MAT1-Modulated Cyclin-Dependent Kinase-Activating Kinase Activity Cross-Regulates Neuroblastoma Cell G₁ Arrest and Neurite Outgrowth

Considerable studies in many cell types have shown that the decision of cells to differentiate is often made in cell cycle G₁ phase and that the transition from proliferation to differentiation is an inversely correlated process at cell cycle exit (1). The discovery that retinoic acid (RA) induces myeloid cell differentiation via RA receptor α (RARα) provides a revolutionized model for studying how differentiation is induced from cancer cell cycle arrest (2, 3). Also, the identification of nuclear superfamilies of RAR isotypes (α, β, and γ) and retinoid X receptor (RXR) isotypes (α, β, and γ; Ref. 4) allows for detailed studies into the mechanisms of RA-induced proliferation/differentiation (P/D) transition. However, how RA induces differentiation from cancer cell cycle arrest is not well understood.

Coordination of P/D transition may result from dual function of central cell cycle regulators to cross-regulate differentiation (1, 2). Numerous studies have been directed toward identification of cell cycle regulators that coordinate proliferation and differentiation. Human cyclin-dependent kinase (CDK)-activating kinase (CAK) is a trimeric enzyme complex consisting of CDK7, cyclin H, and ménage à trois 1 (MAT1; Refs. 5, 6, 7). CAK controls cell cycle progression by catalyzing T-loop phosphorylation of CDKs (8, 9) and regulates transcription by serving as a kinase of general transcription factor IIH to phosphorylate the largest subunit of RNA polymerase II (9, 10). MAT1 assembles CAK (11, 12) and determines the substrate specificity of CAK (13, 14, 15, 16, 17). MAT1-modulated CAK activity regulates cell cycle G₁ exit (16), by which CAK interacts with and phosphorylates retinoblastoma tumor suppressor protein (pRb), a proliferation repressor (18, 19) and a differentiation enhancer (1, 2), in a MAT1-dependent manner (16). Abrogation of MAT1 by retrovirus-MAT1 antisense decreases CAK phosphorylation of pRb and induces G₁ arrest in osteosarcoma cells (16). RARα, a key player in myeloid differentiation (3), is a substrate for CAK, and MAT1 enhances cyclin H-CDK7 phosphorylation of RARα (15). In a recently identified P/D transition characterized as a coordination of G₁ arrest with differentiation activation in human HL60 cells, RA induces a reduction of both MAT1 expression and MAT1 fragmentation to a M_r 30,000 fragment (M30) through a protease pathway (20). Importantly, this reduction of MAT1/M30 is accompanied by decreased CAK activity that switches CAK hyperphosphorylation of RARα (higher CAK activity resulting in active phosphorylation of RARα in proliferating cells) to CAK hypophosphorylation of RARα (decreased CAK activity resulting in significantly declined phosphorylation of RARα in differentiating cells; Ref. 20). Although the above studies about P/D transition have shown overall associations among MAT1 reduction, decreased CAK phosphorylation of differentiation regulators, and terminal differentiation, the basic pathway and precise mechanism for the switch from proliferation to differentiation remain unclear. Therefore, because MAT1 determines the substrate specificity of CAK and CAK phosphorylates key differentiation regulators, we directly addressed the question as to whether MAT1 is a key element to coordinate proliferation inhibition with differentiation activation in neuroblastoma cells responding to 9-cis-RA (9cRA) stimuli (21, 22, 23).

Human CHP126 cells (24) were cultured in RPMI 1640 supplemented with 10% or 2% fetal bovine serum (FBS) depending on the experiment. The E1-transformed human embryonic kidney AD-293 cell line was purchased from Stratagene (La Jolla, CA) and cultured in DMEM with 10% FBS. 9c-RA was from Sigma (St. Louis, MO). All antihuman antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Cell duplication was determined by cell counting as described before (25), and cell cycle profile analysis was performed as described previously (16).

Immunoprecipitation and Western blotting were performed as described previously (16). In vivo CAK phosphorylation of pRb and RXRα was determined by labeling CHP126 cells with [³²P]P_i as described previously (20).

Replication-defective AdEasy pAdtrack-CMV vector fused with green fluorescent protein (GFP; provided by Dr. Vogelstein, Johns Hopkins Oncology Center, Baltimore, MD; Ref. 26) was used as a shuttle vector. MAT1 antisense fragment or MAT1 coding region (25) was cloned into pAdtrack-CMV vector by using established methods (25). The resultant adenovirus-GFP-MAT1 antisense and adenovirus-GFP-MAT1 plasmids were linearized by restriction endonuclease Pme I, and then cotransformed with adenoviral backbone plasmid pAdEasy-1 (provided by Dr. Vogelstein) into Escherichia coli BJ5183 cells (Stratagene). The recombinant selection, restriction digestion/sequencing confirmation, and adenovirus production were performed as described previously (26). Recombinant adenovirus-GFP-vector was prepared in parallel for control. Adenoviruses were purified/concentrated by using BD Adeno-X virus filter (BD Biosciences Clontech, Palo Alto, CA) and stored in aliquots at −70°C. Viral titer was determined by both standard absorbance read at A₂₆₀ and GFP-monitored tissue culture infectious dose 50. Final viral titer is about 5 × 10¹² particles/ml. For viral transduction, CHP126 cells were seeded at 40% confluent density and cultured for at least 24 h. The cells were then transduced at a multiplicity of infection of 500 in the presence of 2% FBS for 2 h.

Nonintegrated adenovirus-GFP will be lost after only a few divisions of active cell proliferation but retained in proliferation-inhibited living cells. Thus, assessing the change of GFP fluorescence with time in adenovirus-transduced cells provides the means to monitor the transition from CHP126 cell proliferation inhibition to neurite outgrowth in dishes under a phase contrast fluorescence microscope. Morphology differentiation of neurite outgrowth was examined under a Nikon phase contrast microscope, and images were photographed under ×10 magnification. Digital overlay of GFP fluorescence image with the respective phase contrast image was accomplished in Adobe Photoshop. The relationship of GFP fluorescence per cell abundance with change in cell confluence and neurite outgrowth was quantified by analyses of phase contrast and fluorescence images using an Olympus DP-11 color digital camera. Fluorescence intensity per cell abundance was determined through color threshold and integrated morphometry analysis by using MetaMorph (version 6.1r2) software (Universal Imaging Corp., Downingtown, PA). The color thresholds for GFP fluorescence were as follows: red, 0.255; green, 24.255; and blue, 0.255. Each fluorescent cell in the color threshold range was measured by total pixel area, total gray level (integrated fluorescence), and average gray level. Small fluorescent debris and instrument noise were excluded by a size classifier of >299 pixels. Data for all cells in each image were logged to Microsoft Excel 2000.

We first investigated at which cell cycle stage 9cRA induces neuron P/D transition and whether MAT1 reduction is associated with neuron differentiation activation in neuroblastoma cells. CHP126 cells that can proliferate either in 10% or 2% FBS (Fig. 1,A) were exposed to different periods of 9cRA. Cell proliferation analyses showed that, like other tumor cell lines in which proliferation inhibition is induced at a low level of serum (27, 28), CHP126 cell proliferation is significantly halted after 3 days of 9cRA treatment in 2% FBS (Fig. 1,A). This proliferation inhibition is correlated to the occurrence of cell cycle G₁ arrest after 3 days of 9cRA stimuli (Fig. 1,B). Differentiation then proceeds as shown by neurite outgrowth (Fig. 1 C). Hence, the results demonstrate that 9cRA-induced CHP126 cell differentiation is induced from neuron G₁ arrest.

To determine whether MAT1 reduction is associated with 9cRA-induced neuron P/D transition, the expression of CAK components was examined in CHP126 cells after different periods of 9cRA exposure. The results from Western analyses showed that, as demonstrated in ATRA-induced HL60 cell P/D transition (20), MAT1 expression and MAT1 fragmentation to a M_r 30,000 fragment (M30) are inhibited after 48 h of 9cRA exposure, whereas the levels of total cyclin H and CDK7 remain steady (Fig. 2,D). By Western analyses of RXRα, a key regulator of neuron differentiation in responding to 9cRA (21, 22, 23, 29, 30), we found that RXRα hypophosphorylation in differentiating cells is induced from RXRα hyperphosphorylation in proliferating cells (Fig. 1,E). pRb, a substrate for CAK (16) and a differentiation enhancer (1, 2), also shows a hypophosphorylated status in differentiating cells and a hyperphosphorylated status in proliferating cells (Fig. 1,E). The total levels of RARα show a reduction immediately after cells are exposed to 9cRA and then remain steady (Fig. 1 F), as similarly demonstrated before in HL60 cells (20). However, the levels of both RARβ and RARγ show no significant change under 9cRA treatment (data not shown). Hence, these results show that MAT1 reduction is associated with decreased phosphorylation of pRb, RARα, and RXRα in 9cRA-induced neuron P/D transition. Because MAT1 determines the substrate specificity of CAK and CAK involves differentiation regulation, the data suggest that decreased phosphorylation of pRb, RARα, and RXRα in 9cRA-induced P/D transition might be related to decreased CAK activity due to MAT1 reduction.

To test that decreased phosphorylation of pRb, RARα, and RXRα with 9cRA treatment is due to decreased CAK activity, we first tested whether CAK interacts with and phosphorylates any of those differentiation regulators in 9cRA-induced neuron P/D transition. After CHP126 cells were exposed to different periods of 9cRA, anti-CDK7 antibodies were used to precipitate these putative CAK-bound molecules. By Western analyses of the precipitates, we found that pRb and RXRα are coprecipitated by anti-CDK7 antibodies (Fig. 2, A and B), whereas there was a negative detection of RARα (data not shown). To define that the coprecipitated pRb and RXRα are bound by CAK, these same blots resulting from CDK7 precipitation were immunoblotted with MAT1 antibodies. We found that both pRb and RXRα indeed bind to CAK as evidenced by their association with MAT1 and that 9cRA also induces MAT1 reduction in the precipitates (Fig. 2, A and B). Thus, the findings that CAK-bound pRb and RXRα are associated with MAT1 reduction may indicate a neuronal characteristic of CAK regulatory network in 9cRA-induced CHP126 cell P/D transition.

CDK7 is auto-phosphorylated in the kinase reaction, and CDK7 auto-phosphorylation is MAT1 dependent (14, 16, 20). Therefore, to determine whether MAT1 reduction is associated with decreased phosphorylation of CAK-bound pRb and RXRα, we next in vivo-labeled CAK signaling with [³²P]P_i after CHP126 cells were exposed to different periods of 9cRA. Then, we used anti-CDK7 antibodies to precipitate CAK-bound pRb and RXRα from these labeled cells. We found that by autoradiography visualization of CAK-RXRα and CAK-pRb signaling, anti-CDK7 antibodies coprecipitate ³²P-labeled RXRα, ³²P-labeled pRb, and ³²P-labeled CDK7 simultaneously at their corresponding molecular weight positions (Fig. 2, C and D). CAK-bound hyperphosphorylated ³²P-labeled pRb and ³²P-labeled RXRα are present in proliferating cells. However, the levels of hyperphosphorylated ³²P-labeled pRb and ³²P-labeled RXRα are diminished by more than 90% after 120 h of 9cRA stimuli (Fig. 2, C and D; densitometer results not shown). These declined levels of ³²P-labeled RXRα and ³²P-labeled pRb correspond well with decreased CAK activity as represented by declined auto-phosphorylated ³²P-labeled CDK7 (Fig. 2, C and D). To define that the specific ³²P bands at M_r 110,000 and 55,000 are indeed CAK-bound pRb and RXRα, these same autoradiography blots were immunoblotted with pRb, RXRα, CDK7, and MAT1 antibodies, respectively. We found that pRb, RXRα, and CDK7 are specifically recognized at their corresponding molecular weight positions of their phosphorylations, whereas the binding of pRb and RXRα to CAK is confirmed by positive detection of CDK7 and MAT1 in the precipitates (Fig. 2, C and D). CDK7 antibodies distinguished both forms of the auto-hyperphosphorylated CDK7 (P-K7) and the auto-hypophosphorylated CDK7 (K7). After 9cRA stimuli, P-K7 diminished gradually whereas K7 increased correspondingly (Fig. 2, C and D). The dynamic levels of P-K7 and K7 corresponded well with the dynamic phosphorylation status of ³²P-labeled RXRα, ³²P-labeled pRb, and ³²P-labeled CDK7 (Fig. 2, C and D). Importantly, declined ³²P-labeled RXRα, ³²P-labeled pRb, and ³²P-labeled CDK7 are correlated with MAT1 reduction in the precipitates (Fig. 2, C and D). Together, these results demonstrate that 9cRA induces a concurrent switch from CAK hyperphosphorylation of pRb and RXRα in proliferating cells to CAK hypophosphorylation of pRb and RXRα in differentiating cells. Because CAK hypophosphorylation of pRb induces G₁ arrest (16) and RXRα involves neuron differentiation (21, 22, 23), concurrent CAK hypophosphorylation of pRb and RXRα may coordinate G₁ arrest with differentiation in 9cRA-induced CHP126 cell P/D transition. Also, because the switch from CAK hyperphosphorylation of pRb and RXRα to CAK hypophosphorylation of pRb and RXRα is accompanied by MAT1 reduction, this suggests that MAT1 reduction may inhibit CAK phosphorylation of pRb and RXRα to coordinate neuron G₁ arrest and differentiation activation.

To test the hypothesis that MAT1 reduction decreases CAK phosphorylation of pRb/RXRα to induce proliferation inhibition and differentiation activation, we diminished MAT1 levels in CHP126 cells by transduction of adenovirus-GFP-MAT1 antisense to mimic 9cRA action in P/D transition. CHP126 cells were transduced with adenovirus-GFP-MAT1 antisense or adenovirus-GFP-vector in the presence or absence of 9cRA. By monitoring cell proliferation, we found that, either in the absence or presence of 9cRA, adenovirus-GFP-MAT1 antisense mimics 9cRA-induced proliferation inhibition (Fig. 3,A). Adenovirus does not integrate into the cellular genome and therefore leads to loss of transiently transduced GFP fluorescence after only a few divisions of active cell replication (Fig. 3,B). This provides an opportunity for us to determine whether overexpression of adenovirus-GFP-MAT1 antisense mimics neuron P/D transition by morphometrically assessing and quantifying the change with time in GFP fluorescence, cell confluence, and neurite outgrowth in dishes. We found that, under a phase contrast fluorescence microscope, the cells under adenovirus-GFP-vector transduction are continuously proliferating, which is evidenced by loss of GFP fluorescence and increasing cell confluence (Fig. 3,C, panels 4–6). In contrast, the cells transduced by adenovirus-GFP-MAT1 antisense showed proliferation inhibition because these cells retained both a high level of GFP fluorescence and a low level of cell confluence (Fig. 3,C, panels 1–3). Interestingly, adenovirus-GFP-MAT1 antisense also significantly induced neurite outgrowth in the absence of 9cRA after 11 days post-transduction (Fig. 3,C, panel 3). This indicates that MAT1 reduction not only inhibits proliferation but also enhances differentiation. Furthermore, to test whether MAT1 reduction cooperates with 9cRA action, CHP126 cells were exposed to either 9cRA alone or addition of 9cRA after 2 h post-transduction of adenovirus. Cells that were transduced by adenovirus-GFP-MAT1 antisense in the presence of 9cRA underwent P/D transition processes, showing that a high population of GFP-positive cells is associated with a low level of cell confluence and a significant neurite outgrowth (Fig. 3,C, panel 10–12). This P/D transition is similar to the cells exposed to 9cRA alone (Fig. 3,C, panels 7–9) and the vector control in the presence of 9cRA (Fig. 3,C, panels 13–15). To determine the relationship of GFP fluorescence per cell abundance with change in cell replication and neurite outgrowth under adenovirus transduction, GFP fluorescence intensity was quantified by digital image analysis. Both phase contrast and fluorescence images were randomly captured in parallel from at least 10 different areas of the dishes using a color digital camera. The fluorescence intensity versus average gray value was analyzed by using MetaMorph software. The results showed that high levels of fluorescence intensity are retained in the cells transduced by adenovirus-GFP-MAT1 antisense alone, or in the cells exposed to 9cRA after 2 h post-transduction of either adenovirus-GFP-MAT1 antisense or adenovirus-GFP-vector (Fig. 3,D, panels 1, 3, and 4). These high levels of fluorescence intensity are unanimously associated with a high population of GFP-positive cells, low levels of cell confluence, and significant neurite outgrowth (Fig. 3,C, panels 2, 3, 11, 12, 14, and 15). However, in the control cells that were transduced by adenovirus-GFP-vector without 9cRA, we found that GFP fluorescence completely vanishes after 11 days post-transduction (Fig. 3,D, panel 2, day 11), which corresponds well with a rapidly increasing GFP-negative population without neurite outgrowth (Fig. 3 C, panels 5 and 6). Hence, these studies show that adenovirus-GFP-MAT1 antisense mimics and cooperates with 9cRA-induced proliferation inhibition and neurite outgrowth.

To further define that MAT1 reduction mimics 9cRA action to induce P/D transition in CHP126 cells, we tested whether overexpression of adenovirus-GFP-MAT1 resists 9cRA-induced effect of MAT1 reduction in both proliferation inhibition and neurite outgrowth. First, we examined whether MAT1 overexpression resists 9cRA action in proliferation inhibition. CHP126 cells were exposed to 9cRA after 2 h post-transduction of adenovirus-GFP-MAT1, and then cell replication was monitored for up to 4 days as described previously (16). We found that MAT1 overexpression resists 9cRA-induced proliferation inhibition (Fig. 4,A). Because transiently transduced adenovirus will lose GFP fluorescence after only a few divisions of active cell replication as illustrated (Fig. 3,B), we then assessed whether adenovirus-GFP-MAT1 overexpression resists 9cRA-induced neuron P/D transition by morphometric analysis and quantification of the change with time in GFP fluorescence, cell confluence, and neurite outgrowth in dishes. CHP126 cells were exposed to 9cRA for up to 11 days after 2 h post-transduction of adenovirus-GFP-MAT1. We found that 9cRA action was reversed in these cells. This was evidenced by: a) rapid loss of GFP fluorescence corresponding well with progressively increasing cell confluence (Fig. 4,B, panels 1–3); and b) a complete inhibition of neurite outgrowth in these cells losing GFP fluorescence (Fig. 4,B, panel 3). However, control cells that were either exposed to 9cRA alone or transduced by adenovirus-GFP-vector in the presence of 9cRA showed high levels of GFP fluorescence, low levels of cell confluence, and significant neurite outgrowth (Fig. 4,B, panels 5, 6, 8, and 9). To assess the relationship of GFP fluorescence per cell abundance with change in cell replication and neurite outgrowth under adenovirus transduction, GFP fluorescence intensity was quantified by digital image analysis. Similarly as explained above, both phase contrast and fluorescence images were randomly captured in parallel from at least 10 different areas of the dishes by using a color digital camera. The fluorescence intensity versus average gray value was analyzed by using MetaMorph software. We found that the cells transduced by adenovirus-GFP-vector in the presence of 9cRA retained high levels of GFP fluorescence after 11 days post-transduction (Fig. 4,C, panel 2), which was accompanied by a high population of GFP-positive cells, low level of cell confluence, and significant neurite outgrowth (Fig. 4,B, panels 5 and 6). In contrast, cells transduced by adenovirus-GFP-MAT1 resist 9cRA action as shown by complete loss of GFP fluorescence after 11 days post-transduction (Fig. 4,C, panel 1, day 11), which corresponds well with rapidly increasing GFP-negative cells and high levels of cell confluence without neurite outgrowth (Fig. 4,B, panels 2 and 3). Hence, these studies show that MAT1 overexpression resists 9cRA-induced neuronal cell P/D transition. Together these results with the above studies that show the effect of adenovirus-GFP-MAT1 antisense (Fig. 3), we demonstrate that the dynamic state of MAT1 plays a key role in mediating cells either to proliferate or to differentiate, i.e., MAT1 reduction mimics 9cRA actions of proliferation inhibition and neurite outgrowth, whereas MAT1 overexpression resists these 9cRA actions.

Because MAT1 is an assembly factor and targeting subunit of CAK, the above results suggest that the dynamic state of MAT1 in control of cells either to proliferate or to differentiate might act through altering CAK activity in phosphorylation of its substrates pRb and RXRα. To test this hypothesis, we assessed the causal effect of either MAT1 reduction or MAT1 overexpression in alteration of the phosphorylation statuses of pRb and RXRα. We found that in CHP126 cells that were post-transduced by adenovirus for 11 days, adenovirus-GFP-MAT1 antisense mimics 9cRA-induced MAT1 reduction at total protein levels (Fig. 5,A, Lanes 3 versus 4) and in the precipitates (Fig. 5,B, Lanes 4 versus 6). However, adenovirus-GFP-MAT1 resists 9cRA action of MAT1 reduction (Fig. 5, A, Lanes 1 versus 4; and B, Lanes 3 versus 6). Furthermore, by Western analyses of pRb and RXRα phosphorylation status at the total protein levels with anti-RXRα and anti-pRb antibodies after 11 days post-transduction, we found that adenovirus-GFP-MAT1 antisense mimics 9cRA-induced hypophosphorylation of RXRα and pRb (Fig. 5, C, Lanes 5 versus 7; and D, Lanes 5 versus 7). However, overexpressing adenovirus-GFP-MAT1 allows cells to resist 9cRA action as shown by hyperphosphorylation of pRb and RXRα (Fig. 5 C, Lanes 1 versus 7; and D, Lanes 1 versus 7). These results demonstrate that MAT1 reduction is crucial for inducing hypophosphorylation of pRb and RXRα in CHP126 cells and suggest that 9cRA-induced MAT1 reduction may inhibit CAK phosphorylation of pRb and RXRα to coordinate CHP126 cell P/D transition.

In conclusion, we have shown that the switch from proliferation to differentiation in 9cRA-induced P/D transition requires MAT1 reduction that decreases CAK phosphorylation of pRb and RXRα. Although intracellular regulation of MAT1 abundance via protease pathway (20) awaits characterization, these studies demonstrate the physiological relevance of CAK in regulating the inversely correlated processes of neuroblastoma cell proliferation and differentiation. It is very likely that changing the balance between MAT1 abundance via protease pathway and MAT1-mediated CAK assembly/CAK substrate specificity is the key for coordinating G₁ arrest with differentiation activation in neuroblastoma cells.

We thank Dr. Vogelstein (Johns Hopkins Oncology Center, Baltimore, MD) for providing replication-defective recombinant adenoviral vectors. We thank Dr. McNamara for assistance with digital image analyses.