Fragile X cortex is characterized by decreased parvalbumin-expressing interneurons

Abstract

Fragile X syndrome is a genetic neurodevelopmental disorder caused by a mutation of the fragile X messenger ribonucleoprotein 1 (FMR1) gene in the X chromosome. Many fragile X syndrome cases present with autism spectrum disorder and fragile X syndrome cases account for up to 5% of all autism spectrum disorder cases. The cellular composition of the fragile X syndrome cortex is not well known. We evaluated alterations in the number of Calbindin, Calretinin, and Parvalbumin expressing interneurons across 5 different cortical areas, medial prefrontal cortex (BA46), primary somatosensory cortex (BA3), primary motor cortex (BA4), superior temporal cortex (BA22), and anterior cingulate cortex (BA24) of fragile X syndrome and neurotypical brains. Compared with neurotypical cases, fragile X syndrome brains displayed a significant reduction in the number of PV⁺ interneurons in all areas and of CR⁺ interneurons in BA22 and BA3. The number of CB⁺ interneurons did not differ. These findings are the first to demonstrate that fragile X syndrome brains are characterized by cortical wide PV⁺ interneuron deficits across multiple cortical areas. These add to the idea that deficits in PV+ interneurons could disrupt the cortical balance and promote clinical deficits in fragile X syndrome patients and help to develop novel therapies for neurodevelopmental disorders like fragile X syndrome and autism spectrum disorder.

Introduction

Fragile X syndrome (FXS) is a neurodevelopmental disorder caused by a mutation in the non-coding region of the fragile X messenger ribonucleoprotein 1 (FMR1) gene in the X chromosome. Silencing of the FMR1 gene is the result of hypermethylation of a CGG trinucleotide expansion of over 200 repeats in the promoter region of this gene. Silencing of the FMR1 gene results in no FMR protein (FMRP) production. FMRP is an RNA-binding protein that regulates translation of hundreds of proteins and its loss in individuals with FXS leads to dysregulation in many genes that are responsible for synaptic plasticity and connectivity in the developing brain. Dysregulation of the glutamatergic-GABAergic, Wingless (Wnt), retinoic acid (RA), mechanistic target of rapamycin (mTOR), extracellular signal related kinase (ERK), and endocannabinoid (eCB) signaling pathways during early neurodevelopment is believed to promote some of the clinical manifestations that characterize FXS patients (Salcedo-Arellano et al. 2021).

FXS is the single most common monogenic cause of autism spectrum disorder (ASD), with 50 to 60% of the individuals with FXS being co-diagnosed with ASD. Individuals with FXS present with many behavioral challenges which may include hyperactivity, impairments in social interactions, obsessive-compulsive traits, particular phobias and aggression, in addition to learning disabilities and the presence of psychological problems, such as anxiety and depression. The clinical overlap between FXS and idiopathic ASD suggests that the cortical wide cellular and molecular alterations that promote the ASD phenotype may have similarities in FXS patients. This includes numerous MRI studies that show neuroanatomical abnormalities of increased brain matter in the caudate nucleus of both ASD and FXS brains (Hollander et al. 2005; Sandoval et al. 2018). Another study found volumetric deficits in regions of the cerebellum for both FXS and ASD groups, although there were additional deficits seen in the parietal and temporal gyri for the FXS group only (Wilson et al. 2009). Another study showed that relative to individuals with only FXS, those with both FXS + ASD had substantially enlarged caudate nucleus volume and smaller amygdala (Hazlett et al. 2009). Behaviorally, it has been shown that relative to individuals with only FXS, those co-diagnosed with ASD (FXS + ASD) present with more severe behavioral and cognitive challenges and had a higher percentage of seizures, sleep problems, like aggressive and disruptive behavior and higher use of α-agonsits and antipsychotics. While the exact pathophysiology of ASD remains unknown and is most likely multifactorial, the excitation/inhibition hypothesis has been proposed to explain how alterations in brain circuitry can lead to promoting clinical manifestations. Given the close behavioral phenotype found between FXS and idiopathic ASD, this hypothesis is believed to apply largely to brains of FXS individuals. Some of these changes are a result from alterations in both glutamatergic and GABAergic signaling. Previous studies in human ASD brains reported a reduction of gamma amino butyric acid (GABA) levels in the frontal lobe (Harada et al. 2011), a decrease in GABA_A and GABA_B receptors in the anterior cingulate cortex and fusiform gyrus, and a reduction in the GABA_ARα2 subunit in the prefrontal cortex (Oblak et al. 2009; Oblak et al. 2010; Mori et al. 2012; Hong et al. 2020). Another study also found a reduction in GABAergic Purkinje cells in the brain with ASD (Whitney et al. 2009). In FXS, the expression of the GABA_A subunit was reduced by half in animal model studies (D’Hulst et al. 2006; D’Hulst et al. 2009). Deficits in GABA signaling, in ASD, at the anatomical level also contribute to deficits in inhibitory signaling (Ariza et al. 2018; Hashemi et al. 2018; Amina et al. 2021). Accumulating evidence in other neurodevelopmental disorders like ASD, schizophrenia, Rett Syndrome, and Down Syndrome suggest that there are interneuron alterations in various cortical regions. For example, animal models of Rett Syndrome (Morello et al. 2018) and Down Syndrome (Pérez-Cremades et al. 2010) showed increases in the amount of Parvalbumin (PV) expression and number of calretinin-immunoreactive cells in the primary somatosensory cortex, respectively. In the Valproic Acid model of Autism, there was a 15% reduction in the density of PV+ interneurons and decreased Pvalb mRNA and PV protein levels in the striatum (Lauber et al. 2016). Moreover, our group previously studied 3 subpopulations of GABAergic interneurons based on expression of the calcium-binding proteins Calbindin (CB), Calretinin (CR), and PV in postmortem human brains of individuals with ASD; and found a reduction of (PV⁺) interneurons across Brodmann areas (BAs) 9, 46, and 47 in the prefrontal cortex (Ariza et al. 2018; Hashemi et al. 2018; Amina et al. 2021). Three independent studies in Fmr1 KO mouse models showed a significant decrease in (PV⁺) interneurons in the somatosensory cortex (Selby et al. 2007; Kourdougli et al. 2023) and the neocortex (Lee et al. 2019).

At the moment, there are no human studies that have evaluated the underlying pathology of cortical-wide interneuron alterations in the cerebral cortex of FXS brains. The present study is the first to determine changes in the GABAergic interneuron population in FXS postmortem tissue. The aim of this study was to evaluate alterations in the CB, CR, and PV expressing interneurons across 5 different cortical BAs sampled from the frontal, temporal, and parietal cortical lobes: the anterior cingulate cortex (BA24), superior temporal cortex (BA22), medial prefrontal cortex (BA46), primary motor cortex (BA4), and primary somatosensory cortex (BA3a, hereafter referred to as BA3). All BAs were chosen due to their functional roles that are believed to be altered at the molecular and cellular levels in individuals with FXS.

Materials and methods

Subjects and tissue collection

Human cortical tissue samples from BA3, BA4, BA22, BA24, and BA46 were obtained from 7 FXS and 7 neurotypical, age and sex-matched cases (Table 1). For FXS cases, the diagnosis of FXS (CGG expansion > 200 repeats) was confirmed by PCR and Southern blot. Our FXS cases consisted of 6 males and 1 female, for a mean age of 69.6 ± 3.48 (±SEM) years. Neurotypical cases did not have history of neurological disorders, mean age was 68.4 ± 2.84. The tissue was obtained from the Fragile X Brain Repository held at the University of California Davis, a node of the Hispano American Brain Bank of Neurodevelopmental Disorders (Dufour et al. 2022). Brain tissue was collected from donors at autopsy, after obtaining informed consent from the next of kin, and with the approval of the UC Davis Institutional Review Board (IRB). Upon brain collection, tissue was immediately preserved in 10% buffered formalin and subsequently cut into 1 cm thick coronal slabs prior to sectioning. All the Neurotypical cases were confirmed to be free of neurological disorders, including ASD and FXS, based on medical records and information gathered at the time of death from the next of kin. FMR1 genotyping was assessed by PCR on all FXS cases to determine CGG repeat length and confirm a clinical diagnosis, as previously described (Tassone et al. 2008; Filipovic-Sadic et al. 2010). Based on Brodmann cortical neuroanatomy (Mai et al. 2015), blocks containing BA46 in the middle prefrontal cortex, BA3 in primary somatosensory cortex, BA4 in primary motor cortex, BA22 in superior temporal cortex, and BA24 in anterior cingulate, were isolated from each of the cases and embedded into optimal cutting temperature compound (OCT) (Fig. 1). This OCT embedded tissue was cut perpendicularly to the pial surface into 14-μm sections on a cryostat and sections were stored at −20°C prior to staining. 2–4 tissue sections per case were used for staining procedures and were selected sequentially based on order in which they were cut, with rostral sections being used first. We used 14 μm thin slices to avoid having overlapping layers of cells that may contribute to potential quantitation errors such as counting 2 identically stained cells as 1.

Table 1.

Clinical characteristics of postmortem neurotypical and fragile X cases, including sex, age, PMI, diagnosis, CGG, hemisphere from which the brain was collected, different BAs analyzed and the known cause of death. NK = not known.

Case ID	Sex	Age	PMI (h)	Diagnosis	CGG repeat Count	Hemisphere analyzed	BAs analyzed	Cause of death
UCD 14-15	M	60	80	Neurotypical	Not Known	Right	BA3, BA4, BA24 BA22, BA46	Pulmonary Emboli
UCD 18-05	M	62	37	Neurotypical	Not Known	Left	BA3, BA4, BA46	Cardiopulmonary Arrest
UCD 15-07	F	64	NK	Neurotypical	Not Known	Left	BA3, BA4, BA24 BA22, BA46	Not Known
UCD 14-01	M	66	48.5	Neurotypical	Not Known	Not Known	BA3, BA4, BA24 BA22,BA46	Acute Renal Failure
UCD 21-10	M	72	66	Neurotypical	Not Known	Right	BA3,BA24, BA46	Hypoxic Respiratory Failure
UCD 20-02	M	74	136	Neurotypical	Not Known	Left	BA3, BA4, BA24 BA22,BA46	Cardiac Arrest
UCD 19-12	M	81	72	Neurotypical	Not Known	Left	BA3, BA4, BA24	Not Known
1031-08-GP	M	57	20	Fragile X	436	Left	BA3, BA4, BA24 BA22,BA46	Multiple Systems Organ Failure due to Choking
1031-09-LZ	M	64	11.5	Fragile X	440	Left	BA3, BA4, BA24, BA46	Primary Liver Neoplasm
1061-19-JB	F	64	30	Fragile X	780	Left	BA3, BA4, BA24 BA22,BA46	Not Known
1005-14-JC	M	65	60	Fragile X	700	Right	BA3, BA4, BA24 BA22,BA46	Congestive Heart Failure and Cardiac Arythmia
1001-18-LD	M	78	6	Fragile X	235	Right	BA3, BA4, BA24 BA22,BA46	Septicemia and Aspiration Pneumonia
1033-08-WS	M	78	17.5	Fragile X	1225	Left	BA3, BA4, BA24, BA46	Suspected Gastrointestinal Bleeding
1007-18-RF	M	80	NK	Fragile X	1000	Right	BA3, BA4, BA24 BA22,BA46	Aspiration Pneumonia

Figure 1.

Cortical areas investigated in the present study. (A) Lateral and medial views of a neurotypical brain that illustrates the 5 cortical BAs analyzed in the present study: BA46 (medial prefrontal cortex), BA22 (superior temporal cortex), BA 4 (primary motor cortex), BA3a (primary somatosensory cortex), BA24 (anterior cingulate cortex). (B) Large box indicating the approximate size and location of where the tissue blocks were dissected for each respective BA, which is represented by darkly shaded region in each coronal slab. Smaller box illustrating an example of a 3 mm wide region that encompasses all cortical layers, from where cell counts were performed for each BA in this study. (C) Nissl-stained sections of B24, BA46, BA4, BA3, and BA22 were used to confirm the cytoarchitectural region matched the von Economo description of each respective cortical area. Dark hash marks denote the boundaries between layers I, II-III, VI-VI, and white matter (WM). Scale bar = 100 μm.

Immunostaining

We performed triple enzymatic immunostaining on postmortem human tissue, using a previously validated protocol previously described (Hashemi et al. 2018; Juarez and Martínez-Cerdeño 2023) to stain for 3 different cortical interneuron populations: CB (brown), CR (blue), and PV (pink). Briefly, 14-μm tissue was dehydrated in successive baths of ethanol (50–100%) and with DIVA (DV2004 LX, MX,Biocare medical, United States of America) in decloaking chamber (Biocare medical, United States of America) at 110°C for 6 min, followed by blocking with 10% donkey based blocking solution (10% Donkey Serum+0.3% 100X-Triton in TBS). Tissue was incubated at room temperature (RT) overnight in a sequential order with primary antibodies against CB (day 1), CR (day 2), and PV (day 3). Tissue was then incubated with secondary antibodies conjugated with biotin or alkaline phosphatase (AP) for 2 h at RT and signal was achieved using a combination of both horseradish peroxidase (HRP) and AP enzymatic-based reactions. Color for each cell population was generated using the insoluble chromogens: Vector Blue (blue), 3,3’-Diaminobenzidine (DAB, brown), and Vector VIP (pink). Immunostained tissues were dehydrated in successive baths of 50–100% ethanol, followed by 15 min in xylene solution, mounted and cover-slipped with Permount. Primary antibodies used for our quantification experiments were: monoclonal mouse anti-CB-D28k (1:250, Swant300, Switzerland), polyclonal rabbit anti-CR (1:250, Swant 7697, Switzerland), and monoclonal mouse anti-PV (1:250, Swant 235, Switzerland). Secondary antibodies included donkey anti-mouse conjugated with biotin and amplified with Avidin-Biotin complex and developed with diaminobenzene (DAB) or VectorVIP (all from Vector, United States of America) for CB and PV detection on days 1 and 3 of the protocol, respectively. Donkey anti-rabbit conjugated with AP and developed with Vector Blue substrate (Vector) was used for CR detection on day 2 of the protocol. An adjacent 14 μm section from the same subjects was also used to perform a Nissl stain as previously described (Falcone et al. 2021). Nissl staining was done to confirm that the cytoarchitectural properties of a tissue region we selected matched exactly the Von Economo descriptions for each of the respective BAs being studied (Fig. 1).

Cytoarchitecture

Cytoarchitectural characteristics of each BA studied is depicted in Fig. 1C and described below (Von Economo 2009). The granular frontal area (F_DΔ) encapsulates the entire anterior one-third of the frontal lobe, including the dorsolateral prefrontal cortex (BA46), and is characterized by a distinctly granular appearance with thick granular layers (II and IV). This part of the cortex diminishes in thickness, as does the size of layer III and V pyramidal neurons. Functional roles in the area are responsible for certain intellectual functions which include attention, will, and emotion.

The Giant Pyramidal Precentral Area (F_Ay) encapsulates the largest component of the precentral gyrus, including primary motor cortex (BA4). Granule cells in layers II and IV become very small in density, and cell size is most consistent with medium to large pyramidal neurons. This results in an abnormally thick pyramidal layer that spans what would normally correspond layers II-V, making the granular layers not perceivable. Functionally, this area corresponds to the electromotor zone, or a prototype of the primary motor cortex, and is the source of origin of the largest part of the pyramidal tract and other descending tracts.

The cingulate agranular anterior limbic area (L_A2) encapsulates the dorsal and external wall of the limbic gyrus and includes the anterior cingulate cortex (BA24). Layer II is not as notable in this area an in its place, there are many triangular and small pyramidal cells of the upper zone of layer III. This same similarity is seen in layer IV, where there is a robust presence of layer V pyramidal cells in its place. Functionally, this area plays an important role in emotional regulation that includes motivation and decision making.

The postcentral giant pyramidal area PA1 (P_B1) covers the floor of the central sulcus of Rolando and rolls over into the lower segment of the anterior wall of the postcentral gyrus, which encapsulates primary somatosensory cortex (BA3), subarea 1, of the primary somatosensory cortex. Layer V is reduced in thickness and increased with the presence of giant cells of Betz and 2 very distinct granular layers rich in cells. Functionally, this is the primary area responsible for processing sensory information from the body.

The anterior temporal superior area (T_A2) encompasses the major part of the superior temporal gyri, including temporal cortex (BA22). Notable architectural features include a thin layer 2 not dense in cells. Medium sized pyramidal cells are arranged into wider sized vertical columns in Layers III and V. Functionally, this area is important for the understanding of verbal audition, verbal cognition, as well as music intelligence.

Quantification

Upon confirming our areas of interest, we quantified the total number of immunopositive cells (PV⁺, CB⁺, and CR⁺) within a 3-mm wide bin parallel to the ventricular surface and that extended perpendicular from the ventricle through the thickness of the cortical gray matter to include all cortical layers. We used a x60 oil objective on a microscope (Olympus BX61 microscope with a Hamamatsu Camera, a Dell Prevision PWS 690, Intel Xeon CPU Computer with Microsoft Windows XP Professional V.2002 system, and MBF Bioscience StereoInvesigator V.9 Software, MicroBrightField, Williston, Vermont) to count immunopositive cells. CB, CR, or PV positive cells were counted as those that had clear cytoplasmic staining. We excluded a small number of immunopostive cells from our analysis that had morphology consistent with a pyramidal cell or that were immunoreactive for more than 1 marker (<0.01%). We defined the total number of interneurons as equal to the sum of all PV⁺, CB⁺, and CR⁺ interneurons.

Statistical analysis

The aim of the statistical analysis was to compare the total number of each interneuron subtype in each specific BA, between FXS and Neurotypical cases. Between 6 and 7 age-matched FXS and neurotypical cases were analyzed for each BA, with exception of BA 22, where 4 FXS cases and 4 neurotypical were analyzed. Number of interneurons was compared between FXS and neurotypical using a linear mixed effects model with fixed effects for group, cell type, BA, all 2- and 3- way interactions among these variables, age, and sex. The model also included a random intercept for subject. Counts were log transformed prior to analysis. Correlations between numbers of PV cells in different BAs were estimated using Spearman correlations. Significance was set at α < 0.05 and a trending change is indicated by 0.051 < α < 0.099. Analyses were conducted using R version 4.2.1 (2022 June 23), with linear mixed effects modeling conducted using the R package nlme, version 3.1-159.

Results

We quantified the total interneuron cell number in 3 mm wide bins with the base parallel to the ventricular surface, within the 6 layers of the cortex (Fig. 2A1–E1). We defined the total number of interneurons as the sum of CB⁺, CR⁺, and PV⁺ cells within a bin. PV+, CB+, and CR+ cells more densely populated the supragranular layers, relative to the infragranular layers (Fig. 3A1–E1).

Figure 2.

Representative images of cortex from all 5 BAs in fragile X (FXS) and Neurotypical (NT) cases. 500 μm wide cortical strips with triple stain illustrating PV+ (pink), CB+ (brown), and CR+ (blue) cell populations in FXS and NT cases in BA46(A1), BA3(B1), BA4(C1), BA24(D1), and BA22 (E1). Layer II-III higher magnification PV+, CB+, and CR+ images of NT (A2-E2) and FXS cases (A3-E3). PV+ cells are highlighted with white arrowheads and CB+ with black arrowheads. Distinct morphology of CR+ (multipolar- A4, bipolar- B4, double bouquet- C4), CB+ (bipolar- D4), and PV+ (bipolar—E4) subtypes can be appreciated (A4-E4). Scale bar is 100 μm in A1-E1 and 20 μm in A2-E2, A3-E3, and A4-E4.

Figure 3.

Interneuron number in fragile X (FXS) and Neurotypical (NT) cases. Representative reconstructions depicting the anatomical distribution of PV+ (pink), CB+ (brown), and CR+ (blue) interneurons over a 3 mm wide cortical strip within BA 46 (A1), BA3 (B1), BA4 (C1), BA24 (D1), and BA 22 (E1), in FXS and NT cases. Total number of PV+, CB+, and CR+ interneuron changes for BA46 (A2), BA3 (B2), BA4 (C2), BA24 (D2), and BA 22 (E2) are illustrated where circles (O) represent NT cases and triangles (▲) represent FXS cases. One red asterisk indicates P < 0.05, and 2 red asterisks indicate P < 0.001. A blue asterisk indicates 0.05 < P < 0.1.

We found a generalized decrease in the number of PV⁺ interneurons, an occasional decreased in CR⁺ interneurons, and no change in CB⁺ interneurons. Compared with neurotypical cases (Fig. 2A2–E2), the FXS cases displayed a significant decrease in the number of PV⁺ interneurons in the 5 areas (Fig. 2A3–E3). BA24 had the largest change with a 79% decrease (72 ± 9.55 vs 15 ± 4.49, P-value 0.001, Fig. 3D2), BA46 showed a 64% decrease (142 ± 21.3 vs 51 ± 10.9, P-value 0.004, Fig. 3A2), BA4 had a 71% decrease (158 ± 38.5 vs 46 ± 11.8, P-value 0.003, Fig. 3C2), BA22 had a 45% decrease (180 ± 15.1 vs 100 ± 26.8, P-value 0.046, Fig. 3E2), while BA3 had a trending decrease of 40% (164 ± 18.9 in CT vs 99 ± 20.1 in FXS, P-value 0.07, Fig. 3B2). A significant change in the number of CR⁺ interneurons was found in BA22 with a 63% reduction (179 ± 11.9 vs 67 ± 12.4, P-value 0.031, Fig. 3E2) while a trending decrease of 31% (159 ± 28.75 vs 110 ± 39.66, P-value 0.059, Fig. 3B2) was observed in BA3.

We then analyzed whether the reduction in PV⁺ and CR⁺ interneurons was due to changes in the supragranular (I-III) or infragranular (V-VI) cortical layers (Fig. 4). There were significantly fewer PV⁺ interneurons in FXS in both the infragranular and supragranular layers for BA46, BA4, and BA24. Significantly fewer CR⁺ interneurons were in FXS in the supragranular layers of BA22. For BA24, there was a 79% reduction of PV+ interneurons in the supragranular layers (45 ± 6.56 vs 10 ± 4.10, P-value 0.042, Fig. 4D1), and 78% in the infragranular layers (27 ± 5.68 vs 6 ± 2.17, P-value 0.003, Fig. 4D2). In BA4, there was a reduction of 65% in the supragranular layers (83 ± 25.41 vs 29 ± 9.56, P-value 0.01, Fig. 4C1), and an 80% reduction in the infragranular layers (61 ± 21.56 vs 12 ± 2.74, P-value 0.01, Fig. 4C2). For BA46, there was a reduction of 61% in the supragranular layers (102 ± 15.8 vs 40 ± 8.26, P-value 0.02, Fig. 4A1) and 73% in the infragranular layers (41 ± 9.17 vs 11.14 ± 2.89, P-value 0.004, Fig. 4A2). There was a trending decrease of 66% in the supragranular layers of BA3 (94 ± 22.36 vs 32 ± 6.91, P-value 0.051, Fig. 4B1) and of 46% in the supragranular layers of BA22 (126 ± 14.6 vs 68 ± 16.2, P-value 0.09, Fig. 4E1) but no reductions of PV+ interneurons were observed in the infragranular layers (Fig. 4B2 and E2). For the CR+ population, we detected a significant 74% reduction in the supragranular layers of BA22 (172 ± 12.18 vs 45 ± 14.32, P-value 0.042, Fig. 4E1), but not in the infragranular layers (Fig. 4E2). Correlational analysis showed that age, sex, PMI, and formalin storage time were not significantly associated with the total number of PV⁺, CB⁺, or CR⁺ cells in any of the 5 areas analyzed.

Figure 4.

Interneuron number in the supragranular and infragranular layers in fragile X (FXS) and Neurotypical (NT) cases. Changes in the total number of PV+ (pink), CB+ (brown), and CR+ (blue) interneurons in the supragranular (I-III) and infragranular layers (V-VI) within BA 46 (A1,A2), BA3 (B1,B2), BA4 (C1,C2), BA24 (D1,D2), and BA 22 (E1,E2) are depicted in FXS and NT cases, where circles (O) represent Neurotypical cases and triangles (▲) represent FXS cases. One red asterisk indicates P < 0.05. A blue asterisk indicates 0.05 < P < 0.1.

Discussion

We reported a reduction in the number of PV⁺ interneurons in all the areas of interest, and a reduction in the number of CR⁺ interneurons in BA24 and BA3. In line with our findings, human studies have shown neuropathological alterations of cellular populations in Fragile X brains, including in interneurons. Neuronal quantification in the cingulate and the temporal FXS cortex (BA23 & BA38) did not show differences when compared with age-matched neurotypical cases (Hinton 1991), while a significant reduction of PV⁺ cells was seen in BA3 (Kourdougli et al. 2023). Another study showed mild reduction in the number of CA4 pyramidal cells was detected in the hippocampus together with a focal Purkinje cell loss in the cerebellum (Sabaratnam 2000). Our quantification of PV⁺, CB⁺, and CR⁺ interneurons in human tissue showed a significant decrease of PV⁺ interneurons in BA3, BA4, BA22, BA24, and BA46 cortices. Similar results in BA46 of ASD individuals were previously reported in ASD (Hashemi et al. 2018), where up to 70% reduction of PV⁺ interneurons was found, while the CB⁺ and CR⁺ subtypes remained unchanged. Another study showed that this PV+ cell loss in BA47 in ASD was associated with restrictive repetitive behavioral deficits and stereotypic motor mannerisms (Dufour et al. 2023). Given the neurobiological overlap between FXS and ASD, and the very similar reductions that we saw in BA46 PV+ interneurons between both conditions, it is plausible to suggest that reductions in PV⁺ interneurons may contribute to exacerbating clinical deficits in FXS. Moreover, our findings may suggest that the reduction of PV⁺ and CR⁺ interneurons that we see in various cortical regions of FXS brains, may be also present in ASD brains. However, it is worth noting that no other data on PV⁺ interneurons in other regions, outside of prefrontal cortex, are available for ASD. Further studies are needed to confirm this hypothesis and confirm the relationships between PV+ cell loss and behavioral deficits in FXS individuals.

In addition to human studies, animal models of FXS have also proven defects in the GABAergic pathway. Studies in Fmr1-knock out (KO) mice have provided insight on altered cortical circuit function, including a substantial decrease of PV⁺ interneurons in the somatosensory cortex (Gibson et al. 2008; Kourdougli et al. 2023) and impaired development of PV⁺ interneurons in the auditory cortex (Wen et al. 2018). Independent studies reported a decrease of PV⁺ neurons across the neocortex (Selby et al. 2007; Lee et al. 2019). In their analysis, Selby and colleagues showed a reduction in PV⁺ neurons in the supragranular layer (II/III) and in layer IV in the somatosensory cortex in Fmr1-KO, and an increased distribution of PV⁺ neurons in the deeper layers (V/VI). Also, optical in vivo approach demonstrated that PV⁺ INs and their immature precursors are hypoactive and transiently decoupled from excitatory neurons in postnatal mouse somatosensory cortex (S1) of Fmr1 KO mice and that increasing cortical PV⁺ interneuron activity improves sensory deficits in FXS mice (Kourdougli et al. 2023). In addition, a recent analysis in a FXS human forebrain organoid model observed a significant reduction of GABAergic inhibitory neurons due to dysregulated neural differentiation (Kang et al. 2021); in accordance with similar findings reported in FXS human pluripotent stem cell (hPSC) lines (Zhang et al. 2022).

Locally-projecting GABA interneurons are the primary source of inhibition in the human brain. PV expressing interneurons comprise 40–50% of all inhibitory neurons in neocortex (Gonchar 1997; Tremblay et al. 2016). This group of interneurons is composed of 2 distinct fast-spiking cells subgroups: (i) the chandelier or axo-axonic cells, targeting the initial segment of pyramidal neurons, and (ii) the basket cells, the largest population of interneurons in the neocortex, that make perisomatic terminals on the soma and proximal dendrites of principal cells and other interneurons. Growing evidence suggests that loss of inhibitory interneurons, primarily in the PV+ class, may contribute to the neural basis of ASD development (Hashemi et al. 2018; Contractor et al. 2021). Our group previously studied the human prefrontal cortex of ASD cases showing a reduced number of PV⁺ cells attributable to a decrease in the number of chandelier cells (Ariza et al. 2018; Amina et al. 2021). PV⁺ fast-spiking GABAergic interneurons modulate important aspects of human learning capabilities and motor skills through their regulatory inhibition. Each chandelier cell controls the firing rate of up to 250 pyramidal neurons and participates in cortical executive functions (Markram et al. 2004). The high abundance of mitochondria needed to sustain fast-spiking makes PV⁺ interneurons highly vulnerable to metabolic stress (Hu et al. 2014) and death by apoptosis.

Experimental and theoretical studies demonstrate that neuronal gamma oscillations crucially depend on interneurons (Keeley et al. 2017) and gamma oscillations could potentially be a physiological biomarker for abnormal functioning of PV+ neurons. An imbalance of excitation/inhibition of pyramidal neurons is believed to underlie atypical gamma oscillations in FXS based on preclinical observations (Hu et al. 2014; Goswami et al. 2019). High frequency gamma oscillations generated by PV⁺ interneuron activity are associated with high level cognitive functions including attention, conscious perception, and working memory (Rodriguez et al. 1999; Fries et al. 2001; Pesaran et al. 2002; Melloni et al. 2007; Colgin et al. 2009; Gregoriou et al. 2009). The disruption of cortical synchrony is associated with cognitive impairments in ASD and schizophrenia (Uhlhaas and Singer 2006; Uhlhaas et al. 2010). EEG studies in individuals with neurodevelopmental disorders have abnormal gamma oscillations (Grice et al. 2001; Sohal 2012). Recent clinical studies in FXS have reported disruptions in the functional connectivity between frontal and temporal cortex that interfere with the brain’s ability to implement sensorimotor adaptation for optimal speech development and contribute to chronic and pervasive expressive language deficits (Schmitt et al. 2020). Furthermore, elevations of gamma activity have been significantly associated with core cognitive and neuropsychiatric symptoms in FXS, such as abnormal speech, hyperactivity, irritability, and lethargy/social withdrawal (Pedapati et al. 2022). It is important to highlight that gamma activity seems to vary significantly based on sex in individuals with FXS which should translate into differences found in the brain. Unfortunately, we were unable to assess these differences during our analysis since only 1 female case was included in our FXS group.

We found that a significant reduction in the number of CR⁺ interneurons in BA22. CR⁺ interneurons are composed of double bouquet, bipolar, and Cajal–Retzius cells, with bipolar cells being the most abundant (Jacobowitz and Winsky 1991; Kubota et al. 1994; Gonchar 1997). CR⁺ interneurons have the capability of integrating excitatory and inhibitory inputs across the cortical layers (Gonchar 1999). In the somatosensory cortex, CR⁺ innervate pyramidal cells and PV⁺ interneurons, while in the hippocampus CR⁺ are considered to exclusively target other interneurons (Acsády et al. 1996a, 1996b). Hippocampal CR⁺ interneurons are highly vulnerable in epilepsy (Suckling et al. 2000; André et al. 2001); a reduction in the number of CR⁺ in the hippocampus has been consistently found in animal models and in epileptic human samples (Maglóczky et al. 2000; André et al. 2001; Tang et al. 2006). The relationship between a reduction of CR+ interneurons in and the common presence of epilepsy in FXS should be further investigated.

Overall, we found that in FXS, PV⁺ interneurons are consistently decreased in the cortex while CR⁺ cells are decrease in specific areas of the cortex.

Authors’ contributions

Pablo Juarez (Formal analysis, Investigation, Methodology, Writing—review & editing), Jimena Salcedo (Writing—review & editing), Brett Dufour (Supervision), Verónica Martínez Cerdeño (Conceptualization, Funding acquisition, Supervision, Writing—review & editing).

Funding

This work was funded and supported by the National Institute of Mental Health grant MH094681 and Shriners Hospitals.

Conflict of interest statement: None declared.