To understand the mechanisms by which Sema3F and Sema3A exert distinct effects on the development of layer V cortical pyramidal neurons, we investigated the localization and function of their obligate co-receptors, Nrp2 and Nrp1, respectively. First, we assessed the localization of Nrp2 and Nrp1 on the surface of COS-7 cells owing to the well-articulated plasma membrane and large circumference of these cells in culture. COS-7 cells transfected with flag-tagged wild-type Nrp2 or Nrp1 expression plasmids were subjected to surface staining with a flag antibody to visualize Nrp localization on the plasma membrane. We observed that Nrp2 is distributed on the COS-7 cell surface in a clustered pattern consisting of numerous discrete puncta (Figure 1A), whereas Nrp1 is evenly distributed over the entire plasma membrane with little evidence of clustering (Figure 1B). Particle analysis (see Materials and methods) provides a quantitative assessment of protein clustering and reveals a marked difference between the membrane localization of Nrp2 and Nrp1 (Figure 1C). This robust Nrp2 clustering is reminiscent of the cell surface clustering displayed by palmitoylated proteins in COS-7 cells (Webb et al., 2000).

To explore the potential role of palmitoylation in Nrp2 surface localization, we used a pharmacological approach employing 2-bromopalmitate, which is a specific and irreversible inhibitor of protein palmitoylation (Jennings et al., 2009; Resh, 2006). We treated COS-7 cells expressing flag-tagged wild-type Nrp2 (Nrp2^WT) with 2-bromopalmitate, or a control solution, and visualized cell surface Nrp2. Following bath application of 2-bromopalmitate, Nrp2 was markedly redistributed compared to the control, displaying a diffuse localization on the cell surface similar to Nrp1, which is localized diffusely following either control or 2-bromopalmitate treatment (Figure 1D and E). 2-Bromopalmitate-induced Nrp2 dispersion is consistent with previous reports showing 2-bromopalmitate-induced diffusion of otherwise clustered palmitoylated proteins in non-neuronal cell lines (Webb et al., 2000) and in neurons (El-Husseini et al., 2002). These results show that Nrp2 and Nrp1 exhibit distinct cell surface compartmental localization and that protein palmitoylation is required for Nrp2 cell surface clustering in heterologous cells in vitro.

Given the significant Nrp2 cell surface distribution perturbations observed upon 2-bromopalmitate bath application to COS-7 cells, we next examined whether Nrps are palmitoylated. Both Nrp2 and Nrp1 harbor cysteine residues in their transmembrane and membrane-proximal (juxtamembrane) domains (Figure 2A, in red). Nrp2 also has a lone cysteine residue, C897 (Figure 2A, in red) carboxy-terminal (C-terminal) to the transmembrane domain and two adjacent C-terminal cysteine residues (Figure 2A, in blue). These Nrp2 and Nrp1 cysteine residues are all highly phylogenetically conserved among vertebrate species (Figure 2—figure supplement 1). The location and conservation of these cysteine residues, coupled with the presence of palmitoyl acceptor amino acid residues predicted using CSS-Palm software (Ren et al., 2008), suggested that these cysteine residues are palmitoyl acceptor sites. To explore this further, we used the acyl-biotin exchange (ABE) assay, a well-established biochemical approach for the detection of palmitoylation on thiol groups of cysteines (S-palmitoylation) (Drisdel and Green, 2004; Kang et al., 2008; Roth et al., 2006; Wan et al., 2007). In this assay, we used the postsynaptic density protein 95 (PSD-95) as a positive palmitoylation control because it is palmitoylated (Topinka and Bredt, 1998), and synapse-associated protein 102 (SAP102) as a negative palmitoylation control because it is not (Kang et al., 2008). We observed that both Nrp2 and Nrp1 are palmitoylated in adult mouse forebrain and in embryonic day 14.5 (E14.5) mouse cerebral cortex (Figure 2B), as are both Nrps in cortical neurons 28 days in vitro (DIV) derived from E14.5 embryos (Figure 2C). The detection of Nrp palmitoylation from early embryonic stages (E14.5) to adulthood raises the possibility that palmitoylation of these co-receptors is required throughout neural development and in the adult to regulate Nrp trafficking and function.

We next sought to identify the Nrp2 and Nrp1 cysteine residues that serve as palmitate acceptors. For this analysis, we generated point mutations that replaced individual Nrp2 and Nrp1 cysteines with a serine residue, which cannot be S-palmitoylated. Cysteine-to-serine (CS) point mutant Nrp expression plasmids were transfected into neuroblastoma-2a (Neuro-2a) cells on which the ABE assay was performed. This analysis revealed that Nrp2 is palmitoylated on all transmembrane and membrane-proximal cysteines, albeit to variable extents, as well as on the C-terminal di-cysteine motif (Figure 2D and E). The triple CS (TCS) Nrp2 mutant C878S/C885S/C887S (Nrp2^TCS: substituting serine for the three transmembrane/juxtamembrane cysteines) and the Full CS Nrp2 mutant C878S/C885S/C887S/C897S/C922S/C923S (Nrp2^{Full CS}: substituting all transmembrane and cytoplasmic cysteines) display very little to no palmitoylation, respectively (Figure 2F and G). A similar analysis for Nrp1 revealed that Nrp1 is palmitoylated on cysteines C875 and C883, but apparently not on C881 (Figure 2H and I). These results show that Nrp2 and Nrp1 display similar palmitoylation patterns in their transmembrane and membrane-proximal segments, and that Nrp2 also harbors C-terminal palmitoylation sites that are not present in Nrp1.

Palmitoylation enables proteins to be anchored onto specialized membrane domains, to interact with other proteins, and to shuttle between the plasma membrane and intracellular organelles (Salaun et al., 2010). The compartmentalized distribution of Nrp2 on the cell surface of COS-7 cells in culture, which is palmitoylation-dependent (Figure 1D and E), prompted us to investigate the role of palmitoylation in Nrp2 localization, trafficking, and function in cortical neurons.

First, we addressed the role of palmitoyl acceptor Nrp2 cysteines in Nrp2 surface localization in cell lines. We expressed wild-type or various CS mutant flag-tagged Nrp2 proteins in COS-7 cells and assessed their plasma membrane distribution with surface staining. This assay revealed that localization of Nrp2^C922S/C923S, which lacks the C-terminal di-cysteine motif, is for the most part punctate with little difference from Nrp2^WT, whereas Nrp2^TCS and Nrp2^{Full CS} proteins are profoundly mislocalized (Figure 3—figure supplement 1A–1C), exhibiting a non-patterned distribution over the entire plasma membrane that resembles Nrp1. For our localization analysis experiments, we also generated pHluorin-tagged Nrp2 wild-type and CS mutants. The performance of the same localization assay with pHluorin-tagged Nrp2 wild-type and CS mutants yielded similar results (Figure 3A and B). The diffuse distribution of palmitoylation-deficient Nrp2 CS mutants on the cell surface is reminiscent of Nrp2^WT protein distribution following treatment with 2-bromopalmitate (Figure 1D). Further, these results are in line with the palmitoylation patterns we observed for these same Nrp2 CS mutants (Figure 2D–G).

Second, we examined Nrp2 cell surface localization in cortical neurons in culture. We transfected cortical neurons with plasmids expressing pHluorin-tagged wild-type Nrp2 or various CS mutants and carried out live imaging. The pH-sensitive GFP variant pHluorin fluoresces robustly at neutral pH, and so the tagged protein is readily visualized when localized at the cell surface (see Materials and methods). These live imaging assessments of surface protein localization showed that Nrp2^WT (Figure 3—figure supplement 2A; Video 1) and Nrp2^C922S/C923S (Figure 3—figure supplement 2C) exhibit a prominent punctate localization along the plasma membrane, consistent with their distribution in COS-7 cells. In contrast, Nrp2^TCS apparently forms larger protein clusters (Figure 3—figure supplement 2B), and Nrp2^{Full CS} is profoundly mislocalized displaying a diffuse distribution over all membranes (Figure 3—figure supplement 2D). These pHluorin-tagged Nrp2 plasmids were also tested with live imaging in Neuro-2a cells, with similar results (data not shown).

Given the surface mislocalization of palmitoylation-deficient Nrp2, we next addressed the early intracellular trafficking defects that lead to aberrant Nrp2 delivery to the cell surface. Since the Golgi apparatus is the main intracellular compartment where palmitoylation enzymes are localized and function (Ohno et al., 2006; Rocks et al., 2010), we asked whether Nrp2 localizes to Golgi membranes. We stained primary cortical neurons in culture with antibodies directed against Nrp2 and the cis-Golgi marker Golgi matrix protein 130 (GM130), which showed that endogenous Nrp2 is robustly associated with somatic Golgi and dendritic Golgi cisternae known as Golgi outposts (Figure 3—figure supplement 3A). Further, Golgi isolation experiments from mouse brain revealed Nrp2 enrichment in the GM130-positive Golgi fraction (Figure 3—figure supplement 3B).

Next, we examined whether palmitoylated Nrp2 cysteine residues are required for the association between Nrp2 and the Golgi apparatus by performing immunofluorescence on Nrp2^-/- cortical neurons transfected with either wild-type Nrp2 or various Nrp2 CS mutants. These experiments show that Nrp2^WT and Nrp2^C922S/C923S proteins display robust association with Golgi membranes, whereas the Nrp2^TCS and Nrp2^{Full CS} mutant proteins are significantly deficient in their association with the Golgi apparatus (Figure 3C, D). The aberrant association of TCS and Full CS Nrp2 proteins with the Golgi apparatus could have major effects on their trafficking and plasma membrane insertion.

To investigate the role of palmitoylation in the spatial and temporal dynamics of Nrp2 trafficking, we focused on the role of transmembrane/juxtamembrane cysteines, which have the greatest effect on Nrp2 surface localization and Golgi association, performing fluorescence recovery after photobleaching (FRAP) in cortical neurons expressing exogenous pHluorin-tagged Nrp2^WT or Nrp2^TCS. The fluorescent signal was bleached over a selected region of interest (ROI) and after photobleaching neurons were imaged for at least 15 min to visualize the recovery of fluorescence in the bleached area, which represents protein molecules either derived from nearby plasma membrane compartments or newly inserted in the cell membrane. We observed that Nrp2^TCS protein displays a markedly different trafficking pattern with a higher mobile fraction compared to Nrp2^WT (Figure 3E and F; Videos 2 and 3). This observation is in accordance with other studies reporting higher diffusibility and an altered intracellular localization of depalmitoylated proteins (Miura et al., 2006; Rocks et al., 2010; Rocks et al., 2005).

Taken together, these experiments support the functional importance of transmembrane/juxtamembrane palmitoyl acceptor Nrp2 cysteines in regulating the localization and trafficking of Nrp2 protein across subcellular compartments and the plasma membrane.

Sema3F binds Nrp2, which forms homo-oligomers and interacts with plexinA3 (PlexA3) to form the Sema3F holoreceptor complex (Giger et al., 1998; Janssen et al., 2012). Since cysteine residues and their posttranslational modification can influence protein structure, we asked whether Nrp2 cysteines control the ability of Nrp2 to associate with PlexA3 or affect the ability of Nrp2 to homo-oligomerize.

First, we performed coimmunoprecipitation experiments using myc-tagged PlexA3 and either Nrp2^WT or Nrp2^{Full CS}. These experiments showed that PlexA3 binds Nrp2^{Full CS} to a similar extent as it does to Nrp2^WT (Figure 3—figure supplement 4A and B). Further investigation into the association between palmitoylation-deficient Nrp2 and PlexA3 using immunofluorescence in primary cortical neurons in culture showed that exogenous Nrp2^{Full CS} exhibits strong co-localization with endogenous PlexA3, similar to exogenously expressed Nrp2^WT (Figure 3—figure supplement 4C). Thus, the membrane-spanning and cytoplasmic Nrp2 cysteines are dispensable for the interaction between Nrp2 and PlexA3.

Next, we addressed whether the transmembrane and cytoplasmic Nrp2 cysteines are required for Nrp2 homo-oligomerization. We performed coimmunoprecipitation experiments using flag-tagged Nrp2^WT and pHluorin-tagged Nrp2^WT or Nrp2^{Full CS}. These experiments revealed that pHluorin-Nrp2^{Full CS} associates with flag-Nrp2^WT to a significantly lesser extent than pHuorin-Nrp2^WT does (Figure 3—figure supplement 4D and E). Moreover, coimmunoprecipitation experiments between flag-Nrp2^WT and pHluorin-Nrp2^WT or Nrp2^TCS demonstrated that pHluorin-Nrp2^TCS associates with flag-Nrp2^WT to a lesser extent than pHluorin-Nrp2^WT does (Figure 3—figure supplement 4F and G), although this difference is not as robust as it is with Nrp2^{Full CS}. These findings show that the Nrp2 cysteine residues that we have shown to be palmitoylated and required for Nrp2 localization are also necessary for Nrp2 homo-oligomerization.

Sema3F/Nrp2 signaling regulates pruning of supernumerary dendritic spines on apical dendrites of deep layer cortical pyramidal neurons (Demyanenko et al., 2014; Duncan et al., 2021; Tran et al., 2009). To investigate a potential role for palmitoyl acceptor Nrp2 cysteines in Sema3F/Nrp2-dependent spine constraint, we employed an in vitro assay aimed at rescuing the inability of Sema3F to decrease dendritic spine density in Nrp2 knockout (Nrp2^-/-) cortical neurons by introducing flag-tagged Nrp2^WT or Nrp2^CS mutant expression plasmids subcloned in a pCIG2-ires-EGFP backbone vector. Nrp2^-/- primary cortical neurons were transfected with backbone vector as a control or a Nrp2 expression plasmid. At DIV21, when dendritic spines are well developed, neurons were treated with 5 nM Sema3F-AP or 5 nM AP (control) for 6 hr, followed by EGFP immunofluorescence to visualize neuronal morphology. The rescue ability of the various Nrp2 plasmids was assessed using two criteria: (1) a comparison of dendritic spine density between AP-treated Nrp2^-/- neurons expressing backbone vector (control) and AP-treated neurons expressing each Nrp-2 plasmid; and (2) a comparison of dendritic spine density between Sema3F-AP-treated and control AP-treated Nrp2^-/- cortical neurons expressing a specific (the same) Nrp2 plasmid. If a plasmid is able to rescue the Nrp2 knockout spine phenotype, it should: (1) constrain spine number compared to the backbone vector (control) in the absence of exogenous Sema3F treatment and/or (2) cause a reduction in spine number upon exogenous Sema3F-AP treatment compared to AP treatment leading to dendritic spine pruning that is similar to that imparted by Nrp2^WT protein.

These experiments show that Nrp2^WT and Nrp2^C922S/C923S can rescue the Nrp2 mutant dendritic spine phenotype (Figure 4A–C). This is in accordance with our finding that the C-terminal Nrp2 cysteines regulate Nrp2 subcellular localization to a rather minor extent (Figure 3A–D, Figure 3—figure supplement 1, Figure 3—figure supplement 2). In contrast, Nrp2^C878S and Nrp2^C887S are not capable of constraining spine number, whereas Nrp2^C885S and Nrp2^C897S can partially rescue the Nrp2 knockout dendritic spine phenotype (Figure 4A–C). The relative requirements for Nrp2 palmitoyl acceptor cysteines in Sema3F/Nrp2-dependent dendritic spine pruning are summarized in Figure 4D. Likewise, Nrp2 mutant rescue experiments in cortical neuron cultures with pHluorin-tagged Nrp2^WT or Nrp2^CS mutant expression plasmids including the Nrp2^TCS, Nrp2^C922S/C923S, and Nrp2^{Full CS} mutants, reveal a requirement for transmembrane/juxtamembrane cysteines, but not for the two C-terminal cysteines, in Sema3F/Nrp2-dependent dendritic spine pruning (Figure 4—figure supplement 1A and B).

To investigate whether Sema3F effects on cortical neuron dendritic spines are associated with Nrp2 palmitoylation, we asked whether Sema3F bath application impacts Nrp2 palmitoylation. We treated wild-type cortical neurons with 5 nM Sema3F-AP, Sema3A-AP, or AP (control) and assessed palmitoylation of endogenous Nrp2 with the ABE assay. Interestingly, Sema3F application enhances Nrp2 palmitoylation in deep layer primary cortical neurons in culture (Figure 4E and F). This shows that Sema3F can upregulate Nrp2 palmitoylation, presumably at nearby subcellular domains and within dendritic spines, and it further suggests that Sema3F effects on cortical neurons are mediated by Nrp2 palmitoylation.

Next, we addressed whether Nrp2 cysteines C878, C885, and C887, which are required for rescuing the Nrp2 mutant dendritic spine phenotype in cortical neurons in vitro, are also required for Sema3F/Nrp2-dependent dendritic spine pruning in vivo. We examined the ability of Nrp2^TCS to constrain excess dendritic spines resulting from Nrp2 deletion in the mouse cortex using an in utero electroporation (IUE) rescue paradigm. To overcome breeding challenges, including reduced viability encountered in the Nrp2^-/- mouse line, we crossed homozygous Nrp2 mutant mice with homozygous Nrp2 floxed (Nrp2^F/F) mice and targeted layer V cortical pyramidal neurons using IUE in Nrp2^F/- embryos of pregnant females at E13.5. To assess the feasibility of this approach, we first carried out two control IUE experiments: one with an EGFP-expressing plasmid to visualize neuronal morphology (control) and a second using a combination of three plasmids: an EGFP-expressing plasmid, a Cre-expressing plasmid to excise the floxed Nrp2 allele, and a LSL-tdTomato-expressing plasmid that serves as a reporter of Cre activity. In neurons electroporated with all three plasmids, we expected Cre to excise the floxed Nrp2 allele rendering neurons Nrp2 knockout (Nrp2^-/-) and also to turn on tdTomato expression (Figure 5A). These experiments revealed increased spine density in neurons lacking Nrp2 compared to control (Figure 5B and C), confirming that this approach is a useful assay for rescue analysis of the Nrp2^-/--associated dendritic spine phenotype. Next, to assess the rescue ability Nrp2^WT and Nrp2^TCS proteins in vivo, we used two additional IUE conditions: one with the plasmid combination [Cre+LSL-tdTomato+flag-Nrp2^WT-ires-EGFP] and one with the plasmid combination [Cre+LSL-tdTomato+flag-Nrp2^TCS-ires-EGFP]. Neurons electroporated with the combination expressing Nrp2^WT showed significantly fewer spines compared to Nrp2^-/- neurons that do not express Nrp2 (electroporated with [Cre+LSL-tdTomato+EGFP]). However, neurons electroporated with the combination of plasmids that includes Nrp2^TCS exhibited a spine density similar to that of Nrp2^-/- neurons (Figure 5B and C).

Taken together, both in vitro and in vivo rescue experiments of the Nrp2^-/- dendritic spine density phenotype reveal an essential role for the transmembrane and membrane-proximal Nrp2 cysteines in mediating Sema3F/Nrp2-dependent spine pruning in layer V cortical pyramidal neurons.

To identify the DHHC enzymes that palmitoylate Nrp2, we carried out a Nrp2 palmitoylation screen in HEK293T cells by co-expressing exogenous wild-type Nrp2 and each of the 23 DHHCs (Fukata et al., 2004) and assessing palmitoylation with the ABE assay. Of the 23 DHHCs, ZDHHC15 (zinc finger DHHC domain-containing protein 15, also known as DHHC15) was among the three DHHCs that most strongly enhanced Nrp2 palmitoylation (data not shown). ZDHHC15, encoded by a gene located at the Xq13.3 genetic locus in humans, is known to palmitoylate PSD-95 (Fukata et al., 2004), is strongly expressed in the mouse cerebral cortex (Mejias et al., 2021) and is enriched in Golgi membranes (Ohno et al., 2006). Therefore, we examined the role of ZDHHC15 in Nrp2 palmitoylation and function in cortical neurons.

We employed a loss-of-function approach using a Zdhhc15 knockout (Zdhhc15-KO) mouse line (Mejias et al., 2021), exploring whether Nrp2 is a ZDHHC15 substrate by performing ABE in wild-type and Zdhhc15-KO primary cortical neurons. In Zdhhc15-KO cortical neurons overall Nrp2 palmitoylation is substantially reduced (~50%) compared to that in wild-type neurons (Figure 6A and B). Unlike Nrp2, Nrp1 palmitoylation in Zdhhc15-KO neurons is comparable to that observed in wild-type neurons (Figure 6A and C). To test for DHHC-substrate specificity, we assessed Nrp2 palmitoylation in the previously characterized Zdhhc8 knockout (Zdhhc8^-/-) mouse line (Mukai et al., 2015; Mukai et al., 2008; Mukai et al., 2004), since ZDHHC8 did not enhance Nrp2 palmitoylation in our screen in 293T cells. Nrp2 palmitoylation levels in Zdhhc8^-/- cortical neurons are very close to those observed in wild-type cortical neurons (Figure 6—figure supplement 1A and B) and, therefore, Nrp2 does not appear to be a palmitoyl substrate of ZDHHC8 in these neurons.

Next, we asked whether ZDHHC15 plays a role in Sema3F/Nrp2-dependent dendritic spine pruning. Wild-type and Zdhhc15-KO deep layer primary cortical neurons were cultured, transfected with an EGFP-expressing plasmid, and at DIV21 they were treated with 5 nM Sema3F-AP or 5 nM AP for 6 hr. They were then subjected to EGFP immunofluorescence, imaged and spine density was quantified along the largest dendrite. Unlike wild-type cortical neurons, which displayed a significant reduction in spine density following treatment with Sema3F-AP compared to AP (Figure 6D and E), Zdhhc15-KO neurons failed to exhibit apical dendritic spine pruning following Sema3F-AP treatment (Figure 6D and F).

To address the specificity of these spine morphology effects, we tested the ability of cortical neurons lacking ZDHHC8, which plays important roles in the nervous system (Mukai et al., 2015; Mukai et al., 2008), to exhibit spine pruning in response to Sema3F. Zdhhc8^-/- cortical neuron cultures were transfected with an EGFP expression plasmid, and at DIV21 they were treated with Sema3F-AP or AP for 6 hr followed by EGFP immunofluorescence. Zdhhc8^-/- cortical neurons responded to Sema3F with significant spine retraction (Figure 6—figure supplement 1C and D), similar to wild-type cortical neurons.

Since we observed no effect of Zdhhc15 loss-of-function on Nrp1 palmitoylation, we investigated whether ZDHHC15 is required for Sema3A/Nrp1-dependent basal dendritic elaboration (Gu et al., 2003). Wild-type and Zdhhc15-KO cortical neuron cultures were treated at DIV12 with 5 nM Sema3A-AP or 5 nM AP for 6 hr. Next, they were subjected to immunofluorescence with an antibody directed against the dendritic marker microtubule-associated protein 2 (MAP2) to visualize dendritic trees, imaged and subjected to Sholl analysis. Wild-type neurons exhibited enhanced elaboration of perisomatic dendrites in response to Sema3A-AP, compared to the AP control (Figure 6G and H), in line with the effect of Sema3A/Nrp1 signaling on basal dendrite elaboration of deep layer cortical pyramidal neurons in the mouse brain (Gu et al., 2003). Likewise, Zdhhc15-KO cortical neurons responded to Sema3A with enhanced elaboration of their basal dendrite arbors following Sema3A-AP treatment as compared to the AP control (Figure 6G and I).

These biochemical and phenotypic experiments in vitro reveal that ZDHHC15 plays essential roles in Nrp2 palmitoylation and Sema3F/Nrp2-dependent dendritic spine pruning, and that ZDHHC15 is dispensable for Nrp1 palmitoylation and Sema3A/Nrp1-dependent basal dendritic elaboration in primary deep layer cortical neurons. These results underscore the exquisite specificity among PATs for their neuronal substrates.

Since we observe that ZDHHC15 is required for Nrp2 palmitoylation and Sema3F/Nrp2-dependent dendritic spine pruning in deep layer cortical neurons in culture, we asked whether ZDHHC15 is also required for Nrp2-mediated functions in cortical pyramidal neurons in the mouse neocortex. We examined the dendritic spine density on the apical dendrite of layer V cortical pyramidal neurons in Zdhhc15-KO and wild-type mice using two different labeling techniques, a genetic labeling approach using the Thy1-EGFP-m mouse line that labels layer V cortical pyramidal neurons and the Golgi staining technique for sparse neuronal labeling. Our analysis of apical dendritic spines in wild-type and Zdhhc15-KO layer V cortical pyramidal neurons labeled with the Thy1-EGFP-m line shows that Zdhhc15-KO apical dendrites harbor more dendritic spines compared to wild-type neurons (Figure 7A and B), phenocopying previously observed effects of Nrp2 loss-of-function in this same neuronal population in vivo (Tran et al., 2009). Analysis of Golgi-stained brains yielded similar results (Figure 7—figure supplement 1A and B). Next, given the lack of an effect of ZDHHC15 on Nrp1 palmitoylation and function, we assessed the Sema3A/Nrp1-dependent basal dendritic arbor elaboration in Golgi-stained wild-type and Zdhhc15-KO brains and observed no difference in dendritic arbor complexity between wild-type and Zdhhc15-KO layer V cortical pyramidal neurons (Figure 7C and D).

We next examined whether ZDHHC15 and Nrp2 exert their effects in the same signaling pathway in layer V cortical pyramidal neurons in vivo. Despite potential issues related to haploinsufficiency (dose-dependent effects) and/or random X-inactivation, we performed a genetic interaction experiment involving Nrp2 and Zdhhc15. For this approach we analyzed littermates of three different genotypes: (1) Nrp2^+/-;Zdhhc15^+/+;Thy1-EGFP, (2) Nrp2^+/+;Zdhhc15^+/-;Thy1-EGFP, and (3) Nrp2^+/-;Zdhhc15^+/-;Thy1-EGFP. Our sample size for each genotype was small (n=2 mice/genotype) due to the difficulty in obtaining all four alleles in the three genetic configurations along with Thy1-GFP expression in littermates. Despite the shortcomings of this experiment, we observed that individual neurons in transheterozygous Nrp2 and Zdhhc15 mutant mice display a higher density of dendritic spines on the apical dendrite of deep layer cortical pyramidal neurons compared to single Nrp2 or Zdhhc15 heterozygotes (Figure 7—figure supplement 1C and D). These observations are in line with our biochemical and phenotypic experiments presented here and support the hypothesis that Nrp2 and ZDHHC15 function in the same signaling pathway in vivo.

We also examined whether ZDHHC15 contributes to Nrp2-dependent developmental guidance of fiber tracts, including the anterior commissure (AC) and select cranial nerves (Giger et al., 2000). Analysis of wild-type and Zdhhc15-KO brains using neurofilament staining revealed an intact AC in Zdhhc15-deficient mice (Figure 7—figure supplement 1E), with none of the hallmark AC defects known to be present in Nrp2 mutants (Chen et al., 2000; Giger et al., 2000). Likewise, the examination of Zdhhc15-KO embryos with whole-mount neurofilament staining revealed normal development and guidance of cranial nerves (data not shown).

Taken together, our in vitro and in vivo data suggest a critical role for select palmitoylated Nrp2 cysteines in Nrp2 subcellular localization and function in deep layer cortical neurons. Further, they reveal that PAT-substrate specificity drives diversification of Nrp2 and Nrp1 palmitoyl substrates to control deep layer cortical pyramidal neuron morphology and, presumably, neural circuit assembly.

Animal procedures were carried out in conformity with the policies and guidelines of the Animal Care and Use Committee (ACUC) of the Johns Hopkins University (protocol # MO20M48), which are established according to the US National Research Council’s Guide to the Care and Use of Laboratory Animals and in compliance with the Animal Welfare Act and Public Health Service Policy. Mice were handled with care and every effort was made to minimize suffering. Wild-type C57BL/6J mice and Thy1-EGFP-m transgenic mice were purchased from the Jackson Laboratory. The Zdhhc15 knockout (Zdhhc15-KO) mouse line was generated in Dr Tao Wang’s laboratory (Johns Hopkins University, Baltimore, MD, USA) (Mejias et al., 2021). The Zdhhc8 knockout mouse line was provided to our lab by Dr Joseph Gogos (Columbia University, New York, NY, USA) and has been characterized (Mukai et al., 2015; Mukai et al., 2008; Mukai et al., 2004).

All cell lines used in this study were obtained from a commercial source and meet the characterization, authentication, and safety standards according to the distributor (American Type Culture Collection-ATCC): HEK293T cells (ATCC, Cat no. CRL-11268 RRID:CVCL_1926), COS-7 cells (ATCC, Cat no. CRL-1651, RRID:CVCL_0224), and Neuroblastoma 2a (Neuro-2a) cells (ATCC, Cat no. CCL-131 RRID:CVCL_0470). Neuro-2a cells were authenticated by ATCC using short tandem repeat (STR) profiling. HEK293T cells were authenticated by the Genetic Resources Core Facility (GRCF) of the Johns Hopkins University by STR profiling. Authentication of COS-7 cells was not performed since STR profiling cannot be provided by any commercial facility. However, verification of COS-7 cells was confirmed by morphological examination and stereotypic behavior of these cells in a variety of cell collapse and protein localization assays, as observed previously in the field. Mycoplasma testing for all three cell lines mentioned above was performed at the Johns Hopkins GRCF, and results were negative for Mycoplasma contamination for all cell lines. The cell line authentication STR profile reports and Mycoplasma testing reports are available at this study's data archiving database (https://doi.org/10.7281/T1/OY2X8T). Cells in culture were maintained in culture media consisting of DMEM (Dulbecco’s Modified Eagle Medium), fetal bovine serum (FBS, 10% final), penicillin and streptomycin (50 U/ml), and Glutamax supplement (1× final, Thermo Fisher Scientific), in a humidified incubator at 37°C with 5% CO₂. Cells were plated in six-well dishes for biochemical experiments or in 12-well or 24-well dishes on glass coverslips for immunofluorescence, at the desired confluency. Cells were transfected using Lipofectamine 2000 reagent (Thermo Fisher Scientific), according to the manufacturer’s instructions, and processed 24–48 hr after transfection based on protein expression.

Timed-pregnant female mice were either obtained from external organizations or generated in-house by plug checks. For deep layer cortical neuronal cultures the hemispheres (excluding ventral structures and the olfactory bulb) were dissected out from embryos of both sexes from timed-pregnant mice at E14.5. During dissection the dissected tissue was kept on ice-cold Leibovitz’s L-15 medium (Thermo Fisher Scientific). Tissue was digested in HBSS (Hank’s balanced salt solution) containing 0.1% trypsin, in a 37°C-water bath for 15 min. Next, cortices were washed twice with HBSS containing 10% FBS to inactivate trypsin and dissociated in neuron growth medium containing 10% FBS by gently passing them several times through a glass Pasteur pipette. Dissociated neurons were plated onto six-well dishes for biochemical experiments or onto 12-well or 24-well dishes on glass coverslips for immunofluorescence, at the desired confluency. Next day, medium was replaced by fresh neuron growth medium and thereafter half of the medium was changed every 1 or 2 days. Neuron growth medium consisted of Neurobasal medium (Thermo Fisher Scientific, Gibco) supplemented with 2% B-27 supplement (Gibco), 2 mM Glutamax (Gibco), and 1× penicillin/streptomycin (Gibco). Neuronal cultures were maintained in a humidified incubator at 37°C with 5% CO₂.

Dish/coverslip preparation: Round glass coverslips were treated with nitric acid overnight, next washed with ddH₂O and ethanol, and stored in 95% ethanol. The day of culture, dishes, with or without coverslips, were coated with 0.1 mg/ml poly-D-lysine (Sigma) diluted in ddH₂O at 37°C for at least 3 hr. Before plating, poly-D-lysine was removed and dishes were washed twice with DPBS (Dulbecco’s phosphate-buffered saline [PBS]).

Plasmids were generated with standard cloning techniques. Briefly, polymerase chain reaction-amplified inserts carrying the appropriate restriction sites on the 5′ and 3′ ends were digested and ligated with the backbone vector. N-terminally flag-tagged Nrp1 and Nrp2 expression plasmids were generated in a pCIG2-ires-EGFP backbone vector containing the preprotrypsin signal peptide followed by the flag epitope tag and the Nrp1 or Nrp2 coding sequences in frame with the above elements. N-terminally pHluorin-tagged Nrp2 expression plasmid was generated by subcloning in frame the endogenous signal peptide of Nrp2, the pHluorin coding sequence and the Nrp2 coding sequence (downstream of its signal peptide) in a pCAGGS-ires-dsRED backbone vector. The GFP variant pHluorin displays strong fluorescence (bright green) at neutral pH, which occurs upon receptor surface localization, whereas it does not fluoresce in vesicles (pH <6.0). As such, it provides a robust assay for cell surface protein visualization in live cells, while upon regular EGFP immunofluorescence on permeabilized cells allows total (surface and intracellular) protein visualization. CS point mutants of Nrp1 and Nrp2 were generated by amplification of the wild-type protein with reverse primers harboring the desired mutation(s). All clones were fully sequenced and protein expression was assessed by western blotting in heterologous cells and immunofluorescence in cell lines and cultured neurons. The CS point mutations did not alter the molecular weight of Nrp2 or Nrp1, as evidenced by western blotting.

Transfection of COS-7 cells, 293T cells, and primary cortical neurons was performed with Lipofectamine 2000 (Thermo Fisher Scientific, Cat no. 11668-019) according to the manufacturer’s guide. In brief, the appropriate amounts of DNA and Lipofectamine were added in separate tubes containing Opti-MEM I medium and incubated at room temperature for 5 min. Next, diluted DNA and diluted Lipofectamine were mixed and co-incubated at room temperature for 20 min. The mix was added to cells and after ~5 hr the medium was replaced by fresh medium. Primary cortical neurons were transfected between DIV7 and DIV9. Transfection of Neuro-2a cells was performed with Metafectene Pro (Biontex, Germany) according to the manufacturer’s instructions.

2-Bromopalmitate (or 2-bromohexadecanoic acid, Sigma-Aldrich, Cat no. 238422) was dissolved in dimethyl sulfoxide (Sigma-Aldrich, Cat no. D2650) for making concentrated stock solution. Stock solution was diluted in the appropriate culture medium to a final concentration of 10 μM. Cells were incubated with either 2-bromopalmitate-containing medium or control medium (the same quantity of solvent diluted in culture medium) overnight, in a humidified chamber at 37°C with 5% CO₂. The use of ethanol as a solvent worked equally well. No particular toxicity on cultured cells was observed following incubation.

ABE was performed according to published protocols (Drisdel and Green, 2004; Wan et al., 2007), with slight modifications as reported in the study of Hayashi et al., 2009. Briefly, the workflow of this assay involves four main steps: (1) Tissue is harvested and lysed in lysis buffer containing S-methyl methanethiosulfonate (Sigma, Cat no. 64306) to block free thiol groups (-SH). (2) Incubation with a buffer containing hydroxylamine (NH₂OH, Fisher Scientific, Cat no. 26103) that cleaves cysteine-linked palmitate thioester bonds or with control buffer that instead of hydroxylamine contains Tris. (3) Samples are incubated with a buffer containing biotin-HPDP (Soltec Ventures, Cat no. B106) that attaches to newly exposed cysteine thiols. Both buffers (for hydroxylamine-treated and control-treated samples) contain biotin-HPDP. Between steps 1–3, protein was precipitated by incubation with –20°C-cold 80% acetone at –20°C overnight and protein pellet was washed several times with the same solution before proceeding to next step. (4) Biotinylated proteins are affinity-purified with streptavidin agarose resin (Thermo Scientific, Cat no. 20349). The use of neutravidin agarose resin yielded the same results. ABE samples obtained with this assay were resolved with SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting with protein-specific antibodies. In this assay, the proteins detected in the plus hydroxylamine (+HA) sample but not in the minus hydroxylamine (-HA) sample (which serves as an internal negative control) represent palmitoylated proteins and collectively comprise the palmitoyl proteome. In all ABE experiments, for each tissue sample, both +HA and -HA samples were processed in parallel and the specificity of +HA signal was confirmed by the complete absence of signal in -HA sample in western blotting.

Samples were mixed with Laemmli sample buffer (Laemmli, Bio-Rad, Cat no. 1610747), boiled at 95°C for 5 min, stored at –20°C and analyzed with SDS-PAGE and western blotting. Briefly, samples were loaded on 4–20% gradient mini-PROTEAN TGX precast protein gels (Bio-Rad, 10-well Cat no. 4561094, or 15-well Cat no. 4561096) and run under standard protein electrophoresis conditions. Proteins on gels were transferred on PVDF membrane (Immobilon-P, Millipore, Cat no. IPVH00010) and blocked with 5% non-fat dry milk (Scientific, Cat no. M0841), diluted in TBS-T (TBS-Tween 20), for 1 hr at room temperature. Next, membranes were incubated with primary antibody in 1% milk/TBS-T, overnight at 4°C, with end-to-end rotation. Next day, membranes were washed with TBS-T three times and incubated with horseradish peroxidase (HRP)-conjugated species-specific secondary antibody in 1% milk, to a final concentration 1:10,000, for 1 hr at room temperature, with end-to-end rotation. Membranes were washed with TBS-T four times. Signal detection was performed with ECL Western Blotting Detection Reagent (GE Healthcare, Cat no. RPN2109) or ECL Prime Western Blotting Detection Reagent (GE Healthcare, Cat no. RPN2232) or Clarity Western ECL substrate (Bio-Rad, Cat no. 1705060). Different exposures were obtained for each experiment to ensure that the detected signals were in the linear range. Western blot densitometry was performed with the gel analysis tool of ImageJ. For ABE experiments, palmitoylated protein was calculated as the ratio of +HA (hydroxylamine) signal to the respective input and subsequently all values (ratios) were expressed as a percentage of the ratio for wild-type protein (as explained in figure legends).

Live imaging was performed to visualize cell surface localization of pHluorin-tagged Nrp2 in primary cortical neurons or Neuro-2a cells. The pHluorin epitope tag serves as a robust fluorescent reporter of protein expressed on the cell surface because of its sensitivity to pH. Cells were cultured in individual glass bottom dishes (MatTek Life Sciences, Cat no. P35G-0-14C) precoated with poly-D-lysine and transfected with appropriate plasmids. One or two days after transfection, they were imaged live at an inverted LSM700 confocal microscope (Zeiss) with a 40× oil lens. Imaging parameters were set prior to the start of these experiments, and were kept constant during image acquisition.

Wild-type cortical neurons were cultured and plated on glass bottom dishes precoated with poly-D-lysine. Between DIV5 and DIV8 neurons were transfected with the indicated pHluorin-Nrp2-ires-dsRED expression plasmids with the method of Lipofectamine. A few days later, they were imaged with an inverted LSM700 microscope (Zeiss) with an incubation chamber (PeCon) (temperature: 37°C, CO₂: 5%), using a 63× NA1.4 oil immersion lens. For the performance of FRAP, we used the following protocol: a ROI containing the main dendritic process was selected for FRAP analysis. Images were acquired every 30 s; five images were acquired prior to bleaching (prebleach time: 2.5 min) and 31 additional images were acquired post-bleaching (total post-bleach time: 15.5 min). Analysis: Fluorescence intensity of time-lapse images was determined by measuring the fluorescence intensity within the ROI and correcting for fluorescence decay and for background fluorescence. Imaging parameters were set prior to image acquisition and were kept constant throughout the course of these experiments.

Cell growth medium was aspirated and lysis buffer (TNE buffer: NP-40, Tris, EDTA, NaCl, protease inhibitor cocktail) was added in cells in the culture plate. Cells were scraped off, mechanically lysed on ice by passage through 1 ml syringes and incubated with rotation at 4°C for 1 hr. Samples were precleared with protein A/G agarose resin (Pierce, Thermo Fisher Scientific, Cat no. 20421) at 4°C for 2 hr. Next, after the resin was discarded, a small fraction of each sample was kept as input and the remaining samples were incubated with the appropriate antibody at 4°C overnight. Next, samples were incubated with protein A/G agarose with end-over-end rotation at 4°C for 2 hr. Resin was washed twice with high-salt buffer (500 mM NaCl) and twice with low-salt buffer (150 mM NaCl). Proteins were eluted from resin with Laemmli sample buffer diluted in TNE buffer, boiled at 95°C for 5 min and analyzed with SDS-PAGE and western blotting.

Alkaline phosphatase (AP)-tagged Sema3A and Sema3F or AP expression plasmids were transfected into 293T cells with the method of Lipofectamine. Cells were allowed to secrete the ligand for 2–5 days. The culture supernatant was collected and transferred to a 50 ml MW filter tube (Centricon filters, Millipore) and was centrifuged at 35 rpm for 10 min at 4°C. A 50 K filter was used for AP concentration and a 100 K filter was used for concentration of Sema3A-AP or Sema3F-AP. The liquid above the filter was transferred to a new tube and ligand concentration was determined with a spectrophotometer by measuring AP activity by the change in absorbance at OD 405 nm. This figure was then converted to nM. The concentrated ligand was aliquoted and stored at –80°C. For each experiment, a new aliquot was thawed and used for culture treatment to avoid protein decay from thawing and refreezing.

For in vitro rescue, E14.5 Nrp2^-/- primary cortical neurons were transfected at DIV7 with the indicated plasmids. At DIV21 neurons were treated with 5 nM Sema3F-AP or AP for 6 hr and afterward they were processed for immunofluorescence. Confocal images (stacks) were taken with an LSM700 with a 63× oil lens. Analysis was performed with ImageJ and dendritic spines were counted along the 0–50 μm apical dendritic segment from the cell body. For in vivo rescue, IUE was performed on E13.5 Nrp2^F/- embryos of timed-pregnant females (detailed below). Nrp2^-/- mice were crossed with Nrp2^F/F mice so that all embryos are Nrp2^F/-.

Statistical significance and the size of samples analyzed are presented in figures and figure legends. Image analysis was performed using ImageJ software with appropriate plugins (Rasband, WS, ImageJ, U.S. National Institutes of Health, Bethesda, MD, USA, https://imagej.nih.gov/ij/, 1997–2016). The prediction of neuropilin palmitoylation sites was performed with the CSS-Palm 3.0 software, which is freely available (Ren et al., 2008). Conservation analysis of Nrp2 and Nrp1 amino acid sequences across species was performed with the ClustalW software. Details for the analysis of individual biochemical and phenotypic experiments are provided in separate sections. Data were collected in Excel and GraphPad Prism. Generation of all graphs and statistical analyses were performed in GraphPad Prism. The GraphPad Prism software was purchased. Figure assembly and generation of schematics presented in Figure 7E and F have been performed in Adobe Illustrator. Statistical analysis tests include two-tailed t test and analysis of variance (ANOVA) followed by the appropriate post hoc test for multiple comparisons (as mentioned in figure legends). The choice of post hoc statistical tests was based on the appropriate comparison-biological question under study in each experiment. For example, when one condition is used as a control and all other conditions are compared with the control, which is the case in many of our experiments, Dunnett’s post hoc test is performed since it is the most appropriate test in these cases. For comparison of predefined pairs, such as in Figure 4C where each Nrp2 protein is separately examined for its ability to respond to Sema3F, Sidak’s post hoc analysis is used since it assumes that each comparison is independent of the others, which is the case in this experiment. For comparison of every mean with every other mean, as it is the case in Figure 5C, Tukey’s test is the most appropriate. Overall, besides selecting post hoc tests based on generally accepted determinations of appropriate statistical tests for the different experimental paradigms in this study, our post hoc test selection was also guided by GraphPad Prism, which provides, based on the particular biological question under study (i.e. sets of comparisons), recommended and non-recommended post hoc analysis methods. Statistical significance is defined as: *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001, ns: not significant. Exact p values are presented either on graphs or in figure legends.