Background
Autism spectrum disorders (ASD) consist of a group of heterogeneous neurodevelopmental disorders, with a high prevalence of ~1/100 children [
1]. Core symptoms of ASD are deficits in social interaction and communication, together with restricted interests and repetitive behaviors [
2‐
4]. The etiology of ASD remains unclear, but a strong genetic component is evident. ASD candidate genes are often implicated in synaptic transmission, are part of synapse formation/maintenance and/or affect the neurodevelopment during particular moments, e.g., during the “critical period” [
5,
6]. Functionally, these changes affect the excitation/inhibition (E/I) balance and subsequently influence network properties [
7,
8].
One of the most important gene families mutated in ASD is the
SHANK gene family [
9], coding for multi-domain scaffolding proteins located in the postsynaptic density of glutamatergic synapses [
10]. In individuals with ASD or schizophrenia patients with ASD traits, mutations were repeatedly reported for all
SHANK gene family members [
4,
9,
11,
12], namely
SHANK1 [
13],
SHANK2 [
14‐
16] and
SHANK3 [
17‐
21]. Moreover,
SHANK3 haploinsufficiency has been found in patients affected by the Phelan-McDermid 22q13 deletion mental retardation syndrome [
22‐
25], often characterized by ASD features [
26]. Importantly, mutations of
SHANK genes were detected in the whole spectrum with a gradient in severity in mental retardation. Specifically,
SHANK1 mutations were found in individuals with ASD and normal intelligence, whereas
SHANK2 and
SHANK3 mutations were associated with mild and severe mental retardation, respectively [
9]. Consistent with the important role of the
SHANK gene family in ASD, genetic
Shank mouse models display behavioral alterations with relevance to all human ASD core symptoms.
Shank1 null mutant mice display social and communication deficits [
27], alterations in repetitive behavior, with elevated self-grooming behavior, particularly in social situations [
28], and a mixed cognitive phenotype resembling aberrant cognitive processing evident in some ASD cases [
29,
30]. Likewise in two
Shank2 models, strong ASD-related behavioral alterations are evident [
17‐
19,
31,
32]. In the various
Shank3 models severity of the ASD phenotype varies with genetic manipulation, with a comparatively mild phenotype in the
Shank3 model lacking the ANK domain [
33,
34], but strong phenotypes in the other models [
35‐
37], see also [
38‐
40].
In the process of brain development, GABAergic signaling plays an essential role. Thus, it is not surprising that its disturbance/disruption has been related to the pathogenesis of ASD [
41,
42]. Investigations on the role of the GABAergic system during neurodevelopment, however, are impeded by the fact that GABAergic interneurons are made up of different subtypes displaying heterogeneous morphological and physiological features; in the hippocampus, up to 21 subtypes have been identified [
43]. One way to classify GABAergic, e.g., cortical interneurons, is based on the expression of Ca
2+-binding proteins such as parvalbumin (PV; gene symbol:
PVALB), calbindin D-28k and calretinin [
44]. Among these specific subpopulations, PV-expressing interneurons seem to be highly impacted in several neuropsychiatric disorders including schizophrenia, bipolar disorder and ASD [
45]. While in none of the previous studies in humans the
PVALB gene itself was found to be mutated, a reduction in the order of 20–25 % in the number of PV-immunoreactive (PV
+) neurons has been reported in ASD individuals [
46] and mouse ASD models [
47], see Table S1 in [
48]. However, so far the question was not thoroughly addressed as to what extent the reduction in PV
+ neurons was the result of improper neurodevelopment (e.g., altered GABA interneuron subtype), neuron loss (neuronal death) or PV down-regulation (mRNA and/or protein). This is of high relevance, since the alternatives likely have distinct and even opposing effects on the excitation/inhibition balance: while loss of the Pvalb neuron subpopulation is likely to result in reduced inhibition, decreased PV expression results in enhanced inhibition. The absence of PV in PV-/- Pvalb neurons does not affect basal synaptic transmission, but enhances facilitation [
49,
50] and shortens delayed transmitter release [
51]. This asynchronous release augmented in the presence of PV is assumed to be important to desynchronize large fractions of local networks and prevent/disrupt excessive synchronized activity [
52]. In line the observed increased regularity of spiking of PV-/- striatal Pvalb FSI in vitro [
53], the appearance of synchronous 160-Hz oscillations in the cerebellum of PV-/- mice in vivo [
54] and facilitation of the GABA
A-ergic current reversal caused by high-frequency stimulation in PV-/- hippocampal Pvalb FSI in vitro [
55], all together provide evidence that PV plays a key role in the regulation of local inhibitory effects on pyramidal neurons, as well as on other interneurons (for more details, see [
56]). Importantly, at the behavioral level, mice with reduced PV expression (PV+/-) or without PV (PV-/-) display a robust ASD-like phenotype [
48]. Although qualitative immunohistochemistry revealed no striking differences with respect to VVA
+ (putatively Pvalb) neurons in PV-/- mice [
55,
57], a quantitative and systematic analysis of Pvalb neurons by unbiased stereology in different brain regions including ASD-implicated regions such as the medial prefrontal cortex (mPFC), somatosensory cortex (SSC) and striatum was still missing. Besides determining the number of Pvalb neurons in PV+/- and PV-/- mice, we quantified the number of this interneuron subpopulation in two well-established ASD mouse models, i.e.,
Shank1-/- and
Shank3B-/- mice, covering the extremes of the spectrum with a gradient in severity in mental retardation.
Discussion
One of the difficulties in gaining knowledge on the mechanisms underlying the pathogenesis of ASD is the large number (>100) of putative ASD risk genes identified in genetic studies in humans and animal models [
4,
78,
79]. An important line of evidence has centered on mutations in genes implicated in synapse structure and/or function including
NRXN,
NLGN and
SHANK family members [
17‐
19,
80]. Such mutations might eventually lead to an alteration of the E/I balance as demonstrated in some ASD mouse models [
7,
8]. Other hypotheses on ASD-associated genes and/or gene/environment relationships come from many computational studies (GWAS, transcriptomic expression network analyses); several ones revealed impaired Ca
2+ signaling, i.e., alterations in the Ca
2+ node in gene networks [
4,
81,
82] to represent a convergence of mechanisms relating to ASD, in line with previous propositions [
83‐
85]. PV plays an important role in the Ca
2+ homeostasis regulating many aspects of neuronal signaling (short-term plasticity, synchronization, precision of spike timing, etc.) in the subpopulation of PV
+ interneurons [
56]. Absence/down-regulation of PV in PV-/- and PV+/- mice, respectively, not only affects the properties of the Pvalb neurons, but also of neurons impacting on Pvalb cells, and leads to a robust ASD-like behavioral phenotype characterized by impaired social interaction behavior, reduced pro-social ultrasonic vocalizations and deficits in reversal learning [
48]. Of relevance, a decrease in the number of PV
+ cells has been reported in many genetic ASD mouse models in various brain regions including mPFC, SSC and striatum; (see Table S1 in [
48]). Also in the few human studies, a decrease in PV
+ cells and/or
PVALB mRNA was reported [
46,
86]. Thus, the networks containing Pvalb neurons were hypothesized to be strongly implicated in ASD [
45].
While an involvement of Pvalb neurons in ASD is rather undisputed, it remains unclear whether the observed reduction of PV
+ cells in a particular brain region is the result of I) a truly decreased number of Pvalb neurons resulting from the many putative mechanisms including an immature or perturbed developmental state (e.g., layer- and/or region-inappropriate localization of Pvalb neurons, increased susceptibility, premature cell death, etc.) or II) alternatively from the down-regulation of PV protein levels or the failure to express adequate levels of the protein. To address this question, one needs to identify Pvalb cells by another means; one of the most common marker is the particular extracellular matrix surrounding Pvalb cells that can be visualized by VVA staining. Of note, the appearance of VVA staining is developmentally and layer-dependent regulated as shown in the mouse visual cortex (V1) [
87] and the barrel cortex [
88]. Moreover, PNNs are regulated by activity and are decreased under certain pathological conditions such as oxidative stress [
89] or Alzheimer’s disease [
90]. Thus, in the first step we ascertained by stereology [
60] that the number of VVA
+ cells was unchanged in mice with reduced (PV+/-) or absent (PV-/-) PV expression in the mPFC, SSC and striatum. In agreement with previous results obtained in the cortex and hippocampus of adult PV-/- mice [
55,
57], there was no indication of a cell decrease/loss of Pvalb neurons. Moreover, in a mouse line expressing EGFP in Pvalb neurons, the number of EGFP
+ cells was found to be the same as for PV
+ and VVA
+ cells in PV+/+ mice indicative of the identification of the essentially same Pvalb cell population, where either morphological or functional abnormalities have been reported in ASD [
91‐
93]. Analyses of double-labeled (VVA
+ and PV
+) cells using either the total of PV
+ or VVA
+ cells for normalization revealed approximately 70–80 % of co-stained cells within the mPFC, SSC and striatum of WT mice. Similar numbers were reported in the mouse cortical V1 region, where 82 % of all VVA
+ cells were found to also express PV [
87]. While the percentage of PV
+ cells enwrapped by PNN (VVA
+) was not different between sections from WT and PV+/- mice, as the result of decreased PV protein expression levels in PV+/- mice, both the total number of PV
+ cells, as well as the percentage of VVA
+ cells expressing PV was significantly decreased in all regions analyzed in our study.
Thus, we wondered, whether the previously reported decrease in PV
+ cells in other canonical ASD mouse models might not be -globally or in part- the result of PV down-regulation. We therefore assessed PV protein expression and VVA as a marker for Pvalb cells in the
Shank1 and
Shank3 mouse models. The selection among the many ASD mouse models available was based on the fact that more than 900 patients with genetic alterations in
SHANK genes were identified, with the
SHANK gene family being the primary gene family implicated in ASD [
9].
Shank1 and
Shank3 mouse models were selected to cover both extremes of the spectrum, namely
SHANK1 mutations found in individuals with ASD and normal intelligence and
SHANK3 mutations associated with severe mental retardation [
9]. The selection was further facilitated by a recent comprehensive neuro-morphological MRI study on 26 ASD mouse models [
75]. Similar to the most consistent finding in humans [
94,
95], an increase in the frontal and parieto-temporal lobes and decreased volume of the cerebellar cortex were observed in one of the three subgroups (“group 1”) identified among the 26 investigated models [
75]. A prominent example of the “group 1” ASD mouse models are
Shank3 null mutant mice. Importantly, however, also the morphological changes in PV-/- mice follow the same pattern, as we recently demonstrated: increased neocortical volume and a decreased size of the cerebellum [
48]. A further link exists between Shank protein expression and Pvalb neurons [
76,
96]. In
Shank1-/- hippocampal PV
+ FSI the absence of Shank1 functionally leads to a decrease in excitatory synaptic inputs and inhibitory synaptic outputs to pyramidal neurons and furthermore to molecular changes including the down-regulation of the postsynaptic proteins GKAP, PSD-95 and gephyrin [
96]. These alterations affect the E/I balance in CA1 pyramidal neurons. Whether a similar situation prevails in
Shank1-/- cortical PV
+ FSI is currently unknown. In
Shank3 knockout mice, a reduction in PV
+ puncta staining (intensity and puncta numbers) around pyramidal cells in the insular cortex of
Shank3B-/- mice was associated with weakened GABAergic circuit function and impaired postnatal pruning [
76]. Moreover, the decrease in PV-ir puncta intensity in
Shank3B-/- hinted towards a PV expression-related phenomenon, although the question of PV expression levels was not directly addressed in this study.
For the
Shank models our interest was focused on regions with high expression of either protein, i.e., SSC for
Shank1-/- and striatum for
Shank3B-/-. In both ASD models, the VVA
+ cell number was unchanged and the number of PV
+ cells was decreased, also evidenced by the decreased percentage of PV
+ cells among all VVA
+ cells. A decrease in PV protein levels and
Pvalb mRNA levels to approximately 50 % of WT in both
Shank mutants are in full support of a down-regulation of PV. In addition, the demonstration that PV levels are decreased may indicate a shift in the E/I balance towards an increased inhibition, taking into account the proven role of PV in synaptic transmission [
52,
53,
55,
97,
98]. In striatal PV
+ FSI the increased facilitation (increased inhibition) between FSI and medium spiny neurons (MSN) caused by the absence of PV in PV-/-mice [
53] is partly compensated by a decrease in the excitatory synaptic input from cortical pyramidal cells, a mechanism hypothesized to compensate for the increased output of PV-/- neurons [
48]. The modification of the E/I balance within the PV-circuitry in PV-/- mice is reminiscent of the situation in schizophrenia, where NMDA receptor hypofunction leads to the reduction of glutamic acid decarboxylase 67 (GAD67) levels and thus GABA synthesis. The concomitant decrease in PV expression levels might be viewed as an adaptive/compensatory response in order to enhance facilitation (inhibition) and to compensate –at least partially– for the decrease in GAD67 [
99‐
101].
How might the absence of either Shank proteins lead to PV down-regulation? The similar magnitude in down-regulation of PV and
Pvalb mRNA in both
Shank mutants is indicative of a regulation at the transcriptional level. Currently little is known on the physiological regulation of PV expression in the brain. The enfeebled GABAergic circuit function reported in
Shank3B-/- [
76] and
Shank1-/- mice [
96] is likely to decrease somatic Ca
2+ signals and subsequently modify the Ca
2+ signaling components of Pvalb neurons; a neuronal activity-related Ca
2+-dependent transcriptional regulation of PV expression was proposed before [
100]. Thus, the impairment in GABAergic function reported in both
Shank models is likely to impact on the Ca
2+-dependent excitability-transcription (E/T) coupling including transcription of the
Pvalb gene. Alterations in other genes implicated in E/T coupling have been observed before in ASD individuals and include Ca
2+ signaling components such as the voltage-dependent channels Ca
v1.2 (
CACNA1C), Ca
v1.3 (
CACNA1D) and the α-δ auxiliary subunit of L-type voltage-gated Ca
2+ channels (
CACNA2D3) [
102,
103]. Evidence has accumulated that mutated ASD risk genes are critical components of activity-regulated signaling networks often controlling synapse development and morphology, as well as structural and functional plasticity [
8]. In summary, the observed decrease in PV expression in
Shank mutant mice might be viewed as an adaptive or compensatory mechanism to possibly restore (increase) synaptic output according to the concept of the Ca
2+ homeostasome [
104].
Our findings might have important implications for novel treatment strategies for ASD, particularly as most current strategies aim to enhance inhibition in order to compensate for a presumed increase in excitatory neurotransmission, e.g., by the GABA
B receptor agonist barbaclofen [
80,
105]. Our findings indicate the exact opposite, namely that an enhancement of excitatory neurotransmission onto PV
+ neurons possibly restoring PV levels and thus PV-modulated functions (e.g., short-term plasticity, synchronization, precision of spike timing) may ameliorate ASD symptoms. Alternatively, one might envisage that direct up-regulation of PV might be a means to ameliorate PV-circuitry function resulting in the attenuation -or in the best case abolition- of the ASD phenotype.
Authors’ contributions
BS conceived the study, participated in the data analyses and in the writing of the manuscript. FF carried out the Stereology experiments, performed the statistical analysis and participated in writing of the manuscript. KJV, AÖS and MW provided and characterized the Shank mutant mice. MW additionally participated in the design of the study and participated in the writing of the manuscript. All authors read and approved the final manuscript.