a molecular mechanism for synapse elimination: novel ... · thrombin inhibition delays synapse...

DEVELOPMENTAL BIOLOGY 179, 447–457 (1996)ARTICLE NO. 0274

A Molecular Mechanism for Synapse Elimination:Novel Inhibition of Locally Generated ThrombinDelays Synapse Loss in Neonatal Mouse Muscle

M. N. Zoubine, J. Y. Ma, I. V. Smirnova,B. A. Citron, and B. W. Festoff1

Neurobiology Research Laboratory, Department of Veterans Affairs Medical Center,Kansas City, Missouri 64128; and Department of Neurology, University of KansasMedical Center, Kansas City, Kansas 66103

Activity-dependent, polyneuronal synapse elimination (ADPSE) is a programmed, regressive event in the development ofthe nervous system and readily studied at the neuromuscular junction, where it is complete 15–20 days after birth. Localexcess, or imbalanced, protease activity is one of several possible underlying mechanisms. In this regard, thrombin mediatesactivity-dependent synapse loss in an in vitro model of ADPSE. To test the involvement of thrombin in vivo, we locallyapplied the leech thrombin-specific inhibitor, hirudin. We monitored neuromuscular behavior, correlated with acetylcholin-esterase and silver nitrate histochemistry at endplates, for changes in the timecourse of in vivo synapse elimination andassayed both thrombin activity and prothrombin expression in developing muscle. Hirudin retarded elimination, withoutaltering motor performance, uniquely at Postnatal Day 5 (P5) and maximally at P9. Reverse transcription–polymerasechain reaction (PCR) showed that neonatal muscle was a source of local prothrombin, with peak expression during thefirst week after birth. A specific chromogenic assay revealed that local thrombin, activated from muscle-derived prothrom-bin, peaked during maximal synapse remodeling. q 1996 Academic Press, Inc.

INTRODUCTION anced, excess local protease activity have been hypothe-sized (Changeux and Danchin, 1976; Connold et al., 1986;Festoff, 1987; O’Brien et al., 1978, 1984; Vrbova and Fisher,The change from perinatal, multineuronal, or polyneuro-1989). Local protease excess has been particularly attractive,nal innervation of skeletal muscle to the adult mononeuro-since this proposed mechanism, implicating degradation ofnal state carries the implication of removal or eliminationone or more critical extracellular matrix (ECM) adhesionof initial synaptic connections. Synapse elimination occursmolecules (Connold et al., 1986; O’Brien et al., 1978), isduring this developmental time period, not only in muscle,consistent with the phenomena of simple retraction with-but in all parts of the nervous system and in all species.out degeneration, previously found by several authorsMost studies point to synaptic competition among ‘‘domi-(Riley, 1977; Vrbova and Fisher, 1989). The proteases sug-nant’’ and ‘‘subordinate’’ synaptic terminals; however, thegested have included a thiol-based, calcium-activated neu-specific molecular mechanism of synapse elimination is nottral protease (Connold et al., 1986; Vrbova and Fisher, 1989)known (Colman and Lichtman, 1993; Lichtman and Balice-and plasminogen activators (PAs), particularly urokinase-Gordon, 1990; Purves and Lichtman, 1980; Van Essen, 1982;like PA (Hantaı et al., 1989).Van Essen et al., 1990). Of the several potential mechanistic

Using an in vitro model, which reproduced the activityprocesses, competition between terminals, based on dwin-dependence and leupeptin sensitivity seen in vivo (Connolddling supplies of essential neurotrophic factors, and imbal-et al., 1986), we unexpectedly found that thrombin (Fenton,1986) serine proteolytic activity was required for activity-dependent synapse elimination (Liu et al., 1994a,b). Demon-1 To whom correspondence should be addressed at Neurobiologystrating specificity for thrombin in this same system, syn-Research Lab (151R), VA Medical Center, 4801 Linwood Boulevard,apse reduction was not blocked by inhibitors of plasmin,Kansas City, MO 64128. Fax: (816) 922-3375. E-mail: serpin@

kuhub.cc.ukans.edu. trypsin, or cysteine proteases (1 mM aprotinin, cystatin) or

447

0012-1606/96 $18.00Copyright q 1996 by Academic Press, Inc.All rights of reproduction in any form reserved.

AID DB 8381 / 6x14$$$161 09-28-96 00:47:55 dba AP: Dev Bio

448 Zoubine et al.

tized with an ip injection of 20 ml/g of rodent cocktail (ketaminetwo calcium-activating neutral protease (calpain I and II)5 mg/ml; 5% v/v acepromazine in PBS), a small dorsal skin incisioninhibitors. Only hirudin (Fenton, 1989) and protease nexinwas made, and a mini-osmotic pump (Alzet, 1007 D; Alza Corp.,I (PNI) (Scott et al., 1985), potent inhibitors of thrombin, atPalo Alto, CA), filled with 100 ml hirudin (100 mM in PBS) or vehicle,nanomolar concentrations, blocked synapse loss.was inserted. In separate preliminary experiments, we confirmed,using thrombin inhibition assays with the specific S-2238 sub-strate, that hirudin maintains its activity over 5 days in a pump at

MATERIALS AND METHODS 377C. The pump was attached to a fine-bore Teflon catheter whichreleased 0.52 { 0.02 ml/hr with the catheter free end directed tothe surface of the left gastrocnemius muscle. The skin incision wasAnimals and animal husbandry. BALB/c female mice 8–9sealed with cyanoacrylate adhesive (Zoubine et al., in preparation)weeks old, weighing 18–20 g, were mated to obtain normal off-and the pup was returned to its mother. Intramuscular injectionsspring. Pregnant females were housed in our Animal Research Fa-were replaced by this pump delivery on P7.cility, kept to a 12-hr light/dark cycle, and allowed rodent chow

Neuromuscular behavior. Spontaneous placing and rightingad libitum. All procedures were approved by the Animal Studiesactivities were carefully observed and recorded before, during, andCommittee and adhered to NIH guidelines.after intramuscular injection and implanting of osmotic pumps andMuscle dissection, endplate counting, and staining. Gastrocne-catheters. Activity was identical for hirudin-treated mice (compar-mius muscles were removed from neonates at Postnatal Days P1,ing right and left sides) and for PBS-injected and nonoperated lit-P5, P9, P13, P15, P18, and P22, mounted, frozen in OCT compoundtermates. There was no difference in the behavior of the mothersin liquid nitrogen, and stored at 0707C until used. Longitudinaltoward treated pups compared with untreated, and all pups weresections (50–60 mm) were cut in a Hacker–Bright cryostat at0187C.handled by examiners in the same way.A drop of EDTA solution was placed on the sections to prevent

Muscle tissue extraction. We performed all assays with frozenmuscle contraction, and slides were air-dried and stored at 0707C.muscle stored at 0707C. Consequently, sufficient samples for theFor staining with anti-synaptophysin antibody (Boehringerentire temporal course (i.e., E14 through P30) were collected priorMannheim, Indianapolis, IN) and tetramethylrhodamine-a-bunga-to assay. For the earlier developmental time points, three to fiverotoxin (TMR-a-BTX, Molecular Probes, Eugene, OR), slides were

treated as previously described (Ma et al., 1994) and then preincu- embryonic samples were pooled. Prior to assay all tissues werebated with 10% donkey serum in phosphate-buffered saline (PBS) thawed on ice, homogenized with a glass–glass conical homoge-for 1 hr at room temperature. Following this, anti-synaptophysin nizer in ice-cold aqueous homogenization buffer (250 mM sucrose,antibody was added at a 1:100 dilution and the slides were placed 1.0 mM EDTA, 0.9% NaCl, 50 mM Tris–HCl, pH 7.4), and centri-at 47C for 18 hr in a humidity chamber. The slides were then fuged at 17,500g in a Beckman G15R refrigerated centrifugewashed in PBS followed by incubation with donkey anti-rabbit IgG equipped with an F2404 rotor for 15 min. The resulting superna-(H / L) conjugated with dichlorotriazinylamino fluorescein (Jack- tants (S1) were saved for assay. Protein concentration was deter-son ImmunoResearch, West Grove, PA) at a 1:300 dilution. After mined using bicinchoninic acid protein assay reagent (Pierce Bio-an additional wash with PBS, slides were incubated with 10 nM chemical, Rockford, IL) according to the manufacturer’s recom-TMR-a-BTX for 5 min. Hoechst AG Mowiol 4-88 medium (Calbio- mendations.chem, La Jolla, CA) was added, coverslips were placed, and slides Thrombin chromogenic (S-2238 amidolytic) assay. We mea-were viewed using epifluorescence with a Leitz Orthoplan micro- sured thrombin activity in extracts from rinsed (embryonic andscope and Ploem filters. early postnatal) or perfused (later postnatal and adult) muscle with

To observe axonal inputs to endplates, sections were fixed in 4% an amidolytic assay by quantifying the thrombin-mediated cleav-paraformaldehyde in phosphate buffer (pH 7.2) for 2 hr at room age of a synthetic tripeptide substrate, D-Phe-Pip-Arg-p-nitroanilidetemperature, washed in double-distilled water, and stained with (pNA) (S-2238) (Kabi Pharmacia Hepar. Inc, Franklin, OH). Eitheracetylthiocholine iodide combined with impregnation of nerves samples (10 ml) or dilutions of standard human a-thrombin (gener-with 1% silver nitrate (Hantaı et al., 1989). The counting of poly- ous gift from Dr. J. W. Fenton, II, Wadsworth Center for Labora-and mononeuronally innervated endplates (EPs) from indepen- tories and Research, Albany, NY), ranging from 1 to 20 nM, weredently coded slides was performed ‘‘blind’’ by three individuals. mixed with 25 ml of S-2238 (1 mg/ml) and adjusted to 200 ml withThree fields in each of six sections per muscle were counted by assay buffer (0.05 M Tris–HCl, pH 8.3, containing 0.15 M NaCl).each individual. The data were decoded, organized in a table, and Samples were incubated at 377C in 96-well, flat-bottomed microti-analyzed by the sign test for statistical significance (Siegel, 1956). ter plates (Imulon 2, Dynatech Laboratories, Inc., Chantilly, VA)The following formula was used: and monitored for the release of pNA at 410 nm with the MR 5000

microplate reader (Dynatech). Controls (including buffer, substrate,and extracts alone) were assayed in parallel wells, with duplicatepercentage polyneuronal innervation (%)wells for each sample, and subtracted from experimental values.The concentration of thrombin in samples was calculated withÅ No. of EPs with 2 or more axon terminals

Total No. of EPs1 100.

reference to a standard curve obtained with purified human a-thrombin that was assayed simultaneously. The amidolytic activ-ity was expressed as units per milligram protein, where one unitHirudin administration. On P5 (day of birth, P0) one intramus-is defined as one nanomole of cleaved substrate per hour. An absorp-cular injection of 15 ml hirudin (Sigma, St. Louis, MO), at 1 mM intion coefficient of 9920 M01 cm01 for pNA was used (Witting etPBS, or vehicle using an ultrathin, 32-gauge steel needle (Hamiltonal., 1987). Values listed represent means { SDs of three separateCo., Reno, NV), was made in posterior hindlimb muscles and re-assays.peated again in 24 hr. Injection was made in the long axis of the

Prothrombin mRNA determination. RNA was extracted in themuscle and only on withdrawal (English and Schwartz, 1995), tominimize injury and leakage of solution. On P7, mice were anesthe- presence of denaturants and verified both by methylene blue stain-

Copyright q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.


449Thrombin Inhibition Delays Synapse Elimination

ing and spectrophotometrically. The PCR primers, 5*GAGAAG- 82 { 12 (SEM)% of the total EPs at P5, 44 { 11 (SEM)% atGGTATCGCCCCCTG and 5*TGTCTGCTTGTCTGGCAAAC, P9, and 11 { 5% at P15, a naturally occurring loss of 87%synthesized by DNA International (Lake Oswego, OR), were chosen of polyneuronal innervation during these 10 days. We calcu-to amplify the mRNA segment 1120–1473 of mouse prothrombin lated the rate of synapse elimination between P5 and P15,(GenBank Accession No. X52308) (Degen et al., 1990). RT-PCR, which showed that 7.1% of excess synapses were lost peroptimized for subplateau signals using GeneAmp EZ rTth polymer-

day (Fig. 3). Three rates can be discerned in natural ADPSEase (Perkin–Elmer, Norwalk, CT) per the manufacturer’s recom-in mouse gastrocnemius: a shallow slope between P1 andmendations, was performed with 0.45 mM 3* and 5* oligonucleo-P5, a much steeper slope between P5 and P15, and a third,tides and 0.5–1 mg total RNA. Initial cDNA synthesis at 607C formuch slower, or no rate of loss, after P15.30 min was followed by incubation at 947C for 2 min and then

40 cycles of amplification (947C denaturation and 607C annealingextension; 1 min each). Products were visualized with ethidium

Effects of Local Hirudin on Synapse Lossbromide after electrophoresis on 3% metaphor agarose (FMC, Rock-land, ME). Having examined naturally occurring ADPSE, we next

evaluated the effects of local application of hirudin (Fentonet al., 1991) on this elimination. We injected the leechthrombin inhibitor intramuscularly for 2 days starting atRESULTSP5, since our data indicated the greatest remodeling afterthis date. Then, beginning on P7, we replaced injectionsTimecourse of Polyneuronal Synapse Losswith rate-controlled, local delivery of hirudin by a subcuta-neously implanted osmotic mini-pump and a fine catheterThe neuromuscular junction (NMJ) is the best understood

vertebrate synapse (Hall and Sanes, 1993) and a useful model placed over the gastrocnemius muscle. The hirudin-treatedanimals appeared no different than untreated littermates.for investigation of synaptic plasticity (Balice-Gordon et al.,

1990, 1993; Balice-Gordon and Lichtman, 1990). Over the Importantly, no impairment of motor function was detectedon the hirudin-treated side compared to the contralateralfirst 3 weeks of life, we observed a maturation of the NMJs

evidenced at low power by an apparent overall increase in leg. Consequently, the results of counting mono- and poly-neuronal-innervated neuromuscular synapses in BALB/cthe presynaptic marker, synaptophysin (Figs. 1A, 1C, and

1E), relative to the acetylcholine receptors (AChRs) stained mice at P9, following 4 days of hirudin treatment, werejudged to be unrelated to the degree or extent of nerve-with TMR-a-BTX (Figs. 1B, 1D, and 1F). Although the num-

ber of synapses at each endplate is decreasing during this evoked muscle activity.We removed leg muscles at P9 and P15, flash-froze them,period, the local levels of at least one presynaptic protein are

increasing, perhaps reflecting the structural and functional and used the silver-AChE technique to obtain the percentageof polyneuronal-innervated EPs after hirudin treatment. Nostate of the developed endplate. Although indistinguishable

at this level, this change would be consistent with the re- evidence of inflammatory cell reaction (lymphocytes or mac-rophages) was found and needle tracks were not seen. Moreplacement of less active presynaptic termini, as well as

other alterations that could affect the staining. In any case, than 4000 EPs were counted in these experiments and theresults indicated a marked effect on the elimination time-the colocalization of the pre- and postsynaptic markers in

the sections under investigation is shown. course by the thrombin inhibitor. We found that muscles satu-rated locally with hirudin had a greater number of polyneuro-For this study of molecular mechanisms of synapse elimi-

nation, we first established the natural timecourse of activ- nal-innervated EPs than did the contralateral control muscles.In a representative series, a total of 1845 EPs were counted atity-dependent, polyneuronal synapse elimination (ADPSE)

in flash-frozen mouse gastrocnemius muscle removed from P9 (n Å 5 mice); 816 hirudin-treated EPs and 1029 EPs werecounted in control, contralateral muscles. The average per-P1 to P22 mice, using the silver-AChE staining technique

(Hantaı et al., 1989; Ma et al., 1994; Tian et al., 1995). This centage of polyneuronal-innervated EPs on the hirudin-treatedside was 80 { 7%. On the contralateral control side, this wastechnique has the advantage of enabling large numbers of

EPs to be examined individually. However, in contrast to 51 { 9% (Fig. 4A). The latter number correlated with the 44{ 11% that we found during the natural postnatal eliminationelectrical measurements, it may include a few axonal

branchings derived from the same primary axon. Although timecourse (Fig. 3). We found that hirudin-treated EPs had56% greater polyneuronal innervation compared to controlin embryos each muscle fiber is innervated by three to four

different axons, after birth, the majority of EPs are doubly EPs, which was statistically significant (P Å 0.05, sign test(Siegel, 1956)). In other experiments (Fig. 4A), we confirmedinnervated (Dennis, 1981). These are then removed to arrive

at a single, mononeuronal innervation. With this technique, that this was dependent on hirudin and not the injection sincePBS, both injected into muscle and delivered by pump, re-poly- and mononeuronal-innervated EPs were clearly distin-

guished during the postnatal period, as shown in the repre- sulted in 40 { 12% polyneuronal innervation. As we foundwith hirudin, saline injection into muscle and infusion bysentative photomicrographs (Fig. 2). We quantified these

results using the formula presented above, and these analy- pump caused no signs of diminished neuromuscular activity.These results further confirmed that pump and catheter place-ses are graphically presented in Figs. 3 and 4. Polyneuronal

synapses, defined as EPs with two or more inputs, represent ment had no effect on nerve-evoked muscle activity.



450 Zoubine et al.

FIG. 1. Endplates in mouse gastrocnemius muscle at Postnatal Days 1 (P1; A, B), P9 (C, D), and P21 (E, F). The synapses are stainedwith antibody to the presynaptic marker anti-synaptophysin (A, C, E) or with TMR-a-bungarotoxin for postsynaptic AChRs (B, D, F).During development, there is progressive intensity, but an apparent reduced frequency, of synaptophysin.

09-28-96 00:47:55 dba AP: Dev Bio


FIG. 2. Naturally occurring polyneuronal synapse elimination in mouse gastrocnemius muscle. (A) Both polyaxonal innervation (arrows)and monoaxonal innervation (.) at EPs are clearly visible at P9. (B) Primarily monoaxonal innervation (.) is found in gastrocnemiusmuscles at P15. Scale bar, 10 mm.

Since the majority of polyneuronally innervated EPs have tion of EPs on P15, after 10 days of hirudin treatment, none-theless showed that hirudin still delayed synapse elimination.already been eliminated by P15, we expected the effect of

hirudin to be less obvious at this stage than at P9. Our control A total of 2102 EPs were counted at P15: 1264 on the hirudin-treated side and 838 in contralateral, control muscles. Thedata, in fact, show that only 11 { 5% of EPs remain multiply

innervated at P15 (see Fig. 3). Results obtained from examina- average percentage of polyneuronal-innervated EPs (nÅ 6 ani-



452 Zoubine et al.

grammed timecourse of polyneuronal synapse eliminationwith the profile observed in the presence of hirudin. The delayin the elimination attributable to the effects of hirudin wasclearly more evident at the earlier time point selected (P9).However, statistically significant differences between treatedand control EPs were still detected at P15. With hirudin, overthe first half of this timecourse (P5–P9), synapse loss was only0.9% per day, compared with 12% lost per day naturally.Comparing the half-maximal decrease in polyneuronally in-nervated EPs in controls to the hirudin-treated muscles indi-cates that the elimination process was delayed approximately4 days (from 8.5 to 12.5) by hirudin.

Prothrombin Gene Expression

The leech protein, hirudin, is a specific and potent throm-FIG. 3. The timecourse of synapse loss in active postnatal mouse bin inhibitor, implicating thrombin activity in any parame-muscle. P9 and P15 synapse counts from muscles infused with the ter affected by hirudin (Fenton et al., 1991). Since hirudinthrombin inhibitor, hirudin (solid line), were compared to control

had a dramatic effect at the time of maximal synapse remod-polyneuronal innervation data (dashed line) collected from un-eling in muscle, and this effect was mediated by inhibitingtreated animals, contralateral muscles, and muscles treated withits cognate protease, thrombin, an important questionPBS.raised by our results is the source of thrombin activity thatis potently inhibited by local hirudin to delay ADPSE inneonatal muscle. Thrombin, as prothrombin, would be ex-pected to be in the muscle vascular bed and might be acti-mals) was 22 { 3% in hirudin-treated muscles while in the

untreated, contralateral controls it was 14 { 3% (Fig. 4B). vated locally by prothrombinase (Benzakour et al., 1995;Fenton, 1988). However, prothrombin expression has pre-The hirudin-treated side at P15 had a 57% greater number of

polyneuronal-innervated NMJs compared to control (PÅ 0.05, viously been detected in rat brain, where it is maximallyexpressed before birth (Dihanich et al., 1991), indicatingsign test). Figure 3 compares our data for the natural pro-

FIG. 4. Effects of hirudin on polyneuronal-innervated EPs in mouse gastrocnemius muscle. Mice were treated with hirudin, beginningat P5. As before (Fig. 1), muscles were removed, frozen, sectioned, and stained with silver-AChE. The formula under Materials and Methodswas used to determine the effects of hirudin on percentage of polyneuronal innervation. EPs of treated muscle, under constant hirudinexposure, were compared with contralateral muscle and with both untreated littermate and PBS-treated controls at (A) P9 and (B) P15.The scale in B is 0–30% polyneuronal endplates. The P values were determined using the sign test (Siegel, 1956).




FIG. 5. Prothrombin RT-PCR of mouse gastrocnemius muscle. Total RNA was extracted from developing muscle at different develop-mental times from Embryonic Day 12 (E12) to P30, reverse transcribed, and amplified by PCR to subplateau levels. N, negative control(no template); and L, positive control (liver RNA).

that tissues other than liver (Fenton, 1986) might also ex- hibitor dramatically delays programmed synapse elimina-tion, but it also offers molecular evidence to specificallypress prothrombin. That muscle may be a source of this

thrombin precursor, and may be implicated in synapse re- support the ‘‘protease hypothesis’’ of synapse eliminationmodeling, is further supported by alterations in measurable (Festoff, 1980; O’Brien et al., 1978). This delay (Fig. 3), with-thrombin activity released from cultured myotubes, af- out concomitant effects on muscle activity, correlates withfected by agents that influence activity (Nelson et al., 1994). both prothrombin gene expression and thrombin activity inFurthermore, since hirudin delayed but did not totally block muscle at a time in postnatal development during extensiveelimination, this suggested that thrombin might be locally synapse remodeling. These effects in the mouse gastrocne-generated in neonatal muscle and may be developmentally mius, as revealed by changes in the morphology of suchregulated. This implies that the muscle fiber itself might remodeling, were greatest between P5 and P9 (Fig. 4A), thebe the source of released prothrombin which is activated to time for maximum synapse elimination in mouse muscle,thrombin in the vicinity of the synaptic cleft. and were relatively complete by P15 (Fig. 4B) (Colman and

With this in mind, and as a first step to assessing changes Lichtman, 1993). Less remodeling is taking place in musclein local thrombin protein levels, we determined whether at this later time, and less prothrombin expression andprothrombin gene expression was detectable in postnatal thrombin activity (Figs. 5, 6), which correlate with less ef-skeletal muscle. We used RT-PCR and found that muscle fect of the thrombin inhibitor.did, indeed, express prothrombin, with peak prothrombin Previous models relating to synapse elimination in mus-mRNA during the final phase of ADPSE (Fig. 5). Expression cle have not focused on specific molecular mechanisms.decreased considerably after P7, but prothrombin mRNA However, the work of Lichtman and colleagues (Balice-Gor-was still detected until P20. don et al., 1990, 1993; Balice-Gordon and Lichtman, 1990,

1994; Colman and Lichtman, 1993; Lichtman and Balice-Gordon, 1990) have emphasized the restricted, local natureThrombin Amidolytic Activityof the elimination process. The contributions of Vrbova and

Since it is thrombin’s proteolytic activity, and not just her colleagues have called attention to the possibility thatprothrombin expression, that is implicated in synapse re- proteases, acting on ECM adhesive macromolecules, maymoval by these experiments, we also assayed amidolytic be important (Connold et al., 1986; O’Brien et al., 1984;activity that is specific for thrombin. We found that such Vrbova and Fisher, 1989). Our own studies have implicatedactivity is detected at P15, but it is much greater at earlier a developmental imbalance of serine proteases and theirpostnatal times. Although thrombin activity was detectable serpin inhibitors in driving the wave of ADPSE in muscleas early as E14, it was biphasic, elevated at E18 and greater (Festoff, 1980; Hantaı et al., 1989). Our recent studies pointat P10, where the maximal peak was found (Fig. 6A). This to an imbalance between the coagulation protease, throm-activity was inhibited by thrombin inhibitors, such as D- bin (Liu et al., 1994a,b), and its principal serpin in tissues,Phe-Pro-Arg-chloromethyl ketone (PPACK) and PNI, in ad- PNI (Akaaboune et al., 1995; Festoff et al., 1991; Verdiere-dition to hirudin (not shown). In addition, prothrombin lev- Sahuque et al., 1996).els, as measured by ecarin activation of muscle extracts The reduced effect of hirudin at P15 (Fig. 4B) indicates(Rosing and Tans, 1988), also revealed developmental regu- that hirudin delays but does not prevent ADPSE. This maylation (Fig. 6B), reflecting local synthesis of the precursor. correlate with prothrombin expression downregulation

after the wave of ADPSE is complete. It also suggests thatthe inhibitor has less effect, since more polyneuronal EPs

DISCUSSION have been eliminated and less protease is locally producedto inhibit, perhaps, in conjunction with reduced hirudinactivity released from the pump over time. Although we didThese results are novel and informative, since not only

is this the first demonstration that a specific thrombin in- not directly estimate residual activity of hirudin in pumps



454 Zoubine et al.

FIG. 6. Thrombin (A) and ecarin-activated prothrombin (B) amidolytic activity during neuromuscular development determined by cleavageof the specific substrate S-2238.

removed from animals, we did do separate experiments to to, and cleavage of, the PART receptor may also initiateother signal transduction/molecular mechanisms underly-determine the stability of hirudin in pumps maintained at

377C in a humid environment. These studies showed no ing the process of selection of active synapses. Such changesmight affect clusters of AChRs, leading to their disaggrega-decrease in hirudin’s potency to inhibit thrombin in the S-

2238 assay (not shown). Thus, in this regard, the neonatal tion and further promoting terminal retraction (Balice-Gor-don and Lichtman, 1994). The current results, demonstra-muscle fiber appears to be a significant source of zymogen

prothrombin, but the ‘‘window of opportunity’’ for a throm- ting the importance of thrombin activity in elimination,are consistent with the localization of PART at the synapse,bin inhibitor appears restricted, based on the availability of

the protease, its precursor and/or activator (prothrombinase) where it may be activated by thrombin-directed proteolysisafter thrombin binding, via its anion exosite I, to the hiru-at the synaptic site.

Most, if not all, of thrombin’s effects in the nervous sys- din-like domain of the receptor. However, this is currentlynot known, but may be easily tested in model systems sincetem are mediated via the seven-transmembrane-domain,

proteolytically activated receptor for thrombin (PART) the many components of the thrombin signaling system arenow available for study.(Hung et al., 1992; Vu et al., 1991), since thrombin receptor

activating peptides produce the same effect as thrombin on For inhibition of released thrombin before its interactionwith the PART, PNI, the most potent mammalian throm-neurite and astrocyte process retraction (Beecher et al.,

1994; Jalink and Moolenaar, 1992; Suidan et al., 1992). Al- bin inhibitor found in tissues (Farrell et al., 1988; Stoneet al., 1987), is likely the physiologic regulator of excessthough hirudin would not be expected to be present at the

mouse NMJ, mouse PART has a hirudin-like domain, with thrombin action in ADPSE. From our prior studies, PNIclosely colocalizes with AChRs at the adult mouse NMJ,the sequence KYEPF-WEDEE (Ishii et al., 1995), which is

expressed in skeletal muscle (Niclou et al., 1994). This do- appearing later than the peak of prothrombin expressionin neonatal muscle and localizing to the synapse after P5main would be free to participate in thrombin binding at

its anion-binding exosite I, after prothrombin secretion and (Akaaboune et al., 1995; Festoff et al., 1991). In this regard,PNI blocked ADPSE in the same nanomolar concentrationactivation. Signal transduction through this receptor is G-

protein-coupled and functions to mobilize [Ca2/]i within the range as hirudin in our in vitro model (Liu et al., 1994a).PNI, a serpin identical to the glia-derived nexin, is regu-respective cells (Pouyssegur et al., 1993). [Ca2/]i mobiliza-

tion might then be able to depolymerize microtubules and lated by factors related to cell activity and injury (Monard,1993; Monard et al., 1990). It is released from glial cells,neurofilaments in distal axons, possibly with the participa-

tion of the small 21-kDa Ras G-protein, Rho (Jalink et al., activated by injury-related cytokines and by vasoactive in-testinal polypeptide (Bleuel et al., 1995; Cunningham et1994). It may also raise intracellular, calcium-activated neu-

tral proteases, such as calpains, to further promote retrac- al., 1993; Festoff et al., 1996), and binds to the cell surfaceand/or the ECM, where it forms SDS-resistant complexestion of subordinate terminals from EPs. This may explain,

in part, the observation by Vrbova and colleagues that milli- with thrombin, which are then internalized and degradedwithin the cell (Baker et al., 1980; Verdiere-Sahuque et al.,molar concentrations of calpain inhibitors also caused retar-

dation of ADPSE (Connold et al., 1986). Binding of thrombin 1996). Binding of PNI to the ECM of the NMJ may be to




rence Brass (University of Pennsylvania), Ann Lohof (Universite de‘‘receptors,’’ such as collagen IV, vitronectin (Donovan etParis VII), Peter Smith (University of Kansas Medical Center), andal., 1994; Rovelli et al., 1990), or others, to inactivate ex-Albert Herrera (University of Southern California) for critical read-cess thrombin. This would function to produce a spatiallying of the manuscript. Part of this work was performed during anand temporally restricted neutralization of thrombin pro-Academie des Sciences Francaise Professorship (B.W.F.), funded byteolytic activity and a sparing or maintenance of the activeElf Aquitaine, Inc. This work was supported, in part, by the Alzhei-

(‘‘dominant’’) presynaptic terminals. mer’s Disease and Related Disorders Association, the Marion Mer-The localization and activity of both thrombin and PNI rell Dow Scientific Education Partnership, and the Medical Re-

support an important role for the balance of these molecules search Service of the Department of Veterans Affairs.in synapse stabilization and homeostasis after synapse re-duction. The release of thrombin could result from activa-tion of the muscle fiber by ACh, as suggested previously REFERENCES(Nelson et al., 1994). Our model predicts that thrombinproteolytic activity would act on the presynaptic terminal, Akaaboune, M., Hantaı, D., Smirnova, I. V., and Festoff, B. W.on ECM macromolecules that adhere pre- to postsynaptic (1995). Expression of protease nexin I in mouse skeletal muscleelements, and/or on postsynaptic receptors (i.e., muscle during development. Soc. Neurosci. Abstr. 21, 800.PARTs). These novel concepts for synapse maturation have Appel, S. H., and Festoff, B. W. (1971). Biochemical dysfunction

and dementia. Contemp. Neurol. Ser. 9, 133–149.previously been advanced for other proteases in the nervousBaker, J. B., Low, D. A., Simmer, R. L., and Cunningham, D. D.system (Connold et al., 1986; Siman et al., 1987). The hiru-

(1980). Protease-nexin: A cellular component that links thrombindin effect suggests that the process is local and not systemicand plasminogen activator and mediates their binding to cells.and depends on the timing of ADPSE. Our current experi-Cell 21, 37–45.ments are consistent with previous experiments demonstra-

Balice-Gordon, R. J., Breedlove, S. M., Bernstein, S., and Lichtman,ting localized regulation of synapse elimination (Balice-Gor-J. W. (1990). Neuromuscular junctions shrink and expand as mus-

don and Lichtman, 1994; Lichtman and Balice-Gordon, cle fiber size is manipulated: In vivo observations in the andro-1990), since only local application of the leech thrombin gen-sensitive bulbocavernosus muscle of mice. J. Neurosci. 10,inhibitor on the treated side delayed ADPSE. 2660–2671.

In summary, hirudin delays naturally occurring ADPSE in Balice-Gordon, R. J., Chua, C. K., Nelson, C. C., and Lichtman,J. W. (1993). Gradual loss of synaptic cartels precedes axon with-developing postnatal mouse muscle in vivo. Since hirudin isdrawal at developing neuromuscular junctions. Neuron 11, 801–a specific inhibitor for thrombin, these results support the815.in vitro evidence for thrombin’s participation in synapse

Balice-Gordon, R. J., and Lichtman, J. W. (1990). In vivo visualiza-elimination (Liu et al., 1994a). The thrombin that acts ontion of the growth of pre- and postsynaptic elements of neuro-susceptible components of the synapse, such as PART ormuscular junctions in the mouse. J. Neurosci. 10, 894–908.other thrombin substrates, is generated locally in devel-

Balice-Gordon, R. J., and Lichtman, J. W. (1994). Long-term synapseoping muscle, where it is expressed and presumably acti- loss induced by focal blockade of postsynaptic receptors. Naturevated for developmental function. PNI is progressively lo- 372, 519–524.calized at adult synapses (Akaaboune et al., 1995; Festoff Beecher, K. L., Andersen, T. T., Fenton, J. W., and Festoff, B. W.et al., 1991), could inhibit excess thrombin that might be (1994). Thrombin receptor peptides induce shape change in neo-

natal murine astrocytes in culture. J. Neurosci. Res. 37, 108–115.generated after the wave of ADPSE, and may actually partic-Benzakour, O., Kanthou, C., Lupu, F., Dennehy, U., Goodwin, C.,ipate in stopping further elimination by inhibiting throm-

Scully, M. F., Kakkar, V. V., and Cooper, D. N. (1995). Prothrom-bin. Although not specifically addressed by these experi-bin cleavage by human vascular smooth muscle cells: A potentialments, it is tempting to speculate that these results mayalternative pathway to the coagulation cascade. J. Cell. Biochem.also be applicable to underlying molecular mechanisms59, 514–528.contributing to synapse loss, leading to neuronal cell death,

Bleuel, A., De Gasparo, M., Whitebread, S., Puttner, I., and Monard,in neurodegenerative diseases such as amyotrophic lateral D. (1995). Regulation of protease nexin-1 expression in culturedsclerosis (Festoff, 1980, Festoff, 1992; Sasaki and Maruyama, Schwann cells is mediated by angiotensin II receptors. J. Neu-1994) or Alzheimer’s disease (Appel and Festoff, 1971; De- rosci. 15, 750–761.Kosky and Scheff, 1990; Ho et al., 1994; Smirnova et al., Changeux, J. P., and Danchin, A. (1976). Selective stabilization of

developing synapses as a mechanism for the specification of neu-1994; Terry et al., 1991). Such information may be usefulronal networks. Nature 264, 705–712.for testing new hypotheses and may allow for drug discovery

Colman, H., and Lichtman, J. W. (1993). Interactions between nervein developing new therapeutic agents for such disorders.and muscle: Synapse elimination at the developing neuromuscu-lar junction. Dev. Biol. 156, 1–10.

Connold, A. L., Evers, J. V., and Vrbova, G. (1986). Effect of lowACKNOWLEDGMENTS calcium and protease inhibitors on synapse elimination during

postnatal development in the rat soleus muscle. Brain Res. 393,99–107.We thank Drs. Huaibao Sheng for expert technical assistance,

J. W. Fenton, II (Wadsworth Center, Albany), for providing a-throm- Cunningham, D. D., Pulliam, L., and Vaughan, P. J. (1993). Proteasenexin-1 and thrombin: Injury-related processes in the brain.bin, Khatab Hassanein (Department of Biometry, University of

Kansas Medical Center) for help with statistical analyses, and Law- Thrombos. Haemostas. 70, 168–171.



456 Zoubine et al.

Degen, S. J., Schaefer, L. A., Jamison, C. S., Grant, S. G., Fitzgibbon, Ishii, K., Gerszten, R., Zheng, Y. W., Welsh, J. B., Turck, C. W.,and Coughlin, S. R. (1995). Determinants of thrombin receptorJ. J., Pai, J. A., and Elliott, R. W. (1990). Characterization of the

cDNA coding for mouse prothrombin and localization of the gene cleavage. Receptor domains involved, specificity, and role of theP3 aspartate. J. Biol. Chem. 270, 16435–16440.on mouse chromosome 2. DNA Cell Biol. 9, 487–498.

DeKosky, S. T., and Scheff, S. W. (1990). Synapse loss in frontal Jalink, K., and Moolenaar, W. H. (1992). Thrombin receptor activa-tion causes rapid neural cell rounding and neurite retraction inde-cortex biopsies in Alzheimer’s disease: Correlation with cogni-

tive severity. Ann. Neurol. 27, 457–464. pendent of classic second messengers. J. Cell Biol. 118, 411–419.Jalink, K., van Corven, E. J., Hengeveld, T., Morii, N., Narumiya,Dennis, M. J. (1981). Development of the neuromuscular junction:

Inductive interactions between cells. Ann. Rev. Neurosci 4, 43. S., and Moolenaar, W. H. (1994). Inhibition of lysophosphatidate-and thrombin-induced neurite retraction and neuronal cellDihanich, M., Kaser, M., Reinhard, E., Cunningham, D., and Mo-rounding by ADP ribosylation of the small GTP-binding proteinnard, D. (1991). Prothrombin mRNA is expressed by cells of theRho. J. Cell Biol. 126, 801–810.nervous system. Neuron 6, 575–581.

Lichtman, J. W., and Balice-Gordon, R. J. (1990). UnderstandingDonovan, F. M., Vaughan, P. J., and Cunningham, D. D. (1994).synaptic competition in theory and in practice [Review]. J. Neuro-Regulation of protease nexin-1 target protease specificity by col-biol. 21, 99–106.lagen type IV. J. Biol. Chem. 269, 17199–17205.

Liu, Y., Fields, R. D., Festoff, B. W., and Nelson, P. G. (1994a).English, A. W., and Schwartz, G. (1995). Both basic fibroblastProteolytic action of thrombin is required for electrical activity-growth factor and ciliary neurotrophic factor promote the reten-dependent synapse reduction. Proc. Natl. Acad. Sci. USA 91,tion of polyneuronal innervation of developing skeletal muscle10300–10304.fibers. Dev. Biol. 169, 57–64.

Liu, Y., Fields, R. D., Fitzgerald, S., Festoff, B. W., and Nelson, P. G.Farrell, D. H., Wagner, S. L., Yuan, R. H., and Cunningham, D. D.(1994b). Proteolytic activity, synapse elimination, and the Hebb(1988). Localization of protease nexin-1 on the fibroblast extracel-synapse. J. Neurobiol. 25, 325–335.lular matrix. J. Cell Physiol. 134, 179–188.

Ma, J., Yang, S. X., Ho, G. J., and Festoff, B. W. (1994). Insulin-Fenton, J. W. (1986). Thrombin. Ann. N.Y. Acad. Sci. 485, 5–15.like growth factor binding protein-1 at mouse neuromuscularFenton, J. W. (1989). Thrombin interactions with hirudin. Semin.synapses. Synapse 17, 225–229.Thromb. Hemostas. 15, 265–268.

Monard, D. (1993). Tinkering with certain blood components canFenton, J. W., Villanueva, G. B., Ofosu, F. A., and Maraganore, J. M.engender distinct functions in the nervous system [Review]. Per-(1991). Thrombin inhibition by hirudin: How hirudin inhibitsspect. Dev. Neurobiol. 1, 165–168.thrombin. Haemostasis 21(Suppl. 1), 27–31.

Monard, D., Reinhard, E., Meier, R., Sommer, J., Farmer, L., Rovelli,Fenton, J. W. (1988). Regulation of thrombin generation and func-tions. Semin. Thromb. Hemostas. 14, 234–240. G., and Ortmann, R. (1990). Steps in establishing a biological

relevance for glia-derived nexin. In ‘‘Serine Proteases and TheirFestoff, B. W. (1980). Neuromuscular junction macromolecules inSerpin Inhibitors in the Nervous System: Regulation in Develop-the pathogenesis of amyotrophic leteral sclerosis. Med. Hypothe-ment and in Degenerative and Malignant Disease’’ (B. W. Festoff,ses 6, 121–131.Ed.), pp. 275–281. Plenum, New York.Festoff, B. W. (1987). Proteases, their inhibitors and the extracellu-

Nelson, P. G., Fields, R. D., Yu, C., and Liu, Y. (1993). Synapselar matrix: Factors in nerve–muscle development and mainte-elimination from the mouse neuromuscular junction in vitro: Anance. Adv. Exp. Med. Biol. 209, 25–39.non-Hebbian activity-dependent process. J. Neurobiol. 24, 1517–Festoff, B. W. (1992). Protease cascade dysregulation and synaptic1530.degeneration in amyotrophic lateral sclerosis. In ‘‘Handbook of

Nelson, P. G., Yadav, K., Fitzgerald, S., and Brenneman, D. E. (1994).Amyotrophic Lateral Sclerosis’’ (R. A. Smith, Ed.), Vol. 12, pp.Activation of mouse skeletal muscle in vitro produces increased661–685. Dekker, New York.secretion of thrombin. Soc. Neurosci. Abstr. 20, 1089.Festoff, B. W., Nelson, P. G., and Brenneman, D. E. (1996). Preven-

Niclou, S., Suidan, H. S., Brown-Luedi, M., and Monard, D. (1994).tion of activity-dependent neuronal death: Vasoactive intestinalExpression of the thrombin receptor mRNA in rat brain. Cell.polypeptide stimulates astrocytes to secrete the thrombin-inhib-Mol. Biol. 40, 421–428.iting, neurotrophic serpin, protease nexin I. J. Neurobiol. 30, 255–

266. O’Brien, R., Ostberg, A., and Vrbova, G. (1978). Observation on theelimination of polyneuronal innervation in developing mamma-Festoff, B. W., Rao, J. S., and Hantaı, D. (1991). Plasminogen activa-lian skeletal muscle. J. Physiol. 282, 571–582.tors and inhibitors in the neuromuscular system. III. The serpin

protease nexin I is synthesized by muscle and localized at neuro- O’Brien, R. A., Ostberg, A. J., and Vrbova, G. (1984). Protease inhibi-tors reduce the loss of nerve terminals induced by activity andmuscular synapses. J. Cell. Physiol. 147, 76–86.calcium in developing rat soleus muscles in vitro. NeuroscienceHall, Z. W., and Sanes, J. R. (1993). Synaptic structure and develop-12, 637–646.ment: The neuromuscular junction. Cell 72(Suppl.), 99–121.

Pouyssegur, J., L’Allemain, G., Lenormand, P., Pages, G., Pavirani,Hantaı, D., Rao, J. S., Kahler, C., and Festoff, B. W. (1989). DecreaseA., Rasmussen, U., Van Obberghen-Schilling, E., and Vouret-in plasminogen activator correlates with synapse eliminationCraviari, V. (1993). Alpha-thrombin receptor and MAP kinasesduring neonatal development of mouse skeletal muscle. Proc.in the control of cell growth. Nouv. Rev. Francaise Hematol. 35,Natl. Acad. Sci. USA 86, 362–366.263–264.Ho, G. J., Smirnova, I. V., Akaaboune, M., Hantai, D., and Festoff,

Purves, D., and Lichtman, J. W. (1980). Elimination of synapses inB. W. (1994). Serine proteases and their serpin inhibitors in Alz-the developing nervous system. Science 210, 153–157.heimer’s disease. Biomed. Pharmacother. 48, 296–304.

Riley, D. (1977). Spontaneous elimination of nerve terminals fromHung, D. T., Vu, T. K., Wheaton, V. I., Ishii, K., and Coughlin,the endplates of developing skeletal myofibers. Brain Res. 134,S. R. (1992). Cloned platelet thrombin receptor is necessary for279–285.thrombin-induced platelet activation. J. Clin. Invest. 89, 1350–

1353. Rosing, J., and Tans, G. (1988). Meizothrombin, a major product of




factor Xa-catalyzed prothrombin activation. Thrombos. Haemo- cognitive alterations in Alzheimer’s disease: Synapse loss is themajor correlate of cognitive impairment. Ann. Neurol. 30, 572–stas. 60, 355–360.580.Rovelli, G., Stone, S. R., Preissner, K. T., and Monard, D. (1990).

Tian, W. H., Festoff, B. W., Blot, S., Diaz, J., and Hantaı, D. (1995).Specific interaction of vitronectin with the cell-secreted proteaseSynaptic transmission blockade increases plasminogen activatorinhibitor glia-derived nexin and its thrombin complex. Eur. J.activity in mouse skeletal muscle poisoned with botulinumBiochem. 192, 797–803.toxin type A. Synapse 20, 24–32.Sasaki, S., and Maruyama, S. (1994). Synapse loss in anterior horn

Van Essen, D. C. (1982). ‘‘Neuromuscular Synapse Elimination,’’neurons in amyotrophic lateral sclerosis. Acta Neuropathol. 88,pp. 333–376. Plenum, New York.222–227.

Van Essen, D. C., Gordon, H., Soha, J. M., and Fraser, S. E. (1990).Scott, R. W., Bergman, B. L., Bajpai, A., Hersh, R. T., Rodriguez,Synaptic dynamics at the neuromuscular junction: MechanismsH., Jones, B. N., Barreda, C., Watts, S., and Baker, J. B. (1985).and models. J. Neurobiol. 21, 223–249.Protease nexin. Properties and a modified purification procedure.

Verdiere-Sahuque, M., Akaaboune, M., Lachkar, S., Festoff, B. W.,J. Biol. Chem. 260, 7029–7034.Jandrot Peruss, M., Garcia, L., Barlovatz-Meimon, G., and Hantaı,Siegel, S. (1956). ‘‘Nonparametric Statistics for the Behavioral Sci-D. (1996). Myoblast fusion promotes the appearance of activeences.’’ McGraw-Hill, New York.protease nexin I on human muscle cell surfaces. Exp. Cell Res.Siman, R., Baudry, M., and Lynch, G. S. (1987). Calcium-activated222, 70–76.protease as possible mediator of synaptic plasticity. In ‘‘Synaptic

Vrbova, G., and Fisher, T. J. (1989). The effect of inhibiting theFunction’’ (G. M. Edelman, W. E. I. Gall, and W. M. Cowan, Eds.),calcium activated neutral protease, on the motor unit size afterpp. 519–548. Wiley, New York.partial denervation of the rat soleus muscle. Eur. J. Neurosci. 1,Smirnova, I. V., Ho, G. J., Fenton, J. W. I., and Festoff, B. W. (1994).616–625.Extravascular proteolysis and the nervous system: Serine prote-

Vu, T. K., Hung, D. T., Wheaton, V. I., and Coughlin, S. R. (1991).ase/serpin balance. Semin. Thromb. Hemostas. 20, 426–432.Molecular cloning of a functional thrombin receptor reveals aStone, S. R., Nick, H., Hofsteenge, J., and Monard, D. (1987). Glial-novel proteolytic mechanism of receptor activation. Cell 64,derived neurite-promoting factor is a slow-binding inhibitor of1057–1068.trypsin, thrombin, and urokinase. Arch. Biochem. Biophys. 252,

Witting, J. I., Miller, T. M., and Fenton, J. W. d. (1987). Human237–244.alpha- and gamma-thrombin specificity with tripeptide p-nitro-Suidan, H. S., Stone, S. R., Hemmings, B. A., and Monard, D. (1992).analide substrates under physiologically relevant conditions.Thrombin causes neurite retraction in neuronal cells throughThromb. Res. 46, 567–574.activation of cell surface receptors. Neuron 8, 363–375.

Terry, R. D., Masliah, E., Salmon, D. P., Butters, N., DeTeresa, R., Received for publication June 28, 1996Accepted August 26, 1996Hill, R., Hansen, L. A., and Katzman, R. (1991). Physical basis of



a molecular mechanism for synapse elimination: novel ... · thrombin inhibition delays synapse...

Documents