Pten heterozygosity restores neuronal morphology in fragile X syndrome mice.

Genetic studies of hippocampal granule neuron development have been used to elucidate cellular functions of Pten and Fmr1. While mutations in each gene cause neurodevelopmental disorders such as autism and fragile X syndrome, how Pten and Fmr1 function alone or together during normal development is not known. Moreover, Pten mRNA is bound by the fragile X mental retardation protein (FMRP) RNA binding protein, but how this physical interaction impinges on phosphatase and tensin homolog protein (PTEN) expression is not known. To understand the interaction of PTEN and FMRP, we investigated the dentate gyrus granule neuron development in Pten and Fmr1 knockout (KO) mice. Interestingly, heterozygosity of Pten restored Fmr1 KO cellular phenotypes, including dendritic arborization, and spine density, while PTEN protein expression was significantly increased in Fmr1 KO animals. However, complete deletion of both Pten and Fmr1 resulted in a dramatic increase in dendritic length, spine density, and spine length. In addition, overexpression of PTEN in Fmr1 KO Pten heterozygous background reduced dendritic length, arborization, spine density, and spine length including pS6 levels. Our findings suggest that PTEN levels are negatively regulated by FMRP, and some Fmr1 KO phenotypes are caused by dysregulation of PTEN protein. These observations provide evidence for the genetic interaction of PTEN and FMRP and a possible mechanistic basis for the pathogenesis of Fmr1-related fragile X neurodevelopmental disorders.

Neurons undergo remarkable structural complexification during development. Developmental disorders like fragile X syndrome (FXS) and some forms of autism spectrum disorder (ASD) are caused by mutations that alter the morphological development of neurons. PTEN is a tumor suppressor gene that encodes for phosphatase and tensin homolog protein (PTEN) and FMR1 is a translational regulator that encodes for fragile X mental retardation protein (FMRP) (1–6). Loss of Pten in rodents causes macrocephaly, seizures, impaired social behaviors, and sensory hypersensitivity (7–11). short hairpin RNA-mediated knockdown of Pten led to an increase in the neuron soma size, dendrite arbor growth, dendritic shaft diameter, and dendritic spine density (12, 13). Cre-mediated Pten knockout (KO) demonstrates that neuronal hypertrophy is accompanied by an increase in inappropriate excitatory synapses and neuronal hyperactivity (5, 14).

FMRP is an RNA binding protein regulating the localization, stability, and translation of a large number of RNAs which are critical for neuronal development (15–17). Unstable CGG trinucleotide repeats (>200) in the 5′-untranslated regions of the FMR1 gene causes hypermethylation and transcriptional silencing resulting in the loss of FMRP expression causing FXS (2, 18–20). In general, animal models and humans with FXS display an increased dendritic spine density with a greater number of filopodial protrusions compared to controls (21). While both PTEN and FMRP loss appear to result in an increased number of dendritic “spines,” PTEN loss results in an increase in the number of stable spines and electrophysiologically defined synapses while FMRP loss results in an increase in unstable filopodial spines and a decrease in functional excitatory synapses.

Both Pten and Fmr1 genes regulate the target of rapamycin (mTOR) signaling network (16, 22–25). FMRP regulates the expression of proteins in the mTORC1 pathway and disruption of mTORC1 signaling has been suggested to give rise to the FXS phenotype (2, 26). For example, overactivation of p70S6 kinase 1 (S6K1) in a FXS mouse model prompts mTORC1 signaling in hippocampal and cortical synapses (6, 27). By contrast, reduction of S6K1 activity in Fmr1 KO mice rectifies the molecular signaling, restores spine morphology, and ASD-like behaviors (1, 28). Tsc2 heterozygous mutations in the Fmr1 KO background rescues the synaptic and behavioral impairments of FXS (29). These observations suggest that cellular pathways in which PTEN and FMRP function may share multiple points of intersection, raising the possibility that ASD severity is modulated by interactions between genetic variants of some of these genes.

To understand the functional role of both PTEN and FMRP in neuronal development, and their genetic interactions, we investigated the dentate gyrus granule neuron development in Pten and Fmr1 KO animals. In the Fmr1 KO background, we introduced Pten heterozygous or homozygous loss, and examined neuron morphological development including dendritic arborization, spine density, and spine length. We show that heterozygous Pten loss could restore granule neuron cellular phenotypes in Fmr1 KO mice. We found increased PTEN expression in the Fmr1 KO mouse that is restored to normal levels by Pten heterozygosity. In addition, overexpression of PTEN in Fmr1 KO Pten heterozygous background reduce dendritic arborization, length, spine density, spine length, and levels of pS6 in cells expressing excess PTEN. Thus, our genetic ablation study reveals FMR1 regulates PTEN expression and that partial genetic loss-of-function in Pten counteracts the up-regulation of PTEN protein in mouse models of FXS.

Results

Loss of FMRP Increases Total PTEN Protein Expression.

Previous work has examined the specific RNA targets bound by the FMRP RNA binding protein (2, 30). We therefore hypothesized that the loss of FMRP affects the expression levels of PTEN protein. We used Pten and Fmr1 germline KO mice derived from B6;129P2-Pten/Mmjax (for PTEN) and B6.129P2-Fmr1tm1Cgr/J (for FMRP) to harvest the whole hippocampal tissue on P67. As expected, PtenHet mice showed a significant reduction in PTEN protein levels (Fig. 1; 50% decrease to wild type; n = 6 for wild type; n = 5 for PtenHet; P = 0.001). Interestingly, PtenHet hippocampal tissue was found to have increased FMRP expression compared to wild-type mice (40% increase to wild type; n = 6 wild type; n = 5 PtenHet; P = 0.001, one-way ANOVA) (Fig. 1B and Dataset S1). Conversely, the hippocampi lysate of Fmr1KO also showed a significant increase in the expression of PTEN protein levels (80% increase relative to wild type; n = 6 for wild type; n = 6 for Fmr1KO; P ≤ 0.0001) (Fig. 1 and Dataset S1). Interestingly, this increase in PTEN protein levels was restored to wild-type levels when Pten gene dosage was reduced to half in Fmr1KO background (n = 6 for wild type; n = 4 for PtenHet/Fmr1KO; P = 0.0581) (Fig. 1 and Dataset S1). These observations indicate that PTEN protein levels are negatively regulated, possibly by FMRP based translational control. While FMRP protein levels are also increased in a PtenHet tissue, the mechanisms through which PTEN may negatively regulate FMRP levels are likely to be indirect.

Fig. 1.

PTEN protein expression is increased in Fmr1 KO animals. (A) Representative Western blot images of whole hippocampal lysate from Fmr1KO, PtenHet, and PtenHet/Fmr1KO mice probed for PTEN and FMRP. (B and C) Quantitative graphs initially normalized to actin and then matched relative protein expression in the wild-type mice. **P < 0.01, ***P < 0.001, ****P < 0.0001, and ns for nonsignificance (P > 0.05) calculated using one-way ANOVA with Tukey multiple comparison post hoc test.

Deletion of Pten or Fmr1 Genes Increase Dendritic Growth of Granule Neurons.

To investigate if Pten and Fmr1 genes interact to regulate dendritic growth, we examined the dentate gyrus granule neuron development in Pten and Fmr1 animals. Retroviruses specifically infect dividing neurons allowing for cre-mediated deletion of floxed alleles and birth-dating of neurons (5). A mixture of Cre+ (pRubiC-T2A-Cre; mCherry) and Cre− (pRubi; green fluorescent protein [GFP]) viruses were injected into the dentate gyrus of Pten (Pten heterozygous/Fmr1 hemizygous) and Ptenflx/flx/Fmr1flx/y (Pten homozygous/Fmr1 hemizygous referred as Pten/Fmr1 DKO) on postnatal day 7 (P7) and granule neurons cellular phenotypes were analyzed at 60 day post injection (60 DPI) (Fig. 2 ). With this system, wild-type cells are labeled by retrovirus expressing GFP (Cre−/Control) and KO cells by retrovirus expressing mCherry (Cre+) (Fig. 2 ). We also analyzed Fmr1flx/y (Fmr1 hemizygous), Ptenflx/wt (Pten heterozygous), and Ptenflx/flx (Pten homozygous referred as Pten KO) (Fig. 2 ) using the same experimental approach.

Fig. 2.

Genetic deletion of Pten and Fmr1 using retrovirus-mediated Cre expression alters granule neuron morphology. (A) Schematic shows the retroviral coinjection of pRubi and pRubiC-T2A-Cre into hippocampal dentate gyrus on P7, and the representative confocal low magnification (10×) image of dentate gyrus expressing GFP and mCherry in granule cells. (Scale bar, 70 µm.) (B) Schematic shows the experimental timelines: Pups were injected on P7 and all the histological analyses were performed 60 DPI (P67). (C and D) Representative (C) high-resolution confocal images and (D) three-dimensional reconstructed 60 DPI granule neurons from Fmr1flx/y, Ptenflx/wt, Ptenflx/wt/Fmr1flx/y, Ptenflx/flx, and Ptenflx/flx/Fmr1flx/y mice. (Scale bars in C, 50 µm.) Throughout this paper, the wild type/control (Cre−) and knockout neurons (Cre+) are represented in green (GFP) and red color (mCherry), respectively. (E) Dot plots with each dot representing an individual neuron showing the quantification of the total dendritic length in raw values and (F) normalized values for Fmr1flx/y, Ptenflx/wt, and Ptenflx/wt/Fmr1flx/y and (G and H) Ptenflx/flx (Pten KO) and Ptenflx/flx/Fmr1flx/y (Pten/Fmr1 DKO) animals. (E and F) Pten heterozygous (Ptenflx/wt) and Fmr1 KOs (Fmr1flx/y) showed a significant increase in the total dendritic length compared to Cre− controls while the Ptenflx/wt/Fmr1flx/y do not. (G and H) Both Pten KO and Pten/Fmr1 DKO neurons showed increased dendritic length than Cre− controls. Data are represented as the mean ± SEM. **P < 0.01, ****P < 0.0001 and ns for nonsignificance (P > 0.05). Refer to for mean ± SEM, n, and P values.

The single-gene KOs of Fmr1 (Fmr1flx/y) and Pten (Ptenflx/wt and Ptenflx/flx) showed increased granule neuron total dendritic length (Fig. 2 ). These observations are consistent with previously described studies (5, 14, 31–33). Compared to single (Fmr1flx/y, Ptenflx/wt, and Ptenflx/flx) or double KO (Ptenflx/flx/Fmr1flx/y) of Pten and Fmr1 gene deletion, gross level morphological defects of the granule neurons are ameliorated in Pten heterozygous/Fmr1 KO combination (Ptenflx/wt/Fmr1flx/y) (Fig. 2 ). This suggested that Fmr1 mutant phenotypes were being caused, at least in part, by the expression of PTEN. Therefore, we quantified the cellular phenotypes of the dentate gyrus granule neurons such as the total dendritic length, dendritic arborization, spine density, and spine length in single and combination KOs of Pten and Fmr1.

Pten Heterozygous, Fmr1 Hemizygous KO Neurons Display Wild-Type Dendritic Arborization.

In 60 DPI animals, the overall dendritic length and its processes revealed an increase in Fmr1flx/y KO neurons (Cre−: 1,165.41 ± 30.09, Cre+: 1,301.92 ± 33.74; P < 0.01). Ptenflx/wt KO neurons also showed a significant increase in the total dendritic length compared to their matched Cre− controls (Cre−: 1,325.09 ± 34.94, Cre+: 1,413.42 ± 50.30; P < 0.0001). Pten heterozygosity in Fmr1 KO background (Ptenflx/wt/Fmr1flx/y) led to a correction in the dendritic length, with the mean dendritic length of KO granule neurons indistinguishable from Cre− control cells (Cre−: 1,394.04 ± 30.7, Cre+: 1,407.20 ± 26.06; P = 0.71) (Fig. 2; ; and Dataset S2). The observed variations in the Cre negative raw values may be due to different genetic backgrounds or animal to animal variability, thus normalizing the KO data with its respective in-tissue wild-type control neurons demonstrate the increased arborization in Pten heterozygotes, Fmr1 hemizygotes and rescue of the single mutant phenotypes when the two gene deletions are combined (Fig. 2).

We observed a significant increase in the total dendritic length of Ptenflx/flx KO neurons (Cre−: 1,230.47 ± 31.51, Cre+: 1,686.12 ± 49.08; P < 0.0001) (Fig. 2 and Dataset S2). Similar to Pten KO, Pten and Fmr1 DKO neurons also showed a significant increase in the dendritic length (Cre−: 1,343.50 ± 57.74, Cre+: 1,927.29 ± 63.27; P < 0.0001) (Fig. 2 ; ; and Dataset S2). To investigate the combinational effect of heterozygous Pten deletion in a hemizygous Fmr1 KO background on the dendritic arborization we performed the Sholl of length and the Sholl of intersection analyses (Fig. 3; ; and Datasets S3 and S7). The Sholl of length analysis between controls and KO neurons within the experimental genotypes showed 1) a significant increase in the dendritic arbor at the radial distance of 170 µm to 210 µm in Fmr1flx/y KO neurons (P < 0.05), suggesting a possible increase in the number of secondary and tertiary branches closer to the soma (Fig. 3; ; and Dataset S3); 2) the Ptenflx/wt mice did not display a significant increase at a specific distance from soma but do exhibit a trend of increase at every distance indicating no variation in the branching pattern (Fig. 3; ; and Dataset S3); and 3) reduced Sholl of length in Ptenflx/wt/Fmr1flx/y KO neurons compared to Cre− controls at the radial distance of 190 µm and 200 µm from the soma (P = 0.002) suggests a decreased number of secondary and tertiary branching compared to control, which may result in reduced arborization of KO neurons (Fig. 3; ; and Dataset S3).

Fig. 3.

Pten KO and Pten/Fmr1 DKO increased dendritic arborization of granule neuron. Sholl of length analysis of (A) Fmr1flx/y, (B) Ptenflx/wt, and (C) Ptenflx/wt/Fmr1flx/y, (D) Ptenflx/flx, and (E) Ptenflx/flx/Fmr1flx/y KO neurons with their respective Cre− controls. Data are represented as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 calculated using two-way ANOVA with Bonferroni multiple comparison post hoc test.

Using the Sholl of intersection analysis, we examined the number of intersection points in each concentric radial junction per 10 μm outwards from the soma. The Sholl of intersection analysis performed between the Cre− and Cre+ cells showed 1) a significant increase in the number of intersections at the radial junction 150 µm to 180 µm in Fmr1flx/y KO neurons indicating an increased neuronal intersection closer to the soma that leads to increased number of dendritic branching points ( and Dataset S7) and 2) the absence of significant change at a given distance from the soma in Ptenflx/wt and Ptenflx/wt/Fmr1flx/y KO neurons compared to their respective controls ( and Dataset S7). However, Ptenflx/wt had more intersections overall than the control while the Ptenflx/wt/Fmr1flx/y displayed a trend of fewer intersections. Taken together, our data suggest that compared to single gene deletion of Pten or Fmr1, deletion of these two gene products in combination is associated with a less complex and more wild-type dendritic arbor.

Complete Loss of Pten and Fmr1 Increases Dendritic Arborization.

We next examined the effect of genetic deletion of both Pten and Fmr1 on the granule neuron dendritic arborization and branching (Fig. 3 ; ; and Datasets S3 and S7). Differences in the Sholl of length between KO and control neurons of Ptenflx/flx and Ptenflx/flx/Fmr1flx/y demonstrate the functional requirement of both Pten and Fmr1 in the dendritic development of the granule neurons. As predicted, Ptenflx/flx KO neurons showed an increase in the dendritic arbor at the radial distance from 20 µm to 60 µm (P < 0.0001, two-way ANOVA) and 110 µm to 190 µm (P < 0.0001) (Fig. 3; and Dataset S7), whereas Ptenflx/flx/Fmr1flx/y KO neurons displayed a significant increase in Sholl length at the radial distance from 20 µm to 70 µm (P < 0.0001) and 100 µm to 180 µm than Cre− controls (P < 0.0001) (Fig. 3E; and Dataset S7).

KO neurons of both Ptenflx/flx and Ptenflx/flx/Fmr1flx/y displayed an increased number of the intersections from the radial distance 10 µm to 160 µm (P < 0.0001, two-way ANOVA) ( and Dataset S7). Based on the above Sholl of length and the Sholl of intersection analysis, we suggest that both Pten and Fmr1 contribute toward the growth and maturation of granule neurons dendritic morphology during the critical period of development in which activity-dependent mechanisms may play a crucial role in the overall dendritic architecture (34, 35). Our data suggest that, either complete lack of Pten or loss of both Pten and Fmr1 gene during critical developmental time window might lead to abnormal cellular phenotypes.

Pten Heterozygosity Restores the Aberrant Dendritic Spines in Fmr1 KO Neurons.

Consistent with the prior studies, compared to Cre− neurons Fmr1flx/y KO neurons showed increased spine density (Cre−: 1.67 ± 0.04, Cre+: 2.36 ± 0.05; P < 0.0001) (Fig. 4 ; ; and Dataset S4) and spine length (Cre−: 0.90 ± 0.01, Cre+: 1.00 ± 0.05; P < 0.004) (Fig. 4 ; ; and Dataset S4). Ptenflx/wt also showed increased spine density (Cre−: 1.44 ± 0.03, Cre+: 1.72 ± 0.04; P < 0.02) (Fig. 4 and Dataset S4) while the spine length remains comparable (Cre−: 0.91 ± 0.1, Cre+: 0.94 ± 0.1; P = 0.35) (Fig. 4 ; ; and Dataset S4). Remarkably, the observed increase in the spine density and spine length of Fmr1flx/y KO neurons was corrected to Cre− control levels in Ptenflx/wt/Fmr1flx/y KO neurons (Cre−: 1.62 ± 0.05, Cre+: 1.70 ± 0.001; P = 0.65) (Fig. 4 ; ; and Dataset S4) (Fig. 4 and Dataset S4). Our data suggest that introducing Pten heterozygous mutation under Fmr1 KO background can neutralize dendritic spine abnormalities associated with deficit of Fmr1.

Fig. 4.

Pten heterozygosity restores the dendritic spine density in Fmr1 KO neurons. (A) Grayscale (Top) and pseudocolored (Bottom; ImageJ fire lookup table) confocal dendritic spine images of all experimental genotypes: Fmr1flx/y, Ptenflx/wt, Ptenflx/wt/Fmr1flx/y, Ptenflx/flx, and Ptenflx/flx/Fmr1flx/y. (B and C) Spine density quantification (B) raw data and (C) normalized values for Fmr1flx/y, Ptenflx/wt, and Ptenflx/wt/Fmr1flx/y granule neurons. Ptenflx/wt/Fmr1flx/y KO neurons showed reduced spine density compared to Fmr1flx/y KO neurons. (D and E) Quantification of the spine density for Pten flx/flx and Pten flx/flx/Fmr1flx/y presented in (D) raw values and (E) normalized data. Compared to Cre− controls, both Pten KO and Pten/Fmr1 DKO neurons showed increased spine density. Data are represented in mean ± SEM. **P < 0.01, ***P < 0.001, ****P < 0.0001, and ns for nonsignificance (P > 0.05). Refer to for mean spine density ± SEM, P values, number of cells, and animals. (F–I) Quantification of the spine length presented with (F and G) raw dataset and (H and I) normalized values. Dots plots in F and G correspond to Fmr1flx/y, Ptenflx/wt, and Ptenflx/wt/Fmr1flx/y and in H and I to Ptenflx/flx and Ptenflx/flx/Fmr1flx/y animals. (H and I) Pten KO and Pten/Fmr1 DKO cells displayed increased spine length than Cre− neurons. Data are represented in mean ± SEM. ***P < 0.001, ****P < 0.0001, and ns for nonsignificance (P > 0.05). Refer to for mean spine length ± SEM, P values, number of cells, and animals. (Scale bars, 3.72 µm.)

Deletion of Both Pten and Fmr1 Enhances the Abnormalities in Dendritic Spine Phenotypes.

We next analyzed cells with both copies of Pten deleted. We found that complete loss of Pten significantly increases the dendritic spine density (Cre−: 1.42 ± 0.04, Cre+: 2.74 ± 0.10; P < 0.0001) (Fig. 4 and Dataset S4) and the spine length (P < 0.0001) in Ptenflx/flx Cre+ neurons compared to Cre− cells (Fig. 4 ; ; and Dataset S4). Similarly, Pten and Fmr1 DKO also showed a massive increase in the spine density (Cre−: 1.35 ± 0.03, Cre+: 4.05 ± 0.12; P < 0.0001) (Fig. 4 and Dataset S4) and spine length (Fig. 4 ; ; and Dataset S4). The observed increase in the spine density and the spine length of Ptenflx/flx/Fmr1flx/y KO neurons is significant compared to Ptenflx/flx KO neurons (P < 0.0001; Fig. 4 ; ; and Dataset S4). Given the functional requirement of both Pten and Fmr1 in the regulation of synaptic structure and function (36), our data suggest that inhibiting both Pten and Fmr1 gene function simultaneously has a deleterious effect on the neurons resulting in abnormal dendritic spine phenotypes.

Overexpression of PTEN in Ptenflx/wt/Fmr1flx/y Cre+ Neurons Reduces Dendritic Length and Arborization.

To investigate the effect of PTEN overexpression in Ptenflx/wt/Fmr1flx/y neurons, we delivered a mixture of retroviruses, pRubiC-T2A-Cre (mCherry, referred as Cre+) and pRubi-GFP-Pten (Cre+ + PTEN), on P7 into the dentate gyrus, and the cellular phenotypes were studied at 60 DPI (Fig. 5). In this experiment, the Ptenflx/wt/Fmr1flx/y neurons overexpressing PTEN are GFP and mCherry positive (Cre+ + PTEN), while all neurons expressing Cre only are labeled with mCherry (Cre+) (Fig. 5 ). Overexpression of GFP-PTEN in Ptenflx/wt/Fmr1flx/y Cre+ neurons resulted in reduction in the dendritic length (Ptenflx/wt/Fmr1flx/y Cre+: 1,591 ± 3.60, Cre+ + PTEN: 1,025 ± 3.31; P < 0.0006) (Fig. 5; ; and Dataset S5). Further, the Sholl of length analysis between Ptenflx/wt/Fmr1flx/y Cre+ and Cre+ + PTEN cells revealed significant decrease in the dendritic arbor at the radial distance of 140 µm to 220 µm (P < 0.05) in the PTEN overexpressing neurons (Fig. 5; ; and Dataset S5). Additionally, PTEN overexpressing neurons showed reduced Sholl of intersection values at the radius distance of 120 µm to 210 µm (P < 0.05) suggesting decrease in the number of secondary and tertiary branch points resulting in the overall reduction in dendritic length and arborization (Fig. 5; and Dataset S5).

Fig. 5.

PTEN overexpression in Fmr1 KO reduce dendritic arborization and length. (A) The representative confocal image of granule neurons in Ptenflx/wt/Fmr1flx/y mice expressing pRubiC-T2A-Cre (Cre+) and pRubi-GFP-PTEN (Cre+ + PTEN) in hippocampal dentate gyrus. (Scale bar, 100 µm.) (B) A 60-DPI three-dimensional reconstruction of Cre+ and Cre+ + PTEN granule neurons from Ptenflx/wt/Fmr1flx/y animals. (C) Dot plots showing the quantification of total dendritic length (raw values). Each dot corresponds to an individual neuron. (D and E) Sholl of (D) length and (E) intersection between Cre+ and Cre+ + PTEN neurons. Data are represented in mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 calculated using the t test (two-tailed) and two-way ANOVA with Bonferroni multiple comparison post hoc test.

PTEN Overexpression Decreases the Spine Density, Spine Length, and pS6 Levels in Fmr1 KO Neurons.

We next investigated the effect of PTEN overexpression by examining phosphorylation of protein targets downstream of mTOR by analyzing pS6 (S235, 236) intensity. The brain tissue slices from Ptenflx/wt/Fmr1flx/y injected with pRubiC-T2A-Cre and pRubi-GFP-PTEN were stained for pS6 and visualized in combination with mCherry and GFP-Pten. PTEN overexpression showed reduced pS6 intensity compared to Ptenflx/wt/Fmr1flx/y Cre+ neurons (Fig. 6 , P < 0.001, and Dataset S6A) suggesting that PTEN overexpression in Fmr1 KO background might be controlling the hyperactivity of mTOR signaling.

Fig. 6.

Overexpression of PTEN decreases spine density, spine length, and pS6 levels in Fmr1 KO. (A) Representative confocal image of Cre+ (red) and Cre+ + PTEN (green) cell body in Ptenflx/wt/Fmr1flx/y mice with pS6 staining (gray). (Scale bar, 20 µm.) (B) Dot plot showing the quantification of pS6 intensity. (C) Pseudocolored (Left; ImageJ fire lookup table) and grayscale (Right) confocal dendritic spine images of Cre+ and Cre+ + PTEN in Ptenflx/wt/Fmr1flx/y. (Scale bars, 3.72 µm.) (D and E) Dot plots showing (D) spine density and (E) spine length quantification. Data are represented in mean ± SEM. *P < 0.05, and ****P < 0.0001 calculated using t test (two-tailed).

In Fmr1 KO background we found that Cre+ + PTEN neurons showed a significant decrease in spine density (Ptenflx/wt/Fmr1flx/y Cre+: 1.37 ± 0.04, Cre+ + PTEN: 1.09 ± 0.03; P < 0.0001) (Fig. 6 and Dataset S6B) and spine length (Ptenflx/wt/Fmr1flx/y Cre+: 1.06 ± 0.01, Cre+ + PTEN: 1.01 ± 0.01; P < 0.01) (Fig. 6 and Dataset S6B) compared to Ptenflx/wt Cre+ neurons. Together our data suggest that PTEN overexpression in Fmr1 KO background, might be beneficial in terms of cellular morphology.

Discussion

Individuals with ASD or FXS display significant alteration in neuronal and synaptic structure, and such alterations are thought to be a causative factor. Genetic studies have removed key regulators in the PI3K/Akt/mTOR pathway to examine the effect on FXS pathogenesis (37). FMRP has been shown to interact directly with Pten mRNA via sequencing of immunoprecipitated crosslinked mRNAs (2). Here, we introduced a Pten heterozygous mutation in Fmr1flx/y mice and found that it restored dendritic length, neuronal arborization, excess branching, and corrected the increased spine density observed independently in Pten and Fmr1 deficient mice. We found that PTEN protein levels were increased in Fmr1 KO hippocampus and the increased PTEN levels were restored in the Pten heterozygous, Fmr1 KO mice (Fig. 1). Paradoxically, complete loss of either Pten or Fmr1 results in increased dendritic complexity and dendritic spines while the heterozygous reduction of Pten in the Fmr1 KO results in neurons with wild-type arborization and spines. This suggests both the presence and proper balance of these two proteins is important since elimination of both does not simply cancel out opposing functions.

Protein synthesis regulation through the mTOR network is essential for synaptic organization and development in the central nervous system (38), and both FMRP and PTEN proteins are known interactors of one or more proteins in the mTOR network (1, 4–6, 33, 39–41). Mutations in a negative regulator of mTORC1 such as Tsc1, Tsc2, Pten, Mnk1, Mnk2, and Fmr1 generally cause abnormal protein levels leading to a common root cause of FXS and a subset of ASDs (22, 24, 42). mTOR signaling pathway and its machinery can localize to synaptic sites where they control synaptic plasticity via regulation of local protein synthesis in dendrites and actin polymerization in spines (43, 44). In a Pten mouse model, the absence of PTEN leads to an increase in phosphatidylinositol (3,4,5)-triphosphate (PIP3) levels causing an excess in phosphorylated AKT and hyperactivation of the mTOR pathway (45). In the mouse model of FXS, levels of FMRP regulate mTOR signaling via PI3K upstream of the mTORC1 complex (4, 6). However, loss of FMRP also triggers hyperactivation of PI3K that causes increased protein synthesis (4). The deletion of either PTEN or FMRP exhibit hyperactivation of mTOR signaling in hippocampal neurons and overactivation of downstream mTOR targets (6, 27).

Pten deleted mice display increased levels of FMRP and phosphorylated FMRP (40). Here, we show an increased level of total PTEN protein expression in 8-wk-old Fmr1 KO mice (Fig. 1 and Dataset S1). Heterozygous Pten deletion in the Fmr1 KO mouse background rectified the increased PTEN protein levels observed in Fmr1 KO mice. In addition, we observed that FMRP levels increased in Pten heterozygous mice, indicating a complex reciprocal regulation. Although, FMRP may directly regulate Pten mRNA translation, FMRP may also indirectly regulate protein synthesis through its interaction with the mTOR network. How loss of PTEN leads to up-regulation of FMRP is less clear, but dysregulation of mTOR via PI3K and AKT may contribute to changes in FMRP levels.

FMRP localizes to both axons and synapses within the dendrites. This localization is critical for regulating neuronal morphology via FMRP transport (46, 47). Interestingly, in Drosophila, the dFmr1 mutants exhibit enlarged presynaptic terminals, excessive axonal branching, and increased architectural complexity (48, 49). Twenty-eight-day-old motor neurons of Fmr1 KO mice displayed increased amounts of dendritic arborization closer to the cell body with an increase in the number of branch points (50). We found that Fmr1 KO neurons display increased dendritic length (Fig. 2 and Dataset S2). Numerous studies conducted on Fmr1 KO mice display excess dendritic spines with a higher proportion of immature filopodial like spines in the different brain regions (1, 31, 36, 51, 52). Variation in brain region, age, fixation protocol, and staining technique may account for the contradictions in the reported studies (53). Here, we report a substantial increase in spine density and length at 60 DPI Fmr1flx/y KO granule neurons (Fig. 4 and Dataset S4).

The Pten gene plays a critical role in the development of dentate granule neurons of the hippocampus, thus determining the synaptic structure and function (36). Previous work from our laboratory and several others have reported that loss of Pten leads to ASD and alteration in neuronal morphology and physiological function (3, 8, 54, 55). Conditional Pten KO using different Cre-recombinase delivery methods in mice led to neuronal hypertrophy, seizures, increased dendritic arborization, and increased spine density (7, 8, 56, 57). Pten haploinsufficiency leads to increased dendritic arborization with ASD-related social and behavioral deficits (58). Consistent with previous studies, we found homozygous Pten KO neurons displayed a dramatic increase in dendritic length, neuronal arborization, and increased dendritic spine density as well as spine length (5, 14, 33, 56, 59). Our study also examined morphological features in Ptenflx/wt (Pten heterozygous) mice showing increases in dendritic length, arbor complexity and spine density (Figs. 2 , 3, and 4 ; ; and Datasets S2, S3, S4, and S7). While the changes are less severe than those seen in the homozygous KOs, the ability to examine Cre− control cells in the same tissue as Cre+ Pten heterozygous cells eliminates variability resulting from animal experience and genetic background allowing us to consistently detect subtle changes.

Our data suggest that Fmr1 and Pten genes interact in way that is more complex than simply opposing one another’s function. It is unclear how Pten heterozygous deletion in the Fmr1 hemizygote results in reduced arborization and spine density when Pten loss alone increases arborization and spine density. Further, exogenous overexpression of PTEN also decreases arborization and spine density in the Ptenflx/wt/Fmr1flx/y Cre+ + PTEN cells (Figs. 5 and 6 and Datasets S5 and S6 A and B). We speculate that FMRP may alter the functionality of endogenously expressed PTEN. However, when overexpressed as a cDNA any alteration in function resulting from mechanisms targeting untranslated regions of the PTEN transcript would no longer be intact. Because FMRP has many mRNA targets, not all the cellular defects would be expected to be ameliorated by restoring PTEN levels. However, the degree of restoration we observe underscores the strength of biological impact that Pten loss has. Finally, it is also notable that complete loss of PTEN in Ptenflx/flx cells does not further ameliorate Fmr1 associated defects (Figs. 2 and 4 and Datasets S2 and S4), consistent with the idea that proper levels of PTEN are critical for cellular development as opposed to PTEN simply antagonizing FMRP functions.

Several independent studies have demonstrated the loss of PTEN or FMRP is a likely mechanism for dysregulation of PI3K, AKT, mTOR, and p70S6K1 including cap-dependent protein translational molecules causing neurodevelopmental or neurological disease (1, 4, 6, 39–41). Our study reveals that in Fmr1 KO mouse, genetic reduction of Pten gene dosage by half corrects excessive PTEN protein levels and rectifies dendritic length, complexity, and spine density. This suggests the amelioration observed in Ptenflx/wt/Fmr1flx/y is mediated by reduction of PTEN levels. This interpretation is strengthened by the observation that complete genetic ablation of both Pten and Fmr1 displayed a greater increase in dendritic length, neuronal complexity, spine density, and length. In a Ptenflx/wt/Fmr1flx/y background increasing PTEN levels by overexpression further reduced multiple morphological deficits (Figs. 5 and 6), consistent with the idea that Fmr1 mutant phenotypes are sensitive to PTEN protein levels. It will be of interest to determine the mechanistic basis through which these two ASD genes interact and of great clinical relevance to better understand how Pten dosage modulates phenotypes caused by loss of Fmr1.

Materials and Methods

Animals.

Procedures involving mice were approved and performed following the Dartmouth College’s Institutional Animal Care and Use Committee and the Association for the Assessment and Accreditation of Laboratory Animal Care Review Board. Conditional male KO mice (cre-lox system) used in this study are Fmr1flx/y (Fmr1 hemizygous), Ptenflx/wt (Pten heterozygous), Ptenflx/wt/Fmr1flx/y (Pten heterozygous and Fmr1 hemizygous), Ptenflx/flx (Pten homozygous), and Ptenflx/flx/Fmr1flx/y (Pten/Fmr1 double homozygous). The parental lines C57BL/6J wild-type controls and B6.129S4-Ptentm1Hwu/J homozygous (Pten floxed) were from the Jackson Laboratory, and Fmr1flx/y hemizygous/homozygous (Fmr1 floxed) was a generous gift from Prof. David Nelson, Baylor College of Medicine, Houston, TX. Both Pten floxed and Fmr1 floxed animals that were on a C57BL/6J genetic background were used to generate the above-mentioned experimental genotypes.

For whole-animal KO alleles, the Jackson Laboratory mice C57BL/6J, B6;129P2-Pten/Mmjax (also known as mPTEN) and B6.129P2-Fmr1/J (Fmr1 KO mice) were bred to generate male PtenHet, Fmr1KO, and PtenHet/Fmr1KO genotypes. All animals were housed with their littermates on a 12-h regular light/dark cycle with chow and water provided ad libitum. Data presented in this study were collected and analyzed with the experimenters blinded to genotypes.

Viral Packaging.

Retrovirus containing GFP only (retrovirus with internal ubiquitin promoter; pRubi; Addgene 66696), pRubi-mCherry-T2A-Cre recombinase (pRubi-C-T2A-Cre; Addgene 66692), and pRubi-GFP-Pten were packaged and produced as previously described (60).

Stereotaxic Brain Injections.

Using previously described stereotactic protocols, the retrovirus was injected into anesthetized male mice brains (61). Similarly, to test the effect of PTEN overexpression in granule neuron, we sterotaxically injected pRubi-C-T2A-Cre (mCherry) and pRubi-GFP-Pten to Ptenflx/wt/Fmr1flx/y (Pten heterozygous and Fmr1 hemizygous) mice at P7 as previously described (61). Briefly, on P7, isoflurane-anesthetized pups were injected with 2 μL of replicative-defective retrovirus into the hippocampal dentate gyrus on both the hemisphere. From lambda, the injection coordinates are y = ±1.55, x = ±1.30, and z = −2.3 to −2.0. At 60 DPI, P67 animals were anesthetized using 2,2,2-tribromoethanol (Sigma Aldrich) and transcardially perfused with 4% paraformaldehyde in phosphate-buffered saline (1× PBS) containing 4% sucrose (12).

Immunohistochemistry.

After perfusion, the hippocampal coronal brain slices (50 µm and 150 µm in thickness) were rinsed with 0.4% PBS containing Triton-X (0.4% PBTx) and blocked with 10% donor horse serum in PBS for 1 h at room temperature (RT). Later, the samples were incubated with primary antibodies for 48 h at 4 °C. Samples were washed with 0.4% PBTx three times, 15 min each, and labeled with fluorophore-tagged secondary antibodies for 48 h at 4 °C. Thus, immunolabeled samples were washed with 0.4% PBTx and mounted in Vectashield (H-1000, Vector Laboratories). Primary and secondary antibodies used in this study were rabbit anti-mCherry (1:5,000, Abcam, #167453), chicken anti-GFP (1:3,000, Abcam, #ab13970), rabbit anti-pS6 (1:200, Cell Signaling, #4858), anti-rabbit Cy3 peroxidase AffiniPure donkey anti-rabbit IgG (1:200, Jackson ImmunoResearch Labs, #711-035-152; RRID: AB_10015282), Alexa Fluor 488 conjugated goat anti-chicken IgY (1:200, Jackson ImmunoResearch Labs, #103-545-155; RRID: AB 2337390), and Alexa Fluor 647 conjugated goat anti-rabbit IgG (1:200, Jackson ImmunoResearch Labs, #111-605-144).

Confocal Image Analysis.

Confocal images acquired using a Zeiss LSM 510 laser scanning microscope (Zeiss) were analyzed for the total dendritic arborization, the dendritic spine density, and the spine length and pS6 intensity levels as previously described (5, 14, 59). For dendritic arborization analysis, cells on the suprapyramidal blade of the dentate gyrus were imaged with a 2-µm interval and 1,024 × 1,024-pixel resolution using a 20× objective. Similarly, for the PTEN overexpressed animals, the images were acquired with a 2-µm interval and 1,024 × 1,024-pixel resolution using a 40× objective. Three-dimensional reconstruction of the whole granule neuron arbors was achieved by semiautomated tracing in Neurolucida 360 software (MBF Biosciences). For dendritic spine analysis, images were acquired with a 0.5-µm interval and 1,024 × 1,024-pixel resolution using a 63× objective and 3× optical zoom. After image deconvolution (Deconvolution Lab) (62), the quantification of dendritic spine density was performed with Neuron Studio software (14, 59, 63). For pS6 fluorescence intensity measurement, z-stacks of supra and infrapyramidal blades of dentate gyrus were acquired at 40× with a 1.3-µm oil immersion lens with 0.7× zoom, at 1,024 × 1,024 resolution with a 2-µm z-step. The Align Master plugin of ImageJ/Fiji (NIH) was utilized to account for variation in staining intensity throughout the z-stack. Further, p-S6 fluorescence intensity was normalized to the average fluorescence intensity of the control cells in the same z-stack.

Western Blot.

On P67, all the experimental genotypes, PtenHet, Fmr1KO, and PtenHet/Fmr1KO, were euthanized, and the hippocampus was dissected. Samples were quickly rinsed in 1× PBS, flash frozen in liquid nitrogen, and then stored at −80 °C for further usage. Whole hippocampal lysate was prepared using a homogenization solution. After adding 600 to 700 µL of lysis buffer (1% sodium dodecyl sulfate [SDS] and 1 mM dithiothreitol in PBS), samples were homogenized with pestles and boiled at 95 °C for 5 min; samples were then cooled to RT and mixed with freshly prepared N-ethylmaleimide (30 mM final) (Thermo Fisher Scientific). Samples were then passed through a G25 needle two to three times, vortexed, and centrifuged for 15 min at 20,000 relative centrifugal force (RFC), and the supernatant was collected into new tubes. Just before SDS-polyacrylamide gel electrophoresis (PAGE), the protein sample was mixed 1:1 with 2× DB (250 mM Tris⋅HCl, pH 6.8, 2 mM ethylenediaminetetraacetic acid, 20% glycerol, 0.8% SDS, 0.02% bromophenol blue, 1,000 mM NaCl, and 4 M urea). Proteins were separated in a 10% SDS-PAGE gel and transferred to a polyvinylidene difluoride membrane (Millipore). The membrane was blocked with 1× TBS-T (20 mM Tris⋅HCl, pH 7.6, 136 mM NaCl, and 0.1% Tween-20) containing 3% bovine serum albumin prepared in 1× TBS-T (MP Biomedicals, LLC) for 1 h at RT. Membranes were then incubated overnight at 4 °C with the primary antibodies diluted in blocking solution. The primary antibodies used were as follows: rabbit anti-PTEN (1:1,000, Cell Signaling Technologies, #9559), rabbit anti-FMRP (1:1,000, Cell Signaling Technologies, #4317), anti-alpha actin (1:5,000, Sigma #T9026), and anti-tubulin (1:10,000, Sigma #T9026). Following the incubation in primary antibodies, membranes were washed with 1× TBS-T buffer (three times each for 5 min) and incubated with fluorescently tagged LI-COR antibodies anti-rabbit IRDye 800CW (1:15,000, #926-32211) and anti-mouse IRDye 680RD (1:15,000, #926-68070) for 1 h at 23 °C. Signals were detected by chemiluminescence or using a LI-COR fluorescent imager (64). The Western blot images were quantified using ImageJ. Equal-sized boxes were drawn over the blots and a plot profile was generated. The area under the curve was then calculated for each band and plotted relative to the loading control.

Statistical Analysis.

Sample numbers and sizes are based on similar previously published studies. To perform the statistical analysis, we used Stata 13 (StataCorp) and Prism 9 (GraphPad). For all the experimental parameters, mean ± SEM values were derived using the average values of Cre− and Cre+ neurons. Total dendritic length, dendritic spine density, spine length, and pS6 data were analyzed as previously described (60, 65). Significance was determined by comparing cre+ and cre− cells clustered within animals using a mixed-effect linear regression model (65). Sholl of length and Sholl of intersection analysis was performed with a two-way repeated-measure ANOVA with Bonferroni multiple comparison. One-way ANOVA with the Tukey multiple comparison test was used for quantification of the Western blot data. A two-tailed t test with Welch corrections was used for quantification of PTEN overexpression total dendritic length, dendritic spine density, and spine length.