Regulation of alternative VEGF-A mRNA splicing is a therapeutic target for analgesia

Abstract

Vascular endothelial growth factor-A (VEGF-A) is best known as a key regulator of the formation of new blood vessels. Neutralization of VEGF-A with anti-VEGF therapy e.g. bevacizumab, can be painful, and this is hypothesized to result from a loss of VEGF-A-mediated neuroprotection. The multiple vegf-a gene products consist of two alternatively spliced families, typified by VEGF-A165a and VEGF-A165b (both contain 165 amino acids), both of which are neuroprotective. Under pathological conditions, such as in inflammation and cancer, the pro-angiogenic VEGF-A165a is upregulated and predominates over the VEGF-A165b isoform. We show here that in rats and mice VEGF-A165a and VEGF-A165b have opposing effects on pain, and that blocking the proximal splicing event - leading to the preferential expression of VEGF-A165b over VEGF165a - prevents pain in vivo. VEGF-A165a sensitizes peripheral nociceptive neurons through actions on VEGFR2 and a TRPV1-dependent mechanism, thus enhancing nociceptive signaling. VEGF-A165b blocks the effect of VEGF-A165a. After nerve injury, the endogenous balance of VEGF-A isoforms switches to greater expression of VEGF-Axxxa compared to VEGF-Axxxb, through an SRPK1-dependent pre-mRNA splicing mechanism. Pharmacological inhibition of SRPK1 after traumatic nerve injury selectively reduced VEGF-Axxxa expression and reversed associated neuropathic pain. Exogenous VEGF-A165b also ameliorated neuropathic pain. We conclude that the relative levels of alternatively spliced VEGF-A isoforms are critical for pain modulation under both normal conditions and in sensory neuropathy. Altering VEGF-Axxxa/VEGF-Axxxb balance by targeting alternative RNA splicing may be a new analgesic strategy.

Keywords: Alternative mRNA splicing; Neuropathy; Nociceptors; Vascular endothelial growth factor A.

Graphical abstract

Fig. 1

VEGF-A gene splice variant isoforms. VEGF-A pre-mRNA is alternatively spliced to form two families of mRNAs: VEGF-A_xxxa and VEGF-A_xxxb. The archetypal forms VEGF-A₁₆₅a and VEGF-A₁₆₅b are shown for illustration. VEGF-A_xxxa proteins are translated from mRNAs that use the proximal splice site (PSS) and include all of exon 8, VEGF-A_xxxb proteins from mRNAs that use the distal splice site (DSS) and contain only the b part of exon 8. The neuropilin-1 (NP-1) co-receptor binding site is located at the distal end of exon 7 and proximal exon 8a.

Fig. 2

VEGF-A isoforms differentially affect pain depending on VEGFR2 activation. A. Intraperitoneal injection of 6 μg/g anti-VEGF-A antibody induced significant mechanical allodynia in mice (n = 5; vehicle n = 6). B. Systemic injection of anti-pan-VEGF-A antibody (6 μg/g) but not vehicle lowered thermal nociceptive withdrawal latency. C. Mechanical allodynia was reproduced by an anti-VEGF-A₁₆₅b antibody (n = 6), shown normalized to the data from  panel A. D. Local blockade of VEGFR2 with 100 nM ZM323881 (specific for VEGFR2) resulted in mechanical allodynia (n = 6/group). E. Systemic injection of PTK787 (30 μg/g) significantly reduced mechanical withdrawal threshold in naïve rats compared to vehicle (saline, n = 6/group). F. rhVEGF-A₁₆₅b (8 ng/g or 20 ng/g) was not painful in normal animals (n = 5/group). Arrowheads denote times of drug administration. G. Neither rhVEGF-A₁₆₅a nor rhVEGF-A₁₆₅b (both 8 ng/g bodyweight) affected thermal hyperalgesia in naïve mice compared to vehicle (saline, n = 5/group). H. rhVEGF-A₁₆₅a (8 ng/g) induced mechanical allodynia. I. rhVEGF-A₁₂₁a administration caused mechanical allodynia whereas rVEGF-A₁₅₉ did not (n = 5/group). J. Comparison of the effects of different VEGF-A isoforms shows that rhVEGF-A_xxxa-evoked allodynia is mediated by the C-terminal 6 amino acids. * = p < 0.05, ** = p < 0.01, *** = p < 0.001 compared with baseline measurements within the same group, ‡ = p < 0.05, ‡‡ = p < 0.01, ‡‡‡ = p < 0.001, between groups, NS = not significantly different. Mean ± SEM for mouse behavior, and median ± IQR for rat behavior.

Fig. 3

Effects of rhVEGF-A isoforms on primary afferent nociceptors. A. VEGFR2 is expressed in nociceptive sensory neurons as determined by double-labeling with the nociceptive markers TrkA (high affinity nerve growth factor receptor) and isolectin B4 (IB4). VEGFR2 expression is upregulated in TrkA + ve nociceptors ipsilateral, and in IB4-binding nociceptors contralateral, to partial saphenous nerve injury (PSNI). B. Photomicrographs of (i) TrkA positive DRG neurons, (ii) VEGFR2 positive neurons and (iii) merged images of (i) and (ii) showing TrkA and VEGFR2 colocalization (scale bar = 50 μm). (iv) Expression of VEGR2 in DRG neurons is much lower in naïve rat DRG compared to (v) animals with PSNI (scale bar = 100 μm). C. Endogenous VEGF-A moderates nociceptor sensitivity, as when VEGFR2 is inhibited by PTK787 mechanical activation threshold of individual nociceptors is reduced within 5 min and over the next 60 min, indicating sensitization. D. Digitized data trace showing the effect of vehicle (saline), VEGF-A₁₆₅a and VEGF-A₁₆₅b on mechanically evoked activity at 5 min, after discharge and ongoing activity in a single afferent nociceptor. rhVEGF-A₁₆₅a sensitized afferents to mechanical stimulation, enhancing after discharge and ongoing activity. Vertical lines are time-compressed action potentials. E. Increased spontaneous ongoing activity was evoked by rhVEGF-A₁₆₅a but not rhVEGF-A₁₆₅b in ~ 50% of mechanonociceptive afferents in rats. (Saline vehicle n = 12, VEGF-A₁₆₅a n = 15, VEGF-A₁₆₅b n = 5). Graphs include data from all neurons, including those in which properties did not change in response to VEGF-A. F. VEGF-A₁₆₅a led to increased ongoing activity in 56% of nociceptive C fibers (OA > 0.1 Hz (Shim et al., 2005)). VEGF-A₁₆₅b did not alter the degree of ongoing activity or number of C fibers that demonstrated ongoing activity, and in addition blocked VEGF-A₁₆₅a-induced ongoing activity. G. rhVEGF-A₁₆₅a reduced primary afferent mechanical threshold 60 min after rhVEGF-A₁₆₅a injection. This was not seen for rhVEGF-A₁₆₅b, and was blocked by its co-administration. H rhVEGF-A₁₆₅a increased primary afferent activity in response to stimulation at suprathreshold force, 5 and 60 min after the injection of rhVEGF-A₁₆₅a, whereas saline and rhVEGF-A₁₆₅b had no effect. * = p < 0.05, ** = p < 0.01, *** = p < 0.001 compared with saline, mean ± SEM.

Fig. 4

Splicing inhibitors that shift the balance of endogenous VEGF-A towards an excess of VEGF-A_xxxb isoforms are anti-nociceptive in normal and nerve injured rats. A. Intraplantar injection of SRPK1 inhibitor SRPIN340 reduced the amount of VEGF-A₁₆₅a mRNA as a proportion of the total VEGF-A mRNA in plantar skin compared to vehicle (saline). B. SRPK inhibition raised mechanical withdrawal thresholds i.e. resulted in hypoalgesia, in mice. C. SRPIN340 did not alter thermal withdrawal latencies. D. VEGF-A_xxxa expression increased as a proportion of total VEGF-A after PSNI. This increase was inhibited by SRPK inhibition. E. Nuclear localization of SRSF1, indicative of SRPK1 activity, is increased in L3/4 DRG neurons following PSNI. F. SRSF1 expression (red) in the cytoplasm of naïve rat DRG sensory neurons (scale bar 50 μm) and SRSF1 expression in the nucleus (stained blue with Hoechst) of rat DRG sensory neurons following PSNI. Note blue staining of nuclei in naïve rats, but purple in PSNI (inset, arrow). G. Phosphorylated (p)Y1175-VEGFR2 (red) staining in naïve and nerve injured mice. H. The number of pY1175-VEGFR2 positive DRG neurons increased after PSNI (*p = 0.019). I. SRPIN340 prevented PSNI-induced mechanical allodynia. J. SRPIN340 reduced SRSF1 activation in DRG containing injured neurons 2 days after nerve injury. ‡, ‡‡‡, p < 0.05, 0.001 respectively compared to baseline; *, *** = p < 0.05, 0.001 respectively compared to other groups.

Fig. 5

Exogenous VEGF-A₁₆₅a exacerbates, and VEGF-A₁₆₅b alleviates neuropathic pain. A. PSNI resulted in ipsilateral mechanical allodynia (NI + Vehicle) compared with sham and baseline. rhVEGF-A₁₆₅b (20 ng/g) was anti-allodynic on days 3 (p < 0.001), 7 (p < 0.01) and 10 (p < 0.0001). Nerve injury on day 0, arrowheads denote drug injection. B. PSNI does not normally result in thermal hyperalgesia (NI + vehicle), but rhVEGF-A₁₆₅a induced hyperalgesia (NI + VEGF-A₁₆₅a) and rhVEGF-A₁₆₅b hypoalgesia. C. rhVEGF-A₁₆₅a (8 ng/g) enhanced ipsilateral mechanical allodynia (filled squares) compared to vehicle (filled circles). D. rhVEGF-A₁₆₅a induced thermal hyperalgesia contralateral to PSNI. rhVEGF-A₁₆₅b again resulted in hypoalgesia. ‡, ‡‡‡, p < 0.05, 0.001 respectively compared to baseline (not shown for mechanical thresholds for clarity as all significant); *, *** p < 0.05, 0.001 respectively compared to vehicle.

Fig. 6

Expression of VEGF-A₁₆₅a and VEGF-A₁₆₅b in rat DRG. A. VEGF-A₁₆₅b represents ~ 70% of total VEGF-A expression in DRG. B. In one human DRG VEGF-A₁₆₅b represented a similar proportion of total VEGF-A expression to that seen in the rat. C. VEGF-A₁₆₅b is expressed in neurons in embryonic human spinal cord and DRG. Higher magnification images are derived from the boxes in the top image and are left: DRG and right: spinal cord ventral horn. D. VEGF-A₁₆₅b is expressed in a proportion of rat DRG neurons (Ai, iii, v), with overlap (arrows) with the nociceptive markers TrkA (Aii, iv, vi) and a small colocalization with IB4 (Aii, iv, vi). Scale bar = 75 μm. High power images of a single neuron showing colocalization of VEGF-A₁₆₅b (green) and TrkA (red). Scale bar = 50 μm.

Fig. 7

VEGF-A isoforms alter nociception in a TRPv1 dependent manner. A. Systemic TRPV1 antagonism with SB366791 in mice resulted in inhibition of rhVEGF-A₁₆₅a-induced mechanical allodynia. Arrows denote time of drug administration. B. TRPV1 knockout mice did not develop rhVEGF-A₁₆₅a-induced mechanical allodynia, in contrast to wild-type strain matched controls. C. TRPV1 was co-expressed with VEGFR2 in sensory dorsal root ganglia sensory neurons (scale bar = 20 μm). D. Local administration of VEGF₁₆₅a + vehicle into the plantar hindpaw resulted in a reduction in mechanical withdrawal values, which was blocked by co-administration of the TRPV1 antagonist SB366791 (TRPV1 antagonist).

Fig. 8

VEGF-A modulated TRPV1-agonist evoked responses in dorsal root ganglion neurons. A. Capsaicin stimulated a concentration-dependent increase in intracellular calcium in DRG neurons. B. This was increased by rhVEGF-A₁₆₅a, and reduced by rhVEGF-A₁₆₅b (mean ± SEM, n = 3–7). C. Treatment of rat DRG neurons with rhVEGF-A₁₆₅a increased capsaicin-stimulated calcium influx (area under the curve of the calcium responses shown in Fig. 8B) compared with capsaicin alone or rhVEGF-A₁₆₅b (2 way ANOVA main effect of drug p = 0.0051). The bell shaped concentration–response curve displays TRPV1 desensitization at higher capsaicin concentrations (5 μM). D. Example of a digitized trace of raw capsaicin-evoked current in the presence (gray) and absence of capsaicin. E. Capsaicin-evoked currents in primary DRG neurons were significantly larger in neurons incubated in VEGF-A₁₆₅a overnight compared to vehicle treated neurons (box and whisker plots showing median, range, min and max). F. rhVEGF-A₁₆₅b treatment enhanced TRPV1 serine phosphorylation in 50B11 immortalized DRG cells. IP of protein with TRPV1 antibody followed by IB with anti-pSer antibody showed rhVEGF-A₁₆₅a, but not rhVEGF-A₁₆₅b-mediated phosphorylation of TRPV1. (NGF treatment = positive control). G. Whereas 0.2 μM capsaicin alone did not alter intracellular calcium itself, overnight treatment with rhVEGF-A₁₆₅a + 0.2 μM capsaicin resulted in a robust sustained increase in response to capsaicin, which was blocked by treatment with the PKC inhibitor BIM1 (2 way ANOVA main effect of drug p = 0.0003). H. Low concentration capsaicin (concentration at terminals ~ 10 nM) led to evoked activity from C fiber nociceptors in vivo. Capsaicin-evoked activity was increased by rhVEGF-A₁₆₅a and blocked by rhVEGF-A₁₆₅b. ‡, ‡‡, ‡‡‡, p < 0.05, 0.01, 0.001 respectively compared to baseline. *, **, *** = p < 0.05, 0.01, 0.001 respectively compared to other groups.

Fig. 9

Downstream targets of the serine–arginine protein kinase SRPK1. The serine–arginine protein kinase is known to have three major downstream targets, the RNA splicing factors SRSF1 (Edmond et al., 2011), SRSF2 (Aubol and Adams, 2011, Ngo et al., 2005, Velazquez-Dones et al., 2005), and the lamin B receptor (Papoutsopoulou et al., 1999). SRPK1 activity results in Hsp90-dependent nuclear translocation of SRSF1 (Zhou et al., 2012). SRSF1 has been reported to control alternative RNA splicing of the proto-oncogene myc, BIM (BCL2L11) (Anczukow et al., 2012), the cation cotransporter SLC39A14 (Thorsen et al., 2011), the tumor suppressors MKNK2 and BIN1 (Das et al., 2012, Karni et al., 2007), the angiogenesis related genes RON (Ghigna et al., 2005) and TEAD1 (Das et al., 2012), and VEGF-A (Amin et al., 2011, Nowak et al., 2010, Nowak et al., 2008). TEAD1 activates VEGF-A expression (Teng et al., 2010). None of the downstream targets of SRPK1 has been implicated in nociception other than VEGF-A.