somites of cytotactin and its proteoglycan ligand is correlated with neural crest cell distribution. Proceedings of the National Academy of Sciences of the United States of America 84:7977-7981, 1987.

Tomaselli, K.J.; Damsky, C.H.; and Reichardt, L.F. Interactions of a neuronal cell line (PC12) with laminin, collagen IV, and fibronectin: Identification of integrin-related glycoproteins involved in attachment and process outgrowth. Journal of Cell Biology 105:2347-2358, 1987.

Tomaselli, K.J.; Damsky, C.H.; and Reichardt, L.F. Purification and characterization of mammalian integrins expressed by a rat neuronal cell line (PC12): Evidence that they function as a/ß heterodimeric receptors for laminin and type IV collagen. Journal of Cell Biology, in press, 1988b.

Tomaselli, K.J.; Neugebauer, K.M.; Bixby, J.L.; Lilien, J.; and Reichardt, L.F. N-cadherin and integrins: Two receptor systems that mediate neuronal process outgrowth on astrocytes. Neuron 1:33-43, 1988a.

Tomaselli, K.J., and Reichardt, L.F. Peripheral motoneuron interactions with laminin and Schwann cell-derived neurite-promoting factors molecules: Developmental regulation of laminin receptor function. Submitted, 1988.

Tomaselli, K.J.; Reichardt, L.F.; and Bixby, J.L. Distinct molecular interactions mediate neuronal process outgrowth on nonneuronal cell surfaces and extracellular matrices. Journal of Cell Biology 103:2659-2672, 1986.

van Noesel, C.; Miedema, F.; Brouwer, M.; de Rie, M.A.; Aarden, L.A.; and van Lier, R.A.W. Regulatory properties of LFA-1 a and ẞ chains in human T-lymphocyte activation. Nature 333:850-851, 1988.

Volk, T.; Cohen, O.; and Geiger, B. Formation of heterotypic adherens-type junctions between L-CAM-containing liver cells and A-CAM-containing lens cells. Cell 50:987-994, 1987.

Wayner, E.A., and Carter, W.G. Identification of multiple cell adhesion receptors for collagen and fibronectin in human fibrosarcoma cells possessing unique a and common ẞ subunits. Journal of Cell Biology 105:1873-1884, 1987.

The Period Gene and Biological

Rhythms in Drosophila

M. Rosbash, H.V. Colot, J. Ewer, X. Liu, G. Petersen, K.K. Siwicki, Q. Yu, D.A. Wheeler, and J.C. Hall

Department of Biology, Brandeis University
Waltham, MA 02254

Many biological and biochemical parameters are known to fluctuate rhythmically, with either circadian (~24-hour) or ultradian (much less than 24hour) periods (Takahashi and Menaker 1984; Schulz and Lavie 1985). Despite the many years devoted to their study, "clocks" remain an almost complete mystery. Over the last 15-20 years, isolation of rhythm mutants has led to the identification of several genetic loci that help control the construction or operation of oscillators in a few different species of eukaryotes. Detailed analyses of one of these genes, the period (per) gene of Drosophila melanogaster, is beginning to yield insights into the molecular nature of biological clocks.

Circadian rhythms have a number of characteristic properties. (1) They will "free run" under constant environmental conditions and, therefore, are not mere responses to changes in light or temperature. (2) Although bona fide rhythms are free running, a previous signal from the environment is required (at least in some invertebrates) to “entrain" the rhythmicity or start the clock; these cues can consist of alternating light/dark or high/low temperature cycles, or even short pulses of light. (3) The period of a circadian rhythm is relatively insensitive to changes in temperature, so it is said to be "temperature-compensated." (4) Biological clocks can be reset, such that the phase of a rhythm can be altered by the same kinds of cues that affect entrainment. Depending on the phase of the cycle at which the cue is given, the rhythm will advance, delay, or not change at all, generating a characteristic phase response curve (PRC).

The genetic approach to understanding the control of biological rhythms has resulted in the isolation of a number of circadian mutants, some of which have turned out to exhibit aberrant ultradian rhythms as well. The two circadian rhythms that have been most studied in Drosophila are periodic

eclosion in populations of flies (i.e., the timing of emergence of adults from pupal cases) and rest/activity cycles in individual adults. Mutants were isolated by treating flies with chemical mutagens and monitoring either eclosion or locomotor activity of their descendants (Konopka and Benzer 1971; Konopka 1986). In D. melanogaster, most of the rhythm mutations discovered so far are at one locus on the X chromosome, the per gene (Konopka and Benzer 1971). Different per alleles (Konopka and Benzer 1971; Konopka 1986) can confer shortened (pers) or lengthened (perL) circadian periods, or apparently abolish rhythmicity altogether (pero).

The main ultradian rhythm that has been studied in Drosophila is a temperature-compensated periodicity found in the courtship song of male flies. The oscillating parameter here is the rate at which the male generates pulses of tone accompanying the wing vibrations he directs toward the female. This rate fluctuates sinusoidally with a period of about a minute in D. melanogaster (Kyriacou and Hall 1980). A closely related Drosophila species sings with a different periodicity; D. simulans males have faster, 30-40 second song periods, superimposed on a rate of overall tone pulse production that is slower than that in the song of D. melanogaster males (Kyriacou and Hall 1980, 1986).

The significance of the per gene's action increased when it was found that mutations at this locus affect the courtship song rhythm as well as circadian rhythmicity (Kyriacou and Hall 1980). In pers males, the period of the song rhythm is considerably shortened, whereas it is dramatically lengthened in perL1; per01 seems to eliminate the song rhythm (Kyriacou and Hall 1980).

The per alleles (of which there are four) and per mutations, caused by deletions of the locus, are superficially equivalent in their effects on circadian rhythms; that is, all these genotypes lead to "arrhythmicity." Yet recent analyses, using special algorithms designed to extract rhythms out of “noisy" data, have revealed that 50-70 percent of per and per flies in fact show underlying ultradian (10-15 hour) periodicities in their locomotor activity (Dowse et al. 1987).

The reference to the per-deleted (per) genotype raises questions about the effects of this gene's dosage on circadian rhythms. Experiments involving altered copy numbers of the normal per allele, achieved by cytogenetic manipulations, showed that increased and decreased per+ dosages cause shortened and lengthened circadian periods, respectively (Smith and Konopka 1981, 1982). For example, females heterozygous for the normal allele and a deletion of the locus exhibit longer than normal periods, as do heterozygotes for the normal allele and a per mutation, or for a long-period per mutation and per+. These results suggest that per mutants can be thought of as "underproducers" of per activity, whereas the pers mutant would be an "overproducer," and per mutants possibly "null" for the action of this gene. This does not necessarily imply, for these mutants, that there are quantitative changes in transcription of a normal per structural gene. Indeed, it is easier

to imagine that pers has resulted in a qualitatively altered gene product. This has turned out to be so, and has been one of the interesting findings from molecular studies of this clock gene.

The per locus was isolated by chromosome walking and jumping from nearby clones and by microexcision experiments which, when augmented by the isolation of further clones from libraries and walking analysis, allowed a minilibrary to be built up whose clones could be said almost certainly to include all of per (Reddy et al. 1984; Bargiello and Young 1984).

One important early finding, which emerged from the analysis of chromosomal breakpoints and roughly indicated the limits of per, was that the genotypes designated per- are indeed homozygous for a deletion of about 10-15 kb of DNA (Reddy et al. 1984; Bargiello and Young 1984). They are, therefore, almost certainly devoid of the entirety of per, proving that the gene is nonvital. P element-mediated germ-line transformation with DNA fragments that include the coding region of the 4.5-kb transcript can restore circadian and ultradian (song) rhythms to arrhythmic hosts expressing either per or per geneotypes (Zehring et al. 1984; Bargiello et al. 1984; Hamblen et al. 1986).

The primary per transcript, which is processed to yield a 4.5-kb mRNA, contains 8 exons. We and others have analyzed per's transcription unit and predicted the nature of its protein products (Jackson et al. 1986; Citri et al. 1987). In general, the different studies have reached similar conclusions, with some minor discrepancies. Recently, it has been found that, in fact, the 4.5-kb transcript is a family of RNAs; all are similar in length but produced by differential splicing of the primary transcript (Citri et al. 1987). These results raise the possibility that the different products of per might have different functional significance – perhaps not surprising, given the different types of rhythms affected by per mutations.

Northern blotting and in situ hybridization (using DNA probes) have been used to determine the tissue and developmental distribution of per gene expression. The 4.5-kb mRNA is expressed in apparently all ganglia in the embryonic nervous system from about 45 to 80 percent of the way through this developmental stage (James et al. 1986). It has also been reported that per is expressed in embryonic salivary glands (Bargiello et al. 1987). Northern blotting indicates that expression wanes to low levels throughout the three larval stages (Reddy et al. 1984; James et al. 1986; Bargiello et al. 1987; Young et al. 1985), rises again during pupation, and continues during adulthood (Reddy et al. 1984; Bargiello et al. 1987), when it is enriched in the head (James et al. 1986).

Since no clocks seem to be "running" in Drosophila embryos, it is intriguing that per is expressed at this stage. Perhaps per participates in building clocks, as well as having a physiological role, later on, in running them. If per participates in the proper development of the embryonic brain and ventral

ganglia, it may be involved in the construction of "neural oscillators" (cf. Takahashi and Menaker 1984).

We suspect, however, that per expression in embryos is unrelated to the construction or running of the adult oscillators. Very recent results from our laboratories suggest that per expression in the adult is both necessary and sufficient for rhythmic behavior (Ewer et al. 1988; Yu et al. 1988). Presumably, the embryo expression contributes to some embryo or larval-specific (inessential) function such as the larval heartbeat rhythm (Livingstone 1981; Dowse et al. 1988).

In the adult, per is expressed in both nervous and nonnervous tissues. Assays on adult expression have used in situ hybridization (Liu et al. 1988), analysis of ß-galactosidase fusion-transformed strains (Liu et al. 1988), and direct localization of per protein with anti-per antibody reagents (Siwicki et al. 1988). The results indicate that the nervous system, the digestive system, and the reproductive system are all sites of per expression (Liu et al. 1988; Siwicki et al. 1988). In the nervous system, the eyes, optic lobes, central brain, and the thoracic and abdominal ganglia all show a characteristic and nonuniform pattern of per expression (Liu et al. 1988; Siwicki et al. 1988).

The brain expression data contain hints concerning the possible whereabouts of a central pacemaker; yet, it would seem that the expression is too widespread to be uniquely accounted for by a single oscillator function (Siwicki et al. 1988). Similarly, per expression in the reproductive and digestive systems may contribute to other functions in addition to possible semiautonomous circadian timers present in these tissues (Liu et al. 1988). At the risk of making an extreme understatement, let us say that the final word on these matters is not yet in.

With regard to biochemical function, we know even less. There is a report that per affects gap junction activity (Bargiello et al. 1987). Subcellular localization studies have not clarified matters, because the per protein appears to be cytoplasmic in some tissues (Siwicki et al. 1988), nuclear in others (Liu et al. 1988; Siwicki et al. 1988), and associated with the cell membrane and/or extracellular matrix in still others (Bargiello et al. 1987).

Three per mutations-per01, pers, and perL1- have been analyzed at the DNA level. The last mutation (the original per-long allele) (Konopka 1986) was pinpointed to a Val→ Asp substitution in the second amino acid-encoding exon (Baylies et al. 1987). The per01 allele is a nonsense mutation; pers is due to a single nucleotide substitution resulting in a Ser→ Asn missense mutation (Baylies et al. 1987; Yu et al. 1987). It was suggested that the Ser residue in the wild-type protein might be phosphorylated as part of the protein's normal function (Yu et al. 1987). This amino acid substitution will undoubtedly be very important in trying to determine just what this domain of the protein is doing and how, when mutated, it can cause a clock to run several hours or several seconds too fast.

The 4.5-kb transcript (and, in fact, the minor, alternatively spliced variants