L1-deficient knockout mice (Dahme et al., 1997) were bred on the SV129 genetic background and housed at 22°C on a 12 h light/dark cycle with ad libitum access to food and water. The L1 gene is on the X chromosome, thus hemizygous males (L1-/y) are null mutants. L1-/y and wild type (WT) male littermates were analyzed in this study. L1Y¹²²⁹H (L1YH) mice, mutated in the Ankyrin binding motif FIGQY (Buhusi et al., 2008), were maintained by crossing WT C57Bl/6 males to heterozygous L1YH females (C57Bl/6) to yield WT and mutant littermate males because L1-/y males have greatly reduced fertility. WT and L1 mutant mice were analyzed in adulthood (postnatal days P50–150) after the most active period of juvenile spine remodeling (Holtmaat et al., 2005; Holtmaat and Svoboda, 2009).

For immunostaining WT Nex1Cre-ERT2:RCE mice containing a loxP-stop-loxP EGFP allele were induced to express EGFP in pyramidal neurons by intraperitoneal injections of tamoxifen (100 mg/kg) at P10–P13 as described (Agarwal et al., 2012). Postnatal tamoxifen induction in Nex1Cre-ERT2:RCE mice has been shown to achieve cell-specific targeting of postmitotic cortical pyramidal neurons with no detectable targeting of interneurons, oligodendroglia, astrocytes, or non-neural cells (Agarwal et al., 2012). All mice were handled according to the University of North Carolina Institutional Animal Care and Use Committee policies in accordance with NIH guidelines.

Adult mice (50–150) were anesthetized with isoflurane, brains were isolated and processed for Golgi impregnation using the FD Rapid Golgi Stain Kit (FD NeuroTechnologies) as described (Mohan et al., 2019b). Coronal vibratome sections (100 μm) containing the medial PFC (cingulate cortex 1 and 2), M1, and V1 were mounted on gelatin-coated microscope slides. Golgi-labeled neurons were imaged under brightfield illumination using an Olympus Neville microscope by scanning optical sections at 60x and generating minimum intensity projections in FIJI. Dendritic spines were analyzed using Neurolucida software (MBF Bioscience) by investigators blind to genotype as reported (Demyanenko et al., 2014; Mohan et al., 2019a,b). Briefly, spines were traced and quantified on 30 μm segments of the first branch of apical or basal dendrites from confocal z-stack images (6–8 mice/genotype; 30–55 neurons/genotype; 10–30 spines/neuron). The mean spine number per 10 μm of dendritic length (density) was calculated. Mean spine densities/10 μm ± SEM were compared by Mann-Whitney 2-tailed tests (unequal variance, p < 0.05), as normal distributions were not assumed. Spine morphologies are classified as mushroom, stubby, and thin spines based on the relative size of the spine head and neck (Peters and Kaiserman-Abramof, 1970; Peters and Harriman, 1990). The density of spines of each morphological type was scored blind to the observer using Neurolucida software on dendritic segments on Golgi-labeled pyramidal neurons of WT, L1-/y, and L1YH mice (P50) as defined (Peters and Kaiserman-Abramof, 1970; Peters and Harriman, 1990) and described previously (Mohan et al., 2019a). The fraction of each spine type (percent of total spines) was calculated and compared for significant differences by one-factor ANOVA with Tukey post-hoc testing (p < 0.05). The proportion of spine morphological types was similarly analyzed in EGFP-labeled cortical neuron cultures.

Dendritic arborization of Golgi-labeled neurons was assessed by Sholl analysis of image stacks captured at lower magnification (20×/0.5 NA, 1 μm z-series sections). The Sholl center was defined as the midpoint of the cell body at a soma detector sensitivity of 1.5 μm, and the automatic tracing mode was used to seed and trace dendritic arbors. Images in DAT format were subjected to Sholl analysis using Neurolucida Explorer with a starting radius of 10 μm and radius increments of 10 μm. Two factor ANOVA was used to assess significant differences in the mean number of crossings at each distance from the soma with significance set at p < 0.05.

Nex1Cre-ERT2: RCE mice were induced with tamoxifen at P10-P13 and brains were harvested at P20. After transcardial perfusion with 4% PFA, fixed brains were isolated, sectioned, permeabilized, and blocked in 10% normal donkey serum containing 0.3% Triton X-100 in PBS. Sections were stained with antibodies directed against GFP (1:250, Abcam #13970, RRID:AB_300798, host chicken), L1 (1:100, monoclonal antibody 324, Millipore-Sigma #MAB5272, RRID:AB_2133200, host rat) and MAP2 (1:100, Abcam 254143, host mouse) for 2 h at 4°C. AlexaFluor-conjugated secondary antibodies (anti-chicken AF488, anti-rat AF555, anti-mouse AF647 (1:250, Thermofisher) were incubated with sections for 2 h at room temperature. Sections were mounted on Superfrost Plus slides with Prolong Glass mountant, cured for 48–72 h, imaged confocally, and deconvolved using AutoQuant 3 software (Media Cybernetics). Microscopy was performed in the UNC Microscopy Services Laboratory (Dr. Pablo Ariel, Director).

For preparation of mouse cortical lysates, forebrains (P30) were homogenized in RIPA buffer and centrifuged at 16,000× g for 10 min. The supernatant was retained, and protein concentration was determined by BCA. For immunoprecipitation, lysates (1 mg) were precleared for 30 min at 4°C using Protein A/G Sepharose beads. Precleared lysates (equal amounts of protein) were incubated with 2.5 μg nonimmune IgG (NIgG) or L1 monoclonal antibody 2C2 (Abcam #ab24345) together with 1.25 μg L1 monoclonal antibody 5G3 (BD/Pharmingen #554273) for 2 h on ice. Protein A/G Sepharose beads were added for 30 min before washing with RIPA buffer. Samples were subjected to SDS-PAGE (6%) and transferred to nitrocellulose. Membranes were blocked in TBST containing 5% nonfat dried milk and incubated overnight with monoclonal antibodies (1:1,000) against Ankyrin B (ThermoFisher #33-3700, RRID:AB_2533115), washed, and incubated with HRP-secondary antibodies (1:5,000) for 1 h. Blots were developed using Western Bright ECL Substrate (Advansta) and exposed to film for times yielding a linear response of the signal.

Cortical neuron cultures were prepared from the forebrains of WT and L1YH embryos at E15.5 as described (Demyanenko et al., 2014; Mohan et al., 2019a). At DIV11 cells were transfected with pCAG-IRES-EGFP to aid in visualizing and quantifying dendritic spines. At DIV14 cultures were treated with purified human Fc or recombinant mouse Sema3F-Fc or Sema3B-Fc fusion proteins (R&D Systems) at 5 nM for 30 min as reported (Demyanenko et al., 2014). Cultures were fixed with 4% paraformaldehyde, quenched with 0.1 M glycine, permeabilized with 0.1% Triton X-100, and blocked with 10% donkey serum. Cells were incubated with chicken anti-GFP and AlexaFluor AF488-conjugated goat anti-chicken secondary antibodies (1:500), washed, and mounted. At least 10 images of apical dendrites of EGFP-labeled pyramidal neurons were captured per condition. Confocal z-stacks were obtained using 0.2 μm optical sections of field size 64.02 × 64.02 μm with a 40× oil objective and 2.4× digital zoom, and deconvolved using Autoquant 3 software (Media Cybernetics). Spines from maximum intensity projections were traced and scored blind to the observer using Neurolucida software. Mean spine densities (no./10 μm ± SEM) were calculated and compared by 2-factor ANOVA followed by Tukey’s post-hoc testing (p < 0.05).

All experiments were designed to provide sufficient power (80%–90%) to discriminate significant differences (p < 0.05) in means between independent controls and experimental subjects as described (Dupont and Plummer, 1990). The type I error probability associated tests of the null hypothesis was set at 0.05.

To investigate a potential role for L1 in dendritic spine regulation we focused primarily on pyramidal neurons in layer 2/3 of the medial prefrontal cortex (PFC, primary cingulate area) due to its importance in cognitive functions (Yizhar, 2012). In addition, motor and sensory cortical areas in prominent pyramidal cell layers were assessed for a broader role: layer 5 of the primary motor (M1) and layer 4 of the visual cortex (V1). Brains of WT and hemizygous L1-null male mice (L1-/y) were sparsely labeled by Golgi-Cox impregnation in early adulthood (P50). Examination of Golgi-impregnated WT and L1-/y cerebral cortices showed typical pyramidal neurons in each cortical area with well-developed, branched apical dendrites that reached layer I, as well as basal dendrites extending from the soma. Only in M1, layer 5 of mutant mice were a minor number of apical dendrites laterally oriented as previously described (Demyanenko et al., 1999), but these were not analyzed. Spine densities on apical dendrites were found to be significantly increased in PFC (layer 2/3), M1 (layer 5), and V1 (layer 4) in L1-/y mice compared to pyramidal neurons of WT mice in each cortical area (2-tailed Mann-Whitney test, p < 0.05; Figures 1A–C). In contrast, spine density on basal dendrites of L1-/y pyramidal neurons did not differ from WT in any cortical area (p > 0.05; Figures 1E–G). Apical dendrites are known to differ from basal dendrites in receiving different synaptic inputs and exhibiting distinct synaptic plasticity functions (Brzdak et al., 2019), potentially affecting spine density. Additional representative images of Golgi-labeled dendritic spines from WT and L1-/y mice are shown in Supplementary Figures 1A,B.

Pyramidal neurons in layer 2/3 of the PFC were further analyzed to determine if loss of L1 affected spine morphology or dendritic arborization. Dendritic protrusions acquire different morphologies classified as mushroom, stubby, and thin spines based on the relative size of the spine, head, and neck (Peters and Kaiserman-Abramof, 1970). Spine morphologies are dynamically interchangeable, and comprise a continuum from thin spines, which can have excitable synapses with postsynaptic densities, to mushroom spines with mature synaptic functions including neurotransmission (Bhatt et al., 2009; Holtmaat and Svoboda, 2009; Berry and Nedivi, 2017). Spines exhibiting mushroom, stubby, or thin morphology were scored on dendrites of Golgi-labeled pyramidal neurons in PFC, M1, and V1 of WT and L1-/y mice (P50). L1-/y pyramidal neurons exhibited no significant differences compared to WT in the fraction of total spines of each spine morphological type on apical or basal dendrites in the PFC (Figures 1D,H), or in M1 and V1 (Supplementary Figure 2).

In summary, an increase in spine density was observed on apical but not basal dendrites of pyramidal neurons in the PFC and other cortical areas (M1, V1) of L1-/y mice compared to WT.

To determine if the Ankyrin binding site on L1 was required for dendritic spine regulation we focused on layer 2/3 pyramidal neurons in the PFC of L1YH knock-in mice, in which a histidine substitution for tyrosine at position 1229 causes deficiency in binding the actin-spectrin scaffold protein Ankyrin (Buhusi et al., 2008). Pyramidal neurons in WT and homozygous L1YH adult mice were sparsely labeled by Golgi-Cox impregnation and examined in layer 2/3 of PFC. WT and L1YH pyramidal neurons displayed normal morphology and distribution (Figure 2A). Spine densities on apical dendrites were found to be significantly increased in L1YH PFC compared to WT (Figures 2B,C). However, spine density on basal dendrites of L1YH pyramidal neurons did not differ from WT (Figures 2B,D). To investigate whether the L1YH mutation affected spine morphology, the fraction of spines with mushroom, stubby, or thin morphology was quantified on apical and basal dendrites of Golgi-labeled pyramidal neurons in PFC layer 2/3 of WT and L1YH mice. No significant differences in spine morphology were observed between genotypes (Figures 2E,F). Additional representative images of Golgi-labeled spines on apical and basal dendrites are shown in Supplementary Figure 1C.

To assess whether an L1 deletion or mutation of the L1-Ankyrin binding motif affected dendritic arborization, Sholl analysis was performed on the entire dendritic tree of Golgi-labeled pyramidal neurons in PFC layer 2/3 of L1YH and L1-/y mice compared to WT. Arborization of apical and basal dendrites cannot be reliably analyzed at increasing distances from the soma of pyramidal neurons in the cortex due to intercrossing of branches, thus total dendritic arborization was analyzed. Dendritic arborization measured by Sholl analysis was not significantly different from WT in either L1YH (2-factor ANOVA, p = 0.95) or L1-/y PFC (p = 0.12; Figures 2G,H).

In summary, increased spine density was observed on apical but not basal dendrites of pyramidal neurons in PFC layer 2/3 of L1YH compared to WT mice. These results suggest that L1 and its ability to recruit Ankyrin play important roles in limiting spines’ number on apical dendrites of cortical pyramidal neurons.

L1 is localized on axons and apical dendrites of pyramidal neurons in the mouse visual cortex at embryonic and postnatal stages, declining in adulthood (Demyanenko et al., 1999). Similarly, in cultures of neuron-induced human embryonic stem cells, L1 localized initially to all neurites, then became restricted to axons upon maturation (Patzke et al., 2016). To determine if L1 was present on spines in the postnatal PFC in vivo, immunofluorescence staining of L1 was carried out on brain sections of WT Nex1Cre-ERT2: RCE mice (P20). Tamoxifen exposure of these mice at P10–P13 induces the expression of EGFP in postmitotic, postmigratory pyramidal neurons, enabling visualization of spines, as well as dendrites, axons, and soma as described (Mohan et al., 2019a). L1 immunolabeling was observed on spine heads in layer 2/3 pyramidal neurons of the PFC, and in a patchy distribution on dendrites identified by immunostaining for the somatodendritic marker MAP2 (Figures 3A,B). To evaluate L1 association with Ankyrin at the FIGQY motif, L1 was immunoprecipitated from cortical lysates of WT and L1YH mice (P30), then immunoblotted for the ubiquitously expressed Ankyrin B isoform of 220 kDa molecular weight (AnkB 220). L1 co-immunoprecipitated with AnkB 220 from WT but not L1YH cortical lysates (Figure 3C). This result was in agreement with co-immunoprecipitation of L1 and AnkB 220 from WT but not L1YH lysates of mouse superior colliculus (P8; Buhusi et al., 2008). The L1YH mutation did not alter AnkB 220 stability, as equal amounts of protein from WT and L1YH cortical lysates (inputs) showed equivalent levels of AnkB 220 protein on immunoblots (Figure 3C).

To investigate whether L1-FIGQY interactions were required for Sema3B- or Sema3F-mediated spine retraction, cortical neuron cultures were prepared from forebrains of WT and L1YH embryos (E15.5) and cultured for 14 days in vitro (DIV) as described (Demyanenko et al., 2014). On DIV11 cells were transfected with pCAG-IRES-EGFP to enhance spine visualization, then treated on DIV14 with Sema3F-Fc, Sema3B-Fc or control Fc proteins (5 nM) for 30 min (Mohan et al., 2019a, b). Spine densities on EGFP-labeled dendrites were quantified, and mean spine densities compared (Figures 3D,E). As shown previously Sema3B-Fc and Sema3F-Fc induced spine retraction on WT neurons, decreasing mean spine density to a significant extent. Spine density was not significantly different in WT and L1YH control cultures treated with Fc protein (2-factor ANOVA with Tukey post-hoc testing, p = 0.99). In L1YH neuronal cultures Sema3B-Fc and Sema3F-Fc also induced a significant degree of spine retraction compared to Fc-treated L1YH neurons (*p = 0.01, p = 0.001, respectively). Spine densities of L1YH neurons treated with Sema3B-Fc (*p > 0.99) or Sema3F-Fc (*p = 0.91) were not significantly different from similarly treated WT neurons. Spine morphology of WT and L1YH cortical neurons was quantified in Fc-treated cultures to determine if mutation of the L1 Ankyrin binding motif altered the proportion of spine types in vitro. There were no significant differences in the fraction of mushroom, stubby, or thin spines relative to total spines in cultured WT neurons (0.33, 0.25, 0.42, respectively compared to L1YH neurons (0.35, 0.25, 0.39; p > 0.05). The relative proportion of spine types in vitro differed somewhat from that in vivo, where mature mushroom spines were more predominant. Cultures contain a diversity of neuronal types that may be altered in their state of differentiation and lack factors present in vivo, which likely influence spine morphology. In summary, these results supported the interpretation that Sema3B- and Sema3F-induced spine retraction is not substantially mediated by L1 interactions at FIGQY in vitro. It should be noted that L1-null (L1-/y) neuronal cultures were not assayed, because breeding requires mating WT males with heterozygous L1 females, yielding litters with a low percentage of L1-/y male embryos.