Background
Autism Spectrum Disorders (ASD) are characterized by a range of sensory abnormalities, in particular in the domain of tactile sensitivity [
1]. Among these, abnormal nociception is strikingly common in ASD and it manifests itself either as hypo-sensitivity or as hyper-sensitivity to painful stimuli. The common occurrence of self-injury, self-mutilation (including cases of self-extraction of teeth) and unreported wounds [
2], supported by clinical and experimental studies [
3‐
5] has been interpreted as evidence of reduced sensitivity to painful stimuli in ASD patients. Conversely, a subset of ASD patients display hyperalgesia and pain hypersensitivity in the form of mechanical/tactile allodynia [
6], static mechanical allodynia (pain in response to light touch/pressure [
7,
8], dynamic mechanical allodynia (pain in response to stroking lightly), reduced threshold for thermal pain [
9], movement allodynia (pain triggered by normal movement of joints or muscles) and chronic pain unrelated to medical conditions [
10‐
12]. Pain hypersensitivity may constitute a major and underappreciated source of discomfort for ASD patients, in particular due to their inefficient communication capabilities that afflict most of the affected ASD patients [
2], resulting in unnecessary medical procedures [
10].
It is worth noting that ASD mouse models, carrying mutations in different genes, may display hyper- or hyposensitivity to pain, suggesting that the variability of the clinical phenotype may be linked to the genetic heterogeneity of ASD (more than 800 genes are linked to ASD [
13]). In fact, each genetic mutation might disrupt a discrete but different node of the nociceptive circuit, leading to mutation-specific phenotypes and mechanisms. In particular, loss of the ASD-related Shank3 protein results in pain hyposensitivity due to the disruption of the scaffold architecture, enabling TRPV1 signaling in dorsal root ganglion (DRG) neurons [
14].
Since disturbed nociception is a major source of discomfort for patients, knowledge of the involved neuronal circuits may provide insights into diagnosis and treatment. Although substantial processing of somatosensory and nociceptive stimuli takes place in circuits in the dorsal spinal cord [
15], we largely ignore the extent of the involvement of these circuits in ASD-associated nociceptive phenotypes, either in terms of the cellular subpopulations involved or in terms of molecular and neurochemical abnormalities at work. One could speculate that, since a large fraction of ASD-associated genes code for synaptic proteins [
16] change in the synaptic architecture, connectivity and excitation/inhibition balance may be altered in the spinal cord of ASD patients and murine models.
Here we consider the ASD model obtained by deleting the gene coding for the postsynaptic density (PSD)-enriched scaffold protein Shank2 [
17]. Indeed, point mutations and missense mutations in Shank2 are responsible for a small but consistent fraction of ASD cases [
18].
Shank2
−/− mice (obtained by targeting exon-6 and exon-7; [
19,
20]), are considered a
bona fide mouse model of ASD. In fact, they display a phenotype characterized by reduced social interaction, increased anxiety and compulsive grooming. Recent work has suggested that Shank2
−/− mice may display abnormal nociception due to alteration of spinal cord circuits [
21], although it must be stressed that deletion of different exons in distinct Shank2
−/− models may also lead to different phenotypes [
22]. Nevertheless, the neuronal subpopulations and the circuit architectural features responsible for such phenotype remain to be investigated.
Here we show that Shank2−/− mice display hypersensitivity to formalin-induced pain; notably, we have identified (in murine and human spinal cord) a subpopulation of glycinergic interneurons characterized by very high expression levels of Shank2. Loss of Shank2 results in the reduction in synaptic NMDAR and in the blunted recruitment of these inhibitory interneurons upon painful stimuli, which results in the over-activation of lamina-I nociceptive projection neurons.
Discussion
Our data show that high levels of Shank2 expression identifies a subpopulation of glycinergic inhibitory interneurons located in the dorsal horn of the spinal cord, receiving inputs from somatosensory afferents; Shank2 loss leads to a decrease in excitatory synapses onto glycinergic interneurons and in net increase of excitation in the dorsal horn, correlated with increased sensitivity to chemically induced pain.
The proposed mechanism involved in pain hypersensitivity of Shank2
−/− mice is therefore related to dis-inhibition of spinal cord nociceptive circuits [
63,
64]. A similar increase in nociception has been observed in several experimental models in which the inhibitory tone has been decreased: pharmacological blockade of glycinergic inhibition produces mechanical allodynia through the dis-inhibition of PKC-γ+ neurons [
65] and silencing of glycinergic interneurons is sufficient to induce allodynia [
59]. This concept has been further supported by recent data obtained in patients with loss-of-function mutations of glycine receptors or glycine transporters (leading to the clinical syndrome of hyperekplexia), who display decreased pain thresholds and amplified pain withdrawal reflexes [
66]. Increased nociception as a consequence of loss of spinal inhibition has been also observed in case of loss of GABAergic inputs [
67]: inactivation of PV+ inhibitory interneurons cause the appearance of tactile allodynia [
68,
69], whereas GABAergic agonists produce anesthesia [
70]. Within the conceptual framework of the “gate-control” theory, inhibitory inputs are well understood to prevent the runaway activation of PKC-γ+ interneurons by touch-evoked inputs as well as other excitatory interneurons in laminae II [
15], which, in turn, would drive the excitation of laminae I (NK1+) interneurons. In agreement with this model, reduced excitation of glycinergic interneurons in Shank2
−/− mice leads to increased activity of laminae I neurons.
Abnormalities in pain processing have been previously reported in Shank2
−/− mice [
21,
71]. Ko et al. reported of an overall decreased basal tactile perception and acute pain response in Shank2
−/− and induction of neuropathic or inflammatory chronic pain. Some of the reported findings agree with our own: the reduced mechanical allodynia at baseline reported by Ko is also visible in our dataset (Fig.
1d-pre) with a response comparable in the two genotypes up to day 7 in the CFA-von Frey test followed by reduced chronic pain at a later timepoint (Fig.
1d). Furthermore, Shank2
−/− did display a longer latency to escape in the texture preference test (Fig.
1n), implying a decrease in tactile sensitivity. In other cases, conditions are not fully comparable: the hot-plate test was performed by Ko at 55 °C, whereas 45 °C was used in this study, and we did not specifically investigate neuropathic pain, which may hinge on different long-term sensitization mechanisms [
72]. Taken together, ours and previously published data suggest that nociceptive abnormalities in Shank2
−/− may be specific of sensory modalities and depending on multiple alterations in synaptic plasticity and circuit function.
Genetic approaches have increasingly revealed the high degree of heterogeneity in neuronal subpopulations in the dorsal spinal cord [
41,
61,
62]. Recently, the genetic diversity of these populations has been demonstrated to be extensive and a large number of subpopulations have been identified in a single-cell transcriptome study [
73]. Nevertheless, neuronal physiology is highly influenced by the quantity and quality of their synaptic inputs, which is strongly dependent upon the composition and architectural organization of postsynaptic structures [
74,
75]. Therefore, a distinct layer of diversity may be identified once the synaptic composition of neuronal subtypes is taken into consideration. Here we provide a first proof of this concept; in fact, Shank2 distinguishes a subclass of glycinergic and parvalbumin interneurons as well as a small population of excitatory interneurons.
Since different members of the Shank family, despite their similarity, are not considered mutually redundant [
19,
76], the function of Shank2
high cells is predicted to be heavily impacted by Shank2 loss, despite their anatomical integrity (i.e., their normal number and positioning). Of note, the Shank2 gene gives rise to several splice variants and isoforms generated using alternative promoters (ranging from 130 to 230 KDa; [
29,
53,
77]). The Shank2
−/− model used in the present study is more precisely a deletion mutant lacking the exon 7 and this mutation should lead to the early termination and non-sense decay of the Shank2 mRNA [
19]. In agreement with the cortex finding, in the spinal cord homogenate of Shank2
−/− mice we observe the loss of the most abundant isoforms (160–220 kDa) whereas only a low abundance, low MW isoform may still be expressed. However, the immunostaining of cortical samples from Shank2
−/− mice reveals a largely complete loss of immunoreactivity and the immunostaining of spinal cord from Shank2
−/− mice reveals an almost complete loss of immunoreactivity (any residual immunoreactivity seen in spinal cord may correspond to the low-abundance 130 KDa isoform, whose functional relevance, if any, remains to be investigated).
DRG are known to express some isoforms of Shank3, which contribute to arrange the arrays of receptors and ion channels in the peripheral projections of the ganglion cells [
14,
54]. However, DRGs appear to express very little of the 160–180 KDa better known Shank2 isoform, and almost none of the 160 kDa isoform which is characteristically eliminated by the deletion of the exon 7. The immunostaining of DRG for Shank2 revealed only a small number of cells with a very faint immunopositivity, in contrast with the abundant and widespread expression of Shank3 [
14]. While it is not possible to fully exclude any DRG or peripheral effect of Shank2 loss on sensory phenotypes observed in our mouse model, the limited expression of Shank2 in DRG suggest that at least a component, possibly a substantial one, of the observed phenotype is due to the loss of Shank2 within the central nervous system (CNS).
In agreement with observations in other neuronal subtypes [
22,
76,
78], we find that loss of Shank2 causes the decrease in NMDAR expression in excitatory synapses on GlyT2+ interneurons. Although baseline neurotransmission in the pain processing circuit appears to be not affected (as the acute phase of the formalin test is comparable in Shank2
−/− mice and WT littermates), the NMDAR-dependent synaptic plasticity that is thought to underlie the second phase of the behavioural response to the formalin test [
79‐
81] may be unbalanced, with insufficient potentiation of the inhibitory circuit. In fact, glycinergic interneurons are strongly activated in WT (as shown by the c-Fos induction), but not in the Shank2
−/− animals. Thus, the resulting decrease in glycinergic transmission in the pain processing circuit would cause excessive excitatory drive, as demonstrated by the increase in the number of c-fos+ neurons in laminae I and II. In fact, dysfunction or silencing of inhibitory interneurons is known to cause pain hypersensitivity in human patients and experimental animals [
59,
66,
82] in particular by dis-inhibiting PKC-γ+ excitatory interneurons [
65]. However, disturbances in excitatory synapses due to Shank2 loss may affect other neuronal elements of the circuit: in fact, pain induced by intrathecal administration of NDMA is reduced in Shank2
−/− mice, suggesting that Shank2 loss may affect differentially multiple sensory and nociceptive modalities.
Thus, circuits involved in nociception may be disturbed across multiple nociceptive stimuli and alteration of somatosensation may co-exist (as previously shown, [
54,
83]). Interestingly, Shank2
−/− mice have been also reported to display a reduced sensitivity to the nociceptive response evoked by intrathecal injection of NMDA [
21]. Since this procedure does not selectively activate one subpopulation of neurons, it is not straightforward to explain it in circuit terms. In fact, Shank2 is highly enriched in glycinergic interneurons, but it is not restricted to these cells (see Fig.
1a, b and Additional file
1: Fig. S1) and even among the Shank2
high cells, a fraction of excitatory neurons (VGluT2+, Prrxl1+) can be identified and their role remains unexplored. Furthermore, the dysfunction of PV interneurons in spinal cord has been directly related to mechanical but not thermal allodynia[
69]; interestingly, only a fraction of PV interneurons appear to be Shank2
high and indeed we detect thermal but not mechanical allodynia.
Thus, the impact of Shank2 loss may affect modality-specific pain processing circuits in a distinct way, depending on the role of different cellular subpopulations.
The insufficient activation of glycinergic interneurons because of disrupted excitatory synapses observed in Shank2
−/− mice is a new mechanism for abnormal pain processing in autism. In fact, reported pain hyposensitivity in Shank3 mice has been linked to the loss of Shank3 in neurons in the dorsal root ganglia and the direct effect of Shank3 absence on TRPV channel expression [
14]. Likewise, autism-related behavioural dysfunctions have been linked to the disturbed sensory input generated by abnormal sensory neurons in dorsal root ganglia [
54,
83]. Conversely, loss of function of the ASD-associated gene Caspr2 is associated with neuropathic pain [
84] through mechanisms involving increased sensitization of neurons in dorsal root ganglia. Although these reports all point to a sensory dysfunction originating in the periphery, our findings suggest that disruption of spinal cord circuits may be a strong contributor to the observed hyper- and hyposensitivity to nociceptive stimuli. Furthermore, one can speculate that the same excitation/inhibition balance disruption that is thought to underlie the ASD spectrum disorder may also manifest itself in spinal circuits to contribute to drive sensory abnormalities.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.