Figure 7A-B). We next made domain-deletion mutants of TRIM13 and found that the canonical protein-protein interaction domains (coiled-coil and B-box) of TRIM13 were predominantly responsible for mediating the interaction with CCNA1 (Figure 7C). We then found that CCNA1 is capable of being ubiquitinated by TRIM13 through most K-linkage conditions with 2 main exceptions, K11 and K33 (Figure 7D; supplemental Figure 6A). In addition, recombinant TRIM13 was able to directly ubiquitinate recombinant CCNA1 in cell-free ubiquitination assay using an E1 and an E2 panel. These experiments revealed that TRIM13 can use multiple E2s to ubiquitinate CCNA1, except for UbcH5a (Figure 7E; supplemental Figure 6B), which may explain the many different K-linked ubiquitin chains that we found built via CCNA1 by TRIM13 in Figure 7D. Because several of these noncanonical ubiquitination chains (including K27-linked) have been previously associated with protein stability and signaling,⁴⁶ and TRIM13 catalytic inactivation was causing reduction in CCNA1 protein levels, we hypothesized that TRIM13-mediated ubiquitination was stabilizing CCNA1. In support of this hypothesis, it was found that the half-life of CCNA1 was reduced in T13RINGdel U937 cells exposed to a cycloheximide chase (Figure 7F-G). To test which degradation pathway of CCNA1 was being prevented by TRIM13, we inhibited the activity of either the proteosome (using MG-132, CTNNB1 as positive control for proteosome inhibitor function) or the lysosome (using Bafilomycin, LC3 as positive control for lysosome inhibitor function) and found that the TRIM13 catalytic function was to protect CCNA1 from lysosomal degradation (Figure 7G-H). We further confirmed lysosomal degradation of CCNA1 in the absence of TRIM13 by rescuing protein levels with chloroquine treatment (supplemental Figure 6C). Finally, we significantly rescued CCNA1 degradation in T13RINGdel U937 cells with lysosomal inhibition (Figure 7J-K). Together, these data demonstrate that TRIM13 stabilizes CCNA1 through ubiquitination, preventing its degradation by the lysosome.

In this study, we found that CHAF1B exerts its proleukemic function by suppressing the transcription of TRIM13 by binding to the promoter (Figures 1 and 2). We then found that TRIM13 is a putative tumor suppressor because overexpression is detrimental to leukemogenesis (Figure 3), and the loss of TRIM13 catalytic activity enhances AML tumorigenicity concurrent with an immature phenotype (Figure 4). We found that TRIM13 has a novel nuclear localization in AML that is necessary for its tumor suppressor function (Figure 5) and that TRIM13 catalyzes the ubiquitination of many different cell cycle regulators (Figure 6). The catalytic ubiquitination activity of TRIM13 on these cell cycle regulators promotes their stability. We then found that TRIM13 can directly interact with CCNA1 and promote a variety of different K-linked ubiquitination events associated with protein stability by preventing lysosomal degradation (Figure 7).

In our AML models, we found that K27-linked ubiquitination of CCNA1 by TRIM13 led to a reduced degradation of the CCNA1 protein product by the lysosome (Figure 7). What was most striking, though, was that (1) CCNA1 could be ubiquitinated by essentially every known K-linkage of ubiquitin and (2) TRIM13 was promiscuous in the E2 required to catalyze CCNA1 ubiquitination in a cell-free system (Figure 7). In other models, when K63 ubiquitinated, TRIM13 has been shown to drive ER-phagy to degrade aggregates of proteins resistant to the normal autophagic degradation by the lysosomal degradation pathway.⁴⁷ In addition, other studies have shown that clonogenicity can be negatively affected by TRIM13 because of ER stress and induction of autophagy via p62 in the adherent cell lines HEK293 and HeLa, though it should be noted that these are quite different from leukemic cells.^43,47 Our proteomics data showed significant changes in the ubiquitination of lysine residues in these proteins that are not canonically associated with degradation, therefore we propose that this is a result of stabilizing ubiquitination by TRIM13 on the other residues (supplemental Table 4). Another possibility is that the ubiquitination of these factors by TRIM13 in the nucleus (instead of in the ER⁴²) is contributing to their stability. In conclusion, we mechanistically demonstrated how the chromatin assembly machinery regulates the transcription of genes associated with cell cycle in a manner that promotes the immature state of leukemic cells.

CHAF1B (and the overall CAF1 complex) are functionally associated with the cell cycle, facilitating the assembly of histone H3/H4 heterodimers at the replication fork during S-phase. However, neither the overexpression nor depletion of CHAF1B alone is sufficient to drive changes in cell cycle in AML before differentiation (Figure 2H).⁵ In addition, these data suggest that CHAF1B loss may not lead to an immediate reduction in cell cycling in AML because the associated upregulation of TRIM13 promotes cell cycle entry and proliferation. This is supported by our observation that TRIM13 loss partially rescues the colony formation capacity of CHAF1B-depleted cells (supplemental Figure 4). When CHAF1B is depleted in cells lacking functional TRIM13, we propose that the progression into cell cycle is slowed, thereby allowing cells to potentially activate parallel antidifferentiation pathways and maintain their state throughout the process of cell division.

Some mutations or modulations observed in AMLs that cause precocious entry into cell cycle (FOXM1 deletion, a recent example) also cause loss of self-renewal and eventual exhaustion of the malignant stem cell pool.^48,49 In our AML models, overexpression of TRIM13 leads to an increase in cell proliferation coupled with a loss of colony formation and in vivo leukemogenesis. Our observation that loss of TRIM13, in contrast, leads to both increased colony formation and a reduction in the number of cells entering cell cycle, suggests that it is beneficial for leukemic cells to control cell cycle entry by downregulating TRIM13. However, our experiments are unable to conclude if this effect is predominating in leukemia stem or progenitor cells. Based on our data, we propose that TRIM13 is detrimental to leukemogenesis in general by promoting premature entry into or progression through cell cycle.

A patient with Trisomy 21 developed acute megakaryoblastic leukemia in the absence of GATA1s, and it was later uncovered that there was a biallelic deletion of 13q14.2 and 13q14.3, the region that includes TRIM13. Though this study attributed acute megakaryoblastic leukemia progression to RB1 loss, which also resides in that region, our studies suggest that the TRIM13 loss could also contribute to the proleukemic state of those cells.⁵⁰ Several studies suggest that TRIM13 promoter methylation is predominantly responsible for its transcriptional repression in some models.^51-53 Although a reduction in TRIM13 expression is seen in the tumor of patients with CLL, the loss of transcriptional activity does not appear to be caused by promoter methylation.⁵⁴ In contrast, a study focusing on pediatric patients with AML FAB M5 who showed specific patterns of hypermethylation around TRIM13, there was no significant change in the TRIM13 (or RFP2/DLEU5 by their notation) expression, although the DLEU1/2 genes were affected.⁵³ We interpret our data considering this study by suggesting that TRIM13 is being repressed by CHAF1B binding to its promoter.

The authors acknowledge Nicolas Nassar for the thoughtful discussions and Charlie Nims and Levi Fox for their participation in the laboratory activities.

This work was supported by a seed grant from Ohio Cancer Research as well as National Institutes of Health grants (National Cancer Institute R00CA230314 and National Institute of General Medical Sciences R35GM142452) (A.G.V.). A.G.V. is supported by a Special Fellow Award from Leukemia and Lymphoma Society as well as a Junior Faculty Scholar Award from American Society of Hematology. All flow cytometric data were acquired using equipment maintained by the Research Flow Cytometry Core in the Division of Rheumatology at Cincinnati Children’s Hospital Medical Center, and the authors would like to acknowledge the support of the Coordinating Center of Excellence in Hematology grant U54DK126108 to Cincinnati Children’s Hospital Medical Center. The Proteomics data were collected on a mass spectrometer that was supported in part by a National Institutes of Health S10 Shared Instrumentation grant (S10OD026717) (K.D.G.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

We dedicate this work to Mattie Louise Martin (1921-2020).

Contribution: S.T.D., C.I., X.Z., S.W., J.K.J., and T.N. contributed to formal analysis, investigation, methodology, and manuscript review and editing; S.H. contributed to the investigation, methodology, and manuscript review; M. Wyder contributed to the investigation, mass spectrometry methodology, and manuscript review; N.S. contributed to the methodology, conceptualization, and RNA sequencing splicing analysis; M. Wunderlich contributed to resource acquisition, methodology, conceptualization, in vivo and PDX investigation, and manuscript review and editing; K.D.G. contributed to the methodology, conceptualization, mass spectrometry investigation, and manuscript review and editing; D.T.S. contributed to the formal analysis, methodology, and manuscript review and editing; and A.G.V. contributed to the funding and resource acquisition, formal analysis, investigation, methodology, writing of the original draft, and project administration.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Andrew G. Volk, Division of Experimental Hematology and Cancer Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229; e-mail: andrew.volk@cchmc.org.