Motor Neuron M3 Controls Pharyngeal Muscle Relaxation Timing in Caenorhabditis Elegans

The nematode Caenorhabditis elegans has two distinct and almost independent nervous systems. Most of the neurons (282/302 in the adult hermaphrodite, of 104 anatomical types) belong to the somatic nervous system, which controls locomotion, egg⁻ laying, formation and recovery of the environmentally resistant dauer larva, defecation, mating and perhaps other behaviors (White et al. 1986). The smaller pharyngeal nervous system contains only 20 neurons and is concerned with a single behavior: feeding. These 20 neurons are divided into 14 types: six bilaterally symmetrical pairs and eight unpaired neurons (Albertson and Thomson, 1976).

The pharynx (Fig. 1) is a neuromuscular organ responsible for concentrating bacteria suspended in the surrounding fluid, grinding them up and passing them back to the intestine (Albertson and Thomson, 1976). The function of the pharynx is superficially very simple, consisting of only two motions (Avery and Horvitz, 1989). The first, isthmus peristalsis, is a peristaltic contraction of the posterior halves of the pm5 muscles, which carries bacteria from the anterior isthmus to the terminal bulb. The second, pumping, consists of a simultaneous contraction of the corpus, anterior isthmus and terminal bulb, followed by a simultaneous relaxation (Fig. 2A). A pump accomplishes two purposes. First, any bacteria in the terminal bulb are processed through the grinder into the intestine (Doncaster, 1962). Second, bacteria suspended in the surrounding fluid are sucked into the corpus and trapped (Seymour et al. 1983).

Bacterial trapping is the most poorly understood of the pharyngeal functions. Corpus and isthmus muscles are radially oriented, attached on one side to the basement membrane that surrounds the pharynx and on the other to the cuticular lining of the pharyngeal lumen (Albertson and Thomson, 1976). When they contract at the beginning of a pump, the muscles pull the lumen open. Since the posterior isthmus is always closed at this time, liquid is sucked into the corpus through the mouth, along with suspended bacteria. Any bacteria already in the corpus are swept posteriorly. When the muscles relax, the lumen closes and the liquid is spat out. The bacteria, however, are not swept out with the liquid, but stay where they are. The net effect is efficient trapping and posterior transport of bacteria within the corpus and anterior isthmus (Avery and Horvitz, 1989).

It is easy to understand how bacteria are sucked into the pharynx and swept posteriorly during contraction. The puzzle is why they stay where they are during the relaxation. Clearly, corpus relaxation is the act by which bacteria are separated from liquid, probably a critical function for surviving in soil, where food may be dilute. (Indeed, even when supplied with abundant food in the laboratory, worms in which this function is disrupted are severely retarded and stunted; Avery and Horvitz, 1989.) How does it work?

Some light was thrown on this problem by Avery and Horvitz (1989). By using a laser microbeam to kill pharyngeal neurons, we showed that pharyngeal muscles are capable of pumping even after the entire pharyngeal nervous system has been killed. We were able to identify three pharyngeal nervous system functions. The motor neuron M4 was necessary for isthmus peristalsis and essential for viability. The sensory MC neurons were necessary for normal stimulation of pumping rate by bacteria. The remaining 12 types of pharyngeal neurons, which we called the GREENs, were needed for efficient trapping and transport of bacteria. When they were killed, the worm continued to pump, but the bacteria, which swept posteriorly as expected during contraction, swept anteriorly again during relaxation, a phenotype I call ‘slippery pharynx’ (Fig. 2B). One explanation of this result is that bacterial trapping depends on precise timing of the relaxation of the pharyngeal muscles and that the GREENs regulate relaxation timing.

Do the GREENs indeed work by regulating the timing of pharyngeal relaxation? What role do individual GREEN neurons play? The purpose of the experiments reported here was to answer these questions. The results show that GREEN neurons, in particular the M3 motor neurons and the sensory neuron I5, do indeed play a critical role in bacterial trapping and that they regulate the timing of pharyngeal muscle relaxation. M3 speeds up relaxation and I5 slows it down, probably by direct inhibition of M3. I propose, although I have not been able to demonstrate this directly, that M3 and I5 effect trapping by regulating relaxation.

Worms were cultured and handled as described by Sulston and Hodgkin (1988). The wild⁻type was N2. Two mutations were used: unc⁻31(e928) IV and unc⁻29(e1072amber) I. All strains came from the collection of the Medical Research Council Laboratory of Molecular Biology in Cambridge, England.

Neurons were killed by the laser ablation technique of White (White and Horvitz, 1979; Sulston and White, 1980). The beam of a nitrogen laser⁻pumped dye laser was directed via a dichroic reflector through a 100X objective of numerical aperture 1.30 or 1.25 as previously described (Avery and Horvitz, 1987). All operations were performed less than 4h after hatching. Behavioral observations were made on the worms after they had reached the L4 or adult stage.

Many of the operations in this paper involved killing large numbers of neurons. Complex operations are time⁻consuming and success rates are low. As in previous papers (Avery and Horvitz, 1987, 1989), all operations were verified 1 day later. Unlike previous papers, I used some data from flawed operations in order to speed progress in the early stages, reasoning that important conclusions would be verified by later operations. Two assumptions were made in interpreting these flawed kills. (1) Destruction of one member of a bilaterally symmetrical pair of neurons will have no effect on function. (2) Destruction of a neuron will never improve function. Of course, exceptions to both assumptions are possible in principle, but I have seen only minor violations of the first and have never yet seen a violation of the second. This approach was quite successful in reducing the work required to focus attention on M3 and I5.

Eggs were transferred to a fresh plate and worms that hatched within 3h were operated on. (Intact controls were mounted for lasering, then recovered after 12min.) After recovery, they were placed at 15°C and remained at that temperature for the remainder of the experiment, except for brief examinations. All worms were kept together at all times. Kills were verified by Nomarski microscopy 37h after operation (Avery and Horvitz, 1987). As they approached adulthood, the worms were examined every 3h. The lower bound for the time to grow to adulthood was calculated as the last time at which a worm was a larva minus the time of the operation. The upper bound was calculated as the first time at which the worm was an adult minus the time when the eggs were transferred. One worm molted while I was examining it, allowing the calculation of tighter bounds.

All pump duration measurements were made in unc⁻29 hermaphrodites. unc⁻29 encodes a nicotinic acetylcholine receptor subunit (J. T. Fleming, H. A. Riina, D. B. Sattelle and J. A. Lewis, personal communication) that functions in body muscle (Lewis et al. 1980,1987). unc⁻29 mutants, lacking this receptor, move slowly and are easier to videotape than wild⁻type worms. Genetic and pharmacological evidence suggests that the receptor of which the unc⁻29 gene product forms a part is not active in the pharynx (Avery and Horvitz, 1990).

unc⁻29 animals in which M3, I5 or both had been killed were allowed to grow to adulthood, then transferred individually to unlabeled Petri dishes seeded with Escherichia coli. An equal number of age⁻matched intact unc⁻29 worms was also placed on individual dishes and the dishes were randomized and coded by a colleague. 1–2ml of mineral oil was added to each place and distributed over the entire agar surface, then excess oil was removed with a Pasteur pipette. (The oil improves optics by providing a flat transparent surface, similar to a coverslip. It also prevents the microscope objective from fogging with the water vapor released by the agar. Oil has very little effect on the worms: they can go through multiple life cycles under oil, although after a few days some of them appear slightly unhealthy.) The dish was then placed on the stage of a Zeiss Axiophot microscope and observed with bright⁻field optics through a 20X objective. When the worm was immersed in the bacterial lawn and pumping rapidly, an approximately 1min videorecording was made of the terminal bulb. Videorecordings were made with a Hitachi KP⁻160 black⁻and⁻white solid⁻state videocamera and a Panasonic AG⁻1960 Super⁻VHS videocassette recorder.

After recording all the worms, I analyzed the tapes. I searched through each 1min segment for a period of a few seconds during which the worm was pumping rapidly and regularly. I then played back this period at the slowest available speed: 2framess⁻¹ (each frame 1/60s) and counted the number of frames from the first movement of the grinder from its relaxed position until it returned to the fully relaxed state. The terminal bulb is stationary between pumps, so this period was well⁻defined. After counting several dozen pumps, I wrote down a subjectively judged typical pump duration for that worm, either an integral or half⁻integral number of frames. To measure relaxation speed in M3⁻ and intact worms, some of the tapes were reanalyzed, this time counting the number of frames from the beginning of forward grinder motion to the end of relaxation. Because the time resolution of the videotapes was 1/60s (17ms), all measured times are likely to carry a systematic error of about that magnitude.

To compare pump durations and grinder relaxation speeds, I wrote a program to co⁻ rank the merged data from the two sets of worms being compared (e.g. intact versus M3⁻), sum the ranks for the smaller data set, then to compute the probability, if the data were randomly reassorted, that this rank sum would be equally or more extreme (i.e. different from the expected mean). In the special case that the values are all different, this would be equivalent to the Wilcoxon two⁻sample test. The purpose of this special test was not to extract statistical significance from marginal data, but to avoid any questionable assumptions about underlying distributions, even at the expense of being conservative. In fact, none of the conclusions of this paper would be changed had Student’s t⁻test (which assumes normal distributions) been used.

I reasoned that killing GREEN motor neurons should have the same effect as killing all the GREENs, since the remaining interneurons and sensory neurons would have no way to contact effectors. Killing the six GREEN motor neurons M1, M2, M3, M5, MI and NSM had, however, only a very slight effect on behavior (see Table 2, row 1). Pumping became very slightly irregular, but this effect was much less striking than the slippery phenotype that results from killing all 12 GREENs. In particular, M1⁻M2⁻M3⁻M5⁻MI⁻NSM⁻ worms, unlike GREEN⁻, did not have slippery pharynxes. The complementary operation, killing the five sensory and interneurons I2, I3, I4, I5 and I6, gave equally unimpressive results (1 worm). (I1 was spared in this and many of the following operations, because it is technically difficult to kill. Animals in which all the GREENs except I1 have been killed have feeding behavior indistinguishable from GREEN⁻.)

Three possible explanations came to mind. First was the possibility that the neurons were not completely destroyed by the laser. However, extensive controls by Avery and Horvitz (1989) showed that pharyngeal neurons are indeed destroyed by laser killing under identical conditions to those used here. Second, it could be that GREEN interneurons and sensory neurons act through non−GREEN neurons MC and M4. While this may be a partial explanation, it is unlikely to fully account for the lack of effect of the GREEN motor neuron kill, since neither MC nor M4 has been observed to have any effect on the slippery phenotype in any combination (Avery and Horvitz, 1989). The third and most plausible explanation was that some of the neurons classified as sensory or interneurons were able to affect muscle directly. Albertson and Thomson (1976) defined as motor neurons those neurons that had muscle as a postsynaptic partner. However, synapses in C. elegans are very simple. Synapses are made en passant, without synaptic boutons or dendritic spines. Indeed, there are no obvious postsynaptic specializations (Albertson and Thomson, 1976; White et al. 1986). The postsynaptic partner was considered to be the cell nearest the presynaptic density. In the pharynx, all nerve cells and processes are embedded in invaginations of the muscle membrane (Albertson and Thomson, 1976). Process bundles are small (four processes or fewer in most parts of the pharynx; Albertson and Thomson, 1976). Every neuron has direct access to the muscle membrane. It is possible, therefore, that some cells defined as sensory or interneurons could release neurotransmitter onto pharyngeal muscle.

An alternative approach to determining the functions of the GREENs was to examine what particular GREEN motor neurons could accomplish by themselves, in the absence of any other GREEN neurons. Three GREEN motor neurons innervate corpus muscles: M1, which innervates the anterior procorpus, M2, which innervates metacorpus and isthmus, and M3, which innervates metacorpus and anterior isthmus (Fig. 1). The results of operations in which all GREEN neurons except some combination of M1, M2 and M3 were killed are shown in Table 1.

M3 had a striking effect: by itself, it was sufficient for essentially normal trapping of bacteria in the entire corpus. M1 by itself improved the operation of the procorpus, but the metacorpus remained slippery. M2 had no effect in any combination. (I tried to kill all the GREENs except M2 in two worms but, because the laser intensity was low, some of the other GREENs survived. In one of these, where I6 and one of the NSMs survived as well as the M2s, pumping was not improved over GREEN⁻ (data not shown). It therefore seems unlikely that M2 can have any effect on corpus function in the absence of other GREENs. See Materials and methods for a discussion of the assumptions used in interpreting such operations.) I conclude that M1 plays a minor role in effecting efficient trapping of bacteria in the procorpus and M3 a major role in bacterial trapping throughout the corpus.

These operations also affected pump duration. The pump duration (i.e. the interval from the beginning of contraction to the end of relaxation) was obviously longer in GREEN⁻ than in intact worms. However, when all GREENs except M3 were killed, pump durations were obviously briefer than in intact worms. These observations suggested that M3 decreases pump duration.

In the second approach to elucidating GREEN motor neuron function, I performed the five operations consisting of killing all six motor neurons M1, M2, M3, M5, MI and NSM, plus one of the sensory or interneurons I1, I2, I4, I5 and I6. I2, I4, I5 and I6 had apparent effects (data not shown). The I2, I4 and I6 effects have yet to be carefully verified. The clearest effect was a slippery isthmus phenotype that resulted when I5 and the GREEN motor neurons were killed (Table 2, row 2).

It seemed unlikely that all the motor neurons were relevant to the I5 effect. To identify the important ones, I killed six of the seven neurons M1, M2, M3, M5, MI, NSM and I5, sparing just one of the six motor neurons (Table 2, rows 3–8). In every case, the worm had a slippery isthmus. I therefore repeated the experiment, but without M5, killing five of the six neurons M1, M2, M3, MI, NSM and I5 (Table 2, rows 9–14). This time there was a clear effect when M3 was spared: isthmus function was normal in M1⁻M2⁻MI⁻NSM⁻I5⁻ worms. Killing just M3 and I5 also produced a slippery isthmus effect, confirming that M3 was the important motor neuron in the I5 effect (Table 2, row 15). In confirmation of previous results (Avery and Horvitz, 1989), neither I5 nor M3 was individually necessary to prevent slippery isthmus (Table 2, rows 16 and 17), although some I5⁻ worms had a weak slippery isthmus effect (Table 2, row 16). The I5⁻M3⁻ effect was confirmed in five additional experiments in which one GREEN motor neuron in addition to M3 and I5 was killed (Table 2, rows 18–22). Killing any single GREEN motor neuron in addition to M3 and I5 had no additional effect on function.

Since the distinction between the weak slippery isthmus defect in some I5⁻ worms (Table 2, row 16) and the strong slippery isthmus effect of all I5⁻M3⁻ worms (Table 2, row 15) was subjective, I attempted to quantify the effects on feeding by measuring the time required for growth from hatching to adulthood in intact, M3⁻, I5 ⁻ and I5⁻M3⁻ worms, on the hypothesis that feeding efficiency might limit growth rate (Fig. 3). The results were consistent with the visual observations of isthmus function, in that the magnitude of the defect was variable, I5⁻M3⁻ worms were more defective than I5⁻ and, with one possible exception, the intact and I5⁻M3⁻ distributions did not overlap. In addition, the results suggest that feeding efficiency may indeed be one factor that limits growth rate, since slight defects in pharyngeal function caused by these operations were sufficient to retard development measurably. The variability in growth rate caused by killing neurons is interesting, suggesting that one reason for the redundancy may be to compensate for variation from worm to worm in the functions of M3 and I5. However, I cannot eliminate the possibility that the variability in growth rate is due to variability in laser killing (how long the neurons live after being shot, for instance).

In conclusion, the GREEN motor neuron M3 and sensory neuron I5 are collectively, but not individually, necessary for trapping of bacteria in the isthmus.

The third approach to GREEN motor neuron function, and the most informative mechanistically, relied on unc−31 mutant worms. Genetic studies suggested that because unc−31 mutations cause rapid, regular pumping (L. Avery, C. I. Bargmann and H. R. Horvitz, in preparation), some nervous system defects that have little effect on wild−type pharyngeal function might produce obvious effects in an unc−31 mutant background. I therefore killed M1, M2, M3 and M5 in unc−31 worms. A visible abnormality in pharyngeal function resulted: pumps were obviously longer−lasting than normal in three out of five M1⁻M2⁻M3⁻M5⁻unc−31 worms. By killing the individual motor neurons, I found that the effect could be entirely accounted for by M3. This visible increase in pump duration could be scored blind: when five M3⁻unc−31 worms were randomly mixed with six M1⁻unc−31 worms from the same experiment, the M3 ⁻ worms could be picked out on the basis of pump duration (P=0.002). Pump duration was also visibly increased in M3⁻ wild−type worms (4/5 worms), although the effect was less obvious than in unc−31 worms. Indeed, this may have been the source of the slight irregularity in pumping in earlier motor neuron kills. The effect of M3 on pump duration was quantified by analysis of terminal bulb motion on videotapes (Fig. 4). Killing M3 increased pump duration by about 26ms. Pumps were 11ms shorter in I5⁻ than in intact worms (Fig. 4). These differences (M3⁻versus intact, I5⁻versus intact and M3⁻versus I5⁻) are statistically highly significant.

I5 has synapses on M3 (Fig. 1; Albertson and Thomson, 1976). Since I5 has an effect on pump duration opposite to that of M3, it seemed likely that I5 might increase pump duration by inhibiting M3. If so, I5⁻M3⁻ worms should have long pumps, indistinguishable from those of M3⁻ and clearly different from those of I5⁻ worms. This prediction was confirmed, both by visual observation (4/4 worms) and by video analysis (Fig. 4).

To find out whether the negative effect of M3 on pump duration might be due to an increase in the speed of relaxation, I reanalyzed videotapes of M3⁻ and intact worms, measuring the time required for the grinder to snap from its contracted to its relaxed position, a subinterval of the entire relaxation. The motion lasted 42ms in intact worms (N=11) and 53ms in M3⁻ worms (N=11) and the difference was statistically significant (P<0.05).

In conclusion, M3 and I5 have opposite effects on pump duration, M3 decreasing and I5 increasing it. I5 probably produces its effect on pump duration by inhibition of M3.

The M3 effect on pump duration is at least partially due to an increase in the speed of pharyngeal relaxation.

The strongest and most direct conclusion of this work is that the motor neuron M3 influences the speed of terminal bulb relaxation. This is based on two measurements. First, the overall duration of a pump, measured as the time from the beginning of terminal bulb contraction to the end of terminal bulb relaxation, increases when M3 is killed (Fig. 4). Second, terminal bulb relaxation, measured as the time spent by the grinder in moving from its contracted to its relaxed position, occurs more slowly when M3 is killed (53 versus 42ms).

I5 also influences overall pump duration (Fig. 4). I5 synapses on M3 (Albertson and Thomson, 1976) and I5 ⁻M3⁻ worms have pump durations indistinguishable from M3⁻ worms. These observations suggest that I5 slows down relaxation by inhibiting M3.

If I5 acts through M3, why do some I5⁻ worms and all I5⁻M3⁻ worms have defects in transport of bacteria in the isthmus (slippery isthmus), unlike M3⁻ worms? The answer must be that, although the effects of I5 on pump duration may be mediated by M3, it has other effects on pharyngeal function not mediated by M3. I5 is known to synapse on M4, a motor neuron that innervates the posterior isthmus (Fig. 1; Albertson and Thomson, 1976) and that is necessary for posterior isthmus function (Avery and Horvitz, 1987). I suspect that the slippery isthmus defect results when M3 is non−functional and M4 is free of the influence of I5.

A second direct conclusion of this work is that M3 and I5 are important for the effective trapping and transport of bacteria in the pharyngeal lumen. This conclusion is based on the observation that, under various conditions, M3 and I5 are necessary for effective bacterial transport in various regions of the pharynx. M3 is sufficient by itself to prevent slippery corpus in the absence of any other of the GREEN neurons. The isthmus becomes slippery when both M3 and I5 are killed.

Since the slippery phenotype appears to be an abnormality of the motion of bacteria in the pharyngeal lumen during relaxation (the bacteria slip anteriorly, rather than staying where they are), it is plausible that slippery pharynx is actually caused by an abnormality in the timing of relaxation. At the moment this hypothesis is supported by a correlation (that M3 and I5 affect both) but by no direct evidence.

Interestingly, a mutation in the gene eat−4 causes long pumps and a slippery isthmus, just as when M3 and I5 are killed in the wild type (M. Avery, in preparation). eat−4 pharynxes still contain M3 and I5 nuclei, so the defect is probably not in the development of these neurons. The neurons may not function properly, or there may be a defect in muscle cells that block their ability to respond.

At first, the question of the function of the pharyngeal nervous system seems simple, perhaps even foolish. Without M4, isthmus peristalsis does not happen, no bacteria reach the intestine and the worm starves (Avery and Horvitz, 1987). Without MC, worms fail to respond to bacteria and, therefore, pump slowly even when there is abundant food, are retarded, stunted and produce few progeny (Avery and Horvitz, 1989). Without the GREENs the pharynx becomes slippery, few bacteria reach the intestine, again resulting in retardation, stunting and low fertility (Avery and Horvitz, 1989). Obviously, the function of the pharyngeal nervous system is to prevent these things.

This answer is not satisfactory. To show why, consider another functional question: why is there isthmus peristalsis? Without isthmus peristalsis the worm swallows no food and starves. But why does the worm swallow no food? Why is food not swallowed during the pump? The answer is that posterior pm5, unique among the large pharyngeal muscles, fails to contract during a pump (Albertson and Thomson, 1976; Avery and Horvitz, 1989). In fact, since each pm5 muscle cell extends the entire length of the isthmus (Albertson and Thomson, 1976), the failure of the posterior isthmus to open during a pump is a consequence of subcellular specialization. Why has evolution produced this seemingly needless elaboration: a subcellular specialization to keep the posterior isthmus closed during a pump and a neuron, M4, whose sole function is to undo the damage by telling the muscles to contract?

A possible answer is that the corpus and terminal bulb do not work well unless they are hydrodynamically isolated from each other during a pump. When the corpus contracts, the luminal pressure decreases and liquid and suspended bacteria are sucked in at the mouth. If the isthmus were open, liquid and bacteria could instead be sucked anteriorly from the terminal bulb. This would interfere both with the corpus function of accumulating bacteria from the outside liquid and with the terminal bulb function of moving bacteria posteriorly to the intestine. Interestingly, the lumen of the isthmus, unlike that of the procorpus, is so shaped that it closes tightly when relaxed (compare Figs 5 and 6 of Albertson and Thomson, 1976). If this explanation is right, the function of isthmus peristalsis, or, more accurately, the function of the design that requires isthmus peristalsis, is to isolate the corpus and terminal bulb while they act yet allow bacteria to pass after the dust has settled.

This example shows that to answer questions about function it is not sufficient to consider what goes wrong when it is eliminated. One must consider why the rest of the system is designed in such a way as to make that function necessary.

What then is the function of the pharyngeal nervous system? A nervous system is to some extent a liability, because it is vulnerable to attack. Many organisms that need to prey on or protect themselves against animals do so by attacking the nervous system: plants produce poisonous alkaloids, scorpions make potassium channel blockers and puffer fish contain a sodium channel blocker. There must be a reason why the pharynx has evolved in such a way as to require nervous system function for efficient operation. I propose that the function of the pharyngeal nervous system is primarily sensory: rather than directly producing pharyngeal motions, the pharyngeal nervous system allows the motions to change in response to a changing environment.

In this regard, the pharyngeal nervous system differs from the lobster stomatogastric ganglion, another small nervous network involved in feeding (Selverston et al. 1976). The function of the stomatogastric ganglion, which contains no sensory neurons, is to generate repetitive, properly phased muscle contractions. Thus, its function is probably carried out by pharyngeal muscles in nematodes. The role of the pharyngeal nervous system is probably more similar to that of the modulatory inputs to the stomatogastric ganglion (Nagy and Moulins, 1987).

The clearest example of a useful pharyngeal sensory neuron is MC, which is necessary for a response to bacteria to occur. Without MC, pumping is slow and irregular, even in the presence of abundant food, and growth is retarded (Avery and Horvitz, 1989). It might have been possible for evolution to produce pharyngeal muscles that pump rapidly all the time without nervous system input, but this would also be unsatisfactory, since it would waste energy unnecessarily when there was no food available, a time when waste of energy can be least afforded. Efficient use of resources requires the ability to sense whether there are bacteria present, an ability provided by MC.

For M4 and the GREENs, I believe the answer is generally similar, but I cannot be as specific. That is, their function is not simply to permit isthmus peristalsis and maximal bacterial trapping, respectively, since there is no obvious reason why the muscles could not be designed so as to do these things constitutively. Rather, they serve to regulate isthmus peristalsis and bacterial trapping in response to changes in the environment or in the pharyngeal lumen. For instance, I5, which has sensory endings in the terminal bulb (Albertson and Thomson, 1976), might sense when the terminal bulb is becoming too full for effective grinder operation. It would respond by inhibiting M3 and M4. Inhibition of M4 would prevent isthmus peristalsis from bringing more bacteria into the already full terminal bulb. M3 inhibition would slow down posterior transport of bacteria in the corpus.

This work was supported by a Basil O’Connor Research Grant from the March of Dimes. I am grateful to Siegfried Hekimi, Flora Katz, Raymond Lee and David Raizen for comments.