Therapeutic Targeting of Nuclear γ-Tubulin in RB1-Negative Tumors

This article is featured in Highlights of This Issue, p. 1057

In each cell, microtubules (MT) form a cytoskeletal network of filaments that maintain cell polarity and execute chromosome partitioning during cell division (1, 2). MTs are composed of α- and β-tubulin dimers and are the target of various chemotherapies that impair microtubule function, cell division, and angiogenesis and thus reduce tumor growth (3, 4). These compounds are widely prescribed antineoplastic drugs for a broad range of malignancies, including lung, breast, gastric, esophageal, bladder and prostate cancer, Kaposi sarcoma and squamous cell carcinoma of the head and neck (3). However, the effectiveness of MT-targeting drugs for cancer therapy has been limited by drug resistance and severe side effects in treated patients (3).

An important regulator of microtubule formation is γ-tubulin, which nucleates αβ-tubulin dimers at the minus end of a growing microtubule (5). We and others have shown that in addition to its cytosolic function, γ-tubulin is a chromatin-associated protein that binds to Rad51, C53, and E2 promoter-binding factor 1 (E2F1; refs. 6–9). In various tumors and cell lines, the localization and expression of γ-tubulin are altered (10–12). In line with these findings, we discovered a mechanism allowing γ-tubulin and retinoblastoma protein (RB1) to moderate each other's expression by direct binding to E2F-binding sites on TUBG and RB1 promoter regions (6). In the absence of γ-tubulin and RB1, E2F activity leads to cellular death, whereas in the presence of RB1, absence of γ-tubulin activity causes an increase in RB1 protein levels (6, 7). As the RB-signaling pathway is one of the most distorted signal pathways in cancer cells (6, 13), the development of drugs that specifically inhibit γ-tubulin activity has the potential to pave the way for chemotherapies that target tumor cells from multiple malignancies but have no impact on healthy cells.

Despite recent efforts to find novel γ-tubulin inhibitors (14–16), to our knowledge there are no reported studies that have looked for inhibitors targeting nuclear γ-tubulin activity. Here, we characterize the effect of various MT-targeting agents on nuclear γ-tubulin. We confirm that citral affects αβ- and γ-tubulin activity (17) and we report that a citral analogue, citral dimethyl acetal (CDA), specifically inhibited the nuclear activity of γ-tubulin without affecting MT dynamics. Furthermore, we present data suggesting that the approved drug dimethyl fumarate (DMF) also interferes with the nuclear activity of γ-tubulin. Finally, in vitro, CDA and DMF killed tumor cells carrying defects in the RB1 signaling pathway, and we provide evidence of CDA's in vivo antitumorigenic activity.

See Supplementary Materials and Methods for detailed methodology.

See Supplementary Materials and Methods for detailed methodology.

All data are expressed as mean ± SD, and the statistical significance of differences between two or more groups was analyzed by paired Student t test (*, P < 0.05; **, P < 0.01) or by two-way ANOVA using Prism 6 software (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). Cell-cycle profiles were assessed using FlowJo (Tree Star, Inc.).

On the basis of the γ-tubulin–E2F–RB1 network (6, 7), inhibition of the nuclear activity of γ-tubulin will affect the activity of E2F (Fig. 1A). Applying this strategy, we examined the effect of the β-tubulin inhibitors paclitaxel (3), colcemid (18), and citral (17) on E2F transcriptional activity. We treated U2OS human osteosarcoma cells that transiently expressed luciferase reporter plasmids containing E2F-binding sites (19, 20) with the various inhibitors (Fig. 1A). Luciferase reporter assays showed that baseline luciferase activity caused by the endogenous activity of E2F was increased in a concentration-dependent manner when U2OS cells were treated with citral, but not with paclitaxel or colcemid (Fig. 1B). Moreover, treatment with citral increased RB1 protein levels (Fig. 1C). This effect was similar to that caused by the reduction of γ-tubulin protein levels by γTUBULIN shRNA (6, 7), suggesting that citral is an inhibitor of the nuclear activity of γ-tubulin.

MT-targeting compounds impair αβ-tubulin dynamics (17), causing a G₂–M transition cell-cycle arrest (3). To exclude possible effects of citral on αβ-tubulin dynamics (17), we tested the effect of citral on cell division, using cells harboring a functional RB1 pathway (murine NIH3T3) or an affected RB1 pathway (U2OS; these cells lack expression of p16^INK4; ref. 21, which causes a constitutive phosphorylation of RB1; Fig. 1D). Treatment of both cell lines with 2 mmol/L citral showed that citral triggered an accumulation of U2OS cells in G₂–M (Fig. 1E). This effect became more prominent when both cell lines were treated with 4 mmol/L citral (Fig. 1E; n = 3, P < 0.001), indicating that citral may affect MT dynamics. As citral is a mixture of two isomers, neral and geranial, we tested the following citral analogues: nerol, geraniol, geranic acid, citronellal, CDA, 6,10-dimethyl-5, 9-undecadien-2-one, α-ionone β-ionone, and all trans-retinal, in search for a more specific γ-tubulin inhibitor (Fig. 1F). Two compounds, citronellal and CDA, had the most significant effect on E2F activity at a concentration of 20 μmol/L (Fig. 1F; n = 8, P < 0.01); however, as with citral, citronellal triggered an accumulation of cells in G₂–M (Fig. 1E and G). Of all tested inhibitors, only citral, CDA, and citronellal appeared to interfere with γ-tubulin activity.

To investigate the effect of CDA on αβ-tubulin dynamics (22), we studied the impact of CDA treatment on astral MT regrowth in U2OS cells (Fig. 2A). Immunofluorescence staining showed that MT regrowth was diminished by colcemid (a known MT destabilizing agent), citral, and citronellal, but unaffected by paclitaxel and CDA (Fig. 2A). To exclude the possibility that CDA has a microtubule-stabilizing effect that hinders mitotic spindle formation, as paclitaxel does (3), we determined the effect of CDA on mitotic progression by immunofluorescence staining (Fig. 2B) and differential interference contrast (DIC)/fluorescence time-lapse microscopy in MCF10A human mammary epithelial and U2OS cells (Fig. 2C and D and movies 1 to 4). Compared with nontreated cells (movies 1 and 3), CDA neither impeded the formation of mitotic spindles in treated U2OS cells (Fig. 2B and C and movies 1 and 2) nor altered progression of MCF10A cells through mitosis (Fig. 2D and movies 3 and 4). These observations indicate that CDA is the only compound of those tested in our study that affects E2F activity without interfering with astral microtubule outgrowth and mitotic progression.

To further investigate the effect of CDA (Fig. 3A) on the RB1–γ-tubulin protein network (6), we treated various cell lines with CDA and studied its effect on cell division (Fig. 3B). Flow-cytometric analysis showed that CDA-treated cells lacking RB1 expression (U1690 and Y79; Fig. 3C) accumulated shortly first in G₂–M before dying in a dose-dependent manner (Fig. 3B and Supplementary Fig. S1), whereas at higher concentrations, CDA induced an accumulation in G₂–M (Fig. 3B; n = 4, P < 0.05) in tumor cells with a phosphorylated RB1 (U2OS; Fig. 1D), but did not affect cells with a functional RB1 pathway (MCF10A, NIH3T3, and A549; Fig. 1D and 3B). Considering the effect of CDA on E2F activity and its toxic effect on RB1-negative tumor cell lines, we studied the effect of CDA on RB1 expression. As expected with a γ-tubulin inhibitor (6), MCF10A and U2OS cells treated with CDA displayed an increase in the level of RB1 protein (Fig. 3D). Together, these findings strongly suggest that CDA targets the nuclear activity of γ-tubulin.

To demonstrate that the cytotoxic effect of CDA is dependent on the cellular levels of RB1 protein, RB1 shRNA-mediated reduction of RB1 in stably transfected MCF10A cells (Fig. 4A) was shown to sensitize MCF10A to CDA treatment (Fig. 4B). Similarly, expression of RB1 in the RB1-deficient cell line U1690 (Fig. 4C) reduced the cytotoxic effect of CDA treatment in comparison with U1690 cells lacking RB1 expression (Fig. 4D). However, the expression of a functional RB1 was transient: after few cell passages, RB1 was phosphorylated and expression levels decreased in the U1690 cells (Fig. 4E), which then became sensitive to CDA treatment (Fig. 4D).

Finally, to demonstrate that CDA targets γ-tubulin activity, we investigated the effect of CDA in U2OS cells expressing varying amounts of γ-tubulin and E2F1 (Fig. 5). We found that the cytotoxic effect of CDA was accentuated in stably transfected γTUBULIN shRNA U2OS cells (γTUBshU2OS) that expressed reduced protein levels of γ-tubulin (Fig. 5A and B). In line with these findings, the cytotoxic effect of CDA was attenuated in U2OS cells stably expressing GFP-γ-tubulin (Fig. 5C) or E2F1 shRNA (Fig. 5D and E). Overall, the data presented here show that CDA is a small-molecule inhibitor that targets the RB1–E2F1–γ-tubulin network (6).

Citral and colcemid impaired microtubule regrowth, but only citral affected E2F activity, suggesting that these compounds associate with different β-tubulin domains (15). In addition, these molecules are set-up for a Michael addition, particularly if hydrolysis occurs before or at the binding site, and thus may bind covalently to a cysteine residue (23). To find the CDA-binding pocket at the surface of γ-tubulin, we used a computational method to simulate CDA binding (Fig. 6A). Several binding pocket prediction algorithms were used and the key-binding site found at the surface of γ-tubulin was the GTP/GDP binding cavity. Sequence alignment of the GTP/magnesium-interacting residues of human α-, β-, and γ-tubulin led us to residue Cys¹³ (Supplementary Fig. S2; ref. 24). Cys¹³ is conserved in β-tubulin, but not in α-tubulin, which has an Ala instead (Supplementary Fig. S2). To determine whether the binding of CDA to γ-tubulin was Cys¹³ mediated, we analyzed the effect of CDA treatment on γTUBshU2OS cells stably expressing the non-reactive Ala¹³-GFP-γ-tubulin_resist (γTUBshU2OS-A¹³γtub_resist) or wild-type GFP-γ-tubulin_resist (γTUBshU2OS-WTγtub_resist; Fig. 6B). We found that the effect of CDA seen in γTUBshU2OS cells expressing GFP-γ-tubulin_resist cells was reduced in γTUBshU2OS-A¹³γtub_resist cells, demonstrating that Cys¹³ is necessary for CDA binding. We then docked CDA into the GTP/GDP-binding pocket of γ-tubulin. Several energetically favorable poses were found, but considering the above experimental findings, we concluded that the acetal moiety of the CDA molecule is sandwiched between Cys¹³ and Phe²²⁵, with hydrogen bonds to Asn²⁰⁷ and Asn²²⁹, in a similar manner to the guanine base of GTP (Fig. 6C; Supplementary Fig. S2). The lipophilic moiety of CDA would also follow the overall orientation of GTP although, being much smaller, it would not reach phosphate-binding subpockets (Fig. 6D). In this position, the dimethyl acetal could be hydrolyzed to an aldehyde and a Michael addition with Cys¹³ can be envisioned (Fig. 6E). This modified compound would thus be covalently bound to Cys¹³-γ-tubulin, and in this way, CDA may interfere with γ-tubulin's GTPase activity.

On the basis of CDA's interactions with Cys¹³-γ-tubulin, ligand-based chemoinformatics approaches were used to reposition approved oral drugs from two databases (25, 26) on γ-tubulin. The predicted top candidate found was dimethyl fumarate (DMF; Supplementary Fig. S3A), an approved drug for the treatment of multiple sclerosis and psoriasis (27, 28). Docking studies predicted DMF to bind to Cys¹³ in γ-tubulin (Supplementary Fig. S3B). To investigate DMF's effect on the RB1–γ-tubulin network, we treated A549 and Y79 cells with DMF and found that DMF killed cells in a RB1-dependent manner (Supplementary Fig. S3C), strongly suggesting that drugs that bind to the GTPase domain of γ-tubulin have a toxic effect on RB1-negative tumor cell lines.

To study the effect of CDA on tumor growth, we performed two initial experiments using mice that were implanted with either U1690 (first and second experiment) or U1690 cells stably expressing RB1 (U1690RB1; second experiment). In these experiments, mice suffered of an initial marked loss in body weight caused by the vehicle (Supplementary Fig. S4). Moreover, we concluded that xenografted U1690/U1690RB1 cells formed fast-growing heterogeneous tumors, and given the size of the tumors, the mice had to be sacrificed before the treatment reached a significant level (Supplementary Fig. S5A and S5B), but still after daily CDA treatment, there was a trend toward decrease in tumor growth independent of the RB1 status (Supplementary Fig. 5B).

To investigate whether this unexpected effect was real, we studied the protein expression of RB1 protein and pRB1 phosphorylated on Ser⁷⁸⁰ (6, 29) in tumors from mice implanted with U1690RB1. An almost undetectable amount of RB1 protein, which was also highly phosphorylated, was detected in both CDA and vehicle-treated tumors (Fig. 7A). To confirm that the detected protein was pRB1, we demonstrated the specificity of the anti-Ser(P)⁷⁸⁰ pRB1 by treating U1690RB1 tumors with λ-phosphatase (Fig. 7B). The altered RB1 activity in the U1690RB1 tumors therefore explained the tumor-inhibitory effect of CDA on these mice. However, we noticed differences in tumor growth between mice implanted with U1690 and U1690RB1, with the latter forming slower growing and more homogeneous tumors (Fig. 7C and Supplementary Fig. S5B).

With this in mind, we planned a third experiment. This time, in order to achieve more homogenous tumor growth, we implanted a mixture of U1690RB1 cells with Matrigel (30) and to avoid unwanted weight loss caused by the vehicle during CDA treatment, the vehicle was administered daily to the animals from day 5 after inoculation. CDA itself was well tolerated compared with the vehicle group and CDA significantly interfered with tumor growth in the treated animals (Fig. 7D and Supplementary Fig. S6). Altogether, these data demonstrate that inhibition of the nuclear activity of γ-tubulin can be considered as a new target for cancer treatment.

There is a need for development of novel chemotherapeutic agents that cause fewer side effects in treated patients. Today, αβ-tubulin–targeting compounds are widely used for the treatment of various malignancies, but these agents cause undesired side effects as they target both tumor and healthy cells. We have previously described that during S-phase, γ-tubulin moderates the activity of E2F1 and absence of γ-tubulin activity is compensated by RB1. Simultaneous reduction of RB1 and γ-tubulin activities inactivates the G₁ to S transition checkpoint and an uncontrolled E2F1 activity leads to a subsequent G₂–M checkpoint activation and cell death (6). Thus, the development of drugs that specifically inhibit γ-tubulin activity has the potential to pave the way for chemotherapies that target only tumor cells. Here, we show that CDA specifically targets the nuclear activity of γ-tubulin without affecting MT dynamics, and variations in the protein levels of γ-tubulin, RB1, or E2F1 alter its cytotoxic effect.

Our data suggest that the cytotoxic effect of CDA on tumors cells is based on its ability to interfere with GTP binding in γ-tubulin's GTPase domain by filling part of the GTP/GDP-binding cavity, an event that affects transcription and leads to cell death in cells lacking functional RB1 (6). In line with this, computational drug design using approved oral drugs led us to DMF that as CDA, is predicted to interact with Cys¹³-γ-tubulin and was then found to kill cells in a RB1-depentdent manner. Finally, we also prove that CDA treatment delays xenograft tumor progression in mice. Together, these data demonstrate that it is possible to affect the function of nuclear γ-tubulin without affecting microtubule dynamics and suggest that drugs developed to target the nuclear activity of γ-tubulin may lead to the design of more efficient and safer chemotherapeutic agents.

The finding that the GTPase domain of γ-tubulin is involved in its nuclear function is a novel and unexpected result. The lack of reaction of CDA with α-tubulin is likely to be due the absence of a free Cys in the GTPase domain pocket, whereas in the case of β-tubulin, the reason for the absence of binding and/or reaction has not yet been elucidated. Comparisons of the GTP/GDP-binding site have been performed (31) and several differences between γ-tubulin and other forms were observed in terms of amino acid composition and flexibility of some of the surrounding loops, which may explain the lack of binding of CDA to β-tubulin.

In this study, we also identified an approved drug, DMF, that we predict binds to γ-tubulin. In addition of being an FDA-approved drug for the treatment of multiple sclerosis and psoriasis (27, 28), DMF is reported to reduce melanoma growth and metastasis in animal models (32). There are several mechanism of action attributed to DMF, which are related to transcriptional regulation of the nuclear-related factor 2 (Nrf2) and nuclear factor kappa-light-chain enhancer of activated B cells (NF-κB; refs. 27, 28). Notably, nuclear γ-tubulin regulates the transcriptional activity of E2F (6, 7) and our findings suggest that γ-tubulin may also regulate the activity of other transcription factors by a yet unknown mechanism that is currently being investigated.

Overall, our results demonstrate that drugs developed to inhibit the nuclear activity of γ-tubulin can potentially act specifically on tumor cells while sparing healthy tissue. The present study may open new avenues for the rational design of novel chemotherapeutic agents for the treatment of various malignancies.

L. Lindström, R. Olsson, and M. Alvarado-Kristensson have filed a patent on CDA and its use in cancer treatment. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Lindström, R. Olsson, M. Alvarado-Kristensson

Development of methodology: L. Lindström, B.O. Villoutreix, S. Lehn, E. Nilsson, M. Alvarado-Kristensson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Lindström, B.O. Villoutreix, S. Lehn, E. Nilsson, M. Alvarado-Kristensson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Lindström, B.O. Villoutreix, R. Hellsten, E. Nilsson, R. Olsson, M. Alvarado-Kristensson

Writing, review, and/or revision of the manuscript: L. Lindström, B.O. Villoutreix, R. Hellsten, R. Olsson, M. Alvarado-Kristensson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Alvarado-Kristensson

Study supervision: E. Crneta, M. Alvarado-Kristensson

Other (computational analysis, drug repositioning, identification and selection of compounds, structural analysis of the target): Bruno O. Villoutreix

The authors thank Urban Gullberg (Lund University), Kristian Helin (Biotech Research & Innovation Centre), and Antoine Royant (Institut de Biologie Structurale) for reagents and Elevate Scientific for editorial assistance.

This work was funded by the Swedish Cancer Society (to M. Alvarado-Kristensson), the Swedish childhood cancer foundation (to M. Alvarado-Kristensson), the Skane University Hospital in Malmö Cancer Research Fund (to M. Alvarado-Kristensson), Gunnar Nilsson (to M. Alvarado-Kristensson), the Royal Physiographic Society in Lund (to L. Lindström), Novo Nordisk foundation (to M. Alvarado-Kristensson), BioCare (to M. Alvarado-Kristensson), MultiPark (to R. Olsson; both Strategic Research Programs at Lund University), and the Inserm institute (to B.O. Villoutreix).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.