Exercise-induced enhancement of synaptic function triggered by the inverse BAR protein, Mtss1L

Exercise is a potent enhancer of learning and memory, yet we know little of the underlying mechanisms that likely include alterations in synaptic efficacy in the hippocampus. To address this issue, we exposed mice to a single episode of voluntary exercise, and permanently marked activated mature hippocampal dentate granule cells using conditional Fos-TRAP mice. Exercise-activated neurons (Fos-TRAPed) showed an input-selective increase in dendritic spines and excitatory postsynaptic currents at 3 days post-exercise, indicative of exercise-induced structural plasticity. Laser-capture microdissection and RNASeq of activated neurons revealed that the most highly induced transcript was Mtss1L, a little-studied I-BAR domain-containing gene, which we hypothesized could be involved in membrane curvature and dendritic spine formation. shRNA-mediated Mtss1L knockdown in vivo prevented the exercise-induced increases in spines and excitatory postsynaptic currents. Our results link short-term effects of exercise to activity-dependent expression of Mtss1L, which we propose as a novel effector of activity-dependent rearrangement of synapses.


Introduction
The beneficial cognitive effects of physical exercise cross the lifespan as well as disease 20 boundaries (1, 2). Exercise alters neural activity in local hippocampal circuits, presumably by enhancing learning and memory through short and long-term changes in synaptic plasticity (3,4). The dentate gyrus is uniquely important in learning and memory, acting as an input stage for encoding contextual and spatial information from multiple brain regions. This circuit is well suited to its biological function because of its sparse coding design, with only a few dentate granule cells active at any one time (5)(6)(7).
These properties also provide an ideal network to investigate how exercise-induced changes in activity-dependent gene expression affect hippocampal structural and synaptic 5 plasticity in vivo.
Most studies have focused on the cognitive effects of sustained exercise (8)(9)(10). However, memory improvements are well documented after acute bouts of exercise (11,12). Thus, the response to a single episode of exercise may be better suited to uncover cellular and 10 molecular cascades that form and rearrange synapses, which evolve over minutes to weeks following exercise. We reasoned that these mechanisms could be unmasked by a single period of voluntary exercise in vivo, followed by functional and molecular analysis of subsequent changes specific to neurons that were activated during the exercise.

15
Results cFos cre/ERT2/Luo transgenic mice (Fos-TRAP) provide valid proxies of neural activity (11 and fig. S1A) and a means to permanently label activated dentate granule cells. During a two-hour exposure to running wheels, mice ran approximately 3 km. We examined activated cells 3 days post-running in Fos-TRAP mice crossed with a TdT reporter line 20 (Fig. 1A). We used Fos immunohistochemistry at 1 hour post-exercise, confirming that this stimulates robust neuronal activity in mature granule cells (Fig. 1B). The increase in Fos expression, assessed by immunohistochemistry, matched the increase in TdTomato-positive cells (TdT + ) measured 3 or 7 day later in Fos-TRAP mice, indicating that activated granule cells were accurately and permanently labeled during the 2 hour time window. We refer to these cells as "exercise-TRAPed". To investigate whether a single bout of exercise activated a specific subset of granule cells, we exercise TRAPed dentate granule cells (TdT + ) and compared this population to an ensemble activated by a 5 subsequent re-exposure to exercise either 1 or 4 days later, as measured by Fos immunohistochemistry at 2 hours after the 2 nd exercise period (Fig. 1D). When the two exercise periods were separated by 24 hours, only 13% of exercise TRAPed cells were activated in the 2 nd exercise period. There was almost no overlap between the two neuronal ensembles (1%) when the two periods were separated by 4 days (Fig. 1D, right 10 panel). These results indicate that our exercise protocol activated stochastic, nonoverlapping sets of granule cells, consistent with the sparse coding design of this circuit.
Higher cortical information converges on the dentate gyrus through laminated perforant path axons from the entorhinal cortex. To examine whether a single exposure to exercise 15 altered structural plasticity in exercise-activated neurons, we analyzed exercise-TRAPed cells in mice 3 or 7 days post-exercise as compared to home cage controls ( Fig. 2A, left panels). Exercise-TRAPed cells showed a nearly 50% increase in dendritic spines at 3 days. This increase was limited to the outer molecular layer (OML), which interestingly, receives contextual and time information via the lateral perforant path (13,14), whereas 20 there was no change in dendritic spines in the middle molecular layer (MML), which receives spatial information via the medial perforant path (15,16). Consistent with a transient response to a single stimulus, spine density in the OML returned to baseline levels by 7 days (Fig. 2A, right panels). Exercise did not affect total dendritic length in OML or MML indicating that the increase in spine density reflected an increase in the number of synapses ( fig. S2).
The increase in spine density in the OML was associated with a corresponding increase in 5 functional synaptic input in exercise-TRAPed cells (Fig. 2B). Three days after exercise, we used acute brain slices to make simultaneous whole-cell voltage clamp recordings from an exercise-TRAPed granule cell (yellow, Fig. 2B, middle) and a neighboring control granule cell (green, Fig 2B, Table S1). These results indicate that dentate granule cells activated by a single bout of exercise show a laminar-specific increase in dendritic spines and in excitatory postsynaptic currents.

20
Activity-dependent gene expression is one of the main drivers of short and long-term changes in synaptic function (17). To identify exercise-activated genes underlying the observed structural plasticity in the dentate gyrus, we isolated RNA from RFP + nuclei of exercise-TRAPed cells (Fig. 3A,B)  We focused on Mtss1-like (metastasis-suppressor 1-like, Mtss1L), a protein that has been little studied in the adult nervous system (20). Mtss1L, the most enriched transcript in our experiments, was 9-fold elevated compared to non-activated neighboring cells, as 15 validated in exercise TRAPed cells by RT-PCR (Fig. 3D, fig. S5). To determine the spatiotemporal pattern of Mtss1L expression, we derived KOMP Mtss1L reporter mice (21) in which the endogenous Mtss1L promoter drives bacterial beta-galactosidase (lacZ) gene expression in a Cre-dependent manner. LacZ expression was undetectable in the dentate gyrus of KOMP Mtss1L +/housed in their homecage, suggesting no expression of 20 Mtss1L under baseline conditions. In contrast, exercise-induced LacZ expression peaked at 3 days post-exercise (Fig. 4), confirming the activity-dependence of Mtss1L expression in the dentate gyrus.
Mtss1L belongs to the inverse-BAR (Bin, Amphiphysin and Rvs, I-BAR) protein family, a class of proteins that function by promoting curvature in membranes. BAR domain proteins such as amphyphysin promote convex curvatures including invaginations that are involved in endocytosis or synaptic vesicles at presynaptic nerve terminals (22). In 5 contrast, I-BAR proteins promote concave structures typically seen in outward-projecting membrane protrusions. Thus, we hypothesized that Mtss1L expression mediated exercise-dependent formation or rearrangement of synapses in postsynaptic dendrites.
As proof of principle, we first examined the localization of endogenous Mtss1L in 10 hippocampal neurons in vitro following exposure to brain-derived neurotrophic factor (BDNF), which is upregulated by exercise and involved in activity-dependent synaptic plasticity (23,24). Mtss1L immunoreactivity was detectable only after BDNF treatment Mtss1L also contain postsynaptic proteins (Fig. 5C). In vivo, overexpression of mCherry-Mtss1lL by DNA electroporation at P0 increased spine density of mature dentate granule cells at P21 (Fig. 5D). These results indicate that Mtss1L is expressed in dendrites and can induce spine-like protrusions, consistent with a putative role in synaptic plasticity. 5 To determine whether Mtss1L was responsible for the increase in synapses in exercise . The shscramble-GFP, injected into the contralateral dentate gyrus of the same mice, had no effect ( Fig. 6B right panel). Viral knockdown of Mtss1L in exercise-TRAPed cells also prevented the increase in evoked EPSCs in simultaneous paired recordings from TdTand TdT + /GFP + granule cells (Fig. 6C-D). The amplitude of EPSCs 20 following viral knockdown in exercise-TRAPed cells was the same as neighboring cells that were not activated by exercise (Fig 6D, right panel). Intrinsic membrane properties were not affected by Mtss1L knockdown (Table S4). Figure 6E includes the results from Importantly, evoked EPSCs in granule cells that only expressed GFP+shRNA were unaffected, demonstrating that the effect of Mtss1L was dependent on activity-dependent 5 expression ( fig. S8).

Discussion
Our experiments were designed to test the cellular and molecular response to acute exercise with an emphasis on time periods during which synapses might form or 10 reorganize. This approach differs from studies of sustained or chronic exercise that likely involve systemic as well as neural mechanisms (3,4,26). Such studies have shown effects of exercise on many aspects of hippocampal function, including neurogenesis, synaptic plasticity, structural changes at synapses and dendritic spines as well as behavioral effects on learning and memory (1, 2). Our goal was to examine the 15 transcriptional response in the days following acute exercise, which led to the identification of a novel effector of structural plasticity, Mtss1L.
Although the Fos-TRAP method is limited by the time required for reporter expression (ca. 1 day post-exercise), the immediate early gene Egr1 was upregulated (Fig. 3C), 20 indicating that our protocol detected activity-dependent gene expression in mature granule cells. Although the presence of the running wheel itself could act as a novel object, there was only a small increase in Fos-TRAPed cells when the wheel was fixed to prevent running (data not shown). It is interesting that adult-generated granule cells in the subgranular zone were not activated by two hours of exercise (data not shown). Although it is well known that chronic exercise increases adult neurogenesis (27), progenitors and newborn neurons do not receive direct excitatory perforant path input for several weeks (28), perhaps explaining their lack of activation in our experiments. 5 Our results provide the first evidence for activity-dependent expression of an I-BAR protein, and suggest that I-BAR proteins can affect both constitutive dendritic spine formation (29) as well as experience-dependent remodeling of synapses. The immediate family members of Mtss1L include MIM/Mtss1 and IRSp53 (30), but their function in 10 the nervous system is just beginning to be explored. For example, Mtss1L expression was previously detected in radial glial cells during development as important for membrane protrusions and end-feet (20). The related I-BAR protein, MIM/Mtss1, is involved in cerebellar synapse formation with signaling including PIP2-dependent membrane curvature and subsequent Arp2/3-mediated action polymerization (29). MIM and another 15 member of the I-BAR subfamily, IRSp53, have also been implicated in aspects of synapse formation during development (31,32).
At a circuit level, the laminar-specific increase in synaptic function suggests that acute exercise had a network-specific effect. Entorhinal inputs to the outer molecular layer, the 20 distal dendrites of mature granule cells, have been previously associated with contextual information and particularly the "how" and "when" aspects of encoding memories (33).
How might these results related to the beneficial effects of exercise? One possibility consistent with our data is that exercise acts as a preconditioning signal that primes exercise-activated neurons for contextual information incoming during the several days following exercise. This represents a broader time window than is usually associated with short-term plasticity. For example, human studies give support for the idea that exercise within 4 hours of a learning task improved memory performance (34). It will be 5 interesting to examine if exercise enhances the pattern of granule cell responses to spatial or context-specific tasks. Our identification of Mtss1L as an activity-dependent I-BAR domain protein makes it ideally suited to act at an early mediator of structural plasticity following neural activity.  Voluntary Exercise Protocol. Both homecage and exercise groups were housed together until 7 days before the experiment, when they were singled housed in an oversized (rat) sedentary cage (43×21.5 cm 2 ) to allow acclimation in the novel environment before tamoxifen administration. At 23hrs post-tamoxifen injection, a running wheel was introduced in the cage of animals in the exercise group and mice had free access to running for 2hrs at the beginning of the dark period, after which the running wheel was removed from the cage. Total distance (km) was measured using an odometer. All mouse 5 groups were handled for 5 days before tamoxifen administration.
Coronal sections (100µm) of the hippocampus were collected and permeabilized in 0.4% Dendritic spines were imaged and analyzed in the middle (MML) and outer molecular 10 (OML) layers of the dentate gyrus. The MML and OML were distinguished based on the pattern of VGluT2 immunofluorescence, which begins at the border between the inner molecular layer and middle molecular layer. MML dendritic segments were therefore imaged at the beginning of the VGluT2 staining nearest the granule cell body layer, whereas OML dendritic segments were imaged at the distal tip of the molecular layer. 15 The same microscope settings (laser intensity and gain) were used for each experimental group analyzed. Slides were coded and imaged by an investigator blinded to experimental condition.
Laser Capture Microdissection (LCM). To obtain nuclei of exercise-TRAPed cells, fresh 20 frozen coronal brain sections (12 µm) were collected by cryostat sectioning on polyethylene napthalate membrane slides. Sections were fixed for 1 min in 75% ethanol, followed by a final wash in 100% ethanol for 1 min. Pools of 150 RFP + and RFP -individual cells were laser captured from the granular cell layer of the dentate gyrus of each mouse using a Leica Microsystem LMD system under a HC PL FL L 63/0.60 XT objective and collected in lysis buffer (50 ul RLT buffer, Qiagen). Total RNA from LCM samples was isolated separately from each animal.

5
RNA extraction and RNAseq library preparation. RNA was extracted from laser-captured materials using RNeasy mini-elute column with a modified protocol. Briefly, cells were picked from laser capture microscopy onto caps of 0.5ml Eppendorf tubes, 50ul RLT buffer was added immediately, and placed on ice. After spinning down the samples, 300ul RLT was added to each sample, mixed well, and vortexed for 30s. After brief spin, 10 2ng tRNA was added as carrier and 625µl 100% ETOH was added into each sample and mixed well. After a quick spin, samples were loaded onto RNeasy mini-elute columns, washed, dried and eluted in 10ul RNAse-free water. 8ul RNA was used for reverse transcription using Superscribe reverse transcriptase 2ul, DTT 0.1M 0.5ul, 5x RT buffer 4ul at 42degrees Celsius for 1.5hr then heat inactivated at 75 °C for 20 minutes (self-15 developed protocol for RNA extraction, superscribe reverse transcriptase was from Clontech, protocol adopted from manufacturer's protocol for smarter single cell RNAseq    Fos-TRAP:Tdtomato mice were injected with Tamoxifen (150 mg/kg) 24hrs prior to exercise. Animals were exposed to a second bout of exercise either 1 or 4 days later. (B) When the two exercise periods were separated by 24 hrs, 12.5±5.6%, (n=3) of the exercise-TRAPed cells were re-activated, whereas with a 4 day separation only 1.3±0.3% 5 (n=4, unpaired t-test, p=0.07) overlapped, consistent with sparse labeling in the dentate gyrus.

This PDF file includes:
Figs. S1 to S8 Tables S1 to S4    Normalized data (blue dots) were used for differential expression comparison.   with 10^5 particles of one or both short hairpin shRNAs against Mtss1L (designated L1 and L2) or shMtss1Lscramble lentivirus, and then treated for 7 d.i.v. with media supplemented by BDNF (25ng/ml). The shScramble virus was ineffective, whereas the combination of the two short hairpins was the most effective. Mtss1L shRNA knockdown efficiency was analyzed by RT-qPCR for Mtss1L expression, using 18S as the housekeeping/control gene (n=3 biological replicates, one way ANOVA with Dunnett's multiple comparison, p<0.001).