The nucleolar δ isoform of adapter protein SH2B1 enhances morphological complexity and function of cultured neurons

ABSTRACT

The adapter protein SH2B1 is recruited to neurotrophin receptors, including TrkB (also known as NTRK2), the receptor for brain-derived neurotrophic factor (BDNF). Herein, we demonstrate that the four alternatively spliced isoforms of SH2B1 (SH2B1α–SH2B1δ) are important determinants of neuronal architecture and neurotrophin-induced gene expression. Primary hippocampal neurons from Sh2b1^−/− [knockout (KO)] mice exhibit decreased neurite complexity and length, and BDNF-induced expression of the synapse-related immediate early genes Egr1 and Arc. Reintroduction of each SH2B1 isoform into KO neurons increases neurite complexity; the brain-specific δ isoform also increases total neurite length. Human obesity-associated variants, when expressed in SH2B1δ, alter neurite complexity, suggesting that a decrease or increase in neurite branching may have deleterious effects that contribute to the severe childhood obesity and neurobehavioral abnormalities associated with these variants. Surprisingly, in contrast to SH2B1α, SH2B1β and SH2B1γ, which localize primarily in the cytoplasm and plasma membrane, SH2B1δ resides primarily in nucleoli. Some SH2B1δ is also present in the plasma membrane and nucleus. Nucleolar localization, driven by two highly basic regions unique to SH2B1δ, is required for SH2B1δ to maximally increase neurite complexity and BDNF-induced expression of Egr1, Arc and FosL1.

Summary: Of the four isoforms of adapter protein SH2B1, only SH2B1δ is nucleolar. SH2B1δ greatly enhances neurite complexity, a function impacted by human obesity-associated variants.

INTRODUCTION

Neurotrophins make critical contributions to the development and maintenance of the nervous system. They regulate neuronal structure and function locally, by modifying the cytoskeleton and signaling proteins, and globally, by impacting gene expression (Huang and Reichardt, 2001). The neurotrophin family comprises nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin-3 and neurotrophin-4 (also known as neurotrophin-5). Of these, NGF and BDNF are the primary regulators of peripheral and central nervous system development, respectively (Snider, 1994). NGF and BDNF bind their respective receptor tyrosine kinases, TrkA and TrkB (also known as NTRK1 and NTRK2, respectively), to activate and induce autophosphorylation of these receptors, which initiates recruitment of signaling proteins and subsequent downstream activities (Huang and Reichardt, 2003; Schlessinger and Ullrich, 1992).

SH2B1 is an adapter protein that is recruited, via its SH2 domain, to phosphotyrosines in activated receptor tyrosine kinases including TrkA and TrkB (Rui, 2014). Ectopically expressed SH2B1 enhances NGF- and BDNF-induced signaling events (Joe et al., 2017; Shih et al., 2013; Wang et al., 2011, 2004), NGF-induced gene expression (Chen et al., 2008; Maures et al., 2009), and reorganization of the actin cytoskeleton (Chen et al., 2015; Diakonova et al., 2002; Lanning et al., 2011; Rider and Diakonova, 2011; Rider et al., 2009). Conversely, knockdown of SH2B1 using shRNA reduces NGF-induced gene expression and NGF- and BDNF-induced neurite outgrowth in PC12 cells (Chen et al., 2008; Maures et al., 2009; Shih et al., 2013), as well as length and branching of axons and/or dendrites, collectively referred to as neurites, in primary neurons (Chen et al., 2015; Shih et al., 2013). The effect of SH2B1 on BDNF-induced gene expression had not been investigated.

Human and mouse studies indicate that disruption or deletion of SH2B1 disturbs mechanisms that regulate feeding and social behaviors. Variants in SH2B1 have been identified in humans with severe, early-onset obesity. Many of these individuals have disproportionate insulin resistance for their level of obesity and/or display maladaptive behaviors including aggression (Doche et al., 2012; Flores et al., 2019; Pearce et al., 2014). Mice null for Sh2b1 [Sh2b1 knockout (KO) mice] similarly exhibit obesity, insulin resistance and aggression (Duan et al., 2004; Jiang et al., 2018; Ren et al., 2005). Although earlier work (Ren et al., 2005) provided evidence that SH2B1 regulates energy balance, at least in part, by modulating leptin signaling, more recent work (Cote et al., 2021) suggests that some isoforms of SH2B1 regulate energy balance through leptin-independent means. Because BDNF/TrkB activity is strongly implicated in the regulation of energy balance and SH2B1 is recruited to activated TrkB, SH2B1 may regulate energy balance, at least in part, by modulating TrkB signaling. This possibility underscores the need to better understand how SH2B1 modulates Trk receptor activity.

Alternative splicing generates four SH2B1 isoforms: SH2B1α, SH2B1β, SH2B1γ and SH2B1δ (Nelms et al., 1999; Yousaf et al., 2001). These isoforms differ only in their C-terminal tails (Fig. 1A). In humans, SH2B1β and SH2B1γ are expressed ubiquitously, whereas SH2B1α and SH2B1δ are expressed almost exclusively in brain (Doche et al., 2012). SH2B1α and SH2B1β localize primarily to the plasma membrane and cytoplasm; SH2B1β, but not SH2B1α, also cycles through the nucleus (Maures et al., 2011; Pearce et al., 2014). Whereas the C-terminal tail of SH2B1α contains a unique tyrosine (Tyr753) whose phosphorylation regulates several isoform-specific functions (Joe et al., 2017), the C-terminal tails of SH2B1β and SH2B1γ do not contain recognizable domains or features. The C-terminal tail of SH2B1δ contains two highly basic regions (Yousaf et al., 2001) yet its subcellular localization had not been determined. SH2B1β, SH2B1γ or both, but not SH2B1α, have been shown to enhance NGF-induced signaling events, gene expression and neurite outgrowth in PC12 cells (Chen et al., 2008; Joe et al., 2017; Maures et al., 2009; Qian et al., 1998; Rui et al., 1999). SH2B1β enhances neurite length and branching in primary hippocampal and cortical neurons (Chen et al., 2015; Shih et al., 2013; Zhang et al., 2006). How the unique C-terminal tail of SH2B1δ modulates SH2B1 regulation of neurotrophin-induced cellular activities was unknown.

Fig. 1.

Sh2b1 KO neurons exhibit decreased neurite complexity and length. (A) Schematic of SH2B1 isoforms modified from Cote et al. (2021). Different colored C-terminal tails denote isoform-specific amino acids. Numbers indicate amino acids in mouse and human sequences. P, proline-rich region; DD, dimerization domain; NLS, nuclear localization sequence; NES, nuclear export sequence; PH, pleckstrin homology domain; SH2, Src homology 2 domain. (B,C) mRNA levels of (B) total Sh2b1 or (C) Sh2b1 isoforms were measured in WT hippocampal neurons by qPCR. The number (n) of biological replicates was: DIV2,5,8, n=2; DIV14, n=3. (D) Representative WT and KO hippocampal neurons transiently expressing GFP (DIV4) and imaged (DIV5) using live-cell fluorescence microscopy. Images inverted. Scale bars: 20 µm. (E–G) Parameters measured using Simple Neurite Tracer on neuron images prepared as in D. The number (n) of neurons from seven distinct experiments was: WT, n=92; KO, n=89. (H–J) Parameters were obtained via Sholl analysis of images used in E–G. H inset, subset of data in H. (K) Proteins in hippocampal neuron lysates were immunoblotted with antibodies to PSD-95 (αPSD-95) or ERK1/2 (αERK1/2). Migration of molecular mass standards is on the right. PSD-95 signals were normalized to ERK1/2 signals and then to WT values. Data are means±s.e.m. Statistics: B,C,E–G,I,J, one-tailed unpaired t-test compared to DIV2 (B,C) and WT (E–G,I,J); H, inset, two-way repeated measures ANOVA with Dunnett's multiple comparisons test, thick line indicates significance; K, one-tailed paired t-test. *P<0.05.

Studies performed in vivo have implicated specific SH2B1 isoforms in SH2B1 regulation of energy balance. For example, reintroduction of SH2B1β into neurons of Sh2b1 KO mice via the neuron-specific enolase promoter results in mice of normal, rather than high, body weight (Ren et al., 2007). We recently demonstrated that deletion of the brain-specific α and δ isoforms results in lean mice that eat less food than control littermates (Cote et al., 2021). These results suggest not only that the α, β and/or δ isoforms of SH2B1 contribute to SH2B1 regulation of energy balance, but also that different SH2B1 isoforms perform distinct functions in vivo. Together, these findings highlight the need to better understand how the different SH2B1 isoforms, particularly the less studied δ isoform, impact neurotrophin-induced cellular activities.

Here, we investigate how the various SH2B1 isoforms regulate neuron structure and function. We show that compared to primary hippocampal neurons from wild-type (WT) mice, neurons from Sh2b1 KO mice exhibit decreased neurite length and arborization and BDNF-induced expression of synapse-related immediate early genes. Reintroduction of each SH2B1 isoform into KO neurons increases neurite length and/or arborization. SH2B1δ, like SH2B1β, enhances NGF-induced signaling events and gene expression in PC12 cells. Similarly, reintroduction of SH2B1δ into KO neurons enhances BDNF-induced gene expression. In contrast to the other three isoforms, SH2B1δ localizes primarily to nucleoli, a process requiring two highly basic regions in its unique C-terminal tail. SH2B1δ must localize to nucleoli to maximally increase neurite complexity and BDNF-dependent gene expression. Together, these data suggest that the different SH2B1 isoforms act together to enhance neurite arborization and length and neurotrophin-induced gene expression. Furthermore, the unique C-terminal tail of SH2B1δ enhances SH2B1 regulation of neuronal development, at least in part, by its ability to target SH2B1 to nucleoli.

RESULTS

Sh2b1 KO neurons exhibit reduced neurite complexity and length

To understand how the various SH2B1 isoforms regulate neuronal function, we first investigated their ability to alter neurite complexity and length. Arborization and lengths of axons and dendrites impact the ability of individual neurons to make appropriate connections with other neurons at synapses (Brown et al., 2008). We used primary hippocampal neuron cultures to assess the effects of SH2B1 isoforms on these parameters because these neurons display elaborate branching patterns, express TrkB, and synthesize and secrete BDNF (Ivanova and Beyer, 2001). In WT neurons, Sh2b1 mRNA was detected by quantitative (q)PCR at day in vitro (DIV) 2, the earliest timepoint taken, and its expression level increased substantially between DIV2 and 14 (Fig. 1B). Neither Sh2b1 mRNA (Fig. S1) nor SH2B1 protein (data not shown) was present in Sh2b1 KO neuron cultures. qPCR data suggest that in WT neurons, α is the most highly expressed SH2B1 isoform, followed by β, δ and γ (Fig. 1C). Expression of SH2B1α increased significantly between DIV2 and 14.

To analyze the contribution of all SH2B1 isoforms to neurite complexity and length, we cultured WT and KO hippocampal neurons and transiently expressed GFP to facilitate visualization of individual neurons (Fig. 1D). Fluorescent images of GFP-positive neurons were traced using Simple Neurite Tracer in FIJI and manually curated. Compared to WT neurons, KO neurons exhibited a decrease in: (1) number of endpoints, indicative of overall neurite complexity; (2) total neurite length, indicative of the combined length of the axon and dendrites; and (3) length of the longest neurite, typically indicative of axon length (Fig. 1E–G).

We next used a computerized Sholl analysis (Sholl, 1953) to obtain a more detailed appreciation of the impact of SH2B1 on the complexity and reach of the axon and dendrites of a neuron. Sholl analysis centers concentric circles of incrementally increasing radii around the soma and determines the number of neurites that intersect each circle. Consistent with the neurite tracing results, KO neurons displayed fewer intersections compared to WT neurons (Fig. 1H). These decreases occurred at multiple distances (between 18 and 112 µm) close to the soma (Fig. 1H), suggesting that KO neurons have decreased dendritic arborization. KO neurons also displayed a decreased peak number of intersections (Fig. 1I), suggesting decreased dendritic complexity. The maximum distance from the soma to the tip of the furthest neurite was also decreased (Fig. 1J), suggesting decreased axonal length.

Because KO neurons exhibited decreased neurite complexity, we predicted that they would have reduced expression of synaptic proteins. Postsynaptic density (PSD) 95 (PSD-95; also known as DLG4) protein is a key synaptic protein that localizes to the PSD within dendritic spines. We measured PSD-95 levels by immunoblotting lysates of primary hippocampal cultures at DIV8, when synapses are beginning to form, and at DIV13, when synapses are mostly formed. PSD-95 expression levels were reduced in KO neurons to a statistically significant extent at DIV13 (Fig. 1K).

All SH2B1 isoforms increase neurite complexity and/or length

Because the absence of all SH2B1 isoforms decreased neurite complexity and length in hippocampal neurons, we investigated the influence of each individual SH2B1 isoform on these parameters. We transiently expressed GFP or a GFP-tagged SH2B1 isoform in Sh2b1 KO neurons and co-expressed mCherry to ensure that neurites were equally visible for all conditions. We analyzed WT neurons for comparison. We imaged individual GFP-positive neurons in the red channel using live-cell fluorescence microscopy. Neurite tracing indicated that SH2B1δ increased the number of endpoints and total neurite length to levels similar to (total neurite length) or greater than (number of endpoints) those seen with WT neurons (Fig. 2A,B). SH2B1α and SH2B1β also increased the number of endpoints (Fig. 2A). None of the SH2B1 isoforms impacted the length of the longest neurite (Fig. 2C). Sholl analysis indicated that SH2B1δ greatly increased neurite complexity near the soma (10 to 124 µm) (Fig. 2D) to levels greater than those seen with WT neurons. SH2B1α, SH2B1β and SH2B1γ also increased neurite complexity near the soma, although these increases were less pronounced. Sholl analysis also indicated that SH2B1δ, but not the other isoforms, increased the maximum number of intersections (Fig. 2E), restoring it to the level seen in WT neurons. None of the isoforms impacted the maximum neurite distance measured by Sholl analysis (Fig. 2F). Thus, whereas SH2B1α, SH2B1β and SH2B1γ each increased some aspects of neurite complexity but not neurite length, SH2B1δ increased both neurite complexity and length. Given that KO neurons expressing SH2B1δ exhibited increased complexity near the soma and increased total neurite length, yet their reach was unchanged, SH2B1δ likely impacts dendritic (not axonal) complexity and length.

Fig. 2.

SH2B1 isoforms increase neurite complexity and/or length. (A–C) Parameters measured using Simple Neurite Tracer on images of Sh2b1 KO hippocampal neurons transiently expressing GFP (−) or GFP–SH2B1 isoforms. Neurons were transfected (DIV4) and imaged (DIV5) using fluorescence microscopy. The number (n) of neurons from three distinct experiments was: GFP, n=44; GFP–SH2B1α, n=45; GFP–SH2B1β, n=48; GFP–SH2B1γ, n=46; GFP–SH2B1δ, n=44. (D–F) Parameters obtained from Sholl analysis of images used in A–C. D inset, subset of data in D. Data are means±s.e.m. Statistics: A–C,E,F, one-tailed unpaired t-test, WT versus KO neurons expressing GFP (^#P<0.05); one-way ANOVA with Dunnett's multiple comparisons test, KO neurons expressing GFP versus each GFP–SH2B1 isoform (*P<0.05); D, inset, two-way repeated measures ANOVA with Dunnett's multiple comparisons test, KO neurons expressing GFP versus each GFP–SH2B1 isoform, thick lines indicate significance (*P<0.05).

Human obesity-associated SH2B1 variants impair the ability of SH2B1δ to increase neurite complexity

We tested whether four rare human obesity-associated SH2B1 variants (P90H, P322S, T546A, R680C) or the common human obesity-associated variant (A484T) (Doche et al., 2012; Flores et al., 2019; Pearce et al., 2014) (Fig. 3A) would disrupt the ability of SH2B1δ to increase neurite complexity and length. These variants exhibited subcellular localizations similar to that of SH2B1δ(WT) in PC12 cells (data not shown). GFP–SH2B1δ(WT) increased the number of endpoints and total neurite length in Sh2b1 KO neurons compared to the GFP control (Fig. 3B,C) as in Fig. 2. None of these variants impacted the ability of SH2B1δ to increase neurite complexity or length in KO neurons as assessed by neurite tracing (Fig. 3B–D). However, these obesity-associated variants impacted, to varying degrees, the ability of SH2B1δ to increase neurite complexity in KO neurons as assessed by Sholl analysis (Fig. 3E). The T546A variant, located in the SH2 domain and, to a much greater extent, the P322S variant, located in the PH domain, disrupted the ability of SH2B1δ to increase the number of intersections close to the soma. The P90H variant, located in a proline-rich region, and the common A484T variant, located in a region of the protein without recognized domains, barely disrupted the ability of SH2B1δ to increase the number of intersections. The nominal disruption from the A484T variant is consistent with our previous observation that this variant does not affect the ability of SH2B1β to enhance NGF-induced neurite outgrowth in PC12 cells (Doche et al., 2012). Interestingly, the R680C variant, located in the unique C-terminal tail of SH2B1δ, had the opposite effect from the other variants, enhancing the ability of SH2B1δ to increase the number of intersections close to the soma. These data suggest that human obesity-associated SH2B1 variants located in the PH and SH2 domains decrease, and the R680C variant in the unique C-terminal tail of SH2B1δ increases, SH2B1δ-induced dendritic complexity.

Fig. 3.

The ability of SH2B1δ to increase neurite complexity is altered by some human obesity-associated SH2B1 variants. (A) Schematic of SH2B1δ showing human obesity-associated variants. (B–D) Parameters measured using Simple Neurite Tracer on images of Sh2b1 KO hippocampal neurons transiently expressing GFP (−) or the indicated GFP–SH2B1δ construct. Neurons were transfected (DIV4) and imaged (DIV5) using fluorescence microscopy. The number (n) of neurons from three distinct experiments was: GFP, n=34; GFP–SH2B1δ(WT), n=45; GFP–SH2B1δ(P90H), n=38; GFP–SH2B1δ(P322S), n=39; GFP–SH2B1δ(A484T), n=38; GFP–SH2B1δ(T546A), n=43; GFP–SH2B1δ(R680C), n=39. (E) Sholl analysis of images used in B-D. Inset: subset of data in E. Data are means±s.e.m. Statistics: B–D, one-tailed unpaired t-test, GFP versus GFP–SH2B1δ(WT) (^#P<0.05); one-way ANOVA with Dunnett's multiple comparisons test, GFP–SH2B1δ(WT) versus each GFP–SH2B1δ variant (no significant differences detected); E, inset, two-way repeated measures ANOVA with Dunnett's multiple comparisons test, GFP–SH2B1δ(WT) versus each GFP–SH2B1δ variant, thick lines indicate significance (*P<0.05).

SH2B1δ localizes to nucleoli and the plasma membrane

The unique C-terminal tail of SH2B1δ (Fig. 1A) contains two highly basic regions that match the canonical sequence for nuclear localization motifs (Yousaf et al., 2001), referred to here as nuclear localization sequence (NLS) 2 and NLS3. Because of these unique motifs and the dramatic enhancement of neuronal complexity induced by SH2B1δ, we examined whether SH2B1δ localizes differently from the other SH2B1 isoforms. We transiently expressed GFP-tagged SH2B1 isoforms in PC12 cells and imaged their steady-state subcellular localization using live-cell confocal microscopy. Consistent with our previous results (Maures et al., 2011; Pearce et al., 2014), SH2B1α and SH2B1β localized primarily in the cytoplasm and at the plasma membrane (Fig. 4A). SH2B1γ localized similarly. Surprisingly, SH2B1δ localized primarily to nucleoli. SH2B1δ also localized at the plasma membrane and in the nucleus, which was observed at an increased photomultiplier gain.

Fig. 4.

SH2B1δ localizes to nucleoli and the plasma membrane. (A) PC12 cells transiently co-expressing GFP–SH2B1 isoforms and mCherry–nucleolin were stained with Alexa Fluor 488–conjugated wheat germ agglutinin (WGA) and imaged using live-cell confocal microscopy. Left five panels: low photomultiplier gain. Right two panels: high photomultiplier gain. Images representative of ≥20 cells/isoform. DIC, differential interference contrast. (B) Sh2b1 KO hippocampal neurons were transiently transfected with the indicated GFP–SH2B1 isoforms (DIV5) and imaged (DIV6) using fluorescence (top panels) or confocal microscopy of fixed neurons stained with antibody to GFP (αGFP) (middle panels) or live-cell confocal microscopy (bottom panels). Images inverted in top and middle panels. Arrowheads (middle panels) denote microstructures protruding from dendritic shafts. Scale bars: 10 µm.

We next tested where the different SH2B1 isoforms localize in primary hippocampal neurons. Fluorescence microscopy of fixed DIV6 neurons stained with antibody to GFP revealed that all four isoforms were present within the soma and throughout neurites (Fig. 4B, top panels). All four isoforms also appeared to be expressed within microstructures protruding from dendritic shafts (Fig. 4B, middle panels). These microstructures were likely filopodia, some of which would have presumably developed into dendritic spines (Yoshihara et al., 2009). Live-cell confocal microscopy confirmed that in hippocampal neurons as in PC12 cells, SH2B1α, SH2B1β and SH2B1γ localized primarily at the plasma membrane and in the cytoplasm, whereas SH2B1δ localized primarily in nucleoli with some at the plasma membrane. Some SH2B1δ was also present in the cytoplasm (Fig. 4B, bottom).

SH2B1δ enhances neurotrophin-induced signaling events

NGF and BDNF activation of their respective Trk receptors induces receptor autophosphorylation. The phosphotyrosines recruit signaling proteins, including SH2B1. Our lab and/or others have shown that SH2B1β enhances and/or prolongs NGF-induced phosphorylation of phospholipase C (PLC)γ (Tyr783) and Akt (Ser473) in PC12 cells (Joe et al., 2017; Wang et al., 2011, 2004). SH2B1β also enhances and prolongs BDNF-induced phosphorylation of these signaling proteins in PC12 cells stably overexpressing TrkB (Shih et al., 2013). Some, but not all, studies have observed that SH2B1β modestly enhances NGF- and BDNF-induced phosphorylation of ERK1 and ERK2 (ERK1/2, also known as MAPK3 and MAPK1, respectively; at Thr202/Tyr204) in PC12 cells (Joe et al., 2017; Maures et al., 2009; Rui et al., 1999; Shih et al., 2013; Wang et al., 2011). All three signaling pathways may contribute to neurite outgrowth (Atwal et al., 2000; Inagaki et al., 1995; Jaworski et al., 2005; Kumar et al., 2005).

Given that SH2B1β and SH2B1δ localize to distinct subcellular compartments, we wondered whether SH2B1δ would, like SH2B1β, enhance neurotrophin-induced signaling events. We used PC12 cells as our model system due to our much greater ability to detect subtle differences in signaling events in a cell line compared to primary hippocampal neurons. We first verified that SH2B1δ, like SH2B1β (Rui et al., 1999), enhances NGF-induced neurite outgrowth (Fig. S2). To assess the effect of SH2B1δ on NGF signaling pathways, we stably expressed GFP, GFP–SH2B1β or GFP–SH2B1δ in PC12 cells, added NGF (25 ng/ml) for 10 or 60 min, and immunoblotted cell lysates to determine phosphorylation of activating amino acids in PLCγ, Akt and ERK1/2. SH2B1δ, like SH2B1β, enhanced and prolonged phosphorylation of PLCγ (Fig. 5A,B), and enhanced rapid phosphorylation of Akt family proteins (Fig. 5A,C). SH2B1δ enhancement of ERK1/2 phosphorylation reached statistical significance, whereas SH2B1β enhancement did not (Fig. 5A,D).

Fig. 5.

SH2B1δ promotes NGF-stimulated signaling activities. (A) Serum-starved PC12 cells stably expressing GFP, GFP–SH2B1β or GFP–SH2B1δ were stimulated with 25 ng/ml NGF as indicated. Proteins in cell lysates were immunoblotted (IB) with the indicated antibodies. Migration of molecular mass standards is on the left. (B–D) Quantification of immunoblots in A plus two additional sets of immunoblots, each performed using a distinct biological sample. Relative signals of phosphorylated proteins were normalized first to the signal for the total amount of that protein and then to the signal for GFP-expressing cells stimulated with NGF for 10 min. Data are means±s.e.m. Statistics: B–D, one-way ANOVA at 10 and 60 min with Tukey's multiple comparisons test to compare GFP versus GFP–SH2B1β (^#P<0.05) and GFP versus GFP–SH2B1δ (*P<0.05).

SH2B1δ enhanced signaling activity to a similar extent as SH2B1β despite its substantially reduced protein expression (Fig. 5A, bottom panel). Therefore, on a per molecule basis, SH2B1δ may be a more potent promoter of NGF-induced signaling activity than SH2B1β. Thus, despite localizing primarily to nucleoli, SH2B1δ enhances several neurotrophin-induced signaling events to an extent similar to, or even greater than, SH2B1β. Interestingly, despite the difference in protein levels, GFP–SH2B1β and GFP–SH2B1δ mRNA were expressed at equal levels as assessed by qPCR using GFP-targeted primers (data not shown), raising the possibility that these two isoforms may have different rates of synthesis and/or degradation.

The bipartite NLS in the C-terminal tail of SH2B1δ is required for SH2B1δ to localize to nucleoli

We next questioned whether nucleolar localization of SH2B1δ contributes to its ability to enhance neurite complexity and length. We used mutational analysis to identify the sequence(s) in SH2B1δ required for its nucleolar localization. We mutated the highly basic regions NLS2 and NLS3, individually and in combination (Fig. 6A). The SH2B1δ tail also contains a serine/tryptophan-rich segment (residues 708–718) (Yousaf et al., 2001); stretches of aromatic amino acids such as tryptophan residues have been implicated in targeting nuclear proteins to nucleoli (Grummitt et al., 2008; Nishimura et al., 2002; Schmidt-Zachmann and Nigg, 1993). We therefore also truncated SH2B1δ after residue 712 (W712X) to delete a stretch containing 4 tryptophan residues (Fig. 6A). We transiently expressed these GFP-tagged SH2B1δ mutants in PC12 cells and assessed their subcellular localization using live-cell confocal microscopy (Fig. 6B,C). Mutating either NLS2 or NLS3 caused SH2B1δ to localize more strongly in the nucleus and at the plasma membrane. However, mutating NLS2 and NLS3 simultaneously (mNLS2+3) caused SH2B1δ to be excluded from nucleoli and localize primarily at the plasma membrane. The W712X truncation did not affect the ability of SH2B1δ to localize in nucleoli. However, surprisingly, it prevented SH2B1δ from localizing at the plasma membrane.

Fig. 6.

Bipartite NLS in C-terminal tail is required for SH2B1δ to localize in nucleoli. (A) Schematic of SH2B1δ with mutations and truncation. Magenta font and/or underline indicates mutation or removal of amino acids. mNLS, mutated NLS. (B) PC12 cells transiently co-expressing the indicated GFP–SH2B1δ mutant and mCherry–nucleolin were stained with Alexa Fluor 467-conjugated WGA and imaged using live-cell confocal microscopy. Left five panels: low photomultiplier gain. Right two panels: high photomultiplier gain. DIC, differential interference contrast. Scale bars: 10 µm. (C) Subcellular localization of the indicated SH2B1δ mutants. Bright, dim and absent refer to the signal intensity in indicated cellular compartment [bright, seen with low gain; dim, seen only with high gain; absent, not seen even with high gain]. The number (n) of cells from two or three distinct experiments were: SH2B1δ(WT), n=28; δ(mNLS2), n=24; δ(mNLS3), n=19; δ(mNLS2+3), n=25; δ(W712X), n=20; δ(mNLS1), n=25. PM, plasma membrane.

All SH2B1 isoforms share a nuclear localization sequence, NLS1, and nuclear export sequence (NES) (Fig. 1A). NLS1 serves a dual purpose of localizing SH2B1β to the plasma membrane and enabling it to cycle through the nucleus (Maures et al., 2011). Mutating NLS1 in SH2B1δ prevented SH2B1δ from localizing to the plasma membrane. However, mutating NLS1 in SH2B1δ did not affect its ability to localize in nucleoli (Fig. 6B,C), presumably because it has additional NLSs in its unique C-terminal tail. Thus, for SH2B1δ to localize in nucleoli, either NLS2 or NLS3 must be functional. For SH2B1δ to localize at the plasma membrane, NLS1 and the tryptophan-rich segment of the C-terminal tail, amino acids 712–724, must be intact.

SH2B1δ localization to nucleoli and the plasma membrane, and its recruitment to an activated tyrosine kinase, are required for SH2B1δ to increase neurite complexity

To determine whether SH2B1δ localization to nucleoli affects neurite complexity and length, we analyzed the architecture, via neurite tracing, of Sh2b1 KO hippocampal neurons transiently expressing GFP-tagged SH2B1δ mutants with distinct localization profiles (Fig. 7A). We first confirmed that the NLS2+3, W712X and NLS1 mutations caused the same alteration in SH2B1δ localization in hippocampal neurons as seen in PC12 cells (Fig. 7B). We then confirmed that GFP–SH2B1δ(WT) increased the number of endpoints and total neurite length compared to the GFP control (Fig. 7C,D). Preventing SH2B1δ from going to nucleoli by mutating both NLS2 and NLS3 disrupted the ability of SH2B1δ to increase the number of endpoints, reducing the number of endpoints to a level similar to that seen with the GFP control (Fig. 7C). Changing the subcellular localization of SH2B1δ via these mutations did not impair the ability of SH2B1δ to increase total neurite length or the length of the longest neurite (Fig. 7D,E).

Fig. 7.

SH2B1δ must localize to nucleoli and the plasma membrane to maximally increase neurite complexity. (A) Schematic of SH2B1δ showing mutations. (B) Sh2b1 KO hippocampal neurons were transiently transfected (DIV5) with cDNA encoding the indicated GFP–SH2B1δ constructs and imaged (DIV6) using live-cell confocal microscopy. Left three panels: low photomultiplier gain. Right panel: high photomultiplier gain. Scale bars: 10 µm. (C–E) Parameters measured using Simple Neurite Tracer on images of KO neurons transiently expressing the indicated GFP–SH2B1δ construct. Neurons were transfected (DIV4) and imaged (DIV5) using fluorescence microscopy. The number (n) of neurons from three distinct experiments was: GFP, n=49; GFP–SH2B1δ(WT), n=58; GFP–SH2B1δ(mNLS2+3), n=41; GFP–SH2B1δ(W712X), n=28; GFP–SH2B1δ(mNLS1), n=27; GFP–SH2B1δ(R555E), n=34. (F) Sholl analysis of neuron images used in C–E. Inset: subset of data in F. Data are means±s.e.m. Statistics: C–E, one-tailed unpaired t-test, GFP versus GFP–SH2B1δ(WT) (^#P<0.05); one-way ANOVA with Dunnett's multiple comparisons test, GFP–SH2B1δ(WT) versus each GFP–SH2B1δ mutant (*P<0.05); F, inset, two-way repeated measures ANOVA with Dunnett's multiple comparisons test, GFP–SH2B1δ(WT) versus each GFP–SH2B1δ mutant, thick lines indicate significance (*P<0.05).

Sholl analysis indicated that mutating both NLS2 and NLS3 in SH2B1δ substantially reduced the ability of SH2B1δ to increase the number of intersections near the soma in KO neurons (Fig. 7F); at multiple distances between 50 and 100 µm, mutating both NLS2 and NLS3 brought the number of intersections down to the GFP level. Truncating W712X or mutating NLS1 in SH2B1δ more modestly disrupted the ability of SH2B1δ to increase the number of intersections near the soma.

To determine whether recruitment to phosphorylated tyrosine residues in activated tyrosine kinases (e.g. TrkB) is required for SH2B1δ to increase neurite complexity and length, we mutated arginine 555 to glutamate (R555E) in the SH2 domain of SH2B1δ (Fig. 7A), a domain shared by all isoforms of SH2B1 (Fig. 1A). This mutation disrupts the SH2 domain such that SH2B1 is unable to be recruited to phosphotyrosines in tyrosine kinases (Rui and Carter-Su, 1998; Rui et al., 1999) and prevents SH2B1β from enhancing NGF-induced neurite outgrowth, signaling pathways and gene expression in PC12 cells (Chen et al., 2008; Rui et al., 1999; Wang et al., 2004). Like SH2B1δ(WT), SH2B1δ(R555E) localized primarily in nucleoli and at the plasma membrane (Fig. S3). Mutating R555E in SH2B1δ reduced the ability of SH2B1δ to increase the number of endpoints (Fig. 7C) and total neurite length (Fig. 7D). Sholl analysis revealed that mutating R555E brought the number of intersections between 12 and 122 µm from the soma down to the GFP level (Fig. 7F). In conjunction with the data above, these results suggest that SH2B1δ needs to localize to, and perhaps cycle between, both nucleoli and the plasma membrane to increase dendritic complexity to its fullest extent. These data also suggest that SH2B1δ recruitment to phosphotyrosines in activated tyrosine kinases (e.g. TrkB) is required for SH2B1δ to fully increase dendritic complexity and length.

BDNF-induced expression of neuronal immediate early genes is partially dependent on SH2B1

SH2B1β enhances expression of a subset of NGF-sensitive genes in PC12 cells (Chen et al., 2008). We predicted that SH2B1δ would similarly enhance neurotrophin-induced expression of these genes. We tested the effect of SH2B1δ on four of these genes: Plaur, encoding urokinase plasminogen activator receptor (uPAR); Mmp3 and Mmp10, encoding matrix metalloproteinases (MMP) 3 and 10; and FosL1, encoding Fos-related antigen 1 (FosL1/Fra1). uPAR, MMP3 and MMP10 are members of a protease cascade implicated in extracellular matrix degradation, neuronal differentiation and cell motility (Basbaum and Werb, 1996; Ossowski and Aguirre-Ghiso, 2000). Plaur is an early response gene critical for NGF-induced neurite outgrowth in PC12 cells (Farias-Eisner et al., 2000), differentiation of cultured mouse dorsal root ganglia neurons (Hayden and Seeds, 1996) and migration of mouse cortical interneurons in vivo (Powell et al., 2001). MMPs, such as MMP3 and MMP10, are critical for the extension of neurites from the soma (McFarlane, 2003). FosL1, an immediate early response gene, encodes a transcription factor that regulates a wave of late response genes (Yap and Greenberg, 2018), which in turn encode proteins that mediate multiple morphological processes including dendritic growth, dendritic spine maturation and synapse elimination (West and Greenberg, 2011).

Like SH2B1β, SH2B1δ enhances NGF-induced expression of Plaur, Mmp3, Mmp10 and FosL1 in PC12 cells (Fig. 8A–D). Despite SH2B1β and SH2B1δ enhancing NGF stimulation of PLCγ, Akt and ERK1/2 to similar extents, SH2B1δ seemed to stimulate NGF-dependent gene expression to a lesser extent than SH2B1β. However, their relative potency is unclear because SH2B1δ protein appeared to be present at decreased levels compared to SH2B1β in these stably transfected cell lines (Fig. 5A). Neither SH2B1β nor SH2B1δ altered expression of these NGF-dependent genes in the absence of NGF. Because deletion of SH2B1 reduced neurite complexity and expression of PSD-95 in hippocampal neurons, we next tested whether deletion of SH2B1 would reduce BDNF-induced expression in hippocampal neurons of two immediate early genes important for synaptic function, early growth response 1 (Egr1) and activity-regulated cytoskeleton-associated gene (Arc) (Alder et al., 2003; Duclot and Kabbaj, 2017; Minatohara et al., 2015; Yin et al., 2002). As expected, BDNF treatment greatly increased expression levels of both genes in WT neuron cultures (Fig. 8E,F). However, in Sh2b1 KO neuron cultures, the BDNF-induced enhancement of the expression of both of these genes was markedly (by ∼50%) decreased. These results suggest that SH2B1 is critical for maximizing the well-documented BDNF-induced expression of immediate early genes Egr1 and Arc.

Fig. 8.

SH2B1δ enhances neurotrophin-induced expression of neuronal immediate early genes; enhancement is partially dependent on nucleolar SH2B1δ. (A–D) PC12 cells stably expressing GFP (−), GFP–SH2B1β or GFP–SH2B1δ were treated with or without NGF (100 ng/ml, 6 h). Relative mRNA levels of (A) Plaur, (B) Mmp3, (C) Mmp10 and (D) FosL1 were determined using qPCR. Data were normalized to NGF-treated GFP–SH2B1β values. n=3 biological replicates. (E,F) WT or KO hippocampal neurons (DIV5) were treated with or without BDNF (50 ng/ml, 1 h). Relative mRNA levels of (E) Egr1 and (F) Arc were determined using qPCR. n=3 biological replicates. (G–I) WT or KO neurons (DIV8) were infected with lentivirus encoding GFP, GFP–SH2B1β or GFP–SH2B1δ and treated with or without BDNF (50 ng/ml, 2 h) (DIV14). Relative mRNA levels of (G) FosL1, (H) Egr1 and (I) Arc were determined using qPCR. n=3−5 biological replicates; one set of biological replicates of GFP–SH2B1β (with or without BDNF) was averaged from duplicate samples. Data are means±s.e.m. Statistics: A–D, one-way ANOVA (randomized block experiment) on NGF-treated samples with Holm–Sidak's multiple comparisons test (*P<0.05); one-way ANOVA (randomized block experiment) on untreated samples (no significant differences detected); E,F, one-tailed paired t-test on BDNF-treated samples (*P<0.05); one-tailed paired t-test on untreated samples (no significant differences detected); G–I, one-way ANOVA (randomized block experiment with mixed effects analysis) with Holm–Sidak's multiple comparisons test on BDNF-treated samples, GFP versus GFP–SH2B1β versus GFP–SH2B1δ (*P<0.05); one-tailed paired t-test on BDNF-treated samples, GFP–SH2B1δ versus GFP–SH2B1δ(mNLS2+3) (^#P<0.05); one-way ANOVA (randomized block experiment with mixed effects analysis) with Holm–Sidak's multiple comparisons test on untreated samples (no significant differences detected).

Nucleolar localization is required for maximal SH2B1δ enhancement of BDNF-induced expression of neuronal immediate early genes

We next questioned whether ectopic expression of SH2B1β or SH2B1δ would increase BDNF-induced gene expression in Sh2b1 KO hippocampal neurons. We introduced GFP, GFP–SH2B1β or GFP–SH2B1δ into KO neuron cultures on DIV8 using lentiviral transduction. By DIV14, fluorescence microscopy indicated expression in nearly 100% of the cells. Therefore, at DIV14, we added 50 ng/ml BDNF for 2 h and assessed by qPCR the expression of Egr1 and Arc. We also measured expression of FosL1, one of the NGF-sensitive genes whose expression was increased by SH2B1δ in PC12 cells. Both SH2B1β and SH2B1δ enhanced BDNF-induced expression of FosL1 (Fig. 8G). However, only SH2B1δ enhanced BDNF-induced expression of Egr1 and Arc (Fig. 8H,I). We then asked whether nucleolar localization of SH2B1δ was required for its stimulatory effect. We found that mutating both NLS2 and NLS3, which prevents nucleolar localization of SH2B1δ, substantially impaired the ability of SH2B1δ to enhance BDNF-induced expression of FosL1, Egr1 and Arc (Fig. 8G–I). qPCR revealed that the observed differences in gene expression could not be explained by differences in level of expression of the different forms of SH2B1 (Fig. S1). None of these SH2B1 constructs altered expression of these BDNF-dependent genes in the absence of BDNF. Taken together, these results suggest that, in primary hippocampal neurons, nucleolar localization is critical for the ability of SH2B1δ to increase neurite complexity and BDNF-induced gene expression.

DISCUSSION

Adapter proteins play an essential role in neurite growth (Vessey and Karra, 2007). Here, we investigated the role of the different alternatively spliced isoforms of adapter protein SH2B1 in neurite complexity and length as well as neurotrophin-induced gene expression in PC12 cells and primary hippocampal neurons.

SH2B1 and neuronal architecture

Our assessment of WT and Sh2b1 KO hippocampal neurons demonstrated that SH2B1 is an important determinant of neuronal architecture. KO neurons exhibited a decrease in the number of endpoints, length of the longest neurite and total neurite length. These results are consistent with previous reports, which indicate that shRNA-driven reduction of SH2B1 levels decreases neurite outgrowth and/or branching in PC12 cells (Maures et al., 2009) and in primary cortical (Shih et al., 2013) and hippocampal (Chen et al., 2015) neurons. Our extensive morphological Sholl analyses revealed that deletion of SH2B1 suppressed arborization of both proximal and distal neurites, and that reintroduction of any individual isoform into KO neurons increased neurite arborization. The α, β and δ isoforms also increased the number of endpoints and the δ isoform additionally increased total neurite length. The ability of SH2B1β to increase the number of endpoints of hippocampal neurons is consistent with previous overexpression studies (Chen et al., 2015; Shih et al., 2013). However, we also found that neither arborization of distal neurites nor length of the longest neurite changed with the reintroduction of any individual isoform into KO neurons, suggesting that in WT neurons, interaction among the different SH2B1 isoforms may be necessary to enhance complexity or length of axons and/or distal dendrites. Thus, the shared region of SH2B1 may provide a baseline ability for the isoforms to facilitate the coordinated regulation of the actin cytoskeleton that ultimately drives morphological changes and neurite outgrowth. The unique C-terminal tails of each isoform may modify how these functions are performed.

In PC12 cells, the effect of SH2B1 isoforms on neurite outgrowth is thought to be mediated primarily through TrkA because SH2B1-induced neurite outgrowth is NGF dependent. Furthermore, mutating the SH2 domain of SH2B1β prevents SH2B1β binding to TrkA and prevents SH2B1β enhancement of NGF-induced signaling, gene expression and neurite outgrowth (Chen et al., 2008; Maures et al., 2009; Rui et al., 1999; Wang et al., 2004). However, primary hippocampal neurons synthesize and secrete BDNF (Ivanova and Beyer, 2001). Our experiments were also performed in the presence of B-27 supplement, which contains insulin. Thus, it is difficult to know to what extent the observed effects of SH2B1 on neuronal architecture are due to SH2B1 interactions with TrkB, the insulin receptor or other SH2B1-interacting proteins implicated in the regulation of the actin cytoskeleton. However, because SH2B1 is recruited to receptor tyrosine kinases, including the receptor for BDNF (TrkB), and because SH2B1 enhances BDNF-dependent expression of genes implicated in neuronal development and function, it seems likely that the actions of SH2B1 on neuronal architecture are a consequence, at least in part, of its recruitment to TrkB. Indeed, when we prevented SH2B1δ from being recruited to phosphorylated tyrosine residues in receptor tyrosine kinases by mutating its SH2 domain (R555E), the ability of SH2B1δ to increase neurite complexity was greatly decreased. However, GFP–SH2B1δ(R555E) was still able to increase neurite complexity to some degree compared to control KO neurons expressing only GFP. This raises the possibility that SH2B1δ increases neurite complexity, at least in part, through mechanisms that are independent of its recruitment to activated tyrosine kinases. One possibility is that SH2B1δ facilitates rearrangement of actin filaments that form proximal dendrites, even in the absence of tyrosine kinase activity. Previous work has shown that SH2B1 promotes actin cytoskeleton rearrangement and cell motility and regulates the number of focal adhesions (Chen et al., 2015; Diakonova et al., 2002; Herrington et al., 2000; Rider and Diakonova, 2011; Rider et al., 2009; Su et al., 2013). Furthermore, SH2B1β enhances actin-based movement of ActA-coated beads in a biomimetic actin-based motility assay (Diakonova et al., 2007). Rac, a small GTPase implicated in membrane ruffling, lamellipodia formation and cell motility, binds to the proline-rich region (residues 85–106) of the shared region of SH2B1 (Diakonova et al., 2002). In primary neurons, SH2B1β–IRSp53 complexes have been implicated in the formation of filopodia, neurite initiation and neurite branching (Chen et al., 2015). IRSp53 (also known as BAIAP2) interacts with and regulates actin cytoskeleton-associated proteins involved in filopodium formation (Zhao et al., 2011). Thus, most of the motifs/sites required for SH2B1β to enhance actin cytoskeleton-related activity (e.g. enhance growth hormone-induced membrane ruffling and cell motility, cross-link actin filaments, or bind to actin-related proteins including Rac and IRSp53) are located in the shared N-terminal region of SH2B1 outside the SH2 domain (Chen et al., 2015; Diakonova et al., 2002; Lanning et al., 2011; Rider and Diakonova, 2011; Rider et al., 2009).

SH2B1 regulation of synapse-related gene expression

Consistent with the ability of all SH2B1 isoforms to enhance neurite complexity in hippocampal neurons, we present substantial evidence that SH2B1 isoforms, including SH2B1δ, are important for synaptic function in neurons. First, PSD-95 expression was decreased in hippocampal neuron cultures from Sh2b1 KO mice. PSD-95 is a scaffold protein that organizes postsynaptic signaling complexes that include glutamate receptors, ion channels, signaling enzymes and adhesion proteins (Fernandez et al., 2009; Husi et al., 2000). Second, the ability of SH2B1β and SH2B1δ to promote NGF-induced expression of Plaur, Mmp3 and Mmp10 in PC12 cells has strong implications for a role of SH2B1 in synaptic plasticity. Extracellular proteolysis, to which these genes contribute, has been shown to be important for plasticity of glutamatergic excitatory synapses (Nagappan-Chettiar et al., 2017; Sonderegger and Matsumoto-Miyai, 2014) and GABAergic inhibitory synapses (Wiera et al., 2020). Third, we found that SH2B1 is critical for BDNF-induced expression of immediate early genes Egr1, Arc and FosL1 in hippocampal neuron cultures. Expression of both Egr1 and Arc was decreased in KO hippocampal neurons, and expressing SH2B1δ in KO hippocampal neurons substantially increased expression of FosL1, Egr1 and Arc. Immediate early genes contribute to a rapid and dynamic response to neuronal activity that can cause lasting change within the cell. FOSL1 transcription factor regulates expression of several proteins that mediate synaptic morphology (West and Greenberg, 2011). EGR1 is a regulatory transcription factor that broadly influences neuronal function by regulating the expression of multiple downstream genes involved in a variety of neuronal processes including synaptic plasticity (Duclot and Kabbaj, 2017). Indeed, EGR1 triggers the transcription of Arc following synaptic activation. Both Arc mRNA and ARC protein accumulate in dendrites following synaptic activity (Moga et al., 2004; Steward et al., 1998). ARC facilitates AMPA receptor endocytosis within the postsynaptic compartment (Chowdhury et al., 2006; Shepherd et al., 2006) and is a critical enhancer of actin cytoskeleton rearrangement in dendritic spines (Messaoudi et al., 2007). Both EGR1 and ARC have been shown to be important for synaptic plasticity, long-term potentiation and memory consolidation (Duclot and Kabbaj, 2017; Minatohara et al., 2015). ERK1/2, Akt and PLCγ have been implicated in the regulation of these genes and in neuronal development. However, because SH2B1 isoforms cause relatively small changes in these signaling proteins, it is unclear whether the combined effect of these four signaling proteins is sufficient to explain the much larger effects of SH2B1β and SH2B1δ on neurotrophin-induced gene expression and neuronal architecture. It has been hypothesized that other proteins/signaling pathways, not yet identified, may be involved (Maures et al., 2009; Rui and Carter-Su, 1999).

Taken together, these findings suggest that SH2B1 is a critical component of neurotrophin-induced gene expression associated with neuronal synapses. Future studies will be necessary to clarify whether all, or only some, SH2B1 isoforms contribute to SH2B1 regulation of neurotrophin-induced gene expression and whether SH2B1 promotes synapse-related gene expression in all, or only some, subsets of hippocampal neurons.

Human obesity-associated SH2B1 variants

Disruption or deletion of SH2B1 in humans and mice negatively impacts feeding and social behaviors (Doche et al., 2012; Flores et al., 2019; Jiang et al., 2018; Pearce et al., 2014; Ren et al., 2005). SH2B1δ (and SH2B1α) are expressed almost exclusively in brain tissue (Doche et al., 2012), deletion of SH2B1α and SH2B1δ decreases body weight in mice (Cote et al., 2021), and the brain is the major regulator of energy balance. Therefore, we predicted that human obesity-associated SH2B1 variants would affect the ability of SH2B1δ to regulate the development and maintenance of neuronal circuitry governing energy balance. Thus, we tested the impact of obesity-associated variants on the ability of SH2B1δ to increase neurite complexity and length in hippocampal neurons. Sholl analysis revealed that the P322S variant located in the shared region of SH2B1 dramatically impaired the ability of SH2B1δ to increase the complexity of proximal neurites, whereas the R680C variant, unique to SH2B1δ, dramatically enhanced the ability of SH2B1δ to increase the complexity of proximal neurites. Thus, while SH2B1 isoforms, particularly SH2B1δ, may increase neurite complexity and/or length important for the development and maintenance of appetite- and behavior-regulating neuronal networks, the opposing effects of the human SH2B1 variants suggest that either a decrease or increase in neurite complexity may have deleterious effects that contribute to the obesity and neurobehavioral abnormalities associated with these human variants.

Localization of SH2B1δ in nucleoli

Somewhat surprisingly, SH2B1δ localizes primarily to nucleoli, with some visible in the nucleus, plasma membrane and, in hippocampal neurons, the cytoplasm. The unique nucleolar localization of SH2B1δ distinguishes it not only from other SH2B1 isoforms but also from most other proteins. Only 7% of all human proteins have been experimentally identified as being present in nucleoli (Thul et al., 2017). Of those, fewer than 50 are seen both at the plasma membrane and in nucleoli. Yousaf and colleagues (2001) identified two highly basic regions (NLS2 and NLS3) and a tryptophan-rich segment within the unique C-terminal tail of SH2B1δ and suggested that the two highly basic regions may target the SH2B1δ isoform to the nucleus. Our mutational analysis indicated that instead, SH2B1δ requires both of these highly basic regions to localize to nucleoli. In addition, as seen with SH2B1β (Maures et al., 2009), the NLS1 domain is required for SH2B1δ to localize to the plasma membrane. Surprisingly, the tryptophan-rich segment (residues 712–724) is also required for SH2B1δ to localize to the plasma membrane. The presence of these specialized sequences suggests that SH2B1δ may shuttle between the plasma membrane and nucleoli.

Experiments using SH2B1δ mutants with different subcellular distributions suggest that SH2B1δ must localize to both nucleoli and the plasma membrane and have a functional SH2 domain to maximally increase neurite complexity. The requirement for SH2B1δ to localize at the plasma membrane and have a functional SH2 domain is consistent with SH2B1δ being recruited to activated Trk receptors at the plasma membrane. Indeed, we showed not only that SH2B1δ promoted NGF-induced gene expression in PC12 cells and BDNF-induced gene expression in Sh2b1 KO hippocampal neurons, but also that BDNF-induced gene expression was greatly impaired in KO hippocampal neurons. When compared with the NLS1 or W712X mutations, disruption of the function of the SH2 domain in SH2B1δ via the R555E mutation resulted in a more dramatic morphological change in hippocampal neuron architecture, greatly reducing the ability of SH2B1δ to increase the number of endpoints, total neurite length and number of intersections. The more disruptive phenotype caused by the R555E mutation suggests that this mutation prevents cellular functions that are not affected by the W712X or NLS1 mutations. One possible explanation for this difference is that the W712X and NLS1 mutations do not completely prevent plasma membrane localization of SH2B1δ, allowing in some cells enough SH2B1δ to access the plasma membrane for recruitment to phosphotyrosines in activated TrkB. While localization at the plasma membrane via NLS1 may enhance the ability of SH2B1δ to bind Trk receptors, SH2B1δ binding to TrkB may only require small or transient amounts of SH2B1δ at the plasma membrane, which might escape detection when visualized during the steady state. Indeed, in a small percentage (<10%) of the PC12 cells expressing either SH2B1δ(W712X) or SH2B1δ(mNLS1), some SH2B1δ was found at the plasma membrane. Alternatively, SH2B1δ may affect some cellular functions as a consequence of binding via its SH2 domain to phosphotyrosines located elsewhere in the cell (e.g. on internalized receptor tyrosine kinases, non-receptor tyrosine kinases or cytoskeletal proteins).

Together, these results suggest that both plasma membrane- and nucleolar-localized SH2B1δ play critical roles in SH2B1δ neuronal function. Based on the function of other nucleolar proteins (Hetman and Pietrzak, 2012; Iarovaia et al., 2019), one possibility for the role of nucleolar SH2B1δ is that it serves as a scaffold protein that facilitates the formation of multiprotein complexes in nucleoli, thereby enhancing their function (e.g. BDNF-induced ribosomal biogenesis). Alternatively, SH2B1δ may transport other proteins to nucleoli for sequestration, similar to Cfi–Net1 complexes sequestering Cdc14 in nucleoli until the onset of anaphase and thereby preventing it from becoming active (Visintin and Amon, 2000).

Conclusion

The current work advances our understanding of cellular actions of SH2B1 isoforms. We demonstrate that the unique C-terminal tail of the brain-specific δ isoform of SH2B1 facilitates its subcellular localization to nucleoli. SH2B1δ promotes neurotrophin-stimulated signaling events and gene expression in PC12 cells and increases neurite arborization and length in neurotrophin-responsive hippocampal neurons. We also demonstrate that SH2B1 isoforms are critical regulators of not only neuronal architecture, but also expression of synapse-related genes. Together, these findings provide novel insight into the cellular functions of the SH2B1 isoforms in the context of nervous system development and function.

MATERIALS AND METHODS

Antibodies

Primary antibodies

Mouse monoclonal antibody to SH2B1 (sc-136065; 1:1000) was from Santa Cruz Biotechnology (Dallas, TX). Rabbit polyclonal antibody to GFP (632592; 1:2000) was from Clontech (San Jose, CA). Rabbit polyclonal antibody to PSD-95 (AB9708) (1:1000) was from Sigma-Aldrich (St Louis, MO). Chicken polyclonal antibody to GFP (AB_2307313; 1:500) was from Aves Labs (Davis, CA). Rabbit polyclonal antibodies to PLCγ (2822; 1:1000), phospho-PLCγ (Y783) (2821; 1:500), doubly phosphorylated ERK1/2 (T202/Y204) (9101S; 1:1000), rabbit monoclonal antibody to phospho-Akt (S473) (4058; 1:1000), and mouse monoclonal antibodies to pan-Akt (2920S; 1:2000) and ERK1/2 (4696S; 1:2000) were from Cell Signaling Technology (Danvers, MA).

Secondary antibodies

Goat anti-chicken-IgY conjugated to Alexa Fluor 488 (Invitrogen A-11039) was from Thermo Fisher Scientific (Waltham, MA). Infrared dye-conjugated goat anti-mouse-IgG 680LT (926-68020; 1:15,000), 680RD (926-68070; 1:15,000), 800CW (926-32210; 1:20,000) and goat anti-rabbit-IgG 680RD (926-68071; 1:15,000) came from Li-Cor Biosciences (Lincoln, NE).

Verification

We previously verified the antibody to SH2B1 by confirming that the band was absent in tissues from Sh2b1 KO mice (Cote et al., 2021; Flores et al., 2019). All other antibodies are well documented antibodies that identify proteins migrating with the size reported for that protein.

Plasmids

cDNAs encoding rat GFP–SH2B1β (GenBank accession #NM_001048180; Herrington et al., 2000) and mouse GFP–SH2B1α (GenBank accession #NM_001289538.1; Joe et al., 2017) were described previously. cDNAs encoding mouse SH2B1γ (GenBank accession #NM_011363.3) and SH2B1δ (GenBank accession #NM_001289541) (Yousaf et al., 2001) (from Dr Heimo Riedel, formerly at West Virginia University, USA) were subcloned into pEGFPC1 (Clontech, San Jose, CA). cDNA encoding mouse GFP–SH2B1β (GenBank accession #NM_001081459.2) was created from cDNA encoding mouse GFP–SH2B1γ by removing the γ/δ-specific exon 8b and increasing the length of exon 10 to include the 19 nucleotides in exon 10 that encode the final 5 amino acids and stop codon (TEHLP*) of SH2B1β using the QuikChange II site-directed mutagenesis kit (200521; Agilent, Santa Clara, CA). Mutations were introduced into SH2B1δ using site-directed mutagenesis (see primers in Table S2). cDNA encoding mCherry-tagged human nucleolin (GenBank accession #NM_005381.2) was created from cDNA encoding GFP-tagged human nucleolin (Addgene plasmid #28176) using AgeI and BsrGI restriction enzymes. All sequences were confirmed by the University of Michigan DNA Sequencing Core or Eurofin Genomics (Louisville, KY).

To create the pLentilox pro-viral plasmids, PfuUltra High-Fidelity DNA Polymerase (#600380; Agilent) was used to produce a PCR fragment containing the desired SH2B1 isoform flanked by AgeI (ACCGGT) and BamHI (GAATTC) restriction sites from GFP-tagged mouse SH2B1β and SH2B1δ (GFP N-terminal and isoform-specific C-terminal tail primers described in Table S3). The PCR product was then cleaved using AgeI and BamHI, and subcloned into the GFP-pLentilox 3.7 pro-viral plasmid (Addgene plasmid #11795; deposited by Luk Parijs; provided by the University of Michigan Vector Core). The presence of this insert was confirmed by DNA sequencing (Eurofins Genomics, Louisville, KY). The resulting GFP-SH2B1 isoform pLentilox pro-viral plasmids were provided to the University of Michigan Vector Core for the production of a 10× concentrated active lentivirus for transduction.

PC12 cell culture and transfection

PC12 cells (CRL-1721; ATCC, Manassas, VA) were grown on 10 cm tissue culture dishes (0877222; Corning, Corning, NY) coated with rat tail type I collagen (354236; Corning, Corning, NY) and maintained at 37°C with 5% CO₂ in normal growth medium (RPMI 1640; A10491-01; Life Technologies, Carlsbad, CA) supplemented with 10% horse serum (HS) (Gibco 16050-122; Thermo Fisher Scientific, Waltham, MA) and 5% fetal bovine serum (FBS) (S11150; Atlanta Biologicals, Miami, FL). PC12 cells were transiently transfected using Lipofectamine LTX (Invitrogen, Thermo Fisher Scientific, Waltham, MA). PC12 cell lines stably expressing GFP, GFP–SH2B1β or GFP–SH2B1δ were prepared and maintained as described previously (Joe et al., 2017). Prior to experiments requiring NGF stimulation, PC12 cells were incubated overnight in deprivation medium [RPMI 1640 containing 1% bovine serum albumin (BSA); 7500804; Proliant Biologicals, Ankeny, IA]. Recently thawed stocks of PC12 cells were periodically tested to confirm the absence of mycoplasma.

Animal care and models

Mice (Mus musculus) were housed in ventilated cages (∼22°C) on a 12-h light–12-h dark cycle (∼6am to 6pm) in a pathogen-free animal facility at the University of Michigan. Food and water were available ad libitum. Breeders were given standard chow containing 6.5% fat (5008; LabDiet, St Louis, MO) or 9% fat (5058; LabDiet, St Louis, MO). The previously described (Duan et al., 2004) Sh2b1 KO mouse strain was obtained from Dr Liangyou Rui (University of Michigan, Ann Arbor, MI). C57BL/6J mice (000664) were obtained from The Jackson Laboratory (Bar Harbor, ME). Experiments were approved by the University of Michigan Institutional Animal Care & Use Committee.

Primary hippocampal neuron culture

Hippocampal neuron cultures were prepared from male and/or female mouse neonates from C57BL/6J×C57BL/6J breeding pairs (Fig. 1B,C) or Sh2b1^+/−×Sh2b1^+/− breeding pairs (all other figures) on postnatal day 0, 1 or 2. The day of neuron plating was considered DIV0. Briefly, hippocampi were dissected in ice-cold dissociation medium (DM) (82 mM Na₂SO₄, 30 mM K₂SO₄, 5.8 mM MgCl₂, 0.25 mM CaCl₂, 1 mM HEPES, 20 mM glucose, 0.001% Phenol Red, pH 7.4) under a stereo microscope (SZ51; Olympus, Center Valley, PA). Isolated hippocampi were treated with L-cysteine (C7352; Sigma-Aldrich, St Louis, MO)-activated papain (pH 7.4) (LK003178; Worthington Biochemical, Lakewood, NJ) for 30 min at 37°C. They were washed first with DM supplemented with 10% FBS followed by DM without FBS and then transferred to neuron growth medium [Gibco Neurobasal Plus medium (A3582901; Thermo Fisher Scientific, Waltham, MA), 2% B-27 Plus (Gibco A3582801; Thermo Fisher Scientific, Waltham, MA), 1% penicillin/streptomycin (Gibco 15140-122; Thermo Fisher Scientific, Waltham, MA)]. Hippocampi were dissociated by triturating the tissue 10–20 times using a 5 ml serological pipette. The cell suspension was placed on ice for 3 min and centrifuged at 600 g for 5 min at 4°C. Pelleted cells were resuspended in neuron growth medium and plated onto coverslips as detailed below. Cultures were maintained at 37°C in 5% CO₂ in neuron growth medium.

Live-cell confocal imaging and analysis

Prior to live-cell imaging, PC12 cells were grown overnight on 35 mm poly-D-lysine-coated glass bottom dishes (P35GC-1.5-15-C; MatTek Life Sciences, Ashland, MA) and then transiently transfected with the indicated constructs. Live cells were imaged in Ringer's buffer (10 mM HEPES, 155 mM NaCl, 2 mM CaCl₂, 5 mM KCl, 1 mM MgCl₂, 10 mM NaH₂PO₄, 10 mM D-glucose, pH 7.2) using a 60× water immersion objective on a Nikon A1 laser-scanning confocal microscope (Nikon Instruments, Melville, NY) and Nikon NIS-Elements software (Microscopy, Imaging, and Cellular Physiology Core of the Michigan Diabetes Research Center). The plasma membrane was labeled with wheat germ agglutinin (WGA) Alexa Fluor 647 conjugate (Invitrogen W32466; Thermo Fisher Scientific, Waltham, MA). Focal planes were chosen to allow for maximum visualization of nucleoli and the plasma membrane, and each cell was imaged at both low and high gain by adjusting the voltage to the photomultiplier tubes on the microscope. The relative levels of GFP–SH2B1 in the nucleoli, nucleus, cytoplasm and plasma membrane were assessed by a researcher who was blind to the experimental condition. Levels were designated ‘bright’ if visible at low gain, ‘dim’ if visible at high gain but not at low gain, or ‘absent’ if not visible even at high gain. Sample size (∼20–30 cells/condition) was chosen based on our previous work (Joe et al., 2017). Cells were excluded from analysis if they exhibited unusually high or low levels of expression or were dead or unhealthy (<10% of cells).

Hippocampal neurons (90,000 cells/dish) were plated onto 35 mm glass bottom dishes (P35G-1.5-14-C; MatTek Life Sciences, Ashland, MA) that had been coated with poly-D-lysine (Gibco A3890401; Thermo Fisher Scientific, Waltham, MA) prior to plating. Neurons were transfected using Lipofectamine 3000 (Invitrogen L3000-008; Thermo Fisher Scientific, Waltham, MA) and 1 μg total of the indicated cDNAs. Live neurons were imaged using 40× and 60× water immersion objectives on a Nikon A1 laser-scanning confocal microscope (Nikon Instruments, Melville, NY). Focal planes were chosen to maximize visualization of the nucleus and plasma membrane.

Immunocytochemistry

Hippocampal neurons (90,000 cells/coverslip) were plated onto 15 mm glass coverslips pre-coated with poly-D-lysine (GG-15-PDL; Neuvitro Corporation, Vancouver, WA) and laminin (L2020-1MG; Sigma-Aldrich). Cultures were fixed using 4% paraformaldehyde (15710; Electron Microscopy Sciences, Hatfield, PA) for 20 min at room temperature and permeabilized using 0.3% Triton X-100 in immunocytochemistry (ICC) blocking buffer [phosphate-buffered saline (PBS) containing 1% BSA]. Following 30 min in ICC blocking buffer, cells were incubated with primary antibody for 1 h at room temperature or overnight at 4°C. Cultures were incubated with secondary antibody for 1 or 2 h at room temperature. Coverslips were mounted on glass microscope slides (22-035813; Fisherbrand, Waltham, MA) using ProLong Gold Antifade Mountant (Invitrogen P10144; Thermo Fisher Scientific, Waltham, MA) and stored at 4°C. Cells were imaged using a Nikon A1 laser-scanning confocal microscope or Eclipse TE200 fluorescence microscope (Nikon Instruments, Melville, NY).

Analysis of neuronal architecture

Hippocampal neurons (90,000 cells/coverslip) were plated as described above for ICC. Neurons were transfected on DIV4 using Lipofectamine 3000 and 1 μg total of the indicated cDNAs. Live neurons were imaged on DIV5 using 10× and 20× objectives on Nikon Eclipse TE200 fluorescence microscope. Neurons chosen for tracing had similar GFP–SH2B1 expression levels with the caveat that it was difficult to visually compare the expression level of GFP–SH2B1δ with that of other isoforms because of the overly bright signal of GFP–SH2B1δ concentrated in nucleoli. For neurons that did not fit in the field of view of the microscope, multiple images were taken and ‘stitched’ together using the Stitching plugin in FIJI2 (Preibisch et al., 2009). A subset of neuron images were selected for analysis from a larger depository of images using a random number generator. Sample size (>25 neurons/condition) was determined from preliminary data analysis (data not shown) and after consultation with the Biostatistical Services of the Michigan Diabetes Research Center. Neurites were traced using FIJI2 with the semi-automated Simple Neurite Tracer plugin (Longair et al., 2011) and then manually checked. Any gaps or errors in the traces were repaired manually. Traces were then analyzed for number of endpoints, total neurite length and length of the longest neurite. Sholl analysis was performed using the Sholl Analysis plugin from Simple Neurite Tracer (Ferreira et al., 2014) on the neuron traces mentioned above. Concentric circles were placed every 2 µm from a point in the center of the soma. The number of neurite intersections at each fixed distance from the soma was determined, as well as the peak number of intersections at any one circle and the maximal distance from the soma reached by the furthest neurite of each neuron. Neurons were excluded from all further analysis if any of their measurements were greater than three standard deviations from the mean. Researchers taking and tracing images were blind to experimental conditions.

Immunoblotting

PC12 cells were treated and lysed as previously described (Joe et al., 2017) using ice-cold modified L-RIPA buffer (20 mM Tris-HCl, 150 mM NaCl, 1% Triton X-100, 1 mM EGTA and 1 mM EDTA, pH 7.4) supplemented with 20 mM NaF, 1 mM Na₃VO₄ and protease inhibitors (1 mM PMSF, 10 μg/ml aprotinin and 10 μg/ml leupeptin). Hippocampal neurons (200,000 cells/coverslip) were plated as described above for ICC. Neurons were washed with Hank's basic salt solution or PBS (10 mM NaPO₄ and 140 mM NaCl, pH 7.4) and lysed using ice-cold modified L-RIPA buffer supplemented with 50 mM α-glycerophosphate, 1 mM Na₃VO₄ and protease inhibitor cocktail (1:100; P8340; Sigma-Aldrich). Cell lysates from 5–8 coverslips/condition were centrifuged at 15,000 g for 15 min at 4°C. Protein concentration of the supernatant was measured using the DC Protein Assay (5000116; Bio-Rad, Hercules, CA).

Equal amounts of proteins were resolved using 4–15% precast Criterion Tris-HCl protein gels (3450027; Bio-Rad, Hercules, CA) and transferred to low-fluorescence PVDF membranes (1620264; Bio-Rad, Hercules, CA). Blots were blocked for 1 h at room temperature in blocking buffer (10 mM Tris-HCl, 150 mM NaCl, pH 7.4, 0.1% Tween 20 and 3% BSA). Blots were incubated with primary antibody diluted in blocking buffer overnight at 4°C followed by secondary antibody diluted in blocking buffer for 1 h at room temperature. Bands were visualized using the Odyssey Infrared Imaging System and quantified using Image Studio Lite 5.2 (Li-Cor Biosciences, Lincoln, NE).

Neurite outgrowth assay

PC12 cells were plated in six-well dishes and transiently transfected with the indicated construct. Cells were treated with 25 ng/ml mouse NGF (BT 5025, Harlan Laboratories, Indianapolis, IN) in RPMI 1640 medium containing 2% HS and 1% FBS. GFP-positive cells were visualized by fluorescence microscopy (20× or 40× objective, Nikon Eclipse TE200) after 2 days. Next, 100 GFP-positive cells in three different areas of each plate were scored for the presence of neurites at least two times the length of the cell body (total of 300 cells per condition per experiment). The percentage of cells with neurites was determined by dividing the number of GFP-positive cells with neurites by the total number of GFP-positive cells counted.

RNA isolation and qPCR

PC12 cells were treated with or without 100 ng/ml NGF for 6 h and relative mRNA levels of Plaur, Mmp3, Mmp10 and FosL1 were determined as described previously (Joe et al., 2017).

Hippocampal neurons (200,000 cells/coverslip) were plated as described above for ICC. For Fig. 8E,F, neurons were treated with or without BDNF (50 ng/ml, 1 h) on DIV5. For Fig. 8G–I, neurons were transduced on DIV8 with lentivirus constructs encoding GFP, or GFP-tagged SH2B1β, SH2B1δ(WT) or SH2B1δ(mNLS2+3). Prior to lentiviral transduction, the medium (‘conditioned medium’, 1.5 ml) was removed from each well and incubated at 37°C (95% humidity, 5% CO₂). An equal volume of Neurobasal medium containing ∼1× viral concentration was added to each well, and cells were incubated at 37°C (95% humidity, 5% CO₂) for ∼20 h. The virus-containing medium was then replaced with 1.5 ml of the original conditioned medium. Transduction efficiency was monitored by fluorescence microscopy and found to be near 100% by DIV14. On DIV14, cells were incubated with or without BDNF (50 ng/ml for 2 h). RNA was isolated on the indicated DIV using the RNeasy Mini Kit (74104; QIAGEN, Germantown, MD). RNA quality was assessed using a Nanodrop spectrophotometer before RNA was reverse-transcribed using iScript cDNA Synthesis Kit (1708891; Bio-Rad, Hercules, CA). Relative mRNA levels were determined using Applied Biosystems' TaqMan Gene Expression Assays (dye, FAM-MGB; Thermo Fisher Scientific, Waltham, MA) (Table S1), which determine mRNA levels with an efficiency of 100±10%. qPCR was performed with an Eppendorf Realplex2 using Mastercycler software and PCR parameters recommended by Applied Biosystems (Waltham, MA). All cDNA samples were analyzed in triplicate and a non-template control was included for each gene of interest. Cycle threshold (Ct) values were normalized to the geometric mean of the Ct values of three reference genes (36b4, Gapdh and Tbp) as previously described (Vandesompele et al., 2002). Expression of the reference genes did not differ between control and experimental samples (data not shown).

Statistical analyses

Statistics were performed using GraphPad Prism 8 or 9. Statistical analyses are indicated in figure legends. Data are means±s.e.m. P<0.05 was considered significant.

Peer review history

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259179.