|Home | About | Journals | Submit | Contact Us | Français|
Itch can be suppressed by painful stimuli, but the underlying neural basis is unknown. We generated conditional null mice in which VGLUT2-dependent synaptic glutamate release from mainly Nav1.8-expressing nociceptors was abolished. These mice showed deficits in pain behaviors including mechanical pain, heat pain, capsaicin-evoked pain, inflammatory pain and neuropathic pain. The pain deficits were accompanied by greatly enhanced itching, as suggested by i) sensitization of both histamine-dependent and histamine-independent itch pathways, and ii) development of spontaneous scratching and skin lesions. Strikingly, intradermal capsaicin injection promotes itch responses in these mutant mice, as opposed to pain responses in control littermates. Consequently, co-injection of capsaicin was no longer able to mask itch evoked by pruritogenic compounds. Our studies suggest that synaptic glutamate release from a group of peripheral nociceptors is required to sense pain and suppress itch. Elimination of VGLUT2 in these nociceptors creates a mouse model of chronic neurogenic itch.
Itch and pain represent two distinct sensations. Moreover, it has been long recognized that there is an antagonistic relationship between pain and itch (Ikoma et al., 2006; Schmelz, 2010). Over 70 years ago, Lewis et al. first reported that itch sensation evoked by histamine injection in humans can be blocked by electrical stimuli (Lewis et al., 1927/1929). Other painful stimuli, such as noxious heat and noxious chemicals (mustard oil or capsaicin), can also suppress itch (Brull et al., 1999; Graham et al., 1951; Ward et al., 1996). Electrophysiological studies show that the firing of spinal itch relay neurons can be suppressed by inputs of pain processing neurons (Andrew and Craig, 2001; Davidson et al., 2009). Conversely, a blockage of pain can induce or enhance itch. For example, pain inhibition by anesthetic compounds can enhance itch evoked by histamine (Atanassoff et al., 1999), and intrathecal injection of opioid analgesics is often associated with itch side effects (Ikoma et al., 2006; Schmelz, 2010).
Several theories have been proposed to explain itch suppression by pain. The “population coding” hypothesis, also called selectivity hypothesis proposes that the senses of itch and pain are processed along specific neural circuits or labeled lines, but the activation of pain-sensing fibers can dominantly mask itch, even if the stimuli activate both pain-sensing and itch-sensing fibers (Handwerker, 2010; McMahon and Koltzenburg, 1992; Schmelz, 2010; Wood et al., 2009). The existence of itch-specific neurons was supported initially by electrophysiological studies in humans and cats (Andrew and Craig, 2001; Schmelz et al., 1997), and subsequently by genetic and cell ablation studies in mice (Liu et al., 2009; Sun and Chen, 2007; Sun et al., 2009). For example, sensory neurons expressing the G-protein coupled receptor Mrgpra3, which represent 4–5% of neurons in dorsal root ganglia (DRG) (Liu et al., 2008), are necessary for itch evoked by chloroquine, but dispensable for pain (Liu et al., 2009). Spinal neurons expressing the Gastrin-releasing peptide receptor (GRPR) are also dedicated to itch (Sun and Chen, 2007; Sun et al., 2009). The “spatial contrast” theory, however, proposes that pain and itch can be encoded without having itch-specific and pain-specific neurons: itch is evoked when a minority of nociceptive fibers are activated in a receptive field, whereas pain is evoked when a majority of fibers are activated (Johanek et al., 2008; LaMotte et al., 2009; Namer et al., 2008; Schmelz, 2010; Sikand et al., 2009). However, this view seems to conflict with the actual existence of itch-specific circuits, as mentioned above.
Thus, the coding of pain versus itch may be best explained by the population-coding hypothesis that highlights both the existence of itch-specific and pain-specific sensory neurons, as well as a dominant suppression of itch by pain. However, while important progress has been made in identifying itch-specific sensory neurons (Andrew and Craig, 2001; Liu et al., 2009; Schmelz et al., 1997; Sun and Chen, 2007; Sun et al., 2009), the neural basis underlying itch suppression by painful stimuli has not yet been characterized.
Vesicular glutamate transporter type 2 (VGLUT2) and the related proteins, VGLUT1 and VGLUT3, belong to a family of transporters that package glutamate into synaptic vesicles and are necessary for most fast excitatory synaptic transmission in the vertebrate nervous system (Fremeau et al., 2004). These three proteins are expressed in a partially overlapping manner in peripheral sensory neurons in DRG (Brumovsky et al., 2007; Seal et al., 2009). By generating and analyzing Vglut2 conditional knockout mice, here we found that VGLUT2-dependent synaptic glutamate release from mainly Nav1.8-expressing nociceptors represents a neuronal component that is critical for pain sensation and itch suppression. Removal of VGLUT2 in these nociceptors leads to i) marked pain deficits, ii) sensitization of multiple itch pathways, and iii) spontaneous development of excessive scratching and skin lesions. Moreover, capsaicin is able to activate a normally “hidden” itch pathway in these mutant mice, and fails to suppress itch evoked by pruritogenic compounds. These studies provide new insight into the neural basis underlying an antagonistic interaction between pain and itch.
In this study, we made Vglut2 conditional knockout mice by using the Nav1.8Cre transgenic mice made by the Kuner group (Agarwal et al., 2004). It had been reported that in Nav1.8Cre transgenic mice, Cre activity is detected in DRG, but not in the central nervous system (Agarwal et al., 2004). To further determine Cre specificity, we crossed Nav1.8Cre mice with a Cre-dependent TauGFP reporter mice (Hippenmeyer et al., 2005), in which Nav1.8Cre-active neurons were marked by the expression of green fluorescent protein (GFP) and can be detected by GFP immunostaining (Figure S1). A double staining of GFP with the pan-neural marker SCG10 (Stein et al., 1988) showed that 81.0% (884/1091) of lumbar DRG neurons expressed GFP. Additional double staining showed that GFP was expressed in all Nav1.8-expressing neurons, but surprisingly, also in a small subset of Nav1.8-negative DRG neurons (Figure S1), which is different from another Nav1.8Cre line made by the Wood group that drives reporter expression only in Nav1.8-expressing neurons (Stirling et al., 2005). We had also crossed Nav1.8Cre with ROSARFP reporter mice (Madisen et al., 2010), with the resulting heterozygous mice referred to as ROSARFP;Nav1.8Cre, in which Nav1.8Cre-active neurons can be directly visualized with the expression of red fluorescent protein or RFP, without the involvement of immunostaining (Figure S2). As described below, ROSARFP;Nav1.8Cre mice were used to examine the expression of other markers in Nav1.8Cre-active DRG neurons.
To determine the distribution of VGLUT2 expression in DRG, we performed VGLUT2 immunostaining in lumbar DRG of ROSARFP;Nav1.8Cre mice. We found that all RFP-expressing, thereby Nav1.8Cre-active, neurons coexpressed VGLUT2, albeit at heterogeneous expression levels (Figure 1A). 88% and 12% of VGLUT2-expressing neurons are RFP positive (Nav1.8Cre-active) and RFP negative (Nav1.8Cre-negative), respectively (Figure 1A).
To make a Vglut2 conditional knockout in the DRG, we crossed mice carrying a conditional null allele of Vglut2 (Vglut2F) with Nav1.8Cre transgenic mice (Agarwal et al., 2004; Tong et al., 2007), with the resulting homozygous conditional knockout mice (Vglut2F/F;Nav1.8Cre) referred to as CKO mice (Figure S3A). The Vglut2F/F littermates were referred to as control mice. Consistent with the observation that 88% of VGLUT2-expressing neurons are Nav1.8Cre active, VGLUT2 expression was greatly reduced in CKO mice (Figure S3B). Double staining showed that in control mice, VGLUT2 expression was expressed in all non-peptidergic nociceptors marked by the binding of the isolectin B4 (IB4; Figure 1B) and peptidergic nociceptors marked by the expression of calcitonin-gene related peptide (CGRP; Figure 1C), albeit at various expression levels; in CKO mice, VGLUT2 expression in these two groups of neurons was reduced to 0.88% and 12.3%, respectively (Figure 1). VGLUT2 expression was detected in 94.2% of TRPV1-expressing neurons in control mice, but reduced to 3.88% in CKO mice (Figure 1D). Thus, in Vglut2 CKO mice, VGLUT2 expression is eliminated in a majority of classical nociceptors.
To determine to what degree the expression of VGLUT1 or VGLUT3 can compensate the loss of VGLUT2, we examined VGLUT1 and VGLUT3 expression in wild type and CKO mice. Using ROSARFP;Nav1.8Cre fate-mapping mice, we found that 15.7% of RFP-positive (Nav1.8Cre-active) neurons in lumbar DRG at postnatal day 30 (P30) coexpressed VGLUT1, and these neurons represent 38.0% of VGLUT1-expressing neurons (Figure 2A). Additional double staining showed that VGLUT1 expression was detected in 25.5% of CGRP-expressing neurons, 2.66% of IB4-positive neurons, and virtually none of TRPV1-expressing neurons (Figure 2B). Importantly, the numbers of VGLUT1-expressing neurons in P30 lumbar DRG were unchanged between control and Vglut2 CKO mice (Figure 2C). By using VGLUT3-GFP reporter mice, it was shown that VGLUT3 was expressed in ~10% of lumbar DRG neurons (Seal et al., 2009). However, VGLUT3 was expressed at levels too low to be detected by our non-radioactive in situ hybridization. We then measured VGLUT3 expression levels in lumbar DRG by quantitative real-time RT-PCR, and found that there was no significant change in CKO mice (data not shown). Thus, there is no compensatory increase in VGLUT1 or VGLUT3 expression in Vglut2 CKO mice.
Based on these expression analyses, we concluded that a majority of Nav1.8Cre-active neurons express VGLUT2, but not VGLUT1 or VGLUT3; glutamate release from these neurons should be eliminated in Vglut2 CKO mice. Synaptic glutamate release in the remaining Nav1.8Cre-active neurons was expected to be attenuated (due to a loss of VGLUT2), but will not be fully eliminated (due to VGLUT1 and/or VGLUT3 expression). Finally, synaptic glutamate release from Nav1.8Cre-negative neurons, representing 19.0% of total DRG neurons, will be unaffected (summarized in Figure 2D).
The loss of VGLUT2 in a large subset of DRG neurons was expected to cause reduced excitatory glutamatergic transmission from DRG neurons onto the dorsal spinal cord. To test this hypothesis, spontaneous excitatory postsynaptic currents (sEPSC) were recorded from neurons in lamina II ex vivo in isolated spinal cord slices from control and CKO mice (Figure 3A). We indeed found that the frequency of sEPSC was reduced by 47% in CKO mice (4.22 ± 0.63 per second, n=11) in comparison with that in control mice (7.86 ±1.16 per second, n=9, p<0.05) (Figure 3B). Since VGLUT2 was eliminated in the DRG but not in the central nervous system (CNS), the reduction of sEPSC in mutant dorsal horn neurons should be caused exclusively by an impairment of excitatory glutamatergic transmission by VGLUT2-dependent primary sensory afferents, irrespective of the direct or indirect connections between recorded dorsal horn neurons and primary sensory fibers. The synaptic transmission defect was further supported by a change of c-Fos induction in specific groups of spinal neurons following intraplantar injection of capsaicin (see below).
We next asked if the attenuation of excitatory synaptic transmission led to molecular and anatomical changes in DRG and the dorsal spinal cord. The number of total DRG neurons, determined by the expression of the pan-neuronal marker SCG10, and the numbers of IB4-positive and CGRP-positive neurons were unchanged in lumbar DRG between CKO and control mice (Figure S3), suggesting that the development and survival of DRG neurons are unaffected. Furthermore, calcium-imaging studies showed that responses to capsaicin and histamine by DRG neurons were unchanged (Figure S4). Lamina-specific projections of IB4-positive, CGRP-positive, and VGLUT1-positive sensory fibers in the dorsal horn of the spinal cord were also unaffected (Figure 3C). Furthermore, molecular identities of dorsal horn neurons were also unaffected (Figure 3D). For example, expression of somatostatin, which is enriched in excitatory neurons in the superficial dorsal horn, and the expression of dynorphin or enkephalin, which is enriched in inhibitory neurons, was grossly unaffected in CKO mice in comparison with control mice (Figure 3D) (Bröhl et al., 2008; Huang et al., 2008; Xu et al., 2008). Thus, the molecular and anatomical features of DRG and dorsal spinal cord neurons are grossly unaffected in Vglut2 CKO mice.
We next assessed how pain behaviors were affected in CKO mice, using Vglut2F/F littermates as a control. In comparison with control mice, CKO mice showed similar paw withdrawal latencies on a 50°C hot plate, but increased latency at 52°C and 55°C (Figure 4A). We also observed a small but significant increase in withdrawal latencies in the Hargreaves radiant heat test (Figure 4B). Increased heat pain deficit at higher temperatures is analogous to the mutant phenotype seen in TRPV1 null mice (Caterina et al., 2000; Davis et al., 2000). Consistently, acute pain induced by capsaicin injection, which is TRPV1-dependent (Caterina et al., 2000; Davis et al., 2000), was again markedly impaired in CKO mice (see below).
The withdrawal threshold in response to von Frey filaments was unchanged in CKO mice (Figure 4C), implying that pain evoked by light noxious mechanical stimuli remains intact. In contrast, responses to intense noxious mechanical stimuli delivered by the Randall-Selitto apparatus were markedly attenuated (Figure 4D).
We next examined two types of chronic pain: inflammatory and neuropathic. Inflammatory pain was induced by intraplantar injection of Complete Freund’s Adjuvant (CFA). CFA-induced mechanical hypersensitivity, measured by reduced mechanical threshold in eliciting painful withdrawal responses, was significantly attenuated in CKO mice (Figure 4E). CFA-induced heat hypersensitivity, measured by reduced latency in response to radiant heat, was largely abolished in CKO mice (Figure 4F). To assess neuropathic pain, we used the spared nerve injury (SNI) model (Decosterd and Woolf, 2000). In control mice, SNI caused a profound mechanical hypersensitivity, as indicated by a marked reduction in withdrawal threshold in response to mechanical stimuli (Figure 4G). Such SNI-induced mechanical hypersensitivity was largely abolished in CKO mice (Figure 4G).
Thus, a loss of VGLUT2-dependent synaptic glutamate release from Nav1.8Cre-active neurons results in deficits in a range of acute and chronic pain, including intense mechanical pain, intense heat pain, capsaicin-induced spontaneous pain (see below), CFA-induced inflammatory pain, and SNI-induced neuropathic pain.
The most noticeable mutant phenotype is that by the time Vglut2 CKO mice reached two months old, a vast majority of them developed skin lesions (Figure 5A). The lesions were most frequently observed around the neck and ears, but also in other parts of body. In contrast, none of the control mice showed such lesions (data not shown). We postulated that the skin lesions were caused by itch-induced scratches. Therefore, we monitored spontaneous scratching at the time when CKO mice showed the earliest sign of hair loss. We found that CKO mice showed 6-fold more spontaneous scratch bouts than control littermates (Figure 5B). In other words, excessive scratching preceded the development of overt skin lesions.
To determine if itching pathways were sensitized in Vglut2 CKO mice, we injected low dosages of pruritogenic agents and monitored scratching responses. The experiments were done on one-month old mice, when most CKO mice had not yet developed excessive spontaneous scratching behavior (Figure 5C–5H). Compound 48/80 activates a histamine-dependent itching pathway (Sugimoto et al., 1998). Nape injection of 2 µg of Compound 48/80, as opposed to 100 µg used in other studies (Sun et al., 2009), induced two-fold increase in scratching responses in CKO mice than in control mice (Figure 5C). Injection of 10 µg of Compound 48/80 also induced more scratching in CKO mice (Figure 5D). The protease-activated receptor 2 (PAR2) is a G-protein coupled receptor and activates a histamine-independent itch pathway (Shimada et al., 2006; Tsujii et al., 2008). Injection of a low dosage [20 µg, as opposed to 100 µg (Sun et al., 2009)] of the PAR2 agonist SLIGRL-NH2 induced modest scratching responses in control mice, but induced three-fold more scratch bouts in CKO mice (Figure 5E). A serotonin derivative, α-Me-5-HT, evokes pure itching responses in mice (Imamachi et al., 2009). Injection of 30 µg of α-Me-5-HT also induced more scratching responses in CKO mice (841 bouts) than in control mice (377 bouts) (Figure 5F). There was an exception. Chloroquine, a compound used to treat malaria in humans, evoked histamine-independent itch through activating the G-protein coupled receptor Mrgpra3 (Liu et al., 2009). We found that chloroquine evoked similar degrees of scratching responses in Vglut2 CKO mice versus control littermates at both low and high dosages (Figure 6G and 6H). These observations suggested that the loss of VGLUT2 in Nav1.8Cre-active neurons in CKO mice results in sensitization of multiple, but not all, itch pathways.
Capsaicin activates the TRPV1 transient receptor potential ion channel, and capsaicin-responsive neurons are essential for the senses of both pain and itch (Basbaum et al., 2009; Imamachi et al., 2009; Liu et al., 2009; Lynn, 1992; Shim and Oh, 2008; Simone et al., 1989). To determine how capsaicin-evoked pain or itch was affected in CKO mice, we used a recently developed behavioral assay that clearly distinguishes pain versus itch: intradermal capsaicin injection in the cheek induces pain-indicative wiping by the forelimb, whereas injection of itching compounds in the cheek induces scratching by the hind limb (Shimada and LaMotte, 2008). Again, one-month old control and CKO mice were used, when they showed no significant difference in baseline scratching or wiping (Figure 6). Following injection of 20 µg of capsaicin, CKO mice exhibited a 56% reduction in the number of wipes compared to control mice, suggesting an attenuation of capsaicin-evoked pain (Figure 6A and 6B). More strikingly, capsaicin induced robust scratching responses in CKO mice, as opposed to minimal responses in control mice (Figure 6C and 6D). Injection of capsaicin at a lower dosage (5 µg) also induced more scratching bouts in CKO mice than control littermates (Figure 6E), although this low dosage evoked less scratching bouts in comparison with the higher dosage of capsaicin (Figure 6E versus 6C). This dramatic behavioral switch suggested that VGLUT2-dependent synaptic glutamate release is necessary for capsaicin-evoked pain, and its removal allows capsaicin to activate a normally “masked” itch pathway.
One interpretation of this behavioral switch is that capsaicin-evoked pain can dominantly inhibit itch evoked by capsaicin-sensitive pruriceptors, and this inhibition is attenuated in CKO mice. To further explore this possibility, we asked if capsaicin can suppress itch evoked by α-Me-5-HT, a potent pruritogenic compound (Figure 5F) that acts through capsaicin-sensitive neurons (Imamachi et al., 2009). In control mice, cheek injection of α-Me-5-HT evoked robust scratching responses (Figure 6F); strikingly, a co-injection of α-Me-5-HT and capsaicin almost completely inhibited scratching response. In contrast, a co-injection of α-Me-5-HT and capsaicin in CKO mice still showed significant amount of scratching response (Figure 6F), with the amount of scratch bouts comparable to that evoked by capsaicin alone in CKO mice (Figure 6C). These studies suggested that a strong pain-inducing stimulus such as capsaicin can dominantly mask a strong itch signal such as α-Me-5-HT, and this pain-induced itch inhibition is markedly attenuated in CKO mice.
The enhanced or normal itching responses seen in Vglut2 CKO mice suggested that VGLUT2 expression in Nav1.8Cre-active sensory neurons is dispensable for itch. To further look into this issue, we examined how VGLUT2 expression was affected in established or putative pruriceptors. Mrgpra3-expressing pruriceptors mediate itch evoked by chloroquine (Liu et al., 2009). We found that these pruriceptors expressed VGLUT2 (Figure 7A), but not VGLUT1 (Figure 7B) in control mice, and VGLUT2 expression in these neurons was eliminated in Vglut2 CKO mice (Figure 7A). Sensory neurons expressing the gastrin-releasing peptide (GRP) are putative pruriceptors (Sun and Chen, 2007). In control mice, GRP was expressed at high levels in a subset of small diameter neurons (Figure 7C, arrows), as reported previously (Sun and Chen, 2007). Low levels of GRP expression were detected in many small and large DRG neurons (Figure 7C, arrowheads), consistent with a more recent report (Liu et al., 2009). In GRPhigh neurons, VGLUT2 expression was detected in control mice, but was eliminated in Vglut2 CKO mice (Figure 7C, arrows). Again, no VGLUT1 expression was detected in GRPhigh neurons (Figure 7D). With enhanced or unaffected itching responses observed in Vglut2 CKO mice, these data suggest that under the Vglut2 CKO genetic background, glutamate release from Mrgpra3-expressing and GRPhigh neurons appears to be dispensable for itch.
Spinal inhibitory neurons have recently been suggested to play a role in preventing itch sensitization (Ross et al., 2010). It has also been known for a while that noxious stimuli, such as intraplantar capsaicin injection, were able to activate spinal inhibitory neurons (Binshtok et al., 2007; Zou et al., 2002). We therefore asked if such activation was affected in Vglut2 CKO mice. We used the induction of c-Fos to identify capsaicin-responsive spinal neurons (Gao and Ji, 2009; Hunt et al., 1987; Zou et al., 2002). Previous studies showed that spinal inhibitory neurons can be identified by the expression of a set of neuropeptides, including neuropeptide Y (NPY) and enkephalin (ENK) (Bröhl et al., 2008; Huang et al., 2008; Xu et al., 2008). We found that, following capsaicin injection in hindpaw, c-Fos expression was detected in 40.1% of NPY-expressing neurons at the ipsilateral side of control mice, but reduced to 11.9% in Vglut2 CKO mice (Figure 8A), a 70.3% reduction. The reduction of c-Fos-positive neurons was not due to a loss of NPY-expressing neurons per se, as suggested by the normal numbers of NPY-expressing neurons in CKO versus control mice (data not shown). Interestingly, c-Fos induction in ENK-expressing neurons was unaffected in CKO mice (Figure 8B). These data suggest that VGLUT2-depedent synaptic glutamate release from Nav1.8Cre-active neurons is required for capsaicin to activate a specific (NPY-expressing) subset of spinal inhibitory neurons.
GRPR-expressing spinal neurons are required to process itching information (Sun et al., 2009). With the finding that capsaicin evokes itch, rather than pain in CKO mice, we next asked if capsaicin injection could differentially activate GRPR-expressing neurons in control versus CKO mice. We found that following intraplantar capsaicin injection, there was indeed a significant, albeit modest, increase of c-Fos induction in GRPR-expressing neurons, from 18.5 ± 0.4% in control mice to 27.2 ±1.5% in CKO mice (p<0.05), a 47% increase (Figure 8C). All together, these studies show that upon removal of VGLUT2 in Nav1.8Cre-active neurons, capsaicin on one hand fails to activate NPY-expressing spinal inhibitory neurons, but on the other hand causes an increase of activation of GPRP-expressing spinal itch relay neurons.
In this study we have generated Vglut2 CKO mice in which Vglut2 was removed from Nav1.8Cre-active DRG neurons. The loss of VGLUT2 in these neurons leads to impaired glutamatergic transmission, as indicated by reduced sEPSC frequency in postsynaptic dorsal horn neurons, although additional electrophysiological recording is needed to determine exact deficits in synaptic transmission in these mice. Behavioral studies show that VGLUT2-dependent glutamate release from Nav1.8Cre-active DRG neurons is required to sense a range of acute and chronic pain, including intense mechanical pain, capsaicin-evoked pain, intense heat pain, inflammatory and neuropathic pain. It has been previously reported that ablation of Nav1.8-expressing neurons causes similar deficits in mechanical and inflammatory pain, but surprisingly, without affecting neuropathic pain or heat pain (Abrahamsen et al., 2008). How could we explain the deficits in neuropathic and heat pain following the loss of VGLUT2 in Nav1.8Cre-active neurons? This discrepancy could be caused by the use of two different Nav1.8Cre mice. The Nav1.8Cre mice used for cell ablation were made by the knock-in strategy (Abrahamsen et al., 2008), and drove reporter expression that matches Nav1.8 expression in DRG (Stirling et al., 2005). However, the Nav1.8Cre mice used in this study were made through a transgenic approach (Agarwal et al., 2004), and drove reporter expression not only in all Nav1.8-expressing neurons, but also in a small subset of Nav1.8-negative neurons (Figure S1). It is possible that the transgenic approach might drive Cre expression at higher levels than the knock-in approach, such that neurons normally expressing low levels of Nav1.8 might contain sufficient Cre activity. The loss of VGLUT2 expression in these extra “Nav1.8-negative” neurons in Vglut2 CKO mice might then contribute to the impairment of neuropathic and heat pain. Alternatively, compensatory mechanisms may develop following ablation of Nav1.8-expressing neurons, but not after removal of VGLUT2.
Neuropathic pain is also impaired in conventional Vglut2 heterozygous null mice (Leo et al., 2009; Moechars et al., 2006). Because VGLUT2 is broadly expressed in the CNS, including the dorsal horn of the spinal cord, the underlying cellular basis of neuropathic pain deficit in those heterozygous null mice cannot be certain. By analyzing Vglut2 CKO mice, we have now clearly shown that VGLUT2-dependent synaptic glutamate release from DRG neurons is essential for neuropathic pain; this association is also consistent with the report that VGLUT2 expression is elevated in a subset of nociceptors following nerve injury (Brumovsky et al., 2007).
VGLUT3-expressing DRG neurons represent low threshold mechanoreceptors (Seal et al., 2009). Interestingly, mice lacking Vglut3 also show a partial defect in sensing intense mechanical pain and a marked deficit in developing persistent mechanical hypersensitivity following inflammation or nerve injury (Seal et al., 2009), suggesting that glutamate release from both VGLUT3-expressing and VGLUT2-expressing neurons is involved in sensing mechanical pain. Meanwhile, the expression patterns of VGLUT1–3 in DRG neurons could explain why inflammatory heat hypersensitivity, which is known to be dependent on TRPV1 (Caterina et al., 2000; Davis et al., 2000), is abolished in Vglut2 CKO mice, but not in Vglut3 mutant mice (Seal et al., 2009). Most TRPV1-expressing neurons show detectable expression of VGLUT2 (Figure 1), but not VGLUT1 (Figure 2) or VGLUT3 (Seal et al., 2009). In Vglut2 CKO mice, only 4% of TRPV1-expressing neurons retain VGLUT2 expression. In other words, excitatory glutamatergic synaptic transmission is eliminated from most TRPV1-expressing neurons in Vglut2 CKO mice, which may render them functionally silent, thereby explaining the impaired inflammatory heat hyperalgesia in these mice.
Our studies provide new insight into the coding of pain versus itch, by showing that VGLUT2-dependent glutamate release from Nav1.8Cre-active peripheral nociceptors represents a neuronal component that is required to sense pain and suppress itch. Removal of this component leads to i) marked pain deficits, ii) sensitization of both histamine-dependent and histamine-independent itch pathways, iii) spontaneous development of excessive scratching and eventual skin lesions, iv) a failure of capsaicin to dominantly mask a strong itch signal, and v) direct paradoxical promotion of itch by capsaicin. Our studies also suggest that neurons that retain glutamate release in Vglut2 CKO mice, including Nav1.8Cre-negative neurons plus Nav1.8Cre-active neurons that express VGLUT1 and/or VGLUT3, are insufficient to prevent itch sensitization, even though they are sufficient to mediate light mechanical pain (measured by von Frey assay) and light heat pain (measured by hot plate at 50°C).
Our studies raise an intriguing question regarding the role of synaptic glutamate release from peripheral pruriceptors in processing itching information. In Vglut2 CKO mice, VGLUT2 was eliminated in 81% of DRG neurons. The loss was not just confined to pain-sensing neurons, but also occurred in Mrgpra3-expressing pruriceptors and in putative pruriceptors marked by GRPhigh expression (Figure 7) (Liu et al., 2009; Sun and Chen, 2007; Sun et al., 2009). Moreover, Mrgpra3-expressing and GRPhigh neurons do not express VGLUT1 (Figure 7), and may also not express VGLUT3 since most of these neurons coexpressed CGRP (Sun and Chen, 2007) (data not shown), and CGRP-expressing neurons do not express VGLUT3 (Seal et al., 2009). With the loss of VGLUT2 and a lack of VGLUT1/3 expression in these established and putative pruriceptors, it is surprising to observe that itch evoked by a range of pruritogenic compounds is either enhanced or unaffected in Vglut2 CKO mice, including chloroquine-evoked itch that is mediated by Mrgpra3-expressing pruriceptors. We envision the following two possibilities. First, glutamate release from pruriceptors might be dispensable for itch. In other words, pruriceptors may use other transmitters such as GRP to mediate itch. Consistent with this, intrathecal injection of GRP is sufficient to evoke itching response and histamine-independent itch is markedly impaired in GRPR mutant mice (Sun and Chen, 2007). Second, glutamate release from pruriceptors may normally act to remove tonic itch inhibition by pain-processing neurons (Andrew and Craig, 2001). In Vglut2 CKO mice, the loss of pain may have already removed such inhibition; as a result, glutamate release from pruriceptors becomes dispensable for the transmission of itching information.
How do the pain loss and itch sensitization observed in Vglut2 CKO mice fit into current itch theories? The “spatial contrast” hypothesis (Johanek et al., 2008; Namer et al., 2008; Schmelz, 2010) proposes that itch is coded when a small subset of nociceptor fibers is activated in a receptive field, whereas pain is encoded when more nociceptor fibers are activated. The loss of VGLUT2 in 81% of DRG neurons in Vglut2 CKO mice will certainly lead to a great reduction of the density of functional nociceptor fibers in the skin, and according to this theory, such reduction should result in enhanced itching. However, the argument by the spatial contrast theory that pain and itch can be coded without having pain-specific and itch-specific fibers conflicts with actual existence of itch-specific sensory neurons, such as Mrgpra3-expressing DRG neurons and GRPR-expressing spinal neurons (Andrew and Craig, 2001; Liu et al., 2009; Schmelz et al., 1997; Sun and Chen, 2007; Sun et al., 2009). Moreover, a reduction in density of functional nociceptor fibers in the skin is not automatically linked with itch sensitization. For example, ablation of Mrgprd-positive polymodal nociceptors that densely innervate the skin epidermis has no impact on itch (Imamachi et al., 2009; Rau et al., 2009; Zylka et al., 2005). Consistently, in Vglut2 CKO mice, a lower dosage of capsaicin, which is supposed to activate fewer capsaicin-sensitive fibers and should enhance itch according to the spatial contrast hypothesis, actually causes reduced scratching responses in comparison with a higher dosage of capsaicin (Figure 6).
Thus, with increasing evidence arguing against the spatial contrast theory, several investigators suggested that the coding of pain versus itch may be best explained by the population-coding hypothesis, which emphasizes both the existence of itch-specific and pain-specific neural components, and dominant suppression of itch by pain (Handwerker, 2010; McMahon and Koltzenburg, 1992; Wood et al., 2009). According to this hypothesis, VGLUT2-dependent glutamate release from Nav1.8Cre-active DRG neurons should represent a neural component that is necessary for pain sensation and itch suppression. For example, in the absence of this component in Vglut2 CKO mice, a co-injection of a strong pain-inducing compound (capsaicin) is no longer able to mask itch evoked by a strong pruritogenic compound (Figure 6). Moreover, because many itch-sensing neurons respond to capsaicin (Imamachi et al., 2009; Liu et al., 2009; Lynn, 1992; Shim and Oh, 2008), the loss of pain-induced inhibition of itch may explain why capsaicin is able to activate a normally “hidden” itch pathway in these mutant mice (Figure 6). The enhanced itching responses observed in Vglut2 CKO mice do raise the following question: why do itching compounds fail to evoke maximum amount of itch in wild type mice? It should be noted that most itching compounds (such as histamine) in fact activate both itch-sensing and pain-sensing neurons (Atanassoff et al., 1999), and the amount of itch evoked by a pruritogenic compound is normally attenuated by this pain component. For example, anesthetic blockage of pain can greatly enhance histamine-evoked itch in humans (Atanassoff et al., 1999). In Vglut2 CKO mice, the loss of pain (and itch inhibition by pain) may mimic anesthetic treatment in humans, thereby allowing itching compounds to evoke enhanced responses.
How could painful stimuli suppress itch? It was proposed that painful stimuli may activate inhibitory neurons in the dorsal spinal cord, which in turn inhibits itch-processing neurons (Andrew and Craig, 2001; Davidson et al., 2009; Handwerker, 2010). This hypothesis is strongly supported by a recent study showing that Bhlhb5-dependent spinal inhibitory neurons are involved in itch suppression, whose developmental impairment leads to sensitization of multiple itch pathways (Ross et al., 2010). Intriguingly, in Vglut2 CKO mice, capsaicin injection fails to activate NPY-expressing inhibitory neurons in the spinal cord; moreover, there is a modest increase of activation of GRPR-expressing itch relay neurons. Future studies will be warranted to determine if capsaicin-responsive pain fibers connect with NPY-expressing inhibitory neurons to suppress itch, and if the development of these inhibitory neurons is dependent on Bhlhb5.
Our studies provide new insight into the coding of pain versus itch. First, VGLUT2-dependent glutamate release from Nav1.8Cre-active neurons is required to sense pain and suppress itch. The itch sensitization observed in Vglut2 CKO mice is likely caused by a loss of pain-induced inhibition of itch. Alternatively, two separate populations of DRG neurons are impaired in these mutant mice: one for pain sensation and one for itch suppression. Second, synaptic glutamate release from established (Mrgpra3-expressing) and putative (GRPhigh) pruriceptors appears to be dispensable for the transmission of itching information in the Vglut2 CKO background, as indicated by the loss of VGLUT2 expression and a lack of VGLUT1/3 expression in these neurons. This observation raises the possibility that pruriceptors may use other transmitters such as GRP to mediate itch. Third, removal of VGLUT2 from Nav1.8Cre-active DRG neurons creates a mouse model of chronic neurogenic itch, as indicated by sensitization of multiple itch pathways and development of excessive spontaneous scratching and skin lesions. Importantly, the mutant phenotypes seen in Vglut2 CKO mice are similar to the symptoms seen in human patients suffering from chronic itch. In such patients, painful stimuli do not just fail to suppress itch, but paradoxically promote itch (Hosogi et al., 2006; Ikoma et al., 2004; Ishiuji et al., 2008; Schmelz, 2010), exactly analogous to capsaicin-evoked itch seen in Vglut2 CKO mice. The creation of a mouse model of chronic itch that shares key features with human patients could be invaluable for future mechanistic and intervention studies.
The generation of mice carrying the floxed Vglut2 allele, Nav1.8Cre transgenic mice, RosaRFP mice, and the Tau-lox-STOP-lox-mGFP-IRES-nlsLacZ-neo mice have been described previously (Agarwal et al., 2004; Hippenmeyer et al., 2005; Madisen et al., 2010; Tong et al., 2007). For histochemical studies, mice at postnatal day 30 (P30) were used. For behavioral analyses, 1–2 month-old mutant and control littermates were used. All behavioral test protocols were approved by the Institutional Animal Care and Use Committee at Dana-Farber Cancer Institute.
In situ hybridization (ISH) procedures and the probes (CGRP, Nav1.8, SCG10, TrkB, Parvalbumin, Somatostatin, Dynorphin, and Enkephalin) have been described previously (Chen et al., 2006; Cheng et al., 2004; Cheng et al., 2005; Ma et al., 1999; Xu et al., 2008). Immunohistochemistry (IHC) using rabbit anti-VGLUT1 (1/1000, Swant, Switzerland), guinea pig anti-VGLUT2 (1/200, Frontier Institute Co., Japan), chicken anti-GFP (1/1000, Invitrogen, USA), rabbit anti-TRPV1 (1/1000, AbCam, USA), mouse anti-NF200 (1/200, Sigma, USA), rabbit anti-c-Fos (1/1000, Santa Cruz, USA) or IB4-biotin (10 µg/ml, Sigma, USA) was carried out as previously described (Chen et al., 2006). The ISH/IHC double staining was performed as previously described (Liu et al., 2008). For RFP/IHC double staining, the RFP fluorescent signal was directly photographed followed by performing single IHC. The fluorescent IHC signals were then merged with the RFP signal.
L4/L5 lumbar DRG were dissected from two to three pairs of mutant and control mice. 3–4 mutant or control DRG were used to prepare eight adjacent sections at 12-µm thickness. Each set was processed for immunostaining or used for ISH with the gene of interest. Only cells containing nuclei and showing levels of expression or staining clearly above background were counted. Averages and standard errors of the mean (SEM) were calculated and the difference between control and mutant samples was subjected to a Student’s t test, with p<0.05 considered significant.
Patch clamp recording in spinal slices has been done as previously reported (Kawasaki et al., 2008), using the lumbar spinal cord (L4-L5) from Vglut2F/F; Nav1.8Cre CKO and Vglut2F/F control littermates (3–5 week old). More detailed procedures are provided as supplemental information.
The spared nerve injury (SNI) model for neuropathic pain was performed on adult mice (P30 to P60) as described for rats (Decosterd and Woolf, 2000) and as we previously did in mice (Chen et al., 2006). More detailed procedures are provided as supplemental information.
Vglut2F/F; Nav1.8Cre CKO and Vglut2F/F control littermates of 1–2 months age were used. All pain and itch behavioral tests were performed as previously described (Chen et al., 2006; Shimada and LaMotte, 2008) with minor modifications. More detailed procedures are provided as supplemental information.
One-month old Vglut2F/F; Nav1.8Cre CKO and Vglut2F/F control littermates were given a 2.5 µg/10 µl intraplantar injection of capsaicin under anesthesia. Two hours later, the L4-L5 spinal cord were dissected and treated as previously described (Liu et al., 2008). C-Fos immunostaining uses 30 min incubation at room temperature for the primary and secondary antibodies, respectively, followed by in situ hybridization with NPY, enkephalin, or GRPR probe as previously described (Liu et al., 2008). 3–5 pairs of control and CKO mice were used for quantitative analysis.
For itching behaviors, the mean number of scratching bouts or wipes and standard errors of the mean during the period were calculated for each group. The difference between the mutant and control group was subjected to a Student’s t test (Two-Sample Assuming Unequal Variance), with p<0.05 considered significant. For acute mechanical and heat pain, data were calculated as the average of two independent tests performed on two consecutive days and subjected to the Student’s t-test. For CFA-induced inflammatory and SNI-induced neuropathic pain, time course measurements were analyzed by both ANOVA (within each group), and two-way repeated ANOVA (R, R Development Core Team, Austria) (to compare groups), with p<0.05 accepted as statistically significant.
We thank Dr. Rohini Kuner for the Nav1.8Cre mice, Dr. Silvia Arber for the TauGFP reporter mice, and Dr. Zhoufeng Chen for the GRPR probe. We thank Drs. Clifford Woolf, Sang-Kyou Han, Charles Stiles, and Fu-chia Yang for critical comments on the manuscript. The work is supported by the NIH grants from NIDCR (1R01DE018025) and NINDS (5P01NS047572).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.