Active presynaptic ribosomes in mammalian brain nerve terminals, and increased transmitter release after protein synthesis inhibition

Presynaptic neuronal activity requires the localization of thousands of proteins that are typically synthesized in the soma and transported to nerve terminals. Local translation for some dendritic proteins occurs, but local translation in mammalian presynaptic nerve terminals is difficult to demonstrate. Here, we present evidence for local presynaptic protein synthesis in the mammalian brain at a glutamatergic nerve terminal. We show an essential ribosomal component, 5.8s rRNA, in terminals. We also show active translation in nerve terminals, in situ, in brain slices demonstrating ongoing presynaptic protein synthesis. After inhibiting translation for ~1 hour, the presynaptic terminal exhibits increased spontaneous release, and increased evoked release with an increase in vesicle recycling during stimulation trains. Postsynaptic response, shape and amplitude were not affected. We conclude that ongoing protein synthesis limits vesicle release at the nerve terminal which reduces the need for presynaptic vesicle replenishment, thus conserving energy required for maintaining synaptic transmission.

Over the past decade, RNA based mechanisms have been discovered that respond to 37 extrinsic signals that affect postsynaptic local translation in dendrites to modify activity at 38 specific regions (Liu-Yesucevitz, et al., 2011;Yoon, et al., 2016). This is possible due to 39 the targeting of coding and non-coding RNA (Vo, et al., 2010) with RNA binding 40 proteins, and the presence of ribosomes that are located in, or moved to specific neuronal 41 regions or compartments (Ostroff, et al., 2002). This allows the neuron to have the 42 necessary components in place to translate specific dendritic proteins on-site, in response 43 to specific signals. The role of local translation in resting and sustained levels of synaptic 44 transmission is a major issue of interest.
Local protein synthesis is thought to provide a faster and more efficient 46 mechanism for neurons to maintain or alter activity levels and respond to rapidly 47 changing inputs. In mammalian central nervous system (CNS) neurons, local 48 postsynaptic protein synthesis in dendrites is well established. In contrast, until recently, 49 most evidence for local presynaptic protein synthesis in axons and nerve terminals came 50 from invertebrates and the mammalian peripheral nervous system (Alvarez, et al., 2000). 51 Evidence for presynaptic protein synthesis in the mammalian brain has been difficult to 52 demonstrate largely due to the difficulties of accessing and imaging CNS presynaptic 53 terminals (Akins, et al., 2009). Despite these issues, presynaptic ribosomes have recently 54 been shown to be present in GABA-ergic interneurons in mature mouse hippocampal 55 neurons, where presynaptic protein synthesis is necessary to induce a long-term 56 depression of synaptic responses (Younts, et al., 2016). Local protein synthesis has also 57 been shown to occur in the axons of developing mammalian brain neurons, and plays a 58 role in establishing nerve terminals (Batista, et al., 2017) and affecting release at recently 59 formed nerve terminals (Taylor, et al., 2013). Although it is still highly debated, recent 60 work provides good evidence that local protein synthesis can occur in nerve terminals in 61 mammalian brains and it can affect presynaptic activity. 62 To better understand the role of presynaptic protein synthesis in the brain, we  (Rodriguez-Contreras, et al., 2008;Korber, et al., 2015). In 70 addition, the basic mechanisms of pre-and postsynaptic responses have been extensively 71 characterized at this synapse (Neher, 2017). This synapse also undergoes significant 72 developmental changes in its morphology and physiological characteristics that involves

91
Evidence for presynaptic ribosomes at the calyx of Held nerve terminal 92 The calyx of Held is a large glutamatergic, monosynaptic nerve terminal located 93 in the medial nucleus of the trapezoid body (MNTB) in the mammalian auditory 94 brainstem ( Figure 1A). Cell bodies in the anterior ventral cochlear nucleus project axons 95 a significant distance to the MNTB, which is ~3 mm in a mouse brain ( Figure 1A). Up to 96 approximately postnatal day (P) 12, the calyx primarily has a spherical or spoon shaped Hoeve, 2012) that begin at ~P10, and allow this synapse to function at the high frequency 106 and fidelity (Taschenberger and von Gersdorff, 2000) that is required for sound 107 localization (Oertel, 1999;Carr, et al., 2001). Accordingly, there is a high level of protein 108 turnover that occurs slightly before and throughout this period. 109 We hypothesized that local protein synthesis could occur at this nerve terminal, 110 particularly due to the long axon length, its size (Borst, et al., 1995) and high frequency

132
G. Nuclease treatment prior to 5.8S rRNA antibody binding eliminates the ribosomal signal.

133
Enhanced contrast further shows the lack of ribosomal RNA after nuclease treatment.

181
To verify the specificity of the 5.8S rRNA signal, we treated the brain slices with 182 nucleases to degrade the 5.8S rRNA prior to antibody labelling and found this eliminated  between puromycin labelling and the presynaptic marker VGLUT1, we used line scans to 219 assess the degree of colocalization ( Figure 2D,E, graphs). We find excellent agreement in 220 the location and relative intensity of the two signals ( Figure 2D

247 248
ribosomes are found in the calyx of Held presynaptic nerve terminal, and they are active 249 under resting conditions. We note that longer puromycin application periods were tested, 250 and this produced a strong pre-and postsynaptic signal, but longer application times also 251 sharply increased the background signal, presumably due to activity from other neurons 252 and glia in the slice (data not shown). We conclude that all of the necessary components 253 (mRNA, tRNA, rRNA, ribosomal proteins and ribosomal binding proteins), which are 254 required to execute translation, must also be located in the nerve terminal.

255
To validate that the SUnSET assay detects active translation, we treated brain 256 slices with the translational inhibitor anisomycin for 1 hour prior to the addition of 257 puromycin ( Figure 2F). Ribosomes must be active for puromycin to be incorporated into 258 a polypeptide chain to be detected by the SUnSET assay. Consistent with this, we find  We find that the initial mEPSC frequency, measured shortly after onset of whole-      Figure 3D, p = 0.14), and a slightly 327 smaller average mEPSC area ( Figure 3E, p = 0.16) but none of these differences were statistically significant. In summary, the absence of an effect on the mEPSC amplitude, 329 and small non-significant effects on the shape of the mEPSC, demonstrate that the 330 postsynaptic response is not significantly affected (within ~1 hour) by inhibiting protein 331 synthesis. However, the differences in the mEPSC frequency demonstrate that inhibiting 332 protein synthesis has a presynaptic effect on the probability of spontaneous release.

334
Enhanced spontaneous release following tetanus eliminates mEPSC frequency 335 differences between control and protein synthesis inhibited neurons. 336 Prolonged stimulation produces a transient elevation in the frequency of 337 spontaneous release events (Habets and Borst, 2006). Given our finding that inhibiting 338 protein synthesis also increases the frequency of spontaneous release events, we 339 determined if the two effects act independently. Accordingly, we measured the frequency 340 of spontaneous release in control and protein synthesis inhibited neurons, before and 341 shortly after a tetanic stimulation at 100 Hz for 4 sec. It is important to note that prior to 342 the pre-tetanus mEPSC recording, the neurons had received several rounds of evoked 343 activity which is discussed in the next section. Although there was a ≥2 minute period 344 without evoked stimulation to allow recovery for the pre-tetanus recording (Fig 4A1), 345 there is still a small increase in the mEPSC frequency in both control (2.5 ± 0.45 Hz) and 346 protein synthesis inhibited neurons (3.6 ± 0.89 Hz) compared to the spontaneous 347 frequency measured before any evoked responses were given (Fig 3A). The difference in            Hz, 4 sec) and measured the mEPSC frequency starting <5 sec after tetanic stimulation.

396
Interestingly, following tetanic stimulation, the frequency of spontaneous release is 397 nearly identical for both control (6.1 ± 0.86 Hz) and protein synthesis inhibited (6.9 ± 1.2   The finding that spontaneous release of synaptic vesicles is affected by inhibiting protein 442 synthesis suggests that evoked responses could also be affected. In our initial tests, we 443 stimulated at a low frequency (0.1 Hz) to determine if the peak amplitude, shape, and        components that could be affected by inhibiting protein synthesis. Therefore, we tested 484 short stimulus trains at 200 Hz for 400 msec which were followed two or more minutes 485 later by a stimulus at 100 Hz. At both frequencies, we observed a tendency for lower         Figure 6A1 and A2). We find paired pulse depression at an interpulse pulse interval (IPI) 509 of 5 msec in control cells (0.72 ± 0.07 SEM, n = 9 cells from 7 animals) but a slight facilitation in translation inhibited cells at the same interval (1.09 ± 0.09 SEM, p = 0.004, 511 n = 8 cells from 6 animals). To determine if this ratio was affected following prolonged 512 activity, we also measured the 5 msec IPI paired pulse ratio > 2 minutes after a tetanic 513 stimulation (100 Hz for 4 sec). The paired pulse ratio was slightly decreased following 514 tetanic stimulation, which eliminated the small facilitation seen in the protein synthesis 515 inhibited cells ( Figure 6A2). Next, we compared EPSCs in control and protein synthesis 516 inhibited conditions throughout a 400 msec train at 200 Hz stimulation ( Figure 6B1).

517
Consistent with the paired pulse results, we observe a reduction in EPSC depression 518 throughout the train in the protein synthesis inhibited responses compared to controls.

519
Accordingly, we directly compared the average responses for the entire train and found 520 that the reduced depression seen after inhibiting protein synthesis occurred throughout 521 the 400 msec train at 200 Hz ( Figure 6B2). In contrast, for the blinded experiments in 522 which DMSO was applied, the treated and untreated responses were very similar 523 ( Supplementary Fig. 1A). Therefore, inhibiting protein synthesis results in higher levels 524 of neurotransmitter release that can be sustained (80 EPSCs) even at a brief 5 msec IPI.

525
Facilitation and depression are affected by the interval between pulses. 526 Accordingly, we also measured the paired pulse ratio at 10 msec IPI ( Figure 6C1 and C2).   (Schneggenburger, et al., 1999;Neher, 2015). Using this method, the average size of the 579 RRP ( Figure 7A2) was the same for control neurons (16.3 ±1.6 nA, n = 9 cells from 7 580 animals) and protein synthesis inhibited neurons (anisomycin, 15.97 ± 1.9 nA, n = 8 from 581 7 animals; p = 0.89). This indicates that the initial capacity for vesicular release of 582 neurotransmitter is not changed by inhibiting protein synthesis. Next, the initial 583 probability of release was measured by dividing the peak amplitude the first EPSC 584 response by the corresponding RRP measured for that train (Schneggenburger,et al.,585 1999). Here, we find the initial release probability (Pr) for control neurons (0.43 ± 0.03) 586 is higher than the value in anisomycin treated neurons (0.29 ± 0.02; p = 0.003), therefore 587 the differences we see in the initial peak response in protein synthesis inhibited neurons is 588 explained by a reduction in the initial probability of release ( Figure 7A3). Lastly, the 589 slope of the steady state of the cumulative response ( Figure 7A1) provides a measurement 590 of the rate of vesicle recycling. We find that the vesicle recycling rate ( Figure 7A4) 591 increases when protein synthesis is inhibited (96.9 ± 1.6 pA/msec), compared to the rate 592 in control neurons (62.2 ± 3.5 pA/msec; p = 0.037). This indicates that an increased rate 593 of vesicle replenishment is responsible for the increased responses following anisomycin 594 treatment ( Figure 6B2, D1, and D3).

595
To provide additional estimates of vesicle properties, we also measured responses 596 from stimulation at 200 Hz ( Figure 7B1). As anticipated, we found that the readily           (Kavalali, 2015). For example, spontaneous release of glutamate can 668 suppress local protein synthesis in dendrites (Sutton, et al., 2006). Furthermore, NMDA 669 receptors that are specifically activated by spontaneous release appear to be responsible 670 for the rapid antidepressant effect produced by ketamine exposure (Autry, et al., 2011). Therefore, over the last decade, accumulating evidence demonstrates that spontaneous 672 release can have significant effects on synaptic function. In addition, vesicles that 673 undergo spontaneous release may preferentially come from a population separate from 674 the vesicles that respond to evoked release (Sara, et al., 2005), and this appears to involve 675 association with specific proteins (Hua, et al., 2011). Finally, vesicles involved in 676 spontaneous release can have different sensitivity to intracellular calcium than vesicles 677 involved in evoked release, and spontaneous release may be able to occur independent of 678 intracellular calcium (Schneggenburger and Rosenmund, 2015). Therefore vesicles that 679 undergo spontaneous release appear to be controlled separately from vesicles that fuse in 680 response to action potentials. Spontaneous release is a highly regulated process that is 681 important in maintaining synaptic function, and we show that ongoing protein synthesis 682 plays a role in limiting the frequency of spontaneous release events.

683
Tetanic stimulation has previously been shown to transiently increase the 684 frequency of spontaneous release, due to residual calcium (Korogod, et al., 2005) and 685 may also involve PKC activation (Korogod, et al., 2007;Fioravante, et al., 2014). In our 686 work, we show that increased frequency of spontaneous release following tetanic In addition to the effects on spontaneous release, we also observe facilitation or 695 reduced depression in paired pulse ratios when protein synthesis is inhibited. 696 Interestingly, although we found that the readily releasable pools are identical in protein 697 synthesis inhibited and control conditions, the initial probability of release during a train 698 of stimulation is lower when protein synthesis is inhibited. However, during prolonged 699 stimulation, a reduced amount of depression is maintained throughout the stimulation.  However, work using mammalian synaptosomes has produced evidence of mRNA 732 transcripts and the ability to generate newly synthesized proteins (Alvarez, et al., 2000), 733 although potential contamination from postsynaptic neurons or glia has been a major 734 concern. Additional evidence comes from mRNA found in axons (Alvarez, et al., 2000) 735 and recently formed nerve terminals (Batista, et al., 2017). These presynaptic transcripts 736 code for a variety of proteins including some that can affect vesicle recycling and fusion 737 such as -catenin, -tubulin, and -actin. In addition, transcripts for nuclear encoded 738 mitochondrial proteins have been found in axons. It is important to note that in our experiments, the calyx nerve terminal is no longer connected to the neuronal cell body 740 because the axons are severed during brain slicing. Given the lack of connection between 741 the cell body and nerve terminal, newly synthesized proteins in the calyx nerve terminal 742 cannot come from the cell body that gives rise to the axon that forms the nerve terminal.

743
Based on our imaging data, we conclude that ongoing protein synthesis is occurring in 744 the presynaptic terminal, although some amount could also occur in the adjacent section 745 of the axon.

746
The work shown here is the first demonstration that local presynaptic translation 747 occurs in established nerve terminals in situ in mammalian brain slices. Our finding that 748 ongoing protein synthesis helps to limit vesicle release has interesting implications for     conditions were determined by 1-2 initial recordings in normal aCSF. If the initial recordings had 823 stable responses that lasted the duration of the stimulus protocols, a minimum of 25 minutes, then 824 the recording solution was switched to a blinded cylinder of recording solution for subsequent 825 recordings which were performed after ~ 1 hour of treatment (45 minutes to 120 minutes) in the 826 absence of fiber stimulation. Since spontaneous action potentials are not present in these 827 recording conditions, only spontaneous release activity occurred during the treatment period. On 828 each day, the blinded cylinder would contain either: 40μM anisomycin (Sigma Aldrich) or 829 DMSO alone (vehicle). To test the effect of the translational inhibitor anisomycin (40μM) on 830 synaptic response characteristics, slices were preincubated for ~1hour in the presence of the drug 831 in the absence of afferent fiber stimulation. At all times, aCSF was continuously circulated using 832 a peristaltic pump; total volume of the solution was 30mL. All recordings, control and test 833 conditions, were made in the presence of 25μm bicuculline and 2μm strychnine to block 834 inhibitory responses. Power analysis to determine the appropriate sample size was performed 835 based on means and standard deviation values of preliminary data. Recordings from 5 cells in 836 each condition was estimated to be adequately powered, for  = 0.5, and a 0.8 power of test. 837 Recordings from MNTB neurons and Data Analysis: Miniature excitatory post-synaptic currents 838 (mEPSCs) were recorded during 30 sec continuous recordings at several times during the 839 stimulation protocol. mEPSCs were analyzed by Mini Analysis Software (Synaptosoft). The 840 following mEPSC search parameters were used: gain, 20; blocks, 3,940; threshold, 10 pA; period 841 to search for a local maximum, 20,000 sec; time before a peak for baseline, 5,000 sec; period 842 to search a decay time, 5,000; fraction of peak to find a decay time, 0.5; period to average a 843 baseline, 2,000 sec; area threshold, 10; number of points to average for peak, 3; direction of 844 peak, negative). Analysis was performed using the above settings, and visually checked to ensure 845 accuracy. Evoked response traces were exported to Igor Pro (Wavemetrics, Portland, OR), and 846 measurements were made manually, or using Taro Tools (Igor macro, Taro Ishikawa) with visual 847 inspection and adjustment as necessary for every measured peak amplitude. 848 Data are presented as mean ± standard error of the mean (SEM). Unless otherwise noted, a 849 Student's t-test was employed to determine statistical significance. Calculated p-values are 850 indicated in relevant figures as follows: p ≤ 0.05 is considered significant (*); p ≤ 0.01 very 851 significant (**); and p ≤ 0.001 highly significant (***). We define biological replicates as each 852 tested cell (number of recordings), and technical replicates as multiple tests on a single cell. In 853 our experiments, a minimum of four recordings, of spontaneous activity, 30 sec each, were made 854 during the recording time. Data were analyzed as initial spontaneous release levels, and 855 spontaneous release following activity as described in the text. Outlier data for spontaneous event 856 recordings resulted in removal of two cells from the data, as determined by Grubb's test with = 857 1%. The two-sample Kolmogorov-Smirnov (KS) test was used to calculate the p-value for the 858 cumulative probabilities of the mEPSC event intervals for two different conditions. Briefly, this 859 non-parametric test uses the maximum vertical difference between two cumulative probability 860 graphs and the total number of measurements to determine the statistical significance of the 861 differences between two cumulative probability distributions. Histograms with identical bin-862 ranges were used to compare the mEPSC intervals for the two different conditions. This 863 calculation was preformed manually, and by the KS function in MATLAB, which gave very 864