Proteasome inhibitors increase tubulin polymerization and stabilization in tissue culture cells

Abstract

Bortezomib (Velcade^®), a proteasome inhibitor, is approved by the FDA for the treatment of multiple myeloma (MM). While effective, its use has been hampered by peripheral neurotoxicity of unexplained etiology. Since proteasome inhibitors alter protein degradation, we speculated that proteins regulating microtubule (MT) stability may be affected after treatment and examined MT polymerization in cells by comparing the distribution of tubulin between polymerized (P) and soluble (S) fractions. We observed increased MT polymerization following treatment of SY5Y and KCNR [neuroblastoma], HCN2 and 8226 [MM] cells, using five proteasome inhibitors; the baseline proportion of total α-tubulin in ‘P’ fractions ranged from ~41–68%, and increased to ~55–99% after treatment. Increased acetylated α-tubulin, a post-translational marker of stabilized MTs, was observed in the neural cell lines HCN1A and HCN2 and this was sustained up to 144 hours after the proteasome inhibitor was removed. Cell cycle analysis of three cell lines after treatment, showed ~50–75% increases in the G₂M phase. Immunofluorescent localization studies of proteasome inhibitor treated cells did not reveal microtubule bundles in contrast to paclitaxel treated, suggesting MT stabilization via a mechanism other than direct drug binding. We examined the levels of microtubule associated proteins and observed a 1.4–3.7 fold increase in the microtubule associated protein MAP2, in HCN2 cells following treatment with proteasome inhibitors. These data provide a plausible explanation for the neurotoxicity observed clinically and raise the possibility that microtubule stabilization contributes to cytotoxicity.

Introduction

Drugs that associate with microtubules and change their dynamics and function have been extensively studied in tissue culture and utilized widely with some success in the treatment of numerous malignancies. The microtubule depolymerizing or polymerizing effects of these drugs disrupt processes necessary for cell survival, including intracellular trafficking, cell motility, maintenance of cell architecture, chromosome segregation and mitosis. A complex assembly of proteins, the proteasome, is responsible for the majority of intracellular proteolysis.^1,2 Via the regulation of proteins involved in transcription, cell cycle, adhesion and apoptosis, the proteasome impacts many cellular processes, including cell proliferation, signaling and differentiation.^3,4 Its enzyme activity effects non-ubiquitinated protein eradication as well as the destruction of misfolded and/or unfolded proteins, which are routed to the proteasome following their ubiquitination.⁵ Its importance is underscored by the fact that malfunctioning ubiquitin/proteasome systems have been implicated in a variety of diseases.^6-8 In addition, when proteasome function is overwhelmed or inhibited, inclusion bodies of accumulated proteins termed aggresomes,⁹ form in the periphery of cells and track on MTs centrally towards the microtubule- organizing center (MTOC). Proteasome inhibitors, such as MG-132, lactacystin¹⁰ and bortezomib [Velcade^®], a drug approved for the treatment of MM,¹¹ inhibit protein degradation by the proteasome. Somewhat reminiscent of chemotherapeutic agents that target the microtubule cytoskeleton, peripheral and autonomic neurotoxicity occurs with bortezomib [Velcade^®].^12-14 Although the microtubule network is crucial in all cells for the range of vital functions previously enumerated, its function in slowly dividing neural cells also encompasses neural signal transmission. Peripheral neuropathy, a side effect which impacts quality of life and is a contraindication for continued treatment, has a poorly defined etiology. It is conceivable that an increase in microtubule stability following drug treatment might impact neurotoxicity. With better understanding of the mechanisms underlying the neuropathy, this side effect may be circumvented, thus permitting more effective treatments. In the present study we observe that a variety of malignant and non-malignant tissue culture cells treated with proteasome inhibitors exhibit an increase in stabilized microtubules. This increase in MT polymerization appeared to be a drug class effect since it occurred for the five proteasome inhibitors used.

Results

Analysis of tubulin polymerization and α-tubulin acetylation. While examining the putative role of VHL in tubulin stabilization, we noted that exponentially growing renal carcinoma cells treated with the proteasome inhibitor MG-132, showed a distinct increase in the amount of tubulin found in the polymerized fraction as compared with untreated cells (unpublished observations). We postulated that this increase in stabilization could occur in cells of neural origin and might contribute to the neurotoxicity that has been reported in patients receiving bortezomib (Velcade^®). Therefore, we examined MT stabilization and tubulin polymerization following drug treatment in slowly dividing neural cells and more rapidly dividing neuroblastoma cells.

Indeed, when both SY5Y and KCNR neuroblastoma cells were treated with any of five different proteasome inhibitors [bortezomib (Velcade^®), MG-132, lactacystin, epoxomycin, or Z-Ile-Glu-(OtBu)-Alu-LeuH(aldehyde)(PSI)], for 18–22 hours, we observed an increase in the amount of tubulin in the polymerized fraction as compared with untreated cells (Fig. 1A). This increase was also apparent for the neural cell line HCN2 (Fig. 1B). The baseline proportion of α-tubulin in the polymerized fraction ranged from ~41–68%, while the polymerized proportion observed after treatment was ~55–99% (Table 1). Similarly, when we examined the effects of the proteasome inhibitors on microtubule stabilization in 8226 MM cells we observed an increase in the amount of tubulin in the polymerized fraction from the baseline level of 51% in untreated cells to levels in excess of 81% in the cells treated with the proteasome inhibitors (Suppl. Fig. 1 and Suppl. Table 1). The latter is comparable to the effect of paclitaxel, emphasizing the marked stabilization of microtubules, and suggesting that this might be a contributing factor to the mechanism of proteasome inhibition and toxicity in MM cells. As noted in ‘Methods’, the concentrations used were sub-lethal for most cells. The median Sub-G₁ fraction was 11% ± 6 for SY5Y cells, 3% ± 1 for K NR cells, and 13% ± 5 for the 8226 MM cells (Table 3). Also, as shown in Figure 5, by light microscopy the cellular architecture and morphology were likewise normal in appearance 24 hours after treatment.

Figure 1.

The fraction of polymerized tubulin is increased in neuroblastoma and cultured neural cells after treatment with proteasome inhibitors. Lysates from neuroblastoma cells, SY5Y and KCNR, (Fig. 1A) and neural derived HCN2 cells (Fig. 1B), were obtained from cells either treated or not for 18–22 hours, with proteasome inhibitors at the indicated concentrations. Lysates were separated into polymerized (P) or soluble (S) fractions by centrifugation at ~15,000 g at 22°C for 10 minutes. Aliquots of equal volume were separated by SDS-PAGE, the blots probed with anti α-tubulin and the percent of polymerized tubulin calculated for each ‘P’ and ‘S’ pair. Vertical lines indicate that tracks from different blots from the same experiment, exposed equivalently, are shown alongside each other, and the results represent that of a typical experiment.

Table 1.

Percent of polymerized tubulin in cell lines treated with proteasome inhibitors

Cell line	Drug	Concentration	% Polymerized tubulin average ± std dev	p-value	# of expts
SY5Y	None		41 ± 6.7		7
	Bortezomib	100 nM	65 ± 14.6	0.005	4
	MG 132	50 μM	82 ± 17.1	0.017	3
	Lactacystin	5 μM	55 ± 2.1	0.009	3
	Epoxomycin	100 nM	56 ± 2.5	0.006	3
	PSI	10 μM	73 ± 15.6	0.002	2
KCNR	None		51 ± 3.4		7
	Bortezomib	100 nM	67 ± 6.8	<0.001	5
	MG132	50 μM	63 ± 6.5	0.002	5
	Lactacystin	5 μM	61 ± 10.1	0.073	5
	Epoxomycin	100 nM	66 ± 4.3	0.006	4
	PSI	10 μM	93 ± 2.1	<0.001	2
HCN2	None		62 ± 9.9		7
	Bortezomib	300 nM	94 ± 1.4	0.001*	2
	Bortezomib	100 nM	77 ± 11.6		4
	MG132	50 μM	88 ± 5		3
	Lactacystin	30 μM	94 ± 6.4		2
	Lactacystin	10 μM	90 ± 0.7		2
	Lactacystin	5 μM	97		1
	Epoxomycin	300 nM	92 ± 6.4		2
	Epoxomycin	100 nM	78 ± 10.6		2
	PSI	30 μM	99.5 ± 0.1		2
	PSI	10 μM	80 ± 7.8		2

^*Because HCN2 cells grow very slowly and fewer experiments were performed, precluding p-value determinations for most drugs, a p-value for all proteasome inhibitors was determined and is shown.

Table 3.

Cell cycle distribution of cells treated with proteasome inhibitors for 24 hrs

Cell type & treatment	% Cell cycle phase*
	G1	S	G2/M	Sub-G1
SY5Y
0	71	23	5
100 nM Bortezomib	37	33	14	16
50 μM MG 132	52	30	10	7
10 μM Lactacystin	40	30	17	14
100 nM Epoxomycin	50	31	16	3
50 μM PSI	40	34	10	16
KCNR
0	67	26	6
100 nM Bortezomib	37	43	17	4
50 μM MG 132	44	41	14	1
10 μM Lactacystin	30	48	19	3
100 nM Epoxomycin	36	35	28	2
50 μM PSI	33	45	18	4
8226
0	39	38	22	3
0	36	38	25	4
Saline Vehicle	36	38	25	3
DMSO Vehicle	39	39	20	3
300 nM Bortezomib	13	37	32	20
100 nM Bortezomib	14	37	37	11
50 μM MG 132	20	43	32	6
20 μM Lactacystin	25	37	27	10
100 nM Epoxomycin	12	33	39	17
10 μM PSI	14	36	33	15
10 nM Paclitaxel	7	38	23	32

^*Percentage of cells in each phase of the cell cycle following 24 hours of drug treatment.

^**Total percent may odd up to other than 100% in some cases secondary to rounding off of numbers and the inherent nature of the program analysis.

Figure 5.

Confocal immunofluroescenl localization of HDAC6, tubulin and vimentin in SY5Y and HCN2 cells after treatment with proteasome inhibitors. For SY5Y cells [upper] or HCN2 cells [lower], which were either untreated (No Drug), treated with poclitaxel (PTX), or proteasome inhibitors (MG-132, bortezomib) at the concentrations indicated in the left margin, separate panels display the immunlocalization of HDAC6, a marker for aggresomes, indicated by the red color (rhodamine secondary antibody) and α-tubulin [center] indicated by the green color (fluorescein secondary antibody). DAPI stain (blue) localizes to cell nuclei. The tricolor localization in each cell line of ‘Tublin/HDAC6/DAPI’ is shown by the superimposition of three confocal images and shown in the third column panels. The superimposition from other samples of three images for the tricolor ‘Vimentin/HDAC6/DAPI’ localization, are shown in the fourth column of panels.

Furthermore, we examined the extent of acetylation of α-tubulin, a post-translational modification indicative of stabilized microtubules, often observed following exposure of cells to MT stabilizing agents such as paclitaxel or epothilone.²⁰ Following treatment with proteasome inhibitors we observed a 1.5–3.5 fold increase in the amount of acetylated α-tubulin in whole cell lysates from HCN1A and HCN2 cells (Fig. 2A and B, displaying data representative of two or more experiments; and Table 2 showing averaged data for HCN2 alongside standard deviations). This range for fold increase is similar to that observed following the addition of paclitaxel (data not shown) and indicative of stabilized microtubules. We would note that because these two neural cell lines have very stable microtubules even in the absence of proteasome inhibitors, the basal level of acetylation is high. Thus, increased tubulin polymerization occurs in a variety of cell types following treatment with five different proteasome inhibitors, consistent with a drug class effect. Statistically significant increases in the percent polymerized tubulin in treated cells compared to the untreated cells were found in all cell lines as summarized in Table 1.

Figure 2.

Acetylated α-tubulin and MAP2 are increased in HCN1A and HCN2 neural cells after treatment with proteasome inhibitors. Cells were treated or not for 24 hours with proteasome inhibitors and total cell lysates were obtained. For each sample, equal amounts of total protein were separated by SDS-PAGE. For HCN1A cells, the western blot was probed with anti acetylated tubulin and GAPDH (Fig. 2A), and for the HCN2 cells, the blots were also probed with anti MAP2 (Fig. 2B). The intensity of each band was quantitated by densitometry, normalized to the value for GAPDH, and the fold increase compared to the untreated sample.

Table 2.

Average fold Increase in acetylated α-tubulin in HCN2 cells treated with proteasome inhibitors

Cell line	Drug	Concentration	Acetylated α-tubulin* ± std dev	# of experiments
HCN2	None
	Bortezomib	100 nM	2.2 ± 0.7	5
	MG132	50 μM	2.3 ± 0.3	2
	Lactacystin	30 μM	2.6	1
	Lactacystin	10 μM	2.1 ± 0.3	2
	Epoxomycin	300 nM	2.8	1
	Epoxomycin	100 nM	2.7 ± 0.7	2
	PSI	30 μM	4
	PSI	10 μM	2.9 ± 0.8	2
Cell line	Drug	Hours washout	Acetylated α-tubulin* ± std dev	# of experiments
HCN2	24 hour	0	2.2 ± 0.7	5
	incubation	5	3.2 ± 1.9	4
	in 100 nM	24	4.3 ± 2.1	4
	Bortezomib	48	3.8 ± 1.7	4
	followed by	72	4.5 ± 2.1	2
	washout	96	6.0
		120	2.9
		144	4.3 ± 2.4	2

^*Normalized to GAPDH.

^*Normalized to GAPDH or Actin.

Further evidence of the effect of proteasome inhibition on microtubule stability was obtained when the time course study was performed, demonstrating the persistence of this effect for up to 144 hours after the removal of bortezomib (Fig. 3). The effect on both microtubule polymerization and acetylation of α-tubulin is shown. The sustained effects observed long after drug removal give insight into how this could confer neurotoxicity. We would note here that marked and sustained effects were observed at both 100 and 25 nM bortezomib, the latter a concentration that in our hands had very minimal effects on these cells. We would also note the tolerability of these concentrations since the cells were still alive six days after administration and undigested protein could be readily retrieved. All of these observations argue strongly against the effect on microtubules being an aberration of excessive drug concentrations.

Figure 3.

The increases in the fraction of polymerized tubulin and acetylated α-tubulin in HCN2 cells treated with Bortezomib are stable during drug washout. HCN2 cells were treated or not for 24 hours with either 100 or 25 nM of bortezomib prior to a rinse in PBS and incubation in drug free medium for 0, 5 , 24, 48, 120 or 144 hours. In the case of the cells incubated in 100 nM bortezomib, lysates were either separated into polymerized (P) or soluble (S) fractions by centrifugation at ~15,000 g at 22°C for 10 minutes. Aliquots of equal volume for each pair were separated by SDS-PAGE, the blot probed with anti α-tubulin and the percent of polymerized tubulin calculated for each ‘P’ and ‘S’ pair is indicated. Data from several washout time points is shown as is that for untreated cells harvested at 24 or 48 hour time points (Fig 3A). Total lysates from HCN2 cells that were treated or not far 24 hours with either 100 or 25 nM of bortezomib prior to incubation in drug free medium far 0, 5, 24, 48, 120 or 144 hours were separated by SDS-PAGE. Western blots were sequentially probed with mouse monoclonal antibodies to either acetylated α-tubulin or actin (Fig. 3B). The intensity of each band was quantitated by densitometry, normalized to the value for actin, and the fold increase compared to the untreated sample.

To exclude the possibility that tubulin polymerization was not a result of “tubulin trapping in the vimentin cages” that have been described following treatment with proteasome inhibitors,^8,9 we examined the extent of vimentin polymerization following proteasome inhibition. When treated or untreated HCN2 cell lysates were separated into ‘P’ and ‘S’ [polymerized and soluble tubulin] fractions, virtually all of the vimentin separated into the ‘P’ fractions whether or not cells were treated with proteasome inhibitors (data not shown). This result emphasizes the specificity of the distribution of tubulin between the two fractions, and demonstrates that vimentin did not entrap tubulin.

Finally, one would expect that cells which exhibit MT polymerization would accumulate in mitosis. Extending published observations we found that after treating SY5Y and KCNR neuroblastoma, and 8226 multiple myeloma cells with bortezomib or any of four other proteasome inhibitors, the fraction of cells in G₂M phase increased ~50–75% compared to untreated cells (Table 3 and Fig. 4).^11,22-26

Figure 4.

The percentage of cells in the G₂M phase of the cell cycle is increased after treatment with proteaseome inhibitors. When compared to cells treated with no drug or with vehicle, an increase in the number of cells in the G₂M phase of the cell cycle was observed for SY5Y [diagonal stripe bars], KCNR [solid black bars] and 8226 cells [stippled bars] which were treated for 24 hours with proteasome inhibitors. The percentage of cells in the G₂M phase is displayed on the y-axis.

Immunofluorescent analysis of the relationship between aggresomes and the cytoskeleton. We next studied the relationship between aggresome formation and cytoskeletal structures by confocal immunofluorescent microscopy. Aggresome assembly occurs following treatment of cells with proteasome inhibitors and microtubules contribute to aggresome formation by transport of protein aggregates to these sites.⁹ Intermediate filaments, specifically vimentin, are also often in close proximity to the aggresomes.⁹ We evaluated the localization of HDAC6, a known marker for aggesomes,²¹ α-tubulin and vimentin, by immunofluorescence in SY5Y and HCN2 cells treated with the same five proteasome inhibitors. In the absence of drug, HDAC6 was fairly uniformly dispersed throughout the cytoplasm, in both HCN2 and SY5Y cells (Fig. 5, no drug, first column). However, after treatment a marked increase in HDAC6 labeled aggregates was observed in both SY5Y and HCN2 cells and often localized more to the center of the cell (Fig. 5, first column, shown for 50 μM MG-132 in SY5Y cells, and both 50 μM MG-132 and 200 nM bortezomib, in HCN2 cells, and in Suppl. Fig. 2 shown for 20 μM lactacystin, 20 μM PSI and 100 nM epoxomycin). Immunolocalization for tubulin in SY5Y or HCN2 cells identified a fine extensive network throughout all cells with only subtle bundling after treatment with proteasome inhibitors for 18–24 hours, despite distinct aggresome formation (Fig. 5). This pattern was in sharp contrast to the microtubule pattern observed in HCN2 cells treated with paclitaxel, where distinct microtubule bundles were formed, cells contracted, and aggresome formation was absent (Fig. 5).

In the absence of proteasome inhibitors, vimentin was observed as an ordered cage-like network that extended throughout the entire cell cytoplasm to the periphery (Fig. 5, no drug, last column). After proteasome inhibition, alterations in vimentin organization were observed although the vimentin network appeared to closely surround the aggresomes rather than to colocalize with them (Fig. 5 and Suppl. Fig. 3).

Analysis of microtubule associated proteins in proteasome treated cells. Despite the increase in the polymerized fraction of tubulin after treatment of cells with the proteasome inhibitors, we did not observe the extent of microtubule bundling typically seen by immunofluorescent microscopy following paclitaxel-induced MT polymer stabilization. In addition, when we examined the ability of the proteasome inhibitors to polymerize purified microtubule protein in solution, we did not find any evidence of polymerization (data not shown). These two observations indicate that the effect observed in cells treated with the proteasome inhibitors is indirect and not a result of microtubule binding as occurs with paclitaxel and other microtubule stabilizing agents. We observed that, unlike agents that bind microtubules, the degree of MT stabilization induced by proteasome inhibitors was variable across different cell lines. Microtubule associated proteins (MAPs), are known modulators of microtubule stability whose composition varies from cell to cell. We postulated that MAPs might play a role in the proteasome mediated microtubule stabilization. To test this hypothesis, we examined the levels of MAP2, MAP4 and tau in whole cell lysates from HCN2 cells and found a 1.4–3.7 fold increase in MAP2 in cells treated with the proteasome inhibitors, the largest increase being for those treated with 50 μM MG-132 (Fig. 2B). In contrast, the levels of MAP4 and tau remained monotonous under these same conditions (data not shown).

Discussion

In the present study we explored the relationship between proteasome inhibition and the microtubule cytoskeleton. We found that treatment of cells with any of five proteasome inhibitors resulted in increased tubulin polymerization and G₂M arrest, changes that are similar to those found with microtubule-stabilizing agents such as the taxanes. Because neurotoxicity and cytotoxicty are observed clinically with the taxanes, the present findings offer one possible explanation for the peripheral neurotoxicity observed with bortezomib (Velcade^®), as well as an additional mechanism to those already implicated for cytotoxicity.

Our original observations were made in renal cell carcinoma cells (786–0, data not shown) and consisted of increased tubulin polymerization following the administration of the proteasome inhibitor MG-132. Because of our interest in tubulin active agents and knowledge that peripheral neurotoxicity, constipation and hypotension have been reported following the administration of Velcade^®, we pursued these observations further.^12-14 As reported herein, we observed marked differences in the extent of tubulin polymerization following drug exposure in 8226 multiple myeloma cells, the neuroblastoma cell lines SY5Y and KCNR, and HCN2 cells of neural origin. An increase in MT polymerization was also observed in 786–0 renal carcinoma cells following treatment with bortezomib (data not shown), extending our original observation using MG-132 and indicating the generality of these findings in a variety of cell types. Differences in basal and proteasome-induced tubulin polymerization among cell types may be a consequence of differences in the composition and expression level of specific microtubule associated proteins (MAPs) that impact microtubule stability and differences in (3-isotype composition. That the changes observed after exposure to a proteasome inhibitor are a result of an increase in MT stabilization is supported by the increase in acetylated α-tubulin, detected using an antibody that recognizes this post-translational modification of lysine-40, the latter a marker of stabilized microtubules.

We observed increased tubulin polymerization with five different compounds (MG-132, bortezomib, lactacystin, P I and epoxomycin), in the class of proteasome inhibitors principally targeting the chymotrypsin activity of the proteasome.²⁷ The concentrations of the proteasome inhibitors used in our experiments (e.g., incubations ~24 hours) were sub-lethal and generally approximated 1C50 values obtained by others in 24, 48 or 72 hour viability experiments or during other functional studies performed for a variety of cell types (lactacystin,^28,29 epoxomycin,^15,29 PSI,^21,30,31 MG-132³² and bortezomib.^23,33-35) Although the concentrations used for some of the proteasome inhibitors in our experiments were higher than those used by others,^22,26 they were concentrations that, in our hands, effected very little cytotoxicity as documented here by the low percentage of cells in Sub-G₁. Cells were similarly viable under the same conditions in the immunofluorescent and biochemical experiments that we performed, and in the washout experiment, were viable 144 hours after drug treatment. This leads us to conclude that the observations herein are valid and specific and not artificial secondary to inordinately high drug concentrations.

If recognized as a class effect, one can expect that the neurotoxicity attributable to microtubule stabilization will be seen with all members of this class of drugs. The molecular mechanism responsible for the increased tubulin polymerization and MT stabilization is likely indirect since the proteasome inhibitors failed to cause increased polymerization when added to purified microtubules in vitro (data not shown). Nevertheless, the outcome might be similar if the end result is a dysfunctional microtubule. Notably, by immunofluorescence microscopy, the microtubules in cells treated with a proteasome inhibitor appeared intact, or displayed subtle bundling. This was in contrast to the taxanes or epothilones, which result in extensive MT bundling. Our observation, along with the hypothesis that the increased MT stabilization is mediated by an increase in the levels of one or several microtubule-associated proteins (MAPs), suggests that the effect of proteasome inhibitors on microtubules is more readily reversible, consistent with clinical observations that neurotoxicity following therapy with Velcade^® usually reverses more readily than that following taxanes.³⁶

Differences in tubulin polymerization and MT stabilization among cell types might also be explained by differences in the β-tubulin isotype composition. In humans, at least seven isotypes of β-tubulin exist at the protein level.³⁷ Sequence differences among the β-isotypes occur primarily in the C-terminal 15 to 20 amino acids, the putative binding site for several microtubule-associated proteins.^38,39 The C-terminal region of the protein has been shown by electron crystallography to lie on the exterior of the microtubule.⁴⁰ Although the functional significance of the differences in β-tubulin isotype expression is unknown, differences in the inherent stability and response to drugs have been demonstrated.^41-43 Expression of some β-tubulin isotypes are tissue specific whereas other β-tubulin isotypes are more universally expressed, resulting in distinct tissue patterns of expression. The class βVI isotype, for example, is found in platelets and constitutes about 90% of the total platelet tubulin (Note: Class βVI is also called Human β1; whereas Class βI is the most abundant isotype in cancer cells and is also called Human M40).⁴⁴ Class βVI tubulin mRNA is induced in megakaryocytes by the transcription factor NF-E2 (nuclear factor-erythroid 2) and organizes into a marginal band at the periphery of the cell consisting of a single 100 micrometer microtubule wrapped 7 to 12 times around itself. Studies have shown that the Class βVI-tubulin isoform is essential for the biogeneis, structure and function of platelets.^45-47 Given the effect demonstrated in this study on tubulin polymerization and the importance that this marginal band has in platelet biogenesis and function, further studies are warranted in an attempt to understand whether the effect of proteasome inhibition on the microtubule might also be responsible for the thrombocytopenia observed with Velcade^®.^12-14

Aggresomes are specialized, dynamic “holding stations” that aid in the disposal of insoluble cellular proteins.^48,49 Viewed by many as an alternate pathway to handle misfolded proteins, aggresomes are found in neurodegenerative diseases and many other pathological states, although their causal relationship with disease remains unclear.⁵⁰ Microtubules play a role in the formation of the aggresome, and also in the autophagy-lysosome pathway that is crucial for the clearance of these aggregated proteins.⁵¹ When proteasome function is either inhibited by drugs or overwhelmed with misfolded proteins, aggregates form and are transported in a microtubule dependent manner to the perinuclear microtubule-organizing center (MTOC) where an aggresome is formed.⁴⁸ HDAC6, a tubulin deacetylase, can serve as one of many adaptors linking protein cargo to the minus end-directed motor, cytoplasmic dynein. Others have shown that the tubulin deacetylase inhibitor (LBH589) and bortezomib are synergistic in MM cells generating apoptosis, diminished aggresome size and abnormal bundles of hyperacetylated α-tubulin.⁵² HDAC6 was used as in our experiments as a marker of aggresomes⁵³ and we observed aggresome formation, although the extent to which the aggregated proteins had coalesced as a major aggresome in the perinuclear region varied front one cell to another, as did changes in the organization of the intermediate filament cyroskeleton, represented by vimentin.⁹

While vincristine as a component of the VAD regimen has been used in the therapy of multiple myeloma for several decades, the role of the taxanes is less clear. Docetaxel has been shown to be inactive, while paclitaxel as a single agent in untreated patients or in combination with gemcitabine in patients with refractory disease has shown some activity.^55-57 However, by either interfering with microtubule trafficking, or by inducing G₂M arrest the increase in MT stabilization and tubulin polymerization could contribute to drug cytotoxicity. Regarding G₂M arrest, we would note that while it is expected that this should occur for a drug that induces tubulin polymerization, we observed only modest increases in G₂M values for the proteasome inhibitors, consistent with multiple effects of these agents.

We would also note that a very recent study using a synthetic lethal screen identified seven proteasome subunits among the 87 genes whose knockdown enhances paclitaxel sensitivity.⁵⁸ As the authors noted, “the probability of this extent of enrichment of proteasome subunits by random chance was 1 in 10”.¹⁰ These results demonstrate that the activities of the proteasome and microtubule targeting agents are interrelated. We would note that our observations support this conclusion, and that the study using the synthetic lethal screen in turn provides support for our work. By interfering with the degradation of microtubule associated proteins that stabilize the microtubules, inhibition or knockdown of the proteasome results in conditions favorable for microtubule stabilization making the effect of paclitaxel easier to achieve.

In summary we report increased tubulin polymerization following the addition of proteasome inhibitors in a variety of cell lines. Given the clinical observation of peripheral neuropathy as well as constipation, abdominal pain/cramping and postural hypotension reported in patients receiving Velcade^® and the similarity of these adverse effects to those observed in patients receiving microtubule targeting agents, we propose that an effect on microtubules might be responsible at least in part for these clinical findings. The possibility that microtubule stabilization might have a role in the thrombocytopenia often observed during Velcade^® therapy should also be considered. Finally, we suggest that in addition to the mechanisms already proposed, microtubule stabilization might be yet another mechanism that contributes to cellular cytotoxicity following proteasome inhibition. Further studies should bring us a greater understanding and also clarify whether this apparent class effect can be dissected and either abrogated or exploited.

Materials and Methods

Cells and reagents.

SY5Y and KCNR neuroblastoma cell lines were obtained from Carol Thiele (NIH, NCI, DBS, POB, Bethesda, MD). HCN2 and HCN1A neuronal cells were purchased from ATCC (Manassas, VA). 8226 multiple myeloma cells were from William Dalton (University of South Florida, Tampa, Florida). Cells were either grown in RPMI (SY5Y, KCNR, 8226), or DMEM (HCN2 and HCN1A) medium supplemented with 10% fetal bovine serum, glutamine, 10 units/ml penicillin and 10 μg/ml streptomycin. Bortezomib [Velcade^® Millenium Pharmaceuticals, Cambridge, MA], MG-132 (Calbiochem, San Diego, CA), Epoxomycin,¹⁵ and Z-Ile-Glu-(OtBu)-Alu-LeuH(aldehyde)(PSI)¹⁶ (Peptides International, Osaka, Japan) and lactacystin (Peptides International or Calbiochem, San Diego, CA) were each reconstituted and stored according to the manufacturers’ instructions.

Proteasome inhibitor treatment and tubulin polymerization assay.

Cells were grown in 6, 12 or 24-well dishes, to 60–80% confluency, and treated with each of the proteasome inhibitors at the concentrations indicated, at 37°C for 18–24 hours prior to analysis. The concentrations chosen were based on data in the literature that provided the initial ranges, as well as results in our own laboratory, such that at the time the cells were harvested or assayed, there was very little cell death compared to untreated controls. Cell death was measured by the Sub-G₁ population and the overall median did not exceed 13% (Table 3). In addition, at the drug concentrations and incubation times used, cellular structures were maintained, as assessed by immunofluorescence (Fig. 5 and Suppl. Figs. 2 and 3).

The non-adherent 8226 cell line was harvested by centrifugation for two minutes before being rinsed in 1X PBS and processed further. For adherent cells, the media was retrieved from each well and any floating cells collected by centrifugation for 2 minutes. These were combined with those harvested directly in the dish following a rinse in 1X PBS. The cells were all incubated at 37°C for 10–15 minutes in hypotonic lysis buffer containing 5 μM paclitaxel, 10 μM Trichostatin-A (Calbiochem, San Diego, CA), 1 mM MgCl₂, 2 mM EGTA, 0.5% Nonidet P-40, 20 mM Tris-HCl, pH 6.8 and protease inhibitors (Roche Diagnostics GmbH, Mannheim, Germany), including aprotinin (200 units/ml), vortexed vigorously and centrifuged at ~15,000 g at 22°C for 10 minutes in an Eppendorf model 5402 temperature controlled centrifuge (Brinkmann Inst., Westbury, NY), to separate polymerized (P) from soluble (S) tubulin. Upon separation, tubes were placed on ice and pellets of polymerized ‘P’ tubulin were resuspended by sonication for 10–20 seconds in a volume of lysis buffer equal to the soluble ‘S’ fraction. Each had gel sample buffer added, equal aliquots were separated by 7.5% SDS-PAGE (Bio-Rad, Hercules, CA), and western blots were obtained. As previously described, the percent ‘P’ fraction was calculated for each pair by dividing ‘P’ by ‘P + S’ and multiplying by 100.¹⁷ An advantage of this assay is that the amount of total protein loaded for each sample is irrelevant since the ‘P’ and ‘S’ fractions are equalized for each pair, and it is the proportion of the polymerized to the soluble tubulin fraction that is measured.

Western blotting and comparison of acetylated α-tubulin. After transfer of the separated polymerized and soluble cell fractions to either Immobilon-P (Millipore Corp., Bedford MA) or to nitrocellulose Protran (Biorad, Hercules, CA), immunoblotting was performed with monoclonal anti α-tubulin (DM1A, Sigma, St. Louis, MO), followed by an HRP-conjugated secondary antibody (Amersham Biosciences UK Limited, UK), as previously described,¹⁸ visualized using SuperSignal Reagents (Pierce Biotechnology, Rockford, IL), and band intensities quantified on film by densitometry using IPLabgel software. Experiments were performed multiple times as indicated in Table 1, and the average percent of polymerized tubulin, or ‘P’ fraction, as described above, for each experimental condition, is shown also for three cell lines alongside standard deviations (and for 8226 cells in Suppl. Fig. 1 and Suppl. Table 1).

Whole cell lysates were also prepared from cells that were either not treated or treated with proteasome inhibitors, as described for the tubulin separations. An aliquot of a mixture containing hypotonic lysis buffer and gel sample buffer at 98–100°C, was added to wells containing cells plated in equal numbers, or to tubes containing non-adherent cells that had been pelleted briefly. Floating cells were retrieved from the medium, as described above, and pooled with the adherent cell lysates. Tubes were vortexed vigorously and the samples boiled for 5 minutes. Equal aliquots of cell lysates were analyzed on either 7.5% or 4–15% SDS-PAGE (Bio-Rad, Hercules, CA). As an alternative method to obtain whole cell lysates, harvested cells were lysed in hypotonic lysis buffer, vortexed, placed on ice and then pulse sonicated to clarity in a bath for 20–30 seconds. Protein concentrations were determined using the Bio-Rad protein assay dye reagent. Equal amounts of total protein were then loaded in each lane of the gel. The proteins were transferred as described and the blots probed with mouse monoclonal antibodies to either acetylated α-tubulin (Clone-6-11B-1, Sigma, St. Louis, MO), GAPDH (Chemicon International, Temecula, CA), MAP2 (Calbiochent, San Diego, CA), Tau (Transduction Laboratories, San Jose, CA), or MAP4 (BD Biosciences, San Jose, CA). For re-probing, blots were stripped in Restore™ reagent according to the manufacturer’s instructions (Pierce Biotechnology, Rockford, IL). Band intensities on film were quantified by densitometry as described above and normalized to the values for either the GAPDH or acrin loading control. The average values of several experiments as indicated, with standard deviations, are displayed in Table 2 for HCN2 cells.

Analysis of the persistence of stabilized tubulin in HCN2 cells following Bortezomib treatment and washout.

H N2 cells were treated or not for 24 hours with either 100 or 25 nM of bortezomib prior to a rinse in PBS and incubation in drug free medium for 0, 5 , 24, 48, 72, 96, 120 or 144 hours. Cells were lysed as described above in buffer that contained 0.4–5 μM paclitaxel. In the case of cells incubated in 100 nM Bortezontib, lysates were either separated into polymerized (P) or soluble (S) fractions by centrifugation at ~15,000 g at 22°C for 10 minutes. Untreated cells were also harvested at 24 or 48 hour time points alongside Bortezomib treated cells. Aliquots of equal volume were separated by SDS-PAGE, the blots probed with anti α-tubulin and the percent of polymerized tubulin calculated for each ‘P’ and ‘S’ pair. Alternatively, total cell lysates were made for HCN2 cells by adding an aliquot of a mixture containing hypotonic lysis buffer and gel sample buffer at 98–100°C to wells which had undergone the drug washout periods after being treated for 24 hours with either 100 nM or 25 nM Bortezomib. Tubes containing lysates were vortexed vigorously, boiled for 5 minutes and analyzed on 4–15% SDS-PAGE as described above. Western blots were sequentially probed with mouse monoclonal antibodies to either acetylated α-tubulin, as described or actin (Chemicon International, Temecula, CA). Variations of this experiment were performed four times.

Immunofluorescence and confocal microscopy.

Cells plated on coverslips in 24 well dishes were incubated either with or without proteasome inhibitors for 20–24 hours, prior to fixation in 100% methanol at −20°C for 10 minutes.¹⁹ After rinsing in PBS the coverslips were incubated for 45 minutes in 20% goat serum in PBS followed by incubation in a primary antibody. A rabbit polyclonal antibody to HDAC6 (H-300, Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and either a mouse monoclonal antibody to vimentin (clone V9, Sigma, St. Louis, MO), or mouse monoclonal anti α-tubulin (DM1A, Sigma, St. Louis, MO) were used. Following rinses in PBS after each primary antibody incubation, the coverslips were incubated in either fluorescein (FITC)-conjugated goat anti mouse IgG or rhodamine (RHOD)-conjugated donkey anti rabbit secondary IgG, (both from Jackson Immuno Research Labs, Inc., West Grove PA) as secondary antibodies. Dapi (Sigma, St. Louis, MO) was used to counterstain nuclei. Stained cells were visualized on a Zeiss Axiovert 100 M microscope equipped with a Plan-NeoFluar 100x/1.3 oil immersion objective and confocal images were generated using a Zeiss LSM510. Images in Results are shown as 3-dimensional maximal projections reconstructed from Z-stacks.

Cell cycle analysis.

Untreated cells or those treated with a proteasome inhibitor for 24 hours, were trypsinized if adherent or pelleted if non-adherent, washed in DPBS, centrifuged and resuspended in 500 μl of a stain solution containing 0.1 mg/ml propidium iodide and 0.6% Triton-X in PBS. Subsequently, 500 μl of RNAse A solution (200 U/ml in PBS) was added and the cells incubated for 30 minutes at 22°C. Prior to analysis, the cells were passed through nylon mesh to remove clumps and aggregates. The cells were analyzed using a FACSort flow cytometer (Becton Dickinson, San José, CA). Excitation was at 488 nm with a 15 mW argon laser and fluorescence was detected with a 585 nm filter. FlowJo software (v.8.3.3) from Tree Star, Inc., (Ashland, OR) was used to determine the percentage of cells in each phase of the cell cycle.