The lost intrinsic fragmentation of MAT1 protein during granulopoiesis promotes the growth and metastasis of leukemic myeloblasts

Abstract

MAT1, an assembly factor and targeting subunit of both cyclin-dependent kinase-activating kinase (CAK) and general transcription factor IIH (TFIIH) kinase, regulates cell cycle and transcription. Previous studies show that expression of intact MAT1 protein is associated with expansion of human hematopoietic stem cells (HSC), whereas intrinsically programmed or retinoic acid (RA)-induced MAT1 fragmentation accompanies granulocytic differentiation of HSC or leukemic myeloblasts. Here we determined that, in humanized mouse microenvironment, MAT1 overexpression resisted intrinsic MAT1 fragmentation to sustain hematopoietic CD34+ cell expansion while preventing granulopoiesis. Conversely, we mimicked MAT1 fragmentation in vitro and in a mouse model by overexpressing a fragmented 81-aa MAT1 polypeptide (pM9) that retains the domain for assembling CAK but cannot affix CAK to TFIIH-core. Our results showed that pM9 formed ΔCAK by competing with MAT1 for CAK assembly to mimic MAT1 fragmentation-depletion of CAK. This resulting ΔCAK acted as a dominant negative to inhibit the growth and metastasis of different leukemic myeloblasts, with or without RA-resistance, by concurrently suppressing CAK and TFIIH kinase activities to inhibit cell cycle and gene transcription. These findings suggest that the intrinsically programmed MAT1 expression and fragmentation regulate granulopoiesis by inversely coordinating CAK and TFIIH activities, whereas pM9 shares a mechanistic resemblance with MAT1 fragmentation in suppressing myeloid leukemogenesis.

Introduction

The human CAK is a trimeric complex consisting of CDK7 [1], cyclin H [2], and MAT1 [3, 4]. CAK was originally implicated in regulating cell cycle by phosphorylation-activation of different CDKs [5, 6] while mediating cell cycle G₁ exit through phosphorylation-inactivation of the retinoblastoma tumor suppressor protein (pRb) [7]. It has now demonstrated that CAK exists in cells either as a free CAK or as the kinase subunit of the TFIIH complex [8-11], where it mediates transcription through phosphorylation of the RNA polymerase II (Pol II) C-terminal domain (CTD) [8, 12-14]. Thus, CAK is considered a crossroads regulator that coordinates cell cycle control with transcriptional regulation [6, 11].

Many studies have shown that CAK-coordinated cell cycle and transcriptional regulation are mediated by MAT1. Egly's group characterizes the distinct regions of MAT1 involved in mediating CAK-dependent cell cycle and transcriptional activities [15]. They find that: a) the MAT1 C-terminus is required to assemble and activate CAK; b) the coiled-coil motif of the median portion of MAT1, by virtue of its interaction with TFIIH-core's XPD and XPB helicases, anchors CAK to the TFIIH-core for enabling CAK to serve as the TFIIH kinase; and c) the N-terminal RING domain of MAT1 is required for the TFIIH kinase to mediate transcription via phosphorylation of Pol II CTD. Indeed, MAT1 is required for assembling CAK [3, 4], determining CAK's substrate-specificity [7, 9, 16-18], shifting CAK substrate-preference from CDK2 to Pol II [16, 17], and mediating TFIIH kinase phosphorylation of Pol II [12]. Moreover, antisense abrogation of MAT1 induces cell cycle G₁ arrest [19], whereas mouse MAT1-/- cells fail to enter S phase and are defective in Pol II phosphorylation [20]. However, despite such emerging knowledge, exactly how the distinct structural regions of MAT1 are coordinated to mediate the cell cycle and transcriptional activities of CAK and TFIIH kinase remains unclear, especially in the content of human diseases.

Others and we have reported that MAT1 mediates CAK activity in cell cycle, transcription, and differentiation [3, 7, 9, 16, 17, 20-23]. We found that in myeloid leukemia cells, RA-induced MAT1 degradation suppresses CAK phosphorylation of pRb [22, 23] and RA receptor alpha (RARα) [21-24], leading to cell cycle arrest, release of transcriptional suppression, and induction of granulocytic differentiation [21, 22, 24]. Such RA-induced MAT1 degradation results in both decreased MAT1 levels and fragmenting of the 37-kDa intact MAT1 to a major 30-kDa fragment (M30), as detected on SDS-PAGE [21]. This M30 is likely cleaved at MAT1 C-terminus so that a smaller 9-kDa fragment (so-called pM9) is consequentially generated [21]. Such pM9 fragment covers the MAT1 C-terminus, which is required for assembling and activating CAK [15]. In fact, pM9 is probably the “minimal” fragment that breaks at the C-terminal 229-aa position and remains associated within the CAK complex after spontaneous degradation of MAT1 [15], indicating a ΔCAK formation by this pM9 peptide. Our earlier studies show that RA-induced fragmentation of MAT1 decreases CAK assembly, leading to suppressing proliferation and inducing differentiation of leukemic myeloblasts [21, 22]. Of note, such RA-induced MAT1 degradation-fragmentation in leukemic myeloblasts is in fact proceeding intrinsically during normal granulopoiesis: MAT1 remains essentially intact in HSC, then is cleaved to generate M30 in common myeloid progenitors (CMP), markedly degraded in granulocyte/monocyte progenitors (GMP), and significantly fragmented to M30 in mature granulocytes [23]. These findings led us to hypothesize that the intrinsic coordination of MAT1 expression vs. fragmentation determines the choice of normal primitive hematopoietic expansion or granulocytic differentiation, whereas loss of MAT1 fragmentation promotes a myeloid leukemogenic state. We substantiate this concept by using a humanized mouse model of granulopoiesis as well as xenotransplantation of leukemic myeloblasts expressing a fragmented C-terminal portion of MAT1, the pM9 peptide that possesses a mechanistic resemblance with MAT1 fragmentation-depletion of CAK assembly by forming of ΔCAK. Our studies discovered that the intrinsically programmed MAT1 expression and C-terminal fragmentation coordinate hematopoietic expansion and granulopoiesis by inversely modulating CAK-dependent cell cycle and transcriptional activities, whereas pM9-mimicked MAT1 fragmentation-depletion of CAK assembly via ΔCAK formation suppresses the growth and metastasis of different leukemic myeloblasts, with or without RA-resistance.

Materials and Methods

Human cells and cell lines

See Supplemental Information for details.

Humanized granulopoiesis mouse model and xenografting of leukemic myeloblasts

Mouse work was performed according to guidelines under protocols approved by the Children's Hospital Los Angeles Institutional Animal Care and Use Committee. Eight-week-old NOD/SCID/IL-2Rγ^null (NSG) mice (Jackson Laboratory, Bar Harbor, ME) were sub-lethally irradiated with 250-cGy 4 hr before transplantation. Each 6 × 10⁵ CD34+ cells expressing lentiviral pCCL-MAT1-GFP or pCCL-GFP were co-transplanted with 2 × 10⁵ human mesenchymal stem cells (hMSC) through intravenously injection of the tail-vein. Daily intraperitoneal injection of RA (2 mg/kg/day) into mice transplanted with CD34+ cells expressing pCCL-GFP vector was served as dual-control to evaluate granulopoiesis while monitoring vector-effect. After 7, 14, and 21 days of transplantation, cells were collected from peripheral blood (PB) and bone marrow (BM), as described [25, 26], for analysis of the GFP+ donor subpopulation, BM reconstitution, CD marker expression, CAK signaling molecules, and granulocytic morphologic differentiation. Xenografting of leukemic myeloblasts is detailed in Supplemental Information.

Immunofluorescence, immunochemistry, immunoprecipitation, western blot (WB) analyses and liquid chromatography-tandem mass spectrometry (LC-MS/MS) identification

See Supplemental Information.

Cell proliferation and morphologic analyses of granulocytic differentiation

The experiments were performed as described [23, 27] and detailed in Supplemental Information.

Quantitative real-time polymerase chain reaction (qRT-PCR)

qRT-PCR analysis was performed, as described [27], on the 7900HT FastqRT-PCR System (Applied Biosystems, Carlsbad, CA). The GAPDH was used as an internal control for normalization of RNA quantity. Primers and amplification conditions are provided in Supplemental Tables 1 and 2.

Flow cytometric analysis and cell sorting

Antibodies against human CD markers (PE-CD34, PE-CD66, APC-CD11b, and APC-CD19) as well as corresponding APC and PE isotypes are from BD Biosciences PharMingen (Franklin Lakes, NJ). The analyses were detailed in Supplemental Information.

Lentiviral vector construction, virion production, and transduction

MAT1 cDNA was cloned into a pCCL-c-MNDU3c-X2-PGK-eGFP lentiviral vector, as described [23]. Virion production, titration, and transduction were described before [23, 24]. The cloning of lentiviral pM9 is outlined in Supplemental Information.

Fluorescence microscopy analysis, HE staining, and statistical analysis

See Supplemental Information.

Results

MAT1 overexpression sustains proliferation while suppressing granulocytic differentiation of CD34+ cells in vitro

We previously found that significant degradation of MAT1 protein to M30 fragment is associated with myeloid differentiation of HSC toward mature granulocytes [23]. To test that intrinsically programmed MAT1 fragmentation is required for normal granulocytic differentiation, we blocked such MAT1 fragmentation-dependent granulopoiesis of hematopoietic precursors by overexpressing lentiviral pCCL-MAT1-GFP in CD34+ cells (Supplemental Figure 1A). The results showed that a higher rate of proliferation and a decreased cell doubling time of CD34+ cells (Figure 1A) underlay increased levels of MAT1 protein (Figure 1B, lanes 5, 6 vs. 7). Moreover, MAT1 was progressively degraded as CD34+ cells proceeding toward granulopoiesis from day 6 to 12 in myeloid medium (Figure 1B, lanes 1 vs. 2 vs. 5), similar to the progressive degradation of MAT1 observed in HSC to CMP to GMP [23]. Moreover, overexpressed MAT1 was associated with the elevated levels of either CDK7 or RARα (Figure 1C, lane 4). Interestingly, in the presence of MAT1 overexpression, only about 7% of cell population was reduced in G₀/G₁ phase together with an increased 21% of G₂/M cells undergoing mitosis (Figure 1D, sections I vs. IV). These data suggest that MAT1 promotes expansion of CD34+ precursors mainly by shortening cell doubling time (Figure 1A, right section). In addition, MAT1 overexpression inhibited granulocytic morphologic differentiation (Figure 1E, 1F) and suppressed expression of CD11b and CD66, the maturation markers of granulocytic differentiation (Figure 1G). Together, these in vitro data demonstrate that higher expression of MAT1 in CD34+ cells sustains hematopoietic expansion while inhibiting granulocytic differentiation.

Figure 1. MAT1 overexpression sustains proliferation while suppressing granulocytic differentiation of CD34+ cells in vitro.

(A) Proliferation and doubling time of CD34+ cells overexpressing MAT1. *, MAT1 vs. control, P<0.05; MAT1 vs. vector, P<0.05. (B) WB analysis of MAT1 in CD34+ cells in the presence or absence of overexpressed MAT1. A significant intrinsic fragmentation of intact MAT1 was observed at day 12. (C) WB depicted levels of CDK7, RARα, and MAT1 in CD34+ cells overexpressing MAT1 or treated with RA. (D) Flow cytometric analysis of cell cycle progression of CD34+ cells overexpressing MAT1 or treated with RA. (E) Granulocytic morphologic differentiation of CD34+ cells assessed by light microscopy. Scale bar, 5 μm. (F) Quantification of granulocytic morphologic differentiation in panel E. *, P<0.05. (G) Expression of granulocytic differentiation markers CD11b and CD66 in CD34+ cells transduced with MAT1 or treated with RA, as determined by flow cytometric analysis.

MAT1 overexpression sustains hematopoietic expansion and suppresses granulocytic differentiation in humanized mouse microenvironment

Increasing evidence has shown that signaling from the osteoblastic HSC niche, derived from hMSC, is required for human granulopoiesis [28-30]. Thus, hMSC engrafted in immuno -deficient mice provide a suitable HSC niche for studying human hematopoiesis [30-32]. We therefore established a humanized mouse microenvironment by co-engrafting CD34+ cells together with hMSC into NSG mice for studying MAT1-modulated granulopoiesis. After 21 days of co-transplantation, the adherent bone-associated cells were liberated and sorted for determining the efficiency of hematopoietic reconstitution of donor CD34+ cells expressing lentiviral pCCL-GFP. We found that in the presence of hMSC, 2.81% of the total BM cells were GFP+, compared to only 0.13% in the samples without hMSC (Figure 2A, left section, panels 4 vs. 8). This model was also responsive to RA stimulation and MAT1 overexpression, as shown by that MAT1 overexpression significantly enhanced BM reconstitution of transplanted GFP+ donor cells, whereas RA inhibited such engraftment (Figure 2A, right section, panels 12 vs. 16; Figure 2B). Furthermore, we collected PB from the host mice on day 7, 14, and 21 to determine their progressive changes in granulopoiesis vs. hematopoietic expansion underlying RA stimuli and MAT1 overexpression. FACS analysis of these cells showed that GFP+ cells were increased in samples co-transplanted with hMSC, peaking on day 14, compared to samples lacking hMSC (Figure 2C, left section). This increase was augmented by MAT1 overexpression but inhibited by RA stimulation (Figure 2C, right section, panels 3, 7, 11 vs. 4, 8, 12), as shown by statistical analysis (Figure 2D). All together, these results demonstrate that the humanized hMSC mouse model can promote BM reconstitution of hematopoietic precursors, enhance expansion of GFP+ donor cells in PB, and is responsive to either RA stimuli or MAT1 overexpression.

Figure 2. Humanized mouse microenvironment enhances bone marrow reconstitution of CD34+ cells and responds to stimulation of MAT1 overexpression or RA treatment.

(A) Fluorescent images showing the sorted GFP+ BM cells (panels 1-3, 5-7, 9-11, 13-15). BM reconstitution of GFP+ donor cells was promoted by hMSC (left section, panels 4 vs. 8). The levels of GFP+ donor cells in BM were decreased with intraperitoneal injection of RA but increased by overexpressing MAT1 (right section, panels 12 vs. 16). (B) Statistical analysis of BM engraftment of GFP+ donor cells in response to MAT1 overexpression or RA stimulation. *, compared to vector or vector + RA sample, P<0.05. (C) Fluorescence visualization of the sorted GFP+ PB cells (top cell-image panel). GFP+ donor cells in PB were increased in the presence of hMSC (left section, panels 1, 5, 9 vs. 2, 6, 10). GFP+ donor cells in PB were decreased with intraperitoneal injection of RA but increased by overexpressing MAT1 (right section, panels 3, 7, 11 vs. 4, 8, 12). (D) Quantification of GFP+ donor cells in PB. *, P<0.05; **, P<0.01.

Using the humanized mouse model described above, we next assessed whether MAT1 overexpression could overcome intrinsic MAT1 degradation to sustain hematopoietic expansion while inhibiting granulopoiesis of CD34+ cells. CD34+ cells transduced with lentiviral pCCL-MAT1-GFP or pCCL-GFP vector were co-transplanted together with hMSC into NSG mice. Knowing of the evident effect of RA on mediating MAT1 degradation to induce granulopoiesis of both normal and malignant hematopoietic precursors [21-24], an intraperitoneal injection of RA for mice co-transplanted with hMSC and CD34+ cells expressing vector was used as a dual-control to examine in vivo granulopoiesis (vector + RA group) vs. hematopoiesis (vector group) vs. hematopoietic expansion (MAT1 group) while monitoring possible vector-effect. PB was collected for sorting of GFP+ donor cells at day 7, 14, and 21 post-transplantation. Flow cytometric analysis of CD marker expression showed that compared to vector and vector + RA samples, those transduced with MAT1 had the highest expression of the primitive CD34 marker while the lowest levels of either CD11b or CD66 at day 7, 14, or 21 (Figure 3A, 3B). The significant differences of these CD marker expressions were progressively developed from day 7 to 14 to 21 (Figure 3B, 3C). Interestingly, the levels of B-lymphocyte antigen CD19 were generally lower in MAT1 samples than those in vector or vector + RA samples and, at day 21, showed significant difference between MAT1 and vector samples (Figure 3B, 3C). Hence, these data indicate that with consistent higher levels of CD34 marker in MAT1 group than those in vector or vector + RA group, MAT1 sustains expansion of hematopoietic precursors while inhibiting either granulopoiesis or hematopoiesis or lymphopoiesis, as shown by significant lower expression of CD11b or CD66 or CD19. Moreover, In contrast to the markedly higher expression of MAT1 mRNA in samples transduced with MAT1, the mRNA levels of both CD11b and CD66 were significantly lower than in samples either treated with RA or expressing vector only (Figure 3D). Furthermore, samples with MAT1 overexpression exhibited significantly less granulocytic morphologic differentiation than did either cells expressing vector only or treated with RA (Figure 3E, 3F). Hence, as found in vitro, MAT1 expression in humanized mouse model repeats to maintain hematopoietic expansion while inhibiting granulopoiesis of hematopoietic precursors.

Figure 3. MAT1 overexpression sustains hematopoietic expansion and suppresses granulocytic differentiation in humanized mouse microenvironment.

(A) Representative FACS analysis of CD34, CD11b, CD66, and CD19 markers on GFP+ donor cells sorted from PB. Corresponding isotype staining of PB from non-transplanted or vector transplanted mice served as controls. (B) Quantification of CD marker expression on GFP+ donor cells sorted from PB. *, P<0.05 at least, compared with vector or vector + RA samples; #, P>0.05 compared with vector or vector + RA samples. (C) Column illustration of data in panel B. *, P<0.05; **, P<0.01; ***, P<0.001. (D) qRT-PCR depicted mRNA levels of MAT1, CD11b, and CD66 markers in GFP+ donor cells sorted from PB. *, P<0.05; **, P<0.01; ***, P<0.001. (E) Representative fluorescence visualization of granulocytic morphologic differentiation of GFP+ donor cells sorted from PB. Scale bar, 5 μm. (F) Statistical analysis of nuclear segmentation of GFP+ donor cells sorted from PB. *, P<0.05; **, P<0.01; ***, P<0.001. (G) Immunofluorescence detection of MAT1 (left section) and p21^Cip/Kip (right section) in GFP+ donor cells sorted from PB. DAPI was used for nuclear staining. Scale bar, 20 μm. (H) Quantification of MAT1 and p21^Cip/Kip levels in panels G. *, P<0.05; ***, P<0.001.

Because granulopoiesis is inhibited in transplanted CD34+ cells overexpressing MAT1 (Figure 3A-F), we determined the level of MAT1 protein in those GFP+ donor cells. Owing to a limited amount of GFP+ donor cells, we used immunofluorescence analysis to determine the levels of MAT1 protein in parallel to p21^Cip/Kip protein, the only known cell cycle inhibitor that its transcriptional expression is directly activated by RARα [33] during RA-induced granulocytic differentiation of malignant and normal hematopoietic precursors [23, 24, 34]. In parallel to vector controls (Supplemental Figure 2), we found that significantly higher level of MAT1 protein was indeed retained in samples transduced with pCCL-MAT1-GFP, compared to samples transduced with vector only (Figure 3G, left section; 3H). The increased level of MAT1 protein was associated with decreased p21^Cip/Kip expression (Figure 3G, right section; 3H), similar to the previous observations determined in an in vitro system [23, 24]. Together, these findings suggest that down-regulation of p21^Cip/Kip expression is one of the mechanisms by which MAT1 inhibits granulopoiesis and promotes the expansion of primitive hematopoietic precursors.

MAT1 is cleaved at C-terminus to generate M30 and pM9 fragments in human cells

Since MAT1 protein sustains hematopoietic expansion (Figures 1-3) whereas MAT1 fragmentation is associated with granulopoietic series [21-23], we asked question: what role does MAT1 fragmentation play in granulocytic differentiation? We first determined whether M30 indeed resulted from a C-terminal cleavage of MAT1 in human cells. LC-MS/MS sequencing analysis of MAT1 protein and M30 polypeptide, isolated from leukemic myeloblasts, showed that MAT1 protein is cleaved around 229-aa of the C-terminus (Supplemental Figures 3, 4). This result agrees with a previous report that a small recombinant MAT1 fragment is cleaved at 229-aa of the C-terminus in Sf9 cells [15], and supports our previous observations that either intrinsically programmed or RA-induced degradation of intact MAT1 generates M30 (covering the N-terminal and median portions of MAT1) and a 9-kDa C-terminal fragment (pM9) in either malignant or normal hematopoietic precursors [21, 23]. Based on the studies of MAT1 structure [15], pM9 is able to assemble CAK but cannot anchor CAK to TFIIH-core, owing to the lack of a coiled-coil motif at the median portion of MAT1 that interacts with the XPD and XPB subunits of TFIIH-core [12, 15].

pM9 forms ΔCAK to inhibit CAK assembly and reduce TFIIH kinase in different subtypes of myeloid leukemic cells

We previously found that either intrinsically programmed or RA-induced MAT1 fragmentation decreases the levels of MAT1 to inhibit MAT1-dependent CAK assembly and CAK activities in normal and malignant hematopoietic precursors, thereby leading to suppressing proliferation and inducing granulocytic differentiation [21-24]. In both cases, the fragmented M30 is consistently decreased whereas pM9 is rarely detectable, likely resulting from elimination of such unneeded forms of M30 and pM9 via ubiquitination-degradation [35]. These studies show that the inhibition of leukemic growth mediated by RA-dependent MAT1 fragmentation primarily results from depletion of MAT1-dependent CAK assembly [21, 22, 24], whereas the structure of the fragmented pM9 possesses the ability to assemble CAK [15]. Thus, pM9 could bypass RA-dependent MAT1 fragmentation to inhibit leukemogenesis by forming ΔCAK to eliminate MAT1-dependent CAK assembly. We therefore first tested whether pM9 forms ΔCAK via competing with endogenous MAT1 for CAK assembly in different subtypes of myeloid leukemia cells. We overexpressed lentiviral pCLS-Flag-pM9-GFP or pCLS-GFP (Supplemental Figure 5) in RA-sensitive AML HL60 cells, RA-resistant AML HL60R cells harboring a RARαΔAF-2 domain [21, 36], and CML JURL-MK1 cells with a BCR/ABL-fused gene [37]. WB analysis showed that in contrast to the barely detectable endogenous pM9, anti-Flag antibody recognized the Flag tag-fused pM9 in those cells (Figure 4A, lanes 1-3), which was associated with decreased levels of endogenous MAT1 compared to the controls (Figure 4A, lanes 4-9). Such positive detection of Flag-pM9 is likely resulting from both its overexpression as well as conformation change in structure due to its fusion with Flag and GFP, leading to resistance of protease-mediated degradation. Furthermore, immunoprecipitation with anti-Flag and anti-MAT1 antibodies in parallel revealed that pM9 forms ΔCAK (Figure 4B, lane 2). Such ΔCAK formation suppressed assembly of wild type CAK in those cells (Figure 4B, lanes 6 vs. 7, 8), likely through competing with MAT1 to consume endogenous CDK7 and cyclin H (Figure 4B, lane 6) while inhibiting endogenous MAT1 expression (Figure 4A, lanes 1-3), leading to reduction of CAK assembly.

Figure 4. pM9-formed ΔCAK inhibits CAK assembly in different subtypes of myeloid leukemic cells.

(A) WB analysis of Flag tag-fused pM9 and endogenous MAT1 in HL60, HL60R and Jurl-MK1 cells at the indicated time points. The band of 42-kDa represents the molecular mass of the fused Flag-pM9-GFP (lanes 1-3). Endogenous 37-kDa fragments of intact MAT1 were detected in all samples. (B) Immunoprecipitation analysis of pM9-anchored ΔCAK and endogenous CAK in parallel by using anti-Flag and anti-MAT1 antibodies, followed by WB detection of CAK components with antibodies against MAT1, cyclin H, and CDK7, respectively. PI, pre-immune IgG.

Because pM9 lacks the coiled-coil motif of the median portion of MAT1, which is required for interaction with TFIIH-core's XPD and XPB subunits to anchors CAK to the TFIIH-core [12, 15], we next examined both the interaction of ΔCAK with TFIIH-core and the effect of ΔCAK formation on assembly of TFIIH kinase. We used anti-Flag and anti-MAT1 antibodies in parallel to immunoprecipitate TFIIH complexes from HL60, HL60R, and Jurl-MK1 cells expressing pM9. WB analysis showed that both p62 and p89, the 2 of 10 subunits of TFIIH complex, were not detectable in the precipitates of anti-Flag antibody against Flag-pM9 (Figure 5A-C, lanes 2). In contrast, we detected both p62 and p89 in the precipitates of anti-MAT1 antibody recognizing both MAT1 and Flag-pM9, with significantly decreased levels in pM9 sample compared to controls (Figure 5A-C, lanes 6 vs. 7, 8). Hence, owing to pM9's lack of the motif affixing CAK with TFIIH-core, pM9-anchored ΔCAK cannot be recruited by TFIIH-core (Figure 5A-C, lanes 2). Moreover, pM9 expression decreases the levels of TFIIH kinase through its competitive effect on inducing ΔCAK formation to suppress MAT1-dependent CAK assembly in those myeloid leukemia cells (Figure 5A-C, lanes 6 vs. 7, 8).

Figure 5. pM9 expression inhibits CAK interaction with TFIIH-core in different subtypes of myeloid leukemic cells.

Immunoprecipitation detecting the interaction of TFIIH-core with either ΔCAK (lane 2) or endogenous CAK (lanes 6-8) using antibodies against Flag-tag and MAT1 in parallel. The levels of TFIIH kinase in the precipitates were determined by WB analysis by using antibodies against TFIIH-core components p62 and p89, respectively, in HL60 (A), HL60R (B), and Jurl-MK1 (C) cells.

Overexpression of pM9 inhibits the proliferation of myeloid leukemic cells by decreasing CAK-coordinated cell cycle and transcriptional activities

RA-resistant leukemic myeloblasts consistently show high levels of intact MAT1 protein [21, 24]. By contrast, MAT1 is typically degraded into M30 and pM9 fragments (Supplemental Figures 3, 4) in RA-sensitive leukemic cells or normal myeloid progenitors undergoing granulocytic differentiation [21-24]. These observations raise the intriguing possibility that in both RA-sensitive and RA-resistant leukemic myeloblasts, loss of MAT1 fragmentation is one of the mechanisms that sustain the leukemic state. Because pM9 not only can consume endogenous cyclin H and CDK7 to form ΔCAK but also can reduce the level of intact MAT1 protein (Figure 4), it may mimic MAT1 fragmentation to inhibit wild-type CAK and TFIIH kinase activities to suppress the growth of leukemic myeloblasts, with or without RA-resistance. To test this prediction, we overexpressed lentiviral pCLS-Flag-pM9-GFP or pCLS-GFP (Supplemental Figure 5) in HL60, HL60R, and Jurl-MK1 cells [21, 36, 37]. As predicted, pM9 significantly inhibited the proliferation of all those cells tested, as indicated by an increased cell doubling time without appreciable cell death (Figure 6A). With at least three times of independent cell cycle analyses showing similar results, we found that regardless of the varied cellular background in these different leukemic cell lines, pM9 inhibits the leukemic state by lengthening cell division time with or without inducing significant arrest of cell populations at specific cell cycle stages (Figure 6A, 6B). To determine if the inhibitory effect of pM9 on cell proliferation is related to decreased CAK activity due to ΔCAK formation, we first assessed changes in CDK7 phosphorylation at the total protein levels. WB analysis showed that expression of pM9 decreased CDK7 phosphorylation, as reflected by increased hypophosphorylation form of CDK7 (Figure 6C, left panel). Because, in addition to the presence of MAT1, CAK can also be activated through CDK7 phosphorylation on Ser164 and/or Thr170 residues in its T-loop [38], we further analyzed CDK7 phosphorylation within the ternary complexes. We found that anti-phosphoserine antibody detected a decreased CDK7 autophosphorylation in cells transduced with pM9 (Figure 6C, middle panel), whereas the phosphorylation of CDK7 Thr170 was also inhibited (Figure 6C, right panel). Of note, phosphorylation of other targeting substrates by CAK, including RARα [18, 21, 22, 24] and CDK1 [6, 39], was inhibited accordingly (Figure 6D). Moreover, because MAT1 is required for transcription mediated by TFIIH kinase-dependent phosphorylation of Pol II CTD [8, 12-14], we also examined whether suppressed expression of endogenous MAT1 and decreased levels of TFIIH kinase due to overexpression of pM9 (Figures 4, 5) alter Pol II CTD-mediated transcription. The results from qRT-PCR analysis showed that similar to findings reported by Mäkelä's group in mouse fibroblasts depleted with MAT1 [12], the mRNA levels of both MAT1 and PNRC2 were decreased, whereas CRABP2 was increased in HL60, HL60R, and Jurl-MK1 cells expressing pM9 (Figure 6E). Interestingly, the up-regulated MMP3 in HL60 cells did not appear in RA-resistant AML HL60R or CML Jurl-MK1 cells, possibly due in part to different cellular background [21, 36, 37]. Moreover, we also observed that phosphorylation of Pol II was inhibited in those cells expressing pM9 (Figure 6F). Together, these experiments show that the inhibitory effect of pM9 on proliferation of different leukemic myeloblasts results from depletion of intact MAT1-dependnet CAK assembly (Figure 4), by which pM9-anchored ΔCAK bypasses the resistance of intrinsic MAT1 fragmentation in these cells to suppress the pathologic activities of CAK and TFIIH kinase (Figure 6).

Figure 6. Overexpression of pM9 inhibits the proliferation of myeloid leukemic cells by decreasing CAK-coordinated cell cycle and transcriptional activities.

(A) Proliferation, death rate, and doubling time of HL60, HL60R and Jurl-MK1 cells expressing pM9 were monitored at different time points. **, P<0.01; ***, P<0.001. (B) Cell cycle progression status of HL60, HL60R and Jurl-MK1 cells at day 6 post-transduction of pM9. (C) WB analysis of phosphorylation status of CDK7 (left panel). CDK7 phosphorylation in the precipitated trimeric complexes was detected with WB using anti-phosphoserine (middle panel) and anti-phosphorylation of CDK7 Thr-170 (p-CDK7 Thr170) antibodies (right panel). Immunoprecipitation of both ΔCAK and endogenous CAK using antibodies against Flag tag and MAT1 was performed in parallel. (D) Expression of CAK substrates RARα and CDK1 as well as their phosphorylation statuses in HL60, HL60R and Jurl-MK1 cells expressing pM9 were determined by WB analysis. p-Serine, phosphoserine; p-CDK1, phosphorylated Thr-161 of CDK1. (E) qRT-PCR analysis of MAT1, CRABP2, PNRC2, and MMP3 mRNA levels in HL60, HL60R and Jurl-MK1 cells at day 6 post-transduction of pM9. *, P<0.05; **, P<0.01; ***, P<0.001. (F) WB analysis of Pol II. p-Pol II, hyperphosphorylated Pol II.

Proliferation and metastasis of leukemic myeloblasts are inhibited by pM9 in vivo

To further evaluate the effect of pM9 on inhibiting proliferation of leukemic myeloblasts, we studied NSG mice bearing xenografts of HL60R cells, which resist RA-induced inhibition of proliferation while maintaining high level of intact MAT1 [21, 24]. HL60R cells (5 × 10⁶) expressing lentiviral pCLS-Flag-pM9-GFP or pCLS-GFP (Supplemental Figure 5) were transplanted into NSG mice. After 34 days of engraftment, PB, BM, spleen, as well as tumors formed in lymph nodes were collected for evaluation of leukemic cell proliferation and metastasis. We found that GFP+ donor cells in PB from vector samples were significantly more than those in pM9 samples, while no significant change in GFP+ donor cells collected from BM or spleen (Figure 7A, 7B). Moreover, the vector-only samples had a much higher prevalence of lymph-node metastases than did the pM9-containing samples, where only a few small metastatic tumors were detected (Figure 7C, 7D). Interestingly, HE staining of the infiltrated lymph node showed multi-nucleated leukemic cells in pM9 but not in vector samples (Figure 7E). Because CDK1 activity mediated by CAK is involved in regulating cell cycle G₂/M progression [6, 39], suggesting that pM9 expression suppresses CDK1 activity in regulating G₂/M transition in the xenografted cells, as we observed in HL60R cells transduced with pM9 in vitro (Figure 6B, 6D). Of note, RARα phosphorylation was decreased in mouse lymph node tissues filtrated with HL60R cells expressing pM9 (Figure 7F). Moreover, by using high-resolution fluorescence microscopy analysis of the consecutive tissue sections in parallel to HE staining (Figure 7E), we found that either pM9-GFP or GFP vector was localized in nuclei of metastasized HL60R cells in the mouse lymph nodes (Figure 7G). Correspondingly, these migrated human cells were recognized by anti-human CD45 antibody (Figure 7H). Together, these in vivo data demonstrate that pM9 can overcome the resistance of MAT1 fragmentation in RA-resistant cells to inhibit proliferation and metastasis.

Figure 7. Proliferation and metastasis of leukemic myeloblasts are inhibited by pM9 in vivo.

(A) After 34 days xenotransplantation of NSG mice with RA-resistant HL60R cells expressing lentiviral pCLS-Flag-pM9-GFP or pCLS-GFP, GFP+ donor cells in PB were examined with fluorescence microscope (top cell-image panels) and analyzed with flow cytometry (bottom FACS panel). Scale bar, 10 μm. (B) Quantification of GFP+ donor cells in PB, BM, and spleen samples at 34 days post-transplantation. *, pM9 vs. vector in PB, P<0.013. (C) Tumors formed in mouse lymph nodes at 34 days post-transplantation. Scale bar, 1 cm. (D) Quantification of lymph node metastases in panel C. *, pM9 vs. vector, P<0.018; **, pM9 vs. vector, P<0.03. (E) HE staining of lymph node tissues infiltrated with HL60R cells expressing vector or pM9. Arrows indicate examples of the multinucleated cells in the pM9 sample. Scale bar, 20 μm. (F) WB analysis of RARα phosphorylation status in mouse lymph nodes filtrated with HL60R cells expressing pM9 or vector. (G) High-resolution fluorescence microscopy analysis of cellular localization of pM9-GFP or vector GFP in metastasized HL60R cells in mouse lymph nodes. Arrows indicate examples of the nuclei. Scale bar, 5 μm. (H) Immunohistochemistry analysis of CD45 expression in metastasized HL60R cells in mouse lymph nodes. Tissues from human tonsil were used as positive control. Scale bar, 10 μm.

Discussion

Proliferation vs. differentiation choice of hematopoietic precursors mediated by MAT1 expression and C-terminal fragmentation

Given its unique structure, intact MAT1 can assemble and activate CAK to mediate cell cycle progression while forming TFIIH kinase to regulate transcription [8-15]. Here, our in vitro and in vivo data demonstrate that MAT1 overexpression resists the intrinsically programmed MAT1 degradation to sustain the restricted expansion of CD34+ cells while preventing granulopoiesis (Figures 1-3). On the other hand, pM9 overexpression can bypass MAT1 fragmentation to inhibit proliferation of different subtypes of myeloid leukemia cells via forming ΔCAK complexes through competing with intact MAT1 to consume endogenous CDK7 and cyclin H. This resultant ΔCAK leads to decreasing CAK assembly, inhibiting TFIIH-core recruitment of CAK, and reducing the activities of both CAK and TFIIH kinase in either RA-sensitive or RA-resistant leukemic myeloblasts (Figures 4-6). The clinical implications of these findings became apparent in experiments with a mouse xenograft model bearing RA-resistant leukemic myeloblasts, where we showed that pM9 overexpression can bypass the resistance of MAT1 fragmentation to inhibit both leukemic cell proliferation and metastasis (Figure 7). This is likely through ΔCAK-dependent suppression of both CAK and TFIIH kinase activities while inducing RARα hypophosphorylation (Figures 6, 7F) to mediate RA signaling of transcription response for myelopoiesis [21-24]. Hence, our studies suggest a novel intrinsic regulation of expression vs. C-terminal fragmentation-depletion of MAT1 that mediates the choice of population expansion and granulopoiesis in hematopoietic precursors, revealing a mechanistic link between loss of programmed MAT1 fragmentation and myeloid leukemogenesis underlying CAK hyperphosphorylation of RARα (Supplemental Figure 6).

pM9-anchored ΔCAK formation mimics RA-induced MAT1 fragmentation-depletion of CAK assembly to inhibit growth and metastasis of leukemic myeloblasts

Our data (Supplemental Figures 3, 4) demonstrate that MAT1 is either intrinsically fragmented to M30 and pM9 pieces during normal granulopoiesis [23], or RA induces such fragmentations in RA-sensitive leukemic myeloblasts [21, 22, 24]. Because pM9 overexpression can reverse the leukemic state by forming of ΔCAK to mimic MAT1 fragmentation-elimination of MAT1-dependent CAK assembly (Figures 4-7), it raises a question: does the fragmented cellular pM9 form ΔCAK endogenously? Our studies eliminate this possibility based on that: a) pM9 is rarely detectable not only because it has the small size of 9-kDa, but also because it is likely eliminated together with M30 consistently through the ubiquitination-degradation [35]; and b) antibody against C-terminal MAT1 fails to precipitate endogenous ΔCAK (data not shown). Of note, MAT1 fragmentation is lost in myeloid leukemic cells, whereas overexpression of pM9 can bypass such lost MAT1 fragmentation to reverse the leukemic state by concurrently targeting CAK-coordinated cell cycle progression and gene transcription (Figures 4-7). This suggests that pM9 peptide shares a mechanistic resemblance with intrinsic MAT1 fragmentation in suppressing MAT1-dependent CAK assembly, thereby possessing therapeutic potential against different subtypes of myeloid leukemia with lost MAT1 fragmentation, in which the leukemic state depends on this mechanism of transformation.

To date, the most successful RA therapy for differentiation-induction of acute promyelocytic leukemia cells has not been extended to the remaining 90% of myeloid leukemia subtypes [40, 41]. Although most AML patients achieve complete remission after standard induction chemotherapy, the majorities subsequently relapse and die of the disease [41-43]. Moreover, even though tyrosine kinase inhibitors have a remarkable therapeutic effect for CML, a substantial portion of the patients failing to respond to such drugs has a higher risk of disease progression [44]. Therefore, it is essential to understand the pathophysiological mechanisms underlying myeloid leukemogenesis in order to discover new modes of treatment for different myeloid leukemia subtypes. Previous studies have shown that in RA-resistant myeloid leukemic cells, active proliferation is associated with consistent high level of intact MAT1 [21, 24]. This suggests that the normal physiologic transition from hematopoietic expansion to granulopoiesis coordinated by MAT1 expression and fragmentation (Figures 1-3) [23] is lost in myeloid leukemia cells. By overexpressing pM9 to induce depletion of MAT1-dependent CAK assembly in different subtypes of myeloid leukemia cells, with or without RA-resistance, we find that pM9-anchored ΔCAK formation not only inhibits both CAK assembly (Figure 4) and TFIIH-core recruitment of CAK (Figure 5), but also instigates a negative feedback of decreased TFIIH kinase activities to inhibit transcriptional expression of MAT1 mRNA (Figure 6E), likely leading to reduction of MAT1 proteins (Figure 4A). The cascade of these events leads to inhibition of proliferation, suppression of tumor formation, and prevention of metastasis of leukemic myeloblasts (Figures 4-7). Interestingly, such effects of pM9 were not observed in several different solid tumor cells that we have tested (data not shown), indicating that pM9-induced modulation of CAK and TFIIH kinase activities are specific to myeloid lineage cells. Together, these findings provide a molecular rationale of using pM9 to reverse the myeloid leukemic state through mimicking MAT1 fragmentation to concurrently inhibit cell cycle and gene transcription. Further experiments favoring this novel therapeutic approach needs to come from: a) evaluating the inhibitory effect of pM9 in a mouse model bearing xenografts of primary leukemic myeloblasts; and b) identifying the core-motif of pM9 for the design of drug-able small molecules.

Serine protease signaling involved in mediating MAT1 fragmentation to deplete MAT1-dependent CAK assembly during granulopoiesis

Rapid development in protease research has resulted in a paradigm shift from the concept of proteases as protein-degrading enzymes to proteases as key signaling molecules, by which the signal is transmitted through the cleavage of substrates resulting in their activation, inactivation, or modulation of function [45]. To date, how protease signaling is involved in control of MAT1 fragmentation in granulopoiesis and myeloid leukemogenesis remains unknown. By using both protein database and PeptideCutter program to analyze MAT1 structure, serine proteases are predicted as potential candidates to catalyze the hydrolytic degradation of MAT1 around 229-aa position. Previous studies show that the level of diisopropylfluorophosphate (DFP) binding proteins, the serine hydrolases, is significantly low in normal hematopoietic precursors but increasing with cell maturation along granulocytic series [46]. Some serine proteases, especially myeloid serine proteases, are involved in mediating of both granulopoiesis and myeloid leukemogenesis [47, 48], while RA-mediated granulopoiesis are associated with changes in expression of serine proteases [49, 50]. A recent study also reveals that Cathepsin G (CatG), a myeloid serine protease, is suppressed in t(8;21) AML, whereas restoration of CatG induces partial differentiation, growth inhibition and apoptosis in AML1-ETO-positive cells [51]. Our data indicate that likely through serine protease signaling pathway, MAT1 fragmentation suppress CAK phosphorylation of RARα, a critical molecular event regulating transactivation of RA target genes to mediate myelopoiesis [21-24]. It is known that the myeloid-specific transcription factor, CCAAT/enhancer binding protein-epsilon (C/EBPε), is a downstream target of RARα [52, 53]. Hence, a possibility whether MAT1 fragmentation induces transcription response for myelopoiesis through CAK-RARα-C/EBPε signaling remains to be tested. Moreover, further studies to determine which serine protease(s) specifically cleave MAT1 are likely to open new opportunities to define the role of serine protease signaling in the control of intrinsically programmed MAT1 expression vs. fragmentation during granulopoiesis, thus providing insight into potential drug targets for disrupting the oncogenic circuitry of the myeloid leukemia.

Conclusion

We present evidence that in normal and malignant hematopoietic precursors, the choice of hematopoietic expansion vs. granulocytic differentiation is mediated in part by MAT1 expression and C-terminal fragmentation that inversely coordinate CAK and TFIIH kinase activities. Of note, pM9-mimicked MAT1 fragmentation inhibits myeloid leukemic cell growth and metastasis. This suggests a novel therapeutic potential of using pM9 to mimic the dual-inhibitory action of MAT1 protein fragmentation, an approach that should block the cell cycle and general transcription while activating the transcriptional response for myelopoiesis.

Supplementary Material

Supp Fig S1-S6

Supplemental Figure 1. (A) The diagrams of lentiviral pCCL-c-MNDU3c-MAT1-PGK-eGFP construct. (B) GFP expression of CD34+ cells was visualized under fluorescence microscope. Scale bar, 5 μm.

Supplemental Figure 2. In parallel controls for immunofluorescence analysis of MAT1 and p21^Cip/Kip expression (Figure 3G) were performed by using PB samples collected from vector-transplanted mice. No 1^st AB, no primary antibody; No 2^nd AB, no secondary antibody. Scale bar, 20 μm.

Supplemental Figure 3. (A) The structure of intact MAT1 protein (top panel) [14] and predicted cleavage sites of M30 (middle and bottom panels). (B-D) NB4 cell lysates were subjected to immunoprecipitation using MAT1 antibody (panel B). The aliquots of the precipitates with an amount of 99/100 (left panel) and 1/100 (right panel) were separated by SDS-PAGE in parallel. The corresponding portion of membrane was cut and then stained with Coomassie blue or depicted with WB analysis using anti-MAT1 antibody, respectively (panel C). The identified MAT1 and M30 bands, which were stained with Coomassie blue while displaying the same molecule weight determined in parallel by WB with MAT1 antibody, were collected for LC-MS/MS analysis (panel D).

Supplemental Figure 4. (A) The diagram of C-terminal cleavage of M30 determined by LC-MS/MS analysis. In contrast to LC-MS/MS sequencing of the intact MAT1 that ended at aa-276, the last detected sequence in M30 was aa-177. The squares in yellow represent the sequences detected by LC-MS/MS. (B) The sequences detected by LC-MS/MS in yellow squares (panel A) are aligned with full length of MAT1 sequence. Arrow indicates the cleavage position of 229-aa that was determined previously by N-terminal sequencing [14]. The underlined M30 sequence was not detected by LC-MS/MS analysis.

Supplemental Figure 5. (A) Primers for amplification of Flag-pM9 cDNA insert. (B) Constructed lentiviral pCLS-Flag-pM9-2A-eGFP (pCLS-Flag-pM9-GFP) linear. (C) GFP expression in different leukemic myeloblasts expressing pCLS-Flag-pM9-GFP or pCLS-GFP vector only was visualized under low-resolution fluorescence microscope. Scale bar, 10 μm.

Supplemental Figure 6. (A) Loss of MAT1 protein fragmentation (MPF) induces leukemic transformation. (B) pM9-formed ΔCAK shares a mechanistic resemblance with MPF-depletion of CAK to reverse the myeloid leukemic state, through inhibiting CAK-coordinated cell cycle and gene transcription while activating RARα-mediated transcription of RA target genes.

Supp Table S1. Supplemental Table 1.

Primers used in qRT-PCR amplification for determining gene expression of GFP+ donor cells isolated from humanized mice model. Based on the manufacturer's instructions (Invitrogen, Grand Island, NY), qRT-PCR was performed with an initial denaturization at 95°C for 15 min, followed by consecutive 40 thermal cycles (94°C, 15 s; 55°C, 30 s; and 72°C, 30 s) and a final dissociation-curve cycle (95°C, 15 s; 60°C, 15 s; and 95°C, 15 s).

Supp Table S2. Supplemental Table 2.

Primers used for qRT-PCR analysis to depict gene expression in HL60, HL60R and Jurl-MK1 cells overexpressing pM9. The amplification condition was setup according to the manufacturer's instructions (Invitrogen, Grand Island, NY), qRT-PCR was performed with an initial denaturization at 95°C for 15 min, followed by consecutive 40 thermal cycles (94°C, 15 s; 59°C, 30 s; and 72°C, 30 s) and a final dissociation-curve cycle (95°C, 15 s; 60°C, 15 s; and 95°C, 15 s).