OUP user menu

Systemic anti-vascular endothelial growth factor therapies induce a painful sensory neuropathy

An Verheyen, Eve Peeraer, Rony Nuydens, Joke Dhondt, Koen Poesen, Isabel Pintelon, Anneleen Daniels, Jean-Pierre Timmermans, Theo Meert, Peter Carmeliet, Diether Lambrechts
DOI: http://dx.doi.org/10.1093/brain/aws145 2629-2641 First published online: 25 June 2012


Systemic vascular endothelial growth factor inhibition, in combination with chemotherapy, improves the outcome of patients with metastatic cancer. Peripheral sensory neuropathies occurring in patients receiving both drugs are attributed to the chemotherapy. Here, we provide unprecedented evidence that vascular endothelial growth factor receptor inhibitors trigger a painful neuropathy and aggravate paclitaxel-induced neuropathies in mice. By using transgenic mice with altered neuronal vascular endothelial growth factor receptor expression, systemic inhibition of vascular endothelial growth factor receptors was shown to interfere with the endogenous neuroprotective activities of vascular endothelial growth factor on sensory neurons. In vitro, vascular endothelial growth factor prevented primary dorsal root ganglion cultures from paclitaxel-induced neuronal stress and cell death by counteracting mitochondrial membrane potential decreases and normalizing hyperacetylation of α-tubulin. In contrast, vascular endothelial growth factor receptor inhibitors exerted opposite effects. Intriguingly, vascular endothelial growth factor or vascular endothelial growth factor receptor inhibitors exerted their effects through a mechanism whereby Hdac6, through Hsp90, controls vascular endothelial growth factor receptor-2-mediated expression of the anti-apoptotic Bcl2. Our observations that systemic anti-vascular endothelial growth factor therapies interfere with the neuroprotective activities of vascular endothelial growth factor may have important implications for the application of anti-vascular endothelial growth factor therapies in cancer patients.

  • anti-angiogenesis
  • histone deacetylase 6
  • neuropathy
  • vascular endothelial growth factor


Tumours critically depend on the formation of new blood vessels for their outgrowth and metastasis. Numerous studies have highlighted vascular endothelial growth factor (VEGF) as a key inducer of this process and have shown that VEGF interference impedes vessel growth and starves tumours (Carmeliet, 2005; Jain et al., 2007). This resulted, in 2004, in the approval of a neutralizing VEGF antibody (bevacizumab) as an effective treatment for advanced colorectal, breast and lung cancer (Hurwitz et al., 2004; Miller et al., 2005; Sandler et al., 2006). Soon thereafter, VEGF receptor inhibitors, such as sorafenib and sunitinib proved equally successful in other cancers (Demetri et al., 2006; Llovet et al., 2008).

Following extensive clinical use of bevacizumab, a distinct profile of side-effects, mostly mild to moderate in severity (hypertension and proteinuria) or occurring rather uncommonly (wound healing complications and gastro-intestinal perforations) was established (Gordon and Cunningham, 2005). Emerging evidence suggests, however, that bevacizumab may also be associated with increased incidence of sensory neuropathies. Indeed, when administered in combination with FOLFOX (fluorouracil, leucovorin and oxaliplatin) the incidence of severe neuropathies induced by bevacizumab increases from 9.3% to 16.3% (Giantonio et al., 2007). Peripheral sensory neuropathies also frequently occur as a common side-effect of chemotherapy and severely impact the patient’s quality of life, leading to dose reduction or premature termination of the cytostatic regimen (Mantyh, 2006). Since bevacizumab is delivered together with chemotherapy and patients receiving both drugs exhibit prolonged progression-free survival, neuropathies occurring in patients receiving chemotherapy and bevacizumab are often attributed to prolonged exposure to chemotherapy. It is unknown, however, whether bevacizumab also directly induces or aggravates sensory neuropathies.

Likewise, sunitinib and sorafenib may also affect the PNS. Both therapies are frequently associated with hand–foot syndrome in 42% and 8.9% of the patients, respectively (Lipworth et al., 2009). Although hand–foot syndrome is considered a cutaneous condition, characterized by palmoplantar erythema, oedema and loss of skin integrity, patients developing hand–foot syndrome usually first note a tingling sensation, which subsequently progresses to a burning pain, suggesting that the PNS is also affected by these inhibitors (Lipworth et al., 2009).

Several studies revealed that VEGF also exerts strong neuroprotective effects. The first neuroprotective effects of VEGF were described by Sondell et al. (2000) who reported that VEGF promotes axonal outgrowth of primary dorsal root ganglia. More convincing in vivo evidence for VEGF-mediated neuroprotection was provided by studies in knock-in mice, in which reduced expression of VEGF caused adult-onset progressive degeneration of motor neurons (Oosthuyse et al., 2001; Storkebaum et al., 2005). Meanwhile, studies injecting VEGF inhibitors directly into the nervous system have also implicated VEGF in other neurological disorders such as depression and Parkinson’s disease (Warner-Schmidt and Duman, 2007). Although these findings triggered some initial concerns about the use of anti-VEGF therapies in cancer patients, none of these studies could convincingly demonstrate that systemic delivery of VEGF (receptor) inhibitors also directly contributes to these disorders.

Since clinical studies suggest that anti-VEGF therapies may increase the incidence of neuropathies in cancer patients, and since evidence that anti-VEGF therapies interfere with VEGF-mediated neuroprotection in cancer patients is still lacking, here we carefully assess the role of VEGF (inhibition) in the sensory nervous system.

Materials and methods

Paclitaxel-induced neuropathy and behavioural testing

Mice were injected intraperitoneally on four alternate days with paclitaxel (Bristol Myers Squibb; 1 mg/kg in saline). All mice were always tested in behavioural assays before paclitaxel injections. The following behavioural assays were used: the Von Frey test to evaluate mechanical (tactile) allodynia and the Hargreaves’ Paw Flick test to evaluate thermal hyperalgesia. A more detailed description of these tests is given in the online Supplementary material. The local ethical committee approved all experiments.

Mice overexpressing a wild-type or truncated murine VEGF receptor-2 or Flk1 transgene in post-natal neurons were generated by pronuclear injection of a mouse Thy1.2 expression cassette. ThyFlk1WT mice were further intercrossed with FVB mice to obtain heterozygous litters, while ThyFlk1DN mice were bred homozygous to achieve maximal expression levels of the Flk1DN transgene.

Isolation and treatment of dorsal root ganglion neurons

Dorsal root ganglia (all levels, unless explicitly specified) were dissected from the spinal column and collected in PBS containing 1 g/l glucose. Ganglia were then enzymatically dissociated by incubating them in medium containing 0.5% collagenase followed by 0.25% trypsin. Subsequently, ganglia were mechanically dissociated into single cells. The cell suspension was placed in a Petri dish coated with foetal calf serum for 90 min at 37°C. Dorsal root ganglion neurons were plated in poly-l-lysine coated 96-well plates in Neurobasal® medium supplemented with B27 (Gibco). Glial cell line-derived neurotrophic factor (GDNF; PeproTech EC), VEGFA (Supplementary material), SU5416 (Sigma-Aldrich) or DC101 were added to the culture medium 4 h prior to paclitaxel addition.


Dorsal root ganglion neurons were fixed using 0.5% Triton™ X-100 (Sigma) and 0.5% glutaraldehyde dissolved in PHEM buffer (buffer containing PIPES, double Hanks, EGTA and MgCl2). Cells were permeabilized with 0.5% Triton™ X-100 and incubated in 1 mg/ml NaBH4 in PHEM. Primary antibodies used are polyclonal anti-ATF3 (Santa Cruz Biotechnology 1:800) and monoclonal anti-neurofilament SMI32 (Sternberger Monoclonals Incorporated, 1:1000). For tubulin stainings, anti-acetylated and anti-total tubulin antibodies (Sigma, 1:300) were combined with anti-rabbit β3-tubulin (Covance; 1:1000). Subsequently, cells were incubated with Alexa Fluor® 488 and Alexa Fluor® 555 secondary antibodies (Invitrogen). The percentage of ATF3-positive neurons was determined using fluorescence microscopy and by analysing >100 cells per well. The percentage of acetylated or total tubulin was determined by general segmentation (KS300 image analysis software) on pictures taken with a fluorescence microscope (Zeiss). Single cell analysis was used to calculate the mean fluorescence of the cytoplasm of β3-tubulin stained neurons. To examine cell death, neurons were paraformaldehyde-fixed and labelled with anti-β3-tubulin followed by TUNEL staining (terminal deoxynucleotidyl transferase dUTP nick end labelling; Roche). The percentage of TUNEL-positive neurons was determined in each specific culture condition by using fluorescence microscopy. More than 100 neurons were counted per well, each well was analysed three times and at least two wells per condition were determined.

Western blot

Whole L4-5-6 dorsal root ganglia were isolated, snap-frozen and proteins were extracted with tissue protein extraction reagent. Dorsal root ganglion cultures were lysed in mammalian protein extraction reagent buffer or directly in SDS sample buffer. Proteins were loaded on Novex Bis–Tris gels (4–12%) (Invitrogen) and blotted on nitrocellulose. Membranes were blocked with PBS-Tween (PBS-T) containing 5% bovine serum albumin and incubated with primary antibodies overnight at 4°C. Primary antibodies used are mouse anti-acetylated or total tubulin (Sigma), rabbit anti-HDAC6 (Millipore), rabbit anti-HSP90 (StressMarq) and mouse anti-BCL2 (Santa Cruz). Primary antibodies were detected using horseradish peroxidase-labelled secondary antibodies via West Dura® enhanced chemiluminescence (Pierce, Thermoscientific). Signals were captured and quantified by a Lumi-imaging system (Roche Diagnostics).

Mitochondrial membrane potential measurement

Mouse dorsal root ganglia were cultured overnight in 96-well plates. The next day, 4 h after preincubation, neurons were loaded with 2 μM JC-1 (Molecular Probes) in PBS (+Ca2+, Mg2+ and 1 g/l glucose) for 30 min at 37°C. After washing, mitochondrial membrane potential was measured with a Zeiss LSM 510 confocal microscope. A total of 15 images was taken, paclitaxel (10 µM, diluted in PBS) or PBS was added after the second image and FCCP (10 µM), as positive control, after Image 8. Red to green ratios were calculated and normalized to the first value of each cell. Pre-values and values after treatment were averaged.


Data are shown as mean ± SEM. To calculate differences between groups, unpaired Student’s t-tests or univariate ANOVA considering equal variances was used. For the Von Frey measurements, overall differences between groups were calculated using the repeated measurement ANOVA test. Significance was defined as P < 0.05.


VEGF receptor inhibitors induce a painful neuropathy

To assess the possibility that VEGF receptor inhibitors trigger adverse effects in the sensory nervous system, we first assessed whether their systemic administration affects sensory nerve function in mice. Two inhibitors were used: a small-molecule VEGF receptor tyrosine-kinase (TK) inhibitor (SU5416) and a monoclonal antibody directed against the murine Flk1 receptor (αFlk1 or DC101) (Tessler et al., 1994). The latter was chosen because the neutralizing monoclonal antibody for murine VEGF is not commercially available. Furthermore, since Flk1 is the main ‘angiogenic’ receptor of VEGF, αFlk1 should closely mimic bevacizumab. When delivering both inhibitors intraperitoneally at doses that successfully inhibit tumour angiogenesis (Klement et al., 2000; Bergers et al., 2003), mice dose-dependently developed tactile allodynia (Fig. 1A and B). Monitoring the hind limb paw by the Von Frey aesthesiometer revealed that allodynia was mild upon administration of a daily dose of 12.5 mg/kg SU5416, but more severe after delivery of 25 mg/kg SU5416 (Fig. 1A). A similar effect was observed when mice were treated with 20 and 40 mg/kg αFlk1 three times per week (Fig. 1B). Notably, tactile allodynia developed rapidly, within days of the first injection, but the effect was only transient, as mice gradually recovered after the last injection. Mice treated with αFlk1 also developed thermal hyperalgesia, as assessed by the Paw Flick test (Supplementary Fig. 1).

Figure 1

Systemic delivery of VEGF receptor inhibitors induces and aggravates a painful neuropathy. (A) SU5416 dose-dependently induces mechanical allodynia, as measured with a Von Frey test [P = 0.49 and P < 0.001 for low and high doses of SU5416 versus DMSO (dimethylsulphoxide); n = 10–10]. *P < 0.05 high dose SU5416 versus DMSO at individual days. (B) Low and high doses of DC101 induce a similar mechanical allodynia as SU5416 (P = 0.073 and P = 0.023 for low and high doses of DC101 versus rat IgG; n >10 per group). *P < 0.05 for high dose of DC101 versus rat IgG at individual days. (C) Co-delivery of paclitaxel and DC101, on alternate days, induces more mechanical allodynia than paclitaxel or rat IgG alone (P < 0.001; n = 16–16). *P < 0.05 for DC101 + paclitaxel versus rat IgG + paclitaxel. (D) Rat IgG and DC101 are delivered to mice, in which systemic paclitaxel was used to induce a painful neuropathy. Mice receiving rat IgG recover very quickly and exhibit a normal withdrawal response 1 day after the first injection (P = 0.15 versus vehicle treatment; n = 8–10). In contrast, DC101-treated mice exhibit a reduced withdrawal response for the total duration of the treatment (P = 0.004 and P = 0.042 after 1 and 2 weeks of DC101 versus rat IgG; n = 8–8). *P < 0.05 DC101 versus rat IgG-treated mice on individual days.

VEGF receptor inhibitors aggravate painful paclitaxel-induced neuropathies

In clinical practice, bevacizumab is combined with a reference chemotherapy such as paclitaxel. The intriguing question therefore arises whether VEGF receptor inhibitors—in particular αFlk1 as it most closely mimics bevacizumab—also aggravate paclitaxel-induced neuropathies. In a first set of experiments, αFlk1 was delivered together with paclitaxel. Mice receiving paclitaxel or αFlk1 alone exhibited a similar withdrawal response in the Von Frey test, whereas mice receiving both paclitaxel and αFlk1 developed more tactile allodynia than mice receiving paclitaxel alone (Fig. 1C). Alternatively, when αFlk1 was given after halting paclitaxel injections, i.e. to mice that had already developed a painful neuropathy due to paclitaxel, mice treated with αFlk1 failed to recover. In contrast, mice receiving paclitaxel and rat IgG quickly recovered after halting paclitaxel injections (Fig. 1D).

VEGF exerts direct neuroprotective effects on isolated dorsal root ganglion neurons

Since VEGF can affect both vessels and nerves, it is possible that VEGF receptor inhibitors—in addition to their established effects on the vasculature—also directly affect dorsal root ganglion neurons. To assess this intriguing hypothesis, we first characterized the neuroprotective effects of VEGF on primary dorsal root ganglion neurons. Since pilot studies revealed that paclitaxel did not induce cell death of dorsal root ganglion neurons, but rather increased expression of the neuronal stress marker ATF3 (activating transcription factor 3, Supplementary Fig. 2A and B), primary dorsal root ganglion cultures were challenged with a low concentration of paclitaxel (10 nM), which increased ATF3 without inducing cell death (Supplementary Fig. 2). VEGF administration to primary dorsal root ganglion neurons exposed to paclitaxel dose-dependently reduced ATF3 immunoreactivity and was as effective as GDNF in exerting these effects (Fig. 2A).

Figure 2

VEGF and VEGF receptor inhibitors directly affect primary dorsal root ganglion neurons through Flk1. (A) Measurement of neuronal stress in primary dorsal root ganglion cultures by quantifying ATF3 reactivity after paclitaxel administration. VEGF dose-dependently reduces paclitaxel-induced ATF3 reactivity (r2 = 0.999 after a sigmoidal fit, n = 6 wells from three rats per condition) similarly as glial cell-derived neurotrophic factor (GDNF). ***P < 0.001 versus vehicle; #P < 0.05, ##P < 0.01 and ###P < 0.001 versus paclitaxel. (B) Paclitaxel reduces the mitochondrial membrane potential (ΔΨm) relative to PBS (n > 20 neurons from three mice). Pretreatment (4 h) of dorsal root ganglion cultures with VEGF protects against a decrease in ΔΨm. **P < 0.01 versus PBS; ##P < 0.01 and ###P < 0.001 versus paclitaxel. (C) Mouse dorsal root ganglion cultures treated with paclitaxel for 24 h contain more dead neurons than untreated dorsal root ganglion cultures. Pretreatment with VEGF protects neurons against paclitaxel-induced cell death (n = 6 wells from three mice). ***P < 0.001 versus vehicle; #P < 0.05 and ##P < 0.01 versus paclitaxel. (D and E) Tubulin acetylation as determined by quantitative microscopy is increased by paclitaxel. This increase is partly counteracted by the addition of VEGF (n > 40 neurons from two mice). ***P < 0.001 versus vehicle; ##P < 0.01 versus paclitaxel. Representative images are shown in (E). Acetylated tubulin (red), β3 tubulin (green) and DAPI (blue) expression in a representative dorsal root ganglion neuron from a primary dorsal root ganglion culture. Treatments are indicated on each panel. Scale bar = 25 µm. (F and G) Western blots (F) for acetylated tubulin on dorsal root ganglion neurons treated with vehicle, VEGF, with and without paclitaxel confirming that VEGF reduces the amount of acetylated tubulin (P = 0.024, n = 4 mice). **P < 0.01 versus PBS and #P < 0.05 versus paclitaxel. (H) At baseline, ATF3 is increased in Thy:Flk1DN neurons. Paclitaxel induces only minimal neuronal stress in Thy:Flk1WT neurons compared with wild-type (WT) and Thy:Flk1DN neurons. VEGF reduces ATF3 levels in wild-type neurons, whereas Thy:Flk1DN neurons do not respond to VEGF treatment (P = 0.84) (n > 6 wells from at least three mice per genotype). *P < 0.05 and ***P < 0.001 versus respective vehicles; #P < 0.05 and ##P < 0.01 versus paclitaxel. (I) VEGF receptor inhibitors dose-dependently increase ATF3 reactivity in primary wild-type dorsal root ganglion cultures (r2 = 0.836 and 0.968 for DC101 and SU5416, respectively; n = 6 wells per condition). **P < 0.01 and ***P < 0.001 versus vehicle. (J and K) Exposure to DC101 or paclitaxel causes hyperacetylation of tubulin in dorsal root ganglion cultures (n > 45 neurons from three mice). Notably, DC101 combined with paclitaxel causes an additional increase in acetylated tubulin. *P < 0.05 and ***P < 0.001 versus vehicle and ###P < 0.001 versus paclitaxel. Representative images are shown (K). Acetylated tubulin (red), β3 tubulin (green) and DAPI (blue) expression in a representative dorsal root ganglion neuron from a primary dorsal root ganglion culture. Treatments are indicated on each panel. Scale bar = 25 µm. (L) Western blot for acetylated tubulin on dorsal root ganglion neurons treated with rat IgG, DC101, with and without paclitaxel confirming that DC101 increases the amount of acetylated tubulin (P = 0.03, n = 3 mice) alone and in combination with paclitaxel (P = 0.018). *P < 0.05 and ***P < 0.001 versus rat IgG and #P < 0.05 versus paclitaxel. (M) Acetylated/total tubulin ratio is increased in L4–L6 dorsal root ganglia from DC101-treated mice compared with dorsal root ganglia from rat IgG-treated mice (n = 4). *P < 0.05 and **P < 0.01 versus rat IgG.

Next, we assessed whether VEGF also counteracts paclitaxel-induced decreases in the mitochondrial membrane potential (ΔΨm) as one of the earliest markers of apoptosis (Bernardi et al., 1999). To monitor ΔΨm, we exposed the cell-permeable mitochondrial probe JC-1 to dorsal root ganglion cultures and applied two-colour confocal microscopy. At a high ΔΨm, JC-1 is detectable as a red fluorescent signal, but as ΔΨm drops, the fluorescence emitted by JC-1 changes to green. The ΔΨm can thus be monitored as the ratio of red over green mean fluorescent intensity. Administration of paclitaxel (10 µM, Supplementary Fig. 2C) caused a significant reduction in ΔΨm and this reduction could effectively be inhibited by pretreating cultures with VEGF (Fig. 2B).

Finally, we also assessed whether VEGF affects paclitaxel-induced cell death of primary dorsal root ganglion neurons. Rather than exposing dorsal root ganglia to low levels of paclitaxel, which were used to assess the effects on ATF3 immunoreactivity, higher levels of paclitaxel were applied to assess the effect on cell death (Supplementary Fig. 2D). As expected, VEGF also protected dorsal root ganglion neurons against 10 µM paclitaxel-induced cell death, as assessed by TUNEL staining (Fig. 2C). Overall, this confirms that VEGF exerts direct neuroprotective effects on primary dorsal root ganglion neurons, by reducing neuronal stress and protecting against cell death.

VEGF counteracts paclitaxel-induced increases in tubulin acetylation

Paclitaxel directly binds to microtubules, thereby preventing depolymerization of α-tubulin and inducing a stable, hyper-acetylated tubular state (Schiff et al., 1979). We therefore also assessed whether VEGF is capable of counteracting these effects, as a measure of protection against paclitaxel-induced neurotoxicity. To this extent, we immunocytochemically quantified fluorescence intensities of acetylated and total α-tubulin in the cytoplasm of dorsal root ganglion neurons using multichannel fluorescence microscopy. As expected, paclitaxel increased acetylated α-tubulin levels dose-dependently (Supplementary Fig. 2E). This increase was counteracted by the addition of VEGF (Fig. 2D). Representative images are shown in Fig. 2E. Furthermore, to observe the general state of tubulin acetylation in dorsal root ganglion cultures after paclitaxel and VEGF, a complimentary western blot analysis for acetylated and total tubulin was performed to confirm these effects (Fig. 2F and G). Since paclitaxel is known to upregulate various isoforms of tubulin, representing the cell’s attempt to replenish the supplies of tubulin monomers that are depleted due to paclitaxel (Stargell et al., 1992), total α-tubulin levels were also quantified. Paclitaxel increased total α-tubulin levels, whereas VEGF failed to affect total α-tubulin levels (Supplementary Fig. 3A–C). Similar results were obtained using an independent method to quantify acetylated and total α-tubulin levels (Supplementary Fig. 3D), thus confirming that VEGF selectively reduces acetylated but not total α-tubulin levels when combined with paclitaxel.

Flk1 mediates the neuroprotective activities of VEGF

Since VEGF induces potent neuroprotective activities on primary dorsal root ganglion neurons and VEGF receptor inhibitors, possibly by interfering with the direct neuroprotective activities of VEGF, induce a painful neuropathy, the role of the VEGF receptors, Flk1 and Flt1, in mediating VEGF’s neuroprotective activities was studied in more detail. To this end, various transgenic mouse strains with altered expression of Flk1 or Flt1 were used, including mice overexpressing, under control of the neuron-specific Thy1.2 promoter, a wild-type (Thy:Flk1WT mice) or dominant-negative Flk1 (Thy:Flk1DN mice). Quantification of Flk1 messenger RNA transcripts confirmed that Flk1 was overexpressed in whole dorsal root ganglia (number of Flk1 copies per 103 β-actin copies: 7.32 ± 0.30, 855.40 ± 81.99 and 3.91 ± 0.28 for ThyFlk1WT, ThyFlk1DN and wild-type mice, respectively, P < 0.005 versus wild-type mice) and primary dorsal root ganglion cultures (number of Flk1 copies per 103 β-actin copies: 11.88 ± 1.04, 134.83 ± 25.45 and 4.34 ± 0.65 for ThyFlk1WT, ThyFlk1DN and wild-type mice, respectively, P < 0.01 versus wild-type mice). Additionally, knock-in mice expressing a TK dead Flt1 receptor (Flt1-TK−/−) were also used.

When isolating primary dorsal root ganglion cultures from wild-type, Thy:Flk1WT and Thy:Flk1DN mice under baseline conditions, ATF3 levels did not differ in wild-type and Thy:Flk1WT dorsal root ganglia. On the other hand, ATF3 levels were slightly increased in Thy:Flk1DN dorsal root ganglia (Fig. 2H), presumably due to a lack of baseline Flk1 activation. Upon challenge with a low concentration of paclitaxel, Thy:Flk1WT dorsal root ganglion cultures exhibited less pronounced increases in ATF3 levels than wild-type or Thy:Flk1DN cultures. When cultures were subsequently treated with VEGF prior to exposure to paclitaxel, ATF3 levels decreased in treated wild-type cultures, but Thy:Flk1DN cultures failed to respond (Fig. 2H). Notably, Thy:Flk1WT, but not Thy:Flk1DN dorsal root ganglion cultures, also exhibited less hyperacetylation of α-tubulin, but not total tubulin, when challenged with paclitaxel (Supplementary Fig. 4). On the other hand, dorsal root ganglion cultures with a defective Flt1 receptor (Flt1-TK−/−) exhibited decreased ATF3 and hyperacetylation of α-tubulin after VEGF exposure (data not shown), indicating that Flt1 was not crucially involved in mediating VEGF’s neuroprotective activities.

VEGF receptor inhibitors negatively affect primary dorsal root ganglion neurons

Since VEGF potently protected primary dorsal root ganglion neurons through Flk1, we assessed whether VEGF receptor inhibitors also directly affect primary dorsal root ganglion neurons and exert similar effects as paclitaxel. When exposing primary dorsal root ganglion neurons to αFlk1 and SU5416, both inhibitors dose-dependently increased ATF3 levels (Fig. 2I) without affecting cell death. Additionally, αFlk1 caused a significant increase in acetylated α-tubulin (Fig. 2J) without affecting total α-tubulin levels (Supplementary Fig. 5). When αFlk1 was combined with paclitaxel, acetylated α-tubulin levels also increased compared to αFlk1 or paclitaxel alone (Fig. 2J). Representative images confirming these effects are shown in Fig. 2K. A complimentary western blot on dorsal root ganglion cultures treated with αFlk1 with or without paclitaxel confirmed this effect (Fig. 2L and Supplementary Fig. 5). Similar effects on acetylated tubulin levels were observed for SU5416 (data not shown). Overall, these data illustrate that VEGF receptor inhibitors exert direct effects on primary dorsal root ganglion neurons and amplify the neurotoxic effects of paclitaxel. Notably, upon systemic delivery, αFlk1 also increased expression of Atf3 in dorsal root ganglia as determined by real-time PCR (RT-PCR) expression analysis of whole dorsal root ganglia. Likewise, we found that αFlk1 increased acetylated versus total α-tubulin levels (Fig. 2M), thereby reinforcing the hypothesis that VEGF receptor inhibitors may directly affect the sensory nervous system.

VEGF receptor inhibitors interfere with the neuroprotective effects of Flk1

To further assess the in vivo relevance of Flk1-mediated neuroprotection, we characterized sensory nerve function in Thy:Flk1WT, Thy:Flk1DN and wild-type mice under normal conditions and after exposure to paclitaxel. Thy:Flk1WT, Thy:Flk1DN and wild-type mice appeared healthy and fertile and exhibited normal densities of PGP9.5+ axons or perfused vessels in their paws (Supplementary Fig. 6). At baseline, Thy:Flk1WT and wild-type mice also responded similarly to mechanical pressure applied by the Von Frey meter. Intriguingly, Thy:Flk1DN mice showed clear signs of tactile allodynia (Fig. 3A). When assessing thermal hyperalgesia, Thy:Flk1DN were also hypersensitive to heat compared to Thy:Flk1WT and wild-type mice, which behaved normally in the Paw Flick assay (Supplementary Fig. 1).

Figure 3

VEGF receptor inhibitors interfere with the neuroprotective effects of Flk1. (A) Systemic paclitaxel-treated Thy:Flk1WT mice are protected against mechanical allodynia [n = 9–11; P = 0.045 for Thy:Flk1WT mice versus wild-type (WT) mice]. Thy:Flk1DN mice have mechanical allodynia at the start of the experiment and do not recover over time (P = 0.005 versus wild-type mice; n = 11–18). *P < 0.05 and #P < 0.05 for Thy:Flk1WT and Thy:Flk1DN mice versus wild-type mice, respectively. (B and C) Systemic paclitaxel increases the acetylated/total tubulin ratio in wild-type and Thy:Flk1DN dorsal root ganglia, but not in Thy:Flk1WT dorsal root ganglia (P = 0.83) (n = 6 mice per group). A representative western blot is shown. *P < 0.05 versus untreated mice. (D) Thy:Flk1WT mice receiving DC101 are resistant to hypersensitivity (P = 0.39 versus rat IgG Thy:Flk1WT mice). *P < 0.05 DC101 Thy:Flk1WT mice versus DC101 wild-type mice at individual days. (E) At baseline, Thy:Flk1DN mice display a mechanical allodynia, which initially is only mild (P = 0.024 at baseline versus wild-type mice) but then aggravates over time (P < 0.001 at Day 22). DC101 does not cause any additional hypersensitivity (P = 0.93 versus rat IgG Thy:Flk1DN mice). Mechanical allodynia of DC101-treated wild-type mice is comparable to that in Thy:Flk1DN mice (P = 0.81 for rat IgG Thy:Flk1DN mice versus DC101 wild-type on Day 8; n = 8–8). *P < 0.05 for rat IgG Thy:Flk1DN mice versus rat IgG wild-type mice on individual days.

When Thy:Flk1WT, Thy:Flk1DN and wild-type mice were subsequently challenged with systemic paclitaxel, Thy:Flk1WT mice developed a less painful neuropathy compared with wild-type mice, as assessed by the Von Frey test (Fig. 3A). In Thy:Flk1DN mice, tactile allodynia did not increase substantially, such that on the last day of paclitaxel injections, Thy:Flk1DN and wild-type mice showed a similar withdrawal response. However, after halting paclitaxel injections, tactile allodynia disappeared in wild-type mice, whereas Thy:Flk1DN mice failed to recover (Fig. 3A). When quantifying acetylated and total α-tubulin levels in whole dorsal root ganglia at the time of the last paclitaxel injection, as a measure of induced neurotoxicity, wild-type and Thy:Flk1DN dorsal root ganglia exhibited increased acetylated levels relative to total α-tubulin levels. On the other hand, these levels remained unaltered in Thy:Flk1WT mice (Fig. 3B and C). Overall, these data indicate that neuron-specific overexpression of Flk1 protects against paclitaxel-induced neuropathy, whereas dominant-negative inhibition of neuronal Flk1, similar to the systemic delivery of αFlk1, induces a painful sensory neuropathy.

In a next set of experiments, we measured tactile allodynia in wild-type, Thy:Flk1WT and Thy:Flk1DN mice treated with systemic αFlk1. Thy:Flk1WT mice did not develop allodynia after αFlk1 (Fig. 3D), presumably due to residual activation of overexpressed Flk1 receptors that could not be blocked by αFlk1. On the other hand, Thy:Flk1DN mice failed to develop additional signs of allodynia and were as sensitive to a tactile stimulus as wild-type mice treated with αFlk1 (Fig. 3E). Similar effects were observed when thermal hyperalgesia was assessed (Supplementary Fig. 1). In conclusion, since Thy:Flk1WT mice treated with αFlk1 did not develop a painful neuropathy, these data indicate that αFlk1 interferes with the neuroprotective activities of Flk1 to sensitize sensory nerves.

The neuroprotective activities of Flk1 are Hdac6-dependent

Since neuronal Flk1 was essential for normal sensory nerve function and potently protected against a paclitaxel-induced neuropathy, we assessed downstream signalling activation of Flk1 after VEGF stimulation. VEGF stimulation and Flk1 overexpression both decreased acetylated α-tubulin levels after paclitaxel exposure, thereby suggesting that histone deacetylase 6 (HDAC6; Hubbert et al., 2002), which is responsible for deacetylation of α-tubulin, is involved in mediating these effects. Two HDAC inhibitors were used to assess this hypothesis: trichostatin A, which inhibits class I and II mammalian HDACs, and tubacin, which specifically inhibits HDAC6. As expected, trichostatin A (30 nM) and tubacin (100 nM) increased acetylated α-tubulin levels in primary dorsal root ganglia (Fig. 4A and B) without affecting total α-tubulin levels. VEGF decreased acetylated α-tubulin levels in paclitaxel-treated cultures but failed to reduce acetylated α-tubulin when combined with trichostatin A or tubacin (Fig. 4A and B). We then assessed whether the neuroprotective activities of VEGF were affected by HDAC6 inhibition. As expected, VEGF was protective against a paclitaxel-induced increase in ATF3. However, in combination with trichostatin A or tubacin, VEGF failed to reduce ATF3 expression (Fig. 4C and D). Likewise, although trichostatin A and tubacin individually did not affect ΔΨm, VEGF combined with either trichostatin A or tubacin failed to prevent a paclitaxel-induced decrease in ΔΨm (Fig. 4E and F). Finally, VEGF also failed to provide protection against paclitaxel-induced cell death when combined with trichostatin A or tubacin (Fig. 4G and H), thus indicating that Hdac6 plays an essential role within the pathway of VEGF-mediated neuroprotection against paclitaxel-induced toxicity.

Figure 4

The neuroprotective effects of Flk1/VEGF are Hdac6-dependent. (A and B) Trichostatin A (TSA) and tubacin cause an increase in acetylated tubulin. Trichostatin A additionally increases acetylated tubulin in combination with paclitaxel. VEGF fails to reduce paclitaxel-induced hyperacetylation in combination with trichostatin A (P = 0.3) or tubacin (P = 0.9) (n > 60 neurons from four mice). *P < 0.05 and ***P < 0.001 versus vehicle; #P < 0.05 and ##P < 0.01 versus paclitaxel. (C and D) Treatment of dorsal root ganglion neurons with trichostatin A or paclitaxel causes neuronal stress. VEGF does not protect against paclitaxel-induced neuronal stress in combination with trichostatin A (P = 0.67) or tubacin (P = 0.35) (n = 6 wells from three rats per condition). **P < 0.01 and ***P < 0.001 versus vehicle; #P < 0.05 versus paclitaxel. (E and F) Pretreatment of dorsal root ganglion cultures with VEGF protects against paclitaxel-induced depolarization, but not when VEGF is combined with trichostatin A or tubacin (P = 0.30 and P = 0.37, respectively; n > 14 neurons from at least three mice). *P < 0.05 and **P < 0.01 versus PBS; #P < 0.05 versus paclitaxel. (G and H) VEGF reduces neuronal cell death caused by paclitaxel, but not in combination with trichostatin A (P = 0.14) or tubacin (P = 0.14) (n = 6–8 wells from three mice). ***P < 0.001 versus vehicle; #P < 0.05 and ##P < 0.01 versus paclitaxel. NS = not significant.

Hsp90 and Bcl2 as mediators of Flk1-driven neuroprotection

To explore how Hdac6 regulates Flk1-mediated neuroprotection, we assessed whether VEGF by directly affecting Hdac6 activity reduces acetylation of α-tubulin and protects against paclitaxel-induced toxicity. We failed, however, to detect any change in Hdac6 messenger RNA or protein expression levels in primary dorsal root ganglion cultures after VEGF exposure (data not shown). Moreover, VEGF did not affect acetylation of α-tubulin in primary dorsal root ganglion neurons not exposed to paclitaxel (Supplementary Fig. 3D), suggesting that there is no direct interaction between VEGF and Hdac6. Overall, this finding supports previous observations that hyperacetylation of α-tubulin represents a marker rather than a cause of microtubule hyperstability (Zhang et al., 2003).

Intriguingly, HDAC6 is also involved in the regulation of various other non-histone proteins. For instance, HDAC6-mediated deacetylation activates Hsp90, which is an important regulator of cell signalling (Aoyagi and Archer, 2005). Additionally, in leukaemia cells, VEGF promotes survival through Hsp90-mediated induction of Bcl2 expression and inhibition of apoptosis (Dias et al., 2002). We therefore assessed whether Hdac6, through the regulation of Hsp90 and Bcl2, could mediate the neuroprotective effects of VEGF against paclitaxel-induced toxicity. Whole dorsal root ganglia isolated from Thy:Flk1WT mice showed increased expression of BCL2 and HSP90 proteins (Fig. 5A). Likewise, VEGF stimulation of primary dorsal root ganglion cultures increased expression of BCL2 and HSP90 (Fig. 5B). However, when VEGF stimulation was combined with the HDAC6-specific inhibitor tubacin, expression of BCL2 failed to be upregulated (Fig. 5B and C), whereas expression of HSP90 was unaffected (Fig. 5B and D), suggesting that failure of deacetylating Hsp90 by Hdac6 inhibits VEGF-induced expression of Bcl2. Since quantification of acetylated HSP90 after immunoprecipitation of total HSP90 was technically not feasible in primary dorsal root ganglion cultures, we could not, however, formally prove that deacetylation of HSP90 by Hdac6 indeed mediated these effects. Application of the Hsp90 inhibitor 17-AAG (0.5 μM) in combination with VEGF and paclitaxel inhibited the neuroprotective effects of VEGF (Fig. 5E and F), thereby reinforcing the role of HSP90. When applying the Bcl2 inhibitor ABT-737 (1 μM), at concentrations that failed to induce cell death (Supplementary Fig. 7), together with VEGF and paclitaxel, the neuroprotective activities of VEGF on cell death (Fig. 5G) and tubulin deacetylation (Fig. 5H) were also inhibited. Overall, these data suggest that VEGF-mediated neuroprotection against paclitaxel-induced toxicity depends at least partially on a mechanism (Fig. 5I), whereby Hdac6 mediates VEGF-induced upregulation of Bcl2.

Figure 5

Hsp90 and Bcl2 mediate VEGF/Flk1-induced neuroprotection. (A) Western blot for BCL2 and HSP90 on whole dorsal root ganglia from Thy:Flk1WT mice and wild-type (WT) littermates. Thy:Flk1WT mice have 47.37 ± 16.02% more BCL2 and 22.42 ± 4.56% more HSP90 in their dorsal root ganglia compared with wild-type mice (n = 3–4 mice per genotype). *P < 0.05 versus wild-type mice. (B–D) Western blot on dorsal root ganglion cultures for BCL2 and HSP90 (B) reveals that VEGF causes a 38.22 ± 11.68% significant increase in BCL2 but not when combined with tubacin (P = 0.22 versus untreated cultures; n = 3). VEGF also increases the expression of HSP90 with 33.49 ± 12.78%, which is not affected by tubacin (P = 0.91 versus VEGF-treated cultures; n = 6). Representative blots for BCL2 (C) and HSP90 (D) experiments are shown. *P < 0.05. (E) VEGF prevents paclitaxel-induced cell death, but not when VEGF is combined with 17-AAG (P = 0.43; n = 8–10 wells from three mice). ***P < 0.001 versus vehicle; #P < 0.05 versus paclitaxel. (F) Paclitaxel-induced hyperacetylation of tubulin is partly prevented by VEGF, but not when VEGF is combined with 17-AAG (P = 0.63; n > 40 neurons from two mice). ***P < 0.001 versus vehicle; ##P < 0.01 versus paclitaxel. (G) VEGF protects against paclitaxel-induced cell death, but not when VEGF is combined with a Bcl2 inhibitor (ABT-737) (P = 0.86) (n = 8–9 wells from three mice). ***P < 0.001 versus vehicle; #P < 0.05 versus paclitaxel. (H) VEGF partly prevents paclitaxel-induced hyperacetylation of tubulin, but not in combination with ABT-737 (P = 0.24; n > 50 neurons from two mice). ***P < 0.001 versus vehicle and #P < 0.05 versus paclitaxel. (I) Proposed model for VEGF-mediated neuroprotection against paclitaxel. VEGF binds Flk1 and increases expression of Hsp90 and Bcl2, thereby protecting neurons from paclitaxel-induced toxicity. NS = not significant.

Finally, we also observed that BCL2 expression was drastically decreased in primary dorsal root ganglion cultures exposed to αFlk1 (41.1 ± 7.8% reduction in BCL2 relative to control cultures by western blotting, P = 0.0016). To show that reduced BCL2 expression after αFlk1 was mediated by HDAC6, we also transiently expressed a HDAC6-GFP construct in endothelioma E2 cells (the latter cells were used since transfection in primary dorsal root ganglion cultures was technically not feasible). Overexpression of HDAC6 at least partially inhibited the increase in acetylated tubulin levels induced by αFlk1 as well as the reduction in BCL2 expression relative to untransfected cells (Supplementary Fig. 8). Similar effects were observed for the other VEGF receptor inhibitor SU5416 (Supplementary Fig. 8), thereby reinforcing the role of HDAC6 in Flk1-mediated neuroprotection and the neurotoxic effects of the VEGF receptor inhibitors.


The most important finding of the study is that systemic application of VEGF (receptor) inhibitors triggers undesired effects in the sensory nervous system. Indeed, when systemically delivering αFlk1 or SU5416, mice dose-dependently developed mechanical allodynia and thermal hyperalgesia, and also suffered from more severe paclitaxel-induced neuropathies. Remarkably, wild-type mice receiving αFlk1 developed an allodynia that was very similar to that of Thy:Flk1DN mice under baseline conditions, whereas Thy:Flk1WT mice were completely resistant to αFlk1. These observations clearly indicate that VEGF receptor inhibitors interfere with the endogenous neuroprotective effects of VEGF.

Are there any reasons to be wary of similar effects when applying anti-VEGF (receptor) therapies to humans? Several observations suggest that caution should be warranted. When delivered simultaneously with systemic paclitaxel, αFlk1 aggravated paclitaxel-induced neuropathy in mice. Combined treatment regimens of bevacizumab and taxanes are effectively being used in the clinic. The fact that increased incidence of sensory neuropathies in combination therapies has already been reported, suggests that anti-VEGF therapies also contribute to neuropathies in humans (Miller et al., 2007). Increasing evidence indicates that anti-VEGF therapies are being delivered for longer periods. Indeed, due to their efficacy in delaying disease progression, cancer patients receive chemotherapy and bevacizumab for much longer periods and neuropathies induced by both substances may therefore accumulate or aggravate over time. Clinical trials assessing the efficacy of adjuvant anti-angiogenic therapies, which implicates prolonged treatment schedules with bevacizumab, have recently also been initiated (Allegra et al., 2011). Although VEGF receptor inhibitors, sorafenib and sunitinib, have proven single-agent activity in patients with solid tumours, combination therapies with chemotherapy are currently also being considered (Hauschild et al., 2009). Based on our findings, the long-term effects of anti-VEGF therapies in the sensory nervous system of patients receiving such prolonged combination treatments should carefully be monitored. Importantly, VEGF is not only neuroprotective for sensory neurons—it also potently affects other types of neurons. Therefore, it is possible that the long-term delivery of anti-VEGF (receptor) therapies will also interfere with these activities, thereby inducing or accelerating the onset of other neurological disorders. In this respect, it is striking that a considerable fraction of glioblastoma patients treated with bevacizumab develop severe optic neuropathies (Sherman et al., 2009).

Since systemic VEGF inhibition—alone or in combination with paclitaxel—interferes with the neuroprotective activities of VEGF, we carefully studied the mechanisms of VEGF-mediated neuroprotection in sensory neurons. We found that Hdac6 regulates Flk1-driven upregulation of the anti-apoptotic BCL2 protein, resulting in protection against paclitaxel-induced toxicity. HDAC6 can regulate deacetylation of various non-histone proteins, including Hsp90 (Kovacs et al., 2005), cortactin (Zhang et al., 2007) and β-catenin (Li et al., 2008). Although we could not provide any formal proof, HDAC6 could participate in the proposed pathway by deacetylating Hsp90. VEGF has been shown to stimulate the survival of various cell types through a Bcl2 dependent mechanism (Nor et al., 1999; Pidgeon et al., 2001; Hwang et al., 2009). Our study is the first, however, to report that VEGF regulates Bcl2 expression through an Hdac6- and Hsp90-dependent mechanism. Interestingly, Bcl2 has also been shown to protect against paclitaxel-induced toxicity by directly interacting with microtubules and preventing polymerization (Nuydens et al., 2000). In addition to this, paclitaxel can also functionally mimic the endogenous Bcl2 ligand (Nur77) and directly bind to Bcl2 (Ferlini et al., 2009). A potential secondary mechanism of VEGF-induced neuroprotection against paclitaxel could thus rely on enhanced neutralization of paclitaxel through upregulation of Bcl2. Since we found that VEGF receptor inhibitors also affect BCL2 expression, paclitaxel and VEGF receptor inhibitors could both exert their neurotoxic activities by affecting BCL2 expression. In the case of paclitaxel, which exerts a much broader neurotoxicity profile than VEGF receptor inhibitors, it should be noted, however, that BCL2 represents only one of the many molecules or pathways that may be affected.

Interestingly, the role of HDAC6-driven deacetylase activity in the context of neurodegenerative diseases is incompletely understood. HDAC6 deficiency may lead to autophagosome maturation failure and protein aggregate build-up, thereby promoting a neurodegenerative disease process (Kawaguchi et al., 2003). Depletion of HDAC6 enhances loss of dopaminergic neurons, retinal degeneration and locomotor dysfunction caused by ectopic expression of alpha-synuclein (Du et al., 2010). HDAC6 activity also promotes outgrowth and regeneration of neurons (Tapia et al., 2010). On the other hand, HDAC6 activity may also trigger a number of detrimental effects in neurons, the precise reasons for these opposite effects still require further elucidation (Dompierre et al., 2007; Parmigiani et al., 2008). Our current findings reveal for the first time that HDAC6 activity is required in the context of neuroprotection. In particular, our data provide unprecedented evidence that HDAC6 regulates VEGF-mediated neuroprotection against paclitaxel. The extent to which this proposed mechanism is also relevant for other neurotrophic growth factors or other causes of neurotoxicity remains outstanding and a most intriguing question.


Institute for the Promotion of Innovation by Science and Technology (IWT) in Flanders (to J.D. and K.P.); Geneeskundige stichting Koningin Elisabeth (to P.C.); ‘Long term structural Methusalem funding’ by the Flemish Government (to P.C.); Fonds Wetenschappelijk Onderzoek in Flanders (G.0210.07 to P.C.) and governmental Institute for Science and Technology (IWT) for the promotion of research between universities (‘Research & Development’ grant to D.L., partial) and industry (‘Research & Development’ grant to T.M. and R.N., partial), Stichting Tegen Kanker (to D.L).

Supplementary material

Supplementary material is available at Brain online.

vascular endothelial growth factor


View Abstract