Chemosensory Neurons with Overlapping Functions Direct - POST

keuron, Vol. 7, 729-742, November, 1991, Copyright 0 1991 by Cell Press

Chemosensory Neurons with Overlapping Functions Direct Chemotaxis to Multiple Chemicals in C. elegans Cornelia I. Bargmann* and H. Robert Horvitz Howard Hughes Medical Institute Department of Biology Massachusetts Institute of Technology Cambridge, Massachusetts 02139

Summary

The functions of the 11 classes of exposed chemosensory neurons of C. elegans were tested by killing cells with a laser microbeam. One pair of neurons, the ASE neurons, is uniquely important for chemotaxis: killing the ASE neurons greatly reduced chemotaxis to CAMP, biotin, Cl-, and Na+. Additional chemosensory function is distributed among several other cell types. Thus, 3 pairs of chemosensory neurons (ADF, ASG, and ASI) contribute to a residual response to CAMP, biotin, Cl-, and Na+after ASE is killed. Chemotaxis to lysine similarly depends on the partly redundant functions of 4 pairs of chemosensory neurons (ASE, ASC, ASI, and ASK). The combined activity of several neuron types that act in parallel might increase the fidelity of chemotaxis.

Introduction

Chemosensation provides an animal with important information about its environment: attractive and re- pulsive molecules are recognized by organisms from bacteria to vertebrates (Adler, 1975; Van Houten et al., 1981; Van Haastert et al., 1982; Finger and Silver, 1987). 1~1 multicellular animals, specialized cells in the nervous system respond to chemical stimuli. It is known tilat different categories of molecules are recognized by different kinds of sensory neurons: for example, in vertebrates volatile molecules are recognized by olfactory neurons, and salts and sugars by taste bud neurons. However, the mechanisms by which these neuronsprocessand integrate manydiversechemical stimuli are not well understood.

The nematode Caenorhabditis elegans is amenable to detailed study of nervous system structure and function. The C. elegans nervous system, consisting of 302 neurons in the adult hermaphrodite, has been analyzed using electron micrographs of serial sec- t:ons(Whiteetal.,1976,1986;AlbertsonandThomson, 1976). This work has led to a description of the num- her, process morphologies, and positions of all of the neurons, as well as identification of each of the mor- phologically identifiable synapses and gap junctions between neurons and between neurons and muscles. In principle, this’wiring diagram”should provide the basis for much of the behavior of this animal.

i l Present address: Department of Anatomy, University of Califor- r ia, San Francisco, San Francisco, California 941420452.

The structural description of a neuronal circuit per- mits analysis of the functional relationships between particular neurons and particular behaviors (Selver- ston and Miller, 1980; Chiel et al., 1988; Nusbaum and Marder, 1989). In C. elegans, this relationship can be elucidated by killing identified cell types with a laser microbeam (Sulston and White, 1980) and observing the effects on behavior (Chalfie and Sulston, 1981; Chalfie et al., 1985; Avery and Horvitz, 1987, 1989). Since C. elegans has few neurons and since most of the neuronscan be unambiguously identified in living animals by morphology and position, it is possible to target particular cells for destruction with a high degree of accuracy. Laser damage to the nucleus of a neuron in young animals results in a loss of the function of that neuron (Chalfie et al., 1985; Avery and Horvitz, 1987, 1989). Thus, hypotheses concerning neuronal function can be tested by killing a cell and subsequently examining the animal for behavioral abnormalities. Using a combination of laser ablation experiments and genetic data, strong evidence for the involvement of mechanosensory neurons in touch sensation (Chalfie and Sulston, 1981; Chalfie et al., 1985) and for the involvement of particular motor neurons in movement (Chalfie et al., 1985; J. White, personal communication; S. Mclntire and R. Horvitz, unpublished data), egg-laying (Trent et al., 1983; Desai and Horvitz, 1989), and feeding (Avery and Horvitz, 1987, 1989) has been gathered.

Among the behaviors that have been described for C. elegans, many are regulated by the presence of bacteria, its food. Bacteria cause increased feeding and egg-laying and decreased movement of the animal (Horvitz et al., 1982; Avery and Horvitz, 1990). In addition, the animal can find food in its environment. C.eleganscanchemotaxtowardthepeakofagradient of a number of small molecules, including positively charged ions such as Na+ and K+, negatively charged ions such as Cl-and S042-, basic pH, and several small organic molecules, including CAMP, cGMP, lysine, cysteine, and histidine (Ward, 1973; Dusenbery, 1974). In addition, using these or other cues, animals can locate bacteria and males can locate hermaphrodites (J. Sulston, R. Horvitz, and E. Jorgensen, unpublished data). Chemotaxis represents a complex response of the animal to its environment: the animal must sense continuously changing external conditions and con- vert sensory information into movement in a particular direction. In this work, we identify the C. elegans chemosensory neurons required for normal responses to a number of chemical attractants.

Results

Assays and Scoring of Chemotaxis Responses Chemotaxis was measured using a modified version of an assay developed by Ward (1973; see also Experi-

Figure 1. Chemotaxis Assay and Scoring

A chemotaxis assay of a single wild-type animal toward Cl- is shown at top, and the interpretation of the animal’s behavior beneath. The animal was placed near the center of the plate at the origin. The labeled plug (at left) was the source of a gradient of Cl-. A control area (unlabeled, at right) is indicated. After an hour, the animal was removed and the tracks the animal made on the surface of the agar plate were photographed in a contact print of the plate. These tracks are visible to the naked eye. According to the interpretation of this assay, the animal travelled to the peak of the gradient immediately after being placed on the plate (track I), left the peak of the gradient, and returned twice(tracks2and 3, which might have occurred in either order). The animal’s pattern of approaching and leaving the attractant several times may reflect sensory adaptation during the assay (Ward, 1973). To score the assay, the number of trips to the Cl- (A) was counted, and the number of trips to the control area (6) was subtracted from (A). If A - B is greater than zero, the animal is scored as responding to the attractant. Here, A - B = 3, so this animal is scored as responding to Cl-.

mental Procedures). Briefly, agradient of an attractant is established in a IO-cm petri plate containing agar by placing an agar plug soaked in attractant in awell near one edge of the plate and allowing the attractant to diffuse for 12-24 hr. A single animal is placed near the center of the plate, about 5 cm from the peak of the gradient. After an hour, the animal is removed and its tracks on the agar are examined.

A wild-type animal responds to an attractant in over

90% of chemotaxis assays. Typically, the animal will move to the peak of the gradient and wander away and back several times in an hour (Figure 1). Occasion- ally an animal will move to the peak of the gradient and remain there for the duration of the assay. A simple scoring method that counted each assay as either positive or negative was devised. Assays were scored by counting the number of times an animal that moved well arrived at the peak of the gradient and subtracting the number of times the animal arrived at a control plug at the opposite side of the plate. If the difference was greater than zero, the assay was scored as positive; if it was less than or equal to zero, the assay was scored as negative.

Table 1 lists the responses of wild-type animals to six different attractants. CAMP, Cl-, and Na+ are each representative members of classes of attractants known to be distinguished from one another by C. elegans (Ward, 1973). Lysine is aweak attractant identified also by Ward (1973). Biotin and serotonin are strong attractants that we have identified (Table IA; also unpublished data). Between 75% and 93% of the animals responded to each of these attractants.

Wild-type animals were observed also on similar plates without added attractants. Twenty-three per- cent of the animals gave an apparent response to the mock attractant in these control assays (see Experi- mental Procedures). Thus, there is a false-positive response rate of 23% in the assay as it is scored. More complex scoring methods that took into account total movement or graded responses were no more effective than the one we used in distinguishing the responses of wild-type animals in the presence or absence of attractants (data not shown).

Anatomical analysis of mutant animals defective in chemotaxis has indicated that exposed ciliated neurons are required for chemosensation (Lewis and Hodgkin,1977; Perkinsetal.,1986).Todeterminewhat component of the chemotaxis measured in our assay reflects the function of exposed ciliated neurons, we examined the mutant the-2(e7033). the-2 animals are chemotaxis-defective (hence the name the) and are also abnormal in other behaviors regulated by sensory stimuli (Lewis and Hodgkin, 1977). the-2animals have structural abnormalities in most of the ciliated neurons of the head, including all of the neurons with endings exposed to the environment that are thought to be chemosensory (Lewis and Hodgkin, 1977). The severe structural defects in the-2 animals should lead to a profound defect in sensory neuron function.

The results of chemotaxis assays with the-2 animals are shown in Table IA and Figure 3A. the-2animals are defective in their ability to chemotax to all attractants tested when compared with wild-type animals (12%- 32% of animals responded to each attractant, p < 0.001 in all cases). An independent estimation of the level of chemotaxis for the-2 animals was made by scoring a number of assays in reverse, i.e., as if the control plug had been the attractant (see Experimental Proce- dures). In this case, the same chemotaxis index was

Chemosensory Neurons in C. elegans i 31

derived for the control plug as for the attractant plug. Thus, the-2 animals show neither chemotaxis toward nor avoidance of these attractants (although there could be a weak residual ability to chemotax to biotin in the-2 animals).

Based on the combined responses of the-2 animals and of wild-type animals in the absence of attractant, 2, negativecontrol value representingzero chemotaxis in the assay was defined (see Experimental Proce- dures). This value corresponds to an apparent response of 20% in the chemotaxis assays.

We were concerned that laser surgery itself might cause a slight decrease in chemotaxis, since surgery involves anesthetization, tissue damage, and trauma to the animal. There is a subtle but statistically significant decrease in the overall chemotaxis efficiency of animals that have undergone any laser surgery (Table IM; also unpublished data). Therefore, we have compared all operated animals with both intact wild-type animals and a control set of ablations in which cells apparently uninvolved in chemotaxis were killed (for each set of experiments, the statistical analysis is described in detail in Experimental Procedures).

Functional Amphid Neurons Are Required for Chemotaxis Twenty-six neurons in C. elegans have been proposed to function as chemosensory cells based on the obser- lation that they have ciliated endings that areexposed to the environment through specialized structures in the cuticle (Ward et al., 1975; Ware et al., 1975; White et al., 1986). The cell bodies of 22 of these cells are located in ganglia at the anterior end of the animal. The processes of these neurons run to the tip of the animal’s nose, where they are presented to the environment by the amphid and inner labial sensilla. The other 4 cells have sensory endings in the phasmid sensilla at the posterior end of the animal. Figure 2 illustrates the positions of the cell bodies located on

the left side of the animal (all cells belong to bilaterally symmetrical pairs) and the locations of the chemosensory process bundles.

There are 2 amphids, 1 each on the right and left sides of the animal, and each amphid contains 1 of a left-right pair of each type of neuron (Ward et al., 1975; Ware et al., 1975; White et al., 1986). The inner labial sensilla have a 6-fold symmetry, with the cells and sensilla radially arrayed around the animal’s mouth. Thus, the chemosensory cells of the head can be considered to be of 9 types: 2 cells each of 8 types of amphidial neurons (ASE, ADF, ASG, ASH, ASI, ASJ, ASK, and ADL), and 6 cells of 1 type of inner labial neuron (IL2). In general, we will refer, for example, to the left and right ADF cells collectively as the ADF cells.

Previous experiments have indicated that phasmid function is unnecessary for chemotaxis (Ward, 1973). \Veconfirmed these results bytestinganimals bearing mutations in the gene /in-77. In lit-77 animals, the phasmid opening is defective, so the phasmid neu-

rons are not exposed to the environment correctly (Sternberg and Horvitz, 1988; E. Hedgecock, personal communication). /in-77 animals were tested and responded as well as wild-type animals to biotin, lysine, Cl-, and serotonin (data not shown). Thus, exposed phasmid neurons are not essential for chemotaxis to these attractants.

The role of other sensory cell types in chemotaxis was assessed by killing particular cell types using a laser microbeam (see Experimental Procedures). Be- cause of the similarity of the left and right representa- tives of each cell type, both cells were killed in a single animal. The cells were killed during the first larval stage, shortly after hatching. Animals were tested as adults about 3 days later. Single animals were usually tested once fortheir response to each attractant. How- ever, since chemotaxis-defective animals sometimes crawlontotheedgeoftheplasticdishanddie,animals were often lost before they could be tested with all attractants.

When all 8 exposed cell types of the amphid were killed, the animals became defective in chemotaxis to all attractants (overall response = 22%; Table IB, II). This result indicates that some of these cells are required for normal responses. However, the large number of cells killed in these animals could poten- tially lead to nonspecific defects in behavior; this con- cern is addressed by experiments presented below.

By contrast, the inner labial IL2cells are not essential for at least some chemotactic responses (overall response = 89%; Table IB, II), and the inner labial sensilla were not studied further.

The ASE Neurons Are Important for Chemotaxis to Several Attractants To facilitate analysis of the amphid cell types, subsets of the amphid cells were killed in single animals, typically about 3 types of cells per animal. Pilot experiments included at least 5 animals in which each type of cell was killed. Despite the large number of cells killed, most animals of this sort were indistinguishable from intact wild-type animals (data not shown). Only animals in which the ASE amphid cell type was killed were noticeably defective in their chemotaxis behaviors.

The involvement of the ASE cells in chemotaxis was examined further by killing only the ASE cells in individual animals. Animals lacking only ASE were defective in their ability to chemotax to CAMP, biotin, Cl-, and Na+compared with wild-typeanimals or operated animals in which nonessential cells were killed (41% of the assays were positive, p < 0.001 in all cases; Table IC, IJ; Figure 3A). Chemotaxis to lysine and serotonin by these animals was also inefficient compared with that in wild-type animals, but was not significantly different from that in animals in which nonessential cells were killed. Thus this effect may not be specific.

Animals in which 1 or more of the other (non-ASE) 7 exposed amphid cell types were killed were also tested for chemotaxis. None of these cells was essen-

Tabl

e 1.

C

hem

otax

is

of

Wild

-Typ

e,

Mut

ant,

and

Ope

rate

d An

imal

s

Indi

vidu

al

Assa

ys

Cal

cula

ted

Valu

es

for

Com

bine

d As

says

CA/

Bid

CA

MP

95

%

n B

iotin

95

%

n C

l- 95

%

n N

a’

95%

n

Lysi

ne

95%

n

Sero

toni

n 95

%

n A

ll 95

%

n C

l /N

a+

95%

n

.86-

.97

96

0.91

.8

3-.9

6 12

7 0.

87

.78-

.93

99

0.75

.6

2-.a

5 68

0.

86

.20-

.46

50

0.15

',b

.06-

.29

48

0.1F

.b

.07-

.37

32

0.14

',b

.05-

.29

37

0.12

'.b

.Ol-.

4a

9 0.

40"

.12-

.74

10

0.11

” .O

l-/la

9

0.0"

-

5 0.

14

0.92

.6

2-.9

9 12

1.

0 7

ND

0.

85

.34-

.70

31

0.38

" .2

4-.5

2 46

0.

37"

.23-

.51

47

0.48

.3

1-.6

5 33

0.

60

.60-

.98

15

0.64

.3

5-.8

7 14

0.

64

.35-

.87

14

ND

0.

64

.3a-

A8

15

0.57

.2

8-.a

2 14

0.

64

.35-

.87

14

ND

0.

64

A.

Con

trols

W

ild-ty

pe

0.92

.8

5-.9

6 10

6 0.

93

the-

2 0.

16".b

.0

8-.2

8 56

0.

32",b

B. A

mph

id

Cel

ls

or

IL2

cells

ki

lled

Cel

ls

kille

d:

EFC

HIJ

KL

0.40

" .l2

-.74

10

0.11

” IL

2 0.

67

- 3

ND

C.

AS

E n

euro

ns

kille

d C

ells

ki

lled:

E

0.

42"

.25-

.59

31

0.52

" E(

left)

0.

73

.45-

.92

15

0.87

E(

right

) 0.

43"

.16-

.69

16

0.67

D.

Oth

er

amph

id

cells

ki

lled

Cel

ls

kille

d:

F 0.

84

.62-

.96

22

0.94

G

0.

81

.54-

.96

16

1.0

H

0.93

.6

6-.9

9 14

1.

0 I

0.81

.6

0-.9

4 24

0.

95

J 0.

91

.58-

.99

11

0.80

K

0.

89

.65-

.98

18

1.0

L 0.

85

.55-

.9a

13

0.83

E. C

ombi

natio

ns

of

amph

id

cells

ki

lled

Cel

ls

kille

d:

EFC

IK

0.10

” .O

l-.43

10

N

D

EHJL

0.

73

.39-

.94

11

0.55

EF

GI

0.34

" .1

5-.5

7 22

0.

14”

EFC

0.

64

.35-

.87

14

0.54

" EF

I 0.

59

.26-

.86

11

0.44

" EC

I 0.

47"

.22-

.74

15

0.33

' FC

I 0.

70

.35-

.93

IO

1.0

ECIK

0.

40"

.12-

.74

10

0.50

" E

CK

0.

33'

.12-

.61

15

0.40

E

IK

0.36

" .1

3-.6

5 14

0.

38"

GIK

0.

91

.58-

.99

11

1.0

FCIK

0.

89

.52-

.99

9 0.

86

EF

0.39

.2

0-.6

2 23

0.

41'

El

0.50

.2

4-.7

6 15

0.

54"

F. “

Non

esse

ntta

l” co

ntro

ls

Cel

ls

kille

d:

Vario

us

cell\

0.

88

.79-

.97

50

0.88

7L.9

9 18

0.

92

.76-

.98

30

0.70

.4

9-.8

6 27

0.

47

.23-

.72

17

0.90

.7

4-1.

0 12

0.

85

.55-

.98

13

0.67

.3

5-.9

1 12

0.

50

.18-

.82

10

0.83

.7

4-1.

0 12

0.

92

.62-

.99

12

0.60

.2

6-.8

8 10

0.

67

.30-

.93

9 0.

80

.75-

.99

21

0.90

.7

1-.9

8 24

0.

62

.38-

.82

21

0.58

.3

3-.a

0 19

0.

86

.44-

.97

10

0.92

.6

2-.9

9 12

0.

67

.35-

.91

12

0.90

.5

5-.9

9 10

0.

91

.77-

1.0

14

0.85

.5

9-.9

7 17

0.

63

.36-

.85

16

0.63

.3

2-.8

7 12

0.

93

.51-

.98

12

0.89

.6

1-.9

9 74

0.

79

.49-

.96

14

0.85

.5

5-.9

8 73

0.

91

o.o-

.31

.23-

.a4

.03-

.38

.26-

.80

.14-

.78

.12-

.61

.18-

.82

.09-

.76

0.0”

11

0.

29"

18

0.23

” 14

0.

75

9 0.

40"

15

0.27

& 7

0.86

10

0.

63

5 0.

30"

8 0.

17"

2 0.

75

7 0.

89

23

0.38

" 14

0.

50"

.22-

.63

.26-

.80

.07-

.47

.43-

.94

.12-

.74

.oa-

.55

.46-

.99

.25-

.92

.07-

.65

.02-

.4a

.35-

.97

.52-

.99

.18-

.62

.23-

.77

10

ND

0.

8 7

0.67

-

6 0.

43

20

0.18

" .0

4-.4

3 17

0.

47

12

0.27

" .0

6-.6

1 11

0.

42

10

0.35

" .0

9-.7

0 10

0.

44

15

0.39

" .1

5-.6

8 14

0.

63

a 0.

50

.16-

.84

8 0.

33

a 0.

4~

.14-

.78

9 0.

0"

10

0.42

" .1

5-.7

3 12

0.

43

12

0.27

" .0

6-.6

1 11

0.

38

8 0.

83

- 6

0.71

9

0.40

5

0.43

20

0.

24"

.08-

A8

19

0.38

14

0.

08"

.Ol-.

38

12

0.42

.79-

.97

48

0.87

.7

8-.9

6 60

0.

83

o.o-

.29

.25-

.71

.15-

.73

.12-

.80

.25-

.92

O&.

31

.18-

.7l

.16-

.65

.13-

.69

.15-

.73

11

0.40

7

0.78

19

0.

50

12

0.82

8

0.50

8

0.61

6

0.84

10

0.

33

14

0.55

16

0.

20"

7 0.

80

7 0.

88

12

0.50

12

0.

47

.73-

.93

54

0.70

.5

6-.a

4 40

0.

66

H.

Con

trols

.7

6-.9

3 88

W

ild-ty

pe

0.88

.8

6-.9

1 58

4 0.

91

.04-

.26

41

the-

2 0.

18’,b

.1

4-.2

3 26

4 0.

20”,b

I. A

mph

id

cells

or

IL

2 ce

lls

kille

d C

ells

ki

lled:

7

EFG

HIJ

KL

0.2P

h .ll

-.33

50

0.26

",b

.54-

.98

13

IL2

0.89

.7

3-.9

7 35

0.

91

J. A

SE

ne

uron

s ki

lled

Cel

ls

kille

d:

k-.7

9 26

E

0.

45"

.38-

.51

214

0.41

" .3

5-.a

7 14

E(

left)

0.

65"

.54-

.75

79

0.71

.35-

.a7

14

E(rig

ht)

0.55

" .4

4-.6

5 84

0.

58"

K. O

ther

am

phid

ce

lls

kille

d C

ells

ki

lled:

.6

4-.9

9 15

F

0.80

.7

3-.8

7 12

9 0.

85

.57-

.9a

12

G

0.79

.6

9-.8

8 75

0.

83

.4l-.

97

10

H

0.84

.7

5-.9

2 67

0.

88

.61-

.97

18

I 0.

79

.72-

.86

127

0.82

.58-

.99

11

J 0.

85

.76-

.93

66

0.82

.6

6-.9

9 14

K

0.

82

.75-

.90

91

0.84

.5

8-.9

9 71

L

0.85

.7

7-.9

3 77

0.

84

L. C

ombi

natio

ns

of

amph

id

cells

ki

lled

Cel

ls

kille

d:

.12-

.74

10

EFG

IK

O.lL

W

.04-

.26

41

0.05

”,”

.4r-.

97

9 EH

JL

0.59

' .4

5-.7

2 51

0.

57"

.27-

.73

19

EFG

I 0.

31”

.23-

.40

115

0.23

","

.,x-.9

7 11

EF

C

0.56

" .4

5-.6

7 74

0.

54"

.16-

.&l

8 EF

I 0.

46"

.32-

.59

56

0.45

"

.32-

.a5

14

EGI

0.43

" .3

2-.5

4 81

0.

36"

7 FC

I 0.

72

.59-

.85

46

0.76

.0

7-.7

0 9

EGIK

0.

38"

.25-

.50

56

0.49

" .2

3-M

11

EG

K 0.

40"

.29-

.52

67

0.36

"

.03-

.56

10

EIK

0.

3W.b

.1

9-.4

0 71

0.

29'"

5 G

IK

0.82

.7

0-.9

4 39

0.

85

.a-.9

9 8

FGIK

0.

76

.63-

.88

45

0.80

.26-

.74

19

EF

0.38

" .3

0-.4

7 11

6 0.

36"

.21-

.74

15

El

0.43

" .3

2-.5

3 82

0.

42'

M.

“Non

esse

ntia

l” co

ntro

ls

Cel

ls

kille

d:

.57-

.a1

41

Vario

us

cells

0.

81

.77-

.86

293

0.86

BE-.9

4 42

8 .1

4-.2

6 18

6

.12-

M

38

.71-

.98

22

.34-

.4a

155

.59-

.83

58

.45-

.71

59

.78-

.92

97

.73-

.93

53

.79-

.97

‘la

.74-

.90

90

.71-

.93

45

,755

.93

65

.74-

.94

53

.Ol-.

25

20

.41-

.73

35

.14-

.31

77

.40-

68

51

.30-

.60

40

.24-

M

59

.61-

.91

33

.33-

.65

37

.21-

.51

42

.16-

.42

45

.66-

.96

27

.62-

.92

30

.26-

.46

85

.29-

.55

55

.81-

.91

212

CJ;emosensory Neurons in C. elegans

tial for normal chemotaxis (Table ID, IK; see Experi- mental Procedures for statistical analysis). Thus ASE is the single most important chemosensory neuron for these chemotactic responses.

Either the left or the right ASE was also killed sepa- rately (Table IC, IJ). The overall defect in chemotaxis observed after killing 1 ASE was about half the magni-

tude of that caused by killing both cells. When the results from all attractants were combined, the effect of killing either the left or right ASE was greater than the effect of killing both members of any other type of amphidial sensory neuron or all 6 IL2 neurons (p < 0.05).

Nonspecific damage caused by laser surgery might kill cells other than those targeted for ablation. To exclude the possibility that a cell adjacent to the ASE neuron was in fact the cell important for chemotaxis, all cellsadjacent tothe ASE neuron were killed with no effect on chemotaxis (see Experimental Procedures). Thus, defective chemotaxis of animals in which the

ASE neurons were laser-ablated is probably due to the deaths of the ASE neurons themselves.

One simple explanation for these results is that the ASE neurons directly sense each of these different attractants. Another possible explanation is that these neurons might integrate sensory information from other neurons, but not have a primary sensory function themselves. We tested this possibility by killing all exposed and nonexposed amphid and phasmid neurons except ASE in 5 animals and comparing their responses with those of 5 animals in which all amphid and phasmid neurons including ASE were killed (Ta- ble IG, IN). When all amphid and phasmid neurons were killed together, the animals were extremely defective in chemotaxis defective (overall response = 18%). By contrast, when only the ASE neurons were allowed to survive, the animals’chemotaxis responses were substantially more efficient (overall response = 69%, p < 0.001). Nine animals that each had a single surviving amphid cell type other than ASE were also tested (the other cell type was ADF [n = 2 animals], ASG [n = I], ASH [n = I], ASI [n = 21, ASJ [n = 21, or ADL [n = I]). These animals were indistinguishable from animals in which all amphid and phasmid cells were killed and were much more defective than the animals in which only the ASE neurons were killed (overall response = 15%, p < 0.001).

This experiment demonstrates that of the amphid and phasmid neurons, the ASE neurons are sufficient as well as necessary for chemotaxis. Although the numbers are too small to determine whether the ASE neurons sense all of the attractants, the data indicate that the ASE neurons sense at least three of the six attractants tested (p < 0.01; see Experimental Proce- dures).

Four Partially Redundant Cell Types Direct Chemotaxis to Many Attractants Although animals in which the ASE neurons were killed were defective in chemotaxis compared with

Neuron 734

Phasmld cells

Figure 2. Positions of the Chemosensory Neurons of C. elegans

(Top) Approximate positions of the nuclei and sensory processes of putative chemosensory neurons. The animal is diagrammed lying on its side, with the dorsal side up and anterior to the left. The sensory processes of the amphid and inner labial neurons are exposed to the environment at the anterior tip of the animal, near the mouth opening (Ward et al., 1975; Ware et al., 1975). The phasmid neuron processes are exposed through posterior sensilla. Only neurons on the left side are shown; a bilaterally symmetrical set of neurons is present on the right side. (Bottom) Positions of the nuclei of the neurons in the head region of a newly hatched Ll, reproduced from Sulston et al. (1983). Again, the left side of the animal is shown, with the dorsal side up and anterior to the left. Most but not all of the neurons have bilaterally symmetrical counterparts on the right side. Putative chemosensory neurons are black. For more details about the anatomy of the head, see Sulston et al. (1983) and White et al. (1986).

wild-type animals, these animals still responded to all attractants at a level higher than the negative control valueof 20% (p<O.Ol in all cases; Figure3A;TablelC). However, the killing of no other single type of cell

gave a defect as great as that observed after killing

ASE. Therefore, more than 1 type of cell must direct chemotaxis. To identify the additional cells responsible for this residual response, we killed ASE together with other types of cells in single animals. We rea- soned that the additional cells might lie in the amphid, since killing all amphid neurons completely elimi- nated the chemotaxis responses.

Animals with the ASE cells and additional cells killed were compared with unoperated animals, with animals with only the ASE cells killed, and with the negative control value for the assay. No single amphid neuron accounted for the residual response after the ASE neurons were killed (Table 1; also data not shown). However, killing the ASE neurons together with the amphid neurons ADF, ASC, and ASI resulted in animals completely defective in chemotaxis toward biotin, Cl-, Na+ , and possibly CAMP (Table IE, IL; Fig- ure 3A).

All combinations of 3 of the 4 cell types ASE, ADF, ASG, and ASI were killed in many individual animals (Table IE, IL; Figure 3B). Only animals in which all 4 cell types were killed had as great a combined defect in chemotaxis to the attractants CAMP, biotin, Cl-, and Na+ as the-2 animals (p < 0.01).

The ability of operated animals to chemotax toward lysine and serotonin was not highly correlated with their ability to chemotax toward the other chemicals (Figure 3C). This difference can be seen most readily in animals in which ASE, ADF, ASC, and ASI have been killed (Figures 3A-3C). The responses to serotonin and lysine were the weakest of those measured in intact animals, but a significant component of these responses was still present after these 4 chemosensory cell types were killed. The responses to amino acids, which ordinarily are the least robust responses, became the most robust chemotaxis responses following laser surgery. These observations indicate that the defects in chemotaxis to CAMP, biotin, Cl-, and Na+ are not likely to be a result of general unhealthiness caused by sensory cell death.

The basis for the residual response to lysine was

Chemosensory Neurons in C. elegans 73!

_ ” N2 ASE- E-F-b-I- toe-2

c

-T T ‘“1

Figure 3. Effects of Laser Killing of Amphidial Neurons on Chemotaxis

‘ihis figure summarizes data presented in Table 1. Error bars indicate the estimated 95% confidence limits (see Experimental Procedures). Y2, wild-type C. elegans var. Bristol; the-2, the-2(e1033) animals. All other entries are wild-type animals in which neurons were killed. rhe cell types killed are abbreviated by using the last letter in their names, thus, in entries that include Em, ASE neurons were killed; :-, ADF neurons were killed; Cm, ASG neurons were killed; I-, ASI neurons were killed; K-, ASK neurons were killed; for controls, cells

+lpparently uninvolved in chemotaxis were killed. The 20% value represents complete elimination of chemotaxis (see text); lower values :ould arise from weak avoidance of the compounds, but they are statistically indistinguishable from zero. A) Chemotaxis of wild-type and the-2 animals and of wild-type animals in which ASE was killed, or ASE, ADF, ASC, and ASI were killed. B) Chemotaxis of animals with combinations of cells ASE, ADF, ASC, and ASI killed toward CAMP, biotin, Cl-., and Na’. The chemotaxis

Index given is calculated by combining the results of assays to all of these attractants fTable 1) to give an average response for the group. C) Chemotaxis of animals with combinations of cells ASE, ADF, ASG, and ASI killed toward lysine and serotonin.

D) Chemotaxis of animals with combinations of cells ASE, ASG, ASI, and ASK killed toward lysine. Since ADF does not contribute to the lysine response (Table I), the numbers for animals with or without ADF have been combined in this figure for entries E-G-l-K-, E-G-l-, and G-l-K-. For example, the animals labeled G-l-K- include 20 animals, 11 of which had only ASG, ASI, and ASK killed and 9 of which had ASC, ASI, ASK, and ADF killed.

sought in a similar set of experiments. When the sensory cells ASE, ASC, ASI, and ASK were killed, the loesponse to lysine was abolished (Figure 3D). All 4 of these cell types had to be killed to delete the response (Figure 3D). The responses to CAMP, biotin, Cl-, and Ua+ in these animals were about the same as those in animals in whichonlyASE had been killed (49% versus 41%; Table IJ, IL). The cells responsible for chemotaxis to serotonin have not been identified.

Redundant Chemosensory Cells Increase the Efficiency of Chemotaxis The results of the above experiments indicate that some redundancy exists among the cells that sense a particular attractant. Nonetheless, in the responses to CAMP, biotin, Cl-, and Na+, the ASE cells appeared to be more important than the other cell types. There are several possible explanations for this observation. For example, the ASE cells and the minor cells might

Neuron 736

on 25 50 100 200 400 800 mM NaCl

have different thresholds for the attractants, in which case a different cell might appear to be most important depending on the concentration of attractant that was used. Alternatively, the minor cells might respond to the same concentration of attractant as do the ASE cells, but be lessefficient at initiating directed movement.

We explored these possibilities by testing animals that had particular cells killed for their responses to different concentrations of attractants. In chemotaxis experiments with populations of intact wild-type animals, we found that NaCl was attractive when concentrations at the peak of a gradient varied from 25 mM to 800 mM and that attractive responses decreased when the concentrations were lower or higher (data not shown). Unoperated animals, animals in which the ASE cells had been killed, and animals in which ADF, ASG, and ASI (the minor cell types involved in chemotaxis to Na+and Cl-) were killed were tested for their responses to each of six concentrations of NaCl (Figure 4).

Animals in which the ASE neurons were killed were significantly defective in their responses to all concentrations of NaCl over 25 mM. These data indicate that the ASE neurons direct chemotaxis responses throughout this range of NaCl concentrations. How- ever, at all NaCl concentrations, a residual response was present after the ASE neuronswere killed, indicat- ing that some minor cells could direct an inefficient response to the attractant. The animals in which ADF, ASG, and ASI were killed were also defective compared with unoperated animals, suggesting that the minor neurons contribute to the normal chemotaxis response. The importance of the minor cells was most marked at 800 mM NaCl (p < O.OOl), but was also apparent at lower concentrations of NaCI. The minor cells might be somewhat more effective in directing a response either to high concentrations of NaCl or to a steeper NaCl gradient.

Figure 4. Responses of Operated Animals to Different Concentrations of NaCl

Either unoperated wild-type animals, animals in which the ASE neurons were killed CASE-), or animals in which the ADF, ASC, and ASI neuronswere killed (F-G-I-) were testedfortheir responses to NaCI.Thecon- centration of NaCl at the peak of the gradient ranged from 25 mM to 800 mM. Chemo- taxis indices at each concentration are derived from 80 or more unoperated animals, 50 or more animals in which ASE was killed and 30 or more animals in which ADF, ASG, and ASI were killed. Error bars indicate the upper limit of the 95% confidence limits for each number (the similar lower limits are omitted for clarity). The 20% valuerepresentscompleteelimination of chemotaxis (see text).

Taken together, these data indicate that the strong response of wild-type animals to NaCl could result from a summation of the less efficient responses of several chemosensory neuron types.

Discussion

The ASE Neurons Are Required for Normal Chemotaxis The ASE neurons are essential for normal chemotaxis to a number of attractants with different chemical structures. The simplest hypothesis explaining the observed requirement for the ASE neurons is that they directly sense CAMP, biotin, Cl-, Na+, and lysine and that sensory information from the ASE neurons is used to orient the animal’s movement.

Several other explanations could account for the effectof killingtheASEneurons.ThedeathsoftheASE neurons might cause an animal to become generally deregulated in movement, or ASE might be necessary to maintain either the viability or the activity of the true chemosensory neurons (for example by atrophic interaction). These alternative explanations would predict that the deaths of some other cells, the true chemosensory neurons, would also lead to defective chemotaxis. However, no other class of exposed sensory neuron was necessary for chemotaxis to CAMP, biotin, Cl-, or Na+ (Table 1). Indeed, even after all amphid and phasmid neurons except ASE were killed together, animals responded to these attractants. Therefore, if ASE is not the true chemosensory neuron, chemosensory function must be highly delocal- ized among the amphid, phasmid, and inner labial sensilla, or found among cell types that are not exposed to the environment.

Our control experiments verify that no cells in the vicinity of ASE were essential for chemotaxis and that this procedure for laser surgery is able to kill individual cell types without damaging adjacent cells. Similar

C-temosensory Neurons in C. elegans 7: 7

specificity was demonstrated using laser surgery to kill cells in the pharyngeal nervous system of C. elegans (Avery and Horvitz, 1987, 1989).

Additional Sensory Neurons Have a Minor Role in Chemotaxis The deaths of the ASE neurons reduced but did not eliminate the abilities of animals to chemotax to each of the attractants, and no single cell type accounted for the residual responses after ASE was killed. Thus, the ability to direct chemotaxis is distributed among multiplechemosensorycell types. We have identified groups of cells that contribute to five of the six responses that were examined.

The residual ability of ASE-killed animals to chemotax required ASC, ASI, and either ADF or ASK, depending on the attractant tested. For the responses to c 4MP, biotin, Cl-, and Na+, at least 2 and probably all 3 of the ADF, ASC, and ASI cell types are involved in the residual chemotaxis following the deaths of the ASE neurons. However, the number of animals scored with respect to each attractant was not sufficient to determine whether all 4cell typesareequally involved in the sensation of all four attractants. Similarly, to eliminate the lysine response, 4 cell types (ASE, ASG, ASI, and ASK) had to be killed together. Since different cells are required for different responses, these chemotaxis defects are most likely explained by specific sensory defects rather than general nervous system abnormalities.

One simple model that explains these data is that the ASE, ASG, and ASI neurons sense CAMP, biotin, C I-, Na+, and lysine; the ASK neurons sense only ly- s:ne; and the ADF neurons sense only the other at-

tractants. However, more complex possibilities are also consistent with our data. Since we examined rela- tivelyfewanimalswith multipleneurons killed, poten- tially interesting regulatory or inhibitory interactions among the neurons could have been obscured by statistical insignificance. Minor chemosensory neurons could have been overlooked if they contribute to a normal response but are unable to direct a response after ASE, ADF, ASG, ASI, and ASK are killed. Alterna- tively, only a subset of the identified cells might nor- mallycontributetochemotaxis:l or moreoftheminor cells might participate in chemotaxis only after regulation, i.e., after it has assumed the function of a killed cell.

Only a subset of the C. elegans candidate chemosensory neurons defined by anatomy (Ward et al., 1975; Ware et al., 1975) were implicated in chemotaxis in these experiments. The other putative chemosensory neurons could be playing minor or regulatory roles in responses to compounds we assayed, sensing chemoattractants that we did not assay, or performing other chemosensory functions.

The complete cell lineage of C. elegans is known (Sulston and Horvitz, 1977; Kimble and Hirsh, 1979; Sulston et al., 1983). Although many chemosensory neurons arise from a particular branch of the lineage, the ASE, ADF, ASG, ASI, and ASK cells are no closer to one another by lineage than they are to other cell types of the amphid. The anatomical connectivities within the nervous system are also known (White et al., 1986). The synaptic targets of ASE, ADF, ASG, ASI, and ASK all include several interneurons of the head, among which are at least 2 of the interconnected neurons AIA, AIB, AIY, AIZ, and RIA. Some or all of these

Development- Neuron type ___) Behavior

Dauer formation- ASG

Chemotaxis: CAMP, biotin, Na +, Cl-

Chemotaxis: lysine

ASK A

Iauer recovery - ASJ

ASH L Chemical avoidance ADL -

i:igure 5. Assigned Functions for Chemosensory Cells of the Amphid

Eknctions for the 8 exposed cell types of the amphid, as derived from these and other studies of chemotaxis (this manuscript), dauer :ormation (Bargmann and Horvitz, 1991), and avoidance of chemical repellents (including high osmotic strength and garlic; Bargmann l:t al., 1990; J. Thomas and R. Horvitz, unpublished data).

Neuron 738

interneurons might receive information from the chemosensory cells for further sensory integration and processing.

Chemosensory Neurons Might Recognize and Distinguish Multiple Attractants These experiments demonstrate multiple functions for single cells and overlapping functions among groups of cells in the C. elegans chemosensory nervous system. Figure 5 lists the 8 exposed amphid cell types and their apparent functions, as derived from these and similar experiments (Bargmann and Hor- vitz, 1991; Bargmann et al., 1990; J. Thomas and R. Horvitz, unpublished data). The identification of functions for these neuron types provides a framework for further behavioral and genetic analysis of these questions in C. elegans.

A single cell type, ASE, is involved in the sensation of several unrelated chemicals, including some that can be distinguished by the animal. For example, under conditions in which the response to Na+ is satu- rated, C. elegans can still chemotax to Cl- (Ward, 1973). In bacteria, similar saturation experiments indicate that a single bacterial cell can sense and distinguish several attractants by expressing and regulating different receptors at the bacterial membrane (Adler, 1975; Goy et al., 1977; Koshland, 1988). One simple way in which multiple attractants could be sensed by the ASE neurons in C. elegans would utilize an analogous mechanism: different attractants might be sensed by different receptors on the ASE sensory endings, with the receptors independently regulated to distinguish among them. Alternatively, C. elegans might distinguish among attractants by comparing the inputs from several different chemosensory neurons.

The properties of these sensory neurons in C. elegans are analogous to those in other animals. Individ- ual olfactory neurons of vertebrates and the fruit fly Drosophila melanogaster respond to many distinct odorants (Sicard and Holley, 1984; Siddiqi, 1987). Sin- gle vertebrate taste bud neurons also respond to chemicallyand psychophysicallydistinguishable molecules (Scott and Chang, 1984). Thus, chemosensory neurons can have multiple reactivities even in animals in which the total number of neurons is much larger than in C. elegans.

Two very different responses to chemical stimuli are apparently initiated by the sensory neurons ADF, ASG, and ASI. Under conditions of crowding and starvation, C. elegans arrests its development in a specialized larval stage called the dauer larva (Cassada and Russell, 1975). Dauer larva formation is regulated by competing chemical stimuli from bacteria and a pheromone (Golden and Riddle, 1984). ADF, ASG, and ASI have a critical developmental role in the initiation of dauer larva formation (Bargmann and Horvitz, 1991), so they might be sensing food or pheromone signals (or both) to regulate development as well as sensing chemoattractants during chemotaxis. Similarly, the mechanosensory neurons of C. elegans apparently

initiate several distinct responses to touch (Chalfie et al., 1985).

Multiple Chemosensory Neurons Contribute to the Fidelity of Chemotaxis The responses of animals to different concentrations of NaCl suggest that parallel activity of the ASE neurons and the ADF, ASG, and ASI neurons might increase the efficiency of this response. Similarly, the left and right ASE neurons are not fully redundant in function, despite their similar morphologies and synaptic connections (White et al., 1986). Unlike in- sects that locate chemical stimuli by comparing their left and right antennae, C. elegans probably does not orient itself by comparing the left and right ASE neurons (Ward, 1973). However, the combined activity of this pair of neurons leads to increased fidelity of chemotaxis.

The C. elegans sensory neurons involved in chemotaxis (this work), dauer formation (Bargmann and Hor- vitz, 1991), mechanosensory responses (Chalfie et al., 1981; Way and Chalfie, 1989), and avoidance of chemical repellents (Bargmann et al., 1990; J. Thomas and R. Horvitz, unpublished data) have been identified. In

each case, a hierarchy of sensory neurons exists in which different cell types respond with varying thresholds or efficiencies to related stimuli: major cell types (like ASE in chemotaxis) are assisted by additional cell typeswith superficiallysimilarfunctions. Distribution of function among multiple neurons and parallel processing are widely observed in sensory systems (Bur- rows, 1987; Heiligenberg, 1988; Lockery and Kristan, 1990). Our data suggest that these strategies are also used in the simple C. elegans nervous system and that the apparent redundancy and functional overlap among chemosensory neurons might serve to increase the robustness and the versatility of the behavioral response.

Experimental Procedures

Strains Nematodes were grown on E. coli strain HBIOI at 20°C (Brenner, 1974). All animals were raised with plentiful food in uncrowded conditions. Wild-type animals were C. elegans variety Bristol, strain N2. Mutant strains the-2 (e70331 X (Lewis and Hodgkin, 1977) and /in-77(n677) I (Sternberg and Horvitz, 1988) were also used.

Chemotaxis Assays Chemotaxis assays were based on the assay developed by Ward (1973). Assay plates were IO-cm tissue culture dishes (Corning) containing 10 ml of 1.6% BBL-agar (Becton-Dickinson) or 2% Difco-agar, 5 mM potassium phosphate (pH 6.0), 1 mM CaCb, 1 mM MgSO+ 10 mM sodium phosphate (pH 7.21, except for assay plates for Na’chemotaxis, in which the sodium phosphate was omitted. An agar plug 6 mm in diameter (volume of 36 ~1) was removed from near one edge of the plate using a cork borer and replaced with a similar plug soaked in attractant. The attractant was permitted to equilibrate in the plate and form a gradient for 12-24 hr. The plug at the peak of the gradient was often removed at the beginning of the assay. A similar plug was removed from the plate at a point diametrically opposite the peak of the gradient; this plug was considered the negative control area (Figure I). Assay plates were air-dried for 1 hr before the

(:hemosensory Neurons in C. elegans i39

beginning of the assay to enhance the visibility of the animal’s tracks on the agar. The initial attractant concentration at the peak of the gradient was 0.2 M CAMP-NH,or biotin-NHI, 0.4 M NH&l or sodium acetate, 0.5 M lysine acetate, or 0.026 M serotonin- < reatinine sulfate complex (all attractants were obtained from SigmaChemical Co.). Theseconcentrations were chosen to max- imize the responses to each attractant. The highest concentrations of CAMP, biotin, lysine, or serotonin that are freely soluble in water were used, and the highest concentrations of NH&I or sodium acetate thatdid not inducean avoidanceof high osmotic strength were used. Attractants were adjusted to neutral pH using either NH40H or acetic acid, since ammonium and acetate ions are relatively unattractive to C. elegans Ward, 1973).

Animals were placed briefly on fresh agar plates to free them of bacteria and then transferred to assay plates on a platinum wire using2% methylcellulose (1500 centipoise; SigmaChemical Co.). This viscous and unattractive compound was used to trans- fer the animals without injuring them, since nematodes adhere to it but can be easily released from it onto the agar. An animal was placed on the assay plate equidistant from the two plugs, slightly displaced from the center of the plate, about 5 cm from the peak of the gradient of attractant. Animals were permitted to move freely for 1 hr, after which they were removed and their tracks were observed as indentations on the agar. Occasionally animals crawled off the edge of the agar surface and onto the plastic surface of the assay plate. These animals were gently removed and replaced at their starting position.

Assays were scored as either positive or negative by counting the number of independent trips to the peak of the gradient (A) and subtracting the number of independent trips to the negative control agar plug (6). If A - B was greater than zero, the assay was scored as positive; if A - B was less than or equal to zero, the assay was scored as negative. If the animal did not move at least 5 cm total in any direction during the assay, the assay was rrotcounted.Ananimalwasusuallysubjected to6assaysin 1 day, in the following order: CAMP, biotin, lysine, Cl-, Na+, serotonin. (There was no clear effect of order on efficiency of chemotaxis, although animals often moved more vigorously in the first two assays.) Some animals were tested only with a subset of these attractants; their responses were similar to those tested with all zttractants. Occasionally a single animal was tested several times \vith the same attractant. If an animal was tested more than once, i’ was given a single final score for the attractant that was the Gverage of the assays run; thus, if 1 assay to CAMP was positive 2nd 1 negative, the animal was scored as 0.5 and considered to have been tested in 1 assay.

Our scoring method is biased in favor of a positive outcome: in animal moving randomly will encounter the experimental plug more often than the control plug in some tests and hence he scored as positive at some frequency. This background fre- ouency was measured for wild-type animals by placing the animals on plates without added attractant. An animal arrived at clne of the two neutral plugs (mock attractants) in the agar in 46% c-f the assays (102/221 assays). Thus the apparent chemotaxis to 2 single plug as a result of this random movement was one-half c>f that, or 23%.

the-2 animals, which have severe chemosensory defects, scored positive in 18% of chemotaxis assays (Table 1). As a com- plementaryapproach to estimate the percentage of false-positive responses in the assays, 216 assays of the-2 animals (29-43 assays per attractant) were scored in reverse, i.e., scoring the assay as though the control plug had been the attractant plug. Of 216 assays, 40 (19%) gave positive results in reverse scoring. This value should approximate the false-positive rate of the-2 animals; it corresponds closely to the chemotaxis of the-2 animals (-48/2&I, or 18%). The results for all attractants were comparable.

The negative control level for the assay was set by combining the values for wild-type animals in the absence of attractant and tilevalues for the-2animals scored both normally and in reverse, TV give a final value of 139/701 assays (20%). This value was used 2s the baseline for all statistical comparisons.

Only adults within 48 hr of the LCadult molt were assayed for tqese experiments, which allowed at least 2.5 days following

laser surgery for the ablated cells to lose function. Chemotaxis has been observed in animals as young as the L2 stage (Ward, 1973).

Assignment of Cell Identities The identities of the 8 amphid chemosensory cells visualized by Nomarski microscopy in younganimalswereassigned by several criteria. First, larvae were soaked in the fluorescent dye fluores- cein isothiocyanate (FITC), which stains 6 of the 8 putative amphid chemosensory cells (ADF, ASH, ASI, ASJ, ASK, and ADL; Hedgecock et al., 1985). The positions of the dye-filling cell bodies were related to the positions of particular nuclei visible by Nomarski microscopy (Figure 2). A series of criteriawere defined based upon which of the nuclei of the dye-filling cells could be unambiguously identified in most animals viewed using Nomar- ski optics. These criteria included but were not limited to the following: the 3 cells ASK, ADL, and ASI are the 3 largest nuclei in the dorsal-medial region; ADF has a smaller nucleus than ASK and lies adjacent to ASK, slightly ventral to and laterally located from ASK; 7 cells, including ASH and ASI, define a Y-shaped lateral group in which ASH and ASJ are 2 of the 4 largest cells and always occupy stereotypic positions within the Y. The positions of ASE and ASC, which do not fill with FITC, were derived from those of the FITC-filling cells in combination with the anatomical data provided by Sulston et al. (1983) and White et al. (1986). ASE has a large nucleus and is adjacent to ASH in the Y-shaped group of cells. ASC was assigned to be a small nucleus that occupies a position between ADL and ASI, slightly dorsal and usually slightly anterior to ASE. All of theamphid chemosensory cells in an animal were identified prior to ablation of any cell; if twisting of the animal or interference from the pharynx caused any of the assignments to be ambiguous, the animal was discarded.

Theaccuracyofthecellassignmentsmadeasdescribedabove was confirmed by FITC filling after laser ablation of candidate nuclei for all of the cells that stain with FITC. These results support the assignments that were made on the basis of position. ASC is a small nucleus that does not fill with FITC, and there are several other nuclei near this cell that are distinguished from ASC only by position, most notably AWA and AWB. If those positions are not invariant, it is possible that AWA or AWB was killed instead of ASG. To address this question, a single putative ASG cell was killed in each of 2 animals and the animals were analyzed by electron microscopy (D. Jacobson, C. Bargmann, and R. Hotvitz, unpublished data). Animals were fixed in 2% paraformaldehyde, 2% glutaraldehyde in 0.1 M HEPES (pH 7.4) for1 hrat20°C,osmicated (1% OsO.,inO.l M HEPES),and embed- ded in agarose, and the anterior 5-10 pm were sectioned. Thin sections were viewed on a JEOL JEM 1200 EXII transmission electron microscope. The identities of the amphid neurons were assigned as by Ward et al. (1975) and Ware et al. (1975). In both animals, a single dendritic profile corresponding in position to the laser-killed ASC neuron sensory process was abnormally electron-dense and shrunken to IO%-2096 of its normal size. All other amphid neuron processes were apparently normal.

The inner labial IL2 neurons were identified using the monoclonal antibody MAb22 (E. Hedgecock and H. Bhatt, personal communication; Okamoto and Thomson, 1985). This antibody recognizes several cell types, including the IL2 neurons. Wild- type animals were doubly stained with MAb22 and the nucleic acid stain diamidinophenolindole (DAPI), using the protocol of Desai et al. (1988). The positions of the nuclei that stained with MAb22 relative to all nuclei stained with DAPI were observed. Candidate IL2 nuclei were then identified in live Ll larvae using Nomarski optics and killed using a laser microbeam. After the animals grew to adulthood, they were tested for chemotaxis and then fixed and stained with MAb22. One IL2 cell stained in 6 animals generated in this way, i.e., the IL2 neurons were identified correctly in 35 of 36 cases (6 cells each in 6 animals).

laser Ablation of Neurons Cells were killed using a laser microbeam as described by Avery and Horvitz (1987, 1989). Animals in the first larval stage were

mounted on slides on 5% agar pads containing 3-5 mM NaN3 as anesthetic. Several different kinds of damage were considered to indicatethatacell had been killed: achangein nuclearappear- ante to the flattened disc shape characteristic of programmed cell death (Sulston and Horvitz, 1977); disappearance or striking alteration in the appearance of the nucleus and appearance of a distinct region assumed to correspond to the cell body; or the appearance of scars that transected the nucleus. We observed that other nuclei could change position following the death of 1 cell, so the cell being killed was watched after surgery until sufficient damage was observed to consider the cell dead (5- 60 min).

Animals were recovered from the slide in M9 buffer (Brenner, 1974) and placed on a bacterial lawn. Chemotaxis assays were conducted on operated animals after they reached adulthood, usually 3 days after surgery. Following the chemotaxis assays, surviving animals were soaked in FITC and visualized to confirm the deaths of FITC-filling cells (Hedgecock et al., 1985; Perkins et al., 1986). The criteria for nuclear death listed above usually led to the disappearance of the corresponding FITC-filling cell.

To confirm the specificity of the effect of killing the ASE neurons, all cells adjacent to ASE were killed and the animals were tested for chemotaxis. The cells that are closest to the ASE cells are ASG, ASJ, ASH, AVB, AWA, and RIB (Figure 2; also unpublished data). The first 3 are amphid cells, the deaths of which have little effect on chemotaxis compared with the deaths of the ASE cells (Table 2). AVB regulates forward movement (Chalfie et al., 1985), but is nonessential for chemotaxis; when AVB was killed in 8 animals, their chemotaxis responses were completely normal: 8/8 responded to CAMP, 7/7 to biotin, 617 to Cl-, and 3/4 to Na+. Similarly, killing RIB did not lead to a chemotaxis defect (5/5animals responded tocAMPand biotin,4/5 responded to Cl- and Na’), nor did killing AWA lead to a defect (2/2 animals responded to all attractants). Thus no other cell near ASE is required for chemotaxis.

When the ADF, ASC, ASI, and AS] neurons are all killed in the same young animals, many of the animals become dauer larvae and fail to develop to adulthood (Bargmann and Horvitz, 1991). For the experiments described here, only those animals that became adults were tested. To increase the proportion of adults among operated animals, these neurons were killed in the late Ll or early L2 stages; dauer larva form only if their development is initiated within the first larval stage(Swanson and Riddle, 1981). When the ADF and ASI neurons are killed and the ASJ neurons are not killed, many animals become dauer larvae transiently and then recover (Bargmann and Horvitz, 1991). Passing through the dauer stagedid not significantly increase or decrease chemotaxis of the resulting adults (data not shown).

Statistical Analysis Confidence limits of 95% for chemotaxis assays were calculated using the normal theory method for obtaining confidence inter- vals for binomial parameters and the formula

95% confidence range for chemotaxis frequency x = x f 1.964(x)(1 -x)/n

where n is the total number of assays from which x is derived. In some cases in which the number of assays was small, this

approximation was not valid and exact confidence limits were obtained using tables (Rosner, 1986).

Most pair-wise combinations between animals werecompared with controls using chi squareanalysis (Rosner, 1986). In all cases, the results for each attractant were considered independently. To identify weaker general effects on chemotaxis, we also combined results from all six attractants and results from CAMP, biotin, Cl-, and Na+ for the different sets of animals. These comparisons are described individually below. Wild-Type and the2 Animals Wild-type animals responded more efficiently to each attractant than to a mock attractant in control experiments (p < 0.001 for each attractant; see Chemotaxis Assays for details). Unoperated wild-type animals and the-2animals were demonstrated to differ

from each other in their responses to each attractant using chi square analysis (p < 0.001 in all cases). The responses of the-2 animals were indistinguishable from the responses of wild-type animals in the absence of attractants, and these values were combined to define the negative control value (see Chemotaxis Assays). Effect of Laser Ablation on Responses In the course of these and other experiments, many animals have been generated with many combinations of cells killed. Generally, the responses of these animals are slightly weaker than those of wild-type animals, regardless of which cells are killed (Table ID, IF, IK, IM; also unpublished data). We defined a set of control animals in which we killed only cells for which we had been unable to demonstrate any significant effect on chemotaxis either alone or in combination with other cells. One or more of the following cell types were killed in these animals: RIH, AVB, RIB, AIY, AIA,AFD, AWA,AWB, AWC, RIM, AUA, BAG, Cl, W, RIF, and SMB. Between 40 and 60 animals in this set were tested for their responses to each attractant. When the data for all assays with these animals were compared with those with wild-type animals, the responses were found to be significantly reduced (p < 0.01). The responses to serotonin of this set of animals were also reduced (p < O.Ol), but no other individual response was significantly lower than the corresponding response of unoperated animals.

All other chemotaxis responses have been compared with this set of responses as a control for effects of laser surgery and with the responses of unoperated animals. However, if any of the cells killed in the control set has a weak or cryptic effect on chemotaxis, the control level determined in these experiments might be inappropriately low. ASE Kills Animals in which ASE was killed were chemotaxis-defective compared with intact wild-type animals (p < 0.01 for all attractants). However, when these animals were compared with animals hav- ing the control ablations, the defects caused by killing ASE were significant only for the attractants CAMP, biotin, Cl-, and Nat (p < 0.01 in each case) and not for lysine and serotonin. Animals in which ASE was killed still had a significant residual response to each attractant (p < 0.01 in all cases) when compared with the negative control values for the assay.

The data from animals in which only ASE was killed were analyzed to determine whether these animals were truly intermedi- ate in their responses, or whether they were composed of a mixture of normal (like intact animals) and fully defective (like the-2) animals. This comparison was done by examining animals that were tested for their responses to all four attractants CAMP, biotin, Cl-, and Na+. These animals were given a score of 0 to 4 according to the number of assays to which they gave a positive response (most intact wild-type animals were scored as 3 or 4; most the-2 animals as 0 or 1). The distribution of scores for ASE- killed animals was then tested to see whether it could be composed of a mixture of intact wild-type and the-2 animals. A com- puter program written by Dr. Rolf E. Bargmann (personal communication) for this purpose generated the best fit for the datafora mixtureofwild-typeand the-2animals and then scored this fit by chi square analysis. ASE-killed animals were not well represented by a mixture of wild-type and the-2 animals (p < 0.01).

Killing either the left or right ASE neuron decreased the combined responses to CAMP, biotin, Cl-, and Na’ compared with either unoperated animals (p < 0.001 for either cell) or laser- operated controls (p < 0.05 for ASE left, p < 0.001 for ASE right, p < 0.001 for the 2 combined). Other Amphid Cell Kills Each of the 7 exposed amphid cells ADF, ASC, ASH, ASI, ASJ, ASK, and ADL was killed in at least IO animals in which ASE was not killed. Many animals of this group had several different cell types killed. For example, 7animals had both ASJ and ADL killed; these animals represent 7 of the 13 animals in which ASJ was killed as well as 7 of the 15 animals in which ADL was killed.

The combined values for all attractants were compared with those for unoperated animals and laser controls. The responses

C%nemosensory Neurons in C. elegans 7z 1

0:’ animals following the deaths of ASH, ASJ, ASK, or ADL were indistinguishablefrom unoperatedanimalsor lasercontrols.The responses of animals in which ADF, AS, or ASI were killed were significantly decreased compared with those of unoperated animals (p < 0.05 for ADF and ASG, p < 0.01 for ASI) but not significantly decreased compared with those of the laser controls. The responses of animals in which any of these cells was killed were much stronger than those of animals in which ASE was killed (p < 0.001 in all cases) or the negative control values (p < 0.001 in all cases). When the responses of these animals to CAMP, brotin,CI-,and Na+werecombined,onlyASI killsdifferedsignifi- cmtly from unoperated animals (p < 0.05) and no animals dif- fered from the laser controls.

The responses to individual attractants were also examined. The following results were significant when compared with unoperated wild-type animals: for the Na+ response, ADF (p < 0.05), ASH (p < 0.05), ASI (p < O.Ol), ASK (p < 0.05); for the lysine re- soonse, ADF (p < 0.05). However, when compared with the laser controls, only the decreased response to Na+of animals in which tire ASI cells were killed was significant (p < 0.05). Kills of All Amphid and Phasmid Neurons Except ASE The 5 animals with all amphid and phasmid neurons killed were compared with 5 animals with all amphid and phasmid neurons except ASE killed. No single response was significantly different between the two groups of operated animals (Table IG). How- ever, when the responses of these animals to all attractants were c’ombined, the animals with ASE alive were clearly more effi- cent at chemotaxis than those in which ASE was killed (p < 0.001; Table IN).

To estimate the minimum number of attractants to which ASE responded, we calculated the values for chemotaxis of these animals to different subsets of the attractants. In each case, animals with all amphid and phasmid cells killed were compared with animals with all except ASE killed. First, we combined the values for each group of five of the six attractants (for example, all assays except the Cl- assay) and compared the two types of cperated animals. In each case, the two groups of animals were cifferent (p< 0.01 in all cases). This result indicated that no single attractant accounted for the difference between the two groups c f animals and therefore that responses to at least two attractants were improved in the presence of ASE.

We then combined the values for each group of four of the SIX attractants (for example, all assays except the Cl- assay and me Na+ assay). Again, the animals were different in all cases (I) < 0.01). This result indicates that the responses to at least t’iree attractants were improved in the presence of ASE. Similar calculations grouping three attractants showed that at least four r,asponses were probably improved in the presence of ASE (p < C.05). Kills of ASE Together with Other Amphid Neurons Animals with multiple cell kills were compared with intact wild- t/pe animals, laser controls, negative controls, and animals in which only ASE was killed. When results with all six attractants c r with the four attractants CAMP, biotin, Cl-, and Na+ were combined, each combination of cells in which ASE was 1 of the cells killed was defectivecompared with either unoperated wild- type animals or laser controls (p < 0.001 in all cases).

Most of these combinations of kills resulted in animals with a residual chemotaxis responsecompared with the negativecon- tool values; only animals in which ASE, ADF, ASG, ASI, and ASK were all killed and animals in which ASE, ASI, and ASK were hilled were indistinguishable from the negative control levels (!a < 0.01 for all other combinations of cells). Since killing ASC together with ASE, ASI, and ASK resulted in animals that didstill respond to the attractants (p < O.Ol), we attribute the results of the ASE, ASI, and ASK kills to the small numbers of animals tested i- nd assume that some other cells must also sense the attractants.

When the results from the four attractants CAMP, biotin, Cl-, ;,nd Na+ were combined, animals in which ASE, ADF, ASG, and ,&I were killed were indistinguishable from the negative control value and significantly defective compared with animals in \vhichonlyASEwas killed(p<0.01).Animalsinwhichanysubset of 3 of those cells had been killed were significantly more effec-

tive at chemotaxis than the negative control values fp < 0.01 in all cases) and more effective than animals in which all 4 cells were killed (p < 0.01, except for kills of ASE, ASG, and ASI, for which p < 0.05; in the latter case, if the data from animals in which ASE, ASG, ASI, and ASK were killed are included, p<O.Ol).

The responses to lysine were analyzed in a similar way. Killing ASE, ASC, ASI, and ASK resulted in animals whose response to lysine was as poor as that of the negative controls, and the effect of killing ASE, ASG, ASI, and ASK was more severe than the effect of killing any 3 of those cell types (p < 0.01).

Except for the responses listed above, no significant differences were seen between animals in which ASE and other neuronswere killed and animals in which onlythe ASE neuronswere killed.

Acknowledgments

We thank Rolf Bargmann for assistance with statistical analysis; Dean Jacobson for sectioning the animals in which ASG was killed; and Leon Avery, Michael Basson, Marty Chalfie, Susan Glass, Erik Jorgensen, Josh Kaplan, Beth Sawin, and one anony- mous reviewer for discussions and critical reading of this manuscript. This work was supported by PHS grant number CM24663. C. I. 6. was supported by the Helen Hay Whitney Foundation and is a Lucille P. Markey Scholar. H. R. H. is an Investigator of the Howard Hughes Medical Institute.

The costs of publication of this article were defrayed in part by the payment of page charages. This article must therefore be hereby marked “advertisement” in accordance with 18 USC Section 1734 solely to indicate this fact.

Received March 8, 1991; revised August 6, 1991

References

Adler, J. (1975). Chemotaxis in bacteria. Annu. Rev. Biochem. 44, 341-356.

Albertson, D. C., and Thomson, J. N. (1976). The pharynx of Caenorhabditis elegans. Phil. Trans. Roy. Sot. (Land.) B 275,299- 325.

Avery, L., and Horvitz, H. R. (1987). A cell that dies during wild- type C. elegans development can function as a neuron in a ted-3 mutant. Cell 57, 1071-1078.

Avery, L., and Horvitz, H. R. (1989). Pharyngeal pumping contin- ues after laser killing of the pharyngeal nervous system of C. elegans. Neuron 3, 473-485.

Avery, L., and Horvitz, H. R. (1990). Effects of starvation and neu- reactive drugs on feeding in Caenorhabditis elegans. J. Exp. Zool. 253, 263-270.

Bargmann, C. I., and Horvitz, H. R. (1991). Control of larval development by chemosensory neurons in Caenorhabditis elegans. Science 257, 1243-1246.

Bargmann, C. I., Thomas, J. T., and Horvitz, H. R. (1990). Chemo- sensory cell function in the behavior and development of Caenorbabditis elegans. Cold Spring Harbor Symp. Quant. Biol. 55, 529-538.

Brenner, S. (1974). The genetics of Caenorhabditis elegans. Ce- netics 7, 71-94.

Burrows, M. (1987). Parallel processing of proprioceptive signals by spiking local interneurons and motor neurons in the locust. J. Neurosci. 7, 1064-1080.

Cassada, R. C., and Russell, R. L. (1975). The dauer larva, a postembryonic developmental variant of the nematode Caenorhab- ditis elegans. Dev. Biol. 46, 326-342.

Chalfie, M., and Sulston, J. (1981). Developmental genetics of the mechanosensory neurons of Caenorhabditis elegans. Dev. Biol. 82, 358-370.

Chalfie, M., Sulston, J. E., White, J. G., Southgate, E., Thomson, J. N., and Brenner, S. (1985). The neural circuit for touch sensitiv- ity in Caenorhabditis elegans. J. Neurosci. 5, 956-964.

742

Chiel, H. J., Weiss, K. R., and Kupfermann, I. (1988). An identified histaminergic neuron modulates feeding motor circuitry in Aplysia. J. Neurosci. 6, 2427-2450.

Desai, C., and Horvitz, H. R. (1989). Caenorhabditis elegans mu- tantsdefective in thefunctioningof the motor neurons responsible for egg laying. Genetics 727, 703-721.

Desai, C., Carriga, G., Mclntire, S. L., and Horvitz, H. R. (1988). A genetic pathway for the development of the Caenorhabditis elegans HSN motor neurons. Nature 336, 638-646.

Dusenbery, D. B. (1974). Analysis of chemotaxis in the nematode Caenorhabditis elegans by countercurrent separation. J. Exp. Zool 788,41-48.

Finger, T. E., and Silver, W. L., eds. (1987). Neurobiology of Taste and Smell (New York: John Wiley and Sons).

Golden, J. W., and Riddle, D. L. (1984). The Caenorhabditis e/e- gans dauer larva, developmental effects of pheromone, food and temperature. Dev. Biol. 702, 368-378.

Coy, M. F., Springer, M. S., and Adler, J. (1977). Sensory transduc- tion in fscherichia co/i: role of a protein methylation reaction in sensory adaptation. Proc. Natl. Acad. Sci. USA 74, 49644968.

Hedgecock, E. M., Culotti, J. C., Thomson, J. N., and Perkins, L. A. (1985). Axonal guidance mutants of Caenorhabditis elegans identified by filling sensory neurons with fluorescent dyes. Dev. Biol. 777, 158-170.

Heiligenberg, W. (1988). Electrosensory maps form a substrate for the distributed and parallel control of behavioral responses in weakly electric fish. Brain Behav. Evol. 37, 6-16.

Horvitz, H. R., Chalfie, M., Trent, C., Sulston, J., and Evans, P. D. (1982). Serotonin and octopamine in the nematode Caenorhab- ditis elegans. Science 276, 1012-1014.

Kimble, J., and Hirsh, D. (1979). The postembryonic cell lineages of the hermaphrodite and male gonads in Caenorhabditis e/e- gans. Dev. Biol. 70, 396-417.

Koshland, D. E. (1988). Chemotaxis as a model second messenger system. Biochemistry 27, 5829-5834.

Lewis, J. A., and Hodgkin, J. A. (1977). Specific neuroanatomical changes in chemosensory mutants of the nematode Caenorhab- ditis elegans. J. Comp. Neurol. 772, 489-510.

Lockery, S. R., and Kristan, W. B. (1990). Distributed processing of sensory information in the leech. 1. Input-output relations of the local bending reflex. J. Neurosci. 70, 1811-1815.

Nusbaum, M. P., and Marder, E. (1989). A modulatory proctolin- containing neuron (MPN). I. Identification and characterization. J. Neurosci 9, 1591-1599.

Okamoto, H., and Thomson, J. N. (1985). Monoclonal antibodies which distinguish certain classes of neuronal and support cells in the nervous tissue of the nematode Caenorhabditis elegans. J. Neurosci. 5, 643-653.

Perkins, L. A., Hedgecock, E. M., Thomson, J. N., and Culotti, J. G. (1986). Mutant sensory cilia in the nematode Caenorhabditis elegans. Dev. Biol. 777, 456-487.

Rosner, B. (1986). Fundamentals of Biostatistics (Boston: PWS Publishers).

Scott, T. R., and Chang, F.-C. T. (1984). The state of gustatory neural coding. Chem. Senses 8, 297-314.

Selverston, A. I., and Miller, J. P. (1980). Mechanisms underlying pattern generation in lobster stomatogastric ganglion as determined by selective inactivation of identified neurons. I. Pyloric system. J. Neurophysiol. 44, 1102-1121.

Sicard, C., and Holley, A. (1984). Receptor cell responses toodor- ants: similarities and differences among odorants. Brain Res. 292, 283-296.

Siddiqi, 0. (1987). Neurogenetics of olfaction in Drosophila me/a- nogaster. Trends Cenet. 3, 137-142.

Sternberg, P. W., and Horvitz, H. R. (1988). lin-77 mutations of Caenorhabditis e/egans disrupt certain asymmetric cell divi- sions. Dev. Biol. 730, 67-73.

Sulston, J. E., and Horvitz, H. R. (1977). Post-embryonic cell lin-

eages of the nematode Caenorhabditis elegans. Dev. Biol. 56, 110-156.

Sulston, J. E., and White, J. C. (1980). Regulation and cell- autonomy during post-embryonic development of Caenorhab- ditis elegans. Dev. Biol. 78, 577-597.

Sulston, J. E., Schierenberg, E., White, 1. G., and Thomson, J. N. (1983). The embryonic cell lineage of the nematode Caenorhab- ditis elegans. Dev. Biol. 700, 64-119.

Swanson, M. M., and Riddle, D. L. (1981). Critical periods in the development of the Caenorhabditis e/egans dauer larva. Dev. Biol. 84, 27-40.

Trent,C.,Tsung, N.,and Horvitz, H. R. (1983). Egg-laying defective mutants of the nematode Caenorhabditis elegans. Genetics 704, 619-647.

Van Haastert, P. J. M., DeWit, R. J. W., and Konijn, T. M. (1982). Antagonists of chemoattractants reveal separate receptors for CAMP, folic acid, and pterin in Dictyostelium. Exp. Cell Res. 140, 453-456.

Van Houten, J., Hauser, D. C. R., and Levandowsky, M. (1981). Chemosensory behavior in protozoa. In Biochemistry and Physi- ology of Protozoa, Volume 4, M. Levandowsky and S. H. Hunter, eds. (New York: Academic Press, Inc.), pp. 67-124.

Ward, S. (1973). Chemotaxis by the nematode Caenorhabditrs elegans: identification of attractants and analysis of the response by use of mutants. Proc. Natl. Acad. Sci. USA 70, 817-821.

Ward, S., Thomson, N., White, J. G., and Brenner, S. (1975). Elec- tron microscopical reconstruction of the anterior sensory anatomy of the nematode Caenorhabditis elegans. J. Comp. Neurol. 760, 313-338.

Ware, R. W., Clark, D., Crossland, K., and Russell, R. L. (1975). The nerve ring of the nematode Caenorhabditis elegans: sensory input and motor output. J. Comp. Neurol. 762, 71-110.

Way, J. C., and Chalfie, M. (1989). The met3 gene of Caenorhab- ditis elegans requires its own product for maintained expression and is expressed in three neuronal cell types. Genes Dev. 3,1823- 1833.

White, J. G., Southgate, E., Thomson, J. N., and Brenner, S. (1976). The structure of the ventral nerve cord of Caenorhabditis e/e- gans. Phil. Trans. Roy. Sot. (Lond.) B 275, 327-348.

White, J. G., Southgate, E., Thomson,J. N., and Brenner, S. (1986). The structure of the nervous system of the nematode Caenorhab- ditis elegans. Phil. Trans. Roy. Sot. (Lond.) B 374, I-340.

Documents

Chemosensory Neurons with Overlapping Functions Direct - POST